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> |
7e75bf3f | 33 | #include <asm/unaligned.h> |
4b4e25f2 | 34 | #include "compat.h" |
eb60ceac CM |
35 | #include "ctree.h" |
36 | #include "disk-io.h" | |
e089f05c | 37 | #include "transaction.h" |
0f7d52f4 | 38 | #include "btrfs_inode.h" |
0b86a832 | 39 | #include "volumes.h" |
db94535d | 40 | #include "print-tree.h" |
8b712842 | 41 | #include "async-thread.h" |
925baedd | 42 | #include "locking.h" |
e02119d5 | 43 | #include "tree-log.h" |
fa9c0d79 | 44 | #include "free-space-cache.h" |
581bb050 | 45 | #include "inode-map.h" |
21adbd5c | 46 | #include "check-integrity.h" |
606686ee | 47 | #include "rcu-string.h" |
eb60ceac | 48 | |
d1310b2e | 49 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 50 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 51 | static void free_fs_root(struct btrfs_root *root); |
fcd1f065 | 52 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 53 | int read_only); |
143bede5 JM |
54 | static void btrfs_destroy_ordered_operations(struct btrfs_root *root); |
55 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); | |
acce952b | 56 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
57 | struct btrfs_root *root); | |
143bede5 JM |
58 | static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t); |
59 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); | |
acce952b | 60 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, |
61 | struct extent_io_tree *dirty_pages, | |
62 | int mark); | |
63 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
64 | struct extent_io_tree *pinned_extents); | |
ce9adaa5 | 65 | |
d352ac68 CM |
66 | /* |
67 | * end_io_wq structs are used to do processing in task context when an IO is | |
68 | * complete. This is used during reads to verify checksums, and it is used | |
69 | * by writes to insert metadata for new file extents after IO is complete. | |
70 | */ | |
ce9adaa5 CM |
71 | struct end_io_wq { |
72 | struct bio *bio; | |
73 | bio_end_io_t *end_io; | |
74 | void *private; | |
75 | struct btrfs_fs_info *info; | |
76 | int error; | |
22c59948 | 77 | int metadata; |
ce9adaa5 | 78 | struct list_head list; |
8b712842 | 79 | struct btrfs_work work; |
ce9adaa5 | 80 | }; |
0da5468f | 81 | |
d352ac68 CM |
82 | /* |
83 | * async submit bios are used to offload expensive checksumming | |
84 | * onto the worker threads. They checksum file and metadata bios | |
85 | * just before they are sent down the IO stack. | |
86 | */ | |
44b8bd7e CM |
87 | struct async_submit_bio { |
88 | struct inode *inode; | |
89 | struct bio *bio; | |
90 | struct list_head list; | |
4a69a410 CM |
91 | extent_submit_bio_hook_t *submit_bio_start; |
92 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
93 | int rw; |
94 | int mirror_num; | |
c8b97818 | 95 | unsigned long bio_flags; |
eaf25d93 CM |
96 | /* |
97 | * bio_offset is optional, can be used if the pages in the bio | |
98 | * can't tell us where in the file the bio should go | |
99 | */ | |
100 | u64 bio_offset; | |
8b712842 | 101 | struct btrfs_work work; |
79787eaa | 102 | int error; |
44b8bd7e CM |
103 | }; |
104 | ||
85d4e461 CM |
105 | /* |
106 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
107 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
108 | * the level the eb occupies in the tree. | |
109 | * | |
110 | * Different roots are used for different purposes and may nest inside each | |
111 | * other and they require separate keysets. As lockdep keys should be | |
112 | * static, assign keysets according to the purpose of the root as indicated | |
113 | * by btrfs_root->objectid. This ensures that all special purpose roots | |
114 | * have separate keysets. | |
4008c04a | 115 | * |
85d4e461 CM |
116 | * Lock-nesting across peer nodes is always done with the immediate parent |
117 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
118 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 119 | * |
85d4e461 CM |
120 | * The key is set by the readpage_end_io_hook after the buffer has passed |
121 | * csum validation but before the pages are unlocked. It is also set by | |
122 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 123 | * |
85d4e461 CM |
124 | * We also add a check to make sure the highest level of the tree is the |
125 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
126 | * needs update as well. | |
4008c04a CM |
127 | */ |
128 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
129 | # if BTRFS_MAX_LEVEL != 8 | |
130 | # error | |
131 | # endif | |
85d4e461 CM |
132 | |
133 | static struct btrfs_lockdep_keyset { | |
134 | u64 id; /* root objectid */ | |
135 | const char *name_stem; /* lock name stem */ | |
136 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
137 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
138 | } btrfs_lockdep_keysets[] = { | |
139 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
140 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
141 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
142 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
143 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
144 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
145 | { .id = BTRFS_ORPHAN_OBJECTID, .name_stem = "orphan" }, | |
146 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, | |
147 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
148 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
149 | { .id = 0, .name_stem = "tree" }, | |
4008c04a | 150 | }; |
85d4e461 CM |
151 | |
152 | void __init btrfs_init_lockdep(void) | |
153 | { | |
154 | int i, j; | |
155 | ||
156 | /* initialize lockdep class names */ | |
157 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
158 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
159 | ||
160 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
161 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
162 | "btrfs-%s-%02d", ks->name_stem, j); | |
163 | } | |
164 | } | |
165 | ||
166 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
167 | int level) | |
168 | { | |
169 | struct btrfs_lockdep_keyset *ks; | |
170 | ||
171 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
172 | ||
173 | /* find the matching keyset, id 0 is the default entry */ | |
174 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
175 | if (ks->id == objectid) | |
176 | break; | |
177 | ||
178 | lockdep_set_class_and_name(&eb->lock, | |
179 | &ks->keys[level], ks->names[level]); | |
180 | } | |
181 | ||
4008c04a CM |
182 | #endif |
183 | ||
d352ac68 CM |
184 | /* |
185 | * extents on the btree inode are pretty simple, there's one extent | |
186 | * that covers the entire device | |
187 | */ | |
b2950863 | 188 | static struct extent_map *btree_get_extent(struct inode *inode, |
306e16ce | 189 | struct page *page, size_t pg_offset, u64 start, u64 len, |
b2950863 | 190 | int create) |
7eccb903 | 191 | { |
5f39d397 CM |
192 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
193 | struct extent_map *em; | |
194 | int ret; | |
195 | ||
890871be | 196 | read_lock(&em_tree->lock); |
d1310b2e | 197 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d CM |
198 | if (em) { |
199 | em->bdev = | |
200 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
890871be | 201 | read_unlock(&em_tree->lock); |
5f39d397 | 202 | goto out; |
a061fc8d | 203 | } |
890871be | 204 | read_unlock(&em_tree->lock); |
7b13b7b1 | 205 | |
172ddd60 | 206 | em = alloc_extent_map(); |
5f39d397 CM |
207 | if (!em) { |
208 | em = ERR_PTR(-ENOMEM); | |
209 | goto out; | |
210 | } | |
211 | em->start = 0; | |
0afbaf8c | 212 | em->len = (u64)-1; |
c8b97818 | 213 | em->block_len = (u64)-1; |
5f39d397 | 214 | em->block_start = 0; |
a061fc8d | 215 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e | 216 | |
890871be | 217 | write_lock(&em_tree->lock); |
5f39d397 CM |
218 | ret = add_extent_mapping(em_tree, em); |
219 | if (ret == -EEXIST) { | |
0afbaf8c CM |
220 | u64 failed_start = em->start; |
221 | u64 failed_len = em->len; | |
222 | ||
5f39d397 | 223 | free_extent_map(em); |
7b13b7b1 | 224 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c | 225 | if (em) { |
7b13b7b1 | 226 | ret = 0; |
0afbaf8c CM |
227 | } else { |
228 | em = lookup_extent_mapping(em_tree, failed_start, | |
229 | failed_len); | |
7b13b7b1 | 230 | ret = -EIO; |
0afbaf8c | 231 | } |
5f39d397 | 232 | } else if (ret) { |
7b13b7b1 CM |
233 | free_extent_map(em); |
234 | em = NULL; | |
5f39d397 | 235 | } |
890871be | 236 | write_unlock(&em_tree->lock); |
7b13b7b1 CM |
237 | |
238 | if (ret) | |
239 | em = ERR_PTR(ret); | |
5f39d397 CM |
240 | out: |
241 | return em; | |
7eccb903 CM |
242 | } |
243 | ||
19c00ddc CM |
244 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
245 | { | |
163e783e | 246 | return crc32c(seed, data, len); |
19c00ddc CM |
247 | } |
248 | ||
249 | void btrfs_csum_final(u32 crc, char *result) | |
250 | { | |
7e75bf3f | 251 | put_unaligned_le32(~crc, result); |
19c00ddc CM |
252 | } |
253 | ||
d352ac68 CM |
254 | /* |
255 | * compute the csum for a btree block, and either verify it or write it | |
256 | * into the csum field of the block. | |
257 | */ | |
19c00ddc CM |
258 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
259 | int verify) | |
260 | { | |
6c41761f | 261 | u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy); |
607d432d | 262 | char *result = NULL; |
19c00ddc CM |
263 | unsigned long len; |
264 | unsigned long cur_len; | |
265 | unsigned long offset = BTRFS_CSUM_SIZE; | |
19c00ddc CM |
266 | char *kaddr; |
267 | unsigned long map_start; | |
268 | unsigned long map_len; | |
269 | int err; | |
270 | u32 crc = ~(u32)0; | |
607d432d | 271 | unsigned long inline_result; |
19c00ddc CM |
272 | |
273 | len = buf->len - offset; | |
d397712b | 274 | while (len > 0) { |
19c00ddc | 275 | err = map_private_extent_buffer(buf, offset, 32, |
a6591715 | 276 | &kaddr, &map_start, &map_len); |
d397712b | 277 | if (err) |
19c00ddc | 278 | return 1; |
19c00ddc CM |
279 | cur_len = min(len, map_len - (offset - map_start)); |
280 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
281 | crc, cur_len); | |
282 | len -= cur_len; | |
283 | offset += cur_len; | |
19c00ddc | 284 | } |
607d432d JB |
285 | if (csum_size > sizeof(inline_result)) { |
286 | result = kzalloc(csum_size * sizeof(char), GFP_NOFS); | |
287 | if (!result) | |
288 | return 1; | |
289 | } else { | |
290 | result = (char *)&inline_result; | |
291 | } | |
292 | ||
19c00ddc CM |
293 | btrfs_csum_final(crc, result); |
294 | ||
295 | if (verify) { | |
607d432d | 296 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
297 | u32 val; |
298 | u32 found = 0; | |
607d432d | 299 | memcpy(&found, result, csum_size); |
e4204ded | 300 | |
607d432d | 301 | read_extent_buffer(buf, &val, 0, csum_size); |
7a36ddec | 302 | printk_ratelimited(KERN_INFO "btrfs: %s checksum verify " |
193f284d CM |
303 | "failed on %llu wanted %X found %X " |
304 | "level %d\n", | |
305 | root->fs_info->sb->s_id, | |
306 | (unsigned long long)buf->start, val, found, | |
307 | btrfs_header_level(buf)); | |
607d432d JB |
308 | if (result != (char *)&inline_result) |
309 | kfree(result); | |
19c00ddc CM |
310 | return 1; |
311 | } | |
312 | } else { | |
607d432d | 313 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 314 | } |
607d432d JB |
315 | if (result != (char *)&inline_result) |
316 | kfree(result); | |
19c00ddc CM |
317 | return 0; |
318 | } | |
319 | ||
d352ac68 CM |
320 | /* |
321 | * we can't consider a given block up to date unless the transid of the | |
322 | * block matches the transid in the parent node's pointer. This is how we | |
323 | * detect blocks that either didn't get written at all or got written | |
324 | * in the wrong place. | |
325 | */ | |
1259ab75 | 326 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
327 | struct extent_buffer *eb, u64 parent_transid, |
328 | int atomic) | |
1259ab75 | 329 | { |
2ac55d41 | 330 | struct extent_state *cached_state = NULL; |
1259ab75 CM |
331 | int ret; |
332 | ||
333 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
334 | return 0; | |
335 | ||
b9fab919 CM |
336 | if (atomic) |
337 | return -EAGAIN; | |
338 | ||
2ac55d41 | 339 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
d0082371 | 340 | 0, &cached_state); |
0b32f4bb | 341 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
342 | btrfs_header_generation(eb) == parent_transid) { |
343 | ret = 0; | |
344 | goto out; | |
345 | } | |
7a36ddec | 346 | printk_ratelimited("parent transid verify failed on %llu wanted %llu " |
193f284d CM |
347 | "found %llu\n", |
348 | (unsigned long long)eb->start, | |
349 | (unsigned long long)parent_transid, | |
350 | (unsigned long long)btrfs_header_generation(eb)); | |
1259ab75 | 351 | ret = 1; |
0b32f4bb | 352 | clear_extent_buffer_uptodate(eb); |
33958dc6 | 353 | out: |
2ac55d41 JB |
354 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
355 | &cached_state, GFP_NOFS); | |
1259ab75 | 356 | return ret; |
1259ab75 CM |
357 | } |
358 | ||
d352ac68 CM |
359 | /* |
360 | * helper to read a given tree block, doing retries as required when | |
361 | * the checksums don't match and we have alternate mirrors to try. | |
362 | */ | |
f188591e CM |
363 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
364 | struct extent_buffer *eb, | |
ca7a79ad | 365 | u64 start, u64 parent_transid) |
f188591e CM |
366 | { |
367 | struct extent_io_tree *io_tree; | |
ea466794 | 368 | int failed = 0; |
f188591e CM |
369 | int ret; |
370 | int num_copies = 0; | |
371 | int mirror_num = 0; | |
ea466794 | 372 | int failed_mirror = 0; |
f188591e | 373 | |
a826d6dc | 374 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
f188591e CM |
375 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
376 | while (1) { | |
bb82ab88 AJ |
377 | ret = read_extent_buffer_pages(io_tree, eb, start, |
378 | WAIT_COMPLETE, | |
f188591e | 379 | btree_get_extent, mirror_num); |
b9fab919 CM |
380 | if (!ret && !verify_parent_transid(io_tree, eb, |
381 | parent_transid, 0)) | |
ea466794 | 382 | break; |
d397712b | 383 | |
a826d6dc JB |
384 | /* |
385 | * This buffer's crc is fine, but its contents are corrupted, so | |
386 | * there is no reason to read the other copies, they won't be | |
387 | * any less wrong. | |
388 | */ | |
389 | if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) | |
ea466794 JB |
390 | break; |
391 | ||
f188591e CM |
392 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
393 | eb->start, eb->len); | |
4235298e | 394 | if (num_copies == 1) |
ea466794 | 395 | break; |
4235298e | 396 | |
5cf1ab56 JB |
397 | if (!failed_mirror) { |
398 | failed = 1; | |
399 | failed_mirror = eb->read_mirror; | |
400 | } | |
401 | ||
f188591e | 402 | mirror_num++; |
ea466794 JB |
403 | if (mirror_num == failed_mirror) |
404 | mirror_num++; | |
405 | ||
4235298e | 406 | if (mirror_num > num_copies) |
ea466794 | 407 | break; |
f188591e | 408 | } |
ea466794 | 409 | |
c0901581 | 410 | if (failed && !ret && failed_mirror) |
ea466794 JB |
411 | repair_eb_io_failure(root, eb, failed_mirror); |
412 | ||
413 | return ret; | |
f188591e | 414 | } |
19c00ddc | 415 | |
d352ac68 | 416 | /* |
d397712b CM |
417 | * checksum a dirty tree block before IO. This has extra checks to make sure |
418 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 419 | */ |
d397712b | 420 | |
b2950863 | 421 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 422 | { |
d1310b2e | 423 | struct extent_io_tree *tree; |
35ebb934 | 424 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc | 425 | u64 found_start; |
19c00ddc | 426 | struct extent_buffer *eb; |
f188591e | 427 | |
d1310b2e | 428 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc | 429 | |
4f2de97a JB |
430 | eb = (struct extent_buffer *)page->private; |
431 | if (page != eb->pages[0]) | |
432 | return 0; | |
19c00ddc CM |
433 | found_start = btrfs_header_bytenr(eb); |
434 | if (found_start != start) { | |
55c69072 | 435 | WARN_ON(1); |
4f2de97a | 436 | return 0; |
55c69072 | 437 | } |
727011e0 | 438 | if (eb->pages[0] != page) { |
55c69072 | 439 | WARN_ON(1); |
4f2de97a | 440 | return 0; |
55c69072 CM |
441 | } |
442 | if (!PageUptodate(page)) { | |
55c69072 | 443 | WARN_ON(1); |
4f2de97a | 444 | return 0; |
19c00ddc | 445 | } |
19c00ddc | 446 | csum_tree_block(root, eb, 0); |
19c00ddc CM |
447 | return 0; |
448 | } | |
449 | ||
2b82032c YZ |
450 | static int check_tree_block_fsid(struct btrfs_root *root, |
451 | struct extent_buffer *eb) | |
452 | { | |
453 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
454 | u8 fsid[BTRFS_UUID_SIZE]; | |
455 | int ret = 1; | |
456 | ||
457 | read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), | |
458 | BTRFS_FSID_SIZE); | |
459 | while (fs_devices) { | |
460 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
461 | ret = 0; | |
462 | break; | |
463 | } | |
464 | fs_devices = fs_devices->seed; | |
465 | } | |
466 | return ret; | |
467 | } | |
468 | ||
a826d6dc JB |
469 | #define CORRUPT(reason, eb, root, slot) \ |
470 | printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \ | |
471 | "root=%llu, slot=%d\n", reason, \ | |
472 | (unsigned long long)btrfs_header_bytenr(eb), \ | |
473 | (unsigned long long)root->objectid, slot) | |
474 | ||
475 | static noinline int check_leaf(struct btrfs_root *root, | |
476 | struct extent_buffer *leaf) | |
477 | { | |
478 | struct btrfs_key key; | |
479 | struct btrfs_key leaf_key; | |
480 | u32 nritems = btrfs_header_nritems(leaf); | |
481 | int slot; | |
482 | ||
483 | if (nritems == 0) | |
484 | return 0; | |
485 | ||
486 | /* Check the 0 item */ | |
487 | if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != | |
488 | BTRFS_LEAF_DATA_SIZE(root)) { | |
489 | CORRUPT("invalid item offset size pair", leaf, root, 0); | |
490 | return -EIO; | |
491 | } | |
492 | ||
493 | /* | |
494 | * Check to make sure each items keys are in the correct order and their | |
495 | * offsets make sense. We only have to loop through nritems-1 because | |
496 | * we check the current slot against the next slot, which verifies the | |
497 | * next slot's offset+size makes sense and that the current's slot | |
498 | * offset is correct. | |
499 | */ | |
500 | for (slot = 0; slot < nritems - 1; slot++) { | |
501 | btrfs_item_key_to_cpu(leaf, &leaf_key, slot); | |
502 | btrfs_item_key_to_cpu(leaf, &key, slot + 1); | |
503 | ||
504 | /* Make sure the keys are in the right order */ | |
505 | if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { | |
506 | CORRUPT("bad key order", leaf, root, slot); | |
507 | return -EIO; | |
508 | } | |
509 | ||
510 | /* | |
511 | * Make sure the offset and ends are right, remember that the | |
512 | * item data starts at the end of the leaf and grows towards the | |
513 | * front. | |
514 | */ | |
515 | if (btrfs_item_offset_nr(leaf, slot) != | |
516 | btrfs_item_end_nr(leaf, slot + 1)) { | |
517 | CORRUPT("slot offset bad", leaf, root, slot); | |
518 | return -EIO; | |
519 | } | |
520 | ||
521 | /* | |
522 | * Check to make sure that we don't point outside of the leaf, | |
523 | * just incase all the items are consistent to eachother, but | |
524 | * all point outside of the leaf. | |
525 | */ | |
526 | if (btrfs_item_end_nr(leaf, slot) > | |
527 | BTRFS_LEAF_DATA_SIZE(root)) { | |
528 | CORRUPT("slot end outside of leaf", leaf, root, slot); | |
529 | return -EIO; | |
530 | } | |
531 | } | |
532 | ||
533 | return 0; | |
534 | } | |
535 | ||
727011e0 CM |
536 | struct extent_buffer *find_eb_for_page(struct extent_io_tree *tree, |
537 | struct page *page, int max_walk) | |
538 | { | |
539 | struct extent_buffer *eb; | |
540 | u64 start = page_offset(page); | |
541 | u64 target = start; | |
542 | u64 min_start; | |
543 | ||
544 | if (start < max_walk) | |
545 | min_start = 0; | |
546 | else | |
547 | min_start = start - max_walk; | |
548 | ||
549 | while (start >= min_start) { | |
550 | eb = find_extent_buffer(tree, start, 0); | |
551 | if (eb) { | |
552 | /* | |
553 | * we found an extent buffer and it contains our page | |
554 | * horray! | |
555 | */ | |
556 | if (eb->start <= target && | |
557 | eb->start + eb->len > target) | |
558 | return eb; | |
559 | ||
560 | /* we found an extent buffer that wasn't for us */ | |
561 | free_extent_buffer(eb); | |
562 | return NULL; | |
563 | } | |
564 | if (start == 0) | |
565 | break; | |
566 | start -= PAGE_CACHE_SIZE; | |
567 | } | |
568 | return NULL; | |
569 | } | |
570 | ||
b2950863 | 571 | static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
5cf1ab56 | 572 | struct extent_state *state, int mirror) |
ce9adaa5 CM |
573 | { |
574 | struct extent_io_tree *tree; | |
575 | u64 found_start; | |
576 | int found_level; | |
ce9adaa5 CM |
577 | struct extent_buffer *eb; |
578 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 579 | int ret = 0; |
727011e0 | 580 | int reads_done; |
ce9adaa5 | 581 | |
ce9adaa5 CM |
582 | if (!page->private) |
583 | goto out; | |
d397712b | 584 | |
727011e0 | 585 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
4f2de97a | 586 | eb = (struct extent_buffer *)page->private; |
d397712b | 587 | |
0b32f4bb JB |
588 | /* the pending IO might have been the only thing that kept this buffer |
589 | * in memory. Make sure we have a ref for all this other checks | |
590 | */ | |
591 | extent_buffer_get(eb); | |
592 | ||
593 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
727011e0 CM |
594 | if (!reads_done) |
595 | goto err; | |
f188591e | 596 | |
5cf1ab56 | 597 | eb->read_mirror = mirror; |
ea466794 JB |
598 | if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) { |
599 | ret = -EIO; | |
600 | goto err; | |
601 | } | |
602 | ||
ce9adaa5 | 603 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 604 | if (found_start != eb->start) { |
7a36ddec | 605 | printk_ratelimited(KERN_INFO "btrfs bad tree block start " |
193f284d CM |
606 | "%llu %llu\n", |
607 | (unsigned long long)found_start, | |
608 | (unsigned long long)eb->start); | |
f188591e | 609 | ret = -EIO; |
ce9adaa5 CM |
610 | goto err; |
611 | } | |
2b82032c | 612 | if (check_tree_block_fsid(root, eb)) { |
7a36ddec | 613 | printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n", |
193f284d | 614 | (unsigned long long)eb->start); |
1259ab75 CM |
615 | ret = -EIO; |
616 | goto err; | |
617 | } | |
ce9adaa5 CM |
618 | found_level = btrfs_header_level(eb); |
619 | ||
85d4e461 CM |
620 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
621 | eb, found_level); | |
4008c04a | 622 | |
ce9adaa5 | 623 | ret = csum_tree_block(root, eb, 1); |
a826d6dc | 624 | if (ret) { |
f188591e | 625 | ret = -EIO; |
a826d6dc JB |
626 | goto err; |
627 | } | |
628 | ||
629 | /* | |
630 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
631 | * that we don't try and read the other copies of this block, just | |
632 | * return -EIO. | |
633 | */ | |
634 | if (found_level == 0 && check_leaf(root, eb)) { | |
635 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); | |
636 | ret = -EIO; | |
637 | } | |
ce9adaa5 | 638 | |
0b32f4bb JB |
639 | if (!ret) |
640 | set_extent_buffer_uptodate(eb); | |
ce9adaa5 | 641 | err: |
4bb31e92 AJ |
642 | if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) { |
643 | clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags); | |
644 | btree_readahead_hook(root, eb, eb->start, ret); | |
645 | } | |
646 | ||
0b32f4bb JB |
647 | if (ret) |
648 | clear_extent_buffer_uptodate(eb); | |
649 | free_extent_buffer(eb); | |
ce9adaa5 | 650 | out: |
f188591e | 651 | return ret; |
ce9adaa5 CM |
652 | } |
653 | ||
ea466794 | 654 | static int btree_io_failed_hook(struct page *page, int failed_mirror) |
4bb31e92 | 655 | { |
4bb31e92 AJ |
656 | struct extent_buffer *eb; |
657 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
658 | ||
4f2de97a | 659 | eb = (struct extent_buffer *)page->private; |
ea466794 | 660 | set_bit(EXTENT_BUFFER_IOERR, &eb->bflags); |
5cf1ab56 | 661 | eb->read_mirror = failed_mirror; |
ea466794 | 662 | if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) |
4bb31e92 | 663 | btree_readahead_hook(root, eb, eb->start, -EIO); |
4bb31e92 AJ |
664 | return -EIO; /* we fixed nothing */ |
665 | } | |
666 | ||
ce9adaa5 | 667 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
668 | { |
669 | struct end_io_wq *end_io_wq = bio->bi_private; | |
670 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 671 | |
ce9adaa5 | 672 | fs_info = end_io_wq->info; |
ce9adaa5 | 673 | end_io_wq->error = err; |
8b712842 CM |
674 | end_io_wq->work.func = end_workqueue_fn; |
675 | end_io_wq->work.flags = 0; | |
d20f7043 | 676 | |
7b6d91da | 677 | if (bio->bi_rw & REQ_WRITE) { |
0cb59c99 | 678 | if (end_io_wq->metadata == 1) |
cad321ad CM |
679 | btrfs_queue_worker(&fs_info->endio_meta_write_workers, |
680 | &end_io_wq->work); | |
0cb59c99 JB |
681 | else if (end_io_wq->metadata == 2) |
682 | btrfs_queue_worker(&fs_info->endio_freespace_worker, | |
683 | &end_io_wq->work); | |
cad321ad CM |
684 | else |
685 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
686 | &end_io_wq->work); | |
d20f7043 CM |
687 | } else { |
688 | if (end_io_wq->metadata) | |
689 | btrfs_queue_worker(&fs_info->endio_meta_workers, | |
690 | &end_io_wq->work); | |
691 | else | |
692 | btrfs_queue_worker(&fs_info->endio_workers, | |
693 | &end_io_wq->work); | |
694 | } | |
ce9adaa5 CM |
695 | } |
696 | ||
0cb59c99 JB |
697 | /* |
698 | * For the metadata arg you want | |
699 | * | |
700 | * 0 - if data | |
701 | * 1 - if normal metadta | |
702 | * 2 - if writing to the free space cache area | |
703 | */ | |
22c59948 CM |
704 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
705 | int metadata) | |
0b86a832 | 706 | { |
ce9adaa5 | 707 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
708 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
709 | if (!end_io_wq) | |
710 | return -ENOMEM; | |
711 | ||
712 | end_io_wq->private = bio->bi_private; | |
713 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 714 | end_io_wq->info = info; |
ce9adaa5 CM |
715 | end_io_wq->error = 0; |
716 | end_io_wq->bio = bio; | |
22c59948 | 717 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
718 | |
719 | bio->bi_private = end_io_wq; | |
720 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
721 | return 0; |
722 | } | |
723 | ||
b64a2851 | 724 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 725 | { |
4854ddd0 CM |
726 | unsigned long limit = min_t(unsigned long, |
727 | info->workers.max_workers, | |
728 | info->fs_devices->open_devices); | |
729 | return 256 * limit; | |
730 | } | |
0986fe9e | 731 | |
4a69a410 CM |
732 | static void run_one_async_start(struct btrfs_work *work) |
733 | { | |
4a69a410 | 734 | struct async_submit_bio *async; |
79787eaa | 735 | int ret; |
4a69a410 CM |
736 | |
737 | async = container_of(work, struct async_submit_bio, work); | |
79787eaa JM |
738 | ret = async->submit_bio_start(async->inode, async->rw, async->bio, |
739 | async->mirror_num, async->bio_flags, | |
740 | async->bio_offset); | |
741 | if (ret) | |
742 | async->error = ret; | |
4a69a410 CM |
743 | } |
744 | ||
745 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
746 | { |
747 | struct btrfs_fs_info *fs_info; | |
748 | struct async_submit_bio *async; | |
4854ddd0 | 749 | int limit; |
8b712842 CM |
750 | |
751 | async = container_of(work, struct async_submit_bio, work); | |
752 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 753 | |
b64a2851 | 754 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
755 | limit = limit * 2 / 3; |
756 | ||
8b712842 | 757 | atomic_dec(&fs_info->nr_async_submits); |
0986fe9e | 758 | |
b64a2851 CM |
759 | if (atomic_read(&fs_info->nr_async_submits) < limit && |
760 | waitqueue_active(&fs_info->async_submit_wait)) | |
4854ddd0 CM |
761 | wake_up(&fs_info->async_submit_wait); |
762 | ||
79787eaa JM |
763 | /* If an error occured we just want to clean up the bio and move on */ |
764 | if (async->error) { | |
765 | bio_endio(async->bio, async->error); | |
766 | return; | |
767 | } | |
768 | ||
4a69a410 | 769 | async->submit_bio_done(async->inode, async->rw, async->bio, |
eaf25d93 CM |
770 | async->mirror_num, async->bio_flags, |
771 | async->bio_offset); | |
4a69a410 CM |
772 | } |
773 | ||
774 | static void run_one_async_free(struct btrfs_work *work) | |
775 | { | |
776 | struct async_submit_bio *async; | |
777 | ||
778 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
779 | kfree(async); |
780 | } | |
781 | ||
44b8bd7e CM |
782 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
783 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 784 | unsigned long bio_flags, |
eaf25d93 | 785 | u64 bio_offset, |
4a69a410 CM |
786 | extent_submit_bio_hook_t *submit_bio_start, |
787 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
788 | { |
789 | struct async_submit_bio *async; | |
790 | ||
791 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
792 | if (!async) | |
793 | return -ENOMEM; | |
794 | ||
795 | async->inode = inode; | |
796 | async->rw = rw; | |
797 | async->bio = bio; | |
798 | async->mirror_num = mirror_num; | |
4a69a410 CM |
799 | async->submit_bio_start = submit_bio_start; |
800 | async->submit_bio_done = submit_bio_done; | |
801 | ||
802 | async->work.func = run_one_async_start; | |
803 | async->work.ordered_func = run_one_async_done; | |
804 | async->work.ordered_free = run_one_async_free; | |
805 | ||
8b712842 | 806 | async->work.flags = 0; |
c8b97818 | 807 | async->bio_flags = bio_flags; |
eaf25d93 | 808 | async->bio_offset = bio_offset; |
8c8bee1d | 809 | |
79787eaa JM |
810 | async->error = 0; |
811 | ||
cb03c743 | 812 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 | 813 | |
7b6d91da | 814 | if (rw & REQ_SYNC) |
d313d7a3 CM |
815 | btrfs_set_work_high_prio(&async->work); |
816 | ||
8b712842 | 817 | btrfs_queue_worker(&fs_info->workers, &async->work); |
9473f16c | 818 | |
d397712b | 819 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
820 | atomic_read(&fs_info->nr_async_submits)) { |
821 | wait_event(fs_info->async_submit_wait, | |
822 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
823 | } | |
824 | ||
44b8bd7e CM |
825 | return 0; |
826 | } | |
827 | ||
ce3ed71a CM |
828 | static int btree_csum_one_bio(struct bio *bio) |
829 | { | |
830 | struct bio_vec *bvec = bio->bi_io_vec; | |
831 | int bio_index = 0; | |
832 | struct btrfs_root *root; | |
79787eaa | 833 | int ret = 0; |
ce3ed71a CM |
834 | |
835 | WARN_ON(bio->bi_vcnt <= 0); | |
d397712b | 836 | while (bio_index < bio->bi_vcnt) { |
ce3ed71a | 837 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
79787eaa JM |
838 | ret = csum_dirty_buffer(root, bvec->bv_page); |
839 | if (ret) | |
840 | break; | |
ce3ed71a CM |
841 | bio_index++; |
842 | bvec++; | |
843 | } | |
79787eaa | 844 | return ret; |
ce3ed71a CM |
845 | } |
846 | ||
4a69a410 CM |
847 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
848 | struct bio *bio, int mirror_num, | |
eaf25d93 CM |
849 | unsigned long bio_flags, |
850 | u64 bio_offset) | |
22c59948 | 851 | { |
8b712842 CM |
852 | /* |
853 | * when we're called for a write, we're already in the async | |
5443be45 | 854 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 855 | */ |
79787eaa | 856 | return btree_csum_one_bio(bio); |
4a69a410 | 857 | } |
22c59948 | 858 | |
4a69a410 | 859 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
860 | int mirror_num, unsigned long bio_flags, |
861 | u64 bio_offset) | |
4a69a410 | 862 | { |
8b712842 | 863 | /* |
4a69a410 CM |
864 | * when we're called for a write, we're already in the async |
865 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 866 | */ |
8b712842 | 867 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
868 | } |
869 | ||
44b8bd7e | 870 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
871 | int mirror_num, unsigned long bio_flags, |
872 | u64 bio_offset) | |
44b8bd7e | 873 | { |
cad321ad CM |
874 | int ret; |
875 | ||
7b6d91da | 876 | if (!(rw & REQ_WRITE)) { |
f3f266ab | 877 | |
4a69a410 CM |
878 | /* |
879 | * called for a read, do the setup so that checksum validation | |
880 | * can happen in the async kernel threads | |
881 | */ | |
f3f266ab CM |
882 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, |
883 | bio, 1); | |
1d4284bd CM |
884 | if (ret) |
885 | return ret; | |
4a69a410 | 886 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, |
6f3577bd | 887 | mirror_num, 0); |
44b8bd7e | 888 | } |
d313d7a3 | 889 | |
cad321ad CM |
890 | /* |
891 | * kthread helpers are used to submit writes so that checksumming | |
892 | * can happen in parallel across all CPUs | |
893 | */ | |
44b8bd7e | 894 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, |
c8b97818 | 895 | inode, rw, bio, mirror_num, 0, |
eaf25d93 | 896 | bio_offset, |
4a69a410 CM |
897 | __btree_submit_bio_start, |
898 | __btree_submit_bio_done); | |
44b8bd7e CM |
899 | } |
900 | ||
3dd1462e | 901 | #ifdef CONFIG_MIGRATION |
784b4e29 | 902 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
903 | struct page *newpage, struct page *page, |
904 | enum migrate_mode mode) | |
784b4e29 CM |
905 | { |
906 | /* | |
907 | * we can't safely write a btree page from here, | |
908 | * we haven't done the locking hook | |
909 | */ | |
910 | if (PageDirty(page)) | |
911 | return -EAGAIN; | |
912 | /* | |
913 | * Buffers may be managed in a filesystem specific way. | |
914 | * We must have no buffers or drop them. | |
915 | */ | |
916 | if (page_has_private(page) && | |
917 | !try_to_release_page(page, GFP_KERNEL)) | |
918 | return -EAGAIN; | |
a6bc32b8 | 919 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 920 | } |
3dd1462e | 921 | #endif |
784b4e29 | 922 | |
0da5468f CM |
923 | |
924 | static int btree_writepages(struct address_space *mapping, | |
925 | struct writeback_control *wbc) | |
926 | { | |
d1310b2e CM |
927 | struct extent_io_tree *tree; |
928 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 929 | if (wbc->sync_mode == WB_SYNC_NONE) { |
b9473439 | 930 | struct btrfs_root *root = BTRFS_I(mapping->host)->root; |
793955bc | 931 | u64 num_dirty; |
24ab9cd8 | 932 | unsigned long thresh = 32 * 1024 * 1024; |
448d640b CM |
933 | |
934 | if (wbc->for_kupdate) | |
935 | return 0; | |
936 | ||
b9473439 CM |
937 | /* this is a bit racy, but that's ok */ |
938 | num_dirty = root->fs_info->dirty_metadata_bytes; | |
d397712b | 939 | if (num_dirty < thresh) |
793955bc | 940 | return 0; |
793955bc | 941 | } |
0b32f4bb | 942 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
943 | } |
944 | ||
b2950863 | 945 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 946 | { |
d1310b2e CM |
947 | struct extent_io_tree *tree; |
948 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 949 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 950 | } |
22b0ebda | 951 | |
70dec807 | 952 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 953 | { |
98509cfc | 954 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 955 | return 0; |
0c4e538b DS |
956 | /* |
957 | * We need to mask out eg. __GFP_HIGHMEM and __GFP_DMA32 as we're doing | |
958 | * slab allocation from alloc_extent_state down the callchain where | |
959 | * it'd hit a BUG_ON as those flags are not allowed. | |
960 | */ | |
961 | gfp_flags &= ~GFP_SLAB_BUG_MASK; | |
962 | ||
3083ee2e | 963 | return try_release_extent_buffer(page, gfp_flags); |
d98237b3 CM |
964 | } |
965 | ||
5f39d397 | 966 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 967 | { |
d1310b2e CM |
968 | struct extent_io_tree *tree; |
969 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
970 | extent_invalidatepage(tree, page, offset); |
971 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 972 | if (PagePrivate(page)) { |
d397712b CM |
973 | printk(KERN_WARNING "btrfs warning page private not zero " |
974 | "on page %llu\n", (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
975 | ClearPagePrivate(page); |
976 | set_page_private(page, 0); | |
977 | page_cache_release(page); | |
978 | } | |
d98237b3 CM |
979 | } |
980 | ||
0b32f4bb JB |
981 | static int btree_set_page_dirty(struct page *page) |
982 | { | |
983 | struct extent_buffer *eb; | |
984 | ||
985 | BUG_ON(!PagePrivate(page)); | |
986 | eb = (struct extent_buffer *)page->private; | |
987 | BUG_ON(!eb); | |
988 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
989 | BUG_ON(!atomic_read(&eb->refs)); | |
990 | btrfs_assert_tree_locked(eb); | |
991 | return __set_page_dirty_nobuffers(page); | |
992 | } | |
993 | ||
7f09410b | 994 | static const struct address_space_operations btree_aops = { |
d98237b3 | 995 | .readpage = btree_readpage, |
0da5468f | 996 | .writepages = btree_writepages, |
5f39d397 CM |
997 | .releasepage = btree_releasepage, |
998 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 999 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1000 | .migratepage = btree_migratepage, |
5a92bc88 | 1001 | #endif |
0b32f4bb | 1002 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1003 | }; |
1004 | ||
ca7a79ad CM |
1005 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1006 | u64 parent_transid) | |
090d1875 | 1007 | { |
5f39d397 CM |
1008 | struct extent_buffer *buf = NULL; |
1009 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 1010 | int ret = 0; |
090d1875 | 1011 | |
db94535d | 1012 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 1013 | if (!buf) |
090d1875 | 1014 | return 0; |
d1310b2e | 1015 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
bb82ab88 | 1016 | buf, 0, WAIT_NONE, btree_get_extent, 0); |
5f39d397 | 1017 | free_extent_buffer(buf); |
de428b63 | 1018 | return ret; |
090d1875 CM |
1019 | } |
1020 | ||
ab0fff03 AJ |
1021 | int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1022 | int mirror_num, struct extent_buffer **eb) | |
1023 | { | |
1024 | struct extent_buffer *buf = NULL; | |
1025 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1026 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; | |
1027 | int ret; | |
1028 | ||
1029 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
1030 | if (!buf) | |
1031 | return 0; | |
1032 | ||
1033 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1034 | ||
1035 | ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, | |
1036 | btree_get_extent, mirror_num); | |
1037 | if (ret) { | |
1038 | free_extent_buffer(buf); | |
1039 | return ret; | |
1040 | } | |
1041 | ||
1042 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1043 | free_extent_buffer(buf); | |
1044 | return -EIO; | |
0b32f4bb | 1045 | } else if (extent_buffer_uptodate(buf)) { |
ab0fff03 AJ |
1046 | *eb = buf; |
1047 | } else { | |
1048 | free_extent_buffer(buf); | |
1049 | } | |
1050 | return 0; | |
1051 | } | |
1052 | ||
0999df54 CM |
1053 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
1054 | u64 bytenr, u32 blocksize) | |
1055 | { | |
1056 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1057 | struct extent_buffer *eb; | |
1058 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
f09d1f60 | 1059 | bytenr, blocksize); |
0999df54 CM |
1060 | return eb; |
1061 | } | |
1062 | ||
1063 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
1064 | u64 bytenr, u32 blocksize) | |
1065 | { | |
1066 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1067 | struct extent_buffer *eb; | |
1068 | ||
1069 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
727011e0 | 1070 | bytenr, blocksize); |
0999df54 CM |
1071 | return eb; |
1072 | } | |
1073 | ||
1074 | ||
e02119d5 CM |
1075 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1076 | { | |
727011e0 | 1077 | return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, |
8aa38c31 | 1078 | buf->start + buf->len - 1); |
e02119d5 CM |
1079 | } |
1080 | ||
1081 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
1082 | { | |
727011e0 | 1083 | return filemap_fdatawait_range(buf->pages[0]->mapping, |
8aa38c31 | 1084 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
1085 | } |
1086 | ||
0999df54 | 1087 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 1088 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
1089 | { |
1090 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1091 | int ret; |
1092 | ||
0999df54 CM |
1093 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
1094 | if (!buf) | |
1095 | return NULL; | |
0999df54 | 1096 | |
ca7a79ad | 1097 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
5f39d397 | 1098 | return buf; |
ce9adaa5 | 1099 | |
eb60ceac CM |
1100 | } |
1101 | ||
d5c13f92 JM |
1102 | void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
1103 | struct extent_buffer *buf) | |
ed2ff2cb | 1104 | { |
55c69072 | 1105 | if (btrfs_header_generation(buf) == |
925baedd | 1106 | root->fs_info->running_transaction->transid) { |
b9447ef8 | 1107 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1108 | |
b9473439 CM |
1109 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
1110 | spin_lock(&root->fs_info->delalloc_lock); | |
1111 | if (root->fs_info->dirty_metadata_bytes >= buf->len) | |
1112 | root->fs_info->dirty_metadata_bytes -= buf->len; | |
d5c13f92 JM |
1113 | else { |
1114 | spin_unlock(&root->fs_info->delalloc_lock); | |
1115 | btrfs_panic(root->fs_info, -EOVERFLOW, | |
1116 | "Can't clear %lu bytes from " | |
1117 | " dirty_mdatadata_bytes (%lu)", | |
1118 | buf->len, | |
1119 | root->fs_info->dirty_metadata_bytes); | |
1120 | } | |
b9473439 CM |
1121 | spin_unlock(&root->fs_info->delalloc_lock); |
1122 | } | |
b4ce94de | 1123 | |
b9473439 CM |
1124 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1125 | btrfs_set_lock_blocking(buf); | |
0b32f4bb | 1126 | clear_extent_buffer_dirty(buf); |
925baedd | 1127 | } |
5f39d397 CM |
1128 | } |
1129 | ||
143bede5 JM |
1130 | static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
1131 | u32 stripesize, struct btrfs_root *root, | |
1132 | struct btrfs_fs_info *fs_info, | |
1133 | u64 objectid) | |
d97e63b6 | 1134 | { |
cfaa7295 | 1135 | root->node = NULL; |
a28ec197 | 1136 | root->commit_root = NULL; |
db94535d CM |
1137 | root->sectorsize = sectorsize; |
1138 | root->nodesize = nodesize; | |
1139 | root->leafsize = leafsize; | |
87ee04eb | 1140 | root->stripesize = stripesize; |
123abc88 | 1141 | root->ref_cows = 0; |
0b86a832 | 1142 | root->track_dirty = 0; |
c71bf099 | 1143 | root->in_radix = 0; |
d68fc57b YZ |
1144 | root->orphan_item_inserted = 0; |
1145 | root->orphan_cleanup_state = 0; | |
0b86a832 | 1146 | |
0f7d52f4 CM |
1147 | root->objectid = objectid; |
1148 | root->last_trans = 0; | |
13a8a7c8 | 1149 | root->highest_objectid = 0; |
58176a96 | 1150 | root->name = NULL; |
6bef4d31 | 1151 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1152 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1153 | root->block_rsv = NULL; |
d68fc57b | 1154 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
1155 | |
1156 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1157 | INIT_LIST_HEAD(&root->root_list); |
d68fc57b | 1158 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 1159 | spin_lock_init(&root->inode_lock); |
f0486c68 | 1160 | spin_lock_init(&root->accounting_lock); |
a2135011 | 1161 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1162 | mutex_init(&root->log_mutex); |
7237f183 YZ |
1163 | init_waitqueue_head(&root->log_writer_wait); |
1164 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1165 | init_waitqueue_head(&root->log_commit_wait[1]); | |
1166 | atomic_set(&root->log_commit[0], 0); | |
1167 | atomic_set(&root->log_commit[1], 0); | |
1168 | atomic_set(&root->log_writers, 0); | |
8a35d95f | 1169 | atomic_set(&root->orphan_inodes, 0); |
7237f183 YZ |
1170 | root->log_batch = 0; |
1171 | root->log_transid = 0; | |
257c62e1 | 1172 | root->last_log_commit = 0; |
d0c803c4 | 1173 | extent_io_tree_init(&root->dirty_log_pages, |
f993c883 | 1174 | fs_info->btree_inode->i_mapping); |
017e5369 | 1175 | |
3768f368 CM |
1176 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1177 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1178 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 1179 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 1180 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 1181 | init_completion(&root->kobj_unregister); |
6702ed49 | 1182 | root->defrag_running = 0; |
4d775673 | 1183 | root->root_key.objectid = objectid; |
0ee5dc67 | 1184 | root->anon_dev = 0; |
8ea05e3a AB |
1185 | |
1186 | spin_lock_init(&root->root_times_lock); | |
3768f368 CM |
1187 | } |
1188 | ||
200a5c17 JM |
1189 | static int __must_check find_and_setup_root(struct btrfs_root *tree_root, |
1190 | struct btrfs_fs_info *fs_info, | |
1191 | u64 objectid, | |
1192 | struct btrfs_root *root) | |
3768f368 CM |
1193 | { |
1194 | int ret; | |
db94535d | 1195 | u32 blocksize; |
84234f3a | 1196 | u64 generation; |
3768f368 | 1197 | |
db94535d | 1198 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1199 | tree_root->sectorsize, tree_root->stripesize, |
1200 | root, fs_info, objectid); | |
3768f368 CM |
1201 | ret = btrfs_find_last_root(tree_root, objectid, |
1202 | &root->root_item, &root->root_key); | |
4df27c4d YZ |
1203 | if (ret > 0) |
1204 | return -ENOENT; | |
200a5c17 JM |
1205 | else if (ret < 0) |
1206 | return ret; | |
3768f368 | 1207 | |
84234f3a | 1208 | generation = btrfs_root_generation(&root->root_item); |
db94535d | 1209 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
af31f5e5 | 1210 | root->commit_root = NULL; |
db94535d | 1211 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), |
84234f3a | 1212 | blocksize, generation); |
b9fab919 | 1213 | if (!root->node || !btrfs_buffer_uptodate(root->node, generation, 0)) { |
68433b73 | 1214 | free_extent_buffer(root->node); |
af31f5e5 | 1215 | root->node = NULL; |
68433b73 CM |
1216 | return -EIO; |
1217 | } | |
4df27c4d | 1218 | root->commit_root = btrfs_root_node(root); |
d97e63b6 CM |
1219 | return 0; |
1220 | } | |
1221 | ||
f84a8bd6 | 1222 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info) |
6f07e42e AV |
1223 | { |
1224 | struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS); | |
1225 | if (root) | |
1226 | root->fs_info = fs_info; | |
1227 | return root; | |
1228 | } | |
1229 | ||
20897f5c AJ |
1230 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
1231 | struct btrfs_fs_info *fs_info, | |
1232 | u64 objectid) | |
1233 | { | |
1234 | struct extent_buffer *leaf; | |
1235 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1236 | struct btrfs_root *root; | |
1237 | struct btrfs_key key; | |
1238 | int ret = 0; | |
1239 | u64 bytenr; | |
1240 | ||
1241 | root = btrfs_alloc_root(fs_info); | |
1242 | if (!root) | |
1243 | return ERR_PTR(-ENOMEM); | |
1244 | ||
1245 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1246 | tree_root->sectorsize, tree_root->stripesize, | |
1247 | root, fs_info, objectid); | |
1248 | root->root_key.objectid = objectid; | |
1249 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1250 | root->root_key.offset = 0; | |
1251 | ||
1252 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, | |
1253 | 0, objectid, NULL, 0, 0, 0); | |
1254 | if (IS_ERR(leaf)) { | |
1255 | ret = PTR_ERR(leaf); | |
1256 | goto fail; | |
1257 | } | |
1258 | ||
1259 | bytenr = leaf->start; | |
1260 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); | |
1261 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1262 | btrfs_set_header_generation(leaf, trans->transid); | |
1263 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1264 | btrfs_set_header_owner(leaf, objectid); | |
1265 | root->node = leaf; | |
1266 | ||
1267 | write_extent_buffer(leaf, fs_info->fsid, | |
1268 | (unsigned long)btrfs_header_fsid(leaf), | |
1269 | BTRFS_FSID_SIZE); | |
1270 | write_extent_buffer(leaf, fs_info->chunk_tree_uuid, | |
1271 | (unsigned long)btrfs_header_chunk_tree_uuid(leaf), | |
1272 | BTRFS_UUID_SIZE); | |
1273 | btrfs_mark_buffer_dirty(leaf); | |
1274 | ||
1275 | root->commit_root = btrfs_root_node(root); | |
1276 | root->track_dirty = 1; | |
1277 | ||
1278 | ||
1279 | root->root_item.flags = 0; | |
1280 | root->root_item.byte_limit = 0; | |
1281 | btrfs_set_root_bytenr(&root->root_item, leaf->start); | |
1282 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1283 | btrfs_set_root_level(&root->root_item, 0); | |
1284 | btrfs_set_root_refs(&root->root_item, 1); | |
1285 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1286 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1287 | btrfs_set_root_dirid(&root->root_item, 0); | |
1288 | root->root_item.drop_level = 0; | |
1289 | ||
1290 | key.objectid = objectid; | |
1291 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1292 | key.offset = 0; | |
1293 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1294 | if (ret) | |
1295 | goto fail; | |
1296 | ||
1297 | btrfs_tree_unlock(leaf); | |
1298 | ||
1299 | fail: | |
1300 | if (ret) | |
1301 | return ERR_PTR(ret); | |
1302 | ||
1303 | return root; | |
1304 | } | |
1305 | ||
7237f183 YZ |
1306 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1307 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1308 | { |
1309 | struct btrfs_root *root; | |
1310 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1311 | struct extent_buffer *leaf; |
e02119d5 | 1312 | |
6f07e42e | 1313 | root = btrfs_alloc_root(fs_info); |
e02119d5 | 1314 | if (!root) |
7237f183 | 1315 | return ERR_PTR(-ENOMEM); |
e02119d5 CM |
1316 | |
1317 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1318 | tree_root->sectorsize, tree_root->stripesize, | |
1319 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1320 | ||
1321 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1322 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1323 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
7237f183 YZ |
1324 | /* |
1325 | * log trees do not get reference counted because they go away | |
1326 | * before a real commit is actually done. They do store pointers | |
1327 | * to file data extents, and those reference counts still get | |
1328 | * updated (along with back refs to the log tree). | |
1329 | */ | |
e02119d5 CM |
1330 | root->ref_cows = 0; |
1331 | ||
5d4f98a2 | 1332 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, |
66d7e7f0 | 1333 | BTRFS_TREE_LOG_OBJECTID, NULL, |
5581a51a | 1334 | 0, 0, 0); |
7237f183 YZ |
1335 | if (IS_ERR(leaf)) { |
1336 | kfree(root); | |
1337 | return ERR_CAST(leaf); | |
1338 | } | |
e02119d5 | 1339 | |
5d4f98a2 YZ |
1340 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1341 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1342 | btrfs_set_header_generation(leaf, trans->transid); | |
1343 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1344 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1345 | root->node = leaf; |
e02119d5 CM |
1346 | |
1347 | write_extent_buffer(root->node, root->fs_info->fsid, | |
1348 | (unsigned long)btrfs_header_fsid(root->node), | |
1349 | BTRFS_FSID_SIZE); | |
1350 | btrfs_mark_buffer_dirty(root->node); | |
1351 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1352 | return root; |
1353 | } | |
1354 | ||
1355 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1356 | struct btrfs_fs_info *fs_info) | |
1357 | { | |
1358 | struct btrfs_root *log_root; | |
1359 | ||
1360 | log_root = alloc_log_tree(trans, fs_info); | |
1361 | if (IS_ERR(log_root)) | |
1362 | return PTR_ERR(log_root); | |
1363 | WARN_ON(fs_info->log_root_tree); | |
1364 | fs_info->log_root_tree = log_root; | |
1365 | return 0; | |
1366 | } | |
1367 | ||
1368 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1369 | struct btrfs_root *root) | |
1370 | { | |
1371 | struct btrfs_root *log_root; | |
1372 | struct btrfs_inode_item *inode_item; | |
1373 | ||
1374 | log_root = alloc_log_tree(trans, root->fs_info); | |
1375 | if (IS_ERR(log_root)) | |
1376 | return PTR_ERR(log_root); | |
1377 | ||
1378 | log_root->last_trans = trans->transid; | |
1379 | log_root->root_key.offset = root->root_key.objectid; | |
1380 | ||
1381 | inode_item = &log_root->root_item.inode; | |
1382 | inode_item->generation = cpu_to_le64(1); | |
1383 | inode_item->size = cpu_to_le64(3); | |
1384 | inode_item->nlink = cpu_to_le32(1); | |
1385 | inode_item->nbytes = cpu_to_le64(root->leafsize); | |
1386 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); | |
1387 | ||
5d4f98a2 | 1388 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1389 | |
1390 | WARN_ON(root->log_root); | |
1391 | root->log_root = log_root; | |
1392 | root->log_transid = 0; | |
257c62e1 | 1393 | root->last_log_commit = 0; |
e02119d5 CM |
1394 | return 0; |
1395 | } | |
1396 | ||
1397 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
1398 | struct btrfs_key *location) | |
1399 | { | |
1400 | struct btrfs_root *root; | |
1401 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1402 | struct btrfs_path *path; |
5f39d397 | 1403 | struct extent_buffer *l; |
84234f3a | 1404 | u64 generation; |
db94535d | 1405 | u32 blocksize; |
0f7d52f4 | 1406 | int ret = 0; |
8ea05e3a | 1407 | int slot; |
0f7d52f4 | 1408 | |
6f07e42e | 1409 | root = btrfs_alloc_root(fs_info); |
0cf6c620 | 1410 | if (!root) |
0f7d52f4 | 1411 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 1412 | if (location->offset == (u64)-1) { |
db94535d | 1413 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
1414 | location->objectid, root); |
1415 | if (ret) { | |
0f7d52f4 CM |
1416 | kfree(root); |
1417 | return ERR_PTR(ret); | |
1418 | } | |
13a8a7c8 | 1419 | goto out; |
0f7d52f4 CM |
1420 | } |
1421 | ||
db94535d | 1422 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1423 | tree_root->sectorsize, tree_root->stripesize, |
1424 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
1425 | |
1426 | path = btrfs_alloc_path(); | |
db5b493a TI |
1427 | if (!path) { |
1428 | kfree(root); | |
1429 | return ERR_PTR(-ENOMEM); | |
1430 | } | |
0f7d52f4 | 1431 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); |
13a8a7c8 YZ |
1432 | if (ret == 0) { |
1433 | l = path->nodes[0]; | |
8ea05e3a AB |
1434 | slot = path->slots[0]; |
1435 | btrfs_read_root_item(tree_root, l, slot, &root->root_item); | |
13a8a7c8 | 1436 | memcpy(&root->root_key, location, sizeof(*location)); |
0f7d52f4 | 1437 | } |
0f7d52f4 CM |
1438 | btrfs_free_path(path); |
1439 | if (ret) { | |
5e540f77 | 1440 | kfree(root); |
13a8a7c8 YZ |
1441 | if (ret > 0) |
1442 | ret = -ENOENT; | |
0f7d52f4 CM |
1443 | return ERR_PTR(ret); |
1444 | } | |
13a8a7c8 | 1445 | |
84234f3a | 1446 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1447 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1448 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1449 | blocksize, generation); |
5d4f98a2 | 1450 | root->commit_root = btrfs_root_node(root); |
79787eaa | 1451 | BUG_ON(!root->node); /* -ENOMEM */ |
13a8a7c8 | 1452 | out: |
08fe4db1 | 1453 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { |
e02119d5 | 1454 | root->ref_cows = 1; |
08fe4db1 LZ |
1455 | btrfs_check_and_init_root_item(&root->root_item); |
1456 | } | |
13a8a7c8 | 1457 | |
5eda7b5e CM |
1458 | return root; |
1459 | } | |
1460 | ||
edbd8d4e CM |
1461 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
1462 | struct btrfs_key *location) | |
5eda7b5e CM |
1463 | { |
1464 | struct btrfs_root *root; | |
1465 | int ret; | |
1466 | ||
edbd8d4e CM |
1467 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1468 | return fs_info->tree_root; | |
1469 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1470 | return fs_info->extent_root; | |
8f18cf13 CM |
1471 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1472 | return fs_info->chunk_root; | |
1473 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1474 | return fs_info->dev_root; | |
0403e47e YZ |
1475 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1476 | return fs_info->csum_root; | |
bcef60f2 AJ |
1477 | if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1478 | return fs_info->quota_root ? fs_info->quota_root : | |
1479 | ERR_PTR(-ENOENT); | |
4df27c4d YZ |
1480 | again: |
1481 | spin_lock(&fs_info->fs_roots_radix_lock); | |
5eda7b5e CM |
1482 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
1483 | (unsigned long)location->objectid); | |
4df27c4d | 1484 | spin_unlock(&fs_info->fs_roots_radix_lock); |
5eda7b5e CM |
1485 | if (root) |
1486 | return root; | |
1487 | ||
e02119d5 | 1488 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
1489 | if (IS_ERR(root)) |
1490 | return root; | |
3394e160 | 1491 | |
581bb050 | 1492 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); |
581bb050 LZ |
1493 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), |
1494 | GFP_NOFS); | |
35a30d7c DS |
1495 | if (!root->free_ino_pinned || !root->free_ino_ctl) { |
1496 | ret = -ENOMEM; | |
581bb050 | 1497 | goto fail; |
35a30d7c | 1498 | } |
581bb050 LZ |
1499 | |
1500 | btrfs_init_free_ino_ctl(root); | |
1501 | mutex_init(&root->fs_commit_mutex); | |
1502 | spin_lock_init(&root->cache_lock); | |
1503 | init_waitqueue_head(&root->cache_wait); | |
1504 | ||
0ee5dc67 | 1505 | ret = get_anon_bdev(&root->anon_dev); |
ac08aedf CM |
1506 | if (ret) |
1507 | goto fail; | |
3394e160 | 1508 | |
d68fc57b YZ |
1509 | if (btrfs_root_refs(&root->root_item) == 0) { |
1510 | ret = -ENOENT; | |
1511 | goto fail; | |
1512 | } | |
1513 | ||
1514 | ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid); | |
1515 | if (ret < 0) | |
1516 | goto fail; | |
1517 | if (ret == 0) | |
1518 | root->orphan_item_inserted = 1; | |
1519 | ||
4df27c4d YZ |
1520 | ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); |
1521 | if (ret) | |
1522 | goto fail; | |
1523 | ||
1524 | spin_lock(&fs_info->fs_roots_radix_lock); | |
2619ba1f CM |
1525 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
1526 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 | 1527 | root); |
d68fc57b | 1528 | if (ret == 0) |
4df27c4d | 1529 | root->in_radix = 1; |
d68fc57b | 1530 | |
4df27c4d YZ |
1531 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1532 | radix_tree_preload_end(); | |
0f7d52f4 | 1533 | if (ret) { |
4df27c4d YZ |
1534 | if (ret == -EEXIST) { |
1535 | free_fs_root(root); | |
1536 | goto again; | |
1537 | } | |
1538 | goto fail; | |
0f7d52f4 | 1539 | } |
4df27c4d YZ |
1540 | |
1541 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
1542 | root->root_key.objectid); | |
1543 | WARN_ON(ret); | |
edbd8d4e | 1544 | return root; |
4df27c4d YZ |
1545 | fail: |
1546 | free_fs_root(root); | |
1547 | return ERR_PTR(ret); | |
edbd8d4e CM |
1548 | } |
1549 | ||
04160088 CM |
1550 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1551 | { | |
1552 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1553 | int ret = 0; | |
04160088 CM |
1554 | struct btrfs_device *device; |
1555 | struct backing_dev_info *bdi; | |
b7967db7 | 1556 | |
1f78160c XG |
1557 | rcu_read_lock(); |
1558 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1559 | if (!device->bdev) |
1560 | continue; | |
04160088 CM |
1561 | bdi = blk_get_backing_dev_info(device->bdev); |
1562 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1563 | ret = 1; | |
1564 | break; | |
1565 | } | |
1566 | } | |
1f78160c | 1567 | rcu_read_unlock(); |
04160088 CM |
1568 | return ret; |
1569 | } | |
1570 | ||
ad081f14 JA |
1571 | /* |
1572 | * If this fails, caller must call bdi_destroy() to get rid of the | |
1573 | * bdi again. | |
1574 | */ | |
04160088 CM |
1575 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1576 | { | |
ad081f14 JA |
1577 | int err; |
1578 | ||
1579 | bdi->capabilities = BDI_CAP_MAP_COPY; | |
e6d086d8 | 1580 | err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY); |
ad081f14 JA |
1581 | if (err) |
1582 | return err; | |
1583 | ||
4575c9cc | 1584 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1585 | bdi->congested_fn = btrfs_congested_fn; |
1586 | bdi->congested_data = info; | |
1587 | return 0; | |
1588 | } | |
1589 | ||
8b712842 CM |
1590 | /* |
1591 | * called by the kthread helper functions to finally call the bio end_io | |
1592 | * functions. This is where read checksum verification actually happens | |
1593 | */ | |
1594 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1595 | { |
ce9adaa5 | 1596 | struct bio *bio; |
8b712842 CM |
1597 | struct end_io_wq *end_io_wq; |
1598 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1599 | int error; |
ce9adaa5 | 1600 | |
8b712842 CM |
1601 | end_io_wq = container_of(work, struct end_io_wq, work); |
1602 | bio = end_io_wq->bio; | |
1603 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1604 | |
8b712842 CM |
1605 | error = end_io_wq->error; |
1606 | bio->bi_private = end_io_wq->private; | |
1607 | bio->bi_end_io = end_io_wq->end_io; | |
1608 | kfree(end_io_wq); | |
8b712842 | 1609 | bio_endio(bio, error); |
44b8bd7e CM |
1610 | } |
1611 | ||
a74a4b97 CM |
1612 | static int cleaner_kthread(void *arg) |
1613 | { | |
1614 | struct btrfs_root *root = arg; | |
1615 | ||
1616 | do { | |
a74a4b97 | 1617 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); |
76dda93c YZ |
1618 | |
1619 | if (!(root->fs_info->sb->s_flags & MS_RDONLY) && | |
1620 | mutex_trylock(&root->fs_info->cleaner_mutex)) { | |
24bbcf04 | 1621 | btrfs_run_delayed_iputs(root); |
76dda93c YZ |
1622 | btrfs_clean_old_snapshots(root); |
1623 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
4cb5300b | 1624 | btrfs_run_defrag_inodes(root->fs_info); |
76dda93c | 1625 | } |
a74a4b97 | 1626 | |
a0acae0e | 1627 | if (!try_to_freeze()) { |
a74a4b97 | 1628 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1629 | if (!kthread_should_stop()) |
1630 | schedule(); | |
a74a4b97 CM |
1631 | __set_current_state(TASK_RUNNING); |
1632 | } | |
1633 | } while (!kthread_should_stop()); | |
1634 | return 0; | |
1635 | } | |
1636 | ||
1637 | static int transaction_kthread(void *arg) | |
1638 | { | |
1639 | struct btrfs_root *root = arg; | |
1640 | struct btrfs_trans_handle *trans; | |
1641 | struct btrfs_transaction *cur; | |
8929ecfa | 1642 | u64 transid; |
a74a4b97 CM |
1643 | unsigned long now; |
1644 | unsigned long delay; | |
914b2007 | 1645 | bool cannot_commit; |
a74a4b97 CM |
1646 | |
1647 | do { | |
914b2007 | 1648 | cannot_commit = false; |
a74a4b97 CM |
1649 | delay = HZ * 30; |
1650 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1651 | mutex_lock(&root->fs_info->transaction_kthread_mutex); | |
1652 | ||
a4abeea4 | 1653 | spin_lock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1654 | cur = root->fs_info->running_transaction; |
1655 | if (!cur) { | |
a4abeea4 | 1656 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1657 | goto sleep; |
1658 | } | |
31153d81 | 1659 | |
a74a4b97 | 1660 | now = get_seconds(); |
8929ecfa YZ |
1661 | if (!cur->blocked && |
1662 | (now < cur->start_time || now - cur->start_time < 30)) { | |
a4abeea4 | 1663 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1664 | delay = HZ * 5; |
1665 | goto sleep; | |
1666 | } | |
8929ecfa | 1667 | transid = cur->transid; |
a4abeea4 | 1668 | spin_unlock(&root->fs_info->trans_lock); |
56bec294 | 1669 | |
79787eaa | 1670 | /* If the file system is aborted, this will always fail. */ |
7a7eaa40 | 1671 | trans = btrfs_join_transaction(root); |
914b2007 JK |
1672 | if (IS_ERR(trans)) { |
1673 | cannot_commit = true; | |
79787eaa | 1674 | goto sleep; |
914b2007 | 1675 | } |
8929ecfa | 1676 | if (transid == trans->transid) { |
79787eaa | 1677 | btrfs_commit_transaction(trans, root); |
8929ecfa YZ |
1678 | } else { |
1679 | btrfs_end_transaction(trans, root); | |
1680 | } | |
a74a4b97 CM |
1681 | sleep: |
1682 | wake_up_process(root->fs_info->cleaner_kthread); | |
1683 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1684 | ||
a0acae0e | 1685 | if (!try_to_freeze()) { |
a74a4b97 | 1686 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa | 1687 | if (!kthread_should_stop() && |
914b2007 JK |
1688 | (!btrfs_transaction_blocked(root->fs_info) || |
1689 | cannot_commit)) | |
8929ecfa | 1690 | schedule_timeout(delay); |
a74a4b97 CM |
1691 | __set_current_state(TASK_RUNNING); |
1692 | } | |
1693 | } while (!kthread_should_stop()); | |
1694 | return 0; | |
1695 | } | |
1696 | ||
af31f5e5 CM |
1697 | /* |
1698 | * this will find the highest generation in the array of | |
1699 | * root backups. The index of the highest array is returned, | |
1700 | * or -1 if we can't find anything. | |
1701 | * | |
1702 | * We check to make sure the array is valid by comparing the | |
1703 | * generation of the latest root in the array with the generation | |
1704 | * in the super block. If they don't match we pitch it. | |
1705 | */ | |
1706 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1707 | { | |
1708 | u64 cur; | |
1709 | int newest_index = -1; | |
1710 | struct btrfs_root_backup *root_backup; | |
1711 | int i; | |
1712 | ||
1713 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1714 | root_backup = info->super_copy->super_roots + i; | |
1715 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1716 | if (cur == newest_gen) | |
1717 | newest_index = i; | |
1718 | } | |
1719 | ||
1720 | /* check to see if we actually wrapped around */ | |
1721 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
1722 | root_backup = info->super_copy->super_roots; | |
1723 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1724 | if (cur == newest_gen) | |
1725 | newest_index = 0; | |
1726 | } | |
1727 | return newest_index; | |
1728 | } | |
1729 | ||
1730 | ||
1731 | /* | |
1732 | * find the oldest backup so we know where to store new entries | |
1733 | * in the backup array. This will set the backup_root_index | |
1734 | * field in the fs_info struct | |
1735 | */ | |
1736 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
1737 | u64 newest_gen) | |
1738 | { | |
1739 | int newest_index = -1; | |
1740 | ||
1741 | newest_index = find_newest_super_backup(info, newest_gen); | |
1742 | /* if there was garbage in there, just move along */ | |
1743 | if (newest_index == -1) { | |
1744 | info->backup_root_index = 0; | |
1745 | } else { | |
1746 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1747 | } | |
1748 | } | |
1749 | ||
1750 | /* | |
1751 | * copy all the root pointers into the super backup array. | |
1752 | * this will bump the backup pointer by one when it is | |
1753 | * done | |
1754 | */ | |
1755 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1756 | { | |
1757 | int next_backup; | |
1758 | struct btrfs_root_backup *root_backup; | |
1759 | int last_backup; | |
1760 | ||
1761 | next_backup = info->backup_root_index; | |
1762 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1763 | BTRFS_NUM_BACKUP_ROOTS; | |
1764 | ||
1765 | /* | |
1766 | * just overwrite the last backup if we're at the same generation | |
1767 | * this happens only at umount | |
1768 | */ | |
1769 | root_backup = info->super_for_commit->super_roots + last_backup; | |
1770 | if (btrfs_backup_tree_root_gen(root_backup) == | |
1771 | btrfs_header_generation(info->tree_root->node)) | |
1772 | next_backup = last_backup; | |
1773 | ||
1774 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1775 | ||
1776 | /* | |
1777 | * make sure all of our padding and empty slots get zero filled | |
1778 | * regardless of which ones we use today | |
1779 | */ | |
1780 | memset(root_backup, 0, sizeof(*root_backup)); | |
1781 | ||
1782 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1783 | ||
1784 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1785 | btrfs_set_backup_tree_root_gen(root_backup, | |
1786 | btrfs_header_generation(info->tree_root->node)); | |
1787 | ||
1788 | btrfs_set_backup_tree_root_level(root_backup, | |
1789 | btrfs_header_level(info->tree_root->node)); | |
1790 | ||
1791 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1792 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1793 | btrfs_header_generation(info->chunk_root->node)); | |
1794 | btrfs_set_backup_chunk_root_level(root_backup, | |
1795 | btrfs_header_level(info->chunk_root->node)); | |
1796 | ||
1797 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
1798 | btrfs_set_backup_extent_root_gen(root_backup, | |
1799 | btrfs_header_generation(info->extent_root->node)); | |
1800 | btrfs_set_backup_extent_root_level(root_backup, | |
1801 | btrfs_header_level(info->extent_root->node)); | |
1802 | ||
7c7e82a7 CM |
1803 | /* |
1804 | * we might commit during log recovery, which happens before we set | |
1805 | * the fs_root. Make sure it is valid before we fill it in. | |
1806 | */ | |
1807 | if (info->fs_root && info->fs_root->node) { | |
1808 | btrfs_set_backup_fs_root(root_backup, | |
1809 | info->fs_root->node->start); | |
1810 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1811 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1812 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1813 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1814 | } |
af31f5e5 CM |
1815 | |
1816 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1817 | btrfs_set_backup_dev_root_gen(root_backup, | |
1818 | btrfs_header_generation(info->dev_root->node)); | |
1819 | btrfs_set_backup_dev_root_level(root_backup, | |
1820 | btrfs_header_level(info->dev_root->node)); | |
1821 | ||
1822 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
1823 | btrfs_set_backup_csum_root_gen(root_backup, | |
1824 | btrfs_header_generation(info->csum_root->node)); | |
1825 | btrfs_set_backup_csum_root_level(root_backup, | |
1826 | btrfs_header_level(info->csum_root->node)); | |
1827 | ||
1828 | btrfs_set_backup_total_bytes(root_backup, | |
1829 | btrfs_super_total_bytes(info->super_copy)); | |
1830 | btrfs_set_backup_bytes_used(root_backup, | |
1831 | btrfs_super_bytes_used(info->super_copy)); | |
1832 | btrfs_set_backup_num_devices(root_backup, | |
1833 | btrfs_super_num_devices(info->super_copy)); | |
1834 | ||
1835 | /* | |
1836 | * if we don't copy this out to the super_copy, it won't get remembered | |
1837 | * for the next commit | |
1838 | */ | |
1839 | memcpy(&info->super_copy->super_roots, | |
1840 | &info->super_for_commit->super_roots, | |
1841 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1842 | } | |
1843 | ||
1844 | /* | |
1845 | * this copies info out of the root backup array and back into | |
1846 | * the in-memory super block. It is meant to help iterate through | |
1847 | * the array, so you send it the number of backups you've already | |
1848 | * tried and the last backup index you used. | |
1849 | * | |
1850 | * this returns -1 when it has tried all the backups | |
1851 | */ | |
1852 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
1853 | struct btrfs_super_block *super, | |
1854 | int *num_backups_tried, int *backup_index) | |
1855 | { | |
1856 | struct btrfs_root_backup *root_backup; | |
1857 | int newest = *backup_index; | |
1858 | ||
1859 | if (*num_backups_tried == 0) { | |
1860 | u64 gen = btrfs_super_generation(super); | |
1861 | ||
1862 | newest = find_newest_super_backup(info, gen); | |
1863 | if (newest == -1) | |
1864 | return -1; | |
1865 | ||
1866 | *backup_index = newest; | |
1867 | *num_backups_tried = 1; | |
1868 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
1869 | /* we've tried all the backups, all done */ | |
1870 | return -1; | |
1871 | } else { | |
1872 | /* jump to the next oldest backup */ | |
1873 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1874 | BTRFS_NUM_BACKUP_ROOTS; | |
1875 | *backup_index = newest; | |
1876 | *num_backups_tried += 1; | |
1877 | } | |
1878 | root_backup = super->super_roots + newest; | |
1879 | ||
1880 | btrfs_set_super_generation(super, | |
1881 | btrfs_backup_tree_root_gen(root_backup)); | |
1882 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
1883 | btrfs_set_super_root_level(super, | |
1884 | btrfs_backup_tree_root_level(root_backup)); | |
1885 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
1886 | ||
1887 | /* | |
1888 | * fixme: the total bytes and num_devices need to match or we should | |
1889 | * need a fsck | |
1890 | */ | |
1891 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
1892 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
1893 | return 0; | |
1894 | } | |
1895 | ||
1896 | /* helper to cleanup tree roots */ | |
1897 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
1898 | { | |
1899 | free_extent_buffer(info->tree_root->node); | |
1900 | free_extent_buffer(info->tree_root->commit_root); | |
1901 | free_extent_buffer(info->dev_root->node); | |
1902 | free_extent_buffer(info->dev_root->commit_root); | |
1903 | free_extent_buffer(info->extent_root->node); | |
1904 | free_extent_buffer(info->extent_root->commit_root); | |
1905 | free_extent_buffer(info->csum_root->node); | |
1906 | free_extent_buffer(info->csum_root->commit_root); | |
bcef60f2 AJ |
1907 | if (info->quota_root) { |
1908 | free_extent_buffer(info->quota_root->node); | |
1909 | free_extent_buffer(info->quota_root->commit_root); | |
1910 | } | |
af31f5e5 CM |
1911 | |
1912 | info->tree_root->node = NULL; | |
1913 | info->tree_root->commit_root = NULL; | |
1914 | info->dev_root->node = NULL; | |
1915 | info->dev_root->commit_root = NULL; | |
1916 | info->extent_root->node = NULL; | |
1917 | info->extent_root->commit_root = NULL; | |
1918 | info->csum_root->node = NULL; | |
1919 | info->csum_root->commit_root = NULL; | |
bcef60f2 AJ |
1920 | if (info->quota_root) { |
1921 | info->quota_root->node = NULL; | |
1922 | info->quota_root->commit_root = NULL; | |
1923 | } | |
af31f5e5 CM |
1924 | |
1925 | if (chunk_root) { | |
1926 | free_extent_buffer(info->chunk_root->node); | |
1927 | free_extent_buffer(info->chunk_root->commit_root); | |
1928 | info->chunk_root->node = NULL; | |
1929 | info->chunk_root->commit_root = NULL; | |
1930 | } | |
1931 | } | |
1932 | ||
1933 | ||
ad2b2c80 AV |
1934 | int open_ctree(struct super_block *sb, |
1935 | struct btrfs_fs_devices *fs_devices, | |
1936 | char *options) | |
2e635a27 | 1937 | { |
db94535d CM |
1938 | u32 sectorsize; |
1939 | u32 nodesize; | |
1940 | u32 leafsize; | |
1941 | u32 blocksize; | |
87ee04eb | 1942 | u32 stripesize; |
84234f3a | 1943 | u64 generation; |
f2b636e8 | 1944 | u64 features; |
3de4586c | 1945 | struct btrfs_key location; |
a061fc8d | 1946 | struct buffer_head *bh; |
4d34b278 | 1947 | struct btrfs_super_block *disk_super; |
815745cf | 1948 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
f84a8bd6 | 1949 | struct btrfs_root *tree_root; |
4d34b278 ID |
1950 | struct btrfs_root *extent_root; |
1951 | struct btrfs_root *csum_root; | |
1952 | struct btrfs_root *chunk_root; | |
1953 | struct btrfs_root *dev_root; | |
bcef60f2 | 1954 | struct btrfs_root *quota_root; |
e02119d5 | 1955 | struct btrfs_root *log_tree_root; |
eb60ceac | 1956 | int ret; |
e58ca020 | 1957 | int err = -EINVAL; |
af31f5e5 CM |
1958 | int num_backups_tried = 0; |
1959 | int backup_index = 0; | |
4543df7e | 1960 | |
f84a8bd6 | 1961 | tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info); |
6f07e42e AV |
1962 | extent_root = fs_info->extent_root = btrfs_alloc_root(fs_info); |
1963 | csum_root = fs_info->csum_root = btrfs_alloc_root(fs_info); | |
1964 | chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info); | |
1965 | dev_root = fs_info->dev_root = btrfs_alloc_root(fs_info); | |
bcef60f2 | 1966 | quota_root = fs_info->quota_root = btrfs_alloc_root(fs_info); |
8790d502 | 1967 | |
f84a8bd6 | 1968 | if (!tree_root || !extent_root || !csum_root || |
bcef60f2 | 1969 | !chunk_root || !dev_root || !quota_root) { |
39279cc3 CM |
1970 | err = -ENOMEM; |
1971 | goto fail; | |
1972 | } | |
76dda93c YZ |
1973 | |
1974 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
1975 | if (ret) { | |
1976 | err = ret; | |
1977 | goto fail; | |
1978 | } | |
1979 | ||
1980 | ret = setup_bdi(fs_info, &fs_info->bdi); | |
1981 | if (ret) { | |
1982 | err = ret; | |
1983 | goto fail_srcu; | |
1984 | } | |
1985 | ||
1986 | fs_info->btree_inode = new_inode(sb); | |
1987 | if (!fs_info->btree_inode) { | |
1988 | err = -ENOMEM; | |
1989 | goto fail_bdi; | |
1990 | } | |
1991 | ||
a6591715 | 1992 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
1561deda | 1993 | |
76dda93c | 1994 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
8fd17795 | 1995 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1996 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 1997 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
19c00ddc | 1998 | INIT_LIST_HEAD(&fs_info->hashers); |
ea8c2819 | 1999 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
5a3f23d5 | 2000 | INIT_LIST_HEAD(&fs_info->ordered_operations); |
11833d66 | 2001 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
1832a6d5 | 2002 | spin_lock_init(&fs_info->delalloc_lock); |
a4abeea4 | 2003 | spin_lock_init(&fs_info->trans_lock); |
31153d81 | 2004 | spin_lock_init(&fs_info->ref_cache_lock); |
76dda93c | 2005 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2006 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2007 | spin_lock_init(&fs_info->defrag_inodes_lock); |
2bf64758 | 2008 | spin_lock_init(&fs_info->free_chunk_lock); |
f29021b2 JS |
2009 | spin_lock_init(&fs_info->tree_mod_seq_lock); |
2010 | rwlock_init(&fs_info->tree_mod_log_lock); | |
7585717f | 2011 | mutex_init(&fs_info->reloc_mutex); |
19c00ddc | 2012 | |
58176a96 | 2013 | init_completion(&fs_info->kobj_unregister); |
0b86a832 | 2014 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2015 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2016 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
0b86a832 | 2017 | btrfs_mapping_init(&fs_info->mapping_tree); |
f0486c68 YZ |
2018 | btrfs_init_block_rsv(&fs_info->global_block_rsv); |
2019 | btrfs_init_block_rsv(&fs_info->delalloc_block_rsv); | |
2020 | btrfs_init_block_rsv(&fs_info->trans_block_rsv); | |
2021 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv); | |
2022 | btrfs_init_block_rsv(&fs_info->empty_block_rsv); | |
6d668dda | 2023 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv); |
cb03c743 | 2024 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 2025 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 2026 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 2027 | atomic_set(&fs_info->nr_async_bios, 0); |
4cb5300b | 2028 | atomic_set(&fs_info->defrag_running, 0); |
f29021b2 | 2029 | atomic_set(&fs_info->tree_mod_seq, 0); |
e20d96d6 | 2030 | fs_info->sb = sb; |
6f568d35 | 2031 | fs_info->max_inline = 8192 * 1024; |
9ed74f2d | 2032 | fs_info->metadata_ratio = 0; |
4cb5300b | 2033 | fs_info->defrag_inodes = RB_ROOT; |
a4abeea4 | 2034 | fs_info->trans_no_join = 0; |
2bf64758 | 2035 | fs_info->free_chunk_space = 0; |
f29021b2 | 2036 | fs_info->tree_mod_log = RB_ROOT; |
c8b97818 | 2037 | |
90519d66 AJ |
2038 | /* readahead state */ |
2039 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); | |
2040 | spin_lock_init(&fs_info->reada_lock); | |
c8b97818 | 2041 | |
b34b086c CM |
2042 | fs_info->thread_pool_size = min_t(unsigned long, |
2043 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2044 | |
3eaa2885 CM |
2045 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
2046 | spin_lock_init(&fs_info->ordered_extent_lock); | |
16cdcec7 MX |
2047 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
2048 | GFP_NOFS); | |
2049 | if (!fs_info->delayed_root) { | |
2050 | err = -ENOMEM; | |
2051 | goto fail_iput; | |
2052 | } | |
2053 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 2054 | |
a2de733c AJ |
2055 | mutex_init(&fs_info->scrub_lock); |
2056 | atomic_set(&fs_info->scrubs_running, 0); | |
2057 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2058 | atomic_set(&fs_info->scrubs_paused, 0); | |
2059 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2060 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
2061 | init_rwsem(&fs_info->scrub_super_lock); | |
2062 | fs_info->scrub_workers_refcnt = 0; | |
21adbd5c SB |
2063 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2064 | fs_info->check_integrity_print_mask = 0; | |
2065 | #endif | |
a2de733c | 2066 | |
c9e9f97b ID |
2067 | spin_lock_init(&fs_info->balance_lock); |
2068 | mutex_init(&fs_info->balance_mutex); | |
837d5b6e ID |
2069 | atomic_set(&fs_info->balance_running, 0); |
2070 | atomic_set(&fs_info->balance_pause_req, 0); | |
a7e99c69 | 2071 | atomic_set(&fs_info->balance_cancel_req, 0); |
c9e9f97b | 2072 | fs_info->balance_ctl = NULL; |
837d5b6e | 2073 | init_waitqueue_head(&fs_info->balance_wait_q); |
a2de733c | 2074 | |
a061fc8d CM |
2075 | sb->s_blocksize = 4096; |
2076 | sb->s_blocksize_bits = blksize_bits(4096); | |
32a88aa1 | 2077 | sb->s_bdi = &fs_info->bdi; |
a061fc8d | 2078 | |
76dda93c | 2079 | fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; |
bfe86848 | 2080 | set_nlink(fs_info->btree_inode, 1); |
0afbaf8c CM |
2081 | /* |
2082 | * we set the i_size on the btree inode to the max possible int. | |
2083 | * the real end of the address space is determined by all of | |
2084 | * the devices in the system | |
2085 | */ | |
2086 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 2087 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
2088 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
2089 | ||
5d4f98a2 | 2090 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); |
d1310b2e | 2091 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
f993c883 | 2092 | fs_info->btree_inode->i_mapping); |
0b32f4bb | 2093 | BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; |
a8067e02 | 2094 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); |
d1310b2e CM |
2095 | |
2096 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 2097 | |
76dda93c YZ |
2098 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
2099 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
2100 | sizeof(struct btrfs_key)); | |
72ac3c0d JB |
2101 | set_bit(BTRFS_INODE_DUMMY, |
2102 | &BTRFS_I(fs_info->btree_inode)->runtime_flags); | |
c65ddb52 | 2103 | insert_inode_hash(fs_info->btree_inode); |
76dda93c | 2104 | |
0f9dd46c | 2105 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2106 | fs_info->block_group_cache_tree = RB_ROOT; |
0f9dd46c | 2107 | |
11833d66 | 2108 | extent_io_tree_init(&fs_info->freed_extents[0], |
f993c883 | 2109 | fs_info->btree_inode->i_mapping); |
11833d66 | 2110 | extent_io_tree_init(&fs_info->freed_extents[1], |
f993c883 | 2111 | fs_info->btree_inode->i_mapping); |
11833d66 | 2112 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
e66f709b | 2113 | fs_info->do_barriers = 1; |
e18e4809 | 2114 | |
39279cc3 | 2115 | |
5a3f23d5 | 2116 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2117 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2118 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2119 | mutex_init(&fs_info->transaction_kthread_mutex); |
2120 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 2121 | mutex_init(&fs_info->volume_mutex); |
276e680d | 2122 | init_rwsem(&fs_info->extent_commit_sem); |
c71bf099 | 2123 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2124 | init_rwsem(&fs_info->subvol_sem); |
fa9c0d79 | 2125 | |
416ac51d AJ |
2126 | spin_lock_init(&fs_info->qgroup_lock); |
2127 | fs_info->qgroup_tree = RB_ROOT; | |
2128 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); | |
2129 | fs_info->qgroup_seq = 1; | |
2130 | fs_info->quota_enabled = 0; | |
2131 | fs_info->pending_quota_state = 0; | |
2132 | ||
fa9c0d79 CM |
2133 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2134 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2135 | ||
e6dcd2dc | 2136 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2137 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2138 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2139 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2140 | |
0b86a832 | 2141 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 2142 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2143 | |
3c4bb26b | 2144 | invalidate_bdev(fs_devices->latest_bdev); |
a512bbf8 | 2145 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
20b45077 DY |
2146 | if (!bh) { |
2147 | err = -EINVAL; | |
16cdcec7 | 2148 | goto fail_alloc; |
20b45077 | 2149 | } |
39279cc3 | 2150 | |
6c41761f DS |
2151 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2152 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2153 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2154 | brelse(bh); |
5f39d397 | 2155 | |
6c41761f | 2156 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2157 | |
6c41761f | 2158 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2159 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2160 | goto fail_alloc; |
0f7d52f4 | 2161 | |
acce952b | 2162 | /* check FS state, whether FS is broken. */ |
2163 | fs_info->fs_state |= btrfs_super_flags(disk_super); | |
2164 | ||
fcd1f065 DS |
2165 | ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); |
2166 | if (ret) { | |
2167 | printk(KERN_ERR "btrfs: superblock contains fatal errors\n"); | |
2168 | err = ret; | |
2169 | goto fail_alloc; | |
2170 | } | |
acce952b | 2171 | |
af31f5e5 CM |
2172 | /* |
2173 | * run through our array of backup supers and setup | |
2174 | * our ring pointer to the oldest one | |
2175 | */ | |
2176 | generation = btrfs_super_generation(disk_super); | |
2177 | find_oldest_super_backup(fs_info, generation); | |
2178 | ||
75e7cb7f LB |
2179 | /* |
2180 | * In the long term, we'll store the compression type in the super | |
2181 | * block, and it'll be used for per file compression control. | |
2182 | */ | |
2183 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2184 | ||
2b82032c YZ |
2185 | ret = btrfs_parse_options(tree_root, options); |
2186 | if (ret) { | |
2187 | err = ret; | |
16cdcec7 | 2188 | goto fail_alloc; |
2b82032c | 2189 | } |
dfe25020 | 2190 | |
f2b636e8 JB |
2191 | features = btrfs_super_incompat_flags(disk_super) & |
2192 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2193 | if (features) { | |
2194 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
2195 | "unsupported optional features (%Lx).\n", | |
21380931 | 2196 | (unsigned long long)features); |
f2b636e8 | 2197 | err = -EINVAL; |
16cdcec7 | 2198 | goto fail_alloc; |
f2b636e8 JB |
2199 | } |
2200 | ||
727011e0 CM |
2201 | if (btrfs_super_leafsize(disk_super) != |
2202 | btrfs_super_nodesize(disk_super)) { | |
2203 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2204 | "blocksizes don't match. node %d leaf %d\n", | |
2205 | btrfs_super_nodesize(disk_super), | |
2206 | btrfs_super_leafsize(disk_super)); | |
2207 | err = -EINVAL; | |
2208 | goto fail_alloc; | |
2209 | } | |
2210 | if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2211 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2212 | "blocksize (%d) was too large\n", | |
2213 | btrfs_super_leafsize(disk_super)); | |
2214 | err = -EINVAL; | |
2215 | goto fail_alloc; | |
2216 | } | |
2217 | ||
5d4f98a2 | 2218 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 2219 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
69e380d1 | 2220 | if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 2221 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
727011e0 CM |
2222 | |
2223 | /* | |
2224 | * flag our filesystem as having big metadata blocks if | |
2225 | * they are bigger than the page size | |
2226 | */ | |
2227 | if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) { | |
2228 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) | |
2229 | printk(KERN_INFO "btrfs flagging fs with big metadata feature\n"); | |
2230 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
2231 | } | |
2232 | ||
bc3f116f CM |
2233 | nodesize = btrfs_super_nodesize(disk_super); |
2234 | leafsize = btrfs_super_leafsize(disk_super); | |
2235 | sectorsize = btrfs_super_sectorsize(disk_super); | |
2236 | stripesize = btrfs_super_stripesize(disk_super); | |
2237 | ||
2238 | /* | |
2239 | * mixed block groups end up with duplicate but slightly offset | |
2240 | * extent buffers for the same range. It leads to corruptions | |
2241 | */ | |
2242 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
2243 | (sectorsize != leafsize)) { | |
2244 | printk(KERN_WARNING "btrfs: unequal leaf/node/sector sizes " | |
2245 | "are not allowed for mixed block groups on %s\n", | |
2246 | sb->s_id); | |
2247 | goto fail_alloc; | |
2248 | } | |
2249 | ||
a6fa6fae | 2250 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 2251 | |
f2b636e8 JB |
2252 | features = btrfs_super_compat_ro_flags(disk_super) & |
2253 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
2254 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
2255 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
2256 | "unsupported option features (%Lx).\n", | |
21380931 | 2257 | (unsigned long long)features); |
f2b636e8 | 2258 | err = -EINVAL; |
16cdcec7 | 2259 | goto fail_alloc; |
f2b636e8 | 2260 | } |
61d92c32 CM |
2261 | |
2262 | btrfs_init_workers(&fs_info->generic_worker, | |
2263 | "genwork", 1, NULL); | |
2264 | ||
5443be45 | 2265 | btrfs_init_workers(&fs_info->workers, "worker", |
61d92c32 CM |
2266 | fs_info->thread_pool_size, |
2267 | &fs_info->generic_worker); | |
c8b97818 | 2268 | |
771ed689 | 2269 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
61d92c32 CM |
2270 | fs_info->thread_pool_size, |
2271 | &fs_info->generic_worker); | |
771ed689 | 2272 | |
5443be45 | 2273 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 | 2274 | min_t(u64, fs_devices->num_devices, |
61d92c32 CM |
2275 | fs_info->thread_pool_size), |
2276 | &fs_info->generic_worker); | |
61b49440 | 2277 | |
bab39bf9 JB |
2278 | btrfs_init_workers(&fs_info->caching_workers, "cache", |
2279 | 2, &fs_info->generic_worker); | |
2280 | ||
61b49440 CM |
2281 | /* a higher idle thresh on the submit workers makes it much more |
2282 | * likely that bios will be send down in a sane order to the | |
2283 | * devices | |
2284 | */ | |
2285 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 2286 | |
771ed689 | 2287 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 2288 | fs_info->workers.ordered = 1; |
61b49440 | 2289 | |
771ed689 CM |
2290 | fs_info->delalloc_workers.idle_thresh = 2; |
2291 | fs_info->delalloc_workers.ordered = 1; | |
2292 | ||
61d92c32 CM |
2293 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1, |
2294 | &fs_info->generic_worker); | |
5443be45 | 2295 | btrfs_init_workers(&fs_info->endio_workers, "endio", |
61d92c32 CM |
2296 | fs_info->thread_pool_size, |
2297 | &fs_info->generic_worker); | |
d20f7043 | 2298 | btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", |
61d92c32 CM |
2299 | fs_info->thread_pool_size, |
2300 | &fs_info->generic_worker); | |
cad321ad | 2301 | btrfs_init_workers(&fs_info->endio_meta_write_workers, |
61d92c32 CM |
2302 | "endio-meta-write", fs_info->thread_pool_size, |
2303 | &fs_info->generic_worker); | |
5443be45 | 2304 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", |
61d92c32 CM |
2305 | fs_info->thread_pool_size, |
2306 | &fs_info->generic_worker); | |
0cb59c99 JB |
2307 | btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write", |
2308 | 1, &fs_info->generic_worker); | |
16cdcec7 MX |
2309 | btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta", |
2310 | fs_info->thread_pool_size, | |
2311 | &fs_info->generic_worker); | |
90519d66 AJ |
2312 | btrfs_init_workers(&fs_info->readahead_workers, "readahead", |
2313 | fs_info->thread_pool_size, | |
2314 | &fs_info->generic_worker); | |
61b49440 CM |
2315 | |
2316 | /* | |
2317 | * endios are largely parallel and should have a very | |
2318 | * low idle thresh | |
2319 | */ | |
2320 | fs_info->endio_workers.idle_thresh = 4; | |
b51912c9 CM |
2321 | fs_info->endio_meta_workers.idle_thresh = 4; |
2322 | ||
9042846b CM |
2323 | fs_info->endio_write_workers.idle_thresh = 2; |
2324 | fs_info->endio_meta_write_workers.idle_thresh = 2; | |
90519d66 | 2325 | fs_info->readahead_workers.idle_thresh = 2; |
9042846b | 2326 | |
0dc3b84a JB |
2327 | /* |
2328 | * btrfs_start_workers can really only fail because of ENOMEM so just | |
2329 | * return -ENOMEM if any of these fail. | |
2330 | */ | |
2331 | ret = btrfs_start_workers(&fs_info->workers); | |
2332 | ret |= btrfs_start_workers(&fs_info->generic_worker); | |
2333 | ret |= btrfs_start_workers(&fs_info->submit_workers); | |
2334 | ret |= btrfs_start_workers(&fs_info->delalloc_workers); | |
2335 | ret |= btrfs_start_workers(&fs_info->fixup_workers); | |
2336 | ret |= btrfs_start_workers(&fs_info->endio_workers); | |
2337 | ret |= btrfs_start_workers(&fs_info->endio_meta_workers); | |
2338 | ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers); | |
2339 | ret |= btrfs_start_workers(&fs_info->endio_write_workers); | |
2340 | ret |= btrfs_start_workers(&fs_info->endio_freespace_worker); | |
2341 | ret |= btrfs_start_workers(&fs_info->delayed_workers); | |
2342 | ret |= btrfs_start_workers(&fs_info->caching_workers); | |
2343 | ret |= btrfs_start_workers(&fs_info->readahead_workers); | |
2344 | if (ret) { | |
fed425c7 | 2345 | err = -ENOMEM; |
0dc3b84a JB |
2346 | goto fail_sb_buffer; |
2347 | } | |
4543df7e | 2348 | |
4575c9cc | 2349 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
2350 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
2351 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 2352 | |
db94535d CM |
2353 | tree_root->nodesize = nodesize; |
2354 | tree_root->leafsize = leafsize; | |
2355 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 2356 | tree_root->stripesize = stripesize; |
a061fc8d CM |
2357 | |
2358 | sb->s_blocksize = sectorsize; | |
2359 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 2360 | |
39279cc3 CM |
2361 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
2362 | sizeof(disk_super->magic))) { | |
d397712b | 2363 | printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
2364 | goto fail_sb_buffer; |
2365 | } | |
19c00ddc | 2366 | |
8d082fb7 LB |
2367 | if (sectorsize != PAGE_SIZE) { |
2368 | printk(KERN_WARNING "btrfs: Incompatible sector size(%lu) " | |
2369 | "found on %s\n", (unsigned long)sectorsize, sb->s_id); | |
941b2ddf KM |
2370 | goto fail_sb_buffer; |
2371 | } | |
2372 | ||
925baedd | 2373 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 2374 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 2375 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2376 | if (ret) { |
d397712b CM |
2377 | printk(KERN_WARNING "btrfs: failed to read the system " |
2378 | "array on %s\n", sb->s_id); | |
5d4f98a2 | 2379 | goto fail_sb_buffer; |
84eed90f | 2380 | } |
0b86a832 CM |
2381 | |
2382 | blocksize = btrfs_level_size(tree_root, | |
2383 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 2384 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
2385 | |
2386 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2387 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
2388 | ||
2389 | chunk_root->node = read_tree_block(chunk_root, | |
2390 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 2391 | blocksize, generation); |
79787eaa | 2392 | BUG_ON(!chunk_root->node); /* -ENOMEM */ |
83121942 DW |
2393 | if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { |
2394 | printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n", | |
2395 | sb->s_id); | |
af31f5e5 | 2396 | goto fail_tree_roots; |
83121942 | 2397 | } |
5d4f98a2 YZ |
2398 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
2399 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 2400 | |
e17cade2 | 2401 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
d397712b CM |
2402 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), |
2403 | BTRFS_UUID_SIZE); | |
e17cade2 | 2404 | |
0b86a832 | 2405 | ret = btrfs_read_chunk_tree(chunk_root); |
2b82032c | 2406 | if (ret) { |
d397712b CM |
2407 | printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", |
2408 | sb->s_id); | |
af31f5e5 | 2409 | goto fail_tree_roots; |
2b82032c | 2410 | } |
0b86a832 | 2411 | |
dfe25020 CM |
2412 | btrfs_close_extra_devices(fs_devices); |
2413 | ||
a6b0d5c8 CM |
2414 | if (!fs_devices->latest_bdev) { |
2415 | printk(KERN_CRIT "btrfs: failed to read devices on %s\n", | |
2416 | sb->s_id); | |
2417 | goto fail_tree_roots; | |
2418 | } | |
2419 | ||
af31f5e5 | 2420 | retry_root_backup: |
db94535d CM |
2421 | blocksize = btrfs_level_size(tree_root, |
2422 | btrfs_super_root_level(disk_super)); | |
84234f3a | 2423 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 2424 | |
e20d96d6 | 2425 | tree_root->node = read_tree_block(tree_root, |
db94535d | 2426 | btrfs_super_root(disk_super), |
84234f3a | 2427 | blocksize, generation); |
af31f5e5 CM |
2428 | if (!tree_root->node || |
2429 | !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { | |
83121942 DW |
2430 | printk(KERN_WARNING "btrfs: failed to read tree root on %s\n", |
2431 | sb->s_id); | |
af31f5e5 CM |
2432 | |
2433 | goto recovery_tree_root; | |
83121942 | 2434 | } |
af31f5e5 | 2435 | |
5d4f98a2 YZ |
2436 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
2437 | tree_root->commit_root = btrfs_root_node(tree_root); | |
db94535d CM |
2438 | |
2439 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 2440 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 2441 | if (ret) |
af31f5e5 | 2442 | goto recovery_tree_root; |
0b86a832 CM |
2443 | extent_root->track_dirty = 1; |
2444 | ||
2445 | ret = find_and_setup_root(tree_root, fs_info, | |
2446 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
0b86a832 | 2447 | if (ret) |
af31f5e5 | 2448 | goto recovery_tree_root; |
5d4f98a2 | 2449 | dev_root->track_dirty = 1; |
3768f368 | 2450 | |
d20f7043 CM |
2451 | ret = find_and_setup_root(tree_root, fs_info, |
2452 | BTRFS_CSUM_TREE_OBJECTID, csum_root); | |
2453 | if (ret) | |
af31f5e5 | 2454 | goto recovery_tree_root; |
d20f7043 CM |
2455 | csum_root->track_dirty = 1; |
2456 | ||
bcef60f2 AJ |
2457 | ret = find_and_setup_root(tree_root, fs_info, |
2458 | BTRFS_QUOTA_TREE_OBJECTID, quota_root); | |
2459 | if (ret) { | |
2460 | kfree(quota_root); | |
2461 | quota_root = fs_info->quota_root = NULL; | |
2462 | } else { | |
2463 | quota_root->track_dirty = 1; | |
2464 | fs_info->quota_enabled = 1; | |
2465 | fs_info->pending_quota_state = 1; | |
2466 | } | |
2467 | ||
8929ecfa YZ |
2468 | fs_info->generation = generation; |
2469 | fs_info->last_trans_committed = generation; | |
8929ecfa | 2470 | |
68310a5e ID |
2471 | ret = btrfs_recover_balance(fs_info); |
2472 | if (ret) { | |
2473 | printk(KERN_WARNING "btrfs: failed to recover balance\n"); | |
2474 | goto fail_block_groups; | |
2475 | } | |
2476 | ||
733f4fbb SB |
2477 | ret = btrfs_init_dev_stats(fs_info); |
2478 | if (ret) { | |
2479 | printk(KERN_ERR "btrfs: failed to init dev_stats: %d\n", | |
2480 | ret); | |
2481 | goto fail_block_groups; | |
2482 | } | |
2483 | ||
c59021f8 | 2484 | ret = btrfs_init_space_info(fs_info); |
2485 | if (ret) { | |
2486 | printk(KERN_ERR "Failed to initial space info: %d\n", ret); | |
2487 | goto fail_block_groups; | |
2488 | } | |
2489 | ||
1b1d1f66 JB |
2490 | ret = btrfs_read_block_groups(extent_root); |
2491 | if (ret) { | |
2492 | printk(KERN_ERR "Failed to read block groups: %d\n", ret); | |
2493 | goto fail_block_groups; | |
2494 | } | |
9078a3e1 | 2495 | |
a74a4b97 CM |
2496 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
2497 | "btrfs-cleaner"); | |
57506d50 | 2498 | if (IS_ERR(fs_info->cleaner_kthread)) |
1b1d1f66 | 2499 | goto fail_block_groups; |
a74a4b97 CM |
2500 | |
2501 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
2502 | tree_root, | |
2503 | "btrfs-transaction"); | |
57506d50 | 2504 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 2505 | goto fail_cleaner; |
a74a4b97 | 2506 | |
c289811c CM |
2507 | if (!btrfs_test_opt(tree_root, SSD) && |
2508 | !btrfs_test_opt(tree_root, NOSSD) && | |
2509 | !fs_info->fs_devices->rotating) { | |
2510 | printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD " | |
2511 | "mode\n"); | |
2512 | btrfs_set_opt(fs_info->mount_opt, SSD); | |
2513 | } | |
2514 | ||
21adbd5c SB |
2515 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2516 | if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { | |
2517 | ret = btrfsic_mount(tree_root, fs_devices, | |
2518 | btrfs_test_opt(tree_root, | |
2519 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? | |
2520 | 1 : 0, | |
2521 | fs_info->check_integrity_print_mask); | |
2522 | if (ret) | |
2523 | printk(KERN_WARNING "btrfs: failed to initialize" | |
2524 | " integrity check module %s\n", sb->s_id); | |
2525 | } | |
2526 | #endif | |
bcef60f2 AJ |
2527 | ret = btrfs_read_qgroup_config(fs_info); |
2528 | if (ret) | |
2529 | goto fail_trans_kthread; | |
21adbd5c | 2530 | |
acce952b | 2531 | /* do not make disk changes in broken FS */ |
2532 | if (btrfs_super_log_root(disk_super) != 0 && | |
2533 | !(fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)) { | |
e02119d5 CM |
2534 | u64 bytenr = btrfs_super_log_root(disk_super); |
2535 | ||
7c2ca468 | 2536 | if (fs_devices->rw_devices == 0) { |
d397712b CM |
2537 | printk(KERN_WARNING "Btrfs log replay required " |
2538 | "on RO media\n"); | |
7c2ca468 | 2539 | err = -EIO; |
bcef60f2 | 2540 | goto fail_qgroup; |
7c2ca468 | 2541 | } |
e02119d5 CM |
2542 | blocksize = |
2543 | btrfs_level_size(tree_root, | |
2544 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 2545 | |
6f07e42e | 2546 | log_tree_root = btrfs_alloc_root(fs_info); |
676e4c86 DC |
2547 | if (!log_tree_root) { |
2548 | err = -ENOMEM; | |
bcef60f2 | 2549 | goto fail_qgroup; |
676e4c86 | 2550 | } |
e02119d5 CM |
2551 | |
2552 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2553 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
2554 | ||
2555 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
2556 | blocksize, |
2557 | generation + 1); | |
79787eaa | 2558 | /* returns with log_tree_root freed on success */ |
e02119d5 | 2559 | ret = btrfs_recover_log_trees(log_tree_root); |
79787eaa JM |
2560 | if (ret) { |
2561 | btrfs_error(tree_root->fs_info, ret, | |
2562 | "Failed to recover log tree"); | |
2563 | free_extent_buffer(log_tree_root->node); | |
2564 | kfree(log_tree_root); | |
2565 | goto fail_trans_kthread; | |
2566 | } | |
e556ce2c YZ |
2567 | |
2568 | if (sb->s_flags & MS_RDONLY) { | |
79787eaa JM |
2569 | ret = btrfs_commit_super(tree_root); |
2570 | if (ret) | |
2571 | goto fail_trans_kthread; | |
e556ce2c | 2572 | } |
e02119d5 | 2573 | } |
1a40e23b | 2574 | |
76dda93c | 2575 | ret = btrfs_find_orphan_roots(tree_root); |
79787eaa JM |
2576 | if (ret) |
2577 | goto fail_trans_kthread; | |
76dda93c | 2578 | |
7c2ca468 | 2579 | if (!(sb->s_flags & MS_RDONLY)) { |
d68fc57b | 2580 | ret = btrfs_cleanup_fs_roots(fs_info); |
44c44af2 ID |
2581 | if (ret) |
2582 | goto fail_trans_kthread; | |
d68fc57b | 2583 | |
5d4f98a2 | 2584 | ret = btrfs_recover_relocation(tree_root); |
d7ce5843 MX |
2585 | if (ret < 0) { |
2586 | printk(KERN_WARNING | |
2587 | "btrfs: failed to recover relocation\n"); | |
2588 | err = -EINVAL; | |
bcef60f2 | 2589 | goto fail_qgroup; |
d7ce5843 | 2590 | } |
7c2ca468 | 2591 | } |
1a40e23b | 2592 | |
3de4586c CM |
2593 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
2594 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2595 | location.offset = (u64)-1; | |
2596 | ||
3de4586c CM |
2597 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
2598 | if (!fs_info->fs_root) | |
bcef60f2 | 2599 | goto fail_qgroup; |
3140c9a3 DC |
2600 | if (IS_ERR(fs_info->fs_root)) { |
2601 | err = PTR_ERR(fs_info->fs_root); | |
bcef60f2 | 2602 | goto fail_qgroup; |
3140c9a3 | 2603 | } |
c289811c | 2604 | |
2b6ba629 ID |
2605 | if (sb->s_flags & MS_RDONLY) |
2606 | return 0; | |
59641015 | 2607 | |
2b6ba629 ID |
2608 | down_read(&fs_info->cleanup_work_sem); |
2609 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
2610 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
e3acc2a6 | 2611 | up_read(&fs_info->cleanup_work_sem); |
2b6ba629 ID |
2612 | close_ctree(tree_root); |
2613 | return ret; | |
2614 | } | |
2615 | up_read(&fs_info->cleanup_work_sem); | |
59641015 | 2616 | |
2b6ba629 ID |
2617 | ret = btrfs_resume_balance_async(fs_info); |
2618 | if (ret) { | |
2619 | printk(KERN_WARNING "btrfs: failed to resume balance\n"); | |
2620 | close_ctree(tree_root); | |
2621 | return ret; | |
e3acc2a6 JB |
2622 | } |
2623 | ||
ad2b2c80 | 2624 | return 0; |
39279cc3 | 2625 | |
bcef60f2 AJ |
2626 | fail_qgroup: |
2627 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
2628 | fail_trans_kthread: |
2629 | kthread_stop(fs_info->transaction_kthread); | |
3f157a2f | 2630 | fail_cleaner: |
a74a4b97 | 2631 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
2632 | |
2633 | /* | |
2634 | * make sure we're done with the btree inode before we stop our | |
2635 | * kthreads | |
2636 | */ | |
2637 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
2638 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
2639 | ||
1b1d1f66 JB |
2640 | fail_block_groups: |
2641 | btrfs_free_block_groups(fs_info); | |
af31f5e5 CM |
2642 | |
2643 | fail_tree_roots: | |
2644 | free_root_pointers(fs_info, 1); | |
2645 | ||
39279cc3 | 2646 | fail_sb_buffer: |
61d92c32 | 2647 | btrfs_stop_workers(&fs_info->generic_worker); |
306c8b68 | 2648 | btrfs_stop_workers(&fs_info->readahead_workers); |
247e743c | 2649 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2650 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2651 | btrfs_stop_workers(&fs_info->workers); |
2652 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 2653 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 2654 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 2655 | btrfs_stop_workers(&fs_info->endio_write_workers); |
0cb59c99 | 2656 | btrfs_stop_workers(&fs_info->endio_freespace_worker); |
1cc127b5 | 2657 | btrfs_stop_workers(&fs_info->submit_workers); |
16cdcec7 | 2658 | btrfs_stop_workers(&fs_info->delayed_workers); |
bab39bf9 | 2659 | btrfs_stop_workers(&fs_info->caching_workers); |
16cdcec7 | 2660 | fail_alloc: |
4543df7e | 2661 | fail_iput: |
586e46e2 ID |
2662 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
2663 | ||
7c2ca468 | 2664 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
4543df7e | 2665 | iput(fs_info->btree_inode); |
ad081f14 | 2666 | fail_bdi: |
7e662854 | 2667 | bdi_destroy(&fs_info->bdi); |
76dda93c YZ |
2668 | fail_srcu: |
2669 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 2670 | fail: |
586e46e2 | 2671 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 2672 | return err; |
af31f5e5 CM |
2673 | |
2674 | recovery_tree_root: | |
af31f5e5 CM |
2675 | if (!btrfs_test_opt(tree_root, RECOVERY)) |
2676 | goto fail_tree_roots; | |
2677 | ||
2678 | free_root_pointers(fs_info, 0); | |
2679 | ||
2680 | /* don't use the log in recovery mode, it won't be valid */ | |
2681 | btrfs_set_super_log_root(disk_super, 0); | |
2682 | ||
2683 | /* we can't trust the free space cache either */ | |
2684 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2685 | ||
2686 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
2687 | &num_backups_tried, &backup_index); | |
2688 | if (ret == -1) | |
2689 | goto fail_block_groups; | |
2690 | goto retry_root_backup; | |
eb60ceac CM |
2691 | } |
2692 | ||
f2984462 CM |
2693 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
2694 | { | |
f2984462 CM |
2695 | if (uptodate) { |
2696 | set_buffer_uptodate(bh); | |
2697 | } else { | |
442a4f63 SB |
2698 | struct btrfs_device *device = (struct btrfs_device *) |
2699 | bh->b_private; | |
2700 | ||
606686ee JB |
2701 | printk_ratelimited_in_rcu(KERN_WARNING "lost page write due to " |
2702 | "I/O error on %s\n", | |
2703 | rcu_str_deref(device->name)); | |
1259ab75 CM |
2704 | /* note, we dont' set_buffer_write_io_error because we have |
2705 | * our own ways of dealing with the IO errors | |
2706 | */ | |
f2984462 | 2707 | clear_buffer_uptodate(bh); |
442a4f63 | 2708 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); |
f2984462 CM |
2709 | } |
2710 | unlock_buffer(bh); | |
2711 | put_bh(bh); | |
2712 | } | |
2713 | ||
a512bbf8 YZ |
2714 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
2715 | { | |
2716 | struct buffer_head *bh; | |
2717 | struct buffer_head *latest = NULL; | |
2718 | struct btrfs_super_block *super; | |
2719 | int i; | |
2720 | u64 transid = 0; | |
2721 | u64 bytenr; | |
2722 | ||
2723 | /* we would like to check all the supers, but that would make | |
2724 | * a btrfs mount succeed after a mkfs from a different FS. | |
2725 | * So, we need to add a special mount option to scan for | |
2726 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
2727 | */ | |
2728 | for (i = 0; i < 1; i++) { | |
2729 | bytenr = btrfs_sb_offset(i); | |
2730 | if (bytenr + 4096 >= i_size_read(bdev->bd_inode)) | |
2731 | break; | |
2732 | bh = __bread(bdev, bytenr / 4096, 4096); | |
2733 | if (!bh) | |
2734 | continue; | |
2735 | ||
2736 | super = (struct btrfs_super_block *)bh->b_data; | |
2737 | if (btrfs_super_bytenr(super) != bytenr || | |
2738 | strncmp((char *)(&super->magic), BTRFS_MAGIC, | |
2739 | sizeof(super->magic))) { | |
2740 | brelse(bh); | |
2741 | continue; | |
2742 | } | |
2743 | ||
2744 | if (!latest || btrfs_super_generation(super) > transid) { | |
2745 | brelse(latest); | |
2746 | latest = bh; | |
2747 | transid = btrfs_super_generation(super); | |
2748 | } else { | |
2749 | brelse(bh); | |
2750 | } | |
2751 | } | |
2752 | return latest; | |
2753 | } | |
2754 | ||
4eedeb75 HH |
2755 | /* |
2756 | * this should be called twice, once with wait == 0 and | |
2757 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
2758 | * we write are pinned. | |
2759 | * | |
2760 | * They are released when wait == 1 is done. | |
2761 | * max_mirrors must be the same for both runs, and it indicates how | |
2762 | * many supers on this one device should be written. | |
2763 | * | |
2764 | * max_mirrors == 0 means to write them all. | |
2765 | */ | |
a512bbf8 YZ |
2766 | static int write_dev_supers(struct btrfs_device *device, |
2767 | struct btrfs_super_block *sb, | |
2768 | int do_barriers, int wait, int max_mirrors) | |
2769 | { | |
2770 | struct buffer_head *bh; | |
2771 | int i; | |
2772 | int ret; | |
2773 | int errors = 0; | |
2774 | u32 crc; | |
2775 | u64 bytenr; | |
a512bbf8 YZ |
2776 | |
2777 | if (max_mirrors == 0) | |
2778 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
2779 | ||
a512bbf8 YZ |
2780 | for (i = 0; i < max_mirrors; i++) { |
2781 | bytenr = btrfs_sb_offset(i); | |
2782 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) | |
2783 | break; | |
2784 | ||
2785 | if (wait) { | |
2786 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
2787 | BTRFS_SUPER_INFO_SIZE); | |
2788 | BUG_ON(!bh); | |
a512bbf8 | 2789 | wait_on_buffer(bh); |
4eedeb75 HH |
2790 | if (!buffer_uptodate(bh)) |
2791 | errors++; | |
2792 | ||
2793 | /* drop our reference */ | |
2794 | brelse(bh); | |
2795 | ||
2796 | /* drop the reference from the wait == 0 run */ | |
2797 | brelse(bh); | |
2798 | continue; | |
a512bbf8 YZ |
2799 | } else { |
2800 | btrfs_set_super_bytenr(sb, bytenr); | |
2801 | ||
2802 | crc = ~(u32)0; | |
2803 | crc = btrfs_csum_data(NULL, (char *)sb + | |
2804 | BTRFS_CSUM_SIZE, crc, | |
2805 | BTRFS_SUPER_INFO_SIZE - | |
2806 | BTRFS_CSUM_SIZE); | |
2807 | btrfs_csum_final(crc, sb->csum); | |
2808 | ||
4eedeb75 HH |
2809 | /* |
2810 | * one reference for us, and we leave it for the | |
2811 | * caller | |
2812 | */ | |
a512bbf8 YZ |
2813 | bh = __getblk(device->bdev, bytenr / 4096, |
2814 | BTRFS_SUPER_INFO_SIZE); | |
2815 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); | |
2816 | ||
4eedeb75 | 2817 | /* one reference for submit_bh */ |
a512bbf8 | 2818 | get_bh(bh); |
4eedeb75 HH |
2819 | |
2820 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
2821 | lock_buffer(bh); |
2822 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
442a4f63 | 2823 | bh->b_private = device; |
a512bbf8 YZ |
2824 | } |
2825 | ||
387125fc CM |
2826 | /* |
2827 | * we fua the first super. The others we allow | |
2828 | * to go down lazy. | |
2829 | */ | |
21adbd5c | 2830 | ret = btrfsic_submit_bh(WRITE_FUA, bh); |
4eedeb75 | 2831 | if (ret) |
a512bbf8 | 2832 | errors++; |
a512bbf8 YZ |
2833 | } |
2834 | return errors < i ? 0 : -1; | |
2835 | } | |
2836 | ||
387125fc CM |
2837 | /* |
2838 | * endio for the write_dev_flush, this will wake anyone waiting | |
2839 | * for the barrier when it is done | |
2840 | */ | |
2841 | static void btrfs_end_empty_barrier(struct bio *bio, int err) | |
2842 | { | |
2843 | if (err) { | |
2844 | if (err == -EOPNOTSUPP) | |
2845 | set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); | |
2846 | clear_bit(BIO_UPTODATE, &bio->bi_flags); | |
2847 | } | |
2848 | if (bio->bi_private) | |
2849 | complete(bio->bi_private); | |
2850 | bio_put(bio); | |
2851 | } | |
2852 | ||
2853 | /* | |
2854 | * trigger flushes for one the devices. If you pass wait == 0, the flushes are | |
2855 | * sent down. With wait == 1, it waits for the previous flush. | |
2856 | * | |
2857 | * any device where the flush fails with eopnotsupp are flagged as not-barrier | |
2858 | * capable | |
2859 | */ | |
2860 | static int write_dev_flush(struct btrfs_device *device, int wait) | |
2861 | { | |
2862 | struct bio *bio; | |
2863 | int ret = 0; | |
2864 | ||
2865 | if (device->nobarriers) | |
2866 | return 0; | |
2867 | ||
2868 | if (wait) { | |
2869 | bio = device->flush_bio; | |
2870 | if (!bio) | |
2871 | return 0; | |
2872 | ||
2873 | wait_for_completion(&device->flush_wait); | |
2874 | ||
2875 | if (bio_flagged(bio, BIO_EOPNOTSUPP)) { | |
606686ee JB |
2876 | printk_in_rcu("btrfs: disabling barriers on dev %s\n", |
2877 | rcu_str_deref(device->name)); | |
387125fc CM |
2878 | device->nobarriers = 1; |
2879 | } | |
2880 | if (!bio_flagged(bio, BIO_UPTODATE)) { | |
2881 | ret = -EIO; | |
442a4f63 SB |
2882 | if (!bio_flagged(bio, BIO_EOPNOTSUPP)) |
2883 | btrfs_dev_stat_inc_and_print(device, | |
2884 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
387125fc CM |
2885 | } |
2886 | ||
2887 | /* drop the reference from the wait == 0 run */ | |
2888 | bio_put(bio); | |
2889 | device->flush_bio = NULL; | |
2890 | ||
2891 | return ret; | |
2892 | } | |
2893 | ||
2894 | /* | |
2895 | * one reference for us, and we leave it for the | |
2896 | * caller | |
2897 | */ | |
9c017abc | 2898 | device->flush_bio = NULL; |
387125fc CM |
2899 | bio = bio_alloc(GFP_NOFS, 0); |
2900 | if (!bio) | |
2901 | return -ENOMEM; | |
2902 | ||
2903 | bio->bi_end_io = btrfs_end_empty_barrier; | |
2904 | bio->bi_bdev = device->bdev; | |
2905 | init_completion(&device->flush_wait); | |
2906 | bio->bi_private = &device->flush_wait; | |
2907 | device->flush_bio = bio; | |
2908 | ||
2909 | bio_get(bio); | |
21adbd5c | 2910 | btrfsic_submit_bio(WRITE_FLUSH, bio); |
387125fc CM |
2911 | |
2912 | return 0; | |
2913 | } | |
2914 | ||
2915 | /* | |
2916 | * send an empty flush down to each device in parallel, | |
2917 | * then wait for them | |
2918 | */ | |
2919 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
2920 | { | |
2921 | struct list_head *head; | |
2922 | struct btrfs_device *dev; | |
2923 | int errors = 0; | |
2924 | int ret; | |
2925 | ||
2926 | /* send down all the barriers */ | |
2927 | head = &info->fs_devices->devices; | |
2928 | list_for_each_entry_rcu(dev, head, dev_list) { | |
2929 | if (!dev->bdev) { | |
2930 | errors++; | |
2931 | continue; | |
2932 | } | |
2933 | if (!dev->in_fs_metadata || !dev->writeable) | |
2934 | continue; | |
2935 | ||
2936 | ret = write_dev_flush(dev, 0); | |
2937 | if (ret) | |
2938 | errors++; | |
2939 | } | |
2940 | ||
2941 | /* wait for all the barriers */ | |
2942 | list_for_each_entry_rcu(dev, head, dev_list) { | |
2943 | if (!dev->bdev) { | |
2944 | errors++; | |
2945 | continue; | |
2946 | } | |
2947 | if (!dev->in_fs_metadata || !dev->writeable) | |
2948 | continue; | |
2949 | ||
2950 | ret = write_dev_flush(dev, 1); | |
2951 | if (ret) | |
2952 | errors++; | |
2953 | } | |
2954 | if (errors) | |
2955 | return -EIO; | |
2956 | return 0; | |
2957 | } | |
2958 | ||
a512bbf8 | 2959 | int write_all_supers(struct btrfs_root *root, int max_mirrors) |
f2984462 | 2960 | { |
e5e9a520 | 2961 | struct list_head *head; |
f2984462 | 2962 | struct btrfs_device *dev; |
a061fc8d | 2963 | struct btrfs_super_block *sb; |
f2984462 | 2964 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
2965 | int ret; |
2966 | int do_barriers; | |
a236aed1 CM |
2967 | int max_errors; |
2968 | int total_errors = 0; | |
a061fc8d | 2969 | u64 flags; |
f2984462 | 2970 | |
6c41761f | 2971 | max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; |
f2984462 | 2972 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
af31f5e5 | 2973 | backup_super_roots(root->fs_info); |
f2984462 | 2974 | |
6c41761f | 2975 | sb = root->fs_info->super_for_commit; |
a061fc8d | 2976 | dev_item = &sb->dev_item; |
e5e9a520 | 2977 | |
174ba509 | 2978 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); |
e5e9a520 | 2979 | head = &root->fs_info->fs_devices->devices; |
387125fc CM |
2980 | |
2981 | if (do_barriers) | |
2982 | barrier_all_devices(root->fs_info); | |
2983 | ||
1f78160c | 2984 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
2985 | if (!dev->bdev) { |
2986 | total_errors++; | |
2987 | continue; | |
2988 | } | |
2b82032c | 2989 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2990 | continue; |
2991 | ||
2b82032c | 2992 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
2993 | btrfs_set_stack_device_type(dev_item, dev->type); |
2994 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
2995 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
2996 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
2997 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
2998 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
2999 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
3000 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 3001 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 3002 | |
a061fc8d CM |
3003 | flags = btrfs_super_flags(sb); |
3004 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
3005 | ||
a512bbf8 | 3006 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
3007 | if (ret) |
3008 | total_errors++; | |
f2984462 | 3009 | } |
a236aed1 | 3010 | if (total_errors > max_errors) { |
d397712b CM |
3011 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
3012 | total_errors); | |
79787eaa JM |
3013 | |
3014 | /* This shouldn't happen. FUA is masked off if unsupported */ | |
a236aed1 CM |
3015 | BUG(); |
3016 | } | |
f2984462 | 3017 | |
a512bbf8 | 3018 | total_errors = 0; |
1f78160c | 3019 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3020 | if (!dev->bdev) |
3021 | continue; | |
2b82032c | 3022 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3023 | continue; |
3024 | ||
a512bbf8 YZ |
3025 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
3026 | if (ret) | |
3027 | total_errors++; | |
f2984462 | 3028 | } |
174ba509 | 3029 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 3030 | if (total_errors > max_errors) { |
79787eaa JM |
3031 | btrfs_error(root->fs_info, -EIO, |
3032 | "%d errors while writing supers", total_errors); | |
3033 | return -EIO; | |
a236aed1 | 3034 | } |
f2984462 CM |
3035 | return 0; |
3036 | } | |
3037 | ||
a512bbf8 YZ |
3038 | int write_ctree_super(struct btrfs_trans_handle *trans, |
3039 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 3040 | { |
e66f709b | 3041 | int ret; |
5f39d397 | 3042 | |
a512bbf8 | 3043 | ret = write_all_supers(root, max_mirrors); |
5f39d397 | 3044 | return ret; |
cfaa7295 CM |
3045 | } |
3046 | ||
143bede5 | 3047 | void btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f | 3048 | { |
4df27c4d | 3049 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
3050 | radix_tree_delete(&fs_info->fs_roots_radix, |
3051 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 3052 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
3053 | |
3054 | if (btrfs_root_refs(&root->root_item) == 0) | |
3055 | synchronize_srcu(&fs_info->subvol_srcu); | |
3056 | ||
581bb050 LZ |
3057 | __btrfs_remove_free_space_cache(root->free_ino_pinned); |
3058 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d | 3059 | free_fs_root(root); |
4df27c4d YZ |
3060 | } |
3061 | ||
3062 | static void free_fs_root(struct btrfs_root *root) | |
3063 | { | |
82d5902d | 3064 | iput(root->cache_inode); |
4df27c4d | 3065 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
0ee5dc67 AV |
3066 | if (root->anon_dev) |
3067 | free_anon_bdev(root->anon_dev); | |
4df27c4d YZ |
3068 | free_extent_buffer(root->node); |
3069 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
3070 | kfree(root->free_ino_ctl); |
3071 | kfree(root->free_ino_pinned); | |
d397712b | 3072 | kfree(root->name); |
2619ba1f | 3073 | kfree(root); |
2619ba1f CM |
3074 | } |
3075 | ||
143bede5 | 3076 | static void del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
3077 | { |
3078 | int ret; | |
3079 | struct btrfs_root *gang[8]; | |
3080 | int i; | |
3081 | ||
76dda93c YZ |
3082 | while (!list_empty(&fs_info->dead_roots)) { |
3083 | gang[0] = list_entry(fs_info->dead_roots.next, | |
3084 | struct btrfs_root, root_list); | |
3085 | list_del(&gang[0]->root_list); | |
3086 | ||
3087 | if (gang[0]->in_radix) { | |
3088 | btrfs_free_fs_root(fs_info, gang[0]); | |
3089 | } else { | |
3090 | free_extent_buffer(gang[0]->node); | |
3091 | free_extent_buffer(gang[0]->commit_root); | |
3092 | kfree(gang[0]); | |
3093 | } | |
3094 | } | |
3095 | ||
d397712b | 3096 | while (1) { |
0f7d52f4 CM |
3097 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
3098 | (void **)gang, 0, | |
3099 | ARRAY_SIZE(gang)); | |
3100 | if (!ret) | |
3101 | break; | |
2619ba1f | 3102 | for (i = 0; i < ret; i++) |
5eda7b5e | 3103 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 | 3104 | } |
0f7d52f4 | 3105 | } |
b4100d64 | 3106 | |
c146afad | 3107 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 3108 | { |
c146afad YZ |
3109 | u64 root_objectid = 0; |
3110 | struct btrfs_root *gang[8]; | |
3111 | int i; | |
3768f368 | 3112 | int ret; |
e089f05c | 3113 | |
c146afad YZ |
3114 | while (1) { |
3115 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
3116 | (void **)gang, root_objectid, | |
3117 | ARRAY_SIZE(gang)); | |
3118 | if (!ret) | |
3119 | break; | |
5d4f98a2 YZ |
3120 | |
3121 | root_objectid = gang[ret - 1]->root_key.objectid + 1; | |
c146afad | 3122 | for (i = 0; i < ret; i++) { |
66b4ffd1 JB |
3123 | int err; |
3124 | ||
c146afad | 3125 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
3126 | err = btrfs_orphan_cleanup(gang[i]); |
3127 | if (err) | |
3128 | return err; | |
c146afad YZ |
3129 | } |
3130 | root_objectid++; | |
3131 | } | |
3132 | return 0; | |
3133 | } | |
a2135011 | 3134 | |
c146afad YZ |
3135 | int btrfs_commit_super(struct btrfs_root *root) |
3136 | { | |
3137 | struct btrfs_trans_handle *trans; | |
3138 | int ret; | |
a74a4b97 | 3139 | |
c146afad | 3140 | mutex_lock(&root->fs_info->cleaner_mutex); |
24bbcf04 | 3141 | btrfs_run_delayed_iputs(root); |
a74a4b97 | 3142 | btrfs_clean_old_snapshots(root); |
c146afad | 3143 | mutex_unlock(&root->fs_info->cleaner_mutex); |
c71bf099 YZ |
3144 | |
3145 | /* wait until ongoing cleanup work done */ | |
3146 | down_write(&root->fs_info->cleanup_work_sem); | |
3147 | up_write(&root->fs_info->cleanup_work_sem); | |
3148 | ||
7a7eaa40 | 3149 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3150 | if (IS_ERR(trans)) |
3151 | return PTR_ERR(trans); | |
54aa1f4d | 3152 | ret = btrfs_commit_transaction(trans, root); |
79787eaa JM |
3153 | if (ret) |
3154 | return ret; | |
c146afad | 3155 | /* run commit again to drop the original snapshot */ |
7a7eaa40 | 3156 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3157 | if (IS_ERR(trans)) |
3158 | return PTR_ERR(trans); | |
79787eaa JM |
3159 | ret = btrfs_commit_transaction(trans, root); |
3160 | if (ret) | |
3161 | return ret; | |
79154b1b | 3162 | ret = btrfs_write_and_wait_transaction(NULL, root); |
79787eaa JM |
3163 | if (ret) { |
3164 | btrfs_error(root->fs_info, ret, | |
3165 | "Failed to sync btree inode to disk."); | |
3166 | return ret; | |
3167 | } | |
d6bfde87 | 3168 | |
a512bbf8 | 3169 | ret = write_ctree_super(NULL, root, 0); |
c146afad YZ |
3170 | return ret; |
3171 | } | |
3172 | ||
3173 | int close_ctree(struct btrfs_root *root) | |
3174 | { | |
3175 | struct btrfs_fs_info *fs_info = root->fs_info; | |
3176 | int ret; | |
3177 | ||
3178 | fs_info->closing = 1; | |
3179 | smp_mb(); | |
3180 | ||
837d5b6e ID |
3181 | /* pause restriper - we want to resume on mount */ |
3182 | btrfs_pause_balance(root->fs_info); | |
3183 | ||
a2de733c | 3184 | btrfs_scrub_cancel(root); |
4cb5300b CM |
3185 | |
3186 | /* wait for any defraggers to finish */ | |
3187 | wait_event(fs_info->transaction_wait, | |
3188 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3189 | ||
3190 | /* clear out the rbtree of defraggable inodes */ | |
e3029d9f | 3191 | btrfs_run_defrag_inodes(fs_info); |
4cb5300b | 3192 | |
acce952b | 3193 | /* |
3194 | * Here come 2 situations when btrfs is broken to flip readonly: | |
3195 | * | |
3196 | * 1. when btrfs flips readonly somewhere else before | |
3197 | * btrfs_commit_super, sb->s_flags has MS_RDONLY flag, | |
3198 | * and btrfs will skip to write sb directly to keep | |
3199 | * ERROR state on disk. | |
3200 | * | |
3201 | * 2. when btrfs flips readonly just in btrfs_commit_super, | |
ae0e47f0 | 3202 | * and in such case, btrfs cannot write sb via btrfs_commit_super, |
acce952b | 3203 | * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag, |
3204 | * btrfs will cleanup all FS resources first and write sb then. | |
3205 | */ | |
c146afad | 3206 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
acce952b | 3207 | ret = btrfs_commit_super(root); |
3208 | if (ret) | |
3209 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
3210 | } | |
3211 | ||
3212 | if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) { | |
3213 | ret = btrfs_error_commit_super(root); | |
d397712b CM |
3214 | if (ret) |
3215 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
c146afad | 3216 | } |
0f7d52f4 | 3217 | |
300e4f8a JB |
3218 | btrfs_put_block_group_cache(fs_info); |
3219 | ||
e3029d9f AV |
3220 | kthread_stop(fs_info->transaction_kthread); |
3221 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 3222 | |
f25784b3 YZ |
3223 | fs_info->closing = 2; |
3224 | smp_mb(); | |
3225 | ||
bcef60f2 AJ |
3226 | btrfs_free_qgroup_config(root->fs_info); |
3227 | ||
b0c68f8b | 3228 | if (fs_info->delalloc_bytes) { |
d397712b | 3229 | printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n", |
21380931 | 3230 | (unsigned long long)fs_info->delalloc_bytes); |
b0c68f8b | 3231 | } |
31153d81 | 3232 | if (fs_info->total_ref_cache_size) { |
d397712b CM |
3233 | printk(KERN_INFO "btrfs: at umount reference cache size %llu\n", |
3234 | (unsigned long long)fs_info->total_ref_cache_size); | |
31153d81 | 3235 | } |
bcc63abb | 3236 | |
5d4f98a2 YZ |
3237 | free_extent_buffer(fs_info->extent_root->node); |
3238 | free_extent_buffer(fs_info->extent_root->commit_root); | |
3239 | free_extent_buffer(fs_info->tree_root->node); | |
3240 | free_extent_buffer(fs_info->tree_root->commit_root); | |
e3029d9f AV |
3241 | free_extent_buffer(fs_info->chunk_root->node); |
3242 | free_extent_buffer(fs_info->chunk_root->commit_root); | |
3243 | free_extent_buffer(fs_info->dev_root->node); | |
3244 | free_extent_buffer(fs_info->dev_root->commit_root); | |
3245 | free_extent_buffer(fs_info->csum_root->node); | |
3246 | free_extent_buffer(fs_info->csum_root->commit_root); | |
bcef60f2 AJ |
3247 | if (fs_info->quota_root) { |
3248 | free_extent_buffer(fs_info->quota_root->node); | |
3249 | free_extent_buffer(fs_info->quota_root->commit_root); | |
3250 | } | |
d20f7043 | 3251 | |
e3029d9f | 3252 | btrfs_free_block_groups(fs_info); |
d10c5f31 | 3253 | |
c146afad | 3254 | del_fs_roots(fs_info); |
d10c5f31 | 3255 | |
c146afad | 3256 | iput(fs_info->btree_inode); |
9ad6b7bc | 3257 | |
61d92c32 | 3258 | btrfs_stop_workers(&fs_info->generic_worker); |
247e743c | 3259 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 3260 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
3261 | btrfs_stop_workers(&fs_info->workers); |
3262 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 3263 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 3264 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 3265 | btrfs_stop_workers(&fs_info->endio_write_workers); |
0cb59c99 | 3266 | btrfs_stop_workers(&fs_info->endio_freespace_worker); |
1cc127b5 | 3267 | btrfs_stop_workers(&fs_info->submit_workers); |
16cdcec7 | 3268 | btrfs_stop_workers(&fs_info->delayed_workers); |
bab39bf9 | 3269 | btrfs_stop_workers(&fs_info->caching_workers); |
90519d66 | 3270 | btrfs_stop_workers(&fs_info->readahead_workers); |
d6bfde87 | 3271 | |
21adbd5c SB |
3272 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3273 | if (btrfs_test_opt(root, CHECK_INTEGRITY)) | |
3274 | btrfsic_unmount(root, fs_info->fs_devices); | |
3275 | #endif | |
3276 | ||
dfe25020 | 3277 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 3278 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 3279 | |
04160088 | 3280 | bdi_destroy(&fs_info->bdi); |
76dda93c | 3281 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 3282 | |
eb60ceac CM |
3283 | return 0; |
3284 | } | |
3285 | ||
b9fab919 CM |
3286 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
3287 | int atomic) | |
5f39d397 | 3288 | { |
1259ab75 | 3289 | int ret; |
727011e0 | 3290 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 3291 | |
0b32f4bb | 3292 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
3293 | if (!ret) |
3294 | return ret; | |
3295 | ||
3296 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
3297 | parent_transid, atomic); |
3298 | if (ret == -EAGAIN) | |
3299 | return ret; | |
1259ab75 | 3300 | return !ret; |
5f39d397 CM |
3301 | } |
3302 | ||
3303 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 3304 | { |
0b32f4bb | 3305 | return set_extent_buffer_uptodate(buf); |
5f39d397 | 3306 | } |
6702ed49 | 3307 | |
5f39d397 CM |
3308 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
3309 | { | |
727011e0 | 3310 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
5f39d397 | 3311 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 3312 | int was_dirty; |
b4ce94de | 3313 | |
b9447ef8 | 3314 | btrfs_assert_tree_locked(buf); |
ccd467d6 | 3315 | if (transid != root->fs_info->generation) { |
d397712b CM |
3316 | printk(KERN_CRIT "btrfs transid mismatch buffer %llu, " |
3317 | "found %llu running %llu\n", | |
db94535d | 3318 | (unsigned long long)buf->start, |
d397712b CM |
3319 | (unsigned long long)transid, |
3320 | (unsigned long long)root->fs_info->generation); | |
ccd467d6 CM |
3321 | WARN_ON(1); |
3322 | } | |
0b32f4bb | 3323 | was_dirty = set_extent_buffer_dirty(buf); |
b9473439 CM |
3324 | if (!was_dirty) { |
3325 | spin_lock(&root->fs_info->delalloc_lock); | |
3326 | root->fs_info->dirty_metadata_bytes += buf->len; | |
3327 | spin_unlock(&root->fs_info->delalloc_lock); | |
3328 | } | |
eb60ceac CM |
3329 | } |
3330 | ||
d3c2fdcf | 3331 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
16cdcec7 MX |
3332 | { |
3333 | /* | |
3334 | * looks as though older kernels can get into trouble with | |
3335 | * this code, they end up stuck in balance_dirty_pages forever | |
3336 | */ | |
3337 | u64 num_dirty; | |
3338 | unsigned long thresh = 32 * 1024 * 1024; | |
3339 | ||
3340 | if (current->flags & PF_MEMALLOC) | |
3341 | return; | |
3342 | ||
3343 | btrfs_balance_delayed_items(root); | |
3344 | ||
3345 | num_dirty = root->fs_info->dirty_metadata_bytes; | |
3346 | ||
3347 | if (num_dirty > thresh) { | |
3348 | balance_dirty_pages_ratelimited_nr( | |
3349 | root->fs_info->btree_inode->i_mapping, 1); | |
3350 | } | |
3351 | return; | |
3352 | } | |
3353 | ||
3354 | void __btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) | |
35b7e476 | 3355 | { |
188de649 CM |
3356 | /* |
3357 | * looks as though older kernels can get into trouble with | |
3358 | * this code, they end up stuck in balance_dirty_pages forever | |
3359 | */ | |
d6bfde87 | 3360 | u64 num_dirty; |
771ed689 | 3361 | unsigned long thresh = 32 * 1024 * 1024; |
d6bfde87 | 3362 | |
6933c02e | 3363 | if (current->flags & PF_MEMALLOC) |
d6bfde87 CM |
3364 | return; |
3365 | ||
585ad2c3 CM |
3366 | num_dirty = root->fs_info->dirty_metadata_bytes; |
3367 | ||
d6bfde87 CM |
3368 | if (num_dirty > thresh) { |
3369 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 3370 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 3371 | } |
188de649 | 3372 | return; |
35b7e476 | 3373 | } |
6b80053d | 3374 | |
ca7a79ad | 3375 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 3376 | { |
727011e0 | 3377 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
0b32f4bb | 3378 | return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
6b80053d | 3379 | } |
0da5468f | 3380 | |
20897f5c | 3381 | int btree_lock_page_hook(struct page *page, void *data, |
01d658f2 | 3382 | void (*flush_fn)(void *)) |
4bef0848 CM |
3383 | { |
3384 | struct inode *inode = page->mapping->host; | |
b9473439 | 3385 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4bef0848 | 3386 | struct extent_buffer *eb; |
4bef0848 | 3387 | |
4f2de97a JB |
3388 | /* |
3389 | * We culled this eb but the page is still hanging out on the mapping, | |
3390 | * carry on. | |
3391 | */ | |
3392 | if (!PagePrivate(page)) | |
4bef0848 CM |
3393 | goto out; |
3394 | ||
4f2de97a JB |
3395 | eb = (struct extent_buffer *)page->private; |
3396 | if (!eb) { | |
3397 | WARN_ON(1); | |
3398 | goto out; | |
3399 | } | |
3400 | if (page != eb->pages[0]) | |
4bef0848 CM |
3401 | goto out; |
3402 | ||
01d658f2 CM |
3403 | if (!btrfs_try_tree_write_lock(eb)) { |
3404 | flush_fn(data); | |
3405 | btrfs_tree_lock(eb); | |
3406 | } | |
4bef0848 | 3407 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); |
b9473439 CM |
3408 | |
3409 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) { | |
3410 | spin_lock(&root->fs_info->delalloc_lock); | |
3411 | if (root->fs_info->dirty_metadata_bytes >= eb->len) | |
3412 | root->fs_info->dirty_metadata_bytes -= eb->len; | |
3413 | else | |
3414 | WARN_ON(1); | |
3415 | spin_unlock(&root->fs_info->delalloc_lock); | |
3416 | } | |
3417 | ||
4bef0848 | 3418 | btrfs_tree_unlock(eb); |
4bef0848 | 3419 | out: |
01d658f2 CM |
3420 | if (!trylock_page(page)) { |
3421 | flush_fn(data); | |
3422 | lock_page(page); | |
3423 | } | |
4bef0848 CM |
3424 | return 0; |
3425 | } | |
3426 | ||
fcd1f065 | 3427 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 3428 | int read_only) |
3429 | { | |
fcd1f065 DS |
3430 | if (btrfs_super_csum_type(fs_info->super_copy) >= ARRAY_SIZE(btrfs_csum_sizes)) { |
3431 | printk(KERN_ERR "btrfs: unsupported checksum algorithm\n"); | |
3432 | return -EINVAL; | |
3433 | } | |
3434 | ||
acce952b | 3435 | if (read_only) |
fcd1f065 | 3436 | return 0; |
acce952b | 3437 | |
fcd1f065 | 3438 | if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) { |
acce952b | 3439 | printk(KERN_WARNING "warning: mount fs with errors, " |
3440 | "running btrfsck is recommended\n"); | |
fcd1f065 DS |
3441 | } |
3442 | ||
3443 | return 0; | |
acce952b | 3444 | } |
3445 | ||
3446 | int btrfs_error_commit_super(struct btrfs_root *root) | |
3447 | { | |
3448 | int ret; | |
3449 | ||
3450 | mutex_lock(&root->fs_info->cleaner_mutex); | |
3451 | btrfs_run_delayed_iputs(root); | |
3452 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
3453 | ||
3454 | down_write(&root->fs_info->cleanup_work_sem); | |
3455 | up_write(&root->fs_info->cleanup_work_sem); | |
3456 | ||
3457 | /* cleanup FS via transaction */ | |
3458 | btrfs_cleanup_transaction(root); | |
3459 | ||
3460 | ret = write_ctree_super(NULL, root, 0); | |
3461 | ||
3462 | return ret; | |
3463 | } | |
3464 | ||
143bede5 | 3465 | static void btrfs_destroy_ordered_operations(struct btrfs_root *root) |
acce952b | 3466 | { |
3467 | struct btrfs_inode *btrfs_inode; | |
3468 | struct list_head splice; | |
3469 | ||
3470 | INIT_LIST_HEAD(&splice); | |
3471 | ||
3472 | mutex_lock(&root->fs_info->ordered_operations_mutex); | |
3473 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3474 | ||
3475 | list_splice_init(&root->fs_info->ordered_operations, &splice); | |
3476 | while (!list_empty(&splice)) { | |
3477 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3478 | ordered_operations); | |
3479 | ||
3480 | list_del_init(&btrfs_inode->ordered_operations); | |
3481 | ||
3482 | btrfs_invalidate_inodes(btrfs_inode->root); | |
3483 | } | |
3484 | ||
3485 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3486 | mutex_unlock(&root->fs_info->ordered_operations_mutex); | |
acce952b | 3487 | } |
3488 | ||
143bede5 | 3489 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 3490 | { |
3491 | struct list_head splice; | |
3492 | struct btrfs_ordered_extent *ordered; | |
3493 | struct inode *inode; | |
3494 | ||
3495 | INIT_LIST_HEAD(&splice); | |
3496 | ||
3497 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3498 | ||
3499 | list_splice_init(&root->fs_info->ordered_extents, &splice); | |
3500 | while (!list_empty(&splice)) { | |
3501 | ordered = list_entry(splice.next, struct btrfs_ordered_extent, | |
3502 | root_extent_list); | |
3503 | ||
3504 | list_del_init(&ordered->root_extent_list); | |
3505 | atomic_inc(&ordered->refs); | |
3506 | ||
3507 | /* the inode may be getting freed (in sys_unlink path). */ | |
3508 | inode = igrab(ordered->inode); | |
3509 | ||
3510 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3511 | if (inode) | |
3512 | iput(inode); | |
3513 | ||
3514 | atomic_set(&ordered->refs, 1); | |
3515 | btrfs_put_ordered_extent(ordered); | |
3516 | ||
3517 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3518 | } | |
3519 | ||
3520 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
acce952b | 3521 | } |
3522 | ||
79787eaa JM |
3523 | int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
3524 | struct btrfs_root *root) | |
acce952b | 3525 | { |
3526 | struct rb_node *node; | |
3527 | struct btrfs_delayed_ref_root *delayed_refs; | |
3528 | struct btrfs_delayed_ref_node *ref; | |
3529 | int ret = 0; | |
3530 | ||
3531 | delayed_refs = &trans->delayed_refs; | |
3532 | ||
3533 | spin_lock(&delayed_refs->lock); | |
3534 | if (delayed_refs->num_entries == 0) { | |
cfece4db | 3535 | spin_unlock(&delayed_refs->lock); |
acce952b | 3536 | printk(KERN_INFO "delayed_refs has NO entry\n"); |
3537 | return ret; | |
3538 | } | |
3539 | ||
b939d1ab | 3540 | while ((node = rb_first(&delayed_refs->root)) != NULL) { |
acce952b | 3541 | ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); |
acce952b | 3542 | |
3543 | atomic_set(&ref->refs, 1); | |
3544 | if (btrfs_delayed_ref_is_head(ref)) { | |
3545 | struct btrfs_delayed_ref_head *head; | |
3546 | ||
3547 | head = btrfs_delayed_node_to_head(ref); | |
b939d1ab JB |
3548 | if (!mutex_trylock(&head->mutex)) { |
3549 | atomic_inc(&ref->refs); | |
3550 | spin_unlock(&delayed_refs->lock); | |
3551 | ||
3552 | /* Need to wait for the delayed ref to run */ | |
3553 | mutex_lock(&head->mutex); | |
3554 | mutex_unlock(&head->mutex); | |
3555 | btrfs_put_delayed_ref(ref); | |
3556 | ||
e18fca73 | 3557 | spin_lock(&delayed_refs->lock); |
b939d1ab JB |
3558 | continue; |
3559 | } | |
3560 | ||
acce952b | 3561 | kfree(head->extent_op); |
3562 | delayed_refs->num_heads--; | |
3563 | if (list_empty(&head->cluster)) | |
3564 | delayed_refs->num_heads_ready--; | |
3565 | list_del_init(&head->cluster); | |
acce952b | 3566 | } |
b939d1ab JB |
3567 | ref->in_tree = 0; |
3568 | rb_erase(&ref->rb_node, &delayed_refs->root); | |
3569 | delayed_refs->num_entries--; | |
3570 | ||
acce952b | 3571 | spin_unlock(&delayed_refs->lock); |
3572 | btrfs_put_delayed_ref(ref); | |
3573 | ||
3574 | cond_resched(); | |
3575 | spin_lock(&delayed_refs->lock); | |
3576 | } | |
3577 | ||
3578 | spin_unlock(&delayed_refs->lock); | |
3579 | ||
3580 | return ret; | |
3581 | } | |
3582 | ||
143bede5 | 3583 | static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t) |
acce952b | 3584 | { |
3585 | struct btrfs_pending_snapshot *snapshot; | |
3586 | struct list_head splice; | |
3587 | ||
3588 | INIT_LIST_HEAD(&splice); | |
3589 | ||
3590 | list_splice_init(&t->pending_snapshots, &splice); | |
3591 | ||
3592 | while (!list_empty(&splice)) { | |
3593 | snapshot = list_entry(splice.next, | |
3594 | struct btrfs_pending_snapshot, | |
3595 | list); | |
3596 | ||
3597 | list_del_init(&snapshot->list); | |
3598 | ||
3599 | kfree(snapshot); | |
3600 | } | |
acce952b | 3601 | } |
3602 | ||
143bede5 | 3603 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 3604 | { |
3605 | struct btrfs_inode *btrfs_inode; | |
3606 | struct list_head splice; | |
3607 | ||
3608 | INIT_LIST_HEAD(&splice); | |
3609 | ||
acce952b | 3610 | spin_lock(&root->fs_info->delalloc_lock); |
5be76758 | 3611 | list_splice_init(&root->fs_info->delalloc_inodes, &splice); |
acce952b | 3612 | |
3613 | while (!list_empty(&splice)) { | |
3614 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3615 | delalloc_inodes); | |
3616 | ||
3617 | list_del_init(&btrfs_inode->delalloc_inodes); | |
3618 | ||
3619 | btrfs_invalidate_inodes(btrfs_inode->root); | |
3620 | } | |
3621 | ||
3622 | spin_unlock(&root->fs_info->delalloc_lock); | |
acce952b | 3623 | } |
3624 | ||
3625 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, | |
3626 | struct extent_io_tree *dirty_pages, | |
3627 | int mark) | |
3628 | { | |
3629 | int ret; | |
3630 | struct page *page; | |
3631 | struct inode *btree_inode = root->fs_info->btree_inode; | |
3632 | struct extent_buffer *eb; | |
3633 | u64 start = 0; | |
3634 | u64 end; | |
3635 | u64 offset; | |
3636 | unsigned long index; | |
3637 | ||
3638 | while (1) { | |
3639 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
3640 | mark); | |
3641 | if (ret) | |
3642 | break; | |
3643 | ||
3644 | clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); | |
3645 | while (start <= end) { | |
3646 | index = start >> PAGE_CACHE_SHIFT; | |
3647 | start = (u64)(index + 1) << PAGE_CACHE_SHIFT; | |
3648 | page = find_get_page(btree_inode->i_mapping, index); | |
3649 | if (!page) | |
3650 | continue; | |
3651 | offset = page_offset(page); | |
3652 | ||
3653 | spin_lock(&dirty_pages->buffer_lock); | |
3654 | eb = radix_tree_lookup( | |
3655 | &(&BTRFS_I(page->mapping->host)->io_tree)->buffer, | |
3656 | offset >> PAGE_CACHE_SHIFT); | |
3657 | spin_unlock(&dirty_pages->buffer_lock); | |
ee670f0a | 3658 | if (eb) |
acce952b | 3659 | ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
3660 | &eb->bflags); | |
acce952b | 3661 | if (PageWriteback(page)) |
3662 | end_page_writeback(page); | |
3663 | ||
3664 | lock_page(page); | |
3665 | if (PageDirty(page)) { | |
3666 | clear_page_dirty_for_io(page); | |
3667 | spin_lock_irq(&page->mapping->tree_lock); | |
3668 | radix_tree_tag_clear(&page->mapping->page_tree, | |
3669 | page_index(page), | |
3670 | PAGECACHE_TAG_DIRTY); | |
3671 | spin_unlock_irq(&page->mapping->tree_lock); | |
3672 | } | |
3673 | ||
acce952b | 3674 | unlock_page(page); |
ee670f0a | 3675 | page_cache_release(page); |
acce952b | 3676 | } |
3677 | } | |
3678 | ||
3679 | return ret; | |
3680 | } | |
3681 | ||
3682 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
3683 | struct extent_io_tree *pinned_extents) | |
3684 | { | |
3685 | struct extent_io_tree *unpin; | |
3686 | u64 start; | |
3687 | u64 end; | |
3688 | int ret; | |
ed0eaa14 | 3689 | bool loop = true; |
acce952b | 3690 | |
3691 | unpin = pinned_extents; | |
ed0eaa14 | 3692 | again: |
acce952b | 3693 | while (1) { |
3694 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
3695 | EXTENT_DIRTY); | |
3696 | if (ret) | |
3697 | break; | |
3698 | ||
3699 | /* opt_discard */ | |
5378e607 LD |
3700 | if (btrfs_test_opt(root, DISCARD)) |
3701 | ret = btrfs_error_discard_extent(root, start, | |
3702 | end + 1 - start, | |
3703 | NULL); | |
acce952b | 3704 | |
3705 | clear_extent_dirty(unpin, start, end, GFP_NOFS); | |
3706 | btrfs_error_unpin_extent_range(root, start, end); | |
3707 | cond_resched(); | |
3708 | } | |
3709 | ||
ed0eaa14 LB |
3710 | if (loop) { |
3711 | if (unpin == &root->fs_info->freed_extents[0]) | |
3712 | unpin = &root->fs_info->freed_extents[1]; | |
3713 | else | |
3714 | unpin = &root->fs_info->freed_extents[0]; | |
3715 | loop = false; | |
3716 | goto again; | |
3717 | } | |
3718 | ||
acce952b | 3719 | return 0; |
3720 | } | |
3721 | ||
49b25e05 JM |
3722 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
3723 | struct btrfs_root *root) | |
3724 | { | |
3725 | btrfs_destroy_delayed_refs(cur_trans, root); | |
3726 | btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv, | |
3727 | cur_trans->dirty_pages.dirty_bytes); | |
3728 | ||
3729 | /* FIXME: cleanup wait for commit */ | |
3730 | cur_trans->in_commit = 1; | |
3731 | cur_trans->blocked = 1; | |
d7096fc3 | 3732 | wake_up(&root->fs_info->transaction_blocked_wait); |
49b25e05 JM |
3733 | |
3734 | cur_trans->blocked = 0; | |
d7096fc3 | 3735 | wake_up(&root->fs_info->transaction_wait); |
49b25e05 JM |
3736 | |
3737 | cur_trans->commit_done = 1; | |
d7096fc3 | 3738 | wake_up(&cur_trans->commit_wait); |
49b25e05 | 3739 | |
67cde344 MX |
3740 | btrfs_destroy_delayed_inodes(root); |
3741 | btrfs_assert_delayed_root_empty(root); | |
49b25e05 JM |
3742 | |
3743 | btrfs_destroy_pending_snapshots(cur_trans); | |
3744 | ||
3745 | btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, | |
3746 | EXTENT_DIRTY); | |
6e841e32 LB |
3747 | btrfs_destroy_pinned_extent(root, |
3748 | root->fs_info->pinned_extents); | |
49b25e05 JM |
3749 | |
3750 | /* | |
3751 | memset(cur_trans, 0, sizeof(*cur_trans)); | |
3752 | kmem_cache_free(btrfs_transaction_cachep, cur_trans); | |
3753 | */ | |
3754 | } | |
3755 | ||
3756 | int btrfs_cleanup_transaction(struct btrfs_root *root) | |
acce952b | 3757 | { |
3758 | struct btrfs_transaction *t; | |
3759 | LIST_HEAD(list); | |
3760 | ||
acce952b | 3761 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
3762 | ||
a4abeea4 | 3763 | spin_lock(&root->fs_info->trans_lock); |
acce952b | 3764 | list_splice_init(&root->fs_info->trans_list, &list); |
a4abeea4 JB |
3765 | root->fs_info->trans_no_join = 1; |
3766 | spin_unlock(&root->fs_info->trans_lock); | |
3767 | ||
acce952b | 3768 | while (!list_empty(&list)) { |
3769 | t = list_entry(list.next, struct btrfs_transaction, list); | |
3770 | if (!t) | |
3771 | break; | |
3772 | ||
3773 | btrfs_destroy_ordered_operations(root); | |
3774 | ||
3775 | btrfs_destroy_ordered_extents(root); | |
3776 | ||
3777 | btrfs_destroy_delayed_refs(t, root); | |
3778 | ||
3779 | btrfs_block_rsv_release(root, | |
3780 | &root->fs_info->trans_block_rsv, | |
3781 | t->dirty_pages.dirty_bytes); | |
3782 | ||
3783 | /* FIXME: cleanup wait for commit */ | |
3784 | t->in_commit = 1; | |
3785 | t->blocked = 1; | |
3786 | if (waitqueue_active(&root->fs_info->transaction_blocked_wait)) | |
3787 | wake_up(&root->fs_info->transaction_blocked_wait); | |
3788 | ||
3789 | t->blocked = 0; | |
3790 | if (waitqueue_active(&root->fs_info->transaction_wait)) | |
3791 | wake_up(&root->fs_info->transaction_wait); | |
acce952b | 3792 | |
acce952b | 3793 | t->commit_done = 1; |
3794 | if (waitqueue_active(&t->commit_wait)) | |
3795 | wake_up(&t->commit_wait); | |
acce952b | 3796 | |
67cde344 MX |
3797 | btrfs_destroy_delayed_inodes(root); |
3798 | btrfs_assert_delayed_root_empty(root); | |
3799 | ||
acce952b | 3800 | btrfs_destroy_pending_snapshots(t); |
3801 | ||
3802 | btrfs_destroy_delalloc_inodes(root); | |
3803 | ||
a4abeea4 | 3804 | spin_lock(&root->fs_info->trans_lock); |
acce952b | 3805 | root->fs_info->running_transaction = NULL; |
a4abeea4 | 3806 | spin_unlock(&root->fs_info->trans_lock); |
acce952b | 3807 | |
3808 | btrfs_destroy_marked_extents(root, &t->dirty_pages, | |
3809 | EXTENT_DIRTY); | |
3810 | ||
3811 | btrfs_destroy_pinned_extent(root, | |
3812 | root->fs_info->pinned_extents); | |
3813 | ||
13c5a93e | 3814 | atomic_set(&t->use_count, 0); |
acce952b | 3815 | list_del_init(&t->list); |
3816 | memset(t, 0, sizeof(*t)); | |
3817 | kmem_cache_free(btrfs_transaction_cachep, t); | |
3818 | } | |
3819 | ||
a4abeea4 JB |
3820 | spin_lock(&root->fs_info->trans_lock); |
3821 | root->fs_info->trans_no_join = 0; | |
3822 | spin_unlock(&root->fs_info->trans_lock); | |
acce952b | 3823 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); |
acce952b | 3824 | |
3825 | return 0; | |
3826 | } | |
3827 | ||
d1310b2e | 3828 | static struct extent_io_ops btree_extent_io_ops = { |
4bef0848 | 3829 | .write_cache_pages_lock_hook = btree_lock_page_hook, |
ce9adaa5 | 3830 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
4bb31e92 | 3831 | .readpage_io_failed_hook = btree_io_failed_hook, |
0b86a832 | 3832 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
3833 | /* note we're sharing with inode.c for the merge bio hook */ |
3834 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 3835 | }; |