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 | |
de0022b9 JB |
49 | #ifdef CONFIG_X86 |
50 | #include <asm/cpufeature.h> | |
51 | #endif | |
52 | ||
d1310b2e | 53 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 54 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 55 | static void free_fs_root(struct btrfs_root *root); |
fcd1f065 | 56 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 57 | int read_only); |
143bede5 JM |
58 | static void btrfs_destroy_ordered_operations(struct btrfs_root *root); |
59 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); | |
acce952b | 60 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
61 | struct btrfs_root *root); | |
143bede5 JM |
62 | static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t); |
63 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); | |
acce952b | 64 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, |
65 | struct extent_io_tree *dirty_pages, | |
66 | int mark); | |
67 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
68 | struct extent_io_tree *pinned_extents); | |
ce9adaa5 | 69 | |
d352ac68 CM |
70 | /* |
71 | * end_io_wq structs are used to do processing in task context when an IO is | |
72 | * complete. This is used during reads to verify checksums, and it is used | |
73 | * by writes to insert metadata for new file extents after IO is complete. | |
74 | */ | |
ce9adaa5 CM |
75 | struct end_io_wq { |
76 | struct bio *bio; | |
77 | bio_end_io_t *end_io; | |
78 | void *private; | |
79 | struct btrfs_fs_info *info; | |
80 | int error; | |
22c59948 | 81 | int metadata; |
ce9adaa5 | 82 | struct list_head list; |
8b712842 | 83 | struct btrfs_work work; |
ce9adaa5 | 84 | }; |
0da5468f | 85 | |
d352ac68 CM |
86 | /* |
87 | * async submit bios are used to offload expensive checksumming | |
88 | * onto the worker threads. They checksum file and metadata bios | |
89 | * just before they are sent down the IO stack. | |
90 | */ | |
44b8bd7e CM |
91 | struct async_submit_bio { |
92 | struct inode *inode; | |
93 | struct bio *bio; | |
94 | struct list_head list; | |
4a69a410 CM |
95 | extent_submit_bio_hook_t *submit_bio_start; |
96 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
97 | int rw; |
98 | int mirror_num; | |
c8b97818 | 99 | unsigned long bio_flags; |
eaf25d93 CM |
100 | /* |
101 | * bio_offset is optional, can be used if the pages in the bio | |
102 | * can't tell us where in the file the bio should go | |
103 | */ | |
104 | u64 bio_offset; | |
8b712842 | 105 | struct btrfs_work work; |
79787eaa | 106 | int error; |
44b8bd7e CM |
107 | }; |
108 | ||
85d4e461 CM |
109 | /* |
110 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
111 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
112 | * the level the eb occupies in the tree. | |
113 | * | |
114 | * Different roots are used for different purposes and may nest inside each | |
115 | * other and they require separate keysets. As lockdep keys should be | |
116 | * static, assign keysets according to the purpose of the root as indicated | |
117 | * by btrfs_root->objectid. This ensures that all special purpose roots | |
118 | * have separate keysets. | |
4008c04a | 119 | * |
85d4e461 CM |
120 | * Lock-nesting across peer nodes is always done with the immediate parent |
121 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
122 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 123 | * |
85d4e461 CM |
124 | * The key is set by the readpage_end_io_hook after the buffer has passed |
125 | * csum validation but before the pages are unlocked. It is also set by | |
126 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 127 | * |
85d4e461 CM |
128 | * We also add a check to make sure the highest level of the tree is the |
129 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
130 | * needs update as well. | |
4008c04a CM |
131 | */ |
132 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
133 | # if BTRFS_MAX_LEVEL != 8 | |
134 | # error | |
135 | # endif | |
85d4e461 CM |
136 | |
137 | static struct btrfs_lockdep_keyset { | |
138 | u64 id; /* root objectid */ | |
139 | const char *name_stem; /* lock name stem */ | |
140 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
141 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
142 | } btrfs_lockdep_keysets[] = { | |
143 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
144 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
145 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
146 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
147 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
148 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
149 | { .id = BTRFS_ORPHAN_OBJECTID, .name_stem = "orphan" }, | |
150 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, | |
151 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
152 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
153 | { .id = 0, .name_stem = "tree" }, | |
4008c04a | 154 | }; |
85d4e461 CM |
155 | |
156 | void __init btrfs_init_lockdep(void) | |
157 | { | |
158 | int i, j; | |
159 | ||
160 | /* initialize lockdep class names */ | |
161 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
162 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
163 | ||
164 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
165 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
166 | "btrfs-%s-%02d", ks->name_stem, j); | |
167 | } | |
168 | } | |
169 | ||
170 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
171 | int level) | |
172 | { | |
173 | struct btrfs_lockdep_keyset *ks; | |
174 | ||
175 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
176 | ||
177 | /* find the matching keyset, id 0 is the default entry */ | |
178 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
179 | if (ks->id == objectid) | |
180 | break; | |
181 | ||
182 | lockdep_set_class_and_name(&eb->lock, | |
183 | &ks->keys[level], ks->names[level]); | |
184 | } | |
185 | ||
4008c04a CM |
186 | #endif |
187 | ||
d352ac68 CM |
188 | /* |
189 | * extents on the btree inode are pretty simple, there's one extent | |
190 | * that covers the entire device | |
191 | */ | |
b2950863 | 192 | static struct extent_map *btree_get_extent(struct inode *inode, |
306e16ce | 193 | struct page *page, size_t pg_offset, u64 start, u64 len, |
b2950863 | 194 | int create) |
7eccb903 | 195 | { |
5f39d397 CM |
196 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
197 | struct extent_map *em; | |
198 | int ret; | |
199 | ||
890871be | 200 | read_lock(&em_tree->lock); |
d1310b2e | 201 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d CM |
202 | if (em) { |
203 | em->bdev = | |
204 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
890871be | 205 | read_unlock(&em_tree->lock); |
5f39d397 | 206 | goto out; |
a061fc8d | 207 | } |
890871be | 208 | read_unlock(&em_tree->lock); |
7b13b7b1 | 209 | |
172ddd60 | 210 | em = alloc_extent_map(); |
5f39d397 CM |
211 | if (!em) { |
212 | em = ERR_PTR(-ENOMEM); | |
213 | goto out; | |
214 | } | |
215 | em->start = 0; | |
0afbaf8c | 216 | em->len = (u64)-1; |
c8b97818 | 217 | em->block_len = (u64)-1; |
5f39d397 | 218 | em->block_start = 0; |
a061fc8d | 219 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e | 220 | |
890871be | 221 | write_lock(&em_tree->lock); |
5f39d397 CM |
222 | ret = add_extent_mapping(em_tree, em); |
223 | if (ret == -EEXIST) { | |
224 | free_extent_map(em); | |
7b13b7b1 | 225 | em = lookup_extent_mapping(em_tree, start, len); |
b4f359ab | 226 | if (!em) |
0433f20d | 227 | em = ERR_PTR(-EIO); |
5f39d397 | 228 | } else if (ret) { |
7b13b7b1 | 229 | free_extent_map(em); |
0433f20d | 230 | em = ERR_PTR(ret); |
5f39d397 | 231 | } |
890871be | 232 | write_unlock(&em_tree->lock); |
7b13b7b1 | 233 | |
5f39d397 CM |
234 | out: |
235 | return em; | |
7eccb903 CM |
236 | } |
237 | ||
19c00ddc CM |
238 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
239 | { | |
163e783e | 240 | return crc32c(seed, data, len); |
19c00ddc CM |
241 | } |
242 | ||
243 | void btrfs_csum_final(u32 crc, char *result) | |
244 | { | |
7e75bf3f | 245 | put_unaligned_le32(~crc, result); |
19c00ddc CM |
246 | } |
247 | ||
d352ac68 CM |
248 | /* |
249 | * compute the csum for a btree block, and either verify it or write it | |
250 | * into the csum field of the block. | |
251 | */ | |
19c00ddc CM |
252 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
253 | int verify) | |
254 | { | |
6c41761f | 255 | u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy); |
607d432d | 256 | char *result = NULL; |
19c00ddc CM |
257 | unsigned long len; |
258 | unsigned long cur_len; | |
259 | unsigned long offset = BTRFS_CSUM_SIZE; | |
19c00ddc CM |
260 | char *kaddr; |
261 | unsigned long map_start; | |
262 | unsigned long map_len; | |
263 | int err; | |
264 | u32 crc = ~(u32)0; | |
607d432d | 265 | unsigned long inline_result; |
19c00ddc CM |
266 | |
267 | len = buf->len - offset; | |
d397712b | 268 | while (len > 0) { |
19c00ddc | 269 | err = map_private_extent_buffer(buf, offset, 32, |
a6591715 | 270 | &kaddr, &map_start, &map_len); |
d397712b | 271 | if (err) |
19c00ddc | 272 | return 1; |
19c00ddc CM |
273 | cur_len = min(len, map_len - (offset - map_start)); |
274 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
275 | crc, cur_len); | |
276 | len -= cur_len; | |
277 | offset += cur_len; | |
19c00ddc | 278 | } |
607d432d JB |
279 | if (csum_size > sizeof(inline_result)) { |
280 | result = kzalloc(csum_size * sizeof(char), GFP_NOFS); | |
281 | if (!result) | |
282 | return 1; | |
283 | } else { | |
284 | result = (char *)&inline_result; | |
285 | } | |
286 | ||
19c00ddc CM |
287 | btrfs_csum_final(crc, result); |
288 | ||
289 | if (verify) { | |
607d432d | 290 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
291 | u32 val; |
292 | u32 found = 0; | |
607d432d | 293 | memcpy(&found, result, csum_size); |
e4204ded | 294 | |
607d432d | 295 | read_extent_buffer(buf, &val, 0, csum_size); |
7a36ddec | 296 | printk_ratelimited(KERN_INFO "btrfs: %s checksum verify " |
193f284d CM |
297 | "failed on %llu wanted %X found %X " |
298 | "level %d\n", | |
299 | root->fs_info->sb->s_id, | |
300 | (unsigned long long)buf->start, val, found, | |
301 | btrfs_header_level(buf)); | |
607d432d JB |
302 | if (result != (char *)&inline_result) |
303 | kfree(result); | |
19c00ddc CM |
304 | return 1; |
305 | } | |
306 | } else { | |
607d432d | 307 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 308 | } |
607d432d JB |
309 | if (result != (char *)&inline_result) |
310 | kfree(result); | |
19c00ddc CM |
311 | return 0; |
312 | } | |
313 | ||
d352ac68 CM |
314 | /* |
315 | * we can't consider a given block up to date unless the transid of the | |
316 | * block matches the transid in the parent node's pointer. This is how we | |
317 | * detect blocks that either didn't get written at all or got written | |
318 | * in the wrong place. | |
319 | */ | |
1259ab75 | 320 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
321 | struct extent_buffer *eb, u64 parent_transid, |
322 | int atomic) | |
1259ab75 | 323 | { |
2ac55d41 | 324 | struct extent_state *cached_state = NULL; |
1259ab75 CM |
325 | int ret; |
326 | ||
327 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
328 | return 0; | |
329 | ||
b9fab919 CM |
330 | if (atomic) |
331 | return -EAGAIN; | |
332 | ||
2ac55d41 | 333 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
d0082371 | 334 | 0, &cached_state); |
0b32f4bb | 335 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
336 | btrfs_header_generation(eb) == parent_transid) { |
337 | ret = 0; | |
338 | goto out; | |
339 | } | |
7a36ddec | 340 | printk_ratelimited("parent transid verify failed on %llu wanted %llu " |
193f284d CM |
341 | "found %llu\n", |
342 | (unsigned long long)eb->start, | |
343 | (unsigned long long)parent_transid, | |
344 | (unsigned long long)btrfs_header_generation(eb)); | |
1259ab75 | 345 | ret = 1; |
0b32f4bb | 346 | clear_extent_buffer_uptodate(eb); |
33958dc6 | 347 | out: |
2ac55d41 JB |
348 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
349 | &cached_state, GFP_NOFS); | |
1259ab75 | 350 | return ret; |
1259ab75 CM |
351 | } |
352 | ||
d352ac68 CM |
353 | /* |
354 | * helper to read a given tree block, doing retries as required when | |
355 | * the checksums don't match and we have alternate mirrors to try. | |
356 | */ | |
f188591e CM |
357 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
358 | struct extent_buffer *eb, | |
ca7a79ad | 359 | u64 start, u64 parent_transid) |
f188591e CM |
360 | { |
361 | struct extent_io_tree *io_tree; | |
ea466794 | 362 | int failed = 0; |
f188591e CM |
363 | int ret; |
364 | int num_copies = 0; | |
365 | int mirror_num = 0; | |
ea466794 | 366 | int failed_mirror = 0; |
f188591e | 367 | |
a826d6dc | 368 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
f188591e CM |
369 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
370 | while (1) { | |
bb82ab88 AJ |
371 | ret = read_extent_buffer_pages(io_tree, eb, start, |
372 | WAIT_COMPLETE, | |
f188591e | 373 | btree_get_extent, mirror_num); |
256dd1bb SB |
374 | if (!ret) { |
375 | if (!verify_parent_transid(io_tree, eb, | |
b9fab919 | 376 | parent_transid, 0)) |
256dd1bb SB |
377 | break; |
378 | else | |
379 | ret = -EIO; | |
380 | } | |
d397712b | 381 | |
a826d6dc JB |
382 | /* |
383 | * This buffer's crc is fine, but its contents are corrupted, so | |
384 | * there is no reason to read the other copies, they won't be | |
385 | * any less wrong. | |
386 | */ | |
387 | if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) | |
ea466794 JB |
388 | break; |
389 | ||
f188591e CM |
390 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
391 | eb->start, eb->len); | |
4235298e | 392 | if (num_copies == 1) |
ea466794 | 393 | break; |
4235298e | 394 | |
5cf1ab56 JB |
395 | if (!failed_mirror) { |
396 | failed = 1; | |
397 | failed_mirror = eb->read_mirror; | |
398 | } | |
399 | ||
f188591e | 400 | mirror_num++; |
ea466794 JB |
401 | if (mirror_num == failed_mirror) |
402 | mirror_num++; | |
403 | ||
4235298e | 404 | if (mirror_num > num_copies) |
ea466794 | 405 | break; |
f188591e | 406 | } |
ea466794 | 407 | |
c0901581 | 408 | if (failed && !ret && failed_mirror) |
ea466794 JB |
409 | repair_eb_io_failure(root, eb, failed_mirror); |
410 | ||
411 | return ret; | |
f188591e | 412 | } |
19c00ddc | 413 | |
d352ac68 | 414 | /* |
d397712b CM |
415 | * checksum a dirty tree block before IO. This has extra checks to make sure |
416 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 417 | */ |
d397712b | 418 | |
b2950863 | 419 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 420 | { |
d1310b2e | 421 | struct extent_io_tree *tree; |
35ebb934 | 422 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc | 423 | u64 found_start; |
19c00ddc | 424 | struct extent_buffer *eb; |
f188591e | 425 | |
d1310b2e | 426 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc | 427 | |
4f2de97a JB |
428 | eb = (struct extent_buffer *)page->private; |
429 | if (page != eb->pages[0]) | |
430 | return 0; | |
19c00ddc CM |
431 | found_start = btrfs_header_bytenr(eb); |
432 | if (found_start != start) { | |
55c69072 | 433 | WARN_ON(1); |
4f2de97a | 434 | return 0; |
55c69072 | 435 | } |
727011e0 | 436 | if (eb->pages[0] != page) { |
55c69072 | 437 | WARN_ON(1); |
4f2de97a | 438 | return 0; |
55c69072 CM |
439 | } |
440 | if (!PageUptodate(page)) { | |
55c69072 | 441 | WARN_ON(1); |
4f2de97a | 442 | return 0; |
19c00ddc | 443 | } |
19c00ddc | 444 | csum_tree_block(root, eb, 0); |
19c00ddc CM |
445 | return 0; |
446 | } | |
447 | ||
2b82032c YZ |
448 | static int check_tree_block_fsid(struct btrfs_root *root, |
449 | struct extent_buffer *eb) | |
450 | { | |
451 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
452 | u8 fsid[BTRFS_UUID_SIZE]; | |
453 | int ret = 1; | |
454 | ||
455 | read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), | |
456 | BTRFS_FSID_SIZE); | |
457 | while (fs_devices) { | |
458 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
459 | ret = 0; | |
460 | break; | |
461 | } | |
462 | fs_devices = fs_devices->seed; | |
463 | } | |
464 | return ret; | |
465 | } | |
466 | ||
a826d6dc JB |
467 | #define CORRUPT(reason, eb, root, slot) \ |
468 | printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \ | |
469 | "root=%llu, slot=%d\n", reason, \ | |
470 | (unsigned long long)btrfs_header_bytenr(eb), \ | |
471 | (unsigned long long)root->objectid, slot) | |
472 | ||
473 | static noinline int check_leaf(struct btrfs_root *root, | |
474 | struct extent_buffer *leaf) | |
475 | { | |
476 | struct btrfs_key key; | |
477 | struct btrfs_key leaf_key; | |
478 | u32 nritems = btrfs_header_nritems(leaf); | |
479 | int slot; | |
480 | ||
481 | if (nritems == 0) | |
482 | return 0; | |
483 | ||
484 | /* Check the 0 item */ | |
485 | if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != | |
486 | BTRFS_LEAF_DATA_SIZE(root)) { | |
487 | CORRUPT("invalid item offset size pair", leaf, root, 0); | |
488 | return -EIO; | |
489 | } | |
490 | ||
491 | /* | |
492 | * Check to make sure each items keys are in the correct order and their | |
493 | * offsets make sense. We only have to loop through nritems-1 because | |
494 | * we check the current slot against the next slot, which verifies the | |
495 | * next slot's offset+size makes sense and that the current's slot | |
496 | * offset is correct. | |
497 | */ | |
498 | for (slot = 0; slot < nritems - 1; slot++) { | |
499 | btrfs_item_key_to_cpu(leaf, &leaf_key, slot); | |
500 | btrfs_item_key_to_cpu(leaf, &key, slot + 1); | |
501 | ||
502 | /* Make sure the keys are in the right order */ | |
503 | if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { | |
504 | CORRUPT("bad key order", leaf, root, slot); | |
505 | return -EIO; | |
506 | } | |
507 | ||
508 | /* | |
509 | * Make sure the offset and ends are right, remember that the | |
510 | * item data starts at the end of the leaf and grows towards the | |
511 | * front. | |
512 | */ | |
513 | if (btrfs_item_offset_nr(leaf, slot) != | |
514 | btrfs_item_end_nr(leaf, slot + 1)) { | |
515 | CORRUPT("slot offset bad", leaf, root, slot); | |
516 | return -EIO; | |
517 | } | |
518 | ||
519 | /* | |
520 | * Check to make sure that we don't point outside of the leaf, | |
521 | * just incase all the items are consistent to eachother, but | |
522 | * all point outside of the leaf. | |
523 | */ | |
524 | if (btrfs_item_end_nr(leaf, slot) > | |
525 | BTRFS_LEAF_DATA_SIZE(root)) { | |
526 | CORRUPT("slot end outside of leaf", leaf, root, slot); | |
527 | return -EIO; | |
528 | } | |
529 | } | |
530 | ||
531 | return 0; | |
532 | } | |
533 | ||
727011e0 CM |
534 | struct extent_buffer *find_eb_for_page(struct extent_io_tree *tree, |
535 | struct page *page, int max_walk) | |
536 | { | |
537 | struct extent_buffer *eb; | |
538 | u64 start = page_offset(page); | |
539 | u64 target = start; | |
540 | u64 min_start; | |
541 | ||
542 | if (start < max_walk) | |
543 | min_start = 0; | |
544 | else | |
545 | min_start = start - max_walk; | |
546 | ||
547 | while (start >= min_start) { | |
548 | eb = find_extent_buffer(tree, start, 0); | |
549 | if (eb) { | |
550 | /* | |
551 | * we found an extent buffer and it contains our page | |
552 | * horray! | |
553 | */ | |
554 | if (eb->start <= target && | |
555 | eb->start + eb->len > target) | |
556 | return eb; | |
557 | ||
558 | /* we found an extent buffer that wasn't for us */ | |
559 | free_extent_buffer(eb); | |
560 | return NULL; | |
561 | } | |
562 | if (start == 0) | |
563 | break; | |
564 | start -= PAGE_CACHE_SIZE; | |
565 | } | |
566 | return NULL; | |
567 | } | |
568 | ||
b2950863 | 569 | static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
5cf1ab56 | 570 | struct extent_state *state, int mirror) |
ce9adaa5 CM |
571 | { |
572 | struct extent_io_tree *tree; | |
573 | u64 found_start; | |
574 | int found_level; | |
ce9adaa5 CM |
575 | struct extent_buffer *eb; |
576 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 577 | int ret = 0; |
727011e0 | 578 | int reads_done; |
ce9adaa5 | 579 | |
ce9adaa5 CM |
580 | if (!page->private) |
581 | goto out; | |
d397712b | 582 | |
727011e0 | 583 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
4f2de97a | 584 | eb = (struct extent_buffer *)page->private; |
d397712b | 585 | |
0b32f4bb JB |
586 | /* the pending IO might have been the only thing that kept this buffer |
587 | * in memory. Make sure we have a ref for all this other checks | |
588 | */ | |
589 | extent_buffer_get(eb); | |
590 | ||
591 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
727011e0 CM |
592 | if (!reads_done) |
593 | goto err; | |
f188591e | 594 | |
5cf1ab56 | 595 | eb->read_mirror = mirror; |
ea466794 JB |
596 | if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) { |
597 | ret = -EIO; | |
598 | goto err; | |
599 | } | |
600 | ||
ce9adaa5 | 601 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 602 | if (found_start != eb->start) { |
7a36ddec | 603 | printk_ratelimited(KERN_INFO "btrfs bad tree block start " |
193f284d CM |
604 | "%llu %llu\n", |
605 | (unsigned long long)found_start, | |
606 | (unsigned long long)eb->start); | |
f188591e | 607 | ret = -EIO; |
ce9adaa5 CM |
608 | goto err; |
609 | } | |
2b82032c | 610 | if (check_tree_block_fsid(root, eb)) { |
7a36ddec | 611 | printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n", |
193f284d | 612 | (unsigned long long)eb->start); |
1259ab75 CM |
613 | ret = -EIO; |
614 | goto err; | |
615 | } | |
ce9adaa5 CM |
616 | found_level = btrfs_header_level(eb); |
617 | ||
85d4e461 CM |
618 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
619 | eb, found_level); | |
4008c04a | 620 | |
ce9adaa5 | 621 | ret = csum_tree_block(root, eb, 1); |
a826d6dc | 622 | if (ret) { |
f188591e | 623 | ret = -EIO; |
a826d6dc JB |
624 | goto err; |
625 | } | |
626 | ||
627 | /* | |
628 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
629 | * that we don't try and read the other copies of this block, just | |
630 | * return -EIO. | |
631 | */ | |
632 | if (found_level == 0 && check_leaf(root, eb)) { | |
633 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); | |
634 | ret = -EIO; | |
635 | } | |
ce9adaa5 | 636 | |
0b32f4bb JB |
637 | if (!ret) |
638 | set_extent_buffer_uptodate(eb); | |
ce9adaa5 | 639 | err: |
4bb31e92 AJ |
640 | if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) { |
641 | clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags); | |
642 | btree_readahead_hook(root, eb, eb->start, ret); | |
643 | } | |
644 | ||
0b32f4bb JB |
645 | if (ret) |
646 | clear_extent_buffer_uptodate(eb); | |
647 | free_extent_buffer(eb); | |
ce9adaa5 | 648 | out: |
f188591e | 649 | return ret; |
ce9adaa5 CM |
650 | } |
651 | ||
ea466794 | 652 | static int btree_io_failed_hook(struct page *page, int failed_mirror) |
4bb31e92 | 653 | { |
4bb31e92 AJ |
654 | struct extent_buffer *eb; |
655 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
656 | ||
4f2de97a | 657 | eb = (struct extent_buffer *)page->private; |
ea466794 | 658 | set_bit(EXTENT_BUFFER_IOERR, &eb->bflags); |
5cf1ab56 | 659 | eb->read_mirror = failed_mirror; |
ea466794 | 660 | if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) |
4bb31e92 | 661 | btree_readahead_hook(root, eb, eb->start, -EIO); |
4bb31e92 AJ |
662 | return -EIO; /* we fixed nothing */ |
663 | } | |
664 | ||
ce9adaa5 | 665 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
666 | { |
667 | struct end_io_wq *end_io_wq = bio->bi_private; | |
668 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 669 | |
ce9adaa5 | 670 | fs_info = end_io_wq->info; |
ce9adaa5 | 671 | end_io_wq->error = err; |
8b712842 CM |
672 | end_io_wq->work.func = end_workqueue_fn; |
673 | end_io_wq->work.flags = 0; | |
d20f7043 | 674 | |
7b6d91da | 675 | if (bio->bi_rw & REQ_WRITE) { |
0cb59c99 | 676 | if (end_io_wq->metadata == 1) |
cad321ad CM |
677 | btrfs_queue_worker(&fs_info->endio_meta_write_workers, |
678 | &end_io_wq->work); | |
0cb59c99 JB |
679 | else if (end_io_wq->metadata == 2) |
680 | btrfs_queue_worker(&fs_info->endio_freespace_worker, | |
681 | &end_io_wq->work); | |
cad321ad CM |
682 | else |
683 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
684 | &end_io_wq->work); | |
d20f7043 CM |
685 | } else { |
686 | if (end_io_wq->metadata) | |
687 | btrfs_queue_worker(&fs_info->endio_meta_workers, | |
688 | &end_io_wq->work); | |
689 | else | |
690 | btrfs_queue_worker(&fs_info->endio_workers, | |
691 | &end_io_wq->work); | |
692 | } | |
ce9adaa5 CM |
693 | } |
694 | ||
0cb59c99 JB |
695 | /* |
696 | * For the metadata arg you want | |
697 | * | |
698 | * 0 - if data | |
699 | * 1 - if normal metadta | |
700 | * 2 - if writing to the free space cache area | |
701 | */ | |
22c59948 CM |
702 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
703 | int metadata) | |
0b86a832 | 704 | { |
ce9adaa5 | 705 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
706 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
707 | if (!end_io_wq) | |
708 | return -ENOMEM; | |
709 | ||
710 | end_io_wq->private = bio->bi_private; | |
711 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 712 | end_io_wq->info = info; |
ce9adaa5 CM |
713 | end_io_wq->error = 0; |
714 | end_io_wq->bio = bio; | |
22c59948 | 715 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
716 | |
717 | bio->bi_private = end_io_wq; | |
718 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
719 | return 0; |
720 | } | |
721 | ||
b64a2851 | 722 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 723 | { |
4854ddd0 CM |
724 | unsigned long limit = min_t(unsigned long, |
725 | info->workers.max_workers, | |
726 | info->fs_devices->open_devices); | |
727 | return 256 * limit; | |
728 | } | |
0986fe9e | 729 | |
4a69a410 CM |
730 | static void run_one_async_start(struct btrfs_work *work) |
731 | { | |
4a69a410 | 732 | struct async_submit_bio *async; |
79787eaa | 733 | int ret; |
4a69a410 CM |
734 | |
735 | async = container_of(work, struct async_submit_bio, work); | |
79787eaa JM |
736 | ret = async->submit_bio_start(async->inode, async->rw, async->bio, |
737 | async->mirror_num, async->bio_flags, | |
738 | async->bio_offset); | |
739 | if (ret) | |
740 | async->error = ret; | |
4a69a410 CM |
741 | } |
742 | ||
743 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
744 | { |
745 | struct btrfs_fs_info *fs_info; | |
746 | struct async_submit_bio *async; | |
4854ddd0 | 747 | int limit; |
8b712842 CM |
748 | |
749 | async = container_of(work, struct async_submit_bio, work); | |
750 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 751 | |
b64a2851 | 752 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
753 | limit = limit * 2 / 3; |
754 | ||
66657b31 | 755 | if (atomic_dec_return(&fs_info->nr_async_submits) < limit && |
b64a2851 | 756 | waitqueue_active(&fs_info->async_submit_wait)) |
4854ddd0 CM |
757 | wake_up(&fs_info->async_submit_wait); |
758 | ||
79787eaa JM |
759 | /* If an error occured we just want to clean up the bio and move on */ |
760 | if (async->error) { | |
761 | bio_endio(async->bio, async->error); | |
762 | return; | |
763 | } | |
764 | ||
4a69a410 | 765 | async->submit_bio_done(async->inode, async->rw, async->bio, |
eaf25d93 CM |
766 | async->mirror_num, async->bio_flags, |
767 | async->bio_offset); | |
4a69a410 CM |
768 | } |
769 | ||
770 | static void run_one_async_free(struct btrfs_work *work) | |
771 | { | |
772 | struct async_submit_bio *async; | |
773 | ||
774 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
775 | kfree(async); |
776 | } | |
777 | ||
44b8bd7e CM |
778 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
779 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 780 | unsigned long bio_flags, |
eaf25d93 | 781 | u64 bio_offset, |
4a69a410 CM |
782 | extent_submit_bio_hook_t *submit_bio_start, |
783 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
784 | { |
785 | struct async_submit_bio *async; | |
786 | ||
787 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
788 | if (!async) | |
789 | return -ENOMEM; | |
790 | ||
791 | async->inode = inode; | |
792 | async->rw = rw; | |
793 | async->bio = bio; | |
794 | async->mirror_num = mirror_num; | |
4a69a410 CM |
795 | async->submit_bio_start = submit_bio_start; |
796 | async->submit_bio_done = submit_bio_done; | |
797 | ||
798 | async->work.func = run_one_async_start; | |
799 | async->work.ordered_func = run_one_async_done; | |
800 | async->work.ordered_free = run_one_async_free; | |
801 | ||
8b712842 | 802 | async->work.flags = 0; |
c8b97818 | 803 | async->bio_flags = bio_flags; |
eaf25d93 | 804 | async->bio_offset = bio_offset; |
8c8bee1d | 805 | |
79787eaa JM |
806 | async->error = 0; |
807 | ||
cb03c743 | 808 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 | 809 | |
7b6d91da | 810 | if (rw & REQ_SYNC) |
d313d7a3 CM |
811 | btrfs_set_work_high_prio(&async->work); |
812 | ||
8b712842 | 813 | btrfs_queue_worker(&fs_info->workers, &async->work); |
9473f16c | 814 | |
d397712b | 815 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
816 | atomic_read(&fs_info->nr_async_submits)) { |
817 | wait_event(fs_info->async_submit_wait, | |
818 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
819 | } | |
820 | ||
44b8bd7e CM |
821 | return 0; |
822 | } | |
823 | ||
ce3ed71a CM |
824 | static int btree_csum_one_bio(struct bio *bio) |
825 | { | |
826 | struct bio_vec *bvec = bio->bi_io_vec; | |
827 | int bio_index = 0; | |
828 | struct btrfs_root *root; | |
79787eaa | 829 | int ret = 0; |
ce3ed71a CM |
830 | |
831 | WARN_ON(bio->bi_vcnt <= 0); | |
d397712b | 832 | while (bio_index < bio->bi_vcnt) { |
ce3ed71a | 833 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
79787eaa JM |
834 | ret = csum_dirty_buffer(root, bvec->bv_page); |
835 | if (ret) | |
836 | break; | |
ce3ed71a CM |
837 | bio_index++; |
838 | bvec++; | |
839 | } | |
79787eaa | 840 | return ret; |
ce3ed71a CM |
841 | } |
842 | ||
4a69a410 CM |
843 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
844 | struct bio *bio, int mirror_num, | |
eaf25d93 CM |
845 | unsigned long bio_flags, |
846 | u64 bio_offset) | |
22c59948 | 847 | { |
8b712842 CM |
848 | /* |
849 | * when we're called for a write, we're already in the async | |
5443be45 | 850 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 851 | */ |
79787eaa | 852 | return btree_csum_one_bio(bio); |
4a69a410 | 853 | } |
22c59948 | 854 | |
4a69a410 | 855 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
856 | int mirror_num, unsigned long bio_flags, |
857 | u64 bio_offset) | |
4a69a410 | 858 | { |
8b712842 | 859 | /* |
4a69a410 CM |
860 | * when we're called for a write, we're already in the async |
861 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 862 | */ |
8b712842 | 863 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
864 | } |
865 | ||
de0022b9 JB |
866 | static int check_async_write(struct inode *inode, unsigned long bio_flags) |
867 | { | |
868 | if (bio_flags & EXTENT_BIO_TREE_LOG) | |
869 | return 0; | |
870 | #ifdef CONFIG_X86 | |
871 | if (cpu_has_xmm4_2) | |
872 | return 0; | |
873 | #endif | |
874 | return 1; | |
875 | } | |
876 | ||
44b8bd7e | 877 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
878 | int mirror_num, unsigned long bio_flags, |
879 | u64 bio_offset) | |
44b8bd7e | 880 | { |
de0022b9 | 881 | int async = check_async_write(inode, bio_flags); |
cad321ad CM |
882 | int ret; |
883 | ||
7b6d91da | 884 | if (!(rw & REQ_WRITE)) { |
f3f266ab | 885 | |
4a69a410 CM |
886 | /* |
887 | * called for a read, do the setup so that checksum validation | |
888 | * can happen in the async kernel threads | |
889 | */ | |
f3f266ab CM |
890 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, |
891 | bio, 1); | |
1d4284bd CM |
892 | if (ret) |
893 | return ret; | |
4a69a410 | 894 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, |
6f3577bd | 895 | mirror_num, 0); |
de0022b9 JB |
896 | } else if (!async) { |
897 | ret = btree_csum_one_bio(bio); | |
898 | if (ret) | |
899 | return ret; | |
900 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
901 | mirror_num, 0); | |
44b8bd7e | 902 | } |
d313d7a3 | 903 | |
cad321ad CM |
904 | /* |
905 | * kthread helpers are used to submit writes so that checksumming | |
906 | * can happen in parallel across all CPUs | |
907 | */ | |
44b8bd7e | 908 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, |
c8b97818 | 909 | inode, rw, bio, mirror_num, 0, |
eaf25d93 | 910 | bio_offset, |
4a69a410 CM |
911 | __btree_submit_bio_start, |
912 | __btree_submit_bio_done); | |
44b8bd7e CM |
913 | } |
914 | ||
3dd1462e | 915 | #ifdef CONFIG_MIGRATION |
784b4e29 | 916 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
917 | struct page *newpage, struct page *page, |
918 | enum migrate_mode mode) | |
784b4e29 CM |
919 | { |
920 | /* | |
921 | * we can't safely write a btree page from here, | |
922 | * we haven't done the locking hook | |
923 | */ | |
924 | if (PageDirty(page)) | |
925 | return -EAGAIN; | |
926 | /* | |
927 | * Buffers may be managed in a filesystem specific way. | |
928 | * We must have no buffers or drop them. | |
929 | */ | |
930 | if (page_has_private(page) && | |
931 | !try_to_release_page(page, GFP_KERNEL)) | |
932 | return -EAGAIN; | |
a6bc32b8 | 933 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 934 | } |
3dd1462e | 935 | #endif |
784b4e29 | 936 | |
0da5468f CM |
937 | |
938 | static int btree_writepages(struct address_space *mapping, | |
939 | struct writeback_control *wbc) | |
940 | { | |
d1310b2e CM |
941 | struct extent_io_tree *tree; |
942 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 943 | if (wbc->sync_mode == WB_SYNC_NONE) { |
b9473439 | 944 | struct btrfs_root *root = BTRFS_I(mapping->host)->root; |
793955bc | 945 | u64 num_dirty; |
24ab9cd8 | 946 | unsigned long thresh = 32 * 1024 * 1024; |
448d640b CM |
947 | |
948 | if (wbc->for_kupdate) | |
949 | return 0; | |
950 | ||
b9473439 CM |
951 | /* this is a bit racy, but that's ok */ |
952 | num_dirty = root->fs_info->dirty_metadata_bytes; | |
d397712b | 953 | if (num_dirty < thresh) |
793955bc | 954 | return 0; |
793955bc | 955 | } |
0b32f4bb | 956 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
957 | } |
958 | ||
b2950863 | 959 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 960 | { |
d1310b2e CM |
961 | struct extent_io_tree *tree; |
962 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 963 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 964 | } |
22b0ebda | 965 | |
70dec807 | 966 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 967 | { |
98509cfc | 968 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 969 | return 0; |
0c4e538b DS |
970 | /* |
971 | * We need to mask out eg. __GFP_HIGHMEM and __GFP_DMA32 as we're doing | |
972 | * slab allocation from alloc_extent_state down the callchain where | |
973 | * it'd hit a BUG_ON as those flags are not allowed. | |
974 | */ | |
975 | gfp_flags &= ~GFP_SLAB_BUG_MASK; | |
976 | ||
3083ee2e | 977 | return try_release_extent_buffer(page, gfp_flags); |
d98237b3 CM |
978 | } |
979 | ||
5f39d397 | 980 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 981 | { |
d1310b2e CM |
982 | struct extent_io_tree *tree; |
983 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
984 | extent_invalidatepage(tree, page, offset); |
985 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 986 | if (PagePrivate(page)) { |
d397712b CM |
987 | printk(KERN_WARNING "btrfs warning page private not zero " |
988 | "on page %llu\n", (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
989 | ClearPagePrivate(page); |
990 | set_page_private(page, 0); | |
991 | page_cache_release(page); | |
992 | } | |
d98237b3 CM |
993 | } |
994 | ||
0b32f4bb JB |
995 | static int btree_set_page_dirty(struct page *page) |
996 | { | |
997 | struct extent_buffer *eb; | |
998 | ||
999 | BUG_ON(!PagePrivate(page)); | |
1000 | eb = (struct extent_buffer *)page->private; | |
1001 | BUG_ON(!eb); | |
1002 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1003 | BUG_ON(!atomic_read(&eb->refs)); | |
1004 | btrfs_assert_tree_locked(eb); | |
1005 | return __set_page_dirty_nobuffers(page); | |
1006 | } | |
1007 | ||
7f09410b | 1008 | static const struct address_space_operations btree_aops = { |
d98237b3 | 1009 | .readpage = btree_readpage, |
0da5468f | 1010 | .writepages = btree_writepages, |
5f39d397 CM |
1011 | .releasepage = btree_releasepage, |
1012 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1013 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1014 | .migratepage = btree_migratepage, |
5a92bc88 | 1015 | #endif |
0b32f4bb | 1016 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1017 | }; |
1018 | ||
ca7a79ad CM |
1019 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1020 | u64 parent_transid) | |
090d1875 | 1021 | { |
5f39d397 CM |
1022 | struct extent_buffer *buf = NULL; |
1023 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 1024 | int ret = 0; |
090d1875 | 1025 | |
db94535d | 1026 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 1027 | if (!buf) |
090d1875 | 1028 | return 0; |
d1310b2e | 1029 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
bb82ab88 | 1030 | buf, 0, WAIT_NONE, btree_get_extent, 0); |
5f39d397 | 1031 | free_extent_buffer(buf); |
de428b63 | 1032 | return ret; |
090d1875 CM |
1033 | } |
1034 | ||
ab0fff03 AJ |
1035 | int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1036 | int mirror_num, struct extent_buffer **eb) | |
1037 | { | |
1038 | struct extent_buffer *buf = NULL; | |
1039 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1040 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; | |
1041 | int ret; | |
1042 | ||
1043 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
1044 | if (!buf) | |
1045 | return 0; | |
1046 | ||
1047 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1048 | ||
1049 | ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, | |
1050 | btree_get_extent, mirror_num); | |
1051 | if (ret) { | |
1052 | free_extent_buffer(buf); | |
1053 | return ret; | |
1054 | } | |
1055 | ||
1056 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1057 | free_extent_buffer(buf); | |
1058 | return -EIO; | |
0b32f4bb | 1059 | } else if (extent_buffer_uptodate(buf)) { |
ab0fff03 AJ |
1060 | *eb = buf; |
1061 | } else { | |
1062 | free_extent_buffer(buf); | |
1063 | } | |
1064 | return 0; | |
1065 | } | |
1066 | ||
0999df54 CM |
1067 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
1068 | u64 bytenr, u32 blocksize) | |
1069 | { | |
1070 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1071 | struct extent_buffer *eb; | |
1072 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
f09d1f60 | 1073 | bytenr, blocksize); |
0999df54 CM |
1074 | return eb; |
1075 | } | |
1076 | ||
1077 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
1078 | u64 bytenr, u32 blocksize) | |
1079 | { | |
1080 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1081 | struct extent_buffer *eb; | |
1082 | ||
1083 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
727011e0 | 1084 | bytenr, blocksize); |
0999df54 CM |
1085 | return eb; |
1086 | } | |
1087 | ||
1088 | ||
e02119d5 CM |
1089 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1090 | { | |
727011e0 | 1091 | return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, |
8aa38c31 | 1092 | buf->start + buf->len - 1); |
e02119d5 CM |
1093 | } |
1094 | ||
1095 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
1096 | { | |
727011e0 | 1097 | return filemap_fdatawait_range(buf->pages[0]->mapping, |
8aa38c31 | 1098 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
1099 | } |
1100 | ||
0999df54 | 1101 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 1102 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
1103 | { |
1104 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1105 | int ret; |
1106 | ||
0999df54 CM |
1107 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
1108 | if (!buf) | |
1109 | return NULL; | |
0999df54 | 1110 | |
ca7a79ad | 1111 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
5f39d397 | 1112 | return buf; |
ce9adaa5 | 1113 | |
eb60ceac CM |
1114 | } |
1115 | ||
d5c13f92 JM |
1116 | void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
1117 | struct extent_buffer *buf) | |
ed2ff2cb | 1118 | { |
55c69072 | 1119 | if (btrfs_header_generation(buf) == |
925baedd | 1120 | root->fs_info->running_transaction->transid) { |
b9447ef8 | 1121 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1122 | |
b9473439 CM |
1123 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
1124 | spin_lock(&root->fs_info->delalloc_lock); | |
1125 | if (root->fs_info->dirty_metadata_bytes >= buf->len) | |
1126 | root->fs_info->dirty_metadata_bytes -= buf->len; | |
d5c13f92 JM |
1127 | else { |
1128 | spin_unlock(&root->fs_info->delalloc_lock); | |
1129 | btrfs_panic(root->fs_info, -EOVERFLOW, | |
1130 | "Can't clear %lu bytes from " | |
533574c6 | 1131 | " dirty_mdatadata_bytes (%llu)", |
d5c13f92 JM |
1132 | buf->len, |
1133 | root->fs_info->dirty_metadata_bytes); | |
1134 | } | |
b9473439 CM |
1135 | spin_unlock(&root->fs_info->delalloc_lock); |
1136 | } | |
b4ce94de | 1137 | |
b9473439 CM |
1138 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1139 | btrfs_set_lock_blocking(buf); | |
0b32f4bb | 1140 | clear_extent_buffer_dirty(buf); |
925baedd | 1141 | } |
5f39d397 CM |
1142 | } |
1143 | ||
143bede5 JM |
1144 | static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
1145 | u32 stripesize, struct btrfs_root *root, | |
1146 | struct btrfs_fs_info *fs_info, | |
1147 | u64 objectid) | |
d97e63b6 | 1148 | { |
cfaa7295 | 1149 | root->node = NULL; |
a28ec197 | 1150 | root->commit_root = NULL; |
db94535d CM |
1151 | root->sectorsize = sectorsize; |
1152 | root->nodesize = nodesize; | |
1153 | root->leafsize = leafsize; | |
87ee04eb | 1154 | root->stripesize = stripesize; |
123abc88 | 1155 | root->ref_cows = 0; |
0b86a832 | 1156 | root->track_dirty = 0; |
c71bf099 | 1157 | root->in_radix = 0; |
d68fc57b YZ |
1158 | root->orphan_item_inserted = 0; |
1159 | root->orphan_cleanup_state = 0; | |
0b86a832 | 1160 | |
0f7d52f4 CM |
1161 | root->objectid = objectid; |
1162 | root->last_trans = 0; | |
13a8a7c8 | 1163 | root->highest_objectid = 0; |
58176a96 | 1164 | root->name = NULL; |
6bef4d31 | 1165 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1166 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1167 | root->block_rsv = NULL; |
d68fc57b | 1168 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
1169 | |
1170 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1171 | INIT_LIST_HEAD(&root->root_list); |
d68fc57b | 1172 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 1173 | spin_lock_init(&root->inode_lock); |
f0486c68 | 1174 | spin_lock_init(&root->accounting_lock); |
a2135011 | 1175 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1176 | mutex_init(&root->log_mutex); |
7237f183 YZ |
1177 | init_waitqueue_head(&root->log_writer_wait); |
1178 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1179 | init_waitqueue_head(&root->log_commit_wait[1]); | |
1180 | atomic_set(&root->log_commit[0], 0); | |
1181 | atomic_set(&root->log_commit[1], 0); | |
1182 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1183 | atomic_set(&root->log_batch, 0); |
8a35d95f | 1184 | atomic_set(&root->orphan_inodes, 0); |
7237f183 | 1185 | root->log_transid = 0; |
257c62e1 | 1186 | root->last_log_commit = 0; |
d0c803c4 | 1187 | extent_io_tree_init(&root->dirty_log_pages, |
f993c883 | 1188 | fs_info->btree_inode->i_mapping); |
017e5369 | 1189 | |
3768f368 CM |
1190 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1191 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1192 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 1193 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 1194 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 1195 | init_completion(&root->kobj_unregister); |
6702ed49 | 1196 | root->defrag_running = 0; |
4d775673 | 1197 | root->root_key.objectid = objectid; |
0ee5dc67 | 1198 | root->anon_dev = 0; |
8ea05e3a AB |
1199 | |
1200 | spin_lock_init(&root->root_times_lock); | |
3768f368 CM |
1201 | } |
1202 | ||
200a5c17 JM |
1203 | static int __must_check find_and_setup_root(struct btrfs_root *tree_root, |
1204 | struct btrfs_fs_info *fs_info, | |
1205 | u64 objectid, | |
1206 | struct btrfs_root *root) | |
3768f368 CM |
1207 | { |
1208 | int ret; | |
db94535d | 1209 | u32 blocksize; |
84234f3a | 1210 | u64 generation; |
3768f368 | 1211 | |
db94535d | 1212 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1213 | tree_root->sectorsize, tree_root->stripesize, |
1214 | root, fs_info, objectid); | |
3768f368 CM |
1215 | ret = btrfs_find_last_root(tree_root, objectid, |
1216 | &root->root_item, &root->root_key); | |
4df27c4d YZ |
1217 | if (ret > 0) |
1218 | return -ENOENT; | |
200a5c17 JM |
1219 | else if (ret < 0) |
1220 | return ret; | |
3768f368 | 1221 | |
84234f3a | 1222 | generation = btrfs_root_generation(&root->root_item); |
db94535d | 1223 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
af31f5e5 | 1224 | root->commit_root = NULL; |
db94535d | 1225 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), |
84234f3a | 1226 | blocksize, generation); |
b9fab919 | 1227 | if (!root->node || !btrfs_buffer_uptodate(root->node, generation, 0)) { |
68433b73 | 1228 | free_extent_buffer(root->node); |
af31f5e5 | 1229 | root->node = NULL; |
68433b73 CM |
1230 | return -EIO; |
1231 | } | |
4df27c4d | 1232 | root->commit_root = btrfs_root_node(root); |
d97e63b6 CM |
1233 | return 0; |
1234 | } | |
1235 | ||
f84a8bd6 | 1236 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info) |
6f07e42e AV |
1237 | { |
1238 | struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS); | |
1239 | if (root) | |
1240 | root->fs_info = fs_info; | |
1241 | return root; | |
1242 | } | |
1243 | ||
20897f5c AJ |
1244 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
1245 | struct btrfs_fs_info *fs_info, | |
1246 | u64 objectid) | |
1247 | { | |
1248 | struct extent_buffer *leaf; | |
1249 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1250 | struct btrfs_root *root; | |
1251 | struct btrfs_key key; | |
1252 | int ret = 0; | |
1253 | u64 bytenr; | |
1254 | ||
1255 | root = btrfs_alloc_root(fs_info); | |
1256 | if (!root) | |
1257 | return ERR_PTR(-ENOMEM); | |
1258 | ||
1259 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1260 | tree_root->sectorsize, tree_root->stripesize, | |
1261 | root, fs_info, objectid); | |
1262 | root->root_key.objectid = objectid; | |
1263 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1264 | root->root_key.offset = 0; | |
1265 | ||
1266 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, | |
1267 | 0, objectid, NULL, 0, 0, 0); | |
1268 | if (IS_ERR(leaf)) { | |
1269 | ret = PTR_ERR(leaf); | |
1270 | goto fail; | |
1271 | } | |
1272 | ||
1273 | bytenr = leaf->start; | |
1274 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); | |
1275 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1276 | btrfs_set_header_generation(leaf, trans->transid); | |
1277 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1278 | btrfs_set_header_owner(leaf, objectid); | |
1279 | root->node = leaf; | |
1280 | ||
1281 | write_extent_buffer(leaf, fs_info->fsid, | |
1282 | (unsigned long)btrfs_header_fsid(leaf), | |
1283 | BTRFS_FSID_SIZE); | |
1284 | write_extent_buffer(leaf, fs_info->chunk_tree_uuid, | |
1285 | (unsigned long)btrfs_header_chunk_tree_uuid(leaf), | |
1286 | BTRFS_UUID_SIZE); | |
1287 | btrfs_mark_buffer_dirty(leaf); | |
1288 | ||
1289 | root->commit_root = btrfs_root_node(root); | |
1290 | root->track_dirty = 1; | |
1291 | ||
1292 | ||
1293 | root->root_item.flags = 0; | |
1294 | root->root_item.byte_limit = 0; | |
1295 | btrfs_set_root_bytenr(&root->root_item, leaf->start); | |
1296 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1297 | btrfs_set_root_level(&root->root_item, 0); | |
1298 | btrfs_set_root_refs(&root->root_item, 1); | |
1299 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1300 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1301 | btrfs_set_root_dirid(&root->root_item, 0); | |
1302 | root->root_item.drop_level = 0; | |
1303 | ||
1304 | key.objectid = objectid; | |
1305 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1306 | key.offset = 0; | |
1307 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1308 | if (ret) | |
1309 | goto fail; | |
1310 | ||
1311 | btrfs_tree_unlock(leaf); | |
1312 | ||
1313 | fail: | |
1314 | if (ret) | |
1315 | return ERR_PTR(ret); | |
1316 | ||
1317 | return root; | |
1318 | } | |
1319 | ||
7237f183 YZ |
1320 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1321 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1322 | { |
1323 | struct btrfs_root *root; | |
1324 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1325 | struct extent_buffer *leaf; |
e02119d5 | 1326 | |
6f07e42e | 1327 | root = btrfs_alloc_root(fs_info); |
e02119d5 | 1328 | if (!root) |
7237f183 | 1329 | return ERR_PTR(-ENOMEM); |
e02119d5 CM |
1330 | |
1331 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1332 | tree_root->sectorsize, tree_root->stripesize, | |
1333 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1334 | ||
1335 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1336 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1337 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
7237f183 YZ |
1338 | /* |
1339 | * log trees do not get reference counted because they go away | |
1340 | * before a real commit is actually done. They do store pointers | |
1341 | * to file data extents, and those reference counts still get | |
1342 | * updated (along with back refs to the log tree). | |
1343 | */ | |
e02119d5 CM |
1344 | root->ref_cows = 0; |
1345 | ||
5d4f98a2 | 1346 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, |
66d7e7f0 | 1347 | BTRFS_TREE_LOG_OBJECTID, NULL, |
5581a51a | 1348 | 0, 0, 0); |
7237f183 YZ |
1349 | if (IS_ERR(leaf)) { |
1350 | kfree(root); | |
1351 | return ERR_CAST(leaf); | |
1352 | } | |
e02119d5 | 1353 | |
5d4f98a2 YZ |
1354 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1355 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1356 | btrfs_set_header_generation(leaf, trans->transid); | |
1357 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1358 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1359 | root->node = leaf; |
e02119d5 CM |
1360 | |
1361 | write_extent_buffer(root->node, root->fs_info->fsid, | |
1362 | (unsigned long)btrfs_header_fsid(root->node), | |
1363 | BTRFS_FSID_SIZE); | |
1364 | btrfs_mark_buffer_dirty(root->node); | |
1365 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1366 | return root; |
1367 | } | |
1368 | ||
1369 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1370 | struct btrfs_fs_info *fs_info) | |
1371 | { | |
1372 | struct btrfs_root *log_root; | |
1373 | ||
1374 | log_root = alloc_log_tree(trans, fs_info); | |
1375 | if (IS_ERR(log_root)) | |
1376 | return PTR_ERR(log_root); | |
1377 | WARN_ON(fs_info->log_root_tree); | |
1378 | fs_info->log_root_tree = log_root; | |
1379 | return 0; | |
1380 | } | |
1381 | ||
1382 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1383 | struct btrfs_root *root) | |
1384 | { | |
1385 | struct btrfs_root *log_root; | |
1386 | struct btrfs_inode_item *inode_item; | |
1387 | ||
1388 | log_root = alloc_log_tree(trans, root->fs_info); | |
1389 | if (IS_ERR(log_root)) | |
1390 | return PTR_ERR(log_root); | |
1391 | ||
1392 | log_root->last_trans = trans->transid; | |
1393 | log_root->root_key.offset = root->root_key.objectid; | |
1394 | ||
1395 | inode_item = &log_root->root_item.inode; | |
1396 | inode_item->generation = cpu_to_le64(1); | |
1397 | inode_item->size = cpu_to_le64(3); | |
1398 | inode_item->nlink = cpu_to_le32(1); | |
1399 | inode_item->nbytes = cpu_to_le64(root->leafsize); | |
1400 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); | |
1401 | ||
5d4f98a2 | 1402 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1403 | |
1404 | WARN_ON(root->log_root); | |
1405 | root->log_root = log_root; | |
1406 | root->log_transid = 0; | |
257c62e1 | 1407 | root->last_log_commit = 0; |
e02119d5 CM |
1408 | return 0; |
1409 | } | |
1410 | ||
1411 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
1412 | struct btrfs_key *location) | |
1413 | { | |
1414 | struct btrfs_root *root; | |
1415 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1416 | struct btrfs_path *path; |
5f39d397 | 1417 | struct extent_buffer *l; |
84234f3a | 1418 | u64 generation; |
db94535d | 1419 | u32 blocksize; |
0f7d52f4 | 1420 | int ret = 0; |
8ea05e3a | 1421 | int slot; |
0f7d52f4 | 1422 | |
6f07e42e | 1423 | root = btrfs_alloc_root(fs_info); |
0cf6c620 | 1424 | if (!root) |
0f7d52f4 | 1425 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 1426 | if (location->offset == (u64)-1) { |
db94535d | 1427 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
1428 | location->objectid, root); |
1429 | if (ret) { | |
0f7d52f4 CM |
1430 | kfree(root); |
1431 | return ERR_PTR(ret); | |
1432 | } | |
13a8a7c8 | 1433 | goto out; |
0f7d52f4 CM |
1434 | } |
1435 | ||
db94535d | 1436 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1437 | tree_root->sectorsize, tree_root->stripesize, |
1438 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
1439 | |
1440 | path = btrfs_alloc_path(); | |
db5b493a TI |
1441 | if (!path) { |
1442 | kfree(root); | |
1443 | return ERR_PTR(-ENOMEM); | |
1444 | } | |
0f7d52f4 | 1445 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); |
13a8a7c8 YZ |
1446 | if (ret == 0) { |
1447 | l = path->nodes[0]; | |
8ea05e3a AB |
1448 | slot = path->slots[0]; |
1449 | btrfs_read_root_item(tree_root, l, slot, &root->root_item); | |
13a8a7c8 | 1450 | memcpy(&root->root_key, location, sizeof(*location)); |
0f7d52f4 | 1451 | } |
0f7d52f4 CM |
1452 | btrfs_free_path(path); |
1453 | if (ret) { | |
5e540f77 | 1454 | kfree(root); |
13a8a7c8 YZ |
1455 | if (ret > 0) |
1456 | ret = -ENOENT; | |
0f7d52f4 CM |
1457 | return ERR_PTR(ret); |
1458 | } | |
13a8a7c8 | 1459 | |
84234f3a | 1460 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1461 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1462 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1463 | blocksize, generation); |
5d4f98a2 | 1464 | root->commit_root = btrfs_root_node(root); |
79787eaa | 1465 | BUG_ON(!root->node); /* -ENOMEM */ |
13a8a7c8 | 1466 | out: |
08fe4db1 | 1467 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { |
e02119d5 | 1468 | root->ref_cows = 1; |
08fe4db1 LZ |
1469 | btrfs_check_and_init_root_item(&root->root_item); |
1470 | } | |
13a8a7c8 | 1471 | |
5eda7b5e CM |
1472 | return root; |
1473 | } | |
1474 | ||
edbd8d4e CM |
1475 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
1476 | struct btrfs_key *location) | |
5eda7b5e CM |
1477 | { |
1478 | struct btrfs_root *root; | |
1479 | int ret; | |
1480 | ||
edbd8d4e CM |
1481 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1482 | return fs_info->tree_root; | |
1483 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1484 | return fs_info->extent_root; | |
8f18cf13 CM |
1485 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1486 | return fs_info->chunk_root; | |
1487 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1488 | return fs_info->dev_root; | |
0403e47e YZ |
1489 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1490 | return fs_info->csum_root; | |
bcef60f2 AJ |
1491 | if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1492 | return fs_info->quota_root ? fs_info->quota_root : | |
1493 | ERR_PTR(-ENOENT); | |
4df27c4d YZ |
1494 | again: |
1495 | spin_lock(&fs_info->fs_roots_radix_lock); | |
5eda7b5e CM |
1496 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
1497 | (unsigned long)location->objectid); | |
4df27c4d | 1498 | spin_unlock(&fs_info->fs_roots_radix_lock); |
5eda7b5e CM |
1499 | if (root) |
1500 | return root; | |
1501 | ||
e02119d5 | 1502 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
1503 | if (IS_ERR(root)) |
1504 | return root; | |
3394e160 | 1505 | |
581bb050 | 1506 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); |
581bb050 LZ |
1507 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), |
1508 | GFP_NOFS); | |
35a30d7c DS |
1509 | if (!root->free_ino_pinned || !root->free_ino_ctl) { |
1510 | ret = -ENOMEM; | |
581bb050 | 1511 | goto fail; |
35a30d7c | 1512 | } |
581bb050 LZ |
1513 | |
1514 | btrfs_init_free_ino_ctl(root); | |
1515 | mutex_init(&root->fs_commit_mutex); | |
1516 | spin_lock_init(&root->cache_lock); | |
1517 | init_waitqueue_head(&root->cache_wait); | |
1518 | ||
0ee5dc67 | 1519 | ret = get_anon_bdev(&root->anon_dev); |
ac08aedf CM |
1520 | if (ret) |
1521 | goto fail; | |
3394e160 | 1522 | |
d68fc57b YZ |
1523 | if (btrfs_root_refs(&root->root_item) == 0) { |
1524 | ret = -ENOENT; | |
1525 | goto fail; | |
1526 | } | |
1527 | ||
1528 | ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid); | |
1529 | if (ret < 0) | |
1530 | goto fail; | |
1531 | if (ret == 0) | |
1532 | root->orphan_item_inserted = 1; | |
1533 | ||
4df27c4d YZ |
1534 | ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); |
1535 | if (ret) | |
1536 | goto fail; | |
1537 | ||
1538 | spin_lock(&fs_info->fs_roots_radix_lock); | |
2619ba1f CM |
1539 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
1540 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 | 1541 | root); |
d68fc57b | 1542 | if (ret == 0) |
4df27c4d | 1543 | root->in_radix = 1; |
d68fc57b | 1544 | |
4df27c4d YZ |
1545 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1546 | radix_tree_preload_end(); | |
0f7d52f4 | 1547 | if (ret) { |
4df27c4d YZ |
1548 | if (ret == -EEXIST) { |
1549 | free_fs_root(root); | |
1550 | goto again; | |
1551 | } | |
1552 | goto fail; | |
0f7d52f4 | 1553 | } |
4df27c4d YZ |
1554 | |
1555 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
1556 | root->root_key.objectid); | |
1557 | WARN_ON(ret); | |
edbd8d4e | 1558 | return root; |
4df27c4d YZ |
1559 | fail: |
1560 | free_fs_root(root); | |
1561 | return ERR_PTR(ret); | |
edbd8d4e CM |
1562 | } |
1563 | ||
04160088 CM |
1564 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1565 | { | |
1566 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1567 | int ret = 0; | |
04160088 CM |
1568 | struct btrfs_device *device; |
1569 | struct backing_dev_info *bdi; | |
b7967db7 | 1570 | |
1f78160c XG |
1571 | rcu_read_lock(); |
1572 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1573 | if (!device->bdev) |
1574 | continue; | |
04160088 CM |
1575 | bdi = blk_get_backing_dev_info(device->bdev); |
1576 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1577 | ret = 1; | |
1578 | break; | |
1579 | } | |
1580 | } | |
1f78160c | 1581 | rcu_read_unlock(); |
04160088 CM |
1582 | return ret; |
1583 | } | |
1584 | ||
ad081f14 JA |
1585 | /* |
1586 | * If this fails, caller must call bdi_destroy() to get rid of the | |
1587 | * bdi again. | |
1588 | */ | |
04160088 CM |
1589 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1590 | { | |
ad081f14 JA |
1591 | int err; |
1592 | ||
1593 | bdi->capabilities = BDI_CAP_MAP_COPY; | |
e6d086d8 | 1594 | err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY); |
ad081f14 JA |
1595 | if (err) |
1596 | return err; | |
1597 | ||
4575c9cc | 1598 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1599 | bdi->congested_fn = btrfs_congested_fn; |
1600 | bdi->congested_data = info; | |
1601 | return 0; | |
1602 | } | |
1603 | ||
8b712842 CM |
1604 | /* |
1605 | * called by the kthread helper functions to finally call the bio end_io | |
1606 | * functions. This is where read checksum verification actually happens | |
1607 | */ | |
1608 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1609 | { |
ce9adaa5 | 1610 | struct bio *bio; |
8b712842 CM |
1611 | struct end_io_wq *end_io_wq; |
1612 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1613 | int error; |
ce9adaa5 | 1614 | |
8b712842 CM |
1615 | end_io_wq = container_of(work, struct end_io_wq, work); |
1616 | bio = end_io_wq->bio; | |
1617 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1618 | |
8b712842 CM |
1619 | error = end_io_wq->error; |
1620 | bio->bi_private = end_io_wq->private; | |
1621 | bio->bi_end_io = end_io_wq->end_io; | |
1622 | kfree(end_io_wq); | |
8b712842 | 1623 | bio_endio(bio, error); |
44b8bd7e CM |
1624 | } |
1625 | ||
a74a4b97 CM |
1626 | static int cleaner_kthread(void *arg) |
1627 | { | |
1628 | struct btrfs_root *root = arg; | |
1629 | ||
1630 | do { | |
76dda93c YZ |
1631 | if (!(root->fs_info->sb->s_flags & MS_RDONLY) && |
1632 | mutex_trylock(&root->fs_info->cleaner_mutex)) { | |
24bbcf04 | 1633 | btrfs_run_delayed_iputs(root); |
76dda93c YZ |
1634 | btrfs_clean_old_snapshots(root); |
1635 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
4cb5300b | 1636 | btrfs_run_defrag_inodes(root->fs_info); |
76dda93c | 1637 | } |
a74a4b97 | 1638 | |
a0acae0e | 1639 | if (!try_to_freeze()) { |
a74a4b97 | 1640 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1641 | if (!kthread_should_stop()) |
1642 | schedule(); | |
a74a4b97 CM |
1643 | __set_current_state(TASK_RUNNING); |
1644 | } | |
1645 | } while (!kthread_should_stop()); | |
1646 | return 0; | |
1647 | } | |
1648 | ||
1649 | static int transaction_kthread(void *arg) | |
1650 | { | |
1651 | struct btrfs_root *root = arg; | |
1652 | struct btrfs_trans_handle *trans; | |
1653 | struct btrfs_transaction *cur; | |
8929ecfa | 1654 | u64 transid; |
a74a4b97 CM |
1655 | unsigned long now; |
1656 | unsigned long delay; | |
914b2007 | 1657 | bool cannot_commit; |
a74a4b97 CM |
1658 | |
1659 | do { | |
914b2007 | 1660 | cannot_commit = false; |
a74a4b97 | 1661 | delay = HZ * 30; |
a74a4b97 CM |
1662 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
1663 | ||
a4abeea4 | 1664 | spin_lock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1665 | cur = root->fs_info->running_transaction; |
1666 | if (!cur) { | |
a4abeea4 | 1667 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1668 | goto sleep; |
1669 | } | |
31153d81 | 1670 | |
a74a4b97 | 1671 | now = get_seconds(); |
8929ecfa YZ |
1672 | if (!cur->blocked && |
1673 | (now < cur->start_time || now - cur->start_time < 30)) { | |
a4abeea4 | 1674 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1675 | delay = HZ * 5; |
1676 | goto sleep; | |
1677 | } | |
8929ecfa | 1678 | transid = cur->transid; |
a4abeea4 | 1679 | spin_unlock(&root->fs_info->trans_lock); |
56bec294 | 1680 | |
79787eaa | 1681 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1682 | trans = btrfs_attach_transaction(root); |
914b2007 | 1683 | if (IS_ERR(trans)) { |
354aa0fb MX |
1684 | if (PTR_ERR(trans) != -ENOENT) |
1685 | cannot_commit = true; | |
79787eaa | 1686 | goto sleep; |
914b2007 | 1687 | } |
8929ecfa | 1688 | if (transid == trans->transid) { |
79787eaa | 1689 | btrfs_commit_transaction(trans, root); |
8929ecfa YZ |
1690 | } else { |
1691 | btrfs_end_transaction(trans, root); | |
1692 | } | |
a74a4b97 CM |
1693 | sleep: |
1694 | wake_up_process(root->fs_info->cleaner_kthread); | |
1695 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1696 | ||
a0acae0e | 1697 | if (!try_to_freeze()) { |
a74a4b97 | 1698 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa | 1699 | if (!kthread_should_stop() && |
914b2007 JK |
1700 | (!btrfs_transaction_blocked(root->fs_info) || |
1701 | cannot_commit)) | |
8929ecfa | 1702 | schedule_timeout(delay); |
a74a4b97 CM |
1703 | __set_current_state(TASK_RUNNING); |
1704 | } | |
1705 | } while (!kthread_should_stop()); | |
1706 | return 0; | |
1707 | } | |
1708 | ||
af31f5e5 CM |
1709 | /* |
1710 | * this will find the highest generation in the array of | |
1711 | * root backups. The index of the highest array is returned, | |
1712 | * or -1 if we can't find anything. | |
1713 | * | |
1714 | * We check to make sure the array is valid by comparing the | |
1715 | * generation of the latest root in the array with the generation | |
1716 | * in the super block. If they don't match we pitch it. | |
1717 | */ | |
1718 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1719 | { | |
1720 | u64 cur; | |
1721 | int newest_index = -1; | |
1722 | struct btrfs_root_backup *root_backup; | |
1723 | int i; | |
1724 | ||
1725 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1726 | root_backup = info->super_copy->super_roots + i; | |
1727 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1728 | if (cur == newest_gen) | |
1729 | newest_index = i; | |
1730 | } | |
1731 | ||
1732 | /* check to see if we actually wrapped around */ | |
1733 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
1734 | root_backup = info->super_copy->super_roots; | |
1735 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1736 | if (cur == newest_gen) | |
1737 | newest_index = 0; | |
1738 | } | |
1739 | return newest_index; | |
1740 | } | |
1741 | ||
1742 | ||
1743 | /* | |
1744 | * find the oldest backup so we know where to store new entries | |
1745 | * in the backup array. This will set the backup_root_index | |
1746 | * field in the fs_info struct | |
1747 | */ | |
1748 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
1749 | u64 newest_gen) | |
1750 | { | |
1751 | int newest_index = -1; | |
1752 | ||
1753 | newest_index = find_newest_super_backup(info, newest_gen); | |
1754 | /* if there was garbage in there, just move along */ | |
1755 | if (newest_index == -1) { | |
1756 | info->backup_root_index = 0; | |
1757 | } else { | |
1758 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1759 | } | |
1760 | } | |
1761 | ||
1762 | /* | |
1763 | * copy all the root pointers into the super backup array. | |
1764 | * this will bump the backup pointer by one when it is | |
1765 | * done | |
1766 | */ | |
1767 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1768 | { | |
1769 | int next_backup; | |
1770 | struct btrfs_root_backup *root_backup; | |
1771 | int last_backup; | |
1772 | ||
1773 | next_backup = info->backup_root_index; | |
1774 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1775 | BTRFS_NUM_BACKUP_ROOTS; | |
1776 | ||
1777 | /* | |
1778 | * just overwrite the last backup if we're at the same generation | |
1779 | * this happens only at umount | |
1780 | */ | |
1781 | root_backup = info->super_for_commit->super_roots + last_backup; | |
1782 | if (btrfs_backup_tree_root_gen(root_backup) == | |
1783 | btrfs_header_generation(info->tree_root->node)) | |
1784 | next_backup = last_backup; | |
1785 | ||
1786 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1787 | ||
1788 | /* | |
1789 | * make sure all of our padding and empty slots get zero filled | |
1790 | * regardless of which ones we use today | |
1791 | */ | |
1792 | memset(root_backup, 0, sizeof(*root_backup)); | |
1793 | ||
1794 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1795 | ||
1796 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1797 | btrfs_set_backup_tree_root_gen(root_backup, | |
1798 | btrfs_header_generation(info->tree_root->node)); | |
1799 | ||
1800 | btrfs_set_backup_tree_root_level(root_backup, | |
1801 | btrfs_header_level(info->tree_root->node)); | |
1802 | ||
1803 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1804 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1805 | btrfs_header_generation(info->chunk_root->node)); | |
1806 | btrfs_set_backup_chunk_root_level(root_backup, | |
1807 | btrfs_header_level(info->chunk_root->node)); | |
1808 | ||
1809 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
1810 | btrfs_set_backup_extent_root_gen(root_backup, | |
1811 | btrfs_header_generation(info->extent_root->node)); | |
1812 | btrfs_set_backup_extent_root_level(root_backup, | |
1813 | btrfs_header_level(info->extent_root->node)); | |
1814 | ||
7c7e82a7 CM |
1815 | /* |
1816 | * we might commit during log recovery, which happens before we set | |
1817 | * the fs_root. Make sure it is valid before we fill it in. | |
1818 | */ | |
1819 | if (info->fs_root && info->fs_root->node) { | |
1820 | btrfs_set_backup_fs_root(root_backup, | |
1821 | info->fs_root->node->start); | |
1822 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1823 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1824 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1825 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1826 | } |
af31f5e5 CM |
1827 | |
1828 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1829 | btrfs_set_backup_dev_root_gen(root_backup, | |
1830 | btrfs_header_generation(info->dev_root->node)); | |
1831 | btrfs_set_backup_dev_root_level(root_backup, | |
1832 | btrfs_header_level(info->dev_root->node)); | |
1833 | ||
1834 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
1835 | btrfs_set_backup_csum_root_gen(root_backup, | |
1836 | btrfs_header_generation(info->csum_root->node)); | |
1837 | btrfs_set_backup_csum_root_level(root_backup, | |
1838 | btrfs_header_level(info->csum_root->node)); | |
1839 | ||
1840 | btrfs_set_backup_total_bytes(root_backup, | |
1841 | btrfs_super_total_bytes(info->super_copy)); | |
1842 | btrfs_set_backup_bytes_used(root_backup, | |
1843 | btrfs_super_bytes_used(info->super_copy)); | |
1844 | btrfs_set_backup_num_devices(root_backup, | |
1845 | btrfs_super_num_devices(info->super_copy)); | |
1846 | ||
1847 | /* | |
1848 | * if we don't copy this out to the super_copy, it won't get remembered | |
1849 | * for the next commit | |
1850 | */ | |
1851 | memcpy(&info->super_copy->super_roots, | |
1852 | &info->super_for_commit->super_roots, | |
1853 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1854 | } | |
1855 | ||
1856 | /* | |
1857 | * this copies info out of the root backup array and back into | |
1858 | * the in-memory super block. It is meant to help iterate through | |
1859 | * the array, so you send it the number of backups you've already | |
1860 | * tried and the last backup index you used. | |
1861 | * | |
1862 | * this returns -1 when it has tried all the backups | |
1863 | */ | |
1864 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
1865 | struct btrfs_super_block *super, | |
1866 | int *num_backups_tried, int *backup_index) | |
1867 | { | |
1868 | struct btrfs_root_backup *root_backup; | |
1869 | int newest = *backup_index; | |
1870 | ||
1871 | if (*num_backups_tried == 0) { | |
1872 | u64 gen = btrfs_super_generation(super); | |
1873 | ||
1874 | newest = find_newest_super_backup(info, gen); | |
1875 | if (newest == -1) | |
1876 | return -1; | |
1877 | ||
1878 | *backup_index = newest; | |
1879 | *num_backups_tried = 1; | |
1880 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
1881 | /* we've tried all the backups, all done */ | |
1882 | return -1; | |
1883 | } else { | |
1884 | /* jump to the next oldest backup */ | |
1885 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1886 | BTRFS_NUM_BACKUP_ROOTS; | |
1887 | *backup_index = newest; | |
1888 | *num_backups_tried += 1; | |
1889 | } | |
1890 | root_backup = super->super_roots + newest; | |
1891 | ||
1892 | btrfs_set_super_generation(super, | |
1893 | btrfs_backup_tree_root_gen(root_backup)); | |
1894 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
1895 | btrfs_set_super_root_level(super, | |
1896 | btrfs_backup_tree_root_level(root_backup)); | |
1897 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
1898 | ||
1899 | /* | |
1900 | * fixme: the total bytes and num_devices need to match or we should | |
1901 | * need a fsck | |
1902 | */ | |
1903 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
1904 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
1905 | return 0; | |
1906 | } | |
1907 | ||
1908 | /* helper to cleanup tree roots */ | |
1909 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
1910 | { | |
1911 | free_extent_buffer(info->tree_root->node); | |
1912 | free_extent_buffer(info->tree_root->commit_root); | |
1913 | free_extent_buffer(info->dev_root->node); | |
1914 | free_extent_buffer(info->dev_root->commit_root); | |
1915 | free_extent_buffer(info->extent_root->node); | |
1916 | free_extent_buffer(info->extent_root->commit_root); | |
1917 | free_extent_buffer(info->csum_root->node); | |
1918 | free_extent_buffer(info->csum_root->commit_root); | |
bcef60f2 AJ |
1919 | if (info->quota_root) { |
1920 | free_extent_buffer(info->quota_root->node); | |
1921 | free_extent_buffer(info->quota_root->commit_root); | |
1922 | } | |
af31f5e5 CM |
1923 | |
1924 | info->tree_root->node = NULL; | |
1925 | info->tree_root->commit_root = NULL; | |
1926 | info->dev_root->node = NULL; | |
1927 | info->dev_root->commit_root = NULL; | |
1928 | info->extent_root->node = NULL; | |
1929 | info->extent_root->commit_root = NULL; | |
1930 | info->csum_root->node = NULL; | |
1931 | info->csum_root->commit_root = NULL; | |
bcef60f2 AJ |
1932 | if (info->quota_root) { |
1933 | info->quota_root->node = NULL; | |
1934 | info->quota_root->commit_root = NULL; | |
1935 | } | |
af31f5e5 CM |
1936 | |
1937 | if (chunk_root) { | |
1938 | free_extent_buffer(info->chunk_root->node); | |
1939 | free_extent_buffer(info->chunk_root->commit_root); | |
1940 | info->chunk_root->node = NULL; | |
1941 | info->chunk_root->commit_root = NULL; | |
1942 | } | |
1943 | } | |
1944 | ||
1945 | ||
ad2b2c80 AV |
1946 | int open_ctree(struct super_block *sb, |
1947 | struct btrfs_fs_devices *fs_devices, | |
1948 | char *options) | |
2e635a27 | 1949 | { |
db94535d CM |
1950 | u32 sectorsize; |
1951 | u32 nodesize; | |
1952 | u32 leafsize; | |
1953 | u32 blocksize; | |
87ee04eb | 1954 | u32 stripesize; |
84234f3a | 1955 | u64 generation; |
f2b636e8 | 1956 | u64 features; |
3de4586c | 1957 | struct btrfs_key location; |
a061fc8d | 1958 | struct buffer_head *bh; |
4d34b278 | 1959 | struct btrfs_super_block *disk_super; |
815745cf | 1960 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
f84a8bd6 | 1961 | struct btrfs_root *tree_root; |
4d34b278 ID |
1962 | struct btrfs_root *extent_root; |
1963 | struct btrfs_root *csum_root; | |
1964 | struct btrfs_root *chunk_root; | |
1965 | struct btrfs_root *dev_root; | |
bcef60f2 | 1966 | struct btrfs_root *quota_root; |
e02119d5 | 1967 | struct btrfs_root *log_tree_root; |
eb60ceac | 1968 | int ret; |
e58ca020 | 1969 | int err = -EINVAL; |
af31f5e5 CM |
1970 | int num_backups_tried = 0; |
1971 | int backup_index = 0; | |
4543df7e | 1972 | |
f84a8bd6 | 1973 | tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info); |
6f07e42e AV |
1974 | extent_root = fs_info->extent_root = btrfs_alloc_root(fs_info); |
1975 | csum_root = fs_info->csum_root = btrfs_alloc_root(fs_info); | |
1976 | chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info); | |
1977 | dev_root = fs_info->dev_root = btrfs_alloc_root(fs_info); | |
bcef60f2 | 1978 | quota_root = fs_info->quota_root = btrfs_alloc_root(fs_info); |
8790d502 | 1979 | |
f84a8bd6 | 1980 | if (!tree_root || !extent_root || !csum_root || |
bcef60f2 | 1981 | !chunk_root || !dev_root || !quota_root) { |
39279cc3 CM |
1982 | err = -ENOMEM; |
1983 | goto fail; | |
1984 | } | |
76dda93c YZ |
1985 | |
1986 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
1987 | if (ret) { | |
1988 | err = ret; | |
1989 | goto fail; | |
1990 | } | |
1991 | ||
1992 | ret = setup_bdi(fs_info, &fs_info->bdi); | |
1993 | if (ret) { | |
1994 | err = ret; | |
1995 | goto fail_srcu; | |
1996 | } | |
1997 | ||
1998 | fs_info->btree_inode = new_inode(sb); | |
1999 | if (!fs_info->btree_inode) { | |
2000 | err = -ENOMEM; | |
2001 | goto fail_bdi; | |
2002 | } | |
2003 | ||
a6591715 | 2004 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
1561deda | 2005 | |
76dda93c | 2006 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
8fd17795 | 2007 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2008 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2009 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
ea8c2819 | 2010 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
5a3f23d5 | 2011 | INIT_LIST_HEAD(&fs_info->ordered_operations); |
11833d66 | 2012 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
1832a6d5 | 2013 | spin_lock_init(&fs_info->delalloc_lock); |
a4abeea4 | 2014 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 2015 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2016 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2017 | spin_lock_init(&fs_info->defrag_inodes_lock); |
2bf64758 | 2018 | spin_lock_init(&fs_info->free_chunk_lock); |
f29021b2 JS |
2019 | spin_lock_init(&fs_info->tree_mod_seq_lock); |
2020 | rwlock_init(&fs_info->tree_mod_log_lock); | |
7585717f | 2021 | mutex_init(&fs_info->reloc_mutex); |
19c00ddc | 2022 | |
58176a96 | 2023 | init_completion(&fs_info->kobj_unregister); |
0b86a832 | 2024 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2025 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2026 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
0b86a832 | 2027 | btrfs_mapping_init(&fs_info->mapping_tree); |
66d8f3dd MX |
2028 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2029 | BTRFS_BLOCK_RSV_GLOBAL); | |
2030 | btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, | |
2031 | BTRFS_BLOCK_RSV_DELALLOC); | |
2032 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); | |
2033 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2034 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2035 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2036 | BTRFS_BLOCK_RSV_DELOPS); | |
cb03c743 | 2037 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 2038 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 2039 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 2040 | atomic_set(&fs_info->nr_async_bios, 0); |
4cb5300b | 2041 | atomic_set(&fs_info->defrag_running, 0); |
f29021b2 | 2042 | atomic_set(&fs_info->tree_mod_seq, 0); |
e20d96d6 | 2043 | fs_info->sb = sb; |
6f568d35 | 2044 | fs_info->max_inline = 8192 * 1024; |
9ed74f2d | 2045 | fs_info->metadata_ratio = 0; |
4cb5300b | 2046 | fs_info->defrag_inodes = RB_ROOT; |
a4abeea4 | 2047 | fs_info->trans_no_join = 0; |
2bf64758 | 2048 | fs_info->free_chunk_space = 0; |
f29021b2 | 2049 | fs_info->tree_mod_log = RB_ROOT; |
c8b97818 | 2050 | |
90519d66 AJ |
2051 | /* readahead state */ |
2052 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); | |
2053 | spin_lock_init(&fs_info->reada_lock); | |
c8b97818 | 2054 | |
b34b086c CM |
2055 | fs_info->thread_pool_size = min_t(unsigned long, |
2056 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2057 | |
3eaa2885 CM |
2058 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
2059 | spin_lock_init(&fs_info->ordered_extent_lock); | |
16cdcec7 MX |
2060 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
2061 | GFP_NOFS); | |
2062 | if (!fs_info->delayed_root) { | |
2063 | err = -ENOMEM; | |
2064 | goto fail_iput; | |
2065 | } | |
2066 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 2067 | |
a2de733c AJ |
2068 | mutex_init(&fs_info->scrub_lock); |
2069 | atomic_set(&fs_info->scrubs_running, 0); | |
2070 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2071 | atomic_set(&fs_info->scrubs_paused, 0); | |
2072 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2073 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
2074 | init_rwsem(&fs_info->scrub_super_lock); | |
2075 | fs_info->scrub_workers_refcnt = 0; | |
21adbd5c SB |
2076 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2077 | fs_info->check_integrity_print_mask = 0; | |
2078 | #endif | |
a2de733c | 2079 | |
c9e9f97b ID |
2080 | spin_lock_init(&fs_info->balance_lock); |
2081 | mutex_init(&fs_info->balance_mutex); | |
837d5b6e ID |
2082 | atomic_set(&fs_info->balance_running, 0); |
2083 | atomic_set(&fs_info->balance_pause_req, 0); | |
a7e99c69 | 2084 | atomic_set(&fs_info->balance_cancel_req, 0); |
c9e9f97b | 2085 | fs_info->balance_ctl = NULL; |
837d5b6e | 2086 | init_waitqueue_head(&fs_info->balance_wait_q); |
a2de733c | 2087 | |
a061fc8d CM |
2088 | sb->s_blocksize = 4096; |
2089 | sb->s_blocksize_bits = blksize_bits(4096); | |
32a88aa1 | 2090 | sb->s_bdi = &fs_info->bdi; |
a061fc8d | 2091 | |
76dda93c | 2092 | fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; |
bfe86848 | 2093 | set_nlink(fs_info->btree_inode, 1); |
0afbaf8c CM |
2094 | /* |
2095 | * we set the i_size on the btree inode to the max possible int. | |
2096 | * the real end of the address space is determined by all of | |
2097 | * the devices in the system | |
2098 | */ | |
2099 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 2100 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
2101 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
2102 | ||
5d4f98a2 | 2103 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); |
d1310b2e | 2104 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
f993c883 | 2105 | fs_info->btree_inode->i_mapping); |
0b32f4bb | 2106 | BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; |
a8067e02 | 2107 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); |
d1310b2e CM |
2108 | |
2109 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 2110 | |
76dda93c YZ |
2111 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
2112 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
2113 | sizeof(struct btrfs_key)); | |
72ac3c0d JB |
2114 | set_bit(BTRFS_INODE_DUMMY, |
2115 | &BTRFS_I(fs_info->btree_inode)->runtime_flags); | |
c65ddb52 | 2116 | insert_inode_hash(fs_info->btree_inode); |
76dda93c | 2117 | |
0f9dd46c | 2118 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2119 | fs_info->block_group_cache_tree = RB_ROOT; |
0f9dd46c | 2120 | |
11833d66 | 2121 | extent_io_tree_init(&fs_info->freed_extents[0], |
f993c883 | 2122 | fs_info->btree_inode->i_mapping); |
11833d66 | 2123 | extent_io_tree_init(&fs_info->freed_extents[1], |
f993c883 | 2124 | fs_info->btree_inode->i_mapping); |
11833d66 | 2125 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
e66f709b | 2126 | fs_info->do_barriers = 1; |
e18e4809 | 2127 | |
39279cc3 | 2128 | |
5a3f23d5 | 2129 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2130 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2131 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2132 | mutex_init(&fs_info->transaction_kthread_mutex); |
2133 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 2134 | mutex_init(&fs_info->volume_mutex); |
276e680d | 2135 | init_rwsem(&fs_info->extent_commit_sem); |
c71bf099 | 2136 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2137 | init_rwsem(&fs_info->subvol_sem); |
fa9c0d79 | 2138 | |
416ac51d AJ |
2139 | spin_lock_init(&fs_info->qgroup_lock); |
2140 | fs_info->qgroup_tree = RB_ROOT; | |
2141 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); | |
2142 | fs_info->qgroup_seq = 1; | |
2143 | fs_info->quota_enabled = 0; | |
2144 | fs_info->pending_quota_state = 0; | |
2145 | ||
fa9c0d79 CM |
2146 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2147 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2148 | ||
e6dcd2dc | 2149 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2150 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2151 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2152 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2153 | |
0b86a832 | 2154 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 2155 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2156 | |
3c4bb26b | 2157 | invalidate_bdev(fs_devices->latest_bdev); |
a512bbf8 | 2158 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
20b45077 DY |
2159 | if (!bh) { |
2160 | err = -EINVAL; | |
16cdcec7 | 2161 | goto fail_alloc; |
20b45077 | 2162 | } |
39279cc3 | 2163 | |
6c41761f DS |
2164 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2165 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2166 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2167 | brelse(bh); |
5f39d397 | 2168 | |
6c41761f | 2169 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2170 | |
6c41761f | 2171 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2172 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2173 | goto fail_alloc; |
0f7d52f4 | 2174 | |
acce952b | 2175 | /* check FS state, whether FS is broken. */ |
2176 | fs_info->fs_state |= btrfs_super_flags(disk_super); | |
2177 | ||
fcd1f065 DS |
2178 | ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); |
2179 | if (ret) { | |
2180 | printk(KERN_ERR "btrfs: superblock contains fatal errors\n"); | |
2181 | err = ret; | |
2182 | goto fail_alloc; | |
2183 | } | |
acce952b | 2184 | |
af31f5e5 CM |
2185 | /* |
2186 | * run through our array of backup supers and setup | |
2187 | * our ring pointer to the oldest one | |
2188 | */ | |
2189 | generation = btrfs_super_generation(disk_super); | |
2190 | find_oldest_super_backup(fs_info, generation); | |
2191 | ||
75e7cb7f LB |
2192 | /* |
2193 | * In the long term, we'll store the compression type in the super | |
2194 | * block, and it'll be used for per file compression control. | |
2195 | */ | |
2196 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2197 | ||
2b82032c YZ |
2198 | ret = btrfs_parse_options(tree_root, options); |
2199 | if (ret) { | |
2200 | err = ret; | |
16cdcec7 | 2201 | goto fail_alloc; |
2b82032c | 2202 | } |
dfe25020 | 2203 | |
f2b636e8 JB |
2204 | features = btrfs_super_incompat_flags(disk_super) & |
2205 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2206 | if (features) { | |
2207 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
2208 | "unsupported optional features (%Lx).\n", | |
21380931 | 2209 | (unsigned long long)features); |
f2b636e8 | 2210 | err = -EINVAL; |
16cdcec7 | 2211 | goto fail_alloc; |
f2b636e8 JB |
2212 | } |
2213 | ||
727011e0 CM |
2214 | if (btrfs_super_leafsize(disk_super) != |
2215 | btrfs_super_nodesize(disk_super)) { | |
2216 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2217 | "blocksizes don't match. node %d leaf %d\n", | |
2218 | btrfs_super_nodesize(disk_super), | |
2219 | btrfs_super_leafsize(disk_super)); | |
2220 | err = -EINVAL; | |
2221 | goto fail_alloc; | |
2222 | } | |
2223 | if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2224 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2225 | "blocksize (%d) was too large\n", | |
2226 | btrfs_super_leafsize(disk_super)); | |
2227 | err = -EINVAL; | |
2228 | goto fail_alloc; | |
2229 | } | |
2230 | ||
5d4f98a2 | 2231 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 2232 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
69e380d1 | 2233 | if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 2234 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
727011e0 CM |
2235 | |
2236 | /* | |
2237 | * flag our filesystem as having big metadata blocks if | |
2238 | * they are bigger than the page size | |
2239 | */ | |
2240 | if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) { | |
2241 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) | |
2242 | printk(KERN_INFO "btrfs flagging fs with big metadata feature\n"); | |
2243 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
2244 | } | |
2245 | ||
bc3f116f CM |
2246 | nodesize = btrfs_super_nodesize(disk_super); |
2247 | leafsize = btrfs_super_leafsize(disk_super); | |
2248 | sectorsize = btrfs_super_sectorsize(disk_super); | |
2249 | stripesize = btrfs_super_stripesize(disk_super); | |
2250 | ||
2251 | /* | |
2252 | * mixed block groups end up with duplicate but slightly offset | |
2253 | * extent buffers for the same range. It leads to corruptions | |
2254 | */ | |
2255 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
2256 | (sectorsize != leafsize)) { | |
2257 | printk(KERN_WARNING "btrfs: unequal leaf/node/sector sizes " | |
2258 | "are not allowed for mixed block groups on %s\n", | |
2259 | sb->s_id); | |
2260 | goto fail_alloc; | |
2261 | } | |
2262 | ||
a6fa6fae | 2263 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 2264 | |
f2b636e8 JB |
2265 | features = btrfs_super_compat_ro_flags(disk_super) & |
2266 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
2267 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
2268 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
2269 | "unsupported option features (%Lx).\n", | |
21380931 | 2270 | (unsigned long long)features); |
f2b636e8 | 2271 | err = -EINVAL; |
16cdcec7 | 2272 | goto fail_alloc; |
f2b636e8 | 2273 | } |
61d92c32 CM |
2274 | |
2275 | btrfs_init_workers(&fs_info->generic_worker, | |
2276 | "genwork", 1, NULL); | |
2277 | ||
5443be45 | 2278 | btrfs_init_workers(&fs_info->workers, "worker", |
61d92c32 CM |
2279 | fs_info->thread_pool_size, |
2280 | &fs_info->generic_worker); | |
c8b97818 | 2281 | |
771ed689 | 2282 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
61d92c32 CM |
2283 | fs_info->thread_pool_size, |
2284 | &fs_info->generic_worker); | |
771ed689 | 2285 | |
5443be45 | 2286 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 | 2287 | min_t(u64, fs_devices->num_devices, |
61d92c32 CM |
2288 | fs_info->thread_pool_size), |
2289 | &fs_info->generic_worker); | |
61b49440 | 2290 | |
bab39bf9 JB |
2291 | btrfs_init_workers(&fs_info->caching_workers, "cache", |
2292 | 2, &fs_info->generic_worker); | |
2293 | ||
61b49440 CM |
2294 | /* a higher idle thresh on the submit workers makes it much more |
2295 | * likely that bios will be send down in a sane order to the | |
2296 | * devices | |
2297 | */ | |
2298 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 2299 | |
771ed689 | 2300 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 2301 | fs_info->workers.ordered = 1; |
61b49440 | 2302 | |
771ed689 CM |
2303 | fs_info->delalloc_workers.idle_thresh = 2; |
2304 | fs_info->delalloc_workers.ordered = 1; | |
2305 | ||
61d92c32 CM |
2306 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1, |
2307 | &fs_info->generic_worker); | |
5443be45 | 2308 | btrfs_init_workers(&fs_info->endio_workers, "endio", |
61d92c32 CM |
2309 | fs_info->thread_pool_size, |
2310 | &fs_info->generic_worker); | |
d20f7043 | 2311 | btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", |
61d92c32 CM |
2312 | fs_info->thread_pool_size, |
2313 | &fs_info->generic_worker); | |
cad321ad | 2314 | btrfs_init_workers(&fs_info->endio_meta_write_workers, |
61d92c32 CM |
2315 | "endio-meta-write", fs_info->thread_pool_size, |
2316 | &fs_info->generic_worker); | |
5443be45 | 2317 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", |
61d92c32 CM |
2318 | fs_info->thread_pool_size, |
2319 | &fs_info->generic_worker); | |
0cb59c99 JB |
2320 | btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write", |
2321 | 1, &fs_info->generic_worker); | |
16cdcec7 MX |
2322 | btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta", |
2323 | fs_info->thread_pool_size, | |
2324 | &fs_info->generic_worker); | |
90519d66 AJ |
2325 | btrfs_init_workers(&fs_info->readahead_workers, "readahead", |
2326 | fs_info->thread_pool_size, | |
2327 | &fs_info->generic_worker); | |
61b49440 CM |
2328 | |
2329 | /* | |
2330 | * endios are largely parallel and should have a very | |
2331 | * low idle thresh | |
2332 | */ | |
2333 | fs_info->endio_workers.idle_thresh = 4; | |
b51912c9 CM |
2334 | fs_info->endio_meta_workers.idle_thresh = 4; |
2335 | ||
9042846b CM |
2336 | fs_info->endio_write_workers.idle_thresh = 2; |
2337 | fs_info->endio_meta_write_workers.idle_thresh = 2; | |
90519d66 | 2338 | fs_info->readahead_workers.idle_thresh = 2; |
9042846b | 2339 | |
0dc3b84a JB |
2340 | /* |
2341 | * btrfs_start_workers can really only fail because of ENOMEM so just | |
2342 | * return -ENOMEM if any of these fail. | |
2343 | */ | |
2344 | ret = btrfs_start_workers(&fs_info->workers); | |
2345 | ret |= btrfs_start_workers(&fs_info->generic_worker); | |
2346 | ret |= btrfs_start_workers(&fs_info->submit_workers); | |
2347 | ret |= btrfs_start_workers(&fs_info->delalloc_workers); | |
2348 | ret |= btrfs_start_workers(&fs_info->fixup_workers); | |
2349 | ret |= btrfs_start_workers(&fs_info->endio_workers); | |
2350 | ret |= btrfs_start_workers(&fs_info->endio_meta_workers); | |
2351 | ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers); | |
2352 | ret |= btrfs_start_workers(&fs_info->endio_write_workers); | |
2353 | ret |= btrfs_start_workers(&fs_info->endio_freespace_worker); | |
2354 | ret |= btrfs_start_workers(&fs_info->delayed_workers); | |
2355 | ret |= btrfs_start_workers(&fs_info->caching_workers); | |
2356 | ret |= btrfs_start_workers(&fs_info->readahead_workers); | |
2357 | if (ret) { | |
fed425c7 | 2358 | err = -ENOMEM; |
0dc3b84a JB |
2359 | goto fail_sb_buffer; |
2360 | } | |
4543df7e | 2361 | |
4575c9cc | 2362 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
2363 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
2364 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 2365 | |
db94535d CM |
2366 | tree_root->nodesize = nodesize; |
2367 | tree_root->leafsize = leafsize; | |
2368 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 2369 | tree_root->stripesize = stripesize; |
a061fc8d CM |
2370 | |
2371 | sb->s_blocksize = sectorsize; | |
2372 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 2373 | |
39279cc3 CM |
2374 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
2375 | sizeof(disk_super->magic))) { | |
d397712b | 2376 | printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
2377 | goto fail_sb_buffer; |
2378 | } | |
19c00ddc | 2379 | |
8d082fb7 LB |
2380 | if (sectorsize != PAGE_SIZE) { |
2381 | printk(KERN_WARNING "btrfs: Incompatible sector size(%lu) " | |
2382 | "found on %s\n", (unsigned long)sectorsize, sb->s_id); | |
941b2ddf KM |
2383 | goto fail_sb_buffer; |
2384 | } | |
2385 | ||
925baedd | 2386 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 2387 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 2388 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2389 | if (ret) { |
d397712b CM |
2390 | printk(KERN_WARNING "btrfs: failed to read the system " |
2391 | "array on %s\n", sb->s_id); | |
5d4f98a2 | 2392 | goto fail_sb_buffer; |
84eed90f | 2393 | } |
0b86a832 CM |
2394 | |
2395 | blocksize = btrfs_level_size(tree_root, | |
2396 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 2397 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
2398 | |
2399 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2400 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
2401 | ||
2402 | chunk_root->node = read_tree_block(chunk_root, | |
2403 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 2404 | blocksize, generation); |
79787eaa | 2405 | BUG_ON(!chunk_root->node); /* -ENOMEM */ |
83121942 DW |
2406 | if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { |
2407 | printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n", | |
2408 | sb->s_id); | |
af31f5e5 | 2409 | goto fail_tree_roots; |
83121942 | 2410 | } |
5d4f98a2 YZ |
2411 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
2412 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 2413 | |
e17cade2 | 2414 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
d397712b CM |
2415 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), |
2416 | BTRFS_UUID_SIZE); | |
e17cade2 | 2417 | |
0b86a832 | 2418 | ret = btrfs_read_chunk_tree(chunk_root); |
2b82032c | 2419 | if (ret) { |
d397712b CM |
2420 | printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", |
2421 | sb->s_id); | |
af31f5e5 | 2422 | goto fail_tree_roots; |
2b82032c | 2423 | } |
0b86a832 | 2424 | |
dfe25020 CM |
2425 | btrfs_close_extra_devices(fs_devices); |
2426 | ||
a6b0d5c8 CM |
2427 | if (!fs_devices->latest_bdev) { |
2428 | printk(KERN_CRIT "btrfs: failed to read devices on %s\n", | |
2429 | sb->s_id); | |
2430 | goto fail_tree_roots; | |
2431 | } | |
2432 | ||
af31f5e5 | 2433 | retry_root_backup: |
db94535d CM |
2434 | blocksize = btrfs_level_size(tree_root, |
2435 | btrfs_super_root_level(disk_super)); | |
84234f3a | 2436 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 2437 | |
e20d96d6 | 2438 | tree_root->node = read_tree_block(tree_root, |
db94535d | 2439 | btrfs_super_root(disk_super), |
84234f3a | 2440 | blocksize, generation); |
af31f5e5 CM |
2441 | if (!tree_root->node || |
2442 | !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { | |
83121942 DW |
2443 | printk(KERN_WARNING "btrfs: failed to read tree root on %s\n", |
2444 | sb->s_id); | |
af31f5e5 CM |
2445 | |
2446 | goto recovery_tree_root; | |
83121942 | 2447 | } |
af31f5e5 | 2448 | |
5d4f98a2 YZ |
2449 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
2450 | tree_root->commit_root = btrfs_root_node(tree_root); | |
db94535d CM |
2451 | |
2452 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 2453 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 2454 | if (ret) |
af31f5e5 | 2455 | goto recovery_tree_root; |
0b86a832 CM |
2456 | extent_root->track_dirty = 1; |
2457 | ||
2458 | ret = find_and_setup_root(tree_root, fs_info, | |
2459 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
0b86a832 | 2460 | if (ret) |
af31f5e5 | 2461 | goto recovery_tree_root; |
5d4f98a2 | 2462 | dev_root->track_dirty = 1; |
3768f368 | 2463 | |
d20f7043 CM |
2464 | ret = find_and_setup_root(tree_root, fs_info, |
2465 | BTRFS_CSUM_TREE_OBJECTID, csum_root); | |
2466 | if (ret) | |
af31f5e5 | 2467 | goto recovery_tree_root; |
d20f7043 CM |
2468 | csum_root->track_dirty = 1; |
2469 | ||
bcef60f2 AJ |
2470 | ret = find_and_setup_root(tree_root, fs_info, |
2471 | BTRFS_QUOTA_TREE_OBJECTID, quota_root); | |
2472 | if (ret) { | |
2473 | kfree(quota_root); | |
2474 | quota_root = fs_info->quota_root = NULL; | |
2475 | } else { | |
2476 | quota_root->track_dirty = 1; | |
2477 | fs_info->quota_enabled = 1; | |
2478 | fs_info->pending_quota_state = 1; | |
2479 | } | |
2480 | ||
8929ecfa YZ |
2481 | fs_info->generation = generation; |
2482 | fs_info->last_trans_committed = generation; | |
8929ecfa | 2483 | |
68310a5e ID |
2484 | ret = btrfs_recover_balance(fs_info); |
2485 | if (ret) { | |
2486 | printk(KERN_WARNING "btrfs: failed to recover balance\n"); | |
2487 | goto fail_block_groups; | |
2488 | } | |
2489 | ||
733f4fbb SB |
2490 | ret = btrfs_init_dev_stats(fs_info); |
2491 | if (ret) { | |
2492 | printk(KERN_ERR "btrfs: failed to init dev_stats: %d\n", | |
2493 | ret); | |
2494 | goto fail_block_groups; | |
2495 | } | |
2496 | ||
c59021f8 | 2497 | ret = btrfs_init_space_info(fs_info); |
2498 | if (ret) { | |
2499 | printk(KERN_ERR "Failed to initial space info: %d\n", ret); | |
2500 | goto fail_block_groups; | |
2501 | } | |
2502 | ||
1b1d1f66 JB |
2503 | ret = btrfs_read_block_groups(extent_root); |
2504 | if (ret) { | |
2505 | printk(KERN_ERR "Failed to read block groups: %d\n", ret); | |
2506 | goto fail_block_groups; | |
2507 | } | |
9078a3e1 | 2508 | |
a74a4b97 CM |
2509 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
2510 | "btrfs-cleaner"); | |
57506d50 | 2511 | if (IS_ERR(fs_info->cleaner_kthread)) |
1b1d1f66 | 2512 | goto fail_block_groups; |
a74a4b97 CM |
2513 | |
2514 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
2515 | tree_root, | |
2516 | "btrfs-transaction"); | |
57506d50 | 2517 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 2518 | goto fail_cleaner; |
a74a4b97 | 2519 | |
c289811c CM |
2520 | if (!btrfs_test_opt(tree_root, SSD) && |
2521 | !btrfs_test_opt(tree_root, NOSSD) && | |
2522 | !fs_info->fs_devices->rotating) { | |
2523 | printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD " | |
2524 | "mode\n"); | |
2525 | btrfs_set_opt(fs_info->mount_opt, SSD); | |
2526 | } | |
2527 | ||
21adbd5c SB |
2528 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2529 | if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { | |
2530 | ret = btrfsic_mount(tree_root, fs_devices, | |
2531 | btrfs_test_opt(tree_root, | |
2532 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? | |
2533 | 1 : 0, | |
2534 | fs_info->check_integrity_print_mask); | |
2535 | if (ret) | |
2536 | printk(KERN_WARNING "btrfs: failed to initialize" | |
2537 | " integrity check module %s\n", sb->s_id); | |
2538 | } | |
2539 | #endif | |
bcef60f2 AJ |
2540 | ret = btrfs_read_qgroup_config(fs_info); |
2541 | if (ret) | |
2542 | goto fail_trans_kthread; | |
21adbd5c | 2543 | |
acce952b | 2544 | /* do not make disk changes in broken FS */ |
68ce9682 | 2545 | if (btrfs_super_log_root(disk_super) != 0) { |
e02119d5 CM |
2546 | u64 bytenr = btrfs_super_log_root(disk_super); |
2547 | ||
7c2ca468 | 2548 | if (fs_devices->rw_devices == 0) { |
d397712b CM |
2549 | printk(KERN_WARNING "Btrfs log replay required " |
2550 | "on RO media\n"); | |
7c2ca468 | 2551 | err = -EIO; |
bcef60f2 | 2552 | goto fail_qgroup; |
7c2ca468 | 2553 | } |
e02119d5 CM |
2554 | blocksize = |
2555 | btrfs_level_size(tree_root, | |
2556 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 2557 | |
6f07e42e | 2558 | log_tree_root = btrfs_alloc_root(fs_info); |
676e4c86 DC |
2559 | if (!log_tree_root) { |
2560 | err = -ENOMEM; | |
bcef60f2 | 2561 | goto fail_qgroup; |
676e4c86 | 2562 | } |
e02119d5 CM |
2563 | |
2564 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2565 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
2566 | ||
2567 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
2568 | blocksize, |
2569 | generation + 1); | |
79787eaa | 2570 | /* returns with log_tree_root freed on success */ |
e02119d5 | 2571 | ret = btrfs_recover_log_trees(log_tree_root); |
79787eaa JM |
2572 | if (ret) { |
2573 | btrfs_error(tree_root->fs_info, ret, | |
2574 | "Failed to recover log tree"); | |
2575 | free_extent_buffer(log_tree_root->node); | |
2576 | kfree(log_tree_root); | |
2577 | goto fail_trans_kthread; | |
2578 | } | |
e556ce2c YZ |
2579 | |
2580 | if (sb->s_flags & MS_RDONLY) { | |
79787eaa JM |
2581 | ret = btrfs_commit_super(tree_root); |
2582 | if (ret) | |
2583 | goto fail_trans_kthread; | |
e556ce2c | 2584 | } |
e02119d5 | 2585 | } |
1a40e23b | 2586 | |
76dda93c | 2587 | ret = btrfs_find_orphan_roots(tree_root); |
79787eaa JM |
2588 | if (ret) |
2589 | goto fail_trans_kthread; | |
76dda93c | 2590 | |
7c2ca468 | 2591 | if (!(sb->s_flags & MS_RDONLY)) { |
d68fc57b | 2592 | ret = btrfs_cleanup_fs_roots(fs_info); |
44c44af2 ID |
2593 | if (ret) |
2594 | goto fail_trans_kthread; | |
d68fc57b | 2595 | |
5d4f98a2 | 2596 | ret = btrfs_recover_relocation(tree_root); |
d7ce5843 MX |
2597 | if (ret < 0) { |
2598 | printk(KERN_WARNING | |
2599 | "btrfs: failed to recover relocation\n"); | |
2600 | err = -EINVAL; | |
bcef60f2 | 2601 | goto fail_qgroup; |
d7ce5843 | 2602 | } |
7c2ca468 | 2603 | } |
1a40e23b | 2604 | |
3de4586c CM |
2605 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
2606 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2607 | location.offset = (u64)-1; | |
2608 | ||
3de4586c CM |
2609 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
2610 | if (!fs_info->fs_root) | |
bcef60f2 | 2611 | goto fail_qgroup; |
3140c9a3 DC |
2612 | if (IS_ERR(fs_info->fs_root)) { |
2613 | err = PTR_ERR(fs_info->fs_root); | |
bcef60f2 | 2614 | goto fail_qgroup; |
3140c9a3 | 2615 | } |
c289811c | 2616 | |
2b6ba629 ID |
2617 | if (sb->s_flags & MS_RDONLY) |
2618 | return 0; | |
59641015 | 2619 | |
2b6ba629 ID |
2620 | down_read(&fs_info->cleanup_work_sem); |
2621 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
2622 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
e3acc2a6 | 2623 | up_read(&fs_info->cleanup_work_sem); |
2b6ba629 ID |
2624 | close_ctree(tree_root); |
2625 | return ret; | |
2626 | } | |
2627 | up_read(&fs_info->cleanup_work_sem); | |
59641015 | 2628 | |
2b6ba629 ID |
2629 | ret = btrfs_resume_balance_async(fs_info); |
2630 | if (ret) { | |
2631 | printk(KERN_WARNING "btrfs: failed to resume balance\n"); | |
2632 | close_ctree(tree_root); | |
2633 | return ret; | |
e3acc2a6 JB |
2634 | } |
2635 | ||
ad2b2c80 | 2636 | return 0; |
39279cc3 | 2637 | |
bcef60f2 AJ |
2638 | fail_qgroup: |
2639 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
2640 | fail_trans_kthread: |
2641 | kthread_stop(fs_info->transaction_kthread); | |
3f157a2f | 2642 | fail_cleaner: |
a74a4b97 | 2643 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
2644 | |
2645 | /* | |
2646 | * make sure we're done with the btree inode before we stop our | |
2647 | * kthreads | |
2648 | */ | |
2649 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
2650 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
2651 | ||
1b1d1f66 JB |
2652 | fail_block_groups: |
2653 | btrfs_free_block_groups(fs_info); | |
af31f5e5 CM |
2654 | |
2655 | fail_tree_roots: | |
2656 | free_root_pointers(fs_info, 1); | |
2657 | ||
39279cc3 | 2658 | fail_sb_buffer: |
61d92c32 | 2659 | btrfs_stop_workers(&fs_info->generic_worker); |
306c8b68 | 2660 | btrfs_stop_workers(&fs_info->readahead_workers); |
247e743c | 2661 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2662 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2663 | btrfs_stop_workers(&fs_info->workers); |
2664 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 2665 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 2666 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 2667 | btrfs_stop_workers(&fs_info->endio_write_workers); |
0cb59c99 | 2668 | btrfs_stop_workers(&fs_info->endio_freespace_worker); |
1cc127b5 | 2669 | btrfs_stop_workers(&fs_info->submit_workers); |
16cdcec7 | 2670 | btrfs_stop_workers(&fs_info->delayed_workers); |
bab39bf9 | 2671 | btrfs_stop_workers(&fs_info->caching_workers); |
16cdcec7 | 2672 | fail_alloc: |
4543df7e | 2673 | fail_iput: |
586e46e2 ID |
2674 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
2675 | ||
7c2ca468 | 2676 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
4543df7e | 2677 | iput(fs_info->btree_inode); |
ad081f14 | 2678 | fail_bdi: |
7e662854 | 2679 | bdi_destroy(&fs_info->bdi); |
76dda93c YZ |
2680 | fail_srcu: |
2681 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 2682 | fail: |
586e46e2 | 2683 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 2684 | return err; |
af31f5e5 CM |
2685 | |
2686 | recovery_tree_root: | |
af31f5e5 CM |
2687 | if (!btrfs_test_opt(tree_root, RECOVERY)) |
2688 | goto fail_tree_roots; | |
2689 | ||
2690 | free_root_pointers(fs_info, 0); | |
2691 | ||
2692 | /* don't use the log in recovery mode, it won't be valid */ | |
2693 | btrfs_set_super_log_root(disk_super, 0); | |
2694 | ||
2695 | /* we can't trust the free space cache either */ | |
2696 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2697 | ||
2698 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
2699 | &num_backups_tried, &backup_index); | |
2700 | if (ret == -1) | |
2701 | goto fail_block_groups; | |
2702 | goto retry_root_backup; | |
eb60ceac CM |
2703 | } |
2704 | ||
f2984462 CM |
2705 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
2706 | { | |
f2984462 CM |
2707 | if (uptodate) { |
2708 | set_buffer_uptodate(bh); | |
2709 | } else { | |
442a4f63 SB |
2710 | struct btrfs_device *device = (struct btrfs_device *) |
2711 | bh->b_private; | |
2712 | ||
606686ee JB |
2713 | printk_ratelimited_in_rcu(KERN_WARNING "lost page write due to " |
2714 | "I/O error on %s\n", | |
2715 | rcu_str_deref(device->name)); | |
1259ab75 CM |
2716 | /* note, we dont' set_buffer_write_io_error because we have |
2717 | * our own ways of dealing with the IO errors | |
2718 | */ | |
f2984462 | 2719 | clear_buffer_uptodate(bh); |
442a4f63 | 2720 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); |
f2984462 CM |
2721 | } |
2722 | unlock_buffer(bh); | |
2723 | put_bh(bh); | |
2724 | } | |
2725 | ||
a512bbf8 YZ |
2726 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
2727 | { | |
2728 | struct buffer_head *bh; | |
2729 | struct buffer_head *latest = NULL; | |
2730 | struct btrfs_super_block *super; | |
2731 | int i; | |
2732 | u64 transid = 0; | |
2733 | u64 bytenr; | |
2734 | ||
2735 | /* we would like to check all the supers, but that would make | |
2736 | * a btrfs mount succeed after a mkfs from a different FS. | |
2737 | * So, we need to add a special mount option to scan for | |
2738 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
2739 | */ | |
2740 | for (i = 0; i < 1; i++) { | |
2741 | bytenr = btrfs_sb_offset(i); | |
2742 | if (bytenr + 4096 >= i_size_read(bdev->bd_inode)) | |
2743 | break; | |
2744 | bh = __bread(bdev, bytenr / 4096, 4096); | |
2745 | if (!bh) | |
2746 | continue; | |
2747 | ||
2748 | super = (struct btrfs_super_block *)bh->b_data; | |
2749 | if (btrfs_super_bytenr(super) != bytenr || | |
2750 | strncmp((char *)(&super->magic), BTRFS_MAGIC, | |
2751 | sizeof(super->magic))) { | |
2752 | brelse(bh); | |
2753 | continue; | |
2754 | } | |
2755 | ||
2756 | if (!latest || btrfs_super_generation(super) > transid) { | |
2757 | brelse(latest); | |
2758 | latest = bh; | |
2759 | transid = btrfs_super_generation(super); | |
2760 | } else { | |
2761 | brelse(bh); | |
2762 | } | |
2763 | } | |
2764 | return latest; | |
2765 | } | |
2766 | ||
4eedeb75 HH |
2767 | /* |
2768 | * this should be called twice, once with wait == 0 and | |
2769 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
2770 | * we write are pinned. | |
2771 | * | |
2772 | * They are released when wait == 1 is done. | |
2773 | * max_mirrors must be the same for both runs, and it indicates how | |
2774 | * many supers on this one device should be written. | |
2775 | * | |
2776 | * max_mirrors == 0 means to write them all. | |
2777 | */ | |
a512bbf8 YZ |
2778 | static int write_dev_supers(struct btrfs_device *device, |
2779 | struct btrfs_super_block *sb, | |
2780 | int do_barriers, int wait, int max_mirrors) | |
2781 | { | |
2782 | struct buffer_head *bh; | |
2783 | int i; | |
2784 | int ret; | |
2785 | int errors = 0; | |
2786 | u32 crc; | |
2787 | u64 bytenr; | |
a512bbf8 YZ |
2788 | |
2789 | if (max_mirrors == 0) | |
2790 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
2791 | ||
a512bbf8 YZ |
2792 | for (i = 0; i < max_mirrors; i++) { |
2793 | bytenr = btrfs_sb_offset(i); | |
2794 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) | |
2795 | break; | |
2796 | ||
2797 | if (wait) { | |
2798 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
2799 | BTRFS_SUPER_INFO_SIZE); | |
2800 | BUG_ON(!bh); | |
a512bbf8 | 2801 | wait_on_buffer(bh); |
4eedeb75 HH |
2802 | if (!buffer_uptodate(bh)) |
2803 | errors++; | |
2804 | ||
2805 | /* drop our reference */ | |
2806 | brelse(bh); | |
2807 | ||
2808 | /* drop the reference from the wait == 0 run */ | |
2809 | brelse(bh); | |
2810 | continue; | |
a512bbf8 YZ |
2811 | } else { |
2812 | btrfs_set_super_bytenr(sb, bytenr); | |
2813 | ||
2814 | crc = ~(u32)0; | |
2815 | crc = btrfs_csum_data(NULL, (char *)sb + | |
2816 | BTRFS_CSUM_SIZE, crc, | |
2817 | BTRFS_SUPER_INFO_SIZE - | |
2818 | BTRFS_CSUM_SIZE); | |
2819 | btrfs_csum_final(crc, sb->csum); | |
2820 | ||
4eedeb75 HH |
2821 | /* |
2822 | * one reference for us, and we leave it for the | |
2823 | * caller | |
2824 | */ | |
a512bbf8 YZ |
2825 | bh = __getblk(device->bdev, bytenr / 4096, |
2826 | BTRFS_SUPER_INFO_SIZE); | |
2827 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); | |
2828 | ||
4eedeb75 | 2829 | /* one reference for submit_bh */ |
a512bbf8 | 2830 | get_bh(bh); |
4eedeb75 HH |
2831 | |
2832 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
2833 | lock_buffer(bh); |
2834 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
442a4f63 | 2835 | bh->b_private = device; |
a512bbf8 YZ |
2836 | } |
2837 | ||
387125fc CM |
2838 | /* |
2839 | * we fua the first super. The others we allow | |
2840 | * to go down lazy. | |
2841 | */ | |
21adbd5c | 2842 | ret = btrfsic_submit_bh(WRITE_FUA, bh); |
4eedeb75 | 2843 | if (ret) |
a512bbf8 | 2844 | errors++; |
a512bbf8 YZ |
2845 | } |
2846 | return errors < i ? 0 : -1; | |
2847 | } | |
2848 | ||
387125fc CM |
2849 | /* |
2850 | * endio for the write_dev_flush, this will wake anyone waiting | |
2851 | * for the barrier when it is done | |
2852 | */ | |
2853 | static void btrfs_end_empty_barrier(struct bio *bio, int err) | |
2854 | { | |
2855 | if (err) { | |
2856 | if (err == -EOPNOTSUPP) | |
2857 | set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); | |
2858 | clear_bit(BIO_UPTODATE, &bio->bi_flags); | |
2859 | } | |
2860 | if (bio->bi_private) | |
2861 | complete(bio->bi_private); | |
2862 | bio_put(bio); | |
2863 | } | |
2864 | ||
2865 | /* | |
2866 | * trigger flushes for one the devices. If you pass wait == 0, the flushes are | |
2867 | * sent down. With wait == 1, it waits for the previous flush. | |
2868 | * | |
2869 | * any device where the flush fails with eopnotsupp are flagged as not-barrier | |
2870 | * capable | |
2871 | */ | |
2872 | static int write_dev_flush(struct btrfs_device *device, int wait) | |
2873 | { | |
2874 | struct bio *bio; | |
2875 | int ret = 0; | |
2876 | ||
2877 | if (device->nobarriers) | |
2878 | return 0; | |
2879 | ||
2880 | if (wait) { | |
2881 | bio = device->flush_bio; | |
2882 | if (!bio) | |
2883 | return 0; | |
2884 | ||
2885 | wait_for_completion(&device->flush_wait); | |
2886 | ||
2887 | if (bio_flagged(bio, BIO_EOPNOTSUPP)) { | |
606686ee JB |
2888 | printk_in_rcu("btrfs: disabling barriers on dev %s\n", |
2889 | rcu_str_deref(device->name)); | |
387125fc CM |
2890 | device->nobarriers = 1; |
2891 | } | |
2892 | if (!bio_flagged(bio, BIO_UPTODATE)) { | |
2893 | ret = -EIO; | |
442a4f63 SB |
2894 | if (!bio_flagged(bio, BIO_EOPNOTSUPP)) |
2895 | btrfs_dev_stat_inc_and_print(device, | |
2896 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
387125fc CM |
2897 | } |
2898 | ||
2899 | /* drop the reference from the wait == 0 run */ | |
2900 | bio_put(bio); | |
2901 | device->flush_bio = NULL; | |
2902 | ||
2903 | return ret; | |
2904 | } | |
2905 | ||
2906 | /* | |
2907 | * one reference for us, and we leave it for the | |
2908 | * caller | |
2909 | */ | |
9c017abc | 2910 | device->flush_bio = NULL; |
387125fc CM |
2911 | bio = bio_alloc(GFP_NOFS, 0); |
2912 | if (!bio) | |
2913 | return -ENOMEM; | |
2914 | ||
2915 | bio->bi_end_io = btrfs_end_empty_barrier; | |
2916 | bio->bi_bdev = device->bdev; | |
2917 | init_completion(&device->flush_wait); | |
2918 | bio->bi_private = &device->flush_wait; | |
2919 | device->flush_bio = bio; | |
2920 | ||
2921 | bio_get(bio); | |
21adbd5c | 2922 | btrfsic_submit_bio(WRITE_FLUSH, bio); |
387125fc CM |
2923 | |
2924 | return 0; | |
2925 | } | |
2926 | ||
2927 | /* | |
2928 | * send an empty flush down to each device in parallel, | |
2929 | * then wait for them | |
2930 | */ | |
2931 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
2932 | { | |
2933 | struct list_head *head; | |
2934 | struct btrfs_device *dev; | |
2935 | int errors = 0; | |
2936 | int ret; | |
2937 | ||
2938 | /* send down all the barriers */ | |
2939 | head = &info->fs_devices->devices; | |
2940 | list_for_each_entry_rcu(dev, head, dev_list) { | |
2941 | if (!dev->bdev) { | |
2942 | errors++; | |
2943 | continue; | |
2944 | } | |
2945 | if (!dev->in_fs_metadata || !dev->writeable) | |
2946 | continue; | |
2947 | ||
2948 | ret = write_dev_flush(dev, 0); | |
2949 | if (ret) | |
2950 | errors++; | |
2951 | } | |
2952 | ||
2953 | /* wait for all the barriers */ | |
2954 | list_for_each_entry_rcu(dev, head, dev_list) { | |
2955 | if (!dev->bdev) { | |
2956 | errors++; | |
2957 | continue; | |
2958 | } | |
2959 | if (!dev->in_fs_metadata || !dev->writeable) | |
2960 | continue; | |
2961 | ||
2962 | ret = write_dev_flush(dev, 1); | |
2963 | if (ret) | |
2964 | errors++; | |
2965 | } | |
2966 | if (errors) | |
2967 | return -EIO; | |
2968 | return 0; | |
2969 | } | |
2970 | ||
a512bbf8 | 2971 | int write_all_supers(struct btrfs_root *root, int max_mirrors) |
f2984462 | 2972 | { |
e5e9a520 | 2973 | struct list_head *head; |
f2984462 | 2974 | struct btrfs_device *dev; |
a061fc8d | 2975 | struct btrfs_super_block *sb; |
f2984462 | 2976 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
2977 | int ret; |
2978 | int do_barriers; | |
a236aed1 CM |
2979 | int max_errors; |
2980 | int total_errors = 0; | |
a061fc8d | 2981 | u64 flags; |
f2984462 | 2982 | |
6c41761f | 2983 | max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; |
f2984462 | 2984 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
af31f5e5 | 2985 | backup_super_roots(root->fs_info); |
f2984462 | 2986 | |
6c41761f | 2987 | sb = root->fs_info->super_for_commit; |
a061fc8d | 2988 | dev_item = &sb->dev_item; |
e5e9a520 | 2989 | |
174ba509 | 2990 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); |
e5e9a520 | 2991 | head = &root->fs_info->fs_devices->devices; |
387125fc CM |
2992 | |
2993 | if (do_barriers) | |
2994 | barrier_all_devices(root->fs_info); | |
2995 | ||
1f78160c | 2996 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
2997 | if (!dev->bdev) { |
2998 | total_errors++; | |
2999 | continue; | |
3000 | } | |
2b82032c | 3001 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3002 | continue; |
3003 | ||
2b82032c | 3004 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
3005 | btrfs_set_stack_device_type(dev_item, dev->type); |
3006 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
3007 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
3008 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
3009 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
3010 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
3011 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
3012 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 3013 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 3014 | |
a061fc8d CM |
3015 | flags = btrfs_super_flags(sb); |
3016 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
3017 | ||
a512bbf8 | 3018 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
3019 | if (ret) |
3020 | total_errors++; | |
f2984462 | 3021 | } |
a236aed1 | 3022 | if (total_errors > max_errors) { |
d397712b CM |
3023 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
3024 | total_errors); | |
79787eaa JM |
3025 | |
3026 | /* This shouldn't happen. FUA is masked off if unsupported */ | |
a236aed1 CM |
3027 | BUG(); |
3028 | } | |
f2984462 | 3029 | |
a512bbf8 | 3030 | total_errors = 0; |
1f78160c | 3031 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3032 | if (!dev->bdev) |
3033 | continue; | |
2b82032c | 3034 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3035 | continue; |
3036 | ||
a512bbf8 YZ |
3037 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
3038 | if (ret) | |
3039 | total_errors++; | |
f2984462 | 3040 | } |
174ba509 | 3041 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 3042 | if (total_errors > max_errors) { |
79787eaa JM |
3043 | btrfs_error(root->fs_info, -EIO, |
3044 | "%d errors while writing supers", total_errors); | |
3045 | return -EIO; | |
a236aed1 | 3046 | } |
f2984462 CM |
3047 | return 0; |
3048 | } | |
3049 | ||
a512bbf8 YZ |
3050 | int write_ctree_super(struct btrfs_trans_handle *trans, |
3051 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 3052 | { |
e66f709b | 3053 | int ret; |
5f39d397 | 3054 | |
a512bbf8 | 3055 | ret = write_all_supers(root, max_mirrors); |
5f39d397 | 3056 | return ret; |
cfaa7295 CM |
3057 | } |
3058 | ||
143bede5 | 3059 | void btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f | 3060 | { |
4df27c4d | 3061 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
3062 | radix_tree_delete(&fs_info->fs_roots_radix, |
3063 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 3064 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
3065 | |
3066 | if (btrfs_root_refs(&root->root_item) == 0) | |
3067 | synchronize_srcu(&fs_info->subvol_srcu); | |
3068 | ||
581bb050 LZ |
3069 | __btrfs_remove_free_space_cache(root->free_ino_pinned); |
3070 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d | 3071 | free_fs_root(root); |
4df27c4d YZ |
3072 | } |
3073 | ||
3074 | static void free_fs_root(struct btrfs_root *root) | |
3075 | { | |
82d5902d | 3076 | iput(root->cache_inode); |
4df27c4d | 3077 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
0ee5dc67 AV |
3078 | if (root->anon_dev) |
3079 | free_anon_bdev(root->anon_dev); | |
4df27c4d YZ |
3080 | free_extent_buffer(root->node); |
3081 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
3082 | kfree(root->free_ino_ctl); |
3083 | kfree(root->free_ino_pinned); | |
d397712b | 3084 | kfree(root->name); |
2619ba1f | 3085 | kfree(root); |
2619ba1f CM |
3086 | } |
3087 | ||
143bede5 | 3088 | static void del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
3089 | { |
3090 | int ret; | |
3091 | struct btrfs_root *gang[8]; | |
3092 | int i; | |
3093 | ||
76dda93c YZ |
3094 | while (!list_empty(&fs_info->dead_roots)) { |
3095 | gang[0] = list_entry(fs_info->dead_roots.next, | |
3096 | struct btrfs_root, root_list); | |
3097 | list_del(&gang[0]->root_list); | |
3098 | ||
3099 | if (gang[0]->in_radix) { | |
3100 | btrfs_free_fs_root(fs_info, gang[0]); | |
3101 | } else { | |
3102 | free_extent_buffer(gang[0]->node); | |
3103 | free_extent_buffer(gang[0]->commit_root); | |
3104 | kfree(gang[0]); | |
3105 | } | |
3106 | } | |
3107 | ||
d397712b | 3108 | while (1) { |
0f7d52f4 CM |
3109 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
3110 | (void **)gang, 0, | |
3111 | ARRAY_SIZE(gang)); | |
3112 | if (!ret) | |
3113 | break; | |
2619ba1f | 3114 | for (i = 0; i < ret; i++) |
5eda7b5e | 3115 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 | 3116 | } |
0f7d52f4 | 3117 | } |
b4100d64 | 3118 | |
c146afad | 3119 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 3120 | { |
c146afad YZ |
3121 | u64 root_objectid = 0; |
3122 | struct btrfs_root *gang[8]; | |
3123 | int i; | |
3768f368 | 3124 | int ret; |
e089f05c | 3125 | |
c146afad YZ |
3126 | while (1) { |
3127 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
3128 | (void **)gang, root_objectid, | |
3129 | ARRAY_SIZE(gang)); | |
3130 | if (!ret) | |
3131 | break; | |
5d4f98a2 YZ |
3132 | |
3133 | root_objectid = gang[ret - 1]->root_key.objectid + 1; | |
c146afad | 3134 | for (i = 0; i < ret; i++) { |
66b4ffd1 JB |
3135 | int err; |
3136 | ||
c146afad | 3137 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
3138 | err = btrfs_orphan_cleanup(gang[i]); |
3139 | if (err) | |
3140 | return err; | |
c146afad YZ |
3141 | } |
3142 | root_objectid++; | |
3143 | } | |
3144 | return 0; | |
3145 | } | |
a2135011 | 3146 | |
c146afad YZ |
3147 | int btrfs_commit_super(struct btrfs_root *root) |
3148 | { | |
3149 | struct btrfs_trans_handle *trans; | |
3150 | int ret; | |
a74a4b97 | 3151 | |
c146afad | 3152 | mutex_lock(&root->fs_info->cleaner_mutex); |
24bbcf04 | 3153 | btrfs_run_delayed_iputs(root); |
a74a4b97 | 3154 | btrfs_clean_old_snapshots(root); |
c146afad | 3155 | mutex_unlock(&root->fs_info->cleaner_mutex); |
c71bf099 YZ |
3156 | |
3157 | /* wait until ongoing cleanup work done */ | |
3158 | down_write(&root->fs_info->cleanup_work_sem); | |
3159 | up_write(&root->fs_info->cleanup_work_sem); | |
3160 | ||
7a7eaa40 | 3161 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3162 | if (IS_ERR(trans)) |
3163 | return PTR_ERR(trans); | |
54aa1f4d | 3164 | ret = btrfs_commit_transaction(trans, root); |
79787eaa JM |
3165 | if (ret) |
3166 | return ret; | |
c146afad | 3167 | /* run commit again to drop the original snapshot */ |
7a7eaa40 | 3168 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3169 | if (IS_ERR(trans)) |
3170 | return PTR_ERR(trans); | |
79787eaa JM |
3171 | ret = btrfs_commit_transaction(trans, root); |
3172 | if (ret) | |
3173 | return ret; | |
79154b1b | 3174 | ret = btrfs_write_and_wait_transaction(NULL, root); |
79787eaa JM |
3175 | if (ret) { |
3176 | btrfs_error(root->fs_info, ret, | |
3177 | "Failed to sync btree inode to disk."); | |
3178 | return ret; | |
3179 | } | |
d6bfde87 | 3180 | |
a512bbf8 | 3181 | ret = write_ctree_super(NULL, root, 0); |
c146afad YZ |
3182 | return ret; |
3183 | } | |
3184 | ||
3185 | int close_ctree(struct btrfs_root *root) | |
3186 | { | |
3187 | struct btrfs_fs_info *fs_info = root->fs_info; | |
3188 | int ret; | |
3189 | ||
3190 | fs_info->closing = 1; | |
3191 | smp_mb(); | |
3192 | ||
837d5b6e ID |
3193 | /* pause restriper - we want to resume on mount */ |
3194 | btrfs_pause_balance(root->fs_info); | |
3195 | ||
a2de733c | 3196 | btrfs_scrub_cancel(root); |
4cb5300b CM |
3197 | |
3198 | /* wait for any defraggers to finish */ | |
3199 | wait_event(fs_info->transaction_wait, | |
3200 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3201 | ||
3202 | /* clear out the rbtree of defraggable inodes */ | |
e3029d9f | 3203 | btrfs_run_defrag_inodes(fs_info); |
4cb5300b | 3204 | |
c146afad | 3205 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
acce952b | 3206 | ret = btrfs_commit_super(root); |
3207 | if (ret) | |
3208 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
3209 | } | |
3210 | ||
68ce9682 SB |
3211 | if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) |
3212 | btrfs_error_commit_super(root); | |
0f7d52f4 | 3213 | |
300e4f8a JB |
3214 | btrfs_put_block_group_cache(fs_info); |
3215 | ||
e3029d9f AV |
3216 | kthread_stop(fs_info->transaction_kthread); |
3217 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 3218 | |
f25784b3 YZ |
3219 | fs_info->closing = 2; |
3220 | smp_mb(); | |
3221 | ||
bcef60f2 AJ |
3222 | btrfs_free_qgroup_config(root->fs_info); |
3223 | ||
b0c68f8b | 3224 | if (fs_info->delalloc_bytes) { |
d397712b | 3225 | printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n", |
21380931 | 3226 | (unsigned long long)fs_info->delalloc_bytes); |
b0c68f8b | 3227 | } |
bcc63abb | 3228 | |
5d4f98a2 YZ |
3229 | free_extent_buffer(fs_info->extent_root->node); |
3230 | free_extent_buffer(fs_info->extent_root->commit_root); | |
3231 | free_extent_buffer(fs_info->tree_root->node); | |
3232 | free_extent_buffer(fs_info->tree_root->commit_root); | |
e3029d9f AV |
3233 | free_extent_buffer(fs_info->chunk_root->node); |
3234 | free_extent_buffer(fs_info->chunk_root->commit_root); | |
3235 | free_extent_buffer(fs_info->dev_root->node); | |
3236 | free_extent_buffer(fs_info->dev_root->commit_root); | |
3237 | free_extent_buffer(fs_info->csum_root->node); | |
3238 | free_extent_buffer(fs_info->csum_root->commit_root); | |
bcef60f2 AJ |
3239 | if (fs_info->quota_root) { |
3240 | free_extent_buffer(fs_info->quota_root->node); | |
3241 | free_extent_buffer(fs_info->quota_root->commit_root); | |
3242 | } | |
d20f7043 | 3243 | |
e3029d9f | 3244 | btrfs_free_block_groups(fs_info); |
d10c5f31 | 3245 | |
c146afad | 3246 | del_fs_roots(fs_info); |
d10c5f31 | 3247 | |
c146afad | 3248 | iput(fs_info->btree_inode); |
9ad6b7bc | 3249 | |
61d92c32 | 3250 | btrfs_stop_workers(&fs_info->generic_worker); |
247e743c | 3251 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 3252 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
3253 | btrfs_stop_workers(&fs_info->workers); |
3254 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 3255 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 3256 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 3257 | btrfs_stop_workers(&fs_info->endio_write_workers); |
0cb59c99 | 3258 | btrfs_stop_workers(&fs_info->endio_freespace_worker); |
1cc127b5 | 3259 | btrfs_stop_workers(&fs_info->submit_workers); |
16cdcec7 | 3260 | btrfs_stop_workers(&fs_info->delayed_workers); |
bab39bf9 | 3261 | btrfs_stop_workers(&fs_info->caching_workers); |
90519d66 | 3262 | btrfs_stop_workers(&fs_info->readahead_workers); |
d6bfde87 | 3263 | |
21adbd5c SB |
3264 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3265 | if (btrfs_test_opt(root, CHECK_INTEGRITY)) | |
3266 | btrfsic_unmount(root, fs_info->fs_devices); | |
3267 | #endif | |
3268 | ||
dfe25020 | 3269 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 3270 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 3271 | |
04160088 | 3272 | bdi_destroy(&fs_info->bdi); |
76dda93c | 3273 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 3274 | |
eb60ceac CM |
3275 | return 0; |
3276 | } | |
3277 | ||
b9fab919 CM |
3278 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
3279 | int atomic) | |
5f39d397 | 3280 | { |
1259ab75 | 3281 | int ret; |
727011e0 | 3282 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 3283 | |
0b32f4bb | 3284 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
3285 | if (!ret) |
3286 | return ret; | |
3287 | ||
3288 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
3289 | parent_transid, atomic); |
3290 | if (ret == -EAGAIN) | |
3291 | return ret; | |
1259ab75 | 3292 | return !ret; |
5f39d397 CM |
3293 | } |
3294 | ||
3295 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 3296 | { |
0b32f4bb | 3297 | return set_extent_buffer_uptodate(buf); |
5f39d397 | 3298 | } |
6702ed49 | 3299 | |
5f39d397 CM |
3300 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
3301 | { | |
727011e0 | 3302 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
5f39d397 | 3303 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 3304 | int was_dirty; |
b4ce94de | 3305 | |
b9447ef8 | 3306 | btrfs_assert_tree_locked(buf); |
ccd467d6 | 3307 | if (transid != root->fs_info->generation) { |
d397712b CM |
3308 | printk(KERN_CRIT "btrfs transid mismatch buffer %llu, " |
3309 | "found %llu running %llu\n", | |
db94535d | 3310 | (unsigned long long)buf->start, |
d397712b CM |
3311 | (unsigned long long)transid, |
3312 | (unsigned long long)root->fs_info->generation); | |
ccd467d6 CM |
3313 | WARN_ON(1); |
3314 | } | |
0b32f4bb | 3315 | was_dirty = set_extent_buffer_dirty(buf); |
b9473439 CM |
3316 | if (!was_dirty) { |
3317 | spin_lock(&root->fs_info->delalloc_lock); | |
3318 | root->fs_info->dirty_metadata_bytes += buf->len; | |
3319 | spin_unlock(&root->fs_info->delalloc_lock); | |
3320 | } | |
eb60ceac CM |
3321 | } |
3322 | ||
d3c2fdcf | 3323 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
16cdcec7 MX |
3324 | { |
3325 | /* | |
3326 | * looks as though older kernels can get into trouble with | |
3327 | * this code, they end up stuck in balance_dirty_pages forever | |
3328 | */ | |
3329 | u64 num_dirty; | |
3330 | unsigned long thresh = 32 * 1024 * 1024; | |
3331 | ||
3332 | if (current->flags & PF_MEMALLOC) | |
3333 | return; | |
3334 | ||
3335 | btrfs_balance_delayed_items(root); | |
3336 | ||
3337 | num_dirty = root->fs_info->dirty_metadata_bytes; | |
3338 | ||
3339 | if (num_dirty > thresh) { | |
3340 | balance_dirty_pages_ratelimited_nr( | |
3341 | root->fs_info->btree_inode->i_mapping, 1); | |
3342 | } | |
3343 | return; | |
3344 | } | |
3345 | ||
3346 | void __btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) | |
35b7e476 | 3347 | { |
188de649 CM |
3348 | /* |
3349 | * looks as though older kernels can get into trouble with | |
3350 | * this code, they end up stuck in balance_dirty_pages forever | |
3351 | */ | |
d6bfde87 | 3352 | u64 num_dirty; |
771ed689 | 3353 | unsigned long thresh = 32 * 1024 * 1024; |
d6bfde87 | 3354 | |
6933c02e | 3355 | if (current->flags & PF_MEMALLOC) |
d6bfde87 CM |
3356 | return; |
3357 | ||
585ad2c3 CM |
3358 | num_dirty = root->fs_info->dirty_metadata_bytes; |
3359 | ||
d6bfde87 CM |
3360 | if (num_dirty > thresh) { |
3361 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 3362 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 3363 | } |
188de649 | 3364 | return; |
35b7e476 | 3365 | } |
6b80053d | 3366 | |
ca7a79ad | 3367 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 3368 | { |
727011e0 | 3369 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
0b32f4bb | 3370 | return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
6b80053d | 3371 | } |
0da5468f | 3372 | |
fcd1f065 | 3373 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 3374 | int read_only) |
3375 | { | |
fcd1f065 DS |
3376 | if (btrfs_super_csum_type(fs_info->super_copy) >= ARRAY_SIZE(btrfs_csum_sizes)) { |
3377 | printk(KERN_ERR "btrfs: unsupported checksum algorithm\n"); | |
3378 | return -EINVAL; | |
3379 | } | |
3380 | ||
acce952b | 3381 | if (read_only) |
fcd1f065 | 3382 | return 0; |
acce952b | 3383 | |
fcd1f065 | 3384 | return 0; |
acce952b | 3385 | } |
3386 | ||
68ce9682 | 3387 | void btrfs_error_commit_super(struct btrfs_root *root) |
acce952b | 3388 | { |
acce952b | 3389 | mutex_lock(&root->fs_info->cleaner_mutex); |
3390 | btrfs_run_delayed_iputs(root); | |
3391 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
3392 | ||
3393 | down_write(&root->fs_info->cleanup_work_sem); | |
3394 | up_write(&root->fs_info->cleanup_work_sem); | |
3395 | ||
3396 | /* cleanup FS via transaction */ | |
3397 | btrfs_cleanup_transaction(root); | |
acce952b | 3398 | } |
3399 | ||
143bede5 | 3400 | static void btrfs_destroy_ordered_operations(struct btrfs_root *root) |
acce952b | 3401 | { |
3402 | struct btrfs_inode *btrfs_inode; | |
3403 | struct list_head splice; | |
3404 | ||
3405 | INIT_LIST_HEAD(&splice); | |
3406 | ||
3407 | mutex_lock(&root->fs_info->ordered_operations_mutex); | |
3408 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3409 | ||
3410 | list_splice_init(&root->fs_info->ordered_operations, &splice); | |
3411 | while (!list_empty(&splice)) { | |
3412 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3413 | ordered_operations); | |
3414 | ||
3415 | list_del_init(&btrfs_inode->ordered_operations); | |
3416 | ||
3417 | btrfs_invalidate_inodes(btrfs_inode->root); | |
3418 | } | |
3419 | ||
3420 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3421 | mutex_unlock(&root->fs_info->ordered_operations_mutex); | |
acce952b | 3422 | } |
3423 | ||
143bede5 | 3424 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 3425 | { |
3426 | struct list_head splice; | |
3427 | struct btrfs_ordered_extent *ordered; | |
3428 | struct inode *inode; | |
3429 | ||
3430 | INIT_LIST_HEAD(&splice); | |
3431 | ||
3432 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3433 | ||
3434 | list_splice_init(&root->fs_info->ordered_extents, &splice); | |
3435 | while (!list_empty(&splice)) { | |
3436 | ordered = list_entry(splice.next, struct btrfs_ordered_extent, | |
3437 | root_extent_list); | |
3438 | ||
3439 | list_del_init(&ordered->root_extent_list); | |
3440 | atomic_inc(&ordered->refs); | |
3441 | ||
3442 | /* the inode may be getting freed (in sys_unlink path). */ | |
3443 | inode = igrab(ordered->inode); | |
3444 | ||
3445 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3446 | if (inode) | |
3447 | iput(inode); | |
3448 | ||
3449 | atomic_set(&ordered->refs, 1); | |
3450 | btrfs_put_ordered_extent(ordered); | |
3451 | ||
3452 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3453 | } | |
3454 | ||
3455 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
acce952b | 3456 | } |
3457 | ||
79787eaa JM |
3458 | int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
3459 | struct btrfs_root *root) | |
acce952b | 3460 | { |
3461 | struct rb_node *node; | |
3462 | struct btrfs_delayed_ref_root *delayed_refs; | |
3463 | struct btrfs_delayed_ref_node *ref; | |
3464 | int ret = 0; | |
3465 | ||
3466 | delayed_refs = &trans->delayed_refs; | |
3467 | ||
3468 | spin_lock(&delayed_refs->lock); | |
3469 | if (delayed_refs->num_entries == 0) { | |
cfece4db | 3470 | spin_unlock(&delayed_refs->lock); |
acce952b | 3471 | printk(KERN_INFO "delayed_refs has NO entry\n"); |
3472 | return ret; | |
3473 | } | |
3474 | ||
b939d1ab | 3475 | while ((node = rb_first(&delayed_refs->root)) != NULL) { |
acce952b | 3476 | ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); |
acce952b | 3477 | |
3478 | atomic_set(&ref->refs, 1); | |
3479 | if (btrfs_delayed_ref_is_head(ref)) { | |
3480 | struct btrfs_delayed_ref_head *head; | |
3481 | ||
3482 | head = btrfs_delayed_node_to_head(ref); | |
b939d1ab JB |
3483 | if (!mutex_trylock(&head->mutex)) { |
3484 | atomic_inc(&ref->refs); | |
3485 | spin_unlock(&delayed_refs->lock); | |
3486 | ||
3487 | /* Need to wait for the delayed ref to run */ | |
3488 | mutex_lock(&head->mutex); | |
3489 | mutex_unlock(&head->mutex); | |
3490 | btrfs_put_delayed_ref(ref); | |
3491 | ||
e18fca73 | 3492 | spin_lock(&delayed_refs->lock); |
b939d1ab JB |
3493 | continue; |
3494 | } | |
3495 | ||
acce952b | 3496 | kfree(head->extent_op); |
3497 | delayed_refs->num_heads--; | |
3498 | if (list_empty(&head->cluster)) | |
3499 | delayed_refs->num_heads_ready--; | |
3500 | list_del_init(&head->cluster); | |
acce952b | 3501 | } |
b939d1ab JB |
3502 | ref->in_tree = 0; |
3503 | rb_erase(&ref->rb_node, &delayed_refs->root); | |
3504 | delayed_refs->num_entries--; | |
3505 | ||
acce952b | 3506 | spin_unlock(&delayed_refs->lock); |
3507 | btrfs_put_delayed_ref(ref); | |
3508 | ||
3509 | cond_resched(); | |
3510 | spin_lock(&delayed_refs->lock); | |
3511 | } | |
3512 | ||
3513 | spin_unlock(&delayed_refs->lock); | |
3514 | ||
3515 | return ret; | |
3516 | } | |
3517 | ||
143bede5 | 3518 | static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t) |
acce952b | 3519 | { |
3520 | struct btrfs_pending_snapshot *snapshot; | |
3521 | struct list_head splice; | |
3522 | ||
3523 | INIT_LIST_HEAD(&splice); | |
3524 | ||
3525 | list_splice_init(&t->pending_snapshots, &splice); | |
3526 | ||
3527 | while (!list_empty(&splice)) { | |
3528 | snapshot = list_entry(splice.next, | |
3529 | struct btrfs_pending_snapshot, | |
3530 | list); | |
3531 | ||
3532 | list_del_init(&snapshot->list); | |
3533 | ||
3534 | kfree(snapshot); | |
3535 | } | |
acce952b | 3536 | } |
3537 | ||
143bede5 | 3538 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 3539 | { |
3540 | struct btrfs_inode *btrfs_inode; | |
3541 | struct list_head splice; | |
3542 | ||
3543 | INIT_LIST_HEAD(&splice); | |
3544 | ||
acce952b | 3545 | spin_lock(&root->fs_info->delalloc_lock); |
5be76758 | 3546 | list_splice_init(&root->fs_info->delalloc_inodes, &splice); |
acce952b | 3547 | |
3548 | while (!list_empty(&splice)) { | |
3549 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3550 | delalloc_inodes); | |
3551 | ||
3552 | list_del_init(&btrfs_inode->delalloc_inodes); | |
3553 | ||
3554 | btrfs_invalidate_inodes(btrfs_inode->root); | |
3555 | } | |
3556 | ||
3557 | spin_unlock(&root->fs_info->delalloc_lock); | |
acce952b | 3558 | } |
3559 | ||
3560 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, | |
3561 | struct extent_io_tree *dirty_pages, | |
3562 | int mark) | |
3563 | { | |
3564 | int ret; | |
3565 | struct page *page; | |
3566 | struct inode *btree_inode = root->fs_info->btree_inode; | |
3567 | struct extent_buffer *eb; | |
3568 | u64 start = 0; | |
3569 | u64 end; | |
3570 | u64 offset; | |
3571 | unsigned long index; | |
3572 | ||
3573 | while (1) { | |
3574 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
3575 | mark); | |
3576 | if (ret) | |
3577 | break; | |
3578 | ||
3579 | clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); | |
3580 | while (start <= end) { | |
3581 | index = start >> PAGE_CACHE_SHIFT; | |
3582 | start = (u64)(index + 1) << PAGE_CACHE_SHIFT; | |
3583 | page = find_get_page(btree_inode->i_mapping, index); | |
3584 | if (!page) | |
3585 | continue; | |
3586 | offset = page_offset(page); | |
3587 | ||
3588 | spin_lock(&dirty_pages->buffer_lock); | |
3589 | eb = radix_tree_lookup( | |
3590 | &(&BTRFS_I(page->mapping->host)->io_tree)->buffer, | |
3591 | offset >> PAGE_CACHE_SHIFT); | |
3592 | spin_unlock(&dirty_pages->buffer_lock); | |
ee670f0a | 3593 | if (eb) |
acce952b | 3594 | ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
3595 | &eb->bflags); | |
acce952b | 3596 | if (PageWriteback(page)) |
3597 | end_page_writeback(page); | |
3598 | ||
3599 | lock_page(page); | |
3600 | if (PageDirty(page)) { | |
3601 | clear_page_dirty_for_io(page); | |
3602 | spin_lock_irq(&page->mapping->tree_lock); | |
3603 | radix_tree_tag_clear(&page->mapping->page_tree, | |
3604 | page_index(page), | |
3605 | PAGECACHE_TAG_DIRTY); | |
3606 | spin_unlock_irq(&page->mapping->tree_lock); | |
3607 | } | |
3608 | ||
acce952b | 3609 | unlock_page(page); |
ee670f0a | 3610 | page_cache_release(page); |
acce952b | 3611 | } |
3612 | } | |
3613 | ||
3614 | return ret; | |
3615 | } | |
3616 | ||
3617 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
3618 | struct extent_io_tree *pinned_extents) | |
3619 | { | |
3620 | struct extent_io_tree *unpin; | |
3621 | u64 start; | |
3622 | u64 end; | |
3623 | int ret; | |
ed0eaa14 | 3624 | bool loop = true; |
acce952b | 3625 | |
3626 | unpin = pinned_extents; | |
ed0eaa14 | 3627 | again: |
acce952b | 3628 | while (1) { |
3629 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
3630 | EXTENT_DIRTY); | |
3631 | if (ret) | |
3632 | break; | |
3633 | ||
3634 | /* opt_discard */ | |
5378e607 LD |
3635 | if (btrfs_test_opt(root, DISCARD)) |
3636 | ret = btrfs_error_discard_extent(root, start, | |
3637 | end + 1 - start, | |
3638 | NULL); | |
acce952b | 3639 | |
3640 | clear_extent_dirty(unpin, start, end, GFP_NOFS); | |
3641 | btrfs_error_unpin_extent_range(root, start, end); | |
3642 | cond_resched(); | |
3643 | } | |
3644 | ||
ed0eaa14 LB |
3645 | if (loop) { |
3646 | if (unpin == &root->fs_info->freed_extents[0]) | |
3647 | unpin = &root->fs_info->freed_extents[1]; | |
3648 | else | |
3649 | unpin = &root->fs_info->freed_extents[0]; | |
3650 | loop = false; | |
3651 | goto again; | |
3652 | } | |
3653 | ||
acce952b | 3654 | return 0; |
3655 | } | |
3656 | ||
49b25e05 JM |
3657 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
3658 | struct btrfs_root *root) | |
3659 | { | |
3660 | btrfs_destroy_delayed_refs(cur_trans, root); | |
3661 | btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv, | |
3662 | cur_trans->dirty_pages.dirty_bytes); | |
3663 | ||
3664 | /* FIXME: cleanup wait for commit */ | |
3665 | cur_trans->in_commit = 1; | |
3666 | cur_trans->blocked = 1; | |
d7096fc3 | 3667 | wake_up(&root->fs_info->transaction_blocked_wait); |
49b25e05 JM |
3668 | |
3669 | cur_trans->blocked = 0; | |
d7096fc3 | 3670 | wake_up(&root->fs_info->transaction_wait); |
49b25e05 JM |
3671 | |
3672 | cur_trans->commit_done = 1; | |
d7096fc3 | 3673 | wake_up(&cur_trans->commit_wait); |
49b25e05 | 3674 | |
67cde344 MX |
3675 | btrfs_destroy_delayed_inodes(root); |
3676 | btrfs_assert_delayed_root_empty(root); | |
49b25e05 JM |
3677 | |
3678 | btrfs_destroy_pending_snapshots(cur_trans); | |
3679 | ||
3680 | btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, | |
3681 | EXTENT_DIRTY); | |
6e841e32 LB |
3682 | btrfs_destroy_pinned_extent(root, |
3683 | root->fs_info->pinned_extents); | |
49b25e05 JM |
3684 | |
3685 | /* | |
3686 | memset(cur_trans, 0, sizeof(*cur_trans)); | |
3687 | kmem_cache_free(btrfs_transaction_cachep, cur_trans); | |
3688 | */ | |
3689 | } | |
3690 | ||
3691 | int btrfs_cleanup_transaction(struct btrfs_root *root) | |
acce952b | 3692 | { |
3693 | struct btrfs_transaction *t; | |
3694 | LIST_HEAD(list); | |
3695 | ||
acce952b | 3696 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
3697 | ||
a4abeea4 | 3698 | spin_lock(&root->fs_info->trans_lock); |
acce952b | 3699 | list_splice_init(&root->fs_info->trans_list, &list); |
a4abeea4 JB |
3700 | root->fs_info->trans_no_join = 1; |
3701 | spin_unlock(&root->fs_info->trans_lock); | |
3702 | ||
acce952b | 3703 | while (!list_empty(&list)) { |
3704 | t = list_entry(list.next, struct btrfs_transaction, list); | |
3705 | if (!t) | |
3706 | break; | |
3707 | ||
3708 | btrfs_destroy_ordered_operations(root); | |
3709 | ||
3710 | btrfs_destroy_ordered_extents(root); | |
3711 | ||
3712 | btrfs_destroy_delayed_refs(t, root); | |
3713 | ||
3714 | btrfs_block_rsv_release(root, | |
3715 | &root->fs_info->trans_block_rsv, | |
3716 | t->dirty_pages.dirty_bytes); | |
3717 | ||
3718 | /* FIXME: cleanup wait for commit */ | |
3719 | t->in_commit = 1; | |
3720 | t->blocked = 1; | |
66657b31 | 3721 | smp_mb(); |
acce952b | 3722 | if (waitqueue_active(&root->fs_info->transaction_blocked_wait)) |
3723 | wake_up(&root->fs_info->transaction_blocked_wait); | |
3724 | ||
3725 | t->blocked = 0; | |
66657b31 | 3726 | smp_mb(); |
acce952b | 3727 | if (waitqueue_active(&root->fs_info->transaction_wait)) |
3728 | wake_up(&root->fs_info->transaction_wait); | |
acce952b | 3729 | |
acce952b | 3730 | t->commit_done = 1; |
66657b31 | 3731 | smp_mb(); |
acce952b | 3732 | if (waitqueue_active(&t->commit_wait)) |
3733 | wake_up(&t->commit_wait); | |
acce952b | 3734 | |
67cde344 MX |
3735 | btrfs_destroy_delayed_inodes(root); |
3736 | btrfs_assert_delayed_root_empty(root); | |
3737 | ||
acce952b | 3738 | btrfs_destroy_pending_snapshots(t); |
3739 | ||
3740 | btrfs_destroy_delalloc_inodes(root); | |
3741 | ||
a4abeea4 | 3742 | spin_lock(&root->fs_info->trans_lock); |
acce952b | 3743 | root->fs_info->running_transaction = NULL; |
a4abeea4 | 3744 | spin_unlock(&root->fs_info->trans_lock); |
acce952b | 3745 | |
3746 | btrfs_destroy_marked_extents(root, &t->dirty_pages, | |
3747 | EXTENT_DIRTY); | |
3748 | ||
3749 | btrfs_destroy_pinned_extent(root, | |
3750 | root->fs_info->pinned_extents); | |
3751 | ||
13c5a93e | 3752 | atomic_set(&t->use_count, 0); |
acce952b | 3753 | list_del_init(&t->list); |
3754 | memset(t, 0, sizeof(*t)); | |
3755 | kmem_cache_free(btrfs_transaction_cachep, t); | |
3756 | } | |
3757 | ||
a4abeea4 JB |
3758 | spin_lock(&root->fs_info->trans_lock); |
3759 | root->fs_info->trans_no_join = 0; | |
3760 | spin_unlock(&root->fs_info->trans_lock); | |
acce952b | 3761 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); |
acce952b | 3762 | |
3763 | return 0; | |
3764 | } | |
3765 | ||
d1310b2e | 3766 | static struct extent_io_ops btree_extent_io_ops = { |
ce9adaa5 | 3767 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
4bb31e92 | 3768 | .readpage_io_failed_hook = btree_io_failed_hook, |
0b86a832 | 3769 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
3770 | /* note we're sharing with inode.c for the merge bio hook */ |
3771 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 3772 | }; |