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" |
8dabb742 | 48 | #include "dev-replace.h" |
eb60ceac | 49 | |
de0022b9 JB |
50 | #ifdef CONFIG_X86 |
51 | #include <asm/cpufeature.h> | |
52 | #endif | |
53 | ||
d1310b2e | 54 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 55 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 56 | static void free_fs_root(struct btrfs_root *root); |
fcd1f065 | 57 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 58 | int read_only); |
569e0f35 JB |
59 | static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, |
60 | struct btrfs_root *root); | |
143bede5 | 61 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); |
acce952b | 62 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
63 | struct btrfs_root *root); | |
143bede5 JM |
64 | static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t); |
65 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); | |
acce952b | 66 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, |
67 | struct extent_io_tree *dirty_pages, | |
68 | int mark); | |
69 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
70 | struct extent_io_tree *pinned_extents); | |
ce9adaa5 | 71 | |
d352ac68 CM |
72 | /* |
73 | * end_io_wq structs are used to do processing in task context when an IO is | |
74 | * complete. This is used during reads to verify checksums, and it is used | |
75 | * by writes to insert metadata for new file extents after IO is complete. | |
76 | */ | |
ce9adaa5 CM |
77 | struct end_io_wq { |
78 | struct bio *bio; | |
79 | bio_end_io_t *end_io; | |
80 | void *private; | |
81 | struct btrfs_fs_info *info; | |
82 | int error; | |
22c59948 | 83 | int metadata; |
ce9adaa5 | 84 | struct list_head list; |
8b712842 | 85 | struct btrfs_work work; |
ce9adaa5 | 86 | }; |
0da5468f | 87 | |
d352ac68 CM |
88 | /* |
89 | * async submit bios are used to offload expensive checksumming | |
90 | * onto the worker threads. They checksum file and metadata bios | |
91 | * just before they are sent down the IO stack. | |
92 | */ | |
44b8bd7e CM |
93 | struct async_submit_bio { |
94 | struct inode *inode; | |
95 | struct bio *bio; | |
96 | struct list_head list; | |
4a69a410 CM |
97 | extent_submit_bio_hook_t *submit_bio_start; |
98 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
99 | int rw; |
100 | int mirror_num; | |
c8b97818 | 101 | unsigned long bio_flags; |
eaf25d93 CM |
102 | /* |
103 | * bio_offset is optional, can be used if the pages in the bio | |
104 | * can't tell us where in the file the bio should go | |
105 | */ | |
106 | u64 bio_offset; | |
8b712842 | 107 | struct btrfs_work work; |
79787eaa | 108 | int error; |
44b8bd7e CM |
109 | }; |
110 | ||
85d4e461 CM |
111 | /* |
112 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
113 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
114 | * the level the eb occupies in the tree. | |
115 | * | |
116 | * Different roots are used for different purposes and may nest inside each | |
117 | * other and they require separate keysets. As lockdep keys should be | |
118 | * static, assign keysets according to the purpose of the root as indicated | |
119 | * by btrfs_root->objectid. This ensures that all special purpose roots | |
120 | * have separate keysets. | |
4008c04a | 121 | * |
85d4e461 CM |
122 | * Lock-nesting across peer nodes is always done with the immediate parent |
123 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
124 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 125 | * |
85d4e461 CM |
126 | * The key is set by the readpage_end_io_hook after the buffer has passed |
127 | * csum validation but before the pages are unlocked. It is also set by | |
128 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 129 | * |
85d4e461 CM |
130 | * We also add a check to make sure the highest level of the tree is the |
131 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
132 | * needs update as well. | |
4008c04a CM |
133 | */ |
134 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
135 | # if BTRFS_MAX_LEVEL != 8 | |
136 | # error | |
137 | # endif | |
85d4e461 CM |
138 | |
139 | static struct btrfs_lockdep_keyset { | |
140 | u64 id; /* root objectid */ | |
141 | const char *name_stem; /* lock name stem */ | |
142 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
143 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
144 | } btrfs_lockdep_keysets[] = { | |
145 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
146 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
147 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
148 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
149 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
150 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
151 | { .id = BTRFS_ORPHAN_OBJECTID, .name_stem = "orphan" }, | |
152 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, | |
153 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
154 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
155 | { .id = 0, .name_stem = "tree" }, | |
4008c04a | 156 | }; |
85d4e461 CM |
157 | |
158 | void __init btrfs_init_lockdep(void) | |
159 | { | |
160 | int i, j; | |
161 | ||
162 | /* initialize lockdep class names */ | |
163 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
164 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
165 | ||
166 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
167 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
168 | "btrfs-%s-%02d", ks->name_stem, j); | |
169 | } | |
170 | } | |
171 | ||
172 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
173 | int level) | |
174 | { | |
175 | struct btrfs_lockdep_keyset *ks; | |
176 | ||
177 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
178 | ||
179 | /* find the matching keyset, id 0 is the default entry */ | |
180 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
181 | if (ks->id == objectid) | |
182 | break; | |
183 | ||
184 | lockdep_set_class_and_name(&eb->lock, | |
185 | &ks->keys[level], ks->names[level]); | |
186 | } | |
187 | ||
4008c04a CM |
188 | #endif |
189 | ||
d352ac68 CM |
190 | /* |
191 | * extents on the btree inode are pretty simple, there's one extent | |
192 | * that covers the entire device | |
193 | */ | |
b2950863 | 194 | static struct extent_map *btree_get_extent(struct inode *inode, |
306e16ce | 195 | struct page *page, size_t pg_offset, u64 start, u64 len, |
b2950863 | 196 | int create) |
7eccb903 | 197 | { |
5f39d397 CM |
198 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
199 | struct extent_map *em; | |
200 | int ret; | |
201 | ||
890871be | 202 | read_lock(&em_tree->lock); |
d1310b2e | 203 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d CM |
204 | if (em) { |
205 | em->bdev = | |
206 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
890871be | 207 | read_unlock(&em_tree->lock); |
5f39d397 | 208 | goto out; |
a061fc8d | 209 | } |
890871be | 210 | read_unlock(&em_tree->lock); |
7b13b7b1 | 211 | |
172ddd60 | 212 | em = alloc_extent_map(); |
5f39d397 CM |
213 | if (!em) { |
214 | em = ERR_PTR(-ENOMEM); | |
215 | goto out; | |
216 | } | |
217 | em->start = 0; | |
0afbaf8c | 218 | em->len = (u64)-1; |
c8b97818 | 219 | em->block_len = (u64)-1; |
5f39d397 | 220 | em->block_start = 0; |
a061fc8d | 221 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e | 222 | |
890871be | 223 | write_lock(&em_tree->lock); |
5f39d397 CM |
224 | ret = add_extent_mapping(em_tree, em); |
225 | if (ret == -EEXIST) { | |
226 | free_extent_map(em); | |
7b13b7b1 | 227 | em = lookup_extent_mapping(em_tree, start, len); |
b4f359ab | 228 | if (!em) |
0433f20d | 229 | em = ERR_PTR(-EIO); |
5f39d397 | 230 | } else if (ret) { |
7b13b7b1 | 231 | free_extent_map(em); |
0433f20d | 232 | em = ERR_PTR(ret); |
5f39d397 | 233 | } |
890871be | 234 | write_unlock(&em_tree->lock); |
7b13b7b1 | 235 | |
5f39d397 CM |
236 | out: |
237 | return em; | |
7eccb903 CM |
238 | } |
239 | ||
19c00ddc CM |
240 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
241 | { | |
163e783e | 242 | return crc32c(seed, data, len); |
19c00ddc CM |
243 | } |
244 | ||
245 | void btrfs_csum_final(u32 crc, char *result) | |
246 | { | |
7e75bf3f | 247 | put_unaligned_le32(~crc, result); |
19c00ddc CM |
248 | } |
249 | ||
d352ac68 CM |
250 | /* |
251 | * compute the csum for a btree block, and either verify it or write it | |
252 | * into the csum field of the block. | |
253 | */ | |
19c00ddc CM |
254 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
255 | int verify) | |
256 | { | |
6c41761f | 257 | u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy); |
607d432d | 258 | char *result = NULL; |
19c00ddc CM |
259 | unsigned long len; |
260 | unsigned long cur_len; | |
261 | unsigned long offset = BTRFS_CSUM_SIZE; | |
19c00ddc CM |
262 | char *kaddr; |
263 | unsigned long map_start; | |
264 | unsigned long map_len; | |
265 | int err; | |
266 | u32 crc = ~(u32)0; | |
607d432d | 267 | unsigned long inline_result; |
19c00ddc CM |
268 | |
269 | len = buf->len - offset; | |
d397712b | 270 | while (len > 0) { |
19c00ddc | 271 | err = map_private_extent_buffer(buf, offset, 32, |
a6591715 | 272 | &kaddr, &map_start, &map_len); |
d397712b | 273 | if (err) |
19c00ddc | 274 | return 1; |
19c00ddc CM |
275 | cur_len = min(len, map_len - (offset - map_start)); |
276 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
277 | crc, cur_len); | |
278 | len -= cur_len; | |
279 | offset += cur_len; | |
19c00ddc | 280 | } |
607d432d JB |
281 | if (csum_size > sizeof(inline_result)) { |
282 | result = kzalloc(csum_size * sizeof(char), GFP_NOFS); | |
283 | if (!result) | |
284 | return 1; | |
285 | } else { | |
286 | result = (char *)&inline_result; | |
287 | } | |
288 | ||
19c00ddc CM |
289 | btrfs_csum_final(crc, result); |
290 | ||
291 | if (verify) { | |
607d432d | 292 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
293 | u32 val; |
294 | u32 found = 0; | |
607d432d | 295 | memcpy(&found, result, csum_size); |
e4204ded | 296 | |
607d432d | 297 | read_extent_buffer(buf, &val, 0, csum_size); |
7a36ddec | 298 | printk_ratelimited(KERN_INFO "btrfs: %s checksum verify " |
193f284d CM |
299 | "failed on %llu wanted %X found %X " |
300 | "level %d\n", | |
301 | root->fs_info->sb->s_id, | |
302 | (unsigned long long)buf->start, val, found, | |
303 | btrfs_header_level(buf)); | |
607d432d JB |
304 | if (result != (char *)&inline_result) |
305 | kfree(result); | |
19c00ddc CM |
306 | return 1; |
307 | } | |
308 | } else { | |
607d432d | 309 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 310 | } |
607d432d JB |
311 | if (result != (char *)&inline_result) |
312 | kfree(result); | |
19c00ddc CM |
313 | return 0; |
314 | } | |
315 | ||
d352ac68 CM |
316 | /* |
317 | * we can't consider a given block up to date unless the transid of the | |
318 | * block matches the transid in the parent node's pointer. This is how we | |
319 | * detect blocks that either didn't get written at all or got written | |
320 | * in the wrong place. | |
321 | */ | |
1259ab75 | 322 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
323 | struct extent_buffer *eb, u64 parent_transid, |
324 | int atomic) | |
1259ab75 | 325 | { |
2ac55d41 | 326 | struct extent_state *cached_state = NULL; |
1259ab75 CM |
327 | int ret; |
328 | ||
329 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
330 | return 0; | |
331 | ||
b9fab919 CM |
332 | if (atomic) |
333 | return -EAGAIN; | |
334 | ||
2ac55d41 | 335 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
d0082371 | 336 | 0, &cached_state); |
0b32f4bb | 337 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
338 | btrfs_header_generation(eb) == parent_transid) { |
339 | ret = 0; | |
340 | goto out; | |
341 | } | |
7a36ddec | 342 | printk_ratelimited("parent transid verify failed on %llu wanted %llu " |
193f284d CM |
343 | "found %llu\n", |
344 | (unsigned long long)eb->start, | |
345 | (unsigned long long)parent_transid, | |
346 | (unsigned long long)btrfs_header_generation(eb)); | |
1259ab75 | 347 | ret = 1; |
0b32f4bb | 348 | clear_extent_buffer_uptodate(eb); |
33958dc6 | 349 | out: |
2ac55d41 JB |
350 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
351 | &cached_state, GFP_NOFS); | |
1259ab75 | 352 | return ret; |
1259ab75 CM |
353 | } |
354 | ||
d352ac68 CM |
355 | /* |
356 | * helper to read a given tree block, doing retries as required when | |
357 | * the checksums don't match and we have alternate mirrors to try. | |
358 | */ | |
f188591e CM |
359 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
360 | struct extent_buffer *eb, | |
ca7a79ad | 361 | u64 start, u64 parent_transid) |
f188591e CM |
362 | { |
363 | struct extent_io_tree *io_tree; | |
ea466794 | 364 | int failed = 0; |
f188591e CM |
365 | int ret; |
366 | int num_copies = 0; | |
367 | int mirror_num = 0; | |
ea466794 | 368 | int failed_mirror = 0; |
f188591e | 369 | |
a826d6dc | 370 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
f188591e CM |
371 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
372 | while (1) { | |
bb82ab88 AJ |
373 | ret = read_extent_buffer_pages(io_tree, eb, start, |
374 | WAIT_COMPLETE, | |
f188591e | 375 | btree_get_extent, mirror_num); |
256dd1bb SB |
376 | if (!ret) { |
377 | if (!verify_parent_transid(io_tree, eb, | |
b9fab919 | 378 | parent_transid, 0)) |
256dd1bb SB |
379 | break; |
380 | else | |
381 | ret = -EIO; | |
382 | } | |
d397712b | 383 | |
a826d6dc JB |
384 | /* |
385 | * This buffer's crc is fine, but its contents are corrupted, so | |
386 | * there is no reason to read the other copies, they won't be | |
387 | * any less wrong. | |
388 | */ | |
389 | if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) | |
ea466794 JB |
390 | break; |
391 | ||
5d964051 | 392 | num_copies = btrfs_num_copies(root->fs_info, |
f188591e | 393 | eb->start, eb->len); |
4235298e | 394 | if (num_copies == 1) |
ea466794 | 395 | break; |
4235298e | 396 | |
5cf1ab56 JB |
397 | if (!failed_mirror) { |
398 | failed = 1; | |
399 | failed_mirror = eb->read_mirror; | |
400 | } | |
401 | ||
f188591e | 402 | mirror_num++; |
ea466794 JB |
403 | if (mirror_num == failed_mirror) |
404 | mirror_num++; | |
405 | ||
4235298e | 406 | if (mirror_num > num_copies) |
ea466794 | 407 | break; |
f188591e | 408 | } |
ea466794 | 409 | |
c0901581 | 410 | if (failed && !ret && failed_mirror) |
ea466794 JB |
411 | repair_eb_io_failure(root, eb, failed_mirror); |
412 | ||
413 | return ret; | |
f188591e | 414 | } |
19c00ddc | 415 | |
d352ac68 | 416 | /* |
d397712b CM |
417 | * checksum a dirty tree block before IO. This has extra checks to make sure |
418 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 419 | */ |
d397712b | 420 | |
b2950863 | 421 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 422 | { |
d1310b2e | 423 | struct extent_io_tree *tree; |
4eee4fa4 | 424 | u64 start = page_offset(page); |
19c00ddc | 425 | u64 found_start; |
19c00ddc | 426 | struct extent_buffer *eb; |
f188591e | 427 | |
d1310b2e | 428 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc | 429 | |
4f2de97a JB |
430 | eb = (struct extent_buffer *)page->private; |
431 | if (page != eb->pages[0]) | |
432 | return 0; | |
19c00ddc CM |
433 | found_start = btrfs_header_bytenr(eb); |
434 | if (found_start != start) { | |
55c69072 | 435 | WARN_ON(1); |
4f2de97a | 436 | return 0; |
55c69072 | 437 | } |
55c69072 | 438 | if (!PageUptodate(page)) { |
55c69072 | 439 | WARN_ON(1); |
4f2de97a | 440 | return 0; |
19c00ddc | 441 | } |
19c00ddc | 442 | csum_tree_block(root, eb, 0); |
19c00ddc CM |
443 | return 0; |
444 | } | |
445 | ||
2b82032c YZ |
446 | static int check_tree_block_fsid(struct btrfs_root *root, |
447 | struct extent_buffer *eb) | |
448 | { | |
449 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
450 | u8 fsid[BTRFS_UUID_SIZE]; | |
451 | int ret = 1; | |
452 | ||
453 | read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), | |
454 | BTRFS_FSID_SIZE); | |
455 | while (fs_devices) { | |
456 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
457 | ret = 0; | |
458 | break; | |
459 | } | |
460 | fs_devices = fs_devices->seed; | |
461 | } | |
462 | return ret; | |
463 | } | |
464 | ||
a826d6dc JB |
465 | #define CORRUPT(reason, eb, root, slot) \ |
466 | printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \ | |
467 | "root=%llu, slot=%d\n", reason, \ | |
468 | (unsigned long long)btrfs_header_bytenr(eb), \ | |
469 | (unsigned long long)root->objectid, slot) | |
470 | ||
471 | static noinline int check_leaf(struct btrfs_root *root, | |
472 | struct extent_buffer *leaf) | |
473 | { | |
474 | struct btrfs_key key; | |
475 | struct btrfs_key leaf_key; | |
476 | u32 nritems = btrfs_header_nritems(leaf); | |
477 | int slot; | |
478 | ||
479 | if (nritems == 0) | |
480 | return 0; | |
481 | ||
482 | /* Check the 0 item */ | |
483 | if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != | |
484 | BTRFS_LEAF_DATA_SIZE(root)) { | |
485 | CORRUPT("invalid item offset size pair", leaf, root, 0); | |
486 | return -EIO; | |
487 | } | |
488 | ||
489 | /* | |
490 | * Check to make sure each items keys are in the correct order and their | |
491 | * offsets make sense. We only have to loop through nritems-1 because | |
492 | * we check the current slot against the next slot, which verifies the | |
493 | * next slot's offset+size makes sense and that the current's slot | |
494 | * offset is correct. | |
495 | */ | |
496 | for (slot = 0; slot < nritems - 1; slot++) { | |
497 | btrfs_item_key_to_cpu(leaf, &leaf_key, slot); | |
498 | btrfs_item_key_to_cpu(leaf, &key, slot + 1); | |
499 | ||
500 | /* Make sure the keys are in the right order */ | |
501 | if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { | |
502 | CORRUPT("bad key order", leaf, root, slot); | |
503 | return -EIO; | |
504 | } | |
505 | ||
506 | /* | |
507 | * Make sure the offset and ends are right, remember that the | |
508 | * item data starts at the end of the leaf and grows towards the | |
509 | * front. | |
510 | */ | |
511 | if (btrfs_item_offset_nr(leaf, slot) != | |
512 | btrfs_item_end_nr(leaf, slot + 1)) { | |
513 | CORRUPT("slot offset bad", leaf, root, slot); | |
514 | return -EIO; | |
515 | } | |
516 | ||
517 | /* | |
518 | * Check to make sure that we don't point outside of the leaf, | |
519 | * just incase all the items are consistent to eachother, but | |
520 | * all point outside of the leaf. | |
521 | */ | |
522 | if (btrfs_item_end_nr(leaf, slot) > | |
523 | BTRFS_LEAF_DATA_SIZE(root)) { | |
524 | CORRUPT("slot end outside of leaf", leaf, root, slot); | |
525 | return -EIO; | |
526 | } | |
527 | } | |
528 | ||
529 | return 0; | |
530 | } | |
531 | ||
727011e0 CM |
532 | struct extent_buffer *find_eb_for_page(struct extent_io_tree *tree, |
533 | struct page *page, int max_walk) | |
534 | { | |
535 | struct extent_buffer *eb; | |
536 | u64 start = page_offset(page); | |
537 | u64 target = start; | |
538 | u64 min_start; | |
539 | ||
540 | if (start < max_walk) | |
541 | min_start = 0; | |
542 | else | |
543 | min_start = start - max_walk; | |
544 | ||
545 | while (start >= min_start) { | |
546 | eb = find_extent_buffer(tree, start, 0); | |
547 | if (eb) { | |
548 | /* | |
549 | * we found an extent buffer and it contains our page | |
550 | * horray! | |
551 | */ | |
552 | if (eb->start <= target && | |
553 | eb->start + eb->len > target) | |
554 | return eb; | |
555 | ||
556 | /* we found an extent buffer that wasn't for us */ | |
557 | free_extent_buffer(eb); | |
558 | return NULL; | |
559 | } | |
560 | if (start == 0) | |
561 | break; | |
562 | start -= PAGE_CACHE_SIZE; | |
563 | } | |
564 | return NULL; | |
565 | } | |
566 | ||
b2950863 | 567 | static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
5cf1ab56 | 568 | struct extent_state *state, int mirror) |
ce9adaa5 CM |
569 | { |
570 | struct extent_io_tree *tree; | |
571 | u64 found_start; | |
572 | int found_level; | |
ce9adaa5 CM |
573 | struct extent_buffer *eb; |
574 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 575 | int ret = 0; |
727011e0 | 576 | int reads_done; |
ce9adaa5 | 577 | |
ce9adaa5 CM |
578 | if (!page->private) |
579 | goto out; | |
d397712b | 580 | |
727011e0 | 581 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
4f2de97a | 582 | eb = (struct extent_buffer *)page->private; |
d397712b | 583 | |
0b32f4bb JB |
584 | /* the pending IO might have been the only thing that kept this buffer |
585 | * in memory. Make sure we have a ref for all this other checks | |
586 | */ | |
587 | extent_buffer_get(eb); | |
588 | ||
589 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
727011e0 CM |
590 | if (!reads_done) |
591 | goto err; | |
f188591e | 592 | |
5cf1ab56 | 593 | eb->read_mirror = mirror; |
ea466794 JB |
594 | if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) { |
595 | ret = -EIO; | |
596 | goto err; | |
597 | } | |
598 | ||
ce9adaa5 | 599 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 600 | if (found_start != eb->start) { |
7a36ddec | 601 | printk_ratelimited(KERN_INFO "btrfs bad tree block start " |
193f284d CM |
602 | "%llu %llu\n", |
603 | (unsigned long long)found_start, | |
604 | (unsigned long long)eb->start); | |
f188591e | 605 | ret = -EIO; |
ce9adaa5 CM |
606 | goto err; |
607 | } | |
2b82032c | 608 | if (check_tree_block_fsid(root, eb)) { |
7a36ddec | 609 | printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n", |
193f284d | 610 | (unsigned long long)eb->start); |
1259ab75 CM |
611 | ret = -EIO; |
612 | goto err; | |
613 | } | |
ce9adaa5 CM |
614 | found_level = btrfs_header_level(eb); |
615 | ||
85d4e461 CM |
616 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
617 | eb, found_level); | |
4008c04a | 618 | |
ce9adaa5 | 619 | ret = csum_tree_block(root, eb, 1); |
a826d6dc | 620 | if (ret) { |
f188591e | 621 | ret = -EIO; |
a826d6dc JB |
622 | goto err; |
623 | } | |
624 | ||
625 | /* | |
626 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
627 | * that we don't try and read the other copies of this block, just | |
628 | * return -EIO. | |
629 | */ | |
630 | if (found_level == 0 && check_leaf(root, eb)) { | |
631 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); | |
632 | ret = -EIO; | |
633 | } | |
ce9adaa5 | 634 | |
0b32f4bb JB |
635 | if (!ret) |
636 | set_extent_buffer_uptodate(eb); | |
ce9adaa5 | 637 | err: |
4bb31e92 AJ |
638 | if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) { |
639 | clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags); | |
640 | btree_readahead_hook(root, eb, eb->start, ret); | |
641 | } | |
642 | ||
0b32f4bb JB |
643 | if (ret) |
644 | clear_extent_buffer_uptodate(eb); | |
645 | free_extent_buffer(eb); | |
ce9adaa5 | 646 | out: |
f188591e | 647 | return ret; |
ce9adaa5 CM |
648 | } |
649 | ||
ea466794 | 650 | static int btree_io_failed_hook(struct page *page, int failed_mirror) |
4bb31e92 | 651 | { |
4bb31e92 AJ |
652 | struct extent_buffer *eb; |
653 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
654 | ||
4f2de97a | 655 | eb = (struct extent_buffer *)page->private; |
ea466794 | 656 | set_bit(EXTENT_BUFFER_IOERR, &eb->bflags); |
5cf1ab56 | 657 | eb->read_mirror = failed_mirror; |
ea466794 | 658 | if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) |
4bb31e92 | 659 | btree_readahead_hook(root, eb, eb->start, -EIO); |
4bb31e92 AJ |
660 | return -EIO; /* we fixed nothing */ |
661 | } | |
662 | ||
ce9adaa5 | 663 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
664 | { |
665 | struct end_io_wq *end_io_wq = bio->bi_private; | |
666 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 667 | |
ce9adaa5 | 668 | fs_info = end_io_wq->info; |
ce9adaa5 | 669 | end_io_wq->error = err; |
8b712842 CM |
670 | end_io_wq->work.func = end_workqueue_fn; |
671 | end_io_wq->work.flags = 0; | |
d20f7043 | 672 | |
7b6d91da | 673 | if (bio->bi_rw & REQ_WRITE) { |
0cb59c99 | 674 | if (end_io_wq->metadata == 1) |
cad321ad CM |
675 | btrfs_queue_worker(&fs_info->endio_meta_write_workers, |
676 | &end_io_wq->work); | |
0cb59c99 JB |
677 | else if (end_io_wq->metadata == 2) |
678 | btrfs_queue_worker(&fs_info->endio_freespace_worker, | |
679 | &end_io_wq->work); | |
cad321ad CM |
680 | else |
681 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
682 | &end_io_wq->work); | |
d20f7043 CM |
683 | } else { |
684 | if (end_io_wq->metadata) | |
685 | btrfs_queue_worker(&fs_info->endio_meta_workers, | |
686 | &end_io_wq->work); | |
687 | else | |
688 | btrfs_queue_worker(&fs_info->endio_workers, | |
689 | &end_io_wq->work); | |
690 | } | |
ce9adaa5 CM |
691 | } |
692 | ||
0cb59c99 JB |
693 | /* |
694 | * For the metadata arg you want | |
695 | * | |
696 | * 0 - if data | |
697 | * 1 - if normal metadta | |
698 | * 2 - if writing to the free space cache area | |
699 | */ | |
22c59948 CM |
700 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
701 | int metadata) | |
0b86a832 | 702 | { |
ce9adaa5 | 703 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
704 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
705 | if (!end_io_wq) | |
706 | return -ENOMEM; | |
707 | ||
708 | end_io_wq->private = bio->bi_private; | |
709 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 710 | end_io_wq->info = info; |
ce9adaa5 CM |
711 | end_io_wq->error = 0; |
712 | end_io_wq->bio = bio; | |
22c59948 | 713 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
714 | |
715 | bio->bi_private = end_io_wq; | |
716 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
717 | return 0; |
718 | } | |
719 | ||
b64a2851 | 720 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 721 | { |
4854ddd0 CM |
722 | unsigned long limit = min_t(unsigned long, |
723 | info->workers.max_workers, | |
724 | info->fs_devices->open_devices); | |
725 | return 256 * limit; | |
726 | } | |
0986fe9e | 727 | |
4a69a410 CM |
728 | static void run_one_async_start(struct btrfs_work *work) |
729 | { | |
4a69a410 | 730 | struct async_submit_bio *async; |
79787eaa | 731 | int ret; |
4a69a410 CM |
732 | |
733 | async = container_of(work, struct async_submit_bio, work); | |
79787eaa JM |
734 | ret = async->submit_bio_start(async->inode, async->rw, async->bio, |
735 | async->mirror_num, async->bio_flags, | |
736 | async->bio_offset); | |
737 | if (ret) | |
738 | async->error = ret; | |
4a69a410 CM |
739 | } |
740 | ||
741 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
742 | { |
743 | struct btrfs_fs_info *fs_info; | |
744 | struct async_submit_bio *async; | |
4854ddd0 | 745 | int limit; |
8b712842 CM |
746 | |
747 | async = container_of(work, struct async_submit_bio, work); | |
748 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 749 | |
b64a2851 | 750 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
751 | limit = limit * 2 / 3; |
752 | ||
66657b31 | 753 | if (atomic_dec_return(&fs_info->nr_async_submits) < limit && |
b64a2851 | 754 | waitqueue_active(&fs_info->async_submit_wait)) |
4854ddd0 CM |
755 | wake_up(&fs_info->async_submit_wait); |
756 | ||
79787eaa JM |
757 | /* If an error occured we just want to clean up the bio and move on */ |
758 | if (async->error) { | |
759 | bio_endio(async->bio, async->error); | |
760 | return; | |
761 | } | |
762 | ||
4a69a410 | 763 | async->submit_bio_done(async->inode, async->rw, async->bio, |
eaf25d93 CM |
764 | async->mirror_num, async->bio_flags, |
765 | async->bio_offset); | |
4a69a410 CM |
766 | } |
767 | ||
768 | static void run_one_async_free(struct btrfs_work *work) | |
769 | { | |
770 | struct async_submit_bio *async; | |
771 | ||
772 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
773 | kfree(async); |
774 | } | |
775 | ||
44b8bd7e CM |
776 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
777 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 778 | unsigned long bio_flags, |
eaf25d93 | 779 | u64 bio_offset, |
4a69a410 CM |
780 | extent_submit_bio_hook_t *submit_bio_start, |
781 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
782 | { |
783 | struct async_submit_bio *async; | |
784 | ||
785 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
786 | if (!async) | |
787 | return -ENOMEM; | |
788 | ||
789 | async->inode = inode; | |
790 | async->rw = rw; | |
791 | async->bio = bio; | |
792 | async->mirror_num = mirror_num; | |
4a69a410 CM |
793 | async->submit_bio_start = submit_bio_start; |
794 | async->submit_bio_done = submit_bio_done; | |
795 | ||
796 | async->work.func = run_one_async_start; | |
797 | async->work.ordered_func = run_one_async_done; | |
798 | async->work.ordered_free = run_one_async_free; | |
799 | ||
8b712842 | 800 | async->work.flags = 0; |
c8b97818 | 801 | async->bio_flags = bio_flags; |
eaf25d93 | 802 | async->bio_offset = bio_offset; |
8c8bee1d | 803 | |
79787eaa JM |
804 | async->error = 0; |
805 | ||
cb03c743 | 806 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 | 807 | |
7b6d91da | 808 | if (rw & REQ_SYNC) |
d313d7a3 CM |
809 | btrfs_set_work_high_prio(&async->work); |
810 | ||
8b712842 | 811 | btrfs_queue_worker(&fs_info->workers, &async->work); |
9473f16c | 812 | |
d397712b | 813 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
814 | atomic_read(&fs_info->nr_async_submits)) { |
815 | wait_event(fs_info->async_submit_wait, | |
816 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
817 | } | |
818 | ||
44b8bd7e CM |
819 | return 0; |
820 | } | |
821 | ||
ce3ed71a CM |
822 | static int btree_csum_one_bio(struct bio *bio) |
823 | { | |
824 | struct bio_vec *bvec = bio->bi_io_vec; | |
825 | int bio_index = 0; | |
826 | struct btrfs_root *root; | |
79787eaa | 827 | int ret = 0; |
ce3ed71a CM |
828 | |
829 | WARN_ON(bio->bi_vcnt <= 0); | |
d397712b | 830 | while (bio_index < bio->bi_vcnt) { |
ce3ed71a | 831 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
79787eaa JM |
832 | ret = csum_dirty_buffer(root, bvec->bv_page); |
833 | if (ret) | |
834 | break; | |
ce3ed71a CM |
835 | bio_index++; |
836 | bvec++; | |
837 | } | |
79787eaa | 838 | return ret; |
ce3ed71a CM |
839 | } |
840 | ||
4a69a410 CM |
841 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
842 | struct bio *bio, int mirror_num, | |
eaf25d93 CM |
843 | unsigned long bio_flags, |
844 | u64 bio_offset) | |
22c59948 | 845 | { |
8b712842 CM |
846 | /* |
847 | * when we're called for a write, we're already in the async | |
5443be45 | 848 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 849 | */ |
79787eaa | 850 | return btree_csum_one_bio(bio); |
4a69a410 | 851 | } |
22c59948 | 852 | |
4a69a410 | 853 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
854 | int mirror_num, unsigned long bio_flags, |
855 | u64 bio_offset) | |
4a69a410 | 856 | { |
61891923 SB |
857 | int ret; |
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 | */ |
61891923 SB |
863 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
864 | if (ret) | |
865 | bio_endio(bio, ret); | |
866 | return ret; | |
0b86a832 CM |
867 | } |
868 | ||
de0022b9 JB |
869 | static int check_async_write(struct inode *inode, unsigned long bio_flags) |
870 | { | |
871 | if (bio_flags & EXTENT_BIO_TREE_LOG) | |
872 | return 0; | |
873 | #ifdef CONFIG_X86 | |
874 | if (cpu_has_xmm4_2) | |
875 | return 0; | |
876 | #endif | |
877 | return 1; | |
878 | } | |
879 | ||
44b8bd7e | 880 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
881 | int mirror_num, unsigned long bio_flags, |
882 | u64 bio_offset) | |
44b8bd7e | 883 | { |
de0022b9 | 884 | int async = check_async_write(inode, bio_flags); |
cad321ad CM |
885 | int ret; |
886 | ||
7b6d91da | 887 | if (!(rw & REQ_WRITE)) { |
4a69a410 CM |
888 | /* |
889 | * called for a read, do the setup so that checksum validation | |
890 | * can happen in the async kernel threads | |
891 | */ | |
f3f266ab CM |
892 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, |
893 | bio, 1); | |
1d4284bd | 894 | if (ret) |
61891923 SB |
895 | goto out_w_error; |
896 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
897 | mirror_num, 0); | |
de0022b9 JB |
898 | } else if (!async) { |
899 | ret = btree_csum_one_bio(bio); | |
900 | if (ret) | |
61891923 SB |
901 | goto out_w_error; |
902 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
903 | mirror_num, 0); | |
904 | } else { | |
905 | /* | |
906 | * kthread helpers are used to submit writes so that | |
907 | * checksumming can happen in parallel across all CPUs | |
908 | */ | |
909 | ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, | |
910 | inode, rw, bio, mirror_num, 0, | |
911 | bio_offset, | |
912 | __btree_submit_bio_start, | |
913 | __btree_submit_bio_done); | |
44b8bd7e | 914 | } |
d313d7a3 | 915 | |
61891923 SB |
916 | if (ret) { |
917 | out_w_error: | |
918 | bio_endio(bio, ret); | |
919 | } | |
920 | return ret; | |
44b8bd7e CM |
921 | } |
922 | ||
3dd1462e | 923 | #ifdef CONFIG_MIGRATION |
784b4e29 | 924 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
925 | struct page *newpage, struct page *page, |
926 | enum migrate_mode mode) | |
784b4e29 CM |
927 | { |
928 | /* | |
929 | * we can't safely write a btree page from here, | |
930 | * we haven't done the locking hook | |
931 | */ | |
932 | if (PageDirty(page)) | |
933 | return -EAGAIN; | |
934 | /* | |
935 | * Buffers may be managed in a filesystem specific way. | |
936 | * We must have no buffers or drop them. | |
937 | */ | |
938 | if (page_has_private(page) && | |
939 | !try_to_release_page(page, GFP_KERNEL)) | |
940 | return -EAGAIN; | |
a6bc32b8 | 941 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 942 | } |
3dd1462e | 943 | #endif |
784b4e29 | 944 | |
0da5468f CM |
945 | |
946 | static int btree_writepages(struct address_space *mapping, | |
947 | struct writeback_control *wbc) | |
948 | { | |
d1310b2e | 949 | struct extent_io_tree *tree; |
e2d84521 MX |
950 | struct btrfs_fs_info *fs_info; |
951 | int ret; | |
952 | ||
d1310b2e | 953 | tree = &BTRFS_I(mapping->host)->io_tree; |
d8d5f3e1 | 954 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
955 | |
956 | if (wbc->for_kupdate) | |
957 | return 0; | |
958 | ||
e2d84521 | 959 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 960 | /* this is a bit racy, but that's ok */ |
e2d84521 MX |
961 | ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
962 | BTRFS_DIRTY_METADATA_THRESH); | |
963 | if (ret < 0) | |
793955bc | 964 | return 0; |
793955bc | 965 | } |
0b32f4bb | 966 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
967 | } |
968 | ||
b2950863 | 969 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 970 | { |
d1310b2e CM |
971 | struct extent_io_tree *tree; |
972 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 973 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 974 | } |
22b0ebda | 975 | |
70dec807 | 976 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 977 | { |
98509cfc | 978 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 979 | return 0; |
0c4e538b DS |
980 | /* |
981 | * We need to mask out eg. __GFP_HIGHMEM and __GFP_DMA32 as we're doing | |
982 | * slab allocation from alloc_extent_state down the callchain where | |
983 | * it'd hit a BUG_ON as those flags are not allowed. | |
984 | */ | |
985 | gfp_flags &= ~GFP_SLAB_BUG_MASK; | |
986 | ||
3083ee2e | 987 | return try_release_extent_buffer(page, gfp_flags); |
d98237b3 CM |
988 | } |
989 | ||
5f39d397 | 990 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 991 | { |
d1310b2e CM |
992 | struct extent_io_tree *tree; |
993 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
994 | extent_invalidatepage(tree, page, offset); |
995 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 996 | if (PagePrivate(page)) { |
d397712b CM |
997 | printk(KERN_WARNING "btrfs warning page private not zero " |
998 | "on page %llu\n", (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
999 | ClearPagePrivate(page); |
1000 | set_page_private(page, 0); | |
1001 | page_cache_release(page); | |
1002 | } | |
d98237b3 CM |
1003 | } |
1004 | ||
0b32f4bb JB |
1005 | static int btree_set_page_dirty(struct page *page) |
1006 | { | |
bb146eb2 | 1007 | #ifdef DEBUG |
0b32f4bb JB |
1008 | struct extent_buffer *eb; |
1009 | ||
1010 | BUG_ON(!PagePrivate(page)); | |
1011 | eb = (struct extent_buffer *)page->private; | |
1012 | BUG_ON(!eb); | |
1013 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1014 | BUG_ON(!atomic_read(&eb->refs)); | |
1015 | btrfs_assert_tree_locked(eb); | |
bb146eb2 | 1016 | #endif |
0b32f4bb JB |
1017 | return __set_page_dirty_nobuffers(page); |
1018 | } | |
1019 | ||
7f09410b | 1020 | static const struct address_space_operations btree_aops = { |
d98237b3 | 1021 | .readpage = btree_readpage, |
0da5468f | 1022 | .writepages = btree_writepages, |
5f39d397 CM |
1023 | .releasepage = btree_releasepage, |
1024 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1025 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1026 | .migratepage = btree_migratepage, |
5a92bc88 | 1027 | #endif |
0b32f4bb | 1028 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1029 | }; |
1030 | ||
ca7a79ad CM |
1031 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1032 | u64 parent_transid) | |
090d1875 | 1033 | { |
5f39d397 CM |
1034 | struct extent_buffer *buf = NULL; |
1035 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 1036 | int ret = 0; |
090d1875 | 1037 | |
db94535d | 1038 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 1039 | if (!buf) |
090d1875 | 1040 | return 0; |
d1310b2e | 1041 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
bb82ab88 | 1042 | buf, 0, WAIT_NONE, btree_get_extent, 0); |
5f39d397 | 1043 | free_extent_buffer(buf); |
de428b63 | 1044 | return ret; |
090d1875 CM |
1045 | } |
1046 | ||
ab0fff03 AJ |
1047 | int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1048 | int mirror_num, struct extent_buffer **eb) | |
1049 | { | |
1050 | struct extent_buffer *buf = NULL; | |
1051 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1052 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; | |
1053 | int ret; | |
1054 | ||
1055 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
1056 | if (!buf) | |
1057 | return 0; | |
1058 | ||
1059 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1060 | ||
1061 | ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, | |
1062 | btree_get_extent, mirror_num); | |
1063 | if (ret) { | |
1064 | free_extent_buffer(buf); | |
1065 | return ret; | |
1066 | } | |
1067 | ||
1068 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1069 | free_extent_buffer(buf); | |
1070 | return -EIO; | |
0b32f4bb | 1071 | } else if (extent_buffer_uptodate(buf)) { |
ab0fff03 AJ |
1072 | *eb = buf; |
1073 | } else { | |
1074 | free_extent_buffer(buf); | |
1075 | } | |
1076 | return 0; | |
1077 | } | |
1078 | ||
0999df54 CM |
1079 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
1080 | u64 bytenr, u32 blocksize) | |
1081 | { | |
1082 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1083 | struct extent_buffer *eb; | |
1084 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
f09d1f60 | 1085 | bytenr, blocksize); |
0999df54 CM |
1086 | return eb; |
1087 | } | |
1088 | ||
1089 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
1090 | u64 bytenr, u32 blocksize) | |
1091 | { | |
1092 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1093 | struct extent_buffer *eb; | |
1094 | ||
1095 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
727011e0 | 1096 | bytenr, blocksize); |
0999df54 CM |
1097 | return eb; |
1098 | } | |
1099 | ||
1100 | ||
e02119d5 CM |
1101 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1102 | { | |
727011e0 | 1103 | return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, |
8aa38c31 | 1104 | buf->start + buf->len - 1); |
e02119d5 CM |
1105 | } |
1106 | ||
1107 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
1108 | { | |
727011e0 | 1109 | return filemap_fdatawait_range(buf->pages[0]->mapping, |
8aa38c31 | 1110 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
1111 | } |
1112 | ||
0999df54 | 1113 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 1114 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
1115 | { |
1116 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1117 | int ret; |
1118 | ||
0999df54 CM |
1119 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
1120 | if (!buf) | |
1121 | return NULL; | |
0999df54 | 1122 | |
ca7a79ad | 1123 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
5f39d397 | 1124 | return buf; |
ce9adaa5 | 1125 | |
eb60ceac CM |
1126 | } |
1127 | ||
d5c13f92 JM |
1128 | void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
1129 | struct extent_buffer *buf) | |
ed2ff2cb | 1130 | { |
e2d84521 MX |
1131 | struct btrfs_fs_info *fs_info = root->fs_info; |
1132 | ||
55c69072 | 1133 | if (btrfs_header_generation(buf) == |
e2d84521 | 1134 | fs_info->running_transaction->transid) { |
b9447ef8 | 1135 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1136 | |
b9473439 | 1137 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
e2d84521 MX |
1138 | __percpu_counter_add(&fs_info->dirty_metadata_bytes, |
1139 | -buf->len, | |
1140 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1141 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1142 | btrfs_set_lock_blocking(buf); | |
1143 | clear_extent_buffer_dirty(buf); | |
1144 | } | |
925baedd | 1145 | } |
5f39d397 CM |
1146 | } |
1147 | ||
143bede5 JM |
1148 | static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
1149 | u32 stripesize, struct btrfs_root *root, | |
1150 | struct btrfs_fs_info *fs_info, | |
1151 | u64 objectid) | |
d97e63b6 | 1152 | { |
cfaa7295 | 1153 | root->node = NULL; |
a28ec197 | 1154 | root->commit_root = NULL; |
db94535d CM |
1155 | root->sectorsize = sectorsize; |
1156 | root->nodesize = nodesize; | |
1157 | root->leafsize = leafsize; | |
87ee04eb | 1158 | root->stripesize = stripesize; |
123abc88 | 1159 | root->ref_cows = 0; |
0b86a832 | 1160 | root->track_dirty = 0; |
c71bf099 | 1161 | root->in_radix = 0; |
d68fc57b YZ |
1162 | root->orphan_item_inserted = 0; |
1163 | root->orphan_cleanup_state = 0; | |
0b86a832 | 1164 | |
0f7d52f4 CM |
1165 | root->objectid = objectid; |
1166 | root->last_trans = 0; | |
13a8a7c8 | 1167 | root->highest_objectid = 0; |
58176a96 | 1168 | root->name = NULL; |
6bef4d31 | 1169 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1170 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1171 | root->block_rsv = NULL; |
d68fc57b | 1172 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
1173 | |
1174 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1175 | INIT_LIST_HEAD(&root->root_list); |
2ab28f32 JB |
1176 | INIT_LIST_HEAD(&root->logged_list[0]); |
1177 | INIT_LIST_HEAD(&root->logged_list[1]); | |
d68fc57b | 1178 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 1179 | spin_lock_init(&root->inode_lock); |
f0486c68 | 1180 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1181 | spin_lock_init(&root->log_extents_lock[0]); |
1182 | spin_lock_init(&root->log_extents_lock[1]); | |
a2135011 | 1183 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1184 | mutex_init(&root->log_mutex); |
7237f183 YZ |
1185 | init_waitqueue_head(&root->log_writer_wait); |
1186 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1187 | init_waitqueue_head(&root->log_commit_wait[1]); | |
1188 | atomic_set(&root->log_commit[0], 0); | |
1189 | atomic_set(&root->log_commit[1], 0); | |
1190 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1191 | atomic_set(&root->log_batch, 0); |
8a35d95f | 1192 | atomic_set(&root->orphan_inodes, 0); |
7237f183 | 1193 | root->log_transid = 0; |
257c62e1 | 1194 | root->last_log_commit = 0; |
d0c803c4 | 1195 | extent_io_tree_init(&root->dirty_log_pages, |
f993c883 | 1196 | fs_info->btree_inode->i_mapping); |
017e5369 | 1197 | |
3768f368 CM |
1198 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1199 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1200 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 1201 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 1202 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 1203 | init_completion(&root->kobj_unregister); |
6702ed49 | 1204 | root->defrag_running = 0; |
4d775673 | 1205 | root->root_key.objectid = objectid; |
0ee5dc67 | 1206 | root->anon_dev = 0; |
8ea05e3a | 1207 | |
5f3ab90a | 1208 | spin_lock_init(&root->root_item_lock); |
3768f368 CM |
1209 | } |
1210 | ||
200a5c17 JM |
1211 | static int __must_check find_and_setup_root(struct btrfs_root *tree_root, |
1212 | struct btrfs_fs_info *fs_info, | |
1213 | u64 objectid, | |
1214 | struct btrfs_root *root) | |
3768f368 CM |
1215 | { |
1216 | int ret; | |
db94535d | 1217 | u32 blocksize; |
84234f3a | 1218 | u64 generation; |
3768f368 | 1219 | |
db94535d | 1220 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1221 | tree_root->sectorsize, tree_root->stripesize, |
1222 | root, fs_info, objectid); | |
3768f368 CM |
1223 | ret = btrfs_find_last_root(tree_root, objectid, |
1224 | &root->root_item, &root->root_key); | |
4df27c4d YZ |
1225 | if (ret > 0) |
1226 | return -ENOENT; | |
200a5c17 JM |
1227 | else if (ret < 0) |
1228 | return ret; | |
3768f368 | 1229 | |
84234f3a | 1230 | generation = btrfs_root_generation(&root->root_item); |
db94535d | 1231 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
af31f5e5 | 1232 | root->commit_root = NULL; |
db94535d | 1233 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), |
84234f3a | 1234 | blocksize, generation); |
b9fab919 | 1235 | if (!root->node || !btrfs_buffer_uptodate(root->node, generation, 0)) { |
68433b73 | 1236 | free_extent_buffer(root->node); |
af31f5e5 | 1237 | root->node = NULL; |
68433b73 CM |
1238 | return -EIO; |
1239 | } | |
4df27c4d | 1240 | root->commit_root = btrfs_root_node(root); |
d97e63b6 CM |
1241 | return 0; |
1242 | } | |
1243 | ||
f84a8bd6 | 1244 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info) |
6f07e42e AV |
1245 | { |
1246 | struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS); | |
1247 | if (root) | |
1248 | root->fs_info = fs_info; | |
1249 | return root; | |
1250 | } | |
1251 | ||
20897f5c AJ |
1252 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
1253 | struct btrfs_fs_info *fs_info, | |
1254 | u64 objectid) | |
1255 | { | |
1256 | struct extent_buffer *leaf; | |
1257 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1258 | struct btrfs_root *root; | |
1259 | struct btrfs_key key; | |
1260 | int ret = 0; | |
1261 | u64 bytenr; | |
1262 | ||
1263 | root = btrfs_alloc_root(fs_info); | |
1264 | if (!root) | |
1265 | return ERR_PTR(-ENOMEM); | |
1266 | ||
1267 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1268 | tree_root->sectorsize, tree_root->stripesize, | |
1269 | root, fs_info, objectid); | |
1270 | root->root_key.objectid = objectid; | |
1271 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1272 | root->root_key.offset = 0; | |
1273 | ||
1274 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, | |
1275 | 0, objectid, NULL, 0, 0, 0); | |
1276 | if (IS_ERR(leaf)) { | |
1277 | ret = PTR_ERR(leaf); | |
1278 | goto fail; | |
1279 | } | |
1280 | ||
1281 | bytenr = leaf->start; | |
1282 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); | |
1283 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1284 | btrfs_set_header_generation(leaf, trans->transid); | |
1285 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1286 | btrfs_set_header_owner(leaf, objectid); | |
1287 | root->node = leaf; | |
1288 | ||
1289 | write_extent_buffer(leaf, fs_info->fsid, | |
1290 | (unsigned long)btrfs_header_fsid(leaf), | |
1291 | BTRFS_FSID_SIZE); | |
1292 | write_extent_buffer(leaf, fs_info->chunk_tree_uuid, | |
1293 | (unsigned long)btrfs_header_chunk_tree_uuid(leaf), | |
1294 | BTRFS_UUID_SIZE); | |
1295 | btrfs_mark_buffer_dirty(leaf); | |
1296 | ||
1297 | root->commit_root = btrfs_root_node(root); | |
1298 | root->track_dirty = 1; | |
1299 | ||
1300 | ||
1301 | root->root_item.flags = 0; | |
1302 | root->root_item.byte_limit = 0; | |
1303 | btrfs_set_root_bytenr(&root->root_item, leaf->start); | |
1304 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1305 | btrfs_set_root_level(&root->root_item, 0); | |
1306 | btrfs_set_root_refs(&root->root_item, 1); | |
1307 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1308 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1309 | btrfs_set_root_dirid(&root->root_item, 0); | |
1310 | root->root_item.drop_level = 0; | |
1311 | ||
1312 | key.objectid = objectid; | |
1313 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1314 | key.offset = 0; | |
1315 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1316 | if (ret) | |
1317 | goto fail; | |
1318 | ||
1319 | btrfs_tree_unlock(leaf); | |
1320 | ||
1321 | fail: | |
1322 | if (ret) | |
1323 | return ERR_PTR(ret); | |
1324 | ||
1325 | return root; | |
1326 | } | |
1327 | ||
7237f183 YZ |
1328 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1329 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1330 | { |
1331 | struct btrfs_root *root; | |
1332 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1333 | struct extent_buffer *leaf; |
e02119d5 | 1334 | |
6f07e42e | 1335 | root = btrfs_alloc_root(fs_info); |
e02119d5 | 1336 | if (!root) |
7237f183 | 1337 | return ERR_PTR(-ENOMEM); |
e02119d5 CM |
1338 | |
1339 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1340 | tree_root->sectorsize, tree_root->stripesize, | |
1341 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1342 | ||
1343 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1344 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1345 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
7237f183 YZ |
1346 | /* |
1347 | * log trees do not get reference counted because they go away | |
1348 | * before a real commit is actually done. They do store pointers | |
1349 | * to file data extents, and those reference counts still get | |
1350 | * updated (along with back refs to the log tree). | |
1351 | */ | |
e02119d5 CM |
1352 | root->ref_cows = 0; |
1353 | ||
5d4f98a2 | 1354 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, |
66d7e7f0 | 1355 | BTRFS_TREE_LOG_OBJECTID, NULL, |
5581a51a | 1356 | 0, 0, 0); |
7237f183 YZ |
1357 | if (IS_ERR(leaf)) { |
1358 | kfree(root); | |
1359 | return ERR_CAST(leaf); | |
1360 | } | |
e02119d5 | 1361 | |
5d4f98a2 YZ |
1362 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1363 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1364 | btrfs_set_header_generation(leaf, trans->transid); | |
1365 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1366 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1367 | root->node = leaf; |
e02119d5 CM |
1368 | |
1369 | write_extent_buffer(root->node, root->fs_info->fsid, | |
1370 | (unsigned long)btrfs_header_fsid(root->node), | |
1371 | BTRFS_FSID_SIZE); | |
1372 | btrfs_mark_buffer_dirty(root->node); | |
1373 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1374 | return root; |
1375 | } | |
1376 | ||
1377 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1378 | struct btrfs_fs_info *fs_info) | |
1379 | { | |
1380 | struct btrfs_root *log_root; | |
1381 | ||
1382 | log_root = alloc_log_tree(trans, fs_info); | |
1383 | if (IS_ERR(log_root)) | |
1384 | return PTR_ERR(log_root); | |
1385 | WARN_ON(fs_info->log_root_tree); | |
1386 | fs_info->log_root_tree = log_root; | |
1387 | return 0; | |
1388 | } | |
1389 | ||
1390 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1391 | struct btrfs_root *root) | |
1392 | { | |
1393 | struct btrfs_root *log_root; | |
1394 | struct btrfs_inode_item *inode_item; | |
1395 | ||
1396 | log_root = alloc_log_tree(trans, root->fs_info); | |
1397 | if (IS_ERR(log_root)) | |
1398 | return PTR_ERR(log_root); | |
1399 | ||
1400 | log_root->last_trans = trans->transid; | |
1401 | log_root->root_key.offset = root->root_key.objectid; | |
1402 | ||
1403 | inode_item = &log_root->root_item.inode; | |
1404 | inode_item->generation = cpu_to_le64(1); | |
1405 | inode_item->size = cpu_to_le64(3); | |
1406 | inode_item->nlink = cpu_to_le32(1); | |
1407 | inode_item->nbytes = cpu_to_le64(root->leafsize); | |
1408 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); | |
1409 | ||
5d4f98a2 | 1410 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1411 | |
1412 | WARN_ON(root->log_root); | |
1413 | root->log_root = log_root; | |
1414 | root->log_transid = 0; | |
257c62e1 | 1415 | root->last_log_commit = 0; |
e02119d5 CM |
1416 | return 0; |
1417 | } | |
1418 | ||
1419 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
1420 | struct btrfs_key *location) | |
1421 | { | |
1422 | struct btrfs_root *root; | |
1423 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1424 | struct btrfs_path *path; |
5f39d397 | 1425 | struct extent_buffer *l; |
84234f3a | 1426 | u64 generation; |
db94535d | 1427 | u32 blocksize; |
0f7d52f4 | 1428 | int ret = 0; |
8ea05e3a | 1429 | int slot; |
0f7d52f4 | 1430 | |
6f07e42e | 1431 | root = btrfs_alloc_root(fs_info); |
0cf6c620 | 1432 | if (!root) |
0f7d52f4 | 1433 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 1434 | if (location->offset == (u64)-1) { |
db94535d | 1435 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
1436 | location->objectid, root); |
1437 | if (ret) { | |
0f7d52f4 CM |
1438 | kfree(root); |
1439 | return ERR_PTR(ret); | |
1440 | } | |
13a8a7c8 | 1441 | goto out; |
0f7d52f4 CM |
1442 | } |
1443 | ||
db94535d | 1444 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1445 | tree_root->sectorsize, tree_root->stripesize, |
1446 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
1447 | |
1448 | path = btrfs_alloc_path(); | |
db5b493a TI |
1449 | if (!path) { |
1450 | kfree(root); | |
1451 | return ERR_PTR(-ENOMEM); | |
1452 | } | |
0f7d52f4 | 1453 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); |
13a8a7c8 YZ |
1454 | if (ret == 0) { |
1455 | l = path->nodes[0]; | |
8ea05e3a AB |
1456 | slot = path->slots[0]; |
1457 | btrfs_read_root_item(tree_root, l, slot, &root->root_item); | |
13a8a7c8 | 1458 | memcpy(&root->root_key, location, sizeof(*location)); |
0f7d52f4 | 1459 | } |
0f7d52f4 CM |
1460 | btrfs_free_path(path); |
1461 | if (ret) { | |
5e540f77 | 1462 | kfree(root); |
13a8a7c8 YZ |
1463 | if (ret > 0) |
1464 | ret = -ENOENT; | |
0f7d52f4 CM |
1465 | return ERR_PTR(ret); |
1466 | } | |
13a8a7c8 | 1467 | |
84234f3a | 1468 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1469 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1470 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1471 | blocksize, generation); |
5d4f98a2 | 1472 | root->commit_root = btrfs_root_node(root); |
79787eaa | 1473 | BUG_ON(!root->node); /* -ENOMEM */ |
13a8a7c8 | 1474 | out: |
08fe4db1 | 1475 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { |
e02119d5 | 1476 | root->ref_cows = 1; |
08fe4db1 LZ |
1477 | btrfs_check_and_init_root_item(&root->root_item); |
1478 | } | |
13a8a7c8 | 1479 | |
5eda7b5e CM |
1480 | return root; |
1481 | } | |
1482 | ||
edbd8d4e CM |
1483 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
1484 | struct btrfs_key *location) | |
5eda7b5e CM |
1485 | { |
1486 | struct btrfs_root *root; | |
1487 | int ret; | |
1488 | ||
edbd8d4e CM |
1489 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1490 | return fs_info->tree_root; | |
1491 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1492 | return fs_info->extent_root; | |
8f18cf13 CM |
1493 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1494 | return fs_info->chunk_root; | |
1495 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1496 | return fs_info->dev_root; | |
0403e47e YZ |
1497 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1498 | return fs_info->csum_root; | |
bcef60f2 AJ |
1499 | if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1500 | return fs_info->quota_root ? fs_info->quota_root : | |
1501 | ERR_PTR(-ENOENT); | |
4df27c4d YZ |
1502 | again: |
1503 | spin_lock(&fs_info->fs_roots_radix_lock); | |
5eda7b5e CM |
1504 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
1505 | (unsigned long)location->objectid); | |
4df27c4d | 1506 | spin_unlock(&fs_info->fs_roots_radix_lock); |
5eda7b5e CM |
1507 | if (root) |
1508 | return root; | |
1509 | ||
e02119d5 | 1510 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
1511 | if (IS_ERR(root)) |
1512 | return root; | |
3394e160 | 1513 | |
581bb050 | 1514 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); |
581bb050 LZ |
1515 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), |
1516 | GFP_NOFS); | |
35a30d7c DS |
1517 | if (!root->free_ino_pinned || !root->free_ino_ctl) { |
1518 | ret = -ENOMEM; | |
581bb050 | 1519 | goto fail; |
35a30d7c | 1520 | } |
581bb050 LZ |
1521 | |
1522 | btrfs_init_free_ino_ctl(root); | |
1523 | mutex_init(&root->fs_commit_mutex); | |
1524 | spin_lock_init(&root->cache_lock); | |
1525 | init_waitqueue_head(&root->cache_wait); | |
1526 | ||
0ee5dc67 | 1527 | ret = get_anon_bdev(&root->anon_dev); |
ac08aedf CM |
1528 | if (ret) |
1529 | goto fail; | |
3394e160 | 1530 | |
d68fc57b YZ |
1531 | if (btrfs_root_refs(&root->root_item) == 0) { |
1532 | ret = -ENOENT; | |
1533 | goto fail; | |
1534 | } | |
1535 | ||
1536 | ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid); | |
1537 | if (ret < 0) | |
1538 | goto fail; | |
1539 | if (ret == 0) | |
1540 | root->orphan_item_inserted = 1; | |
1541 | ||
4df27c4d YZ |
1542 | ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); |
1543 | if (ret) | |
1544 | goto fail; | |
1545 | ||
1546 | spin_lock(&fs_info->fs_roots_radix_lock); | |
2619ba1f CM |
1547 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
1548 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 | 1549 | root); |
d68fc57b | 1550 | if (ret == 0) |
4df27c4d | 1551 | root->in_radix = 1; |
d68fc57b | 1552 | |
4df27c4d YZ |
1553 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1554 | radix_tree_preload_end(); | |
0f7d52f4 | 1555 | if (ret) { |
4df27c4d YZ |
1556 | if (ret == -EEXIST) { |
1557 | free_fs_root(root); | |
1558 | goto again; | |
1559 | } | |
1560 | goto fail; | |
0f7d52f4 | 1561 | } |
4df27c4d YZ |
1562 | |
1563 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
1564 | root->root_key.objectid); | |
1565 | WARN_ON(ret); | |
edbd8d4e | 1566 | return root; |
4df27c4d YZ |
1567 | fail: |
1568 | free_fs_root(root); | |
1569 | return ERR_PTR(ret); | |
edbd8d4e CM |
1570 | } |
1571 | ||
04160088 CM |
1572 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1573 | { | |
1574 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1575 | int ret = 0; | |
04160088 CM |
1576 | struct btrfs_device *device; |
1577 | struct backing_dev_info *bdi; | |
b7967db7 | 1578 | |
1f78160c XG |
1579 | rcu_read_lock(); |
1580 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1581 | if (!device->bdev) |
1582 | continue; | |
04160088 CM |
1583 | bdi = blk_get_backing_dev_info(device->bdev); |
1584 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1585 | ret = 1; | |
1586 | break; | |
1587 | } | |
1588 | } | |
1f78160c | 1589 | rcu_read_unlock(); |
04160088 CM |
1590 | return ret; |
1591 | } | |
1592 | ||
ad081f14 JA |
1593 | /* |
1594 | * If this fails, caller must call bdi_destroy() to get rid of the | |
1595 | * bdi again. | |
1596 | */ | |
04160088 CM |
1597 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1598 | { | |
ad081f14 JA |
1599 | int err; |
1600 | ||
1601 | bdi->capabilities = BDI_CAP_MAP_COPY; | |
e6d086d8 | 1602 | err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY); |
ad081f14 JA |
1603 | if (err) |
1604 | return err; | |
1605 | ||
4575c9cc | 1606 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1607 | bdi->congested_fn = btrfs_congested_fn; |
1608 | bdi->congested_data = info; | |
1609 | return 0; | |
1610 | } | |
1611 | ||
8b712842 CM |
1612 | /* |
1613 | * called by the kthread helper functions to finally call the bio end_io | |
1614 | * functions. This is where read checksum verification actually happens | |
1615 | */ | |
1616 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1617 | { |
ce9adaa5 | 1618 | struct bio *bio; |
8b712842 CM |
1619 | struct end_io_wq *end_io_wq; |
1620 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1621 | int error; |
ce9adaa5 | 1622 | |
8b712842 CM |
1623 | end_io_wq = container_of(work, struct end_io_wq, work); |
1624 | bio = end_io_wq->bio; | |
1625 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1626 | |
8b712842 CM |
1627 | error = end_io_wq->error; |
1628 | bio->bi_private = end_io_wq->private; | |
1629 | bio->bi_end_io = end_io_wq->end_io; | |
1630 | kfree(end_io_wq); | |
8b712842 | 1631 | bio_endio(bio, error); |
44b8bd7e CM |
1632 | } |
1633 | ||
a74a4b97 CM |
1634 | static int cleaner_kthread(void *arg) |
1635 | { | |
1636 | struct btrfs_root *root = arg; | |
1637 | ||
1638 | do { | |
76dda93c YZ |
1639 | if (!(root->fs_info->sb->s_flags & MS_RDONLY) && |
1640 | mutex_trylock(&root->fs_info->cleaner_mutex)) { | |
24bbcf04 | 1641 | btrfs_run_delayed_iputs(root); |
76dda93c YZ |
1642 | btrfs_clean_old_snapshots(root); |
1643 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
4cb5300b | 1644 | btrfs_run_defrag_inodes(root->fs_info); |
76dda93c | 1645 | } |
a74a4b97 | 1646 | |
a0acae0e | 1647 | if (!try_to_freeze()) { |
a74a4b97 | 1648 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1649 | if (!kthread_should_stop()) |
1650 | schedule(); | |
a74a4b97 CM |
1651 | __set_current_state(TASK_RUNNING); |
1652 | } | |
1653 | } while (!kthread_should_stop()); | |
1654 | return 0; | |
1655 | } | |
1656 | ||
1657 | static int transaction_kthread(void *arg) | |
1658 | { | |
1659 | struct btrfs_root *root = arg; | |
1660 | struct btrfs_trans_handle *trans; | |
1661 | struct btrfs_transaction *cur; | |
8929ecfa | 1662 | u64 transid; |
a74a4b97 CM |
1663 | unsigned long now; |
1664 | unsigned long delay; | |
914b2007 | 1665 | bool cannot_commit; |
a74a4b97 CM |
1666 | |
1667 | do { | |
914b2007 | 1668 | cannot_commit = false; |
a74a4b97 | 1669 | delay = HZ * 30; |
a74a4b97 CM |
1670 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
1671 | ||
a4abeea4 | 1672 | spin_lock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1673 | cur = root->fs_info->running_transaction; |
1674 | if (!cur) { | |
a4abeea4 | 1675 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1676 | goto sleep; |
1677 | } | |
31153d81 | 1678 | |
a74a4b97 | 1679 | now = get_seconds(); |
8929ecfa YZ |
1680 | if (!cur->blocked && |
1681 | (now < cur->start_time || now - cur->start_time < 30)) { | |
a4abeea4 | 1682 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1683 | delay = HZ * 5; |
1684 | goto sleep; | |
1685 | } | |
8929ecfa | 1686 | transid = cur->transid; |
a4abeea4 | 1687 | spin_unlock(&root->fs_info->trans_lock); |
56bec294 | 1688 | |
79787eaa | 1689 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1690 | trans = btrfs_attach_transaction(root); |
914b2007 | 1691 | if (IS_ERR(trans)) { |
354aa0fb MX |
1692 | if (PTR_ERR(trans) != -ENOENT) |
1693 | cannot_commit = true; | |
79787eaa | 1694 | goto sleep; |
914b2007 | 1695 | } |
8929ecfa | 1696 | if (transid == trans->transid) { |
79787eaa | 1697 | btrfs_commit_transaction(trans, root); |
8929ecfa YZ |
1698 | } else { |
1699 | btrfs_end_transaction(trans, root); | |
1700 | } | |
a74a4b97 CM |
1701 | sleep: |
1702 | wake_up_process(root->fs_info->cleaner_kthread); | |
1703 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1704 | ||
a0acae0e | 1705 | if (!try_to_freeze()) { |
a74a4b97 | 1706 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa | 1707 | if (!kthread_should_stop() && |
914b2007 JK |
1708 | (!btrfs_transaction_blocked(root->fs_info) || |
1709 | cannot_commit)) | |
8929ecfa | 1710 | schedule_timeout(delay); |
a74a4b97 CM |
1711 | __set_current_state(TASK_RUNNING); |
1712 | } | |
1713 | } while (!kthread_should_stop()); | |
1714 | return 0; | |
1715 | } | |
1716 | ||
af31f5e5 CM |
1717 | /* |
1718 | * this will find the highest generation in the array of | |
1719 | * root backups. The index of the highest array is returned, | |
1720 | * or -1 if we can't find anything. | |
1721 | * | |
1722 | * We check to make sure the array is valid by comparing the | |
1723 | * generation of the latest root in the array with the generation | |
1724 | * in the super block. If they don't match we pitch it. | |
1725 | */ | |
1726 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1727 | { | |
1728 | u64 cur; | |
1729 | int newest_index = -1; | |
1730 | struct btrfs_root_backup *root_backup; | |
1731 | int i; | |
1732 | ||
1733 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1734 | root_backup = info->super_copy->super_roots + i; | |
1735 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1736 | if (cur == newest_gen) | |
1737 | newest_index = i; | |
1738 | } | |
1739 | ||
1740 | /* check to see if we actually wrapped around */ | |
1741 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
1742 | root_backup = info->super_copy->super_roots; | |
1743 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1744 | if (cur == newest_gen) | |
1745 | newest_index = 0; | |
1746 | } | |
1747 | return newest_index; | |
1748 | } | |
1749 | ||
1750 | ||
1751 | /* | |
1752 | * find the oldest backup so we know where to store new entries | |
1753 | * in the backup array. This will set the backup_root_index | |
1754 | * field in the fs_info struct | |
1755 | */ | |
1756 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
1757 | u64 newest_gen) | |
1758 | { | |
1759 | int newest_index = -1; | |
1760 | ||
1761 | newest_index = find_newest_super_backup(info, newest_gen); | |
1762 | /* if there was garbage in there, just move along */ | |
1763 | if (newest_index == -1) { | |
1764 | info->backup_root_index = 0; | |
1765 | } else { | |
1766 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1767 | } | |
1768 | } | |
1769 | ||
1770 | /* | |
1771 | * copy all the root pointers into the super backup array. | |
1772 | * this will bump the backup pointer by one when it is | |
1773 | * done | |
1774 | */ | |
1775 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1776 | { | |
1777 | int next_backup; | |
1778 | struct btrfs_root_backup *root_backup; | |
1779 | int last_backup; | |
1780 | ||
1781 | next_backup = info->backup_root_index; | |
1782 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1783 | BTRFS_NUM_BACKUP_ROOTS; | |
1784 | ||
1785 | /* | |
1786 | * just overwrite the last backup if we're at the same generation | |
1787 | * this happens only at umount | |
1788 | */ | |
1789 | root_backup = info->super_for_commit->super_roots + last_backup; | |
1790 | if (btrfs_backup_tree_root_gen(root_backup) == | |
1791 | btrfs_header_generation(info->tree_root->node)) | |
1792 | next_backup = last_backup; | |
1793 | ||
1794 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1795 | ||
1796 | /* | |
1797 | * make sure all of our padding and empty slots get zero filled | |
1798 | * regardless of which ones we use today | |
1799 | */ | |
1800 | memset(root_backup, 0, sizeof(*root_backup)); | |
1801 | ||
1802 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1803 | ||
1804 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1805 | btrfs_set_backup_tree_root_gen(root_backup, | |
1806 | btrfs_header_generation(info->tree_root->node)); | |
1807 | ||
1808 | btrfs_set_backup_tree_root_level(root_backup, | |
1809 | btrfs_header_level(info->tree_root->node)); | |
1810 | ||
1811 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1812 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1813 | btrfs_header_generation(info->chunk_root->node)); | |
1814 | btrfs_set_backup_chunk_root_level(root_backup, | |
1815 | btrfs_header_level(info->chunk_root->node)); | |
1816 | ||
1817 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
1818 | btrfs_set_backup_extent_root_gen(root_backup, | |
1819 | btrfs_header_generation(info->extent_root->node)); | |
1820 | btrfs_set_backup_extent_root_level(root_backup, | |
1821 | btrfs_header_level(info->extent_root->node)); | |
1822 | ||
7c7e82a7 CM |
1823 | /* |
1824 | * we might commit during log recovery, which happens before we set | |
1825 | * the fs_root. Make sure it is valid before we fill it in. | |
1826 | */ | |
1827 | if (info->fs_root && info->fs_root->node) { | |
1828 | btrfs_set_backup_fs_root(root_backup, | |
1829 | info->fs_root->node->start); | |
1830 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1831 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1832 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1833 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1834 | } |
af31f5e5 CM |
1835 | |
1836 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1837 | btrfs_set_backup_dev_root_gen(root_backup, | |
1838 | btrfs_header_generation(info->dev_root->node)); | |
1839 | btrfs_set_backup_dev_root_level(root_backup, | |
1840 | btrfs_header_level(info->dev_root->node)); | |
1841 | ||
1842 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
1843 | btrfs_set_backup_csum_root_gen(root_backup, | |
1844 | btrfs_header_generation(info->csum_root->node)); | |
1845 | btrfs_set_backup_csum_root_level(root_backup, | |
1846 | btrfs_header_level(info->csum_root->node)); | |
1847 | ||
1848 | btrfs_set_backup_total_bytes(root_backup, | |
1849 | btrfs_super_total_bytes(info->super_copy)); | |
1850 | btrfs_set_backup_bytes_used(root_backup, | |
1851 | btrfs_super_bytes_used(info->super_copy)); | |
1852 | btrfs_set_backup_num_devices(root_backup, | |
1853 | btrfs_super_num_devices(info->super_copy)); | |
1854 | ||
1855 | /* | |
1856 | * if we don't copy this out to the super_copy, it won't get remembered | |
1857 | * for the next commit | |
1858 | */ | |
1859 | memcpy(&info->super_copy->super_roots, | |
1860 | &info->super_for_commit->super_roots, | |
1861 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1862 | } | |
1863 | ||
1864 | /* | |
1865 | * this copies info out of the root backup array and back into | |
1866 | * the in-memory super block. It is meant to help iterate through | |
1867 | * the array, so you send it the number of backups you've already | |
1868 | * tried and the last backup index you used. | |
1869 | * | |
1870 | * this returns -1 when it has tried all the backups | |
1871 | */ | |
1872 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
1873 | struct btrfs_super_block *super, | |
1874 | int *num_backups_tried, int *backup_index) | |
1875 | { | |
1876 | struct btrfs_root_backup *root_backup; | |
1877 | int newest = *backup_index; | |
1878 | ||
1879 | if (*num_backups_tried == 0) { | |
1880 | u64 gen = btrfs_super_generation(super); | |
1881 | ||
1882 | newest = find_newest_super_backup(info, gen); | |
1883 | if (newest == -1) | |
1884 | return -1; | |
1885 | ||
1886 | *backup_index = newest; | |
1887 | *num_backups_tried = 1; | |
1888 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
1889 | /* we've tried all the backups, all done */ | |
1890 | return -1; | |
1891 | } else { | |
1892 | /* jump to the next oldest backup */ | |
1893 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1894 | BTRFS_NUM_BACKUP_ROOTS; | |
1895 | *backup_index = newest; | |
1896 | *num_backups_tried += 1; | |
1897 | } | |
1898 | root_backup = super->super_roots + newest; | |
1899 | ||
1900 | btrfs_set_super_generation(super, | |
1901 | btrfs_backup_tree_root_gen(root_backup)); | |
1902 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
1903 | btrfs_set_super_root_level(super, | |
1904 | btrfs_backup_tree_root_level(root_backup)); | |
1905 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
1906 | ||
1907 | /* | |
1908 | * fixme: the total bytes and num_devices need to match or we should | |
1909 | * need a fsck | |
1910 | */ | |
1911 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
1912 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
1913 | return 0; | |
1914 | } | |
1915 | ||
1916 | /* helper to cleanup tree roots */ | |
1917 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
1918 | { | |
1919 | free_extent_buffer(info->tree_root->node); | |
1920 | free_extent_buffer(info->tree_root->commit_root); | |
1921 | free_extent_buffer(info->dev_root->node); | |
1922 | free_extent_buffer(info->dev_root->commit_root); | |
1923 | free_extent_buffer(info->extent_root->node); | |
1924 | free_extent_buffer(info->extent_root->commit_root); | |
1925 | free_extent_buffer(info->csum_root->node); | |
1926 | free_extent_buffer(info->csum_root->commit_root); | |
bcef60f2 AJ |
1927 | if (info->quota_root) { |
1928 | free_extent_buffer(info->quota_root->node); | |
1929 | free_extent_buffer(info->quota_root->commit_root); | |
1930 | } | |
af31f5e5 CM |
1931 | |
1932 | info->tree_root->node = NULL; | |
1933 | info->tree_root->commit_root = NULL; | |
1934 | info->dev_root->node = NULL; | |
1935 | info->dev_root->commit_root = NULL; | |
1936 | info->extent_root->node = NULL; | |
1937 | info->extent_root->commit_root = NULL; | |
1938 | info->csum_root->node = NULL; | |
1939 | info->csum_root->commit_root = NULL; | |
bcef60f2 AJ |
1940 | if (info->quota_root) { |
1941 | info->quota_root->node = NULL; | |
1942 | info->quota_root->commit_root = NULL; | |
1943 | } | |
af31f5e5 CM |
1944 | |
1945 | if (chunk_root) { | |
1946 | free_extent_buffer(info->chunk_root->node); | |
1947 | free_extent_buffer(info->chunk_root->commit_root); | |
1948 | info->chunk_root->node = NULL; | |
1949 | info->chunk_root->commit_root = NULL; | |
1950 | } | |
1951 | } | |
1952 | ||
1953 | ||
ad2b2c80 AV |
1954 | int open_ctree(struct super_block *sb, |
1955 | struct btrfs_fs_devices *fs_devices, | |
1956 | char *options) | |
2e635a27 | 1957 | { |
db94535d CM |
1958 | u32 sectorsize; |
1959 | u32 nodesize; | |
1960 | u32 leafsize; | |
1961 | u32 blocksize; | |
87ee04eb | 1962 | u32 stripesize; |
84234f3a | 1963 | u64 generation; |
f2b636e8 | 1964 | u64 features; |
3de4586c | 1965 | struct btrfs_key location; |
a061fc8d | 1966 | struct buffer_head *bh; |
4d34b278 | 1967 | struct btrfs_super_block *disk_super; |
815745cf | 1968 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
f84a8bd6 | 1969 | struct btrfs_root *tree_root; |
4d34b278 ID |
1970 | struct btrfs_root *extent_root; |
1971 | struct btrfs_root *csum_root; | |
1972 | struct btrfs_root *chunk_root; | |
1973 | struct btrfs_root *dev_root; | |
bcef60f2 | 1974 | struct btrfs_root *quota_root; |
e02119d5 | 1975 | struct btrfs_root *log_tree_root; |
eb60ceac | 1976 | int ret; |
e58ca020 | 1977 | int err = -EINVAL; |
af31f5e5 CM |
1978 | int num_backups_tried = 0; |
1979 | int backup_index = 0; | |
4543df7e | 1980 | |
f84a8bd6 | 1981 | tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info); |
6f07e42e AV |
1982 | extent_root = fs_info->extent_root = btrfs_alloc_root(fs_info); |
1983 | csum_root = fs_info->csum_root = btrfs_alloc_root(fs_info); | |
1984 | chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info); | |
1985 | dev_root = fs_info->dev_root = btrfs_alloc_root(fs_info); | |
bcef60f2 | 1986 | quota_root = fs_info->quota_root = btrfs_alloc_root(fs_info); |
8790d502 | 1987 | |
f84a8bd6 | 1988 | if (!tree_root || !extent_root || !csum_root || |
bcef60f2 | 1989 | !chunk_root || !dev_root || !quota_root) { |
39279cc3 CM |
1990 | err = -ENOMEM; |
1991 | goto fail; | |
1992 | } | |
76dda93c YZ |
1993 | |
1994 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
1995 | if (ret) { | |
1996 | err = ret; | |
1997 | goto fail; | |
1998 | } | |
1999 | ||
2000 | ret = setup_bdi(fs_info, &fs_info->bdi); | |
2001 | if (ret) { | |
2002 | err = ret; | |
2003 | goto fail_srcu; | |
2004 | } | |
2005 | ||
e2d84521 MX |
2006 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0); |
2007 | if (ret) { | |
2008 | err = ret; | |
2009 | goto fail_bdi; | |
2010 | } | |
2011 | fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE * | |
2012 | (1 + ilog2(nr_cpu_ids)); | |
2013 | ||
963d678b MX |
2014 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0); |
2015 | if (ret) { | |
2016 | err = ret; | |
2017 | goto fail_dirty_metadata_bytes; | |
2018 | } | |
2019 | ||
76dda93c YZ |
2020 | fs_info->btree_inode = new_inode(sb); |
2021 | if (!fs_info->btree_inode) { | |
2022 | err = -ENOMEM; | |
963d678b | 2023 | goto fail_delalloc_bytes; |
76dda93c YZ |
2024 | } |
2025 | ||
a6591715 | 2026 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
1561deda | 2027 | |
76dda93c | 2028 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
8fd17795 | 2029 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2030 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2031 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
ea8c2819 | 2032 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
11833d66 | 2033 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
1832a6d5 | 2034 | spin_lock_init(&fs_info->delalloc_lock); |
a4abeea4 | 2035 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 2036 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2037 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2038 | spin_lock_init(&fs_info->defrag_inodes_lock); |
2bf64758 | 2039 | spin_lock_init(&fs_info->free_chunk_lock); |
f29021b2 JS |
2040 | spin_lock_init(&fs_info->tree_mod_seq_lock); |
2041 | rwlock_init(&fs_info->tree_mod_log_lock); | |
7585717f | 2042 | mutex_init(&fs_info->reloc_mutex); |
de98ced9 | 2043 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2044 | |
58176a96 | 2045 | init_completion(&fs_info->kobj_unregister); |
0b86a832 | 2046 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2047 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2048 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
0b86a832 | 2049 | btrfs_mapping_init(&fs_info->mapping_tree); |
66d8f3dd MX |
2050 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2051 | BTRFS_BLOCK_RSV_GLOBAL); | |
2052 | btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, | |
2053 | BTRFS_BLOCK_RSV_DELALLOC); | |
2054 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); | |
2055 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2056 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2057 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2058 | BTRFS_BLOCK_RSV_DELOPS); | |
cb03c743 | 2059 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 2060 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 2061 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 2062 | atomic_set(&fs_info->nr_async_bios, 0); |
4cb5300b | 2063 | atomic_set(&fs_info->defrag_running, 0); |
f29021b2 | 2064 | atomic_set(&fs_info->tree_mod_seq, 0); |
e20d96d6 | 2065 | fs_info->sb = sb; |
6f568d35 | 2066 | fs_info->max_inline = 8192 * 1024; |
9ed74f2d | 2067 | fs_info->metadata_ratio = 0; |
4cb5300b | 2068 | fs_info->defrag_inodes = RB_ROOT; |
a4abeea4 | 2069 | fs_info->trans_no_join = 0; |
2bf64758 | 2070 | fs_info->free_chunk_space = 0; |
f29021b2 | 2071 | fs_info->tree_mod_log = RB_ROOT; |
c8b97818 | 2072 | |
90519d66 AJ |
2073 | /* readahead state */ |
2074 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); | |
2075 | spin_lock_init(&fs_info->reada_lock); | |
c8b97818 | 2076 | |
b34b086c CM |
2077 | fs_info->thread_pool_size = min_t(unsigned long, |
2078 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2079 | |
3eaa2885 CM |
2080 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
2081 | spin_lock_init(&fs_info->ordered_extent_lock); | |
16cdcec7 MX |
2082 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
2083 | GFP_NOFS); | |
2084 | if (!fs_info->delayed_root) { | |
2085 | err = -ENOMEM; | |
2086 | goto fail_iput; | |
2087 | } | |
2088 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 2089 | |
a2de733c AJ |
2090 | mutex_init(&fs_info->scrub_lock); |
2091 | atomic_set(&fs_info->scrubs_running, 0); | |
2092 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2093 | atomic_set(&fs_info->scrubs_paused, 0); | |
2094 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2095 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
2096 | init_rwsem(&fs_info->scrub_super_lock); | |
2097 | fs_info->scrub_workers_refcnt = 0; | |
21adbd5c SB |
2098 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2099 | fs_info->check_integrity_print_mask = 0; | |
2100 | #endif | |
a2de733c | 2101 | |
c9e9f97b ID |
2102 | spin_lock_init(&fs_info->balance_lock); |
2103 | mutex_init(&fs_info->balance_mutex); | |
837d5b6e ID |
2104 | atomic_set(&fs_info->balance_running, 0); |
2105 | atomic_set(&fs_info->balance_pause_req, 0); | |
a7e99c69 | 2106 | atomic_set(&fs_info->balance_cancel_req, 0); |
c9e9f97b | 2107 | fs_info->balance_ctl = NULL; |
837d5b6e | 2108 | init_waitqueue_head(&fs_info->balance_wait_q); |
a2de733c | 2109 | |
a061fc8d CM |
2110 | sb->s_blocksize = 4096; |
2111 | sb->s_blocksize_bits = blksize_bits(4096); | |
32a88aa1 | 2112 | sb->s_bdi = &fs_info->bdi; |
a061fc8d | 2113 | |
76dda93c | 2114 | fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; |
bfe86848 | 2115 | set_nlink(fs_info->btree_inode, 1); |
0afbaf8c CM |
2116 | /* |
2117 | * we set the i_size on the btree inode to the max possible int. | |
2118 | * the real end of the address space is determined by all of | |
2119 | * the devices in the system | |
2120 | */ | |
2121 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 2122 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
2123 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
2124 | ||
5d4f98a2 | 2125 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); |
d1310b2e | 2126 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
f993c883 | 2127 | fs_info->btree_inode->i_mapping); |
0b32f4bb | 2128 | BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; |
a8067e02 | 2129 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); |
d1310b2e CM |
2130 | |
2131 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 2132 | |
76dda93c YZ |
2133 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
2134 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
2135 | sizeof(struct btrfs_key)); | |
72ac3c0d JB |
2136 | set_bit(BTRFS_INODE_DUMMY, |
2137 | &BTRFS_I(fs_info->btree_inode)->runtime_flags); | |
c65ddb52 | 2138 | insert_inode_hash(fs_info->btree_inode); |
76dda93c | 2139 | |
0f9dd46c | 2140 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2141 | fs_info->block_group_cache_tree = RB_ROOT; |
a1897fdd | 2142 | fs_info->first_logical_byte = (u64)-1; |
0f9dd46c | 2143 | |
11833d66 | 2144 | extent_io_tree_init(&fs_info->freed_extents[0], |
f993c883 | 2145 | fs_info->btree_inode->i_mapping); |
11833d66 | 2146 | extent_io_tree_init(&fs_info->freed_extents[1], |
f993c883 | 2147 | fs_info->btree_inode->i_mapping); |
11833d66 | 2148 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
e66f709b | 2149 | fs_info->do_barriers = 1; |
e18e4809 | 2150 | |
39279cc3 | 2151 | |
5a3f23d5 | 2152 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2153 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2154 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2155 | mutex_init(&fs_info->transaction_kthread_mutex); |
2156 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 2157 | mutex_init(&fs_info->volume_mutex); |
276e680d | 2158 | init_rwsem(&fs_info->extent_commit_sem); |
c71bf099 | 2159 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2160 | init_rwsem(&fs_info->subvol_sem); |
e922e087 SB |
2161 | fs_info->dev_replace.lock_owner = 0; |
2162 | atomic_set(&fs_info->dev_replace.nesting_level, 0); | |
2163 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); | |
2164 | mutex_init(&fs_info->dev_replace.lock_management_lock); | |
2165 | mutex_init(&fs_info->dev_replace.lock); | |
fa9c0d79 | 2166 | |
416ac51d AJ |
2167 | spin_lock_init(&fs_info->qgroup_lock); |
2168 | fs_info->qgroup_tree = RB_ROOT; | |
2169 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); | |
2170 | fs_info->qgroup_seq = 1; | |
2171 | fs_info->quota_enabled = 0; | |
2172 | fs_info->pending_quota_state = 0; | |
2173 | ||
fa9c0d79 CM |
2174 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2175 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2176 | ||
e6dcd2dc | 2177 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2178 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2179 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2180 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2181 | |
0b86a832 | 2182 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 2183 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2184 | |
3c4bb26b | 2185 | invalidate_bdev(fs_devices->latest_bdev); |
a512bbf8 | 2186 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
20b45077 DY |
2187 | if (!bh) { |
2188 | err = -EINVAL; | |
16cdcec7 | 2189 | goto fail_alloc; |
20b45077 | 2190 | } |
39279cc3 | 2191 | |
6c41761f DS |
2192 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2193 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2194 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2195 | brelse(bh); |
5f39d397 | 2196 | |
6c41761f | 2197 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2198 | |
6c41761f | 2199 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2200 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2201 | goto fail_alloc; |
0f7d52f4 | 2202 | |
acce952b | 2203 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
2204 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
2205 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 2206 | |
fcd1f065 DS |
2207 | ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); |
2208 | if (ret) { | |
2209 | printk(KERN_ERR "btrfs: superblock contains fatal errors\n"); | |
2210 | err = ret; | |
2211 | goto fail_alloc; | |
2212 | } | |
acce952b | 2213 | |
af31f5e5 CM |
2214 | /* |
2215 | * run through our array of backup supers and setup | |
2216 | * our ring pointer to the oldest one | |
2217 | */ | |
2218 | generation = btrfs_super_generation(disk_super); | |
2219 | find_oldest_super_backup(fs_info, generation); | |
2220 | ||
75e7cb7f LB |
2221 | /* |
2222 | * In the long term, we'll store the compression type in the super | |
2223 | * block, and it'll be used for per file compression control. | |
2224 | */ | |
2225 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2226 | ||
2b82032c YZ |
2227 | ret = btrfs_parse_options(tree_root, options); |
2228 | if (ret) { | |
2229 | err = ret; | |
16cdcec7 | 2230 | goto fail_alloc; |
2b82032c | 2231 | } |
dfe25020 | 2232 | |
f2b636e8 JB |
2233 | features = btrfs_super_incompat_flags(disk_super) & |
2234 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2235 | if (features) { | |
2236 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
2237 | "unsupported optional features (%Lx).\n", | |
21380931 | 2238 | (unsigned long long)features); |
f2b636e8 | 2239 | err = -EINVAL; |
16cdcec7 | 2240 | goto fail_alloc; |
f2b636e8 JB |
2241 | } |
2242 | ||
727011e0 CM |
2243 | if (btrfs_super_leafsize(disk_super) != |
2244 | btrfs_super_nodesize(disk_super)) { | |
2245 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2246 | "blocksizes don't match. node %d leaf %d\n", | |
2247 | btrfs_super_nodesize(disk_super), | |
2248 | btrfs_super_leafsize(disk_super)); | |
2249 | err = -EINVAL; | |
2250 | goto fail_alloc; | |
2251 | } | |
2252 | if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2253 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2254 | "blocksize (%d) was too large\n", | |
2255 | btrfs_super_leafsize(disk_super)); | |
2256 | err = -EINVAL; | |
2257 | goto fail_alloc; | |
2258 | } | |
2259 | ||
5d4f98a2 | 2260 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 2261 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
69e380d1 | 2262 | if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 2263 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
727011e0 CM |
2264 | |
2265 | /* | |
2266 | * flag our filesystem as having big metadata blocks if | |
2267 | * they are bigger than the page size | |
2268 | */ | |
2269 | if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) { | |
2270 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) | |
2271 | printk(KERN_INFO "btrfs flagging fs with big metadata feature\n"); | |
2272 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
2273 | } | |
2274 | ||
bc3f116f CM |
2275 | nodesize = btrfs_super_nodesize(disk_super); |
2276 | leafsize = btrfs_super_leafsize(disk_super); | |
2277 | sectorsize = btrfs_super_sectorsize(disk_super); | |
2278 | stripesize = btrfs_super_stripesize(disk_super); | |
e2d84521 | 2279 | fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids)); |
963d678b | 2280 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); |
bc3f116f CM |
2281 | |
2282 | /* | |
2283 | * mixed block groups end up with duplicate but slightly offset | |
2284 | * extent buffers for the same range. It leads to corruptions | |
2285 | */ | |
2286 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
2287 | (sectorsize != leafsize)) { | |
2288 | printk(KERN_WARNING "btrfs: unequal leaf/node/sector sizes " | |
2289 | "are not allowed for mixed block groups on %s\n", | |
2290 | sb->s_id); | |
2291 | goto fail_alloc; | |
2292 | } | |
2293 | ||
a6fa6fae | 2294 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 2295 | |
f2b636e8 JB |
2296 | features = btrfs_super_compat_ro_flags(disk_super) & |
2297 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
2298 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
2299 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
2300 | "unsupported option features (%Lx).\n", | |
21380931 | 2301 | (unsigned long long)features); |
f2b636e8 | 2302 | err = -EINVAL; |
16cdcec7 | 2303 | goto fail_alloc; |
f2b636e8 | 2304 | } |
61d92c32 CM |
2305 | |
2306 | btrfs_init_workers(&fs_info->generic_worker, | |
2307 | "genwork", 1, NULL); | |
2308 | ||
5443be45 | 2309 | btrfs_init_workers(&fs_info->workers, "worker", |
61d92c32 CM |
2310 | fs_info->thread_pool_size, |
2311 | &fs_info->generic_worker); | |
c8b97818 | 2312 | |
771ed689 | 2313 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
61d92c32 CM |
2314 | fs_info->thread_pool_size, |
2315 | &fs_info->generic_worker); | |
771ed689 | 2316 | |
8ccf6f19 MX |
2317 | btrfs_init_workers(&fs_info->flush_workers, "flush_delalloc", |
2318 | fs_info->thread_pool_size, | |
2319 | &fs_info->generic_worker); | |
2320 | ||
5443be45 | 2321 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 | 2322 | min_t(u64, fs_devices->num_devices, |
61d92c32 CM |
2323 | fs_info->thread_pool_size), |
2324 | &fs_info->generic_worker); | |
61b49440 | 2325 | |
bab39bf9 JB |
2326 | btrfs_init_workers(&fs_info->caching_workers, "cache", |
2327 | 2, &fs_info->generic_worker); | |
2328 | ||
61b49440 CM |
2329 | /* a higher idle thresh on the submit workers makes it much more |
2330 | * likely that bios will be send down in a sane order to the | |
2331 | * devices | |
2332 | */ | |
2333 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 2334 | |
771ed689 | 2335 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 2336 | fs_info->workers.ordered = 1; |
61b49440 | 2337 | |
771ed689 CM |
2338 | fs_info->delalloc_workers.idle_thresh = 2; |
2339 | fs_info->delalloc_workers.ordered = 1; | |
2340 | ||
61d92c32 CM |
2341 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1, |
2342 | &fs_info->generic_worker); | |
5443be45 | 2343 | btrfs_init_workers(&fs_info->endio_workers, "endio", |
61d92c32 CM |
2344 | fs_info->thread_pool_size, |
2345 | &fs_info->generic_worker); | |
d20f7043 | 2346 | btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", |
61d92c32 CM |
2347 | fs_info->thread_pool_size, |
2348 | &fs_info->generic_worker); | |
cad321ad | 2349 | btrfs_init_workers(&fs_info->endio_meta_write_workers, |
61d92c32 CM |
2350 | "endio-meta-write", fs_info->thread_pool_size, |
2351 | &fs_info->generic_worker); | |
5443be45 | 2352 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", |
61d92c32 CM |
2353 | fs_info->thread_pool_size, |
2354 | &fs_info->generic_worker); | |
0cb59c99 JB |
2355 | btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write", |
2356 | 1, &fs_info->generic_worker); | |
16cdcec7 MX |
2357 | btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta", |
2358 | fs_info->thread_pool_size, | |
2359 | &fs_info->generic_worker); | |
90519d66 AJ |
2360 | btrfs_init_workers(&fs_info->readahead_workers, "readahead", |
2361 | fs_info->thread_pool_size, | |
2362 | &fs_info->generic_worker); | |
61b49440 CM |
2363 | |
2364 | /* | |
2365 | * endios are largely parallel and should have a very | |
2366 | * low idle thresh | |
2367 | */ | |
2368 | fs_info->endio_workers.idle_thresh = 4; | |
b51912c9 CM |
2369 | fs_info->endio_meta_workers.idle_thresh = 4; |
2370 | ||
9042846b CM |
2371 | fs_info->endio_write_workers.idle_thresh = 2; |
2372 | fs_info->endio_meta_write_workers.idle_thresh = 2; | |
90519d66 | 2373 | fs_info->readahead_workers.idle_thresh = 2; |
9042846b | 2374 | |
0dc3b84a JB |
2375 | /* |
2376 | * btrfs_start_workers can really only fail because of ENOMEM so just | |
2377 | * return -ENOMEM if any of these fail. | |
2378 | */ | |
2379 | ret = btrfs_start_workers(&fs_info->workers); | |
2380 | ret |= btrfs_start_workers(&fs_info->generic_worker); | |
2381 | ret |= btrfs_start_workers(&fs_info->submit_workers); | |
2382 | ret |= btrfs_start_workers(&fs_info->delalloc_workers); | |
2383 | ret |= btrfs_start_workers(&fs_info->fixup_workers); | |
2384 | ret |= btrfs_start_workers(&fs_info->endio_workers); | |
2385 | ret |= btrfs_start_workers(&fs_info->endio_meta_workers); | |
2386 | ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers); | |
2387 | ret |= btrfs_start_workers(&fs_info->endio_write_workers); | |
2388 | ret |= btrfs_start_workers(&fs_info->endio_freespace_worker); | |
2389 | ret |= btrfs_start_workers(&fs_info->delayed_workers); | |
2390 | ret |= btrfs_start_workers(&fs_info->caching_workers); | |
2391 | ret |= btrfs_start_workers(&fs_info->readahead_workers); | |
8ccf6f19 | 2392 | ret |= btrfs_start_workers(&fs_info->flush_workers); |
0dc3b84a | 2393 | if (ret) { |
fed425c7 | 2394 | err = -ENOMEM; |
0dc3b84a JB |
2395 | goto fail_sb_buffer; |
2396 | } | |
4543df7e | 2397 | |
4575c9cc | 2398 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
2399 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
2400 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 2401 | |
db94535d CM |
2402 | tree_root->nodesize = nodesize; |
2403 | tree_root->leafsize = leafsize; | |
2404 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 2405 | tree_root->stripesize = stripesize; |
a061fc8d CM |
2406 | |
2407 | sb->s_blocksize = sectorsize; | |
2408 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 2409 | |
39279cc3 CM |
2410 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
2411 | sizeof(disk_super->magic))) { | |
d397712b | 2412 | printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
2413 | goto fail_sb_buffer; |
2414 | } | |
19c00ddc | 2415 | |
8d082fb7 LB |
2416 | if (sectorsize != PAGE_SIZE) { |
2417 | printk(KERN_WARNING "btrfs: Incompatible sector size(%lu) " | |
2418 | "found on %s\n", (unsigned long)sectorsize, sb->s_id); | |
941b2ddf KM |
2419 | goto fail_sb_buffer; |
2420 | } | |
2421 | ||
925baedd | 2422 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 2423 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 2424 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2425 | if (ret) { |
d397712b CM |
2426 | printk(KERN_WARNING "btrfs: failed to read the system " |
2427 | "array on %s\n", sb->s_id); | |
5d4f98a2 | 2428 | goto fail_sb_buffer; |
84eed90f | 2429 | } |
0b86a832 CM |
2430 | |
2431 | blocksize = btrfs_level_size(tree_root, | |
2432 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 2433 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
2434 | |
2435 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2436 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
2437 | ||
2438 | chunk_root->node = read_tree_block(chunk_root, | |
2439 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 2440 | blocksize, generation); |
79787eaa | 2441 | BUG_ON(!chunk_root->node); /* -ENOMEM */ |
83121942 DW |
2442 | if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { |
2443 | printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n", | |
2444 | sb->s_id); | |
af31f5e5 | 2445 | goto fail_tree_roots; |
83121942 | 2446 | } |
5d4f98a2 YZ |
2447 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
2448 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 2449 | |
e17cade2 | 2450 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
d397712b CM |
2451 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), |
2452 | BTRFS_UUID_SIZE); | |
e17cade2 | 2453 | |
0b86a832 | 2454 | ret = btrfs_read_chunk_tree(chunk_root); |
2b82032c | 2455 | if (ret) { |
d397712b CM |
2456 | printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", |
2457 | sb->s_id); | |
af31f5e5 | 2458 | goto fail_tree_roots; |
2b82032c | 2459 | } |
0b86a832 | 2460 | |
8dabb742 SB |
2461 | /* |
2462 | * keep the device that is marked to be the target device for the | |
2463 | * dev_replace procedure | |
2464 | */ | |
2465 | btrfs_close_extra_devices(fs_info, fs_devices, 0); | |
dfe25020 | 2466 | |
a6b0d5c8 CM |
2467 | if (!fs_devices->latest_bdev) { |
2468 | printk(KERN_CRIT "btrfs: failed to read devices on %s\n", | |
2469 | sb->s_id); | |
2470 | goto fail_tree_roots; | |
2471 | } | |
2472 | ||
af31f5e5 | 2473 | retry_root_backup: |
db94535d CM |
2474 | blocksize = btrfs_level_size(tree_root, |
2475 | btrfs_super_root_level(disk_super)); | |
84234f3a | 2476 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 2477 | |
e20d96d6 | 2478 | tree_root->node = read_tree_block(tree_root, |
db94535d | 2479 | btrfs_super_root(disk_super), |
84234f3a | 2480 | blocksize, generation); |
af31f5e5 CM |
2481 | if (!tree_root->node || |
2482 | !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { | |
83121942 DW |
2483 | printk(KERN_WARNING "btrfs: failed to read tree root on %s\n", |
2484 | sb->s_id); | |
af31f5e5 CM |
2485 | |
2486 | goto recovery_tree_root; | |
83121942 | 2487 | } |
af31f5e5 | 2488 | |
5d4f98a2 YZ |
2489 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
2490 | tree_root->commit_root = btrfs_root_node(tree_root); | |
db94535d CM |
2491 | |
2492 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 2493 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 2494 | if (ret) |
af31f5e5 | 2495 | goto recovery_tree_root; |
0b86a832 CM |
2496 | extent_root->track_dirty = 1; |
2497 | ||
2498 | ret = find_and_setup_root(tree_root, fs_info, | |
2499 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
0b86a832 | 2500 | if (ret) |
af31f5e5 | 2501 | goto recovery_tree_root; |
5d4f98a2 | 2502 | dev_root->track_dirty = 1; |
3768f368 | 2503 | |
d20f7043 CM |
2504 | ret = find_and_setup_root(tree_root, fs_info, |
2505 | BTRFS_CSUM_TREE_OBJECTID, csum_root); | |
2506 | if (ret) | |
af31f5e5 | 2507 | goto recovery_tree_root; |
d20f7043 CM |
2508 | csum_root->track_dirty = 1; |
2509 | ||
bcef60f2 AJ |
2510 | ret = find_and_setup_root(tree_root, fs_info, |
2511 | BTRFS_QUOTA_TREE_OBJECTID, quota_root); | |
2512 | if (ret) { | |
2513 | kfree(quota_root); | |
2514 | quota_root = fs_info->quota_root = NULL; | |
2515 | } else { | |
2516 | quota_root->track_dirty = 1; | |
2517 | fs_info->quota_enabled = 1; | |
2518 | fs_info->pending_quota_state = 1; | |
2519 | } | |
2520 | ||
8929ecfa YZ |
2521 | fs_info->generation = generation; |
2522 | fs_info->last_trans_committed = generation; | |
8929ecfa | 2523 | |
68310a5e ID |
2524 | ret = btrfs_recover_balance(fs_info); |
2525 | if (ret) { | |
2526 | printk(KERN_WARNING "btrfs: failed to recover balance\n"); | |
2527 | goto fail_block_groups; | |
2528 | } | |
2529 | ||
733f4fbb SB |
2530 | ret = btrfs_init_dev_stats(fs_info); |
2531 | if (ret) { | |
2532 | printk(KERN_ERR "btrfs: failed to init dev_stats: %d\n", | |
2533 | ret); | |
2534 | goto fail_block_groups; | |
2535 | } | |
2536 | ||
8dabb742 SB |
2537 | ret = btrfs_init_dev_replace(fs_info); |
2538 | if (ret) { | |
2539 | pr_err("btrfs: failed to init dev_replace: %d\n", ret); | |
2540 | goto fail_block_groups; | |
2541 | } | |
2542 | ||
2543 | btrfs_close_extra_devices(fs_info, fs_devices, 1); | |
2544 | ||
c59021f8 | 2545 | ret = btrfs_init_space_info(fs_info); |
2546 | if (ret) { | |
2547 | printk(KERN_ERR "Failed to initial space info: %d\n", ret); | |
2548 | goto fail_block_groups; | |
2549 | } | |
2550 | ||
1b1d1f66 JB |
2551 | ret = btrfs_read_block_groups(extent_root); |
2552 | if (ret) { | |
2553 | printk(KERN_ERR "Failed to read block groups: %d\n", ret); | |
2554 | goto fail_block_groups; | |
2555 | } | |
5af3e8cc SB |
2556 | fs_info->num_tolerated_disk_barrier_failures = |
2557 | btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); | |
292fd7fc SB |
2558 | if (fs_info->fs_devices->missing_devices > |
2559 | fs_info->num_tolerated_disk_barrier_failures && | |
2560 | !(sb->s_flags & MS_RDONLY)) { | |
2561 | printk(KERN_WARNING | |
2562 | "Btrfs: too many missing devices, writeable mount is not allowed\n"); | |
2563 | goto fail_block_groups; | |
2564 | } | |
9078a3e1 | 2565 | |
a74a4b97 CM |
2566 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
2567 | "btrfs-cleaner"); | |
57506d50 | 2568 | if (IS_ERR(fs_info->cleaner_kthread)) |
1b1d1f66 | 2569 | goto fail_block_groups; |
a74a4b97 CM |
2570 | |
2571 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
2572 | tree_root, | |
2573 | "btrfs-transaction"); | |
57506d50 | 2574 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 2575 | goto fail_cleaner; |
a74a4b97 | 2576 | |
c289811c CM |
2577 | if (!btrfs_test_opt(tree_root, SSD) && |
2578 | !btrfs_test_opt(tree_root, NOSSD) && | |
2579 | !fs_info->fs_devices->rotating) { | |
2580 | printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD " | |
2581 | "mode\n"); | |
2582 | btrfs_set_opt(fs_info->mount_opt, SSD); | |
2583 | } | |
2584 | ||
21adbd5c SB |
2585 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2586 | if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { | |
2587 | ret = btrfsic_mount(tree_root, fs_devices, | |
2588 | btrfs_test_opt(tree_root, | |
2589 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? | |
2590 | 1 : 0, | |
2591 | fs_info->check_integrity_print_mask); | |
2592 | if (ret) | |
2593 | printk(KERN_WARNING "btrfs: failed to initialize" | |
2594 | " integrity check module %s\n", sb->s_id); | |
2595 | } | |
2596 | #endif | |
bcef60f2 AJ |
2597 | ret = btrfs_read_qgroup_config(fs_info); |
2598 | if (ret) | |
2599 | goto fail_trans_kthread; | |
21adbd5c | 2600 | |
acce952b | 2601 | /* do not make disk changes in broken FS */ |
68ce9682 | 2602 | if (btrfs_super_log_root(disk_super) != 0) { |
e02119d5 CM |
2603 | u64 bytenr = btrfs_super_log_root(disk_super); |
2604 | ||
7c2ca468 | 2605 | if (fs_devices->rw_devices == 0) { |
d397712b CM |
2606 | printk(KERN_WARNING "Btrfs log replay required " |
2607 | "on RO media\n"); | |
7c2ca468 | 2608 | err = -EIO; |
bcef60f2 | 2609 | goto fail_qgroup; |
7c2ca468 | 2610 | } |
e02119d5 CM |
2611 | blocksize = |
2612 | btrfs_level_size(tree_root, | |
2613 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 2614 | |
6f07e42e | 2615 | log_tree_root = btrfs_alloc_root(fs_info); |
676e4c86 DC |
2616 | if (!log_tree_root) { |
2617 | err = -ENOMEM; | |
bcef60f2 | 2618 | goto fail_qgroup; |
676e4c86 | 2619 | } |
e02119d5 CM |
2620 | |
2621 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2622 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
2623 | ||
2624 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
2625 | blocksize, |
2626 | generation + 1); | |
79787eaa | 2627 | /* returns with log_tree_root freed on success */ |
e02119d5 | 2628 | ret = btrfs_recover_log_trees(log_tree_root); |
79787eaa JM |
2629 | if (ret) { |
2630 | btrfs_error(tree_root->fs_info, ret, | |
2631 | "Failed to recover log tree"); | |
2632 | free_extent_buffer(log_tree_root->node); | |
2633 | kfree(log_tree_root); | |
2634 | goto fail_trans_kthread; | |
2635 | } | |
e556ce2c YZ |
2636 | |
2637 | if (sb->s_flags & MS_RDONLY) { | |
79787eaa JM |
2638 | ret = btrfs_commit_super(tree_root); |
2639 | if (ret) | |
2640 | goto fail_trans_kthread; | |
e556ce2c | 2641 | } |
e02119d5 | 2642 | } |
1a40e23b | 2643 | |
76dda93c | 2644 | ret = btrfs_find_orphan_roots(tree_root); |
79787eaa JM |
2645 | if (ret) |
2646 | goto fail_trans_kthread; | |
76dda93c | 2647 | |
7c2ca468 | 2648 | if (!(sb->s_flags & MS_RDONLY)) { |
d68fc57b | 2649 | ret = btrfs_cleanup_fs_roots(fs_info); |
44c44af2 ID |
2650 | if (ret) |
2651 | goto fail_trans_kthread; | |
d68fc57b | 2652 | |
5d4f98a2 | 2653 | ret = btrfs_recover_relocation(tree_root); |
d7ce5843 MX |
2654 | if (ret < 0) { |
2655 | printk(KERN_WARNING | |
2656 | "btrfs: failed to recover relocation\n"); | |
2657 | err = -EINVAL; | |
bcef60f2 | 2658 | goto fail_qgroup; |
d7ce5843 | 2659 | } |
7c2ca468 | 2660 | } |
1a40e23b | 2661 | |
3de4586c CM |
2662 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
2663 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2664 | location.offset = (u64)-1; | |
2665 | ||
3de4586c CM |
2666 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
2667 | if (!fs_info->fs_root) | |
bcef60f2 | 2668 | goto fail_qgroup; |
3140c9a3 DC |
2669 | if (IS_ERR(fs_info->fs_root)) { |
2670 | err = PTR_ERR(fs_info->fs_root); | |
bcef60f2 | 2671 | goto fail_qgroup; |
3140c9a3 | 2672 | } |
c289811c | 2673 | |
2b6ba629 ID |
2674 | if (sb->s_flags & MS_RDONLY) |
2675 | return 0; | |
59641015 | 2676 | |
2b6ba629 ID |
2677 | down_read(&fs_info->cleanup_work_sem); |
2678 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
2679 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
e3acc2a6 | 2680 | up_read(&fs_info->cleanup_work_sem); |
2b6ba629 ID |
2681 | close_ctree(tree_root); |
2682 | return ret; | |
2683 | } | |
2684 | up_read(&fs_info->cleanup_work_sem); | |
59641015 | 2685 | |
2b6ba629 ID |
2686 | ret = btrfs_resume_balance_async(fs_info); |
2687 | if (ret) { | |
2688 | printk(KERN_WARNING "btrfs: failed to resume balance\n"); | |
2689 | close_ctree(tree_root); | |
2690 | return ret; | |
e3acc2a6 JB |
2691 | } |
2692 | ||
8dabb742 SB |
2693 | ret = btrfs_resume_dev_replace_async(fs_info); |
2694 | if (ret) { | |
2695 | pr_warn("btrfs: failed to resume dev_replace\n"); | |
2696 | close_ctree(tree_root); | |
2697 | return ret; | |
2698 | } | |
2699 | ||
ad2b2c80 | 2700 | return 0; |
39279cc3 | 2701 | |
bcef60f2 AJ |
2702 | fail_qgroup: |
2703 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
2704 | fail_trans_kthread: |
2705 | kthread_stop(fs_info->transaction_kthread); | |
3f157a2f | 2706 | fail_cleaner: |
a74a4b97 | 2707 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
2708 | |
2709 | /* | |
2710 | * make sure we're done with the btree inode before we stop our | |
2711 | * kthreads | |
2712 | */ | |
2713 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 2714 | |
1b1d1f66 JB |
2715 | fail_block_groups: |
2716 | btrfs_free_block_groups(fs_info); | |
af31f5e5 CM |
2717 | |
2718 | fail_tree_roots: | |
2719 | free_root_pointers(fs_info, 1); | |
2b8195bb | 2720 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 2721 | |
39279cc3 | 2722 | fail_sb_buffer: |
61d92c32 | 2723 | btrfs_stop_workers(&fs_info->generic_worker); |
306c8b68 | 2724 | btrfs_stop_workers(&fs_info->readahead_workers); |
247e743c | 2725 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2726 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2727 | btrfs_stop_workers(&fs_info->workers); |
2728 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 2729 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 2730 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 2731 | btrfs_stop_workers(&fs_info->endio_write_workers); |
0cb59c99 | 2732 | btrfs_stop_workers(&fs_info->endio_freespace_worker); |
1cc127b5 | 2733 | btrfs_stop_workers(&fs_info->submit_workers); |
16cdcec7 | 2734 | btrfs_stop_workers(&fs_info->delayed_workers); |
bab39bf9 | 2735 | btrfs_stop_workers(&fs_info->caching_workers); |
8ccf6f19 | 2736 | btrfs_stop_workers(&fs_info->flush_workers); |
16cdcec7 | 2737 | fail_alloc: |
4543df7e | 2738 | fail_iput: |
586e46e2 ID |
2739 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
2740 | ||
4543df7e | 2741 | iput(fs_info->btree_inode); |
963d678b MX |
2742 | fail_delalloc_bytes: |
2743 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
e2d84521 MX |
2744 | fail_dirty_metadata_bytes: |
2745 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); | |
ad081f14 | 2746 | fail_bdi: |
7e662854 | 2747 | bdi_destroy(&fs_info->bdi); |
76dda93c YZ |
2748 | fail_srcu: |
2749 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 2750 | fail: |
586e46e2 | 2751 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 2752 | return err; |
af31f5e5 CM |
2753 | |
2754 | recovery_tree_root: | |
af31f5e5 CM |
2755 | if (!btrfs_test_opt(tree_root, RECOVERY)) |
2756 | goto fail_tree_roots; | |
2757 | ||
2758 | free_root_pointers(fs_info, 0); | |
2759 | ||
2760 | /* don't use the log in recovery mode, it won't be valid */ | |
2761 | btrfs_set_super_log_root(disk_super, 0); | |
2762 | ||
2763 | /* we can't trust the free space cache either */ | |
2764 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2765 | ||
2766 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
2767 | &num_backups_tried, &backup_index); | |
2768 | if (ret == -1) | |
2769 | goto fail_block_groups; | |
2770 | goto retry_root_backup; | |
eb60ceac CM |
2771 | } |
2772 | ||
f2984462 CM |
2773 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
2774 | { | |
f2984462 CM |
2775 | if (uptodate) { |
2776 | set_buffer_uptodate(bh); | |
2777 | } else { | |
442a4f63 SB |
2778 | struct btrfs_device *device = (struct btrfs_device *) |
2779 | bh->b_private; | |
2780 | ||
606686ee JB |
2781 | printk_ratelimited_in_rcu(KERN_WARNING "lost page write due to " |
2782 | "I/O error on %s\n", | |
2783 | rcu_str_deref(device->name)); | |
1259ab75 CM |
2784 | /* note, we dont' set_buffer_write_io_error because we have |
2785 | * our own ways of dealing with the IO errors | |
2786 | */ | |
f2984462 | 2787 | clear_buffer_uptodate(bh); |
442a4f63 | 2788 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); |
f2984462 CM |
2789 | } |
2790 | unlock_buffer(bh); | |
2791 | put_bh(bh); | |
2792 | } | |
2793 | ||
a512bbf8 YZ |
2794 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
2795 | { | |
2796 | struct buffer_head *bh; | |
2797 | struct buffer_head *latest = NULL; | |
2798 | struct btrfs_super_block *super; | |
2799 | int i; | |
2800 | u64 transid = 0; | |
2801 | u64 bytenr; | |
2802 | ||
2803 | /* we would like to check all the supers, but that would make | |
2804 | * a btrfs mount succeed after a mkfs from a different FS. | |
2805 | * So, we need to add a special mount option to scan for | |
2806 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
2807 | */ | |
2808 | for (i = 0; i < 1; i++) { | |
2809 | bytenr = btrfs_sb_offset(i); | |
2810 | if (bytenr + 4096 >= i_size_read(bdev->bd_inode)) | |
2811 | break; | |
2812 | bh = __bread(bdev, bytenr / 4096, 4096); | |
2813 | if (!bh) | |
2814 | continue; | |
2815 | ||
2816 | super = (struct btrfs_super_block *)bh->b_data; | |
2817 | if (btrfs_super_bytenr(super) != bytenr || | |
2818 | strncmp((char *)(&super->magic), BTRFS_MAGIC, | |
2819 | sizeof(super->magic))) { | |
2820 | brelse(bh); | |
2821 | continue; | |
2822 | } | |
2823 | ||
2824 | if (!latest || btrfs_super_generation(super) > transid) { | |
2825 | brelse(latest); | |
2826 | latest = bh; | |
2827 | transid = btrfs_super_generation(super); | |
2828 | } else { | |
2829 | brelse(bh); | |
2830 | } | |
2831 | } | |
2832 | return latest; | |
2833 | } | |
2834 | ||
4eedeb75 HH |
2835 | /* |
2836 | * this should be called twice, once with wait == 0 and | |
2837 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
2838 | * we write are pinned. | |
2839 | * | |
2840 | * They are released when wait == 1 is done. | |
2841 | * max_mirrors must be the same for both runs, and it indicates how | |
2842 | * many supers on this one device should be written. | |
2843 | * | |
2844 | * max_mirrors == 0 means to write them all. | |
2845 | */ | |
a512bbf8 YZ |
2846 | static int write_dev_supers(struct btrfs_device *device, |
2847 | struct btrfs_super_block *sb, | |
2848 | int do_barriers, int wait, int max_mirrors) | |
2849 | { | |
2850 | struct buffer_head *bh; | |
2851 | int i; | |
2852 | int ret; | |
2853 | int errors = 0; | |
2854 | u32 crc; | |
2855 | u64 bytenr; | |
a512bbf8 YZ |
2856 | |
2857 | if (max_mirrors == 0) | |
2858 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
2859 | ||
a512bbf8 YZ |
2860 | for (i = 0; i < max_mirrors; i++) { |
2861 | bytenr = btrfs_sb_offset(i); | |
2862 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) | |
2863 | break; | |
2864 | ||
2865 | if (wait) { | |
2866 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
2867 | BTRFS_SUPER_INFO_SIZE); | |
2868 | BUG_ON(!bh); | |
a512bbf8 | 2869 | wait_on_buffer(bh); |
4eedeb75 HH |
2870 | if (!buffer_uptodate(bh)) |
2871 | errors++; | |
2872 | ||
2873 | /* drop our reference */ | |
2874 | brelse(bh); | |
2875 | ||
2876 | /* drop the reference from the wait == 0 run */ | |
2877 | brelse(bh); | |
2878 | continue; | |
a512bbf8 YZ |
2879 | } else { |
2880 | btrfs_set_super_bytenr(sb, bytenr); | |
2881 | ||
2882 | crc = ~(u32)0; | |
2883 | crc = btrfs_csum_data(NULL, (char *)sb + | |
2884 | BTRFS_CSUM_SIZE, crc, | |
2885 | BTRFS_SUPER_INFO_SIZE - | |
2886 | BTRFS_CSUM_SIZE); | |
2887 | btrfs_csum_final(crc, sb->csum); | |
2888 | ||
4eedeb75 HH |
2889 | /* |
2890 | * one reference for us, and we leave it for the | |
2891 | * caller | |
2892 | */ | |
a512bbf8 YZ |
2893 | bh = __getblk(device->bdev, bytenr / 4096, |
2894 | BTRFS_SUPER_INFO_SIZE); | |
2895 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); | |
2896 | ||
4eedeb75 | 2897 | /* one reference for submit_bh */ |
a512bbf8 | 2898 | get_bh(bh); |
4eedeb75 HH |
2899 | |
2900 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
2901 | lock_buffer(bh); |
2902 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
442a4f63 | 2903 | bh->b_private = device; |
a512bbf8 YZ |
2904 | } |
2905 | ||
387125fc CM |
2906 | /* |
2907 | * we fua the first super. The others we allow | |
2908 | * to go down lazy. | |
2909 | */ | |
21adbd5c | 2910 | ret = btrfsic_submit_bh(WRITE_FUA, bh); |
4eedeb75 | 2911 | if (ret) |
a512bbf8 | 2912 | errors++; |
a512bbf8 YZ |
2913 | } |
2914 | return errors < i ? 0 : -1; | |
2915 | } | |
2916 | ||
387125fc CM |
2917 | /* |
2918 | * endio for the write_dev_flush, this will wake anyone waiting | |
2919 | * for the barrier when it is done | |
2920 | */ | |
2921 | static void btrfs_end_empty_barrier(struct bio *bio, int err) | |
2922 | { | |
2923 | if (err) { | |
2924 | if (err == -EOPNOTSUPP) | |
2925 | set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); | |
2926 | clear_bit(BIO_UPTODATE, &bio->bi_flags); | |
2927 | } | |
2928 | if (bio->bi_private) | |
2929 | complete(bio->bi_private); | |
2930 | bio_put(bio); | |
2931 | } | |
2932 | ||
2933 | /* | |
2934 | * trigger flushes for one the devices. If you pass wait == 0, the flushes are | |
2935 | * sent down. With wait == 1, it waits for the previous flush. | |
2936 | * | |
2937 | * any device where the flush fails with eopnotsupp are flagged as not-barrier | |
2938 | * capable | |
2939 | */ | |
2940 | static int write_dev_flush(struct btrfs_device *device, int wait) | |
2941 | { | |
2942 | struct bio *bio; | |
2943 | int ret = 0; | |
2944 | ||
2945 | if (device->nobarriers) | |
2946 | return 0; | |
2947 | ||
2948 | if (wait) { | |
2949 | bio = device->flush_bio; | |
2950 | if (!bio) | |
2951 | return 0; | |
2952 | ||
2953 | wait_for_completion(&device->flush_wait); | |
2954 | ||
2955 | if (bio_flagged(bio, BIO_EOPNOTSUPP)) { | |
606686ee JB |
2956 | printk_in_rcu("btrfs: disabling barriers on dev %s\n", |
2957 | rcu_str_deref(device->name)); | |
387125fc | 2958 | device->nobarriers = 1; |
5af3e8cc | 2959 | } else if (!bio_flagged(bio, BIO_UPTODATE)) { |
387125fc | 2960 | ret = -EIO; |
5af3e8cc SB |
2961 | btrfs_dev_stat_inc_and_print(device, |
2962 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
387125fc CM |
2963 | } |
2964 | ||
2965 | /* drop the reference from the wait == 0 run */ | |
2966 | bio_put(bio); | |
2967 | device->flush_bio = NULL; | |
2968 | ||
2969 | return ret; | |
2970 | } | |
2971 | ||
2972 | /* | |
2973 | * one reference for us, and we leave it for the | |
2974 | * caller | |
2975 | */ | |
9c017abc | 2976 | device->flush_bio = NULL; |
387125fc CM |
2977 | bio = bio_alloc(GFP_NOFS, 0); |
2978 | if (!bio) | |
2979 | return -ENOMEM; | |
2980 | ||
2981 | bio->bi_end_io = btrfs_end_empty_barrier; | |
2982 | bio->bi_bdev = device->bdev; | |
2983 | init_completion(&device->flush_wait); | |
2984 | bio->bi_private = &device->flush_wait; | |
2985 | device->flush_bio = bio; | |
2986 | ||
2987 | bio_get(bio); | |
21adbd5c | 2988 | btrfsic_submit_bio(WRITE_FLUSH, bio); |
387125fc CM |
2989 | |
2990 | return 0; | |
2991 | } | |
2992 | ||
2993 | /* | |
2994 | * send an empty flush down to each device in parallel, | |
2995 | * then wait for them | |
2996 | */ | |
2997 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
2998 | { | |
2999 | struct list_head *head; | |
3000 | struct btrfs_device *dev; | |
5af3e8cc SB |
3001 | int errors_send = 0; |
3002 | int errors_wait = 0; | |
387125fc CM |
3003 | int ret; |
3004 | ||
3005 | /* send down all the barriers */ | |
3006 | head = &info->fs_devices->devices; | |
3007 | list_for_each_entry_rcu(dev, head, dev_list) { | |
3008 | if (!dev->bdev) { | |
5af3e8cc | 3009 | errors_send++; |
387125fc CM |
3010 | continue; |
3011 | } | |
3012 | if (!dev->in_fs_metadata || !dev->writeable) | |
3013 | continue; | |
3014 | ||
3015 | ret = write_dev_flush(dev, 0); | |
3016 | if (ret) | |
5af3e8cc | 3017 | errors_send++; |
387125fc CM |
3018 | } |
3019 | ||
3020 | /* wait for all the barriers */ | |
3021 | list_for_each_entry_rcu(dev, head, dev_list) { | |
3022 | if (!dev->bdev) { | |
5af3e8cc | 3023 | errors_wait++; |
387125fc CM |
3024 | continue; |
3025 | } | |
3026 | if (!dev->in_fs_metadata || !dev->writeable) | |
3027 | continue; | |
3028 | ||
3029 | ret = write_dev_flush(dev, 1); | |
3030 | if (ret) | |
5af3e8cc | 3031 | errors_wait++; |
387125fc | 3032 | } |
5af3e8cc SB |
3033 | if (errors_send > info->num_tolerated_disk_barrier_failures || |
3034 | errors_wait > info->num_tolerated_disk_barrier_failures) | |
387125fc CM |
3035 | return -EIO; |
3036 | return 0; | |
3037 | } | |
3038 | ||
5af3e8cc SB |
3039 | int btrfs_calc_num_tolerated_disk_barrier_failures( |
3040 | struct btrfs_fs_info *fs_info) | |
3041 | { | |
3042 | struct btrfs_ioctl_space_info space; | |
3043 | struct btrfs_space_info *sinfo; | |
3044 | u64 types[] = {BTRFS_BLOCK_GROUP_DATA, | |
3045 | BTRFS_BLOCK_GROUP_SYSTEM, | |
3046 | BTRFS_BLOCK_GROUP_METADATA, | |
3047 | BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; | |
3048 | int num_types = 4; | |
3049 | int i; | |
3050 | int c; | |
3051 | int num_tolerated_disk_barrier_failures = | |
3052 | (int)fs_info->fs_devices->num_devices; | |
3053 | ||
3054 | for (i = 0; i < num_types; i++) { | |
3055 | struct btrfs_space_info *tmp; | |
3056 | ||
3057 | sinfo = NULL; | |
3058 | rcu_read_lock(); | |
3059 | list_for_each_entry_rcu(tmp, &fs_info->space_info, list) { | |
3060 | if (tmp->flags == types[i]) { | |
3061 | sinfo = tmp; | |
3062 | break; | |
3063 | } | |
3064 | } | |
3065 | rcu_read_unlock(); | |
3066 | ||
3067 | if (!sinfo) | |
3068 | continue; | |
3069 | ||
3070 | down_read(&sinfo->groups_sem); | |
3071 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
3072 | if (!list_empty(&sinfo->block_groups[c])) { | |
3073 | u64 flags; | |
3074 | ||
3075 | btrfs_get_block_group_info( | |
3076 | &sinfo->block_groups[c], &space); | |
3077 | if (space.total_bytes == 0 || | |
3078 | space.used_bytes == 0) | |
3079 | continue; | |
3080 | flags = space.flags; | |
3081 | /* | |
3082 | * return | |
3083 | * 0: if dup, single or RAID0 is configured for | |
3084 | * any of metadata, system or data, else | |
3085 | * 1: if RAID5 is configured, or if RAID1 or | |
3086 | * RAID10 is configured and only two mirrors | |
3087 | * are used, else | |
3088 | * 2: if RAID6 is configured, else | |
3089 | * num_mirrors - 1: if RAID1 or RAID10 is | |
3090 | * configured and more than | |
3091 | * 2 mirrors are used. | |
3092 | */ | |
3093 | if (num_tolerated_disk_barrier_failures > 0 && | |
3094 | ((flags & (BTRFS_BLOCK_GROUP_DUP | | |
3095 | BTRFS_BLOCK_GROUP_RAID0)) || | |
3096 | ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) | |
3097 | == 0))) | |
3098 | num_tolerated_disk_barrier_failures = 0; | |
3099 | else if (num_tolerated_disk_barrier_failures > 1 | |
3100 | && | |
3101 | (flags & (BTRFS_BLOCK_GROUP_RAID1 | | |
3102 | BTRFS_BLOCK_GROUP_RAID10))) | |
3103 | num_tolerated_disk_barrier_failures = 1; | |
3104 | } | |
3105 | } | |
3106 | up_read(&sinfo->groups_sem); | |
3107 | } | |
3108 | ||
3109 | return num_tolerated_disk_barrier_failures; | |
3110 | } | |
3111 | ||
a512bbf8 | 3112 | int write_all_supers(struct btrfs_root *root, int max_mirrors) |
f2984462 | 3113 | { |
e5e9a520 | 3114 | struct list_head *head; |
f2984462 | 3115 | struct btrfs_device *dev; |
a061fc8d | 3116 | struct btrfs_super_block *sb; |
f2984462 | 3117 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
3118 | int ret; |
3119 | int do_barriers; | |
a236aed1 CM |
3120 | int max_errors; |
3121 | int total_errors = 0; | |
a061fc8d | 3122 | u64 flags; |
f2984462 | 3123 | |
6c41761f | 3124 | max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; |
f2984462 | 3125 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
af31f5e5 | 3126 | backup_super_roots(root->fs_info); |
f2984462 | 3127 | |
6c41761f | 3128 | sb = root->fs_info->super_for_commit; |
a061fc8d | 3129 | dev_item = &sb->dev_item; |
e5e9a520 | 3130 | |
174ba509 | 3131 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); |
e5e9a520 | 3132 | head = &root->fs_info->fs_devices->devices; |
387125fc | 3133 | |
5af3e8cc SB |
3134 | if (do_barriers) { |
3135 | ret = barrier_all_devices(root->fs_info); | |
3136 | if (ret) { | |
3137 | mutex_unlock( | |
3138 | &root->fs_info->fs_devices->device_list_mutex); | |
3139 | btrfs_error(root->fs_info, ret, | |
3140 | "errors while submitting device barriers."); | |
3141 | return ret; | |
3142 | } | |
3143 | } | |
387125fc | 3144 | |
1f78160c | 3145 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3146 | if (!dev->bdev) { |
3147 | total_errors++; | |
3148 | continue; | |
3149 | } | |
2b82032c | 3150 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3151 | continue; |
3152 | ||
2b82032c | 3153 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
3154 | btrfs_set_stack_device_type(dev_item, dev->type); |
3155 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
3156 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
3157 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
3158 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
3159 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
3160 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
3161 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 3162 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 3163 | |
a061fc8d CM |
3164 | flags = btrfs_super_flags(sb); |
3165 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
3166 | ||
a512bbf8 | 3167 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
3168 | if (ret) |
3169 | total_errors++; | |
f2984462 | 3170 | } |
a236aed1 | 3171 | if (total_errors > max_errors) { |
d397712b CM |
3172 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
3173 | total_errors); | |
79787eaa JM |
3174 | |
3175 | /* This shouldn't happen. FUA is masked off if unsupported */ | |
a236aed1 CM |
3176 | BUG(); |
3177 | } | |
f2984462 | 3178 | |
a512bbf8 | 3179 | total_errors = 0; |
1f78160c | 3180 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3181 | if (!dev->bdev) |
3182 | continue; | |
2b82032c | 3183 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3184 | continue; |
3185 | ||
a512bbf8 YZ |
3186 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
3187 | if (ret) | |
3188 | total_errors++; | |
f2984462 | 3189 | } |
174ba509 | 3190 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 3191 | if (total_errors > max_errors) { |
79787eaa JM |
3192 | btrfs_error(root->fs_info, -EIO, |
3193 | "%d errors while writing supers", total_errors); | |
3194 | return -EIO; | |
a236aed1 | 3195 | } |
f2984462 CM |
3196 | return 0; |
3197 | } | |
3198 | ||
a512bbf8 YZ |
3199 | int write_ctree_super(struct btrfs_trans_handle *trans, |
3200 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 3201 | { |
e66f709b | 3202 | int ret; |
5f39d397 | 3203 | |
a512bbf8 | 3204 | ret = write_all_supers(root, max_mirrors); |
5f39d397 | 3205 | return ret; |
cfaa7295 CM |
3206 | } |
3207 | ||
143bede5 | 3208 | void btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f | 3209 | { |
4df27c4d | 3210 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
3211 | radix_tree_delete(&fs_info->fs_roots_radix, |
3212 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 3213 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
3214 | |
3215 | if (btrfs_root_refs(&root->root_item) == 0) | |
3216 | synchronize_srcu(&fs_info->subvol_srcu); | |
3217 | ||
581bb050 LZ |
3218 | __btrfs_remove_free_space_cache(root->free_ino_pinned); |
3219 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d | 3220 | free_fs_root(root); |
4df27c4d YZ |
3221 | } |
3222 | ||
3223 | static void free_fs_root(struct btrfs_root *root) | |
3224 | { | |
82d5902d | 3225 | iput(root->cache_inode); |
4df27c4d | 3226 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
0ee5dc67 AV |
3227 | if (root->anon_dev) |
3228 | free_anon_bdev(root->anon_dev); | |
4df27c4d YZ |
3229 | free_extent_buffer(root->node); |
3230 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
3231 | kfree(root->free_ino_ctl); |
3232 | kfree(root->free_ino_pinned); | |
d397712b | 3233 | kfree(root->name); |
2619ba1f | 3234 | kfree(root); |
2619ba1f CM |
3235 | } |
3236 | ||
143bede5 | 3237 | static void del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
3238 | { |
3239 | int ret; | |
3240 | struct btrfs_root *gang[8]; | |
3241 | int i; | |
3242 | ||
76dda93c YZ |
3243 | while (!list_empty(&fs_info->dead_roots)) { |
3244 | gang[0] = list_entry(fs_info->dead_roots.next, | |
3245 | struct btrfs_root, root_list); | |
3246 | list_del(&gang[0]->root_list); | |
3247 | ||
3248 | if (gang[0]->in_radix) { | |
3249 | btrfs_free_fs_root(fs_info, gang[0]); | |
3250 | } else { | |
3251 | free_extent_buffer(gang[0]->node); | |
3252 | free_extent_buffer(gang[0]->commit_root); | |
3253 | kfree(gang[0]); | |
3254 | } | |
3255 | } | |
3256 | ||
d397712b | 3257 | while (1) { |
0f7d52f4 CM |
3258 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
3259 | (void **)gang, 0, | |
3260 | ARRAY_SIZE(gang)); | |
3261 | if (!ret) | |
3262 | break; | |
2619ba1f | 3263 | for (i = 0; i < ret; i++) |
5eda7b5e | 3264 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 | 3265 | } |
0f7d52f4 | 3266 | } |
b4100d64 | 3267 | |
c146afad | 3268 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 3269 | { |
c146afad YZ |
3270 | u64 root_objectid = 0; |
3271 | struct btrfs_root *gang[8]; | |
3272 | int i; | |
3768f368 | 3273 | int ret; |
e089f05c | 3274 | |
c146afad YZ |
3275 | while (1) { |
3276 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
3277 | (void **)gang, root_objectid, | |
3278 | ARRAY_SIZE(gang)); | |
3279 | if (!ret) | |
3280 | break; | |
5d4f98a2 YZ |
3281 | |
3282 | root_objectid = gang[ret - 1]->root_key.objectid + 1; | |
c146afad | 3283 | for (i = 0; i < ret; i++) { |
66b4ffd1 JB |
3284 | int err; |
3285 | ||
c146afad | 3286 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
3287 | err = btrfs_orphan_cleanup(gang[i]); |
3288 | if (err) | |
3289 | return err; | |
c146afad YZ |
3290 | } |
3291 | root_objectid++; | |
3292 | } | |
3293 | return 0; | |
3294 | } | |
a2135011 | 3295 | |
c146afad YZ |
3296 | int btrfs_commit_super(struct btrfs_root *root) |
3297 | { | |
3298 | struct btrfs_trans_handle *trans; | |
3299 | int ret; | |
a74a4b97 | 3300 | |
c146afad | 3301 | mutex_lock(&root->fs_info->cleaner_mutex); |
24bbcf04 | 3302 | btrfs_run_delayed_iputs(root); |
a74a4b97 | 3303 | btrfs_clean_old_snapshots(root); |
c146afad | 3304 | mutex_unlock(&root->fs_info->cleaner_mutex); |
c71bf099 YZ |
3305 | |
3306 | /* wait until ongoing cleanup work done */ | |
3307 | down_write(&root->fs_info->cleanup_work_sem); | |
3308 | up_write(&root->fs_info->cleanup_work_sem); | |
3309 | ||
7a7eaa40 | 3310 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3311 | if (IS_ERR(trans)) |
3312 | return PTR_ERR(trans); | |
54aa1f4d | 3313 | ret = btrfs_commit_transaction(trans, root); |
79787eaa JM |
3314 | if (ret) |
3315 | return ret; | |
c146afad | 3316 | /* run commit again to drop the original snapshot */ |
7a7eaa40 | 3317 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3318 | if (IS_ERR(trans)) |
3319 | return PTR_ERR(trans); | |
79787eaa JM |
3320 | ret = btrfs_commit_transaction(trans, root); |
3321 | if (ret) | |
3322 | return ret; | |
79154b1b | 3323 | ret = btrfs_write_and_wait_transaction(NULL, root); |
79787eaa JM |
3324 | if (ret) { |
3325 | btrfs_error(root->fs_info, ret, | |
3326 | "Failed to sync btree inode to disk."); | |
3327 | return ret; | |
3328 | } | |
d6bfde87 | 3329 | |
a512bbf8 | 3330 | ret = write_ctree_super(NULL, root, 0); |
c146afad YZ |
3331 | return ret; |
3332 | } | |
3333 | ||
3334 | int close_ctree(struct btrfs_root *root) | |
3335 | { | |
3336 | struct btrfs_fs_info *fs_info = root->fs_info; | |
3337 | int ret; | |
3338 | ||
3339 | fs_info->closing = 1; | |
3340 | smp_mb(); | |
3341 | ||
837d5b6e | 3342 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 3343 | btrfs_pause_balance(fs_info); |
837d5b6e | 3344 | |
8dabb742 SB |
3345 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
3346 | ||
aa1b8cd4 | 3347 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
3348 | |
3349 | /* wait for any defraggers to finish */ | |
3350 | wait_event(fs_info->transaction_wait, | |
3351 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3352 | ||
3353 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 3354 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 3355 | |
c146afad | 3356 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
acce952b | 3357 | ret = btrfs_commit_super(root); |
3358 | if (ret) | |
3359 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
3360 | } | |
3361 | ||
87533c47 | 3362 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) |
68ce9682 | 3363 | btrfs_error_commit_super(root); |
0f7d52f4 | 3364 | |
300e4f8a JB |
3365 | btrfs_put_block_group_cache(fs_info); |
3366 | ||
e3029d9f AV |
3367 | kthread_stop(fs_info->transaction_kthread); |
3368 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 3369 | |
f25784b3 YZ |
3370 | fs_info->closing = 2; |
3371 | smp_mb(); | |
3372 | ||
bcef60f2 AJ |
3373 | btrfs_free_qgroup_config(root->fs_info); |
3374 | ||
963d678b MX |
3375 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
3376 | printk(KERN_INFO "btrfs: at unmount delalloc count %lld\n", | |
3377 | percpu_counter_sum(&fs_info->delalloc_bytes)); | |
b0c68f8b | 3378 | } |
bcc63abb | 3379 | |
5d4f98a2 YZ |
3380 | free_extent_buffer(fs_info->extent_root->node); |
3381 | free_extent_buffer(fs_info->extent_root->commit_root); | |
3382 | free_extent_buffer(fs_info->tree_root->node); | |
3383 | free_extent_buffer(fs_info->tree_root->commit_root); | |
e3029d9f AV |
3384 | free_extent_buffer(fs_info->chunk_root->node); |
3385 | free_extent_buffer(fs_info->chunk_root->commit_root); | |
3386 | free_extent_buffer(fs_info->dev_root->node); | |
3387 | free_extent_buffer(fs_info->dev_root->commit_root); | |
3388 | free_extent_buffer(fs_info->csum_root->node); | |
3389 | free_extent_buffer(fs_info->csum_root->commit_root); | |
bcef60f2 AJ |
3390 | if (fs_info->quota_root) { |
3391 | free_extent_buffer(fs_info->quota_root->node); | |
3392 | free_extent_buffer(fs_info->quota_root->commit_root); | |
3393 | } | |
d20f7043 | 3394 | |
e3029d9f | 3395 | btrfs_free_block_groups(fs_info); |
d10c5f31 | 3396 | |
c146afad | 3397 | del_fs_roots(fs_info); |
d10c5f31 | 3398 | |
c146afad | 3399 | iput(fs_info->btree_inode); |
9ad6b7bc | 3400 | |
61d92c32 | 3401 | btrfs_stop_workers(&fs_info->generic_worker); |
247e743c | 3402 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 3403 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
3404 | btrfs_stop_workers(&fs_info->workers); |
3405 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 3406 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 3407 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 3408 | btrfs_stop_workers(&fs_info->endio_write_workers); |
0cb59c99 | 3409 | btrfs_stop_workers(&fs_info->endio_freespace_worker); |
1cc127b5 | 3410 | btrfs_stop_workers(&fs_info->submit_workers); |
16cdcec7 | 3411 | btrfs_stop_workers(&fs_info->delayed_workers); |
bab39bf9 | 3412 | btrfs_stop_workers(&fs_info->caching_workers); |
90519d66 | 3413 | btrfs_stop_workers(&fs_info->readahead_workers); |
8ccf6f19 | 3414 | btrfs_stop_workers(&fs_info->flush_workers); |
d6bfde87 | 3415 | |
21adbd5c SB |
3416 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3417 | if (btrfs_test_opt(root, CHECK_INTEGRITY)) | |
3418 | btrfsic_unmount(root, fs_info->fs_devices); | |
3419 | #endif | |
3420 | ||
dfe25020 | 3421 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 3422 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 3423 | |
e2d84521 | 3424 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
963d678b | 3425 | percpu_counter_destroy(&fs_info->delalloc_bytes); |
04160088 | 3426 | bdi_destroy(&fs_info->bdi); |
76dda93c | 3427 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 3428 | |
eb60ceac CM |
3429 | return 0; |
3430 | } | |
3431 | ||
b9fab919 CM |
3432 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
3433 | int atomic) | |
5f39d397 | 3434 | { |
1259ab75 | 3435 | int ret; |
727011e0 | 3436 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 3437 | |
0b32f4bb | 3438 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
3439 | if (!ret) |
3440 | return ret; | |
3441 | ||
3442 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
3443 | parent_transid, atomic); |
3444 | if (ret == -EAGAIN) | |
3445 | return ret; | |
1259ab75 | 3446 | return !ret; |
5f39d397 CM |
3447 | } |
3448 | ||
3449 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 3450 | { |
0b32f4bb | 3451 | return set_extent_buffer_uptodate(buf); |
5f39d397 | 3452 | } |
6702ed49 | 3453 | |
5f39d397 CM |
3454 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
3455 | { | |
727011e0 | 3456 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
5f39d397 | 3457 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 3458 | int was_dirty; |
b4ce94de | 3459 | |
b9447ef8 | 3460 | btrfs_assert_tree_locked(buf); |
31b1a2bd JL |
3461 | if (transid != root->fs_info->generation) |
3462 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, " | |
d397712b | 3463 | "found %llu running %llu\n", |
db94535d | 3464 | (unsigned long long)buf->start, |
d397712b CM |
3465 | (unsigned long long)transid, |
3466 | (unsigned long long)root->fs_info->generation); | |
0b32f4bb | 3467 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 MX |
3468 | if (!was_dirty) |
3469 | __percpu_counter_add(&root->fs_info->dirty_metadata_bytes, | |
3470 | buf->len, | |
3471 | root->fs_info->dirty_metadata_batch); | |
eb60ceac CM |
3472 | } |
3473 | ||
b53d3f5d LB |
3474 | static void __btrfs_btree_balance_dirty(struct btrfs_root *root, |
3475 | int flush_delayed) | |
16cdcec7 MX |
3476 | { |
3477 | /* | |
3478 | * looks as though older kernels can get into trouble with | |
3479 | * this code, they end up stuck in balance_dirty_pages forever | |
3480 | */ | |
e2d84521 | 3481 | int ret; |
16cdcec7 MX |
3482 | |
3483 | if (current->flags & PF_MEMALLOC) | |
3484 | return; | |
3485 | ||
b53d3f5d LB |
3486 | if (flush_delayed) |
3487 | btrfs_balance_delayed_items(root); | |
16cdcec7 | 3488 | |
e2d84521 MX |
3489 | ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes, |
3490 | BTRFS_DIRTY_METADATA_THRESH); | |
3491 | if (ret > 0) { | |
16cdcec7 MX |
3492 | balance_dirty_pages_ratelimited_nr( |
3493 | root->fs_info->btree_inode->i_mapping, 1); | |
3494 | } | |
3495 | return; | |
3496 | } | |
3497 | ||
b53d3f5d | 3498 | void btrfs_btree_balance_dirty(struct btrfs_root *root) |
35b7e476 | 3499 | { |
b53d3f5d LB |
3500 | __btrfs_btree_balance_dirty(root, 1); |
3501 | } | |
585ad2c3 | 3502 | |
b53d3f5d LB |
3503 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root) |
3504 | { | |
3505 | __btrfs_btree_balance_dirty(root, 0); | |
35b7e476 | 3506 | } |
6b80053d | 3507 | |
ca7a79ad | 3508 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 3509 | { |
727011e0 | 3510 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
0b32f4bb | 3511 | return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
6b80053d | 3512 | } |
0da5468f | 3513 | |
fcd1f065 | 3514 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 3515 | int read_only) |
3516 | { | |
fcd1f065 DS |
3517 | if (btrfs_super_csum_type(fs_info->super_copy) >= ARRAY_SIZE(btrfs_csum_sizes)) { |
3518 | printk(KERN_ERR "btrfs: unsupported checksum algorithm\n"); | |
3519 | return -EINVAL; | |
3520 | } | |
3521 | ||
acce952b | 3522 | if (read_only) |
fcd1f065 | 3523 | return 0; |
acce952b | 3524 | |
fcd1f065 | 3525 | return 0; |
acce952b | 3526 | } |
3527 | ||
68ce9682 | 3528 | void btrfs_error_commit_super(struct btrfs_root *root) |
acce952b | 3529 | { |
acce952b | 3530 | mutex_lock(&root->fs_info->cleaner_mutex); |
3531 | btrfs_run_delayed_iputs(root); | |
3532 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
3533 | ||
3534 | down_write(&root->fs_info->cleanup_work_sem); | |
3535 | up_write(&root->fs_info->cleanup_work_sem); | |
3536 | ||
3537 | /* cleanup FS via transaction */ | |
3538 | btrfs_cleanup_transaction(root); | |
acce952b | 3539 | } |
3540 | ||
569e0f35 JB |
3541 | static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, |
3542 | struct btrfs_root *root) | |
acce952b | 3543 | { |
3544 | struct btrfs_inode *btrfs_inode; | |
3545 | struct list_head splice; | |
3546 | ||
3547 | INIT_LIST_HEAD(&splice); | |
3548 | ||
3549 | mutex_lock(&root->fs_info->ordered_operations_mutex); | |
3550 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3551 | ||
569e0f35 | 3552 | list_splice_init(&t->ordered_operations, &splice); |
acce952b | 3553 | while (!list_empty(&splice)) { |
3554 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3555 | ordered_operations); | |
3556 | ||
3557 | list_del_init(&btrfs_inode->ordered_operations); | |
3558 | ||
3559 | btrfs_invalidate_inodes(btrfs_inode->root); | |
3560 | } | |
3561 | ||
3562 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3563 | mutex_unlock(&root->fs_info->ordered_operations_mutex); | |
acce952b | 3564 | } |
3565 | ||
143bede5 | 3566 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 3567 | { |
acce952b | 3568 | struct btrfs_ordered_extent *ordered; |
acce952b | 3569 | |
3570 | spin_lock(&root->fs_info->ordered_extent_lock); | |
779880ef JB |
3571 | /* |
3572 | * This will just short circuit the ordered completion stuff which will | |
3573 | * make sure the ordered extent gets properly cleaned up. | |
3574 | */ | |
3575 | list_for_each_entry(ordered, &root->fs_info->ordered_extents, | |
3576 | root_extent_list) | |
3577 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
acce952b | 3578 | spin_unlock(&root->fs_info->ordered_extent_lock); |
acce952b | 3579 | } |
3580 | ||
79787eaa JM |
3581 | int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
3582 | struct btrfs_root *root) | |
acce952b | 3583 | { |
3584 | struct rb_node *node; | |
3585 | struct btrfs_delayed_ref_root *delayed_refs; | |
3586 | struct btrfs_delayed_ref_node *ref; | |
3587 | int ret = 0; | |
3588 | ||
3589 | delayed_refs = &trans->delayed_refs; | |
3590 | ||
3591 | spin_lock(&delayed_refs->lock); | |
3592 | if (delayed_refs->num_entries == 0) { | |
cfece4db | 3593 | spin_unlock(&delayed_refs->lock); |
acce952b | 3594 | printk(KERN_INFO "delayed_refs has NO entry\n"); |
3595 | return ret; | |
3596 | } | |
3597 | ||
b939d1ab | 3598 | while ((node = rb_first(&delayed_refs->root)) != NULL) { |
eb12db69 | 3599 | struct btrfs_delayed_ref_head *head = NULL; |
acce952b | 3600 | |
eb12db69 | 3601 | ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); |
acce952b | 3602 | atomic_set(&ref->refs, 1); |
3603 | if (btrfs_delayed_ref_is_head(ref)) { | |
acce952b | 3604 | |
3605 | head = btrfs_delayed_node_to_head(ref); | |
b939d1ab JB |
3606 | if (!mutex_trylock(&head->mutex)) { |
3607 | atomic_inc(&ref->refs); | |
3608 | spin_unlock(&delayed_refs->lock); | |
3609 | ||
3610 | /* Need to wait for the delayed ref to run */ | |
3611 | mutex_lock(&head->mutex); | |
3612 | mutex_unlock(&head->mutex); | |
3613 | btrfs_put_delayed_ref(ref); | |
3614 | ||
e18fca73 | 3615 | spin_lock(&delayed_refs->lock); |
b939d1ab JB |
3616 | continue; |
3617 | } | |
3618 | ||
78a6184a | 3619 | btrfs_free_delayed_extent_op(head->extent_op); |
acce952b | 3620 | delayed_refs->num_heads--; |
3621 | if (list_empty(&head->cluster)) | |
3622 | delayed_refs->num_heads_ready--; | |
3623 | list_del_init(&head->cluster); | |
acce952b | 3624 | } |
eb12db69 | 3625 | |
b939d1ab JB |
3626 | ref->in_tree = 0; |
3627 | rb_erase(&ref->rb_node, &delayed_refs->root); | |
3628 | delayed_refs->num_entries--; | |
eb12db69 JB |
3629 | if (head) |
3630 | mutex_unlock(&head->mutex); | |
acce952b | 3631 | spin_unlock(&delayed_refs->lock); |
3632 | btrfs_put_delayed_ref(ref); | |
3633 | ||
3634 | cond_resched(); | |
3635 | spin_lock(&delayed_refs->lock); | |
3636 | } | |
3637 | ||
3638 | spin_unlock(&delayed_refs->lock); | |
3639 | ||
3640 | return ret; | |
3641 | } | |
3642 | ||
143bede5 | 3643 | static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t) |
acce952b | 3644 | { |
3645 | struct btrfs_pending_snapshot *snapshot; | |
3646 | struct list_head splice; | |
3647 | ||
3648 | INIT_LIST_HEAD(&splice); | |
3649 | ||
3650 | list_splice_init(&t->pending_snapshots, &splice); | |
3651 | ||
3652 | while (!list_empty(&splice)) { | |
3653 | snapshot = list_entry(splice.next, | |
3654 | struct btrfs_pending_snapshot, | |
3655 | list); | |
3656 | ||
3657 | list_del_init(&snapshot->list); | |
3658 | ||
3659 | kfree(snapshot); | |
3660 | } | |
acce952b | 3661 | } |
3662 | ||
143bede5 | 3663 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 3664 | { |
3665 | struct btrfs_inode *btrfs_inode; | |
3666 | struct list_head splice; | |
3667 | ||
3668 | INIT_LIST_HEAD(&splice); | |
3669 | ||
acce952b | 3670 | spin_lock(&root->fs_info->delalloc_lock); |
5be76758 | 3671 | list_splice_init(&root->fs_info->delalloc_inodes, &splice); |
acce952b | 3672 | |
3673 | while (!list_empty(&splice)) { | |
3674 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3675 | delalloc_inodes); | |
3676 | ||
3677 | list_del_init(&btrfs_inode->delalloc_inodes); | |
df0af1a5 MX |
3678 | clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, |
3679 | &btrfs_inode->runtime_flags); | |
acce952b | 3680 | |
3681 | btrfs_invalidate_inodes(btrfs_inode->root); | |
3682 | } | |
3683 | ||
3684 | spin_unlock(&root->fs_info->delalloc_lock); | |
acce952b | 3685 | } |
3686 | ||
3687 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, | |
3688 | struct extent_io_tree *dirty_pages, | |
3689 | int mark) | |
3690 | { | |
3691 | int ret; | |
3692 | struct page *page; | |
3693 | struct inode *btree_inode = root->fs_info->btree_inode; | |
3694 | struct extent_buffer *eb; | |
3695 | u64 start = 0; | |
3696 | u64 end; | |
3697 | u64 offset; | |
3698 | unsigned long index; | |
3699 | ||
3700 | while (1) { | |
3701 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 3702 | mark, NULL); |
acce952b | 3703 | if (ret) |
3704 | break; | |
3705 | ||
3706 | clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); | |
3707 | while (start <= end) { | |
3708 | index = start >> PAGE_CACHE_SHIFT; | |
3709 | start = (u64)(index + 1) << PAGE_CACHE_SHIFT; | |
3710 | page = find_get_page(btree_inode->i_mapping, index); | |
3711 | if (!page) | |
3712 | continue; | |
3713 | offset = page_offset(page); | |
3714 | ||
3715 | spin_lock(&dirty_pages->buffer_lock); | |
3716 | eb = radix_tree_lookup( | |
3717 | &(&BTRFS_I(page->mapping->host)->io_tree)->buffer, | |
3718 | offset >> PAGE_CACHE_SHIFT); | |
3719 | spin_unlock(&dirty_pages->buffer_lock); | |
ee670f0a | 3720 | if (eb) |
acce952b | 3721 | ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
3722 | &eb->bflags); | |
acce952b | 3723 | if (PageWriteback(page)) |
3724 | end_page_writeback(page); | |
3725 | ||
3726 | lock_page(page); | |
3727 | if (PageDirty(page)) { | |
3728 | clear_page_dirty_for_io(page); | |
3729 | spin_lock_irq(&page->mapping->tree_lock); | |
3730 | radix_tree_tag_clear(&page->mapping->page_tree, | |
3731 | page_index(page), | |
3732 | PAGECACHE_TAG_DIRTY); | |
3733 | spin_unlock_irq(&page->mapping->tree_lock); | |
3734 | } | |
3735 | ||
acce952b | 3736 | unlock_page(page); |
ee670f0a | 3737 | page_cache_release(page); |
acce952b | 3738 | } |
3739 | } | |
3740 | ||
3741 | return ret; | |
3742 | } | |
3743 | ||
3744 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
3745 | struct extent_io_tree *pinned_extents) | |
3746 | { | |
3747 | struct extent_io_tree *unpin; | |
3748 | u64 start; | |
3749 | u64 end; | |
3750 | int ret; | |
ed0eaa14 | 3751 | bool loop = true; |
acce952b | 3752 | |
3753 | unpin = pinned_extents; | |
ed0eaa14 | 3754 | again: |
acce952b | 3755 | while (1) { |
3756 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
e6138876 | 3757 | EXTENT_DIRTY, NULL); |
acce952b | 3758 | if (ret) |
3759 | break; | |
3760 | ||
3761 | /* opt_discard */ | |
5378e607 LD |
3762 | if (btrfs_test_opt(root, DISCARD)) |
3763 | ret = btrfs_error_discard_extent(root, start, | |
3764 | end + 1 - start, | |
3765 | NULL); | |
acce952b | 3766 | |
3767 | clear_extent_dirty(unpin, start, end, GFP_NOFS); | |
3768 | btrfs_error_unpin_extent_range(root, start, end); | |
3769 | cond_resched(); | |
3770 | } | |
3771 | ||
ed0eaa14 LB |
3772 | if (loop) { |
3773 | if (unpin == &root->fs_info->freed_extents[0]) | |
3774 | unpin = &root->fs_info->freed_extents[1]; | |
3775 | else | |
3776 | unpin = &root->fs_info->freed_extents[0]; | |
3777 | loop = false; | |
3778 | goto again; | |
3779 | } | |
3780 | ||
acce952b | 3781 | return 0; |
3782 | } | |
3783 | ||
49b25e05 JM |
3784 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
3785 | struct btrfs_root *root) | |
3786 | { | |
3787 | btrfs_destroy_delayed_refs(cur_trans, root); | |
3788 | btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv, | |
3789 | cur_trans->dirty_pages.dirty_bytes); | |
3790 | ||
3791 | /* FIXME: cleanup wait for commit */ | |
3792 | cur_trans->in_commit = 1; | |
3793 | cur_trans->blocked = 1; | |
d7096fc3 | 3794 | wake_up(&root->fs_info->transaction_blocked_wait); |
49b25e05 JM |
3795 | |
3796 | cur_trans->blocked = 0; | |
d7096fc3 | 3797 | wake_up(&root->fs_info->transaction_wait); |
49b25e05 JM |
3798 | |
3799 | cur_trans->commit_done = 1; | |
d7096fc3 | 3800 | wake_up(&cur_trans->commit_wait); |
49b25e05 | 3801 | |
67cde344 MX |
3802 | btrfs_destroy_delayed_inodes(root); |
3803 | btrfs_assert_delayed_root_empty(root); | |
49b25e05 JM |
3804 | |
3805 | btrfs_destroy_pending_snapshots(cur_trans); | |
3806 | ||
3807 | btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, | |
3808 | EXTENT_DIRTY); | |
6e841e32 LB |
3809 | btrfs_destroy_pinned_extent(root, |
3810 | root->fs_info->pinned_extents); | |
49b25e05 JM |
3811 | |
3812 | /* | |
3813 | memset(cur_trans, 0, sizeof(*cur_trans)); | |
3814 | kmem_cache_free(btrfs_transaction_cachep, cur_trans); | |
3815 | */ | |
3816 | } | |
3817 | ||
3818 | int btrfs_cleanup_transaction(struct btrfs_root *root) | |
acce952b | 3819 | { |
3820 | struct btrfs_transaction *t; | |
3821 | LIST_HEAD(list); | |
3822 | ||
acce952b | 3823 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
3824 | ||
a4abeea4 | 3825 | spin_lock(&root->fs_info->trans_lock); |
acce952b | 3826 | list_splice_init(&root->fs_info->trans_list, &list); |
a4abeea4 JB |
3827 | root->fs_info->trans_no_join = 1; |
3828 | spin_unlock(&root->fs_info->trans_lock); | |
3829 | ||
acce952b | 3830 | while (!list_empty(&list)) { |
3831 | t = list_entry(list.next, struct btrfs_transaction, list); | |
acce952b | 3832 | |
569e0f35 | 3833 | btrfs_destroy_ordered_operations(t, root); |
acce952b | 3834 | |
3835 | btrfs_destroy_ordered_extents(root); | |
3836 | ||
3837 | btrfs_destroy_delayed_refs(t, root); | |
3838 | ||
3839 | btrfs_block_rsv_release(root, | |
3840 | &root->fs_info->trans_block_rsv, | |
3841 | t->dirty_pages.dirty_bytes); | |
3842 | ||
3843 | /* FIXME: cleanup wait for commit */ | |
3844 | t->in_commit = 1; | |
3845 | t->blocked = 1; | |
66657b31 | 3846 | smp_mb(); |
acce952b | 3847 | if (waitqueue_active(&root->fs_info->transaction_blocked_wait)) |
3848 | wake_up(&root->fs_info->transaction_blocked_wait); | |
3849 | ||
3850 | t->blocked = 0; | |
66657b31 | 3851 | smp_mb(); |
acce952b | 3852 | if (waitqueue_active(&root->fs_info->transaction_wait)) |
3853 | wake_up(&root->fs_info->transaction_wait); | |
acce952b | 3854 | |
acce952b | 3855 | t->commit_done = 1; |
66657b31 | 3856 | smp_mb(); |
acce952b | 3857 | if (waitqueue_active(&t->commit_wait)) |
3858 | wake_up(&t->commit_wait); | |
acce952b | 3859 | |
67cde344 MX |
3860 | btrfs_destroy_delayed_inodes(root); |
3861 | btrfs_assert_delayed_root_empty(root); | |
3862 | ||
acce952b | 3863 | btrfs_destroy_pending_snapshots(t); |
3864 | ||
3865 | btrfs_destroy_delalloc_inodes(root); | |
3866 | ||
a4abeea4 | 3867 | spin_lock(&root->fs_info->trans_lock); |
acce952b | 3868 | root->fs_info->running_transaction = NULL; |
a4abeea4 | 3869 | spin_unlock(&root->fs_info->trans_lock); |
acce952b | 3870 | |
3871 | btrfs_destroy_marked_extents(root, &t->dirty_pages, | |
3872 | EXTENT_DIRTY); | |
3873 | ||
3874 | btrfs_destroy_pinned_extent(root, | |
3875 | root->fs_info->pinned_extents); | |
3876 | ||
13c5a93e | 3877 | atomic_set(&t->use_count, 0); |
acce952b | 3878 | list_del_init(&t->list); |
3879 | memset(t, 0, sizeof(*t)); | |
3880 | kmem_cache_free(btrfs_transaction_cachep, t); | |
3881 | } | |
3882 | ||
a4abeea4 JB |
3883 | spin_lock(&root->fs_info->trans_lock); |
3884 | root->fs_info->trans_no_join = 0; | |
3885 | spin_unlock(&root->fs_info->trans_lock); | |
acce952b | 3886 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); |
acce952b | 3887 | |
3888 | return 0; | |
3889 | } | |
3890 | ||
d1310b2e | 3891 | static struct extent_io_ops btree_extent_io_ops = { |
ce9adaa5 | 3892 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
4bb31e92 | 3893 | .readpage_io_failed_hook = btree_io_failed_hook, |
0b86a832 | 3894 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
3895 | /* note we're sharing with inode.c for the merge bio hook */ |
3896 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 3897 | }; |