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> |
6463fe58 | 33 | #include <linux/uuid.h> |
803b2f54 | 34 | #include <linux/semaphore.h> |
7e75bf3f | 35 | #include <asm/unaligned.h> |
eb60ceac CM |
36 | #include "ctree.h" |
37 | #include "disk-io.h" | |
e089f05c | 38 | #include "transaction.h" |
0f7d52f4 | 39 | #include "btrfs_inode.h" |
0b86a832 | 40 | #include "volumes.h" |
db94535d | 41 | #include "print-tree.h" |
8b712842 | 42 | #include "async-thread.h" |
925baedd | 43 | #include "locking.h" |
e02119d5 | 44 | #include "tree-log.h" |
fa9c0d79 | 45 | #include "free-space-cache.h" |
581bb050 | 46 | #include "inode-map.h" |
21adbd5c | 47 | #include "check-integrity.h" |
606686ee | 48 | #include "rcu-string.h" |
8dabb742 | 49 | #include "dev-replace.h" |
53b381b3 | 50 | #include "raid56.h" |
5ac1d209 | 51 | #include "sysfs.h" |
eb60ceac | 52 | |
de0022b9 JB |
53 | #ifdef CONFIG_X86 |
54 | #include <asm/cpufeature.h> | |
55 | #endif | |
56 | ||
d1310b2e | 57 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 58 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 59 | static void free_fs_root(struct btrfs_root *root); |
fcd1f065 | 60 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 61 | int read_only); |
569e0f35 JB |
62 | static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, |
63 | struct btrfs_root *root); | |
143bede5 | 64 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); |
acce952b | 65 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
66 | struct btrfs_root *root); | |
143bede5 | 67 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); |
acce952b | 68 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, |
69 | struct extent_io_tree *dirty_pages, | |
70 | int mark); | |
71 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
72 | struct extent_io_tree *pinned_extents); | |
48a3b636 ES |
73 | static int btrfs_cleanup_transaction(struct btrfs_root *root); |
74 | static void btrfs_error_commit_super(struct btrfs_root *root); | |
ce9adaa5 | 75 | |
d352ac68 CM |
76 | /* |
77 | * end_io_wq structs are used to do processing in task context when an IO is | |
78 | * complete. This is used during reads to verify checksums, and it is used | |
79 | * by writes to insert metadata for new file extents after IO is complete. | |
80 | */ | |
ce9adaa5 CM |
81 | struct end_io_wq { |
82 | struct bio *bio; | |
83 | bio_end_io_t *end_io; | |
84 | void *private; | |
85 | struct btrfs_fs_info *info; | |
86 | int error; | |
22c59948 | 87 | int metadata; |
ce9adaa5 | 88 | struct list_head list; |
8b712842 | 89 | struct btrfs_work work; |
ce9adaa5 | 90 | }; |
0da5468f | 91 | |
d352ac68 CM |
92 | /* |
93 | * async submit bios are used to offload expensive checksumming | |
94 | * onto the worker threads. They checksum file and metadata bios | |
95 | * just before they are sent down the IO stack. | |
96 | */ | |
44b8bd7e CM |
97 | struct async_submit_bio { |
98 | struct inode *inode; | |
99 | struct bio *bio; | |
100 | struct list_head list; | |
4a69a410 CM |
101 | extent_submit_bio_hook_t *submit_bio_start; |
102 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
103 | int rw; |
104 | int mirror_num; | |
c8b97818 | 105 | unsigned long bio_flags; |
eaf25d93 CM |
106 | /* |
107 | * bio_offset is optional, can be used if the pages in the bio | |
108 | * can't tell us where in the file the bio should go | |
109 | */ | |
110 | u64 bio_offset; | |
8b712842 | 111 | struct btrfs_work work; |
79787eaa | 112 | int error; |
44b8bd7e CM |
113 | }; |
114 | ||
85d4e461 CM |
115 | /* |
116 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
117 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
118 | * the level the eb occupies in the tree. | |
119 | * | |
120 | * Different roots are used for different purposes and may nest inside each | |
121 | * other and they require separate keysets. As lockdep keys should be | |
122 | * static, assign keysets according to the purpose of the root as indicated | |
123 | * by btrfs_root->objectid. This ensures that all special purpose roots | |
124 | * have separate keysets. | |
4008c04a | 125 | * |
85d4e461 CM |
126 | * Lock-nesting across peer nodes is always done with the immediate parent |
127 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
128 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 129 | * |
85d4e461 CM |
130 | * The key is set by the readpage_end_io_hook after the buffer has passed |
131 | * csum validation but before the pages are unlocked. It is also set by | |
132 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 133 | * |
85d4e461 CM |
134 | * We also add a check to make sure the highest level of the tree is the |
135 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
136 | * needs update as well. | |
4008c04a CM |
137 | */ |
138 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
139 | # if BTRFS_MAX_LEVEL != 8 | |
140 | # error | |
141 | # endif | |
85d4e461 CM |
142 | |
143 | static struct btrfs_lockdep_keyset { | |
144 | u64 id; /* root objectid */ | |
145 | const char *name_stem; /* lock name stem */ | |
146 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
147 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
148 | } btrfs_lockdep_keysets[] = { | |
149 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
150 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
151 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
152 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
153 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
154 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
60b62978 | 155 | { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, |
85d4e461 CM |
156 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, |
157 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
158 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
13fd8da9 | 159 | { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, |
85d4e461 | 160 | { .id = 0, .name_stem = "tree" }, |
4008c04a | 161 | }; |
85d4e461 CM |
162 | |
163 | void __init btrfs_init_lockdep(void) | |
164 | { | |
165 | int i, j; | |
166 | ||
167 | /* initialize lockdep class names */ | |
168 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
169 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
170 | ||
171 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
172 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
173 | "btrfs-%s-%02d", ks->name_stem, j); | |
174 | } | |
175 | } | |
176 | ||
177 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
178 | int level) | |
179 | { | |
180 | struct btrfs_lockdep_keyset *ks; | |
181 | ||
182 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
183 | ||
184 | /* find the matching keyset, id 0 is the default entry */ | |
185 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
186 | if (ks->id == objectid) | |
187 | break; | |
188 | ||
189 | lockdep_set_class_and_name(&eb->lock, | |
190 | &ks->keys[level], ks->names[level]); | |
191 | } | |
192 | ||
4008c04a CM |
193 | #endif |
194 | ||
d352ac68 CM |
195 | /* |
196 | * extents on the btree inode are pretty simple, there's one extent | |
197 | * that covers the entire device | |
198 | */ | |
b2950863 | 199 | static struct extent_map *btree_get_extent(struct inode *inode, |
306e16ce | 200 | struct page *page, size_t pg_offset, u64 start, u64 len, |
b2950863 | 201 | int create) |
7eccb903 | 202 | { |
5f39d397 CM |
203 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
204 | struct extent_map *em; | |
205 | int ret; | |
206 | ||
890871be | 207 | read_lock(&em_tree->lock); |
d1310b2e | 208 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d CM |
209 | if (em) { |
210 | em->bdev = | |
211 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
890871be | 212 | read_unlock(&em_tree->lock); |
5f39d397 | 213 | goto out; |
a061fc8d | 214 | } |
890871be | 215 | read_unlock(&em_tree->lock); |
7b13b7b1 | 216 | |
172ddd60 | 217 | em = alloc_extent_map(); |
5f39d397 CM |
218 | if (!em) { |
219 | em = ERR_PTR(-ENOMEM); | |
220 | goto out; | |
221 | } | |
222 | em->start = 0; | |
0afbaf8c | 223 | em->len = (u64)-1; |
c8b97818 | 224 | em->block_len = (u64)-1; |
5f39d397 | 225 | em->block_start = 0; |
a061fc8d | 226 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e | 227 | |
890871be | 228 | write_lock(&em_tree->lock); |
09a2a8f9 | 229 | ret = add_extent_mapping(em_tree, em, 0); |
5f39d397 CM |
230 | if (ret == -EEXIST) { |
231 | free_extent_map(em); | |
7b13b7b1 | 232 | em = lookup_extent_mapping(em_tree, start, len); |
b4f359ab | 233 | if (!em) |
0433f20d | 234 | em = ERR_PTR(-EIO); |
5f39d397 | 235 | } else if (ret) { |
7b13b7b1 | 236 | free_extent_map(em); |
0433f20d | 237 | em = ERR_PTR(ret); |
5f39d397 | 238 | } |
890871be | 239 | write_unlock(&em_tree->lock); |
7b13b7b1 | 240 | |
5f39d397 CM |
241 | out: |
242 | return em; | |
7eccb903 CM |
243 | } |
244 | ||
b0496686 | 245 | u32 btrfs_csum_data(char *data, u32 seed, size_t len) |
19c00ddc | 246 | { |
163e783e | 247 | return crc32c(seed, data, len); |
19c00ddc CM |
248 | } |
249 | ||
250 | void btrfs_csum_final(u32 crc, char *result) | |
251 | { | |
7e75bf3f | 252 | put_unaligned_le32(~crc, result); |
19c00ddc CM |
253 | } |
254 | ||
d352ac68 CM |
255 | /* |
256 | * compute the csum for a btree block, and either verify it or write it | |
257 | * into the csum field of the block. | |
258 | */ | |
19c00ddc CM |
259 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
260 | int verify) | |
261 | { | |
6c41761f | 262 | u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy); |
607d432d | 263 | char *result = NULL; |
19c00ddc CM |
264 | unsigned long len; |
265 | unsigned long cur_len; | |
266 | unsigned long offset = BTRFS_CSUM_SIZE; | |
19c00ddc CM |
267 | char *kaddr; |
268 | unsigned long map_start; | |
269 | unsigned long map_len; | |
270 | int err; | |
271 | u32 crc = ~(u32)0; | |
607d432d | 272 | unsigned long inline_result; |
19c00ddc CM |
273 | |
274 | len = buf->len - offset; | |
d397712b | 275 | while (len > 0) { |
19c00ddc | 276 | err = map_private_extent_buffer(buf, offset, 32, |
a6591715 | 277 | &kaddr, &map_start, &map_len); |
d397712b | 278 | if (err) |
19c00ddc | 279 | return 1; |
19c00ddc | 280 | cur_len = min(len, map_len - (offset - map_start)); |
b0496686 | 281 | crc = btrfs_csum_data(kaddr + offset - map_start, |
19c00ddc CM |
282 | crc, cur_len); |
283 | len -= cur_len; | |
284 | offset += cur_len; | |
19c00ddc | 285 | } |
607d432d JB |
286 | if (csum_size > sizeof(inline_result)) { |
287 | result = kzalloc(csum_size * sizeof(char), GFP_NOFS); | |
288 | if (!result) | |
289 | return 1; | |
290 | } else { | |
291 | result = (char *)&inline_result; | |
292 | } | |
293 | ||
19c00ddc CM |
294 | btrfs_csum_final(crc, result); |
295 | ||
296 | if (verify) { | |
607d432d | 297 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
298 | u32 val; |
299 | u32 found = 0; | |
607d432d | 300 | memcpy(&found, result, csum_size); |
e4204ded | 301 | |
607d432d | 302 | read_extent_buffer(buf, &val, 0, csum_size); |
7a36ddec | 303 | printk_ratelimited(KERN_INFO "btrfs: %s checksum verify " |
193f284d CM |
304 | "failed on %llu wanted %X found %X " |
305 | "level %d\n", | |
c1c9ff7c GU |
306 | root->fs_info->sb->s_id, buf->start, |
307 | val, found, btrfs_header_level(buf)); | |
607d432d JB |
308 | if (result != (char *)&inline_result) |
309 | kfree(result); | |
19c00ddc CM |
310 | return 1; |
311 | } | |
312 | } else { | |
607d432d | 313 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 314 | } |
607d432d JB |
315 | if (result != (char *)&inline_result) |
316 | kfree(result); | |
19c00ddc CM |
317 | return 0; |
318 | } | |
319 | ||
d352ac68 CM |
320 | /* |
321 | * we can't consider a given block up to date unless the transid of the | |
322 | * block matches the transid in the parent node's pointer. This is how we | |
323 | * detect blocks that either didn't get written at all or got written | |
324 | * in the wrong place. | |
325 | */ | |
1259ab75 | 326 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
327 | struct extent_buffer *eb, u64 parent_transid, |
328 | int atomic) | |
1259ab75 | 329 | { |
2ac55d41 | 330 | struct extent_state *cached_state = NULL; |
1259ab75 CM |
331 | int ret; |
332 | ||
333 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
334 | return 0; | |
335 | ||
b9fab919 CM |
336 | if (atomic) |
337 | return -EAGAIN; | |
338 | ||
2ac55d41 | 339 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
d0082371 | 340 | 0, &cached_state); |
0b32f4bb | 341 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
342 | btrfs_header_generation(eb) == parent_transid) { |
343 | ret = 0; | |
344 | goto out; | |
345 | } | |
7a36ddec | 346 | printk_ratelimited("parent transid verify failed on %llu wanted %llu " |
193f284d | 347 | "found %llu\n", |
c1c9ff7c | 348 | eb->start, parent_transid, btrfs_header_generation(eb)); |
1259ab75 | 349 | ret = 1; |
0b32f4bb | 350 | clear_extent_buffer_uptodate(eb); |
33958dc6 | 351 | out: |
2ac55d41 JB |
352 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
353 | &cached_state, GFP_NOFS); | |
1259ab75 | 354 | return ret; |
1259ab75 CM |
355 | } |
356 | ||
1104a885 DS |
357 | /* |
358 | * Return 0 if the superblock checksum type matches the checksum value of that | |
359 | * algorithm. Pass the raw disk superblock data. | |
360 | */ | |
361 | static int btrfs_check_super_csum(char *raw_disk_sb) | |
362 | { | |
363 | struct btrfs_super_block *disk_sb = | |
364 | (struct btrfs_super_block *)raw_disk_sb; | |
365 | u16 csum_type = btrfs_super_csum_type(disk_sb); | |
366 | int ret = 0; | |
367 | ||
368 | if (csum_type == BTRFS_CSUM_TYPE_CRC32) { | |
369 | u32 crc = ~(u32)0; | |
370 | const int csum_size = sizeof(crc); | |
371 | char result[csum_size]; | |
372 | ||
373 | /* | |
374 | * The super_block structure does not span the whole | |
375 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space | |
376 | * is filled with zeros and is included in the checkum. | |
377 | */ | |
378 | crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE, | |
379 | crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
380 | btrfs_csum_final(crc, result); | |
381 | ||
382 | if (memcmp(raw_disk_sb, result, csum_size)) | |
383 | ret = 1; | |
667e7d94 CM |
384 | |
385 | if (ret && btrfs_super_generation(disk_sb) < 10) { | |
386 | printk(KERN_WARNING "btrfs: super block crcs don't match, older mkfs detected\n"); | |
387 | ret = 0; | |
388 | } | |
1104a885 DS |
389 | } |
390 | ||
391 | if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) { | |
392 | printk(KERN_ERR "btrfs: unsupported checksum algorithm %u\n", | |
393 | csum_type); | |
394 | ret = 1; | |
395 | } | |
396 | ||
397 | return ret; | |
398 | } | |
399 | ||
d352ac68 CM |
400 | /* |
401 | * helper to read a given tree block, doing retries as required when | |
402 | * the checksums don't match and we have alternate mirrors to try. | |
403 | */ | |
f188591e CM |
404 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
405 | struct extent_buffer *eb, | |
ca7a79ad | 406 | u64 start, u64 parent_transid) |
f188591e CM |
407 | { |
408 | struct extent_io_tree *io_tree; | |
ea466794 | 409 | int failed = 0; |
f188591e CM |
410 | int ret; |
411 | int num_copies = 0; | |
412 | int mirror_num = 0; | |
ea466794 | 413 | int failed_mirror = 0; |
f188591e | 414 | |
a826d6dc | 415 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
f188591e CM |
416 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
417 | while (1) { | |
bb82ab88 AJ |
418 | ret = read_extent_buffer_pages(io_tree, eb, start, |
419 | WAIT_COMPLETE, | |
f188591e | 420 | btree_get_extent, mirror_num); |
256dd1bb SB |
421 | if (!ret) { |
422 | if (!verify_parent_transid(io_tree, eb, | |
b9fab919 | 423 | parent_transid, 0)) |
256dd1bb SB |
424 | break; |
425 | else | |
426 | ret = -EIO; | |
427 | } | |
d397712b | 428 | |
a826d6dc JB |
429 | /* |
430 | * This buffer's crc is fine, but its contents are corrupted, so | |
431 | * there is no reason to read the other copies, they won't be | |
432 | * any less wrong. | |
433 | */ | |
434 | if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) | |
ea466794 JB |
435 | break; |
436 | ||
5d964051 | 437 | num_copies = btrfs_num_copies(root->fs_info, |
f188591e | 438 | eb->start, eb->len); |
4235298e | 439 | if (num_copies == 1) |
ea466794 | 440 | break; |
4235298e | 441 | |
5cf1ab56 JB |
442 | if (!failed_mirror) { |
443 | failed = 1; | |
444 | failed_mirror = eb->read_mirror; | |
445 | } | |
446 | ||
f188591e | 447 | mirror_num++; |
ea466794 JB |
448 | if (mirror_num == failed_mirror) |
449 | mirror_num++; | |
450 | ||
4235298e | 451 | if (mirror_num > num_copies) |
ea466794 | 452 | break; |
f188591e | 453 | } |
ea466794 | 454 | |
c0901581 | 455 | if (failed && !ret && failed_mirror) |
ea466794 JB |
456 | repair_eb_io_failure(root, eb, failed_mirror); |
457 | ||
458 | return ret; | |
f188591e | 459 | } |
19c00ddc | 460 | |
d352ac68 | 461 | /* |
d397712b CM |
462 | * checksum a dirty tree block before IO. This has extra checks to make sure |
463 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 464 | */ |
d397712b | 465 | |
b2950863 | 466 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 467 | { |
4eee4fa4 | 468 | u64 start = page_offset(page); |
19c00ddc | 469 | u64 found_start; |
19c00ddc | 470 | struct extent_buffer *eb; |
f188591e | 471 | |
4f2de97a JB |
472 | eb = (struct extent_buffer *)page->private; |
473 | if (page != eb->pages[0]) | |
474 | return 0; | |
19c00ddc | 475 | found_start = btrfs_header_bytenr(eb); |
fae7f21c | 476 | if (WARN_ON(found_start != start || !PageUptodate(page))) |
4f2de97a | 477 | return 0; |
19c00ddc | 478 | csum_tree_block(root, eb, 0); |
19c00ddc CM |
479 | return 0; |
480 | } | |
481 | ||
2b82032c YZ |
482 | static int check_tree_block_fsid(struct btrfs_root *root, |
483 | struct extent_buffer *eb) | |
484 | { | |
485 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
486 | u8 fsid[BTRFS_UUID_SIZE]; | |
487 | int ret = 1; | |
488 | ||
0a4e5586 | 489 | read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); |
2b82032c YZ |
490 | while (fs_devices) { |
491 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
492 | ret = 0; | |
493 | break; | |
494 | } | |
495 | fs_devices = fs_devices->seed; | |
496 | } | |
497 | return ret; | |
498 | } | |
499 | ||
a826d6dc JB |
500 | #define CORRUPT(reason, eb, root, slot) \ |
501 | printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \ | |
502 | "root=%llu, slot=%d\n", reason, \ | |
c1c9ff7c | 503 | btrfs_header_bytenr(eb), root->objectid, slot) |
a826d6dc JB |
504 | |
505 | static noinline int check_leaf(struct btrfs_root *root, | |
506 | struct extent_buffer *leaf) | |
507 | { | |
508 | struct btrfs_key key; | |
509 | struct btrfs_key leaf_key; | |
510 | u32 nritems = btrfs_header_nritems(leaf); | |
511 | int slot; | |
512 | ||
513 | if (nritems == 0) | |
514 | return 0; | |
515 | ||
516 | /* Check the 0 item */ | |
517 | if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != | |
518 | BTRFS_LEAF_DATA_SIZE(root)) { | |
519 | CORRUPT("invalid item offset size pair", leaf, root, 0); | |
520 | return -EIO; | |
521 | } | |
522 | ||
523 | /* | |
524 | * Check to make sure each items keys are in the correct order and their | |
525 | * offsets make sense. We only have to loop through nritems-1 because | |
526 | * we check the current slot against the next slot, which verifies the | |
527 | * next slot's offset+size makes sense and that the current's slot | |
528 | * offset is correct. | |
529 | */ | |
530 | for (slot = 0; slot < nritems - 1; slot++) { | |
531 | btrfs_item_key_to_cpu(leaf, &leaf_key, slot); | |
532 | btrfs_item_key_to_cpu(leaf, &key, slot + 1); | |
533 | ||
534 | /* Make sure the keys are in the right order */ | |
535 | if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { | |
536 | CORRUPT("bad key order", leaf, root, slot); | |
537 | return -EIO; | |
538 | } | |
539 | ||
540 | /* | |
541 | * Make sure the offset and ends are right, remember that the | |
542 | * item data starts at the end of the leaf and grows towards the | |
543 | * front. | |
544 | */ | |
545 | if (btrfs_item_offset_nr(leaf, slot) != | |
546 | btrfs_item_end_nr(leaf, slot + 1)) { | |
547 | CORRUPT("slot offset bad", leaf, root, slot); | |
548 | return -EIO; | |
549 | } | |
550 | ||
551 | /* | |
552 | * Check to make sure that we don't point outside of the leaf, | |
553 | * just incase all the items are consistent to eachother, but | |
554 | * all point outside of the leaf. | |
555 | */ | |
556 | if (btrfs_item_end_nr(leaf, slot) > | |
557 | BTRFS_LEAF_DATA_SIZE(root)) { | |
558 | CORRUPT("slot end outside of leaf", leaf, root, slot); | |
559 | return -EIO; | |
560 | } | |
561 | } | |
562 | ||
563 | return 0; | |
564 | } | |
565 | ||
facc8a22 MX |
566 | static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, |
567 | u64 phy_offset, struct page *page, | |
568 | u64 start, u64 end, int mirror) | |
ce9adaa5 | 569 | { |
ce9adaa5 CM |
570 | u64 found_start; |
571 | int found_level; | |
ce9adaa5 CM |
572 | struct extent_buffer *eb; |
573 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 574 | int ret = 0; |
727011e0 | 575 | int reads_done; |
ce9adaa5 | 576 | |
ce9adaa5 CM |
577 | if (!page->private) |
578 | goto out; | |
d397712b | 579 | |
4f2de97a | 580 | eb = (struct extent_buffer *)page->private; |
d397712b | 581 | |
0b32f4bb JB |
582 | /* the pending IO might have been the only thing that kept this buffer |
583 | * in memory. Make sure we have a ref for all this other checks | |
584 | */ | |
585 | extent_buffer_get(eb); | |
586 | ||
587 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
727011e0 CM |
588 | if (!reads_done) |
589 | goto err; | |
f188591e | 590 | |
5cf1ab56 | 591 | eb->read_mirror = mirror; |
ea466794 JB |
592 | if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) { |
593 | ret = -EIO; | |
594 | goto err; | |
595 | } | |
596 | ||
ce9adaa5 | 597 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 598 | if (found_start != eb->start) { |
7a36ddec | 599 | printk_ratelimited(KERN_INFO "btrfs bad tree block start " |
193f284d | 600 | "%llu %llu\n", |
c1c9ff7c | 601 | found_start, eb->start); |
f188591e | 602 | ret = -EIO; |
ce9adaa5 CM |
603 | goto err; |
604 | } | |
2b82032c | 605 | if (check_tree_block_fsid(root, eb)) { |
7a36ddec | 606 | printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n", |
c1c9ff7c | 607 | eb->start); |
1259ab75 CM |
608 | ret = -EIO; |
609 | goto err; | |
610 | } | |
ce9adaa5 | 611 | found_level = btrfs_header_level(eb); |
1c24c3ce JB |
612 | if (found_level >= BTRFS_MAX_LEVEL) { |
613 | btrfs_info(root->fs_info, "bad tree block level %d\n", | |
614 | (int)btrfs_header_level(eb)); | |
615 | ret = -EIO; | |
616 | goto err; | |
617 | } | |
ce9adaa5 | 618 | |
85d4e461 CM |
619 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
620 | eb, found_level); | |
4008c04a | 621 | |
ce9adaa5 | 622 | ret = csum_tree_block(root, eb, 1); |
a826d6dc | 623 | if (ret) { |
f188591e | 624 | ret = -EIO; |
a826d6dc JB |
625 | goto err; |
626 | } | |
627 | ||
628 | /* | |
629 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
630 | * that we don't try and read the other copies of this block, just | |
631 | * return -EIO. | |
632 | */ | |
633 | if (found_level == 0 && check_leaf(root, eb)) { | |
634 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); | |
635 | ret = -EIO; | |
636 | } | |
ce9adaa5 | 637 | |
0b32f4bb JB |
638 | if (!ret) |
639 | set_extent_buffer_uptodate(eb); | |
ce9adaa5 | 640 | err: |
79fb65a1 JB |
641 | if (reads_done && |
642 | test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
4bb31e92 | 643 | btree_readahead_hook(root, eb, eb->start, ret); |
4bb31e92 | 644 | |
53b381b3 DW |
645 | if (ret) { |
646 | /* | |
647 | * our io error hook is going to dec the io pages | |
648 | * again, we have to make sure it has something | |
649 | * to decrement | |
650 | */ | |
651 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 652 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 653 | } |
0b32f4bb | 654 | free_extent_buffer(eb); |
ce9adaa5 | 655 | out: |
f188591e | 656 | return ret; |
ce9adaa5 CM |
657 | } |
658 | ||
ea466794 | 659 | static int btree_io_failed_hook(struct page *page, int failed_mirror) |
4bb31e92 | 660 | { |
4bb31e92 AJ |
661 | struct extent_buffer *eb; |
662 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
663 | ||
4f2de97a | 664 | eb = (struct extent_buffer *)page->private; |
ea466794 | 665 | set_bit(EXTENT_BUFFER_IOERR, &eb->bflags); |
5cf1ab56 | 666 | eb->read_mirror = failed_mirror; |
53b381b3 | 667 | atomic_dec(&eb->io_pages); |
ea466794 | 668 | if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) |
4bb31e92 | 669 | btree_readahead_hook(root, eb, eb->start, -EIO); |
4bb31e92 AJ |
670 | return -EIO; /* we fixed nothing */ |
671 | } | |
672 | ||
ce9adaa5 | 673 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
674 | { |
675 | struct end_io_wq *end_io_wq = bio->bi_private; | |
676 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 677 | |
ce9adaa5 | 678 | fs_info = end_io_wq->info; |
ce9adaa5 | 679 | end_io_wq->error = err; |
8b712842 CM |
680 | end_io_wq->work.func = end_workqueue_fn; |
681 | end_io_wq->work.flags = 0; | |
d20f7043 | 682 | |
7b6d91da | 683 | if (bio->bi_rw & REQ_WRITE) { |
53b381b3 | 684 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) |
cad321ad CM |
685 | btrfs_queue_worker(&fs_info->endio_meta_write_workers, |
686 | &end_io_wq->work); | |
53b381b3 | 687 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) |
0cb59c99 JB |
688 | btrfs_queue_worker(&fs_info->endio_freespace_worker, |
689 | &end_io_wq->work); | |
53b381b3 DW |
690 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
691 | btrfs_queue_worker(&fs_info->endio_raid56_workers, | |
692 | &end_io_wq->work); | |
cad321ad CM |
693 | else |
694 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
695 | &end_io_wq->work); | |
d20f7043 | 696 | } else { |
53b381b3 DW |
697 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
698 | btrfs_queue_worker(&fs_info->endio_raid56_workers, | |
699 | &end_io_wq->work); | |
700 | else if (end_io_wq->metadata) | |
d20f7043 CM |
701 | btrfs_queue_worker(&fs_info->endio_meta_workers, |
702 | &end_io_wq->work); | |
703 | else | |
704 | btrfs_queue_worker(&fs_info->endio_workers, | |
705 | &end_io_wq->work); | |
706 | } | |
ce9adaa5 CM |
707 | } |
708 | ||
0cb59c99 JB |
709 | /* |
710 | * For the metadata arg you want | |
711 | * | |
712 | * 0 - if data | |
713 | * 1 - if normal metadta | |
714 | * 2 - if writing to the free space cache area | |
53b381b3 | 715 | * 3 - raid parity work |
0cb59c99 | 716 | */ |
22c59948 CM |
717 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
718 | int metadata) | |
0b86a832 | 719 | { |
ce9adaa5 | 720 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
721 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
722 | if (!end_io_wq) | |
723 | return -ENOMEM; | |
724 | ||
725 | end_io_wq->private = bio->bi_private; | |
726 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 727 | end_io_wq->info = info; |
ce9adaa5 CM |
728 | end_io_wq->error = 0; |
729 | end_io_wq->bio = bio; | |
22c59948 | 730 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
731 | |
732 | bio->bi_private = end_io_wq; | |
733 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
734 | return 0; |
735 | } | |
736 | ||
b64a2851 | 737 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 738 | { |
4854ddd0 CM |
739 | unsigned long limit = min_t(unsigned long, |
740 | info->workers.max_workers, | |
741 | info->fs_devices->open_devices); | |
742 | return 256 * limit; | |
743 | } | |
0986fe9e | 744 | |
4a69a410 CM |
745 | static void run_one_async_start(struct btrfs_work *work) |
746 | { | |
4a69a410 | 747 | struct async_submit_bio *async; |
79787eaa | 748 | int ret; |
4a69a410 CM |
749 | |
750 | async = container_of(work, struct async_submit_bio, work); | |
79787eaa JM |
751 | ret = async->submit_bio_start(async->inode, async->rw, async->bio, |
752 | async->mirror_num, async->bio_flags, | |
753 | async->bio_offset); | |
754 | if (ret) | |
755 | async->error = ret; | |
4a69a410 CM |
756 | } |
757 | ||
758 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
759 | { |
760 | struct btrfs_fs_info *fs_info; | |
761 | struct async_submit_bio *async; | |
4854ddd0 | 762 | int limit; |
8b712842 CM |
763 | |
764 | async = container_of(work, struct async_submit_bio, work); | |
765 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 766 | |
b64a2851 | 767 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
768 | limit = limit * 2 / 3; |
769 | ||
66657b31 | 770 | if (atomic_dec_return(&fs_info->nr_async_submits) < limit && |
b64a2851 | 771 | waitqueue_active(&fs_info->async_submit_wait)) |
4854ddd0 CM |
772 | wake_up(&fs_info->async_submit_wait); |
773 | ||
79787eaa JM |
774 | /* If an error occured we just want to clean up the bio and move on */ |
775 | if (async->error) { | |
776 | bio_endio(async->bio, async->error); | |
777 | return; | |
778 | } | |
779 | ||
4a69a410 | 780 | async->submit_bio_done(async->inode, async->rw, async->bio, |
eaf25d93 CM |
781 | async->mirror_num, async->bio_flags, |
782 | async->bio_offset); | |
4a69a410 CM |
783 | } |
784 | ||
785 | static void run_one_async_free(struct btrfs_work *work) | |
786 | { | |
787 | struct async_submit_bio *async; | |
788 | ||
789 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
790 | kfree(async); |
791 | } | |
792 | ||
44b8bd7e CM |
793 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
794 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 795 | unsigned long bio_flags, |
eaf25d93 | 796 | u64 bio_offset, |
4a69a410 CM |
797 | extent_submit_bio_hook_t *submit_bio_start, |
798 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
799 | { |
800 | struct async_submit_bio *async; | |
801 | ||
802 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
803 | if (!async) | |
804 | return -ENOMEM; | |
805 | ||
806 | async->inode = inode; | |
807 | async->rw = rw; | |
808 | async->bio = bio; | |
809 | async->mirror_num = mirror_num; | |
4a69a410 CM |
810 | async->submit_bio_start = submit_bio_start; |
811 | async->submit_bio_done = submit_bio_done; | |
812 | ||
813 | async->work.func = run_one_async_start; | |
814 | async->work.ordered_func = run_one_async_done; | |
815 | async->work.ordered_free = run_one_async_free; | |
816 | ||
8b712842 | 817 | async->work.flags = 0; |
c8b97818 | 818 | async->bio_flags = bio_flags; |
eaf25d93 | 819 | async->bio_offset = bio_offset; |
8c8bee1d | 820 | |
79787eaa JM |
821 | async->error = 0; |
822 | ||
cb03c743 | 823 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 | 824 | |
7b6d91da | 825 | if (rw & REQ_SYNC) |
d313d7a3 CM |
826 | btrfs_set_work_high_prio(&async->work); |
827 | ||
8b712842 | 828 | btrfs_queue_worker(&fs_info->workers, &async->work); |
9473f16c | 829 | |
d397712b | 830 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
831 | atomic_read(&fs_info->nr_async_submits)) { |
832 | wait_event(fs_info->async_submit_wait, | |
833 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
834 | } | |
835 | ||
44b8bd7e CM |
836 | return 0; |
837 | } | |
838 | ||
ce3ed71a CM |
839 | static int btree_csum_one_bio(struct bio *bio) |
840 | { | |
841 | struct bio_vec *bvec = bio->bi_io_vec; | |
842 | int bio_index = 0; | |
843 | struct btrfs_root *root; | |
79787eaa | 844 | int ret = 0; |
ce3ed71a CM |
845 | |
846 | WARN_ON(bio->bi_vcnt <= 0); | |
d397712b | 847 | while (bio_index < bio->bi_vcnt) { |
ce3ed71a | 848 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
79787eaa JM |
849 | ret = csum_dirty_buffer(root, bvec->bv_page); |
850 | if (ret) | |
851 | break; | |
ce3ed71a CM |
852 | bio_index++; |
853 | bvec++; | |
854 | } | |
79787eaa | 855 | return ret; |
ce3ed71a CM |
856 | } |
857 | ||
4a69a410 CM |
858 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
859 | struct bio *bio, int mirror_num, | |
eaf25d93 CM |
860 | unsigned long bio_flags, |
861 | u64 bio_offset) | |
22c59948 | 862 | { |
8b712842 CM |
863 | /* |
864 | * when we're called for a write, we're already in the async | |
5443be45 | 865 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 866 | */ |
79787eaa | 867 | return btree_csum_one_bio(bio); |
4a69a410 | 868 | } |
22c59948 | 869 | |
4a69a410 | 870 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
871 | int mirror_num, unsigned long bio_flags, |
872 | u64 bio_offset) | |
4a69a410 | 873 | { |
61891923 SB |
874 | int ret; |
875 | ||
8b712842 | 876 | /* |
4a69a410 CM |
877 | * when we're called for a write, we're already in the async |
878 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 879 | */ |
61891923 SB |
880 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
881 | if (ret) | |
882 | bio_endio(bio, ret); | |
883 | return ret; | |
0b86a832 CM |
884 | } |
885 | ||
de0022b9 JB |
886 | static int check_async_write(struct inode *inode, unsigned long bio_flags) |
887 | { | |
888 | if (bio_flags & EXTENT_BIO_TREE_LOG) | |
889 | return 0; | |
890 | #ifdef CONFIG_X86 | |
891 | if (cpu_has_xmm4_2) | |
892 | return 0; | |
893 | #endif | |
894 | return 1; | |
895 | } | |
896 | ||
44b8bd7e | 897 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
898 | int mirror_num, unsigned long bio_flags, |
899 | u64 bio_offset) | |
44b8bd7e | 900 | { |
de0022b9 | 901 | int async = check_async_write(inode, bio_flags); |
cad321ad CM |
902 | int ret; |
903 | ||
7b6d91da | 904 | if (!(rw & REQ_WRITE)) { |
4a69a410 CM |
905 | /* |
906 | * called for a read, do the setup so that checksum validation | |
907 | * can happen in the async kernel threads | |
908 | */ | |
f3f266ab CM |
909 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, |
910 | bio, 1); | |
1d4284bd | 911 | if (ret) |
61891923 SB |
912 | goto out_w_error; |
913 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
914 | mirror_num, 0); | |
de0022b9 JB |
915 | } else if (!async) { |
916 | ret = btree_csum_one_bio(bio); | |
917 | if (ret) | |
61891923 SB |
918 | goto out_w_error; |
919 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
920 | mirror_num, 0); | |
921 | } else { | |
922 | /* | |
923 | * kthread helpers are used to submit writes so that | |
924 | * checksumming can happen in parallel across all CPUs | |
925 | */ | |
926 | ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, | |
927 | inode, rw, bio, mirror_num, 0, | |
928 | bio_offset, | |
929 | __btree_submit_bio_start, | |
930 | __btree_submit_bio_done); | |
44b8bd7e | 931 | } |
d313d7a3 | 932 | |
61891923 SB |
933 | if (ret) { |
934 | out_w_error: | |
935 | bio_endio(bio, ret); | |
936 | } | |
937 | return ret; | |
44b8bd7e CM |
938 | } |
939 | ||
3dd1462e | 940 | #ifdef CONFIG_MIGRATION |
784b4e29 | 941 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
942 | struct page *newpage, struct page *page, |
943 | enum migrate_mode mode) | |
784b4e29 CM |
944 | { |
945 | /* | |
946 | * we can't safely write a btree page from here, | |
947 | * we haven't done the locking hook | |
948 | */ | |
949 | if (PageDirty(page)) | |
950 | return -EAGAIN; | |
951 | /* | |
952 | * Buffers may be managed in a filesystem specific way. | |
953 | * We must have no buffers or drop them. | |
954 | */ | |
955 | if (page_has_private(page) && | |
956 | !try_to_release_page(page, GFP_KERNEL)) | |
957 | return -EAGAIN; | |
a6bc32b8 | 958 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 959 | } |
3dd1462e | 960 | #endif |
784b4e29 | 961 | |
0da5468f CM |
962 | |
963 | static int btree_writepages(struct address_space *mapping, | |
964 | struct writeback_control *wbc) | |
965 | { | |
e2d84521 MX |
966 | struct btrfs_fs_info *fs_info; |
967 | int ret; | |
968 | ||
d8d5f3e1 | 969 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
970 | |
971 | if (wbc->for_kupdate) | |
972 | return 0; | |
973 | ||
e2d84521 | 974 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 975 | /* this is a bit racy, but that's ok */ |
e2d84521 MX |
976 | ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
977 | BTRFS_DIRTY_METADATA_THRESH); | |
978 | if (ret < 0) | |
793955bc | 979 | return 0; |
793955bc | 980 | } |
0b32f4bb | 981 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
982 | } |
983 | ||
b2950863 | 984 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 985 | { |
d1310b2e CM |
986 | struct extent_io_tree *tree; |
987 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 988 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 989 | } |
22b0ebda | 990 | |
70dec807 | 991 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 992 | { |
98509cfc | 993 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 994 | return 0; |
0c4e538b | 995 | |
f7a52a40 | 996 | return try_release_extent_buffer(page); |
d98237b3 CM |
997 | } |
998 | ||
d47992f8 LC |
999 | static void btree_invalidatepage(struct page *page, unsigned int offset, |
1000 | unsigned int length) | |
d98237b3 | 1001 | { |
d1310b2e CM |
1002 | struct extent_io_tree *tree; |
1003 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
1004 | extent_invalidatepage(tree, page, offset); |
1005 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 1006 | if (PagePrivate(page)) { |
d397712b CM |
1007 | printk(KERN_WARNING "btrfs warning page private not zero " |
1008 | "on page %llu\n", (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
1009 | ClearPagePrivate(page); |
1010 | set_page_private(page, 0); | |
1011 | page_cache_release(page); | |
1012 | } | |
d98237b3 CM |
1013 | } |
1014 | ||
0b32f4bb JB |
1015 | static int btree_set_page_dirty(struct page *page) |
1016 | { | |
bb146eb2 | 1017 | #ifdef DEBUG |
0b32f4bb JB |
1018 | struct extent_buffer *eb; |
1019 | ||
1020 | BUG_ON(!PagePrivate(page)); | |
1021 | eb = (struct extent_buffer *)page->private; | |
1022 | BUG_ON(!eb); | |
1023 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1024 | BUG_ON(!atomic_read(&eb->refs)); | |
1025 | btrfs_assert_tree_locked(eb); | |
bb146eb2 | 1026 | #endif |
0b32f4bb JB |
1027 | return __set_page_dirty_nobuffers(page); |
1028 | } | |
1029 | ||
7f09410b | 1030 | static const struct address_space_operations btree_aops = { |
d98237b3 | 1031 | .readpage = btree_readpage, |
0da5468f | 1032 | .writepages = btree_writepages, |
5f39d397 CM |
1033 | .releasepage = btree_releasepage, |
1034 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1035 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1036 | .migratepage = btree_migratepage, |
5a92bc88 | 1037 | #endif |
0b32f4bb | 1038 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1039 | }; |
1040 | ||
ca7a79ad CM |
1041 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1042 | u64 parent_transid) | |
090d1875 | 1043 | { |
5f39d397 CM |
1044 | struct extent_buffer *buf = NULL; |
1045 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 1046 | int ret = 0; |
090d1875 | 1047 | |
db94535d | 1048 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 1049 | if (!buf) |
090d1875 | 1050 | return 0; |
d1310b2e | 1051 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
bb82ab88 | 1052 | buf, 0, WAIT_NONE, btree_get_extent, 0); |
5f39d397 | 1053 | free_extent_buffer(buf); |
de428b63 | 1054 | return ret; |
090d1875 CM |
1055 | } |
1056 | ||
ab0fff03 AJ |
1057 | int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1058 | int mirror_num, struct extent_buffer **eb) | |
1059 | { | |
1060 | struct extent_buffer *buf = NULL; | |
1061 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1062 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; | |
1063 | int ret; | |
1064 | ||
1065 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
1066 | if (!buf) | |
1067 | return 0; | |
1068 | ||
1069 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1070 | ||
1071 | ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, | |
1072 | btree_get_extent, mirror_num); | |
1073 | if (ret) { | |
1074 | free_extent_buffer(buf); | |
1075 | return ret; | |
1076 | } | |
1077 | ||
1078 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1079 | free_extent_buffer(buf); | |
1080 | return -EIO; | |
0b32f4bb | 1081 | } else if (extent_buffer_uptodate(buf)) { |
ab0fff03 AJ |
1082 | *eb = buf; |
1083 | } else { | |
1084 | free_extent_buffer(buf); | |
1085 | } | |
1086 | return 0; | |
1087 | } | |
1088 | ||
0999df54 CM |
1089 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
1090 | u64 bytenr, u32 blocksize) | |
1091 | { | |
f28491e0 | 1092 | return find_extent_buffer(root->fs_info, bytenr); |
0999df54 CM |
1093 | } |
1094 | ||
1095 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
1096 | u64 bytenr, u32 blocksize) | |
1097 | { | |
f28491e0 | 1098 | return alloc_extent_buffer(root->fs_info, bytenr, blocksize); |
0999df54 CM |
1099 | } |
1100 | ||
1101 | ||
e02119d5 CM |
1102 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1103 | { | |
727011e0 | 1104 | return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, |
8aa38c31 | 1105 | buf->start + buf->len - 1); |
e02119d5 CM |
1106 | } |
1107 | ||
1108 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
1109 | { | |
727011e0 | 1110 | return filemap_fdatawait_range(buf->pages[0]->mapping, |
8aa38c31 | 1111 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
1112 | } |
1113 | ||
0999df54 | 1114 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 1115 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
1116 | { |
1117 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1118 | int ret; |
1119 | ||
0999df54 CM |
1120 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
1121 | if (!buf) | |
1122 | return NULL; | |
0999df54 | 1123 | |
ca7a79ad | 1124 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
0f0fe8f7 FDBM |
1125 | if (ret) { |
1126 | free_extent_buffer(buf); | |
1127 | return NULL; | |
1128 | } | |
5f39d397 | 1129 | return buf; |
ce9adaa5 | 1130 | |
eb60ceac CM |
1131 | } |
1132 | ||
d5c13f92 JM |
1133 | void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
1134 | struct extent_buffer *buf) | |
ed2ff2cb | 1135 | { |
e2d84521 MX |
1136 | struct btrfs_fs_info *fs_info = root->fs_info; |
1137 | ||
55c69072 | 1138 | if (btrfs_header_generation(buf) == |
e2d84521 | 1139 | fs_info->running_transaction->transid) { |
b9447ef8 | 1140 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1141 | |
b9473439 | 1142 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
e2d84521 MX |
1143 | __percpu_counter_add(&fs_info->dirty_metadata_bytes, |
1144 | -buf->len, | |
1145 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1146 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1147 | btrfs_set_lock_blocking(buf); | |
1148 | clear_extent_buffer_dirty(buf); | |
1149 | } | |
925baedd | 1150 | } |
5f39d397 CM |
1151 | } |
1152 | ||
143bede5 JM |
1153 | static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
1154 | u32 stripesize, struct btrfs_root *root, | |
1155 | struct btrfs_fs_info *fs_info, | |
1156 | u64 objectid) | |
d97e63b6 | 1157 | { |
cfaa7295 | 1158 | root->node = NULL; |
a28ec197 | 1159 | root->commit_root = NULL; |
db94535d CM |
1160 | root->sectorsize = sectorsize; |
1161 | root->nodesize = nodesize; | |
1162 | root->leafsize = leafsize; | |
87ee04eb | 1163 | root->stripesize = stripesize; |
123abc88 | 1164 | root->ref_cows = 0; |
0b86a832 | 1165 | root->track_dirty = 0; |
c71bf099 | 1166 | root->in_radix = 0; |
d68fc57b YZ |
1167 | root->orphan_item_inserted = 0; |
1168 | root->orphan_cleanup_state = 0; | |
0b86a832 | 1169 | |
0f7d52f4 CM |
1170 | root->objectid = objectid; |
1171 | root->last_trans = 0; | |
13a8a7c8 | 1172 | root->highest_objectid = 0; |
eb73c1b7 | 1173 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1174 | root->nr_ordered_extents = 0; |
58176a96 | 1175 | root->name = NULL; |
6bef4d31 | 1176 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1177 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1178 | root->block_rsv = NULL; |
d68fc57b | 1179 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
1180 | |
1181 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1182 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1183 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1184 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1185 | INIT_LIST_HEAD(&root->ordered_extents); |
1186 | INIT_LIST_HEAD(&root->ordered_root); | |
2ab28f32 JB |
1187 | INIT_LIST_HEAD(&root->logged_list[0]); |
1188 | INIT_LIST_HEAD(&root->logged_list[1]); | |
d68fc57b | 1189 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 1190 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1191 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1192 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1193 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1194 | spin_lock_init(&root->log_extents_lock[0]); |
1195 | spin_lock_init(&root->log_extents_lock[1]); | |
a2135011 | 1196 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1197 | mutex_init(&root->log_mutex); |
7237f183 YZ |
1198 | init_waitqueue_head(&root->log_writer_wait); |
1199 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1200 | init_waitqueue_head(&root->log_commit_wait[1]); | |
1201 | atomic_set(&root->log_commit[0], 0); | |
1202 | atomic_set(&root->log_commit[1], 0); | |
1203 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1204 | atomic_set(&root->log_batch, 0); |
8a35d95f | 1205 | atomic_set(&root->orphan_inodes, 0); |
b0feb9d9 | 1206 | atomic_set(&root->refs, 1); |
7237f183 | 1207 | root->log_transid = 0; |
257c62e1 | 1208 | root->last_log_commit = 0; |
06ea65a3 JB |
1209 | if (fs_info) |
1210 | extent_io_tree_init(&root->dirty_log_pages, | |
1211 | fs_info->btree_inode->i_mapping); | |
017e5369 | 1212 | |
3768f368 CM |
1213 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1214 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1215 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 1216 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
06ea65a3 JB |
1217 | if (fs_info) |
1218 | root->defrag_trans_start = fs_info->generation; | |
1219 | else | |
1220 | root->defrag_trans_start = 0; | |
58176a96 | 1221 | init_completion(&root->kobj_unregister); |
6702ed49 | 1222 | root->defrag_running = 0; |
4d775673 | 1223 | root->root_key.objectid = objectid; |
0ee5dc67 | 1224 | root->anon_dev = 0; |
8ea05e3a | 1225 | |
5f3ab90a | 1226 | spin_lock_init(&root->root_item_lock); |
3768f368 CM |
1227 | } |
1228 | ||
f84a8bd6 | 1229 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info) |
6f07e42e AV |
1230 | { |
1231 | struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS); | |
1232 | if (root) | |
1233 | root->fs_info = fs_info; | |
1234 | return root; | |
1235 | } | |
1236 | ||
06ea65a3 JB |
1237 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1238 | /* Should only be used by the testing infrastructure */ | |
1239 | struct btrfs_root *btrfs_alloc_dummy_root(void) | |
1240 | { | |
1241 | struct btrfs_root *root; | |
1242 | ||
1243 | root = btrfs_alloc_root(NULL); | |
1244 | if (!root) | |
1245 | return ERR_PTR(-ENOMEM); | |
1246 | __setup_root(4096, 4096, 4096, 4096, root, NULL, 1); | |
1247 | root->dummy_root = 1; | |
1248 | ||
1249 | return root; | |
1250 | } | |
1251 | #endif | |
1252 | ||
20897f5c AJ |
1253 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
1254 | struct btrfs_fs_info *fs_info, | |
1255 | u64 objectid) | |
1256 | { | |
1257 | struct extent_buffer *leaf; | |
1258 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1259 | struct btrfs_root *root; | |
1260 | struct btrfs_key key; | |
1261 | int ret = 0; | |
6463fe58 | 1262 | uuid_le uuid; |
20897f5c AJ |
1263 | |
1264 | root = btrfs_alloc_root(fs_info); | |
1265 | if (!root) | |
1266 | return ERR_PTR(-ENOMEM); | |
1267 | ||
1268 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1269 | tree_root->sectorsize, tree_root->stripesize, | |
1270 | root, fs_info, objectid); | |
1271 | root->root_key.objectid = objectid; | |
1272 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1273 | root->root_key.offset = 0; | |
1274 | ||
1275 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, | |
1276 | 0, objectid, NULL, 0, 0, 0); | |
1277 | if (IS_ERR(leaf)) { | |
1278 | ret = PTR_ERR(leaf); | |
1dd05682 | 1279 | leaf = NULL; |
20897f5c AJ |
1280 | goto fail; |
1281 | } | |
1282 | ||
20897f5c AJ |
1283 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1284 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1285 | btrfs_set_header_generation(leaf, trans->transid); | |
1286 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1287 | btrfs_set_header_owner(leaf, objectid); | |
1288 | root->node = leaf; | |
1289 | ||
0a4e5586 | 1290 | write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(), |
20897f5c AJ |
1291 | BTRFS_FSID_SIZE); |
1292 | write_extent_buffer(leaf, fs_info->chunk_tree_uuid, | |
b308bc2f | 1293 | btrfs_header_chunk_tree_uuid(leaf), |
20897f5c AJ |
1294 | BTRFS_UUID_SIZE); |
1295 | btrfs_mark_buffer_dirty(leaf); | |
1296 | ||
1297 | root->commit_root = btrfs_root_node(root); | |
1298 | root->track_dirty = 1; | |
1299 | ||
1300 | ||
1301 | root->root_item.flags = 0; | |
1302 | root->root_item.byte_limit = 0; | |
1303 | btrfs_set_root_bytenr(&root->root_item, leaf->start); | |
1304 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1305 | btrfs_set_root_level(&root->root_item, 0); | |
1306 | btrfs_set_root_refs(&root->root_item, 1); | |
1307 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1308 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1309 | btrfs_set_root_dirid(&root->root_item, 0); | |
6463fe58 SB |
1310 | uuid_le_gen(&uuid); |
1311 | memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); | |
20897f5c AJ |
1312 | root->root_item.drop_level = 0; |
1313 | ||
1314 | key.objectid = objectid; | |
1315 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1316 | key.offset = 0; | |
1317 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1318 | if (ret) | |
1319 | goto fail; | |
1320 | ||
1321 | btrfs_tree_unlock(leaf); | |
1322 | ||
1dd05682 TI |
1323 | return root; |
1324 | ||
20897f5c | 1325 | fail: |
1dd05682 TI |
1326 | if (leaf) { |
1327 | btrfs_tree_unlock(leaf); | |
1328 | free_extent_buffer(leaf); | |
1329 | } | |
1330 | kfree(root); | |
20897f5c | 1331 | |
1dd05682 | 1332 | return ERR_PTR(ret); |
20897f5c AJ |
1333 | } |
1334 | ||
7237f183 YZ |
1335 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1336 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1337 | { |
1338 | struct btrfs_root *root; | |
1339 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1340 | struct extent_buffer *leaf; |
e02119d5 | 1341 | |
6f07e42e | 1342 | root = btrfs_alloc_root(fs_info); |
e02119d5 | 1343 | if (!root) |
7237f183 | 1344 | return ERR_PTR(-ENOMEM); |
e02119d5 CM |
1345 | |
1346 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1347 | tree_root->sectorsize, tree_root->stripesize, | |
1348 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1349 | ||
1350 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1351 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1352 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
7237f183 YZ |
1353 | /* |
1354 | * log trees do not get reference counted because they go away | |
1355 | * before a real commit is actually done. They do store pointers | |
1356 | * to file data extents, and those reference counts still get | |
1357 | * updated (along with back refs to the log tree). | |
1358 | */ | |
e02119d5 CM |
1359 | root->ref_cows = 0; |
1360 | ||
5d4f98a2 | 1361 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, |
66d7e7f0 | 1362 | BTRFS_TREE_LOG_OBJECTID, NULL, |
5581a51a | 1363 | 0, 0, 0); |
7237f183 YZ |
1364 | if (IS_ERR(leaf)) { |
1365 | kfree(root); | |
1366 | return ERR_CAST(leaf); | |
1367 | } | |
e02119d5 | 1368 | |
5d4f98a2 YZ |
1369 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1370 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1371 | btrfs_set_header_generation(leaf, trans->transid); | |
1372 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1373 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1374 | root->node = leaf; |
e02119d5 CM |
1375 | |
1376 | write_extent_buffer(root->node, root->fs_info->fsid, | |
0a4e5586 | 1377 | btrfs_header_fsid(), BTRFS_FSID_SIZE); |
e02119d5 CM |
1378 | btrfs_mark_buffer_dirty(root->node); |
1379 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1380 | return root; |
1381 | } | |
1382 | ||
1383 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1384 | struct btrfs_fs_info *fs_info) | |
1385 | { | |
1386 | struct btrfs_root *log_root; | |
1387 | ||
1388 | log_root = alloc_log_tree(trans, fs_info); | |
1389 | if (IS_ERR(log_root)) | |
1390 | return PTR_ERR(log_root); | |
1391 | WARN_ON(fs_info->log_root_tree); | |
1392 | fs_info->log_root_tree = log_root; | |
1393 | return 0; | |
1394 | } | |
1395 | ||
1396 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1397 | struct btrfs_root *root) | |
1398 | { | |
1399 | struct btrfs_root *log_root; | |
1400 | struct btrfs_inode_item *inode_item; | |
1401 | ||
1402 | log_root = alloc_log_tree(trans, root->fs_info); | |
1403 | if (IS_ERR(log_root)) | |
1404 | return PTR_ERR(log_root); | |
1405 | ||
1406 | log_root->last_trans = trans->transid; | |
1407 | log_root->root_key.offset = root->root_key.objectid; | |
1408 | ||
1409 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1410 | btrfs_set_stack_inode_generation(inode_item, 1); |
1411 | btrfs_set_stack_inode_size(inode_item, 3); | |
1412 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
1413 | btrfs_set_stack_inode_nbytes(inode_item, root->leafsize); | |
1414 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); | |
7237f183 | 1415 | |
5d4f98a2 | 1416 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1417 | |
1418 | WARN_ON(root->log_root); | |
1419 | root->log_root = log_root; | |
1420 | root->log_transid = 0; | |
257c62e1 | 1421 | root->last_log_commit = 0; |
e02119d5 CM |
1422 | return 0; |
1423 | } | |
1424 | ||
35a3621b SB |
1425 | static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, |
1426 | struct btrfs_key *key) | |
e02119d5 CM |
1427 | { |
1428 | struct btrfs_root *root; | |
1429 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1430 | struct btrfs_path *path; |
84234f3a | 1431 | u64 generation; |
db94535d | 1432 | u32 blocksize; |
cb517eab | 1433 | int ret; |
0f7d52f4 | 1434 | |
cb517eab MX |
1435 | path = btrfs_alloc_path(); |
1436 | if (!path) | |
0f7d52f4 | 1437 | return ERR_PTR(-ENOMEM); |
cb517eab MX |
1438 | |
1439 | root = btrfs_alloc_root(fs_info); | |
1440 | if (!root) { | |
1441 | ret = -ENOMEM; | |
1442 | goto alloc_fail; | |
0f7d52f4 CM |
1443 | } |
1444 | ||
db94535d | 1445 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb | 1446 | tree_root->sectorsize, tree_root->stripesize, |
cb517eab | 1447 | root, fs_info, key->objectid); |
0f7d52f4 | 1448 | |
cb517eab MX |
1449 | ret = btrfs_find_root(tree_root, key, path, |
1450 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1451 | if (ret) { |
13a8a7c8 YZ |
1452 | if (ret > 0) |
1453 | ret = -ENOENT; | |
cb517eab | 1454 | goto find_fail; |
0f7d52f4 | 1455 | } |
13a8a7c8 | 1456 | |
84234f3a | 1457 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1458 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1459 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1460 | blocksize, generation); |
cb517eab MX |
1461 | if (!root->node) { |
1462 | ret = -ENOMEM; | |
1463 | goto find_fail; | |
1464 | } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
1465 | ret = -EIO; | |
1466 | goto read_fail; | |
416bc658 | 1467 | } |
5d4f98a2 | 1468 | root->commit_root = btrfs_root_node(root); |
13a8a7c8 | 1469 | out: |
cb517eab MX |
1470 | btrfs_free_path(path); |
1471 | return root; | |
1472 | ||
1473 | read_fail: | |
1474 | free_extent_buffer(root->node); | |
1475 | find_fail: | |
1476 | kfree(root); | |
1477 | alloc_fail: | |
1478 | root = ERR_PTR(ret); | |
1479 | goto out; | |
1480 | } | |
1481 | ||
1482 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, | |
1483 | struct btrfs_key *location) | |
1484 | { | |
1485 | struct btrfs_root *root; | |
1486 | ||
1487 | root = btrfs_read_tree_root(tree_root, location); | |
1488 | if (IS_ERR(root)) | |
1489 | return root; | |
1490 | ||
1491 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { | |
e02119d5 | 1492 | root->ref_cows = 1; |
08fe4db1 LZ |
1493 | btrfs_check_and_init_root_item(&root->root_item); |
1494 | } | |
13a8a7c8 | 1495 | |
5eda7b5e CM |
1496 | return root; |
1497 | } | |
1498 | ||
cb517eab MX |
1499 | int btrfs_init_fs_root(struct btrfs_root *root) |
1500 | { | |
1501 | int ret; | |
1502 | ||
1503 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); | |
1504 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), | |
1505 | GFP_NOFS); | |
1506 | if (!root->free_ino_pinned || !root->free_ino_ctl) { | |
1507 | ret = -ENOMEM; | |
1508 | goto fail; | |
1509 | } | |
1510 | ||
1511 | btrfs_init_free_ino_ctl(root); | |
1512 | mutex_init(&root->fs_commit_mutex); | |
1513 | spin_lock_init(&root->cache_lock); | |
1514 | init_waitqueue_head(&root->cache_wait); | |
1515 | ||
1516 | ret = get_anon_bdev(&root->anon_dev); | |
1517 | if (ret) | |
1518 | goto fail; | |
1519 | return 0; | |
1520 | fail: | |
1521 | kfree(root->free_ino_ctl); | |
1522 | kfree(root->free_ino_pinned); | |
1523 | return ret; | |
1524 | } | |
1525 | ||
171170c1 ST |
1526 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1527 | u64 root_id) | |
cb517eab MX |
1528 | { |
1529 | struct btrfs_root *root; | |
1530 | ||
1531 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1532 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1533 | (unsigned long)root_id); | |
1534 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1535 | return root; | |
1536 | } | |
1537 | ||
1538 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, | |
1539 | struct btrfs_root *root) | |
1540 | { | |
1541 | int ret; | |
1542 | ||
1543 | ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); | |
1544 | if (ret) | |
1545 | return ret; | |
1546 | ||
1547 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1548 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1549 | (unsigned long)root->root_key.objectid, | |
1550 | root); | |
1551 | if (ret == 0) | |
1552 | root->in_radix = 1; | |
1553 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1554 | radix_tree_preload_end(); | |
1555 | ||
1556 | return ret; | |
1557 | } | |
1558 | ||
c00869f1 MX |
1559 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, |
1560 | struct btrfs_key *location, | |
1561 | bool check_ref) | |
5eda7b5e CM |
1562 | { |
1563 | struct btrfs_root *root; | |
1564 | int ret; | |
1565 | ||
edbd8d4e CM |
1566 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1567 | return fs_info->tree_root; | |
1568 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1569 | return fs_info->extent_root; | |
8f18cf13 CM |
1570 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1571 | return fs_info->chunk_root; | |
1572 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1573 | return fs_info->dev_root; | |
0403e47e YZ |
1574 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1575 | return fs_info->csum_root; | |
bcef60f2 AJ |
1576 | if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1577 | return fs_info->quota_root ? fs_info->quota_root : | |
1578 | ERR_PTR(-ENOENT); | |
f7a81ea4 SB |
1579 | if (location->objectid == BTRFS_UUID_TREE_OBJECTID) |
1580 | return fs_info->uuid_root ? fs_info->uuid_root : | |
1581 | ERR_PTR(-ENOENT); | |
4df27c4d | 1582 | again: |
cb517eab | 1583 | root = btrfs_lookup_fs_root(fs_info, location->objectid); |
48475471 | 1584 | if (root) { |
c00869f1 | 1585 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) |
48475471 | 1586 | return ERR_PTR(-ENOENT); |
5eda7b5e | 1587 | return root; |
48475471 | 1588 | } |
5eda7b5e | 1589 | |
cb517eab | 1590 | root = btrfs_read_fs_root(fs_info->tree_root, location); |
5eda7b5e CM |
1591 | if (IS_ERR(root)) |
1592 | return root; | |
3394e160 | 1593 | |
c00869f1 | 1594 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1595 | ret = -ENOENT; |
581bb050 | 1596 | goto fail; |
35a30d7c | 1597 | } |
581bb050 | 1598 | |
cb517eab | 1599 | ret = btrfs_init_fs_root(root); |
ac08aedf CM |
1600 | if (ret) |
1601 | goto fail; | |
3394e160 | 1602 | |
3f870c28 KN |
1603 | ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID, |
1604 | location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL); | |
d68fc57b YZ |
1605 | if (ret < 0) |
1606 | goto fail; | |
1607 | if (ret == 0) | |
1608 | root->orphan_item_inserted = 1; | |
1609 | ||
cb517eab | 1610 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1611 | if (ret) { |
4df27c4d YZ |
1612 | if (ret == -EEXIST) { |
1613 | free_fs_root(root); | |
1614 | goto again; | |
1615 | } | |
1616 | goto fail; | |
0f7d52f4 | 1617 | } |
edbd8d4e | 1618 | return root; |
4df27c4d YZ |
1619 | fail: |
1620 | free_fs_root(root); | |
1621 | return ERR_PTR(ret); | |
edbd8d4e CM |
1622 | } |
1623 | ||
04160088 CM |
1624 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1625 | { | |
1626 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1627 | int ret = 0; | |
04160088 CM |
1628 | struct btrfs_device *device; |
1629 | struct backing_dev_info *bdi; | |
b7967db7 | 1630 | |
1f78160c XG |
1631 | rcu_read_lock(); |
1632 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1633 | if (!device->bdev) |
1634 | continue; | |
04160088 CM |
1635 | bdi = blk_get_backing_dev_info(device->bdev); |
1636 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1637 | ret = 1; | |
1638 | break; | |
1639 | } | |
1640 | } | |
1f78160c | 1641 | rcu_read_unlock(); |
04160088 CM |
1642 | return ret; |
1643 | } | |
1644 | ||
ad081f14 JA |
1645 | /* |
1646 | * If this fails, caller must call bdi_destroy() to get rid of the | |
1647 | * bdi again. | |
1648 | */ | |
04160088 CM |
1649 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1650 | { | |
ad081f14 JA |
1651 | int err; |
1652 | ||
1653 | bdi->capabilities = BDI_CAP_MAP_COPY; | |
e6d086d8 | 1654 | err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY); |
ad081f14 JA |
1655 | if (err) |
1656 | return err; | |
1657 | ||
4575c9cc | 1658 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1659 | bdi->congested_fn = btrfs_congested_fn; |
1660 | bdi->congested_data = info; | |
1661 | return 0; | |
1662 | } | |
1663 | ||
8b712842 CM |
1664 | /* |
1665 | * called by the kthread helper functions to finally call the bio end_io | |
1666 | * functions. This is where read checksum verification actually happens | |
1667 | */ | |
1668 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1669 | { |
ce9adaa5 | 1670 | struct bio *bio; |
8b712842 | 1671 | struct end_io_wq *end_io_wq; |
ce9adaa5 | 1672 | int error; |
ce9adaa5 | 1673 | |
8b712842 CM |
1674 | end_io_wq = container_of(work, struct end_io_wq, work); |
1675 | bio = end_io_wq->bio; | |
ce9adaa5 | 1676 | |
8b712842 CM |
1677 | error = end_io_wq->error; |
1678 | bio->bi_private = end_io_wq->private; | |
1679 | bio->bi_end_io = end_io_wq->end_io; | |
1680 | kfree(end_io_wq); | |
8b712842 | 1681 | bio_endio(bio, error); |
44b8bd7e CM |
1682 | } |
1683 | ||
a74a4b97 CM |
1684 | static int cleaner_kthread(void *arg) |
1685 | { | |
1686 | struct btrfs_root *root = arg; | |
d0278245 | 1687 | int again; |
a74a4b97 CM |
1688 | |
1689 | do { | |
d0278245 | 1690 | again = 0; |
a74a4b97 | 1691 | |
d0278245 | 1692 | /* Make the cleaner go to sleep early. */ |
babbf170 | 1693 | if (btrfs_need_cleaner_sleep(root)) |
d0278245 MX |
1694 | goto sleep; |
1695 | ||
1696 | if (!mutex_trylock(&root->fs_info->cleaner_mutex)) | |
1697 | goto sleep; | |
1698 | ||
dc7f370c MX |
1699 | /* |
1700 | * Avoid the problem that we change the status of the fs | |
1701 | * during the above check and trylock. | |
1702 | */ | |
babbf170 | 1703 | if (btrfs_need_cleaner_sleep(root)) { |
dc7f370c MX |
1704 | mutex_unlock(&root->fs_info->cleaner_mutex); |
1705 | goto sleep; | |
76dda93c | 1706 | } |
a74a4b97 | 1707 | |
d0278245 MX |
1708 | btrfs_run_delayed_iputs(root); |
1709 | again = btrfs_clean_one_deleted_snapshot(root); | |
1710 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
1711 | ||
1712 | /* | |
05323cd1 MX |
1713 | * The defragger has dealt with the R/O remount and umount, |
1714 | * needn't do anything special here. | |
d0278245 MX |
1715 | */ |
1716 | btrfs_run_defrag_inodes(root->fs_info); | |
1717 | sleep: | |
9d1a2a3a | 1718 | if (!try_to_freeze() && !again) { |
a74a4b97 | 1719 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1720 | if (!kthread_should_stop()) |
1721 | schedule(); | |
a74a4b97 CM |
1722 | __set_current_state(TASK_RUNNING); |
1723 | } | |
1724 | } while (!kthread_should_stop()); | |
1725 | return 0; | |
1726 | } | |
1727 | ||
1728 | static int transaction_kthread(void *arg) | |
1729 | { | |
1730 | struct btrfs_root *root = arg; | |
1731 | struct btrfs_trans_handle *trans; | |
1732 | struct btrfs_transaction *cur; | |
8929ecfa | 1733 | u64 transid; |
a74a4b97 CM |
1734 | unsigned long now; |
1735 | unsigned long delay; | |
914b2007 | 1736 | bool cannot_commit; |
a74a4b97 CM |
1737 | |
1738 | do { | |
914b2007 | 1739 | cannot_commit = false; |
8b87dc17 | 1740 | delay = HZ * root->fs_info->commit_interval; |
a74a4b97 CM |
1741 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
1742 | ||
a4abeea4 | 1743 | spin_lock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1744 | cur = root->fs_info->running_transaction; |
1745 | if (!cur) { | |
a4abeea4 | 1746 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1747 | goto sleep; |
1748 | } | |
31153d81 | 1749 | |
a74a4b97 | 1750 | now = get_seconds(); |
4a9d8bde | 1751 | if (cur->state < TRANS_STATE_BLOCKED && |
8b87dc17 DS |
1752 | (now < cur->start_time || |
1753 | now - cur->start_time < root->fs_info->commit_interval)) { | |
a4abeea4 | 1754 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1755 | delay = HZ * 5; |
1756 | goto sleep; | |
1757 | } | |
8929ecfa | 1758 | transid = cur->transid; |
a4abeea4 | 1759 | spin_unlock(&root->fs_info->trans_lock); |
56bec294 | 1760 | |
79787eaa | 1761 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1762 | trans = btrfs_attach_transaction(root); |
914b2007 | 1763 | if (IS_ERR(trans)) { |
354aa0fb MX |
1764 | if (PTR_ERR(trans) != -ENOENT) |
1765 | cannot_commit = true; | |
79787eaa | 1766 | goto sleep; |
914b2007 | 1767 | } |
8929ecfa | 1768 | if (transid == trans->transid) { |
79787eaa | 1769 | btrfs_commit_transaction(trans, root); |
8929ecfa YZ |
1770 | } else { |
1771 | btrfs_end_transaction(trans, root); | |
1772 | } | |
a74a4b97 CM |
1773 | sleep: |
1774 | wake_up_process(root->fs_info->cleaner_kthread); | |
1775 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1776 | ||
4e121c06 JB |
1777 | if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, |
1778 | &root->fs_info->fs_state))) | |
1779 | btrfs_cleanup_transaction(root); | |
a0acae0e | 1780 | if (!try_to_freeze()) { |
a74a4b97 | 1781 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa | 1782 | if (!kthread_should_stop() && |
914b2007 JK |
1783 | (!btrfs_transaction_blocked(root->fs_info) || |
1784 | cannot_commit)) | |
8929ecfa | 1785 | schedule_timeout(delay); |
a74a4b97 CM |
1786 | __set_current_state(TASK_RUNNING); |
1787 | } | |
1788 | } while (!kthread_should_stop()); | |
1789 | return 0; | |
1790 | } | |
1791 | ||
af31f5e5 CM |
1792 | /* |
1793 | * this will find the highest generation in the array of | |
1794 | * root backups. The index of the highest array is returned, | |
1795 | * or -1 if we can't find anything. | |
1796 | * | |
1797 | * We check to make sure the array is valid by comparing the | |
1798 | * generation of the latest root in the array with the generation | |
1799 | * in the super block. If they don't match we pitch it. | |
1800 | */ | |
1801 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1802 | { | |
1803 | u64 cur; | |
1804 | int newest_index = -1; | |
1805 | struct btrfs_root_backup *root_backup; | |
1806 | int i; | |
1807 | ||
1808 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1809 | root_backup = info->super_copy->super_roots + i; | |
1810 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1811 | if (cur == newest_gen) | |
1812 | newest_index = i; | |
1813 | } | |
1814 | ||
1815 | /* check to see if we actually wrapped around */ | |
1816 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
1817 | root_backup = info->super_copy->super_roots; | |
1818 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1819 | if (cur == newest_gen) | |
1820 | newest_index = 0; | |
1821 | } | |
1822 | return newest_index; | |
1823 | } | |
1824 | ||
1825 | ||
1826 | /* | |
1827 | * find the oldest backup so we know where to store new entries | |
1828 | * in the backup array. This will set the backup_root_index | |
1829 | * field in the fs_info struct | |
1830 | */ | |
1831 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
1832 | u64 newest_gen) | |
1833 | { | |
1834 | int newest_index = -1; | |
1835 | ||
1836 | newest_index = find_newest_super_backup(info, newest_gen); | |
1837 | /* if there was garbage in there, just move along */ | |
1838 | if (newest_index == -1) { | |
1839 | info->backup_root_index = 0; | |
1840 | } else { | |
1841 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1842 | } | |
1843 | } | |
1844 | ||
1845 | /* | |
1846 | * copy all the root pointers into the super backup array. | |
1847 | * this will bump the backup pointer by one when it is | |
1848 | * done | |
1849 | */ | |
1850 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1851 | { | |
1852 | int next_backup; | |
1853 | struct btrfs_root_backup *root_backup; | |
1854 | int last_backup; | |
1855 | ||
1856 | next_backup = info->backup_root_index; | |
1857 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1858 | BTRFS_NUM_BACKUP_ROOTS; | |
1859 | ||
1860 | /* | |
1861 | * just overwrite the last backup if we're at the same generation | |
1862 | * this happens only at umount | |
1863 | */ | |
1864 | root_backup = info->super_for_commit->super_roots + last_backup; | |
1865 | if (btrfs_backup_tree_root_gen(root_backup) == | |
1866 | btrfs_header_generation(info->tree_root->node)) | |
1867 | next_backup = last_backup; | |
1868 | ||
1869 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1870 | ||
1871 | /* | |
1872 | * make sure all of our padding and empty slots get zero filled | |
1873 | * regardless of which ones we use today | |
1874 | */ | |
1875 | memset(root_backup, 0, sizeof(*root_backup)); | |
1876 | ||
1877 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1878 | ||
1879 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1880 | btrfs_set_backup_tree_root_gen(root_backup, | |
1881 | btrfs_header_generation(info->tree_root->node)); | |
1882 | ||
1883 | btrfs_set_backup_tree_root_level(root_backup, | |
1884 | btrfs_header_level(info->tree_root->node)); | |
1885 | ||
1886 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1887 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1888 | btrfs_header_generation(info->chunk_root->node)); | |
1889 | btrfs_set_backup_chunk_root_level(root_backup, | |
1890 | btrfs_header_level(info->chunk_root->node)); | |
1891 | ||
1892 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
1893 | btrfs_set_backup_extent_root_gen(root_backup, | |
1894 | btrfs_header_generation(info->extent_root->node)); | |
1895 | btrfs_set_backup_extent_root_level(root_backup, | |
1896 | btrfs_header_level(info->extent_root->node)); | |
1897 | ||
7c7e82a7 CM |
1898 | /* |
1899 | * we might commit during log recovery, which happens before we set | |
1900 | * the fs_root. Make sure it is valid before we fill it in. | |
1901 | */ | |
1902 | if (info->fs_root && info->fs_root->node) { | |
1903 | btrfs_set_backup_fs_root(root_backup, | |
1904 | info->fs_root->node->start); | |
1905 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1906 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1907 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1908 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1909 | } |
af31f5e5 CM |
1910 | |
1911 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1912 | btrfs_set_backup_dev_root_gen(root_backup, | |
1913 | btrfs_header_generation(info->dev_root->node)); | |
1914 | btrfs_set_backup_dev_root_level(root_backup, | |
1915 | btrfs_header_level(info->dev_root->node)); | |
1916 | ||
1917 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
1918 | btrfs_set_backup_csum_root_gen(root_backup, | |
1919 | btrfs_header_generation(info->csum_root->node)); | |
1920 | btrfs_set_backup_csum_root_level(root_backup, | |
1921 | btrfs_header_level(info->csum_root->node)); | |
1922 | ||
1923 | btrfs_set_backup_total_bytes(root_backup, | |
1924 | btrfs_super_total_bytes(info->super_copy)); | |
1925 | btrfs_set_backup_bytes_used(root_backup, | |
1926 | btrfs_super_bytes_used(info->super_copy)); | |
1927 | btrfs_set_backup_num_devices(root_backup, | |
1928 | btrfs_super_num_devices(info->super_copy)); | |
1929 | ||
1930 | /* | |
1931 | * if we don't copy this out to the super_copy, it won't get remembered | |
1932 | * for the next commit | |
1933 | */ | |
1934 | memcpy(&info->super_copy->super_roots, | |
1935 | &info->super_for_commit->super_roots, | |
1936 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1937 | } | |
1938 | ||
1939 | /* | |
1940 | * this copies info out of the root backup array and back into | |
1941 | * the in-memory super block. It is meant to help iterate through | |
1942 | * the array, so you send it the number of backups you've already | |
1943 | * tried and the last backup index you used. | |
1944 | * | |
1945 | * this returns -1 when it has tried all the backups | |
1946 | */ | |
1947 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
1948 | struct btrfs_super_block *super, | |
1949 | int *num_backups_tried, int *backup_index) | |
1950 | { | |
1951 | struct btrfs_root_backup *root_backup; | |
1952 | int newest = *backup_index; | |
1953 | ||
1954 | if (*num_backups_tried == 0) { | |
1955 | u64 gen = btrfs_super_generation(super); | |
1956 | ||
1957 | newest = find_newest_super_backup(info, gen); | |
1958 | if (newest == -1) | |
1959 | return -1; | |
1960 | ||
1961 | *backup_index = newest; | |
1962 | *num_backups_tried = 1; | |
1963 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
1964 | /* we've tried all the backups, all done */ | |
1965 | return -1; | |
1966 | } else { | |
1967 | /* jump to the next oldest backup */ | |
1968 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1969 | BTRFS_NUM_BACKUP_ROOTS; | |
1970 | *backup_index = newest; | |
1971 | *num_backups_tried += 1; | |
1972 | } | |
1973 | root_backup = super->super_roots + newest; | |
1974 | ||
1975 | btrfs_set_super_generation(super, | |
1976 | btrfs_backup_tree_root_gen(root_backup)); | |
1977 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
1978 | btrfs_set_super_root_level(super, | |
1979 | btrfs_backup_tree_root_level(root_backup)); | |
1980 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
1981 | ||
1982 | /* | |
1983 | * fixme: the total bytes and num_devices need to match or we should | |
1984 | * need a fsck | |
1985 | */ | |
1986 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
1987 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
1988 | return 0; | |
1989 | } | |
1990 | ||
7abadb64 LB |
1991 | /* helper to cleanup workers */ |
1992 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
1993 | { | |
1994 | btrfs_stop_workers(&fs_info->generic_worker); | |
1995 | btrfs_stop_workers(&fs_info->fixup_workers); | |
1996 | btrfs_stop_workers(&fs_info->delalloc_workers); | |
1997 | btrfs_stop_workers(&fs_info->workers); | |
1998 | btrfs_stop_workers(&fs_info->endio_workers); | |
1999 | btrfs_stop_workers(&fs_info->endio_meta_workers); | |
2000 | btrfs_stop_workers(&fs_info->endio_raid56_workers); | |
2001 | btrfs_stop_workers(&fs_info->rmw_workers); | |
2002 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); | |
2003 | btrfs_stop_workers(&fs_info->endio_write_workers); | |
2004 | btrfs_stop_workers(&fs_info->endio_freespace_worker); | |
2005 | btrfs_stop_workers(&fs_info->submit_workers); | |
2006 | btrfs_stop_workers(&fs_info->delayed_workers); | |
2007 | btrfs_stop_workers(&fs_info->caching_workers); | |
2008 | btrfs_stop_workers(&fs_info->readahead_workers); | |
2009 | btrfs_stop_workers(&fs_info->flush_workers); | |
2f232036 | 2010 | btrfs_stop_workers(&fs_info->qgroup_rescan_workers); |
7abadb64 LB |
2011 | } |
2012 | ||
2e9f5954 R |
2013 | static void free_root_extent_buffers(struct btrfs_root *root) |
2014 | { | |
2015 | if (root) { | |
2016 | free_extent_buffer(root->node); | |
2017 | free_extent_buffer(root->commit_root); | |
2018 | root->node = NULL; | |
2019 | root->commit_root = NULL; | |
2020 | } | |
2021 | } | |
2022 | ||
af31f5e5 CM |
2023 | /* helper to cleanup tree roots */ |
2024 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
2025 | { | |
2e9f5954 | 2026 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2027 | |
2e9f5954 R |
2028 | free_root_extent_buffers(info->dev_root); |
2029 | free_root_extent_buffers(info->extent_root); | |
2030 | free_root_extent_buffers(info->csum_root); | |
2031 | free_root_extent_buffers(info->quota_root); | |
2032 | free_root_extent_buffers(info->uuid_root); | |
2033 | if (chunk_root) | |
2034 | free_root_extent_buffers(info->chunk_root); | |
af31f5e5 CM |
2035 | } |
2036 | ||
171f6537 JB |
2037 | static void del_fs_roots(struct btrfs_fs_info *fs_info) |
2038 | { | |
2039 | int ret; | |
2040 | struct btrfs_root *gang[8]; | |
2041 | int i; | |
2042 | ||
2043 | while (!list_empty(&fs_info->dead_roots)) { | |
2044 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2045 | struct btrfs_root, root_list); | |
2046 | list_del(&gang[0]->root_list); | |
2047 | ||
2048 | if (gang[0]->in_radix) { | |
cb517eab | 2049 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); |
171f6537 JB |
2050 | } else { |
2051 | free_extent_buffer(gang[0]->node); | |
2052 | free_extent_buffer(gang[0]->commit_root); | |
b0feb9d9 | 2053 | btrfs_put_fs_root(gang[0]); |
171f6537 JB |
2054 | } |
2055 | } | |
2056 | ||
2057 | while (1) { | |
2058 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2059 | (void **)gang, 0, | |
2060 | ARRAY_SIZE(gang)); | |
2061 | if (!ret) | |
2062 | break; | |
2063 | for (i = 0; i < ret; i++) | |
cb517eab | 2064 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); |
171f6537 JB |
2065 | } |
2066 | } | |
af31f5e5 | 2067 | |
ad2b2c80 AV |
2068 | int open_ctree(struct super_block *sb, |
2069 | struct btrfs_fs_devices *fs_devices, | |
2070 | char *options) | |
2e635a27 | 2071 | { |
db94535d CM |
2072 | u32 sectorsize; |
2073 | u32 nodesize; | |
2074 | u32 leafsize; | |
2075 | u32 blocksize; | |
87ee04eb | 2076 | u32 stripesize; |
84234f3a | 2077 | u64 generation; |
f2b636e8 | 2078 | u64 features; |
3de4586c | 2079 | struct btrfs_key location; |
a061fc8d | 2080 | struct buffer_head *bh; |
4d34b278 | 2081 | struct btrfs_super_block *disk_super; |
815745cf | 2082 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
f84a8bd6 | 2083 | struct btrfs_root *tree_root; |
4d34b278 ID |
2084 | struct btrfs_root *extent_root; |
2085 | struct btrfs_root *csum_root; | |
2086 | struct btrfs_root *chunk_root; | |
2087 | struct btrfs_root *dev_root; | |
bcef60f2 | 2088 | struct btrfs_root *quota_root; |
f7a81ea4 | 2089 | struct btrfs_root *uuid_root; |
e02119d5 | 2090 | struct btrfs_root *log_tree_root; |
eb60ceac | 2091 | int ret; |
e58ca020 | 2092 | int err = -EINVAL; |
af31f5e5 CM |
2093 | int num_backups_tried = 0; |
2094 | int backup_index = 0; | |
70f80175 SB |
2095 | bool create_uuid_tree; |
2096 | bool check_uuid_tree; | |
4543df7e | 2097 | |
f84a8bd6 | 2098 | tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info); |
6f07e42e | 2099 | chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info); |
cb517eab | 2100 | if (!tree_root || !chunk_root) { |
39279cc3 CM |
2101 | err = -ENOMEM; |
2102 | goto fail; | |
2103 | } | |
76dda93c YZ |
2104 | |
2105 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
2106 | if (ret) { | |
2107 | err = ret; | |
2108 | goto fail; | |
2109 | } | |
2110 | ||
2111 | ret = setup_bdi(fs_info, &fs_info->bdi); | |
2112 | if (ret) { | |
2113 | err = ret; | |
2114 | goto fail_srcu; | |
2115 | } | |
2116 | ||
e2d84521 MX |
2117 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0); |
2118 | if (ret) { | |
2119 | err = ret; | |
2120 | goto fail_bdi; | |
2121 | } | |
2122 | fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE * | |
2123 | (1 + ilog2(nr_cpu_ids)); | |
2124 | ||
963d678b MX |
2125 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0); |
2126 | if (ret) { | |
2127 | err = ret; | |
2128 | goto fail_dirty_metadata_bytes; | |
2129 | } | |
2130 | ||
76dda93c YZ |
2131 | fs_info->btree_inode = new_inode(sb); |
2132 | if (!fs_info->btree_inode) { | |
2133 | err = -ENOMEM; | |
963d678b | 2134 | goto fail_delalloc_bytes; |
76dda93c YZ |
2135 | } |
2136 | ||
a6591715 | 2137 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
1561deda | 2138 | |
76dda93c | 2139 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
f28491e0 | 2140 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2141 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2142 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2143 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2144 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2145 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2146 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2147 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 2148 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2149 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2150 | spin_lock_init(&fs_info->defrag_inodes_lock); |
2bf64758 | 2151 | spin_lock_init(&fs_info->free_chunk_lock); |
f29021b2 | 2152 | spin_lock_init(&fs_info->tree_mod_seq_lock); |
ceda0864 | 2153 | spin_lock_init(&fs_info->super_lock); |
f28491e0 | 2154 | spin_lock_init(&fs_info->buffer_lock); |
f29021b2 | 2155 | rwlock_init(&fs_info->tree_mod_log_lock); |
7585717f | 2156 | mutex_init(&fs_info->reloc_mutex); |
de98ced9 | 2157 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2158 | |
58176a96 | 2159 | init_completion(&fs_info->kobj_unregister); |
0b86a832 | 2160 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2161 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2162 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
0b86a832 | 2163 | btrfs_mapping_init(&fs_info->mapping_tree); |
66d8f3dd MX |
2164 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2165 | BTRFS_BLOCK_RSV_GLOBAL); | |
2166 | btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, | |
2167 | BTRFS_BLOCK_RSV_DELALLOC); | |
2168 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); | |
2169 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2170 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2171 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2172 | BTRFS_BLOCK_RSV_DELOPS); | |
cb03c743 | 2173 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 2174 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 2175 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 2176 | atomic_set(&fs_info->nr_async_bios, 0); |
4cb5300b | 2177 | atomic_set(&fs_info->defrag_running, 0); |
fc36ed7e | 2178 | atomic64_set(&fs_info->tree_mod_seq, 0); |
e20d96d6 | 2179 | fs_info->sb = sb; |
6f568d35 | 2180 | fs_info->max_inline = 8192 * 1024; |
9ed74f2d | 2181 | fs_info->metadata_ratio = 0; |
4cb5300b | 2182 | fs_info->defrag_inodes = RB_ROOT; |
2bf64758 | 2183 | fs_info->free_chunk_space = 0; |
f29021b2 | 2184 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2185 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
c8b97818 | 2186 | |
90519d66 AJ |
2187 | /* readahead state */ |
2188 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); | |
2189 | spin_lock_init(&fs_info->reada_lock); | |
c8b97818 | 2190 | |
b34b086c CM |
2191 | fs_info->thread_pool_size = min_t(unsigned long, |
2192 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2193 | |
199c2a9c MX |
2194 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2195 | spin_lock_init(&fs_info->ordered_root_lock); | |
16cdcec7 MX |
2196 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
2197 | GFP_NOFS); | |
2198 | if (!fs_info->delayed_root) { | |
2199 | err = -ENOMEM; | |
2200 | goto fail_iput; | |
2201 | } | |
2202 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 2203 | |
a2de733c AJ |
2204 | mutex_init(&fs_info->scrub_lock); |
2205 | atomic_set(&fs_info->scrubs_running, 0); | |
2206 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2207 | atomic_set(&fs_info->scrubs_paused, 0); | |
2208 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2209 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
a2de733c | 2210 | fs_info->scrub_workers_refcnt = 0; |
21adbd5c SB |
2211 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2212 | fs_info->check_integrity_print_mask = 0; | |
2213 | #endif | |
a2de733c | 2214 | |
c9e9f97b ID |
2215 | spin_lock_init(&fs_info->balance_lock); |
2216 | mutex_init(&fs_info->balance_mutex); | |
837d5b6e ID |
2217 | atomic_set(&fs_info->balance_running, 0); |
2218 | atomic_set(&fs_info->balance_pause_req, 0); | |
a7e99c69 | 2219 | atomic_set(&fs_info->balance_cancel_req, 0); |
c9e9f97b | 2220 | fs_info->balance_ctl = NULL; |
837d5b6e | 2221 | init_waitqueue_head(&fs_info->balance_wait_q); |
a2de733c | 2222 | |
a061fc8d CM |
2223 | sb->s_blocksize = 4096; |
2224 | sb->s_blocksize_bits = blksize_bits(4096); | |
32a88aa1 | 2225 | sb->s_bdi = &fs_info->bdi; |
a061fc8d | 2226 | |
76dda93c | 2227 | fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; |
bfe86848 | 2228 | set_nlink(fs_info->btree_inode, 1); |
0afbaf8c CM |
2229 | /* |
2230 | * we set the i_size on the btree inode to the max possible int. | |
2231 | * the real end of the address space is determined by all of | |
2232 | * the devices in the system | |
2233 | */ | |
2234 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 2235 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
2236 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
2237 | ||
5d4f98a2 | 2238 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); |
d1310b2e | 2239 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
f993c883 | 2240 | fs_info->btree_inode->i_mapping); |
0b32f4bb | 2241 | BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; |
a8067e02 | 2242 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); |
d1310b2e CM |
2243 | |
2244 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 2245 | |
76dda93c YZ |
2246 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
2247 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
2248 | sizeof(struct btrfs_key)); | |
72ac3c0d JB |
2249 | set_bit(BTRFS_INODE_DUMMY, |
2250 | &BTRFS_I(fs_info->btree_inode)->runtime_flags); | |
778ba82b | 2251 | btrfs_insert_inode_hash(fs_info->btree_inode); |
76dda93c | 2252 | |
0f9dd46c | 2253 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2254 | fs_info->block_group_cache_tree = RB_ROOT; |
a1897fdd | 2255 | fs_info->first_logical_byte = (u64)-1; |
0f9dd46c | 2256 | |
11833d66 | 2257 | extent_io_tree_init(&fs_info->freed_extents[0], |
f993c883 | 2258 | fs_info->btree_inode->i_mapping); |
11833d66 | 2259 | extent_io_tree_init(&fs_info->freed_extents[1], |
f993c883 | 2260 | fs_info->btree_inode->i_mapping); |
11833d66 | 2261 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
e66f709b | 2262 | fs_info->do_barriers = 1; |
e18e4809 | 2263 | |
39279cc3 | 2264 | |
5a3f23d5 | 2265 | mutex_init(&fs_info->ordered_operations_mutex); |
9ffba8cd | 2266 | mutex_init(&fs_info->ordered_extent_flush_mutex); |
e02119d5 | 2267 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2268 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2269 | mutex_init(&fs_info->transaction_kthread_mutex); |
2270 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 2271 | mutex_init(&fs_info->volume_mutex); |
276e680d | 2272 | init_rwsem(&fs_info->extent_commit_sem); |
c71bf099 | 2273 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2274 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2275 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
e922e087 SB |
2276 | fs_info->dev_replace.lock_owner = 0; |
2277 | atomic_set(&fs_info->dev_replace.nesting_level, 0); | |
2278 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); | |
2279 | mutex_init(&fs_info->dev_replace.lock_management_lock); | |
2280 | mutex_init(&fs_info->dev_replace.lock); | |
fa9c0d79 | 2281 | |
416ac51d | 2282 | spin_lock_init(&fs_info->qgroup_lock); |
f2f6ed3d | 2283 | mutex_init(&fs_info->qgroup_ioctl_lock); |
416ac51d AJ |
2284 | fs_info->qgroup_tree = RB_ROOT; |
2285 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); | |
2286 | fs_info->qgroup_seq = 1; | |
2287 | fs_info->quota_enabled = 0; | |
2288 | fs_info->pending_quota_state = 0; | |
1e8f9158 | 2289 | fs_info->qgroup_ulist = NULL; |
2f232036 | 2290 | mutex_init(&fs_info->qgroup_rescan_lock); |
416ac51d | 2291 | |
fa9c0d79 CM |
2292 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2293 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2294 | ||
e6dcd2dc | 2295 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2296 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2297 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2298 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2299 | |
53b381b3 DW |
2300 | ret = btrfs_alloc_stripe_hash_table(fs_info); |
2301 | if (ret) { | |
83c8266a | 2302 | err = ret; |
53b381b3 DW |
2303 | goto fail_alloc; |
2304 | } | |
2305 | ||
0b86a832 | 2306 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 2307 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2308 | |
3c4bb26b | 2309 | invalidate_bdev(fs_devices->latest_bdev); |
1104a885 DS |
2310 | |
2311 | /* | |
2312 | * Read super block and check the signature bytes only | |
2313 | */ | |
a512bbf8 | 2314 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
20b45077 DY |
2315 | if (!bh) { |
2316 | err = -EINVAL; | |
16cdcec7 | 2317 | goto fail_alloc; |
20b45077 | 2318 | } |
39279cc3 | 2319 | |
1104a885 DS |
2320 | /* |
2321 | * We want to check superblock checksum, the type is stored inside. | |
2322 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
2323 | */ | |
2324 | if (btrfs_check_super_csum(bh->b_data)) { | |
2325 | printk(KERN_ERR "btrfs: superblock checksum mismatch\n"); | |
2326 | err = -EINVAL; | |
2327 | goto fail_alloc; | |
2328 | } | |
2329 | ||
2330 | /* | |
2331 | * super_copy is zeroed at allocation time and we never touch the | |
2332 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
2333 | * the whole block of INFO_SIZE | |
2334 | */ | |
6c41761f DS |
2335 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2336 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2337 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2338 | brelse(bh); |
5f39d397 | 2339 | |
6c41761f | 2340 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2341 | |
1104a885 DS |
2342 | ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); |
2343 | if (ret) { | |
2344 | printk(KERN_ERR "btrfs: superblock contains fatal errors\n"); | |
2345 | err = -EINVAL; | |
2346 | goto fail_alloc; | |
2347 | } | |
2348 | ||
6c41761f | 2349 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2350 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2351 | goto fail_alloc; |
0f7d52f4 | 2352 | |
acce952b | 2353 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
2354 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
2355 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 2356 | |
af31f5e5 CM |
2357 | /* |
2358 | * run through our array of backup supers and setup | |
2359 | * our ring pointer to the oldest one | |
2360 | */ | |
2361 | generation = btrfs_super_generation(disk_super); | |
2362 | find_oldest_super_backup(fs_info, generation); | |
2363 | ||
75e7cb7f LB |
2364 | /* |
2365 | * In the long term, we'll store the compression type in the super | |
2366 | * block, and it'll be used for per file compression control. | |
2367 | */ | |
2368 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2369 | ||
2b82032c YZ |
2370 | ret = btrfs_parse_options(tree_root, options); |
2371 | if (ret) { | |
2372 | err = ret; | |
16cdcec7 | 2373 | goto fail_alloc; |
2b82032c | 2374 | } |
dfe25020 | 2375 | |
f2b636e8 JB |
2376 | features = btrfs_super_incompat_flags(disk_super) & |
2377 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2378 | if (features) { | |
2379 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
2380 | "unsupported optional features (%Lx).\n", | |
c1c9ff7c | 2381 | features); |
f2b636e8 | 2382 | err = -EINVAL; |
16cdcec7 | 2383 | goto fail_alloc; |
f2b636e8 JB |
2384 | } |
2385 | ||
727011e0 CM |
2386 | if (btrfs_super_leafsize(disk_super) != |
2387 | btrfs_super_nodesize(disk_super)) { | |
2388 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2389 | "blocksizes don't match. node %d leaf %d\n", | |
2390 | btrfs_super_nodesize(disk_super), | |
2391 | btrfs_super_leafsize(disk_super)); | |
2392 | err = -EINVAL; | |
2393 | goto fail_alloc; | |
2394 | } | |
2395 | if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2396 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2397 | "blocksize (%d) was too large\n", | |
2398 | btrfs_super_leafsize(disk_super)); | |
2399 | err = -EINVAL; | |
2400 | goto fail_alloc; | |
2401 | } | |
2402 | ||
5d4f98a2 | 2403 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 2404 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
69e380d1 | 2405 | if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 2406 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
727011e0 | 2407 | |
3173a18f JB |
2408 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
2409 | printk(KERN_ERR "btrfs: has skinny extents\n"); | |
2410 | ||
727011e0 CM |
2411 | /* |
2412 | * flag our filesystem as having big metadata blocks if | |
2413 | * they are bigger than the page size | |
2414 | */ | |
2415 | if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) { | |
2416 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) | |
2417 | printk(KERN_INFO "btrfs flagging fs with big metadata feature\n"); | |
2418 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
2419 | } | |
2420 | ||
bc3f116f CM |
2421 | nodesize = btrfs_super_nodesize(disk_super); |
2422 | leafsize = btrfs_super_leafsize(disk_super); | |
2423 | sectorsize = btrfs_super_sectorsize(disk_super); | |
2424 | stripesize = btrfs_super_stripesize(disk_super); | |
e2d84521 | 2425 | fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids)); |
963d678b | 2426 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); |
bc3f116f CM |
2427 | |
2428 | /* | |
2429 | * mixed block groups end up with duplicate but slightly offset | |
2430 | * extent buffers for the same range. It leads to corruptions | |
2431 | */ | |
2432 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
2433 | (sectorsize != leafsize)) { | |
2434 | printk(KERN_WARNING "btrfs: unequal leaf/node/sector sizes " | |
2435 | "are not allowed for mixed block groups on %s\n", | |
2436 | sb->s_id); | |
2437 | goto fail_alloc; | |
2438 | } | |
2439 | ||
ceda0864 MX |
2440 | /* |
2441 | * Needn't use the lock because there is no other task which will | |
2442 | * update the flag. | |
2443 | */ | |
a6fa6fae | 2444 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 2445 | |
f2b636e8 JB |
2446 | features = btrfs_super_compat_ro_flags(disk_super) & |
2447 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
2448 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
2449 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
2450 | "unsupported option features (%Lx).\n", | |
c1c9ff7c | 2451 | features); |
f2b636e8 | 2452 | err = -EINVAL; |
16cdcec7 | 2453 | goto fail_alloc; |
f2b636e8 | 2454 | } |
61d92c32 CM |
2455 | |
2456 | btrfs_init_workers(&fs_info->generic_worker, | |
2457 | "genwork", 1, NULL); | |
2458 | ||
5443be45 | 2459 | btrfs_init_workers(&fs_info->workers, "worker", |
61d92c32 CM |
2460 | fs_info->thread_pool_size, |
2461 | &fs_info->generic_worker); | |
c8b97818 | 2462 | |
771ed689 | 2463 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
45d5fd14 | 2464 | fs_info->thread_pool_size, NULL); |
771ed689 | 2465 | |
8ccf6f19 | 2466 | btrfs_init_workers(&fs_info->flush_workers, "flush_delalloc", |
45d5fd14 | 2467 | fs_info->thread_pool_size, NULL); |
8ccf6f19 | 2468 | |
5443be45 | 2469 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 | 2470 | min_t(u64, fs_devices->num_devices, |
45d5fd14 | 2471 | fs_info->thread_pool_size), NULL); |
61b49440 | 2472 | |
bab39bf9 | 2473 | btrfs_init_workers(&fs_info->caching_workers, "cache", |
45d5fd14 | 2474 | fs_info->thread_pool_size, NULL); |
bab39bf9 | 2475 | |
61b49440 CM |
2476 | /* a higher idle thresh on the submit workers makes it much more |
2477 | * likely that bios will be send down in a sane order to the | |
2478 | * devices | |
2479 | */ | |
2480 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 2481 | |
771ed689 | 2482 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 2483 | fs_info->workers.ordered = 1; |
61b49440 | 2484 | |
771ed689 CM |
2485 | fs_info->delalloc_workers.idle_thresh = 2; |
2486 | fs_info->delalloc_workers.ordered = 1; | |
2487 | ||
61d92c32 CM |
2488 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1, |
2489 | &fs_info->generic_worker); | |
5443be45 | 2490 | btrfs_init_workers(&fs_info->endio_workers, "endio", |
61d92c32 CM |
2491 | fs_info->thread_pool_size, |
2492 | &fs_info->generic_worker); | |
d20f7043 | 2493 | btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", |
61d92c32 CM |
2494 | fs_info->thread_pool_size, |
2495 | &fs_info->generic_worker); | |
cad321ad | 2496 | btrfs_init_workers(&fs_info->endio_meta_write_workers, |
61d92c32 CM |
2497 | "endio-meta-write", fs_info->thread_pool_size, |
2498 | &fs_info->generic_worker); | |
53b381b3 DW |
2499 | btrfs_init_workers(&fs_info->endio_raid56_workers, |
2500 | "endio-raid56", fs_info->thread_pool_size, | |
2501 | &fs_info->generic_worker); | |
2502 | btrfs_init_workers(&fs_info->rmw_workers, | |
2503 | "rmw", fs_info->thread_pool_size, | |
2504 | &fs_info->generic_worker); | |
5443be45 | 2505 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", |
61d92c32 CM |
2506 | fs_info->thread_pool_size, |
2507 | &fs_info->generic_worker); | |
0cb59c99 JB |
2508 | btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write", |
2509 | 1, &fs_info->generic_worker); | |
16cdcec7 MX |
2510 | btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta", |
2511 | fs_info->thread_pool_size, | |
2512 | &fs_info->generic_worker); | |
90519d66 AJ |
2513 | btrfs_init_workers(&fs_info->readahead_workers, "readahead", |
2514 | fs_info->thread_pool_size, | |
2515 | &fs_info->generic_worker); | |
2f232036 JS |
2516 | btrfs_init_workers(&fs_info->qgroup_rescan_workers, "qgroup-rescan", 1, |
2517 | &fs_info->generic_worker); | |
61b49440 CM |
2518 | |
2519 | /* | |
2520 | * endios are largely parallel and should have a very | |
2521 | * low idle thresh | |
2522 | */ | |
2523 | fs_info->endio_workers.idle_thresh = 4; | |
b51912c9 | 2524 | fs_info->endio_meta_workers.idle_thresh = 4; |
53b381b3 DW |
2525 | fs_info->endio_raid56_workers.idle_thresh = 4; |
2526 | fs_info->rmw_workers.idle_thresh = 2; | |
b51912c9 | 2527 | |
9042846b CM |
2528 | fs_info->endio_write_workers.idle_thresh = 2; |
2529 | fs_info->endio_meta_write_workers.idle_thresh = 2; | |
90519d66 | 2530 | fs_info->readahead_workers.idle_thresh = 2; |
9042846b | 2531 | |
0dc3b84a JB |
2532 | /* |
2533 | * btrfs_start_workers can really only fail because of ENOMEM so just | |
2534 | * return -ENOMEM if any of these fail. | |
2535 | */ | |
2536 | ret = btrfs_start_workers(&fs_info->workers); | |
2537 | ret |= btrfs_start_workers(&fs_info->generic_worker); | |
2538 | ret |= btrfs_start_workers(&fs_info->submit_workers); | |
2539 | ret |= btrfs_start_workers(&fs_info->delalloc_workers); | |
2540 | ret |= btrfs_start_workers(&fs_info->fixup_workers); | |
2541 | ret |= btrfs_start_workers(&fs_info->endio_workers); | |
2542 | ret |= btrfs_start_workers(&fs_info->endio_meta_workers); | |
53b381b3 DW |
2543 | ret |= btrfs_start_workers(&fs_info->rmw_workers); |
2544 | ret |= btrfs_start_workers(&fs_info->endio_raid56_workers); | |
0dc3b84a JB |
2545 | ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers); |
2546 | ret |= btrfs_start_workers(&fs_info->endio_write_workers); | |
2547 | ret |= btrfs_start_workers(&fs_info->endio_freespace_worker); | |
2548 | ret |= btrfs_start_workers(&fs_info->delayed_workers); | |
2549 | ret |= btrfs_start_workers(&fs_info->caching_workers); | |
2550 | ret |= btrfs_start_workers(&fs_info->readahead_workers); | |
8ccf6f19 | 2551 | ret |= btrfs_start_workers(&fs_info->flush_workers); |
2f232036 | 2552 | ret |= btrfs_start_workers(&fs_info->qgroup_rescan_workers); |
0dc3b84a | 2553 | if (ret) { |
fed425c7 | 2554 | err = -ENOMEM; |
0dc3b84a JB |
2555 | goto fail_sb_buffer; |
2556 | } | |
4543df7e | 2557 | |
4575c9cc | 2558 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
2559 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
2560 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 2561 | |
db94535d CM |
2562 | tree_root->nodesize = nodesize; |
2563 | tree_root->leafsize = leafsize; | |
2564 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 2565 | tree_root->stripesize = stripesize; |
a061fc8d CM |
2566 | |
2567 | sb->s_blocksize = sectorsize; | |
2568 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 2569 | |
3cae210f | 2570 | if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) { |
d397712b | 2571 | printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
2572 | goto fail_sb_buffer; |
2573 | } | |
19c00ddc | 2574 | |
8d082fb7 LB |
2575 | if (sectorsize != PAGE_SIZE) { |
2576 | printk(KERN_WARNING "btrfs: Incompatible sector size(%lu) " | |
2577 | "found on %s\n", (unsigned long)sectorsize, sb->s_id); | |
941b2ddf KM |
2578 | goto fail_sb_buffer; |
2579 | } | |
2580 | ||
925baedd | 2581 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 2582 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 2583 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2584 | if (ret) { |
d397712b CM |
2585 | printk(KERN_WARNING "btrfs: failed to read the system " |
2586 | "array on %s\n", sb->s_id); | |
5d4f98a2 | 2587 | goto fail_sb_buffer; |
84eed90f | 2588 | } |
0b86a832 CM |
2589 | |
2590 | blocksize = btrfs_level_size(tree_root, | |
2591 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 2592 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
2593 | |
2594 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2595 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
2596 | ||
2597 | chunk_root->node = read_tree_block(chunk_root, | |
2598 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 2599 | blocksize, generation); |
416bc658 JB |
2600 | if (!chunk_root->node || |
2601 | !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { | |
83121942 DW |
2602 | printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n", |
2603 | sb->s_id); | |
af31f5e5 | 2604 | goto fail_tree_roots; |
83121942 | 2605 | } |
5d4f98a2 YZ |
2606 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
2607 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 2608 | |
e17cade2 | 2609 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
b308bc2f | 2610 | btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); |
e17cade2 | 2611 | |
0b86a832 | 2612 | ret = btrfs_read_chunk_tree(chunk_root); |
2b82032c | 2613 | if (ret) { |
d397712b CM |
2614 | printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", |
2615 | sb->s_id); | |
af31f5e5 | 2616 | goto fail_tree_roots; |
2b82032c | 2617 | } |
0b86a832 | 2618 | |
8dabb742 SB |
2619 | /* |
2620 | * keep the device that is marked to be the target device for the | |
2621 | * dev_replace procedure | |
2622 | */ | |
2623 | btrfs_close_extra_devices(fs_info, fs_devices, 0); | |
dfe25020 | 2624 | |
a6b0d5c8 CM |
2625 | if (!fs_devices->latest_bdev) { |
2626 | printk(KERN_CRIT "btrfs: failed to read devices on %s\n", | |
2627 | sb->s_id); | |
2628 | goto fail_tree_roots; | |
2629 | } | |
2630 | ||
af31f5e5 | 2631 | retry_root_backup: |
db94535d CM |
2632 | blocksize = btrfs_level_size(tree_root, |
2633 | btrfs_super_root_level(disk_super)); | |
84234f3a | 2634 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 2635 | |
e20d96d6 | 2636 | tree_root->node = read_tree_block(tree_root, |
db94535d | 2637 | btrfs_super_root(disk_super), |
84234f3a | 2638 | blocksize, generation); |
af31f5e5 CM |
2639 | if (!tree_root->node || |
2640 | !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { | |
83121942 DW |
2641 | printk(KERN_WARNING "btrfs: failed to read tree root on %s\n", |
2642 | sb->s_id); | |
af31f5e5 CM |
2643 | |
2644 | goto recovery_tree_root; | |
83121942 | 2645 | } |
af31f5e5 | 2646 | |
5d4f98a2 YZ |
2647 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
2648 | tree_root->commit_root = btrfs_root_node(tree_root); | |
69e9c6c6 | 2649 | btrfs_set_root_refs(&tree_root->root_item, 1); |
db94535d | 2650 | |
cb517eab MX |
2651 | location.objectid = BTRFS_EXTENT_TREE_OBJECTID; |
2652 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2653 | location.offset = 0; | |
2654 | ||
2655 | extent_root = btrfs_read_tree_root(tree_root, &location); | |
2656 | if (IS_ERR(extent_root)) { | |
2657 | ret = PTR_ERR(extent_root); | |
af31f5e5 | 2658 | goto recovery_tree_root; |
cb517eab | 2659 | } |
0b86a832 | 2660 | extent_root->track_dirty = 1; |
cb517eab | 2661 | fs_info->extent_root = extent_root; |
0b86a832 | 2662 | |
cb517eab MX |
2663 | location.objectid = BTRFS_DEV_TREE_OBJECTID; |
2664 | dev_root = btrfs_read_tree_root(tree_root, &location); | |
2665 | if (IS_ERR(dev_root)) { | |
2666 | ret = PTR_ERR(dev_root); | |
af31f5e5 | 2667 | goto recovery_tree_root; |
cb517eab | 2668 | } |
5d4f98a2 | 2669 | dev_root->track_dirty = 1; |
cb517eab MX |
2670 | fs_info->dev_root = dev_root; |
2671 | btrfs_init_devices_late(fs_info); | |
3768f368 | 2672 | |
cb517eab MX |
2673 | location.objectid = BTRFS_CSUM_TREE_OBJECTID; |
2674 | csum_root = btrfs_read_tree_root(tree_root, &location); | |
2675 | if (IS_ERR(csum_root)) { | |
2676 | ret = PTR_ERR(csum_root); | |
af31f5e5 | 2677 | goto recovery_tree_root; |
cb517eab | 2678 | } |
d20f7043 | 2679 | csum_root->track_dirty = 1; |
cb517eab | 2680 | fs_info->csum_root = csum_root; |
d20f7043 | 2681 | |
cb517eab MX |
2682 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; |
2683 | quota_root = btrfs_read_tree_root(tree_root, &location); | |
2684 | if (!IS_ERR(quota_root)) { | |
bcef60f2 AJ |
2685 | quota_root->track_dirty = 1; |
2686 | fs_info->quota_enabled = 1; | |
2687 | fs_info->pending_quota_state = 1; | |
cb517eab | 2688 | fs_info->quota_root = quota_root; |
bcef60f2 AJ |
2689 | } |
2690 | ||
f7a81ea4 SB |
2691 | location.objectid = BTRFS_UUID_TREE_OBJECTID; |
2692 | uuid_root = btrfs_read_tree_root(tree_root, &location); | |
2693 | if (IS_ERR(uuid_root)) { | |
2694 | ret = PTR_ERR(uuid_root); | |
2695 | if (ret != -ENOENT) | |
2696 | goto recovery_tree_root; | |
2697 | create_uuid_tree = true; | |
70f80175 | 2698 | check_uuid_tree = false; |
f7a81ea4 SB |
2699 | } else { |
2700 | uuid_root->track_dirty = 1; | |
2701 | fs_info->uuid_root = uuid_root; | |
70f80175 SB |
2702 | create_uuid_tree = false; |
2703 | check_uuid_tree = | |
2704 | generation != btrfs_super_uuid_tree_generation(disk_super); | |
f7a81ea4 SB |
2705 | } |
2706 | ||
8929ecfa YZ |
2707 | fs_info->generation = generation; |
2708 | fs_info->last_trans_committed = generation; | |
8929ecfa | 2709 | |
68310a5e ID |
2710 | ret = btrfs_recover_balance(fs_info); |
2711 | if (ret) { | |
2712 | printk(KERN_WARNING "btrfs: failed to recover balance\n"); | |
2713 | goto fail_block_groups; | |
2714 | } | |
2715 | ||
733f4fbb SB |
2716 | ret = btrfs_init_dev_stats(fs_info); |
2717 | if (ret) { | |
2718 | printk(KERN_ERR "btrfs: failed to init dev_stats: %d\n", | |
2719 | ret); | |
2720 | goto fail_block_groups; | |
2721 | } | |
2722 | ||
8dabb742 SB |
2723 | ret = btrfs_init_dev_replace(fs_info); |
2724 | if (ret) { | |
2725 | pr_err("btrfs: failed to init dev_replace: %d\n", ret); | |
2726 | goto fail_block_groups; | |
2727 | } | |
2728 | ||
2729 | btrfs_close_extra_devices(fs_info, fs_devices, 1); | |
2730 | ||
5ac1d209 JM |
2731 | ret = btrfs_sysfs_add_one(fs_info); |
2732 | if (ret) { | |
2733 | pr_err("btrfs: failed to init sysfs interface: %d\n", ret); | |
2734 | goto fail_block_groups; | |
2735 | } | |
2736 | ||
c59021f8 | 2737 | ret = btrfs_init_space_info(fs_info); |
2738 | if (ret) { | |
2739 | printk(KERN_ERR "Failed to initial space info: %d\n", ret); | |
2740 | goto fail_block_groups; | |
2741 | } | |
2742 | ||
1b1d1f66 JB |
2743 | ret = btrfs_read_block_groups(extent_root); |
2744 | if (ret) { | |
2745 | printk(KERN_ERR "Failed to read block groups: %d\n", ret); | |
2746 | goto fail_block_groups; | |
2747 | } | |
5af3e8cc SB |
2748 | fs_info->num_tolerated_disk_barrier_failures = |
2749 | btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); | |
292fd7fc SB |
2750 | if (fs_info->fs_devices->missing_devices > |
2751 | fs_info->num_tolerated_disk_barrier_failures && | |
2752 | !(sb->s_flags & MS_RDONLY)) { | |
2753 | printk(KERN_WARNING | |
2754 | "Btrfs: too many missing devices, writeable mount is not allowed\n"); | |
2755 | goto fail_block_groups; | |
2756 | } | |
9078a3e1 | 2757 | |
a74a4b97 CM |
2758 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
2759 | "btrfs-cleaner"); | |
57506d50 | 2760 | if (IS_ERR(fs_info->cleaner_kthread)) |
1b1d1f66 | 2761 | goto fail_block_groups; |
a74a4b97 CM |
2762 | |
2763 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
2764 | tree_root, | |
2765 | "btrfs-transaction"); | |
57506d50 | 2766 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 2767 | goto fail_cleaner; |
a74a4b97 | 2768 | |
c289811c CM |
2769 | if (!btrfs_test_opt(tree_root, SSD) && |
2770 | !btrfs_test_opt(tree_root, NOSSD) && | |
2771 | !fs_info->fs_devices->rotating) { | |
2772 | printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD " | |
2773 | "mode\n"); | |
2774 | btrfs_set_opt(fs_info->mount_opt, SSD); | |
2775 | } | |
2776 | ||
21adbd5c SB |
2777 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2778 | if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { | |
2779 | ret = btrfsic_mount(tree_root, fs_devices, | |
2780 | btrfs_test_opt(tree_root, | |
2781 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? | |
2782 | 1 : 0, | |
2783 | fs_info->check_integrity_print_mask); | |
2784 | if (ret) | |
2785 | printk(KERN_WARNING "btrfs: failed to initialize" | |
2786 | " integrity check module %s\n", sb->s_id); | |
2787 | } | |
2788 | #endif | |
bcef60f2 AJ |
2789 | ret = btrfs_read_qgroup_config(fs_info); |
2790 | if (ret) | |
2791 | goto fail_trans_kthread; | |
21adbd5c | 2792 | |
acce952b | 2793 | /* do not make disk changes in broken FS */ |
68ce9682 | 2794 | if (btrfs_super_log_root(disk_super) != 0) { |
e02119d5 CM |
2795 | u64 bytenr = btrfs_super_log_root(disk_super); |
2796 | ||
7c2ca468 | 2797 | if (fs_devices->rw_devices == 0) { |
d397712b CM |
2798 | printk(KERN_WARNING "Btrfs log replay required " |
2799 | "on RO media\n"); | |
7c2ca468 | 2800 | err = -EIO; |
bcef60f2 | 2801 | goto fail_qgroup; |
7c2ca468 | 2802 | } |
e02119d5 CM |
2803 | blocksize = |
2804 | btrfs_level_size(tree_root, | |
2805 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 2806 | |
6f07e42e | 2807 | log_tree_root = btrfs_alloc_root(fs_info); |
676e4c86 DC |
2808 | if (!log_tree_root) { |
2809 | err = -ENOMEM; | |
bcef60f2 | 2810 | goto fail_qgroup; |
676e4c86 | 2811 | } |
e02119d5 CM |
2812 | |
2813 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2814 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
2815 | ||
2816 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
2817 | blocksize, |
2818 | generation + 1); | |
416bc658 JB |
2819 | if (!log_tree_root->node || |
2820 | !extent_buffer_uptodate(log_tree_root->node)) { | |
2821 | printk(KERN_ERR "btrfs: failed to read log tree\n"); | |
2822 | free_extent_buffer(log_tree_root->node); | |
2823 | kfree(log_tree_root); | |
2824 | goto fail_trans_kthread; | |
2825 | } | |
79787eaa | 2826 | /* returns with log_tree_root freed on success */ |
e02119d5 | 2827 | ret = btrfs_recover_log_trees(log_tree_root); |
79787eaa JM |
2828 | if (ret) { |
2829 | btrfs_error(tree_root->fs_info, ret, | |
2830 | "Failed to recover log tree"); | |
2831 | free_extent_buffer(log_tree_root->node); | |
2832 | kfree(log_tree_root); | |
2833 | goto fail_trans_kthread; | |
2834 | } | |
e556ce2c YZ |
2835 | |
2836 | if (sb->s_flags & MS_RDONLY) { | |
79787eaa JM |
2837 | ret = btrfs_commit_super(tree_root); |
2838 | if (ret) | |
2839 | goto fail_trans_kthread; | |
e556ce2c | 2840 | } |
e02119d5 | 2841 | } |
1a40e23b | 2842 | |
76dda93c | 2843 | ret = btrfs_find_orphan_roots(tree_root); |
79787eaa JM |
2844 | if (ret) |
2845 | goto fail_trans_kthread; | |
76dda93c | 2846 | |
7c2ca468 | 2847 | if (!(sb->s_flags & MS_RDONLY)) { |
d68fc57b | 2848 | ret = btrfs_cleanup_fs_roots(fs_info); |
44c44af2 ID |
2849 | if (ret) |
2850 | goto fail_trans_kthread; | |
d68fc57b | 2851 | |
5d4f98a2 | 2852 | ret = btrfs_recover_relocation(tree_root); |
d7ce5843 MX |
2853 | if (ret < 0) { |
2854 | printk(KERN_WARNING | |
2855 | "btrfs: failed to recover relocation\n"); | |
2856 | err = -EINVAL; | |
bcef60f2 | 2857 | goto fail_qgroup; |
d7ce5843 | 2858 | } |
7c2ca468 | 2859 | } |
1a40e23b | 2860 | |
3de4586c CM |
2861 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
2862 | location.type = BTRFS_ROOT_ITEM_KEY; | |
cb517eab | 2863 | location.offset = 0; |
3de4586c | 2864 | |
3de4586c | 2865 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
3140c9a3 DC |
2866 | if (IS_ERR(fs_info->fs_root)) { |
2867 | err = PTR_ERR(fs_info->fs_root); | |
bcef60f2 | 2868 | goto fail_qgroup; |
3140c9a3 | 2869 | } |
c289811c | 2870 | |
2b6ba629 ID |
2871 | if (sb->s_flags & MS_RDONLY) |
2872 | return 0; | |
59641015 | 2873 | |
2b6ba629 ID |
2874 | down_read(&fs_info->cleanup_work_sem); |
2875 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
2876 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
e3acc2a6 | 2877 | up_read(&fs_info->cleanup_work_sem); |
2b6ba629 ID |
2878 | close_ctree(tree_root); |
2879 | return ret; | |
2880 | } | |
2881 | up_read(&fs_info->cleanup_work_sem); | |
59641015 | 2882 | |
2b6ba629 ID |
2883 | ret = btrfs_resume_balance_async(fs_info); |
2884 | if (ret) { | |
2885 | printk(KERN_WARNING "btrfs: failed to resume balance\n"); | |
2886 | close_ctree(tree_root); | |
2887 | return ret; | |
e3acc2a6 JB |
2888 | } |
2889 | ||
8dabb742 SB |
2890 | ret = btrfs_resume_dev_replace_async(fs_info); |
2891 | if (ret) { | |
2892 | pr_warn("btrfs: failed to resume dev_replace\n"); | |
2893 | close_ctree(tree_root); | |
2894 | return ret; | |
2895 | } | |
2896 | ||
b382a324 JS |
2897 | btrfs_qgroup_rescan_resume(fs_info); |
2898 | ||
f7a81ea4 SB |
2899 | if (create_uuid_tree) { |
2900 | pr_info("btrfs: creating UUID tree\n"); | |
2901 | ret = btrfs_create_uuid_tree(fs_info); | |
2902 | if (ret) { | |
2903 | pr_warn("btrfs: failed to create the UUID tree %d\n", | |
2904 | ret); | |
2905 | close_ctree(tree_root); | |
2906 | return ret; | |
2907 | } | |
f420ee1e SB |
2908 | } else if (check_uuid_tree || |
2909 | btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) { | |
70f80175 SB |
2910 | pr_info("btrfs: checking UUID tree\n"); |
2911 | ret = btrfs_check_uuid_tree(fs_info); | |
2912 | if (ret) { | |
2913 | pr_warn("btrfs: failed to check the UUID tree %d\n", | |
2914 | ret); | |
2915 | close_ctree(tree_root); | |
2916 | return ret; | |
2917 | } | |
2918 | } else { | |
2919 | fs_info->update_uuid_tree_gen = 1; | |
f7a81ea4 SB |
2920 | } |
2921 | ||
ad2b2c80 | 2922 | return 0; |
39279cc3 | 2923 | |
bcef60f2 AJ |
2924 | fail_qgroup: |
2925 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
2926 | fail_trans_kthread: |
2927 | kthread_stop(fs_info->transaction_kthread); | |
54067ae9 | 2928 | btrfs_cleanup_transaction(fs_info->tree_root); |
7b5ff90e | 2929 | del_fs_roots(fs_info); |
3f157a2f | 2930 | fail_cleaner: |
a74a4b97 | 2931 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
2932 | |
2933 | /* | |
2934 | * make sure we're done with the btree inode before we stop our | |
2935 | * kthreads | |
2936 | */ | |
2937 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 2938 | |
1b1d1f66 | 2939 | fail_block_groups: |
54067ae9 | 2940 | btrfs_put_block_group_cache(fs_info); |
1b1d1f66 | 2941 | btrfs_free_block_groups(fs_info); |
af31f5e5 CM |
2942 | |
2943 | fail_tree_roots: | |
2944 | free_root_pointers(fs_info, 1); | |
2b8195bb | 2945 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 2946 | |
39279cc3 | 2947 | fail_sb_buffer: |
7abadb64 | 2948 | btrfs_stop_all_workers(fs_info); |
16cdcec7 | 2949 | fail_alloc: |
4543df7e | 2950 | fail_iput: |
586e46e2 ID |
2951 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
2952 | ||
4543df7e | 2953 | iput(fs_info->btree_inode); |
963d678b MX |
2954 | fail_delalloc_bytes: |
2955 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
e2d84521 MX |
2956 | fail_dirty_metadata_bytes: |
2957 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); | |
ad081f14 | 2958 | fail_bdi: |
7e662854 | 2959 | bdi_destroy(&fs_info->bdi); |
76dda93c YZ |
2960 | fail_srcu: |
2961 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 2962 | fail: |
53b381b3 | 2963 | btrfs_free_stripe_hash_table(fs_info); |
586e46e2 | 2964 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 2965 | return err; |
af31f5e5 CM |
2966 | |
2967 | recovery_tree_root: | |
af31f5e5 CM |
2968 | if (!btrfs_test_opt(tree_root, RECOVERY)) |
2969 | goto fail_tree_roots; | |
2970 | ||
2971 | free_root_pointers(fs_info, 0); | |
2972 | ||
2973 | /* don't use the log in recovery mode, it won't be valid */ | |
2974 | btrfs_set_super_log_root(disk_super, 0); | |
2975 | ||
2976 | /* we can't trust the free space cache either */ | |
2977 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2978 | ||
2979 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
2980 | &num_backups_tried, &backup_index); | |
2981 | if (ret == -1) | |
2982 | goto fail_block_groups; | |
2983 | goto retry_root_backup; | |
eb60ceac CM |
2984 | } |
2985 | ||
f2984462 CM |
2986 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
2987 | { | |
f2984462 CM |
2988 | if (uptodate) { |
2989 | set_buffer_uptodate(bh); | |
2990 | } else { | |
442a4f63 SB |
2991 | struct btrfs_device *device = (struct btrfs_device *) |
2992 | bh->b_private; | |
2993 | ||
606686ee JB |
2994 | printk_ratelimited_in_rcu(KERN_WARNING "lost page write due to " |
2995 | "I/O error on %s\n", | |
2996 | rcu_str_deref(device->name)); | |
1259ab75 CM |
2997 | /* note, we dont' set_buffer_write_io_error because we have |
2998 | * our own ways of dealing with the IO errors | |
2999 | */ | |
f2984462 | 3000 | clear_buffer_uptodate(bh); |
442a4f63 | 3001 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); |
f2984462 CM |
3002 | } |
3003 | unlock_buffer(bh); | |
3004 | put_bh(bh); | |
3005 | } | |
3006 | ||
a512bbf8 YZ |
3007 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
3008 | { | |
3009 | struct buffer_head *bh; | |
3010 | struct buffer_head *latest = NULL; | |
3011 | struct btrfs_super_block *super; | |
3012 | int i; | |
3013 | u64 transid = 0; | |
3014 | u64 bytenr; | |
3015 | ||
3016 | /* we would like to check all the supers, but that would make | |
3017 | * a btrfs mount succeed after a mkfs from a different FS. | |
3018 | * So, we need to add a special mount option to scan for | |
3019 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3020 | */ | |
3021 | for (i = 0; i < 1; i++) { | |
3022 | bytenr = btrfs_sb_offset(i); | |
8068a47e AJ |
3023 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3024 | i_size_read(bdev->bd_inode)) | |
a512bbf8 | 3025 | break; |
8068a47e AJ |
3026 | bh = __bread(bdev, bytenr / 4096, |
3027 | BTRFS_SUPER_INFO_SIZE); | |
a512bbf8 YZ |
3028 | if (!bh) |
3029 | continue; | |
3030 | ||
3031 | super = (struct btrfs_super_block *)bh->b_data; | |
3032 | if (btrfs_super_bytenr(super) != bytenr || | |
3cae210f | 3033 | btrfs_super_magic(super) != BTRFS_MAGIC) { |
a512bbf8 YZ |
3034 | brelse(bh); |
3035 | continue; | |
3036 | } | |
3037 | ||
3038 | if (!latest || btrfs_super_generation(super) > transid) { | |
3039 | brelse(latest); | |
3040 | latest = bh; | |
3041 | transid = btrfs_super_generation(super); | |
3042 | } else { | |
3043 | brelse(bh); | |
3044 | } | |
3045 | } | |
3046 | return latest; | |
3047 | } | |
3048 | ||
4eedeb75 HH |
3049 | /* |
3050 | * this should be called twice, once with wait == 0 and | |
3051 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
3052 | * we write are pinned. | |
3053 | * | |
3054 | * They are released when wait == 1 is done. | |
3055 | * max_mirrors must be the same for both runs, and it indicates how | |
3056 | * many supers on this one device should be written. | |
3057 | * | |
3058 | * max_mirrors == 0 means to write them all. | |
3059 | */ | |
a512bbf8 YZ |
3060 | static int write_dev_supers(struct btrfs_device *device, |
3061 | struct btrfs_super_block *sb, | |
3062 | int do_barriers, int wait, int max_mirrors) | |
3063 | { | |
3064 | struct buffer_head *bh; | |
3065 | int i; | |
3066 | int ret; | |
3067 | int errors = 0; | |
3068 | u32 crc; | |
3069 | u64 bytenr; | |
a512bbf8 YZ |
3070 | |
3071 | if (max_mirrors == 0) | |
3072 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3073 | ||
a512bbf8 YZ |
3074 | for (i = 0; i < max_mirrors; i++) { |
3075 | bytenr = btrfs_sb_offset(i); | |
3076 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) | |
3077 | break; | |
3078 | ||
3079 | if (wait) { | |
3080 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
3081 | BTRFS_SUPER_INFO_SIZE); | |
634554dc JB |
3082 | if (!bh) { |
3083 | errors++; | |
3084 | continue; | |
3085 | } | |
a512bbf8 | 3086 | wait_on_buffer(bh); |
4eedeb75 HH |
3087 | if (!buffer_uptodate(bh)) |
3088 | errors++; | |
3089 | ||
3090 | /* drop our reference */ | |
3091 | brelse(bh); | |
3092 | ||
3093 | /* drop the reference from the wait == 0 run */ | |
3094 | brelse(bh); | |
3095 | continue; | |
a512bbf8 YZ |
3096 | } else { |
3097 | btrfs_set_super_bytenr(sb, bytenr); | |
3098 | ||
3099 | crc = ~(u32)0; | |
b0496686 | 3100 | crc = btrfs_csum_data((char *)sb + |
a512bbf8 YZ |
3101 | BTRFS_CSUM_SIZE, crc, |
3102 | BTRFS_SUPER_INFO_SIZE - | |
3103 | BTRFS_CSUM_SIZE); | |
3104 | btrfs_csum_final(crc, sb->csum); | |
3105 | ||
4eedeb75 HH |
3106 | /* |
3107 | * one reference for us, and we leave it for the | |
3108 | * caller | |
3109 | */ | |
a512bbf8 YZ |
3110 | bh = __getblk(device->bdev, bytenr / 4096, |
3111 | BTRFS_SUPER_INFO_SIZE); | |
634554dc JB |
3112 | if (!bh) { |
3113 | printk(KERN_ERR "btrfs: couldn't get super " | |
3114 | "buffer head for bytenr %Lu\n", bytenr); | |
3115 | errors++; | |
3116 | continue; | |
3117 | } | |
3118 | ||
a512bbf8 YZ |
3119 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
3120 | ||
4eedeb75 | 3121 | /* one reference for submit_bh */ |
a512bbf8 | 3122 | get_bh(bh); |
4eedeb75 HH |
3123 | |
3124 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
3125 | lock_buffer(bh); |
3126 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
442a4f63 | 3127 | bh->b_private = device; |
a512bbf8 YZ |
3128 | } |
3129 | ||
387125fc CM |
3130 | /* |
3131 | * we fua the first super. The others we allow | |
3132 | * to go down lazy. | |
3133 | */ | |
21adbd5c | 3134 | ret = btrfsic_submit_bh(WRITE_FUA, bh); |
4eedeb75 | 3135 | if (ret) |
a512bbf8 | 3136 | errors++; |
a512bbf8 YZ |
3137 | } |
3138 | return errors < i ? 0 : -1; | |
3139 | } | |
3140 | ||
387125fc CM |
3141 | /* |
3142 | * endio for the write_dev_flush, this will wake anyone waiting | |
3143 | * for the barrier when it is done | |
3144 | */ | |
3145 | static void btrfs_end_empty_barrier(struct bio *bio, int err) | |
3146 | { | |
3147 | if (err) { | |
3148 | if (err == -EOPNOTSUPP) | |
3149 | set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); | |
3150 | clear_bit(BIO_UPTODATE, &bio->bi_flags); | |
3151 | } | |
3152 | if (bio->bi_private) | |
3153 | complete(bio->bi_private); | |
3154 | bio_put(bio); | |
3155 | } | |
3156 | ||
3157 | /* | |
3158 | * trigger flushes for one the devices. If you pass wait == 0, the flushes are | |
3159 | * sent down. With wait == 1, it waits for the previous flush. | |
3160 | * | |
3161 | * any device where the flush fails with eopnotsupp are flagged as not-barrier | |
3162 | * capable | |
3163 | */ | |
3164 | static int write_dev_flush(struct btrfs_device *device, int wait) | |
3165 | { | |
3166 | struct bio *bio; | |
3167 | int ret = 0; | |
3168 | ||
3169 | if (device->nobarriers) | |
3170 | return 0; | |
3171 | ||
3172 | if (wait) { | |
3173 | bio = device->flush_bio; | |
3174 | if (!bio) | |
3175 | return 0; | |
3176 | ||
3177 | wait_for_completion(&device->flush_wait); | |
3178 | ||
3179 | if (bio_flagged(bio, BIO_EOPNOTSUPP)) { | |
606686ee JB |
3180 | printk_in_rcu("btrfs: disabling barriers on dev %s\n", |
3181 | rcu_str_deref(device->name)); | |
387125fc | 3182 | device->nobarriers = 1; |
5af3e8cc | 3183 | } else if (!bio_flagged(bio, BIO_UPTODATE)) { |
387125fc | 3184 | ret = -EIO; |
5af3e8cc SB |
3185 | btrfs_dev_stat_inc_and_print(device, |
3186 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
387125fc CM |
3187 | } |
3188 | ||
3189 | /* drop the reference from the wait == 0 run */ | |
3190 | bio_put(bio); | |
3191 | device->flush_bio = NULL; | |
3192 | ||
3193 | return ret; | |
3194 | } | |
3195 | ||
3196 | /* | |
3197 | * one reference for us, and we leave it for the | |
3198 | * caller | |
3199 | */ | |
9c017abc | 3200 | device->flush_bio = NULL; |
9be3395b | 3201 | bio = btrfs_io_bio_alloc(GFP_NOFS, 0); |
387125fc CM |
3202 | if (!bio) |
3203 | return -ENOMEM; | |
3204 | ||
3205 | bio->bi_end_io = btrfs_end_empty_barrier; | |
3206 | bio->bi_bdev = device->bdev; | |
3207 | init_completion(&device->flush_wait); | |
3208 | bio->bi_private = &device->flush_wait; | |
3209 | device->flush_bio = bio; | |
3210 | ||
3211 | bio_get(bio); | |
21adbd5c | 3212 | btrfsic_submit_bio(WRITE_FLUSH, bio); |
387125fc CM |
3213 | |
3214 | return 0; | |
3215 | } | |
3216 | ||
3217 | /* | |
3218 | * send an empty flush down to each device in parallel, | |
3219 | * then wait for them | |
3220 | */ | |
3221 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
3222 | { | |
3223 | struct list_head *head; | |
3224 | struct btrfs_device *dev; | |
5af3e8cc SB |
3225 | int errors_send = 0; |
3226 | int errors_wait = 0; | |
387125fc CM |
3227 | int ret; |
3228 | ||
3229 | /* send down all the barriers */ | |
3230 | head = &info->fs_devices->devices; | |
3231 | list_for_each_entry_rcu(dev, head, dev_list) { | |
3232 | if (!dev->bdev) { | |
5af3e8cc | 3233 | errors_send++; |
387125fc CM |
3234 | continue; |
3235 | } | |
3236 | if (!dev->in_fs_metadata || !dev->writeable) | |
3237 | continue; | |
3238 | ||
3239 | ret = write_dev_flush(dev, 0); | |
3240 | if (ret) | |
5af3e8cc | 3241 | errors_send++; |
387125fc CM |
3242 | } |
3243 | ||
3244 | /* wait for all the barriers */ | |
3245 | list_for_each_entry_rcu(dev, head, dev_list) { | |
3246 | if (!dev->bdev) { | |
5af3e8cc | 3247 | errors_wait++; |
387125fc CM |
3248 | continue; |
3249 | } | |
3250 | if (!dev->in_fs_metadata || !dev->writeable) | |
3251 | continue; | |
3252 | ||
3253 | ret = write_dev_flush(dev, 1); | |
3254 | if (ret) | |
5af3e8cc | 3255 | errors_wait++; |
387125fc | 3256 | } |
5af3e8cc SB |
3257 | if (errors_send > info->num_tolerated_disk_barrier_failures || |
3258 | errors_wait > info->num_tolerated_disk_barrier_failures) | |
387125fc CM |
3259 | return -EIO; |
3260 | return 0; | |
3261 | } | |
3262 | ||
5af3e8cc SB |
3263 | int btrfs_calc_num_tolerated_disk_barrier_failures( |
3264 | struct btrfs_fs_info *fs_info) | |
3265 | { | |
3266 | struct btrfs_ioctl_space_info space; | |
3267 | struct btrfs_space_info *sinfo; | |
3268 | u64 types[] = {BTRFS_BLOCK_GROUP_DATA, | |
3269 | BTRFS_BLOCK_GROUP_SYSTEM, | |
3270 | BTRFS_BLOCK_GROUP_METADATA, | |
3271 | BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; | |
3272 | int num_types = 4; | |
3273 | int i; | |
3274 | int c; | |
3275 | int num_tolerated_disk_barrier_failures = | |
3276 | (int)fs_info->fs_devices->num_devices; | |
3277 | ||
3278 | for (i = 0; i < num_types; i++) { | |
3279 | struct btrfs_space_info *tmp; | |
3280 | ||
3281 | sinfo = NULL; | |
3282 | rcu_read_lock(); | |
3283 | list_for_each_entry_rcu(tmp, &fs_info->space_info, list) { | |
3284 | if (tmp->flags == types[i]) { | |
3285 | sinfo = tmp; | |
3286 | break; | |
3287 | } | |
3288 | } | |
3289 | rcu_read_unlock(); | |
3290 | ||
3291 | if (!sinfo) | |
3292 | continue; | |
3293 | ||
3294 | down_read(&sinfo->groups_sem); | |
3295 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
3296 | if (!list_empty(&sinfo->block_groups[c])) { | |
3297 | u64 flags; | |
3298 | ||
3299 | btrfs_get_block_group_info( | |
3300 | &sinfo->block_groups[c], &space); | |
3301 | if (space.total_bytes == 0 || | |
3302 | space.used_bytes == 0) | |
3303 | continue; | |
3304 | flags = space.flags; | |
3305 | /* | |
3306 | * return | |
3307 | * 0: if dup, single or RAID0 is configured for | |
3308 | * any of metadata, system or data, else | |
3309 | * 1: if RAID5 is configured, or if RAID1 or | |
3310 | * RAID10 is configured and only two mirrors | |
3311 | * are used, else | |
3312 | * 2: if RAID6 is configured, else | |
3313 | * num_mirrors - 1: if RAID1 or RAID10 is | |
3314 | * configured and more than | |
3315 | * 2 mirrors are used. | |
3316 | */ | |
3317 | if (num_tolerated_disk_barrier_failures > 0 && | |
3318 | ((flags & (BTRFS_BLOCK_GROUP_DUP | | |
3319 | BTRFS_BLOCK_GROUP_RAID0)) || | |
3320 | ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) | |
3321 | == 0))) | |
3322 | num_tolerated_disk_barrier_failures = 0; | |
53b381b3 DW |
3323 | else if (num_tolerated_disk_barrier_failures > 1) { |
3324 | if (flags & (BTRFS_BLOCK_GROUP_RAID1 | | |
3325 | BTRFS_BLOCK_GROUP_RAID5 | | |
3326 | BTRFS_BLOCK_GROUP_RAID10)) { | |
3327 | num_tolerated_disk_barrier_failures = 1; | |
3328 | } else if (flags & | |
15b0a89d | 3329 | BTRFS_BLOCK_GROUP_RAID6) { |
53b381b3 DW |
3330 | num_tolerated_disk_barrier_failures = 2; |
3331 | } | |
3332 | } | |
5af3e8cc SB |
3333 | } |
3334 | } | |
3335 | up_read(&sinfo->groups_sem); | |
3336 | } | |
3337 | ||
3338 | return num_tolerated_disk_barrier_failures; | |
3339 | } | |
3340 | ||
48a3b636 | 3341 | static int write_all_supers(struct btrfs_root *root, int max_mirrors) |
f2984462 | 3342 | { |
e5e9a520 | 3343 | struct list_head *head; |
f2984462 | 3344 | struct btrfs_device *dev; |
a061fc8d | 3345 | struct btrfs_super_block *sb; |
f2984462 | 3346 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
3347 | int ret; |
3348 | int do_barriers; | |
a236aed1 CM |
3349 | int max_errors; |
3350 | int total_errors = 0; | |
a061fc8d | 3351 | u64 flags; |
f2984462 CM |
3352 | |
3353 | do_barriers = !btrfs_test_opt(root, NOBARRIER); | |
af31f5e5 | 3354 | backup_super_roots(root->fs_info); |
f2984462 | 3355 | |
6c41761f | 3356 | sb = root->fs_info->super_for_commit; |
a061fc8d | 3357 | dev_item = &sb->dev_item; |
e5e9a520 | 3358 | |
174ba509 | 3359 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); |
e5e9a520 | 3360 | head = &root->fs_info->fs_devices->devices; |
d7306801 | 3361 | max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; |
387125fc | 3362 | |
5af3e8cc SB |
3363 | if (do_barriers) { |
3364 | ret = barrier_all_devices(root->fs_info); | |
3365 | if (ret) { | |
3366 | mutex_unlock( | |
3367 | &root->fs_info->fs_devices->device_list_mutex); | |
3368 | btrfs_error(root->fs_info, ret, | |
3369 | "errors while submitting device barriers."); | |
3370 | return ret; | |
3371 | } | |
3372 | } | |
387125fc | 3373 | |
1f78160c | 3374 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3375 | if (!dev->bdev) { |
3376 | total_errors++; | |
3377 | continue; | |
3378 | } | |
2b82032c | 3379 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3380 | continue; |
3381 | ||
2b82032c | 3382 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
3383 | btrfs_set_stack_device_type(dev_item, dev->type); |
3384 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
3385 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
3386 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
3387 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
3388 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
3389 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
3390 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 3391 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 3392 | |
a061fc8d CM |
3393 | flags = btrfs_super_flags(sb); |
3394 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
3395 | ||
a512bbf8 | 3396 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
3397 | if (ret) |
3398 | total_errors++; | |
f2984462 | 3399 | } |
a236aed1 | 3400 | if (total_errors > max_errors) { |
d397712b CM |
3401 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
3402 | total_errors); | |
a724b436 | 3403 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
79787eaa | 3404 | |
9d565ba4 SB |
3405 | /* FUA is masked off if unsupported and can't be the reason */ |
3406 | btrfs_error(root->fs_info, -EIO, | |
3407 | "%d errors while writing supers", total_errors); | |
3408 | return -EIO; | |
a236aed1 | 3409 | } |
f2984462 | 3410 | |
a512bbf8 | 3411 | total_errors = 0; |
1f78160c | 3412 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3413 | if (!dev->bdev) |
3414 | continue; | |
2b82032c | 3415 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3416 | continue; |
3417 | ||
a512bbf8 YZ |
3418 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
3419 | if (ret) | |
3420 | total_errors++; | |
f2984462 | 3421 | } |
174ba509 | 3422 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 3423 | if (total_errors > max_errors) { |
79787eaa JM |
3424 | btrfs_error(root->fs_info, -EIO, |
3425 | "%d errors while writing supers", total_errors); | |
3426 | return -EIO; | |
a236aed1 | 3427 | } |
f2984462 CM |
3428 | return 0; |
3429 | } | |
3430 | ||
a512bbf8 YZ |
3431 | int write_ctree_super(struct btrfs_trans_handle *trans, |
3432 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 3433 | { |
f570e757 | 3434 | return write_all_supers(root, max_mirrors); |
cfaa7295 CM |
3435 | } |
3436 | ||
cb517eab MX |
3437 | /* Drop a fs root from the radix tree and free it. */ |
3438 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
3439 | struct btrfs_root *root) | |
2619ba1f | 3440 | { |
4df27c4d | 3441 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
3442 | radix_tree_delete(&fs_info->fs_roots_radix, |
3443 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 3444 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
3445 | |
3446 | if (btrfs_root_refs(&root->root_item) == 0) | |
3447 | synchronize_srcu(&fs_info->subvol_srcu); | |
3448 | ||
d7634482 | 3449 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { |
3321719e LB |
3450 | btrfs_free_log(NULL, root); |
3451 | btrfs_free_log_root_tree(NULL, fs_info); | |
3452 | } | |
3453 | ||
581bb050 LZ |
3454 | __btrfs_remove_free_space_cache(root->free_ino_pinned); |
3455 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d | 3456 | free_fs_root(root); |
4df27c4d YZ |
3457 | } |
3458 | ||
3459 | static void free_fs_root(struct btrfs_root *root) | |
3460 | { | |
82d5902d | 3461 | iput(root->cache_inode); |
4df27c4d | 3462 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
1cb048f5 FDBM |
3463 | btrfs_free_block_rsv(root, root->orphan_block_rsv); |
3464 | root->orphan_block_rsv = NULL; | |
0ee5dc67 AV |
3465 | if (root->anon_dev) |
3466 | free_anon_bdev(root->anon_dev); | |
4df27c4d YZ |
3467 | free_extent_buffer(root->node); |
3468 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
3469 | kfree(root->free_ino_ctl); |
3470 | kfree(root->free_ino_pinned); | |
d397712b | 3471 | kfree(root->name); |
b0feb9d9 | 3472 | btrfs_put_fs_root(root); |
2619ba1f CM |
3473 | } |
3474 | ||
cb517eab MX |
3475 | void btrfs_free_fs_root(struct btrfs_root *root) |
3476 | { | |
3477 | free_fs_root(root); | |
2619ba1f CM |
3478 | } |
3479 | ||
c146afad | 3480 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 3481 | { |
c146afad YZ |
3482 | u64 root_objectid = 0; |
3483 | struct btrfs_root *gang[8]; | |
3484 | int i; | |
3768f368 | 3485 | int ret; |
e089f05c | 3486 | |
c146afad YZ |
3487 | while (1) { |
3488 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
3489 | (void **)gang, root_objectid, | |
3490 | ARRAY_SIZE(gang)); | |
3491 | if (!ret) | |
3492 | break; | |
5d4f98a2 YZ |
3493 | |
3494 | root_objectid = gang[ret - 1]->root_key.objectid + 1; | |
c146afad | 3495 | for (i = 0; i < ret; i++) { |
66b4ffd1 JB |
3496 | int err; |
3497 | ||
c146afad | 3498 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
3499 | err = btrfs_orphan_cleanup(gang[i]); |
3500 | if (err) | |
3501 | return err; | |
c146afad YZ |
3502 | } |
3503 | root_objectid++; | |
3504 | } | |
3505 | return 0; | |
3506 | } | |
a2135011 | 3507 | |
c146afad YZ |
3508 | int btrfs_commit_super(struct btrfs_root *root) |
3509 | { | |
3510 | struct btrfs_trans_handle *trans; | |
a74a4b97 | 3511 | |
c146afad | 3512 | mutex_lock(&root->fs_info->cleaner_mutex); |
24bbcf04 | 3513 | btrfs_run_delayed_iputs(root); |
c146afad | 3514 | mutex_unlock(&root->fs_info->cleaner_mutex); |
9d1a2a3a | 3515 | wake_up_process(root->fs_info->cleaner_kthread); |
c71bf099 YZ |
3516 | |
3517 | /* wait until ongoing cleanup work done */ | |
3518 | down_write(&root->fs_info->cleanup_work_sem); | |
3519 | up_write(&root->fs_info->cleanup_work_sem); | |
3520 | ||
7a7eaa40 | 3521 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3522 | if (IS_ERR(trans)) |
3523 | return PTR_ERR(trans); | |
d52c1bcc | 3524 | return btrfs_commit_transaction(trans, root); |
c146afad YZ |
3525 | } |
3526 | ||
3527 | int close_ctree(struct btrfs_root *root) | |
3528 | { | |
3529 | struct btrfs_fs_info *fs_info = root->fs_info; | |
3530 | int ret; | |
3531 | ||
3532 | fs_info->closing = 1; | |
3533 | smp_mb(); | |
3534 | ||
803b2f54 SB |
3535 | /* wait for the uuid_scan task to finish */ |
3536 | down(&fs_info->uuid_tree_rescan_sem); | |
3537 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
3538 | up(&fs_info->uuid_tree_rescan_sem); | |
3539 | ||
837d5b6e | 3540 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 3541 | btrfs_pause_balance(fs_info); |
837d5b6e | 3542 | |
8dabb742 SB |
3543 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
3544 | ||
aa1b8cd4 | 3545 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
3546 | |
3547 | /* wait for any defraggers to finish */ | |
3548 | wait_event(fs_info->transaction_wait, | |
3549 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3550 | ||
3551 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 3552 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 3553 | |
c146afad | 3554 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
acce952b | 3555 | ret = btrfs_commit_super(root); |
3556 | if (ret) | |
3557 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
3558 | } | |
3559 | ||
87533c47 | 3560 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) |
68ce9682 | 3561 | btrfs_error_commit_super(root); |
0f7d52f4 | 3562 | |
300e4f8a JB |
3563 | btrfs_put_block_group_cache(fs_info); |
3564 | ||
e3029d9f AV |
3565 | kthread_stop(fs_info->transaction_kthread); |
3566 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 3567 | |
f25784b3 YZ |
3568 | fs_info->closing = 2; |
3569 | smp_mb(); | |
3570 | ||
bcef60f2 AJ |
3571 | btrfs_free_qgroup_config(root->fs_info); |
3572 | ||
963d678b MX |
3573 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
3574 | printk(KERN_INFO "btrfs: at unmount delalloc count %lld\n", | |
3575 | percpu_counter_sum(&fs_info->delalloc_bytes)); | |
b0c68f8b | 3576 | } |
bcc63abb | 3577 | |
5ac1d209 JM |
3578 | btrfs_sysfs_remove_one(fs_info); |
3579 | ||
c146afad | 3580 | del_fs_roots(fs_info); |
d10c5f31 | 3581 | |
2b1360da JB |
3582 | btrfs_free_block_groups(fs_info); |
3583 | ||
96192499 JB |
3584 | btrfs_stop_all_workers(fs_info); |
3585 | ||
13e6c37b | 3586 | free_root_pointers(fs_info, 1); |
9ad6b7bc | 3587 | |
13e6c37b | 3588 | iput(fs_info->btree_inode); |
d6bfde87 | 3589 | |
21adbd5c SB |
3590 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3591 | if (btrfs_test_opt(root, CHECK_INTEGRITY)) | |
3592 | btrfsic_unmount(root, fs_info->fs_devices); | |
3593 | #endif | |
3594 | ||
dfe25020 | 3595 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 3596 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 3597 | |
e2d84521 | 3598 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
963d678b | 3599 | percpu_counter_destroy(&fs_info->delalloc_bytes); |
04160088 | 3600 | bdi_destroy(&fs_info->bdi); |
76dda93c | 3601 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 3602 | |
53b381b3 DW |
3603 | btrfs_free_stripe_hash_table(fs_info); |
3604 | ||
1cb048f5 FDBM |
3605 | btrfs_free_block_rsv(root, root->orphan_block_rsv); |
3606 | root->orphan_block_rsv = NULL; | |
3607 | ||
eb60ceac CM |
3608 | return 0; |
3609 | } | |
3610 | ||
b9fab919 CM |
3611 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
3612 | int atomic) | |
5f39d397 | 3613 | { |
1259ab75 | 3614 | int ret; |
727011e0 | 3615 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 3616 | |
0b32f4bb | 3617 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
3618 | if (!ret) |
3619 | return ret; | |
3620 | ||
3621 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
3622 | parent_transid, atomic); |
3623 | if (ret == -EAGAIN) | |
3624 | return ret; | |
1259ab75 | 3625 | return !ret; |
5f39d397 CM |
3626 | } |
3627 | ||
3628 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 3629 | { |
0b32f4bb | 3630 | return set_extent_buffer_uptodate(buf); |
5f39d397 | 3631 | } |
6702ed49 | 3632 | |
5f39d397 CM |
3633 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
3634 | { | |
06ea65a3 | 3635 | struct btrfs_root *root; |
5f39d397 | 3636 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 3637 | int was_dirty; |
b4ce94de | 3638 | |
06ea65a3 JB |
3639 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
3640 | /* | |
3641 | * This is a fast path so only do this check if we have sanity tests | |
3642 | * enabled. Normal people shouldn't be marking dummy buffers as dirty | |
3643 | * outside of the sanity tests. | |
3644 | */ | |
3645 | if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) | |
3646 | return; | |
3647 | #endif | |
3648 | root = BTRFS_I(buf->pages[0]->mapping->host)->root; | |
b9447ef8 | 3649 | btrfs_assert_tree_locked(buf); |
31b1a2bd JL |
3650 | if (transid != root->fs_info->generation) |
3651 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, " | |
d397712b | 3652 | "found %llu running %llu\n", |
c1c9ff7c | 3653 | buf->start, transid, root->fs_info->generation); |
0b32f4bb | 3654 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 MX |
3655 | if (!was_dirty) |
3656 | __percpu_counter_add(&root->fs_info->dirty_metadata_bytes, | |
3657 | buf->len, | |
3658 | root->fs_info->dirty_metadata_batch); | |
eb60ceac CM |
3659 | } |
3660 | ||
b53d3f5d LB |
3661 | static void __btrfs_btree_balance_dirty(struct btrfs_root *root, |
3662 | int flush_delayed) | |
16cdcec7 MX |
3663 | { |
3664 | /* | |
3665 | * looks as though older kernels can get into trouble with | |
3666 | * this code, they end up stuck in balance_dirty_pages forever | |
3667 | */ | |
e2d84521 | 3668 | int ret; |
16cdcec7 MX |
3669 | |
3670 | if (current->flags & PF_MEMALLOC) | |
3671 | return; | |
3672 | ||
b53d3f5d LB |
3673 | if (flush_delayed) |
3674 | btrfs_balance_delayed_items(root); | |
16cdcec7 | 3675 | |
e2d84521 MX |
3676 | ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes, |
3677 | BTRFS_DIRTY_METADATA_THRESH); | |
3678 | if (ret > 0) { | |
d0e1d66b NJ |
3679 | balance_dirty_pages_ratelimited( |
3680 | root->fs_info->btree_inode->i_mapping); | |
16cdcec7 MX |
3681 | } |
3682 | return; | |
3683 | } | |
3684 | ||
b53d3f5d | 3685 | void btrfs_btree_balance_dirty(struct btrfs_root *root) |
35b7e476 | 3686 | { |
b53d3f5d LB |
3687 | __btrfs_btree_balance_dirty(root, 1); |
3688 | } | |
585ad2c3 | 3689 | |
b53d3f5d LB |
3690 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root) |
3691 | { | |
3692 | __btrfs_btree_balance_dirty(root, 0); | |
35b7e476 | 3693 | } |
6b80053d | 3694 | |
ca7a79ad | 3695 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 3696 | { |
727011e0 | 3697 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
0b32f4bb | 3698 | return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
6b80053d | 3699 | } |
0da5468f | 3700 | |
fcd1f065 | 3701 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 3702 | int read_only) |
3703 | { | |
1104a885 DS |
3704 | /* |
3705 | * Placeholder for checks | |
3706 | */ | |
fcd1f065 | 3707 | return 0; |
acce952b | 3708 | } |
3709 | ||
48a3b636 | 3710 | static void btrfs_error_commit_super(struct btrfs_root *root) |
acce952b | 3711 | { |
acce952b | 3712 | mutex_lock(&root->fs_info->cleaner_mutex); |
3713 | btrfs_run_delayed_iputs(root); | |
3714 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
3715 | ||
3716 | down_write(&root->fs_info->cleanup_work_sem); | |
3717 | up_write(&root->fs_info->cleanup_work_sem); | |
3718 | ||
3719 | /* cleanup FS via transaction */ | |
3720 | btrfs_cleanup_transaction(root); | |
acce952b | 3721 | } |
3722 | ||
569e0f35 JB |
3723 | static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, |
3724 | struct btrfs_root *root) | |
acce952b | 3725 | { |
3726 | struct btrfs_inode *btrfs_inode; | |
3727 | struct list_head splice; | |
3728 | ||
3729 | INIT_LIST_HEAD(&splice); | |
3730 | ||
3731 | mutex_lock(&root->fs_info->ordered_operations_mutex); | |
199c2a9c | 3732 | spin_lock(&root->fs_info->ordered_root_lock); |
acce952b | 3733 | |
569e0f35 | 3734 | list_splice_init(&t->ordered_operations, &splice); |
acce952b | 3735 | while (!list_empty(&splice)) { |
3736 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3737 | ordered_operations); | |
3738 | ||
3739 | list_del_init(&btrfs_inode->ordered_operations); | |
199c2a9c | 3740 | spin_unlock(&root->fs_info->ordered_root_lock); |
acce952b | 3741 | |
3742 | btrfs_invalidate_inodes(btrfs_inode->root); | |
b216cbfb | 3743 | |
199c2a9c | 3744 | spin_lock(&root->fs_info->ordered_root_lock); |
acce952b | 3745 | } |
3746 | ||
199c2a9c | 3747 | spin_unlock(&root->fs_info->ordered_root_lock); |
acce952b | 3748 | mutex_unlock(&root->fs_info->ordered_operations_mutex); |
acce952b | 3749 | } |
3750 | ||
143bede5 | 3751 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 3752 | { |
acce952b | 3753 | struct btrfs_ordered_extent *ordered; |
acce952b | 3754 | |
199c2a9c | 3755 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
3756 | /* |
3757 | * This will just short circuit the ordered completion stuff which will | |
3758 | * make sure the ordered extent gets properly cleaned up. | |
3759 | */ | |
199c2a9c | 3760 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
3761 | root_extent_list) |
3762 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
3763 | spin_unlock(&root->ordered_extent_lock); |
3764 | } | |
3765 | ||
3766 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
3767 | { | |
3768 | struct btrfs_root *root; | |
3769 | struct list_head splice; | |
3770 | ||
3771 | INIT_LIST_HEAD(&splice); | |
3772 | ||
3773 | spin_lock(&fs_info->ordered_root_lock); | |
3774 | list_splice_init(&fs_info->ordered_roots, &splice); | |
3775 | while (!list_empty(&splice)) { | |
3776 | root = list_first_entry(&splice, struct btrfs_root, | |
3777 | ordered_root); | |
1de2cfde JB |
3778 | list_move_tail(&root->ordered_root, |
3779 | &fs_info->ordered_roots); | |
199c2a9c MX |
3780 | |
3781 | btrfs_destroy_ordered_extents(root); | |
3782 | ||
3783 | cond_resched_lock(&fs_info->ordered_root_lock); | |
3784 | } | |
3785 | spin_unlock(&fs_info->ordered_root_lock); | |
acce952b | 3786 | } |
3787 | ||
35a3621b SB |
3788 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
3789 | struct btrfs_root *root) | |
acce952b | 3790 | { |
3791 | struct rb_node *node; | |
3792 | struct btrfs_delayed_ref_root *delayed_refs; | |
3793 | struct btrfs_delayed_ref_node *ref; | |
3794 | int ret = 0; | |
3795 | ||
3796 | delayed_refs = &trans->delayed_refs; | |
3797 | ||
3798 | spin_lock(&delayed_refs->lock); | |
3799 | if (delayed_refs->num_entries == 0) { | |
cfece4db | 3800 | spin_unlock(&delayed_refs->lock); |
acce952b | 3801 | printk(KERN_INFO "delayed_refs has NO entry\n"); |
3802 | return ret; | |
3803 | } | |
3804 | ||
b939d1ab | 3805 | while ((node = rb_first(&delayed_refs->root)) != NULL) { |
eb12db69 | 3806 | struct btrfs_delayed_ref_head *head = NULL; |
e78417d1 | 3807 | bool pin_bytes = false; |
acce952b | 3808 | |
eb12db69 | 3809 | ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); |
acce952b | 3810 | atomic_set(&ref->refs, 1); |
3811 | if (btrfs_delayed_ref_is_head(ref)) { | |
acce952b | 3812 | |
3813 | head = btrfs_delayed_node_to_head(ref); | |
b939d1ab JB |
3814 | if (!mutex_trylock(&head->mutex)) { |
3815 | atomic_inc(&ref->refs); | |
3816 | spin_unlock(&delayed_refs->lock); | |
3817 | ||
3818 | /* Need to wait for the delayed ref to run */ | |
3819 | mutex_lock(&head->mutex); | |
3820 | mutex_unlock(&head->mutex); | |
3821 | btrfs_put_delayed_ref(ref); | |
3822 | ||
e18fca73 | 3823 | spin_lock(&delayed_refs->lock); |
b939d1ab JB |
3824 | continue; |
3825 | } | |
3826 | ||
54067ae9 | 3827 | if (head->must_insert_reserved) |
e78417d1 | 3828 | pin_bytes = true; |
78a6184a | 3829 | btrfs_free_delayed_extent_op(head->extent_op); |
acce952b | 3830 | delayed_refs->num_heads--; |
3831 | if (list_empty(&head->cluster)) | |
3832 | delayed_refs->num_heads_ready--; | |
3833 | list_del_init(&head->cluster); | |
acce952b | 3834 | } |
eb12db69 | 3835 | |
b939d1ab JB |
3836 | ref->in_tree = 0; |
3837 | rb_erase(&ref->rb_node, &delayed_refs->root); | |
c46effa6 LB |
3838 | if (head) |
3839 | rb_erase(&head->href_node, &delayed_refs->href_root); | |
3840 | ||
b939d1ab | 3841 | delayed_refs->num_entries--; |
acce952b | 3842 | spin_unlock(&delayed_refs->lock); |
e78417d1 JB |
3843 | if (head) { |
3844 | if (pin_bytes) | |
3845 | btrfs_pin_extent(root, ref->bytenr, | |
3846 | ref->num_bytes, 1); | |
3847 | mutex_unlock(&head->mutex); | |
3848 | } | |
acce952b | 3849 | btrfs_put_delayed_ref(ref); |
3850 | ||
3851 | cond_resched(); | |
3852 | spin_lock(&delayed_refs->lock); | |
3853 | } | |
3854 | ||
3855 | spin_unlock(&delayed_refs->lock); | |
3856 | ||
3857 | return ret; | |
3858 | } | |
3859 | ||
143bede5 | 3860 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 3861 | { |
3862 | struct btrfs_inode *btrfs_inode; | |
3863 | struct list_head splice; | |
3864 | ||
3865 | INIT_LIST_HEAD(&splice); | |
3866 | ||
eb73c1b7 MX |
3867 | spin_lock(&root->delalloc_lock); |
3868 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 3869 | |
3870 | while (!list_empty(&splice)) { | |
eb73c1b7 MX |
3871 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
3872 | delalloc_inodes); | |
acce952b | 3873 | |
3874 | list_del_init(&btrfs_inode->delalloc_inodes); | |
df0af1a5 MX |
3875 | clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, |
3876 | &btrfs_inode->runtime_flags); | |
eb73c1b7 | 3877 | spin_unlock(&root->delalloc_lock); |
acce952b | 3878 | |
3879 | btrfs_invalidate_inodes(btrfs_inode->root); | |
b216cbfb | 3880 | |
eb73c1b7 | 3881 | spin_lock(&root->delalloc_lock); |
acce952b | 3882 | } |
3883 | ||
eb73c1b7 MX |
3884 | spin_unlock(&root->delalloc_lock); |
3885 | } | |
3886 | ||
3887 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
3888 | { | |
3889 | struct btrfs_root *root; | |
3890 | struct list_head splice; | |
3891 | ||
3892 | INIT_LIST_HEAD(&splice); | |
3893 | ||
3894 | spin_lock(&fs_info->delalloc_root_lock); | |
3895 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
3896 | while (!list_empty(&splice)) { | |
3897 | root = list_first_entry(&splice, struct btrfs_root, | |
3898 | delalloc_root); | |
3899 | list_del_init(&root->delalloc_root); | |
3900 | root = btrfs_grab_fs_root(root); | |
3901 | BUG_ON(!root); | |
3902 | spin_unlock(&fs_info->delalloc_root_lock); | |
3903 | ||
3904 | btrfs_destroy_delalloc_inodes(root); | |
3905 | btrfs_put_fs_root(root); | |
3906 | ||
3907 | spin_lock(&fs_info->delalloc_root_lock); | |
3908 | } | |
3909 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 3910 | } |
3911 | ||
3912 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, | |
3913 | struct extent_io_tree *dirty_pages, | |
3914 | int mark) | |
3915 | { | |
3916 | int ret; | |
acce952b | 3917 | struct extent_buffer *eb; |
3918 | u64 start = 0; | |
3919 | u64 end; | |
acce952b | 3920 | |
3921 | while (1) { | |
3922 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 3923 | mark, NULL); |
acce952b | 3924 | if (ret) |
3925 | break; | |
3926 | ||
3927 | clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); | |
3928 | while (start <= end) { | |
fd8b2b61 JB |
3929 | eb = btrfs_find_tree_block(root, start, |
3930 | root->leafsize); | |
69a85bd8 | 3931 | start += root->leafsize; |
fd8b2b61 | 3932 | if (!eb) |
acce952b | 3933 | continue; |
fd8b2b61 | 3934 | wait_on_extent_buffer_writeback(eb); |
acce952b | 3935 | |
fd8b2b61 JB |
3936 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
3937 | &eb->bflags)) | |
3938 | clear_extent_buffer_dirty(eb); | |
3939 | free_extent_buffer_stale(eb); | |
acce952b | 3940 | } |
3941 | } | |
3942 | ||
3943 | return ret; | |
3944 | } | |
3945 | ||
3946 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
3947 | struct extent_io_tree *pinned_extents) | |
3948 | { | |
3949 | struct extent_io_tree *unpin; | |
3950 | u64 start; | |
3951 | u64 end; | |
3952 | int ret; | |
ed0eaa14 | 3953 | bool loop = true; |
acce952b | 3954 | |
3955 | unpin = pinned_extents; | |
ed0eaa14 | 3956 | again: |
acce952b | 3957 | while (1) { |
3958 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
e6138876 | 3959 | EXTENT_DIRTY, NULL); |
acce952b | 3960 | if (ret) |
3961 | break; | |
3962 | ||
3963 | /* opt_discard */ | |
5378e607 LD |
3964 | if (btrfs_test_opt(root, DISCARD)) |
3965 | ret = btrfs_error_discard_extent(root, start, | |
3966 | end + 1 - start, | |
3967 | NULL); | |
acce952b | 3968 | |
3969 | clear_extent_dirty(unpin, start, end, GFP_NOFS); | |
3970 | btrfs_error_unpin_extent_range(root, start, end); | |
3971 | cond_resched(); | |
3972 | } | |
3973 | ||
ed0eaa14 LB |
3974 | if (loop) { |
3975 | if (unpin == &root->fs_info->freed_extents[0]) | |
3976 | unpin = &root->fs_info->freed_extents[1]; | |
3977 | else | |
3978 | unpin = &root->fs_info->freed_extents[0]; | |
3979 | loop = false; | |
3980 | goto again; | |
3981 | } | |
3982 | ||
acce952b | 3983 | return 0; |
3984 | } | |
3985 | ||
49b25e05 JM |
3986 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
3987 | struct btrfs_root *root) | |
3988 | { | |
724e2315 JB |
3989 | btrfs_destroy_ordered_operations(cur_trans, root); |
3990 | ||
49b25e05 | 3991 | btrfs_destroy_delayed_refs(cur_trans, root); |
49b25e05 | 3992 | |
4a9d8bde | 3993 | cur_trans->state = TRANS_STATE_COMMIT_START; |
d7096fc3 | 3994 | wake_up(&root->fs_info->transaction_blocked_wait); |
49b25e05 | 3995 | |
4a9d8bde | 3996 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
d7096fc3 | 3997 | wake_up(&root->fs_info->transaction_wait); |
49b25e05 | 3998 | |
67cde344 MX |
3999 | btrfs_destroy_delayed_inodes(root); |
4000 | btrfs_assert_delayed_root_empty(root); | |
49b25e05 | 4001 | |
49b25e05 JM |
4002 | btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, |
4003 | EXTENT_DIRTY); | |
6e841e32 LB |
4004 | btrfs_destroy_pinned_extent(root, |
4005 | root->fs_info->pinned_extents); | |
49b25e05 | 4006 | |
4a9d8bde MX |
4007 | cur_trans->state =TRANS_STATE_COMPLETED; |
4008 | wake_up(&cur_trans->commit_wait); | |
4009 | ||
49b25e05 JM |
4010 | /* |
4011 | memset(cur_trans, 0, sizeof(*cur_trans)); | |
4012 | kmem_cache_free(btrfs_transaction_cachep, cur_trans); | |
4013 | */ | |
4014 | } | |
4015 | ||
48a3b636 | 4016 | static int btrfs_cleanup_transaction(struct btrfs_root *root) |
acce952b | 4017 | { |
4018 | struct btrfs_transaction *t; | |
acce952b | 4019 | |
acce952b | 4020 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
4021 | ||
a4abeea4 | 4022 | spin_lock(&root->fs_info->trans_lock); |
724e2315 JB |
4023 | while (!list_empty(&root->fs_info->trans_list)) { |
4024 | t = list_first_entry(&root->fs_info->trans_list, | |
4025 | struct btrfs_transaction, list); | |
4026 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
4027 | atomic_inc(&t->use_count); | |
4028 | spin_unlock(&root->fs_info->trans_lock); | |
4029 | btrfs_wait_for_commit(root, t->transid); | |
4030 | btrfs_put_transaction(t); | |
4031 | spin_lock(&root->fs_info->trans_lock); | |
4032 | continue; | |
4033 | } | |
4034 | if (t == root->fs_info->running_transaction) { | |
4035 | t->state = TRANS_STATE_COMMIT_DOING; | |
4036 | spin_unlock(&root->fs_info->trans_lock); | |
4037 | /* | |
4038 | * We wait for 0 num_writers since we don't hold a trans | |
4039 | * handle open currently for this transaction. | |
4040 | */ | |
4041 | wait_event(t->writer_wait, | |
4042 | atomic_read(&t->num_writers) == 0); | |
4043 | } else { | |
4044 | spin_unlock(&root->fs_info->trans_lock); | |
4045 | } | |
4046 | btrfs_cleanup_one_transaction(t, root); | |
4a9d8bde | 4047 | |
724e2315 JB |
4048 | spin_lock(&root->fs_info->trans_lock); |
4049 | if (t == root->fs_info->running_transaction) | |
4050 | root->fs_info->running_transaction = NULL; | |
acce952b | 4051 | list_del_init(&t->list); |
724e2315 | 4052 | spin_unlock(&root->fs_info->trans_lock); |
acce952b | 4053 | |
724e2315 JB |
4054 | btrfs_put_transaction(t); |
4055 | trace_btrfs_transaction_commit(root); | |
4056 | spin_lock(&root->fs_info->trans_lock); | |
4057 | } | |
4058 | spin_unlock(&root->fs_info->trans_lock); | |
4059 | btrfs_destroy_all_ordered_extents(root->fs_info); | |
4060 | btrfs_destroy_delayed_inodes(root); | |
4061 | btrfs_assert_delayed_root_empty(root); | |
4062 | btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents); | |
4063 | btrfs_destroy_all_delalloc_inodes(root->fs_info); | |
acce952b | 4064 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); |
acce952b | 4065 | |
4066 | return 0; | |
4067 | } | |
4068 | ||
d1310b2e | 4069 | static struct extent_io_ops btree_extent_io_ops = { |
ce9adaa5 | 4070 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
4bb31e92 | 4071 | .readpage_io_failed_hook = btree_io_failed_hook, |
0b86a832 | 4072 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
4073 | /* note we're sharing with inode.c for the merge bio hook */ |
4074 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 4075 | }; |