Commit | Line | Data |
---|---|---|
6cbd5570 CM |
1 | /* |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
e20d96d6 | 19 | #include <linux/fs.h> |
d98237b3 | 20 | #include <linux/blkdev.h> |
87cbda5c | 21 | #include <linux/scatterlist.h> |
22b0ebda | 22 | #include <linux/swap.h> |
0f7d52f4 | 23 | #include <linux/radix-tree.h> |
35b7e476 | 24 | #include <linux/writeback.h> |
d397712b | 25 | #include <linux/buffer_head.h> |
ce9adaa5 | 26 | #include <linux/workqueue.h> |
a74a4b97 | 27 | #include <linux/kthread.h> |
4b4e25f2 | 28 | #include <linux/freezer.h> |
163e783e | 29 | #include <linux/crc32c.h> |
5a0e3ad6 | 30 | #include <linux/slab.h> |
784b4e29 | 31 | #include <linux/migrate.h> |
7a36ddec | 32 | #include <linux/ratelimit.h> |
7e75bf3f | 33 | #include <asm/unaligned.h> |
4b4e25f2 | 34 | #include "compat.h" |
eb60ceac CM |
35 | #include "ctree.h" |
36 | #include "disk-io.h" | |
e089f05c | 37 | #include "transaction.h" |
0f7d52f4 | 38 | #include "btrfs_inode.h" |
0b86a832 | 39 | #include "volumes.h" |
db94535d | 40 | #include "print-tree.h" |
8b712842 | 41 | #include "async-thread.h" |
925baedd | 42 | #include "locking.h" |
e02119d5 | 43 | #include "tree-log.h" |
fa9c0d79 | 44 | #include "free-space-cache.h" |
581bb050 | 45 | #include "inode-map.h" |
21adbd5c | 46 | #include "check-integrity.h" |
eb60ceac | 47 | |
d1310b2e | 48 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 49 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 50 | static void free_fs_root(struct btrfs_root *root); |
fcd1f065 | 51 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 52 | int read_only); |
143bede5 JM |
53 | static void btrfs_destroy_ordered_operations(struct btrfs_root *root); |
54 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); | |
acce952b | 55 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
56 | struct btrfs_root *root); | |
143bede5 JM |
57 | static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t); |
58 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); | |
acce952b | 59 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, |
60 | struct extent_io_tree *dirty_pages, | |
61 | int mark); | |
62 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
63 | struct extent_io_tree *pinned_extents); | |
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; |
79787eaa | 101 | int error; |
44b8bd7e CM |
102 | }; |
103 | ||
85d4e461 CM |
104 | /* |
105 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
106 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
107 | * the level the eb occupies in the tree. | |
108 | * | |
109 | * Different roots are used for different purposes and may nest inside each | |
110 | * other and they require separate keysets. As lockdep keys should be | |
111 | * static, assign keysets according to the purpose of the root as indicated | |
112 | * by btrfs_root->objectid. This ensures that all special purpose roots | |
113 | * have separate keysets. | |
4008c04a | 114 | * |
85d4e461 CM |
115 | * Lock-nesting across peer nodes is always done with the immediate parent |
116 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
117 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 118 | * |
85d4e461 CM |
119 | * The key is set by the readpage_end_io_hook after the buffer has passed |
120 | * csum validation but before the pages are unlocked. It is also set by | |
121 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 122 | * |
85d4e461 CM |
123 | * We also add a check to make sure the highest level of the tree is the |
124 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
125 | * needs update as well. | |
4008c04a CM |
126 | */ |
127 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
128 | # if BTRFS_MAX_LEVEL != 8 | |
129 | # error | |
130 | # endif | |
85d4e461 CM |
131 | |
132 | static struct btrfs_lockdep_keyset { | |
133 | u64 id; /* root objectid */ | |
134 | const char *name_stem; /* lock name stem */ | |
135 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
136 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
137 | } btrfs_lockdep_keysets[] = { | |
138 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
139 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
140 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
141 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
142 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
143 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
144 | { .id = BTRFS_ORPHAN_OBJECTID, .name_stem = "orphan" }, | |
145 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, | |
146 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
147 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
148 | { .id = 0, .name_stem = "tree" }, | |
4008c04a | 149 | }; |
85d4e461 CM |
150 | |
151 | void __init btrfs_init_lockdep(void) | |
152 | { | |
153 | int i, j; | |
154 | ||
155 | /* initialize lockdep class names */ | |
156 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
157 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
158 | ||
159 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
160 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
161 | "btrfs-%s-%02d", ks->name_stem, j); | |
162 | } | |
163 | } | |
164 | ||
165 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
166 | int level) | |
167 | { | |
168 | struct btrfs_lockdep_keyset *ks; | |
169 | ||
170 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
171 | ||
172 | /* find the matching keyset, id 0 is the default entry */ | |
173 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
174 | if (ks->id == objectid) | |
175 | break; | |
176 | ||
177 | lockdep_set_class_and_name(&eb->lock, | |
178 | &ks->keys[level], ks->names[level]); | |
179 | } | |
180 | ||
4008c04a CM |
181 | #endif |
182 | ||
d352ac68 CM |
183 | /* |
184 | * extents on the btree inode are pretty simple, there's one extent | |
185 | * that covers the entire device | |
186 | */ | |
b2950863 | 187 | static struct extent_map *btree_get_extent(struct inode *inode, |
306e16ce | 188 | struct page *page, size_t pg_offset, u64 start, u64 len, |
b2950863 | 189 | int create) |
7eccb903 | 190 | { |
5f39d397 CM |
191 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
192 | struct extent_map *em; | |
193 | int ret; | |
194 | ||
890871be | 195 | read_lock(&em_tree->lock); |
d1310b2e | 196 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d CM |
197 | if (em) { |
198 | em->bdev = | |
199 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
890871be | 200 | read_unlock(&em_tree->lock); |
5f39d397 | 201 | goto out; |
a061fc8d | 202 | } |
890871be | 203 | read_unlock(&em_tree->lock); |
7b13b7b1 | 204 | |
172ddd60 | 205 | em = alloc_extent_map(); |
5f39d397 CM |
206 | if (!em) { |
207 | em = ERR_PTR(-ENOMEM); | |
208 | goto out; | |
209 | } | |
210 | em->start = 0; | |
0afbaf8c | 211 | em->len = (u64)-1; |
c8b97818 | 212 | em->block_len = (u64)-1; |
5f39d397 | 213 | em->block_start = 0; |
a061fc8d | 214 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e | 215 | |
890871be | 216 | write_lock(&em_tree->lock); |
5f39d397 CM |
217 | ret = add_extent_mapping(em_tree, em); |
218 | if (ret == -EEXIST) { | |
0afbaf8c CM |
219 | u64 failed_start = em->start; |
220 | u64 failed_len = em->len; | |
221 | ||
5f39d397 | 222 | free_extent_map(em); |
7b13b7b1 | 223 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c | 224 | if (em) { |
7b13b7b1 | 225 | ret = 0; |
0afbaf8c CM |
226 | } else { |
227 | em = lookup_extent_mapping(em_tree, failed_start, | |
228 | failed_len); | |
7b13b7b1 | 229 | ret = -EIO; |
0afbaf8c | 230 | } |
5f39d397 | 231 | } else if (ret) { |
7b13b7b1 CM |
232 | free_extent_map(em); |
233 | em = NULL; | |
5f39d397 | 234 | } |
890871be | 235 | write_unlock(&em_tree->lock); |
7b13b7b1 CM |
236 | |
237 | if (ret) | |
238 | em = ERR_PTR(ret); | |
5f39d397 CM |
239 | out: |
240 | return em; | |
7eccb903 CM |
241 | } |
242 | ||
19c00ddc CM |
243 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
244 | { | |
163e783e | 245 | return crc32c(seed, data, len); |
19c00ddc CM |
246 | } |
247 | ||
248 | void btrfs_csum_final(u32 crc, char *result) | |
249 | { | |
7e75bf3f | 250 | put_unaligned_le32(~crc, result); |
19c00ddc CM |
251 | } |
252 | ||
d352ac68 CM |
253 | /* |
254 | * compute the csum for a btree block, and either verify it or write it | |
255 | * into the csum field of the block. | |
256 | */ | |
19c00ddc CM |
257 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
258 | int verify) | |
259 | { | |
6c41761f | 260 | u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy); |
607d432d | 261 | char *result = NULL; |
19c00ddc CM |
262 | unsigned long len; |
263 | unsigned long cur_len; | |
264 | unsigned long offset = BTRFS_CSUM_SIZE; | |
19c00ddc CM |
265 | char *kaddr; |
266 | unsigned long map_start; | |
267 | unsigned long map_len; | |
268 | int err; | |
269 | u32 crc = ~(u32)0; | |
607d432d | 270 | unsigned long inline_result; |
19c00ddc CM |
271 | |
272 | len = buf->len - offset; | |
d397712b | 273 | while (len > 0) { |
19c00ddc | 274 | err = map_private_extent_buffer(buf, offset, 32, |
a6591715 | 275 | &kaddr, &map_start, &map_len); |
d397712b | 276 | if (err) |
19c00ddc | 277 | return 1; |
19c00ddc CM |
278 | cur_len = min(len, map_len - (offset - map_start)); |
279 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
280 | crc, cur_len); | |
281 | len -= cur_len; | |
282 | offset += cur_len; | |
19c00ddc | 283 | } |
607d432d JB |
284 | if (csum_size > sizeof(inline_result)) { |
285 | result = kzalloc(csum_size * sizeof(char), GFP_NOFS); | |
286 | if (!result) | |
287 | return 1; | |
288 | } else { | |
289 | result = (char *)&inline_result; | |
290 | } | |
291 | ||
19c00ddc CM |
292 | btrfs_csum_final(crc, result); |
293 | ||
294 | if (verify) { | |
607d432d | 295 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
296 | u32 val; |
297 | u32 found = 0; | |
607d432d | 298 | memcpy(&found, result, csum_size); |
e4204ded | 299 | |
607d432d | 300 | read_extent_buffer(buf, &val, 0, csum_size); |
7a36ddec | 301 | printk_ratelimited(KERN_INFO "btrfs: %s checksum verify " |
193f284d CM |
302 | "failed on %llu wanted %X found %X " |
303 | "level %d\n", | |
304 | root->fs_info->sb->s_id, | |
305 | (unsigned long long)buf->start, val, found, | |
306 | btrfs_header_level(buf)); | |
607d432d JB |
307 | if (result != (char *)&inline_result) |
308 | kfree(result); | |
19c00ddc CM |
309 | return 1; |
310 | } | |
311 | } else { | |
607d432d | 312 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 313 | } |
607d432d JB |
314 | if (result != (char *)&inline_result) |
315 | kfree(result); | |
19c00ddc CM |
316 | return 0; |
317 | } | |
318 | ||
d352ac68 CM |
319 | /* |
320 | * we can't consider a given block up to date unless the transid of the | |
321 | * block matches the transid in the parent node's pointer. This is how we | |
322 | * detect blocks that either didn't get written at all or got written | |
323 | * in the wrong place. | |
324 | */ | |
1259ab75 CM |
325 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
326 | struct extent_buffer *eb, u64 parent_transid) | |
327 | { | |
2ac55d41 | 328 | struct extent_state *cached_state = NULL; |
1259ab75 CM |
329 | int ret; |
330 | ||
331 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
332 | return 0; | |
333 | ||
2ac55d41 | 334 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
d0082371 | 335 | 0, &cached_state); |
0b32f4bb | 336 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
337 | btrfs_header_generation(eb) == parent_transid) { |
338 | ret = 0; | |
339 | goto out; | |
340 | } | |
7a36ddec | 341 | printk_ratelimited("parent transid verify failed on %llu wanted %llu " |
193f284d CM |
342 | "found %llu\n", |
343 | (unsigned long long)eb->start, | |
344 | (unsigned long long)parent_transid, | |
345 | (unsigned long long)btrfs_header_generation(eb)); | |
1259ab75 | 346 | ret = 1; |
0b32f4bb | 347 | clear_extent_buffer_uptodate(eb); |
33958dc6 | 348 | out: |
2ac55d41 JB |
349 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
350 | &cached_state, GFP_NOFS); | |
1259ab75 | 351 | return ret; |
1259ab75 CM |
352 | } |
353 | ||
d352ac68 CM |
354 | /* |
355 | * helper to read a given tree block, doing retries as required when | |
356 | * the checksums don't match and we have alternate mirrors to try. | |
357 | */ | |
f188591e CM |
358 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
359 | struct extent_buffer *eb, | |
ca7a79ad | 360 | u64 start, u64 parent_transid) |
f188591e CM |
361 | { |
362 | struct extent_io_tree *io_tree; | |
ea466794 | 363 | int failed = 0; |
f188591e CM |
364 | int ret; |
365 | int num_copies = 0; | |
366 | int mirror_num = 0; | |
ea466794 | 367 | int failed_mirror = 0; |
f188591e | 368 | |
a826d6dc | 369 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
f188591e CM |
370 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
371 | while (1) { | |
bb82ab88 AJ |
372 | ret = read_extent_buffer_pages(io_tree, eb, start, |
373 | WAIT_COMPLETE, | |
f188591e | 374 | btree_get_extent, mirror_num); |
727011e0 | 375 | if (!ret && !verify_parent_transid(io_tree, eb, parent_transid)) |
ea466794 | 376 | break; |
d397712b | 377 | |
a826d6dc JB |
378 | /* |
379 | * This buffer's crc is fine, but its contents are corrupted, so | |
380 | * there is no reason to read the other copies, they won't be | |
381 | * any less wrong. | |
382 | */ | |
383 | if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) | |
ea466794 JB |
384 | break; |
385 | ||
386 | if (!failed_mirror) { | |
387 | failed = 1; | |
388 | printk(KERN_ERR "failed mirror was %d\n", eb->failed_mirror); | |
389 | failed_mirror = eb->failed_mirror; | |
390 | } | |
a826d6dc | 391 | |
f188591e CM |
392 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
393 | eb->start, eb->len); | |
4235298e | 394 | if (num_copies == 1) |
ea466794 | 395 | break; |
4235298e | 396 | |
f188591e | 397 | mirror_num++; |
ea466794 JB |
398 | if (mirror_num == failed_mirror) |
399 | mirror_num++; | |
400 | ||
4235298e | 401 | if (mirror_num > num_copies) |
ea466794 | 402 | break; |
f188591e | 403 | } |
ea466794 JB |
404 | |
405 | if (failed && !ret) | |
406 | repair_eb_io_failure(root, eb, failed_mirror); | |
407 | ||
408 | return ret; | |
f188591e | 409 | } |
19c00ddc | 410 | |
d352ac68 | 411 | /* |
d397712b CM |
412 | * checksum a dirty tree block before IO. This has extra checks to make sure |
413 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 414 | */ |
d397712b | 415 | |
b2950863 | 416 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 417 | { |
d1310b2e | 418 | struct extent_io_tree *tree; |
35ebb934 | 419 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc | 420 | u64 found_start; |
19c00ddc | 421 | struct extent_buffer *eb; |
f188591e | 422 | |
d1310b2e | 423 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc | 424 | |
4f2de97a JB |
425 | eb = (struct extent_buffer *)page->private; |
426 | if (page != eb->pages[0]) | |
427 | return 0; | |
19c00ddc CM |
428 | found_start = btrfs_header_bytenr(eb); |
429 | if (found_start != start) { | |
55c69072 | 430 | WARN_ON(1); |
4f2de97a | 431 | return 0; |
55c69072 | 432 | } |
727011e0 | 433 | if (eb->pages[0] != page) { |
55c69072 | 434 | WARN_ON(1); |
4f2de97a | 435 | return 0; |
55c69072 CM |
436 | } |
437 | if (!PageUptodate(page)) { | |
55c69072 | 438 | WARN_ON(1); |
4f2de97a | 439 | return 0; |
19c00ddc | 440 | } |
19c00ddc | 441 | csum_tree_block(root, eb, 0); |
19c00ddc CM |
442 | return 0; |
443 | } | |
444 | ||
2b82032c YZ |
445 | static int check_tree_block_fsid(struct btrfs_root *root, |
446 | struct extent_buffer *eb) | |
447 | { | |
448 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
449 | u8 fsid[BTRFS_UUID_SIZE]; | |
450 | int ret = 1; | |
451 | ||
452 | read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), | |
453 | BTRFS_FSID_SIZE); | |
454 | while (fs_devices) { | |
455 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
456 | ret = 0; | |
457 | break; | |
458 | } | |
459 | fs_devices = fs_devices->seed; | |
460 | } | |
461 | return ret; | |
462 | } | |
463 | ||
a826d6dc JB |
464 | #define CORRUPT(reason, eb, root, slot) \ |
465 | printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \ | |
466 | "root=%llu, slot=%d\n", reason, \ | |
467 | (unsigned long long)btrfs_header_bytenr(eb), \ | |
468 | (unsigned long long)root->objectid, slot) | |
469 | ||
470 | static noinline int check_leaf(struct btrfs_root *root, | |
471 | struct extent_buffer *leaf) | |
472 | { | |
473 | struct btrfs_key key; | |
474 | struct btrfs_key leaf_key; | |
475 | u32 nritems = btrfs_header_nritems(leaf); | |
476 | int slot; | |
477 | ||
478 | if (nritems == 0) | |
479 | return 0; | |
480 | ||
481 | /* Check the 0 item */ | |
482 | if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != | |
483 | BTRFS_LEAF_DATA_SIZE(root)) { | |
484 | CORRUPT("invalid item offset size pair", leaf, root, 0); | |
485 | return -EIO; | |
486 | } | |
487 | ||
488 | /* | |
489 | * Check to make sure each items keys are in the correct order and their | |
490 | * offsets make sense. We only have to loop through nritems-1 because | |
491 | * we check the current slot against the next slot, which verifies the | |
492 | * next slot's offset+size makes sense and that the current's slot | |
493 | * offset is correct. | |
494 | */ | |
495 | for (slot = 0; slot < nritems - 1; slot++) { | |
496 | btrfs_item_key_to_cpu(leaf, &leaf_key, slot); | |
497 | btrfs_item_key_to_cpu(leaf, &key, slot + 1); | |
498 | ||
499 | /* Make sure the keys are in the right order */ | |
500 | if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { | |
501 | CORRUPT("bad key order", leaf, root, slot); | |
502 | return -EIO; | |
503 | } | |
504 | ||
505 | /* | |
506 | * Make sure the offset and ends are right, remember that the | |
507 | * item data starts at the end of the leaf and grows towards the | |
508 | * front. | |
509 | */ | |
510 | if (btrfs_item_offset_nr(leaf, slot) != | |
511 | btrfs_item_end_nr(leaf, slot + 1)) { | |
512 | CORRUPT("slot offset bad", leaf, root, slot); | |
513 | return -EIO; | |
514 | } | |
515 | ||
516 | /* | |
517 | * Check to make sure that we don't point outside of the leaf, | |
518 | * just incase all the items are consistent to eachother, but | |
519 | * all point outside of the leaf. | |
520 | */ | |
521 | if (btrfs_item_end_nr(leaf, slot) > | |
522 | BTRFS_LEAF_DATA_SIZE(root)) { | |
523 | CORRUPT("slot end outside of leaf", leaf, root, slot); | |
524 | return -EIO; | |
525 | } | |
526 | } | |
527 | ||
528 | return 0; | |
529 | } | |
530 | ||
727011e0 CM |
531 | struct extent_buffer *find_eb_for_page(struct extent_io_tree *tree, |
532 | struct page *page, int max_walk) | |
533 | { | |
534 | struct extent_buffer *eb; | |
535 | u64 start = page_offset(page); | |
536 | u64 target = start; | |
537 | u64 min_start; | |
538 | ||
539 | if (start < max_walk) | |
540 | min_start = 0; | |
541 | else | |
542 | min_start = start - max_walk; | |
543 | ||
544 | while (start >= min_start) { | |
545 | eb = find_extent_buffer(tree, start, 0); | |
546 | if (eb) { | |
547 | /* | |
548 | * we found an extent buffer and it contains our page | |
549 | * horray! | |
550 | */ | |
551 | if (eb->start <= target && | |
552 | eb->start + eb->len > target) | |
553 | return eb; | |
554 | ||
555 | /* we found an extent buffer that wasn't for us */ | |
556 | free_extent_buffer(eb); | |
557 | return NULL; | |
558 | } | |
559 | if (start == 0) | |
560 | break; | |
561 | start -= PAGE_CACHE_SIZE; | |
562 | } | |
563 | return NULL; | |
564 | } | |
565 | ||
b2950863 | 566 | static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
ce9adaa5 CM |
567 | struct extent_state *state) |
568 | { | |
569 | struct extent_io_tree *tree; | |
570 | u64 found_start; | |
571 | int found_level; | |
ce9adaa5 CM |
572 | struct extent_buffer *eb; |
573 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 574 | int ret = 0; |
727011e0 | 575 | int reads_done; |
ce9adaa5 | 576 | |
ce9adaa5 CM |
577 | if (!page->private) |
578 | goto out; | |
d397712b | 579 | |
727011e0 | 580 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
4f2de97a | 581 | eb = (struct extent_buffer *)page->private; |
d397712b | 582 | |
0b32f4bb JB |
583 | /* the pending IO might have been the only thing that kept this buffer |
584 | * in memory. Make sure we have a ref for all this other checks | |
585 | */ | |
586 | extent_buffer_get(eb); | |
587 | ||
588 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
727011e0 CM |
589 | if (!reads_done) |
590 | goto err; | |
f188591e | 591 | |
ea466794 JB |
592 | if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) { |
593 | ret = -EIO; | |
594 | goto err; | |
595 | } | |
596 | ||
ce9adaa5 | 597 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 598 | if (found_start != eb->start) { |
7a36ddec | 599 | printk_ratelimited(KERN_INFO "btrfs bad tree block start " |
193f284d CM |
600 | "%llu %llu\n", |
601 | (unsigned long long)found_start, | |
602 | (unsigned long long)eb->start); | |
f188591e | 603 | ret = -EIO; |
ce9adaa5 CM |
604 | goto err; |
605 | } | |
2b82032c | 606 | if (check_tree_block_fsid(root, eb)) { |
7a36ddec | 607 | printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n", |
193f284d | 608 | (unsigned long long)eb->start); |
1259ab75 CM |
609 | ret = -EIO; |
610 | goto err; | |
611 | } | |
ce9adaa5 CM |
612 | found_level = btrfs_header_level(eb); |
613 | ||
85d4e461 CM |
614 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
615 | eb, found_level); | |
4008c04a | 616 | |
ce9adaa5 | 617 | ret = csum_tree_block(root, eb, 1); |
a826d6dc | 618 | if (ret) { |
f188591e | 619 | ret = -EIO; |
a826d6dc JB |
620 | goto err; |
621 | } | |
622 | ||
623 | /* | |
624 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
625 | * that we don't try and read the other copies of this block, just | |
626 | * return -EIO. | |
627 | */ | |
628 | if (found_level == 0 && check_leaf(root, eb)) { | |
629 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); | |
630 | ret = -EIO; | |
631 | } | |
ce9adaa5 | 632 | |
0b32f4bb JB |
633 | if (!ret) |
634 | set_extent_buffer_uptodate(eb); | |
ce9adaa5 | 635 | err: |
4bb31e92 AJ |
636 | if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) { |
637 | clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags); | |
638 | btree_readahead_hook(root, eb, eb->start, ret); | |
639 | } | |
640 | ||
0b32f4bb JB |
641 | if (ret) |
642 | clear_extent_buffer_uptodate(eb); | |
643 | free_extent_buffer(eb); | |
ce9adaa5 | 644 | out: |
f188591e | 645 | return ret; |
ce9adaa5 CM |
646 | } |
647 | ||
ea466794 | 648 | static int btree_io_failed_hook(struct page *page, int failed_mirror) |
4bb31e92 | 649 | { |
4bb31e92 AJ |
650 | struct extent_buffer *eb; |
651 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
652 | ||
4f2de97a | 653 | eb = (struct extent_buffer *)page->private; |
ea466794 JB |
654 | set_bit(EXTENT_BUFFER_IOERR, &eb->bflags); |
655 | eb->failed_mirror = failed_mirror; | |
656 | if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
4bb31e92 | 657 | btree_readahead_hook(root, eb, eb->start, -EIO); |
4bb31e92 AJ |
658 | return -EIO; /* we fixed nothing */ |
659 | } | |
660 | ||
ce9adaa5 | 661 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
662 | { |
663 | struct end_io_wq *end_io_wq = bio->bi_private; | |
664 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 665 | |
ce9adaa5 | 666 | fs_info = end_io_wq->info; |
ce9adaa5 | 667 | end_io_wq->error = err; |
8b712842 CM |
668 | end_io_wq->work.func = end_workqueue_fn; |
669 | end_io_wq->work.flags = 0; | |
d20f7043 | 670 | |
7b6d91da | 671 | if (bio->bi_rw & REQ_WRITE) { |
0cb59c99 | 672 | if (end_io_wq->metadata == 1) |
cad321ad CM |
673 | btrfs_queue_worker(&fs_info->endio_meta_write_workers, |
674 | &end_io_wq->work); | |
0cb59c99 JB |
675 | else if (end_io_wq->metadata == 2) |
676 | btrfs_queue_worker(&fs_info->endio_freespace_worker, | |
677 | &end_io_wq->work); | |
cad321ad CM |
678 | else |
679 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
680 | &end_io_wq->work); | |
d20f7043 CM |
681 | } else { |
682 | if (end_io_wq->metadata) | |
683 | btrfs_queue_worker(&fs_info->endio_meta_workers, | |
684 | &end_io_wq->work); | |
685 | else | |
686 | btrfs_queue_worker(&fs_info->endio_workers, | |
687 | &end_io_wq->work); | |
688 | } | |
ce9adaa5 CM |
689 | } |
690 | ||
0cb59c99 JB |
691 | /* |
692 | * For the metadata arg you want | |
693 | * | |
694 | * 0 - if data | |
695 | * 1 - if normal metadta | |
696 | * 2 - if writing to the free space cache area | |
697 | */ | |
22c59948 CM |
698 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
699 | int metadata) | |
0b86a832 | 700 | { |
ce9adaa5 | 701 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
702 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
703 | if (!end_io_wq) | |
704 | return -ENOMEM; | |
705 | ||
706 | end_io_wq->private = bio->bi_private; | |
707 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 708 | end_io_wq->info = info; |
ce9adaa5 CM |
709 | end_io_wq->error = 0; |
710 | end_io_wq->bio = bio; | |
22c59948 | 711 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
712 | |
713 | bio->bi_private = end_io_wq; | |
714 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
715 | return 0; |
716 | } | |
717 | ||
b64a2851 | 718 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 719 | { |
4854ddd0 CM |
720 | unsigned long limit = min_t(unsigned long, |
721 | info->workers.max_workers, | |
722 | info->fs_devices->open_devices); | |
723 | return 256 * limit; | |
724 | } | |
0986fe9e | 725 | |
4a69a410 CM |
726 | static void run_one_async_start(struct btrfs_work *work) |
727 | { | |
4a69a410 | 728 | struct async_submit_bio *async; |
79787eaa | 729 | int ret; |
4a69a410 CM |
730 | |
731 | async = container_of(work, struct async_submit_bio, work); | |
79787eaa JM |
732 | ret = async->submit_bio_start(async->inode, async->rw, async->bio, |
733 | async->mirror_num, async->bio_flags, | |
734 | async->bio_offset); | |
735 | if (ret) | |
736 | async->error = ret; | |
4a69a410 CM |
737 | } |
738 | ||
739 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
740 | { |
741 | struct btrfs_fs_info *fs_info; | |
742 | struct async_submit_bio *async; | |
4854ddd0 | 743 | int limit; |
8b712842 CM |
744 | |
745 | async = container_of(work, struct async_submit_bio, work); | |
746 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 747 | |
b64a2851 | 748 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
749 | limit = limit * 2 / 3; |
750 | ||
8b712842 | 751 | atomic_dec(&fs_info->nr_async_submits); |
0986fe9e | 752 | |
b64a2851 CM |
753 | if (atomic_read(&fs_info->nr_async_submits) < limit && |
754 | waitqueue_active(&fs_info->async_submit_wait)) | |
4854ddd0 CM |
755 | wake_up(&fs_info->async_submit_wait); |
756 | ||
79787eaa JM |
757 | /* If an error occured we just want to clean up the bio and move on */ |
758 | if (async->error) { | |
759 | bio_endio(async->bio, async->error); | |
760 | return; | |
761 | } | |
762 | ||
4a69a410 | 763 | async->submit_bio_done(async->inode, async->rw, async->bio, |
eaf25d93 CM |
764 | async->mirror_num, async->bio_flags, |
765 | async->bio_offset); | |
4a69a410 CM |
766 | } |
767 | ||
768 | static void run_one_async_free(struct btrfs_work *work) | |
769 | { | |
770 | struct async_submit_bio *async; | |
771 | ||
772 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
773 | kfree(async); |
774 | } | |
775 | ||
44b8bd7e CM |
776 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
777 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 778 | unsigned long bio_flags, |
eaf25d93 | 779 | u64 bio_offset, |
4a69a410 CM |
780 | extent_submit_bio_hook_t *submit_bio_start, |
781 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
782 | { |
783 | struct async_submit_bio *async; | |
784 | ||
785 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
786 | if (!async) | |
787 | return -ENOMEM; | |
788 | ||
789 | async->inode = inode; | |
790 | async->rw = rw; | |
791 | async->bio = bio; | |
792 | async->mirror_num = mirror_num; | |
4a69a410 CM |
793 | async->submit_bio_start = submit_bio_start; |
794 | async->submit_bio_done = submit_bio_done; | |
795 | ||
796 | async->work.func = run_one_async_start; | |
797 | async->work.ordered_func = run_one_async_done; | |
798 | async->work.ordered_free = run_one_async_free; | |
799 | ||
8b712842 | 800 | async->work.flags = 0; |
c8b97818 | 801 | async->bio_flags = bio_flags; |
eaf25d93 | 802 | async->bio_offset = bio_offset; |
8c8bee1d | 803 | |
79787eaa JM |
804 | async->error = 0; |
805 | ||
cb03c743 | 806 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 | 807 | |
7b6d91da | 808 | if (rw & REQ_SYNC) |
d313d7a3 CM |
809 | btrfs_set_work_high_prio(&async->work); |
810 | ||
8b712842 | 811 | btrfs_queue_worker(&fs_info->workers, &async->work); |
9473f16c | 812 | |
d397712b | 813 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
814 | atomic_read(&fs_info->nr_async_submits)) { |
815 | wait_event(fs_info->async_submit_wait, | |
816 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
817 | } | |
818 | ||
44b8bd7e CM |
819 | return 0; |
820 | } | |
821 | ||
ce3ed71a CM |
822 | static int btree_csum_one_bio(struct bio *bio) |
823 | { | |
824 | struct bio_vec *bvec = bio->bi_io_vec; | |
825 | int bio_index = 0; | |
826 | struct btrfs_root *root; | |
79787eaa | 827 | int ret = 0; |
ce3ed71a CM |
828 | |
829 | WARN_ON(bio->bi_vcnt <= 0); | |
d397712b | 830 | while (bio_index < bio->bi_vcnt) { |
ce3ed71a | 831 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
79787eaa JM |
832 | ret = csum_dirty_buffer(root, bvec->bv_page); |
833 | if (ret) | |
834 | break; | |
ce3ed71a CM |
835 | bio_index++; |
836 | bvec++; | |
837 | } | |
79787eaa | 838 | return ret; |
ce3ed71a CM |
839 | } |
840 | ||
4a69a410 CM |
841 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
842 | struct bio *bio, int mirror_num, | |
eaf25d93 CM |
843 | unsigned long bio_flags, |
844 | u64 bio_offset) | |
22c59948 | 845 | { |
8b712842 CM |
846 | /* |
847 | * when we're called for a write, we're already in the async | |
5443be45 | 848 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 849 | */ |
79787eaa | 850 | return btree_csum_one_bio(bio); |
4a69a410 | 851 | } |
22c59948 | 852 | |
4a69a410 | 853 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
854 | int mirror_num, unsigned long bio_flags, |
855 | u64 bio_offset) | |
4a69a410 | 856 | { |
8b712842 | 857 | /* |
4a69a410 CM |
858 | * when we're called for a write, we're already in the async |
859 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 860 | */ |
8b712842 | 861 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
862 | } |
863 | ||
44b8bd7e | 864 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
865 | int mirror_num, unsigned long bio_flags, |
866 | u64 bio_offset) | |
44b8bd7e | 867 | { |
cad321ad CM |
868 | int ret; |
869 | ||
7b6d91da | 870 | if (!(rw & REQ_WRITE)) { |
f3f266ab | 871 | |
4a69a410 CM |
872 | /* |
873 | * called for a read, do the setup so that checksum validation | |
874 | * can happen in the async kernel threads | |
875 | */ | |
f3f266ab CM |
876 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, |
877 | bio, 1); | |
1d4284bd CM |
878 | if (ret) |
879 | return ret; | |
4a69a410 | 880 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, |
6f3577bd | 881 | mirror_num, 0); |
44b8bd7e | 882 | } |
d313d7a3 | 883 | |
cad321ad CM |
884 | /* |
885 | * kthread helpers are used to submit writes so that checksumming | |
886 | * can happen in parallel across all CPUs | |
887 | */ | |
44b8bd7e | 888 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, |
c8b97818 | 889 | inode, rw, bio, mirror_num, 0, |
eaf25d93 | 890 | bio_offset, |
4a69a410 CM |
891 | __btree_submit_bio_start, |
892 | __btree_submit_bio_done); | |
44b8bd7e CM |
893 | } |
894 | ||
3dd1462e | 895 | #ifdef CONFIG_MIGRATION |
784b4e29 | 896 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
897 | struct page *newpage, struct page *page, |
898 | enum migrate_mode mode) | |
784b4e29 CM |
899 | { |
900 | /* | |
901 | * we can't safely write a btree page from here, | |
902 | * we haven't done the locking hook | |
903 | */ | |
904 | if (PageDirty(page)) | |
905 | return -EAGAIN; | |
906 | /* | |
907 | * Buffers may be managed in a filesystem specific way. | |
908 | * We must have no buffers or drop them. | |
909 | */ | |
910 | if (page_has_private(page) && | |
911 | !try_to_release_page(page, GFP_KERNEL)) | |
912 | return -EAGAIN; | |
a6bc32b8 | 913 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 914 | } |
3dd1462e | 915 | #endif |
784b4e29 | 916 | |
0da5468f CM |
917 | |
918 | static int btree_writepages(struct address_space *mapping, | |
919 | struct writeback_control *wbc) | |
920 | { | |
d1310b2e CM |
921 | struct extent_io_tree *tree; |
922 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 923 | if (wbc->sync_mode == WB_SYNC_NONE) { |
b9473439 | 924 | struct btrfs_root *root = BTRFS_I(mapping->host)->root; |
793955bc | 925 | u64 num_dirty; |
24ab9cd8 | 926 | unsigned long thresh = 32 * 1024 * 1024; |
448d640b CM |
927 | |
928 | if (wbc->for_kupdate) | |
929 | return 0; | |
930 | ||
b9473439 CM |
931 | /* this is a bit racy, but that's ok */ |
932 | num_dirty = root->fs_info->dirty_metadata_bytes; | |
d397712b | 933 | if (num_dirty < thresh) |
793955bc | 934 | return 0; |
793955bc | 935 | } |
0b32f4bb | 936 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
937 | } |
938 | ||
b2950863 | 939 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 940 | { |
d1310b2e CM |
941 | struct extent_io_tree *tree; |
942 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 943 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 944 | } |
22b0ebda | 945 | |
70dec807 | 946 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 947 | { |
98509cfc | 948 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 949 | return 0; |
0c4e538b DS |
950 | /* |
951 | * We need to mask out eg. __GFP_HIGHMEM and __GFP_DMA32 as we're doing | |
952 | * slab allocation from alloc_extent_state down the callchain where | |
953 | * it'd hit a BUG_ON as those flags are not allowed. | |
954 | */ | |
955 | gfp_flags &= ~GFP_SLAB_BUG_MASK; | |
956 | ||
3083ee2e | 957 | return try_release_extent_buffer(page, gfp_flags); |
d98237b3 CM |
958 | } |
959 | ||
5f39d397 | 960 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 961 | { |
d1310b2e CM |
962 | struct extent_io_tree *tree; |
963 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
964 | extent_invalidatepage(tree, page, offset); |
965 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 966 | if (PagePrivate(page)) { |
d397712b CM |
967 | printk(KERN_WARNING "btrfs warning page private not zero " |
968 | "on page %llu\n", (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
969 | ClearPagePrivate(page); |
970 | set_page_private(page, 0); | |
971 | page_cache_release(page); | |
972 | } | |
d98237b3 CM |
973 | } |
974 | ||
0b32f4bb JB |
975 | static int btree_set_page_dirty(struct page *page) |
976 | { | |
977 | struct extent_buffer *eb; | |
978 | ||
979 | BUG_ON(!PagePrivate(page)); | |
980 | eb = (struct extent_buffer *)page->private; | |
981 | BUG_ON(!eb); | |
982 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
983 | BUG_ON(!atomic_read(&eb->refs)); | |
984 | btrfs_assert_tree_locked(eb); | |
985 | return __set_page_dirty_nobuffers(page); | |
986 | } | |
987 | ||
7f09410b | 988 | static const struct address_space_operations btree_aops = { |
d98237b3 | 989 | .readpage = btree_readpage, |
0da5468f | 990 | .writepages = btree_writepages, |
5f39d397 CM |
991 | .releasepage = btree_releasepage, |
992 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 993 | #ifdef CONFIG_MIGRATION |
784b4e29 | 994 | .migratepage = btree_migratepage, |
5a92bc88 | 995 | #endif |
0b32f4bb | 996 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
997 | }; |
998 | ||
ca7a79ad CM |
999 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1000 | u64 parent_transid) | |
090d1875 | 1001 | { |
5f39d397 CM |
1002 | struct extent_buffer *buf = NULL; |
1003 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 1004 | int ret = 0; |
090d1875 | 1005 | |
db94535d | 1006 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 1007 | if (!buf) |
090d1875 | 1008 | return 0; |
d1310b2e | 1009 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
bb82ab88 | 1010 | buf, 0, WAIT_NONE, btree_get_extent, 0); |
5f39d397 | 1011 | free_extent_buffer(buf); |
de428b63 | 1012 | return ret; |
090d1875 CM |
1013 | } |
1014 | ||
ab0fff03 AJ |
1015 | int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1016 | int mirror_num, struct extent_buffer **eb) | |
1017 | { | |
1018 | struct extent_buffer *buf = NULL; | |
1019 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1020 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; | |
1021 | int ret; | |
1022 | ||
1023 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
1024 | if (!buf) | |
1025 | return 0; | |
1026 | ||
1027 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1028 | ||
1029 | ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, | |
1030 | btree_get_extent, mirror_num); | |
1031 | if (ret) { | |
1032 | free_extent_buffer(buf); | |
1033 | return ret; | |
1034 | } | |
1035 | ||
1036 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1037 | free_extent_buffer(buf); | |
1038 | return -EIO; | |
0b32f4bb | 1039 | } else if (extent_buffer_uptodate(buf)) { |
ab0fff03 AJ |
1040 | *eb = buf; |
1041 | } else { | |
1042 | free_extent_buffer(buf); | |
1043 | } | |
1044 | return 0; | |
1045 | } | |
1046 | ||
0999df54 CM |
1047 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
1048 | u64 bytenr, u32 blocksize) | |
1049 | { | |
1050 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1051 | struct extent_buffer *eb; | |
1052 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
f09d1f60 | 1053 | bytenr, blocksize); |
0999df54 CM |
1054 | return eb; |
1055 | } | |
1056 | ||
1057 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
1058 | u64 bytenr, u32 blocksize) | |
1059 | { | |
1060 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1061 | struct extent_buffer *eb; | |
1062 | ||
1063 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
727011e0 | 1064 | bytenr, blocksize); |
0999df54 CM |
1065 | return eb; |
1066 | } | |
1067 | ||
1068 | ||
e02119d5 CM |
1069 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1070 | { | |
727011e0 | 1071 | return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, |
8aa38c31 | 1072 | buf->start + buf->len - 1); |
e02119d5 CM |
1073 | } |
1074 | ||
1075 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
1076 | { | |
727011e0 | 1077 | return filemap_fdatawait_range(buf->pages[0]->mapping, |
8aa38c31 | 1078 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
1079 | } |
1080 | ||
0999df54 | 1081 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 1082 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
1083 | { |
1084 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1085 | int ret; |
1086 | ||
0999df54 CM |
1087 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
1088 | if (!buf) | |
1089 | return NULL; | |
0999df54 | 1090 | |
ca7a79ad | 1091 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
5f39d397 | 1092 | return buf; |
ce9adaa5 | 1093 | |
eb60ceac CM |
1094 | } |
1095 | ||
d5c13f92 JM |
1096 | void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
1097 | struct extent_buffer *buf) | |
ed2ff2cb | 1098 | { |
55c69072 | 1099 | if (btrfs_header_generation(buf) == |
925baedd | 1100 | root->fs_info->running_transaction->transid) { |
b9447ef8 | 1101 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1102 | |
b9473439 CM |
1103 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
1104 | spin_lock(&root->fs_info->delalloc_lock); | |
1105 | if (root->fs_info->dirty_metadata_bytes >= buf->len) | |
1106 | root->fs_info->dirty_metadata_bytes -= buf->len; | |
d5c13f92 JM |
1107 | else { |
1108 | spin_unlock(&root->fs_info->delalloc_lock); | |
1109 | btrfs_panic(root->fs_info, -EOVERFLOW, | |
1110 | "Can't clear %lu bytes from " | |
1111 | " dirty_mdatadata_bytes (%lu)", | |
1112 | buf->len, | |
1113 | root->fs_info->dirty_metadata_bytes); | |
1114 | } | |
b9473439 CM |
1115 | spin_unlock(&root->fs_info->delalloc_lock); |
1116 | } | |
b4ce94de | 1117 | |
b9473439 CM |
1118 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1119 | btrfs_set_lock_blocking(buf); | |
0b32f4bb | 1120 | clear_extent_buffer_dirty(buf); |
925baedd | 1121 | } |
5f39d397 CM |
1122 | } |
1123 | ||
143bede5 JM |
1124 | static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
1125 | u32 stripesize, struct btrfs_root *root, | |
1126 | struct btrfs_fs_info *fs_info, | |
1127 | u64 objectid) | |
d97e63b6 | 1128 | { |
cfaa7295 | 1129 | root->node = NULL; |
a28ec197 | 1130 | root->commit_root = NULL; |
db94535d CM |
1131 | root->sectorsize = sectorsize; |
1132 | root->nodesize = nodesize; | |
1133 | root->leafsize = leafsize; | |
87ee04eb | 1134 | root->stripesize = stripesize; |
123abc88 | 1135 | root->ref_cows = 0; |
0b86a832 | 1136 | root->track_dirty = 0; |
c71bf099 | 1137 | root->in_radix = 0; |
d68fc57b YZ |
1138 | root->orphan_item_inserted = 0; |
1139 | root->orphan_cleanup_state = 0; | |
0b86a832 | 1140 | |
0f7d52f4 CM |
1141 | root->objectid = objectid; |
1142 | root->last_trans = 0; | |
13a8a7c8 | 1143 | root->highest_objectid = 0; |
58176a96 | 1144 | root->name = NULL; |
6bef4d31 | 1145 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1146 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1147 | root->block_rsv = NULL; |
d68fc57b | 1148 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
1149 | |
1150 | INIT_LIST_HEAD(&root->dirty_list); | |
7b128766 | 1151 | INIT_LIST_HEAD(&root->orphan_list); |
5d4f98a2 | 1152 | INIT_LIST_HEAD(&root->root_list); |
d68fc57b | 1153 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 1154 | spin_lock_init(&root->inode_lock); |
f0486c68 | 1155 | spin_lock_init(&root->accounting_lock); |
a2135011 | 1156 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1157 | mutex_init(&root->log_mutex); |
7237f183 YZ |
1158 | init_waitqueue_head(&root->log_writer_wait); |
1159 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1160 | init_waitqueue_head(&root->log_commit_wait[1]); | |
1161 | atomic_set(&root->log_commit[0], 0); | |
1162 | atomic_set(&root->log_commit[1], 0); | |
1163 | atomic_set(&root->log_writers, 0); | |
1164 | root->log_batch = 0; | |
1165 | root->log_transid = 0; | |
257c62e1 | 1166 | root->last_log_commit = 0; |
d0c803c4 | 1167 | extent_io_tree_init(&root->dirty_log_pages, |
f993c883 | 1168 | fs_info->btree_inode->i_mapping); |
017e5369 | 1169 | |
3768f368 CM |
1170 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1171 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1172 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 1173 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 1174 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 1175 | init_completion(&root->kobj_unregister); |
6702ed49 | 1176 | root->defrag_running = 0; |
4d775673 | 1177 | root->root_key.objectid = objectid; |
0ee5dc67 | 1178 | root->anon_dev = 0; |
3768f368 CM |
1179 | } |
1180 | ||
200a5c17 JM |
1181 | static int __must_check find_and_setup_root(struct btrfs_root *tree_root, |
1182 | struct btrfs_fs_info *fs_info, | |
1183 | u64 objectid, | |
1184 | struct btrfs_root *root) | |
3768f368 CM |
1185 | { |
1186 | int ret; | |
db94535d | 1187 | u32 blocksize; |
84234f3a | 1188 | u64 generation; |
3768f368 | 1189 | |
db94535d | 1190 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1191 | tree_root->sectorsize, tree_root->stripesize, |
1192 | root, fs_info, objectid); | |
3768f368 CM |
1193 | ret = btrfs_find_last_root(tree_root, objectid, |
1194 | &root->root_item, &root->root_key); | |
4df27c4d YZ |
1195 | if (ret > 0) |
1196 | return -ENOENT; | |
200a5c17 JM |
1197 | else if (ret < 0) |
1198 | return ret; | |
3768f368 | 1199 | |
84234f3a | 1200 | generation = btrfs_root_generation(&root->root_item); |
db94535d | 1201 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
af31f5e5 | 1202 | root->commit_root = NULL; |
db94535d | 1203 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), |
84234f3a | 1204 | blocksize, generation); |
68433b73 CM |
1205 | if (!root->node || !btrfs_buffer_uptodate(root->node, generation)) { |
1206 | free_extent_buffer(root->node); | |
af31f5e5 | 1207 | root->node = NULL; |
68433b73 CM |
1208 | return -EIO; |
1209 | } | |
4df27c4d | 1210 | root->commit_root = btrfs_root_node(root); |
d97e63b6 CM |
1211 | return 0; |
1212 | } | |
1213 | ||
f84a8bd6 | 1214 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info) |
6f07e42e AV |
1215 | { |
1216 | struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS); | |
1217 | if (root) | |
1218 | root->fs_info = fs_info; | |
1219 | return root; | |
1220 | } | |
1221 | ||
7237f183 YZ |
1222 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1223 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1224 | { |
1225 | struct btrfs_root *root; | |
1226 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1227 | struct extent_buffer *leaf; |
e02119d5 | 1228 | |
6f07e42e | 1229 | root = btrfs_alloc_root(fs_info); |
e02119d5 | 1230 | if (!root) |
7237f183 | 1231 | return ERR_PTR(-ENOMEM); |
e02119d5 CM |
1232 | |
1233 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1234 | tree_root->sectorsize, tree_root->stripesize, | |
1235 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1236 | ||
1237 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1238 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1239 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
7237f183 YZ |
1240 | /* |
1241 | * log trees do not get reference counted because they go away | |
1242 | * before a real commit is actually done. They do store pointers | |
1243 | * to file data extents, and those reference counts still get | |
1244 | * updated (along with back refs to the log tree). | |
1245 | */ | |
e02119d5 CM |
1246 | root->ref_cows = 0; |
1247 | ||
5d4f98a2 | 1248 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, |
66d7e7f0 AJ |
1249 | BTRFS_TREE_LOG_OBJECTID, NULL, |
1250 | 0, 0, 0, 0); | |
7237f183 YZ |
1251 | if (IS_ERR(leaf)) { |
1252 | kfree(root); | |
1253 | return ERR_CAST(leaf); | |
1254 | } | |
e02119d5 | 1255 | |
5d4f98a2 YZ |
1256 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1257 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1258 | btrfs_set_header_generation(leaf, trans->transid); | |
1259 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1260 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1261 | root->node = leaf; |
e02119d5 CM |
1262 | |
1263 | write_extent_buffer(root->node, root->fs_info->fsid, | |
1264 | (unsigned long)btrfs_header_fsid(root->node), | |
1265 | BTRFS_FSID_SIZE); | |
1266 | btrfs_mark_buffer_dirty(root->node); | |
1267 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1268 | return root; |
1269 | } | |
1270 | ||
1271 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1272 | struct btrfs_fs_info *fs_info) | |
1273 | { | |
1274 | struct btrfs_root *log_root; | |
1275 | ||
1276 | log_root = alloc_log_tree(trans, fs_info); | |
1277 | if (IS_ERR(log_root)) | |
1278 | return PTR_ERR(log_root); | |
1279 | WARN_ON(fs_info->log_root_tree); | |
1280 | fs_info->log_root_tree = log_root; | |
1281 | return 0; | |
1282 | } | |
1283 | ||
1284 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1285 | struct btrfs_root *root) | |
1286 | { | |
1287 | struct btrfs_root *log_root; | |
1288 | struct btrfs_inode_item *inode_item; | |
1289 | ||
1290 | log_root = alloc_log_tree(trans, root->fs_info); | |
1291 | if (IS_ERR(log_root)) | |
1292 | return PTR_ERR(log_root); | |
1293 | ||
1294 | log_root->last_trans = trans->transid; | |
1295 | log_root->root_key.offset = root->root_key.objectid; | |
1296 | ||
1297 | inode_item = &log_root->root_item.inode; | |
1298 | inode_item->generation = cpu_to_le64(1); | |
1299 | inode_item->size = cpu_to_le64(3); | |
1300 | inode_item->nlink = cpu_to_le32(1); | |
1301 | inode_item->nbytes = cpu_to_le64(root->leafsize); | |
1302 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); | |
1303 | ||
5d4f98a2 | 1304 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1305 | |
1306 | WARN_ON(root->log_root); | |
1307 | root->log_root = log_root; | |
1308 | root->log_transid = 0; | |
257c62e1 | 1309 | root->last_log_commit = 0; |
e02119d5 CM |
1310 | return 0; |
1311 | } | |
1312 | ||
1313 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
1314 | struct btrfs_key *location) | |
1315 | { | |
1316 | struct btrfs_root *root; | |
1317 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1318 | struct btrfs_path *path; |
5f39d397 | 1319 | struct extent_buffer *l; |
84234f3a | 1320 | u64 generation; |
db94535d | 1321 | u32 blocksize; |
0f7d52f4 CM |
1322 | int ret = 0; |
1323 | ||
6f07e42e | 1324 | root = btrfs_alloc_root(fs_info); |
0cf6c620 | 1325 | if (!root) |
0f7d52f4 | 1326 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 1327 | if (location->offset == (u64)-1) { |
db94535d | 1328 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
1329 | location->objectid, root); |
1330 | if (ret) { | |
0f7d52f4 CM |
1331 | kfree(root); |
1332 | return ERR_PTR(ret); | |
1333 | } | |
13a8a7c8 | 1334 | goto out; |
0f7d52f4 CM |
1335 | } |
1336 | ||
db94535d | 1337 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1338 | tree_root->sectorsize, tree_root->stripesize, |
1339 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
1340 | |
1341 | path = btrfs_alloc_path(); | |
db5b493a TI |
1342 | if (!path) { |
1343 | kfree(root); | |
1344 | return ERR_PTR(-ENOMEM); | |
1345 | } | |
0f7d52f4 | 1346 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); |
13a8a7c8 YZ |
1347 | if (ret == 0) { |
1348 | l = path->nodes[0]; | |
1349 | read_extent_buffer(l, &root->root_item, | |
1350 | btrfs_item_ptr_offset(l, path->slots[0]), | |
1351 | sizeof(root->root_item)); | |
1352 | memcpy(&root->root_key, location, sizeof(*location)); | |
0f7d52f4 | 1353 | } |
0f7d52f4 CM |
1354 | btrfs_free_path(path); |
1355 | if (ret) { | |
5e540f77 | 1356 | kfree(root); |
13a8a7c8 YZ |
1357 | if (ret > 0) |
1358 | ret = -ENOENT; | |
0f7d52f4 CM |
1359 | return ERR_PTR(ret); |
1360 | } | |
13a8a7c8 | 1361 | |
84234f3a | 1362 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1363 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1364 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1365 | blocksize, generation); |
5d4f98a2 | 1366 | root->commit_root = btrfs_root_node(root); |
79787eaa | 1367 | BUG_ON(!root->node); /* -ENOMEM */ |
13a8a7c8 | 1368 | out: |
08fe4db1 | 1369 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { |
e02119d5 | 1370 | root->ref_cows = 1; |
08fe4db1 LZ |
1371 | btrfs_check_and_init_root_item(&root->root_item); |
1372 | } | |
13a8a7c8 | 1373 | |
5eda7b5e CM |
1374 | return root; |
1375 | } | |
1376 | ||
edbd8d4e CM |
1377 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
1378 | struct btrfs_key *location) | |
5eda7b5e CM |
1379 | { |
1380 | struct btrfs_root *root; | |
1381 | int ret; | |
1382 | ||
edbd8d4e CM |
1383 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1384 | return fs_info->tree_root; | |
1385 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1386 | return fs_info->extent_root; | |
8f18cf13 CM |
1387 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1388 | return fs_info->chunk_root; | |
1389 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1390 | return fs_info->dev_root; | |
0403e47e YZ |
1391 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1392 | return fs_info->csum_root; | |
4df27c4d YZ |
1393 | again: |
1394 | spin_lock(&fs_info->fs_roots_radix_lock); | |
5eda7b5e CM |
1395 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
1396 | (unsigned long)location->objectid); | |
4df27c4d | 1397 | spin_unlock(&fs_info->fs_roots_radix_lock); |
5eda7b5e CM |
1398 | if (root) |
1399 | return root; | |
1400 | ||
e02119d5 | 1401 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
1402 | if (IS_ERR(root)) |
1403 | return root; | |
3394e160 | 1404 | |
581bb050 | 1405 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); |
581bb050 LZ |
1406 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), |
1407 | GFP_NOFS); | |
35a30d7c DS |
1408 | if (!root->free_ino_pinned || !root->free_ino_ctl) { |
1409 | ret = -ENOMEM; | |
581bb050 | 1410 | goto fail; |
35a30d7c | 1411 | } |
581bb050 LZ |
1412 | |
1413 | btrfs_init_free_ino_ctl(root); | |
1414 | mutex_init(&root->fs_commit_mutex); | |
1415 | spin_lock_init(&root->cache_lock); | |
1416 | init_waitqueue_head(&root->cache_wait); | |
1417 | ||
0ee5dc67 | 1418 | ret = get_anon_bdev(&root->anon_dev); |
ac08aedf CM |
1419 | if (ret) |
1420 | goto fail; | |
3394e160 | 1421 | |
d68fc57b YZ |
1422 | if (btrfs_root_refs(&root->root_item) == 0) { |
1423 | ret = -ENOENT; | |
1424 | goto fail; | |
1425 | } | |
1426 | ||
1427 | ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid); | |
1428 | if (ret < 0) | |
1429 | goto fail; | |
1430 | if (ret == 0) | |
1431 | root->orphan_item_inserted = 1; | |
1432 | ||
4df27c4d YZ |
1433 | ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); |
1434 | if (ret) | |
1435 | goto fail; | |
1436 | ||
1437 | spin_lock(&fs_info->fs_roots_radix_lock); | |
2619ba1f CM |
1438 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
1439 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 | 1440 | root); |
d68fc57b | 1441 | if (ret == 0) |
4df27c4d | 1442 | root->in_radix = 1; |
d68fc57b | 1443 | |
4df27c4d YZ |
1444 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1445 | radix_tree_preload_end(); | |
0f7d52f4 | 1446 | if (ret) { |
4df27c4d YZ |
1447 | if (ret == -EEXIST) { |
1448 | free_fs_root(root); | |
1449 | goto again; | |
1450 | } | |
1451 | goto fail; | |
0f7d52f4 | 1452 | } |
4df27c4d YZ |
1453 | |
1454 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
1455 | root->root_key.objectid); | |
1456 | WARN_ON(ret); | |
edbd8d4e | 1457 | return root; |
4df27c4d YZ |
1458 | fail: |
1459 | free_fs_root(root); | |
1460 | return ERR_PTR(ret); | |
edbd8d4e CM |
1461 | } |
1462 | ||
04160088 CM |
1463 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1464 | { | |
1465 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1466 | int ret = 0; | |
04160088 CM |
1467 | struct btrfs_device *device; |
1468 | struct backing_dev_info *bdi; | |
b7967db7 | 1469 | |
1f78160c XG |
1470 | rcu_read_lock(); |
1471 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1472 | if (!device->bdev) |
1473 | continue; | |
04160088 CM |
1474 | bdi = blk_get_backing_dev_info(device->bdev); |
1475 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1476 | ret = 1; | |
1477 | break; | |
1478 | } | |
1479 | } | |
1f78160c | 1480 | rcu_read_unlock(); |
04160088 CM |
1481 | return ret; |
1482 | } | |
1483 | ||
ad081f14 JA |
1484 | /* |
1485 | * If this fails, caller must call bdi_destroy() to get rid of the | |
1486 | * bdi again. | |
1487 | */ | |
04160088 CM |
1488 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1489 | { | |
ad081f14 JA |
1490 | int err; |
1491 | ||
1492 | bdi->capabilities = BDI_CAP_MAP_COPY; | |
e6d086d8 | 1493 | err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY); |
ad081f14 JA |
1494 | if (err) |
1495 | return err; | |
1496 | ||
4575c9cc | 1497 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1498 | bdi->congested_fn = btrfs_congested_fn; |
1499 | bdi->congested_data = info; | |
1500 | return 0; | |
1501 | } | |
1502 | ||
8b712842 CM |
1503 | /* |
1504 | * called by the kthread helper functions to finally call the bio end_io | |
1505 | * functions. This is where read checksum verification actually happens | |
1506 | */ | |
1507 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1508 | { |
ce9adaa5 | 1509 | struct bio *bio; |
8b712842 CM |
1510 | struct end_io_wq *end_io_wq; |
1511 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1512 | int error; |
ce9adaa5 | 1513 | |
8b712842 CM |
1514 | end_io_wq = container_of(work, struct end_io_wq, work); |
1515 | bio = end_io_wq->bio; | |
1516 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1517 | |
8b712842 CM |
1518 | error = end_io_wq->error; |
1519 | bio->bi_private = end_io_wq->private; | |
1520 | bio->bi_end_io = end_io_wq->end_io; | |
1521 | kfree(end_io_wq); | |
8b712842 | 1522 | bio_endio(bio, error); |
44b8bd7e CM |
1523 | } |
1524 | ||
a74a4b97 CM |
1525 | static int cleaner_kthread(void *arg) |
1526 | { | |
1527 | struct btrfs_root *root = arg; | |
1528 | ||
1529 | do { | |
a74a4b97 | 1530 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); |
76dda93c YZ |
1531 | |
1532 | if (!(root->fs_info->sb->s_flags & MS_RDONLY) && | |
1533 | mutex_trylock(&root->fs_info->cleaner_mutex)) { | |
24bbcf04 | 1534 | btrfs_run_delayed_iputs(root); |
76dda93c YZ |
1535 | btrfs_clean_old_snapshots(root); |
1536 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
4cb5300b | 1537 | btrfs_run_defrag_inodes(root->fs_info); |
76dda93c | 1538 | } |
a74a4b97 | 1539 | |
a0acae0e | 1540 | if (!try_to_freeze()) { |
a74a4b97 | 1541 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1542 | if (!kthread_should_stop()) |
1543 | schedule(); | |
a74a4b97 CM |
1544 | __set_current_state(TASK_RUNNING); |
1545 | } | |
1546 | } while (!kthread_should_stop()); | |
1547 | return 0; | |
1548 | } | |
1549 | ||
1550 | static int transaction_kthread(void *arg) | |
1551 | { | |
1552 | struct btrfs_root *root = arg; | |
1553 | struct btrfs_trans_handle *trans; | |
1554 | struct btrfs_transaction *cur; | |
8929ecfa | 1555 | u64 transid; |
a74a4b97 CM |
1556 | unsigned long now; |
1557 | unsigned long delay; | |
914b2007 | 1558 | bool cannot_commit; |
a74a4b97 CM |
1559 | |
1560 | do { | |
914b2007 | 1561 | cannot_commit = false; |
a74a4b97 CM |
1562 | delay = HZ * 30; |
1563 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1564 | mutex_lock(&root->fs_info->transaction_kthread_mutex); | |
1565 | ||
a4abeea4 | 1566 | spin_lock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1567 | cur = root->fs_info->running_transaction; |
1568 | if (!cur) { | |
a4abeea4 | 1569 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1570 | goto sleep; |
1571 | } | |
31153d81 | 1572 | |
a74a4b97 | 1573 | now = get_seconds(); |
8929ecfa YZ |
1574 | if (!cur->blocked && |
1575 | (now < cur->start_time || now - cur->start_time < 30)) { | |
a4abeea4 | 1576 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1577 | delay = HZ * 5; |
1578 | goto sleep; | |
1579 | } | |
8929ecfa | 1580 | transid = cur->transid; |
a4abeea4 | 1581 | spin_unlock(&root->fs_info->trans_lock); |
56bec294 | 1582 | |
79787eaa | 1583 | /* If the file system is aborted, this will always fail. */ |
7a7eaa40 | 1584 | trans = btrfs_join_transaction(root); |
914b2007 JK |
1585 | if (IS_ERR(trans)) { |
1586 | cannot_commit = true; | |
79787eaa | 1587 | goto sleep; |
914b2007 | 1588 | } |
8929ecfa | 1589 | if (transid == trans->transid) { |
79787eaa | 1590 | btrfs_commit_transaction(trans, root); |
8929ecfa YZ |
1591 | } else { |
1592 | btrfs_end_transaction(trans, root); | |
1593 | } | |
a74a4b97 CM |
1594 | sleep: |
1595 | wake_up_process(root->fs_info->cleaner_kthread); | |
1596 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1597 | ||
a0acae0e | 1598 | if (!try_to_freeze()) { |
a74a4b97 | 1599 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa | 1600 | if (!kthread_should_stop() && |
914b2007 JK |
1601 | (!btrfs_transaction_blocked(root->fs_info) || |
1602 | cannot_commit)) | |
8929ecfa | 1603 | schedule_timeout(delay); |
a74a4b97 CM |
1604 | __set_current_state(TASK_RUNNING); |
1605 | } | |
1606 | } while (!kthread_should_stop()); | |
1607 | return 0; | |
1608 | } | |
1609 | ||
af31f5e5 CM |
1610 | /* |
1611 | * this will find the highest generation in the array of | |
1612 | * root backups. The index of the highest array is returned, | |
1613 | * or -1 if we can't find anything. | |
1614 | * | |
1615 | * We check to make sure the array is valid by comparing the | |
1616 | * generation of the latest root in the array with the generation | |
1617 | * in the super block. If they don't match we pitch it. | |
1618 | */ | |
1619 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1620 | { | |
1621 | u64 cur; | |
1622 | int newest_index = -1; | |
1623 | struct btrfs_root_backup *root_backup; | |
1624 | int i; | |
1625 | ||
1626 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1627 | root_backup = info->super_copy->super_roots + i; | |
1628 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1629 | if (cur == newest_gen) | |
1630 | newest_index = i; | |
1631 | } | |
1632 | ||
1633 | /* check to see if we actually wrapped around */ | |
1634 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
1635 | root_backup = info->super_copy->super_roots; | |
1636 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1637 | if (cur == newest_gen) | |
1638 | newest_index = 0; | |
1639 | } | |
1640 | return newest_index; | |
1641 | } | |
1642 | ||
1643 | ||
1644 | /* | |
1645 | * find the oldest backup so we know where to store new entries | |
1646 | * in the backup array. This will set the backup_root_index | |
1647 | * field in the fs_info struct | |
1648 | */ | |
1649 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
1650 | u64 newest_gen) | |
1651 | { | |
1652 | int newest_index = -1; | |
1653 | ||
1654 | newest_index = find_newest_super_backup(info, newest_gen); | |
1655 | /* if there was garbage in there, just move along */ | |
1656 | if (newest_index == -1) { | |
1657 | info->backup_root_index = 0; | |
1658 | } else { | |
1659 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1660 | } | |
1661 | } | |
1662 | ||
1663 | /* | |
1664 | * copy all the root pointers into the super backup array. | |
1665 | * this will bump the backup pointer by one when it is | |
1666 | * done | |
1667 | */ | |
1668 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1669 | { | |
1670 | int next_backup; | |
1671 | struct btrfs_root_backup *root_backup; | |
1672 | int last_backup; | |
1673 | ||
1674 | next_backup = info->backup_root_index; | |
1675 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1676 | BTRFS_NUM_BACKUP_ROOTS; | |
1677 | ||
1678 | /* | |
1679 | * just overwrite the last backup if we're at the same generation | |
1680 | * this happens only at umount | |
1681 | */ | |
1682 | root_backup = info->super_for_commit->super_roots + last_backup; | |
1683 | if (btrfs_backup_tree_root_gen(root_backup) == | |
1684 | btrfs_header_generation(info->tree_root->node)) | |
1685 | next_backup = last_backup; | |
1686 | ||
1687 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1688 | ||
1689 | /* | |
1690 | * make sure all of our padding and empty slots get zero filled | |
1691 | * regardless of which ones we use today | |
1692 | */ | |
1693 | memset(root_backup, 0, sizeof(*root_backup)); | |
1694 | ||
1695 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1696 | ||
1697 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1698 | btrfs_set_backup_tree_root_gen(root_backup, | |
1699 | btrfs_header_generation(info->tree_root->node)); | |
1700 | ||
1701 | btrfs_set_backup_tree_root_level(root_backup, | |
1702 | btrfs_header_level(info->tree_root->node)); | |
1703 | ||
1704 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1705 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1706 | btrfs_header_generation(info->chunk_root->node)); | |
1707 | btrfs_set_backup_chunk_root_level(root_backup, | |
1708 | btrfs_header_level(info->chunk_root->node)); | |
1709 | ||
1710 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
1711 | btrfs_set_backup_extent_root_gen(root_backup, | |
1712 | btrfs_header_generation(info->extent_root->node)); | |
1713 | btrfs_set_backup_extent_root_level(root_backup, | |
1714 | btrfs_header_level(info->extent_root->node)); | |
1715 | ||
7c7e82a7 CM |
1716 | /* |
1717 | * we might commit during log recovery, which happens before we set | |
1718 | * the fs_root. Make sure it is valid before we fill it in. | |
1719 | */ | |
1720 | if (info->fs_root && info->fs_root->node) { | |
1721 | btrfs_set_backup_fs_root(root_backup, | |
1722 | info->fs_root->node->start); | |
1723 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1724 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1725 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1726 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1727 | } |
af31f5e5 CM |
1728 | |
1729 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1730 | btrfs_set_backup_dev_root_gen(root_backup, | |
1731 | btrfs_header_generation(info->dev_root->node)); | |
1732 | btrfs_set_backup_dev_root_level(root_backup, | |
1733 | btrfs_header_level(info->dev_root->node)); | |
1734 | ||
1735 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
1736 | btrfs_set_backup_csum_root_gen(root_backup, | |
1737 | btrfs_header_generation(info->csum_root->node)); | |
1738 | btrfs_set_backup_csum_root_level(root_backup, | |
1739 | btrfs_header_level(info->csum_root->node)); | |
1740 | ||
1741 | btrfs_set_backup_total_bytes(root_backup, | |
1742 | btrfs_super_total_bytes(info->super_copy)); | |
1743 | btrfs_set_backup_bytes_used(root_backup, | |
1744 | btrfs_super_bytes_used(info->super_copy)); | |
1745 | btrfs_set_backup_num_devices(root_backup, | |
1746 | btrfs_super_num_devices(info->super_copy)); | |
1747 | ||
1748 | /* | |
1749 | * if we don't copy this out to the super_copy, it won't get remembered | |
1750 | * for the next commit | |
1751 | */ | |
1752 | memcpy(&info->super_copy->super_roots, | |
1753 | &info->super_for_commit->super_roots, | |
1754 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1755 | } | |
1756 | ||
1757 | /* | |
1758 | * this copies info out of the root backup array and back into | |
1759 | * the in-memory super block. It is meant to help iterate through | |
1760 | * the array, so you send it the number of backups you've already | |
1761 | * tried and the last backup index you used. | |
1762 | * | |
1763 | * this returns -1 when it has tried all the backups | |
1764 | */ | |
1765 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
1766 | struct btrfs_super_block *super, | |
1767 | int *num_backups_tried, int *backup_index) | |
1768 | { | |
1769 | struct btrfs_root_backup *root_backup; | |
1770 | int newest = *backup_index; | |
1771 | ||
1772 | if (*num_backups_tried == 0) { | |
1773 | u64 gen = btrfs_super_generation(super); | |
1774 | ||
1775 | newest = find_newest_super_backup(info, gen); | |
1776 | if (newest == -1) | |
1777 | return -1; | |
1778 | ||
1779 | *backup_index = newest; | |
1780 | *num_backups_tried = 1; | |
1781 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
1782 | /* we've tried all the backups, all done */ | |
1783 | return -1; | |
1784 | } else { | |
1785 | /* jump to the next oldest backup */ | |
1786 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1787 | BTRFS_NUM_BACKUP_ROOTS; | |
1788 | *backup_index = newest; | |
1789 | *num_backups_tried += 1; | |
1790 | } | |
1791 | root_backup = super->super_roots + newest; | |
1792 | ||
1793 | btrfs_set_super_generation(super, | |
1794 | btrfs_backup_tree_root_gen(root_backup)); | |
1795 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
1796 | btrfs_set_super_root_level(super, | |
1797 | btrfs_backup_tree_root_level(root_backup)); | |
1798 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
1799 | ||
1800 | /* | |
1801 | * fixme: the total bytes and num_devices need to match or we should | |
1802 | * need a fsck | |
1803 | */ | |
1804 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
1805 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
1806 | return 0; | |
1807 | } | |
1808 | ||
1809 | /* helper to cleanup tree roots */ | |
1810 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
1811 | { | |
1812 | free_extent_buffer(info->tree_root->node); | |
1813 | free_extent_buffer(info->tree_root->commit_root); | |
1814 | free_extent_buffer(info->dev_root->node); | |
1815 | free_extent_buffer(info->dev_root->commit_root); | |
1816 | free_extent_buffer(info->extent_root->node); | |
1817 | free_extent_buffer(info->extent_root->commit_root); | |
1818 | free_extent_buffer(info->csum_root->node); | |
1819 | free_extent_buffer(info->csum_root->commit_root); | |
1820 | ||
1821 | info->tree_root->node = NULL; | |
1822 | info->tree_root->commit_root = NULL; | |
1823 | info->dev_root->node = NULL; | |
1824 | info->dev_root->commit_root = NULL; | |
1825 | info->extent_root->node = NULL; | |
1826 | info->extent_root->commit_root = NULL; | |
1827 | info->csum_root->node = NULL; | |
1828 | info->csum_root->commit_root = NULL; | |
1829 | ||
1830 | if (chunk_root) { | |
1831 | free_extent_buffer(info->chunk_root->node); | |
1832 | free_extent_buffer(info->chunk_root->commit_root); | |
1833 | info->chunk_root->node = NULL; | |
1834 | info->chunk_root->commit_root = NULL; | |
1835 | } | |
1836 | } | |
1837 | ||
1838 | ||
ad2b2c80 AV |
1839 | int open_ctree(struct super_block *sb, |
1840 | struct btrfs_fs_devices *fs_devices, | |
1841 | char *options) | |
2e635a27 | 1842 | { |
db94535d CM |
1843 | u32 sectorsize; |
1844 | u32 nodesize; | |
1845 | u32 leafsize; | |
1846 | u32 blocksize; | |
87ee04eb | 1847 | u32 stripesize; |
84234f3a | 1848 | u64 generation; |
f2b636e8 | 1849 | u64 features; |
3de4586c | 1850 | struct btrfs_key location; |
a061fc8d | 1851 | struct buffer_head *bh; |
4d34b278 | 1852 | struct btrfs_super_block *disk_super; |
815745cf | 1853 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
f84a8bd6 | 1854 | struct btrfs_root *tree_root; |
4d34b278 ID |
1855 | struct btrfs_root *extent_root; |
1856 | struct btrfs_root *csum_root; | |
1857 | struct btrfs_root *chunk_root; | |
1858 | struct btrfs_root *dev_root; | |
e02119d5 | 1859 | struct btrfs_root *log_tree_root; |
eb60ceac | 1860 | int ret; |
e58ca020 | 1861 | int err = -EINVAL; |
af31f5e5 CM |
1862 | int num_backups_tried = 0; |
1863 | int backup_index = 0; | |
4543df7e | 1864 | |
f84a8bd6 | 1865 | tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info); |
6f07e42e AV |
1866 | extent_root = fs_info->extent_root = btrfs_alloc_root(fs_info); |
1867 | csum_root = fs_info->csum_root = btrfs_alloc_root(fs_info); | |
1868 | chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info); | |
1869 | dev_root = fs_info->dev_root = btrfs_alloc_root(fs_info); | |
8790d502 | 1870 | |
f84a8bd6 AV |
1871 | if (!tree_root || !extent_root || !csum_root || |
1872 | !chunk_root || !dev_root) { | |
39279cc3 CM |
1873 | err = -ENOMEM; |
1874 | goto fail; | |
1875 | } | |
76dda93c YZ |
1876 | |
1877 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
1878 | if (ret) { | |
1879 | err = ret; | |
1880 | goto fail; | |
1881 | } | |
1882 | ||
1883 | ret = setup_bdi(fs_info, &fs_info->bdi); | |
1884 | if (ret) { | |
1885 | err = ret; | |
1886 | goto fail_srcu; | |
1887 | } | |
1888 | ||
1889 | fs_info->btree_inode = new_inode(sb); | |
1890 | if (!fs_info->btree_inode) { | |
1891 | err = -ENOMEM; | |
1892 | goto fail_bdi; | |
1893 | } | |
1894 | ||
a6591715 | 1895 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
1561deda | 1896 | |
76dda93c | 1897 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
8fd17795 | 1898 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1899 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 1900 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
19c00ddc | 1901 | INIT_LIST_HEAD(&fs_info->hashers); |
ea8c2819 | 1902 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
5a3f23d5 | 1903 | INIT_LIST_HEAD(&fs_info->ordered_operations); |
11833d66 | 1904 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
1832a6d5 | 1905 | spin_lock_init(&fs_info->delalloc_lock); |
a4abeea4 | 1906 | spin_lock_init(&fs_info->trans_lock); |
31153d81 | 1907 | spin_lock_init(&fs_info->ref_cache_lock); |
76dda93c | 1908 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 1909 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 1910 | spin_lock_init(&fs_info->defrag_inodes_lock); |
2bf64758 | 1911 | spin_lock_init(&fs_info->free_chunk_lock); |
7585717f | 1912 | mutex_init(&fs_info->reloc_mutex); |
19c00ddc | 1913 | |
58176a96 | 1914 | init_completion(&fs_info->kobj_unregister); |
0b86a832 | 1915 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1916 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1917 | btrfs_mapping_init(&fs_info->mapping_tree); |
f0486c68 YZ |
1918 | btrfs_init_block_rsv(&fs_info->global_block_rsv); |
1919 | btrfs_init_block_rsv(&fs_info->delalloc_block_rsv); | |
1920 | btrfs_init_block_rsv(&fs_info->trans_block_rsv); | |
1921 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv); | |
1922 | btrfs_init_block_rsv(&fs_info->empty_block_rsv); | |
6d668dda | 1923 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv); |
cb03c743 | 1924 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 1925 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 1926 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 1927 | atomic_set(&fs_info->nr_async_bios, 0); |
4cb5300b | 1928 | atomic_set(&fs_info->defrag_running, 0); |
e20d96d6 | 1929 | fs_info->sb = sb; |
6f568d35 | 1930 | fs_info->max_inline = 8192 * 1024; |
9ed74f2d | 1931 | fs_info->metadata_ratio = 0; |
4cb5300b | 1932 | fs_info->defrag_inodes = RB_ROOT; |
a4abeea4 | 1933 | fs_info->trans_no_join = 0; |
2bf64758 | 1934 | fs_info->free_chunk_space = 0; |
c8b97818 | 1935 | |
90519d66 AJ |
1936 | /* readahead state */ |
1937 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); | |
1938 | spin_lock_init(&fs_info->reada_lock); | |
c8b97818 | 1939 | |
b34b086c CM |
1940 | fs_info->thread_pool_size = min_t(unsigned long, |
1941 | num_online_cpus() + 2, 8); | |
0afbaf8c | 1942 | |
3eaa2885 CM |
1943 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
1944 | spin_lock_init(&fs_info->ordered_extent_lock); | |
16cdcec7 MX |
1945 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
1946 | GFP_NOFS); | |
1947 | if (!fs_info->delayed_root) { | |
1948 | err = -ENOMEM; | |
1949 | goto fail_iput; | |
1950 | } | |
1951 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 1952 | |
a2de733c AJ |
1953 | mutex_init(&fs_info->scrub_lock); |
1954 | atomic_set(&fs_info->scrubs_running, 0); | |
1955 | atomic_set(&fs_info->scrub_pause_req, 0); | |
1956 | atomic_set(&fs_info->scrubs_paused, 0); | |
1957 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
1958 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
1959 | init_rwsem(&fs_info->scrub_super_lock); | |
1960 | fs_info->scrub_workers_refcnt = 0; | |
21adbd5c SB |
1961 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
1962 | fs_info->check_integrity_print_mask = 0; | |
1963 | #endif | |
a2de733c | 1964 | |
c9e9f97b ID |
1965 | spin_lock_init(&fs_info->balance_lock); |
1966 | mutex_init(&fs_info->balance_mutex); | |
837d5b6e ID |
1967 | atomic_set(&fs_info->balance_running, 0); |
1968 | atomic_set(&fs_info->balance_pause_req, 0); | |
a7e99c69 | 1969 | atomic_set(&fs_info->balance_cancel_req, 0); |
c9e9f97b | 1970 | fs_info->balance_ctl = NULL; |
837d5b6e | 1971 | init_waitqueue_head(&fs_info->balance_wait_q); |
a2de733c | 1972 | |
a061fc8d CM |
1973 | sb->s_blocksize = 4096; |
1974 | sb->s_blocksize_bits = blksize_bits(4096); | |
32a88aa1 | 1975 | sb->s_bdi = &fs_info->bdi; |
a061fc8d | 1976 | |
76dda93c | 1977 | fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; |
bfe86848 | 1978 | set_nlink(fs_info->btree_inode, 1); |
0afbaf8c CM |
1979 | /* |
1980 | * we set the i_size on the btree inode to the max possible int. | |
1981 | * the real end of the address space is determined by all of | |
1982 | * the devices in the system | |
1983 | */ | |
1984 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 1985 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
1986 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
1987 | ||
5d4f98a2 | 1988 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); |
d1310b2e | 1989 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
f993c883 | 1990 | fs_info->btree_inode->i_mapping); |
0b32f4bb | 1991 | BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; |
a8067e02 | 1992 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); |
d1310b2e CM |
1993 | |
1994 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 1995 | |
76dda93c YZ |
1996 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
1997 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
1998 | sizeof(struct btrfs_key)); | |
1999 | BTRFS_I(fs_info->btree_inode)->dummy_inode = 1; | |
c65ddb52 | 2000 | insert_inode_hash(fs_info->btree_inode); |
76dda93c | 2001 | |
0f9dd46c | 2002 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2003 | fs_info->block_group_cache_tree = RB_ROOT; |
0f9dd46c | 2004 | |
11833d66 | 2005 | extent_io_tree_init(&fs_info->freed_extents[0], |
f993c883 | 2006 | fs_info->btree_inode->i_mapping); |
11833d66 | 2007 | extent_io_tree_init(&fs_info->freed_extents[1], |
f993c883 | 2008 | fs_info->btree_inode->i_mapping); |
11833d66 | 2009 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
e66f709b | 2010 | fs_info->do_barriers = 1; |
e18e4809 | 2011 | |
39279cc3 | 2012 | |
5a3f23d5 | 2013 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2014 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2015 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2016 | mutex_init(&fs_info->transaction_kthread_mutex); |
2017 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 2018 | mutex_init(&fs_info->volume_mutex); |
276e680d | 2019 | init_rwsem(&fs_info->extent_commit_sem); |
c71bf099 | 2020 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2021 | init_rwsem(&fs_info->subvol_sem); |
fa9c0d79 CM |
2022 | |
2023 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); | |
2024 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2025 | ||
e6dcd2dc | 2026 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2027 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2028 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2029 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2030 | |
0b86a832 | 2031 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 2032 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2033 | |
3c4bb26b | 2034 | invalidate_bdev(fs_devices->latest_bdev); |
a512bbf8 | 2035 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
20b45077 DY |
2036 | if (!bh) { |
2037 | err = -EINVAL; | |
16cdcec7 | 2038 | goto fail_alloc; |
20b45077 | 2039 | } |
39279cc3 | 2040 | |
6c41761f DS |
2041 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2042 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2043 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2044 | brelse(bh); |
5f39d397 | 2045 | |
6c41761f | 2046 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2047 | |
6c41761f | 2048 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2049 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2050 | goto fail_alloc; |
0f7d52f4 | 2051 | |
acce952b | 2052 | /* check FS state, whether FS is broken. */ |
2053 | fs_info->fs_state |= btrfs_super_flags(disk_super); | |
2054 | ||
fcd1f065 DS |
2055 | ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); |
2056 | if (ret) { | |
2057 | printk(KERN_ERR "btrfs: superblock contains fatal errors\n"); | |
2058 | err = ret; | |
2059 | goto fail_alloc; | |
2060 | } | |
acce952b | 2061 | |
af31f5e5 CM |
2062 | /* |
2063 | * run through our array of backup supers and setup | |
2064 | * our ring pointer to the oldest one | |
2065 | */ | |
2066 | generation = btrfs_super_generation(disk_super); | |
2067 | find_oldest_super_backup(fs_info, generation); | |
2068 | ||
75e7cb7f LB |
2069 | /* |
2070 | * In the long term, we'll store the compression type in the super | |
2071 | * block, and it'll be used for per file compression control. | |
2072 | */ | |
2073 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2074 | ||
2b82032c YZ |
2075 | ret = btrfs_parse_options(tree_root, options); |
2076 | if (ret) { | |
2077 | err = ret; | |
16cdcec7 | 2078 | goto fail_alloc; |
2b82032c | 2079 | } |
dfe25020 | 2080 | |
f2b636e8 JB |
2081 | features = btrfs_super_incompat_flags(disk_super) & |
2082 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2083 | if (features) { | |
2084 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
2085 | "unsupported optional features (%Lx).\n", | |
21380931 | 2086 | (unsigned long long)features); |
f2b636e8 | 2087 | err = -EINVAL; |
16cdcec7 | 2088 | goto fail_alloc; |
f2b636e8 JB |
2089 | } |
2090 | ||
727011e0 CM |
2091 | if (btrfs_super_leafsize(disk_super) != |
2092 | btrfs_super_nodesize(disk_super)) { | |
2093 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2094 | "blocksizes don't match. node %d leaf %d\n", | |
2095 | btrfs_super_nodesize(disk_super), | |
2096 | btrfs_super_leafsize(disk_super)); | |
2097 | err = -EINVAL; | |
2098 | goto fail_alloc; | |
2099 | } | |
2100 | if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2101 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2102 | "blocksize (%d) was too large\n", | |
2103 | btrfs_super_leafsize(disk_super)); | |
2104 | err = -EINVAL; | |
2105 | goto fail_alloc; | |
2106 | } | |
2107 | ||
5d4f98a2 | 2108 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae LZ |
2109 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
2110 | if (tree_root->fs_info->compress_type & BTRFS_COMPRESS_LZO) | |
2111 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; | |
727011e0 CM |
2112 | |
2113 | /* | |
2114 | * flag our filesystem as having big metadata blocks if | |
2115 | * they are bigger than the page size | |
2116 | */ | |
2117 | if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) { | |
2118 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) | |
2119 | printk(KERN_INFO "btrfs flagging fs with big metadata feature\n"); | |
2120 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
2121 | } | |
2122 | ||
a6fa6fae | 2123 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 2124 | |
f2b636e8 JB |
2125 | features = btrfs_super_compat_ro_flags(disk_super) & |
2126 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
2127 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
2128 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
2129 | "unsupported option features (%Lx).\n", | |
21380931 | 2130 | (unsigned long long)features); |
f2b636e8 | 2131 | err = -EINVAL; |
16cdcec7 | 2132 | goto fail_alloc; |
f2b636e8 | 2133 | } |
61d92c32 CM |
2134 | |
2135 | btrfs_init_workers(&fs_info->generic_worker, | |
2136 | "genwork", 1, NULL); | |
2137 | ||
5443be45 | 2138 | btrfs_init_workers(&fs_info->workers, "worker", |
61d92c32 CM |
2139 | fs_info->thread_pool_size, |
2140 | &fs_info->generic_worker); | |
c8b97818 | 2141 | |
771ed689 | 2142 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
61d92c32 CM |
2143 | fs_info->thread_pool_size, |
2144 | &fs_info->generic_worker); | |
771ed689 | 2145 | |
5443be45 | 2146 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 | 2147 | min_t(u64, fs_devices->num_devices, |
61d92c32 CM |
2148 | fs_info->thread_pool_size), |
2149 | &fs_info->generic_worker); | |
61b49440 | 2150 | |
bab39bf9 JB |
2151 | btrfs_init_workers(&fs_info->caching_workers, "cache", |
2152 | 2, &fs_info->generic_worker); | |
2153 | ||
61b49440 CM |
2154 | /* a higher idle thresh on the submit workers makes it much more |
2155 | * likely that bios will be send down in a sane order to the | |
2156 | * devices | |
2157 | */ | |
2158 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 2159 | |
771ed689 | 2160 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 2161 | fs_info->workers.ordered = 1; |
61b49440 | 2162 | |
771ed689 CM |
2163 | fs_info->delalloc_workers.idle_thresh = 2; |
2164 | fs_info->delalloc_workers.ordered = 1; | |
2165 | ||
61d92c32 CM |
2166 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1, |
2167 | &fs_info->generic_worker); | |
5443be45 | 2168 | btrfs_init_workers(&fs_info->endio_workers, "endio", |
61d92c32 CM |
2169 | fs_info->thread_pool_size, |
2170 | &fs_info->generic_worker); | |
d20f7043 | 2171 | btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", |
61d92c32 CM |
2172 | fs_info->thread_pool_size, |
2173 | &fs_info->generic_worker); | |
cad321ad | 2174 | btrfs_init_workers(&fs_info->endio_meta_write_workers, |
61d92c32 CM |
2175 | "endio-meta-write", fs_info->thread_pool_size, |
2176 | &fs_info->generic_worker); | |
5443be45 | 2177 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", |
61d92c32 CM |
2178 | fs_info->thread_pool_size, |
2179 | &fs_info->generic_worker); | |
0cb59c99 JB |
2180 | btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write", |
2181 | 1, &fs_info->generic_worker); | |
16cdcec7 MX |
2182 | btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta", |
2183 | fs_info->thread_pool_size, | |
2184 | &fs_info->generic_worker); | |
90519d66 AJ |
2185 | btrfs_init_workers(&fs_info->readahead_workers, "readahead", |
2186 | fs_info->thread_pool_size, | |
2187 | &fs_info->generic_worker); | |
61b49440 CM |
2188 | |
2189 | /* | |
2190 | * endios are largely parallel and should have a very | |
2191 | * low idle thresh | |
2192 | */ | |
2193 | fs_info->endio_workers.idle_thresh = 4; | |
b51912c9 CM |
2194 | fs_info->endio_meta_workers.idle_thresh = 4; |
2195 | ||
9042846b CM |
2196 | fs_info->endio_write_workers.idle_thresh = 2; |
2197 | fs_info->endio_meta_write_workers.idle_thresh = 2; | |
90519d66 | 2198 | fs_info->readahead_workers.idle_thresh = 2; |
9042846b | 2199 | |
0dc3b84a JB |
2200 | /* |
2201 | * btrfs_start_workers can really only fail because of ENOMEM so just | |
2202 | * return -ENOMEM if any of these fail. | |
2203 | */ | |
2204 | ret = btrfs_start_workers(&fs_info->workers); | |
2205 | ret |= btrfs_start_workers(&fs_info->generic_worker); | |
2206 | ret |= btrfs_start_workers(&fs_info->submit_workers); | |
2207 | ret |= btrfs_start_workers(&fs_info->delalloc_workers); | |
2208 | ret |= btrfs_start_workers(&fs_info->fixup_workers); | |
2209 | ret |= btrfs_start_workers(&fs_info->endio_workers); | |
2210 | ret |= btrfs_start_workers(&fs_info->endio_meta_workers); | |
2211 | ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers); | |
2212 | ret |= btrfs_start_workers(&fs_info->endio_write_workers); | |
2213 | ret |= btrfs_start_workers(&fs_info->endio_freespace_worker); | |
2214 | ret |= btrfs_start_workers(&fs_info->delayed_workers); | |
2215 | ret |= btrfs_start_workers(&fs_info->caching_workers); | |
2216 | ret |= btrfs_start_workers(&fs_info->readahead_workers); | |
2217 | if (ret) { | |
2218 | ret = -ENOMEM; | |
2219 | goto fail_sb_buffer; | |
2220 | } | |
4543df7e | 2221 | |
4575c9cc | 2222 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
2223 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
2224 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 2225 | |
db94535d CM |
2226 | nodesize = btrfs_super_nodesize(disk_super); |
2227 | leafsize = btrfs_super_leafsize(disk_super); | |
2228 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 2229 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
2230 | tree_root->nodesize = nodesize; |
2231 | tree_root->leafsize = leafsize; | |
2232 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 2233 | tree_root->stripesize = stripesize; |
a061fc8d CM |
2234 | |
2235 | sb->s_blocksize = sectorsize; | |
2236 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 2237 | |
39279cc3 CM |
2238 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
2239 | sizeof(disk_super->magic))) { | |
d397712b | 2240 | printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
2241 | goto fail_sb_buffer; |
2242 | } | |
19c00ddc | 2243 | |
941b2ddf KM |
2244 | if (sectorsize < PAGE_SIZE) { |
2245 | printk(KERN_WARNING "btrfs: Incompatible sector size " | |
2246 | "found on %s\n", sb->s_id); | |
2247 | goto fail_sb_buffer; | |
2248 | } | |
2249 | ||
65139ed9 DS |
2250 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && |
2251 | (leafsize != nodesize || sectorsize != nodesize)) { | |
2252 | printk(KERN_WARNING "btrfs: unequal leaf/node/sector sizes " | |
2253 | "are not allowed for mixed block groups on %s\n", | |
2254 | sb->s_id); | |
2255 | goto fail_sb_buffer; | |
2256 | } | |
2257 | ||
925baedd | 2258 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 2259 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 2260 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2261 | if (ret) { |
d397712b CM |
2262 | printk(KERN_WARNING "btrfs: failed to read the system " |
2263 | "array on %s\n", sb->s_id); | |
5d4f98a2 | 2264 | goto fail_sb_buffer; |
84eed90f | 2265 | } |
0b86a832 CM |
2266 | |
2267 | blocksize = btrfs_level_size(tree_root, | |
2268 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 2269 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
2270 | |
2271 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2272 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
2273 | ||
2274 | chunk_root->node = read_tree_block(chunk_root, | |
2275 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 2276 | blocksize, generation); |
79787eaa | 2277 | BUG_ON(!chunk_root->node); /* -ENOMEM */ |
83121942 DW |
2278 | if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { |
2279 | printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n", | |
2280 | sb->s_id); | |
af31f5e5 | 2281 | goto fail_tree_roots; |
83121942 | 2282 | } |
5d4f98a2 YZ |
2283 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
2284 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 2285 | |
e17cade2 | 2286 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
d397712b CM |
2287 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), |
2288 | BTRFS_UUID_SIZE); | |
e17cade2 | 2289 | |
0b86a832 | 2290 | ret = btrfs_read_chunk_tree(chunk_root); |
2b82032c | 2291 | if (ret) { |
d397712b CM |
2292 | printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", |
2293 | sb->s_id); | |
af31f5e5 | 2294 | goto fail_tree_roots; |
2b82032c | 2295 | } |
0b86a832 | 2296 | |
dfe25020 CM |
2297 | btrfs_close_extra_devices(fs_devices); |
2298 | ||
a6b0d5c8 CM |
2299 | if (!fs_devices->latest_bdev) { |
2300 | printk(KERN_CRIT "btrfs: failed to read devices on %s\n", | |
2301 | sb->s_id); | |
2302 | goto fail_tree_roots; | |
2303 | } | |
2304 | ||
af31f5e5 | 2305 | retry_root_backup: |
db94535d CM |
2306 | blocksize = btrfs_level_size(tree_root, |
2307 | btrfs_super_root_level(disk_super)); | |
84234f3a | 2308 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 2309 | |
e20d96d6 | 2310 | tree_root->node = read_tree_block(tree_root, |
db94535d | 2311 | btrfs_super_root(disk_super), |
84234f3a | 2312 | blocksize, generation); |
af31f5e5 CM |
2313 | if (!tree_root->node || |
2314 | !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { | |
83121942 DW |
2315 | printk(KERN_WARNING "btrfs: failed to read tree root on %s\n", |
2316 | sb->s_id); | |
af31f5e5 CM |
2317 | |
2318 | goto recovery_tree_root; | |
83121942 | 2319 | } |
af31f5e5 | 2320 | |
5d4f98a2 YZ |
2321 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
2322 | tree_root->commit_root = btrfs_root_node(tree_root); | |
db94535d CM |
2323 | |
2324 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 2325 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 2326 | if (ret) |
af31f5e5 | 2327 | goto recovery_tree_root; |
0b86a832 CM |
2328 | extent_root->track_dirty = 1; |
2329 | ||
2330 | ret = find_and_setup_root(tree_root, fs_info, | |
2331 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
0b86a832 | 2332 | if (ret) |
af31f5e5 | 2333 | goto recovery_tree_root; |
5d4f98a2 | 2334 | dev_root->track_dirty = 1; |
3768f368 | 2335 | |
d20f7043 CM |
2336 | ret = find_and_setup_root(tree_root, fs_info, |
2337 | BTRFS_CSUM_TREE_OBJECTID, csum_root); | |
2338 | if (ret) | |
af31f5e5 | 2339 | goto recovery_tree_root; |
d20f7043 CM |
2340 | |
2341 | csum_root->track_dirty = 1; | |
2342 | ||
8929ecfa YZ |
2343 | fs_info->generation = generation; |
2344 | fs_info->last_trans_committed = generation; | |
8929ecfa | 2345 | |
c59021f8 | 2346 | ret = btrfs_init_space_info(fs_info); |
2347 | if (ret) { | |
2348 | printk(KERN_ERR "Failed to initial space info: %d\n", ret); | |
2349 | goto fail_block_groups; | |
2350 | } | |
2351 | ||
1b1d1f66 JB |
2352 | ret = btrfs_read_block_groups(extent_root); |
2353 | if (ret) { | |
2354 | printk(KERN_ERR "Failed to read block groups: %d\n", ret); | |
2355 | goto fail_block_groups; | |
2356 | } | |
9078a3e1 | 2357 | |
a74a4b97 CM |
2358 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
2359 | "btrfs-cleaner"); | |
57506d50 | 2360 | if (IS_ERR(fs_info->cleaner_kthread)) |
1b1d1f66 | 2361 | goto fail_block_groups; |
a74a4b97 CM |
2362 | |
2363 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
2364 | tree_root, | |
2365 | "btrfs-transaction"); | |
57506d50 | 2366 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 2367 | goto fail_cleaner; |
a74a4b97 | 2368 | |
c289811c CM |
2369 | if (!btrfs_test_opt(tree_root, SSD) && |
2370 | !btrfs_test_opt(tree_root, NOSSD) && | |
2371 | !fs_info->fs_devices->rotating) { | |
2372 | printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD " | |
2373 | "mode\n"); | |
2374 | btrfs_set_opt(fs_info->mount_opt, SSD); | |
2375 | } | |
2376 | ||
21adbd5c SB |
2377 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2378 | if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { | |
2379 | ret = btrfsic_mount(tree_root, fs_devices, | |
2380 | btrfs_test_opt(tree_root, | |
2381 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? | |
2382 | 1 : 0, | |
2383 | fs_info->check_integrity_print_mask); | |
2384 | if (ret) | |
2385 | printk(KERN_WARNING "btrfs: failed to initialize" | |
2386 | " integrity check module %s\n", sb->s_id); | |
2387 | } | |
2388 | #endif | |
2389 | ||
acce952b | 2390 | /* do not make disk changes in broken FS */ |
2391 | if (btrfs_super_log_root(disk_super) != 0 && | |
2392 | !(fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)) { | |
e02119d5 CM |
2393 | u64 bytenr = btrfs_super_log_root(disk_super); |
2394 | ||
7c2ca468 | 2395 | if (fs_devices->rw_devices == 0) { |
d397712b CM |
2396 | printk(KERN_WARNING "Btrfs log replay required " |
2397 | "on RO media\n"); | |
7c2ca468 CM |
2398 | err = -EIO; |
2399 | goto fail_trans_kthread; | |
2400 | } | |
e02119d5 CM |
2401 | blocksize = |
2402 | btrfs_level_size(tree_root, | |
2403 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 2404 | |
6f07e42e | 2405 | log_tree_root = btrfs_alloc_root(fs_info); |
676e4c86 DC |
2406 | if (!log_tree_root) { |
2407 | err = -ENOMEM; | |
2408 | goto fail_trans_kthread; | |
2409 | } | |
e02119d5 CM |
2410 | |
2411 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2412 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
2413 | ||
2414 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
2415 | blocksize, |
2416 | generation + 1); | |
79787eaa | 2417 | /* returns with log_tree_root freed on success */ |
e02119d5 | 2418 | ret = btrfs_recover_log_trees(log_tree_root); |
79787eaa JM |
2419 | if (ret) { |
2420 | btrfs_error(tree_root->fs_info, ret, | |
2421 | "Failed to recover log tree"); | |
2422 | free_extent_buffer(log_tree_root->node); | |
2423 | kfree(log_tree_root); | |
2424 | goto fail_trans_kthread; | |
2425 | } | |
e556ce2c YZ |
2426 | |
2427 | if (sb->s_flags & MS_RDONLY) { | |
79787eaa JM |
2428 | ret = btrfs_commit_super(tree_root); |
2429 | if (ret) | |
2430 | goto fail_trans_kthread; | |
e556ce2c | 2431 | } |
e02119d5 | 2432 | } |
1a40e23b | 2433 | |
76dda93c | 2434 | ret = btrfs_find_orphan_roots(tree_root); |
79787eaa JM |
2435 | if (ret) |
2436 | goto fail_trans_kthread; | |
76dda93c | 2437 | |
7c2ca468 | 2438 | if (!(sb->s_flags & MS_RDONLY)) { |
d68fc57b | 2439 | ret = btrfs_cleanup_fs_roots(fs_info); |
79787eaa JM |
2440 | if (ret) { |
2441 | } | |
d68fc57b | 2442 | |
5d4f98a2 | 2443 | ret = btrfs_recover_relocation(tree_root); |
d7ce5843 MX |
2444 | if (ret < 0) { |
2445 | printk(KERN_WARNING | |
2446 | "btrfs: failed to recover relocation\n"); | |
2447 | err = -EINVAL; | |
2448 | goto fail_trans_kthread; | |
2449 | } | |
7c2ca468 | 2450 | } |
1a40e23b | 2451 | |
3de4586c CM |
2452 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
2453 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2454 | location.offset = (u64)-1; | |
2455 | ||
3de4586c CM |
2456 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
2457 | if (!fs_info->fs_root) | |
7c2ca468 | 2458 | goto fail_trans_kthread; |
3140c9a3 DC |
2459 | if (IS_ERR(fs_info->fs_root)) { |
2460 | err = PTR_ERR(fs_info->fs_root); | |
2461 | goto fail_trans_kthread; | |
2462 | } | |
c289811c | 2463 | |
e3acc2a6 JB |
2464 | if (!(sb->s_flags & MS_RDONLY)) { |
2465 | down_read(&fs_info->cleanup_work_sem); | |
66b4ffd1 JB |
2466 | err = btrfs_orphan_cleanup(fs_info->fs_root); |
2467 | if (!err) | |
2468 | err = btrfs_orphan_cleanup(fs_info->tree_root); | |
e3acc2a6 | 2469 | up_read(&fs_info->cleanup_work_sem); |
59641015 ID |
2470 | |
2471 | if (!err) | |
2472 | err = btrfs_recover_balance(fs_info->tree_root); | |
2473 | ||
66b4ffd1 JB |
2474 | if (err) { |
2475 | close_ctree(tree_root); | |
ad2b2c80 | 2476 | return err; |
66b4ffd1 | 2477 | } |
e3acc2a6 JB |
2478 | } |
2479 | ||
ad2b2c80 | 2480 | return 0; |
39279cc3 | 2481 | |
7c2ca468 CM |
2482 | fail_trans_kthread: |
2483 | kthread_stop(fs_info->transaction_kthread); | |
3f157a2f | 2484 | fail_cleaner: |
a74a4b97 | 2485 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
2486 | |
2487 | /* | |
2488 | * make sure we're done with the btree inode before we stop our | |
2489 | * kthreads | |
2490 | */ | |
2491 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
2492 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
2493 | ||
1b1d1f66 JB |
2494 | fail_block_groups: |
2495 | btrfs_free_block_groups(fs_info); | |
af31f5e5 CM |
2496 | |
2497 | fail_tree_roots: | |
2498 | free_root_pointers(fs_info, 1); | |
2499 | ||
39279cc3 | 2500 | fail_sb_buffer: |
61d92c32 | 2501 | btrfs_stop_workers(&fs_info->generic_worker); |
306c8b68 | 2502 | btrfs_stop_workers(&fs_info->readahead_workers); |
247e743c | 2503 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2504 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2505 | btrfs_stop_workers(&fs_info->workers); |
2506 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 2507 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 2508 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 2509 | btrfs_stop_workers(&fs_info->endio_write_workers); |
0cb59c99 | 2510 | btrfs_stop_workers(&fs_info->endio_freespace_worker); |
1cc127b5 | 2511 | btrfs_stop_workers(&fs_info->submit_workers); |
16cdcec7 | 2512 | btrfs_stop_workers(&fs_info->delayed_workers); |
bab39bf9 | 2513 | btrfs_stop_workers(&fs_info->caching_workers); |
16cdcec7 | 2514 | fail_alloc: |
4543df7e | 2515 | fail_iput: |
586e46e2 ID |
2516 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
2517 | ||
7c2ca468 | 2518 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
4543df7e | 2519 | iput(fs_info->btree_inode); |
ad081f14 | 2520 | fail_bdi: |
7e662854 | 2521 | bdi_destroy(&fs_info->bdi); |
76dda93c YZ |
2522 | fail_srcu: |
2523 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 2524 | fail: |
586e46e2 | 2525 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 2526 | return err; |
af31f5e5 CM |
2527 | |
2528 | recovery_tree_root: | |
af31f5e5 CM |
2529 | if (!btrfs_test_opt(tree_root, RECOVERY)) |
2530 | goto fail_tree_roots; | |
2531 | ||
2532 | free_root_pointers(fs_info, 0); | |
2533 | ||
2534 | /* don't use the log in recovery mode, it won't be valid */ | |
2535 | btrfs_set_super_log_root(disk_super, 0); | |
2536 | ||
2537 | /* we can't trust the free space cache either */ | |
2538 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2539 | ||
2540 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
2541 | &num_backups_tried, &backup_index); | |
2542 | if (ret == -1) | |
2543 | goto fail_block_groups; | |
2544 | goto retry_root_backup; | |
eb60ceac CM |
2545 | } |
2546 | ||
f2984462 CM |
2547 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
2548 | { | |
2549 | char b[BDEVNAME_SIZE]; | |
2550 | ||
2551 | if (uptodate) { | |
2552 | set_buffer_uptodate(bh); | |
2553 | } else { | |
7a36ddec | 2554 | printk_ratelimited(KERN_WARNING "lost page write due to " |
f2984462 CM |
2555 | "I/O error on %s\n", |
2556 | bdevname(bh->b_bdev, b)); | |
1259ab75 CM |
2557 | /* note, we dont' set_buffer_write_io_error because we have |
2558 | * our own ways of dealing with the IO errors | |
2559 | */ | |
f2984462 CM |
2560 | clear_buffer_uptodate(bh); |
2561 | } | |
2562 | unlock_buffer(bh); | |
2563 | put_bh(bh); | |
2564 | } | |
2565 | ||
a512bbf8 YZ |
2566 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
2567 | { | |
2568 | struct buffer_head *bh; | |
2569 | struct buffer_head *latest = NULL; | |
2570 | struct btrfs_super_block *super; | |
2571 | int i; | |
2572 | u64 transid = 0; | |
2573 | u64 bytenr; | |
2574 | ||
2575 | /* we would like to check all the supers, but that would make | |
2576 | * a btrfs mount succeed after a mkfs from a different FS. | |
2577 | * So, we need to add a special mount option to scan for | |
2578 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
2579 | */ | |
2580 | for (i = 0; i < 1; i++) { | |
2581 | bytenr = btrfs_sb_offset(i); | |
2582 | if (bytenr + 4096 >= i_size_read(bdev->bd_inode)) | |
2583 | break; | |
2584 | bh = __bread(bdev, bytenr / 4096, 4096); | |
2585 | if (!bh) | |
2586 | continue; | |
2587 | ||
2588 | super = (struct btrfs_super_block *)bh->b_data; | |
2589 | if (btrfs_super_bytenr(super) != bytenr || | |
2590 | strncmp((char *)(&super->magic), BTRFS_MAGIC, | |
2591 | sizeof(super->magic))) { | |
2592 | brelse(bh); | |
2593 | continue; | |
2594 | } | |
2595 | ||
2596 | if (!latest || btrfs_super_generation(super) > transid) { | |
2597 | brelse(latest); | |
2598 | latest = bh; | |
2599 | transid = btrfs_super_generation(super); | |
2600 | } else { | |
2601 | brelse(bh); | |
2602 | } | |
2603 | } | |
2604 | return latest; | |
2605 | } | |
2606 | ||
4eedeb75 HH |
2607 | /* |
2608 | * this should be called twice, once with wait == 0 and | |
2609 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
2610 | * we write are pinned. | |
2611 | * | |
2612 | * They are released when wait == 1 is done. | |
2613 | * max_mirrors must be the same for both runs, and it indicates how | |
2614 | * many supers on this one device should be written. | |
2615 | * | |
2616 | * max_mirrors == 0 means to write them all. | |
2617 | */ | |
a512bbf8 YZ |
2618 | static int write_dev_supers(struct btrfs_device *device, |
2619 | struct btrfs_super_block *sb, | |
2620 | int do_barriers, int wait, int max_mirrors) | |
2621 | { | |
2622 | struct buffer_head *bh; | |
2623 | int i; | |
2624 | int ret; | |
2625 | int errors = 0; | |
2626 | u32 crc; | |
2627 | u64 bytenr; | |
a512bbf8 YZ |
2628 | |
2629 | if (max_mirrors == 0) | |
2630 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
2631 | ||
a512bbf8 YZ |
2632 | for (i = 0; i < max_mirrors; i++) { |
2633 | bytenr = btrfs_sb_offset(i); | |
2634 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) | |
2635 | break; | |
2636 | ||
2637 | if (wait) { | |
2638 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
2639 | BTRFS_SUPER_INFO_SIZE); | |
2640 | BUG_ON(!bh); | |
a512bbf8 | 2641 | wait_on_buffer(bh); |
4eedeb75 HH |
2642 | if (!buffer_uptodate(bh)) |
2643 | errors++; | |
2644 | ||
2645 | /* drop our reference */ | |
2646 | brelse(bh); | |
2647 | ||
2648 | /* drop the reference from the wait == 0 run */ | |
2649 | brelse(bh); | |
2650 | continue; | |
a512bbf8 YZ |
2651 | } else { |
2652 | btrfs_set_super_bytenr(sb, bytenr); | |
2653 | ||
2654 | crc = ~(u32)0; | |
2655 | crc = btrfs_csum_data(NULL, (char *)sb + | |
2656 | BTRFS_CSUM_SIZE, crc, | |
2657 | BTRFS_SUPER_INFO_SIZE - | |
2658 | BTRFS_CSUM_SIZE); | |
2659 | btrfs_csum_final(crc, sb->csum); | |
2660 | ||
4eedeb75 HH |
2661 | /* |
2662 | * one reference for us, and we leave it for the | |
2663 | * caller | |
2664 | */ | |
a512bbf8 YZ |
2665 | bh = __getblk(device->bdev, bytenr / 4096, |
2666 | BTRFS_SUPER_INFO_SIZE); | |
2667 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); | |
2668 | ||
4eedeb75 | 2669 | /* one reference for submit_bh */ |
a512bbf8 | 2670 | get_bh(bh); |
4eedeb75 HH |
2671 | |
2672 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
2673 | lock_buffer(bh); |
2674 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
2675 | } | |
2676 | ||
387125fc CM |
2677 | /* |
2678 | * we fua the first super. The others we allow | |
2679 | * to go down lazy. | |
2680 | */ | |
21adbd5c | 2681 | ret = btrfsic_submit_bh(WRITE_FUA, bh); |
4eedeb75 | 2682 | if (ret) |
a512bbf8 | 2683 | errors++; |
a512bbf8 YZ |
2684 | } |
2685 | return errors < i ? 0 : -1; | |
2686 | } | |
2687 | ||
387125fc CM |
2688 | /* |
2689 | * endio for the write_dev_flush, this will wake anyone waiting | |
2690 | * for the barrier when it is done | |
2691 | */ | |
2692 | static void btrfs_end_empty_barrier(struct bio *bio, int err) | |
2693 | { | |
2694 | if (err) { | |
2695 | if (err == -EOPNOTSUPP) | |
2696 | set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); | |
2697 | clear_bit(BIO_UPTODATE, &bio->bi_flags); | |
2698 | } | |
2699 | if (bio->bi_private) | |
2700 | complete(bio->bi_private); | |
2701 | bio_put(bio); | |
2702 | } | |
2703 | ||
2704 | /* | |
2705 | * trigger flushes for one the devices. If you pass wait == 0, the flushes are | |
2706 | * sent down. With wait == 1, it waits for the previous flush. | |
2707 | * | |
2708 | * any device where the flush fails with eopnotsupp are flagged as not-barrier | |
2709 | * capable | |
2710 | */ | |
2711 | static int write_dev_flush(struct btrfs_device *device, int wait) | |
2712 | { | |
2713 | struct bio *bio; | |
2714 | int ret = 0; | |
2715 | ||
2716 | if (device->nobarriers) | |
2717 | return 0; | |
2718 | ||
2719 | if (wait) { | |
2720 | bio = device->flush_bio; | |
2721 | if (!bio) | |
2722 | return 0; | |
2723 | ||
2724 | wait_for_completion(&device->flush_wait); | |
2725 | ||
2726 | if (bio_flagged(bio, BIO_EOPNOTSUPP)) { | |
2727 | printk("btrfs: disabling barriers on dev %s\n", | |
2728 | device->name); | |
2729 | device->nobarriers = 1; | |
2730 | } | |
2731 | if (!bio_flagged(bio, BIO_UPTODATE)) { | |
2732 | ret = -EIO; | |
2733 | } | |
2734 | ||
2735 | /* drop the reference from the wait == 0 run */ | |
2736 | bio_put(bio); | |
2737 | device->flush_bio = NULL; | |
2738 | ||
2739 | return ret; | |
2740 | } | |
2741 | ||
2742 | /* | |
2743 | * one reference for us, and we leave it for the | |
2744 | * caller | |
2745 | */ | |
2746 | device->flush_bio = NULL;; | |
2747 | bio = bio_alloc(GFP_NOFS, 0); | |
2748 | if (!bio) | |
2749 | return -ENOMEM; | |
2750 | ||
2751 | bio->bi_end_io = btrfs_end_empty_barrier; | |
2752 | bio->bi_bdev = device->bdev; | |
2753 | init_completion(&device->flush_wait); | |
2754 | bio->bi_private = &device->flush_wait; | |
2755 | device->flush_bio = bio; | |
2756 | ||
2757 | bio_get(bio); | |
21adbd5c | 2758 | btrfsic_submit_bio(WRITE_FLUSH, bio); |
387125fc CM |
2759 | |
2760 | return 0; | |
2761 | } | |
2762 | ||
2763 | /* | |
2764 | * send an empty flush down to each device in parallel, | |
2765 | * then wait for them | |
2766 | */ | |
2767 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
2768 | { | |
2769 | struct list_head *head; | |
2770 | struct btrfs_device *dev; | |
2771 | int errors = 0; | |
2772 | int ret; | |
2773 | ||
2774 | /* send down all the barriers */ | |
2775 | head = &info->fs_devices->devices; | |
2776 | list_for_each_entry_rcu(dev, head, dev_list) { | |
2777 | if (!dev->bdev) { | |
2778 | errors++; | |
2779 | continue; | |
2780 | } | |
2781 | if (!dev->in_fs_metadata || !dev->writeable) | |
2782 | continue; | |
2783 | ||
2784 | ret = write_dev_flush(dev, 0); | |
2785 | if (ret) | |
2786 | errors++; | |
2787 | } | |
2788 | ||
2789 | /* wait for all the barriers */ | |
2790 | list_for_each_entry_rcu(dev, head, dev_list) { | |
2791 | if (!dev->bdev) { | |
2792 | errors++; | |
2793 | continue; | |
2794 | } | |
2795 | if (!dev->in_fs_metadata || !dev->writeable) | |
2796 | continue; | |
2797 | ||
2798 | ret = write_dev_flush(dev, 1); | |
2799 | if (ret) | |
2800 | errors++; | |
2801 | } | |
2802 | if (errors) | |
2803 | return -EIO; | |
2804 | return 0; | |
2805 | } | |
2806 | ||
a512bbf8 | 2807 | int write_all_supers(struct btrfs_root *root, int max_mirrors) |
f2984462 | 2808 | { |
e5e9a520 | 2809 | struct list_head *head; |
f2984462 | 2810 | struct btrfs_device *dev; |
a061fc8d | 2811 | struct btrfs_super_block *sb; |
f2984462 | 2812 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
2813 | int ret; |
2814 | int do_barriers; | |
a236aed1 CM |
2815 | int max_errors; |
2816 | int total_errors = 0; | |
a061fc8d | 2817 | u64 flags; |
f2984462 | 2818 | |
6c41761f | 2819 | max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; |
f2984462 | 2820 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
af31f5e5 | 2821 | backup_super_roots(root->fs_info); |
f2984462 | 2822 | |
6c41761f | 2823 | sb = root->fs_info->super_for_commit; |
a061fc8d | 2824 | dev_item = &sb->dev_item; |
e5e9a520 | 2825 | |
174ba509 | 2826 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); |
e5e9a520 | 2827 | head = &root->fs_info->fs_devices->devices; |
387125fc CM |
2828 | |
2829 | if (do_barriers) | |
2830 | barrier_all_devices(root->fs_info); | |
2831 | ||
1f78160c | 2832 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
2833 | if (!dev->bdev) { |
2834 | total_errors++; | |
2835 | continue; | |
2836 | } | |
2b82032c | 2837 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2838 | continue; |
2839 | ||
2b82032c | 2840 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
2841 | btrfs_set_stack_device_type(dev_item, dev->type); |
2842 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
2843 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
2844 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
2845 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
2846 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
2847 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
2848 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 2849 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 2850 | |
a061fc8d CM |
2851 | flags = btrfs_super_flags(sb); |
2852 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
2853 | ||
a512bbf8 | 2854 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
2855 | if (ret) |
2856 | total_errors++; | |
f2984462 | 2857 | } |
a236aed1 | 2858 | if (total_errors > max_errors) { |
d397712b CM |
2859 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2860 | total_errors); | |
79787eaa JM |
2861 | |
2862 | /* This shouldn't happen. FUA is masked off if unsupported */ | |
a236aed1 CM |
2863 | BUG(); |
2864 | } | |
f2984462 | 2865 | |
a512bbf8 | 2866 | total_errors = 0; |
1f78160c | 2867 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
2868 | if (!dev->bdev) |
2869 | continue; | |
2b82032c | 2870 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2871 | continue; |
2872 | ||
a512bbf8 YZ |
2873 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
2874 | if (ret) | |
2875 | total_errors++; | |
f2984462 | 2876 | } |
174ba509 | 2877 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 2878 | if (total_errors > max_errors) { |
79787eaa JM |
2879 | btrfs_error(root->fs_info, -EIO, |
2880 | "%d errors while writing supers", total_errors); | |
2881 | return -EIO; | |
a236aed1 | 2882 | } |
f2984462 CM |
2883 | return 0; |
2884 | } | |
2885 | ||
a512bbf8 YZ |
2886 | int write_ctree_super(struct btrfs_trans_handle *trans, |
2887 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 2888 | { |
e66f709b | 2889 | int ret; |
5f39d397 | 2890 | |
a512bbf8 | 2891 | ret = write_all_supers(root, max_mirrors); |
5f39d397 | 2892 | return ret; |
cfaa7295 CM |
2893 | } |
2894 | ||
49b25e05 JM |
2895 | /* Kill all outstanding I/O */ |
2896 | void btrfs_abort_devices(struct btrfs_root *root) | |
2897 | { | |
2898 | struct list_head *head; | |
2899 | struct btrfs_device *dev; | |
2900 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); | |
2901 | head = &root->fs_info->fs_devices->devices; | |
2902 | list_for_each_entry_rcu(dev, head, dev_list) { | |
2903 | blk_abort_queue(dev->bdev->bd_disk->queue); | |
2904 | } | |
2905 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); | |
2906 | } | |
2907 | ||
143bede5 | 2908 | void btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f | 2909 | { |
4df27c4d | 2910 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
2911 | radix_tree_delete(&fs_info->fs_roots_radix, |
2912 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 2913 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
2914 | |
2915 | if (btrfs_root_refs(&root->root_item) == 0) | |
2916 | synchronize_srcu(&fs_info->subvol_srcu); | |
2917 | ||
581bb050 LZ |
2918 | __btrfs_remove_free_space_cache(root->free_ino_pinned); |
2919 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d | 2920 | free_fs_root(root); |
4df27c4d YZ |
2921 | } |
2922 | ||
2923 | static void free_fs_root(struct btrfs_root *root) | |
2924 | { | |
82d5902d | 2925 | iput(root->cache_inode); |
4df27c4d | 2926 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
0ee5dc67 AV |
2927 | if (root->anon_dev) |
2928 | free_anon_bdev(root->anon_dev); | |
4df27c4d YZ |
2929 | free_extent_buffer(root->node); |
2930 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
2931 | kfree(root->free_ino_ctl); |
2932 | kfree(root->free_ino_pinned); | |
d397712b | 2933 | kfree(root->name); |
2619ba1f | 2934 | kfree(root); |
2619ba1f CM |
2935 | } |
2936 | ||
143bede5 | 2937 | static void del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
2938 | { |
2939 | int ret; | |
2940 | struct btrfs_root *gang[8]; | |
2941 | int i; | |
2942 | ||
76dda93c YZ |
2943 | while (!list_empty(&fs_info->dead_roots)) { |
2944 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2945 | struct btrfs_root, root_list); | |
2946 | list_del(&gang[0]->root_list); | |
2947 | ||
2948 | if (gang[0]->in_radix) { | |
2949 | btrfs_free_fs_root(fs_info, gang[0]); | |
2950 | } else { | |
2951 | free_extent_buffer(gang[0]->node); | |
2952 | free_extent_buffer(gang[0]->commit_root); | |
2953 | kfree(gang[0]); | |
2954 | } | |
2955 | } | |
2956 | ||
d397712b | 2957 | while (1) { |
0f7d52f4 CM |
2958 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
2959 | (void **)gang, 0, | |
2960 | ARRAY_SIZE(gang)); | |
2961 | if (!ret) | |
2962 | break; | |
2619ba1f | 2963 | for (i = 0; i < ret; i++) |
5eda7b5e | 2964 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 | 2965 | } |
0f7d52f4 | 2966 | } |
b4100d64 | 2967 | |
c146afad | 2968 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 2969 | { |
c146afad YZ |
2970 | u64 root_objectid = 0; |
2971 | struct btrfs_root *gang[8]; | |
2972 | int i; | |
3768f368 | 2973 | int ret; |
e089f05c | 2974 | |
c146afad YZ |
2975 | while (1) { |
2976 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2977 | (void **)gang, root_objectid, | |
2978 | ARRAY_SIZE(gang)); | |
2979 | if (!ret) | |
2980 | break; | |
5d4f98a2 YZ |
2981 | |
2982 | root_objectid = gang[ret - 1]->root_key.objectid + 1; | |
c146afad | 2983 | for (i = 0; i < ret; i++) { |
66b4ffd1 JB |
2984 | int err; |
2985 | ||
c146afad | 2986 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
2987 | err = btrfs_orphan_cleanup(gang[i]); |
2988 | if (err) | |
2989 | return err; | |
c146afad YZ |
2990 | } |
2991 | root_objectid++; | |
2992 | } | |
2993 | return 0; | |
2994 | } | |
a2135011 | 2995 | |
c146afad YZ |
2996 | int btrfs_commit_super(struct btrfs_root *root) |
2997 | { | |
2998 | struct btrfs_trans_handle *trans; | |
2999 | int ret; | |
a74a4b97 | 3000 | |
c146afad | 3001 | mutex_lock(&root->fs_info->cleaner_mutex); |
24bbcf04 | 3002 | btrfs_run_delayed_iputs(root); |
a74a4b97 | 3003 | btrfs_clean_old_snapshots(root); |
c146afad | 3004 | mutex_unlock(&root->fs_info->cleaner_mutex); |
c71bf099 YZ |
3005 | |
3006 | /* wait until ongoing cleanup work done */ | |
3007 | down_write(&root->fs_info->cleanup_work_sem); | |
3008 | up_write(&root->fs_info->cleanup_work_sem); | |
3009 | ||
7a7eaa40 | 3010 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3011 | if (IS_ERR(trans)) |
3012 | return PTR_ERR(trans); | |
54aa1f4d | 3013 | ret = btrfs_commit_transaction(trans, root); |
79787eaa JM |
3014 | if (ret) |
3015 | return ret; | |
c146afad | 3016 | /* run commit again to drop the original snapshot */ |
7a7eaa40 | 3017 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3018 | if (IS_ERR(trans)) |
3019 | return PTR_ERR(trans); | |
79787eaa JM |
3020 | ret = btrfs_commit_transaction(trans, root); |
3021 | if (ret) | |
3022 | return ret; | |
79154b1b | 3023 | ret = btrfs_write_and_wait_transaction(NULL, root); |
79787eaa JM |
3024 | if (ret) { |
3025 | btrfs_error(root->fs_info, ret, | |
3026 | "Failed to sync btree inode to disk."); | |
3027 | return ret; | |
3028 | } | |
d6bfde87 | 3029 | |
a512bbf8 | 3030 | ret = write_ctree_super(NULL, root, 0); |
c146afad YZ |
3031 | return ret; |
3032 | } | |
3033 | ||
3034 | int close_ctree(struct btrfs_root *root) | |
3035 | { | |
3036 | struct btrfs_fs_info *fs_info = root->fs_info; | |
3037 | int ret; | |
3038 | ||
3039 | fs_info->closing = 1; | |
3040 | smp_mb(); | |
3041 | ||
837d5b6e ID |
3042 | /* pause restriper - we want to resume on mount */ |
3043 | btrfs_pause_balance(root->fs_info); | |
3044 | ||
a2de733c | 3045 | btrfs_scrub_cancel(root); |
4cb5300b CM |
3046 | |
3047 | /* wait for any defraggers to finish */ | |
3048 | wait_event(fs_info->transaction_wait, | |
3049 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3050 | ||
3051 | /* clear out the rbtree of defraggable inodes */ | |
e3029d9f | 3052 | btrfs_run_defrag_inodes(fs_info); |
4cb5300b | 3053 | |
acce952b | 3054 | /* |
3055 | * Here come 2 situations when btrfs is broken to flip readonly: | |
3056 | * | |
3057 | * 1. when btrfs flips readonly somewhere else before | |
3058 | * btrfs_commit_super, sb->s_flags has MS_RDONLY flag, | |
3059 | * and btrfs will skip to write sb directly to keep | |
3060 | * ERROR state on disk. | |
3061 | * | |
3062 | * 2. when btrfs flips readonly just in btrfs_commit_super, | |
ae0e47f0 | 3063 | * and in such case, btrfs cannot write sb via btrfs_commit_super, |
acce952b | 3064 | * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag, |
3065 | * btrfs will cleanup all FS resources first and write sb then. | |
3066 | */ | |
c146afad | 3067 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
acce952b | 3068 | ret = btrfs_commit_super(root); |
3069 | if (ret) | |
3070 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
3071 | } | |
3072 | ||
3073 | if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) { | |
3074 | ret = btrfs_error_commit_super(root); | |
d397712b CM |
3075 | if (ret) |
3076 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
c146afad | 3077 | } |
0f7d52f4 | 3078 | |
300e4f8a JB |
3079 | btrfs_put_block_group_cache(fs_info); |
3080 | ||
e3029d9f AV |
3081 | kthread_stop(fs_info->transaction_kthread); |
3082 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 3083 | |
f25784b3 YZ |
3084 | fs_info->closing = 2; |
3085 | smp_mb(); | |
3086 | ||
b0c68f8b | 3087 | if (fs_info->delalloc_bytes) { |
d397712b | 3088 | printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n", |
21380931 | 3089 | (unsigned long long)fs_info->delalloc_bytes); |
b0c68f8b | 3090 | } |
31153d81 | 3091 | if (fs_info->total_ref_cache_size) { |
d397712b CM |
3092 | printk(KERN_INFO "btrfs: at umount reference cache size %llu\n", |
3093 | (unsigned long long)fs_info->total_ref_cache_size); | |
31153d81 | 3094 | } |
bcc63abb | 3095 | |
5d4f98a2 YZ |
3096 | free_extent_buffer(fs_info->extent_root->node); |
3097 | free_extent_buffer(fs_info->extent_root->commit_root); | |
3098 | free_extent_buffer(fs_info->tree_root->node); | |
3099 | free_extent_buffer(fs_info->tree_root->commit_root); | |
e3029d9f AV |
3100 | free_extent_buffer(fs_info->chunk_root->node); |
3101 | free_extent_buffer(fs_info->chunk_root->commit_root); | |
3102 | free_extent_buffer(fs_info->dev_root->node); | |
3103 | free_extent_buffer(fs_info->dev_root->commit_root); | |
3104 | free_extent_buffer(fs_info->csum_root->node); | |
3105 | free_extent_buffer(fs_info->csum_root->commit_root); | |
d20f7043 | 3106 | |
e3029d9f | 3107 | btrfs_free_block_groups(fs_info); |
d10c5f31 | 3108 | |
c146afad | 3109 | del_fs_roots(fs_info); |
d10c5f31 | 3110 | |
c146afad | 3111 | iput(fs_info->btree_inode); |
9ad6b7bc | 3112 | |
61d92c32 | 3113 | btrfs_stop_workers(&fs_info->generic_worker); |
247e743c | 3114 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 3115 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
3116 | btrfs_stop_workers(&fs_info->workers); |
3117 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 3118 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 3119 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 3120 | btrfs_stop_workers(&fs_info->endio_write_workers); |
0cb59c99 | 3121 | btrfs_stop_workers(&fs_info->endio_freespace_worker); |
1cc127b5 | 3122 | btrfs_stop_workers(&fs_info->submit_workers); |
16cdcec7 | 3123 | btrfs_stop_workers(&fs_info->delayed_workers); |
bab39bf9 | 3124 | btrfs_stop_workers(&fs_info->caching_workers); |
90519d66 | 3125 | btrfs_stop_workers(&fs_info->readahead_workers); |
d6bfde87 | 3126 | |
21adbd5c SB |
3127 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3128 | if (btrfs_test_opt(root, CHECK_INTEGRITY)) | |
3129 | btrfsic_unmount(root, fs_info->fs_devices); | |
3130 | #endif | |
3131 | ||
dfe25020 | 3132 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 3133 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 3134 | |
04160088 | 3135 | bdi_destroy(&fs_info->bdi); |
76dda93c | 3136 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 3137 | |
eb60ceac CM |
3138 | return 0; |
3139 | } | |
3140 | ||
1259ab75 | 3141 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 3142 | { |
1259ab75 | 3143 | int ret; |
727011e0 | 3144 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 3145 | |
0b32f4bb | 3146 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
3147 | if (!ret) |
3148 | return ret; | |
3149 | ||
3150 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
3151 | parent_transid); | |
3152 | return !ret; | |
5f39d397 CM |
3153 | } |
3154 | ||
3155 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 3156 | { |
0b32f4bb | 3157 | return set_extent_buffer_uptodate(buf); |
5f39d397 | 3158 | } |
6702ed49 | 3159 | |
5f39d397 CM |
3160 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
3161 | { | |
727011e0 | 3162 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
5f39d397 | 3163 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 3164 | int was_dirty; |
b4ce94de | 3165 | |
b9447ef8 | 3166 | btrfs_assert_tree_locked(buf); |
ccd467d6 | 3167 | if (transid != root->fs_info->generation) { |
d397712b CM |
3168 | printk(KERN_CRIT "btrfs transid mismatch buffer %llu, " |
3169 | "found %llu running %llu\n", | |
db94535d | 3170 | (unsigned long long)buf->start, |
d397712b CM |
3171 | (unsigned long long)transid, |
3172 | (unsigned long long)root->fs_info->generation); | |
ccd467d6 CM |
3173 | WARN_ON(1); |
3174 | } | |
0b32f4bb | 3175 | was_dirty = set_extent_buffer_dirty(buf); |
b9473439 CM |
3176 | if (!was_dirty) { |
3177 | spin_lock(&root->fs_info->delalloc_lock); | |
3178 | root->fs_info->dirty_metadata_bytes += buf->len; | |
3179 | spin_unlock(&root->fs_info->delalloc_lock); | |
3180 | } | |
eb60ceac CM |
3181 | } |
3182 | ||
d3c2fdcf | 3183 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
16cdcec7 MX |
3184 | { |
3185 | /* | |
3186 | * looks as though older kernels can get into trouble with | |
3187 | * this code, they end up stuck in balance_dirty_pages forever | |
3188 | */ | |
3189 | u64 num_dirty; | |
3190 | unsigned long thresh = 32 * 1024 * 1024; | |
3191 | ||
3192 | if (current->flags & PF_MEMALLOC) | |
3193 | return; | |
3194 | ||
3195 | btrfs_balance_delayed_items(root); | |
3196 | ||
3197 | num_dirty = root->fs_info->dirty_metadata_bytes; | |
3198 | ||
3199 | if (num_dirty > thresh) { | |
3200 | balance_dirty_pages_ratelimited_nr( | |
3201 | root->fs_info->btree_inode->i_mapping, 1); | |
3202 | } | |
3203 | return; | |
3204 | } | |
3205 | ||
3206 | void __btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) | |
35b7e476 | 3207 | { |
188de649 CM |
3208 | /* |
3209 | * looks as though older kernels can get into trouble with | |
3210 | * this code, they end up stuck in balance_dirty_pages forever | |
3211 | */ | |
d6bfde87 | 3212 | u64 num_dirty; |
771ed689 | 3213 | unsigned long thresh = 32 * 1024 * 1024; |
d6bfde87 | 3214 | |
6933c02e | 3215 | if (current->flags & PF_MEMALLOC) |
d6bfde87 CM |
3216 | return; |
3217 | ||
585ad2c3 CM |
3218 | num_dirty = root->fs_info->dirty_metadata_bytes; |
3219 | ||
d6bfde87 CM |
3220 | if (num_dirty > thresh) { |
3221 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 3222 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 3223 | } |
188de649 | 3224 | return; |
35b7e476 | 3225 | } |
6b80053d | 3226 | |
ca7a79ad | 3227 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 3228 | { |
727011e0 | 3229 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
0b32f4bb | 3230 | return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
6b80053d | 3231 | } |
0da5468f | 3232 | |
01d658f2 CM |
3233 | static int btree_lock_page_hook(struct page *page, void *data, |
3234 | void (*flush_fn)(void *)) | |
4bef0848 CM |
3235 | { |
3236 | struct inode *inode = page->mapping->host; | |
b9473439 | 3237 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4bef0848 | 3238 | struct extent_buffer *eb; |
4bef0848 | 3239 | |
4f2de97a JB |
3240 | /* |
3241 | * We culled this eb but the page is still hanging out on the mapping, | |
3242 | * carry on. | |
3243 | */ | |
3244 | if (!PagePrivate(page)) | |
4bef0848 CM |
3245 | goto out; |
3246 | ||
4f2de97a JB |
3247 | eb = (struct extent_buffer *)page->private; |
3248 | if (!eb) { | |
3249 | WARN_ON(1); | |
3250 | goto out; | |
3251 | } | |
3252 | if (page != eb->pages[0]) | |
4bef0848 CM |
3253 | goto out; |
3254 | ||
01d658f2 CM |
3255 | if (!btrfs_try_tree_write_lock(eb)) { |
3256 | flush_fn(data); | |
3257 | btrfs_tree_lock(eb); | |
3258 | } | |
4bef0848 | 3259 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); |
b9473439 CM |
3260 | |
3261 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) { | |
3262 | spin_lock(&root->fs_info->delalloc_lock); | |
3263 | if (root->fs_info->dirty_metadata_bytes >= eb->len) | |
3264 | root->fs_info->dirty_metadata_bytes -= eb->len; | |
3265 | else | |
3266 | WARN_ON(1); | |
3267 | spin_unlock(&root->fs_info->delalloc_lock); | |
3268 | } | |
3269 | ||
4bef0848 | 3270 | btrfs_tree_unlock(eb); |
4bef0848 | 3271 | out: |
01d658f2 CM |
3272 | if (!trylock_page(page)) { |
3273 | flush_fn(data); | |
3274 | lock_page(page); | |
3275 | } | |
4bef0848 CM |
3276 | return 0; |
3277 | } | |
3278 | ||
fcd1f065 | 3279 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 3280 | int read_only) |
3281 | { | |
fcd1f065 DS |
3282 | if (btrfs_super_csum_type(fs_info->super_copy) >= ARRAY_SIZE(btrfs_csum_sizes)) { |
3283 | printk(KERN_ERR "btrfs: unsupported checksum algorithm\n"); | |
3284 | return -EINVAL; | |
3285 | } | |
3286 | ||
acce952b | 3287 | if (read_only) |
fcd1f065 | 3288 | return 0; |
acce952b | 3289 | |
fcd1f065 | 3290 | if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) { |
acce952b | 3291 | printk(KERN_WARNING "warning: mount fs with errors, " |
3292 | "running btrfsck is recommended\n"); | |
fcd1f065 DS |
3293 | } |
3294 | ||
3295 | return 0; | |
acce952b | 3296 | } |
3297 | ||
3298 | int btrfs_error_commit_super(struct btrfs_root *root) | |
3299 | { | |
3300 | int ret; | |
3301 | ||
3302 | mutex_lock(&root->fs_info->cleaner_mutex); | |
3303 | btrfs_run_delayed_iputs(root); | |
3304 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
3305 | ||
3306 | down_write(&root->fs_info->cleanup_work_sem); | |
3307 | up_write(&root->fs_info->cleanup_work_sem); | |
3308 | ||
3309 | /* cleanup FS via transaction */ | |
3310 | btrfs_cleanup_transaction(root); | |
3311 | ||
3312 | ret = write_ctree_super(NULL, root, 0); | |
3313 | ||
3314 | return ret; | |
3315 | } | |
3316 | ||
143bede5 | 3317 | static void btrfs_destroy_ordered_operations(struct btrfs_root *root) |
acce952b | 3318 | { |
3319 | struct btrfs_inode *btrfs_inode; | |
3320 | struct list_head splice; | |
3321 | ||
3322 | INIT_LIST_HEAD(&splice); | |
3323 | ||
3324 | mutex_lock(&root->fs_info->ordered_operations_mutex); | |
3325 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3326 | ||
3327 | list_splice_init(&root->fs_info->ordered_operations, &splice); | |
3328 | while (!list_empty(&splice)) { | |
3329 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3330 | ordered_operations); | |
3331 | ||
3332 | list_del_init(&btrfs_inode->ordered_operations); | |
3333 | ||
3334 | btrfs_invalidate_inodes(btrfs_inode->root); | |
3335 | } | |
3336 | ||
3337 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3338 | mutex_unlock(&root->fs_info->ordered_operations_mutex); | |
acce952b | 3339 | } |
3340 | ||
143bede5 | 3341 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 3342 | { |
3343 | struct list_head splice; | |
3344 | struct btrfs_ordered_extent *ordered; | |
3345 | struct inode *inode; | |
3346 | ||
3347 | INIT_LIST_HEAD(&splice); | |
3348 | ||
3349 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3350 | ||
3351 | list_splice_init(&root->fs_info->ordered_extents, &splice); | |
3352 | while (!list_empty(&splice)) { | |
3353 | ordered = list_entry(splice.next, struct btrfs_ordered_extent, | |
3354 | root_extent_list); | |
3355 | ||
3356 | list_del_init(&ordered->root_extent_list); | |
3357 | atomic_inc(&ordered->refs); | |
3358 | ||
3359 | /* the inode may be getting freed (in sys_unlink path). */ | |
3360 | inode = igrab(ordered->inode); | |
3361 | ||
3362 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3363 | if (inode) | |
3364 | iput(inode); | |
3365 | ||
3366 | atomic_set(&ordered->refs, 1); | |
3367 | btrfs_put_ordered_extent(ordered); | |
3368 | ||
3369 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3370 | } | |
3371 | ||
3372 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
acce952b | 3373 | } |
3374 | ||
79787eaa JM |
3375 | int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
3376 | struct btrfs_root *root) | |
acce952b | 3377 | { |
3378 | struct rb_node *node; | |
3379 | struct btrfs_delayed_ref_root *delayed_refs; | |
3380 | struct btrfs_delayed_ref_node *ref; | |
3381 | int ret = 0; | |
3382 | ||
3383 | delayed_refs = &trans->delayed_refs; | |
3384 | ||
79787eaa | 3385 | again: |
acce952b | 3386 | spin_lock(&delayed_refs->lock); |
3387 | if (delayed_refs->num_entries == 0) { | |
cfece4db | 3388 | spin_unlock(&delayed_refs->lock); |
acce952b | 3389 | printk(KERN_INFO "delayed_refs has NO entry\n"); |
3390 | return ret; | |
3391 | } | |
3392 | ||
3393 | node = rb_first(&delayed_refs->root); | |
3394 | while (node) { | |
3395 | ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); | |
3396 | node = rb_next(node); | |
3397 | ||
3398 | ref->in_tree = 0; | |
3399 | rb_erase(&ref->rb_node, &delayed_refs->root); | |
3400 | delayed_refs->num_entries--; | |
3401 | ||
3402 | atomic_set(&ref->refs, 1); | |
3403 | if (btrfs_delayed_ref_is_head(ref)) { | |
3404 | struct btrfs_delayed_ref_head *head; | |
3405 | ||
3406 | head = btrfs_delayed_node_to_head(ref); | |
79787eaa | 3407 | spin_unlock(&delayed_refs->lock); |
acce952b | 3408 | mutex_lock(&head->mutex); |
3409 | kfree(head->extent_op); | |
3410 | delayed_refs->num_heads--; | |
3411 | if (list_empty(&head->cluster)) | |
3412 | delayed_refs->num_heads_ready--; | |
3413 | list_del_init(&head->cluster); | |
3414 | mutex_unlock(&head->mutex); | |
79787eaa JM |
3415 | btrfs_put_delayed_ref(ref); |
3416 | goto again; | |
acce952b | 3417 | } |
acce952b | 3418 | spin_unlock(&delayed_refs->lock); |
3419 | btrfs_put_delayed_ref(ref); | |
3420 | ||
3421 | cond_resched(); | |
3422 | spin_lock(&delayed_refs->lock); | |
3423 | } | |
3424 | ||
3425 | spin_unlock(&delayed_refs->lock); | |
3426 | ||
3427 | return ret; | |
3428 | } | |
3429 | ||
143bede5 | 3430 | static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t) |
acce952b | 3431 | { |
3432 | struct btrfs_pending_snapshot *snapshot; | |
3433 | struct list_head splice; | |
3434 | ||
3435 | INIT_LIST_HEAD(&splice); | |
3436 | ||
3437 | list_splice_init(&t->pending_snapshots, &splice); | |
3438 | ||
3439 | while (!list_empty(&splice)) { | |
3440 | snapshot = list_entry(splice.next, | |
3441 | struct btrfs_pending_snapshot, | |
3442 | list); | |
3443 | ||
3444 | list_del_init(&snapshot->list); | |
3445 | ||
3446 | kfree(snapshot); | |
3447 | } | |
acce952b | 3448 | } |
3449 | ||
143bede5 | 3450 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 3451 | { |
3452 | struct btrfs_inode *btrfs_inode; | |
3453 | struct list_head splice; | |
3454 | ||
3455 | INIT_LIST_HEAD(&splice); | |
3456 | ||
acce952b | 3457 | spin_lock(&root->fs_info->delalloc_lock); |
5be76758 | 3458 | list_splice_init(&root->fs_info->delalloc_inodes, &splice); |
acce952b | 3459 | |
3460 | while (!list_empty(&splice)) { | |
3461 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3462 | delalloc_inodes); | |
3463 | ||
3464 | list_del_init(&btrfs_inode->delalloc_inodes); | |
3465 | ||
3466 | btrfs_invalidate_inodes(btrfs_inode->root); | |
3467 | } | |
3468 | ||
3469 | spin_unlock(&root->fs_info->delalloc_lock); | |
acce952b | 3470 | } |
3471 | ||
3472 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, | |
3473 | struct extent_io_tree *dirty_pages, | |
3474 | int mark) | |
3475 | { | |
3476 | int ret; | |
3477 | struct page *page; | |
3478 | struct inode *btree_inode = root->fs_info->btree_inode; | |
3479 | struct extent_buffer *eb; | |
3480 | u64 start = 0; | |
3481 | u64 end; | |
3482 | u64 offset; | |
3483 | unsigned long index; | |
3484 | ||
3485 | while (1) { | |
3486 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
3487 | mark); | |
3488 | if (ret) | |
3489 | break; | |
3490 | ||
3491 | clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); | |
3492 | while (start <= end) { | |
3493 | index = start >> PAGE_CACHE_SHIFT; | |
3494 | start = (u64)(index + 1) << PAGE_CACHE_SHIFT; | |
3495 | page = find_get_page(btree_inode->i_mapping, index); | |
3496 | if (!page) | |
3497 | continue; | |
3498 | offset = page_offset(page); | |
3499 | ||
3500 | spin_lock(&dirty_pages->buffer_lock); | |
3501 | eb = radix_tree_lookup( | |
3502 | &(&BTRFS_I(page->mapping->host)->io_tree)->buffer, | |
3503 | offset >> PAGE_CACHE_SHIFT); | |
3504 | spin_unlock(&dirty_pages->buffer_lock); | |
3505 | if (eb) { | |
3506 | ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY, | |
3507 | &eb->bflags); | |
3508 | atomic_set(&eb->refs, 1); | |
3509 | } | |
3510 | if (PageWriteback(page)) | |
3511 | end_page_writeback(page); | |
3512 | ||
3513 | lock_page(page); | |
3514 | if (PageDirty(page)) { | |
3515 | clear_page_dirty_for_io(page); | |
3516 | spin_lock_irq(&page->mapping->tree_lock); | |
3517 | radix_tree_tag_clear(&page->mapping->page_tree, | |
3518 | page_index(page), | |
3519 | PAGECACHE_TAG_DIRTY); | |
3520 | spin_unlock_irq(&page->mapping->tree_lock); | |
3521 | } | |
3522 | ||
3523 | page->mapping->a_ops->invalidatepage(page, 0); | |
3524 | unlock_page(page); | |
3525 | } | |
3526 | } | |
3527 | ||
3528 | return ret; | |
3529 | } | |
3530 | ||
3531 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
3532 | struct extent_io_tree *pinned_extents) | |
3533 | { | |
3534 | struct extent_io_tree *unpin; | |
3535 | u64 start; | |
3536 | u64 end; | |
3537 | int ret; | |
3538 | ||
3539 | unpin = pinned_extents; | |
3540 | while (1) { | |
3541 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
3542 | EXTENT_DIRTY); | |
3543 | if (ret) | |
3544 | break; | |
3545 | ||
3546 | /* opt_discard */ | |
5378e607 LD |
3547 | if (btrfs_test_opt(root, DISCARD)) |
3548 | ret = btrfs_error_discard_extent(root, start, | |
3549 | end + 1 - start, | |
3550 | NULL); | |
acce952b | 3551 | |
3552 | clear_extent_dirty(unpin, start, end, GFP_NOFS); | |
3553 | btrfs_error_unpin_extent_range(root, start, end); | |
3554 | cond_resched(); | |
3555 | } | |
3556 | ||
3557 | return 0; | |
3558 | } | |
3559 | ||
49b25e05 JM |
3560 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
3561 | struct btrfs_root *root) | |
3562 | { | |
3563 | btrfs_destroy_delayed_refs(cur_trans, root); | |
3564 | btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv, | |
3565 | cur_trans->dirty_pages.dirty_bytes); | |
3566 | ||
3567 | /* FIXME: cleanup wait for commit */ | |
3568 | cur_trans->in_commit = 1; | |
3569 | cur_trans->blocked = 1; | |
3570 | if (waitqueue_active(&root->fs_info->transaction_blocked_wait)) | |
3571 | wake_up(&root->fs_info->transaction_blocked_wait); | |
3572 | ||
3573 | cur_trans->blocked = 0; | |
3574 | if (waitqueue_active(&root->fs_info->transaction_wait)) | |
3575 | wake_up(&root->fs_info->transaction_wait); | |
3576 | ||
3577 | cur_trans->commit_done = 1; | |
3578 | if (waitqueue_active(&cur_trans->commit_wait)) | |
3579 | wake_up(&cur_trans->commit_wait); | |
3580 | ||
3581 | btrfs_destroy_pending_snapshots(cur_trans); | |
3582 | ||
3583 | btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, | |
3584 | EXTENT_DIRTY); | |
3585 | ||
3586 | /* | |
3587 | memset(cur_trans, 0, sizeof(*cur_trans)); | |
3588 | kmem_cache_free(btrfs_transaction_cachep, cur_trans); | |
3589 | */ | |
3590 | } | |
3591 | ||
3592 | int btrfs_cleanup_transaction(struct btrfs_root *root) | |
acce952b | 3593 | { |
3594 | struct btrfs_transaction *t; | |
3595 | LIST_HEAD(list); | |
3596 | ||
acce952b | 3597 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
3598 | ||
a4abeea4 | 3599 | spin_lock(&root->fs_info->trans_lock); |
acce952b | 3600 | list_splice_init(&root->fs_info->trans_list, &list); |
a4abeea4 JB |
3601 | root->fs_info->trans_no_join = 1; |
3602 | spin_unlock(&root->fs_info->trans_lock); | |
3603 | ||
acce952b | 3604 | while (!list_empty(&list)) { |
3605 | t = list_entry(list.next, struct btrfs_transaction, list); | |
3606 | if (!t) | |
3607 | break; | |
3608 | ||
3609 | btrfs_destroy_ordered_operations(root); | |
3610 | ||
3611 | btrfs_destroy_ordered_extents(root); | |
3612 | ||
3613 | btrfs_destroy_delayed_refs(t, root); | |
3614 | ||
3615 | btrfs_block_rsv_release(root, | |
3616 | &root->fs_info->trans_block_rsv, | |
3617 | t->dirty_pages.dirty_bytes); | |
3618 | ||
3619 | /* FIXME: cleanup wait for commit */ | |
3620 | t->in_commit = 1; | |
3621 | t->blocked = 1; | |
3622 | if (waitqueue_active(&root->fs_info->transaction_blocked_wait)) | |
3623 | wake_up(&root->fs_info->transaction_blocked_wait); | |
3624 | ||
3625 | t->blocked = 0; | |
3626 | if (waitqueue_active(&root->fs_info->transaction_wait)) | |
3627 | wake_up(&root->fs_info->transaction_wait); | |
acce952b | 3628 | |
acce952b | 3629 | t->commit_done = 1; |
3630 | if (waitqueue_active(&t->commit_wait)) | |
3631 | wake_up(&t->commit_wait); | |
acce952b | 3632 | |
3633 | btrfs_destroy_pending_snapshots(t); | |
3634 | ||
3635 | btrfs_destroy_delalloc_inodes(root); | |
3636 | ||
a4abeea4 | 3637 | spin_lock(&root->fs_info->trans_lock); |
acce952b | 3638 | root->fs_info->running_transaction = NULL; |
a4abeea4 | 3639 | spin_unlock(&root->fs_info->trans_lock); |
acce952b | 3640 | |
3641 | btrfs_destroy_marked_extents(root, &t->dirty_pages, | |
3642 | EXTENT_DIRTY); | |
3643 | ||
3644 | btrfs_destroy_pinned_extent(root, | |
3645 | root->fs_info->pinned_extents); | |
3646 | ||
13c5a93e | 3647 | atomic_set(&t->use_count, 0); |
acce952b | 3648 | list_del_init(&t->list); |
3649 | memset(t, 0, sizeof(*t)); | |
3650 | kmem_cache_free(btrfs_transaction_cachep, t); | |
3651 | } | |
3652 | ||
a4abeea4 JB |
3653 | spin_lock(&root->fs_info->trans_lock); |
3654 | root->fs_info->trans_no_join = 0; | |
3655 | spin_unlock(&root->fs_info->trans_lock); | |
acce952b | 3656 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); |
acce952b | 3657 | |
3658 | return 0; | |
3659 | } | |
3660 | ||
79787eaa JM |
3661 | static int btree_writepage_io_failed_hook(struct bio *bio, struct page *page, |
3662 | u64 start, u64 end, | |
3663 | struct extent_state *state) | |
3664 | { | |
3665 | struct super_block *sb = page->mapping->host->i_sb; | |
3666 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
3667 | btrfs_error(fs_info, -EIO, | |
3668 | "Error occured while writing out btree at %llu", start); | |
3669 | return -EIO; | |
3670 | } | |
3671 | ||
d1310b2e | 3672 | static struct extent_io_ops btree_extent_io_ops = { |
4bef0848 | 3673 | .write_cache_pages_lock_hook = btree_lock_page_hook, |
ce9adaa5 | 3674 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
4bb31e92 | 3675 | .readpage_io_failed_hook = btree_io_failed_hook, |
0b86a832 | 3676 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
3677 | /* note we're sharing with inode.c for the merge bio hook */ |
3678 | .merge_bio_hook = btrfs_merge_bio_hook, | |
79787eaa | 3679 | .writepage_io_failed_hook = btree_writepage_io_failed_hook, |
0da5468f | 3680 | }; |