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> |
163e783e | 29 | #include <linux/crc32c.h> |
5a0e3ad6 | 30 | #include <linux/slab.h> |
784b4e29 | 31 | #include <linux/migrate.h> |
7a36ddec | 32 | #include <linux/ratelimit.h> |
6463fe58 | 33 | #include <linux/uuid.h> |
7e75bf3f | 34 | #include <asm/unaligned.h> |
4b4e25f2 | 35 | #include "compat.h" |
eb60ceac CM |
36 | #include "ctree.h" |
37 | #include "disk-io.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" |
eb60ceac | 51 | |
de0022b9 JB |
52 | #ifdef CONFIG_X86 |
53 | #include <asm/cpufeature.h> | |
54 | #endif | |
55 | ||
d1310b2e | 56 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 57 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 58 | static void free_fs_root(struct btrfs_root *root); |
fcd1f065 | 59 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 60 | int read_only); |
569e0f35 JB |
61 | static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, |
62 | struct btrfs_root *root); | |
143bede5 | 63 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); |
acce952b | 64 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
65 | struct btrfs_root *root); | |
aec8030a | 66 | static void btrfs_evict_pending_snapshots(struct btrfs_transaction *t); |
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); | |
ce9adaa5 | 73 | |
d352ac68 CM |
74 | /* |
75 | * end_io_wq structs are used to do processing in task context when an IO is | |
76 | * complete. This is used during reads to verify checksums, and it is used | |
77 | * by writes to insert metadata for new file extents after IO is complete. | |
78 | */ | |
ce9adaa5 CM |
79 | struct end_io_wq { |
80 | struct bio *bio; | |
81 | bio_end_io_t *end_io; | |
82 | void *private; | |
83 | struct btrfs_fs_info *info; | |
84 | int error; | |
22c59948 | 85 | int metadata; |
ce9adaa5 | 86 | struct list_head list; |
8b712842 | 87 | struct btrfs_work work; |
ce9adaa5 | 88 | }; |
0da5468f | 89 | |
d352ac68 CM |
90 | /* |
91 | * async submit bios are used to offload expensive checksumming | |
92 | * onto the worker threads. They checksum file and metadata bios | |
93 | * just before they are sent down the IO stack. | |
94 | */ | |
44b8bd7e CM |
95 | struct async_submit_bio { |
96 | struct inode *inode; | |
97 | struct bio *bio; | |
98 | struct list_head list; | |
4a69a410 CM |
99 | extent_submit_bio_hook_t *submit_bio_start; |
100 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
101 | int rw; |
102 | int mirror_num; | |
c8b97818 | 103 | unsigned long bio_flags; |
eaf25d93 CM |
104 | /* |
105 | * bio_offset is optional, can be used if the pages in the bio | |
106 | * can't tell us where in the file the bio should go | |
107 | */ | |
108 | u64 bio_offset; | |
8b712842 | 109 | struct btrfs_work work; |
79787eaa | 110 | int error; |
44b8bd7e CM |
111 | }; |
112 | ||
85d4e461 CM |
113 | /* |
114 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
115 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
116 | * the level the eb occupies in the tree. | |
117 | * | |
118 | * Different roots are used for different purposes and may nest inside each | |
119 | * other and they require separate keysets. As lockdep keys should be | |
120 | * static, assign keysets according to the purpose of the root as indicated | |
121 | * by btrfs_root->objectid. This ensures that all special purpose roots | |
122 | * have separate keysets. | |
4008c04a | 123 | * |
85d4e461 CM |
124 | * Lock-nesting across peer nodes is always done with the immediate parent |
125 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
126 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 127 | * |
85d4e461 CM |
128 | * The key is set by the readpage_end_io_hook after the buffer has passed |
129 | * csum validation but before the pages are unlocked. It is also set by | |
130 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 131 | * |
85d4e461 CM |
132 | * We also add a check to make sure the highest level of the tree is the |
133 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
134 | * needs update as well. | |
4008c04a CM |
135 | */ |
136 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
137 | # if BTRFS_MAX_LEVEL != 8 | |
138 | # error | |
139 | # endif | |
85d4e461 CM |
140 | |
141 | static struct btrfs_lockdep_keyset { | |
142 | u64 id; /* root objectid */ | |
143 | const char *name_stem; /* lock name stem */ | |
144 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
145 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
146 | } btrfs_lockdep_keysets[] = { | |
147 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
148 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
149 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
150 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
151 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
152 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
153 | { .id = BTRFS_ORPHAN_OBJECTID, .name_stem = "orphan" }, | |
154 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, | |
155 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
156 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
157 | { .id = 0, .name_stem = "tree" }, | |
4008c04a | 158 | }; |
85d4e461 CM |
159 | |
160 | void __init btrfs_init_lockdep(void) | |
161 | { | |
162 | int i, j; | |
163 | ||
164 | /* initialize lockdep class names */ | |
165 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
166 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
167 | ||
168 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
169 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
170 | "btrfs-%s-%02d", ks->name_stem, j); | |
171 | } | |
172 | } | |
173 | ||
174 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
175 | int level) | |
176 | { | |
177 | struct btrfs_lockdep_keyset *ks; | |
178 | ||
179 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
180 | ||
181 | /* find the matching keyset, id 0 is the default entry */ | |
182 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
183 | if (ks->id == objectid) | |
184 | break; | |
185 | ||
186 | lockdep_set_class_and_name(&eb->lock, | |
187 | &ks->keys[level], ks->names[level]); | |
188 | } | |
189 | ||
4008c04a CM |
190 | #endif |
191 | ||
d352ac68 CM |
192 | /* |
193 | * extents on the btree inode are pretty simple, there's one extent | |
194 | * that covers the entire device | |
195 | */ | |
b2950863 | 196 | static struct extent_map *btree_get_extent(struct inode *inode, |
306e16ce | 197 | struct page *page, size_t pg_offset, u64 start, u64 len, |
b2950863 | 198 | int create) |
7eccb903 | 199 | { |
5f39d397 CM |
200 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
201 | struct extent_map *em; | |
202 | int ret; | |
203 | ||
890871be | 204 | read_lock(&em_tree->lock); |
d1310b2e | 205 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d CM |
206 | if (em) { |
207 | em->bdev = | |
208 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
890871be | 209 | read_unlock(&em_tree->lock); |
5f39d397 | 210 | goto out; |
a061fc8d | 211 | } |
890871be | 212 | read_unlock(&em_tree->lock); |
7b13b7b1 | 213 | |
172ddd60 | 214 | em = alloc_extent_map(); |
5f39d397 CM |
215 | if (!em) { |
216 | em = ERR_PTR(-ENOMEM); | |
217 | goto out; | |
218 | } | |
219 | em->start = 0; | |
0afbaf8c | 220 | em->len = (u64)-1; |
c8b97818 | 221 | em->block_len = (u64)-1; |
5f39d397 | 222 | em->block_start = 0; |
a061fc8d | 223 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e | 224 | |
890871be | 225 | write_lock(&em_tree->lock); |
09a2a8f9 | 226 | ret = add_extent_mapping(em_tree, em, 0); |
5f39d397 CM |
227 | if (ret == -EEXIST) { |
228 | free_extent_map(em); | |
7b13b7b1 | 229 | em = lookup_extent_mapping(em_tree, start, len); |
b4f359ab | 230 | if (!em) |
0433f20d | 231 | em = ERR_PTR(-EIO); |
5f39d397 | 232 | } else if (ret) { |
7b13b7b1 | 233 | free_extent_map(em); |
0433f20d | 234 | em = ERR_PTR(ret); |
5f39d397 | 235 | } |
890871be | 236 | write_unlock(&em_tree->lock); |
7b13b7b1 | 237 | |
5f39d397 CM |
238 | out: |
239 | return em; | |
7eccb903 CM |
240 | } |
241 | ||
b0496686 | 242 | u32 btrfs_csum_data(char *data, u32 seed, size_t len) |
19c00ddc | 243 | { |
163e783e | 244 | return crc32c(seed, data, len); |
19c00ddc CM |
245 | } |
246 | ||
247 | void btrfs_csum_final(u32 crc, char *result) | |
248 | { | |
7e75bf3f | 249 | put_unaligned_le32(~crc, result); |
19c00ddc CM |
250 | } |
251 | ||
d352ac68 CM |
252 | /* |
253 | * compute the csum for a btree block, and either verify it or write it | |
254 | * into the csum field of the block. | |
255 | */ | |
19c00ddc CM |
256 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
257 | int verify) | |
258 | { | |
6c41761f | 259 | u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy); |
607d432d | 260 | char *result = NULL; |
19c00ddc CM |
261 | unsigned long len; |
262 | unsigned long cur_len; | |
263 | unsigned long offset = BTRFS_CSUM_SIZE; | |
19c00ddc CM |
264 | char *kaddr; |
265 | unsigned long map_start; | |
266 | unsigned long map_len; | |
267 | int err; | |
268 | u32 crc = ~(u32)0; | |
607d432d | 269 | unsigned long inline_result; |
19c00ddc CM |
270 | |
271 | len = buf->len - offset; | |
d397712b | 272 | while (len > 0) { |
19c00ddc | 273 | err = map_private_extent_buffer(buf, offset, 32, |
a6591715 | 274 | &kaddr, &map_start, &map_len); |
d397712b | 275 | if (err) |
19c00ddc | 276 | return 1; |
19c00ddc | 277 | cur_len = min(len, map_len - (offset - map_start)); |
b0496686 | 278 | crc = btrfs_csum_data(kaddr + offset - map_start, |
19c00ddc CM |
279 | crc, cur_len); |
280 | len -= cur_len; | |
281 | offset += cur_len; | |
19c00ddc | 282 | } |
607d432d JB |
283 | if (csum_size > sizeof(inline_result)) { |
284 | result = kzalloc(csum_size * sizeof(char), GFP_NOFS); | |
285 | if (!result) | |
286 | return 1; | |
287 | } else { | |
288 | result = (char *)&inline_result; | |
289 | } | |
290 | ||
19c00ddc CM |
291 | btrfs_csum_final(crc, result); |
292 | ||
293 | if (verify) { | |
607d432d | 294 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
295 | u32 val; |
296 | u32 found = 0; | |
607d432d | 297 | memcpy(&found, result, csum_size); |
e4204ded | 298 | |
607d432d | 299 | read_extent_buffer(buf, &val, 0, csum_size); |
7a36ddec | 300 | printk_ratelimited(KERN_INFO "btrfs: %s checksum verify " |
193f284d CM |
301 | "failed on %llu wanted %X found %X " |
302 | "level %d\n", | |
303 | root->fs_info->sb->s_id, | |
304 | (unsigned long long)buf->start, val, found, | |
305 | btrfs_header_level(buf)); | |
607d432d JB |
306 | if (result != (char *)&inline_result) |
307 | kfree(result); | |
19c00ddc CM |
308 | return 1; |
309 | } | |
310 | } else { | |
607d432d | 311 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 312 | } |
607d432d JB |
313 | if (result != (char *)&inline_result) |
314 | kfree(result); | |
19c00ddc CM |
315 | return 0; |
316 | } | |
317 | ||
d352ac68 CM |
318 | /* |
319 | * we can't consider a given block up to date unless the transid of the | |
320 | * block matches the transid in the parent node's pointer. This is how we | |
321 | * detect blocks that either didn't get written at all or got written | |
322 | * in the wrong place. | |
323 | */ | |
1259ab75 | 324 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
325 | struct extent_buffer *eb, u64 parent_transid, |
326 | int atomic) | |
1259ab75 | 327 | { |
2ac55d41 | 328 | struct extent_state *cached_state = NULL; |
1259ab75 CM |
329 | int ret; |
330 | ||
331 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
332 | return 0; | |
333 | ||
b9fab919 CM |
334 | if (atomic) |
335 | return -EAGAIN; | |
336 | ||
2ac55d41 | 337 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
d0082371 | 338 | 0, &cached_state); |
0b32f4bb | 339 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
340 | btrfs_header_generation(eb) == parent_transid) { |
341 | ret = 0; | |
342 | goto out; | |
343 | } | |
7a36ddec | 344 | printk_ratelimited("parent transid verify failed on %llu wanted %llu " |
193f284d CM |
345 | "found %llu\n", |
346 | (unsigned long long)eb->start, | |
347 | (unsigned long long)parent_transid, | |
348 | (unsigned long long)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 | ||
d352ac68 CM |
357 | /* |
358 | * helper to read a given tree block, doing retries as required when | |
359 | * the checksums don't match and we have alternate mirrors to try. | |
360 | */ | |
f188591e CM |
361 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
362 | struct extent_buffer *eb, | |
ca7a79ad | 363 | u64 start, u64 parent_transid) |
f188591e CM |
364 | { |
365 | struct extent_io_tree *io_tree; | |
ea466794 | 366 | int failed = 0; |
f188591e CM |
367 | int ret; |
368 | int num_copies = 0; | |
369 | int mirror_num = 0; | |
ea466794 | 370 | int failed_mirror = 0; |
f188591e | 371 | |
a826d6dc | 372 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
f188591e CM |
373 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
374 | while (1) { | |
bb82ab88 AJ |
375 | ret = read_extent_buffer_pages(io_tree, eb, start, |
376 | WAIT_COMPLETE, | |
f188591e | 377 | btree_get_extent, mirror_num); |
256dd1bb SB |
378 | if (!ret) { |
379 | if (!verify_parent_transid(io_tree, eb, | |
b9fab919 | 380 | parent_transid, 0)) |
256dd1bb SB |
381 | break; |
382 | else | |
383 | ret = -EIO; | |
384 | } | |
d397712b | 385 | |
a826d6dc JB |
386 | /* |
387 | * This buffer's crc is fine, but its contents are corrupted, so | |
388 | * there is no reason to read the other copies, they won't be | |
389 | * any less wrong. | |
390 | */ | |
391 | if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) | |
ea466794 JB |
392 | break; |
393 | ||
5d964051 | 394 | num_copies = btrfs_num_copies(root->fs_info, |
f188591e | 395 | eb->start, eb->len); |
4235298e | 396 | if (num_copies == 1) |
ea466794 | 397 | break; |
4235298e | 398 | |
5cf1ab56 JB |
399 | if (!failed_mirror) { |
400 | failed = 1; | |
401 | failed_mirror = eb->read_mirror; | |
402 | } | |
403 | ||
f188591e | 404 | mirror_num++; |
ea466794 JB |
405 | if (mirror_num == failed_mirror) |
406 | mirror_num++; | |
407 | ||
4235298e | 408 | if (mirror_num > num_copies) |
ea466794 | 409 | break; |
f188591e | 410 | } |
ea466794 | 411 | |
c0901581 | 412 | if (failed && !ret && failed_mirror) |
ea466794 JB |
413 | repair_eb_io_failure(root, eb, failed_mirror); |
414 | ||
415 | return ret; | |
f188591e | 416 | } |
19c00ddc | 417 | |
d352ac68 | 418 | /* |
d397712b CM |
419 | * checksum a dirty tree block before IO. This has extra checks to make sure |
420 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 421 | */ |
d397712b | 422 | |
b2950863 | 423 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 424 | { |
d1310b2e | 425 | struct extent_io_tree *tree; |
4eee4fa4 | 426 | u64 start = page_offset(page); |
19c00ddc | 427 | u64 found_start; |
19c00ddc | 428 | struct extent_buffer *eb; |
f188591e | 429 | |
d1310b2e | 430 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc | 431 | |
4f2de97a JB |
432 | eb = (struct extent_buffer *)page->private; |
433 | if (page != eb->pages[0]) | |
434 | return 0; | |
19c00ddc CM |
435 | found_start = btrfs_header_bytenr(eb); |
436 | if (found_start != start) { | |
55c69072 | 437 | WARN_ON(1); |
4f2de97a | 438 | return 0; |
55c69072 | 439 | } |
55c69072 | 440 | if (!PageUptodate(page)) { |
55c69072 | 441 | WARN_ON(1); |
4f2de97a | 442 | return 0; |
19c00ddc | 443 | } |
19c00ddc | 444 | csum_tree_block(root, eb, 0); |
19c00ddc CM |
445 | return 0; |
446 | } | |
447 | ||
2b82032c YZ |
448 | static int check_tree_block_fsid(struct btrfs_root *root, |
449 | struct extent_buffer *eb) | |
450 | { | |
451 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
452 | u8 fsid[BTRFS_UUID_SIZE]; | |
453 | int ret = 1; | |
454 | ||
455 | read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), | |
456 | BTRFS_FSID_SIZE); | |
457 | while (fs_devices) { | |
458 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
459 | ret = 0; | |
460 | break; | |
461 | } | |
462 | fs_devices = fs_devices->seed; | |
463 | } | |
464 | return ret; | |
465 | } | |
466 | ||
a826d6dc JB |
467 | #define CORRUPT(reason, eb, root, slot) \ |
468 | printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \ | |
469 | "root=%llu, slot=%d\n", reason, \ | |
470 | (unsigned long long)btrfs_header_bytenr(eb), \ | |
471 | (unsigned long long)root->objectid, slot) | |
472 | ||
473 | static noinline int check_leaf(struct btrfs_root *root, | |
474 | struct extent_buffer *leaf) | |
475 | { | |
476 | struct btrfs_key key; | |
477 | struct btrfs_key leaf_key; | |
478 | u32 nritems = btrfs_header_nritems(leaf); | |
479 | int slot; | |
480 | ||
481 | if (nritems == 0) | |
482 | return 0; | |
483 | ||
484 | /* Check the 0 item */ | |
485 | if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != | |
486 | BTRFS_LEAF_DATA_SIZE(root)) { | |
487 | CORRUPT("invalid item offset size pair", leaf, root, 0); | |
488 | return -EIO; | |
489 | } | |
490 | ||
491 | /* | |
492 | * Check to make sure each items keys are in the correct order and their | |
493 | * offsets make sense. We only have to loop through nritems-1 because | |
494 | * we check the current slot against the next slot, which verifies the | |
495 | * next slot's offset+size makes sense and that the current's slot | |
496 | * offset is correct. | |
497 | */ | |
498 | for (slot = 0; slot < nritems - 1; slot++) { | |
499 | btrfs_item_key_to_cpu(leaf, &leaf_key, slot); | |
500 | btrfs_item_key_to_cpu(leaf, &key, slot + 1); | |
501 | ||
502 | /* Make sure the keys are in the right order */ | |
503 | if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { | |
504 | CORRUPT("bad key order", leaf, root, slot); | |
505 | return -EIO; | |
506 | } | |
507 | ||
508 | /* | |
509 | * Make sure the offset and ends are right, remember that the | |
510 | * item data starts at the end of the leaf and grows towards the | |
511 | * front. | |
512 | */ | |
513 | if (btrfs_item_offset_nr(leaf, slot) != | |
514 | btrfs_item_end_nr(leaf, slot + 1)) { | |
515 | CORRUPT("slot offset bad", leaf, root, slot); | |
516 | return -EIO; | |
517 | } | |
518 | ||
519 | /* | |
520 | * Check to make sure that we don't point outside of the leaf, | |
521 | * just incase all the items are consistent to eachother, but | |
522 | * all point outside of the leaf. | |
523 | */ | |
524 | if (btrfs_item_end_nr(leaf, slot) > | |
525 | BTRFS_LEAF_DATA_SIZE(root)) { | |
526 | CORRUPT("slot end outside of leaf", leaf, root, slot); | |
527 | return -EIO; | |
528 | } | |
529 | } | |
530 | ||
531 | return 0; | |
532 | } | |
533 | ||
727011e0 CM |
534 | struct extent_buffer *find_eb_for_page(struct extent_io_tree *tree, |
535 | struct page *page, int max_walk) | |
536 | { | |
537 | struct extent_buffer *eb; | |
538 | u64 start = page_offset(page); | |
539 | u64 target = start; | |
540 | u64 min_start; | |
541 | ||
542 | if (start < max_walk) | |
543 | min_start = 0; | |
544 | else | |
545 | min_start = start - max_walk; | |
546 | ||
547 | while (start >= min_start) { | |
548 | eb = find_extent_buffer(tree, start, 0); | |
549 | if (eb) { | |
550 | /* | |
551 | * we found an extent buffer and it contains our page | |
552 | * horray! | |
553 | */ | |
554 | if (eb->start <= target && | |
555 | eb->start + eb->len > target) | |
556 | return eb; | |
557 | ||
558 | /* we found an extent buffer that wasn't for us */ | |
559 | free_extent_buffer(eb); | |
560 | return NULL; | |
561 | } | |
562 | if (start == 0) | |
563 | break; | |
564 | start -= PAGE_CACHE_SIZE; | |
565 | } | |
566 | return NULL; | |
567 | } | |
568 | ||
b2950863 | 569 | static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
5cf1ab56 | 570 | struct extent_state *state, int mirror) |
ce9adaa5 CM |
571 | { |
572 | struct extent_io_tree *tree; | |
573 | u64 found_start; | |
574 | int found_level; | |
ce9adaa5 CM |
575 | struct extent_buffer *eb; |
576 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 577 | int ret = 0; |
727011e0 | 578 | int reads_done; |
ce9adaa5 | 579 | |
ce9adaa5 CM |
580 | if (!page->private) |
581 | goto out; | |
d397712b | 582 | |
727011e0 | 583 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
4f2de97a | 584 | eb = (struct extent_buffer *)page->private; |
d397712b | 585 | |
0b32f4bb JB |
586 | /* the pending IO might have been the only thing that kept this buffer |
587 | * in memory. Make sure we have a ref for all this other checks | |
588 | */ | |
589 | extent_buffer_get(eb); | |
590 | ||
591 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
727011e0 CM |
592 | if (!reads_done) |
593 | goto err; | |
f188591e | 594 | |
5cf1ab56 | 595 | eb->read_mirror = mirror; |
ea466794 JB |
596 | if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) { |
597 | ret = -EIO; | |
598 | goto err; | |
599 | } | |
600 | ||
ce9adaa5 | 601 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 602 | if (found_start != eb->start) { |
7a36ddec | 603 | printk_ratelimited(KERN_INFO "btrfs bad tree block start " |
193f284d CM |
604 | "%llu %llu\n", |
605 | (unsigned long long)found_start, | |
606 | (unsigned long long)eb->start); | |
f188591e | 607 | ret = -EIO; |
ce9adaa5 CM |
608 | goto err; |
609 | } | |
2b82032c | 610 | if (check_tree_block_fsid(root, eb)) { |
7a36ddec | 611 | printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n", |
193f284d | 612 | (unsigned long long)eb->start); |
1259ab75 CM |
613 | ret = -EIO; |
614 | goto err; | |
615 | } | |
ce9adaa5 | 616 | found_level = btrfs_header_level(eb); |
1c24c3ce JB |
617 | if (found_level >= BTRFS_MAX_LEVEL) { |
618 | btrfs_info(root->fs_info, "bad tree block level %d\n", | |
619 | (int)btrfs_header_level(eb)); | |
620 | ret = -EIO; | |
621 | goto err; | |
622 | } | |
ce9adaa5 | 623 | |
85d4e461 CM |
624 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
625 | eb, found_level); | |
4008c04a | 626 | |
ce9adaa5 | 627 | ret = csum_tree_block(root, eb, 1); |
a826d6dc | 628 | if (ret) { |
f188591e | 629 | ret = -EIO; |
a826d6dc JB |
630 | goto err; |
631 | } | |
632 | ||
633 | /* | |
634 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
635 | * that we don't try and read the other copies of this block, just | |
636 | * return -EIO. | |
637 | */ | |
638 | if (found_level == 0 && check_leaf(root, eb)) { | |
639 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); | |
640 | ret = -EIO; | |
641 | } | |
ce9adaa5 | 642 | |
0b32f4bb JB |
643 | if (!ret) |
644 | set_extent_buffer_uptodate(eb); | |
ce9adaa5 | 645 | err: |
79fb65a1 JB |
646 | if (reads_done && |
647 | test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
4bb31e92 | 648 | btree_readahead_hook(root, eb, eb->start, ret); |
4bb31e92 | 649 | |
53b381b3 DW |
650 | if (ret) { |
651 | /* | |
652 | * our io error hook is going to dec the io pages | |
653 | * again, we have to make sure it has something | |
654 | * to decrement | |
655 | */ | |
656 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 657 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 658 | } |
0b32f4bb | 659 | free_extent_buffer(eb); |
ce9adaa5 | 660 | out: |
f188591e | 661 | return ret; |
ce9adaa5 CM |
662 | } |
663 | ||
ea466794 | 664 | static int btree_io_failed_hook(struct page *page, int failed_mirror) |
4bb31e92 | 665 | { |
4bb31e92 AJ |
666 | struct extent_buffer *eb; |
667 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
668 | ||
4f2de97a | 669 | eb = (struct extent_buffer *)page->private; |
ea466794 | 670 | set_bit(EXTENT_BUFFER_IOERR, &eb->bflags); |
5cf1ab56 | 671 | eb->read_mirror = failed_mirror; |
53b381b3 | 672 | atomic_dec(&eb->io_pages); |
ea466794 | 673 | if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) |
4bb31e92 | 674 | btree_readahead_hook(root, eb, eb->start, -EIO); |
4bb31e92 AJ |
675 | return -EIO; /* we fixed nothing */ |
676 | } | |
677 | ||
ce9adaa5 | 678 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
679 | { |
680 | struct end_io_wq *end_io_wq = bio->bi_private; | |
681 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 682 | |
ce9adaa5 | 683 | fs_info = end_io_wq->info; |
ce9adaa5 | 684 | end_io_wq->error = err; |
8b712842 CM |
685 | end_io_wq->work.func = end_workqueue_fn; |
686 | end_io_wq->work.flags = 0; | |
d20f7043 | 687 | |
7b6d91da | 688 | if (bio->bi_rw & REQ_WRITE) { |
53b381b3 | 689 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) |
cad321ad CM |
690 | btrfs_queue_worker(&fs_info->endio_meta_write_workers, |
691 | &end_io_wq->work); | |
53b381b3 | 692 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) |
0cb59c99 JB |
693 | btrfs_queue_worker(&fs_info->endio_freespace_worker, |
694 | &end_io_wq->work); | |
53b381b3 DW |
695 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
696 | btrfs_queue_worker(&fs_info->endio_raid56_workers, | |
697 | &end_io_wq->work); | |
cad321ad CM |
698 | else |
699 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
700 | &end_io_wq->work); | |
d20f7043 | 701 | } else { |
53b381b3 DW |
702 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
703 | btrfs_queue_worker(&fs_info->endio_raid56_workers, | |
704 | &end_io_wq->work); | |
705 | else if (end_io_wq->metadata) | |
d20f7043 CM |
706 | btrfs_queue_worker(&fs_info->endio_meta_workers, |
707 | &end_io_wq->work); | |
708 | else | |
709 | btrfs_queue_worker(&fs_info->endio_workers, | |
710 | &end_io_wq->work); | |
711 | } | |
ce9adaa5 CM |
712 | } |
713 | ||
0cb59c99 JB |
714 | /* |
715 | * For the metadata arg you want | |
716 | * | |
717 | * 0 - if data | |
718 | * 1 - if normal metadta | |
719 | * 2 - if writing to the free space cache area | |
53b381b3 | 720 | * 3 - raid parity work |
0cb59c99 | 721 | */ |
22c59948 CM |
722 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
723 | int metadata) | |
0b86a832 | 724 | { |
ce9adaa5 | 725 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
726 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
727 | if (!end_io_wq) | |
728 | return -ENOMEM; | |
729 | ||
730 | end_io_wq->private = bio->bi_private; | |
731 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 732 | end_io_wq->info = info; |
ce9adaa5 CM |
733 | end_io_wq->error = 0; |
734 | end_io_wq->bio = bio; | |
22c59948 | 735 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
736 | |
737 | bio->bi_private = end_io_wq; | |
738 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
739 | return 0; |
740 | } | |
741 | ||
b64a2851 | 742 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 743 | { |
4854ddd0 CM |
744 | unsigned long limit = min_t(unsigned long, |
745 | info->workers.max_workers, | |
746 | info->fs_devices->open_devices); | |
747 | return 256 * limit; | |
748 | } | |
0986fe9e | 749 | |
4a69a410 CM |
750 | static void run_one_async_start(struct btrfs_work *work) |
751 | { | |
4a69a410 | 752 | struct async_submit_bio *async; |
79787eaa | 753 | int ret; |
4a69a410 CM |
754 | |
755 | async = container_of(work, struct async_submit_bio, work); | |
79787eaa JM |
756 | ret = async->submit_bio_start(async->inode, async->rw, async->bio, |
757 | async->mirror_num, async->bio_flags, | |
758 | async->bio_offset); | |
759 | if (ret) | |
760 | async->error = ret; | |
4a69a410 CM |
761 | } |
762 | ||
763 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
764 | { |
765 | struct btrfs_fs_info *fs_info; | |
766 | struct async_submit_bio *async; | |
4854ddd0 | 767 | int limit; |
8b712842 CM |
768 | |
769 | async = container_of(work, struct async_submit_bio, work); | |
770 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 771 | |
b64a2851 | 772 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
773 | limit = limit * 2 / 3; |
774 | ||
66657b31 | 775 | if (atomic_dec_return(&fs_info->nr_async_submits) < limit && |
b64a2851 | 776 | waitqueue_active(&fs_info->async_submit_wait)) |
4854ddd0 CM |
777 | wake_up(&fs_info->async_submit_wait); |
778 | ||
79787eaa JM |
779 | /* If an error occured we just want to clean up the bio and move on */ |
780 | if (async->error) { | |
781 | bio_endio(async->bio, async->error); | |
782 | return; | |
783 | } | |
784 | ||
4a69a410 | 785 | async->submit_bio_done(async->inode, async->rw, async->bio, |
eaf25d93 CM |
786 | async->mirror_num, async->bio_flags, |
787 | async->bio_offset); | |
4a69a410 CM |
788 | } |
789 | ||
790 | static void run_one_async_free(struct btrfs_work *work) | |
791 | { | |
792 | struct async_submit_bio *async; | |
793 | ||
794 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
795 | kfree(async); |
796 | } | |
797 | ||
44b8bd7e CM |
798 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
799 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 800 | unsigned long bio_flags, |
eaf25d93 | 801 | u64 bio_offset, |
4a69a410 CM |
802 | extent_submit_bio_hook_t *submit_bio_start, |
803 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
804 | { |
805 | struct async_submit_bio *async; | |
806 | ||
807 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
808 | if (!async) | |
809 | return -ENOMEM; | |
810 | ||
811 | async->inode = inode; | |
812 | async->rw = rw; | |
813 | async->bio = bio; | |
814 | async->mirror_num = mirror_num; | |
4a69a410 CM |
815 | async->submit_bio_start = submit_bio_start; |
816 | async->submit_bio_done = submit_bio_done; | |
817 | ||
818 | async->work.func = run_one_async_start; | |
819 | async->work.ordered_func = run_one_async_done; | |
820 | async->work.ordered_free = run_one_async_free; | |
821 | ||
8b712842 | 822 | async->work.flags = 0; |
c8b97818 | 823 | async->bio_flags = bio_flags; |
eaf25d93 | 824 | async->bio_offset = bio_offset; |
8c8bee1d | 825 | |
79787eaa JM |
826 | async->error = 0; |
827 | ||
cb03c743 | 828 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 | 829 | |
7b6d91da | 830 | if (rw & REQ_SYNC) |
d313d7a3 CM |
831 | btrfs_set_work_high_prio(&async->work); |
832 | ||
8b712842 | 833 | btrfs_queue_worker(&fs_info->workers, &async->work); |
9473f16c | 834 | |
d397712b | 835 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
836 | atomic_read(&fs_info->nr_async_submits)) { |
837 | wait_event(fs_info->async_submit_wait, | |
838 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
839 | } | |
840 | ||
44b8bd7e CM |
841 | return 0; |
842 | } | |
843 | ||
ce3ed71a CM |
844 | static int btree_csum_one_bio(struct bio *bio) |
845 | { | |
846 | struct bio_vec *bvec = bio->bi_io_vec; | |
847 | int bio_index = 0; | |
848 | struct btrfs_root *root; | |
79787eaa | 849 | int ret = 0; |
ce3ed71a CM |
850 | |
851 | WARN_ON(bio->bi_vcnt <= 0); | |
d397712b | 852 | while (bio_index < bio->bi_vcnt) { |
ce3ed71a | 853 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
79787eaa JM |
854 | ret = csum_dirty_buffer(root, bvec->bv_page); |
855 | if (ret) | |
856 | break; | |
ce3ed71a CM |
857 | bio_index++; |
858 | bvec++; | |
859 | } | |
79787eaa | 860 | return ret; |
ce3ed71a CM |
861 | } |
862 | ||
4a69a410 CM |
863 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
864 | struct bio *bio, int mirror_num, | |
eaf25d93 CM |
865 | unsigned long bio_flags, |
866 | u64 bio_offset) | |
22c59948 | 867 | { |
8b712842 CM |
868 | /* |
869 | * when we're called for a write, we're already in the async | |
5443be45 | 870 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 871 | */ |
79787eaa | 872 | return btree_csum_one_bio(bio); |
4a69a410 | 873 | } |
22c59948 | 874 | |
4a69a410 | 875 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
876 | int mirror_num, unsigned long bio_flags, |
877 | u64 bio_offset) | |
4a69a410 | 878 | { |
61891923 SB |
879 | int ret; |
880 | ||
8b712842 | 881 | /* |
4a69a410 CM |
882 | * when we're called for a write, we're already in the async |
883 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 884 | */ |
61891923 SB |
885 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
886 | if (ret) | |
887 | bio_endio(bio, ret); | |
888 | return ret; | |
0b86a832 CM |
889 | } |
890 | ||
de0022b9 JB |
891 | static int check_async_write(struct inode *inode, unsigned long bio_flags) |
892 | { | |
893 | if (bio_flags & EXTENT_BIO_TREE_LOG) | |
894 | return 0; | |
895 | #ifdef CONFIG_X86 | |
896 | if (cpu_has_xmm4_2) | |
897 | return 0; | |
898 | #endif | |
899 | return 1; | |
900 | } | |
901 | ||
44b8bd7e | 902 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
903 | int mirror_num, unsigned long bio_flags, |
904 | u64 bio_offset) | |
44b8bd7e | 905 | { |
de0022b9 | 906 | int async = check_async_write(inode, bio_flags); |
cad321ad CM |
907 | int ret; |
908 | ||
7b6d91da | 909 | if (!(rw & REQ_WRITE)) { |
4a69a410 CM |
910 | /* |
911 | * called for a read, do the setup so that checksum validation | |
912 | * can happen in the async kernel threads | |
913 | */ | |
f3f266ab CM |
914 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, |
915 | bio, 1); | |
1d4284bd | 916 | if (ret) |
61891923 SB |
917 | goto out_w_error; |
918 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
919 | mirror_num, 0); | |
de0022b9 JB |
920 | } else if (!async) { |
921 | ret = btree_csum_one_bio(bio); | |
922 | if (ret) | |
61891923 SB |
923 | goto out_w_error; |
924 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
925 | mirror_num, 0); | |
926 | } else { | |
927 | /* | |
928 | * kthread helpers are used to submit writes so that | |
929 | * checksumming can happen in parallel across all CPUs | |
930 | */ | |
931 | ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, | |
932 | inode, rw, bio, mirror_num, 0, | |
933 | bio_offset, | |
934 | __btree_submit_bio_start, | |
935 | __btree_submit_bio_done); | |
44b8bd7e | 936 | } |
d313d7a3 | 937 | |
61891923 SB |
938 | if (ret) { |
939 | out_w_error: | |
940 | bio_endio(bio, ret); | |
941 | } | |
942 | return ret; | |
44b8bd7e CM |
943 | } |
944 | ||
3dd1462e | 945 | #ifdef CONFIG_MIGRATION |
784b4e29 | 946 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
947 | struct page *newpage, struct page *page, |
948 | enum migrate_mode mode) | |
784b4e29 CM |
949 | { |
950 | /* | |
951 | * we can't safely write a btree page from here, | |
952 | * we haven't done the locking hook | |
953 | */ | |
954 | if (PageDirty(page)) | |
955 | return -EAGAIN; | |
956 | /* | |
957 | * Buffers may be managed in a filesystem specific way. | |
958 | * We must have no buffers or drop them. | |
959 | */ | |
960 | if (page_has_private(page) && | |
961 | !try_to_release_page(page, GFP_KERNEL)) | |
962 | return -EAGAIN; | |
a6bc32b8 | 963 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 964 | } |
3dd1462e | 965 | #endif |
784b4e29 | 966 | |
0da5468f CM |
967 | |
968 | static int btree_writepages(struct address_space *mapping, | |
969 | struct writeback_control *wbc) | |
970 | { | |
d1310b2e | 971 | struct extent_io_tree *tree; |
e2d84521 MX |
972 | struct btrfs_fs_info *fs_info; |
973 | int ret; | |
974 | ||
d1310b2e | 975 | tree = &BTRFS_I(mapping->host)->io_tree; |
d8d5f3e1 | 976 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
977 | |
978 | if (wbc->for_kupdate) | |
979 | return 0; | |
980 | ||
e2d84521 | 981 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 982 | /* this is a bit racy, but that's ok */ |
e2d84521 MX |
983 | ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
984 | BTRFS_DIRTY_METADATA_THRESH); | |
985 | if (ret < 0) | |
793955bc | 986 | return 0; |
793955bc | 987 | } |
0b32f4bb | 988 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
989 | } |
990 | ||
b2950863 | 991 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 992 | { |
d1310b2e CM |
993 | struct extent_io_tree *tree; |
994 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 995 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 996 | } |
22b0ebda | 997 | |
70dec807 | 998 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 999 | { |
98509cfc | 1000 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 1001 | return 0; |
0c4e538b DS |
1002 | /* |
1003 | * We need to mask out eg. __GFP_HIGHMEM and __GFP_DMA32 as we're doing | |
1004 | * slab allocation from alloc_extent_state down the callchain where | |
1005 | * it'd hit a BUG_ON as those flags are not allowed. | |
1006 | */ | |
1007 | gfp_flags &= ~GFP_SLAB_BUG_MASK; | |
1008 | ||
3083ee2e | 1009 | return try_release_extent_buffer(page, gfp_flags); |
d98237b3 CM |
1010 | } |
1011 | ||
5f39d397 | 1012 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 1013 | { |
d1310b2e CM |
1014 | struct extent_io_tree *tree; |
1015 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
1016 | extent_invalidatepage(tree, page, offset); |
1017 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 1018 | if (PagePrivate(page)) { |
d397712b CM |
1019 | printk(KERN_WARNING "btrfs warning page private not zero " |
1020 | "on page %llu\n", (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
1021 | ClearPagePrivate(page); |
1022 | set_page_private(page, 0); | |
1023 | page_cache_release(page); | |
1024 | } | |
d98237b3 CM |
1025 | } |
1026 | ||
0b32f4bb JB |
1027 | static int btree_set_page_dirty(struct page *page) |
1028 | { | |
bb146eb2 | 1029 | #ifdef DEBUG |
0b32f4bb JB |
1030 | struct extent_buffer *eb; |
1031 | ||
1032 | BUG_ON(!PagePrivate(page)); | |
1033 | eb = (struct extent_buffer *)page->private; | |
1034 | BUG_ON(!eb); | |
1035 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1036 | BUG_ON(!atomic_read(&eb->refs)); | |
1037 | btrfs_assert_tree_locked(eb); | |
bb146eb2 | 1038 | #endif |
0b32f4bb JB |
1039 | return __set_page_dirty_nobuffers(page); |
1040 | } | |
1041 | ||
7f09410b | 1042 | static const struct address_space_operations btree_aops = { |
d98237b3 | 1043 | .readpage = btree_readpage, |
0da5468f | 1044 | .writepages = btree_writepages, |
5f39d397 CM |
1045 | .releasepage = btree_releasepage, |
1046 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1047 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1048 | .migratepage = btree_migratepage, |
5a92bc88 | 1049 | #endif |
0b32f4bb | 1050 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1051 | }; |
1052 | ||
ca7a79ad CM |
1053 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1054 | u64 parent_transid) | |
090d1875 | 1055 | { |
5f39d397 CM |
1056 | struct extent_buffer *buf = NULL; |
1057 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 1058 | int ret = 0; |
090d1875 | 1059 | |
db94535d | 1060 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 1061 | if (!buf) |
090d1875 | 1062 | return 0; |
d1310b2e | 1063 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
bb82ab88 | 1064 | buf, 0, WAIT_NONE, btree_get_extent, 0); |
5f39d397 | 1065 | free_extent_buffer(buf); |
de428b63 | 1066 | return ret; |
090d1875 CM |
1067 | } |
1068 | ||
ab0fff03 AJ |
1069 | int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1070 | int mirror_num, struct extent_buffer **eb) | |
1071 | { | |
1072 | struct extent_buffer *buf = NULL; | |
1073 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1074 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; | |
1075 | int ret; | |
1076 | ||
1077 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
1078 | if (!buf) | |
1079 | return 0; | |
1080 | ||
1081 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1082 | ||
1083 | ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, | |
1084 | btree_get_extent, mirror_num); | |
1085 | if (ret) { | |
1086 | free_extent_buffer(buf); | |
1087 | return ret; | |
1088 | } | |
1089 | ||
1090 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1091 | free_extent_buffer(buf); | |
1092 | return -EIO; | |
0b32f4bb | 1093 | } else if (extent_buffer_uptodate(buf)) { |
ab0fff03 AJ |
1094 | *eb = buf; |
1095 | } else { | |
1096 | free_extent_buffer(buf); | |
1097 | } | |
1098 | return 0; | |
1099 | } | |
1100 | ||
0999df54 CM |
1101 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
1102 | u64 bytenr, u32 blocksize) | |
1103 | { | |
1104 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1105 | struct extent_buffer *eb; | |
1106 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
f09d1f60 | 1107 | bytenr, blocksize); |
0999df54 CM |
1108 | return eb; |
1109 | } | |
1110 | ||
1111 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
1112 | u64 bytenr, u32 blocksize) | |
1113 | { | |
1114 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1115 | struct extent_buffer *eb; | |
1116 | ||
1117 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
727011e0 | 1118 | bytenr, blocksize); |
0999df54 CM |
1119 | return eb; |
1120 | } | |
1121 | ||
1122 | ||
e02119d5 CM |
1123 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1124 | { | |
727011e0 | 1125 | return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, |
8aa38c31 | 1126 | buf->start + buf->len - 1); |
e02119d5 CM |
1127 | } |
1128 | ||
1129 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
1130 | { | |
727011e0 | 1131 | return filemap_fdatawait_range(buf->pages[0]->mapping, |
8aa38c31 | 1132 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
1133 | } |
1134 | ||
0999df54 | 1135 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 1136 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
1137 | { |
1138 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1139 | int ret; |
1140 | ||
0999df54 CM |
1141 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
1142 | if (!buf) | |
1143 | return NULL; | |
0999df54 | 1144 | |
ca7a79ad | 1145 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
5f39d397 | 1146 | return buf; |
ce9adaa5 | 1147 | |
eb60ceac CM |
1148 | } |
1149 | ||
d5c13f92 JM |
1150 | void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
1151 | struct extent_buffer *buf) | |
ed2ff2cb | 1152 | { |
e2d84521 MX |
1153 | struct btrfs_fs_info *fs_info = root->fs_info; |
1154 | ||
55c69072 | 1155 | if (btrfs_header_generation(buf) == |
e2d84521 | 1156 | fs_info->running_transaction->transid) { |
b9447ef8 | 1157 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1158 | |
b9473439 | 1159 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
e2d84521 MX |
1160 | __percpu_counter_add(&fs_info->dirty_metadata_bytes, |
1161 | -buf->len, | |
1162 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1163 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1164 | btrfs_set_lock_blocking(buf); | |
1165 | clear_extent_buffer_dirty(buf); | |
1166 | } | |
925baedd | 1167 | } |
5f39d397 CM |
1168 | } |
1169 | ||
143bede5 JM |
1170 | static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
1171 | u32 stripesize, struct btrfs_root *root, | |
1172 | struct btrfs_fs_info *fs_info, | |
1173 | u64 objectid) | |
d97e63b6 | 1174 | { |
cfaa7295 | 1175 | root->node = NULL; |
a28ec197 | 1176 | root->commit_root = NULL; |
db94535d CM |
1177 | root->sectorsize = sectorsize; |
1178 | root->nodesize = nodesize; | |
1179 | root->leafsize = leafsize; | |
87ee04eb | 1180 | root->stripesize = stripesize; |
123abc88 | 1181 | root->ref_cows = 0; |
0b86a832 | 1182 | root->track_dirty = 0; |
c71bf099 | 1183 | root->in_radix = 0; |
d68fc57b YZ |
1184 | root->orphan_item_inserted = 0; |
1185 | root->orphan_cleanup_state = 0; | |
0b86a832 | 1186 | |
0f7d52f4 CM |
1187 | root->objectid = objectid; |
1188 | root->last_trans = 0; | |
13a8a7c8 | 1189 | root->highest_objectid = 0; |
58176a96 | 1190 | root->name = NULL; |
6bef4d31 | 1191 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1192 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1193 | root->block_rsv = NULL; |
d68fc57b | 1194 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
1195 | |
1196 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1197 | INIT_LIST_HEAD(&root->root_list); |
2ab28f32 JB |
1198 | INIT_LIST_HEAD(&root->logged_list[0]); |
1199 | INIT_LIST_HEAD(&root->logged_list[1]); | |
d68fc57b | 1200 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 1201 | spin_lock_init(&root->inode_lock); |
f0486c68 | 1202 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1203 | spin_lock_init(&root->log_extents_lock[0]); |
1204 | spin_lock_init(&root->log_extents_lock[1]); | |
a2135011 | 1205 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1206 | mutex_init(&root->log_mutex); |
7237f183 YZ |
1207 | init_waitqueue_head(&root->log_writer_wait); |
1208 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1209 | init_waitqueue_head(&root->log_commit_wait[1]); | |
1210 | atomic_set(&root->log_commit[0], 0); | |
1211 | atomic_set(&root->log_commit[1], 0); | |
1212 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1213 | atomic_set(&root->log_batch, 0); |
8a35d95f | 1214 | atomic_set(&root->orphan_inodes, 0); |
7237f183 | 1215 | root->log_transid = 0; |
257c62e1 | 1216 | root->last_log_commit = 0; |
d0c803c4 | 1217 | extent_io_tree_init(&root->dirty_log_pages, |
f993c883 | 1218 | fs_info->btree_inode->i_mapping); |
017e5369 | 1219 | |
3768f368 CM |
1220 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1221 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1222 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 1223 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 1224 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 1225 | init_completion(&root->kobj_unregister); |
6702ed49 | 1226 | root->defrag_running = 0; |
4d775673 | 1227 | root->root_key.objectid = objectid; |
0ee5dc67 | 1228 | root->anon_dev = 0; |
8ea05e3a | 1229 | |
5f3ab90a | 1230 | spin_lock_init(&root->root_item_lock); |
3768f368 CM |
1231 | } |
1232 | ||
200a5c17 JM |
1233 | static int __must_check find_and_setup_root(struct btrfs_root *tree_root, |
1234 | struct btrfs_fs_info *fs_info, | |
1235 | u64 objectid, | |
1236 | struct btrfs_root *root) | |
3768f368 CM |
1237 | { |
1238 | int ret; | |
db94535d | 1239 | u32 blocksize; |
84234f3a | 1240 | u64 generation; |
3768f368 | 1241 | |
db94535d | 1242 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1243 | tree_root->sectorsize, tree_root->stripesize, |
1244 | root, fs_info, objectid); | |
3768f368 CM |
1245 | ret = btrfs_find_last_root(tree_root, objectid, |
1246 | &root->root_item, &root->root_key); | |
4df27c4d YZ |
1247 | if (ret > 0) |
1248 | return -ENOENT; | |
200a5c17 JM |
1249 | else if (ret < 0) |
1250 | return ret; | |
3768f368 | 1251 | |
84234f3a | 1252 | generation = btrfs_root_generation(&root->root_item); |
db94535d | 1253 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
af31f5e5 | 1254 | root->commit_root = NULL; |
db94535d | 1255 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), |
84234f3a | 1256 | blocksize, generation); |
b9fab919 | 1257 | if (!root->node || !btrfs_buffer_uptodate(root->node, generation, 0)) { |
68433b73 | 1258 | free_extent_buffer(root->node); |
af31f5e5 | 1259 | root->node = NULL; |
68433b73 CM |
1260 | return -EIO; |
1261 | } | |
4df27c4d | 1262 | root->commit_root = btrfs_root_node(root); |
d97e63b6 CM |
1263 | return 0; |
1264 | } | |
1265 | ||
f84a8bd6 | 1266 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info) |
6f07e42e AV |
1267 | { |
1268 | struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS); | |
1269 | if (root) | |
1270 | root->fs_info = fs_info; | |
1271 | return root; | |
1272 | } | |
1273 | ||
20897f5c AJ |
1274 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
1275 | struct btrfs_fs_info *fs_info, | |
1276 | u64 objectid) | |
1277 | { | |
1278 | struct extent_buffer *leaf; | |
1279 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1280 | struct btrfs_root *root; | |
1281 | struct btrfs_key key; | |
1282 | int ret = 0; | |
1283 | u64 bytenr; | |
6463fe58 | 1284 | uuid_le uuid; |
20897f5c AJ |
1285 | |
1286 | root = btrfs_alloc_root(fs_info); | |
1287 | if (!root) | |
1288 | return ERR_PTR(-ENOMEM); | |
1289 | ||
1290 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1291 | tree_root->sectorsize, tree_root->stripesize, | |
1292 | root, fs_info, objectid); | |
1293 | root->root_key.objectid = objectid; | |
1294 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1295 | root->root_key.offset = 0; | |
1296 | ||
1297 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, | |
1298 | 0, objectid, NULL, 0, 0, 0); | |
1299 | if (IS_ERR(leaf)) { | |
1300 | ret = PTR_ERR(leaf); | |
1dd05682 | 1301 | leaf = NULL; |
20897f5c AJ |
1302 | goto fail; |
1303 | } | |
1304 | ||
1305 | bytenr = leaf->start; | |
1306 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); | |
1307 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1308 | btrfs_set_header_generation(leaf, trans->transid); | |
1309 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1310 | btrfs_set_header_owner(leaf, objectid); | |
1311 | root->node = leaf; | |
1312 | ||
1313 | write_extent_buffer(leaf, fs_info->fsid, | |
1314 | (unsigned long)btrfs_header_fsid(leaf), | |
1315 | BTRFS_FSID_SIZE); | |
1316 | write_extent_buffer(leaf, fs_info->chunk_tree_uuid, | |
1317 | (unsigned long)btrfs_header_chunk_tree_uuid(leaf), | |
1318 | BTRFS_UUID_SIZE); | |
1319 | btrfs_mark_buffer_dirty(leaf); | |
1320 | ||
1321 | root->commit_root = btrfs_root_node(root); | |
1322 | root->track_dirty = 1; | |
1323 | ||
1324 | ||
1325 | root->root_item.flags = 0; | |
1326 | root->root_item.byte_limit = 0; | |
1327 | btrfs_set_root_bytenr(&root->root_item, leaf->start); | |
1328 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1329 | btrfs_set_root_level(&root->root_item, 0); | |
1330 | btrfs_set_root_refs(&root->root_item, 1); | |
1331 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1332 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1333 | btrfs_set_root_dirid(&root->root_item, 0); | |
6463fe58 SB |
1334 | uuid_le_gen(&uuid); |
1335 | memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); | |
20897f5c AJ |
1336 | root->root_item.drop_level = 0; |
1337 | ||
1338 | key.objectid = objectid; | |
1339 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1340 | key.offset = 0; | |
1341 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1342 | if (ret) | |
1343 | goto fail; | |
1344 | ||
1345 | btrfs_tree_unlock(leaf); | |
1346 | ||
1dd05682 TI |
1347 | return root; |
1348 | ||
20897f5c | 1349 | fail: |
1dd05682 TI |
1350 | if (leaf) { |
1351 | btrfs_tree_unlock(leaf); | |
1352 | free_extent_buffer(leaf); | |
1353 | } | |
1354 | kfree(root); | |
20897f5c | 1355 | |
1dd05682 | 1356 | return ERR_PTR(ret); |
20897f5c AJ |
1357 | } |
1358 | ||
7237f183 YZ |
1359 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1360 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1361 | { |
1362 | struct btrfs_root *root; | |
1363 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1364 | struct extent_buffer *leaf; |
e02119d5 | 1365 | |
6f07e42e | 1366 | root = btrfs_alloc_root(fs_info); |
e02119d5 | 1367 | if (!root) |
7237f183 | 1368 | return ERR_PTR(-ENOMEM); |
e02119d5 CM |
1369 | |
1370 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1371 | tree_root->sectorsize, tree_root->stripesize, | |
1372 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1373 | ||
1374 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1375 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1376 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
7237f183 YZ |
1377 | /* |
1378 | * log trees do not get reference counted because they go away | |
1379 | * before a real commit is actually done. They do store pointers | |
1380 | * to file data extents, and those reference counts still get | |
1381 | * updated (along with back refs to the log tree). | |
1382 | */ | |
e02119d5 CM |
1383 | root->ref_cows = 0; |
1384 | ||
5d4f98a2 | 1385 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, |
66d7e7f0 | 1386 | BTRFS_TREE_LOG_OBJECTID, NULL, |
5581a51a | 1387 | 0, 0, 0); |
7237f183 YZ |
1388 | if (IS_ERR(leaf)) { |
1389 | kfree(root); | |
1390 | return ERR_CAST(leaf); | |
1391 | } | |
e02119d5 | 1392 | |
5d4f98a2 YZ |
1393 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1394 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1395 | btrfs_set_header_generation(leaf, trans->transid); | |
1396 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1397 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1398 | root->node = leaf; |
e02119d5 CM |
1399 | |
1400 | write_extent_buffer(root->node, root->fs_info->fsid, | |
1401 | (unsigned long)btrfs_header_fsid(root->node), | |
1402 | BTRFS_FSID_SIZE); | |
1403 | btrfs_mark_buffer_dirty(root->node); | |
1404 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1405 | return root; |
1406 | } | |
1407 | ||
1408 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1409 | struct btrfs_fs_info *fs_info) | |
1410 | { | |
1411 | struct btrfs_root *log_root; | |
1412 | ||
1413 | log_root = alloc_log_tree(trans, fs_info); | |
1414 | if (IS_ERR(log_root)) | |
1415 | return PTR_ERR(log_root); | |
1416 | WARN_ON(fs_info->log_root_tree); | |
1417 | fs_info->log_root_tree = log_root; | |
1418 | return 0; | |
1419 | } | |
1420 | ||
1421 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1422 | struct btrfs_root *root) | |
1423 | { | |
1424 | struct btrfs_root *log_root; | |
1425 | struct btrfs_inode_item *inode_item; | |
1426 | ||
1427 | log_root = alloc_log_tree(trans, root->fs_info); | |
1428 | if (IS_ERR(log_root)) | |
1429 | return PTR_ERR(log_root); | |
1430 | ||
1431 | log_root->last_trans = trans->transid; | |
1432 | log_root->root_key.offset = root->root_key.objectid; | |
1433 | ||
1434 | inode_item = &log_root->root_item.inode; | |
1435 | inode_item->generation = cpu_to_le64(1); | |
1436 | inode_item->size = cpu_to_le64(3); | |
1437 | inode_item->nlink = cpu_to_le32(1); | |
1438 | inode_item->nbytes = cpu_to_le64(root->leafsize); | |
1439 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); | |
1440 | ||
5d4f98a2 | 1441 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1442 | |
1443 | WARN_ON(root->log_root); | |
1444 | root->log_root = log_root; | |
1445 | root->log_transid = 0; | |
257c62e1 | 1446 | root->last_log_commit = 0; |
e02119d5 CM |
1447 | return 0; |
1448 | } | |
1449 | ||
1450 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
1451 | struct btrfs_key *location) | |
1452 | { | |
1453 | struct btrfs_root *root; | |
1454 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1455 | struct btrfs_path *path; |
5f39d397 | 1456 | struct extent_buffer *l; |
84234f3a | 1457 | u64 generation; |
db94535d | 1458 | u32 blocksize; |
0f7d52f4 | 1459 | int ret = 0; |
8ea05e3a | 1460 | int slot; |
0f7d52f4 | 1461 | |
6f07e42e | 1462 | root = btrfs_alloc_root(fs_info); |
0cf6c620 | 1463 | if (!root) |
0f7d52f4 | 1464 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 1465 | if (location->offset == (u64)-1) { |
db94535d | 1466 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
1467 | location->objectid, root); |
1468 | if (ret) { | |
0f7d52f4 CM |
1469 | kfree(root); |
1470 | return ERR_PTR(ret); | |
1471 | } | |
13a8a7c8 | 1472 | goto out; |
0f7d52f4 CM |
1473 | } |
1474 | ||
db94535d | 1475 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1476 | tree_root->sectorsize, tree_root->stripesize, |
1477 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
1478 | |
1479 | path = btrfs_alloc_path(); | |
db5b493a TI |
1480 | if (!path) { |
1481 | kfree(root); | |
1482 | return ERR_PTR(-ENOMEM); | |
1483 | } | |
0f7d52f4 | 1484 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); |
13a8a7c8 YZ |
1485 | if (ret == 0) { |
1486 | l = path->nodes[0]; | |
8ea05e3a | 1487 | slot = path->slots[0]; |
5fbf83c1 | 1488 | btrfs_read_root_item(l, slot, &root->root_item); |
13a8a7c8 | 1489 | memcpy(&root->root_key, location, sizeof(*location)); |
0f7d52f4 | 1490 | } |
0f7d52f4 CM |
1491 | btrfs_free_path(path); |
1492 | if (ret) { | |
5e540f77 | 1493 | kfree(root); |
13a8a7c8 YZ |
1494 | if (ret > 0) |
1495 | ret = -ENOENT; | |
0f7d52f4 CM |
1496 | return ERR_PTR(ret); |
1497 | } | |
13a8a7c8 | 1498 | |
84234f3a | 1499 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1500 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1501 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1502 | blocksize, generation); |
416bc658 JB |
1503 | if (!root->node || !extent_buffer_uptodate(root->node)) { |
1504 | ret = (!root->node) ? -ENOMEM : -EIO; | |
1505 | ||
1506 | free_extent_buffer(root->node); | |
1507 | kfree(root); | |
1508 | return ERR_PTR(ret); | |
1509 | } | |
1510 | ||
5d4f98a2 | 1511 | root->commit_root = btrfs_root_node(root); |
79787eaa | 1512 | BUG_ON(!root->node); /* -ENOMEM */ |
13a8a7c8 | 1513 | out: |
08fe4db1 | 1514 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { |
e02119d5 | 1515 | root->ref_cows = 1; |
08fe4db1 LZ |
1516 | btrfs_check_and_init_root_item(&root->root_item); |
1517 | } | |
13a8a7c8 | 1518 | |
5eda7b5e CM |
1519 | return root; |
1520 | } | |
1521 | ||
edbd8d4e CM |
1522 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
1523 | struct btrfs_key *location) | |
5eda7b5e CM |
1524 | { |
1525 | struct btrfs_root *root; | |
1526 | int ret; | |
1527 | ||
edbd8d4e CM |
1528 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1529 | return fs_info->tree_root; | |
1530 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1531 | return fs_info->extent_root; | |
8f18cf13 CM |
1532 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1533 | return fs_info->chunk_root; | |
1534 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1535 | return fs_info->dev_root; | |
0403e47e YZ |
1536 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1537 | return fs_info->csum_root; | |
bcef60f2 AJ |
1538 | if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1539 | return fs_info->quota_root ? fs_info->quota_root : | |
1540 | ERR_PTR(-ENOENT); | |
4df27c4d YZ |
1541 | again: |
1542 | spin_lock(&fs_info->fs_roots_radix_lock); | |
5eda7b5e CM |
1543 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
1544 | (unsigned long)location->objectid); | |
4df27c4d | 1545 | spin_unlock(&fs_info->fs_roots_radix_lock); |
5eda7b5e CM |
1546 | if (root) |
1547 | return root; | |
1548 | ||
e02119d5 | 1549 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
1550 | if (IS_ERR(root)) |
1551 | return root; | |
3394e160 | 1552 | |
581bb050 | 1553 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); |
581bb050 LZ |
1554 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), |
1555 | GFP_NOFS); | |
35a30d7c DS |
1556 | if (!root->free_ino_pinned || !root->free_ino_ctl) { |
1557 | ret = -ENOMEM; | |
581bb050 | 1558 | goto fail; |
35a30d7c | 1559 | } |
581bb050 LZ |
1560 | |
1561 | btrfs_init_free_ino_ctl(root); | |
1562 | mutex_init(&root->fs_commit_mutex); | |
1563 | spin_lock_init(&root->cache_lock); | |
1564 | init_waitqueue_head(&root->cache_wait); | |
1565 | ||
0ee5dc67 | 1566 | ret = get_anon_bdev(&root->anon_dev); |
ac08aedf CM |
1567 | if (ret) |
1568 | goto fail; | |
3394e160 | 1569 | |
d68fc57b YZ |
1570 | if (btrfs_root_refs(&root->root_item) == 0) { |
1571 | ret = -ENOENT; | |
1572 | goto fail; | |
1573 | } | |
1574 | ||
1575 | ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid); | |
1576 | if (ret < 0) | |
1577 | goto fail; | |
1578 | if (ret == 0) | |
1579 | root->orphan_item_inserted = 1; | |
1580 | ||
4df27c4d YZ |
1581 | ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); |
1582 | if (ret) | |
1583 | goto fail; | |
1584 | ||
1585 | spin_lock(&fs_info->fs_roots_radix_lock); | |
2619ba1f CM |
1586 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
1587 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 | 1588 | root); |
d68fc57b | 1589 | if (ret == 0) |
4df27c4d | 1590 | root->in_radix = 1; |
d68fc57b | 1591 | |
4df27c4d YZ |
1592 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1593 | radix_tree_preload_end(); | |
0f7d52f4 | 1594 | if (ret) { |
4df27c4d YZ |
1595 | if (ret == -EEXIST) { |
1596 | free_fs_root(root); | |
1597 | goto again; | |
1598 | } | |
1599 | goto fail; | |
0f7d52f4 | 1600 | } |
4df27c4d YZ |
1601 | |
1602 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
1603 | root->root_key.objectid); | |
1604 | WARN_ON(ret); | |
edbd8d4e | 1605 | return root; |
4df27c4d YZ |
1606 | fail: |
1607 | free_fs_root(root); | |
1608 | return ERR_PTR(ret); | |
edbd8d4e CM |
1609 | } |
1610 | ||
04160088 CM |
1611 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1612 | { | |
1613 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1614 | int ret = 0; | |
04160088 CM |
1615 | struct btrfs_device *device; |
1616 | struct backing_dev_info *bdi; | |
b7967db7 | 1617 | |
1f78160c XG |
1618 | rcu_read_lock(); |
1619 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1620 | if (!device->bdev) |
1621 | continue; | |
04160088 CM |
1622 | bdi = blk_get_backing_dev_info(device->bdev); |
1623 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1624 | ret = 1; | |
1625 | break; | |
1626 | } | |
1627 | } | |
1f78160c | 1628 | rcu_read_unlock(); |
04160088 CM |
1629 | return ret; |
1630 | } | |
1631 | ||
ad081f14 JA |
1632 | /* |
1633 | * If this fails, caller must call bdi_destroy() to get rid of the | |
1634 | * bdi again. | |
1635 | */ | |
04160088 CM |
1636 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1637 | { | |
ad081f14 JA |
1638 | int err; |
1639 | ||
1640 | bdi->capabilities = BDI_CAP_MAP_COPY; | |
e6d086d8 | 1641 | err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY); |
ad081f14 JA |
1642 | if (err) |
1643 | return err; | |
1644 | ||
4575c9cc | 1645 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1646 | bdi->congested_fn = btrfs_congested_fn; |
1647 | bdi->congested_data = info; | |
1648 | return 0; | |
1649 | } | |
1650 | ||
8b712842 CM |
1651 | /* |
1652 | * called by the kthread helper functions to finally call the bio end_io | |
1653 | * functions. This is where read checksum verification actually happens | |
1654 | */ | |
1655 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1656 | { |
ce9adaa5 | 1657 | struct bio *bio; |
8b712842 CM |
1658 | struct end_io_wq *end_io_wq; |
1659 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1660 | int error; |
ce9adaa5 | 1661 | |
8b712842 CM |
1662 | end_io_wq = container_of(work, struct end_io_wq, work); |
1663 | bio = end_io_wq->bio; | |
1664 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1665 | |
8b712842 CM |
1666 | error = end_io_wq->error; |
1667 | bio->bi_private = end_io_wq->private; | |
1668 | bio->bi_end_io = end_io_wq->end_io; | |
1669 | kfree(end_io_wq); | |
8b712842 | 1670 | bio_endio(bio, error); |
44b8bd7e CM |
1671 | } |
1672 | ||
a74a4b97 CM |
1673 | static int cleaner_kthread(void *arg) |
1674 | { | |
1675 | struct btrfs_root *root = arg; | |
1676 | ||
1677 | do { | |
9d1a2a3a DS |
1678 | int again = 0; |
1679 | ||
76dda93c | 1680 | if (!(root->fs_info->sb->s_flags & MS_RDONLY) && |
9d1a2a3a DS |
1681 | down_read_trylock(&root->fs_info->sb->s_umount)) { |
1682 | if (mutex_trylock(&root->fs_info->cleaner_mutex)) { | |
1683 | btrfs_run_delayed_iputs(root); | |
1684 | again = btrfs_clean_one_deleted_snapshot(root); | |
1685 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
1686 | } | |
4cb5300b | 1687 | btrfs_run_defrag_inodes(root->fs_info); |
9d1a2a3a | 1688 | up_read(&root->fs_info->sb->s_umount); |
76dda93c | 1689 | } |
a74a4b97 | 1690 | |
9d1a2a3a | 1691 | if (!try_to_freeze() && !again) { |
a74a4b97 | 1692 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1693 | if (!kthread_should_stop()) |
1694 | schedule(); | |
a74a4b97 CM |
1695 | __set_current_state(TASK_RUNNING); |
1696 | } | |
1697 | } while (!kthread_should_stop()); | |
1698 | return 0; | |
1699 | } | |
1700 | ||
1701 | static int transaction_kthread(void *arg) | |
1702 | { | |
1703 | struct btrfs_root *root = arg; | |
1704 | struct btrfs_trans_handle *trans; | |
1705 | struct btrfs_transaction *cur; | |
8929ecfa | 1706 | u64 transid; |
a74a4b97 CM |
1707 | unsigned long now; |
1708 | unsigned long delay; | |
914b2007 | 1709 | bool cannot_commit; |
a74a4b97 CM |
1710 | |
1711 | do { | |
914b2007 | 1712 | cannot_commit = false; |
a74a4b97 | 1713 | delay = HZ * 30; |
a74a4b97 CM |
1714 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
1715 | ||
a4abeea4 | 1716 | spin_lock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1717 | cur = root->fs_info->running_transaction; |
1718 | if (!cur) { | |
a4abeea4 | 1719 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1720 | goto sleep; |
1721 | } | |
31153d81 | 1722 | |
a74a4b97 | 1723 | now = get_seconds(); |
8929ecfa YZ |
1724 | if (!cur->blocked && |
1725 | (now < cur->start_time || now - cur->start_time < 30)) { | |
a4abeea4 | 1726 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1727 | delay = HZ * 5; |
1728 | goto sleep; | |
1729 | } | |
8929ecfa | 1730 | transid = cur->transid; |
a4abeea4 | 1731 | spin_unlock(&root->fs_info->trans_lock); |
56bec294 | 1732 | |
79787eaa | 1733 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1734 | trans = btrfs_attach_transaction(root); |
914b2007 | 1735 | if (IS_ERR(trans)) { |
354aa0fb MX |
1736 | if (PTR_ERR(trans) != -ENOENT) |
1737 | cannot_commit = true; | |
79787eaa | 1738 | goto sleep; |
914b2007 | 1739 | } |
8929ecfa | 1740 | if (transid == trans->transid) { |
79787eaa | 1741 | btrfs_commit_transaction(trans, root); |
8929ecfa YZ |
1742 | } else { |
1743 | btrfs_end_transaction(trans, root); | |
1744 | } | |
a74a4b97 CM |
1745 | sleep: |
1746 | wake_up_process(root->fs_info->cleaner_kthread); | |
1747 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1748 | ||
a0acae0e | 1749 | if (!try_to_freeze()) { |
a74a4b97 | 1750 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa | 1751 | if (!kthread_should_stop() && |
914b2007 JK |
1752 | (!btrfs_transaction_blocked(root->fs_info) || |
1753 | cannot_commit)) | |
8929ecfa | 1754 | schedule_timeout(delay); |
a74a4b97 CM |
1755 | __set_current_state(TASK_RUNNING); |
1756 | } | |
1757 | } while (!kthread_should_stop()); | |
1758 | return 0; | |
1759 | } | |
1760 | ||
af31f5e5 CM |
1761 | /* |
1762 | * this will find the highest generation in the array of | |
1763 | * root backups. The index of the highest array is returned, | |
1764 | * or -1 if we can't find anything. | |
1765 | * | |
1766 | * We check to make sure the array is valid by comparing the | |
1767 | * generation of the latest root in the array with the generation | |
1768 | * in the super block. If they don't match we pitch it. | |
1769 | */ | |
1770 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1771 | { | |
1772 | u64 cur; | |
1773 | int newest_index = -1; | |
1774 | struct btrfs_root_backup *root_backup; | |
1775 | int i; | |
1776 | ||
1777 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1778 | root_backup = info->super_copy->super_roots + i; | |
1779 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1780 | if (cur == newest_gen) | |
1781 | newest_index = i; | |
1782 | } | |
1783 | ||
1784 | /* check to see if we actually wrapped around */ | |
1785 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
1786 | root_backup = info->super_copy->super_roots; | |
1787 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1788 | if (cur == newest_gen) | |
1789 | newest_index = 0; | |
1790 | } | |
1791 | return newest_index; | |
1792 | } | |
1793 | ||
1794 | ||
1795 | /* | |
1796 | * find the oldest backup so we know where to store new entries | |
1797 | * in the backup array. This will set the backup_root_index | |
1798 | * field in the fs_info struct | |
1799 | */ | |
1800 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
1801 | u64 newest_gen) | |
1802 | { | |
1803 | int newest_index = -1; | |
1804 | ||
1805 | newest_index = find_newest_super_backup(info, newest_gen); | |
1806 | /* if there was garbage in there, just move along */ | |
1807 | if (newest_index == -1) { | |
1808 | info->backup_root_index = 0; | |
1809 | } else { | |
1810 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1811 | } | |
1812 | } | |
1813 | ||
1814 | /* | |
1815 | * copy all the root pointers into the super backup array. | |
1816 | * this will bump the backup pointer by one when it is | |
1817 | * done | |
1818 | */ | |
1819 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1820 | { | |
1821 | int next_backup; | |
1822 | struct btrfs_root_backup *root_backup; | |
1823 | int last_backup; | |
1824 | ||
1825 | next_backup = info->backup_root_index; | |
1826 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1827 | BTRFS_NUM_BACKUP_ROOTS; | |
1828 | ||
1829 | /* | |
1830 | * just overwrite the last backup if we're at the same generation | |
1831 | * this happens only at umount | |
1832 | */ | |
1833 | root_backup = info->super_for_commit->super_roots + last_backup; | |
1834 | if (btrfs_backup_tree_root_gen(root_backup) == | |
1835 | btrfs_header_generation(info->tree_root->node)) | |
1836 | next_backup = last_backup; | |
1837 | ||
1838 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1839 | ||
1840 | /* | |
1841 | * make sure all of our padding and empty slots get zero filled | |
1842 | * regardless of which ones we use today | |
1843 | */ | |
1844 | memset(root_backup, 0, sizeof(*root_backup)); | |
1845 | ||
1846 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1847 | ||
1848 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1849 | btrfs_set_backup_tree_root_gen(root_backup, | |
1850 | btrfs_header_generation(info->tree_root->node)); | |
1851 | ||
1852 | btrfs_set_backup_tree_root_level(root_backup, | |
1853 | btrfs_header_level(info->tree_root->node)); | |
1854 | ||
1855 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1856 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1857 | btrfs_header_generation(info->chunk_root->node)); | |
1858 | btrfs_set_backup_chunk_root_level(root_backup, | |
1859 | btrfs_header_level(info->chunk_root->node)); | |
1860 | ||
1861 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
1862 | btrfs_set_backup_extent_root_gen(root_backup, | |
1863 | btrfs_header_generation(info->extent_root->node)); | |
1864 | btrfs_set_backup_extent_root_level(root_backup, | |
1865 | btrfs_header_level(info->extent_root->node)); | |
1866 | ||
7c7e82a7 CM |
1867 | /* |
1868 | * we might commit during log recovery, which happens before we set | |
1869 | * the fs_root. Make sure it is valid before we fill it in. | |
1870 | */ | |
1871 | if (info->fs_root && info->fs_root->node) { | |
1872 | btrfs_set_backup_fs_root(root_backup, | |
1873 | info->fs_root->node->start); | |
1874 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1875 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1876 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1877 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1878 | } |
af31f5e5 CM |
1879 | |
1880 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1881 | btrfs_set_backup_dev_root_gen(root_backup, | |
1882 | btrfs_header_generation(info->dev_root->node)); | |
1883 | btrfs_set_backup_dev_root_level(root_backup, | |
1884 | btrfs_header_level(info->dev_root->node)); | |
1885 | ||
1886 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
1887 | btrfs_set_backup_csum_root_gen(root_backup, | |
1888 | btrfs_header_generation(info->csum_root->node)); | |
1889 | btrfs_set_backup_csum_root_level(root_backup, | |
1890 | btrfs_header_level(info->csum_root->node)); | |
1891 | ||
1892 | btrfs_set_backup_total_bytes(root_backup, | |
1893 | btrfs_super_total_bytes(info->super_copy)); | |
1894 | btrfs_set_backup_bytes_used(root_backup, | |
1895 | btrfs_super_bytes_used(info->super_copy)); | |
1896 | btrfs_set_backup_num_devices(root_backup, | |
1897 | btrfs_super_num_devices(info->super_copy)); | |
1898 | ||
1899 | /* | |
1900 | * if we don't copy this out to the super_copy, it won't get remembered | |
1901 | * for the next commit | |
1902 | */ | |
1903 | memcpy(&info->super_copy->super_roots, | |
1904 | &info->super_for_commit->super_roots, | |
1905 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1906 | } | |
1907 | ||
1908 | /* | |
1909 | * this copies info out of the root backup array and back into | |
1910 | * the in-memory super block. It is meant to help iterate through | |
1911 | * the array, so you send it the number of backups you've already | |
1912 | * tried and the last backup index you used. | |
1913 | * | |
1914 | * this returns -1 when it has tried all the backups | |
1915 | */ | |
1916 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
1917 | struct btrfs_super_block *super, | |
1918 | int *num_backups_tried, int *backup_index) | |
1919 | { | |
1920 | struct btrfs_root_backup *root_backup; | |
1921 | int newest = *backup_index; | |
1922 | ||
1923 | if (*num_backups_tried == 0) { | |
1924 | u64 gen = btrfs_super_generation(super); | |
1925 | ||
1926 | newest = find_newest_super_backup(info, gen); | |
1927 | if (newest == -1) | |
1928 | return -1; | |
1929 | ||
1930 | *backup_index = newest; | |
1931 | *num_backups_tried = 1; | |
1932 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
1933 | /* we've tried all the backups, all done */ | |
1934 | return -1; | |
1935 | } else { | |
1936 | /* jump to the next oldest backup */ | |
1937 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1938 | BTRFS_NUM_BACKUP_ROOTS; | |
1939 | *backup_index = newest; | |
1940 | *num_backups_tried += 1; | |
1941 | } | |
1942 | root_backup = super->super_roots + newest; | |
1943 | ||
1944 | btrfs_set_super_generation(super, | |
1945 | btrfs_backup_tree_root_gen(root_backup)); | |
1946 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
1947 | btrfs_set_super_root_level(super, | |
1948 | btrfs_backup_tree_root_level(root_backup)); | |
1949 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
1950 | ||
1951 | /* | |
1952 | * fixme: the total bytes and num_devices need to match or we should | |
1953 | * need a fsck | |
1954 | */ | |
1955 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
1956 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
1957 | return 0; | |
1958 | } | |
1959 | ||
7abadb64 LB |
1960 | /* helper to cleanup workers */ |
1961 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
1962 | { | |
1963 | btrfs_stop_workers(&fs_info->generic_worker); | |
1964 | btrfs_stop_workers(&fs_info->fixup_workers); | |
1965 | btrfs_stop_workers(&fs_info->delalloc_workers); | |
1966 | btrfs_stop_workers(&fs_info->workers); | |
1967 | btrfs_stop_workers(&fs_info->endio_workers); | |
1968 | btrfs_stop_workers(&fs_info->endio_meta_workers); | |
1969 | btrfs_stop_workers(&fs_info->endio_raid56_workers); | |
1970 | btrfs_stop_workers(&fs_info->rmw_workers); | |
1971 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); | |
1972 | btrfs_stop_workers(&fs_info->endio_write_workers); | |
1973 | btrfs_stop_workers(&fs_info->endio_freespace_worker); | |
1974 | btrfs_stop_workers(&fs_info->submit_workers); | |
1975 | btrfs_stop_workers(&fs_info->delayed_workers); | |
1976 | btrfs_stop_workers(&fs_info->caching_workers); | |
1977 | btrfs_stop_workers(&fs_info->readahead_workers); | |
1978 | btrfs_stop_workers(&fs_info->flush_workers); | |
1979 | } | |
1980 | ||
af31f5e5 CM |
1981 | /* helper to cleanup tree roots */ |
1982 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
1983 | { | |
1984 | free_extent_buffer(info->tree_root->node); | |
1985 | free_extent_buffer(info->tree_root->commit_root); | |
1986 | free_extent_buffer(info->dev_root->node); | |
1987 | free_extent_buffer(info->dev_root->commit_root); | |
1988 | free_extent_buffer(info->extent_root->node); | |
1989 | free_extent_buffer(info->extent_root->commit_root); | |
1990 | free_extent_buffer(info->csum_root->node); | |
1991 | free_extent_buffer(info->csum_root->commit_root); | |
bcef60f2 AJ |
1992 | if (info->quota_root) { |
1993 | free_extent_buffer(info->quota_root->node); | |
1994 | free_extent_buffer(info->quota_root->commit_root); | |
1995 | } | |
af31f5e5 CM |
1996 | |
1997 | info->tree_root->node = NULL; | |
1998 | info->tree_root->commit_root = NULL; | |
1999 | info->dev_root->node = NULL; | |
2000 | info->dev_root->commit_root = NULL; | |
2001 | info->extent_root->node = NULL; | |
2002 | info->extent_root->commit_root = NULL; | |
2003 | info->csum_root->node = NULL; | |
2004 | info->csum_root->commit_root = NULL; | |
bcef60f2 AJ |
2005 | if (info->quota_root) { |
2006 | info->quota_root->node = NULL; | |
2007 | info->quota_root->commit_root = NULL; | |
2008 | } | |
af31f5e5 CM |
2009 | |
2010 | if (chunk_root) { | |
2011 | free_extent_buffer(info->chunk_root->node); | |
2012 | free_extent_buffer(info->chunk_root->commit_root); | |
2013 | info->chunk_root->node = NULL; | |
2014 | info->chunk_root->commit_root = NULL; | |
2015 | } | |
2016 | } | |
2017 | ||
171f6537 JB |
2018 | static void del_fs_roots(struct btrfs_fs_info *fs_info) |
2019 | { | |
2020 | int ret; | |
2021 | struct btrfs_root *gang[8]; | |
2022 | int i; | |
2023 | ||
2024 | while (!list_empty(&fs_info->dead_roots)) { | |
2025 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2026 | struct btrfs_root, root_list); | |
2027 | list_del(&gang[0]->root_list); | |
2028 | ||
2029 | if (gang[0]->in_radix) { | |
2030 | btrfs_free_fs_root(fs_info, gang[0]); | |
2031 | } else { | |
2032 | free_extent_buffer(gang[0]->node); | |
2033 | free_extent_buffer(gang[0]->commit_root); | |
2034 | kfree(gang[0]); | |
2035 | } | |
2036 | } | |
2037 | ||
2038 | while (1) { | |
2039 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2040 | (void **)gang, 0, | |
2041 | ARRAY_SIZE(gang)); | |
2042 | if (!ret) | |
2043 | break; | |
2044 | for (i = 0; i < ret; i++) | |
2045 | btrfs_free_fs_root(fs_info, gang[i]); | |
2046 | } | |
2047 | } | |
af31f5e5 | 2048 | |
ad2b2c80 AV |
2049 | int open_ctree(struct super_block *sb, |
2050 | struct btrfs_fs_devices *fs_devices, | |
2051 | char *options) | |
2e635a27 | 2052 | { |
db94535d CM |
2053 | u32 sectorsize; |
2054 | u32 nodesize; | |
2055 | u32 leafsize; | |
2056 | u32 blocksize; | |
87ee04eb | 2057 | u32 stripesize; |
84234f3a | 2058 | u64 generation; |
f2b636e8 | 2059 | u64 features; |
3de4586c | 2060 | struct btrfs_key location; |
a061fc8d | 2061 | struct buffer_head *bh; |
4d34b278 | 2062 | struct btrfs_super_block *disk_super; |
815745cf | 2063 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
f84a8bd6 | 2064 | struct btrfs_root *tree_root; |
4d34b278 ID |
2065 | struct btrfs_root *extent_root; |
2066 | struct btrfs_root *csum_root; | |
2067 | struct btrfs_root *chunk_root; | |
2068 | struct btrfs_root *dev_root; | |
bcef60f2 | 2069 | struct btrfs_root *quota_root; |
e02119d5 | 2070 | struct btrfs_root *log_tree_root; |
eb60ceac | 2071 | int ret; |
e58ca020 | 2072 | int err = -EINVAL; |
af31f5e5 CM |
2073 | int num_backups_tried = 0; |
2074 | int backup_index = 0; | |
4543df7e | 2075 | |
f84a8bd6 | 2076 | tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info); |
6f07e42e AV |
2077 | extent_root = fs_info->extent_root = btrfs_alloc_root(fs_info); |
2078 | csum_root = fs_info->csum_root = btrfs_alloc_root(fs_info); | |
2079 | chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info); | |
2080 | dev_root = fs_info->dev_root = btrfs_alloc_root(fs_info); | |
bcef60f2 | 2081 | quota_root = fs_info->quota_root = btrfs_alloc_root(fs_info); |
8790d502 | 2082 | |
f84a8bd6 | 2083 | if (!tree_root || !extent_root || !csum_root || |
bcef60f2 | 2084 | !chunk_root || !dev_root || !quota_root) { |
39279cc3 CM |
2085 | err = -ENOMEM; |
2086 | goto fail; | |
2087 | } | |
76dda93c YZ |
2088 | |
2089 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
2090 | if (ret) { | |
2091 | err = ret; | |
2092 | goto fail; | |
2093 | } | |
2094 | ||
2095 | ret = setup_bdi(fs_info, &fs_info->bdi); | |
2096 | if (ret) { | |
2097 | err = ret; | |
2098 | goto fail_srcu; | |
2099 | } | |
2100 | ||
e2d84521 MX |
2101 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0); |
2102 | if (ret) { | |
2103 | err = ret; | |
2104 | goto fail_bdi; | |
2105 | } | |
2106 | fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE * | |
2107 | (1 + ilog2(nr_cpu_ids)); | |
2108 | ||
963d678b MX |
2109 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0); |
2110 | if (ret) { | |
2111 | err = ret; | |
2112 | goto fail_dirty_metadata_bytes; | |
2113 | } | |
2114 | ||
76dda93c YZ |
2115 | fs_info->btree_inode = new_inode(sb); |
2116 | if (!fs_info->btree_inode) { | |
2117 | err = -ENOMEM; | |
963d678b | 2118 | goto fail_delalloc_bytes; |
76dda93c YZ |
2119 | } |
2120 | ||
a6591715 | 2121 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
1561deda | 2122 | |
76dda93c | 2123 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
8fd17795 | 2124 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2125 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2126 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
ea8c2819 | 2127 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
11833d66 | 2128 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
1832a6d5 | 2129 | spin_lock_init(&fs_info->delalloc_lock); |
a4abeea4 | 2130 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 2131 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2132 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2133 | spin_lock_init(&fs_info->defrag_inodes_lock); |
2bf64758 | 2134 | spin_lock_init(&fs_info->free_chunk_lock); |
f29021b2 | 2135 | spin_lock_init(&fs_info->tree_mod_seq_lock); |
ceda0864 | 2136 | spin_lock_init(&fs_info->super_lock); |
f29021b2 | 2137 | rwlock_init(&fs_info->tree_mod_log_lock); |
7585717f | 2138 | mutex_init(&fs_info->reloc_mutex); |
de98ced9 | 2139 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2140 | |
58176a96 | 2141 | init_completion(&fs_info->kobj_unregister); |
0b86a832 | 2142 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2143 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2144 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
0b86a832 | 2145 | btrfs_mapping_init(&fs_info->mapping_tree); |
66d8f3dd MX |
2146 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2147 | BTRFS_BLOCK_RSV_GLOBAL); | |
2148 | btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, | |
2149 | BTRFS_BLOCK_RSV_DELALLOC); | |
2150 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); | |
2151 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2152 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2153 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2154 | BTRFS_BLOCK_RSV_DELOPS); | |
cb03c743 | 2155 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 2156 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 2157 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 2158 | atomic_set(&fs_info->nr_async_bios, 0); |
4cb5300b | 2159 | atomic_set(&fs_info->defrag_running, 0); |
f29021b2 | 2160 | atomic_set(&fs_info->tree_mod_seq, 0); |
e20d96d6 | 2161 | fs_info->sb = sb; |
6f568d35 | 2162 | fs_info->max_inline = 8192 * 1024; |
9ed74f2d | 2163 | fs_info->metadata_ratio = 0; |
4cb5300b | 2164 | fs_info->defrag_inodes = RB_ROOT; |
a4abeea4 | 2165 | fs_info->trans_no_join = 0; |
2bf64758 | 2166 | fs_info->free_chunk_space = 0; |
f29021b2 | 2167 | fs_info->tree_mod_log = RB_ROOT; |
c8b97818 | 2168 | |
90519d66 AJ |
2169 | /* readahead state */ |
2170 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); | |
2171 | spin_lock_init(&fs_info->reada_lock); | |
c8b97818 | 2172 | |
b34b086c CM |
2173 | fs_info->thread_pool_size = min_t(unsigned long, |
2174 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2175 | |
3eaa2885 CM |
2176 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
2177 | spin_lock_init(&fs_info->ordered_extent_lock); | |
16cdcec7 MX |
2178 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
2179 | GFP_NOFS); | |
2180 | if (!fs_info->delayed_root) { | |
2181 | err = -ENOMEM; | |
2182 | goto fail_iput; | |
2183 | } | |
2184 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 2185 | |
a2de733c AJ |
2186 | mutex_init(&fs_info->scrub_lock); |
2187 | atomic_set(&fs_info->scrubs_running, 0); | |
2188 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2189 | atomic_set(&fs_info->scrubs_paused, 0); | |
2190 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2191 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
2192 | init_rwsem(&fs_info->scrub_super_lock); | |
2193 | fs_info->scrub_workers_refcnt = 0; | |
21adbd5c SB |
2194 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2195 | fs_info->check_integrity_print_mask = 0; | |
2196 | #endif | |
a2de733c | 2197 | |
c9e9f97b ID |
2198 | spin_lock_init(&fs_info->balance_lock); |
2199 | mutex_init(&fs_info->balance_mutex); | |
837d5b6e ID |
2200 | atomic_set(&fs_info->balance_running, 0); |
2201 | atomic_set(&fs_info->balance_pause_req, 0); | |
a7e99c69 | 2202 | atomic_set(&fs_info->balance_cancel_req, 0); |
c9e9f97b | 2203 | fs_info->balance_ctl = NULL; |
837d5b6e | 2204 | init_waitqueue_head(&fs_info->balance_wait_q); |
a2de733c | 2205 | |
a061fc8d CM |
2206 | sb->s_blocksize = 4096; |
2207 | sb->s_blocksize_bits = blksize_bits(4096); | |
32a88aa1 | 2208 | sb->s_bdi = &fs_info->bdi; |
a061fc8d | 2209 | |
76dda93c | 2210 | fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; |
bfe86848 | 2211 | set_nlink(fs_info->btree_inode, 1); |
0afbaf8c CM |
2212 | /* |
2213 | * we set the i_size on the btree inode to the max possible int. | |
2214 | * the real end of the address space is determined by all of | |
2215 | * the devices in the system | |
2216 | */ | |
2217 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 2218 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
2219 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
2220 | ||
5d4f98a2 | 2221 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); |
d1310b2e | 2222 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
f993c883 | 2223 | fs_info->btree_inode->i_mapping); |
0b32f4bb | 2224 | BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; |
a8067e02 | 2225 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); |
d1310b2e CM |
2226 | |
2227 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 2228 | |
76dda93c YZ |
2229 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
2230 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
2231 | sizeof(struct btrfs_key)); | |
72ac3c0d JB |
2232 | set_bit(BTRFS_INODE_DUMMY, |
2233 | &BTRFS_I(fs_info->btree_inode)->runtime_flags); | |
c65ddb52 | 2234 | insert_inode_hash(fs_info->btree_inode); |
76dda93c | 2235 | |
0f9dd46c | 2236 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2237 | fs_info->block_group_cache_tree = RB_ROOT; |
a1897fdd | 2238 | fs_info->first_logical_byte = (u64)-1; |
0f9dd46c | 2239 | |
11833d66 | 2240 | extent_io_tree_init(&fs_info->freed_extents[0], |
f993c883 | 2241 | fs_info->btree_inode->i_mapping); |
11833d66 | 2242 | extent_io_tree_init(&fs_info->freed_extents[1], |
f993c883 | 2243 | fs_info->btree_inode->i_mapping); |
11833d66 | 2244 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
e66f709b | 2245 | fs_info->do_barriers = 1; |
e18e4809 | 2246 | |
39279cc3 | 2247 | |
5a3f23d5 | 2248 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2249 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2250 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2251 | mutex_init(&fs_info->transaction_kthread_mutex); |
2252 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 2253 | mutex_init(&fs_info->volume_mutex); |
276e680d | 2254 | init_rwsem(&fs_info->extent_commit_sem); |
c71bf099 | 2255 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2256 | init_rwsem(&fs_info->subvol_sem); |
e922e087 SB |
2257 | fs_info->dev_replace.lock_owner = 0; |
2258 | atomic_set(&fs_info->dev_replace.nesting_level, 0); | |
2259 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); | |
2260 | mutex_init(&fs_info->dev_replace.lock_management_lock); | |
2261 | mutex_init(&fs_info->dev_replace.lock); | |
fa9c0d79 | 2262 | |
416ac51d | 2263 | spin_lock_init(&fs_info->qgroup_lock); |
f2f6ed3d | 2264 | mutex_init(&fs_info->qgroup_ioctl_lock); |
416ac51d AJ |
2265 | fs_info->qgroup_tree = RB_ROOT; |
2266 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); | |
2267 | fs_info->qgroup_seq = 1; | |
2268 | fs_info->quota_enabled = 0; | |
2269 | fs_info->pending_quota_state = 0; | |
2270 | ||
fa9c0d79 CM |
2271 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2272 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2273 | ||
e6dcd2dc | 2274 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2275 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2276 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2277 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2278 | |
53b381b3 DW |
2279 | ret = btrfs_alloc_stripe_hash_table(fs_info); |
2280 | if (ret) { | |
83c8266a | 2281 | err = ret; |
53b381b3 DW |
2282 | goto fail_alloc; |
2283 | } | |
2284 | ||
0b86a832 | 2285 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 2286 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2287 | |
3c4bb26b | 2288 | invalidate_bdev(fs_devices->latest_bdev); |
a512bbf8 | 2289 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
20b45077 DY |
2290 | if (!bh) { |
2291 | err = -EINVAL; | |
16cdcec7 | 2292 | goto fail_alloc; |
20b45077 | 2293 | } |
39279cc3 | 2294 | |
6c41761f DS |
2295 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2296 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2297 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2298 | brelse(bh); |
5f39d397 | 2299 | |
6c41761f | 2300 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2301 | |
6c41761f | 2302 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2303 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2304 | goto fail_alloc; |
0f7d52f4 | 2305 | |
acce952b | 2306 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
2307 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
2308 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 2309 | |
fcd1f065 DS |
2310 | ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); |
2311 | if (ret) { | |
2312 | printk(KERN_ERR "btrfs: superblock contains fatal errors\n"); | |
2313 | err = ret; | |
2314 | goto fail_alloc; | |
2315 | } | |
acce952b | 2316 | |
af31f5e5 CM |
2317 | /* |
2318 | * run through our array of backup supers and setup | |
2319 | * our ring pointer to the oldest one | |
2320 | */ | |
2321 | generation = btrfs_super_generation(disk_super); | |
2322 | find_oldest_super_backup(fs_info, generation); | |
2323 | ||
75e7cb7f LB |
2324 | /* |
2325 | * In the long term, we'll store the compression type in the super | |
2326 | * block, and it'll be used for per file compression control. | |
2327 | */ | |
2328 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2329 | ||
2b82032c YZ |
2330 | ret = btrfs_parse_options(tree_root, options); |
2331 | if (ret) { | |
2332 | err = ret; | |
16cdcec7 | 2333 | goto fail_alloc; |
2b82032c | 2334 | } |
dfe25020 | 2335 | |
f2b636e8 JB |
2336 | features = btrfs_super_incompat_flags(disk_super) & |
2337 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2338 | if (features) { | |
2339 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
2340 | "unsupported optional features (%Lx).\n", | |
21380931 | 2341 | (unsigned long long)features); |
f2b636e8 | 2342 | err = -EINVAL; |
16cdcec7 | 2343 | goto fail_alloc; |
f2b636e8 JB |
2344 | } |
2345 | ||
727011e0 CM |
2346 | if (btrfs_super_leafsize(disk_super) != |
2347 | btrfs_super_nodesize(disk_super)) { | |
2348 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2349 | "blocksizes don't match. node %d leaf %d\n", | |
2350 | btrfs_super_nodesize(disk_super), | |
2351 | btrfs_super_leafsize(disk_super)); | |
2352 | err = -EINVAL; | |
2353 | goto fail_alloc; | |
2354 | } | |
2355 | if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2356 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2357 | "blocksize (%d) was too large\n", | |
2358 | btrfs_super_leafsize(disk_super)); | |
2359 | err = -EINVAL; | |
2360 | goto fail_alloc; | |
2361 | } | |
2362 | ||
5d4f98a2 | 2363 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 2364 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
69e380d1 | 2365 | if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 2366 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
727011e0 | 2367 | |
3173a18f JB |
2368 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
2369 | printk(KERN_ERR "btrfs: has skinny extents\n"); | |
2370 | ||
727011e0 CM |
2371 | /* |
2372 | * flag our filesystem as having big metadata blocks if | |
2373 | * they are bigger than the page size | |
2374 | */ | |
2375 | if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) { | |
2376 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) | |
2377 | printk(KERN_INFO "btrfs flagging fs with big metadata feature\n"); | |
2378 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
2379 | } | |
2380 | ||
bc3f116f CM |
2381 | nodesize = btrfs_super_nodesize(disk_super); |
2382 | leafsize = btrfs_super_leafsize(disk_super); | |
2383 | sectorsize = btrfs_super_sectorsize(disk_super); | |
2384 | stripesize = btrfs_super_stripesize(disk_super); | |
e2d84521 | 2385 | fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids)); |
963d678b | 2386 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); |
bc3f116f CM |
2387 | |
2388 | /* | |
2389 | * mixed block groups end up with duplicate but slightly offset | |
2390 | * extent buffers for the same range. It leads to corruptions | |
2391 | */ | |
2392 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
2393 | (sectorsize != leafsize)) { | |
2394 | printk(KERN_WARNING "btrfs: unequal leaf/node/sector sizes " | |
2395 | "are not allowed for mixed block groups on %s\n", | |
2396 | sb->s_id); | |
2397 | goto fail_alloc; | |
2398 | } | |
2399 | ||
ceda0864 MX |
2400 | /* |
2401 | * Needn't use the lock because there is no other task which will | |
2402 | * update the flag. | |
2403 | */ | |
a6fa6fae | 2404 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 2405 | |
f2b636e8 JB |
2406 | features = btrfs_super_compat_ro_flags(disk_super) & |
2407 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
2408 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
2409 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
2410 | "unsupported option features (%Lx).\n", | |
21380931 | 2411 | (unsigned long long)features); |
f2b636e8 | 2412 | err = -EINVAL; |
16cdcec7 | 2413 | goto fail_alloc; |
f2b636e8 | 2414 | } |
61d92c32 CM |
2415 | |
2416 | btrfs_init_workers(&fs_info->generic_worker, | |
2417 | "genwork", 1, NULL); | |
2418 | ||
5443be45 | 2419 | btrfs_init_workers(&fs_info->workers, "worker", |
61d92c32 CM |
2420 | fs_info->thread_pool_size, |
2421 | &fs_info->generic_worker); | |
c8b97818 | 2422 | |
771ed689 | 2423 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
61d92c32 CM |
2424 | fs_info->thread_pool_size, |
2425 | &fs_info->generic_worker); | |
771ed689 | 2426 | |
8ccf6f19 MX |
2427 | btrfs_init_workers(&fs_info->flush_workers, "flush_delalloc", |
2428 | fs_info->thread_pool_size, | |
2429 | &fs_info->generic_worker); | |
2430 | ||
5443be45 | 2431 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 | 2432 | min_t(u64, fs_devices->num_devices, |
61d92c32 CM |
2433 | fs_info->thread_pool_size), |
2434 | &fs_info->generic_worker); | |
61b49440 | 2435 | |
bab39bf9 JB |
2436 | btrfs_init_workers(&fs_info->caching_workers, "cache", |
2437 | 2, &fs_info->generic_worker); | |
2438 | ||
61b49440 CM |
2439 | /* a higher idle thresh on the submit workers makes it much more |
2440 | * likely that bios will be send down in a sane order to the | |
2441 | * devices | |
2442 | */ | |
2443 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 2444 | |
771ed689 | 2445 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 2446 | fs_info->workers.ordered = 1; |
61b49440 | 2447 | |
771ed689 CM |
2448 | fs_info->delalloc_workers.idle_thresh = 2; |
2449 | fs_info->delalloc_workers.ordered = 1; | |
2450 | ||
61d92c32 CM |
2451 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1, |
2452 | &fs_info->generic_worker); | |
5443be45 | 2453 | btrfs_init_workers(&fs_info->endio_workers, "endio", |
61d92c32 CM |
2454 | fs_info->thread_pool_size, |
2455 | &fs_info->generic_worker); | |
d20f7043 | 2456 | btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", |
61d92c32 CM |
2457 | fs_info->thread_pool_size, |
2458 | &fs_info->generic_worker); | |
cad321ad | 2459 | btrfs_init_workers(&fs_info->endio_meta_write_workers, |
61d92c32 CM |
2460 | "endio-meta-write", fs_info->thread_pool_size, |
2461 | &fs_info->generic_worker); | |
53b381b3 DW |
2462 | btrfs_init_workers(&fs_info->endio_raid56_workers, |
2463 | "endio-raid56", fs_info->thread_pool_size, | |
2464 | &fs_info->generic_worker); | |
2465 | btrfs_init_workers(&fs_info->rmw_workers, | |
2466 | "rmw", fs_info->thread_pool_size, | |
2467 | &fs_info->generic_worker); | |
5443be45 | 2468 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", |
61d92c32 CM |
2469 | fs_info->thread_pool_size, |
2470 | &fs_info->generic_worker); | |
0cb59c99 JB |
2471 | btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write", |
2472 | 1, &fs_info->generic_worker); | |
16cdcec7 MX |
2473 | btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta", |
2474 | fs_info->thread_pool_size, | |
2475 | &fs_info->generic_worker); | |
90519d66 AJ |
2476 | btrfs_init_workers(&fs_info->readahead_workers, "readahead", |
2477 | fs_info->thread_pool_size, | |
2478 | &fs_info->generic_worker); | |
61b49440 CM |
2479 | |
2480 | /* | |
2481 | * endios are largely parallel and should have a very | |
2482 | * low idle thresh | |
2483 | */ | |
2484 | fs_info->endio_workers.idle_thresh = 4; | |
b51912c9 | 2485 | fs_info->endio_meta_workers.idle_thresh = 4; |
53b381b3 DW |
2486 | fs_info->endio_raid56_workers.idle_thresh = 4; |
2487 | fs_info->rmw_workers.idle_thresh = 2; | |
b51912c9 | 2488 | |
9042846b CM |
2489 | fs_info->endio_write_workers.idle_thresh = 2; |
2490 | fs_info->endio_meta_write_workers.idle_thresh = 2; | |
90519d66 | 2491 | fs_info->readahead_workers.idle_thresh = 2; |
9042846b | 2492 | |
0dc3b84a JB |
2493 | /* |
2494 | * btrfs_start_workers can really only fail because of ENOMEM so just | |
2495 | * return -ENOMEM if any of these fail. | |
2496 | */ | |
2497 | ret = btrfs_start_workers(&fs_info->workers); | |
2498 | ret |= btrfs_start_workers(&fs_info->generic_worker); | |
2499 | ret |= btrfs_start_workers(&fs_info->submit_workers); | |
2500 | ret |= btrfs_start_workers(&fs_info->delalloc_workers); | |
2501 | ret |= btrfs_start_workers(&fs_info->fixup_workers); | |
2502 | ret |= btrfs_start_workers(&fs_info->endio_workers); | |
2503 | ret |= btrfs_start_workers(&fs_info->endio_meta_workers); | |
53b381b3 DW |
2504 | ret |= btrfs_start_workers(&fs_info->rmw_workers); |
2505 | ret |= btrfs_start_workers(&fs_info->endio_raid56_workers); | |
0dc3b84a JB |
2506 | ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers); |
2507 | ret |= btrfs_start_workers(&fs_info->endio_write_workers); | |
2508 | ret |= btrfs_start_workers(&fs_info->endio_freespace_worker); | |
2509 | ret |= btrfs_start_workers(&fs_info->delayed_workers); | |
2510 | ret |= btrfs_start_workers(&fs_info->caching_workers); | |
2511 | ret |= btrfs_start_workers(&fs_info->readahead_workers); | |
8ccf6f19 | 2512 | ret |= btrfs_start_workers(&fs_info->flush_workers); |
0dc3b84a | 2513 | if (ret) { |
fed425c7 | 2514 | err = -ENOMEM; |
0dc3b84a JB |
2515 | goto fail_sb_buffer; |
2516 | } | |
4543df7e | 2517 | |
4575c9cc | 2518 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
2519 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
2520 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 2521 | |
db94535d CM |
2522 | tree_root->nodesize = nodesize; |
2523 | tree_root->leafsize = leafsize; | |
2524 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 2525 | tree_root->stripesize = stripesize; |
a061fc8d CM |
2526 | |
2527 | sb->s_blocksize = sectorsize; | |
2528 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 2529 | |
cdb4c574 | 2530 | if (disk_super->magic != cpu_to_le64(BTRFS_MAGIC)) { |
d397712b | 2531 | printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
2532 | goto fail_sb_buffer; |
2533 | } | |
19c00ddc | 2534 | |
8d082fb7 LB |
2535 | if (sectorsize != PAGE_SIZE) { |
2536 | printk(KERN_WARNING "btrfs: Incompatible sector size(%lu) " | |
2537 | "found on %s\n", (unsigned long)sectorsize, sb->s_id); | |
941b2ddf KM |
2538 | goto fail_sb_buffer; |
2539 | } | |
2540 | ||
925baedd | 2541 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 2542 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 2543 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2544 | if (ret) { |
d397712b CM |
2545 | printk(KERN_WARNING "btrfs: failed to read the system " |
2546 | "array on %s\n", sb->s_id); | |
5d4f98a2 | 2547 | goto fail_sb_buffer; |
84eed90f | 2548 | } |
0b86a832 CM |
2549 | |
2550 | blocksize = btrfs_level_size(tree_root, | |
2551 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 2552 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
2553 | |
2554 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2555 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
2556 | ||
2557 | chunk_root->node = read_tree_block(chunk_root, | |
2558 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 2559 | blocksize, generation); |
416bc658 JB |
2560 | if (!chunk_root->node || |
2561 | !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { | |
83121942 DW |
2562 | printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n", |
2563 | sb->s_id); | |
af31f5e5 | 2564 | goto fail_tree_roots; |
83121942 | 2565 | } |
5d4f98a2 YZ |
2566 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
2567 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 2568 | |
e17cade2 | 2569 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
d397712b CM |
2570 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), |
2571 | BTRFS_UUID_SIZE); | |
e17cade2 | 2572 | |
0b86a832 | 2573 | ret = btrfs_read_chunk_tree(chunk_root); |
2b82032c | 2574 | if (ret) { |
d397712b CM |
2575 | printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", |
2576 | sb->s_id); | |
af31f5e5 | 2577 | goto fail_tree_roots; |
2b82032c | 2578 | } |
0b86a832 | 2579 | |
8dabb742 SB |
2580 | /* |
2581 | * keep the device that is marked to be the target device for the | |
2582 | * dev_replace procedure | |
2583 | */ | |
2584 | btrfs_close_extra_devices(fs_info, fs_devices, 0); | |
dfe25020 | 2585 | |
a6b0d5c8 CM |
2586 | if (!fs_devices->latest_bdev) { |
2587 | printk(KERN_CRIT "btrfs: failed to read devices on %s\n", | |
2588 | sb->s_id); | |
2589 | goto fail_tree_roots; | |
2590 | } | |
2591 | ||
af31f5e5 | 2592 | retry_root_backup: |
db94535d CM |
2593 | blocksize = btrfs_level_size(tree_root, |
2594 | btrfs_super_root_level(disk_super)); | |
84234f3a | 2595 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 2596 | |
e20d96d6 | 2597 | tree_root->node = read_tree_block(tree_root, |
db94535d | 2598 | btrfs_super_root(disk_super), |
84234f3a | 2599 | blocksize, generation); |
af31f5e5 CM |
2600 | if (!tree_root->node || |
2601 | !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { | |
83121942 DW |
2602 | printk(KERN_WARNING "btrfs: failed to read tree root on %s\n", |
2603 | sb->s_id); | |
af31f5e5 CM |
2604 | |
2605 | goto recovery_tree_root; | |
83121942 | 2606 | } |
af31f5e5 | 2607 | |
5d4f98a2 YZ |
2608 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
2609 | tree_root->commit_root = btrfs_root_node(tree_root); | |
db94535d CM |
2610 | |
2611 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 2612 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 2613 | if (ret) |
af31f5e5 | 2614 | goto recovery_tree_root; |
0b86a832 CM |
2615 | extent_root->track_dirty = 1; |
2616 | ||
2617 | ret = find_and_setup_root(tree_root, fs_info, | |
2618 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
0b86a832 | 2619 | if (ret) |
af31f5e5 | 2620 | goto recovery_tree_root; |
5d4f98a2 | 2621 | dev_root->track_dirty = 1; |
3768f368 | 2622 | |
d20f7043 CM |
2623 | ret = find_and_setup_root(tree_root, fs_info, |
2624 | BTRFS_CSUM_TREE_OBJECTID, csum_root); | |
2625 | if (ret) | |
af31f5e5 | 2626 | goto recovery_tree_root; |
d20f7043 CM |
2627 | csum_root->track_dirty = 1; |
2628 | ||
bcef60f2 AJ |
2629 | ret = find_and_setup_root(tree_root, fs_info, |
2630 | BTRFS_QUOTA_TREE_OBJECTID, quota_root); | |
2631 | if (ret) { | |
2632 | kfree(quota_root); | |
2633 | quota_root = fs_info->quota_root = NULL; | |
2634 | } else { | |
2635 | quota_root->track_dirty = 1; | |
2636 | fs_info->quota_enabled = 1; | |
2637 | fs_info->pending_quota_state = 1; | |
2638 | } | |
2639 | ||
8929ecfa YZ |
2640 | fs_info->generation = generation; |
2641 | fs_info->last_trans_committed = generation; | |
8929ecfa | 2642 | |
68310a5e ID |
2643 | ret = btrfs_recover_balance(fs_info); |
2644 | if (ret) { | |
2645 | printk(KERN_WARNING "btrfs: failed to recover balance\n"); | |
2646 | goto fail_block_groups; | |
2647 | } | |
2648 | ||
733f4fbb SB |
2649 | ret = btrfs_init_dev_stats(fs_info); |
2650 | if (ret) { | |
2651 | printk(KERN_ERR "btrfs: failed to init dev_stats: %d\n", | |
2652 | ret); | |
2653 | goto fail_block_groups; | |
2654 | } | |
2655 | ||
8dabb742 SB |
2656 | ret = btrfs_init_dev_replace(fs_info); |
2657 | if (ret) { | |
2658 | pr_err("btrfs: failed to init dev_replace: %d\n", ret); | |
2659 | goto fail_block_groups; | |
2660 | } | |
2661 | ||
2662 | btrfs_close_extra_devices(fs_info, fs_devices, 1); | |
2663 | ||
c59021f8 | 2664 | ret = btrfs_init_space_info(fs_info); |
2665 | if (ret) { | |
2666 | printk(KERN_ERR "Failed to initial space info: %d\n", ret); | |
2667 | goto fail_block_groups; | |
2668 | } | |
2669 | ||
1b1d1f66 JB |
2670 | ret = btrfs_read_block_groups(extent_root); |
2671 | if (ret) { | |
2672 | printk(KERN_ERR "Failed to read block groups: %d\n", ret); | |
2673 | goto fail_block_groups; | |
2674 | } | |
5af3e8cc SB |
2675 | fs_info->num_tolerated_disk_barrier_failures = |
2676 | btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); | |
292fd7fc SB |
2677 | if (fs_info->fs_devices->missing_devices > |
2678 | fs_info->num_tolerated_disk_barrier_failures && | |
2679 | !(sb->s_flags & MS_RDONLY)) { | |
2680 | printk(KERN_WARNING | |
2681 | "Btrfs: too many missing devices, writeable mount is not allowed\n"); | |
2682 | goto fail_block_groups; | |
2683 | } | |
9078a3e1 | 2684 | |
a74a4b97 CM |
2685 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
2686 | "btrfs-cleaner"); | |
57506d50 | 2687 | if (IS_ERR(fs_info->cleaner_kthread)) |
1b1d1f66 | 2688 | goto fail_block_groups; |
a74a4b97 CM |
2689 | |
2690 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
2691 | tree_root, | |
2692 | "btrfs-transaction"); | |
57506d50 | 2693 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 2694 | goto fail_cleaner; |
a74a4b97 | 2695 | |
c289811c CM |
2696 | if (!btrfs_test_opt(tree_root, SSD) && |
2697 | !btrfs_test_opt(tree_root, NOSSD) && | |
2698 | !fs_info->fs_devices->rotating) { | |
2699 | printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD " | |
2700 | "mode\n"); | |
2701 | btrfs_set_opt(fs_info->mount_opt, SSD); | |
2702 | } | |
2703 | ||
21adbd5c SB |
2704 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2705 | if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { | |
2706 | ret = btrfsic_mount(tree_root, fs_devices, | |
2707 | btrfs_test_opt(tree_root, | |
2708 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? | |
2709 | 1 : 0, | |
2710 | fs_info->check_integrity_print_mask); | |
2711 | if (ret) | |
2712 | printk(KERN_WARNING "btrfs: failed to initialize" | |
2713 | " integrity check module %s\n", sb->s_id); | |
2714 | } | |
2715 | #endif | |
bcef60f2 AJ |
2716 | ret = btrfs_read_qgroup_config(fs_info); |
2717 | if (ret) | |
2718 | goto fail_trans_kthread; | |
21adbd5c | 2719 | |
acce952b | 2720 | /* do not make disk changes in broken FS */ |
68ce9682 | 2721 | if (btrfs_super_log_root(disk_super) != 0) { |
e02119d5 CM |
2722 | u64 bytenr = btrfs_super_log_root(disk_super); |
2723 | ||
7c2ca468 | 2724 | if (fs_devices->rw_devices == 0) { |
d397712b CM |
2725 | printk(KERN_WARNING "Btrfs log replay required " |
2726 | "on RO media\n"); | |
7c2ca468 | 2727 | err = -EIO; |
bcef60f2 | 2728 | goto fail_qgroup; |
7c2ca468 | 2729 | } |
e02119d5 CM |
2730 | blocksize = |
2731 | btrfs_level_size(tree_root, | |
2732 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 2733 | |
6f07e42e | 2734 | log_tree_root = btrfs_alloc_root(fs_info); |
676e4c86 DC |
2735 | if (!log_tree_root) { |
2736 | err = -ENOMEM; | |
bcef60f2 | 2737 | goto fail_qgroup; |
676e4c86 | 2738 | } |
e02119d5 CM |
2739 | |
2740 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2741 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
2742 | ||
2743 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
2744 | blocksize, |
2745 | generation + 1); | |
416bc658 JB |
2746 | if (!log_tree_root->node || |
2747 | !extent_buffer_uptodate(log_tree_root->node)) { | |
2748 | printk(KERN_ERR "btrfs: failed to read log tree\n"); | |
2749 | free_extent_buffer(log_tree_root->node); | |
2750 | kfree(log_tree_root); | |
2751 | goto fail_trans_kthread; | |
2752 | } | |
79787eaa | 2753 | /* returns with log_tree_root freed on success */ |
e02119d5 | 2754 | ret = btrfs_recover_log_trees(log_tree_root); |
79787eaa JM |
2755 | if (ret) { |
2756 | btrfs_error(tree_root->fs_info, ret, | |
2757 | "Failed to recover log tree"); | |
2758 | free_extent_buffer(log_tree_root->node); | |
2759 | kfree(log_tree_root); | |
2760 | goto fail_trans_kthread; | |
2761 | } | |
e556ce2c YZ |
2762 | |
2763 | if (sb->s_flags & MS_RDONLY) { | |
79787eaa JM |
2764 | ret = btrfs_commit_super(tree_root); |
2765 | if (ret) | |
2766 | goto fail_trans_kthread; | |
e556ce2c | 2767 | } |
e02119d5 | 2768 | } |
1a40e23b | 2769 | |
76dda93c | 2770 | ret = btrfs_find_orphan_roots(tree_root); |
79787eaa JM |
2771 | if (ret) |
2772 | goto fail_trans_kthread; | |
76dda93c | 2773 | |
7c2ca468 | 2774 | if (!(sb->s_flags & MS_RDONLY)) { |
d68fc57b | 2775 | ret = btrfs_cleanup_fs_roots(fs_info); |
44c44af2 ID |
2776 | if (ret) |
2777 | goto fail_trans_kthread; | |
d68fc57b | 2778 | |
5d4f98a2 | 2779 | ret = btrfs_recover_relocation(tree_root); |
d7ce5843 MX |
2780 | if (ret < 0) { |
2781 | printk(KERN_WARNING | |
2782 | "btrfs: failed to recover relocation\n"); | |
2783 | err = -EINVAL; | |
bcef60f2 | 2784 | goto fail_qgroup; |
d7ce5843 | 2785 | } |
7c2ca468 | 2786 | } |
1a40e23b | 2787 | |
3de4586c CM |
2788 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
2789 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2790 | location.offset = (u64)-1; | |
2791 | ||
3de4586c CM |
2792 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
2793 | if (!fs_info->fs_root) | |
bcef60f2 | 2794 | goto fail_qgroup; |
3140c9a3 DC |
2795 | if (IS_ERR(fs_info->fs_root)) { |
2796 | err = PTR_ERR(fs_info->fs_root); | |
bcef60f2 | 2797 | goto fail_qgroup; |
3140c9a3 | 2798 | } |
c289811c | 2799 | |
2b6ba629 ID |
2800 | if (sb->s_flags & MS_RDONLY) |
2801 | return 0; | |
59641015 | 2802 | |
2b6ba629 ID |
2803 | down_read(&fs_info->cleanup_work_sem); |
2804 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
2805 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
e3acc2a6 | 2806 | up_read(&fs_info->cleanup_work_sem); |
2b6ba629 ID |
2807 | close_ctree(tree_root); |
2808 | return ret; | |
2809 | } | |
2810 | up_read(&fs_info->cleanup_work_sem); | |
59641015 | 2811 | |
2b6ba629 ID |
2812 | ret = btrfs_resume_balance_async(fs_info); |
2813 | if (ret) { | |
2814 | printk(KERN_WARNING "btrfs: failed to resume balance\n"); | |
2815 | close_ctree(tree_root); | |
2816 | return ret; | |
e3acc2a6 JB |
2817 | } |
2818 | ||
8dabb742 SB |
2819 | ret = btrfs_resume_dev_replace_async(fs_info); |
2820 | if (ret) { | |
2821 | pr_warn("btrfs: failed to resume dev_replace\n"); | |
2822 | close_ctree(tree_root); | |
2823 | return ret; | |
2824 | } | |
2825 | ||
ad2b2c80 | 2826 | return 0; |
39279cc3 | 2827 | |
bcef60f2 AJ |
2828 | fail_qgroup: |
2829 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
2830 | fail_trans_kthread: |
2831 | kthread_stop(fs_info->transaction_kthread); | |
171f6537 | 2832 | del_fs_roots(fs_info); |
54067ae9 | 2833 | btrfs_cleanup_transaction(fs_info->tree_root); |
3f157a2f | 2834 | fail_cleaner: |
a74a4b97 | 2835 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
2836 | |
2837 | /* | |
2838 | * make sure we're done with the btree inode before we stop our | |
2839 | * kthreads | |
2840 | */ | |
2841 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 2842 | |
1b1d1f66 | 2843 | fail_block_groups: |
54067ae9 | 2844 | btrfs_put_block_group_cache(fs_info); |
1b1d1f66 | 2845 | btrfs_free_block_groups(fs_info); |
af31f5e5 CM |
2846 | |
2847 | fail_tree_roots: | |
2848 | free_root_pointers(fs_info, 1); | |
2b8195bb | 2849 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 2850 | |
39279cc3 | 2851 | fail_sb_buffer: |
7abadb64 | 2852 | btrfs_stop_all_workers(fs_info); |
16cdcec7 | 2853 | fail_alloc: |
4543df7e | 2854 | fail_iput: |
586e46e2 ID |
2855 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
2856 | ||
4543df7e | 2857 | iput(fs_info->btree_inode); |
963d678b MX |
2858 | fail_delalloc_bytes: |
2859 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
e2d84521 MX |
2860 | fail_dirty_metadata_bytes: |
2861 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); | |
ad081f14 | 2862 | fail_bdi: |
7e662854 | 2863 | bdi_destroy(&fs_info->bdi); |
76dda93c YZ |
2864 | fail_srcu: |
2865 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 2866 | fail: |
53b381b3 | 2867 | btrfs_free_stripe_hash_table(fs_info); |
586e46e2 | 2868 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 2869 | return err; |
af31f5e5 CM |
2870 | |
2871 | recovery_tree_root: | |
af31f5e5 CM |
2872 | if (!btrfs_test_opt(tree_root, RECOVERY)) |
2873 | goto fail_tree_roots; | |
2874 | ||
2875 | free_root_pointers(fs_info, 0); | |
2876 | ||
2877 | /* don't use the log in recovery mode, it won't be valid */ | |
2878 | btrfs_set_super_log_root(disk_super, 0); | |
2879 | ||
2880 | /* we can't trust the free space cache either */ | |
2881 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2882 | ||
2883 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
2884 | &num_backups_tried, &backup_index); | |
2885 | if (ret == -1) | |
2886 | goto fail_block_groups; | |
2887 | goto retry_root_backup; | |
eb60ceac CM |
2888 | } |
2889 | ||
f2984462 CM |
2890 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
2891 | { | |
f2984462 CM |
2892 | if (uptodate) { |
2893 | set_buffer_uptodate(bh); | |
2894 | } else { | |
442a4f63 SB |
2895 | struct btrfs_device *device = (struct btrfs_device *) |
2896 | bh->b_private; | |
2897 | ||
606686ee JB |
2898 | printk_ratelimited_in_rcu(KERN_WARNING "lost page write due to " |
2899 | "I/O error on %s\n", | |
2900 | rcu_str_deref(device->name)); | |
1259ab75 CM |
2901 | /* note, we dont' set_buffer_write_io_error because we have |
2902 | * our own ways of dealing with the IO errors | |
2903 | */ | |
f2984462 | 2904 | clear_buffer_uptodate(bh); |
442a4f63 | 2905 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); |
f2984462 CM |
2906 | } |
2907 | unlock_buffer(bh); | |
2908 | put_bh(bh); | |
2909 | } | |
2910 | ||
a512bbf8 YZ |
2911 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
2912 | { | |
2913 | struct buffer_head *bh; | |
2914 | struct buffer_head *latest = NULL; | |
2915 | struct btrfs_super_block *super; | |
2916 | int i; | |
2917 | u64 transid = 0; | |
2918 | u64 bytenr; | |
2919 | ||
2920 | /* we would like to check all the supers, but that would make | |
2921 | * a btrfs mount succeed after a mkfs from a different FS. | |
2922 | * So, we need to add a special mount option to scan for | |
2923 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
2924 | */ | |
2925 | for (i = 0; i < 1; i++) { | |
2926 | bytenr = btrfs_sb_offset(i); | |
2927 | if (bytenr + 4096 >= i_size_read(bdev->bd_inode)) | |
2928 | break; | |
2929 | bh = __bread(bdev, bytenr / 4096, 4096); | |
2930 | if (!bh) | |
2931 | continue; | |
2932 | ||
2933 | super = (struct btrfs_super_block *)bh->b_data; | |
2934 | if (btrfs_super_bytenr(super) != bytenr || | |
cdb4c574 | 2935 | super->magic != cpu_to_le64(BTRFS_MAGIC)) { |
a512bbf8 YZ |
2936 | brelse(bh); |
2937 | continue; | |
2938 | } | |
2939 | ||
2940 | if (!latest || btrfs_super_generation(super) > transid) { | |
2941 | brelse(latest); | |
2942 | latest = bh; | |
2943 | transid = btrfs_super_generation(super); | |
2944 | } else { | |
2945 | brelse(bh); | |
2946 | } | |
2947 | } | |
2948 | return latest; | |
2949 | } | |
2950 | ||
4eedeb75 HH |
2951 | /* |
2952 | * this should be called twice, once with wait == 0 and | |
2953 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
2954 | * we write are pinned. | |
2955 | * | |
2956 | * They are released when wait == 1 is done. | |
2957 | * max_mirrors must be the same for both runs, and it indicates how | |
2958 | * many supers on this one device should be written. | |
2959 | * | |
2960 | * max_mirrors == 0 means to write them all. | |
2961 | */ | |
a512bbf8 YZ |
2962 | static int write_dev_supers(struct btrfs_device *device, |
2963 | struct btrfs_super_block *sb, | |
2964 | int do_barriers, int wait, int max_mirrors) | |
2965 | { | |
2966 | struct buffer_head *bh; | |
2967 | int i; | |
2968 | int ret; | |
2969 | int errors = 0; | |
2970 | u32 crc; | |
2971 | u64 bytenr; | |
a512bbf8 YZ |
2972 | |
2973 | if (max_mirrors == 0) | |
2974 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
2975 | ||
a512bbf8 YZ |
2976 | for (i = 0; i < max_mirrors; i++) { |
2977 | bytenr = btrfs_sb_offset(i); | |
2978 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) | |
2979 | break; | |
2980 | ||
2981 | if (wait) { | |
2982 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
2983 | BTRFS_SUPER_INFO_SIZE); | |
2984 | BUG_ON(!bh); | |
a512bbf8 | 2985 | wait_on_buffer(bh); |
4eedeb75 HH |
2986 | if (!buffer_uptodate(bh)) |
2987 | errors++; | |
2988 | ||
2989 | /* drop our reference */ | |
2990 | brelse(bh); | |
2991 | ||
2992 | /* drop the reference from the wait == 0 run */ | |
2993 | brelse(bh); | |
2994 | continue; | |
a512bbf8 YZ |
2995 | } else { |
2996 | btrfs_set_super_bytenr(sb, bytenr); | |
2997 | ||
2998 | crc = ~(u32)0; | |
b0496686 | 2999 | crc = btrfs_csum_data((char *)sb + |
a512bbf8 YZ |
3000 | BTRFS_CSUM_SIZE, crc, |
3001 | BTRFS_SUPER_INFO_SIZE - | |
3002 | BTRFS_CSUM_SIZE); | |
3003 | btrfs_csum_final(crc, sb->csum); | |
3004 | ||
4eedeb75 HH |
3005 | /* |
3006 | * one reference for us, and we leave it for the | |
3007 | * caller | |
3008 | */ | |
a512bbf8 YZ |
3009 | bh = __getblk(device->bdev, bytenr / 4096, |
3010 | BTRFS_SUPER_INFO_SIZE); | |
3011 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); | |
3012 | ||
4eedeb75 | 3013 | /* one reference for submit_bh */ |
a512bbf8 | 3014 | get_bh(bh); |
4eedeb75 HH |
3015 | |
3016 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
3017 | lock_buffer(bh); |
3018 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
442a4f63 | 3019 | bh->b_private = device; |
a512bbf8 YZ |
3020 | } |
3021 | ||
387125fc CM |
3022 | /* |
3023 | * we fua the first super. The others we allow | |
3024 | * to go down lazy. | |
3025 | */ | |
21adbd5c | 3026 | ret = btrfsic_submit_bh(WRITE_FUA, bh); |
4eedeb75 | 3027 | if (ret) |
a512bbf8 | 3028 | errors++; |
a512bbf8 YZ |
3029 | } |
3030 | return errors < i ? 0 : -1; | |
3031 | } | |
3032 | ||
387125fc CM |
3033 | /* |
3034 | * endio for the write_dev_flush, this will wake anyone waiting | |
3035 | * for the barrier when it is done | |
3036 | */ | |
3037 | static void btrfs_end_empty_barrier(struct bio *bio, int err) | |
3038 | { | |
3039 | if (err) { | |
3040 | if (err == -EOPNOTSUPP) | |
3041 | set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); | |
3042 | clear_bit(BIO_UPTODATE, &bio->bi_flags); | |
3043 | } | |
3044 | if (bio->bi_private) | |
3045 | complete(bio->bi_private); | |
3046 | bio_put(bio); | |
3047 | } | |
3048 | ||
3049 | /* | |
3050 | * trigger flushes for one the devices. If you pass wait == 0, the flushes are | |
3051 | * sent down. With wait == 1, it waits for the previous flush. | |
3052 | * | |
3053 | * any device where the flush fails with eopnotsupp are flagged as not-barrier | |
3054 | * capable | |
3055 | */ | |
3056 | static int write_dev_flush(struct btrfs_device *device, int wait) | |
3057 | { | |
3058 | struct bio *bio; | |
3059 | int ret = 0; | |
3060 | ||
3061 | if (device->nobarriers) | |
3062 | return 0; | |
3063 | ||
3064 | if (wait) { | |
3065 | bio = device->flush_bio; | |
3066 | if (!bio) | |
3067 | return 0; | |
3068 | ||
3069 | wait_for_completion(&device->flush_wait); | |
3070 | ||
3071 | if (bio_flagged(bio, BIO_EOPNOTSUPP)) { | |
606686ee JB |
3072 | printk_in_rcu("btrfs: disabling barriers on dev %s\n", |
3073 | rcu_str_deref(device->name)); | |
387125fc | 3074 | device->nobarriers = 1; |
5af3e8cc | 3075 | } else if (!bio_flagged(bio, BIO_UPTODATE)) { |
387125fc | 3076 | ret = -EIO; |
5af3e8cc SB |
3077 | btrfs_dev_stat_inc_and_print(device, |
3078 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
387125fc CM |
3079 | } |
3080 | ||
3081 | /* drop the reference from the wait == 0 run */ | |
3082 | bio_put(bio); | |
3083 | device->flush_bio = NULL; | |
3084 | ||
3085 | return ret; | |
3086 | } | |
3087 | ||
3088 | /* | |
3089 | * one reference for us, and we leave it for the | |
3090 | * caller | |
3091 | */ | |
9c017abc | 3092 | device->flush_bio = NULL; |
387125fc CM |
3093 | bio = bio_alloc(GFP_NOFS, 0); |
3094 | if (!bio) | |
3095 | return -ENOMEM; | |
3096 | ||
3097 | bio->bi_end_io = btrfs_end_empty_barrier; | |
3098 | bio->bi_bdev = device->bdev; | |
3099 | init_completion(&device->flush_wait); | |
3100 | bio->bi_private = &device->flush_wait; | |
3101 | device->flush_bio = bio; | |
3102 | ||
3103 | bio_get(bio); | |
21adbd5c | 3104 | btrfsic_submit_bio(WRITE_FLUSH, bio); |
387125fc CM |
3105 | |
3106 | return 0; | |
3107 | } | |
3108 | ||
3109 | /* | |
3110 | * send an empty flush down to each device in parallel, | |
3111 | * then wait for them | |
3112 | */ | |
3113 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
3114 | { | |
3115 | struct list_head *head; | |
3116 | struct btrfs_device *dev; | |
5af3e8cc SB |
3117 | int errors_send = 0; |
3118 | int errors_wait = 0; | |
387125fc CM |
3119 | int ret; |
3120 | ||
3121 | /* send down all the barriers */ | |
3122 | head = &info->fs_devices->devices; | |
3123 | list_for_each_entry_rcu(dev, head, dev_list) { | |
3124 | if (!dev->bdev) { | |
5af3e8cc | 3125 | errors_send++; |
387125fc CM |
3126 | continue; |
3127 | } | |
3128 | if (!dev->in_fs_metadata || !dev->writeable) | |
3129 | continue; | |
3130 | ||
3131 | ret = write_dev_flush(dev, 0); | |
3132 | if (ret) | |
5af3e8cc | 3133 | errors_send++; |
387125fc CM |
3134 | } |
3135 | ||
3136 | /* wait for all the barriers */ | |
3137 | list_for_each_entry_rcu(dev, head, dev_list) { | |
3138 | if (!dev->bdev) { | |
5af3e8cc | 3139 | errors_wait++; |
387125fc CM |
3140 | continue; |
3141 | } | |
3142 | if (!dev->in_fs_metadata || !dev->writeable) | |
3143 | continue; | |
3144 | ||
3145 | ret = write_dev_flush(dev, 1); | |
3146 | if (ret) | |
5af3e8cc | 3147 | errors_wait++; |
387125fc | 3148 | } |
5af3e8cc SB |
3149 | if (errors_send > info->num_tolerated_disk_barrier_failures || |
3150 | errors_wait > info->num_tolerated_disk_barrier_failures) | |
387125fc CM |
3151 | return -EIO; |
3152 | return 0; | |
3153 | } | |
3154 | ||
5af3e8cc SB |
3155 | int btrfs_calc_num_tolerated_disk_barrier_failures( |
3156 | struct btrfs_fs_info *fs_info) | |
3157 | { | |
3158 | struct btrfs_ioctl_space_info space; | |
3159 | struct btrfs_space_info *sinfo; | |
3160 | u64 types[] = {BTRFS_BLOCK_GROUP_DATA, | |
3161 | BTRFS_BLOCK_GROUP_SYSTEM, | |
3162 | BTRFS_BLOCK_GROUP_METADATA, | |
3163 | BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; | |
3164 | int num_types = 4; | |
3165 | int i; | |
3166 | int c; | |
3167 | int num_tolerated_disk_barrier_failures = | |
3168 | (int)fs_info->fs_devices->num_devices; | |
3169 | ||
3170 | for (i = 0; i < num_types; i++) { | |
3171 | struct btrfs_space_info *tmp; | |
3172 | ||
3173 | sinfo = NULL; | |
3174 | rcu_read_lock(); | |
3175 | list_for_each_entry_rcu(tmp, &fs_info->space_info, list) { | |
3176 | if (tmp->flags == types[i]) { | |
3177 | sinfo = tmp; | |
3178 | break; | |
3179 | } | |
3180 | } | |
3181 | rcu_read_unlock(); | |
3182 | ||
3183 | if (!sinfo) | |
3184 | continue; | |
3185 | ||
3186 | down_read(&sinfo->groups_sem); | |
3187 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
3188 | if (!list_empty(&sinfo->block_groups[c])) { | |
3189 | u64 flags; | |
3190 | ||
3191 | btrfs_get_block_group_info( | |
3192 | &sinfo->block_groups[c], &space); | |
3193 | if (space.total_bytes == 0 || | |
3194 | space.used_bytes == 0) | |
3195 | continue; | |
3196 | flags = space.flags; | |
3197 | /* | |
3198 | * return | |
3199 | * 0: if dup, single or RAID0 is configured for | |
3200 | * any of metadata, system or data, else | |
3201 | * 1: if RAID5 is configured, or if RAID1 or | |
3202 | * RAID10 is configured and only two mirrors | |
3203 | * are used, else | |
3204 | * 2: if RAID6 is configured, else | |
3205 | * num_mirrors - 1: if RAID1 or RAID10 is | |
3206 | * configured and more than | |
3207 | * 2 mirrors are used. | |
3208 | */ | |
3209 | if (num_tolerated_disk_barrier_failures > 0 && | |
3210 | ((flags & (BTRFS_BLOCK_GROUP_DUP | | |
3211 | BTRFS_BLOCK_GROUP_RAID0)) || | |
3212 | ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) | |
3213 | == 0))) | |
3214 | num_tolerated_disk_barrier_failures = 0; | |
53b381b3 DW |
3215 | else if (num_tolerated_disk_barrier_failures > 1) { |
3216 | if (flags & (BTRFS_BLOCK_GROUP_RAID1 | | |
3217 | BTRFS_BLOCK_GROUP_RAID5 | | |
3218 | BTRFS_BLOCK_GROUP_RAID10)) { | |
3219 | num_tolerated_disk_barrier_failures = 1; | |
3220 | } else if (flags & | |
3221 | BTRFS_BLOCK_GROUP_RAID5) { | |
3222 | num_tolerated_disk_barrier_failures = 2; | |
3223 | } | |
3224 | } | |
5af3e8cc SB |
3225 | } |
3226 | } | |
3227 | up_read(&sinfo->groups_sem); | |
3228 | } | |
3229 | ||
3230 | return num_tolerated_disk_barrier_failures; | |
3231 | } | |
3232 | ||
a512bbf8 | 3233 | int write_all_supers(struct btrfs_root *root, int max_mirrors) |
f2984462 | 3234 | { |
e5e9a520 | 3235 | struct list_head *head; |
f2984462 | 3236 | struct btrfs_device *dev; |
a061fc8d | 3237 | struct btrfs_super_block *sb; |
f2984462 | 3238 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
3239 | int ret; |
3240 | int do_barriers; | |
a236aed1 CM |
3241 | int max_errors; |
3242 | int total_errors = 0; | |
a061fc8d | 3243 | u64 flags; |
f2984462 | 3244 | |
6c41761f | 3245 | max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; |
f2984462 | 3246 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
af31f5e5 | 3247 | backup_super_roots(root->fs_info); |
f2984462 | 3248 | |
6c41761f | 3249 | sb = root->fs_info->super_for_commit; |
a061fc8d | 3250 | dev_item = &sb->dev_item; |
e5e9a520 | 3251 | |
174ba509 | 3252 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); |
e5e9a520 | 3253 | head = &root->fs_info->fs_devices->devices; |
387125fc | 3254 | |
5af3e8cc SB |
3255 | if (do_barriers) { |
3256 | ret = barrier_all_devices(root->fs_info); | |
3257 | if (ret) { | |
3258 | mutex_unlock( | |
3259 | &root->fs_info->fs_devices->device_list_mutex); | |
3260 | btrfs_error(root->fs_info, ret, | |
3261 | "errors while submitting device barriers."); | |
3262 | return ret; | |
3263 | } | |
3264 | } | |
387125fc | 3265 | |
1f78160c | 3266 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3267 | if (!dev->bdev) { |
3268 | total_errors++; | |
3269 | continue; | |
3270 | } | |
2b82032c | 3271 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3272 | continue; |
3273 | ||
2b82032c | 3274 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
3275 | btrfs_set_stack_device_type(dev_item, dev->type); |
3276 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
3277 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
3278 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
3279 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
3280 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
3281 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
3282 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 3283 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 3284 | |
a061fc8d CM |
3285 | flags = btrfs_super_flags(sb); |
3286 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
3287 | ||
a512bbf8 | 3288 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
3289 | if (ret) |
3290 | total_errors++; | |
f2984462 | 3291 | } |
a236aed1 | 3292 | if (total_errors > max_errors) { |
d397712b CM |
3293 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
3294 | total_errors); | |
79787eaa JM |
3295 | |
3296 | /* This shouldn't happen. FUA is masked off if unsupported */ | |
a236aed1 CM |
3297 | BUG(); |
3298 | } | |
f2984462 | 3299 | |
a512bbf8 | 3300 | total_errors = 0; |
1f78160c | 3301 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3302 | if (!dev->bdev) |
3303 | continue; | |
2b82032c | 3304 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3305 | continue; |
3306 | ||
a512bbf8 YZ |
3307 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
3308 | if (ret) | |
3309 | total_errors++; | |
f2984462 | 3310 | } |
174ba509 | 3311 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 3312 | if (total_errors > max_errors) { |
79787eaa JM |
3313 | btrfs_error(root->fs_info, -EIO, |
3314 | "%d errors while writing supers", total_errors); | |
3315 | return -EIO; | |
a236aed1 | 3316 | } |
f2984462 CM |
3317 | return 0; |
3318 | } | |
3319 | ||
a512bbf8 YZ |
3320 | int write_ctree_super(struct btrfs_trans_handle *trans, |
3321 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 3322 | { |
e66f709b | 3323 | int ret; |
5f39d397 | 3324 | |
a512bbf8 | 3325 | ret = write_all_supers(root, max_mirrors); |
5f39d397 | 3326 | return ret; |
cfaa7295 CM |
3327 | } |
3328 | ||
143bede5 | 3329 | void btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f | 3330 | { |
4df27c4d | 3331 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
3332 | radix_tree_delete(&fs_info->fs_roots_radix, |
3333 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 3334 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
3335 | |
3336 | if (btrfs_root_refs(&root->root_item) == 0) | |
3337 | synchronize_srcu(&fs_info->subvol_srcu); | |
3338 | ||
d7634482 | 3339 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { |
3321719e LB |
3340 | btrfs_free_log(NULL, root); |
3341 | btrfs_free_log_root_tree(NULL, fs_info); | |
3342 | } | |
3343 | ||
581bb050 LZ |
3344 | __btrfs_remove_free_space_cache(root->free_ino_pinned); |
3345 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d | 3346 | free_fs_root(root); |
4df27c4d YZ |
3347 | } |
3348 | ||
3349 | static void free_fs_root(struct btrfs_root *root) | |
3350 | { | |
82d5902d | 3351 | iput(root->cache_inode); |
4df27c4d | 3352 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
0ee5dc67 AV |
3353 | if (root->anon_dev) |
3354 | free_anon_bdev(root->anon_dev); | |
4df27c4d YZ |
3355 | free_extent_buffer(root->node); |
3356 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
3357 | kfree(root->free_ino_ctl); |
3358 | kfree(root->free_ino_pinned); | |
d397712b | 3359 | kfree(root->name); |
2619ba1f | 3360 | kfree(root); |
2619ba1f CM |
3361 | } |
3362 | ||
c146afad | 3363 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 3364 | { |
c146afad YZ |
3365 | u64 root_objectid = 0; |
3366 | struct btrfs_root *gang[8]; | |
3367 | int i; | |
3768f368 | 3368 | int ret; |
e089f05c | 3369 | |
c146afad YZ |
3370 | while (1) { |
3371 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
3372 | (void **)gang, root_objectid, | |
3373 | ARRAY_SIZE(gang)); | |
3374 | if (!ret) | |
3375 | break; | |
5d4f98a2 YZ |
3376 | |
3377 | root_objectid = gang[ret - 1]->root_key.objectid + 1; | |
c146afad | 3378 | for (i = 0; i < ret; i++) { |
66b4ffd1 JB |
3379 | int err; |
3380 | ||
c146afad | 3381 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
3382 | err = btrfs_orphan_cleanup(gang[i]); |
3383 | if (err) | |
3384 | return err; | |
c146afad YZ |
3385 | } |
3386 | root_objectid++; | |
3387 | } | |
3388 | return 0; | |
3389 | } | |
a2135011 | 3390 | |
c146afad YZ |
3391 | int btrfs_commit_super(struct btrfs_root *root) |
3392 | { | |
3393 | struct btrfs_trans_handle *trans; | |
3394 | int ret; | |
a74a4b97 | 3395 | |
c146afad | 3396 | mutex_lock(&root->fs_info->cleaner_mutex); |
24bbcf04 | 3397 | btrfs_run_delayed_iputs(root); |
c146afad | 3398 | mutex_unlock(&root->fs_info->cleaner_mutex); |
9d1a2a3a | 3399 | wake_up_process(root->fs_info->cleaner_kthread); |
c71bf099 YZ |
3400 | |
3401 | /* wait until ongoing cleanup work done */ | |
3402 | down_write(&root->fs_info->cleanup_work_sem); | |
3403 | up_write(&root->fs_info->cleanup_work_sem); | |
3404 | ||
7a7eaa40 | 3405 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3406 | if (IS_ERR(trans)) |
3407 | return PTR_ERR(trans); | |
54aa1f4d | 3408 | ret = btrfs_commit_transaction(trans, root); |
79787eaa JM |
3409 | if (ret) |
3410 | return ret; | |
c146afad | 3411 | /* run commit again to drop the original snapshot */ |
7a7eaa40 | 3412 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3413 | if (IS_ERR(trans)) |
3414 | return PTR_ERR(trans); | |
79787eaa JM |
3415 | ret = btrfs_commit_transaction(trans, root); |
3416 | if (ret) | |
3417 | return ret; | |
79154b1b | 3418 | ret = btrfs_write_and_wait_transaction(NULL, root); |
79787eaa JM |
3419 | if (ret) { |
3420 | btrfs_error(root->fs_info, ret, | |
3421 | "Failed to sync btree inode to disk."); | |
3422 | return ret; | |
3423 | } | |
d6bfde87 | 3424 | |
a512bbf8 | 3425 | ret = write_ctree_super(NULL, root, 0); |
c146afad YZ |
3426 | return ret; |
3427 | } | |
3428 | ||
3429 | int close_ctree(struct btrfs_root *root) | |
3430 | { | |
3431 | struct btrfs_fs_info *fs_info = root->fs_info; | |
3432 | int ret; | |
3433 | ||
3434 | fs_info->closing = 1; | |
3435 | smp_mb(); | |
3436 | ||
837d5b6e | 3437 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 3438 | btrfs_pause_balance(fs_info); |
837d5b6e | 3439 | |
8dabb742 SB |
3440 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
3441 | ||
aa1b8cd4 | 3442 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
3443 | |
3444 | /* wait for any defraggers to finish */ | |
3445 | wait_event(fs_info->transaction_wait, | |
3446 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3447 | ||
3448 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 3449 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 3450 | |
c146afad | 3451 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
acce952b | 3452 | ret = btrfs_commit_super(root); |
3453 | if (ret) | |
3454 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
3455 | } | |
3456 | ||
87533c47 | 3457 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) |
68ce9682 | 3458 | btrfs_error_commit_super(root); |
0f7d52f4 | 3459 | |
300e4f8a JB |
3460 | btrfs_put_block_group_cache(fs_info); |
3461 | ||
e3029d9f AV |
3462 | kthread_stop(fs_info->transaction_kthread); |
3463 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 3464 | |
f25784b3 YZ |
3465 | fs_info->closing = 2; |
3466 | smp_mb(); | |
3467 | ||
bcef60f2 AJ |
3468 | btrfs_free_qgroup_config(root->fs_info); |
3469 | ||
963d678b MX |
3470 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
3471 | printk(KERN_INFO "btrfs: at unmount delalloc count %lld\n", | |
3472 | percpu_counter_sum(&fs_info->delalloc_bytes)); | |
b0c68f8b | 3473 | } |
bcc63abb | 3474 | |
be283b2e | 3475 | free_root_pointers(fs_info, 1); |
d20f7043 | 3476 | |
e3029d9f | 3477 | btrfs_free_block_groups(fs_info); |
d10c5f31 | 3478 | |
c146afad | 3479 | del_fs_roots(fs_info); |
d10c5f31 | 3480 | |
c146afad | 3481 | iput(fs_info->btree_inode); |
9ad6b7bc | 3482 | |
7abadb64 | 3483 | btrfs_stop_all_workers(fs_info); |
d6bfde87 | 3484 | |
21adbd5c SB |
3485 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3486 | if (btrfs_test_opt(root, CHECK_INTEGRITY)) | |
3487 | btrfsic_unmount(root, fs_info->fs_devices); | |
3488 | #endif | |
3489 | ||
dfe25020 | 3490 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 3491 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 3492 | |
e2d84521 | 3493 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
963d678b | 3494 | percpu_counter_destroy(&fs_info->delalloc_bytes); |
04160088 | 3495 | bdi_destroy(&fs_info->bdi); |
76dda93c | 3496 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 3497 | |
53b381b3 DW |
3498 | btrfs_free_stripe_hash_table(fs_info); |
3499 | ||
eb60ceac CM |
3500 | return 0; |
3501 | } | |
3502 | ||
b9fab919 CM |
3503 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
3504 | int atomic) | |
5f39d397 | 3505 | { |
1259ab75 | 3506 | int ret; |
727011e0 | 3507 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 3508 | |
0b32f4bb | 3509 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
3510 | if (!ret) |
3511 | return ret; | |
3512 | ||
3513 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
3514 | parent_transid, atomic); |
3515 | if (ret == -EAGAIN) | |
3516 | return ret; | |
1259ab75 | 3517 | return !ret; |
5f39d397 CM |
3518 | } |
3519 | ||
3520 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 3521 | { |
0b32f4bb | 3522 | return set_extent_buffer_uptodate(buf); |
5f39d397 | 3523 | } |
6702ed49 | 3524 | |
5f39d397 CM |
3525 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
3526 | { | |
727011e0 | 3527 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
5f39d397 | 3528 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 3529 | int was_dirty; |
b4ce94de | 3530 | |
b9447ef8 | 3531 | btrfs_assert_tree_locked(buf); |
31b1a2bd JL |
3532 | if (transid != root->fs_info->generation) |
3533 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, " | |
d397712b | 3534 | "found %llu running %llu\n", |
db94535d | 3535 | (unsigned long long)buf->start, |
d397712b CM |
3536 | (unsigned long long)transid, |
3537 | (unsigned long long)root->fs_info->generation); | |
0b32f4bb | 3538 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 MX |
3539 | if (!was_dirty) |
3540 | __percpu_counter_add(&root->fs_info->dirty_metadata_bytes, | |
3541 | buf->len, | |
3542 | root->fs_info->dirty_metadata_batch); | |
eb60ceac CM |
3543 | } |
3544 | ||
b53d3f5d LB |
3545 | static void __btrfs_btree_balance_dirty(struct btrfs_root *root, |
3546 | int flush_delayed) | |
16cdcec7 MX |
3547 | { |
3548 | /* | |
3549 | * looks as though older kernels can get into trouble with | |
3550 | * this code, they end up stuck in balance_dirty_pages forever | |
3551 | */ | |
e2d84521 | 3552 | int ret; |
16cdcec7 MX |
3553 | |
3554 | if (current->flags & PF_MEMALLOC) | |
3555 | return; | |
3556 | ||
b53d3f5d LB |
3557 | if (flush_delayed) |
3558 | btrfs_balance_delayed_items(root); | |
16cdcec7 | 3559 | |
e2d84521 MX |
3560 | ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes, |
3561 | BTRFS_DIRTY_METADATA_THRESH); | |
3562 | if (ret > 0) { | |
d0e1d66b NJ |
3563 | balance_dirty_pages_ratelimited( |
3564 | root->fs_info->btree_inode->i_mapping); | |
16cdcec7 MX |
3565 | } |
3566 | return; | |
3567 | } | |
3568 | ||
b53d3f5d | 3569 | void btrfs_btree_balance_dirty(struct btrfs_root *root) |
35b7e476 | 3570 | { |
b53d3f5d LB |
3571 | __btrfs_btree_balance_dirty(root, 1); |
3572 | } | |
585ad2c3 | 3573 | |
b53d3f5d LB |
3574 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root) |
3575 | { | |
3576 | __btrfs_btree_balance_dirty(root, 0); | |
35b7e476 | 3577 | } |
6b80053d | 3578 | |
ca7a79ad | 3579 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 3580 | { |
727011e0 | 3581 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
0b32f4bb | 3582 | return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
6b80053d | 3583 | } |
0da5468f | 3584 | |
fcd1f065 | 3585 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 3586 | int read_only) |
3587 | { | |
fcd1f065 DS |
3588 | if (btrfs_super_csum_type(fs_info->super_copy) >= ARRAY_SIZE(btrfs_csum_sizes)) { |
3589 | printk(KERN_ERR "btrfs: unsupported checksum algorithm\n"); | |
3590 | return -EINVAL; | |
3591 | } | |
3592 | ||
acce952b | 3593 | if (read_only) |
fcd1f065 | 3594 | return 0; |
acce952b | 3595 | |
fcd1f065 | 3596 | return 0; |
acce952b | 3597 | } |
3598 | ||
68ce9682 | 3599 | void btrfs_error_commit_super(struct btrfs_root *root) |
acce952b | 3600 | { |
acce952b | 3601 | mutex_lock(&root->fs_info->cleaner_mutex); |
3602 | btrfs_run_delayed_iputs(root); | |
3603 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
3604 | ||
3605 | down_write(&root->fs_info->cleanup_work_sem); | |
3606 | up_write(&root->fs_info->cleanup_work_sem); | |
3607 | ||
3608 | /* cleanup FS via transaction */ | |
3609 | btrfs_cleanup_transaction(root); | |
acce952b | 3610 | } |
3611 | ||
569e0f35 JB |
3612 | static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, |
3613 | struct btrfs_root *root) | |
acce952b | 3614 | { |
3615 | struct btrfs_inode *btrfs_inode; | |
3616 | struct list_head splice; | |
3617 | ||
3618 | INIT_LIST_HEAD(&splice); | |
3619 | ||
3620 | mutex_lock(&root->fs_info->ordered_operations_mutex); | |
3621 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3622 | ||
569e0f35 | 3623 | list_splice_init(&t->ordered_operations, &splice); |
acce952b | 3624 | while (!list_empty(&splice)) { |
3625 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3626 | ordered_operations); | |
3627 | ||
3628 | list_del_init(&btrfs_inode->ordered_operations); | |
3629 | ||
3630 | btrfs_invalidate_inodes(btrfs_inode->root); | |
3631 | } | |
3632 | ||
3633 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3634 | mutex_unlock(&root->fs_info->ordered_operations_mutex); | |
acce952b | 3635 | } |
3636 | ||
143bede5 | 3637 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 3638 | { |
acce952b | 3639 | struct btrfs_ordered_extent *ordered; |
acce952b | 3640 | |
3641 | spin_lock(&root->fs_info->ordered_extent_lock); | |
779880ef JB |
3642 | /* |
3643 | * This will just short circuit the ordered completion stuff which will | |
3644 | * make sure the ordered extent gets properly cleaned up. | |
3645 | */ | |
3646 | list_for_each_entry(ordered, &root->fs_info->ordered_extents, | |
3647 | root_extent_list) | |
3648 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
acce952b | 3649 | spin_unlock(&root->fs_info->ordered_extent_lock); |
acce952b | 3650 | } |
3651 | ||
79787eaa JM |
3652 | int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
3653 | struct btrfs_root *root) | |
acce952b | 3654 | { |
3655 | struct rb_node *node; | |
3656 | struct btrfs_delayed_ref_root *delayed_refs; | |
3657 | struct btrfs_delayed_ref_node *ref; | |
3658 | int ret = 0; | |
3659 | ||
3660 | delayed_refs = &trans->delayed_refs; | |
3661 | ||
3662 | spin_lock(&delayed_refs->lock); | |
3663 | if (delayed_refs->num_entries == 0) { | |
cfece4db | 3664 | spin_unlock(&delayed_refs->lock); |
acce952b | 3665 | printk(KERN_INFO "delayed_refs has NO entry\n"); |
3666 | return ret; | |
3667 | } | |
3668 | ||
b939d1ab | 3669 | while ((node = rb_first(&delayed_refs->root)) != NULL) { |
eb12db69 | 3670 | struct btrfs_delayed_ref_head *head = NULL; |
acce952b | 3671 | |
eb12db69 | 3672 | ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); |
acce952b | 3673 | atomic_set(&ref->refs, 1); |
3674 | if (btrfs_delayed_ref_is_head(ref)) { | |
acce952b | 3675 | |
3676 | head = btrfs_delayed_node_to_head(ref); | |
b939d1ab JB |
3677 | if (!mutex_trylock(&head->mutex)) { |
3678 | atomic_inc(&ref->refs); | |
3679 | spin_unlock(&delayed_refs->lock); | |
3680 | ||
3681 | /* Need to wait for the delayed ref to run */ | |
3682 | mutex_lock(&head->mutex); | |
3683 | mutex_unlock(&head->mutex); | |
3684 | btrfs_put_delayed_ref(ref); | |
3685 | ||
e18fca73 | 3686 | spin_lock(&delayed_refs->lock); |
b939d1ab JB |
3687 | continue; |
3688 | } | |
3689 | ||
54067ae9 JB |
3690 | if (head->must_insert_reserved) |
3691 | btrfs_pin_extent(root, ref->bytenr, | |
3692 | ref->num_bytes, 1); | |
78a6184a | 3693 | btrfs_free_delayed_extent_op(head->extent_op); |
acce952b | 3694 | delayed_refs->num_heads--; |
3695 | if (list_empty(&head->cluster)) | |
3696 | delayed_refs->num_heads_ready--; | |
3697 | list_del_init(&head->cluster); | |
acce952b | 3698 | } |
eb12db69 | 3699 | |
b939d1ab JB |
3700 | ref->in_tree = 0; |
3701 | rb_erase(&ref->rb_node, &delayed_refs->root); | |
3702 | delayed_refs->num_entries--; | |
eb12db69 JB |
3703 | if (head) |
3704 | mutex_unlock(&head->mutex); | |
acce952b | 3705 | spin_unlock(&delayed_refs->lock); |
3706 | btrfs_put_delayed_ref(ref); | |
3707 | ||
3708 | cond_resched(); | |
3709 | spin_lock(&delayed_refs->lock); | |
3710 | } | |
3711 | ||
3712 | spin_unlock(&delayed_refs->lock); | |
3713 | ||
3714 | return ret; | |
3715 | } | |
3716 | ||
aec8030a | 3717 | static void btrfs_evict_pending_snapshots(struct btrfs_transaction *t) |
acce952b | 3718 | { |
3719 | struct btrfs_pending_snapshot *snapshot; | |
3720 | struct list_head splice; | |
3721 | ||
3722 | INIT_LIST_HEAD(&splice); | |
3723 | ||
3724 | list_splice_init(&t->pending_snapshots, &splice); | |
3725 | ||
3726 | while (!list_empty(&splice)) { | |
3727 | snapshot = list_entry(splice.next, | |
3728 | struct btrfs_pending_snapshot, | |
3729 | list); | |
aec8030a | 3730 | snapshot->error = -ECANCELED; |
acce952b | 3731 | list_del_init(&snapshot->list); |
acce952b | 3732 | } |
acce952b | 3733 | } |
3734 | ||
143bede5 | 3735 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 3736 | { |
3737 | struct btrfs_inode *btrfs_inode; | |
3738 | struct list_head splice; | |
3739 | ||
3740 | INIT_LIST_HEAD(&splice); | |
3741 | ||
acce952b | 3742 | spin_lock(&root->fs_info->delalloc_lock); |
5be76758 | 3743 | list_splice_init(&root->fs_info->delalloc_inodes, &splice); |
acce952b | 3744 | |
3745 | while (!list_empty(&splice)) { | |
3746 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3747 | delalloc_inodes); | |
3748 | ||
3749 | list_del_init(&btrfs_inode->delalloc_inodes); | |
df0af1a5 MX |
3750 | clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, |
3751 | &btrfs_inode->runtime_flags); | |
acce952b | 3752 | |
3753 | btrfs_invalidate_inodes(btrfs_inode->root); | |
3754 | } | |
3755 | ||
3756 | spin_unlock(&root->fs_info->delalloc_lock); | |
acce952b | 3757 | } |
3758 | ||
3759 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, | |
3760 | struct extent_io_tree *dirty_pages, | |
3761 | int mark) | |
3762 | { | |
3763 | int ret; | |
acce952b | 3764 | struct extent_buffer *eb; |
3765 | u64 start = 0; | |
3766 | u64 end; | |
acce952b | 3767 | |
3768 | while (1) { | |
3769 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 3770 | mark, NULL); |
acce952b | 3771 | if (ret) |
3772 | break; | |
3773 | ||
3774 | clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); | |
3775 | while (start <= end) { | |
fd8b2b61 JB |
3776 | eb = btrfs_find_tree_block(root, start, |
3777 | root->leafsize); | |
3778 | start += eb->len; | |
3779 | if (!eb) | |
acce952b | 3780 | continue; |
fd8b2b61 | 3781 | wait_on_extent_buffer_writeback(eb); |
acce952b | 3782 | |
fd8b2b61 JB |
3783 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
3784 | &eb->bflags)) | |
3785 | clear_extent_buffer_dirty(eb); | |
3786 | free_extent_buffer_stale(eb); | |
acce952b | 3787 | } |
3788 | } | |
3789 | ||
3790 | return ret; | |
3791 | } | |
3792 | ||
3793 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
3794 | struct extent_io_tree *pinned_extents) | |
3795 | { | |
3796 | struct extent_io_tree *unpin; | |
3797 | u64 start; | |
3798 | u64 end; | |
3799 | int ret; | |
ed0eaa14 | 3800 | bool loop = true; |
acce952b | 3801 | |
3802 | unpin = pinned_extents; | |
ed0eaa14 | 3803 | again: |
acce952b | 3804 | while (1) { |
3805 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
e6138876 | 3806 | EXTENT_DIRTY, NULL); |
acce952b | 3807 | if (ret) |
3808 | break; | |
3809 | ||
3810 | /* opt_discard */ | |
5378e607 LD |
3811 | if (btrfs_test_opt(root, DISCARD)) |
3812 | ret = btrfs_error_discard_extent(root, start, | |
3813 | end + 1 - start, | |
3814 | NULL); | |
acce952b | 3815 | |
3816 | clear_extent_dirty(unpin, start, end, GFP_NOFS); | |
3817 | btrfs_error_unpin_extent_range(root, start, end); | |
3818 | cond_resched(); | |
3819 | } | |
3820 | ||
ed0eaa14 LB |
3821 | if (loop) { |
3822 | if (unpin == &root->fs_info->freed_extents[0]) | |
3823 | unpin = &root->fs_info->freed_extents[1]; | |
3824 | else | |
3825 | unpin = &root->fs_info->freed_extents[0]; | |
3826 | loop = false; | |
3827 | goto again; | |
3828 | } | |
3829 | ||
acce952b | 3830 | return 0; |
3831 | } | |
3832 | ||
49b25e05 JM |
3833 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
3834 | struct btrfs_root *root) | |
3835 | { | |
3836 | btrfs_destroy_delayed_refs(cur_trans, root); | |
3837 | btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv, | |
3838 | cur_trans->dirty_pages.dirty_bytes); | |
3839 | ||
3840 | /* FIXME: cleanup wait for commit */ | |
3841 | cur_trans->in_commit = 1; | |
3842 | cur_trans->blocked = 1; | |
d7096fc3 | 3843 | wake_up(&root->fs_info->transaction_blocked_wait); |
49b25e05 | 3844 | |
aec8030a MX |
3845 | btrfs_evict_pending_snapshots(cur_trans); |
3846 | ||
49b25e05 | 3847 | cur_trans->blocked = 0; |
d7096fc3 | 3848 | wake_up(&root->fs_info->transaction_wait); |
49b25e05 JM |
3849 | |
3850 | cur_trans->commit_done = 1; | |
d7096fc3 | 3851 | wake_up(&cur_trans->commit_wait); |
49b25e05 | 3852 | |
67cde344 MX |
3853 | btrfs_destroy_delayed_inodes(root); |
3854 | btrfs_assert_delayed_root_empty(root); | |
49b25e05 | 3855 | |
49b25e05 JM |
3856 | btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, |
3857 | EXTENT_DIRTY); | |
6e841e32 LB |
3858 | btrfs_destroy_pinned_extent(root, |
3859 | root->fs_info->pinned_extents); | |
49b25e05 JM |
3860 | |
3861 | /* | |
3862 | memset(cur_trans, 0, sizeof(*cur_trans)); | |
3863 | kmem_cache_free(btrfs_transaction_cachep, cur_trans); | |
3864 | */ | |
3865 | } | |
3866 | ||
3867 | int btrfs_cleanup_transaction(struct btrfs_root *root) | |
acce952b | 3868 | { |
3869 | struct btrfs_transaction *t; | |
3870 | LIST_HEAD(list); | |
3871 | ||
acce952b | 3872 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
3873 | ||
a4abeea4 | 3874 | spin_lock(&root->fs_info->trans_lock); |
acce952b | 3875 | list_splice_init(&root->fs_info->trans_list, &list); |
a4abeea4 JB |
3876 | root->fs_info->trans_no_join = 1; |
3877 | spin_unlock(&root->fs_info->trans_lock); | |
3878 | ||
acce952b | 3879 | while (!list_empty(&list)) { |
3880 | t = list_entry(list.next, struct btrfs_transaction, list); | |
acce952b | 3881 | |
569e0f35 | 3882 | btrfs_destroy_ordered_operations(t, root); |
acce952b | 3883 | |
3884 | btrfs_destroy_ordered_extents(root); | |
3885 | ||
3886 | btrfs_destroy_delayed_refs(t, root); | |
3887 | ||
acce952b | 3888 | /* FIXME: cleanup wait for commit */ |
3889 | t->in_commit = 1; | |
3890 | t->blocked = 1; | |
66657b31 | 3891 | smp_mb(); |
acce952b | 3892 | if (waitqueue_active(&root->fs_info->transaction_blocked_wait)) |
3893 | wake_up(&root->fs_info->transaction_blocked_wait); | |
3894 | ||
aec8030a MX |
3895 | btrfs_evict_pending_snapshots(t); |
3896 | ||
acce952b | 3897 | t->blocked = 0; |
66657b31 | 3898 | smp_mb(); |
acce952b | 3899 | if (waitqueue_active(&root->fs_info->transaction_wait)) |
3900 | wake_up(&root->fs_info->transaction_wait); | |
acce952b | 3901 | |
acce952b | 3902 | t->commit_done = 1; |
66657b31 | 3903 | smp_mb(); |
acce952b | 3904 | if (waitqueue_active(&t->commit_wait)) |
3905 | wake_up(&t->commit_wait); | |
acce952b | 3906 | |
67cde344 MX |
3907 | btrfs_destroy_delayed_inodes(root); |
3908 | btrfs_assert_delayed_root_empty(root); | |
3909 | ||
acce952b | 3910 | btrfs_destroy_delalloc_inodes(root); |
3911 | ||
a4abeea4 | 3912 | spin_lock(&root->fs_info->trans_lock); |
acce952b | 3913 | root->fs_info->running_transaction = NULL; |
a4abeea4 | 3914 | spin_unlock(&root->fs_info->trans_lock); |
acce952b | 3915 | |
3916 | btrfs_destroy_marked_extents(root, &t->dirty_pages, | |
3917 | EXTENT_DIRTY); | |
3918 | ||
3919 | btrfs_destroy_pinned_extent(root, | |
3920 | root->fs_info->pinned_extents); | |
3921 | ||
13c5a93e | 3922 | atomic_set(&t->use_count, 0); |
acce952b | 3923 | list_del_init(&t->list); |
3924 | memset(t, 0, sizeof(*t)); | |
3925 | kmem_cache_free(btrfs_transaction_cachep, t); | |
3926 | } | |
3927 | ||
a4abeea4 JB |
3928 | spin_lock(&root->fs_info->trans_lock); |
3929 | root->fs_info->trans_no_join = 0; | |
3930 | spin_unlock(&root->fs_info->trans_lock); | |
acce952b | 3931 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); |
acce952b | 3932 | |
3933 | return 0; | |
3934 | } | |
3935 | ||
d1310b2e | 3936 | static struct extent_io_ops btree_extent_io_ops = { |
ce9adaa5 | 3937 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
4bb31e92 | 3938 | .readpage_io_failed_hook = btree_io_failed_hook, |
0b86a832 | 3939 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
3940 | /* note we're sharing with inode.c for the merge bio hook */ |
3941 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 3942 | }; |