]>
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" |
eb60ceac | 46 | |
d1310b2e | 47 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 48 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 49 | static void free_fs_root(struct btrfs_root *root); |
acce952b | 50 | static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
51 | int read_only); | |
52 | static int btrfs_destroy_ordered_operations(struct btrfs_root *root); | |
53 | static int btrfs_destroy_ordered_extents(struct btrfs_root *root); | |
54 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, | |
55 | struct btrfs_root *root); | |
56 | static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t); | |
57 | static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root); | |
58 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, | |
59 | struct extent_io_tree *dirty_pages, | |
60 | int mark); | |
61 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
62 | struct extent_io_tree *pinned_extents); | |
63 | static int btrfs_cleanup_transaction(struct btrfs_root *root); | |
ce9adaa5 | 64 | |
d352ac68 CM |
65 | /* |
66 | * end_io_wq structs are used to do processing in task context when an IO is | |
67 | * complete. This is used during reads to verify checksums, and it is used | |
68 | * by writes to insert metadata for new file extents after IO is complete. | |
69 | */ | |
ce9adaa5 CM |
70 | struct end_io_wq { |
71 | struct bio *bio; | |
72 | bio_end_io_t *end_io; | |
73 | void *private; | |
74 | struct btrfs_fs_info *info; | |
75 | int error; | |
22c59948 | 76 | int metadata; |
ce9adaa5 | 77 | struct list_head list; |
8b712842 | 78 | struct btrfs_work work; |
ce9adaa5 | 79 | }; |
0da5468f | 80 | |
d352ac68 CM |
81 | /* |
82 | * async submit bios are used to offload expensive checksumming | |
83 | * onto the worker threads. They checksum file and metadata bios | |
84 | * just before they are sent down the IO stack. | |
85 | */ | |
44b8bd7e CM |
86 | struct async_submit_bio { |
87 | struct inode *inode; | |
88 | struct bio *bio; | |
89 | struct list_head list; | |
4a69a410 CM |
90 | extent_submit_bio_hook_t *submit_bio_start; |
91 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
92 | int rw; |
93 | int mirror_num; | |
c8b97818 | 94 | unsigned long bio_flags; |
eaf25d93 CM |
95 | /* |
96 | * bio_offset is optional, can be used if the pages in the bio | |
97 | * can't tell us where in the file the bio should go | |
98 | */ | |
99 | u64 bio_offset; | |
8b712842 | 100 | struct btrfs_work work; |
44b8bd7e CM |
101 | }; |
102 | ||
85d4e461 CM |
103 | /* |
104 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
105 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
106 | * the level the eb occupies in the tree. | |
107 | * | |
108 | * Different roots are used for different purposes and may nest inside each | |
109 | * other and they require separate keysets. As lockdep keys should be | |
110 | * static, assign keysets according to the purpose of the root as indicated | |
111 | * by btrfs_root->objectid. This ensures that all special purpose roots | |
112 | * have separate keysets. | |
4008c04a | 113 | * |
85d4e461 CM |
114 | * Lock-nesting across peer nodes is always done with the immediate parent |
115 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
116 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 117 | * |
85d4e461 CM |
118 | * The key is set by the readpage_end_io_hook after the buffer has passed |
119 | * csum validation but before the pages are unlocked. It is also set by | |
120 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 121 | * |
85d4e461 CM |
122 | * We also add a check to make sure the highest level of the tree is the |
123 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
124 | * needs update as well. | |
4008c04a CM |
125 | */ |
126 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
127 | # if BTRFS_MAX_LEVEL != 8 | |
128 | # error | |
129 | # endif | |
85d4e461 CM |
130 | |
131 | static struct btrfs_lockdep_keyset { | |
132 | u64 id; /* root objectid */ | |
133 | const char *name_stem; /* lock name stem */ | |
134 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
135 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
136 | } btrfs_lockdep_keysets[] = { | |
137 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
138 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
139 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
140 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
141 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
142 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
143 | { .id = BTRFS_ORPHAN_OBJECTID, .name_stem = "orphan" }, | |
144 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, | |
145 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
146 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
147 | { .id = 0, .name_stem = "tree" }, | |
4008c04a | 148 | }; |
85d4e461 CM |
149 | |
150 | void __init btrfs_init_lockdep(void) | |
151 | { | |
152 | int i, j; | |
153 | ||
154 | /* initialize lockdep class names */ | |
155 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
156 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
157 | ||
158 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
159 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
160 | "btrfs-%s-%02d", ks->name_stem, j); | |
161 | } | |
162 | } | |
163 | ||
164 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
165 | int level) | |
166 | { | |
167 | struct btrfs_lockdep_keyset *ks; | |
168 | ||
169 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
170 | ||
171 | /* find the matching keyset, id 0 is the default entry */ | |
172 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
173 | if (ks->id == objectid) | |
174 | break; | |
175 | ||
176 | lockdep_set_class_and_name(&eb->lock, | |
177 | &ks->keys[level], ks->names[level]); | |
178 | } | |
179 | ||
4008c04a CM |
180 | #endif |
181 | ||
d352ac68 CM |
182 | /* |
183 | * extents on the btree inode are pretty simple, there's one extent | |
184 | * that covers the entire device | |
185 | */ | |
b2950863 | 186 | static struct extent_map *btree_get_extent(struct inode *inode, |
306e16ce | 187 | struct page *page, size_t pg_offset, u64 start, u64 len, |
b2950863 | 188 | int create) |
7eccb903 | 189 | { |
5f39d397 CM |
190 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
191 | struct extent_map *em; | |
192 | int ret; | |
193 | ||
890871be | 194 | read_lock(&em_tree->lock); |
d1310b2e | 195 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d CM |
196 | if (em) { |
197 | em->bdev = | |
198 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
890871be | 199 | read_unlock(&em_tree->lock); |
5f39d397 | 200 | goto out; |
a061fc8d | 201 | } |
890871be | 202 | read_unlock(&em_tree->lock); |
7b13b7b1 | 203 | |
172ddd60 | 204 | em = alloc_extent_map(); |
5f39d397 CM |
205 | if (!em) { |
206 | em = ERR_PTR(-ENOMEM); | |
207 | goto out; | |
208 | } | |
209 | em->start = 0; | |
0afbaf8c | 210 | em->len = (u64)-1; |
c8b97818 | 211 | em->block_len = (u64)-1; |
5f39d397 | 212 | em->block_start = 0; |
a061fc8d | 213 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e | 214 | |
890871be | 215 | write_lock(&em_tree->lock); |
5f39d397 CM |
216 | ret = add_extent_mapping(em_tree, em); |
217 | if (ret == -EEXIST) { | |
0afbaf8c CM |
218 | u64 failed_start = em->start; |
219 | u64 failed_len = em->len; | |
220 | ||
5f39d397 | 221 | free_extent_map(em); |
7b13b7b1 | 222 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c | 223 | if (em) { |
7b13b7b1 | 224 | ret = 0; |
0afbaf8c CM |
225 | } else { |
226 | em = lookup_extent_mapping(em_tree, failed_start, | |
227 | failed_len); | |
7b13b7b1 | 228 | ret = -EIO; |
0afbaf8c | 229 | } |
5f39d397 | 230 | } else if (ret) { |
7b13b7b1 CM |
231 | free_extent_map(em); |
232 | em = NULL; | |
5f39d397 | 233 | } |
890871be | 234 | write_unlock(&em_tree->lock); |
7b13b7b1 CM |
235 | |
236 | if (ret) | |
237 | em = ERR_PTR(ret); | |
5f39d397 CM |
238 | out: |
239 | return em; | |
7eccb903 CM |
240 | } |
241 | ||
19c00ddc CM |
242 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
243 | { | |
163e783e | 244 | return crc32c(seed, data, len); |
19c00ddc CM |
245 | } |
246 | ||
247 | void btrfs_csum_final(u32 crc, char *result) | |
248 | { | |
7e75bf3f | 249 | put_unaligned_le32(~crc, result); |
19c00ddc CM |
250 | } |
251 | ||
d352ac68 CM |
252 | /* |
253 | * compute the csum for a btree block, and either verify it or write it | |
254 | * into the csum field of the block. | |
255 | */ | |
19c00ddc CM |
256 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
257 | int verify) | |
258 | { | |
6c41761f | 259 | u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy); |
607d432d | 260 | char *result = NULL; |
19c00ddc CM |
261 | unsigned long len; |
262 | unsigned long cur_len; | |
263 | unsigned long offset = BTRFS_CSUM_SIZE; | |
19c00ddc CM |
264 | char *kaddr; |
265 | unsigned long map_start; | |
266 | unsigned long map_len; | |
267 | int err; | |
268 | u32 crc = ~(u32)0; | |
607d432d | 269 | unsigned long inline_result; |
19c00ddc CM |
270 | |
271 | len = buf->len - offset; | |
d397712b | 272 | while (len > 0) { |
19c00ddc | 273 | err = map_private_extent_buffer(buf, offset, 32, |
a6591715 | 274 | &kaddr, &map_start, &map_len); |
d397712b | 275 | if (err) |
19c00ddc | 276 | return 1; |
19c00ddc CM |
277 | cur_len = min(len, map_len - (offset - map_start)); |
278 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
279 | crc, cur_len); | |
280 | len -= cur_len; | |
281 | offset += cur_len; | |
19c00ddc | 282 | } |
607d432d JB |
283 | if (csum_size > sizeof(inline_result)) { |
284 | result = kzalloc(csum_size * sizeof(char), GFP_NOFS); | |
285 | if (!result) | |
286 | return 1; | |
287 | } else { | |
288 | result = (char *)&inline_result; | |
289 | } | |
290 | ||
19c00ddc CM |
291 | btrfs_csum_final(crc, result); |
292 | ||
293 | if (verify) { | |
607d432d | 294 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
295 | u32 val; |
296 | u32 found = 0; | |
607d432d | 297 | memcpy(&found, result, csum_size); |
e4204ded | 298 | |
607d432d | 299 | read_extent_buffer(buf, &val, 0, csum_size); |
7a36ddec | 300 | printk_ratelimited(KERN_INFO "btrfs: %s checksum verify " |
193f284d CM |
301 | "failed on %llu wanted %X found %X " |
302 | "level %d\n", | |
303 | root->fs_info->sb->s_id, | |
304 | (unsigned long long)buf->start, val, found, | |
305 | btrfs_header_level(buf)); | |
607d432d JB |
306 | if (result != (char *)&inline_result) |
307 | kfree(result); | |
19c00ddc CM |
308 | return 1; |
309 | } | |
310 | } else { | |
607d432d | 311 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 312 | } |
607d432d JB |
313 | if (result != (char *)&inline_result) |
314 | kfree(result); | |
19c00ddc CM |
315 | return 0; |
316 | } | |
317 | ||
d352ac68 CM |
318 | /* |
319 | * we can't consider a given block up to date unless the transid of the | |
320 | * block matches the transid in the parent node's pointer. This is how we | |
321 | * detect blocks that either didn't get written at all or got written | |
322 | * in the wrong place. | |
323 | */ | |
1259ab75 CM |
324 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
325 | struct extent_buffer *eb, u64 parent_transid) | |
326 | { | |
2ac55d41 | 327 | struct extent_state *cached_state = NULL; |
1259ab75 CM |
328 | int ret; |
329 | ||
330 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
331 | return 0; | |
332 | ||
2ac55d41 JB |
333 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
334 | 0, &cached_state, GFP_NOFS); | |
335 | if (extent_buffer_uptodate(io_tree, eb, cached_state) && | |
1259ab75 CM |
336 | btrfs_header_generation(eb) == parent_transid) { |
337 | ret = 0; | |
338 | goto out; | |
339 | } | |
7a36ddec | 340 | printk_ratelimited("parent transid verify failed on %llu wanted %llu " |
193f284d CM |
341 | "found %llu\n", |
342 | (unsigned long long)eb->start, | |
343 | (unsigned long long)parent_transid, | |
344 | (unsigned long long)btrfs_header_generation(eb)); | |
1259ab75 | 345 | ret = 1; |
2ac55d41 | 346 | clear_extent_buffer_uptodate(io_tree, eb, &cached_state); |
33958dc6 | 347 | out: |
2ac55d41 JB |
348 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
349 | &cached_state, GFP_NOFS); | |
1259ab75 | 350 | return ret; |
1259ab75 CM |
351 | } |
352 | ||
d352ac68 CM |
353 | /* |
354 | * helper to read a given tree block, doing retries as required when | |
355 | * the checksums don't match and we have alternate mirrors to try. | |
356 | */ | |
f188591e CM |
357 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
358 | struct extent_buffer *eb, | |
ca7a79ad | 359 | u64 start, u64 parent_transid) |
f188591e CM |
360 | { |
361 | struct extent_io_tree *io_tree; | |
362 | int ret; | |
363 | int num_copies = 0; | |
364 | int mirror_num = 0; | |
365 | ||
a826d6dc | 366 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
f188591e CM |
367 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
368 | while (1) { | |
bb82ab88 AJ |
369 | ret = read_extent_buffer_pages(io_tree, eb, start, |
370 | WAIT_COMPLETE, | |
f188591e | 371 | btree_get_extent, mirror_num); |
1259ab75 CM |
372 | if (!ret && |
373 | !verify_parent_transid(io_tree, eb, parent_transid)) | |
f188591e | 374 | return ret; |
d397712b | 375 | |
a826d6dc JB |
376 | /* |
377 | * This buffer's crc is fine, but its contents are corrupted, so | |
378 | * there is no reason to read the other copies, they won't be | |
379 | * any less wrong. | |
380 | */ | |
381 | if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) | |
382 | return ret; | |
383 | ||
f188591e CM |
384 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
385 | eb->start, eb->len); | |
4235298e | 386 | if (num_copies == 1) |
f188591e | 387 | return ret; |
4235298e | 388 | |
f188591e | 389 | mirror_num++; |
4235298e | 390 | if (mirror_num > num_copies) |
f188591e | 391 | return ret; |
f188591e | 392 | } |
f188591e CM |
393 | return -EIO; |
394 | } | |
19c00ddc | 395 | |
d352ac68 | 396 | /* |
d397712b CM |
397 | * checksum a dirty tree block before IO. This has extra checks to make sure |
398 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 399 | */ |
d397712b | 400 | |
b2950863 | 401 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 402 | { |
d1310b2e | 403 | struct extent_io_tree *tree; |
35ebb934 | 404 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc | 405 | u64 found_start; |
19c00ddc CM |
406 | unsigned long len; |
407 | struct extent_buffer *eb; | |
f188591e CM |
408 | int ret; |
409 | ||
d1310b2e | 410 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc | 411 | |
eb14ab8e CM |
412 | if (page->private == EXTENT_PAGE_PRIVATE) { |
413 | WARN_ON(1); | |
19c00ddc | 414 | goto out; |
eb14ab8e CM |
415 | } |
416 | if (!page->private) { | |
417 | WARN_ON(1); | |
19c00ddc | 418 | goto out; |
eb14ab8e | 419 | } |
19c00ddc | 420 | len = page->private >> 2; |
d397712b CM |
421 | WARN_ON(len == 0); |
422 | ||
ba144192 | 423 | eb = alloc_extent_buffer(tree, start, len, page); |
91ca338d TI |
424 | if (eb == NULL) { |
425 | WARN_ON(1); | |
426 | goto out; | |
427 | } | |
ca7a79ad CM |
428 | ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, |
429 | btrfs_header_generation(eb)); | |
f188591e | 430 | BUG_ON(ret); |
784b4e29 CM |
431 | WARN_ON(!btrfs_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN)); |
432 | ||
19c00ddc CM |
433 | found_start = btrfs_header_bytenr(eb); |
434 | if (found_start != start) { | |
55c69072 CM |
435 | WARN_ON(1); |
436 | goto err; | |
437 | } | |
438 | if (eb->first_page != page) { | |
55c69072 CM |
439 | WARN_ON(1); |
440 | goto err; | |
441 | } | |
442 | if (!PageUptodate(page)) { | |
55c69072 CM |
443 | WARN_ON(1); |
444 | goto err; | |
19c00ddc | 445 | } |
19c00ddc | 446 | csum_tree_block(root, eb, 0); |
55c69072 | 447 | err: |
19c00ddc CM |
448 | free_extent_buffer(eb); |
449 | out: | |
450 | return 0; | |
451 | } | |
452 | ||
2b82032c YZ |
453 | static int check_tree_block_fsid(struct btrfs_root *root, |
454 | struct extent_buffer *eb) | |
455 | { | |
456 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
457 | u8 fsid[BTRFS_UUID_SIZE]; | |
458 | int ret = 1; | |
459 | ||
460 | read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), | |
461 | BTRFS_FSID_SIZE); | |
462 | while (fs_devices) { | |
463 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
464 | ret = 0; | |
465 | break; | |
466 | } | |
467 | fs_devices = fs_devices->seed; | |
468 | } | |
469 | return ret; | |
470 | } | |
471 | ||
a826d6dc JB |
472 | #define CORRUPT(reason, eb, root, slot) \ |
473 | printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \ | |
474 | "root=%llu, slot=%d\n", reason, \ | |
475 | (unsigned long long)btrfs_header_bytenr(eb), \ | |
476 | (unsigned long long)root->objectid, slot) | |
477 | ||
478 | static noinline int check_leaf(struct btrfs_root *root, | |
479 | struct extent_buffer *leaf) | |
480 | { | |
481 | struct btrfs_key key; | |
482 | struct btrfs_key leaf_key; | |
483 | u32 nritems = btrfs_header_nritems(leaf); | |
484 | int slot; | |
485 | ||
486 | if (nritems == 0) | |
487 | return 0; | |
488 | ||
489 | /* Check the 0 item */ | |
490 | if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != | |
491 | BTRFS_LEAF_DATA_SIZE(root)) { | |
492 | CORRUPT("invalid item offset size pair", leaf, root, 0); | |
493 | return -EIO; | |
494 | } | |
495 | ||
496 | /* | |
497 | * Check to make sure each items keys are in the correct order and their | |
498 | * offsets make sense. We only have to loop through nritems-1 because | |
499 | * we check the current slot against the next slot, which verifies the | |
500 | * next slot's offset+size makes sense and that the current's slot | |
501 | * offset is correct. | |
502 | */ | |
503 | for (slot = 0; slot < nritems - 1; slot++) { | |
504 | btrfs_item_key_to_cpu(leaf, &leaf_key, slot); | |
505 | btrfs_item_key_to_cpu(leaf, &key, slot + 1); | |
506 | ||
507 | /* Make sure the keys are in the right order */ | |
508 | if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { | |
509 | CORRUPT("bad key order", leaf, root, slot); | |
510 | return -EIO; | |
511 | } | |
512 | ||
513 | /* | |
514 | * Make sure the offset and ends are right, remember that the | |
515 | * item data starts at the end of the leaf and grows towards the | |
516 | * front. | |
517 | */ | |
518 | if (btrfs_item_offset_nr(leaf, slot) != | |
519 | btrfs_item_end_nr(leaf, slot + 1)) { | |
520 | CORRUPT("slot offset bad", leaf, root, slot); | |
521 | return -EIO; | |
522 | } | |
523 | ||
524 | /* | |
525 | * Check to make sure that we don't point outside of the leaf, | |
526 | * just incase all the items are consistent to eachother, but | |
527 | * all point outside of the leaf. | |
528 | */ | |
529 | if (btrfs_item_end_nr(leaf, slot) > | |
530 | BTRFS_LEAF_DATA_SIZE(root)) { | |
531 | CORRUPT("slot end outside of leaf", leaf, root, slot); | |
532 | return -EIO; | |
533 | } | |
534 | } | |
535 | ||
536 | return 0; | |
537 | } | |
538 | ||
b2950863 | 539 | static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
ce9adaa5 CM |
540 | struct extent_state *state) |
541 | { | |
542 | struct extent_io_tree *tree; | |
543 | u64 found_start; | |
544 | int found_level; | |
545 | unsigned long len; | |
546 | struct extent_buffer *eb; | |
547 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 548 | int ret = 0; |
ce9adaa5 CM |
549 | |
550 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
551 | if (page->private == EXTENT_PAGE_PRIVATE) | |
552 | goto out; | |
553 | if (!page->private) | |
554 | goto out; | |
d397712b | 555 | |
ce9adaa5 | 556 | len = page->private >> 2; |
d397712b CM |
557 | WARN_ON(len == 0); |
558 | ||
ba144192 | 559 | eb = alloc_extent_buffer(tree, start, len, page); |
91ca338d TI |
560 | if (eb == NULL) { |
561 | ret = -EIO; | |
562 | goto out; | |
563 | } | |
f188591e | 564 | |
ce9adaa5 | 565 | found_start = btrfs_header_bytenr(eb); |
23a07867 | 566 | if (found_start != start) { |
7a36ddec | 567 | printk_ratelimited(KERN_INFO "btrfs bad tree block start " |
193f284d CM |
568 | "%llu %llu\n", |
569 | (unsigned long long)found_start, | |
570 | (unsigned long long)eb->start); | |
f188591e | 571 | ret = -EIO; |
ce9adaa5 CM |
572 | goto err; |
573 | } | |
574 | if (eb->first_page != page) { | |
d397712b CM |
575 | printk(KERN_INFO "btrfs bad first page %lu %lu\n", |
576 | eb->first_page->index, page->index); | |
ce9adaa5 | 577 | WARN_ON(1); |
f188591e | 578 | ret = -EIO; |
ce9adaa5 CM |
579 | goto err; |
580 | } | |
2b82032c | 581 | if (check_tree_block_fsid(root, eb)) { |
7a36ddec | 582 | printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n", |
193f284d | 583 | (unsigned long long)eb->start); |
1259ab75 CM |
584 | ret = -EIO; |
585 | goto err; | |
586 | } | |
ce9adaa5 CM |
587 | found_level = btrfs_header_level(eb); |
588 | ||
85d4e461 CM |
589 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
590 | eb, found_level); | |
4008c04a | 591 | |
ce9adaa5 | 592 | ret = csum_tree_block(root, eb, 1); |
a826d6dc | 593 | if (ret) { |
f188591e | 594 | ret = -EIO; |
a826d6dc JB |
595 | goto err; |
596 | } | |
597 | ||
598 | /* | |
599 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
600 | * that we don't try and read the other copies of this block, just | |
601 | * return -EIO. | |
602 | */ | |
603 | if (found_level == 0 && check_leaf(root, eb)) { | |
604 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); | |
605 | ret = -EIO; | |
606 | } | |
ce9adaa5 CM |
607 | |
608 | end = min_t(u64, eb->len, PAGE_CACHE_SIZE); | |
609 | end = eb->start + end - 1; | |
ce9adaa5 | 610 | err: |
4bb31e92 AJ |
611 | if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) { |
612 | clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags); | |
613 | btree_readahead_hook(root, eb, eb->start, ret); | |
614 | } | |
615 | ||
ce9adaa5 CM |
616 | free_extent_buffer(eb); |
617 | out: | |
f188591e | 618 | return ret; |
ce9adaa5 CM |
619 | } |
620 | ||
4bb31e92 AJ |
621 | static int btree_io_failed_hook(struct bio *failed_bio, |
622 | struct page *page, u64 start, u64 end, | |
32240a91 | 623 | int mirror_num, struct extent_state *state) |
4bb31e92 AJ |
624 | { |
625 | struct extent_io_tree *tree; | |
626 | unsigned long len; | |
627 | struct extent_buffer *eb; | |
628 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
629 | ||
630 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
631 | if (page->private == EXTENT_PAGE_PRIVATE) | |
632 | goto out; | |
633 | if (!page->private) | |
634 | goto out; | |
635 | ||
636 | len = page->private >> 2; | |
637 | WARN_ON(len == 0); | |
638 | ||
639 | eb = alloc_extent_buffer(tree, start, len, page); | |
640 | if (eb == NULL) | |
641 | goto out; | |
642 | ||
643 | if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) { | |
644 | clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags); | |
645 | btree_readahead_hook(root, eb, eb->start, -EIO); | |
646 | } | |
c674e04e | 647 | free_extent_buffer(eb); |
4bb31e92 AJ |
648 | |
649 | out: | |
650 | return -EIO; /* we fixed nothing */ | |
651 | } | |
652 | ||
ce9adaa5 | 653 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
654 | { |
655 | struct end_io_wq *end_io_wq = bio->bi_private; | |
656 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 657 | |
ce9adaa5 | 658 | fs_info = end_io_wq->info; |
ce9adaa5 | 659 | end_io_wq->error = err; |
8b712842 CM |
660 | end_io_wq->work.func = end_workqueue_fn; |
661 | end_io_wq->work.flags = 0; | |
d20f7043 | 662 | |
7b6d91da | 663 | if (bio->bi_rw & REQ_WRITE) { |
0cb59c99 | 664 | if (end_io_wq->metadata == 1) |
cad321ad CM |
665 | btrfs_queue_worker(&fs_info->endio_meta_write_workers, |
666 | &end_io_wq->work); | |
0cb59c99 JB |
667 | else if (end_io_wq->metadata == 2) |
668 | btrfs_queue_worker(&fs_info->endio_freespace_worker, | |
669 | &end_io_wq->work); | |
cad321ad CM |
670 | else |
671 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
672 | &end_io_wq->work); | |
d20f7043 CM |
673 | } else { |
674 | if (end_io_wq->metadata) | |
675 | btrfs_queue_worker(&fs_info->endio_meta_workers, | |
676 | &end_io_wq->work); | |
677 | else | |
678 | btrfs_queue_worker(&fs_info->endio_workers, | |
679 | &end_io_wq->work); | |
680 | } | |
ce9adaa5 CM |
681 | } |
682 | ||
0cb59c99 JB |
683 | /* |
684 | * For the metadata arg you want | |
685 | * | |
686 | * 0 - if data | |
687 | * 1 - if normal metadta | |
688 | * 2 - if writing to the free space cache area | |
689 | */ | |
22c59948 CM |
690 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
691 | int metadata) | |
0b86a832 | 692 | { |
ce9adaa5 | 693 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
694 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
695 | if (!end_io_wq) | |
696 | return -ENOMEM; | |
697 | ||
698 | end_io_wq->private = bio->bi_private; | |
699 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 700 | end_io_wq->info = info; |
ce9adaa5 CM |
701 | end_io_wq->error = 0; |
702 | end_io_wq->bio = bio; | |
22c59948 | 703 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
704 | |
705 | bio->bi_private = end_io_wq; | |
706 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
707 | return 0; |
708 | } | |
709 | ||
b64a2851 | 710 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 711 | { |
4854ddd0 CM |
712 | unsigned long limit = min_t(unsigned long, |
713 | info->workers.max_workers, | |
714 | info->fs_devices->open_devices); | |
715 | return 256 * limit; | |
716 | } | |
0986fe9e | 717 | |
4a69a410 CM |
718 | static void run_one_async_start(struct btrfs_work *work) |
719 | { | |
4a69a410 CM |
720 | struct async_submit_bio *async; |
721 | ||
722 | async = container_of(work, struct async_submit_bio, work); | |
4a69a410 | 723 | async->submit_bio_start(async->inode, async->rw, async->bio, |
eaf25d93 CM |
724 | async->mirror_num, async->bio_flags, |
725 | async->bio_offset); | |
4a69a410 CM |
726 | } |
727 | ||
728 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
729 | { |
730 | struct btrfs_fs_info *fs_info; | |
731 | struct async_submit_bio *async; | |
4854ddd0 | 732 | int limit; |
8b712842 CM |
733 | |
734 | async = container_of(work, struct async_submit_bio, work); | |
735 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 736 | |
b64a2851 | 737 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
738 | limit = limit * 2 / 3; |
739 | ||
8b712842 | 740 | atomic_dec(&fs_info->nr_async_submits); |
0986fe9e | 741 | |
b64a2851 CM |
742 | if (atomic_read(&fs_info->nr_async_submits) < limit && |
743 | waitqueue_active(&fs_info->async_submit_wait)) | |
4854ddd0 CM |
744 | wake_up(&fs_info->async_submit_wait); |
745 | ||
4a69a410 | 746 | async->submit_bio_done(async->inode, async->rw, async->bio, |
eaf25d93 CM |
747 | async->mirror_num, async->bio_flags, |
748 | async->bio_offset); | |
4a69a410 CM |
749 | } |
750 | ||
751 | static void run_one_async_free(struct btrfs_work *work) | |
752 | { | |
753 | struct async_submit_bio *async; | |
754 | ||
755 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
756 | kfree(async); |
757 | } | |
758 | ||
44b8bd7e CM |
759 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
760 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 761 | unsigned long bio_flags, |
eaf25d93 | 762 | u64 bio_offset, |
4a69a410 CM |
763 | extent_submit_bio_hook_t *submit_bio_start, |
764 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
765 | { |
766 | struct async_submit_bio *async; | |
767 | ||
768 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
769 | if (!async) | |
770 | return -ENOMEM; | |
771 | ||
772 | async->inode = inode; | |
773 | async->rw = rw; | |
774 | async->bio = bio; | |
775 | async->mirror_num = mirror_num; | |
4a69a410 CM |
776 | async->submit_bio_start = submit_bio_start; |
777 | async->submit_bio_done = submit_bio_done; | |
778 | ||
779 | async->work.func = run_one_async_start; | |
780 | async->work.ordered_func = run_one_async_done; | |
781 | async->work.ordered_free = run_one_async_free; | |
782 | ||
8b712842 | 783 | async->work.flags = 0; |
c8b97818 | 784 | async->bio_flags = bio_flags; |
eaf25d93 | 785 | async->bio_offset = bio_offset; |
8c8bee1d | 786 | |
cb03c743 | 787 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 | 788 | |
7b6d91da | 789 | if (rw & REQ_SYNC) |
d313d7a3 CM |
790 | btrfs_set_work_high_prio(&async->work); |
791 | ||
8b712842 | 792 | btrfs_queue_worker(&fs_info->workers, &async->work); |
9473f16c | 793 | |
d397712b | 794 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
795 | atomic_read(&fs_info->nr_async_submits)) { |
796 | wait_event(fs_info->async_submit_wait, | |
797 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
798 | } | |
799 | ||
44b8bd7e CM |
800 | return 0; |
801 | } | |
802 | ||
ce3ed71a CM |
803 | static int btree_csum_one_bio(struct bio *bio) |
804 | { | |
805 | struct bio_vec *bvec = bio->bi_io_vec; | |
806 | int bio_index = 0; | |
807 | struct btrfs_root *root; | |
808 | ||
809 | WARN_ON(bio->bi_vcnt <= 0); | |
d397712b | 810 | while (bio_index < bio->bi_vcnt) { |
ce3ed71a CM |
811 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
812 | csum_dirty_buffer(root, bvec->bv_page); | |
813 | bio_index++; | |
814 | bvec++; | |
815 | } | |
816 | return 0; | |
817 | } | |
818 | ||
4a69a410 CM |
819 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
820 | struct bio *bio, int mirror_num, | |
eaf25d93 CM |
821 | unsigned long bio_flags, |
822 | u64 bio_offset) | |
22c59948 | 823 | { |
8b712842 CM |
824 | /* |
825 | * when we're called for a write, we're already in the async | |
5443be45 | 826 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 827 | */ |
4a69a410 CM |
828 | btree_csum_one_bio(bio); |
829 | return 0; | |
830 | } | |
22c59948 | 831 | |
4a69a410 | 832 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
833 | int mirror_num, unsigned long bio_flags, |
834 | u64 bio_offset) | |
4a69a410 | 835 | { |
8b712842 | 836 | /* |
4a69a410 CM |
837 | * when we're called for a write, we're already in the async |
838 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 839 | */ |
8b712842 | 840 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
841 | } |
842 | ||
44b8bd7e | 843 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
844 | int mirror_num, unsigned long bio_flags, |
845 | u64 bio_offset) | |
44b8bd7e | 846 | { |
cad321ad CM |
847 | int ret; |
848 | ||
849 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, | |
850 | bio, 1); | |
851 | BUG_ON(ret); | |
852 | ||
7b6d91da | 853 | if (!(rw & REQ_WRITE)) { |
4a69a410 CM |
854 | /* |
855 | * called for a read, do the setup so that checksum validation | |
856 | * can happen in the async kernel threads | |
857 | */ | |
4a69a410 | 858 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, |
6f3577bd | 859 | mirror_num, 0); |
44b8bd7e | 860 | } |
d313d7a3 | 861 | |
cad321ad CM |
862 | /* |
863 | * kthread helpers are used to submit writes so that checksumming | |
864 | * can happen in parallel across all CPUs | |
865 | */ | |
44b8bd7e | 866 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, |
c8b97818 | 867 | inode, rw, bio, mirror_num, 0, |
eaf25d93 | 868 | bio_offset, |
4a69a410 CM |
869 | __btree_submit_bio_start, |
870 | __btree_submit_bio_done); | |
44b8bd7e CM |
871 | } |
872 | ||
3dd1462e | 873 | #ifdef CONFIG_MIGRATION |
784b4e29 CM |
874 | static int btree_migratepage(struct address_space *mapping, |
875 | struct page *newpage, struct page *page) | |
876 | { | |
877 | /* | |
878 | * we can't safely write a btree page from here, | |
879 | * we haven't done the locking hook | |
880 | */ | |
881 | if (PageDirty(page)) | |
882 | return -EAGAIN; | |
883 | /* | |
884 | * Buffers may be managed in a filesystem specific way. | |
885 | * We must have no buffers or drop them. | |
886 | */ | |
887 | if (page_has_private(page) && | |
888 | !try_to_release_page(page, GFP_KERNEL)) | |
889 | return -EAGAIN; | |
784b4e29 CM |
890 | return migrate_page(mapping, newpage, page); |
891 | } | |
3dd1462e | 892 | #endif |
784b4e29 | 893 | |
0da5468f CM |
894 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
895 | { | |
d1310b2e | 896 | struct extent_io_tree *tree; |
b9473439 CM |
897 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
898 | struct extent_buffer *eb; | |
899 | int was_dirty; | |
900 | ||
d1310b2e | 901 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
b9473439 CM |
902 | if (!(current->flags & PF_MEMALLOC)) { |
903 | return extent_write_full_page(tree, page, | |
904 | btree_get_extent, wbc); | |
905 | } | |
5443be45 | 906 | |
b9473439 | 907 | redirty_page_for_writepage(wbc, page); |
784b4e29 | 908 | eb = btrfs_find_tree_block(root, page_offset(page), PAGE_CACHE_SIZE); |
b9473439 CM |
909 | WARN_ON(!eb); |
910 | ||
911 | was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags); | |
912 | if (!was_dirty) { | |
913 | spin_lock(&root->fs_info->delalloc_lock); | |
914 | root->fs_info->dirty_metadata_bytes += PAGE_CACHE_SIZE; | |
915 | spin_unlock(&root->fs_info->delalloc_lock); | |
5443be45 | 916 | } |
b9473439 CM |
917 | free_extent_buffer(eb); |
918 | ||
919 | unlock_page(page); | |
920 | return 0; | |
5f39d397 | 921 | } |
0da5468f CM |
922 | |
923 | static int btree_writepages(struct address_space *mapping, | |
924 | struct writeback_control *wbc) | |
925 | { | |
d1310b2e CM |
926 | struct extent_io_tree *tree; |
927 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 928 | if (wbc->sync_mode == WB_SYNC_NONE) { |
b9473439 | 929 | struct btrfs_root *root = BTRFS_I(mapping->host)->root; |
793955bc | 930 | u64 num_dirty; |
24ab9cd8 | 931 | unsigned long thresh = 32 * 1024 * 1024; |
448d640b CM |
932 | |
933 | if (wbc->for_kupdate) | |
934 | return 0; | |
935 | ||
b9473439 CM |
936 | /* this is a bit racy, but that's ok */ |
937 | num_dirty = root->fs_info->dirty_metadata_bytes; | |
d397712b | 938 | if (num_dirty < thresh) |
793955bc | 939 | return 0; |
793955bc | 940 | } |
0da5468f CM |
941 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
942 | } | |
943 | ||
b2950863 | 944 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 945 | { |
d1310b2e CM |
946 | struct extent_io_tree *tree; |
947 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 948 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 949 | } |
22b0ebda | 950 | |
70dec807 | 951 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 952 | { |
d1310b2e CM |
953 | struct extent_io_tree *tree; |
954 | struct extent_map_tree *map; | |
5f39d397 | 955 | int ret; |
d98237b3 | 956 | |
98509cfc | 957 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 958 | return 0; |
98509cfc | 959 | |
d1310b2e CM |
960 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
961 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
6af118ce | 962 | |
7b13b7b1 | 963 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
d397712b | 964 | if (!ret) |
6af118ce | 965 | return 0; |
6af118ce CM |
966 | |
967 | ret = try_release_extent_buffer(tree, page); | |
5f39d397 CM |
968 | if (ret == 1) { |
969 | ClearPagePrivate(page); | |
970 | set_page_private(page, 0); | |
971 | page_cache_release(page); | |
972 | } | |
6af118ce | 973 | |
d98237b3 CM |
974 | return ret; |
975 | } | |
976 | ||
5f39d397 | 977 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 978 | { |
d1310b2e CM |
979 | struct extent_io_tree *tree; |
980 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
981 | extent_invalidatepage(tree, page, offset); |
982 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 983 | if (PagePrivate(page)) { |
d397712b CM |
984 | printk(KERN_WARNING "btrfs warning page private not zero " |
985 | "on page %llu\n", (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
986 | ClearPagePrivate(page); |
987 | set_page_private(page, 0); | |
988 | page_cache_release(page); | |
989 | } | |
d98237b3 CM |
990 | } |
991 | ||
7f09410b | 992 | static const struct address_space_operations btree_aops = { |
d98237b3 CM |
993 | .readpage = btree_readpage, |
994 | .writepage = btree_writepage, | |
0da5468f | 995 | .writepages = btree_writepages, |
5f39d397 CM |
996 | .releasepage = btree_releasepage, |
997 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 998 | #ifdef CONFIG_MIGRATION |
784b4e29 | 999 | .migratepage = btree_migratepage, |
5a92bc88 | 1000 | #endif |
d98237b3 CM |
1001 | }; |
1002 | ||
ca7a79ad CM |
1003 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1004 | u64 parent_transid) | |
090d1875 | 1005 | { |
5f39d397 CM |
1006 | struct extent_buffer *buf = NULL; |
1007 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 1008 | int ret = 0; |
090d1875 | 1009 | |
db94535d | 1010 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 1011 | if (!buf) |
090d1875 | 1012 | return 0; |
d1310b2e | 1013 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
bb82ab88 | 1014 | buf, 0, WAIT_NONE, btree_get_extent, 0); |
5f39d397 | 1015 | free_extent_buffer(buf); |
de428b63 | 1016 | return ret; |
090d1875 CM |
1017 | } |
1018 | ||
ab0fff03 AJ |
1019 | int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1020 | int mirror_num, struct extent_buffer **eb) | |
1021 | { | |
1022 | struct extent_buffer *buf = NULL; | |
1023 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1024 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; | |
1025 | int ret; | |
1026 | ||
1027 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
1028 | if (!buf) | |
1029 | return 0; | |
1030 | ||
1031 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1032 | ||
1033 | ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, | |
1034 | btree_get_extent, mirror_num); | |
1035 | if (ret) { | |
1036 | free_extent_buffer(buf); | |
1037 | return ret; | |
1038 | } | |
1039 | ||
1040 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1041 | free_extent_buffer(buf); | |
1042 | return -EIO; | |
1043 | } else if (extent_buffer_uptodate(io_tree, buf, NULL)) { | |
1044 | *eb = buf; | |
1045 | } else { | |
1046 | free_extent_buffer(buf); | |
1047 | } | |
1048 | return 0; | |
1049 | } | |
1050 | ||
0999df54 CM |
1051 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
1052 | u64 bytenr, u32 blocksize) | |
1053 | { | |
1054 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1055 | struct extent_buffer *eb; | |
1056 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
f09d1f60 | 1057 | bytenr, blocksize); |
0999df54 CM |
1058 | return eb; |
1059 | } | |
1060 | ||
1061 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
1062 | u64 bytenr, u32 blocksize) | |
1063 | { | |
1064 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1065 | struct extent_buffer *eb; | |
1066 | ||
1067 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
ba144192 | 1068 | bytenr, blocksize, NULL); |
0999df54 CM |
1069 | return eb; |
1070 | } | |
1071 | ||
1072 | ||
e02119d5 CM |
1073 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1074 | { | |
8aa38c31 CH |
1075 | return filemap_fdatawrite_range(buf->first_page->mapping, buf->start, |
1076 | buf->start + buf->len - 1); | |
e02119d5 CM |
1077 | } |
1078 | ||
1079 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
1080 | { | |
8aa38c31 CH |
1081 | return filemap_fdatawait_range(buf->first_page->mapping, |
1082 | buf->start, buf->start + buf->len - 1); | |
e02119d5 CM |
1083 | } |
1084 | ||
0999df54 | 1085 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 1086 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
1087 | { |
1088 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1089 | int ret; |
1090 | ||
0999df54 CM |
1091 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
1092 | if (!buf) | |
1093 | return NULL; | |
0999df54 | 1094 | |
ca7a79ad | 1095 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 | 1096 | |
d397712b | 1097 | if (ret == 0) |
b4ce94de | 1098 | set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags); |
5f39d397 | 1099 | return buf; |
ce9adaa5 | 1100 | |
eb60ceac CM |
1101 | } |
1102 | ||
e089f05c | 1103 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 1104 | struct extent_buffer *buf) |
ed2ff2cb | 1105 | { |
5f39d397 | 1106 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 | 1107 | if (btrfs_header_generation(buf) == |
925baedd | 1108 | root->fs_info->running_transaction->transid) { |
b9447ef8 | 1109 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1110 | |
b9473439 CM |
1111 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
1112 | spin_lock(&root->fs_info->delalloc_lock); | |
1113 | if (root->fs_info->dirty_metadata_bytes >= buf->len) | |
1114 | root->fs_info->dirty_metadata_bytes -= buf->len; | |
1115 | else | |
1116 | WARN_ON(1); | |
1117 | spin_unlock(&root->fs_info->delalloc_lock); | |
1118 | } | |
b4ce94de | 1119 | |
b9473439 CM |
1120 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1121 | btrfs_set_lock_blocking(buf); | |
d1310b2e | 1122 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 1123 | buf); |
925baedd | 1124 | } |
5f39d397 CM |
1125 | return 0; |
1126 | } | |
1127 | ||
db94535d | 1128 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 1129 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 1130 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 1131 | u64 objectid) |
d97e63b6 | 1132 | { |
cfaa7295 | 1133 | root->node = NULL; |
a28ec197 | 1134 | root->commit_root = NULL; |
db94535d CM |
1135 | root->sectorsize = sectorsize; |
1136 | root->nodesize = nodesize; | |
1137 | root->leafsize = leafsize; | |
87ee04eb | 1138 | root->stripesize = stripesize; |
123abc88 | 1139 | root->ref_cows = 0; |
0b86a832 | 1140 | root->track_dirty = 0; |
c71bf099 | 1141 | root->in_radix = 0; |
d68fc57b YZ |
1142 | root->orphan_item_inserted = 0; |
1143 | root->orphan_cleanup_state = 0; | |
0b86a832 | 1144 | |
0f7d52f4 CM |
1145 | root->objectid = objectid; |
1146 | root->last_trans = 0; | |
13a8a7c8 | 1147 | root->highest_objectid = 0; |
58176a96 | 1148 | root->name = NULL; |
6bef4d31 | 1149 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1150 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1151 | root->block_rsv = NULL; |
d68fc57b | 1152 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
1153 | |
1154 | INIT_LIST_HEAD(&root->dirty_list); | |
7b128766 | 1155 | INIT_LIST_HEAD(&root->orphan_list); |
5d4f98a2 | 1156 | INIT_LIST_HEAD(&root->root_list); |
d68fc57b | 1157 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 1158 | spin_lock_init(&root->inode_lock); |
f0486c68 | 1159 | spin_lock_init(&root->accounting_lock); |
a2135011 | 1160 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1161 | mutex_init(&root->log_mutex); |
7237f183 YZ |
1162 | init_waitqueue_head(&root->log_writer_wait); |
1163 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1164 | init_waitqueue_head(&root->log_commit_wait[1]); | |
1165 | atomic_set(&root->log_commit[0], 0); | |
1166 | atomic_set(&root->log_commit[1], 0); | |
1167 | atomic_set(&root->log_writers, 0); | |
1168 | root->log_batch = 0; | |
1169 | root->log_transid = 0; | |
257c62e1 | 1170 | root->last_log_commit = 0; |
d0c803c4 | 1171 | extent_io_tree_init(&root->dirty_log_pages, |
f993c883 | 1172 | fs_info->btree_inode->i_mapping); |
017e5369 | 1173 | |
3768f368 CM |
1174 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1175 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1176 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 1177 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 1178 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 1179 | init_completion(&root->kobj_unregister); |
6702ed49 | 1180 | root->defrag_running = 0; |
4d775673 | 1181 | root->root_key.objectid = objectid; |
0ee5dc67 | 1182 | root->anon_dev = 0; |
3768f368 CM |
1183 | return 0; |
1184 | } | |
1185 | ||
db94535d | 1186 | static int find_and_setup_root(struct btrfs_root *tree_root, |
9f5fae2f CM |
1187 | struct btrfs_fs_info *fs_info, |
1188 | u64 objectid, | |
e20d96d6 | 1189 | struct btrfs_root *root) |
3768f368 CM |
1190 | { |
1191 | int ret; | |
db94535d | 1192 | u32 blocksize; |
84234f3a | 1193 | u64 generation; |
3768f368 | 1194 | |
db94535d | 1195 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1196 | tree_root->sectorsize, tree_root->stripesize, |
1197 | root, fs_info, objectid); | |
3768f368 CM |
1198 | ret = btrfs_find_last_root(tree_root, objectid, |
1199 | &root->root_item, &root->root_key); | |
4df27c4d YZ |
1200 | if (ret > 0) |
1201 | return -ENOENT; | |
3768f368 CM |
1202 | BUG_ON(ret); |
1203 | ||
84234f3a | 1204 | generation = btrfs_root_generation(&root->root_item); |
db94535d | 1205 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
af31f5e5 | 1206 | root->commit_root = NULL; |
db94535d | 1207 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), |
84234f3a | 1208 | blocksize, generation); |
68433b73 CM |
1209 | if (!root->node || !btrfs_buffer_uptodate(root->node, generation)) { |
1210 | free_extent_buffer(root->node); | |
af31f5e5 | 1211 | root->node = NULL; |
68433b73 CM |
1212 | return -EIO; |
1213 | } | |
4df27c4d | 1214 | root->commit_root = btrfs_root_node(root); |
d97e63b6 CM |
1215 | return 0; |
1216 | } | |
1217 | ||
6f07e42e AV |
1218 | struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info) |
1219 | { | |
1220 | struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS); | |
1221 | if (root) | |
1222 | root->fs_info = fs_info; | |
1223 | return root; | |
1224 | } | |
1225 | ||
7237f183 YZ |
1226 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1227 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1228 | { |
1229 | struct btrfs_root *root; | |
1230 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1231 | struct extent_buffer *leaf; |
e02119d5 | 1232 | |
6f07e42e | 1233 | root = btrfs_alloc_root(fs_info); |
e02119d5 | 1234 | if (!root) |
7237f183 | 1235 | return ERR_PTR(-ENOMEM); |
e02119d5 CM |
1236 | |
1237 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1238 | tree_root->sectorsize, tree_root->stripesize, | |
1239 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1240 | ||
1241 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1242 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1243 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
7237f183 YZ |
1244 | /* |
1245 | * log trees do not get reference counted because they go away | |
1246 | * before a real commit is actually done. They do store pointers | |
1247 | * to file data extents, and those reference counts still get | |
1248 | * updated (along with back refs to the log tree). | |
1249 | */ | |
e02119d5 CM |
1250 | root->ref_cows = 0; |
1251 | ||
5d4f98a2 YZ |
1252 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, |
1253 | BTRFS_TREE_LOG_OBJECTID, NULL, 0, 0, 0); | |
7237f183 YZ |
1254 | if (IS_ERR(leaf)) { |
1255 | kfree(root); | |
1256 | return ERR_CAST(leaf); | |
1257 | } | |
e02119d5 | 1258 | |
5d4f98a2 YZ |
1259 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1260 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1261 | btrfs_set_header_generation(leaf, trans->transid); | |
1262 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1263 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1264 | root->node = leaf; |
e02119d5 CM |
1265 | |
1266 | write_extent_buffer(root->node, root->fs_info->fsid, | |
1267 | (unsigned long)btrfs_header_fsid(root->node), | |
1268 | BTRFS_FSID_SIZE); | |
1269 | btrfs_mark_buffer_dirty(root->node); | |
1270 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1271 | return root; |
1272 | } | |
1273 | ||
1274 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1275 | struct btrfs_fs_info *fs_info) | |
1276 | { | |
1277 | struct btrfs_root *log_root; | |
1278 | ||
1279 | log_root = alloc_log_tree(trans, fs_info); | |
1280 | if (IS_ERR(log_root)) | |
1281 | return PTR_ERR(log_root); | |
1282 | WARN_ON(fs_info->log_root_tree); | |
1283 | fs_info->log_root_tree = log_root; | |
1284 | return 0; | |
1285 | } | |
1286 | ||
1287 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1288 | struct btrfs_root *root) | |
1289 | { | |
1290 | struct btrfs_root *log_root; | |
1291 | struct btrfs_inode_item *inode_item; | |
1292 | ||
1293 | log_root = alloc_log_tree(trans, root->fs_info); | |
1294 | if (IS_ERR(log_root)) | |
1295 | return PTR_ERR(log_root); | |
1296 | ||
1297 | log_root->last_trans = trans->transid; | |
1298 | log_root->root_key.offset = root->root_key.objectid; | |
1299 | ||
1300 | inode_item = &log_root->root_item.inode; | |
1301 | inode_item->generation = cpu_to_le64(1); | |
1302 | inode_item->size = cpu_to_le64(3); | |
1303 | inode_item->nlink = cpu_to_le32(1); | |
1304 | inode_item->nbytes = cpu_to_le64(root->leafsize); | |
1305 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); | |
1306 | ||
5d4f98a2 | 1307 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1308 | |
1309 | WARN_ON(root->log_root); | |
1310 | root->log_root = log_root; | |
1311 | root->log_transid = 0; | |
257c62e1 | 1312 | root->last_log_commit = 0; |
e02119d5 CM |
1313 | return 0; |
1314 | } | |
1315 | ||
1316 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
1317 | struct btrfs_key *location) | |
1318 | { | |
1319 | struct btrfs_root *root; | |
1320 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1321 | struct btrfs_path *path; |
5f39d397 | 1322 | struct extent_buffer *l; |
84234f3a | 1323 | u64 generation; |
db94535d | 1324 | u32 blocksize; |
0f7d52f4 CM |
1325 | int ret = 0; |
1326 | ||
6f07e42e | 1327 | root = btrfs_alloc_root(fs_info); |
0cf6c620 | 1328 | if (!root) |
0f7d52f4 | 1329 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 1330 | if (location->offset == (u64)-1) { |
db94535d | 1331 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
1332 | location->objectid, root); |
1333 | if (ret) { | |
0f7d52f4 CM |
1334 | kfree(root); |
1335 | return ERR_PTR(ret); | |
1336 | } | |
13a8a7c8 | 1337 | goto out; |
0f7d52f4 CM |
1338 | } |
1339 | ||
db94535d | 1340 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1341 | tree_root->sectorsize, tree_root->stripesize, |
1342 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
1343 | |
1344 | path = btrfs_alloc_path(); | |
db5b493a TI |
1345 | if (!path) { |
1346 | kfree(root); | |
1347 | return ERR_PTR(-ENOMEM); | |
1348 | } | |
0f7d52f4 | 1349 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); |
13a8a7c8 YZ |
1350 | if (ret == 0) { |
1351 | l = path->nodes[0]; | |
1352 | read_extent_buffer(l, &root->root_item, | |
1353 | btrfs_item_ptr_offset(l, path->slots[0]), | |
1354 | sizeof(root->root_item)); | |
1355 | memcpy(&root->root_key, location, sizeof(*location)); | |
0f7d52f4 | 1356 | } |
0f7d52f4 CM |
1357 | btrfs_free_path(path); |
1358 | if (ret) { | |
5e540f77 | 1359 | kfree(root); |
13a8a7c8 YZ |
1360 | if (ret > 0) |
1361 | ret = -ENOENT; | |
0f7d52f4 CM |
1362 | return ERR_PTR(ret); |
1363 | } | |
13a8a7c8 | 1364 | |
84234f3a | 1365 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1366 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1367 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1368 | blocksize, generation); |
5d4f98a2 | 1369 | root->commit_root = btrfs_root_node(root); |
0f7d52f4 | 1370 | BUG_ON(!root->node); |
13a8a7c8 | 1371 | out: |
08fe4db1 | 1372 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { |
e02119d5 | 1373 | root->ref_cows = 1; |
08fe4db1 LZ |
1374 | btrfs_check_and_init_root_item(&root->root_item); |
1375 | } | |
13a8a7c8 | 1376 | |
5eda7b5e CM |
1377 | return root; |
1378 | } | |
1379 | ||
edbd8d4e CM |
1380 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
1381 | struct btrfs_key *location) | |
5eda7b5e CM |
1382 | { |
1383 | struct btrfs_root *root; | |
1384 | int ret; | |
1385 | ||
edbd8d4e CM |
1386 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1387 | return fs_info->tree_root; | |
1388 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1389 | return fs_info->extent_root; | |
8f18cf13 CM |
1390 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1391 | return fs_info->chunk_root; | |
1392 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1393 | return fs_info->dev_root; | |
0403e47e YZ |
1394 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1395 | return fs_info->csum_root; | |
4df27c4d YZ |
1396 | again: |
1397 | spin_lock(&fs_info->fs_roots_radix_lock); | |
5eda7b5e CM |
1398 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
1399 | (unsigned long)location->objectid); | |
4df27c4d | 1400 | spin_unlock(&fs_info->fs_roots_radix_lock); |
5eda7b5e CM |
1401 | if (root) |
1402 | return root; | |
1403 | ||
e02119d5 | 1404 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
1405 | if (IS_ERR(root)) |
1406 | return root; | |
3394e160 | 1407 | |
581bb050 | 1408 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); |
581bb050 LZ |
1409 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), |
1410 | GFP_NOFS); | |
35a30d7c DS |
1411 | if (!root->free_ino_pinned || !root->free_ino_ctl) { |
1412 | ret = -ENOMEM; | |
581bb050 | 1413 | goto fail; |
35a30d7c | 1414 | } |
581bb050 LZ |
1415 | |
1416 | btrfs_init_free_ino_ctl(root); | |
1417 | mutex_init(&root->fs_commit_mutex); | |
1418 | spin_lock_init(&root->cache_lock); | |
1419 | init_waitqueue_head(&root->cache_wait); | |
1420 | ||
0ee5dc67 | 1421 | ret = get_anon_bdev(&root->anon_dev); |
ac08aedf CM |
1422 | if (ret) |
1423 | goto fail; | |
3394e160 | 1424 | |
d68fc57b YZ |
1425 | if (btrfs_root_refs(&root->root_item) == 0) { |
1426 | ret = -ENOENT; | |
1427 | goto fail; | |
1428 | } | |
1429 | ||
1430 | ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid); | |
1431 | if (ret < 0) | |
1432 | goto fail; | |
1433 | if (ret == 0) | |
1434 | root->orphan_item_inserted = 1; | |
1435 | ||
4df27c4d YZ |
1436 | ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); |
1437 | if (ret) | |
1438 | goto fail; | |
1439 | ||
1440 | spin_lock(&fs_info->fs_roots_radix_lock); | |
2619ba1f CM |
1441 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
1442 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 | 1443 | root); |
d68fc57b | 1444 | if (ret == 0) |
4df27c4d | 1445 | root->in_radix = 1; |
d68fc57b | 1446 | |
4df27c4d YZ |
1447 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1448 | radix_tree_preload_end(); | |
0f7d52f4 | 1449 | if (ret) { |
4df27c4d YZ |
1450 | if (ret == -EEXIST) { |
1451 | free_fs_root(root); | |
1452 | goto again; | |
1453 | } | |
1454 | goto fail; | |
0f7d52f4 | 1455 | } |
4df27c4d YZ |
1456 | |
1457 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
1458 | root->root_key.objectid); | |
1459 | WARN_ON(ret); | |
edbd8d4e | 1460 | return root; |
4df27c4d YZ |
1461 | fail: |
1462 | free_fs_root(root); | |
1463 | return ERR_PTR(ret); | |
edbd8d4e CM |
1464 | } |
1465 | ||
04160088 CM |
1466 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1467 | { | |
1468 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1469 | int ret = 0; | |
04160088 CM |
1470 | struct btrfs_device *device; |
1471 | struct backing_dev_info *bdi; | |
b7967db7 | 1472 | |
1f78160c XG |
1473 | rcu_read_lock(); |
1474 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1475 | if (!device->bdev) |
1476 | continue; | |
04160088 CM |
1477 | bdi = blk_get_backing_dev_info(device->bdev); |
1478 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1479 | ret = 1; | |
1480 | break; | |
1481 | } | |
1482 | } | |
1f78160c | 1483 | rcu_read_unlock(); |
04160088 CM |
1484 | return ret; |
1485 | } | |
1486 | ||
ad081f14 JA |
1487 | /* |
1488 | * If this fails, caller must call bdi_destroy() to get rid of the | |
1489 | * bdi again. | |
1490 | */ | |
04160088 CM |
1491 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1492 | { | |
ad081f14 JA |
1493 | int err; |
1494 | ||
1495 | bdi->capabilities = BDI_CAP_MAP_COPY; | |
e6d086d8 | 1496 | err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY); |
ad081f14 JA |
1497 | if (err) |
1498 | return err; | |
1499 | ||
4575c9cc | 1500 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1501 | bdi->congested_fn = btrfs_congested_fn; |
1502 | bdi->congested_data = info; | |
1503 | return 0; | |
1504 | } | |
1505 | ||
ce9adaa5 CM |
1506 | static int bio_ready_for_csum(struct bio *bio) |
1507 | { | |
1508 | u64 length = 0; | |
1509 | u64 buf_len = 0; | |
1510 | u64 start = 0; | |
1511 | struct page *page; | |
1512 | struct extent_io_tree *io_tree = NULL; | |
ce9adaa5 CM |
1513 | struct bio_vec *bvec; |
1514 | int i; | |
1515 | int ret; | |
1516 | ||
1517 | bio_for_each_segment(bvec, bio, i) { | |
1518 | page = bvec->bv_page; | |
1519 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
1520 | length += bvec->bv_len; | |
1521 | continue; | |
1522 | } | |
1523 | if (!page->private) { | |
1524 | length += bvec->bv_len; | |
1525 | continue; | |
1526 | } | |
1527 | length = bvec->bv_len; | |
1528 | buf_len = page->private >> 2; | |
1529 | start = page_offset(page) + bvec->bv_offset; | |
1530 | io_tree = &BTRFS_I(page->mapping->host)->io_tree; | |
ce9adaa5 CM |
1531 | } |
1532 | /* are we fully contained in this bio? */ | |
1533 | if (buf_len <= length) | |
1534 | return 1; | |
1535 | ||
1536 | ret = extent_range_uptodate(io_tree, start + length, | |
1537 | start + buf_len - 1); | |
ce9adaa5 CM |
1538 | return ret; |
1539 | } | |
1540 | ||
8b712842 CM |
1541 | /* |
1542 | * called by the kthread helper functions to finally call the bio end_io | |
1543 | * functions. This is where read checksum verification actually happens | |
1544 | */ | |
1545 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1546 | { |
ce9adaa5 | 1547 | struct bio *bio; |
8b712842 CM |
1548 | struct end_io_wq *end_io_wq; |
1549 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1550 | int error; |
ce9adaa5 | 1551 | |
8b712842 CM |
1552 | end_io_wq = container_of(work, struct end_io_wq, work); |
1553 | bio = end_io_wq->bio; | |
1554 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1555 | |
cad321ad | 1556 | /* metadata bio reads are special because the whole tree block must |
8b712842 CM |
1557 | * be checksummed at once. This makes sure the entire block is in |
1558 | * ram and up to date before trying to verify things. For | |
1559 | * blocksize <= pagesize, it is basically a noop | |
1560 | */ | |
7b6d91da | 1561 | if (!(bio->bi_rw & REQ_WRITE) && end_io_wq->metadata && |
cad321ad | 1562 | !bio_ready_for_csum(bio)) { |
d20f7043 | 1563 | btrfs_queue_worker(&fs_info->endio_meta_workers, |
8b712842 CM |
1564 | &end_io_wq->work); |
1565 | return; | |
1566 | } | |
1567 | error = end_io_wq->error; | |
1568 | bio->bi_private = end_io_wq->private; | |
1569 | bio->bi_end_io = end_io_wq->end_io; | |
1570 | kfree(end_io_wq); | |
8b712842 | 1571 | bio_endio(bio, error); |
44b8bd7e CM |
1572 | } |
1573 | ||
a74a4b97 CM |
1574 | static int cleaner_kthread(void *arg) |
1575 | { | |
1576 | struct btrfs_root *root = arg; | |
1577 | ||
1578 | do { | |
a74a4b97 | 1579 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); |
76dda93c YZ |
1580 | |
1581 | if (!(root->fs_info->sb->s_flags & MS_RDONLY) && | |
1582 | mutex_trylock(&root->fs_info->cleaner_mutex)) { | |
24bbcf04 | 1583 | btrfs_run_delayed_iputs(root); |
76dda93c YZ |
1584 | btrfs_clean_old_snapshots(root); |
1585 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
4cb5300b | 1586 | btrfs_run_defrag_inodes(root->fs_info); |
76dda93c | 1587 | } |
a74a4b97 | 1588 | |
a0acae0e | 1589 | if (!try_to_freeze()) { |
a74a4b97 | 1590 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1591 | if (!kthread_should_stop()) |
1592 | schedule(); | |
a74a4b97 CM |
1593 | __set_current_state(TASK_RUNNING); |
1594 | } | |
1595 | } while (!kthread_should_stop()); | |
1596 | return 0; | |
1597 | } | |
1598 | ||
1599 | static int transaction_kthread(void *arg) | |
1600 | { | |
1601 | struct btrfs_root *root = arg; | |
1602 | struct btrfs_trans_handle *trans; | |
1603 | struct btrfs_transaction *cur; | |
8929ecfa | 1604 | u64 transid; |
a74a4b97 CM |
1605 | unsigned long now; |
1606 | unsigned long delay; | |
1607 | int ret; | |
1608 | ||
1609 | do { | |
a74a4b97 CM |
1610 | delay = HZ * 30; |
1611 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1612 | mutex_lock(&root->fs_info->transaction_kthread_mutex); | |
1613 | ||
a4abeea4 | 1614 | spin_lock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1615 | cur = root->fs_info->running_transaction; |
1616 | if (!cur) { | |
a4abeea4 | 1617 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1618 | goto sleep; |
1619 | } | |
31153d81 | 1620 | |
a74a4b97 | 1621 | now = get_seconds(); |
8929ecfa YZ |
1622 | if (!cur->blocked && |
1623 | (now < cur->start_time || now - cur->start_time < 30)) { | |
a4abeea4 | 1624 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1625 | delay = HZ * 5; |
1626 | goto sleep; | |
1627 | } | |
8929ecfa | 1628 | transid = cur->transid; |
a4abeea4 | 1629 | spin_unlock(&root->fs_info->trans_lock); |
56bec294 | 1630 | |
7a7eaa40 | 1631 | trans = btrfs_join_transaction(root); |
3612b495 | 1632 | BUG_ON(IS_ERR(trans)); |
8929ecfa YZ |
1633 | if (transid == trans->transid) { |
1634 | ret = btrfs_commit_transaction(trans, root); | |
1635 | BUG_ON(ret); | |
1636 | } else { | |
1637 | btrfs_end_transaction(trans, root); | |
1638 | } | |
a74a4b97 CM |
1639 | sleep: |
1640 | wake_up_process(root->fs_info->cleaner_kthread); | |
1641 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1642 | ||
a0acae0e | 1643 | if (!try_to_freeze()) { |
a74a4b97 | 1644 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1645 | if (!kthread_should_stop() && |
1646 | !btrfs_transaction_blocked(root->fs_info)) | |
1647 | schedule_timeout(delay); | |
a74a4b97 CM |
1648 | __set_current_state(TASK_RUNNING); |
1649 | } | |
1650 | } while (!kthread_should_stop()); | |
1651 | return 0; | |
1652 | } | |
1653 | ||
af31f5e5 CM |
1654 | /* |
1655 | * this will find the highest generation in the array of | |
1656 | * root backups. The index of the highest array is returned, | |
1657 | * or -1 if we can't find anything. | |
1658 | * | |
1659 | * We check to make sure the array is valid by comparing the | |
1660 | * generation of the latest root in the array with the generation | |
1661 | * in the super block. If they don't match we pitch it. | |
1662 | */ | |
1663 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1664 | { | |
1665 | u64 cur; | |
1666 | int newest_index = -1; | |
1667 | struct btrfs_root_backup *root_backup; | |
1668 | int i; | |
1669 | ||
1670 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1671 | root_backup = info->super_copy->super_roots + i; | |
1672 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1673 | if (cur == newest_gen) | |
1674 | newest_index = i; | |
1675 | } | |
1676 | ||
1677 | /* check to see if we actually wrapped around */ | |
1678 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
1679 | root_backup = info->super_copy->super_roots; | |
1680 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1681 | if (cur == newest_gen) | |
1682 | newest_index = 0; | |
1683 | } | |
1684 | return newest_index; | |
1685 | } | |
1686 | ||
1687 | ||
1688 | /* | |
1689 | * find the oldest backup so we know where to store new entries | |
1690 | * in the backup array. This will set the backup_root_index | |
1691 | * field in the fs_info struct | |
1692 | */ | |
1693 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
1694 | u64 newest_gen) | |
1695 | { | |
1696 | int newest_index = -1; | |
1697 | ||
1698 | newest_index = find_newest_super_backup(info, newest_gen); | |
1699 | /* if there was garbage in there, just move along */ | |
1700 | if (newest_index == -1) { | |
1701 | info->backup_root_index = 0; | |
1702 | } else { | |
1703 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1704 | } | |
1705 | } | |
1706 | ||
1707 | /* | |
1708 | * copy all the root pointers into the super backup array. | |
1709 | * this will bump the backup pointer by one when it is | |
1710 | * done | |
1711 | */ | |
1712 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1713 | { | |
1714 | int next_backup; | |
1715 | struct btrfs_root_backup *root_backup; | |
1716 | int last_backup; | |
1717 | ||
1718 | next_backup = info->backup_root_index; | |
1719 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1720 | BTRFS_NUM_BACKUP_ROOTS; | |
1721 | ||
1722 | /* | |
1723 | * just overwrite the last backup if we're at the same generation | |
1724 | * this happens only at umount | |
1725 | */ | |
1726 | root_backup = info->super_for_commit->super_roots + last_backup; | |
1727 | if (btrfs_backup_tree_root_gen(root_backup) == | |
1728 | btrfs_header_generation(info->tree_root->node)) | |
1729 | next_backup = last_backup; | |
1730 | ||
1731 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1732 | ||
1733 | /* | |
1734 | * make sure all of our padding and empty slots get zero filled | |
1735 | * regardless of which ones we use today | |
1736 | */ | |
1737 | memset(root_backup, 0, sizeof(*root_backup)); | |
1738 | ||
1739 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1740 | ||
1741 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1742 | btrfs_set_backup_tree_root_gen(root_backup, | |
1743 | btrfs_header_generation(info->tree_root->node)); | |
1744 | ||
1745 | btrfs_set_backup_tree_root_level(root_backup, | |
1746 | btrfs_header_level(info->tree_root->node)); | |
1747 | ||
1748 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1749 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1750 | btrfs_header_generation(info->chunk_root->node)); | |
1751 | btrfs_set_backup_chunk_root_level(root_backup, | |
1752 | btrfs_header_level(info->chunk_root->node)); | |
1753 | ||
1754 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
1755 | btrfs_set_backup_extent_root_gen(root_backup, | |
1756 | btrfs_header_generation(info->extent_root->node)); | |
1757 | btrfs_set_backup_extent_root_level(root_backup, | |
1758 | btrfs_header_level(info->extent_root->node)); | |
1759 | ||
7c7e82a7 CM |
1760 | /* |
1761 | * we might commit during log recovery, which happens before we set | |
1762 | * the fs_root. Make sure it is valid before we fill it in. | |
1763 | */ | |
1764 | if (info->fs_root && info->fs_root->node) { | |
1765 | btrfs_set_backup_fs_root(root_backup, | |
1766 | info->fs_root->node->start); | |
1767 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1768 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1769 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1770 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1771 | } |
af31f5e5 CM |
1772 | |
1773 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1774 | btrfs_set_backup_dev_root_gen(root_backup, | |
1775 | btrfs_header_generation(info->dev_root->node)); | |
1776 | btrfs_set_backup_dev_root_level(root_backup, | |
1777 | btrfs_header_level(info->dev_root->node)); | |
1778 | ||
1779 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
1780 | btrfs_set_backup_csum_root_gen(root_backup, | |
1781 | btrfs_header_generation(info->csum_root->node)); | |
1782 | btrfs_set_backup_csum_root_level(root_backup, | |
1783 | btrfs_header_level(info->csum_root->node)); | |
1784 | ||
1785 | btrfs_set_backup_total_bytes(root_backup, | |
1786 | btrfs_super_total_bytes(info->super_copy)); | |
1787 | btrfs_set_backup_bytes_used(root_backup, | |
1788 | btrfs_super_bytes_used(info->super_copy)); | |
1789 | btrfs_set_backup_num_devices(root_backup, | |
1790 | btrfs_super_num_devices(info->super_copy)); | |
1791 | ||
1792 | /* | |
1793 | * if we don't copy this out to the super_copy, it won't get remembered | |
1794 | * for the next commit | |
1795 | */ | |
1796 | memcpy(&info->super_copy->super_roots, | |
1797 | &info->super_for_commit->super_roots, | |
1798 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1799 | } | |
1800 | ||
1801 | /* | |
1802 | * this copies info out of the root backup array and back into | |
1803 | * the in-memory super block. It is meant to help iterate through | |
1804 | * the array, so you send it the number of backups you've already | |
1805 | * tried and the last backup index you used. | |
1806 | * | |
1807 | * this returns -1 when it has tried all the backups | |
1808 | */ | |
1809 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
1810 | struct btrfs_super_block *super, | |
1811 | int *num_backups_tried, int *backup_index) | |
1812 | { | |
1813 | struct btrfs_root_backup *root_backup; | |
1814 | int newest = *backup_index; | |
1815 | ||
1816 | if (*num_backups_tried == 0) { | |
1817 | u64 gen = btrfs_super_generation(super); | |
1818 | ||
1819 | newest = find_newest_super_backup(info, gen); | |
1820 | if (newest == -1) | |
1821 | return -1; | |
1822 | ||
1823 | *backup_index = newest; | |
1824 | *num_backups_tried = 1; | |
1825 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
1826 | /* we've tried all the backups, all done */ | |
1827 | return -1; | |
1828 | } else { | |
1829 | /* jump to the next oldest backup */ | |
1830 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1831 | BTRFS_NUM_BACKUP_ROOTS; | |
1832 | *backup_index = newest; | |
1833 | *num_backups_tried += 1; | |
1834 | } | |
1835 | root_backup = super->super_roots + newest; | |
1836 | ||
1837 | btrfs_set_super_generation(super, | |
1838 | btrfs_backup_tree_root_gen(root_backup)); | |
1839 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
1840 | btrfs_set_super_root_level(super, | |
1841 | btrfs_backup_tree_root_level(root_backup)); | |
1842 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
1843 | ||
1844 | /* | |
1845 | * fixme: the total bytes and num_devices need to match or we should | |
1846 | * need a fsck | |
1847 | */ | |
1848 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
1849 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
1850 | return 0; | |
1851 | } | |
1852 | ||
1853 | /* helper to cleanup tree roots */ | |
1854 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
1855 | { | |
1856 | free_extent_buffer(info->tree_root->node); | |
1857 | free_extent_buffer(info->tree_root->commit_root); | |
1858 | free_extent_buffer(info->dev_root->node); | |
1859 | free_extent_buffer(info->dev_root->commit_root); | |
1860 | free_extent_buffer(info->extent_root->node); | |
1861 | free_extent_buffer(info->extent_root->commit_root); | |
1862 | free_extent_buffer(info->csum_root->node); | |
1863 | free_extent_buffer(info->csum_root->commit_root); | |
1864 | ||
1865 | info->tree_root->node = NULL; | |
1866 | info->tree_root->commit_root = NULL; | |
1867 | info->dev_root->node = NULL; | |
1868 | info->dev_root->commit_root = NULL; | |
1869 | info->extent_root->node = NULL; | |
1870 | info->extent_root->commit_root = NULL; | |
1871 | info->csum_root->node = NULL; | |
1872 | info->csum_root->commit_root = NULL; | |
1873 | ||
1874 | if (chunk_root) { | |
1875 | free_extent_buffer(info->chunk_root->node); | |
1876 | free_extent_buffer(info->chunk_root->commit_root); | |
1877 | info->chunk_root->node = NULL; | |
1878 | info->chunk_root->commit_root = NULL; | |
1879 | } | |
1880 | } | |
1881 | ||
1882 | ||
ad2b2c80 AV |
1883 | int open_ctree(struct super_block *sb, |
1884 | struct btrfs_fs_devices *fs_devices, | |
1885 | char *options) | |
2e635a27 | 1886 | { |
db94535d CM |
1887 | u32 sectorsize; |
1888 | u32 nodesize; | |
1889 | u32 leafsize; | |
1890 | u32 blocksize; | |
87ee04eb | 1891 | u32 stripesize; |
84234f3a | 1892 | u64 generation; |
f2b636e8 | 1893 | u64 features; |
3de4586c | 1894 | struct btrfs_key location; |
a061fc8d | 1895 | struct buffer_head *bh; |
4d34b278 | 1896 | struct btrfs_super_block *disk_super; |
815745cf AV |
1897 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
1898 | struct btrfs_root *tree_root = fs_info->tree_root; | |
4d34b278 ID |
1899 | struct btrfs_root *extent_root; |
1900 | struct btrfs_root *csum_root; | |
1901 | struct btrfs_root *chunk_root; | |
1902 | struct btrfs_root *dev_root; | |
e02119d5 | 1903 | struct btrfs_root *log_tree_root; |
eb60ceac | 1904 | int ret; |
e58ca020 | 1905 | int err = -EINVAL; |
af31f5e5 CM |
1906 | int num_backups_tried = 0; |
1907 | int backup_index = 0; | |
4543df7e | 1908 | |
6f07e42e AV |
1909 | extent_root = fs_info->extent_root = btrfs_alloc_root(fs_info); |
1910 | csum_root = fs_info->csum_root = btrfs_alloc_root(fs_info); | |
1911 | chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info); | |
1912 | dev_root = fs_info->dev_root = btrfs_alloc_root(fs_info); | |
8790d502 | 1913 | |
4d34b278 | 1914 | if (!extent_root || !csum_root || !chunk_root || !dev_root) { |
39279cc3 CM |
1915 | err = -ENOMEM; |
1916 | goto fail; | |
1917 | } | |
76dda93c YZ |
1918 | |
1919 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
1920 | if (ret) { | |
1921 | err = ret; | |
1922 | goto fail; | |
1923 | } | |
1924 | ||
1925 | ret = setup_bdi(fs_info, &fs_info->bdi); | |
1926 | if (ret) { | |
1927 | err = ret; | |
1928 | goto fail_srcu; | |
1929 | } | |
1930 | ||
1931 | fs_info->btree_inode = new_inode(sb); | |
1932 | if (!fs_info->btree_inode) { | |
1933 | err = -ENOMEM; | |
1934 | goto fail_bdi; | |
1935 | } | |
1936 | ||
a6591715 | 1937 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
1561deda | 1938 | |
76dda93c | 1939 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
8fd17795 | 1940 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1941 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 1942 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
19c00ddc | 1943 | INIT_LIST_HEAD(&fs_info->hashers); |
ea8c2819 | 1944 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
5a3f23d5 | 1945 | INIT_LIST_HEAD(&fs_info->ordered_operations); |
11833d66 | 1946 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
1832a6d5 | 1947 | spin_lock_init(&fs_info->delalloc_lock); |
a4abeea4 | 1948 | spin_lock_init(&fs_info->trans_lock); |
31153d81 | 1949 | spin_lock_init(&fs_info->ref_cache_lock); |
76dda93c | 1950 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 1951 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 1952 | spin_lock_init(&fs_info->defrag_inodes_lock); |
2bf64758 | 1953 | spin_lock_init(&fs_info->free_chunk_lock); |
7585717f | 1954 | mutex_init(&fs_info->reloc_mutex); |
19c00ddc | 1955 | |
58176a96 | 1956 | init_completion(&fs_info->kobj_unregister); |
0b86a832 | 1957 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1958 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1959 | btrfs_mapping_init(&fs_info->mapping_tree); |
f0486c68 YZ |
1960 | btrfs_init_block_rsv(&fs_info->global_block_rsv); |
1961 | btrfs_init_block_rsv(&fs_info->delalloc_block_rsv); | |
1962 | btrfs_init_block_rsv(&fs_info->trans_block_rsv); | |
1963 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv); | |
1964 | btrfs_init_block_rsv(&fs_info->empty_block_rsv); | |
6d668dda | 1965 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv); |
cb03c743 | 1966 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 1967 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 1968 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 1969 | atomic_set(&fs_info->nr_async_bios, 0); |
4cb5300b | 1970 | atomic_set(&fs_info->defrag_running, 0); |
e20d96d6 | 1971 | fs_info->sb = sb; |
6f568d35 | 1972 | fs_info->max_inline = 8192 * 1024; |
9ed74f2d | 1973 | fs_info->metadata_ratio = 0; |
4cb5300b | 1974 | fs_info->defrag_inodes = RB_ROOT; |
a4abeea4 | 1975 | fs_info->trans_no_join = 0; |
2bf64758 | 1976 | fs_info->free_chunk_space = 0; |
c8b97818 | 1977 | |
90519d66 AJ |
1978 | /* readahead state */ |
1979 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); | |
1980 | spin_lock_init(&fs_info->reada_lock); | |
c8b97818 | 1981 | |
b34b086c CM |
1982 | fs_info->thread_pool_size = min_t(unsigned long, |
1983 | num_online_cpus() + 2, 8); | |
0afbaf8c | 1984 | |
3eaa2885 CM |
1985 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
1986 | spin_lock_init(&fs_info->ordered_extent_lock); | |
16cdcec7 MX |
1987 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
1988 | GFP_NOFS); | |
1989 | if (!fs_info->delayed_root) { | |
1990 | err = -ENOMEM; | |
1991 | goto fail_iput; | |
1992 | } | |
1993 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 1994 | |
a2de733c AJ |
1995 | mutex_init(&fs_info->scrub_lock); |
1996 | atomic_set(&fs_info->scrubs_running, 0); | |
1997 | atomic_set(&fs_info->scrub_pause_req, 0); | |
1998 | atomic_set(&fs_info->scrubs_paused, 0); | |
1999 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2000 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
2001 | init_rwsem(&fs_info->scrub_super_lock); | |
2002 | fs_info->scrub_workers_refcnt = 0; | |
a2de733c | 2003 | |
a061fc8d CM |
2004 | sb->s_blocksize = 4096; |
2005 | sb->s_blocksize_bits = blksize_bits(4096); | |
32a88aa1 | 2006 | sb->s_bdi = &fs_info->bdi; |
a061fc8d | 2007 | |
76dda93c | 2008 | fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; |
bfe86848 | 2009 | set_nlink(fs_info->btree_inode, 1); |
0afbaf8c CM |
2010 | /* |
2011 | * we set the i_size on the btree inode to the max possible int. | |
2012 | * the real end of the address space is determined by all of | |
2013 | * the devices in the system | |
2014 | */ | |
2015 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 2016 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
2017 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
2018 | ||
5d4f98a2 | 2019 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); |
d1310b2e | 2020 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
f993c883 | 2021 | fs_info->btree_inode->i_mapping); |
a8067e02 | 2022 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); |
d1310b2e CM |
2023 | |
2024 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 2025 | |
76dda93c YZ |
2026 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
2027 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
2028 | sizeof(struct btrfs_key)); | |
2029 | BTRFS_I(fs_info->btree_inode)->dummy_inode = 1; | |
c65ddb52 | 2030 | insert_inode_hash(fs_info->btree_inode); |
76dda93c | 2031 | |
0f9dd46c | 2032 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2033 | fs_info->block_group_cache_tree = RB_ROOT; |
0f9dd46c | 2034 | |
11833d66 | 2035 | extent_io_tree_init(&fs_info->freed_extents[0], |
f993c883 | 2036 | fs_info->btree_inode->i_mapping); |
11833d66 | 2037 | extent_io_tree_init(&fs_info->freed_extents[1], |
f993c883 | 2038 | fs_info->btree_inode->i_mapping); |
11833d66 | 2039 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
e66f709b | 2040 | fs_info->do_barriers = 1; |
e18e4809 | 2041 | |
39279cc3 | 2042 | |
5a3f23d5 | 2043 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2044 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2045 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2046 | mutex_init(&fs_info->transaction_kthread_mutex); |
2047 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 2048 | mutex_init(&fs_info->volume_mutex); |
276e680d | 2049 | init_rwsem(&fs_info->extent_commit_sem); |
c71bf099 | 2050 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2051 | init_rwsem(&fs_info->subvol_sem); |
fa9c0d79 CM |
2052 | |
2053 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); | |
2054 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2055 | ||
e6dcd2dc | 2056 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2057 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2058 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2059 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2060 | |
0b86a832 | 2061 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 2062 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2063 | |
a512bbf8 | 2064 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
20b45077 DY |
2065 | if (!bh) { |
2066 | err = -EINVAL; | |
16cdcec7 | 2067 | goto fail_alloc; |
20b45077 | 2068 | } |
39279cc3 | 2069 | |
6c41761f DS |
2070 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2071 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2072 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2073 | brelse(bh); |
5f39d397 | 2074 | |
6c41761f | 2075 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2076 | |
6c41761f | 2077 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2078 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2079 | goto fail_alloc; |
0f7d52f4 | 2080 | |
acce952b | 2081 | /* check FS state, whether FS is broken. */ |
2082 | fs_info->fs_state |= btrfs_super_flags(disk_super); | |
2083 | ||
2084 | btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); | |
2085 | ||
af31f5e5 CM |
2086 | /* |
2087 | * run through our array of backup supers and setup | |
2088 | * our ring pointer to the oldest one | |
2089 | */ | |
2090 | generation = btrfs_super_generation(disk_super); | |
2091 | find_oldest_super_backup(fs_info, generation); | |
2092 | ||
75e7cb7f LB |
2093 | /* |
2094 | * In the long term, we'll store the compression type in the super | |
2095 | * block, and it'll be used for per file compression control. | |
2096 | */ | |
2097 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2098 | ||
2b82032c YZ |
2099 | ret = btrfs_parse_options(tree_root, options); |
2100 | if (ret) { | |
2101 | err = ret; | |
16cdcec7 | 2102 | goto fail_alloc; |
2b82032c | 2103 | } |
dfe25020 | 2104 | |
f2b636e8 JB |
2105 | features = btrfs_super_incompat_flags(disk_super) & |
2106 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2107 | if (features) { | |
2108 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
2109 | "unsupported optional features (%Lx).\n", | |
21380931 | 2110 | (unsigned long long)features); |
f2b636e8 | 2111 | err = -EINVAL; |
16cdcec7 | 2112 | goto fail_alloc; |
f2b636e8 JB |
2113 | } |
2114 | ||
5d4f98a2 | 2115 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae LZ |
2116 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
2117 | if (tree_root->fs_info->compress_type & BTRFS_COMPRESS_LZO) | |
2118 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; | |
2119 | btrfs_set_super_incompat_flags(disk_super, features); | |
5d4f98a2 | 2120 | |
f2b636e8 JB |
2121 | features = btrfs_super_compat_ro_flags(disk_super) & |
2122 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
2123 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
2124 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
2125 | "unsupported option features (%Lx).\n", | |
21380931 | 2126 | (unsigned long long)features); |
f2b636e8 | 2127 | err = -EINVAL; |
16cdcec7 | 2128 | goto fail_alloc; |
f2b636e8 | 2129 | } |
61d92c32 CM |
2130 | |
2131 | btrfs_init_workers(&fs_info->generic_worker, | |
2132 | "genwork", 1, NULL); | |
2133 | ||
5443be45 | 2134 | btrfs_init_workers(&fs_info->workers, "worker", |
61d92c32 CM |
2135 | fs_info->thread_pool_size, |
2136 | &fs_info->generic_worker); | |
c8b97818 | 2137 | |
771ed689 | 2138 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
61d92c32 CM |
2139 | fs_info->thread_pool_size, |
2140 | &fs_info->generic_worker); | |
771ed689 | 2141 | |
5443be45 | 2142 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 | 2143 | min_t(u64, fs_devices->num_devices, |
61d92c32 CM |
2144 | fs_info->thread_pool_size), |
2145 | &fs_info->generic_worker); | |
61b49440 | 2146 | |
bab39bf9 JB |
2147 | btrfs_init_workers(&fs_info->caching_workers, "cache", |
2148 | 2, &fs_info->generic_worker); | |
2149 | ||
61b49440 CM |
2150 | /* a higher idle thresh on the submit workers makes it much more |
2151 | * likely that bios will be send down in a sane order to the | |
2152 | * devices | |
2153 | */ | |
2154 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 2155 | |
771ed689 | 2156 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 2157 | fs_info->workers.ordered = 1; |
61b49440 | 2158 | |
771ed689 CM |
2159 | fs_info->delalloc_workers.idle_thresh = 2; |
2160 | fs_info->delalloc_workers.ordered = 1; | |
2161 | ||
61d92c32 CM |
2162 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1, |
2163 | &fs_info->generic_worker); | |
5443be45 | 2164 | btrfs_init_workers(&fs_info->endio_workers, "endio", |
61d92c32 CM |
2165 | fs_info->thread_pool_size, |
2166 | &fs_info->generic_worker); | |
d20f7043 | 2167 | btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", |
61d92c32 CM |
2168 | fs_info->thread_pool_size, |
2169 | &fs_info->generic_worker); | |
cad321ad | 2170 | btrfs_init_workers(&fs_info->endio_meta_write_workers, |
61d92c32 CM |
2171 | "endio-meta-write", fs_info->thread_pool_size, |
2172 | &fs_info->generic_worker); | |
5443be45 | 2173 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", |
61d92c32 CM |
2174 | fs_info->thread_pool_size, |
2175 | &fs_info->generic_worker); | |
0cb59c99 JB |
2176 | btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write", |
2177 | 1, &fs_info->generic_worker); | |
16cdcec7 MX |
2178 | btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta", |
2179 | fs_info->thread_pool_size, | |
2180 | &fs_info->generic_worker); | |
90519d66 AJ |
2181 | btrfs_init_workers(&fs_info->readahead_workers, "readahead", |
2182 | fs_info->thread_pool_size, | |
2183 | &fs_info->generic_worker); | |
61b49440 CM |
2184 | |
2185 | /* | |
2186 | * endios are largely parallel and should have a very | |
2187 | * low idle thresh | |
2188 | */ | |
2189 | fs_info->endio_workers.idle_thresh = 4; | |
b51912c9 CM |
2190 | fs_info->endio_meta_workers.idle_thresh = 4; |
2191 | ||
9042846b CM |
2192 | fs_info->endio_write_workers.idle_thresh = 2; |
2193 | fs_info->endio_meta_write_workers.idle_thresh = 2; | |
90519d66 | 2194 | fs_info->readahead_workers.idle_thresh = 2; |
9042846b | 2195 | |
0dc3b84a JB |
2196 | /* |
2197 | * btrfs_start_workers can really only fail because of ENOMEM so just | |
2198 | * return -ENOMEM if any of these fail. | |
2199 | */ | |
2200 | ret = btrfs_start_workers(&fs_info->workers); | |
2201 | ret |= btrfs_start_workers(&fs_info->generic_worker); | |
2202 | ret |= btrfs_start_workers(&fs_info->submit_workers); | |
2203 | ret |= btrfs_start_workers(&fs_info->delalloc_workers); | |
2204 | ret |= btrfs_start_workers(&fs_info->fixup_workers); | |
2205 | ret |= btrfs_start_workers(&fs_info->endio_workers); | |
2206 | ret |= btrfs_start_workers(&fs_info->endio_meta_workers); | |
2207 | ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers); | |
2208 | ret |= btrfs_start_workers(&fs_info->endio_write_workers); | |
2209 | ret |= btrfs_start_workers(&fs_info->endio_freespace_worker); | |
2210 | ret |= btrfs_start_workers(&fs_info->delayed_workers); | |
2211 | ret |= btrfs_start_workers(&fs_info->caching_workers); | |
2212 | ret |= btrfs_start_workers(&fs_info->readahead_workers); | |
2213 | if (ret) { | |
2214 | ret = -ENOMEM; | |
2215 | goto fail_sb_buffer; | |
2216 | } | |
4543df7e | 2217 | |
4575c9cc | 2218 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
2219 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
2220 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 2221 | |
db94535d CM |
2222 | nodesize = btrfs_super_nodesize(disk_super); |
2223 | leafsize = btrfs_super_leafsize(disk_super); | |
2224 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 2225 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
2226 | tree_root->nodesize = nodesize; |
2227 | tree_root->leafsize = leafsize; | |
2228 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 2229 | tree_root->stripesize = stripesize; |
a061fc8d CM |
2230 | |
2231 | sb->s_blocksize = sectorsize; | |
2232 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 2233 | |
39279cc3 CM |
2234 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
2235 | sizeof(disk_super->magic))) { | |
d397712b | 2236 | printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
2237 | goto fail_sb_buffer; |
2238 | } | |
19c00ddc | 2239 | |
925baedd | 2240 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 2241 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 2242 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2243 | if (ret) { |
d397712b CM |
2244 | printk(KERN_WARNING "btrfs: failed to read the system " |
2245 | "array on %s\n", sb->s_id); | |
5d4f98a2 | 2246 | goto fail_sb_buffer; |
84eed90f | 2247 | } |
0b86a832 CM |
2248 | |
2249 | blocksize = btrfs_level_size(tree_root, | |
2250 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 2251 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
2252 | |
2253 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2254 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
2255 | ||
2256 | chunk_root->node = read_tree_block(chunk_root, | |
2257 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 2258 | blocksize, generation); |
0b86a832 | 2259 | BUG_ON(!chunk_root->node); |
83121942 DW |
2260 | if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { |
2261 | printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n", | |
2262 | sb->s_id); | |
af31f5e5 | 2263 | goto fail_tree_roots; |
83121942 | 2264 | } |
5d4f98a2 YZ |
2265 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
2266 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 2267 | |
e17cade2 | 2268 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
d397712b CM |
2269 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), |
2270 | BTRFS_UUID_SIZE); | |
e17cade2 | 2271 | |
925baedd | 2272 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 2273 | ret = btrfs_read_chunk_tree(chunk_root); |
925baedd | 2274 | mutex_unlock(&fs_info->chunk_mutex); |
2b82032c | 2275 | if (ret) { |
d397712b CM |
2276 | printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", |
2277 | sb->s_id); | |
af31f5e5 | 2278 | goto fail_tree_roots; |
2b82032c | 2279 | } |
0b86a832 | 2280 | |
dfe25020 CM |
2281 | btrfs_close_extra_devices(fs_devices); |
2282 | ||
af31f5e5 | 2283 | retry_root_backup: |
db94535d CM |
2284 | blocksize = btrfs_level_size(tree_root, |
2285 | btrfs_super_root_level(disk_super)); | |
84234f3a | 2286 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 2287 | |
e20d96d6 | 2288 | tree_root->node = read_tree_block(tree_root, |
db94535d | 2289 | btrfs_super_root(disk_super), |
84234f3a | 2290 | blocksize, generation); |
af31f5e5 CM |
2291 | if (!tree_root->node || |
2292 | !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { | |
83121942 DW |
2293 | printk(KERN_WARNING "btrfs: failed to read tree root on %s\n", |
2294 | sb->s_id); | |
af31f5e5 CM |
2295 | |
2296 | goto recovery_tree_root; | |
83121942 | 2297 | } |
af31f5e5 | 2298 | |
5d4f98a2 YZ |
2299 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
2300 | tree_root->commit_root = btrfs_root_node(tree_root); | |
db94535d CM |
2301 | |
2302 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 2303 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 2304 | if (ret) |
af31f5e5 | 2305 | goto recovery_tree_root; |
0b86a832 CM |
2306 | extent_root->track_dirty = 1; |
2307 | ||
2308 | ret = find_and_setup_root(tree_root, fs_info, | |
2309 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
0b86a832 | 2310 | if (ret) |
af31f5e5 | 2311 | goto recovery_tree_root; |
5d4f98a2 | 2312 | dev_root->track_dirty = 1; |
3768f368 | 2313 | |
d20f7043 CM |
2314 | ret = find_and_setup_root(tree_root, fs_info, |
2315 | BTRFS_CSUM_TREE_OBJECTID, csum_root); | |
2316 | if (ret) | |
af31f5e5 | 2317 | goto recovery_tree_root; |
d20f7043 CM |
2318 | |
2319 | csum_root->track_dirty = 1; | |
2320 | ||
8929ecfa YZ |
2321 | fs_info->generation = generation; |
2322 | fs_info->last_trans_committed = generation; | |
2323 | fs_info->data_alloc_profile = (u64)-1; | |
2324 | fs_info->metadata_alloc_profile = (u64)-1; | |
2325 | fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; | |
2326 | ||
c59021f8 | 2327 | ret = btrfs_init_space_info(fs_info); |
2328 | if (ret) { | |
2329 | printk(KERN_ERR "Failed to initial space info: %d\n", ret); | |
2330 | goto fail_block_groups; | |
2331 | } | |
2332 | ||
1b1d1f66 JB |
2333 | ret = btrfs_read_block_groups(extent_root); |
2334 | if (ret) { | |
2335 | printk(KERN_ERR "Failed to read block groups: %d\n", ret); | |
2336 | goto fail_block_groups; | |
2337 | } | |
9078a3e1 | 2338 | |
a74a4b97 CM |
2339 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
2340 | "btrfs-cleaner"); | |
57506d50 | 2341 | if (IS_ERR(fs_info->cleaner_kthread)) |
1b1d1f66 | 2342 | goto fail_block_groups; |
a74a4b97 CM |
2343 | |
2344 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
2345 | tree_root, | |
2346 | "btrfs-transaction"); | |
57506d50 | 2347 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 2348 | goto fail_cleaner; |
a74a4b97 | 2349 | |
c289811c CM |
2350 | if (!btrfs_test_opt(tree_root, SSD) && |
2351 | !btrfs_test_opt(tree_root, NOSSD) && | |
2352 | !fs_info->fs_devices->rotating) { | |
2353 | printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD " | |
2354 | "mode\n"); | |
2355 | btrfs_set_opt(fs_info->mount_opt, SSD); | |
2356 | } | |
2357 | ||
acce952b | 2358 | /* do not make disk changes in broken FS */ |
2359 | if (btrfs_super_log_root(disk_super) != 0 && | |
2360 | !(fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)) { | |
e02119d5 CM |
2361 | u64 bytenr = btrfs_super_log_root(disk_super); |
2362 | ||
7c2ca468 | 2363 | if (fs_devices->rw_devices == 0) { |
d397712b CM |
2364 | printk(KERN_WARNING "Btrfs log replay required " |
2365 | "on RO media\n"); | |
7c2ca468 CM |
2366 | err = -EIO; |
2367 | goto fail_trans_kthread; | |
2368 | } | |
e02119d5 CM |
2369 | blocksize = |
2370 | btrfs_level_size(tree_root, | |
2371 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 2372 | |
6f07e42e | 2373 | log_tree_root = btrfs_alloc_root(fs_info); |
676e4c86 DC |
2374 | if (!log_tree_root) { |
2375 | err = -ENOMEM; | |
2376 | goto fail_trans_kthread; | |
2377 | } | |
e02119d5 CM |
2378 | |
2379 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2380 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
2381 | ||
2382 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
2383 | blocksize, |
2384 | generation + 1); | |
e02119d5 CM |
2385 | ret = btrfs_recover_log_trees(log_tree_root); |
2386 | BUG_ON(ret); | |
e556ce2c YZ |
2387 | |
2388 | if (sb->s_flags & MS_RDONLY) { | |
2389 | ret = btrfs_commit_super(tree_root); | |
2390 | BUG_ON(ret); | |
2391 | } | |
e02119d5 | 2392 | } |
1a40e23b | 2393 | |
76dda93c YZ |
2394 | ret = btrfs_find_orphan_roots(tree_root); |
2395 | BUG_ON(ret); | |
2396 | ||
7c2ca468 | 2397 | if (!(sb->s_flags & MS_RDONLY)) { |
d68fc57b YZ |
2398 | ret = btrfs_cleanup_fs_roots(fs_info); |
2399 | BUG_ON(ret); | |
2400 | ||
5d4f98a2 | 2401 | ret = btrfs_recover_relocation(tree_root); |
d7ce5843 MX |
2402 | if (ret < 0) { |
2403 | printk(KERN_WARNING | |
2404 | "btrfs: failed to recover relocation\n"); | |
2405 | err = -EINVAL; | |
2406 | goto fail_trans_kthread; | |
2407 | } | |
7c2ca468 | 2408 | } |
1a40e23b | 2409 | |
3de4586c CM |
2410 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
2411 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2412 | location.offset = (u64)-1; | |
2413 | ||
3de4586c CM |
2414 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
2415 | if (!fs_info->fs_root) | |
7c2ca468 | 2416 | goto fail_trans_kthread; |
3140c9a3 DC |
2417 | if (IS_ERR(fs_info->fs_root)) { |
2418 | err = PTR_ERR(fs_info->fs_root); | |
2419 | goto fail_trans_kthread; | |
2420 | } | |
c289811c | 2421 | |
e3acc2a6 JB |
2422 | if (!(sb->s_flags & MS_RDONLY)) { |
2423 | down_read(&fs_info->cleanup_work_sem); | |
66b4ffd1 JB |
2424 | err = btrfs_orphan_cleanup(fs_info->fs_root); |
2425 | if (!err) | |
2426 | err = btrfs_orphan_cleanup(fs_info->tree_root); | |
e3acc2a6 | 2427 | up_read(&fs_info->cleanup_work_sem); |
66b4ffd1 JB |
2428 | if (err) { |
2429 | close_ctree(tree_root); | |
ad2b2c80 | 2430 | return err; |
66b4ffd1 | 2431 | } |
e3acc2a6 JB |
2432 | } |
2433 | ||
ad2b2c80 | 2434 | return 0; |
39279cc3 | 2435 | |
7c2ca468 CM |
2436 | fail_trans_kthread: |
2437 | kthread_stop(fs_info->transaction_kthread); | |
3f157a2f | 2438 | fail_cleaner: |
a74a4b97 | 2439 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
2440 | |
2441 | /* | |
2442 | * make sure we're done with the btree inode before we stop our | |
2443 | * kthreads | |
2444 | */ | |
2445 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
2446 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
2447 | ||
1b1d1f66 JB |
2448 | fail_block_groups: |
2449 | btrfs_free_block_groups(fs_info); | |
af31f5e5 CM |
2450 | |
2451 | fail_tree_roots: | |
2452 | free_root_pointers(fs_info, 1); | |
2453 | ||
39279cc3 | 2454 | fail_sb_buffer: |
61d92c32 | 2455 | btrfs_stop_workers(&fs_info->generic_worker); |
306c8b68 | 2456 | btrfs_stop_workers(&fs_info->readahead_workers); |
247e743c | 2457 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2458 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2459 | btrfs_stop_workers(&fs_info->workers); |
2460 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 2461 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 2462 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 2463 | btrfs_stop_workers(&fs_info->endio_write_workers); |
0cb59c99 | 2464 | btrfs_stop_workers(&fs_info->endio_freespace_worker); |
1cc127b5 | 2465 | btrfs_stop_workers(&fs_info->submit_workers); |
16cdcec7 | 2466 | btrfs_stop_workers(&fs_info->delayed_workers); |
bab39bf9 | 2467 | btrfs_stop_workers(&fs_info->caching_workers); |
16cdcec7 | 2468 | fail_alloc: |
4543df7e | 2469 | fail_iput: |
586e46e2 ID |
2470 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
2471 | ||
7c2ca468 | 2472 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
4543df7e | 2473 | iput(fs_info->btree_inode); |
ad081f14 | 2474 | fail_bdi: |
7e662854 | 2475 | bdi_destroy(&fs_info->bdi); |
76dda93c YZ |
2476 | fail_srcu: |
2477 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 2478 | fail: |
586e46e2 | 2479 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 2480 | return err; |
af31f5e5 CM |
2481 | |
2482 | recovery_tree_root: | |
af31f5e5 CM |
2483 | if (!btrfs_test_opt(tree_root, RECOVERY)) |
2484 | goto fail_tree_roots; | |
2485 | ||
2486 | free_root_pointers(fs_info, 0); | |
2487 | ||
2488 | /* don't use the log in recovery mode, it won't be valid */ | |
2489 | btrfs_set_super_log_root(disk_super, 0); | |
2490 | ||
2491 | /* we can't trust the free space cache either */ | |
2492 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2493 | ||
2494 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
2495 | &num_backups_tried, &backup_index); | |
2496 | if (ret == -1) | |
2497 | goto fail_block_groups; | |
2498 | goto retry_root_backup; | |
eb60ceac CM |
2499 | } |
2500 | ||
f2984462 CM |
2501 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
2502 | { | |
2503 | char b[BDEVNAME_SIZE]; | |
2504 | ||
2505 | if (uptodate) { | |
2506 | set_buffer_uptodate(bh); | |
2507 | } else { | |
7a36ddec | 2508 | printk_ratelimited(KERN_WARNING "lost page write due to " |
f2984462 CM |
2509 | "I/O error on %s\n", |
2510 | bdevname(bh->b_bdev, b)); | |
1259ab75 CM |
2511 | /* note, we dont' set_buffer_write_io_error because we have |
2512 | * our own ways of dealing with the IO errors | |
2513 | */ | |
f2984462 CM |
2514 | clear_buffer_uptodate(bh); |
2515 | } | |
2516 | unlock_buffer(bh); | |
2517 | put_bh(bh); | |
2518 | } | |
2519 | ||
a512bbf8 YZ |
2520 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
2521 | { | |
2522 | struct buffer_head *bh; | |
2523 | struct buffer_head *latest = NULL; | |
2524 | struct btrfs_super_block *super; | |
2525 | int i; | |
2526 | u64 transid = 0; | |
2527 | u64 bytenr; | |
2528 | ||
2529 | /* we would like to check all the supers, but that would make | |
2530 | * a btrfs mount succeed after a mkfs from a different FS. | |
2531 | * So, we need to add a special mount option to scan for | |
2532 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
2533 | */ | |
2534 | for (i = 0; i < 1; i++) { | |
2535 | bytenr = btrfs_sb_offset(i); | |
2536 | if (bytenr + 4096 >= i_size_read(bdev->bd_inode)) | |
2537 | break; | |
2538 | bh = __bread(bdev, bytenr / 4096, 4096); | |
2539 | if (!bh) | |
2540 | continue; | |
2541 | ||
2542 | super = (struct btrfs_super_block *)bh->b_data; | |
2543 | if (btrfs_super_bytenr(super) != bytenr || | |
2544 | strncmp((char *)(&super->magic), BTRFS_MAGIC, | |
2545 | sizeof(super->magic))) { | |
2546 | brelse(bh); | |
2547 | continue; | |
2548 | } | |
2549 | ||
2550 | if (!latest || btrfs_super_generation(super) > transid) { | |
2551 | brelse(latest); | |
2552 | latest = bh; | |
2553 | transid = btrfs_super_generation(super); | |
2554 | } else { | |
2555 | brelse(bh); | |
2556 | } | |
2557 | } | |
2558 | return latest; | |
2559 | } | |
2560 | ||
4eedeb75 HH |
2561 | /* |
2562 | * this should be called twice, once with wait == 0 and | |
2563 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
2564 | * we write are pinned. | |
2565 | * | |
2566 | * They are released when wait == 1 is done. | |
2567 | * max_mirrors must be the same for both runs, and it indicates how | |
2568 | * many supers on this one device should be written. | |
2569 | * | |
2570 | * max_mirrors == 0 means to write them all. | |
2571 | */ | |
a512bbf8 YZ |
2572 | static int write_dev_supers(struct btrfs_device *device, |
2573 | struct btrfs_super_block *sb, | |
2574 | int do_barriers, int wait, int max_mirrors) | |
2575 | { | |
2576 | struct buffer_head *bh; | |
2577 | int i; | |
2578 | int ret; | |
2579 | int errors = 0; | |
2580 | u32 crc; | |
2581 | u64 bytenr; | |
a512bbf8 YZ |
2582 | |
2583 | if (max_mirrors == 0) | |
2584 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
2585 | ||
a512bbf8 YZ |
2586 | for (i = 0; i < max_mirrors; i++) { |
2587 | bytenr = btrfs_sb_offset(i); | |
2588 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) | |
2589 | break; | |
2590 | ||
2591 | if (wait) { | |
2592 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
2593 | BTRFS_SUPER_INFO_SIZE); | |
2594 | BUG_ON(!bh); | |
a512bbf8 | 2595 | wait_on_buffer(bh); |
4eedeb75 HH |
2596 | if (!buffer_uptodate(bh)) |
2597 | errors++; | |
2598 | ||
2599 | /* drop our reference */ | |
2600 | brelse(bh); | |
2601 | ||
2602 | /* drop the reference from the wait == 0 run */ | |
2603 | brelse(bh); | |
2604 | continue; | |
a512bbf8 YZ |
2605 | } else { |
2606 | btrfs_set_super_bytenr(sb, bytenr); | |
2607 | ||
2608 | crc = ~(u32)0; | |
2609 | crc = btrfs_csum_data(NULL, (char *)sb + | |
2610 | BTRFS_CSUM_SIZE, crc, | |
2611 | BTRFS_SUPER_INFO_SIZE - | |
2612 | BTRFS_CSUM_SIZE); | |
2613 | btrfs_csum_final(crc, sb->csum); | |
2614 | ||
4eedeb75 HH |
2615 | /* |
2616 | * one reference for us, and we leave it for the | |
2617 | * caller | |
2618 | */ | |
a512bbf8 YZ |
2619 | bh = __getblk(device->bdev, bytenr / 4096, |
2620 | BTRFS_SUPER_INFO_SIZE); | |
2621 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); | |
2622 | ||
4eedeb75 | 2623 | /* one reference for submit_bh */ |
a512bbf8 | 2624 | get_bh(bh); |
4eedeb75 HH |
2625 | |
2626 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
2627 | lock_buffer(bh); |
2628 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
2629 | } | |
2630 | ||
387125fc CM |
2631 | /* |
2632 | * we fua the first super. The others we allow | |
2633 | * to go down lazy. | |
2634 | */ | |
2635 | ret = submit_bh(WRITE_FUA, bh); | |
4eedeb75 | 2636 | if (ret) |
a512bbf8 | 2637 | errors++; |
a512bbf8 YZ |
2638 | } |
2639 | return errors < i ? 0 : -1; | |
2640 | } | |
2641 | ||
387125fc CM |
2642 | /* |
2643 | * endio for the write_dev_flush, this will wake anyone waiting | |
2644 | * for the barrier when it is done | |
2645 | */ | |
2646 | static void btrfs_end_empty_barrier(struct bio *bio, int err) | |
2647 | { | |
2648 | if (err) { | |
2649 | if (err == -EOPNOTSUPP) | |
2650 | set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); | |
2651 | clear_bit(BIO_UPTODATE, &bio->bi_flags); | |
2652 | } | |
2653 | if (bio->bi_private) | |
2654 | complete(bio->bi_private); | |
2655 | bio_put(bio); | |
2656 | } | |
2657 | ||
2658 | /* | |
2659 | * trigger flushes for one the devices. If you pass wait == 0, the flushes are | |
2660 | * sent down. With wait == 1, it waits for the previous flush. | |
2661 | * | |
2662 | * any device where the flush fails with eopnotsupp are flagged as not-barrier | |
2663 | * capable | |
2664 | */ | |
2665 | static int write_dev_flush(struct btrfs_device *device, int wait) | |
2666 | { | |
2667 | struct bio *bio; | |
2668 | int ret = 0; | |
2669 | ||
2670 | if (device->nobarriers) | |
2671 | return 0; | |
2672 | ||
2673 | if (wait) { | |
2674 | bio = device->flush_bio; | |
2675 | if (!bio) | |
2676 | return 0; | |
2677 | ||
2678 | wait_for_completion(&device->flush_wait); | |
2679 | ||
2680 | if (bio_flagged(bio, BIO_EOPNOTSUPP)) { | |
2681 | printk("btrfs: disabling barriers on dev %s\n", | |
2682 | device->name); | |
2683 | device->nobarriers = 1; | |
2684 | } | |
2685 | if (!bio_flagged(bio, BIO_UPTODATE)) { | |
2686 | ret = -EIO; | |
2687 | } | |
2688 | ||
2689 | /* drop the reference from the wait == 0 run */ | |
2690 | bio_put(bio); | |
2691 | device->flush_bio = NULL; | |
2692 | ||
2693 | return ret; | |
2694 | } | |
2695 | ||
2696 | /* | |
2697 | * one reference for us, and we leave it for the | |
2698 | * caller | |
2699 | */ | |
2700 | device->flush_bio = NULL;; | |
2701 | bio = bio_alloc(GFP_NOFS, 0); | |
2702 | if (!bio) | |
2703 | return -ENOMEM; | |
2704 | ||
2705 | bio->bi_end_io = btrfs_end_empty_barrier; | |
2706 | bio->bi_bdev = device->bdev; | |
2707 | init_completion(&device->flush_wait); | |
2708 | bio->bi_private = &device->flush_wait; | |
2709 | device->flush_bio = bio; | |
2710 | ||
2711 | bio_get(bio); | |
2712 | submit_bio(WRITE_FLUSH, bio); | |
2713 | ||
2714 | return 0; | |
2715 | } | |
2716 | ||
2717 | /* | |
2718 | * send an empty flush down to each device in parallel, | |
2719 | * then wait for them | |
2720 | */ | |
2721 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
2722 | { | |
2723 | struct list_head *head; | |
2724 | struct btrfs_device *dev; | |
2725 | int errors = 0; | |
2726 | int ret; | |
2727 | ||
2728 | /* send down all the barriers */ | |
2729 | head = &info->fs_devices->devices; | |
2730 | list_for_each_entry_rcu(dev, head, dev_list) { | |
2731 | if (!dev->bdev) { | |
2732 | errors++; | |
2733 | continue; | |
2734 | } | |
2735 | if (!dev->in_fs_metadata || !dev->writeable) | |
2736 | continue; | |
2737 | ||
2738 | ret = write_dev_flush(dev, 0); | |
2739 | if (ret) | |
2740 | errors++; | |
2741 | } | |
2742 | ||
2743 | /* wait for all the barriers */ | |
2744 | list_for_each_entry_rcu(dev, head, dev_list) { | |
2745 | if (!dev->bdev) { | |
2746 | errors++; | |
2747 | continue; | |
2748 | } | |
2749 | if (!dev->in_fs_metadata || !dev->writeable) | |
2750 | continue; | |
2751 | ||
2752 | ret = write_dev_flush(dev, 1); | |
2753 | if (ret) | |
2754 | errors++; | |
2755 | } | |
2756 | if (errors) | |
2757 | return -EIO; | |
2758 | return 0; | |
2759 | } | |
2760 | ||
a512bbf8 | 2761 | int write_all_supers(struct btrfs_root *root, int max_mirrors) |
f2984462 | 2762 | { |
e5e9a520 | 2763 | struct list_head *head; |
f2984462 | 2764 | struct btrfs_device *dev; |
a061fc8d | 2765 | struct btrfs_super_block *sb; |
f2984462 | 2766 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
2767 | int ret; |
2768 | int do_barriers; | |
a236aed1 CM |
2769 | int max_errors; |
2770 | int total_errors = 0; | |
a061fc8d | 2771 | u64 flags; |
f2984462 | 2772 | |
6c41761f | 2773 | max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; |
f2984462 | 2774 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
af31f5e5 | 2775 | backup_super_roots(root->fs_info); |
f2984462 | 2776 | |
6c41761f | 2777 | sb = root->fs_info->super_for_commit; |
a061fc8d | 2778 | dev_item = &sb->dev_item; |
e5e9a520 | 2779 | |
174ba509 | 2780 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); |
e5e9a520 | 2781 | head = &root->fs_info->fs_devices->devices; |
387125fc CM |
2782 | |
2783 | if (do_barriers) | |
2784 | barrier_all_devices(root->fs_info); | |
2785 | ||
1f78160c | 2786 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
2787 | if (!dev->bdev) { |
2788 | total_errors++; | |
2789 | continue; | |
2790 | } | |
2b82032c | 2791 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2792 | continue; |
2793 | ||
2b82032c | 2794 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
2795 | btrfs_set_stack_device_type(dev_item, dev->type); |
2796 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
2797 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
2798 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
2799 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
2800 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
2801 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
2802 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 2803 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 2804 | |
a061fc8d CM |
2805 | flags = btrfs_super_flags(sb); |
2806 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
2807 | ||
a512bbf8 | 2808 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
2809 | if (ret) |
2810 | total_errors++; | |
f2984462 | 2811 | } |
a236aed1 | 2812 | if (total_errors > max_errors) { |
d397712b CM |
2813 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2814 | total_errors); | |
a236aed1 CM |
2815 | BUG(); |
2816 | } | |
f2984462 | 2817 | |
a512bbf8 | 2818 | total_errors = 0; |
1f78160c | 2819 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
2820 | if (!dev->bdev) |
2821 | continue; | |
2b82032c | 2822 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2823 | continue; |
2824 | ||
a512bbf8 YZ |
2825 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
2826 | if (ret) | |
2827 | total_errors++; | |
f2984462 | 2828 | } |
174ba509 | 2829 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 2830 | if (total_errors > max_errors) { |
d397712b CM |
2831 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2832 | total_errors); | |
a236aed1 CM |
2833 | BUG(); |
2834 | } | |
f2984462 CM |
2835 | return 0; |
2836 | } | |
2837 | ||
a512bbf8 YZ |
2838 | int write_ctree_super(struct btrfs_trans_handle *trans, |
2839 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 2840 | { |
e66f709b | 2841 | int ret; |
5f39d397 | 2842 | |
a512bbf8 | 2843 | ret = write_all_supers(root, max_mirrors); |
5f39d397 | 2844 | return ret; |
cfaa7295 CM |
2845 | } |
2846 | ||
5eda7b5e | 2847 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f | 2848 | { |
4df27c4d | 2849 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
2850 | radix_tree_delete(&fs_info->fs_roots_radix, |
2851 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 2852 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
2853 | |
2854 | if (btrfs_root_refs(&root->root_item) == 0) | |
2855 | synchronize_srcu(&fs_info->subvol_srcu); | |
2856 | ||
581bb050 LZ |
2857 | __btrfs_remove_free_space_cache(root->free_ino_pinned); |
2858 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d YZ |
2859 | free_fs_root(root); |
2860 | return 0; | |
2861 | } | |
2862 | ||
2863 | static void free_fs_root(struct btrfs_root *root) | |
2864 | { | |
82d5902d | 2865 | iput(root->cache_inode); |
4df27c4d | 2866 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
0ee5dc67 AV |
2867 | if (root->anon_dev) |
2868 | free_anon_bdev(root->anon_dev); | |
4df27c4d YZ |
2869 | free_extent_buffer(root->node); |
2870 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
2871 | kfree(root->free_ino_ctl); |
2872 | kfree(root->free_ino_pinned); | |
d397712b | 2873 | kfree(root->name); |
2619ba1f | 2874 | kfree(root); |
2619ba1f CM |
2875 | } |
2876 | ||
35b7e476 | 2877 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
2878 | { |
2879 | int ret; | |
2880 | struct btrfs_root *gang[8]; | |
2881 | int i; | |
2882 | ||
76dda93c YZ |
2883 | while (!list_empty(&fs_info->dead_roots)) { |
2884 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2885 | struct btrfs_root, root_list); | |
2886 | list_del(&gang[0]->root_list); | |
2887 | ||
2888 | if (gang[0]->in_radix) { | |
2889 | btrfs_free_fs_root(fs_info, gang[0]); | |
2890 | } else { | |
2891 | free_extent_buffer(gang[0]->node); | |
2892 | free_extent_buffer(gang[0]->commit_root); | |
2893 | kfree(gang[0]); | |
2894 | } | |
2895 | } | |
2896 | ||
d397712b | 2897 | while (1) { |
0f7d52f4 CM |
2898 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
2899 | (void **)gang, 0, | |
2900 | ARRAY_SIZE(gang)); | |
2901 | if (!ret) | |
2902 | break; | |
2619ba1f | 2903 | for (i = 0; i < ret; i++) |
5eda7b5e | 2904 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
2905 | } |
2906 | return 0; | |
2907 | } | |
b4100d64 | 2908 | |
c146afad | 2909 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 2910 | { |
c146afad YZ |
2911 | u64 root_objectid = 0; |
2912 | struct btrfs_root *gang[8]; | |
2913 | int i; | |
3768f368 | 2914 | int ret; |
e089f05c | 2915 | |
c146afad YZ |
2916 | while (1) { |
2917 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2918 | (void **)gang, root_objectid, | |
2919 | ARRAY_SIZE(gang)); | |
2920 | if (!ret) | |
2921 | break; | |
5d4f98a2 YZ |
2922 | |
2923 | root_objectid = gang[ret - 1]->root_key.objectid + 1; | |
c146afad | 2924 | for (i = 0; i < ret; i++) { |
66b4ffd1 JB |
2925 | int err; |
2926 | ||
c146afad | 2927 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
2928 | err = btrfs_orphan_cleanup(gang[i]); |
2929 | if (err) | |
2930 | return err; | |
c146afad YZ |
2931 | } |
2932 | root_objectid++; | |
2933 | } | |
2934 | return 0; | |
2935 | } | |
a2135011 | 2936 | |
c146afad YZ |
2937 | int btrfs_commit_super(struct btrfs_root *root) |
2938 | { | |
2939 | struct btrfs_trans_handle *trans; | |
2940 | int ret; | |
a74a4b97 | 2941 | |
c146afad | 2942 | mutex_lock(&root->fs_info->cleaner_mutex); |
24bbcf04 | 2943 | btrfs_run_delayed_iputs(root); |
a74a4b97 | 2944 | btrfs_clean_old_snapshots(root); |
c146afad | 2945 | mutex_unlock(&root->fs_info->cleaner_mutex); |
c71bf099 YZ |
2946 | |
2947 | /* wait until ongoing cleanup work done */ | |
2948 | down_write(&root->fs_info->cleanup_work_sem); | |
2949 | up_write(&root->fs_info->cleanup_work_sem); | |
2950 | ||
7a7eaa40 | 2951 | trans = btrfs_join_transaction(root); |
3612b495 TI |
2952 | if (IS_ERR(trans)) |
2953 | return PTR_ERR(trans); | |
54aa1f4d | 2954 | ret = btrfs_commit_transaction(trans, root); |
c146afad YZ |
2955 | BUG_ON(ret); |
2956 | /* run commit again to drop the original snapshot */ | |
7a7eaa40 | 2957 | trans = btrfs_join_transaction(root); |
3612b495 TI |
2958 | if (IS_ERR(trans)) |
2959 | return PTR_ERR(trans); | |
79154b1b CM |
2960 | btrfs_commit_transaction(trans, root); |
2961 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 2962 | BUG_ON(ret); |
d6bfde87 | 2963 | |
a512bbf8 | 2964 | ret = write_ctree_super(NULL, root, 0); |
c146afad YZ |
2965 | return ret; |
2966 | } | |
2967 | ||
2968 | int close_ctree(struct btrfs_root *root) | |
2969 | { | |
2970 | struct btrfs_fs_info *fs_info = root->fs_info; | |
2971 | int ret; | |
2972 | ||
2973 | fs_info->closing = 1; | |
2974 | smp_mb(); | |
2975 | ||
a2de733c | 2976 | btrfs_scrub_cancel(root); |
4cb5300b CM |
2977 | |
2978 | /* wait for any defraggers to finish */ | |
2979 | wait_event(fs_info->transaction_wait, | |
2980 | (atomic_read(&fs_info->defrag_running) == 0)); | |
2981 | ||
2982 | /* clear out the rbtree of defraggable inodes */ | |
e3029d9f | 2983 | btrfs_run_defrag_inodes(fs_info); |
4cb5300b | 2984 | |
acce952b | 2985 | /* |
2986 | * Here come 2 situations when btrfs is broken to flip readonly: | |
2987 | * | |
2988 | * 1. when btrfs flips readonly somewhere else before | |
2989 | * btrfs_commit_super, sb->s_flags has MS_RDONLY flag, | |
2990 | * and btrfs will skip to write sb directly to keep | |
2991 | * ERROR state on disk. | |
2992 | * | |
2993 | * 2. when btrfs flips readonly just in btrfs_commit_super, | |
ae0e47f0 | 2994 | * and in such case, btrfs cannot write sb via btrfs_commit_super, |
acce952b | 2995 | * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag, |
2996 | * btrfs will cleanup all FS resources first and write sb then. | |
2997 | */ | |
c146afad | 2998 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
acce952b | 2999 | ret = btrfs_commit_super(root); |
3000 | if (ret) | |
3001 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
3002 | } | |
3003 | ||
3004 | if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) { | |
3005 | ret = btrfs_error_commit_super(root); | |
d397712b CM |
3006 | if (ret) |
3007 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
c146afad | 3008 | } |
0f7d52f4 | 3009 | |
300e4f8a JB |
3010 | btrfs_put_block_group_cache(fs_info); |
3011 | ||
e3029d9f AV |
3012 | kthread_stop(fs_info->transaction_kthread); |
3013 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 3014 | |
f25784b3 YZ |
3015 | fs_info->closing = 2; |
3016 | smp_mb(); | |
3017 | ||
b0c68f8b | 3018 | if (fs_info->delalloc_bytes) { |
d397712b | 3019 | printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n", |
21380931 | 3020 | (unsigned long long)fs_info->delalloc_bytes); |
b0c68f8b | 3021 | } |
31153d81 | 3022 | if (fs_info->total_ref_cache_size) { |
d397712b CM |
3023 | printk(KERN_INFO "btrfs: at umount reference cache size %llu\n", |
3024 | (unsigned long long)fs_info->total_ref_cache_size); | |
31153d81 | 3025 | } |
bcc63abb | 3026 | |
5d4f98a2 YZ |
3027 | free_extent_buffer(fs_info->extent_root->node); |
3028 | free_extent_buffer(fs_info->extent_root->commit_root); | |
3029 | free_extent_buffer(fs_info->tree_root->node); | |
3030 | free_extent_buffer(fs_info->tree_root->commit_root); | |
e3029d9f AV |
3031 | free_extent_buffer(fs_info->chunk_root->node); |
3032 | free_extent_buffer(fs_info->chunk_root->commit_root); | |
3033 | free_extent_buffer(fs_info->dev_root->node); | |
3034 | free_extent_buffer(fs_info->dev_root->commit_root); | |
3035 | free_extent_buffer(fs_info->csum_root->node); | |
3036 | free_extent_buffer(fs_info->csum_root->commit_root); | |
3037 | ||
3038 | btrfs_free_block_groups(fs_info); | |
d10c5f31 | 3039 | |
c146afad | 3040 | del_fs_roots(fs_info); |
d10c5f31 | 3041 | |
c146afad | 3042 | iput(fs_info->btree_inode); |
9ad6b7bc | 3043 | |
61d92c32 | 3044 | btrfs_stop_workers(&fs_info->generic_worker); |
247e743c | 3045 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 3046 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
3047 | btrfs_stop_workers(&fs_info->workers); |
3048 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 3049 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 3050 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 3051 | btrfs_stop_workers(&fs_info->endio_write_workers); |
0cb59c99 | 3052 | btrfs_stop_workers(&fs_info->endio_freespace_worker); |
1cc127b5 | 3053 | btrfs_stop_workers(&fs_info->submit_workers); |
16cdcec7 | 3054 | btrfs_stop_workers(&fs_info->delayed_workers); |
bab39bf9 | 3055 | btrfs_stop_workers(&fs_info->caching_workers); |
90519d66 | 3056 | btrfs_stop_workers(&fs_info->readahead_workers); |
d6bfde87 | 3057 | |
dfe25020 | 3058 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 3059 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 3060 | |
04160088 | 3061 | bdi_destroy(&fs_info->bdi); |
76dda93c | 3062 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 3063 | |
eb60ceac CM |
3064 | return 0; |
3065 | } | |
3066 | ||
1259ab75 | 3067 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 3068 | { |
1259ab75 | 3069 | int ret; |
810191ff | 3070 | struct inode *btree_inode = buf->first_page->mapping->host; |
1259ab75 | 3071 | |
2ac55d41 JB |
3072 | ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf, |
3073 | NULL); | |
1259ab75 CM |
3074 | if (!ret) |
3075 | return ret; | |
3076 | ||
3077 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
3078 | parent_transid); | |
3079 | return !ret; | |
5f39d397 CM |
3080 | } |
3081 | ||
3082 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 3083 | { |
810191ff | 3084 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 3085 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
3086 | buf); |
3087 | } | |
6702ed49 | 3088 | |
5f39d397 CM |
3089 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
3090 | { | |
810191ff | 3091 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
5f39d397 CM |
3092 | u64 transid = btrfs_header_generation(buf); |
3093 | struct inode *btree_inode = root->fs_info->btree_inode; | |
b9473439 | 3094 | int was_dirty; |
b4ce94de | 3095 | |
b9447ef8 | 3096 | btrfs_assert_tree_locked(buf); |
ccd467d6 | 3097 | if (transid != root->fs_info->generation) { |
d397712b CM |
3098 | printk(KERN_CRIT "btrfs transid mismatch buffer %llu, " |
3099 | "found %llu running %llu\n", | |
db94535d | 3100 | (unsigned long long)buf->start, |
d397712b CM |
3101 | (unsigned long long)transid, |
3102 | (unsigned long long)root->fs_info->generation); | |
ccd467d6 CM |
3103 | WARN_ON(1); |
3104 | } | |
b9473439 CM |
3105 | was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
3106 | buf); | |
3107 | if (!was_dirty) { | |
3108 | spin_lock(&root->fs_info->delalloc_lock); | |
3109 | root->fs_info->dirty_metadata_bytes += buf->len; | |
3110 | spin_unlock(&root->fs_info->delalloc_lock); | |
3111 | } | |
eb60ceac CM |
3112 | } |
3113 | ||
d3c2fdcf | 3114 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
16cdcec7 MX |
3115 | { |
3116 | /* | |
3117 | * looks as though older kernels can get into trouble with | |
3118 | * this code, they end up stuck in balance_dirty_pages forever | |
3119 | */ | |
3120 | u64 num_dirty; | |
3121 | unsigned long thresh = 32 * 1024 * 1024; | |
3122 | ||
3123 | if (current->flags & PF_MEMALLOC) | |
3124 | return; | |
3125 | ||
3126 | btrfs_balance_delayed_items(root); | |
3127 | ||
3128 | num_dirty = root->fs_info->dirty_metadata_bytes; | |
3129 | ||
3130 | if (num_dirty > thresh) { | |
3131 | balance_dirty_pages_ratelimited_nr( | |
3132 | root->fs_info->btree_inode->i_mapping, 1); | |
3133 | } | |
3134 | return; | |
3135 | } | |
3136 | ||
3137 | void __btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) | |
35b7e476 | 3138 | { |
188de649 CM |
3139 | /* |
3140 | * looks as though older kernels can get into trouble with | |
3141 | * this code, they end up stuck in balance_dirty_pages forever | |
3142 | */ | |
d6bfde87 | 3143 | u64 num_dirty; |
771ed689 | 3144 | unsigned long thresh = 32 * 1024 * 1024; |
d6bfde87 | 3145 | |
6933c02e | 3146 | if (current->flags & PF_MEMALLOC) |
d6bfde87 CM |
3147 | return; |
3148 | ||
585ad2c3 CM |
3149 | num_dirty = root->fs_info->dirty_metadata_bytes; |
3150 | ||
d6bfde87 CM |
3151 | if (num_dirty > thresh) { |
3152 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 3153 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 3154 | } |
188de649 | 3155 | return; |
35b7e476 | 3156 | } |
6b80053d | 3157 | |
ca7a79ad | 3158 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 3159 | { |
810191ff | 3160 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
ce9adaa5 | 3161 | int ret; |
ca7a79ad | 3162 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
d397712b | 3163 | if (ret == 0) |
b4ce94de | 3164 | set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags); |
ce9adaa5 | 3165 | return ret; |
6b80053d | 3166 | } |
0da5468f | 3167 | |
01d658f2 CM |
3168 | static int btree_lock_page_hook(struct page *page, void *data, |
3169 | void (*flush_fn)(void *)) | |
4bef0848 CM |
3170 | { |
3171 | struct inode *inode = page->mapping->host; | |
b9473439 | 3172 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4bef0848 CM |
3173 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
3174 | struct extent_buffer *eb; | |
3175 | unsigned long len; | |
3176 | u64 bytenr = page_offset(page); | |
3177 | ||
3178 | if (page->private == EXTENT_PAGE_PRIVATE) | |
3179 | goto out; | |
3180 | ||
3181 | len = page->private >> 2; | |
f09d1f60 | 3182 | eb = find_extent_buffer(io_tree, bytenr, len); |
4bef0848 CM |
3183 | if (!eb) |
3184 | goto out; | |
3185 | ||
01d658f2 CM |
3186 | if (!btrfs_try_tree_write_lock(eb)) { |
3187 | flush_fn(data); | |
3188 | btrfs_tree_lock(eb); | |
3189 | } | |
4bef0848 | 3190 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); |
b9473439 CM |
3191 | |
3192 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) { | |
3193 | spin_lock(&root->fs_info->delalloc_lock); | |
3194 | if (root->fs_info->dirty_metadata_bytes >= eb->len) | |
3195 | root->fs_info->dirty_metadata_bytes -= eb->len; | |
3196 | else | |
3197 | WARN_ON(1); | |
3198 | spin_unlock(&root->fs_info->delalloc_lock); | |
3199 | } | |
3200 | ||
4bef0848 CM |
3201 | btrfs_tree_unlock(eb); |
3202 | free_extent_buffer(eb); | |
3203 | out: | |
01d658f2 CM |
3204 | if (!trylock_page(page)) { |
3205 | flush_fn(data); | |
3206 | lock_page(page); | |
3207 | } | |
4bef0848 CM |
3208 | return 0; |
3209 | } | |
3210 | ||
acce952b | 3211 | static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
3212 | int read_only) | |
3213 | { | |
3214 | if (read_only) | |
3215 | return; | |
3216 | ||
3217 | if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) | |
3218 | printk(KERN_WARNING "warning: mount fs with errors, " | |
3219 | "running btrfsck is recommended\n"); | |
3220 | } | |
3221 | ||
3222 | int btrfs_error_commit_super(struct btrfs_root *root) | |
3223 | { | |
3224 | int ret; | |
3225 | ||
3226 | mutex_lock(&root->fs_info->cleaner_mutex); | |
3227 | btrfs_run_delayed_iputs(root); | |
3228 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
3229 | ||
3230 | down_write(&root->fs_info->cleanup_work_sem); | |
3231 | up_write(&root->fs_info->cleanup_work_sem); | |
3232 | ||
3233 | /* cleanup FS via transaction */ | |
3234 | btrfs_cleanup_transaction(root); | |
3235 | ||
3236 | ret = write_ctree_super(NULL, root, 0); | |
3237 | ||
3238 | return ret; | |
3239 | } | |
3240 | ||
3241 | static int btrfs_destroy_ordered_operations(struct btrfs_root *root) | |
3242 | { | |
3243 | struct btrfs_inode *btrfs_inode; | |
3244 | struct list_head splice; | |
3245 | ||
3246 | INIT_LIST_HEAD(&splice); | |
3247 | ||
3248 | mutex_lock(&root->fs_info->ordered_operations_mutex); | |
3249 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3250 | ||
3251 | list_splice_init(&root->fs_info->ordered_operations, &splice); | |
3252 | while (!list_empty(&splice)) { | |
3253 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3254 | ordered_operations); | |
3255 | ||
3256 | list_del_init(&btrfs_inode->ordered_operations); | |
3257 | ||
3258 | btrfs_invalidate_inodes(btrfs_inode->root); | |
3259 | } | |
3260 | ||
3261 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3262 | mutex_unlock(&root->fs_info->ordered_operations_mutex); | |
3263 | ||
3264 | return 0; | |
3265 | } | |
3266 | ||
3267 | static int btrfs_destroy_ordered_extents(struct btrfs_root *root) | |
3268 | { | |
3269 | struct list_head splice; | |
3270 | struct btrfs_ordered_extent *ordered; | |
3271 | struct inode *inode; | |
3272 | ||
3273 | INIT_LIST_HEAD(&splice); | |
3274 | ||
3275 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3276 | ||
3277 | list_splice_init(&root->fs_info->ordered_extents, &splice); | |
3278 | while (!list_empty(&splice)) { | |
3279 | ordered = list_entry(splice.next, struct btrfs_ordered_extent, | |
3280 | root_extent_list); | |
3281 | ||
3282 | list_del_init(&ordered->root_extent_list); | |
3283 | atomic_inc(&ordered->refs); | |
3284 | ||
3285 | /* the inode may be getting freed (in sys_unlink path). */ | |
3286 | inode = igrab(ordered->inode); | |
3287 | ||
3288 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3289 | if (inode) | |
3290 | iput(inode); | |
3291 | ||
3292 | atomic_set(&ordered->refs, 1); | |
3293 | btrfs_put_ordered_extent(ordered); | |
3294 | ||
3295 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3296 | } | |
3297 | ||
3298 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3299 | ||
3300 | return 0; | |
3301 | } | |
3302 | ||
3303 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, | |
3304 | struct btrfs_root *root) | |
3305 | { | |
3306 | struct rb_node *node; | |
3307 | struct btrfs_delayed_ref_root *delayed_refs; | |
3308 | struct btrfs_delayed_ref_node *ref; | |
3309 | int ret = 0; | |
3310 | ||
3311 | delayed_refs = &trans->delayed_refs; | |
3312 | ||
3313 | spin_lock(&delayed_refs->lock); | |
3314 | if (delayed_refs->num_entries == 0) { | |
cfece4db | 3315 | spin_unlock(&delayed_refs->lock); |
acce952b | 3316 | printk(KERN_INFO "delayed_refs has NO entry\n"); |
3317 | return ret; | |
3318 | } | |
3319 | ||
3320 | node = rb_first(&delayed_refs->root); | |
3321 | while (node) { | |
3322 | ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); | |
3323 | node = rb_next(node); | |
3324 | ||
3325 | ref->in_tree = 0; | |
3326 | rb_erase(&ref->rb_node, &delayed_refs->root); | |
3327 | delayed_refs->num_entries--; | |
3328 | ||
3329 | atomic_set(&ref->refs, 1); | |
3330 | if (btrfs_delayed_ref_is_head(ref)) { | |
3331 | struct btrfs_delayed_ref_head *head; | |
3332 | ||
3333 | head = btrfs_delayed_node_to_head(ref); | |
3334 | mutex_lock(&head->mutex); | |
3335 | kfree(head->extent_op); | |
3336 | delayed_refs->num_heads--; | |
3337 | if (list_empty(&head->cluster)) | |
3338 | delayed_refs->num_heads_ready--; | |
3339 | list_del_init(&head->cluster); | |
3340 | mutex_unlock(&head->mutex); | |
3341 | } | |
3342 | ||
3343 | spin_unlock(&delayed_refs->lock); | |
3344 | btrfs_put_delayed_ref(ref); | |
3345 | ||
3346 | cond_resched(); | |
3347 | spin_lock(&delayed_refs->lock); | |
3348 | } | |
3349 | ||
3350 | spin_unlock(&delayed_refs->lock); | |
3351 | ||
3352 | return ret; | |
3353 | } | |
3354 | ||
3355 | static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t) | |
3356 | { | |
3357 | struct btrfs_pending_snapshot *snapshot; | |
3358 | struct list_head splice; | |
3359 | ||
3360 | INIT_LIST_HEAD(&splice); | |
3361 | ||
3362 | list_splice_init(&t->pending_snapshots, &splice); | |
3363 | ||
3364 | while (!list_empty(&splice)) { | |
3365 | snapshot = list_entry(splice.next, | |
3366 | struct btrfs_pending_snapshot, | |
3367 | list); | |
3368 | ||
3369 | list_del_init(&snapshot->list); | |
3370 | ||
3371 | kfree(snapshot); | |
3372 | } | |
3373 | ||
3374 | return 0; | |
3375 | } | |
3376 | ||
3377 | static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root) | |
3378 | { | |
3379 | struct btrfs_inode *btrfs_inode; | |
3380 | struct list_head splice; | |
3381 | ||
3382 | INIT_LIST_HEAD(&splice); | |
3383 | ||
acce952b | 3384 | spin_lock(&root->fs_info->delalloc_lock); |
5be76758 | 3385 | list_splice_init(&root->fs_info->delalloc_inodes, &splice); |
acce952b | 3386 | |
3387 | while (!list_empty(&splice)) { | |
3388 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3389 | delalloc_inodes); | |
3390 | ||
3391 | list_del_init(&btrfs_inode->delalloc_inodes); | |
3392 | ||
3393 | btrfs_invalidate_inodes(btrfs_inode->root); | |
3394 | } | |
3395 | ||
3396 | spin_unlock(&root->fs_info->delalloc_lock); | |
3397 | ||
3398 | return 0; | |
3399 | } | |
3400 | ||
3401 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, | |
3402 | struct extent_io_tree *dirty_pages, | |
3403 | int mark) | |
3404 | { | |
3405 | int ret; | |
3406 | struct page *page; | |
3407 | struct inode *btree_inode = root->fs_info->btree_inode; | |
3408 | struct extent_buffer *eb; | |
3409 | u64 start = 0; | |
3410 | u64 end; | |
3411 | u64 offset; | |
3412 | unsigned long index; | |
3413 | ||
3414 | while (1) { | |
3415 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
3416 | mark); | |
3417 | if (ret) | |
3418 | break; | |
3419 | ||
3420 | clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); | |
3421 | while (start <= end) { | |
3422 | index = start >> PAGE_CACHE_SHIFT; | |
3423 | start = (u64)(index + 1) << PAGE_CACHE_SHIFT; | |
3424 | page = find_get_page(btree_inode->i_mapping, index); | |
3425 | if (!page) | |
3426 | continue; | |
3427 | offset = page_offset(page); | |
3428 | ||
3429 | spin_lock(&dirty_pages->buffer_lock); | |
3430 | eb = radix_tree_lookup( | |
3431 | &(&BTRFS_I(page->mapping->host)->io_tree)->buffer, | |
3432 | offset >> PAGE_CACHE_SHIFT); | |
3433 | spin_unlock(&dirty_pages->buffer_lock); | |
3434 | if (eb) { | |
3435 | ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY, | |
3436 | &eb->bflags); | |
3437 | atomic_set(&eb->refs, 1); | |
3438 | } | |
3439 | if (PageWriteback(page)) | |
3440 | end_page_writeback(page); | |
3441 | ||
3442 | lock_page(page); | |
3443 | if (PageDirty(page)) { | |
3444 | clear_page_dirty_for_io(page); | |
3445 | spin_lock_irq(&page->mapping->tree_lock); | |
3446 | radix_tree_tag_clear(&page->mapping->page_tree, | |
3447 | page_index(page), | |
3448 | PAGECACHE_TAG_DIRTY); | |
3449 | spin_unlock_irq(&page->mapping->tree_lock); | |
3450 | } | |
3451 | ||
3452 | page->mapping->a_ops->invalidatepage(page, 0); | |
3453 | unlock_page(page); | |
3454 | } | |
3455 | } | |
3456 | ||
3457 | return ret; | |
3458 | } | |
3459 | ||
3460 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
3461 | struct extent_io_tree *pinned_extents) | |
3462 | { | |
3463 | struct extent_io_tree *unpin; | |
3464 | u64 start; | |
3465 | u64 end; | |
3466 | int ret; | |
3467 | ||
3468 | unpin = pinned_extents; | |
3469 | while (1) { | |
3470 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
3471 | EXTENT_DIRTY); | |
3472 | if (ret) | |
3473 | break; | |
3474 | ||
3475 | /* opt_discard */ | |
5378e607 LD |
3476 | if (btrfs_test_opt(root, DISCARD)) |
3477 | ret = btrfs_error_discard_extent(root, start, | |
3478 | end + 1 - start, | |
3479 | NULL); | |
acce952b | 3480 | |
3481 | clear_extent_dirty(unpin, start, end, GFP_NOFS); | |
3482 | btrfs_error_unpin_extent_range(root, start, end); | |
3483 | cond_resched(); | |
3484 | } | |
3485 | ||
3486 | return 0; | |
3487 | } | |
3488 | ||
3489 | static int btrfs_cleanup_transaction(struct btrfs_root *root) | |
3490 | { | |
3491 | struct btrfs_transaction *t; | |
3492 | LIST_HEAD(list); | |
3493 | ||
3494 | WARN_ON(1); | |
3495 | ||
acce952b | 3496 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
3497 | ||
a4abeea4 | 3498 | spin_lock(&root->fs_info->trans_lock); |
acce952b | 3499 | list_splice_init(&root->fs_info->trans_list, &list); |
a4abeea4 JB |
3500 | root->fs_info->trans_no_join = 1; |
3501 | spin_unlock(&root->fs_info->trans_lock); | |
3502 | ||
acce952b | 3503 | while (!list_empty(&list)) { |
3504 | t = list_entry(list.next, struct btrfs_transaction, list); | |
3505 | if (!t) | |
3506 | break; | |
3507 | ||
3508 | btrfs_destroy_ordered_operations(root); | |
3509 | ||
3510 | btrfs_destroy_ordered_extents(root); | |
3511 | ||
3512 | btrfs_destroy_delayed_refs(t, root); | |
3513 | ||
3514 | btrfs_block_rsv_release(root, | |
3515 | &root->fs_info->trans_block_rsv, | |
3516 | t->dirty_pages.dirty_bytes); | |
3517 | ||
3518 | /* FIXME: cleanup wait for commit */ | |
3519 | t->in_commit = 1; | |
3520 | t->blocked = 1; | |
3521 | if (waitqueue_active(&root->fs_info->transaction_blocked_wait)) | |
3522 | wake_up(&root->fs_info->transaction_blocked_wait); | |
3523 | ||
3524 | t->blocked = 0; | |
3525 | if (waitqueue_active(&root->fs_info->transaction_wait)) | |
3526 | wake_up(&root->fs_info->transaction_wait); | |
acce952b | 3527 | |
acce952b | 3528 | t->commit_done = 1; |
3529 | if (waitqueue_active(&t->commit_wait)) | |
3530 | wake_up(&t->commit_wait); | |
acce952b | 3531 | |
3532 | btrfs_destroy_pending_snapshots(t); | |
3533 | ||
3534 | btrfs_destroy_delalloc_inodes(root); | |
3535 | ||
a4abeea4 | 3536 | spin_lock(&root->fs_info->trans_lock); |
acce952b | 3537 | root->fs_info->running_transaction = NULL; |
a4abeea4 | 3538 | spin_unlock(&root->fs_info->trans_lock); |
acce952b | 3539 | |
3540 | btrfs_destroy_marked_extents(root, &t->dirty_pages, | |
3541 | EXTENT_DIRTY); | |
3542 | ||
3543 | btrfs_destroy_pinned_extent(root, | |
3544 | root->fs_info->pinned_extents); | |
3545 | ||
13c5a93e | 3546 | atomic_set(&t->use_count, 0); |
acce952b | 3547 | list_del_init(&t->list); |
3548 | memset(t, 0, sizeof(*t)); | |
3549 | kmem_cache_free(btrfs_transaction_cachep, t); | |
3550 | } | |
3551 | ||
a4abeea4 JB |
3552 | spin_lock(&root->fs_info->trans_lock); |
3553 | root->fs_info->trans_no_join = 0; | |
3554 | spin_unlock(&root->fs_info->trans_lock); | |
acce952b | 3555 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); |
acce952b | 3556 | |
3557 | return 0; | |
3558 | } | |
3559 | ||
d1310b2e | 3560 | static struct extent_io_ops btree_extent_io_ops = { |
4bef0848 | 3561 | .write_cache_pages_lock_hook = btree_lock_page_hook, |
ce9adaa5 | 3562 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
4bb31e92 | 3563 | .readpage_io_failed_hook = btree_io_failed_hook, |
0b86a832 | 3564 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
3565 | /* note we're sharing with inode.c for the merge bio hook */ |
3566 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 3567 | }; |