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); |
acce952b | 51 | static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
52 | int read_only); | |
53 | static int btrfs_destroy_ordered_operations(struct btrfs_root *root); | |
54 | static int btrfs_destroy_ordered_extents(struct btrfs_root *root); | |
55 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, | |
56 | struct btrfs_root *root); | |
57 | static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t); | |
58 | static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root); | |
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); | |
64 | static int btrfs_cleanup_transaction(struct btrfs_root *root); | |
ce9adaa5 | 65 | |
d352ac68 CM |
66 | /* |
67 | * end_io_wq structs are used to do processing in task context when an IO is | |
68 | * complete. This is used during reads to verify checksums, and it is used | |
69 | * by writes to insert metadata for new file extents after IO is complete. | |
70 | */ | |
ce9adaa5 CM |
71 | struct end_io_wq { |
72 | struct bio *bio; | |
73 | bio_end_io_t *end_io; | |
74 | void *private; | |
75 | struct btrfs_fs_info *info; | |
76 | int error; | |
22c59948 | 77 | int metadata; |
ce9adaa5 | 78 | struct list_head list; |
8b712842 | 79 | struct btrfs_work work; |
ce9adaa5 | 80 | }; |
0da5468f | 81 | |
d352ac68 CM |
82 | /* |
83 | * async submit bios are used to offload expensive checksumming | |
84 | * onto the worker threads. They checksum file and metadata bios | |
85 | * just before they are sent down the IO stack. | |
86 | */ | |
44b8bd7e CM |
87 | struct async_submit_bio { |
88 | struct inode *inode; | |
89 | struct bio *bio; | |
90 | struct list_head list; | |
4a69a410 CM |
91 | extent_submit_bio_hook_t *submit_bio_start; |
92 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
93 | int rw; |
94 | int mirror_num; | |
c8b97818 | 95 | unsigned long bio_flags; |
eaf25d93 CM |
96 | /* |
97 | * bio_offset is optional, can be used if the pages in the bio | |
98 | * can't tell us where in the file the bio should go | |
99 | */ | |
100 | u64 bio_offset; | |
8b712842 | 101 | struct btrfs_work work; |
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 JB |
334 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
335 | 0, &cached_state, GFP_NOFS); | |
336 | if (extent_buffer_uptodate(io_tree, eb, cached_state) && | |
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; |
2ac55d41 | 347 | clear_extent_buffer_uptodate(io_tree, eb, &cached_state); |
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; | |
363 | int ret; | |
364 | int num_copies = 0; | |
365 | int mirror_num = 0; | |
366 | ||
a826d6dc | 367 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
f188591e CM |
368 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
369 | while (1) { | |
bb82ab88 AJ |
370 | ret = read_extent_buffer_pages(io_tree, eb, start, |
371 | WAIT_COMPLETE, | |
f188591e | 372 | btree_get_extent, mirror_num); |
727011e0 | 373 | if (!ret && !verify_parent_transid(io_tree, eb, parent_transid)) |
f188591e | 374 | return ret; |
d397712b | 375 | |
a826d6dc JB |
376 | /* |
377 | * This buffer's crc is fine, but its contents are corrupted, so | |
378 | * there is no reason to read the other copies, they won't be | |
379 | * any less wrong. | |
380 | */ | |
381 | if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) | |
382 | return ret; | |
383 | ||
f188591e CM |
384 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
385 | eb->start, eb->len); | |
4235298e | 386 | if (num_copies == 1) |
f188591e | 387 | return ret; |
4235298e | 388 | |
f188591e | 389 | mirror_num++; |
4235298e | 390 | if (mirror_num > num_copies) |
f188591e | 391 | return ret; |
f188591e | 392 | } |
f188591e CM |
393 | return -EIO; |
394 | } | |
19c00ddc | 395 | |
d352ac68 | 396 | /* |
d397712b CM |
397 | * checksum a dirty tree block before IO. This has extra checks to make sure |
398 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 399 | */ |
d397712b | 400 | |
b2950863 | 401 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 402 | { |
d1310b2e | 403 | struct extent_io_tree *tree; |
35ebb934 | 404 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc | 405 | u64 found_start; |
19c00ddc CM |
406 | unsigned long len; |
407 | struct extent_buffer *eb; | |
f188591e | 408 | |
d1310b2e | 409 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc | 410 | |
727011e0 | 411 | if (page->private == EXTENT_PAGE_PRIVATE) |
19c00ddc | 412 | goto out; |
eb14ab8e CM |
413 | if (!page->private) { |
414 | WARN_ON(1); | |
19c00ddc | 415 | goto out; |
eb14ab8e | 416 | } |
19c00ddc | 417 | len = page->private >> 2; |
d397712b CM |
418 | WARN_ON(len == 0); |
419 | ||
727011e0 | 420 | eb = find_extent_buffer(tree, start, len); |
784b4e29 | 421 | |
19c00ddc CM |
422 | found_start = btrfs_header_bytenr(eb); |
423 | if (found_start != start) { | |
55c69072 CM |
424 | WARN_ON(1); |
425 | goto err; | |
426 | } | |
727011e0 | 427 | if (eb->pages[0] != page) { |
55c69072 CM |
428 | WARN_ON(1); |
429 | goto err; | |
430 | } | |
431 | if (!PageUptodate(page)) { | |
55c69072 CM |
432 | WARN_ON(1); |
433 | goto err; | |
19c00ddc | 434 | } |
19c00ddc | 435 | csum_tree_block(root, eb, 0); |
55c69072 | 436 | err: |
19c00ddc CM |
437 | free_extent_buffer(eb); |
438 | out: | |
439 | return 0; | |
440 | } | |
441 | ||
2b82032c YZ |
442 | static int check_tree_block_fsid(struct btrfs_root *root, |
443 | struct extent_buffer *eb) | |
444 | { | |
445 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
446 | u8 fsid[BTRFS_UUID_SIZE]; | |
447 | int ret = 1; | |
448 | ||
449 | read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), | |
450 | BTRFS_FSID_SIZE); | |
451 | while (fs_devices) { | |
452 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
453 | ret = 0; | |
454 | break; | |
455 | } | |
456 | fs_devices = fs_devices->seed; | |
457 | } | |
458 | return ret; | |
459 | } | |
460 | ||
a826d6dc JB |
461 | #define CORRUPT(reason, eb, root, slot) \ |
462 | printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \ | |
463 | "root=%llu, slot=%d\n", reason, \ | |
464 | (unsigned long long)btrfs_header_bytenr(eb), \ | |
465 | (unsigned long long)root->objectid, slot) | |
466 | ||
467 | static noinline int check_leaf(struct btrfs_root *root, | |
468 | struct extent_buffer *leaf) | |
469 | { | |
470 | struct btrfs_key key; | |
471 | struct btrfs_key leaf_key; | |
472 | u32 nritems = btrfs_header_nritems(leaf); | |
473 | int slot; | |
474 | ||
475 | if (nritems == 0) | |
476 | return 0; | |
477 | ||
478 | /* Check the 0 item */ | |
479 | if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != | |
480 | BTRFS_LEAF_DATA_SIZE(root)) { | |
481 | CORRUPT("invalid item offset size pair", leaf, root, 0); | |
482 | return -EIO; | |
483 | } | |
484 | ||
485 | /* | |
486 | * Check to make sure each items keys are in the correct order and their | |
487 | * offsets make sense. We only have to loop through nritems-1 because | |
488 | * we check the current slot against the next slot, which verifies the | |
489 | * next slot's offset+size makes sense and that the current's slot | |
490 | * offset is correct. | |
491 | */ | |
492 | for (slot = 0; slot < nritems - 1; slot++) { | |
493 | btrfs_item_key_to_cpu(leaf, &leaf_key, slot); | |
494 | btrfs_item_key_to_cpu(leaf, &key, slot + 1); | |
495 | ||
496 | /* Make sure the keys are in the right order */ | |
497 | if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { | |
498 | CORRUPT("bad key order", leaf, root, slot); | |
499 | return -EIO; | |
500 | } | |
501 | ||
502 | /* | |
503 | * Make sure the offset and ends are right, remember that the | |
504 | * item data starts at the end of the leaf and grows towards the | |
505 | * front. | |
506 | */ | |
507 | if (btrfs_item_offset_nr(leaf, slot) != | |
508 | btrfs_item_end_nr(leaf, slot + 1)) { | |
509 | CORRUPT("slot offset bad", leaf, root, slot); | |
510 | return -EIO; | |
511 | } | |
512 | ||
513 | /* | |
514 | * Check to make sure that we don't point outside of the leaf, | |
515 | * just incase all the items are consistent to eachother, but | |
516 | * all point outside of the leaf. | |
517 | */ | |
518 | if (btrfs_item_end_nr(leaf, slot) > | |
519 | BTRFS_LEAF_DATA_SIZE(root)) { | |
520 | CORRUPT("slot end outside of leaf", leaf, root, slot); | |
521 | return -EIO; | |
522 | } | |
523 | } | |
524 | ||
525 | return 0; | |
526 | } | |
527 | ||
727011e0 CM |
528 | struct extent_buffer *find_eb_for_page(struct extent_io_tree *tree, |
529 | struct page *page, int max_walk) | |
530 | { | |
531 | struct extent_buffer *eb; | |
532 | u64 start = page_offset(page); | |
533 | u64 target = start; | |
534 | u64 min_start; | |
535 | ||
536 | if (start < max_walk) | |
537 | min_start = 0; | |
538 | else | |
539 | min_start = start - max_walk; | |
540 | ||
541 | while (start >= min_start) { | |
542 | eb = find_extent_buffer(tree, start, 0); | |
543 | if (eb) { | |
544 | /* | |
545 | * we found an extent buffer and it contains our page | |
546 | * horray! | |
547 | */ | |
548 | if (eb->start <= target && | |
549 | eb->start + eb->len > target) | |
550 | return eb; | |
551 | ||
552 | /* we found an extent buffer that wasn't for us */ | |
553 | free_extent_buffer(eb); | |
554 | return NULL; | |
555 | } | |
556 | if (start == 0) | |
557 | break; | |
558 | start -= PAGE_CACHE_SIZE; | |
559 | } | |
560 | return NULL; | |
561 | } | |
562 | ||
b2950863 | 563 | static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
ce9adaa5 CM |
564 | struct extent_state *state) |
565 | { | |
566 | struct extent_io_tree *tree; | |
567 | u64 found_start; | |
568 | int found_level; | |
569 | unsigned long len; | |
570 | struct extent_buffer *eb; | |
571 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 572 | int ret = 0; |
727011e0 | 573 | int reads_done; |
ce9adaa5 | 574 | |
ce9adaa5 CM |
575 | if (!page->private) |
576 | goto out; | |
d397712b | 577 | |
727011e0 | 578 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
ce9adaa5 | 579 | len = page->private >> 2; |
d397712b | 580 | |
727011e0 CM |
581 | eb = find_eb_for_page(tree, page, max(root->leafsize, root->nodesize)); |
582 | if (!eb) { | |
91ca338d TI |
583 | ret = -EIO; |
584 | goto out; | |
585 | } | |
727011e0 CM |
586 | reads_done = atomic_dec_and_test(&eb->pages_reading); |
587 | if (!reads_done) | |
588 | goto err; | |
f188591e | 589 | |
ce9adaa5 | 590 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 591 | if (found_start != eb->start) { |
7a36ddec | 592 | printk_ratelimited(KERN_INFO "btrfs bad tree block start " |
193f284d CM |
593 | "%llu %llu\n", |
594 | (unsigned long long)found_start, | |
595 | (unsigned long long)eb->start); | |
f188591e | 596 | ret = -EIO; |
ce9adaa5 CM |
597 | goto err; |
598 | } | |
2b82032c | 599 | if (check_tree_block_fsid(root, eb)) { |
7a36ddec | 600 | printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n", |
193f284d | 601 | (unsigned long long)eb->start); |
1259ab75 CM |
602 | ret = -EIO; |
603 | goto err; | |
604 | } | |
ce9adaa5 CM |
605 | found_level = btrfs_header_level(eb); |
606 | ||
85d4e461 CM |
607 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
608 | eb, found_level); | |
4008c04a | 609 | |
ce9adaa5 | 610 | ret = csum_tree_block(root, eb, 1); |
a826d6dc | 611 | if (ret) { |
f188591e | 612 | ret = -EIO; |
a826d6dc JB |
613 | goto err; |
614 | } | |
615 | ||
616 | /* | |
617 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
618 | * that we don't try and read the other copies of this block, just | |
619 | * return -EIO. | |
620 | */ | |
621 | if (found_level == 0 && check_leaf(root, eb)) { | |
622 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); | |
623 | ret = -EIO; | |
624 | } | |
ce9adaa5 | 625 | |
ce9adaa5 | 626 | err: |
4bb31e92 AJ |
627 | if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) { |
628 | clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags); | |
629 | btree_readahead_hook(root, eb, eb->start, ret); | |
630 | } | |
631 | ||
727011e0 CM |
632 | if (ret && eb) |
633 | clear_extent_buffer_uptodate(tree, eb, NULL); | |
ce9adaa5 CM |
634 | free_extent_buffer(eb); |
635 | out: | |
f188591e | 636 | return ret; |
ce9adaa5 CM |
637 | } |
638 | ||
4bb31e92 AJ |
639 | static int btree_io_failed_hook(struct bio *failed_bio, |
640 | struct page *page, u64 start, u64 end, | |
32240a91 | 641 | int mirror_num, struct extent_state *state) |
4bb31e92 AJ |
642 | { |
643 | struct extent_io_tree *tree; | |
644 | unsigned long len; | |
645 | struct extent_buffer *eb; | |
646 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
647 | ||
648 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
649 | if (page->private == EXTENT_PAGE_PRIVATE) | |
650 | goto out; | |
651 | if (!page->private) | |
652 | goto out; | |
653 | ||
654 | len = page->private >> 2; | |
655 | WARN_ON(len == 0); | |
656 | ||
727011e0 | 657 | eb = alloc_extent_buffer(tree, start, len); |
4bb31e92 AJ |
658 | if (eb == NULL) |
659 | goto out; | |
660 | ||
661 | if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) { | |
662 | clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags); | |
663 | btree_readahead_hook(root, eb, eb->start, -EIO); | |
664 | } | |
c674e04e | 665 | free_extent_buffer(eb); |
4bb31e92 AJ |
666 | |
667 | out: | |
668 | return -EIO; /* we fixed nothing */ | |
669 | } | |
670 | ||
ce9adaa5 | 671 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
672 | { |
673 | struct end_io_wq *end_io_wq = bio->bi_private; | |
674 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 675 | |
ce9adaa5 | 676 | fs_info = end_io_wq->info; |
ce9adaa5 | 677 | end_io_wq->error = err; |
8b712842 CM |
678 | end_io_wq->work.func = end_workqueue_fn; |
679 | end_io_wq->work.flags = 0; | |
d20f7043 | 680 | |
7b6d91da | 681 | if (bio->bi_rw & REQ_WRITE) { |
0cb59c99 | 682 | if (end_io_wq->metadata == 1) |
cad321ad CM |
683 | btrfs_queue_worker(&fs_info->endio_meta_write_workers, |
684 | &end_io_wq->work); | |
0cb59c99 JB |
685 | else if (end_io_wq->metadata == 2) |
686 | btrfs_queue_worker(&fs_info->endio_freespace_worker, | |
687 | &end_io_wq->work); | |
cad321ad CM |
688 | else |
689 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
690 | &end_io_wq->work); | |
d20f7043 CM |
691 | } else { |
692 | if (end_io_wq->metadata) | |
693 | btrfs_queue_worker(&fs_info->endio_meta_workers, | |
694 | &end_io_wq->work); | |
695 | else | |
696 | btrfs_queue_worker(&fs_info->endio_workers, | |
697 | &end_io_wq->work); | |
698 | } | |
ce9adaa5 CM |
699 | } |
700 | ||
0cb59c99 JB |
701 | /* |
702 | * For the metadata arg you want | |
703 | * | |
704 | * 0 - if data | |
705 | * 1 - if normal metadta | |
706 | * 2 - if writing to the free space cache area | |
707 | */ | |
22c59948 CM |
708 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
709 | int metadata) | |
0b86a832 | 710 | { |
ce9adaa5 | 711 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
712 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
713 | if (!end_io_wq) | |
714 | return -ENOMEM; | |
715 | ||
716 | end_io_wq->private = bio->bi_private; | |
717 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 718 | end_io_wq->info = info; |
ce9adaa5 CM |
719 | end_io_wq->error = 0; |
720 | end_io_wq->bio = bio; | |
22c59948 | 721 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
722 | |
723 | bio->bi_private = end_io_wq; | |
724 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
725 | return 0; |
726 | } | |
727 | ||
b64a2851 | 728 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 729 | { |
4854ddd0 CM |
730 | unsigned long limit = min_t(unsigned long, |
731 | info->workers.max_workers, | |
732 | info->fs_devices->open_devices); | |
733 | return 256 * limit; | |
734 | } | |
0986fe9e | 735 | |
4a69a410 CM |
736 | static void run_one_async_start(struct btrfs_work *work) |
737 | { | |
4a69a410 CM |
738 | struct async_submit_bio *async; |
739 | ||
740 | async = container_of(work, struct async_submit_bio, work); | |
4a69a410 | 741 | async->submit_bio_start(async->inode, async->rw, async->bio, |
eaf25d93 CM |
742 | async->mirror_num, async->bio_flags, |
743 | async->bio_offset); | |
4a69a410 CM |
744 | } |
745 | ||
746 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
747 | { |
748 | struct btrfs_fs_info *fs_info; | |
749 | struct async_submit_bio *async; | |
4854ddd0 | 750 | int limit; |
8b712842 CM |
751 | |
752 | async = container_of(work, struct async_submit_bio, work); | |
753 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 754 | |
b64a2851 | 755 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
756 | limit = limit * 2 / 3; |
757 | ||
8b712842 | 758 | atomic_dec(&fs_info->nr_async_submits); |
0986fe9e | 759 | |
b64a2851 CM |
760 | if (atomic_read(&fs_info->nr_async_submits) < limit && |
761 | waitqueue_active(&fs_info->async_submit_wait)) | |
4854ddd0 CM |
762 | wake_up(&fs_info->async_submit_wait); |
763 | ||
4a69a410 | 764 | async->submit_bio_done(async->inode, async->rw, async->bio, |
eaf25d93 CM |
765 | async->mirror_num, async->bio_flags, |
766 | async->bio_offset); | |
4a69a410 CM |
767 | } |
768 | ||
769 | static void run_one_async_free(struct btrfs_work *work) | |
770 | { | |
771 | struct async_submit_bio *async; | |
772 | ||
773 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
774 | kfree(async); |
775 | } | |
776 | ||
44b8bd7e CM |
777 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
778 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 779 | unsigned long bio_flags, |
eaf25d93 | 780 | u64 bio_offset, |
4a69a410 CM |
781 | extent_submit_bio_hook_t *submit_bio_start, |
782 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
783 | { |
784 | struct async_submit_bio *async; | |
785 | ||
786 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
787 | if (!async) | |
788 | return -ENOMEM; | |
789 | ||
790 | async->inode = inode; | |
791 | async->rw = rw; | |
792 | async->bio = bio; | |
793 | async->mirror_num = mirror_num; | |
4a69a410 CM |
794 | async->submit_bio_start = submit_bio_start; |
795 | async->submit_bio_done = submit_bio_done; | |
796 | ||
797 | async->work.func = run_one_async_start; | |
798 | async->work.ordered_func = run_one_async_done; | |
799 | async->work.ordered_free = run_one_async_free; | |
800 | ||
8b712842 | 801 | async->work.flags = 0; |
c8b97818 | 802 | async->bio_flags = bio_flags; |
eaf25d93 | 803 | async->bio_offset = bio_offset; |
8c8bee1d | 804 | |
cb03c743 | 805 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 | 806 | |
7b6d91da | 807 | if (rw & REQ_SYNC) |
d313d7a3 CM |
808 | btrfs_set_work_high_prio(&async->work); |
809 | ||
8b712842 | 810 | btrfs_queue_worker(&fs_info->workers, &async->work); |
9473f16c | 811 | |
d397712b | 812 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
813 | atomic_read(&fs_info->nr_async_submits)) { |
814 | wait_event(fs_info->async_submit_wait, | |
815 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
816 | } | |
817 | ||
44b8bd7e CM |
818 | return 0; |
819 | } | |
820 | ||
ce3ed71a CM |
821 | static int btree_csum_one_bio(struct bio *bio) |
822 | { | |
823 | struct bio_vec *bvec = bio->bi_io_vec; | |
824 | int bio_index = 0; | |
825 | struct btrfs_root *root; | |
826 | ||
827 | WARN_ON(bio->bi_vcnt <= 0); | |
d397712b | 828 | while (bio_index < bio->bi_vcnt) { |
ce3ed71a CM |
829 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
830 | csum_dirty_buffer(root, bvec->bv_page); | |
831 | bio_index++; | |
832 | bvec++; | |
833 | } | |
834 | return 0; | |
835 | } | |
836 | ||
4a69a410 CM |
837 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
838 | struct bio *bio, int mirror_num, | |
eaf25d93 CM |
839 | unsigned long bio_flags, |
840 | u64 bio_offset) | |
22c59948 | 841 | { |
8b712842 CM |
842 | /* |
843 | * when we're called for a write, we're already in the async | |
5443be45 | 844 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 845 | */ |
4a69a410 CM |
846 | btree_csum_one_bio(bio); |
847 | return 0; | |
848 | } | |
22c59948 | 849 | |
4a69a410 | 850 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
851 | int mirror_num, unsigned long bio_flags, |
852 | u64 bio_offset) | |
4a69a410 | 853 | { |
8b712842 | 854 | /* |
4a69a410 CM |
855 | * when we're called for a write, we're already in the async |
856 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 857 | */ |
8b712842 | 858 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
859 | } |
860 | ||
44b8bd7e | 861 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
862 | int mirror_num, unsigned long bio_flags, |
863 | u64 bio_offset) | |
44b8bd7e | 864 | { |
cad321ad CM |
865 | int ret; |
866 | ||
867 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, | |
868 | bio, 1); | |
869 | BUG_ON(ret); | |
870 | ||
7b6d91da | 871 | if (!(rw & REQ_WRITE)) { |
4a69a410 CM |
872 | /* |
873 | * called for a read, do the setup so that checksum validation | |
874 | * can happen in the async kernel threads | |
875 | */ | |
4a69a410 | 876 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, |
6f3577bd | 877 | mirror_num, 0); |
44b8bd7e | 878 | } |
d313d7a3 | 879 | |
cad321ad CM |
880 | /* |
881 | * kthread helpers are used to submit writes so that checksumming | |
882 | * can happen in parallel across all CPUs | |
883 | */ | |
44b8bd7e | 884 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, |
c8b97818 | 885 | inode, rw, bio, mirror_num, 0, |
eaf25d93 | 886 | bio_offset, |
4a69a410 CM |
887 | __btree_submit_bio_start, |
888 | __btree_submit_bio_done); | |
44b8bd7e CM |
889 | } |
890 | ||
3dd1462e | 891 | #ifdef CONFIG_MIGRATION |
784b4e29 | 892 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
893 | struct page *newpage, struct page *page, |
894 | enum migrate_mode mode) | |
784b4e29 CM |
895 | { |
896 | /* | |
897 | * we can't safely write a btree page from here, | |
898 | * we haven't done the locking hook | |
899 | */ | |
900 | if (PageDirty(page)) | |
901 | return -EAGAIN; | |
902 | /* | |
903 | * Buffers may be managed in a filesystem specific way. | |
904 | * We must have no buffers or drop them. | |
905 | */ | |
906 | if (page_has_private(page) && | |
907 | !try_to_release_page(page, GFP_KERNEL)) | |
908 | return -EAGAIN; | |
a6bc32b8 | 909 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 910 | } |
3dd1462e | 911 | #endif |
784b4e29 | 912 | |
0da5468f CM |
913 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
914 | { | |
d1310b2e CM |
915 | struct extent_io_tree *tree; |
916 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
727011e0 | 917 | |
b9473439 CM |
918 | if (!(current->flags & PF_MEMALLOC)) { |
919 | return extent_write_full_page(tree, page, | |
920 | btree_get_extent, wbc); | |
921 | } | |
5443be45 | 922 | |
b9473439 | 923 | redirty_page_for_writepage(wbc, page); |
b9473439 CM |
924 | unlock_page(page); |
925 | return 0; | |
5f39d397 | 926 | } |
0da5468f CM |
927 | |
928 | static int btree_writepages(struct address_space *mapping, | |
929 | struct writeback_control *wbc) | |
930 | { | |
d1310b2e CM |
931 | struct extent_io_tree *tree; |
932 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 933 | if (wbc->sync_mode == WB_SYNC_NONE) { |
b9473439 | 934 | struct btrfs_root *root = BTRFS_I(mapping->host)->root; |
793955bc | 935 | u64 num_dirty; |
24ab9cd8 | 936 | unsigned long thresh = 32 * 1024 * 1024; |
448d640b CM |
937 | |
938 | if (wbc->for_kupdate) | |
939 | return 0; | |
940 | ||
b9473439 CM |
941 | /* this is a bit racy, but that's ok */ |
942 | num_dirty = root->fs_info->dirty_metadata_bytes; | |
d397712b | 943 | if (num_dirty < thresh) |
793955bc | 944 | return 0; |
793955bc | 945 | } |
0da5468f CM |
946 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
947 | } | |
948 | ||
b2950863 | 949 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 950 | { |
d1310b2e CM |
951 | struct extent_io_tree *tree; |
952 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 953 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 954 | } |
22b0ebda | 955 | |
70dec807 | 956 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 957 | { |
d1310b2e CM |
958 | struct extent_io_tree *tree; |
959 | struct extent_map_tree *map; | |
727011e0 CM |
960 | struct extent_buffer *eb; |
961 | struct btrfs_root *root; | |
5f39d397 | 962 | int ret; |
d98237b3 | 963 | |
98509cfc | 964 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 965 | return 0; |
98509cfc | 966 | |
d1310b2e CM |
967 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
968 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
6af118ce | 969 | |
727011e0 CM |
970 | root = BTRFS_I(page->mapping->host)->root; |
971 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
972 | eb = find_eb_for_page(tree, page, max(root->leafsize, root->nodesize)); | |
973 | free_extent_buffer(eb); | |
974 | if (eb) | |
975 | return 0; | |
976 | } | |
0c4e538b DS |
977 | /* |
978 | * We need to mask out eg. __GFP_HIGHMEM and __GFP_DMA32 as we're doing | |
979 | * slab allocation from alloc_extent_state down the callchain where | |
980 | * it'd hit a BUG_ON as those flags are not allowed. | |
981 | */ | |
982 | gfp_flags &= ~GFP_SLAB_BUG_MASK; | |
983 | ||
7b13b7b1 | 984 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
d397712b | 985 | if (!ret) |
6af118ce | 986 | return 0; |
6af118ce CM |
987 | |
988 | ret = try_release_extent_buffer(tree, page); | |
5f39d397 CM |
989 | if (ret == 1) { |
990 | ClearPagePrivate(page); | |
991 | set_page_private(page, 0); | |
992 | page_cache_release(page); | |
993 | } | |
6af118ce | 994 | |
d98237b3 CM |
995 | return ret; |
996 | } | |
997 | ||
5f39d397 | 998 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 999 | { |
d1310b2e CM |
1000 | struct extent_io_tree *tree; |
1001 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
1002 | extent_invalidatepage(tree, page, offset); |
1003 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 1004 | if (PagePrivate(page)) { |
d397712b CM |
1005 | printk(KERN_WARNING "btrfs warning page private not zero " |
1006 | "on page %llu\n", (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
1007 | ClearPagePrivate(page); |
1008 | set_page_private(page, 0); | |
1009 | page_cache_release(page); | |
1010 | } | |
d98237b3 CM |
1011 | } |
1012 | ||
7f09410b | 1013 | static const struct address_space_operations btree_aops = { |
d98237b3 CM |
1014 | .readpage = btree_readpage, |
1015 | .writepage = btree_writepage, | |
0da5468f | 1016 | .writepages = btree_writepages, |
5f39d397 CM |
1017 | .releasepage = btree_releasepage, |
1018 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1019 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1020 | .migratepage = btree_migratepage, |
5a92bc88 | 1021 | #endif |
d98237b3 CM |
1022 | }; |
1023 | ||
ca7a79ad CM |
1024 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1025 | u64 parent_transid) | |
090d1875 | 1026 | { |
5f39d397 CM |
1027 | struct extent_buffer *buf = NULL; |
1028 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 1029 | int ret = 0; |
090d1875 | 1030 | |
db94535d | 1031 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 1032 | if (!buf) |
090d1875 | 1033 | return 0; |
d1310b2e | 1034 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
bb82ab88 | 1035 | buf, 0, WAIT_NONE, btree_get_extent, 0); |
5f39d397 | 1036 | free_extent_buffer(buf); |
de428b63 | 1037 | return ret; |
090d1875 CM |
1038 | } |
1039 | ||
ab0fff03 AJ |
1040 | int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1041 | int mirror_num, struct extent_buffer **eb) | |
1042 | { | |
1043 | struct extent_buffer *buf = NULL; | |
1044 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1045 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; | |
1046 | int ret; | |
1047 | ||
1048 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
1049 | if (!buf) | |
1050 | return 0; | |
1051 | ||
1052 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1053 | ||
1054 | ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, | |
1055 | btree_get_extent, mirror_num); | |
1056 | if (ret) { | |
1057 | free_extent_buffer(buf); | |
1058 | return ret; | |
1059 | } | |
1060 | ||
1061 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1062 | free_extent_buffer(buf); | |
1063 | return -EIO; | |
1064 | } else if (extent_buffer_uptodate(io_tree, buf, NULL)) { | |
1065 | *eb = buf; | |
1066 | } else { | |
1067 | free_extent_buffer(buf); | |
1068 | } | |
1069 | return 0; | |
1070 | } | |
1071 | ||
0999df54 CM |
1072 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
1073 | u64 bytenr, u32 blocksize) | |
1074 | { | |
1075 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1076 | struct extent_buffer *eb; | |
1077 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
f09d1f60 | 1078 | bytenr, blocksize); |
0999df54 CM |
1079 | return eb; |
1080 | } | |
1081 | ||
1082 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
1083 | u64 bytenr, u32 blocksize) | |
1084 | { | |
1085 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1086 | struct extent_buffer *eb; | |
1087 | ||
1088 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
727011e0 | 1089 | bytenr, blocksize); |
0999df54 CM |
1090 | return eb; |
1091 | } | |
1092 | ||
1093 | ||
e02119d5 CM |
1094 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1095 | { | |
727011e0 | 1096 | return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, |
8aa38c31 | 1097 | buf->start + buf->len - 1); |
e02119d5 CM |
1098 | } |
1099 | ||
1100 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
1101 | { | |
727011e0 | 1102 | return filemap_fdatawait_range(buf->pages[0]->mapping, |
8aa38c31 | 1103 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
1104 | } |
1105 | ||
0999df54 | 1106 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 1107 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
1108 | { |
1109 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1110 | int ret; |
1111 | ||
0999df54 CM |
1112 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
1113 | if (!buf) | |
1114 | return NULL; | |
0999df54 | 1115 | |
ca7a79ad | 1116 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 | 1117 | |
d397712b | 1118 | if (ret == 0) |
b4ce94de | 1119 | set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags); |
5f39d397 | 1120 | return buf; |
ce9adaa5 | 1121 | |
eb60ceac CM |
1122 | } |
1123 | ||
e089f05c | 1124 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 1125 | struct extent_buffer *buf) |
ed2ff2cb | 1126 | { |
5f39d397 | 1127 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 | 1128 | if (btrfs_header_generation(buf) == |
925baedd | 1129 | root->fs_info->running_transaction->transid) { |
b9447ef8 | 1130 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1131 | |
b9473439 CM |
1132 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
1133 | spin_lock(&root->fs_info->delalloc_lock); | |
1134 | if (root->fs_info->dirty_metadata_bytes >= buf->len) | |
1135 | root->fs_info->dirty_metadata_bytes -= buf->len; | |
1136 | else | |
1137 | WARN_ON(1); | |
1138 | spin_unlock(&root->fs_info->delalloc_lock); | |
1139 | } | |
b4ce94de | 1140 | |
b9473439 CM |
1141 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1142 | btrfs_set_lock_blocking(buf); | |
d1310b2e | 1143 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 1144 | buf); |
925baedd | 1145 | } |
5f39d397 CM |
1146 | return 0; |
1147 | } | |
1148 | ||
db94535d | 1149 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 1150 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 1151 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 1152 | u64 objectid) |
d97e63b6 | 1153 | { |
cfaa7295 | 1154 | root->node = NULL; |
a28ec197 | 1155 | root->commit_root = NULL; |
db94535d CM |
1156 | root->sectorsize = sectorsize; |
1157 | root->nodesize = nodesize; | |
1158 | root->leafsize = leafsize; | |
87ee04eb | 1159 | root->stripesize = stripesize; |
123abc88 | 1160 | root->ref_cows = 0; |
0b86a832 | 1161 | root->track_dirty = 0; |
c71bf099 | 1162 | root->in_radix = 0; |
d68fc57b YZ |
1163 | root->orphan_item_inserted = 0; |
1164 | root->orphan_cleanup_state = 0; | |
0b86a832 | 1165 | |
0f7d52f4 CM |
1166 | root->objectid = objectid; |
1167 | root->last_trans = 0; | |
13a8a7c8 | 1168 | root->highest_objectid = 0; |
58176a96 | 1169 | root->name = NULL; |
6bef4d31 | 1170 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1171 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1172 | root->block_rsv = NULL; |
d68fc57b | 1173 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
1174 | |
1175 | INIT_LIST_HEAD(&root->dirty_list); | |
7b128766 | 1176 | INIT_LIST_HEAD(&root->orphan_list); |
5d4f98a2 | 1177 | INIT_LIST_HEAD(&root->root_list); |
d68fc57b | 1178 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 1179 | spin_lock_init(&root->inode_lock); |
f0486c68 | 1180 | spin_lock_init(&root->accounting_lock); |
a2135011 | 1181 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1182 | mutex_init(&root->log_mutex); |
7237f183 YZ |
1183 | init_waitqueue_head(&root->log_writer_wait); |
1184 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1185 | init_waitqueue_head(&root->log_commit_wait[1]); | |
1186 | atomic_set(&root->log_commit[0], 0); | |
1187 | atomic_set(&root->log_commit[1], 0); | |
1188 | atomic_set(&root->log_writers, 0); | |
1189 | root->log_batch = 0; | |
1190 | root->log_transid = 0; | |
257c62e1 | 1191 | root->last_log_commit = 0; |
d0c803c4 | 1192 | extent_io_tree_init(&root->dirty_log_pages, |
f993c883 | 1193 | fs_info->btree_inode->i_mapping); |
017e5369 | 1194 | |
3768f368 CM |
1195 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1196 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1197 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 1198 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 1199 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 1200 | init_completion(&root->kobj_unregister); |
6702ed49 | 1201 | root->defrag_running = 0; |
4d775673 | 1202 | root->root_key.objectid = objectid; |
0ee5dc67 | 1203 | root->anon_dev = 0; |
3768f368 CM |
1204 | return 0; |
1205 | } | |
1206 | ||
db94535d | 1207 | static int find_and_setup_root(struct btrfs_root *tree_root, |
9f5fae2f CM |
1208 | struct btrfs_fs_info *fs_info, |
1209 | u64 objectid, | |
e20d96d6 | 1210 | struct btrfs_root *root) |
3768f368 CM |
1211 | { |
1212 | int ret; | |
db94535d | 1213 | u32 blocksize; |
84234f3a | 1214 | u64 generation; |
3768f368 | 1215 | |
db94535d | 1216 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1217 | tree_root->sectorsize, tree_root->stripesize, |
1218 | root, fs_info, objectid); | |
3768f368 CM |
1219 | ret = btrfs_find_last_root(tree_root, objectid, |
1220 | &root->root_item, &root->root_key); | |
4df27c4d YZ |
1221 | if (ret > 0) |
1222 | return -ENOENT; | |
3768f368 CM |
1223 | BUG_ON(ret); |
1224 | ||
84234f3a | 1225 | generation = btrfs_root_generation(&root->root_item); |
db94535d | 1226 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
af31f5e5 | 1227 | root->commit_root = NULL; |
db94535d | 1228 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), |
84234f3a | 1229 | blocksize, generation); |
68433b73 CM |
1230 | if (!root->node || !btrfs_buffer_uptodate(root->node, generation)) { |
1231 | free_extent_buffer(root->node); | |
af31f5e5 | 1232 | root->node = NULL; |
68433b73 CM |
1233 | return -EIO; |
1234 | } | |
4df27c4d | 1235 | root->commit_root = btrfs_root_node(root); |
d97e63b6 CM |
1236 | return 0; |
1237 | } | |
1238 | ||
f84a8bd6 | 1239 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info) |
6f07e42e AV |
1240 | { |
1241 | struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS); | |
1242 | if (root) | |
1243 | root->fs_info = fs_info; | |
1244 | return root; | |
1245 | } | |
1246 | ||
7237f183 YZ |
1247 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1248 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1249 | { |
1250 | struct btrfs_root *root; | |
1251 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1252 | struct extent_buffer *leaf; |
e02119d5 | 1253 | |
6f07e42e | 1254 | root = btrfs_alloc_root(fs_info); |
e02119d5 | 1255 | if (!root) |
7237f183 | 1256 | return ERR_PTR(-ENOMEM); |
e02119d5 CM |
1257 | |
1258 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1259 | tree_root->sectorsize, tree_root->stripesize, | |
1260 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1261 | ||
1262 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1263 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1264 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
7237f183 YZ |
1265 | /* |
1266 | * log trees do not get reference counted because they go away | |
1267 | * before a real commit is actually done. They do store pointers | |
1268 | * to file data extents, and those reference counts still get | |
1269 | * updated (along with back refs to the log tree). | |
1270 | */ | |
e02119d5 CM |
1271 | root->ref_cows = 0; |
1272 | ||
5d4f98a2 | 1273 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, |
66d7e7f0 AJ |
1274 | BTRFS_TREE_LOG_OBJECTID, NULL, |
1275 | 0, 0, 0, 0); | |
7237f183 YZ |
1276 | if (IS_ERR(leaf)) { |
1277 | kfree(root); | |
1278 | return ERR_CAST(leaf); | |
1279 | } | |
e02119d5 | 1280 | |
5d4f98a2 YZ |
1281 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1282 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1283 | btrfs_set_header_generation(leaf, trans->transid); | |
1284 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1285 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1286 | root->node = leaf; |
e02119d5 CM |
1287 | |
1288 | write_extent_buffer(root->node, root->fs_info->fsid, | |
1289 | (unsigned long)btrfs_header_fsid(root->node), | |
1290 | BTRFS_FSID_SIZE); | |
1291 | btrfs_mark_buffer_dirty(root->node); | |
1292 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1293 | return root; |
1294 | } | |
1295 | ||
1296 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1297 | struct btrfs_fs_info *fs_info) | |
1298 | { | |
1299 | struct btrfs_root *log_root; | |
1300 | ||
1301 | log_root = alloc_log_tree(trans, fs_info); | |
1302 | if (IS_ERR(log_root)) | |
1303 | return PTR_ERR(log_root); | |
1304 | WARN_ON(fs_info->log_root_tree); | |
1305 | fs_info->log_root_tree = log_root; | |
1306 | return 0; | |
1307 | } | |
1308 | ||
1309 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1310 | struct btrfs_root *root) | |
1311 | { | |
1312 | struct btrfs_root *log_root; | |
1313 | struct btrfs_inode_item *inode_item; | |
1314 | ||
1315 | log_root = alloc_log_tree(trans, root->fs_info); | |
1316 | if (IS_ERR(log_root)) | |
1317 | return PTR_ERR(log_root); | |
1318 | ||
1319 | log_root->last_trans = trans->transid; | |
1320 | log_root->root_key.offset = root->root_key.objectid; | |
1321 | ||
1322 | inode_item = &log_root->root_item.inode; | |
1323 | inode_item->generation = cpu_to_le64(1); | |
1324 | inode_item->size = cpu_to_le64(3); | |
1325 | inode_item->nlink = cpu_to_le32(1); | |
1326 | inode_item->nbytes = cpu_to_le64(root->leafsize); | |
1327 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); | |
1328 | ||
5d4f98a2 | 1329 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1330 | |
1331 | WARN_ON(root->log_root); | |
1332 | root->log_root = log_root; | |
1333 | root->log_transid = 0; | |
257c62e1 | 1334 | root->last_log_commit = 0; |
e02119d5 CM |
1335 | return 0; |
1336 | } | |
1337 | ||
1338 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
1339 | struct btrfs_key *location) | |
1340 | { | |
1341 | struct btrfs_root *root; | |
1342 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1343 | struct btrfs_path *path; |
5f39d397 | 1344 | struct extent_buffer *l; |
84234f3a | 1345 | u64 generation; |
db94535d | 1346 | u32 blocksize; |
0f7d52f4 CM |
1347 | int ret = 0; |
1348 | ||
6f07e42e | 1349 | root = btrfs_alloc_root(fs_info); |
0cf6c620 | 1350 | if (!root) |
0f7d52f4 | 1351 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 1352 | if (location->offset == (u64)-1) { |
db94535d | 1353 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
1354 | location->objectid, root); |
1355 | if (ret) { | |
0f7d52f4 CM |
1356 | kfree(root); |
1357 | return ERR_PTR(ret); | |
1358 | } | |
13a8a7c8 | 1359 | goto out; |
0f7d52f4 CM |
1360 | } |
1361 | ||
db94535d | 1362 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1363 | tree_root->sectorsize, tree_root->stripesize, |
1364 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
1365 | |
1366 | path = btrfs_alloc_path(); | |
db5b493a TI |
1367 | if (!path) { |
1368 | kfree(root); | |
1369 | return ERR_PTR(-ENOMEM); | |
1370 | } | |
0f7d52f4 | 1371 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); |
13a8a7c8 YZ |
1372 | if (ret == 0) { |
1373 | l = path->nodes[0]; | |
1374 | read_extent_buffer(l, &root->root_item, | |
1375 | btrfs_item_ptr_offset(l, path->slots[0]), | |
1376 | sizeof(root->root_item)); | |
1377 | memcpy(&root->root_key, location, sizeof(*location)); | |
0f7d52f4 | 1378 | } |
0f7d52f4 CM |
1379 | btrfs_free_path(path); |
1380 | if (ret) { | |
5e540f77 | 1381 | kfree(root); |
13a8a7c8 YZ |
1382 | if (ret > 0) |
1383 | ret = -ENOENT; | |
0f7d52f4 CM |
1384 | return ERR_PTR(ret); |
1385 | } | |
13a8a7c8 | 1386 | |
84234f3a | 1387 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1388 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1389 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1390 | blocksize, generation); |
5d4f98a2 | 1391 | root->commit_root = btrfs_root_node(root); |
0f7d52f4 | 1392 | BUG_ON(!root->node); |
13a8a7c8 | 1393 | out: |
08fe4db1 | 1394 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { |
e02119d5 | 1395 | root->ref_cows = 1; |
08fe4db1 LZ |
1396 | btrfs_check_and_init_root_item(&root->root_item); |
1397 | } | |
13a8a7c8 | 1398 | |
5eda7b5e CM |
1399 | return root; |
1400 | } | |
1401 | ||
edbd8d4e CM |
1402 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
1403 | struct btrfs_key *location) | |
5eda7b5e CM |
1404 | { |
1405 | struct btrfs_root *root; | |
1406 | int ret; | |
1407 | ||
edbd8d4e CM |
1408 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1409 | return fs_info->tree_root; | |
1410 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1411 | return fs_info->extent_root; | |
8f18cf13 CM |
1412 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1413 | return fs_info->chunk_root; | |
1414 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1415 | return fs_info->dev_root; | |
0403e47e YZ |
1416 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1417 | return fs_info->csum_root; | |
4df27c4d YZ |
1418 | again: |
1419 | spin_lock(&fs_info->fs_roots_radix_lock); | |
5eda7b5e CM |
1420 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
1421 | (unsigned long)location->objectid); | |
4df27c4d | 1422 | spin_unlock(&fs_info->fs_roots_radix_lock); |
5eda7b5e CM |
1423 | if (root) |
1424 | return root; | |
1425 | ||
e02119d5 | 1426 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
1427 | if (IS_ERR(root)) |
1428 | return root; | |
3394e160 | 1429 | |
581bb050 | 1430 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); |
581bb050 LZ |
1431 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), |
1432 | GFP_NOFS); | |
35a30d7c DS |
1433 | if (!root->free_ino_pinned || !root->free_ino_ctl) { |
1434 | ret = -ENOMEM; | |
581bb050 | 1435 | goto fail; |
35a30d7c | 1436 | } |
581bb050 LZ |
1437 | |
1438 | btrfs_init_free_ino_ctl(root); | |
1439 | mutex_init(&root->fs_commit_mutex); | |
1440 | spin_lock_init(&root->cache_lock); | |
1441 | init_waitqueue_head(&root->cache_wait); | |
1442 | ||
0ee5dc67 | 1443 | ret = get_anon_bdev(&root->anon_dev); |
ac08aedf CM |
1444 | if (ret) |
1445 | goto fail; | |
3394e160 | 1446 | |
d68fc57b YZ |
1447 | if (btrfs_root_refs(&root->root_item) == 0) { |
1448 | ret = -ENOENT; | |
1449 | goto fail; | |
1450 | } | |
1451 | ||
1452 | ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid); | |
1453 | if (ret < 0) | |
1454 | goto fail; | |
1455 | if (ret == 0) | |
1456 | root->orphan_item_inserted = 1; | |
1457 | ||
4df27c4d YZ |
1458 | ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); |
1459 | if (ret) | |
1460 | goto fail; | |
1461 | ||
1462 | spin_lock(&fs_info->fs_roots_radix_lock); | |
2619ba1f CM |
1463 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
1464 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 | 1465 | root); |
d68fc57b | 1466 | if (ret == 0) |
4df27c4d | 1467 | root->in_radix = 1; |
d68fc57b | 1468 | |
4df27c4d YZ |
1469 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1470 | radix_tree_preload_end(); | |
0f7d52f4 | 1471 | if (ret) { |
4df27c4d YZ |
1472 | if (ret == -EEXIST) { |
1473 | free_fs_root(root); | |
1474 | goto again; | |
1475 | } | |
1476 | goto fail; | |
0f7d52f4 | 1477 | } |
4df27c4d YZ |
1478 | |
1479 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
1480 | root->root_key.objectid); | |
1481 | WARN_ON(ret); | |
edbd8d4e | 1482 | return root; |
4df27c4d YZ |
1483 | fail: |
1484 | free_fs_root(root); | |
1485 | return ERR_PTR(ret); | |
edbd8d4e CM |
1486 | } |
1487 | ||
04160088 CM |
1488 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1489 | { | |
1490 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1491 | int ret = 0; | |
04160088 CM |
1492 | struct btrfs_device *device; |
1493 | struct backing_dev_info *bdi; | |
b7967db7 | 1494 | |
1f78160c XG |
1495 | rcu_read_lock(); |
1496 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1497 | if (!device->bdev) |
1498 | continue; | |
04160088 CM |
1499 | bdi = blk_get_backing_dev_info(device->bdev); |
1500 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1501 | ret = 1; | |
1502 | break; | |
1503 | } | |
1504 | } | |
1f78160c | 1505 | rcu_read_unlock(); |
04160088 CM |
1506 | return ret; |
1507 | } | |
1508 | ||
ad081f14 JA |
1509 | /* |
1510 | * If this fails, caller must call bdi_destroy() to get rid of the | |
1511 | * bdi again. | |
1512 | */ | |
04160088 CM |
1513 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1514 | { | |
ad081f14 JA |
1515 | int err; |
1516 | ||
1517 | bdi->capabilities = BDI_CAP_MAP_COPY; | |
e6d086d8 | 1518 | err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY); |
ad081f14 JA |
1519 | if (err) |
1520 | return err; | |
1521 | ||
4575c9cc | 1522 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1523 | bdi->congested_fn = btrfs_congested_fn; |
1524 | bdi->congested_data = info; | |
1525 | return 0; | |
1526 | } | |
1527 | ||
8b712842 CM |
1528 | /* |
1529 | * called by the kthread helper functions to finally call the bio end_io | |
1530 | * functions. This is where read checksum verification actually happens | |
1531 | */ | |
1532 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1533 | { |
ce9adaa5 | 1534 | struct bio *bio; |
8b712842 CM |
1535 | struct end_io_wq *end_io_wq; |
1536 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1537 | int error; |
ce9adaa5 | 1538 | |
8b712842 CM |
1539 | end_io_wq = container_of(work, struct end_io_wq, work); |
1540 | bio = end_io_wq->bio; | |
1541 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1542 | |
8b712842 CM |
1543 | error = end_io_wq->error; |
1544 | bio->bi_private = end_io_wq->private; | |
1545 | bio->bi_end_io = end_io_wq->end_io; | |
1546 | kfree(end_io_wq); | |
8b712842 | 1547 | bio_endio(bio, error); |
44b8bd7e CM |
1548 | } |
1549 | ||
a74a4b97 CM |
1550 | static int cleaner_kthread(void *arg) |
1551 | { | |
1552 | struct btrfs_root *root = arg; | |
1553 | ||
1554 | do { | |
a74a4b97 | 1555 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); |
76dda93c YZ |
1556 | |
1557 | if (!(root->fs_info->sb->s_flags & MS_RDONLY) && | |
1558 | mutex_trylock(&root->fs_info->cleaner_mutex)) { | |
24bbcf04 | 1559 | btrfs_run_delayed_iputs(root); |
76dda93c YZ |
1560 | btrfs_clean_old_snapshots(root); |
1561 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
4cb5300b | 1562 | btrfs_run_defrag_inodes(root->fs_info); |
76dda93c | 1563 | } |
a74a4b97 | 1564 | |
a0acae0e | 1565 | if (!try_to_freeze()) { |
a74a4b97 | 1566 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1567 | if (!kthread_should_stop()) |
1568 | schedule(); | |
a74a4b97 CM |
1569 | __set_current_state(TASK_RUNNING); |
1570 | } | |
1571 | } while (!kthread_should_stop()); | |
1572 | return 0; | |
1573 | } | |
1574 | ||
1575 | static int transaction_kthread(void *arg) | |
1576 | { | |
1577 | struct btrfs_root *root = arg; | |
1578 | struct btrfs_trans_handle *trans; | |
1579 | struct btrfs_transaction *cur; | |
8929ecfa | 1580 | u64 transid; |
a74a4b97 CM |
1581 | unsigned long now; |
1582 | unsigned long delay; | |
1583 | int ret; | |
1584 | ||
1585 | do { | |
a74a4b97 CM |
1586 | delay = HZ * 30; |
1587 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1588 | mutex_lock(&root->fs_info->transaction_kthread_mutex); | |
1589 | ||
a4abeea4 | 1590 | spin_lock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1591 | cur = root->fs_info->running_transaction; |
1592 | if (!cur) { | |
a4abeea4 | 1593 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1594 | goto sleep; |
1595 | } | |
31153d81 | 1596 | |
a74a4b97 | 1597 | now = get_seconds(); |
8929ecfa YZ |
1598 | if (!cur->blocked && |
1599 | (now < cur->start_time || now - cur->start_time < 30)) { | |
a4abeea4 | 1600 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1601 | delay = HZ * 5; |
1602 | goto sleep; | |
1603 | } | |
8929ecfa | 1604 | transid = cur->transid; |
a4abeea4 | 1605 | spin_unlock(&root->fs_info->trans_lock); |
56bec294 | 1606 | |
7a7eaa40 | 1607 | trans = btrfs_join_transaction(root); |
3612b495 | 1608 | BUG_ON(IS_ERR(trans)); |
8929ecfa YZ |
1609 | if (transid == trans->transid) { |
1610 | ret = btrfs_commit_transaction(trans, root); | |
1611 | BUG_ON(ret); | |
1612 | } else { | |
1613 | btrfs_end_transaction(trans, root); | |
1614 | } | |
a74a4b97 CM |
1615 | sleep: |
1616 | wake_up_process(root->fs_info->cleaner_kthread); | |
1617 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1618 | ||
a0acae0e | 1619 | if (!try_to_freeze()) { |
a74a4b97 | 1620 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1621 | if (!kthread_should_stop() && |
1622 | !btrfs_transaction_blocked(root->fs_info)) | |
1623 | schedule_timeout(delay); | |
a74a4b97 CM |
1624 | __set_current_state(TASK_RUNNING); |
1625 | } | |
1626 | } while (!kthread_should_stop()); | |
1627 | return 0; | |
1628 | } | |
1629 | ||
af31f5e5 CM |
1630 | /* |
1631 | * this will find the highest generation in the array of | |
1632 | * root backups. The index of the highest array is returned, | |
1633 | * or -1 if we can't find anything. | |
1634 | * | |
1635 | * We check to make sure the array is valid by comparing the | |
1636 | * generation of the latest root in the array with the generation | |
1637 | * in the super block. If they don't match we pitch it. | |
1638 | */ | |
1639 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1640 | { | |
1641 | u64 cur; | |
1642 | int newest_index = -1; | |
1643 | struct btrfs_root_backup *root_backup; | |
1644 | int i; | |
1645 | ||
1646 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1647 | root_backup = info->super_copy->super_roots + i; | |
1648 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1649 | if (cur == newest_gen) | |
1650 | newest_index = i; | |
1651 | } | |
1652 | ||
1653 | /* check to see if we actually wrapped around */ | |
1654 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
1655 | root_backup = info->super_copy->super_roots; | |
1656 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1657 | if (cur == newest_gen) | |
1658 | newest_index = 0; | |
1659 | } | |
1660 | return newest_index; | |
1661 | } | |
1662 | ||
1663 | ||
1664 | /* | |
1665 | * find the oldest backup so we know where to store new entries | |
1666 | * in the backup array. This will set the backup_root_index | |
1667 | * field in the fs_info struct | |
1668 | */ | |
1669 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
1670 | u64 newest_gen) | |
1671 | { | |
1672 | int newest_index = -1; | |
1673 | ||
1674 | newest_index = find_newest_super_backup(info, newest_gen); | |
1675 | /* if there was garbage in there, just move along */ | |
1676 | if (newest_index == -1) { | |
1677 | info->backup_root_index = 0; | |
1678 | } else { | |
1679 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1680 | } | |
1681 | } | |
1682 | ||
1683 | /* | |
1684 | * copy all the root pointers into the super backup array. | |
1685 | * this will bump the backup pointer by one when it is | |
1686 | * done | |
1687 | */ | |
1688 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1689 | { | |
1690 | int next_backup; | |
1691 | struct btrfs_root_backup *root_backup; | |
1692 | int last_backup; | |
1693 | ||
1694 | next_backup = info->backup_root_index; | |
1695 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1696 | BTRFS_NUM_BACKUP_ROOTS; | |
1697 | ||
1698 | /* | |
1699 | * just overwrite the last backup if we're at the same generation | |
1700 | * this happens only at umount | |
1701 | */ | |
1702 | root_backup = info->super_for_commit->super_roots + last_backup; | |
1703 | if (btrfs_backup_tree_root_gen(root_backup) == | |
1704 | btrfs_header_generation(info->tree_root->node)) | |
1705 | next_backup = last_backup; | |
1706 | ||
1707 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1708 | ||
1709 | /* | |
1710 | * make sure all of our padding and empty slots get zero filled | |
1711 | * regardless of which ones we use today | |
1712 | */ | |
1713 | memset(root_backup, 0, sizeof(*root_backup)); | |
1714 | ||
1715 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1716 | ||
1717 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1718 | btrfs_set_backup_tree_root_gen(root_backup, | |
1719 | btrfs_header_generation(info->tree_root->node)); | |
1720 | ||
1721 | btrfs_set_backup_tree_root_level(root_backup, | |
1722 | btrfs_header_level(info->tree_root->node)); | |
1723 | ||
1724 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1725 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1726 | btrfs_header_generation(info->chunk_root->node)); | |
1727 | btrfs_set_backup_chunk_root_level(root_backup, | |
1728 | btrfs_header_level(info->chunk_root->node)); | |
1729 | ||
1730 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
1731 | btrfs_set_backup_extent_root_gen(root_backup, | |
1732 | btrfs_header_generation(info->extent_root->node)); | |
1733 | btrfs_set_backup_extent_root_level(root_backup, | |
1734 | btrfs_header_level(info->extent_root->node)); | |
1735 | ||
7c7e82a7 CM |
1736 | /* |
1737 | * we might commit during log recovery, which happens before we set | |
1738 | * the fs_root. Make sure it is valid before we fill it in. | |
1739 | */ | |
1740 | if (info->fs_root && info->fs_root->node) { | |
1741 | btrfs_set_backup_fs_root(root_backup, | |
1742 | info->fs_root->node->start); | |
1743 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1744 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1745 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1746 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1747 | } |
af31f5e5 CM |
1748 | |
1749 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1750 | btrfs_set_backup_dev_root_gen(root_backup, | |
1751 | btrfs_header_generation(info->dev_root->node)); | |
1752 | btrfs_set_backup_dev_root_level(root_backup, | |
1753 | btrfs_header_level(info->dev_root->node)); | |
1754 | ||
1755 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
1756 | btrfs_set_backup_csum_root_gen(root_backup, | |
1757 | btrfs_header_generation(info->csum_root->node)); | |
1758 | btrfs_set_backup_csum_root_level(root_backup, | |
1759 | btrfs_header_level(info->csum_root->node)); | |
1760 | ||
1761 | btrfs_set_backup_total_bytes(root_backup, | |
1762 | btrfs_super_total_bytes(info->super_copy)); | |
1763 | btrfs_set_backup_bytes_used(root_backup, | |
1764 | btrfs_super_bytes_used(info->super_copy)); | |
1765 | btrfs_set_backup_num_devices(root_backup, | |
1766 | btrfs_super_num_devices(info->super_copy)); | |
1767 | ||
1768 | /* | |
1769 | * if we don't copy this out to the super_copy, it won't get remembered | |
1770 | * for the next commit | |
1771 | */ | |
1772 | memcpy(&info->super_copy->super_roots, | |
1773 | &info->super_for_commit->super_roots, | |
1774 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1775 | } | |
1776 | ||
1777 | /* | |
1778 | * this copies info out of the root backup array and back into | |
1779 | * the in-memory super block. It is meant to help iterate through | |
1780 | * the array, so you send it the number of backups you've already | |
1781 | * tried and the last backup index you used. | |
1782 | * | |
1783 | * this returns -1 when it has tried all the backups | |
1784 | */ | |
1785 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
1786 | struct btrfs_super_block *super, | |
1787 | int *num_backups_tried, int *backup_index) | |
1788 | { | |
1789 | struct btrfs_root_backup *root_backup; | |
1790 | int newest = *backup_index; | |
1791 | ||
1792 | if (*num_backups_tried == 0) { | |
1793 | u64 gen = btrfs_super_generation(super); | |
1794 | ||
1795 | newest = find_newest_super_backup(info, gen); | |
1796 | if (newest == -1) | |
1797 | return -1; | |
1798 | ||
1799 | *backup_index = newest; | |
1800 | *num_backups_tried = 1; | |
1801 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
1802 | /* we've tried all the backups, all done */ | |
1803 | return -1; | |
1804 | } else { | |
1805 | /* jump to the next oldest backup */ | |
1806 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1807 | BTRFS_NUM_BACKUP_ROOTS; | |
1808 | *backup_index = newest; | |
1809 | *num_backups_tried += 1; | |
1810 | } | |
1811 | root_backup = super->super_roots + newest; | |
1812 | ||
1813 | btrfs_set_super_generation(super, | |
1814 | btrfs_backup_tree_root_gen(root_backup)); | |
1815 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
1816 | btrfs_set_super_root_level(super, | |
1817 | btrfs_backup_tree_root_level(root_backup)); | |
1818 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
1819 | ||
1820 | /* | |
1821 | * fixme: the total bytes and num_devices need to match or we should | |
1822 | * need a fsck | |
1823 | */ | |
1824 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
1825 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
1826 | return 0; | |
1827 | } | |
1828 | ||
1829 | /* helper to cleanup tree roots */ | |
1830 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
1831 | { | |
1832 | free_extent_buffer(info->tree_root->node); | |
1833 | free_extent_buffer(info->tree_root->commit_root); | |
1834 | free_extent_buffer(info->dev_root->node); | |
1835 | free_extent_buffer(info->dev_root->commit_root); | |
1836 | free_extent_buffer(info->extent_root->node); | |
1837 | free_extent_buffer(info->extent_root->commit_root); | |
1838 | free_extent_buffer(info->csum_root->node); | |
1839 | free_extent_buffer(info->csum_root->commit_root); | |
1840 | ||
1841 | info->tree_root->node = NULL; | |
1842 | info->tree_root->commit_root = NULL; | |
1843 | info->dev_root->node = NULL; | |
1844 | info->dev_root->commit_root = NULL; | |
1845 | info->extent_root->node = NULL; | |
1846 | info->extent_root->commit_root = NULL; | |
1847 | info->csum_root->node = NULL; | |
1848 | info->csum_root->commit_root = NULL; | |
1849 | ||
1850 | if (chunk_root) { | |
1851 | free_extent_buffer(info->chunk_root->node); | |
1852 | free_extent_buffer(info->chunk_root->commit_root); | |
1853 | info->chunk_root->node = NULL; | |
1854 | info->chunk_root->commit_root = NULL; | |
1855 | } | |
1856 | } | |
1857 | ||
1858 | ||
ad2b2c80 AV |
1859 | int open_ctree(struct super_block *sb, |
1860 | struct btrfs_fs_devices *fs_devices, | |
1861 | char *options) | |
2e635a27 | 1862 | { |
db94535d CM |
1863 | u32 sectorsize; |
1864 | u32 nodesize; | |
1865 | u32 leafsize; | |
1866 | u32 blocksize; | |
87ee04eb | 1867 | u32 stripesize; |
84234f3a | 1868 | u64 generation; |
f2b636e8 | 1869 | u64 features; |
3de4586c | 1870 | struct btrfs_key location; |
a061fc8d | 1871 | struct buffer_head *bh; |
4d34b278 | 1872 | struct btrfs_super_block *disk_super; |
815745cf | 1873 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
f84a8bd6 | 1874 | struct btrfs_root *tree_root; |
4d34b278 ID |
1875 | struct btrfs_root *extent_root; |
1876 | struct btrfs_root *csum_root; | |
1877 | struct btrfs_root *chunk_root; | |
1878 | struct btrfs_root *dev_root; | |
e02119d5 | 1879 | struct btrfs_root *log_tree_root; |
eb60ceac | 1880 | int ret; |
e58ca020 | 1881 | int err = -EINVAL; |
af31f5e5 CM |
1882 | int num_backups_tried = 0; |
1883 | int backup_index = 0; | |
4543df7e | 1884 | |
f84a8bd6 | 1885 | tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info); |
6f07e42e AV |
1886 | extent_root = fs_info->extent_root = btrfs_alloc_root(fs_info); |
1887 | csum_root = fs_info->csum_root = btrfs_alloc_root(fs_info); | |
1888 | chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info); | |
1889 | dev_root = fs_info->dev_root = btrfs_alloc_root(fs_info); | |
8790d502 | 1890 | |
f84a8bd6 AV |
1891 | if (!tree_root || !extent_root || !csum_root || |
1892 | !chunk_root || !dev_root) { | |
39279cc3 CM |
1893 | err = -ENOMEM; |
1894 | goto fail; | |
1895 | } | |
76dda93c YZ |
1896 | |
1897 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
1898 | if (ret) { | |
1899 | err = ret; | |
1900 | goto fail; | |
1901 | } | |
1902 | ||
1903 | ret = setup_bdi(fs_info, &fs_info->bdi); | |
1904 | if (ret) { | |
1905 | err = ret; | |
1906 | goto fail_srcu; | |
1907 | } | |
1908 | ||
1909 | fs_info->btree_inode = new_inode(sb); | |
1910 | if (!fs_info->btree_inode) { | |
1911 | err = -ENOMEM; | |
1912 | goto fail_bdi; | |
1913 | } | |
1914 | ||
a6591715 | 1915 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
1561deda | 1916 | |
76dda93c | 1917 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
8fd17795 | 1918 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1919 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 1920 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
19c00ddc | 1921 | INIT_LIST_HEAD(&fs_info->hashers); |
ea8c2819 | 1922 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
5a3f23d5 | 1923 | INIT_LIST_HEAD(&fs_info->ordered_operations); |
11833d66 | 1924 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
1832a6d5 | 1925 | spin_lock_init(&fs_info->delalloc_lock); |
a4abeea4 | 1926 | spin_lock_init(&fs_info->trans_lock); |
31153d81 | 1927 | spin_lock_init(&fs_info->ref_cache_lock); |
76dda93c | 1928 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 1929 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 1930 | spin_lock_init(&fs_info->defrag_inodes_lock); |
2bf64758 | 1931 | spin_lock_init(&fs_info->free_chunk_lock); |
7585717f | 1932 | mutex_init(&fs_info->reloc_mutex); |
19c00ddc | 1933 | |
58176a96 | 1934 | init_completion(&fs_info->kobj_unregister); |
0b86a832 | 1935 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1936 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1937 | btrfs_mapping_init(&fs_info->mapping_tree); |
f0486c68 YZ |
1938 | btrfs_init_block_rsv(&fs_info->global_block_rsv); |
1939 | btrfs_init_block_rsv(&fs_info->delalloc_block_rsv); | |
1940 | btrfs_init_block_rsv(&fs_info->trans_block_rsv); | |
1941 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv); | |
1942 | btrfs_init_block_rsv(&fs_info->empty_block_rsv); | |
6d668dda | 1943 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv); |
cb03c743 | 1944 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 1945 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 1946 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 1947 | atomic_set(&fs_info->nr_async_bios, 0); |
4cb5300b | 1948 | atomic_set(&fs_info->defrag_running, 0); |
e20d96d6 | 1949 | fs_info->sb = sb; |
6f568d35 | 1950 | fs_info->max_inline = 8192 * 1024; |
9ed74f2d | 1951 | fs_info->metadata_ratio = 0; |
4cb5300b | 1952 | fs_info->defrag_inodes = RB_ROOT; |
a4abeea4 | 1953 | fs_info->trans_no_join = 0; |
2bf64758 | 1954 | fs_info->free_chunk_space = 0; |
c8b97818 | 1955 | |
90519d66 AJ |
1956 | /* readahead state */ |
1957 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); | |
1958 | spin_lock_init(&fs_info->reada_lock); | |
c8b97818 | 1959 | |
b34b086c CM |
1960 | fs_info->thread_pool_size = min_t(unsigned long, |
1961 | num_online_cpus() + 2, 8); | |
0afbaf8c | 1962 | |
3eaa2885 CM |
1963 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
1964 | spin_lock_init(&fs_info->ordered_extent_lock); | |
16cdcec7 MX |
1965 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
1966 | GFP_NOFS); | |
1967 | if (!fs_info->delayed_root) { | |
1968 | err = -ENOMEM; | |
1969 | goto fail_iput; | |
1970 | } | |
1971 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 1972 | |
a2de733c AJ |
1973 | mutex_init(&fs_info->scrub_lock); |
1974 | atomic_set(&fs_info->scrubs_running, 0); | |
1975 | atomic_set(&fs_info->scrub_pause_req, 0); | |
1976 | atomic_set(&fs_info->scrubs_paused, 0); | |
1977 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
1978 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
1979 | init_rwsem(&fs_info->scrub_super_lock); | |
1980 | fs_info->scrub_workers_refcnt = 0; | |
21adbd5c SB |
1981 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
1982 | fs_info->check_integrity_print_mask = 0; | |
1983 | #endif | |
a2de733c | 1984 | |
c9e9f97b ID |
1985 | spin_lock_init(&fs_info->balance_lock); |
1986 | mutex_init(&fs_info->balance_mutex); | |
837d5b6e ID |
1987 | atomic_set(&fs_info->balance_running, 0); |
1988 | atomic_set(&fs_info->balance_pause_req, 0); | |
a7e99c69 | 1989 | atomic_set(&fs_info->balance_cancel_req, 0); |
c9e9f97b | 1990 | fs_info->balance_ctl = NULL; |
837d5b6e | 1991 | init_waitqueue_head(&fs_info->balance_wait_q); |
a2de733c | 1992 | |
a061fc8d CM |
1993 | sb->s_blocksize = 4096; |
1994 | sb->s_blocksize_bits = blksize_bits(4096); | |
32a88aa1 | 1995 | sb->s_bdi = &fs_info->bdi; |
a061fc8d | 1996 | |
76dda93c | 1997 | fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; |
bfe86848 | 1998 | set_nlink(fs_info->btree_inode, 1); |
0afbaf8c CM |
1999 | /* |
2000 | * we set the i_size on the btree inode to the max possible int. | |
2001 | * the real end of the address space is determined by all of | |
2002 | * the devices in the system | |
2003 | */ | |
2004 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 2005 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
2006 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
2007 | ||
5d4f98a2 | 2008 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); |
d1310b2e | 2009 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
f993c883 | 2010 | fs_info->btree_inode->i_mapping); |
a8067e02 | 2011 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); |
d1310b2e CM |
2012 | |
2013 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 2014 | |
76dda93c YZ |
2015 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
2016 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
2017 | sizeof(struct btrfs_key)); | |
2018 | BTRFS_I(fs_info->btree_inode)->dummy_inode = 1; | |
c65ddb52 | 2019 | insert_inode_hash(fs_info->btree_inode); |
76dda93c | 2020 | |
0f9dd46c | 2021 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2022 | fs_info->block_group_cache_tree = RB_ROOT; |
0f9dd46c | 2023 | |
11833d66 | 2024 | extent_io_tree_init(&fs_info->freed_extents[0], |
f993c883 | 2025 | fs_info->btree_inode->i_mapping); |
11833d66 | 2026 | extent_io_tree_init(&fs_info->freed_extents[1], |
f993c883 | 2027 | fs_info->btree_inode->i_mapping); |
11833d66 | 2028 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
e66f709b | 2029 | fs_info->do_barriers = 1; |
e18e4809 | 2030 | |
39279cc3 | 2031 | |
5a3f23d5 | 2032 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2033 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2034 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2035 | mutex_init(&fs_info->transaction_kthread_mutex); |
2036 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 2037 | mutex_init(&fs_info->volume_mutex); |
276e680d | 2038 | init_rwsem(&fs_info->extent_commit_sem); |
c71bf099 | 2039 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2040 | init_rwsem(&fs_info->subvol_sem); |
fa9c0d79 CM |
2041 | |
2042 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); | |
2043 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2044 | ||
e6dcd2dc | 2045 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2046 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2047 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2048 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2049 | |
0b86a832 | 2050 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 2051 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2052 | |
a512bbf8 | 2053 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
20b45077 DY |
2054 | if (!bh) { |
2055 | err = -EINVAL; | |
16cdcec7 | 2056 | goto fail_alloc; |
20b45077 | 2057 | } |
39279cc3 | 2058 | |
6c41761f DS |
2059 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2060 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2061 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2062 | brelse(bh); |
5f39d397 | 2063 | |
6c41761f | 2064 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2065 | |
6c41761f | 2066 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2067 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2068 | goto fail_alloc; |
0f7d52f4 | 2069 | |
acce952b | 2070 | /* check FS state, whether FS is broken. */ |
2071 | fs_info->fs_state |= btrfs_super_flags(disk_super); | |
2072 | ||
2073 | btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); | |
2074 | ||
af31f5e5 CM |
2075 | /* |
2076 | * run through our array of backup supers and setup | |
2077 | * our ring pointer to the oldest one | |
2078 | */ | |
2079 | generation = btrfs_super_generation(disk_super); | |
2080 | find_oldest_super_backup(fs_info, generation); | |
2081 | ||
75e7cb7f LB |
2082 | /* |
2083 | * In the long term, we'll store the compression type in the super | |
2084 | * block, and it'll be used for per file compression control. | |
2085 | */ | |
2086 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2087 | ||
2b82032c YZ |
2088 | ret = btrfs_parse_options(tree_root, options); |
2089 | if (ret) { | |
2090 | err = ret; | |
16cdcec7 | 2091 | goto fail_alloc; |
2b82032c | 2092 | } |
dfe25020 | 2093 | |
f2b636e8 JB |
2094 | features = btrfs_super_incompat_flags(disk_super) & |
2095 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2096 | if (features) { | |
2097 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
2098 | "unsupported optional features (%Lx).\n", | |
21380931 | 2099 | (unsigned long long)features); |
f2b636e8 | 2100 | err = -EINVAL; |
16cdcec7 | 2101 | goto fail_alloc; |
f2b636e8 JB |
2102 | } |
2103 | ||
727011e0 CM |
2104 | if (btrfs_super_leafsize(disk_super) != |
2105 | btrfs_super_nodesize(disk_super)) { | |
2106 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2107 | "blocksizes don't match. node %d leaf %d\n", | |
2108 | btrfs_super_nodesize(disk_super), | |
2109 | btrfs_super_leafsize(disk_super)); | |
2110 | err = -EINVAL; | |
2111 | goto fail_alloc; | |
2112 | } | |
2113 | if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2114 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2115 | "blocksize (%d) was too large\n", | |
2116 | btrfs_super_leafsize(disk_super)); | |
2117 | err = -EINVAL; | |
2118 | goto fail_alloc; | |
2119 | } | |
2120 | ||
5d4f98a2 | 2121 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae LZ |
2122 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
2123 | if (tree_root->fs_info->compress_type & BTRFS_COMPRESS_LZO) | |
2124 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; | |
727011e0 CM |
2125 | |
2126 | /* | |
2127 | * flag our filesystem as having big metadata blocks if | |
2128 | * they are bigger than the page size | |
2129 | */ | |
2130 | if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) { | |
2131 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) | |
2132 | printk(KERN_INFO "btrfs flagging fs with big metadata feature\n"); | |
2133 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
2134 | } | |
2135 | ||
a6fa6fae | 2136 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 2137 | |
f2b636e8 JB |
2138 | features = btrfs_super_compat_ro_flags(disk_super) & |
2139 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
2140 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
2141 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
2142 | "unsupported option features (%Lx).\n", | |
21380931 | 2143 | (unsigned long long)features); |
f2b636e8 | 2144 | err = -EINVAL; |
16cdcec7 | 2145 | goto fail_alloc; |
f2b636e8 | 2146 | } |
61d92c32 CM |
2147 | |
2148 | btrfs_init_workers(&fs_info->generic_worker, | |
2149 | "genwork", 1, NULL); | |
2150 | ||
5443be45 | 2151 | btrfs_init_workers(&fs_info->workers, "worker", |
61d92c32 CM |
2152 | fs_info->thread_pool_size, |
2153 | &fs_info->generic_worker); | |
c8b97818 | 2154 | |
771ed689 | 2155 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
61d92c32 CM |
2156 | fs_info->thread_pool_size, |
2157 | &fs_info->generic_worker); | |
771ed689 | 2158 | |
5443be45 | 2159 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 | 2160 | min_t(u64, fs_devices->num_devices, |
61d92c32 CM |
2161 | fs_info->thread_pool_size), |
2162 | &fs_info->generic_worker); | |
61b49440 | 2163 | |
bab39bf9 JB |
2164 | btrfs_init_workers(&fs_info->caching_workers, "cache", |
2165 | 2, &fs_info->generic_worker); | |
2166 | ||
61b49440 CM |
2167 | /* a higher idle thresh on the submit workers makes it much more |
2168 | * likely that bios will be send down in a sane order to the | |
2169 | * devices | |
2170 | */ | |
2171 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 2172 | |
771ed689 | 2173 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 2174 | fs_info->workers.ordered = 1; |
61b49440 | 2175 | |
771ed689 CM |
2176 | fs_info->delalloc_workers.idle_thresh = 2; |
2177 | fs_info->delalloc_workers.ordered = 1; | |
2178 | ||
61d92c32 CM |
2179 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1, |
2180 | &fs_info->generic_worker); | |
5443be45 | 2181 | btrfs_init_workers(&fs_info->endio_workers, "endio", |
61d92c32 CM |
2182 | fs_info->thread_pool_size, |
2183 | &fs_info->generic_worker); | |
d20f7043 | 2184 | btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", |
61d92c32 CM |
2185 | fs_info->thread_pool_size, |
2186 | &fs_info->generic_worker); | |
cad321ad | 2187 | btrfs_init_workers(&fs_info->endio_meta_write_workers, |
61d92c32 CM |
2188 | "endio-meta-write", fs_info->thread_pool_size, |
2189 | &fs_info->generic_worker); | |
5443be45 | 2190 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", |
61d92c32 CM |
2191 | fs_info->thread_pool_size, |
2192 | &fs_info->generic_worker); | |
0cb59c99 JB |
2193 | btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write", |
2194 | 1, &fs_info->generic_worker); | |
16cdcec7 MX |
2195 | btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta", |
2196 | fs_info->thread_pool_size, | |
2197 | &fs_info->generic_worker); | |
90519d66 AJ |
2198 | btrfs_init_workers(&fs_info->readahead_workers, "readahead", |
2199 | fs_info->thread_pool_size, | |
2200 | &fs_info->generic_worker); | |
61b49440 CM |
2201 | |
2202 | /* | |
2203 | * endios are largely parallel and should have a very | |
2204 | * low idle thresh | |
2205 | */ | |
2206 | fs_info->endio_workers.idle_thresh = 4; | |
b51912c9 CM |
2207 | fs_info->endio_meta_workers.idle_thresh = 4; |
2208 | ||
9042846b CM |
2209 | fs_info->endio_write_workers.idle_thresh = 2; |
2210 | fs_info->endio_meta_write_workers.idle_thresh = 2; | |
90519d66 | 2211 | fs_info->readahead_workers.idle_thresh = 2; |
9042846b | 2212 | |
0dc3b84a JB |
2213 | /* |
2214 | * btrfs_start_workers can really only fail because of ENOMEM so just | |
2215 | * return -ENOMEM if any of these fail. | |
2216 | */ | |
2217 | ret = btrfs_start_workers(&fs_info->workers); | |
2218 | ret |= btrfs_start_workers(&fs_info->generic_worker); | |
2219 | ret |= btrfs_start_workers(&fs_info->submit_workers); | |
2220 | ret |= btrfs_start_workers(&fs_info->delalloc_workers); | |
2221 | ret |= btrfs_start_workers(&fs_info->fixup_workers); | |
2222 | ret |= btrfs_start_workers(&fs_info->endio_workers); | |
2223 | ret |= btrfs_start_workers(&fs_info->endio_meta_workers); | |
2224 | ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers); | |
2225 | ret |= btrfs_start_workers(&fs_info->endio_write_workers); | |
2226 | ret |= btrfs_start_workers(&fs_info->endio_freespace_worker); | |
2227 | ret |= btrfs_start_workers(&fs_info->delayed_workers); | |
2228 | ret |= btrfs_start_workers(&fs_info->caching_workers); | |
2229 | ret |= btrfs_start_workers(&fs_info->readahead_workers); | |
2230 | if (ret) { | |
2231 | ret = -ENOMEM; | |
2232 | goto fail_sb_buffer; | |
2233 | } | |
4543df7e | 2234 | |
4575c9cc | 2235 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
2236 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
2237 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 2238 | |
db94535d CM |
2239 | nodesize = btrfs_super_nodesize(disk_super); |
2240 | leafsize = btrfs_super_leafsize(disk_super); | |
2241 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 2242 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
2243 | tree_root->nodesize = nodesize; |
2244 | tree_root->leafsize = leafsize; | |
2245 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 2246 | tree_root->stripesize = stripesize; |
a061fc8d CM |
2247 | |
2248 | sb->s_blocksize = sectorsize; | |
2249 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 2250 | |
39279cc3 CM |
2251 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
2252 | sizeof(disk_super->magic))) { | |
d397712b | 2253 | printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
2254 | goto fail_sb_buffer; |
2255 | } | |
19c00ddc | 2256 | |
941b2ddf KM |
2257 | if (sectorsize < PAGE_SIZE) { |
2258 | printk(KERN_WARNING "btrfs: Incompatible sector size " | |
2259 | "found on %s\n", sb->s_id); | |
2260 | goto fail_sb_buffer; | |
2261 | } | |
2262 | ||
925baedd | 2263 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 2264 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 2265 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2266 | if (ret) { |
d397712b CM |
2267 | printk(KERN_WARNING "btrfs: failed to read the system " |
2268 | "array on %s\n", sb->s_id); | |
5d4f98a2 | 2269 | goto fail_sb_buffer; |
84eed90f | 2270 | } |
0b86a832 CM |
2271 | |
2272 | blocksize = btrfs_level_size(tree_root, | |
2273 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 2274 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
2275 | |
2276 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2277 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
2278 | ||
2279 | chunk_root->node = read_tree_block(chunk_root, | |
2280 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 2281 | blocksize, generation); |
0b86a832 | 2282 | BUG_ON(!chunk_root->node); |
83121942 DW |
2283 | if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { |
2284 | printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n", | |
2285 | sb->s_id); | |
af31f5e5 | 2286 | goto fail_tree_roots; |
83121942 | 2287 | } |
5d4f98a2 YZ |
2288 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
2289 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 2290 | |
e17cade2 | 2291 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
d397712b CM |
2292 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), |
2293 | BTRFS_UUID_SIZE); | |
e17cade2 | 2294 | |
0b86a832 | 2295 | ret = btrfs_read_chunk_tree(chunk_root); |
2b82032c | 2296 | if (ret) { |
d397712b CM |
2297 | printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", |
2298 | sb->s_id); | |
af31f5e5 | 2299 | goto fail_tree_roots; |
2b82032c | 2300 | } |
0b86a832 | 2301 | |
dfe25020 CM |
2302 | btrfs_close_extra_devices(fs_devices); |
2303 | ||
a6b0d5c8 CM |
2304 | if (!fs_devices->latest_bdev) { |
2305 | printk(KERN_CRIT "btrfs: failed to read devices on %s\n", | |
2306 | sb->s_id); | |
2307 | goto fail_tree_roots; | |
2308 | } | |
2309 | ||
af31f5e5 | 2310 | retry_root_backup: |
db94535d CM |
2311 | blocksize = btrfs_level_size(tree_root, |
2312 | btrfs_super_root_level(disk_super)); | |
84234f3a | 2313 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 2314 | |
e20d96d6 | 2315 | tree_root->node = read_tree_block(tree_root, |
db94535d | 2316 | btrfs_super_root(disk_super), |
84234f3a | 2317 | blocksize, generation); |
af31f5e5 CM |
2318 | if (!tree_root->node || |
2319 | !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { | |
83121942 DW |
2320 | printk(KERN_WARNING "btrfs: failed to read tree root on %s\n", |
2321 | sb->s_id); | |
af31f5e5 CM |
2322 | |
2323 | goto recovery_tree_root; | |
83121942 | 2324 | } |
af31f5e5 | 2325 | |
5d4f98a2 YZ |
2326 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
2327 | tree_root->commit_root = btrfs_root_node(tree_root); | |
db94535d CM |
2328 | |
2329 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 2330 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 2331 | if (ret) |
af31f5e5 | 2332 | goto recovery_tree_root; |
0b86a832 CM |
2333 | extent_root->track_dirty = 1; |
2334 | ||
2335 | ret = find_and_setup_root(tree_root, fs_info, | |
2336 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
0b86a832 | 2337 | if (ret) |
af31f5e5 | 2338 | goto recovery_tree_root; |
5d4f98a2 | 2339 | dev_root->track_dirty = 1; |
3768f368 | 2340 | |
d20f7043 CM |
2341 | ret = find_and_setup_root(tree_root, fs_info, |
2342 | BTRFS_CSUM_TREE_OBJECTID, csum_root); | |
2343 | if (ret) | |
af31f5e5 | 2344 | goto recovery_tree_root; |
d20f7043 CM |
2345 | |
2346 | csum_root->track_dirty = 1; | |
2347 | ||
8929ecfa YZ |
2348 | fs_info->generation = generation; |
2349 | fs_info->last_trans_committed = generation; | |
8929ecfa | 2350 | |
c59021f8 | 2351 | ret = btrfs_init_space_info(fs_info); |
2352 | if (ret) { | |
2353 | printk(KERN_ERR "Failed to initial space info: %d\n", ret); | |
2354 | goto fail_block_groups; | |
2355 | } | |
2356 | ||
1b1d1f66 JB |
2357 | ret = btrfs_read_block_groups(extent_root); |
2358 | if (ret) { | |
2359 | printk(KERN_ERR "Failed to read block groups: %d\n", ret); | |
2360 | goto fail_block_groups; | |
2361 | } | |
9078a3e1 | 2362 | |
a74a4b97 CM |
2363 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
2364 | "btrfs-cleaner"); | |
57506d50 | 2365 | if (IS_ERR(fs_info->cleaner_kthread)) |
1b1d1f66 | 2366 | goto fail_block_groups; |
a74a4b97 CM |
2367 | |
2368 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
2369 | tree_root, | |
2370 | "btrfs-transaction"); | |
57506d50 | 2371 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 2372 | goto fail_cleaner; |
a74a4b97 | 2373 | |
c289811c CM |
2374 | if (!btrfs_test_opt(tree_root, SSD) && |
2375 | !btrfs_test_opt(tree_root, NOSSD) && | |
2376 | !fs_info->fs_devices->rotating) { | |
2377 | printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD " | |
2378 | "mode\n"); | |
2379 | btrfs_set_opt(fs_info->mount_opt, SSD); | |
2380 | } | |
2381 | ||
21adbd5c SB |
2382 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2383 | if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { | |
2384 | ret = btrfsic_mount(tree_root, fs_devices, | |
2385 | btrfs_test_opt(tree_root, | |
2386 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? | |
2387 | 1 : 0, | |
2388 | fs_info->check_integrity_print_mask); | |
2389 | if (ret) | |
2390 | printk(KERN_WARNING "btrfs: failed to initialize" | |
2391 | " integrity check module %s\n", sb->s_id); | |
2392 | } | |
2393 | #endif | |
2394 | ||
acce952b | 2395 | /* do not make disk changes in broken FS */ |
2396 | if (btrfs_super_log_root(disk_super) != 0 && | |
2397 | !(fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)) { | |
e02119d5 CM |
2398 | u64 bytenr = btrfs_super_log_root(disk_super); |
2399 | ||
7c2ca468 | 2400 | if (fs_devices->rw_devices == 0) { |
d397712b CM |
2401 | printk(KERN_WARNING "Btrfs log replay required " |
2402 | "on RO media\n"); | |
7c2ca468 CM |
2403 | err = -EIO; |
2404 | goto fail_trans_kthread; | |
2405 | } | |
e02119d5 CM |
2406 | blocksize = |
2407 | btrfs_level_size(tree_root, | |
2408 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 2409 | |
6f07e42e | 2410 | log_tree_root = btrfs_alloc_root(fs_info); |
676e4c86 DC |
2411 | if (!log_tree_root) { |
2412 | err = -ENOMEM; | |
2413 | goto fail_trans_kthread; | |
2414 | } | |
e02119d5 CM |
2415 | |
2416 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2417 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
2418 | ||
2419 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
2420 | blocksize, |
2421 | generation + 1); | |
e02119d5 CM |
2422 | ret = btrfs_recover_log_trees(log_tree_root); |
2423 | BUG_ON(ret); | |
e556ce2c YZ |
2424 | |
2425 | if (sb->s_flags & MS_RDONLY) { | |
2426 | ret = btrfs_commit_super(tree_root); | |
2427 | BUG_ON(ret); | |
2428 | } | |
e02119d5 | 2429 | } |
1a40e23b | 2430 | |
76dda93c YZ |
2431 | ret = btrfs_find_orphan_roots(tree_root); |
2432 | BUG_ON(ret); | |
2433 | ||
7c2ca468 | 2434 | if (!(sb->s_flags & MS_RDONLY)) { |
d68fc57b YZ |
2435 | ret = btrfs_cleanup_fs_roots(fs_info); |
2436 | BUG_ON(ret); | |
2437 | ||
5d4f98a2 | 2438 | ret = btrfs_recover_relocation(tree_root); |
d7ce5843 MX |
2439 | if (ret < 0) { |
2440 | printk(KERN_WARNING | |
2441 | "btrfs: failed to recover relocation\n"); | |
2442 | err = -EINVAL; | |
2443 | goto fail_trans_kthread; | |
2444 | } | |
7c2ca468 | 2445 | } |
1a40e23b | 2446 | |
3de4586c CM |
2447 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
2448 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2449 | location.offset = (u64)-1; | |
2450 | ||
3de4586c CM |
2451 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
2452 | if (!fs_info->fs_root) | |
7c2ca468 | 2453 | goto fail_trans_kthread; |
3140c9a3 DC |
2454 | if (IS_ERR(fs_info->fs_root)) { |
2455 | err = PTR_ERR(fs_info->fs_root); | |
2456 | goto fail_trans_kthread; | |
2457 | } | |
c289811c | 2458 | |
e3acc2a6 JB |
2459 | if (!(sb->s_flags & MS_RDONLY)) { |
2460 | down_read(&fs_info->cleanup_work_sem); | |
66b4ffd1 JB |
2461 | err = btrfs_orphan_cleanup(fs_info->fs_root); |
2462 | if (!err) | |
2463 | err = btrfs_orphan_cleanup(fs_info->tree_root); | |
e3acc2a6 | 2464 | up_read(&fs_info->cleanup_work_sem); |
59641015 ID |
2465 | |
2466 | if (!err) | |
2467 | err = btrfs_recover_balance(fs_info->tree_root); | |
2468 | ||
66b4ffd1 JB |
2469 | if (err) { |
2470 | close_ctree(tree_root); | |
ad2b2c80 | 2471 | return err; |
66b4ffd1 | 2472 | } |
e3acc2a6 JB |
2473 | } |
2474 | ||
ad2b2c80 | 2475 | return 0; |
39279cc3 | 2476 | |
7c2ca468 CM |
2477 | fail_trans_kthread: |
2478 | kthread_stop(fs_info->transaction_kthread); | |
3f157a2f | 2479 | fail_cleaner: |
a74a4b97 | 2480 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
2481 | |
2482 | /* | |
2483 | * make sure we're done with the btree inode before we stop our | |
2484 | * kthreads | |
2485 | */ | |
2486 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
2487 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
2488 | ||
1b1d1f66 JB |
2489 | fail_block_groups: |
2490 | btrfs_free_block_groups(fs_info); | |
af31f5e5 CM |
2491 | |
2492 | fail_tree_roots: | |
2493 | free_root_pointers(fs_info, 1); | |
2494 | ||
39279cc3 | 2495 | fail_sb_buffer: |
61d92c32 | 2496 | btrfs_stop_workers(&fs_info->generic_worker); |
306c8b68 | 2497 | btrfs_stop_workers(&fs_info->readahead_workers); |
247e743c | 2498 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2499 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2500 | btrfs_stop_workers(&fs_info->workers); |
2501 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 2502 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 2503 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 2504 | btrfs_stop_workers(&fs_info->endio_write_workers); |
0cb59c99 | 2505 | btrfs_stop_workers(&fs_info->endio_freespace_worker); |
1cc127b5 | 2506 | btrfs_stop_workers(&fs_info->submit_workers); |
16cdcec7 | 2507 | btrfs_stop_workers(&fs_info->delayed_workers); |
bab39bf9 | 2508 | btrfs_stop_workers(&fs_info->caching_workers); |
16cdcec7 | 2509 | fail_alloc: |
4543df7e | 2510 | fail_iput: |
586e46e2 ID |
2511 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
2512 | ||
7c2ca468 | 2513 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
4543df7e | 2514 | iput(fs_info->btree_inode); |
ad081f14 | 2515 | fail_bdi: |
7e662854 | 2516 | bdi_destroy(&fs_info->bdi); |
76dda93c YZ |
2517 | fail_srcu: |
2518 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 2519 | fail: |
586e46e2 | 2520 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 2521 | return err; |
af31f5e5 CM |
2522 | |
2523 | recovery_tree_root: | |
af31f5e5 CM |
2524 | if (!btrfs_test_opt(tree_root, RECOVERY)) |
2525 | goto fail_tree_roots; | |
2526 | ||
2527 | free_root_pointers(fs_info, 0); | |
2528 | ||
2529 | /* don't use the log in recovery mode, it won't be valid */ | |
2530 | btrfs_set_super_log_root(disk_super, 0); | |
2531 | ||
2532 | /* we can't trust the free space cache either */ | |
2533 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2534 | ||
2535 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
2536 | &num_backups_tried, &backup_index); | |
2537 | if (ret == -1) | |
2538 | goto fail_block_groups; | |
2539 | goto retry_root_backup; | |
eb60ceac CM |
2540 | } |
2541 | ||
f2984462 CM |
2542 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
2543 | { | |
2544 | char b[BDEVNAME_SIZE]; | |
2545 | ||
2546 | if (uptodate) { | |
2547 | set_buffer_uptodate(bh); | |
2548 | } else { | |
7a36ddec | 2549 | printk_ratelimited(KERN_WARNING "lost page write due to " |
f2984462 CM |
2550 | "I/O error on %s\n", |
2551 | bdevname(bh->b_bdev, b)); | |
1259ab75 CM |
2552 | /* note, we dont' set_buffer_write_io_error because we have |
2553 | * our own ways of dealing with the IO errors | |
2554 | */ | |
f2984462 CM |
2555 | clear_buffer_uptodate(bh); |
2556 | } | |
2557 | unlock_buffer(bh); | |
2558 | put_bh(bh); | |
2559 | } | |
2560 | ||
a512bbf8 YZ |
2561 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
2562 | { | |
2563 | struct buffer_head *bh; | |
2564 | struct buffer_head *latest = NULL; | |
2565 | struct btrfs_super_block *super; | |
2566 | int i; | |
2567 | u64 transid = 0; | |
2568 | u64 bytenr; | |
2569 | ||
2570 | /* we would like to check all the supers, but that would make | |
2571 | * a btrfs mount succeed after a mkfs from a different FS. | |
2572 | * So, we need to add a special mount option to scan for | |
2573 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
2574 | */ | |
2575 | for (i = 0; i < 1; i++) { | |
2576 | bytenr = btrfs_sb_offset(i); | |
2577 | if (bytenr + 4096 >= i_size_read(bdev->bd_inode)) | |
2578 | break; | |
2579 | bh = __bread(bdev, bytenr / 4096, 4096); | |
2580 | if (!bh) | |
2581 | continue; | |
2582 | ||
2583 | super = (struct btrfs_super_block *)bh->b_data; | |
2584 | if (btrfs_super_bytenr(super) != bytenr || | |
2585 | strncmp((char *)(&super->magic), BTRFS_MAGIC, | |
2586 | sizeof(super->magic))) { | |
2587 | brelse(bh); | |
2588 | continue; | |
2589 | } | |
2590 | ||
2591 | if (!latest || btrfs_super_generation(super) > transid) { | |
2592 | brelse(latest); | |
2593 | latest = bh; | |
2594 | transid = btrfs_super_generation(super); | |
2595 | } else { | |
2596 | brelse(bh); | |
2597 | } | |
2598 | } | |
2599 | return latest; | |
2600 | } | |
2601 | ||
4eedeb75 HH |
2602 | /* |
2603 | * this should be called twice, once with wait == 0 and | |
2604 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
2605 | * we write are pinned. | |
2606 | * | |
2607 | * They are released when wait == 1 is done. | |
2608 | * max_mirrors must be the same for both runs, and it indicates how | |
2609 | * many supers on this one device should be written. | |
2610 | * | |
2611 | * max_mirrors == 0 means to write them all. | |
2612 | */ | |
a512bbf8 YZ |
2613 | static int write_dev_supers(struct btrfs_device *device, |
2614 | struct btrfs_super_block *sb, | |
2615 | int do_barriers, int wait, int max_mirrors) | |
2616 | { | |
2617 | struct buffer_head *bh; | |
2618 | int i; | |
2619 | int ret; | |
2620 | int errors = 0; | |
2621 | u32 crc; | |
2622 | u64 bytenr; | |
a512bbf8 YZ |
2623 | |
2624 | if (max_mirrors == 0) | |
2625 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
2626 | ||
a512bbf8 YZ |
2627 | for (i = 0; i < max_mirrors; i++) { |
2628 | bytenr = btrfs_sb_offset(i); | |
2629 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) | |
2630 | break; | |
2631 | ||
2632 | if (wait) { | |
2633 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
2634 | BTRFS_SUPER_INFO_SIZE); | |
2635 | BUG_ON(!bh); | |
a512bbf8 | 2636 | wait_on_buffer(bh); |
4eedeb75 HH |
2637 | if (!buffer_uptodate(bh)) |
2638 | errors++; | |
2639 | ||
2640 | /* drop our reference */ | |
2641 | brelse(bh); | |
2642 | ||
2643 | /* drop the reference from the wait == 0 run */ | |
2644 | brelse(bh); | |
2645 | continue; | |
a512bbf8 YZ |
2646 | } else { |
2647 | btrfs_set_super_bytenr(sb, bytenr); | |
2648 | ||
2649 | crc = ~(u32)0; | |
2650 | crc = btrfs_csum_data(NULL, (char *)sb + | |
2651 | BTRFS_CSUM_SIZE, crc, | |
2652 | BTRFS_SUPER_INFO_SIZE - | |
2653 | BTRFS_CSUM_SIZE); | |
2654 | btrfs_csum_final(crc, sb->csum); | |
2655 | ||
4eedeb75 HH |
2656 | /* |
2657 | * one reference for us, and we leave it for the | |
2658 | * caller | |
2659 | */ | |
a512bbf8 YZ |
2660 | bh = __getblk(device->bdev, bytenr / 4096, |
2661 | BTRFS_SUPER_INFO_SIZE); | |
2662 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); | |
2663 | ||
4eedeb75 | 2664 | /* one reference for submit_bh */ |
a512bbf8 | 2665 | get_bh(bh); |
4eedeb75 HH |
2666 | |
2667 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
2668 | lock_buffer(bh); |
2669 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
2670 | } | |
2671 | ||
387125fc CM |
2672 | /* |
2673 | * we fua the first super. The others we allow | |
2674 | * to go down lazy. | |
2675 | */ | |
21adbd5c | 2676 | ret = btrfsic_submit_bh(WRITE_FUA, bh); |
4eedeb75 | 2677 | if (ret) |
a512bbf8 | 2678 | errors++; |
a512bbf8 YZ |
2679 | } |
2680 | return errors < i ? 0 : -1; | |
2681 | } | |
2682 | ||
387125fc CM |
2683 | /* |
2684 | * endio for the write_dev_flush, this will wake anyone waiting | |
2685 | * for the barrier when it is done | |
2686 | */ | |
2687 | static void btrfs_end_empty_barrier(struct bio *bio, int err) | |
2688 | { | |
2689 | if (err) { | |
2690 | if (err == -EOPNOTSUPP) | |
2691 | set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); | |
2692 | clear_bit(BIO_UPTODATE, &bio->bi_flags); | |
2693 | } | |
2694 | if (bio->bi_private) | |
2695 | complete(bio->bi_private); | |
2696 | bio_put(bio); | |
2697 | } | |
2698 | ||
2699 | /* | |
2700 | * trigger flushes for one the devices. If you pass wait == 0, the flushes are | |
2701 | * sent down. With wait == 1, it waits for the previous flush. | |
2702 | * | |
2703 | * any device where the flush fails with eopnotsupp are flagged as not-barrier | |
2704 | * capable | |
2705 | */ | |
2706 | static int write_dev_flush(struct btrfs_device *device, int wait) | |
2707 | { | |
2708 | struct bio *bio; | |
2709 | int ret = 0; | |
2710 | ||
2711 | if (device->nobarriers) | |
2712 | return 0; | |
2713 | ||
2714 | if (wait) { | |
2715 | bio = device->flush_bio; | |
2716 | if (!bio) | |
2717 | return 0; | |
2718 | ||
2719 | wait_for_completion(&device->flush_wait); | |
2720 | ||
2721 | if (bio_flagged(bio, BIO_EOPNOTSUPP)) { | |
2722 | printk("btrfs: disabling barriers on dev %s\n", | |
2723 | device->name); | |
2724 | device->nobarriers = 1; | |
2725 | } | |
2726 | if (!bio_flagged(bio, BIO_UPTODATE)) { | |
2727 | ret = -EIO; | |
2728 | } | |
2729 | ||
2730 | /* drop the reference from the wait == 0 run */ | |
2731 | bio_put(bio); | |
2732 | device->flush_bio = NULL; | |
2733 | ||
2734 | return ret; | |
2735 | } | |
2736 | ||
2737 | /* | |
2738 | * one reference for us, and we leave it for the | |
2739 | * caller | |
2740 | */ | |
2741 | device->flush_bio = NULL;; | |
2742 | bio = bio_alloc(GFP_NOFS, 0); | |
2743 | if (!bio) | |
2744 | return -ENOMEM; | |
2745 | ||
2746 | bio->bi_end_io = btrfs_end_empty_barrier; | |
2747 | bio->bi_bdev = device->bdev; | |
2748 | init_completion(&device->flush_wait); | |
2749 | bio->bi_private = &device->flush_wait; | |
2750 | device->flush_bio = bio; | |
2751 | ||
2752 | bio_get(bio); | |
21adbd5c | 2753 | btrfsic_submit_bio(WRITE_FLUSH, bio); |
387125fc CM |
2754 | |
2755 | return 0; | |
2756 | } | |
2757 | ||
2758 | /* | |
2759 | * send an empty flush down to each device in parallel, | |
2760 | * then wait for them | |
2761 | */ | |
2762 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
2763 | { | |
2764 | struct list_head *head; | |
2765 | struct btrfs_device *dev; | |
2766 | int errors = 0; | |
2767 | int ret; | |
2768 | ||
2769 | /* send down all the barriers */ | |
2770 | head = &info->fs_devices->devices; | |
2771 | list_for_each_entry_rcu(dev, head, dev_list) { | |
2772 | if (!dev->bdev) { | |
2773 | errors++; | |
2774 | continue; | |
2775 | } | |
2776 | if (!dev->in_fs_metadata || !dev->writeable) | |
2777 | continue; | |
2778 | ||
2779 | ret = write_dev_flush(dev, 0); | |
2780 | if (ret) | |
2781 | errors++; | |
2782 | } | |
2783 | ||
2784 | /* wait for all the barriers */ | |
2785 | list_for_each_entry_rcu(dev, head, dev_list) { | |
2786 | if (!dev->bdev) { | |
2787 | errors++; | |
2788 | continue; | |
2789 | } | |
2790 | if (!dev->in_fs_metadata || !dev->writeable) | |
2791 | continue; | |
2792 | ||
2793 | ret = write_dev_flush(dev, 1); | |
2794 | if (ret) | |
2795 | errors++; | |
2796 | } | |
2797 | if (errors) | |
2798 | return -EIO; | |
2799 | return 0; | |
2800 | } | |
2801 | ||
a512bbf8 | 2802 | int write_all_supers(struct btrfs_root *root, int max_mirrors) |
f2984462 | 2803 | { |
e5e9a520 | 2804 | struct list_head *head; |
f2984462 | 2805 | struct btrfs_device *dev; |
a061fc8d | 2806 | struct btrfs_super_block *sb; |
f2984462 | 2807 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
2808 | int ret; |
2809 | int do_barriers; | |
a236aed1 CM |
2810 | int max_errors; |
2811 | int total_errors = 0; | |
a061fc8d | 2812 | u64 flags; |
f2984462 | 2813 | |
6c41761f | 2814 | max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; |
f2984462 | 2815 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
af31f5e5 | 2816 | backup_super_roots(root->fs_info); |
f2984462 | 2817 | |
6c41761f | 2818 | sb = root->fs_info->super_for_commit; |
a061fc8d | 2819 | dev_item = &sb->dev_item; |
e5e9a520 | 2820 | |
174ba509 | 2821 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); |
e5e9a520 | 2822 | head = &root->fs_info->fs_devices->devices; |
387125fc CM |
2823 | |
2824 | if (do_barriers) | |
2825 | barrier_all_devices(root->fs_info); | |
2826 | ||
1f78160c | 2827 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
2828 | if (!dev->bdev) { |
2829 | total_errors++; | |
2830 | continue; | |
2831 | } | |
2b82032c | 2832 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2833 | continue; |
2834 | ||
2b82032c | 2835 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
2836 | btrfs_set_stack_device_type(dev_item, dev->type); |
2837 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
2838 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
2839 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
2840 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
2841 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
2842 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
2843 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 2844 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 2845 | |
a061fc8d CM |
2846 | flags = btrfs_super_flags(sb); |
2847 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
2848 | ||
a512bbf8 | 2849 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
2850 | if (ret) |
2851 | total_errors++; | |
f2984462 | 2852 | } |
a236aed1 | 2853 | if (total_errors > max_errors) { |
d397712b CM |
2854 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2855 | total_errors); | |
a236aed1 CM |
2856 | BUG(); |
2857 | } | |
f2984462 | 2858 | |
a512bbf8 | 2859 | total_errors = 0; |
1f78160c | 2860 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
2861 | if (!dev->bdev) |
2862 | continue; | |
2b82032c | 2863 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2864 | continue; |
2865 | ||
a512bbf8 YZ |
2866 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
2867 | if (ret) | |
2868 | total_errors++; | |
f2984462 | 2869 | } |
174ba509 | 2870 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 2871 | if (total_errors > max_errors) { |
d397712b CM |
2872 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2873 | total_errors); | |
a236aed1 CM |
2874 | BUG(); |
2875 | } | |
f2984462 CM |
2876 | return 0; |
2877 | } | |
2878 | ||
a512bbf8 YZ |
2879 | int write_ctree_super(struct btrfs_trans_handle *trans, |
2880 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 2881 | { |
e66f709b | 2882 | int ret; |
5f39d397 | 2883 | |
a512bbf8 | 2884 | ret = write_all_supers(root, max_mirrors); |
5f39d397 | 2885 | return ret; |
cfaa7295 CM |
2886 | } |
2887 | ||
5eda7b5e | 2888 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f | 2889 | { |
4df27c4d | 2890 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
2891 | radix_tree_delete(&fs_info->fs_roots_radix, |
2892 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 2893 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
2894 | |
2895 | if (btrfs_root_refs(&root->root_item) == 0) | |
2896 | synchronize_srcu(&fs_info->subvol_srcu); | |
2897 | ||
581bb050 LZ |
2898 | __btrfs_remove_free_space_cache(root->free_ino_pinned); |
2899 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d YZ |
2900 | free_fs_root(root); |
2901 | return 0; | |
2902 | } | |
2903 | ||
2904 | static void free_fs_root(struct btrfs_root *root) | |
2905 | { | |
82d5902d | 2906 | iput(root->cache_inode); |
4df27c4d | 2907 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
0ee5dc67 AV |
2908 | if (root->anon_dev) |
2909 | free_anon_bdev(root->anon_dev); | |
4df27c4d YZ |
2910 | free_extent_buffer(root->node); |
2911 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
2912 | kfree(root->free_ino_ctl); |
2913 | kfree(root->free_ino_pinned); | |
d397712b | 2914 | kfree(root->name); |
2619ba1f | 2915 | kfree(root); |
2619ba1f CM |
2916 | } |
2917 | ||
35b7e476 | 2918 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
2919 | { |
2920 | int ret; | |
2921 | struct btrfs_root *gang[8]; | |
2922 | int i; | |
2923 | ||
76dda93c YZ |
2924 | while (!list_empty(&fs_info->dead_roots)) { |
2925 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2926 | struct btrfs_root, root_list); | |
2927 | list_del(&gang[0]->root_list); | |
2928 | ||
2929 | if (gang[0]->in_radix) { | |
2930 | btrfs_free_fs_root(fs_info, gang[0]); | |
2931 | } else { | |
2932 | free_extent_buffer(gang[0]->node); | |
2933 | free_extent_buffer(gang[0]->commit_root); | |
2934 | kfree(gang[0]); | |
2935 | } | |
2936 | } | |
2937 | ||
d397712b | 2938 | while (1) { |
0f7d52f4 CM |
2939 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
2940 | (void **)gang, 0, | |
2941 | ARRAY_SIZE(gang)); | |
2942 | if (!ret) | |
2943 | break; | |
2619ba1f | 2944 | for (i = 0; i < ret; i++) |
5eda7b5e | 2945 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
2946 | } |
2947 | return 0; | |
2948 | } | |
b4100d64 | 2949 | |
c146afad | 2950 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 2951 | { |
c146afad YZ |
2952 | u64 root_objectid = 0; |
2953 | struct btrfs_root *gang[8]; | |
2954 | int i; | |
3768f368 | 2955 | int ret; |
e089f05c | 2956 | |
c146afad YZ |
2957 | while (1) { |
2958 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2959 | (void **)gang, root_objectid, | |
2960 | ARRAY_SIZE(gang)); | |
2961 | if (!ret) | |
2962 | break; | |
5d4f98a2 YZ |
2963 | |
2964 | root_objectid = gang[ret - 1]->root_key.objectid + 1; | |
c146afad | 2965 | for (i = 0; i < ret; i++) { |
66b4ffd1 JB |
2966 | int err; |
2967 | ||
c146afad | 2968 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
2969 | err = btrfs_orphan_cleanup(gang[i]); |
2970 | if (err) | |
2971 | return err; | |
c146afad YZ |
2972 | } |
2973 | root_objectid++; | |
2974 | } | |
2975 | return 0; | |
2976 | } | |
a2135011 | 2977 | |
c146afad YZ |
2978 | int btrfs_commit_super(struct btrfs_root *root) |
2979 | { | |
2980 | struct btrfs_trans_handle *trans; | |
2981 | int ret; | |
a74a4b97 | 2982 | |
c146afad | 2983 | mutex_lock(&root->fs_info->cleaner_mutex); |
24bbcf04 | 2984 | btrfs_run_delayed_iputs(root); |
a74a4b97 | 2985 | btrfs_clean_old_snapshots(root); |
c146afad | 2986 | mutex_unlock(&root->fs_info->cleaner_mutex); |
c71bf099 YZ |
2987 | |
2988 | /* wait until ongoing cleanup work done */ | |
2989 | down_write(&root->fs_info->cleanup_work_sem); | |
2990 | up_write(&root->fs_info->cleanup_work_sem); | |
2991 | ||
7a7eaa40 | 2992 | trans = btrfs_join_transaction(root); |
3612b495 TI |
2993 | if (IS_ERR(trans)) |
2994 | return PTR_ERR(trans); | |
54aa1f4d | 2995 | ret = btrfs_commit_transaction(trans, root); |
c146afad YZ |
2996 | BUG_ON(ret); |
2997 | /* run commit again to drop the original snapshot */ | |
7a7eaa40 | 2998 | trans = btrfs_join_transaction(root); |
3612b495 TI |
2999 | if (IS_ERR(trans)) |
3000 | return PTR_ERR(trans); | |
79154b1b CM |
3001 | btrfs_commit_transaction(trans, root); |
3002 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 3003 | BUG_ON(ret); |
d6bfde87 | 3004 | |
a512bbf8 | 3005 | ret = write_ctree_super(NULL, root, 0); |
c146afad YZ |
3006 | return ret; |
3007 | } | |
3008 | ||
3009 | int close_ctree(struct btrfs_root *root) | |
3010 | { | |
3011 | struct btrfs_fs_info *fs_info = root->fs_info; | |
3012 | int ret; | |
3013 | ||
3014 | fs_info->closing = 1; | |
3015 | smp_mb(); | |
3016 | ||
837d5b6e ID |
3017 | /* pause restriper - we want to resume on mount */ |
3018 | btrfs_pause_balance(root->fs_info); | |
3019 | ||
a2de733c | 3020 | btrfs_scrub_cancel(root); |
4cb5300b CM |
3021 | |
3022 | /* wait for any defraggers to finish */ | |
3023 | wait_event(fs_info->transaction_wait, | |
3024 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3025 | ||
3026 | /* clear out the rbtree of defraggable inodes */ | |
e3029d9f | 3027 | btrfs_run_defrag_inodes(fs_info); |
4cb5300b | 3028 | |
acce952b | 3029 | /* |
3030 | * Here come 2 situations when btrfs is broken to flip readonly: | |
3031 | * | |
3032 | * 1. when btrfs flips readonly somewhere else before | |
3033 | * btrfs_commit_super, sb->s_flags has MS_RDONLY flag, | |
3034 | * and btrfs will skip to write sb directly to keep | |
3035 | * ERROR state on disk. | |
3036 | * | |
3037 | * 2. when btrfs flips readonly just in btrfs_commit_super, | |
ae0e47f0 | 3038 | * and in such case, btrfs cannot write sb via btrfs_commit_super, |
acce952b | 3039 | * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag, |
3040 | * btrfs will cleanup all FS resources first and write sb then. | |
3041 | */ | |
c146afad | 3042 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
acce952b | 3043 | ret = btrfs_commit_super(root); |
3044 | if (ret) | |
3045 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
3046 | } | |
3047 | ||
3048 | if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) { | |
3049 | ret = btrfs_error_commit_super(root); | |
d397712b CM |
3050 | if (ret) |
3051 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
c146afad | 3052 | } |
0f7d52f4 | 3053 | |
300e4f8a JB |
3054 | btrfs_put_block_group_cache(fs_info); |
3055 | ||
e3029d9f AV |
3056 | kthread_stop(fs_info->transaction_kthread); |
3057 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 3058 | |
f25784b3 YZ |
3059 | fs_info->closing = 2; |
3060 | smp_mb(); | |
3061 | ||
b0c68f8b | 3062 | if (fs_info->delalloc_bytes) { |
d397712b | 3063 | printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n", |
21380931 | 3064 | (unsigned long long)fs_info->delalloc_bytes); |
b0c68f8b | 3065 | } |
31153d81 | 3066 | if (fs_info->total_ref_cache_size) { |
d397712b CM |
3067 | printk(KERN_INFO "btrfs: at umount reference cache size %llu\n", |
3068 | (unsigned long long)fs_info->total_ref_cache_size); | |
31153d81 | 3069 | } |
bcc63abb | 3070 | |
5d4f98a2 YZ |
3071 | free_extent_buffer(fs_info->extent_root->node); |
3072 | free_extent_buffer(fs_info->extent_root->commit_root); | |
3073 | free_extent_buffer(fs_info->tree_root->node); | |
3074 | free_extent_buffer(fs_info->tree_root->commit_root); | |
e3029d9f AV |
3075 | free_extent_buffer(fs_info->chunk_root->node); |
3076 | free_extent_buffer(fs_info->chunk_root->commit_root); | |
3077 | free_extent_buffer(fs_info->dev_root->node); | |
3078 | free_extent_buffer(fs_info->dev_root->commit_root); | |
3079 | free_extent_buffer(fs_info->csum_root->node); | |
3080 | free_extent_buffer(fs_info->csum_root->commit_root); | |
d20f7043 | 3081 | |
e3029d9f | 3082 | btrfs_free_block_groups(fs_info); |
d10c5f31 | 3083 | |
c146afad | 3084 | del_fs_roots(fs_info); |
d10c5f31 | 3085 | |
c146afad | 3086 | iput(fs_info->btree_inode); |
9ad6b7bc | 3087 | |
61d92c32 | 3088 | btrfs_stop_workers(&fs_info->generic_worker); |
247e743c | 3089 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 3090 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
3091 | btrfs_stop_workers(&fs_info->workers); |
3092 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 3093 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 3094 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 3095 | btrfs_stop_workers(&fs_info->endio_write_workers); |
0cb59c99 | 3096 | btrfs_stop_workers(&fs_info->endio_freespace_worker); |
1cc127b5 | 3097 | btrfs_stop_workers(&fs_info->submit_workers); |
16cdcec7 | 3098 | btrfs_stop_workers(&fs_info->delayed_workers); |
bab39bf9 | 3099 | btrfs_stop_workers(&fs_info->caching_workers); |
90519d66 | 3100 | btrfs_stop_workers(&fs_info->readahead_workers); |
d6bfde87 | 3101 | |
21adbd5c SB |
3102 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3103 | if (btrfs_test_opt(root, CHECK_INTEGRITY)) | |
3104 | btrfsic_unmount(root, fs_info->fs_devices); | |
3105 | #endif | |
3106 | ||
dfe25020 | 3107 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 3108 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 3109 | |
04160088 | 3110 | bdi_destroy(&fs_info->bdi); |
76dda93c | 3111 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 3112 | |
eb60ceac CM |
3113 | return 0; |
3114 | } | |
3115 | ||
1259ab75 | 3116 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 3117 | { |
1259ab75 | 3118 | int ret; |
727011e0 | 3119 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 3120 | |
2ac55d41 JB |
3121 | ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf, |
3122 | NULL); | |
1259ab75 CM |
3123 | if (!ret) |
3124 | return ret; | |
3125 | ||
3126 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
3127 | parent_transid); | |
3128 | return !ret; | |
5f39d397 CM |
3129 | } |
3130 | ||
3131 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 3132 | { |
727011e0 | 3133 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
d1310b2e | 3134 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
3135 | buf); |
3136 | } | |
6702ed49 | 3137 | |
5f39d397 CM |
3138 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
3139 | { | |
727011e0 | 3140 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
5f39d397 CM |
3141 | u64 transid = btrfs_header_generation(buf); |
3142 | struct inode *btree_inode = root->fs_info->btree_inode; | |
b9473439 | 3143 | int was_dirty; |
b4ce94de | 3144 | |
b9447ef8 | 3145 | btrfs_assert_tree_locked(buf); |
ccd467d6 | 3146 | if (transid != root->fs_info->generation) { |
d397712b CM |
3147 | printk(KERN_CRIT "btrfs transid mismatch buffer %llu, " |
3148 | "found %llu running %llu\n", | |
db94535d | 3149 | (unsigned long long)buf->start, |
d397712b CM |
3150 | (unsigned long long)transid, |
3151 | (unsigned long long)root->fs_info->generation); | |
ccd467d6 CM |
3152 | WARN_ON(1); |
3153 | } | |
b9473439 CM |
3154 | was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
3155 | buf); | |
3156 | if (!was_dirty) { | |
3157 | spin_lock(&root->fs_info->delalloc_lock); | |
3158 | root->fs_info->dirty_metadata_bytes += buf->len; | |
3159 | spin_unlock(&root->fs_info->delalloc_lock); | |
3160 | } | |
eb60ceac CM |
3161 | } |
3162 | ||
d3c2fdcf | 3163 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
16cdcec7 MX |
3164 | { |
3165 | /* | |
3166 | * looks as though older kernels can get into trouble with | |
3167 | * this code, they end up stuck in balance_dirty_pages forever | |
3168 | */ | |
3169 | u64 num_dirty; | |
3170 | unsigned long thresh = 32 * 1024 * 1024; | |
3171 | ||
3172 | if (current->flags & PF_MEMALLOC) | |
3173 | return; | |
3174 | ||
3175 | btrfs_balance_delayed_items(root); | |
3176 | ||
3177 | num_dirty = root->fs_info->dirty_metadata_bytes; | |
3178 | ||
3179 | if (num_dirty > thresh) { | |
3180 | balance_dirty_pages_ratelimited_nr( | |
3181 | root->fs_info->btree_inode->i_mapping, 1); | |
3182 | } | |
3183 | return; | |
3184 | } | |
3185 | ||
3186 | void __btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) | |
35b7e476 | 3187 | { |
188de649 CM |
3188 | /* |
3189 | * looks as though older kernels can get into trouble with | |
3190 | * this code, they end up stuck in balance_dirty_pages forever | |
3191 | */ | |
d6bfde87 | 3192 | u64 num_dirty; |
771ed689 | 3193 | unsigned long thresh = 32 * 1024 * 1024; |
d6bfde87 | 3194 | |
6933c02e | 3195 | if (current->flags & PF_MEMALLOC) |
d6bfde87 CM |
3196 | return; |
3197 | ||
585ad2c3 CM |
3198 | num_dirty = root->fs_info->dirty_metadata_bytes; |
3199 | ||
d6bfde87 CM |
3200 | if (num_dirty > thresh) { |
3201 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 3202 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 3203 | } |
188de649 | 3204 | return; |
35b7e476 | 3205 | } |
6b80053d | 3206 | |
ca7a79ad | 3207 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 3208 | { |
727011e0 | 3209 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
ce9adaa5 | 3210 | int ret; |
ca7a79ad | 3211 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
d397712b | 3212 | if (ret == 0) |
b4ce94de | 3213 | set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags); |
ce9adaa5 | 3214 | return ret; |
6b80053d | 3215 | } |
0da5468f | 3216 | |
01d658f2 CM |
3217 | static int btree_lock_page_hook(struct page *page, void *data, |
3218 | void (*flush_fn)(void *)) | |
4bef0848 CM |
3219 | { |
3220 | struct inode *inode = page->mapping->host; | |
b9473439 | 3221 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4bef0848 CM |
3222 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
3223 | struct extent_buffer *eb; | |
3224 | unsigned long len; | |
3225 | u64 bytenr = page_offset(page); | |
3226 | ||
3227 | if (page->private == EXTENT_PAGE_PRIVATE) | |
3228 | goto out; | |
3229 | ||
3230 | len = page->private >> 2; | |
f09d1f60 | 3231 | eb = find_extent_buffer(io_tree, bytenr, len); |
4bef0848 CM |
3232 | if (!eb) |
3233 | goto out; | |
3234 | ||
01d658f2 CM |
3235 | if (!btrfs_try_tree_write_lock(eb)) { |
3236 | flush_fn(data); | |
3237 | btrfs_tree_lock(eb); | |
3238 | } | |
4bef0848 | 3239 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); |
b9473439 CM |
3240 | |
3241 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) { | |
3242 | spin_lock(&root->fs_info->delalloc_lock); | |
3243 | if (root->fs_info->dirty_metadata_bytes >= eb->len) | |
3244 | root->fs_info->dirty_metadata_bytes -= eb->len; | |
3245 | else | |
3246 | WARN_ON(1); | |
3247 | spin_unlock(&root->fs_info->delalloc_lock); | |
3248 | } | |
3249 | ||
4bef0848 CM |
3250 | btrfs_tree_unlock(eb); |
3251 | free_extent_buffer(eb); | |
3252 | out: | |
01d658f2 CM |
3253 | if (!trylock_page(page)) { |
3254 | flush_fn(data); | |
3255 | lock_page(page); | |
3256 | } | |
4bef0848 CM |
3257 | return 0; |
3258 | } | |
3259 | ||
acce952b | 3260 | static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
3261 | int read_only) | |
3262 | { | |
3263 | if (read_only) | |
3264 | return; | |
3265 | ||
3266 | if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) | |
3267 | printk(KERN_WARNING "warning: mount fs with errors, " | |
3268 | "running btrfsck is recommended\n"); | |
3269 | } | |
3270 | ||
3271 | int btrfs_error_commit_super(struct btrfs_root *root) | |
3272 | { | |
3273 | int ret; | |
3274 | ||
3275 | mutex_lock(&root->fs_info->cleaner_mutex); | |
3276 | btrfs_run_delayed_iputs(root); | |
3277 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
3278 | ||
3279 | down_write(&root->fs_info->cleanup_work_sem); | |
3280 | up_write(&root->fs_info->cleanup_work_sem); | |
3281 | ||
3282 | /* cleanup FS via transaction */ | |
3283 | btrfs_cleanup_transaction(root); | |
3284 | ||
3285 | ret = write_ctree_super(NULL, root, 0); | |
3286 | ||
3287 | return ret; | |
3288 | } | |
3289 | ||
3290 | static int btrfs_destroy_ordered_operations(struct btrfs_root *root) | |
3291 | { | |
3292 | struct btrfs_inode *btrfs_inode; | |
3293 | struct list_head splice; | |
3294 | ||
3295 | INIT_LIST_HEAD(&splice); | |
3296 | ||
3297 | mutex_lock(&root->fs_info->ordered_operations_mutex); | |
3298 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3299 | ||
3300 | list_splice_init(&root->fs_info->ordered_operations, &splice); | |
3301 | while (!list_empty(&splice)) { | |
3302 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3303 | ordered_operations); | |
3304 | ||
3305 | list_del_init(&btrfs_inode->ordered_operations); | |
3306 | ||
3307 | btrfs_invalidate_inodes(btrfs_inode->root); | |
3308 | } | |
3309 | ||
3310 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3311 | mutex_unlock(&root->fs_info->ordered_operations_mutex); | |
3312 | ||
3313 | return 0; | |
3314 | } | |
3315 | ||
3316 | static int btrfs_destroy_ordered_extents(struct btrfs_root *root) | |
3317 | { | |
3318 | struct list_head splice; | |
3319 | struct btrfs_ordered_extent *ordered; | |
3320 | struct inode *inode; | |
3321 | ||
3322 | INIT_LIST_HEAD(&splice); | |
3323 | ||
3324 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3325 | ||
3326 | list_splice_init(&root->fs_info->ordered_extents, &splice); | |
3327 | while (!list_empty(&splice)) { | |
3328 | ordered = list_entry(splice.next, struct btrfs_ordered_extent, | |
3329 | root_extent_list); | |
3330 | ||
3331 | list_del_init(&ordered->root_extent_list); | |
3332 | atomic_inc(&ordered->refs); | |
3333 | ||
3334 | /* the inode may be getting freed (in sys_unlink path). */ | |
3335 | inode = igrab(ordered->inode); | |
3336 | ||
3337 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3338 | if (inode) | |
3339 | iput(inode); | |
3340 | ||
3341 | atomic_set(&ordered->refs, 1); | |
3342 | btrfs_put_ordered_extent(ordered); | |
3343 | ||
3344 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3345 | } | |
3346 | ||
3347 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3348 | ||
3349 | return 0; | |
3350 | } | |
3351 | ||
3352 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, | |
3353 | struct btrfs_root *root) | |
3354 | { | |
3355 | struct rb_node *node; | |
3356 | struct btrfs_delayed_ref_root *delayed_refs; | |
3357 | struct btrfs_delayed_ref_node *ref; | |
3358 | int ret = 0; | |
3359 | ||
3360 | delayed_refs = &trans->delayed_refs; | |
3361 | ||
3362 | spin_lock(&delayed_refs->lock); | |
3363 | if (delayed_refs->num_entries == 0) { | |
cfece4db | 3364 | spin_unlock(&delayed_refs->lock); |
acce952b | 3365 | printk(KERN_INFO "delayed_refs has NO entry\n"); |
3366 | return ret; | |
3367 | } | |
3368 | ||
3369 | node = rb_first(&delayed_refs->root); | |
3370 | while (node) { | |
3371 | ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); | |
3372 | node = rb_next(node); | |
3373 | ||
3374 | ref->in_tree = 0; | |
3375 | rb_erase(&ref->rb_node, &delayed_refs->root); | |
3376 | delayed_refs->num_entries--; | |
3377 | ||
3378 | atomic_set(&ref->refs, 1); | |
3379 | if (btrfs_delayed_ref_is_head(ref)) { | |
3380 | struct btrfs_delayed_ref_head *head; | |
3381 | ||
3382 | head = btrfs_delayed_node_to_head(ref); | |
3383 | mutex_lock(&head->mutex); | |
3384 | kfree(head->extent_op); | |
3385 | delayed_refs->num_heads--; | |
3386 | if (list_empty(&head->cluster)) | |
3387 | delayed_refs->num_heads_ready--; | |
3388 | list_del_init(&head->cluster); | |
3389 | mutex_unlock(&head->mutex); | |
3390 | } | |
3391 | ||
3392 | spin_unlock(&delayed_refs->lock); | |
3393 | btrfs_put_delayed_ref(ref); | |
3394 | ||
3395 | cond_resched(); | |
3396 | spin_lock(&delayed_refs->lock); | |
3397 | } | |
3398 | ||
3399 | spin_unlock(&delayed_refs->lock); | |
3400 | ||
3401 | return ret; | |
3402 | } | |
3403 | ||
3404 | static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t) | |
3405 | { | |
3406 | struct btrfs_pending_snapshot *snapshot; | |
3407 | struct list_head splice; | |
3408 | ||
3409 | INIT_LIST_HEAD(&splice); | |
3410 | ||
3411 | list_splice_init(&t->pending_snapshots, &splice); | |
3412 | ||
3413 | while (!list_empty(&splice)) { | |
3414 | snapshot = list_entry(splice.next, | |
3415 | struct btrfs_pending_snapshot, | |
3416 | list); | |
3417 | ||
3418 | list_del_init(&snapshot->list); | |
3419 | ||
3420 | kfree(snapshot); | |
3421 | } | |
3422 | ||
3423 | return 0; | |
3424 | } | |
3425 | ||
3426 | static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root) | |
3427 | { | |
3428 | struct btrfs_inode *btrfs_inode; | |
3429 | struct list_head splice; | |
3430 | ||
3431 | INIT_LIST_HEAD(&splice); | |
3432 | ||
acce952b | 3433 | spin_lock(&root->fs_info->delalloc_lock); |
5be76758 | 3434 | list_splice_init(&root->fs_info->delalloc_inodes, &splice); |
acce952b | 3435 | |
3436 | while (!list_empty(&splice)) { | |
3437 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3438 | delalloc_inodes); | |
3439 | ||
3440 | list_del_init(&btrfs_inode->delalloc_inodes); | |
3441 | ||
3442 | btrfs_invalidate_inodes(btrfs_inode->root); | |
3443 | } | |
3444 | ||
3445 | spin_unlock(&root->fs_info->delalloc_lock); | |
3446 | ||
3447 | return 0; | |
3448 | } | |
3449 | ||
3450 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, | |
3451 | struct extent_io_tree *dirty_pages, | |
3452 | int mark) | |
3453 | { | |
3454 | int ret; | |
3455 | struct page *page; | |
3456 | struct inode *btree_inode = root->fs_info->btree_inode; | |
3457 | struct extent_buffer *eb; | |
3458 | u64 start = 0; | |
3459 | u64 end; | |
3460 | u64 offset; | |
3461 | unsigned long index; | |
3462 | ||
3463 | while (1) { | |
3464 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
3465 | mark); | |
3466 | if (ret) | |
3467 | break; | |
3468 | ||
3469 | clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); | |
3470 | while (start <= end) { | |
3471 | index = start >> PAGE_CACHE_SHIFT; | |
3472 | start = (u64)(index + 1) << PAGE_CACHE_SHIFT; | |
3473 | page = find_get_page(btree_inode->i_mapping, index); | |
3474 | if (!page) | |
3475 | continue; | |
3476 | offset = page_offset(page); | |
3477 | ||
3478 | spin_lock(&dirty_pages->buffer_lock); | |
3479 | eb = radix_tree_lookup( | |
3480 | &(&BTRFS_I(page->mapping->host)->io_tree)->buffer, | |
3481 | offset >> PAGE_CACHE_SHIFT); | |
3482 | spin_unlock(&dirty_pages->buffer_lock); | |
3483 | if (eb) { | |
3484 | ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY, | |
3485 | &eb->bflags); | |
3486 | atomic_set(&eb->refs, 1); | |
3487 | } | |
3488 | if (PageWriteback(page)) | |
3489 | end_page_writeback(page); | |
3490 | ||
3491 | lock_page(page); | |
3492 | if (PageDirty(page)) { | |
3493 | clear_page_dirty_for_io(page); | |
3494 | spin_lock_irq(&page->mapping->tree_lock); | |
3495 | radix_tree_tag_clear(&page->mapping->page_tree, | |
3496 | page_index(page), | |
3497 | PAGECACHE_TAG_DIRTY); | |
3498 | spin_unlock_irq(&page->mapping->tree_lock); | |
3499 | } | |
3500 | ||
3501 | page->mapping->a_ops->invalidatepage(page, 0); | |
3502 | unlock_page(page); | |
3503 | } | |
3504 | } | |
3505 | ||
3506 | return ret; | |
3507 | } | |
3508 | ||
3509 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
3510 | struct extent_io_tree *pinned_extents) | |
3511 | { | |
3512 | struct extent_io_tree *unpin; | |
3513 | u64 start; | |
3514 | u64 end; | |
3515 | int ret; | |
3516 | ||
3517 | unpin = pinned_extents; | |
3518 | while (1) { | |
3519 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
3520 | EXTENT_DIRTY); | |
3521 | if (ret) | |
3522 | break; | |
3523 | ||
3524 | /* opt_discard */ | |
5378e607 LD |
3525 | if (btrfs_test_opt(root, DISCARD)) |
3526 | ret = btrfs_error_discard_extent(root, start, | |
3527 | end + 1 - start, | |
3528 | NULL); | |
acce952b | 3529 | |
3530 | clear_extent_dirty(unpin, start, end, GFP_NOFS); | |
3531 | btrfs_error_unpin_extent_range(root, start, end); | |
3532 | cond_resched(); | |
3533 | } | |
3534 | ||
3535 | return 0; | |
3536 | } | |
3537 | ||
3538 | static int btrfs_cleanup_transaction(struct btrfs_root *root) | |
3539 | { | |
3540 | struct btrfs_transaction *t; | |
3541 | LIST_HEAD(list); | |
3542 | ||
3543 | WARN_ON(1); | |
3544 | ||
acce952b | 3545 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
3546 | ||
a4abeea4 | 3547 | spin_lock(&root->fs_info->trans_lock); |
acce952b | 3548 | list_splice_init(&root->fs_info->trans_list, &list); |
a4abeea4 JB |
3549 | root->fs_info->trans_no_join = 1; |
3550 | spin_unlock(&root->fs_info->trans_lock); | |
3551 | ||
acce952b | 3552 | while (!list_empty(&list)) { |
3553 | t = list_entry(list.next, struct btrfs_transaction, list); | |
3554 | if (!t) | |
3555 | break; | |
3556 | ||
3557 | btrfs_destroy_ordered_operations(root); | |
3558 | ||
3559 | btrfs_destroy_ordered_extents(root); | |
3560 | ||
3561 | btrfs_destroy_delayed_refs(t, root); | |
3562 | ||
3563 | btrfs_block_rsv_release(root, | |
3564 | &root->fs_info->trans_block_rsv, | |
3565 | t->dirty_pages.dirty_bytes); | |
3566 | ||
3567 | /* FIXME: cleanup wait for commit */ | |
3568 | t->in_commit = 1; | |
3569 | t->blocked = 1; | |
3570 | if (waitqueue_active(&root->fs_info->transaction_blocked_wait)) | |
3571 | wake_up(&root->fs_info->transaction_blocked_wait); | |
3572 | ||
3573 | t->blocked = 0; | |
3574 | if (waitqueue_active(&root->fs_info->transaction_wait)) | |
3575 | wake_up(&root->fs_info->transaction_wait); | |
acce952b | 3576 | |
acce952b | 3577 | t->commit_done = 1; |
3578 | if (waitqueue_active(&t->commit_wait)) | |
3579 | wake_up(&t->commit_wait); | |
acce952b | 3580 | |
3581 | btrfs_destroy_pending_snapshots(t); | |
3582 | ||
3583 | btrfs_destroy_delalloc_inodes(root); | |
3584 | ||
a4abeea4 | 3585 | spin_lock(&root->fs_info->trans_lock); |
acce952b | 3586 | root->fs_info->running_transaction = NULL; |
a4abeea4 | 3587 | spin_unlock(&root->fs_info->trans_lock); |
acce952b | 3588 | |
3589 | btrfs_destroy_marked_extents(root, &t->dirty_pages, | |
3590 | EXTENT_DIRTY); | |
3591 | ||
3592 | btrfs_destroy_pinned_extent(root, | |
3593 | root->fs_info->pinned_extents); | |
3594 | ||
13c5a93e | 3595 | atomic_set(&t->use_count, 0); |
acce952b | 3596 | list_del_init(&t->list); |
3597 | memset(t, 0, sizeof(*t)); | |
3598 | kmem_cache_free(btrfs_transaction_cachep, t); | |
3599 | } | |
3600 | ||
a4abeea4 JB |
3601 | spin_lock(&root->fs_info->trans_lock); |
3602 | root->fs_info->trans_no_join = 0; | |
3603 | spin_unlock(&root->fs_info->trans_lock); | |
acce952b | 3604 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); |
acce952b | 3605 | |
3606 | return 0; | |
3607 | } | |
3608 | ||
d1310b2e | 3609 | static struct extent_io_ops btree_extent_io_ops = { |
4bef0848 | 3610 | .write_cache_pages_lock_hook = btree_lock_page_hook, |
ce9adaa5 | 3611 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
4bb31e92 | 3612 | .readpage_io_failed_hook = btree_io_failed_hook, |
0b86a832 | 3613 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
3614 | /* note we're sharing with inode.c for the merge bio hook */ |
3615 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 3616 | }; |