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> |
5a0e3ad6 | 28 | #include <linux/slab.h> |
784b4e29 | 29 | #include <linux/migrate.h> |
7a36ddec | 30 | #include <linux/ratelimit.h> |
6463fe58 | 31 | #include <linux/uuid.h> |
803b2f54 | 32 | #include <linux/semaphore.h> |
7e75bf3f | 33 | #include <asm/unaligned.h> |
eb60ceac CM |
34 | #include "ctree.h" |
35 | #include "disk-io.h" | |
0b947aff | 36 | #include "hash.h" |
e089f05c | 37 | #include "transaction.h" |
0f7d52f4 | 38 | #include "btrfs_inode.h" |
0b86a832 | 39 | #include "volumes.h" |
db94535d | 40 | #include "print-tree.h" |
925baedd | 41 | #include "locking.h" |
e02119d5 | 42 | #include "tree-log.h" |
fa9c0d79 | 43 | #include "free-space-cache.h" |
70f6d82e | 44 | #include "free-space-tree.h" |
581bb050 | 45 | #include "inode-map.h" |
21adbd5c | 46 | #include "check-integrity.h" |
606686ee | 47 | #include "rcu-string.h" |
8dabb742 | 48 | #include "dev-replace.h" |
53b381b3 | 49 | #include "raid56.h" |
5ac1d209 | 50 | #include "sysfs.h" |
fcebe456 | 51 | #include "qgroup.h" |
ebb8765b | 52 | #include "compression.h" |
eb60ceac | 53 | |
de0022b9 JB |
54 | #ifdef CONFIG_X86 |
55 | #include <asm/cpufeature.h> | |
56 | #endif | |
57 | ||
319e4d06 QW |
58 | #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ |
59 | BTRFS_HEADER_FLAG_RELOC |\ | |
60 | BTRFS_SUPER_FLAG_ERROR |\ | |
61 | BTRFS_SUPER_FLAG_SEEDING |\ | |
62 | BTRFS_SUPER_FLAG_METADUMP) | |
63 | ||
e8c9f186 | 64 | static const struct extent_io_ops btree_extent_io_ops; |
8b712842 | 65 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 66 | static void free_fs_root(struct btrfs_root *root); |
fcd1f065 | 67 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 68 | int read_only); |
143bede5 | 69 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); |
acce952b | 70 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
71 | struct btrfs_root *root); | |
143bede5 | 72 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); |
acce952b | 73 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, |
74 | struct extent_io_tree *dirty_pages, | |
75 | int mark); | |
76 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
77 | struct extent_io_tree *pinned_extents); | |
48a3b636 ES |
78 | static int btrfs_cleanup_transaction(struct btrfs_root *root); |
79 | static void btrfs_error_commit_super(struct btrfs_root *root); | |
ce9adaa5 | 80 | |
d352ac68 | 81 | /* |
97eb6b69 DS |
82 | * btrfs_end_io_wq structs are used to do processing in task context when an IO |
83 | * is complete. This is used during reads to verify checksums, and it is used | |
d352ac68 CM |
84 | * by writes to insert metadata for new file extents after IO is complete. |
85 | */ | |
97eb6b69 | 86 | struct btrfs_end_io_wq { |
ce9adaa5 CM |
87 | struct bio *bio; |
88 | bio_end_io_t *end_io; | |
89 | void *private; | |
90 | struct btrfs_fs_info *info; | |
91 | int error; | |
bfebd8b5 | 92 | enum btrfs_wq_endio_type metadata; |
ce9adaa5 | 93 | struct list_head list; |
8b712842 | 94 | struct btrfs_work work; |
ce9adaa5 | 95 | }; |
0da5468f | 96 | |
97eb6b69 DS |
97 | static struct kmem_cache *btrfs_end_io_wq_cache; |
98 | ||
99 | int __init btrfs_end_io_wq_init(void) | |
100 | { | |
101 | btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq", | |
102 | sizeof(struct btrfs_end_io_wq), | |
103 | 0, | |
fba4b697 | 104 | SLAB_MEM_SPREAD, |
97eb6b69 DS |
105 | NULL); |
106 | if (!btrfs_end_io_wq_cache) | |
107 | return -ENOMEM; | |
108 | return 0; | |
109 | } | |
110 | ||
111 | void btrfs_end_io_wq_exit(void) | |
112 | { | |
5598e900 | 113 | kmem_cache_destroy(btrfs_end_io_wq_cache); |
97eb6b69 DS |
114 | } |
115 | ||
d352ac68 CM |
116 | /* |
117 | * async submit bios are used to offload expensive checksumming | |
118 | * onto the worker threads. They checksum file and metadata bios | |
119 | * just before they are sent down the IO stack. | |
120 | */ | |
44b8bd7e CM |
121 | struct async_submit_bio { |
122 | struct inode *inode; | |
123 | struct bio *bio; | |
124 | struct list_head list; | |
4a69a410 CM |
125 | extent_submit_bio_hook_t *submit_bio_start; |
126 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e | 127 | int mirror_num; |
c8b97818 | 128 | unsigned long bio_flags; |
eaf25d93 CM |
129 | /* |
130 | * bio_offset is optional, can be used if the pages in the bio | |
131 | * can't tell us where in the file the bio should go | |
132 | */ | |
133 | u64 bio_offset; | |
8b712842 | 134 | struct btrfs_work work; |
79787eaa | 135 | int error; |
44b8bd7e CM |
136 | }; |
137 | ||
85d4e461 CM |
138 | /* |
139 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
140 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
141 | * the level the eb occupies in the tree. | |
142 | * | |
143 | * Different roots are used for different purposes and may nest inside each | |
144 | * other and they require separate keysets. As lockdep keys should be | |
145 | * static, assign keysets according to the purpose of the root as indicated | |
146 | * by btrfs_root->objectid. This ensures that all special purpose roots | |
147 | * have separate keysets. | |
4008c04a | 148 | * |
85d4e461 CM |
149 | * Lock-nesting across peer nodes is always done with the immediate parent |
150 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
151 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 152 | * |
85d4e461 CM |
153 | * The key is set by the readpage_end_io_hook after the buffer has passed |
154 | * csum validation but before the pages are unlocked. It is also set by | |
155 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 156 | * |
85d4e461 CM |
157 | * We also add a check to make sure the highest level of the tree is the |
158 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
159 | * needs update as well. | |
4008c04a CM |
160 | */ |
161 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
162 | # if BTRFS_MAX_LEVEL != 8 | |
163 | # error | |
164 | # endif | |
85d4e461 CM |
165 | |
166 | static struct btrfs_lockdep_keyset { | |
167 | u64 id; /* root objectid */ | |
168 | const char *name_stem; /* lock name stem */ | |
169 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
170 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
171 | } btrfs_lockdep_keysets[] = { | |
172 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
173 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
174 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
175 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
176 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
177 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
60b62978 | 178 | { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, |
85d4e461 CM |
179 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, |
180 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
181 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
13fd8da9 | 182 | { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, |
6b20e0ad | 183 | { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" }, |
85d4e461 | 184 | { .id = 0, .name_stem = "tree" }, |
4008c04a | 185 | }; |
85d4e461 CM |
186 | |
187 | void __init btrfs_init_lockdep(void) | |
188 | { | |
189 | int i, j; | |
190 | ||
191 | /* initialize lockdep class names */ | |
192 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
193 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
194 | ||
195 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
196 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
197 | "btrfs-%s-%02d", ks->name_stem, j); | |
198 | } | |
199 | } | |
200 | ||
201 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
202 | int level) | |
203 | { | |
204 | struct btrfs_lockdep_keyset *ks; | |
205 | ||
206 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
207 | ||
208 | /* find the matching keyset, id 0 is the default entry */ | |
209 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
210 | if (ks->id == objectid) | |
211 | break; | |
212 | ||
213 | lockdep_set_class_and_name(&eb->lock, | |
214 | &ks->keys[level], ks->names[level]); | |
215 | } | |
216 | ||
4008c04a CM |
217 | #endif |
218 | ||
d352ac68 CM |
219 | /* |
220 | * extents on the btree inode are pretty simple, there's one extent | |
221 | * that covers the entire device | |
222 | */ | |
b2950863 | 223 | static struct extent_map *btree_get_extent(struct inode *inode, |
306e16ce | 224 | struct page *page, size_t pg_offset, u64 start, u64 len, |
b2950863 | 225 | int create) |
7eccb903 | 226 | { |
5f39d397 CM |
227 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
228 | struct extent_map *em; | |
229 | int ret; | |
230 | ||
890871be | 231 | read_lock(&em_tree->lock); |
d1310b2e | 232 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d CM |
233 | if (em) { |
234 | em->bdev = | |
235 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
890871be | 236 | read_unlock(&em_tree->lock); |
5f39d397 | 237 | goto out; |
a061fc8d | 238 | } |
890871be | 239 | read_unlock(&em_tree->lock); |
7b13b7b1 | 240 | |
172ddd60 | 241 | em = alloc_extent_map(); |
5f39d397 CM |
242 | if (!em) { |
243 | em = ERR_PTR(-ENOMEM); | |
244 | goto out; | |
245 | } | |
246 | em->start = 0; | |
0afbaf8c | 247 | em->len = (u64)-1; |
c8b97818 | 248 | em->block_len = (u64)-1; |
5f39d397 | 249 | em->block_start = 0; |
a061fc8d | 250 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e | 251 | |
890871be | 252 | write_lock(&em_tree->lock); |
09a2a8f9 | 253 | ret = add_extent_mapping(em_tree, em, 0); |
5f39d397 CM |
254 | if (ret == -EEXIST) { |
255 | free_extent_map(em); | |
7b13b7b1 | 256 | em = lookup_extent_mapping(em_tree, start, len); |
b4f359ab | 257 | if (!em) |
0433f20d | 258 | em = ERR_PTR(-EIO); |
5f39d397 | 259 | } else if (ret) { |
7b13b7b1 | 260 | free_extent_map(em); |
0433f20d | 261 | em = ERR_PTR(ret); |
5f39d397 | 262 | } |
890871be | 263 | write_unlock(&em_tree->lock); |
7b13b7b1 | 264 | |
5f39d397 CM |
265 | out: |
266 | return em; | |
7eccb903 CM |
267 | } |
268 | ||
b0496686 | 269 | u32 btrfs_csum_data(char *data, u32 seed, size_t len) |
19c00ddc | 270 | { |
0b947aff | 271 | return btrfs_crc32c(seed, data, len); |
19c00ddc CM |
272 | } |
273 | ||
274 | void btrfs_csum_final(u32 crc, char *result) | |
275 | { | |
7e75bf3f | 276 | put_unaligned_le32(~crc, result); |
19c00ddc CM |
277 | } |
278 | ||
d352ac68 CM |
279 | /* |
280 | * compute the csum for a btree block, and either verify it or write it | |
281 | * into the csum field of the block. | |
282 | */ | |
01d58472 DD |
283 | static int csum_tree_block(struct btrfs_fs_info *fs_info, |
284 | struct extent_buffer *buf, | |
19c00ddc CM |
285 | int verify) |
286 | { | |
01d58472 | 287 | u16 csum_size = btrfs_super_csum_size(fs_info->super_copy); |
607d432d | 288 | char *result = NULL; |
19c00ddc CM |
289 | unsigned long len; |
290 | unsigned long cur_len; | |
291 | unsigned long offset = BTRFS_CSUM_SIZE; | |
19c00ddc CM |
292 | char *kaddr; |
293 | unsigned long map_start; | |
294 | unsigned long map_len; | |
295 | int err; | |
296 | u32 crc = ~(u32)0; | |
607d432d | 297 | unsigned long inline_result; |
19c00ddc CM |
298 | |
299 | len = buf->len - offset; | |
d397712b | 300 | while (len > 0) { |
19c00ddc | 301 | err = map_private_extent_buffer(buf, offset, 32, |
a6591715 | 302 | &kaddr, &map_start, &map_len); |
d397712b | 303 | if (err) |
8bd98f0e | 304 | return err; |
19c00ddc | 305 | cur_len = min(len, map_len - (offset - map_start)); |
b0496686 | 306 | crc = btrfs_csum_data(kaddr + offset - map_start, |
19c00ddc CM |
307 | crc, cur_len); |
308 | len -= cur_len; | |
309 | offset += cur_len; | |
19c00ddc | 310 | } |
607d432d | 311 | if (csum_size > sizeof(inline_result)) { |
31e818fe | 312 | result = kzalloc(csum_size, GFP_NOFS); |
607d432d | 313 | if (!result) |
8bd98f0e | 314 | return -ENOMEM; |
607d432d JB |
315 | } else { |
316 | result = (char *)&inline_result; | |
317 | } | |
318 | ||
19c00ddc CM |
319 | btrfs_csum_final(crc, result); |
320 | ||
321 | if (verify) { | |
607d432d | 322 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
323 | u32 val; |
324 | u32 found = 0; | |
607d432d | 325 | memcpy(&found, result, csum_size); |
e4204ded | 326 | |
607d432d | 327 | read_extent_buffer(buf, &val, 0, csum_size); |
94647322 DS |
328 | btrfs_warn_rl(fs_info, |
329 | "%s checksum verify failed on %llu wanted %X found %X " | |
330 | "level %d", | |
01d58472 | 331 | fs_info->sb->s_id, buf->start, |
efe120a0 | 332 | val, found, btrfs_header_level(buf)); |
607d432d JB |
333 | if (result != (char *)&inline_result) |
334 | kfree(result); | |
8bd98f0e | 335 | return -EUCLEAN; |
19c00ddc CM |
336 | } |
337 | } else { | |
607d432d | 338 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 339 | } |
607d432d JB |
340 | if (result != (char *)&inline_result) |
341 | kfree(result); | |
19c00ddc CM |
342 | return 0; |
343 | } | |
344 | ||
d352ac68 CM |
345 | /* |
346 | * we can't consider a given block up to date unless the transid of the | |
347 | * block matches the transid in the parent node's pointer. This is how we | |
348 | * detect blocks that either didn't get written at all or got written | |
349 | * in the wrong place. | |
350 | */ | |
1259ab75 | 351 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
352 | struct extent_buffer *eb, u64 parent_transid, |
353 | int atomic) | |
1259ab75 | 354 | { |
2ac55d41 | 355 | struct extent_state *cached_state = NULL; |
1259ab75 | 356 | int ret; |
2755a0de | 357 | bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB); |
1259ab75 CM |
358 | |
359 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
360 | return 0; | |
361 | ||
b9fab919 CM |
362 | if (atomic) |
363 | return -EAGAIN; | |
364 | ||
a26e8c9f JB |
365 | if (need_lock) { |
366 | btrfs_tree_read_lock(eb); | |
367 | btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); | |
368 | } | |
369 | ||
2ac55d41 | 370 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
ff13db41 | 371 | &cached_state); |
0b32f4bb | 372 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
373 | btrfs_header_generation(eb) == parent_transid) { |
374 | ret = 0; | |
375 | goto out; | |
376 | } | |
94647322 DS |
377 | btrfs_err_rl(eb->fs_info, |
378 | "parent transid verify failed on %llu wanted %llu found %llu", | |
379 | eb->start, | |
29549aec | 380 | parent_transid, btrfs_header_generation(eb)); |
1259ab75 | 381 | ret = 1; |
a26e8c9f JB |
382 | |
383 | /* | |
384 | * Things reading via commit roots that don't have normal protection, | |
385 | * like send, can have a really old block in cache that may point at a | |
01327610 | 386 | * block that has been freed and re-allocated. So don't clear uptodate |
a26e8c9f JB |
387 | * if we find an eb that is under IO (dirty/writeback) because we could |
388 | * end up reading in the stale data and then writing it back out and | |
389 | * making everybody very sad. | |
390 | */ | |
391 | if (!extent_buffer_under_io(eb)) | |
392 | clear_extent_buffer_uptodate(eb); | |
33958dc6 | 393 | out: |
2ac55d41 JB |
394 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
395 | &cached_state, GFP_NOFS); | |
472b909f JB |
396 | if (need_lock) |
397 | btrfs_tree_read_unlock_blocking(eb); | |
1259ab75 | 398 | return ret; |
1259ab75 CM |
399 | } |
400 | ||
1104a885 DS |
401 | /* |
402 | * Return 0 if the superblock checksum type matches the checksum value of that | |
403 | * algorithm. Pass the raw disk superblock data. | |
404 | */ | |
405 | static int btrfs_check_super_csum(char *raw_disk_sb) | |
406 | { | |
407 | struct btrfs_super_block *disk_sb = | |
408 | (struct btrfs_super_block *)raw_disk_sb; | |
409 | u16 csum_type = btrfs_super_csum_type(disk_sb); | |
410 | int ret = 0; | |
411 | ||
412 | if (csum_type == BTRFS_CSUM_TYPE_CRC32) { | |
413 | u32 crc = ~(u32)0; | |
414 | const int csum_size = sizeof(crc); | |
415 | char result[csum_size]; | |
416 | ||
417 | /* | |
418 | * The super_block structure does not span the whole | |
419 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space | |
01327610 | 420 | * is filled with zeros and is included in the checksum. |
1104a885 DS |
421 | */ |
422 | crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE, | |
423 | crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
424 | btrfs_csum_final(crc, result); | |
425 | ||
426 | if (memcmp(raw_disk_sb, result, csum_size)) | |
427 | ret = 1; | |
428 | } | |
429 | ||
430 | if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) { | |
efe120a0 | 431 | printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n", |
1104a885 DS |
432 | csum_type); |
433 | ret = 1; | |
434 | } | |
435 | ||
436 | return ret; | |
437 | } | |
438 | ||
d352ac68 CM |
439 | /* |
440 | * helper to read a given tree block, doing retries as required when | |
441 | * the checksums don't match and we have alternate mirrors to try. | |
442 | */ | |
f188591e CM |
443 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
444 | struct extent_buffer *eb, | |
ca7a79ad | 445 | u64 start, u64 parent_transid) |
f188591e CM |
446 | { |
447 | struct extent_io_tree *io_tree; | |
ea466794 | 448 | int failed = 0; |
f188591e CM |
449 | int ret; |
450 | int num_copies = 0; | |
451 | int mirror_num = 0; | |
ea466794 | 452 | int failed_mirror = 0; |
f188591e | 453 | |
a826d6dc | 454 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
f188591e CM |
455 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
456 | while (1) { | |
bb82ab88 AJ |
457 | ret = read_extent_buffer_pages(io_tree, eb, start, |
458 | WAIT_COMPLETE, | |
f188591e | 459 | btree_get_extent, mirror_num); |
256dd1bb SB |
460 | if (!ret) { |
461 | if (!verify_parent_transid(io_tree, eb, | |
b9fab919 | 462 | parent_transid, 0)) |
256dd1bb SB |
463 | break; |
464 | else | |
465 | ret = -EIO; | |
466 | } | |
d397712b | 467 | |
a826d6dc JB |
468 | /* |
469 | * This buffer's crc is fine, but its contents are corrupted, so | |
470 | * there is no reason to read the other copies, they won't be | |
471 | * any less wrong. | |
472 | */ | |
473 | if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) | |
ea466794 JB |
474 | break; |
475 | ||
5d964051 | 476 | num_copies = btrfs_num_copies(root->fs_info, |
f188591e | 477 | eb->start, eb->len); |
4235298e | 478 | if (num_copies == 1) |
ea466794 | 479 | break; |
4235298e | 480 | |
5cf1ab56 JB |
481 | if (!failed_mirror) { |
482 | failed = 1; | |
483 | failed_mirror = eb->read_mirror; | |
484 | } | |
485 | ||
f188591e | 486 | mirror_num++; |
ea466794 JB |
487 | if (mirror_num == failed_mirror) |
488 | mirror_num++; | |
489 | ||
4235298e | 490 | if (mirror_num > num_copies) |
ea466794 | 491 | break; |
f188591e | 492 | } |
ea466794 | 493 | |
c0901581 | 494 | if (failed && !ret && failed_mirror) |
ea466794 JB |
495 | repair_eb_io_failure(root, eb, failed_mirror); |
496 | ||
497 | return ret; | |
f188591e | 498 | } |
19c00ddc | 499 | |
d352ac68 | 500 | /* |
d397712b CM |
501 | * checksum a dirty tree block before IO. This has extra checks to make sure |
502 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 503 | */ |
d397712b | 504 | |
01d58472 | 505 | static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page) |
19c00ddc | 506 | { |
4eee4fa4 | 507 | u64 start = page_offset(page); |
19c00ddc | 508 | u64 found_start; |
19c00ddc | 509 | struct extent_buffer *eb; |
f188591e | 510 | |
4f2de97a JB |
511 | eb = (struct extent_buffer *)page->private; |
512 | if (page != eb->pages[0]) | |
513 | return 0; | |
0f805531 | 514 | |
19c00ddc | 515 | found_start = btrfs_header_bytenr(eb); |
0f805531 AL |
516 | /* |
517 | * Please do not consolidate these warnings into a single if. | |
518 | * It is useful to know what went wrong. | |
519 | */ | |
520 | if (WARN_ON(found_start != start)) | |
521 | return -EUCLEAN; | |
522 | if (WARN_ON(!PageUptodate(page))) | |
523 | return -EUCLEAN; | |
524 | ||
525 | ASSERT(memcmp_extent_buffer(eb, fs_info->fsid, | |
526 | btrfs_header_fsid(), BTRFS_FSID_SIZE) == 0); | |
527 | ||
8bd98f0e | 528 | return csum_tree_block(fs_info, eb, 0); |
19c00ddc CM |
529 | } |
530 | ||
01d58472 | 531 | static int check_tree_block_fsid(struct btrfs_fs_info *fs_info, |
2b82032c YZ |
532 | struct extent_buffer *eb) |
533 | { | |
01d58472 | 534 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; |
2b82032c YZ |
535 | u8 fsid[BTRFS_UUID_SIZE]; |
536 | int ret = 1; | |
537 | ||
0a4e5586 | 538 | read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); |
2b82032c YZ |
539 | while (fs_devices) { |
540 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
541 | ret = 0; | |
542 | break; | |
543 | } | |
544 | fs_devices = fs_devices->seed; | |
545 | } | |
546 | return ret; | |
547 | } | |
548 | ||
a826d6dc | 549 | #define CORRUPT(reason, eb, root, slot) \ |
efe120a0 FH |
550 | btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \ |
551 | "root=%llu, slot=%d", reason, \ | |
c1c9ff7c | 552 | btrfs_header_bytenr(eb), root->objectid, slot) |
a826d6dc JB |
553 | |
554 | static noinline int check_leaf(struct btrfs_root *root, | |
555 | struct extent_buffer *leaf) | |
556 | { | |
557 | struct btrfs_key key; | |
558 | struct btrfs_key leaf_key; | |
559 | u32 nritems = btrfs_header_nritems(leaf); | |
560 | int slot; | |
561 | ||
1ba98d08 LB |
562 | if (nritems == 0) { |
563 | struct btrfs_root *check_root; | |
564 | ||
565 | key.objectid = btrfs_header_owner(leaf); | |
566 | key.type = BTRFS_ROOT_ITEM_KEY; | |
567 | key.offset = (u64)-1; | |
568 | ||
569 | check_root = btrfs_get_fs_root(root->fs_info, &key, false); | |
570 | /* | |
571 | * The only reason we also check NULL here is that during | |
572 | * open_ctree() some roots has not yet been set up. | |
573 | */ | |
574 | if (!IS_ERR_OR_NULL(check_root)) { | |
575 | /* if leaf is the root, then it's fine */ | |
576 | if (leaf->start != | |
577 | btrfs_root_bytenr(&check_root->root_item)) { | |
578 | CORRUPT("non-root leaf's nritems is 0", | |
579 | leaf, root, 0); | |
580 | return -EIO; | |
581 | } | |
582 | } | |
a826d6dc | 583 | return 0; |
1ba98d08 | 584 | } |
a826d6dc JB |
585 | |
586 | /* Check the 0 item */ | |
587 | if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != | |
588 | BTRFS_LEAF_DATA_SIZE(root)) { | |
589 | CORRUPT("invalid item offset size pair", leaf, root, 0); | |
590 | return -EIO; | |
591 | } | |
592 | ||
593 | /* | |
594 | * Check to make sure each items keys are in the correct order and their | |
595 | * offsets make sense. We only have to loop through nritems-1 because | |
596 | * we check the current slot against the next slot, which verifies the | |
597 | * next slot's offset+size makes sense and that the current's slot | |
598 | * offset is correct. | |
599 | */ | |
600 | for (slot = 0; slot < nritems - 1; slot++) { | |
601 | btrfs_item_key_to_cpu(leaf, &leaf_key, slot); | |
602 | btrfs_item_key_to_cpu(leaf, &key, slot + 1); | |
603 | ||
604 | /* Make sure the keys are in the right order */ | |
605 | if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { | |
606 | CORRUPT("bad key order", leaf, root, slot); | |
607 | return -EIO; | |
608 | } | |
609 | ||
610 | /* | |
611 | * Make sure the offset and ends are right, remember that the | |
612 | * item data starts at the end of the leaf and grows towards the | |
613 | * front. | |
614 | */ | |
615 | if (btrfs_item_offset_nr(leaf, slot) != | |
616 | btrfs_item_end_nr(leaf, slot + 1)) { | |
617 | CORRUPT("slot offset bad", leaf, root, slot); | |
618 | return -EIO; | |
619 | } | |
620 | ||
621 | /* | |
622 | * Check to make sure that we don't point outside of the leaf, | |
01327610 | 623 | * just in case all the items are consistent to each other, but |
a826d6dc JB |
624 | * all point outside of the leaf. |
625 | */ | |
626 | if (btrfs_item_end_nr(leaf, slot) > | |
627 | BTRFS_LEAF_DATA_SIZE(root)) { | |
628 | CORRUPT("slot end outside of leaf", leaf, root, slot); | |
629 | return -EIO; | |
630 | } | |
631 | } | |
632 | ||
633 | return 0; | |
634 | } | |
635 | ||
053ab70f LB |
636 | static int check_node(struct btrfs_root *root, struct extent_buffer *node) |
637 | { | |
638 | unsigned long nr = btrfs_header_nritems(node); | |
639 | ||
640 | if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(root)) { | |
641 | btrfs_crit(root->fs_info, | |
642 | "corrupt node: block %llu root %llu nritems %lu", | |
643 | node->start, root->objectid, nr); | |
644 | return -EIO; | |
645 | } | |
646 | return 0; | |
647 | } | |
648 | ||
facc8a22 MX |
649 | static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, |
650 | u64 phy_offset, struct page *page, | |
651 | u64 start, u64 end, int mirror) | |
ce9adaa5 | 652 | { |
ce9adaa5 CM |
653 | u64 found_start; |
654 | int found_level; | |
ce9adaa5 CM |
655 | struct extent_buffer *eb; |
656 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
02873e43 | 657 | struct btrfs_fs_info *fs_info = root->fs_info; |
f188591e | 658 | int ret = 0; |
727011e0 | 659 | int reads_done; |
ce9adaa5 | 660 | |
ce9adaa5 CM |
661 | if (!page->private) |
662 | goto out; | |
d397712b | 663 | |
4f2de97a | 664 | eb = (struct extent_buffer *)page->private; |
d397712b | 665 | |
0b32f4bb JB |
666 | /* the pending IO might have been the only thing that kept this buffer |
667 | * in memory. Make sure we have a ref for all this other checks | |
668 | */ | |
669 | extent_buffer_get(eb); | |
670 | ||
671 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
727011e0 CM |
672 | if (!reads_done) |
673 | goto err; | |
f188591e | 674 | |
5cf1ab56 | 675 | eb->read_mirror = mirror; |
656f30db | 676 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { |
ea466794 JB |
677 | ret = -EIO; |
678 | goto err; | |
679 | } | |
680 | ||
ce9adaa5 | 681 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 682 | if (found_start != eb->start) { |
02873e43 ZL |
683 | btrfs_err_rl(fs_info, "bad tree block start %llu %llu", |
684 | found_start, eb->start); | |
f188591e | 685 | ret = -EIO; |
ce9adaa5 CM |
686 | goto err; |
687 | } | |
02873e43 ZL |
688 | if (check_tree_block_fsid(fs_info, eb)) { |
689 | btrfs_err_rl(fs_info, "bad fsid on block %llu", | |
690 | eb->start); | |
1259ab75 CM |
691 | ret = -EIO; |
692 | goto err; | |
693 | } | |
ce9adaa5 | 694 | found_level = btrfs_header_level(eb); |
1c24c3ce | 695 | if (found_level >= BTRFS_MAX_LEVEL) { |
02873e43 ZL |
696 | btrfs_err(fs_info, "bad tree block level %d", |
697 | (int)btrfs_header_level(eb)); | |
1c24c3ce JB |
698 | ret = -EIO; |
699 | goto err; | |
700 | } | |
ce9adaa5 | 701 | |
85d4e461 CM |
702 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
703 | eb, found_level); | |
4008c04a | 704 | |
02873e43 | 705 | ret = csum_tree_block(fs_info, eb, 1); |
8bd98f0e | 706 | if (ret) |
a826d6dc | 707 | goto err; |
a826d6dc JB |
708 | |
709 | /* | |
710 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
711 | * that we don't try and read the other copies of this block, just | |
712 | * return -EIO. | |
713 | */ | |
714 | if (found_level == 0 && check_leaf(root, eb)) { | |
715 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); | |
716 | ret = -EIO; | |
717 | } | |
ce9adaa5 | 718 | |
053ab70f LB |
719 | if (found_level > 0 && check_node(root, eb)) |
720 | ret = -EIO; | |
721 | ||
0b32f4bb JB |
722 | if (!ret) |
723 | set_extent_buffer_uptodate(eb); | |
ce9adaa5 | 724 | err: |
79fb65a1 JB |
725 | if (reads_done && |
726 | test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
02873e43 | 727 | btree_readahead_hook(fs_info, eb, eb->start, ret); |
4bb31e92 | 728 | |
53b381b3 DW |
729 | if (ret) { |
730 | /* | |
731 | * our io error hook is going to dec the io pages | |
732 | * again, we have to make sure it has something | |
733 | * to decrement | |
734 | */ | |
735 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 736 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 737 | } |
0b32f4bb | 738 | free_extent_buffer(eb); |
ce9adaa5 | 739 | out: |
f188591e | 740 | return ret; |
ce9adaa5 CM |
741 | } |
742 | ||
ea466794 | 743 | static int btree_io_failed_hook(struct page *page, int failed_mirror) |
4bb31e92 | 744 | { |
4bb31e92 | 745 | struct extent_buffer *eb; |
4bb31e92 | 746 | |
4f2de97a | 747 | eb = (struct extent_buffer *)page->private; |
656f30db | 748 | set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags); |
5cf1ab56 | 749 | eb->read_mirror = failed_mirror; |
53b381b3 | 750 | atomic_dec(&eb->io_pages); |
ea466794 | 751 | if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) |
02873e43 | 752 | btree_readahead_hook(eb->fs_info, eb, eb->start, -EIO); |
4bb31e92 AJ |
753 | return -EIO; /* we fixed nothing */ |
754 | } | |
755 | ||
4246a0b6 | 756 | static void end_workqueue_bio(struct bio *bio) |
ce9adaa5 | 757 | { |
97eb6b69 | 758 | struct btrfs_end_io_wq *end_io_wq = bio->bi_private; |
ce9adaa5 | 759 | struct btrfs_fs_info *fs_info; |
9e0af237 LB |
760 | struct btrfs_workqueue *wq; |
761 | btrfs_work_func_t func; | |
ce9adaa5 | 762 | |
ce9adaa5 | 763 | fs_info = end_io_wq->info; |
4246a0b6 | 764 | end_io_wq->error = bio->bi_error; |
d20f7043 | 765 | |
37226b21 | 766 | if (bio_op(bio) == REQ_OP_WRITE) { |
9e0af237 LB |
767 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) { |
768 | wq = fs_info->endio_meta_write_workers; | |
769 | func = btrfs_endio_meta_write_helper; | |
770 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) { | |
771 | wq = fs_info->endio_freespace_worker; | |
772 | func = btrfs_freespace_write_helper; | |
773 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { | |
774 | wq = fs_info->endio_raid56_workers; | |
775 | func = btrfs_endio_raid56_helper; | |
776 | } else { | |
777 | wq = fs_info->endio_write_workers; | |
778 | func = btrfs_endio_write_helper; | |
779 | } | |
d20f7043 | 780 | } else { |
8b110e39 MX |
781 | if (unlikely(end_io_wq->metadata == |
782 | BTRFS_WQ_ENDIO_DIO_REPAIR)) { | |
783 | wq = fs_info->endio_repair_workers; | |
784 | func = btrfs_endio_repair_helper; | |
785 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { | |
9e0af237 LB |
786 | wq = fs_info->endio_raid56_workers; |
787 | func = btrfs_endio_raid56_helper; | |
788 | } else if (end_io_wq->metadata) { | |
789 | wq = fs_info->endio_meta_workers; | |
790 | func = btrfs_endio_meta_helper; | |
791 | } else { | |
792 | wq = fs_info->endio_workers; | |
793 | func = btrfs_endio_helper; | |
794 | } | |
d20f7043 | 795 | } |
9e0af237 LB |
796 | |
797 | btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL); | |
798 | btrfs_queue_work(wq, &end_io_wq->work); | |
ce9adaa5 CM |
799 | } |
800 | ||
22c59948 | 801 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
bfebd8b5 | 802 | enum btrfs_wq_endio_type metadata) |
0b86a832 | 803 | { |
97eb6b69 | 804 | struct btrfs_end_io_wq *end_io_wq; |
8b110e39 | 805 | |
97eb6b69 | 806 | end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); |
ce9adaa5 CM |
807 | if (!end_io_wq) |
808 | return -ENOMEM; | |
809 | ||
810 | end_io_wq->private = bio->bi_private; | |
811 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 812 | end_io_wq->info = info; |
ce9adaa5 CM |
813 | end_io_wq->error = 0; |
814 | end_io_wq->bio = bio; | |
22c59948 | 815 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
816 | |
817 | bio->bi_private = end_io_wq; | |
818 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
819 | return 0; |
820 | } | |
821 | ||
b64a2851 | 822 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 823 | { |
4854ddd0 | 824 | unsigned long limit = min_t(unsigned long, |
5cdc7ad3 | 825 | info->thread_pool_size, |
4854ddd0 CM |
826 | info->fs_devices->open_devices); |
827 | return 256 * limit; | |
828 | } | |
0986fe9e | 829 | |
4a69a410 CM |
830 | static void run_one_async_start(struct btrfs_work *work) |
831 | { | |
4a69a410 | 832 | struct async_submit_bio *async; |
79787eaa | 833 | int ret; |
4a69a410 CM |
834 | |
835 | async = container_of(work, struct async_submit_bio, work); | |
81a75f67 | 836 | ret = async->submit_bio_start(async->inode, async->bio, |
79787eaa JM |
837 | async->mirror_num, async->bio_flags, |
838 | async->bio_offset); | |
839 | if (ret) | |
840 | async->error = ret; | |
4a69a410 CM |
841 | } |
842 | ||
843 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
844 | { |
845 | struct btrfs_fs_info *fs_info; | |
846 | struct async_submit_bio *async; | |
4854ddd0 | 847 | int limit; |
8b712842 CM |
848 | |
849 | async = container_of(work, struct async_submit_bio, work); | |
850 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 851 | |
b64a2851 | 852 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
853 | limit = limit * 2 / 3; |
854 | ||
ee863954 DS |
855 | /* |
856 | * atomic_dec_return implies a barrier for waitqueue_active | |
857 | */ | |
66657b31 | 858 | if (atomic_dec_return(&fs_info->nr_async_submits) < limit && |
b64a2851 | 859 | waitqueue_active(&fs_info->async_submit_wait)) |
4854ddd0 CM |
860 | wake_up(&fs_info->async_submit_wait); |
861 | ||
bb7ab3b9 | 862 | /* If an error occurred we just want to clean up the bio and move on */ |
79787eaa | 863 | if (async->error) { |
4246a0b6 CH |
864 | async->bio->bi_error = async->error; |
865 | bio_endio(async->bio); | |
79787eaa JM |
866 | return; |
867 | } | |
868 | ||
81a75f67 MC |
869 | async->submit_bio_done(async->inode, async->bio, async->mirror_num, |
870 | async->bio_flags, async->bio_offset); | |
4a69a410 CM |
871 | } |
872 | ||
873 | static void run_one_async_free(struct btrfs_work *work) | |
874 | { | |
875 | struct async_submit_bio *async; | |
876 | ||
877 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
878 | kfree(async); |
879 | } | |
880 | ||
44b8bd7e | 881 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
81a75f67 | 882 | struct bio *bio, int mirror_num, |
c8b97818 | 883 | unsigned long bio_flags, |
eaf25d93 | 884 | u64 bio_offset, |
4a69a410 CM |
885 | extent_submit_bio_hook_t *submit_bio_start, |
886 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
887 | { |
888 | struct async_submit_bio *async; | |
889 | ||
890 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
891 | if (!async) | |
892 | return -ENOMEM; | |
893 | ||
894 | async->inode = inode; | |
44b8bd7e CM |
895 | async->bio = bio; |
896 | async->mirror_num = mirror_num; | |
4a69a410 CM |
897 | async->submit_bio_start = submit_bio_start; |
898 | async->submit_bio_done = submit_bio_done; | |
899 | ||
9e0af237 | 900 | btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start, |
5cdc7ad3 | 901 | run_one_async_done, run_one_async_free); |
4a69a410 | 902 | |
c8b97818 | 903 | async->bio_flags = bio_flags; |
eaf25d93 | 904 | async->bio_offset = bio_offset; |
8c8bee1d | 905 | |
79787eaa JM |
906 | async->error = 0; |
907 | ||
cb03c743 | 908 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 | 909 | |
1eff9d32 | 910 | if (bio->bi_opf & REQ_SYNC) |
5cdc7ad3 | 911 | btrfs_set_work_high_priority(&async->work); |
d313d7a3 | 912 | |
5cdc7ad3 | 913 | btrfs_queue_work(fs_info->workers, &async->work); |
9473f16c | 914 | |
d397712b | 915 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
916 | atomic_read(&fs_info->nr_async_submits)) { |
917 | wait_event(fs_info->async_submit_wait, | |
918 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
919 | } | |
920 | ||
44b8bd7e CM |
921 | return 0; |
922 | } | |
923 | ||
ce3ed71a CM |
924 | static int btree_csum_one_bio(struct bio *bio) |
925 | { | |
2c30c71b | 926 | struct bio_vec *bvec; |
ce3ed71a | 927 | struct btrfs_root *root; |
2c30c71b | 928 | int i, ret = 0; |
ce3ed71a | 929 | |
2c30c71b | 930 | bio_for_each_segment_all(bvec, bio, i) { |
ce3ed71a | 931 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
01d58472 | 932 | ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); |
79787eaa JM |
933 | if (ret) |
934 | break; | |
ce3ed71a | 935 | } |
2c30c71b | 936 | |
79787eaa | 937 | return ret; |
ce3ed71a CM |
938 | } |
939 | ||
81a75f67 MC |
940 | static int __btree_submit_bio_start(struct inode *inode, struct bio *bio, |
941 | int mirror_num, unsigned long bio_flags, | |
eaf25d93 | 942 | u64 bio_offset) |
22c59948 | 943 | { |
8b712842 CM |
944 | /* |
945 | * when we're called for a write, we're already in the async | |
5443be45 | 946 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 947 | */ |
79787eaa | 948 | return btree_csum_one_bio(bio); |
4a69a410 | 949 | } |
22c59948 | 950 | |
81a75f67 | 951 | static int __btree_submit_bio_done(struct inode *inode, struct bio *bio, |
eaf25d93 CM |
952 | int mirror_num, unsigned long bio_flags, |
953 | u64 bio_offset) | |
4a69a410 | 954 | { |
61891923 SB |
955 | int ret; |
956 | ||
8b712842 | 957 | /* |
4a69a410 CM |
958 | * when we're called for a write, we're already in the async |
959 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 960 | */ |
81a75f67 | 961 | ret = btrfs_map_bio(BTRFS_I(inode)->root, bio, mirror_num, 1); |
4246a0b6 CH |
962 | if (ret) { |
963 | bio->bi_error = ret; | |
964 | bio_endio(bio); | |
965 | } | |
61891923 | 966 | return ret; |
0b86a832 CM |
967 | } |
968 | ||
de0022b9 JB |
969 | static int check_async_write(struct inode *inode, unsigned long bio_flags) |
970 | { | |
971 | if (bio_flags & EXTENT_BIO_TREE_LOG) | |
972 | return 0; | |
973 | #ifdef CONFIG_X86 | |
bc696ca0 | 974 | if (static_cpu_has(X86_FEATURE_XMM4_2)) |
de0022b9 JB |
975 | return 0; |
976 | #endif | |
977 | return 1; | |
978 | } | |
979 | ||
81a75f67 | 980 | static int btree_submit_bio_hook(struct inode *inode, struct bio *bio, |
eaf25d93 CM |
981 | int mirror_num, unsigned long bio_flags, |
982 | u64 bio_offset) | |
44b8bd7e | 983 | { |
de0022b9 | 984 | int async = check_async_write(inode, bio_flags); |
cad321ad CM |
985 | int ret; |
986 | ||
37226b21 | 987 | if (bio_op(bio) != REQ_OP_WRITE) { |
4a69a410 CM |
988 | /* |
989 | * called for a read, do the setup so that checksum validation | |
990 | * can happen in the async kernel threads | |
991 | */ | |
f3f266ab | 992 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, |
bfebd8b5 | 993 | bio, BTRFS_WQ_ENDIO_METADATA); |
1d4284bd | 994 | if (ret) |
61891923 | 995 | goto out_w_error; |
81a75f67 | 996 | ret = btrfs_map_bio(BTRFS_I(inode)->root, bio, mirror_num, 0); |
de0022b9 JB |
997 | } else if (!async) { |
998 | ret = btree_csum_one_bio(bio); | |
999 | if (ret) | |
61891923 | 1000 | goto out_w_error; |
81a75f67 | 1001 | ret = btrfs_map_bio(BTRFS_I(inode)->root, bio, mirror_num, 0); |
61891923 SB |
1002 | } else { |
1003 | /* | |
1004 | * kthread helpers are used to submit writes so that | |
1005 | * checksumming can happen in parallel across all CPUs | |
1006 | */ | |
1007 | ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, | |
81a75f67 | 1008 | inode, bio, mirror_num, 0, |
61891923 SB |
1009 | bio_offset, |
1010 | __btree_submit_bio_start, | |
1011 | __btree_submit_bio_done); | |
44b8bd7e | 1012 | } |
d313d7a3 | 1013 | |
4246a0b6 CH |
1014 | if (ret) |
1015 | goto out_w_error; | |
1016 | return 0; | |
1017 | ||
61891923 | 1018 | out_w_error: |
4246a0b6 CH |
1019 | bio->bi_error = ret; |
1020 | bio_endio(bio); | |
61891923 | 1021 | return ret; |
44b8bd7e CM |
1022 | } |
1023 | ||
3dd1462e | 1024 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1025 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
1026 | struct page *newpage, struct page *page, |
1027 | enum migrate_mode mode) | |
784b4e29 CM |
1028 | { |
1029 | /* | |
1030 | * we can't safely write a btree page from here, | |
1031 | * we haven't done the locking hook | |
1032 | */ | |
1033 | if (PageDirty(page)) | |
1034 | return -EAGAIN; | |
1035 | /* | |
1036 | * Buffers may be managed in a filesystem specific way. | |
1037 | * We must have no buffers or drop them. | |
1038 | */ | |
1039 | if (page_has_private(page) && | |
1040 | !try_to_release_page(page, GFP_KERNEL)) | |
1041 | return -EAGAIN; | |
a6bc32b8 | 1042 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 1043 | } |
3dd1462e | 1044 | #endif |
784b4e29 | 1045 | |
0da5468f CM |
1046 | |
1047 | static int btree_writepages(struct address_space *mapping, | |
1048 | struct writeback_control *wbc) | |
1049 | { | |
e2d84521 MX |
1050 | struct btrfs_fs_info *fs_info; |
1051 | int ret; | |
1052 | ||
d8d5f3e1 | 1053 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
1054 | |
1055 | if (wbc->for_kupdate) | |
1056 | return 0; | |
1057 | ||
e2d84521 | 1058 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 1059 | /* this is a bit racy, but that's ok */ |
e2d84521 MX |
1060 | ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
1061 | BTRFS_DIRTY_METADATA_THRESH); | |
1062 | if (ret < 0) | |
793955bc | 1063 | return 0; |
793955bc | 1064 | } |
0b32f4bb | 1065 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
1066 | } |
1067 | ||
b2950863 | 1068 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 1069 | { |
d1310b2e CM |
1070 | struct extent_io_tree *tree; |
1071 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 1072 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 1073 | } |
22b0ebda | 1074 | |
70dec807 | 1075 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 1076 | { |
98509cfc | 1077 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 1078 | return 0; |
0c4e538b | 1079 | |
f7a52a40 | 1080 | return try_release_extent_buffer(page); |
d98237b3 CM |
1081 | } |
1082 | ||
d47992f8 LC |
1083 | static void btree_invalidatepage(struct page *page, unsigned int offset, |
1084 | unsigned int length) | |
d98237b3 | 1085 | { |
d1310b2e CM |
1086 | struct extent_io_tree *tree; |
1087 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
1088 | extent_invalidatepage(tree, page, offset); |
1089 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 1090 | if (PagePrivate(page)) { |
efe120a0 FH |
1091 | btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, |
1092 | "page private not zero on page %llu", | |
1093 | (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
1094 | ClearPagePrivate(page); |
1095 | set_page_private(page, 0); | |
09cbfeaf | 1096 | put_page(page); |
9ad6b7bc | 1097 | } |
d98237b3 CM |
1098 | } |
1099 | ||
0b32f4bb JB |
1100 | static int btree_set_page_dirty(struct page *page) |
1101 | { | |
bb146eb2 | 1102 | #ifdef DEBUG |
0b32f4bb JB |
1103 | struct extent_buffer *eb; |
1104 | ||
1105 | BUG_ON(!PagePrivate(page)); | |
1106 | eb = (struct extent_buffer *)page->private; | |
1107 | BUG_ON(!eb); | |
1108 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1109 | BUG_ON(!atomic_read(&eb->refs)); | |
1110 | btrfs_assert_tree_locked(eb); | |
bb146eb2 | 1111 | #endif |
0b32f4bb JB |
1112 | return __set_page_dirty_nobuffers(page); |
1113 | } | |
1114 | ||
7f09410b | 1115 | static const struct address_space_operations btree_aops = { |
d98237b3 | 1116 | .readpage = btree_readpage, |
0da5468f | 1117 | .writepages = btree_writepages, |
5f39d397 CM |
1118 | .releasepage = btree_releasepage, |
1119 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1120 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1121 | .migratepage = btree_migratepage, |
5a92bc88 | 1122 | #endif |
0b32f4bb | 1123 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1124 | }; |
1125 | ||
d3e46fea | 1126 | void readahead_tree_block(struct btrfs_root *root, u64 bytenr) |
090d1875 | 1127 | { |
5f39d397 CM |
1128 | struct extent_buffer *buf = NULL; |
1129 | struct inode *btree_inode = root->fs_info->btree_inode; | |
090d1875 | 1130 | |
a83fffb7 | 1131 | buf = btrfs_find_create_tree_block(root, bytenr); |
c871b0f2 | 1132 | if (IS_ERR(buf)) |
6197d86e | 1133 | return; |
d1310b2e | 1134 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
bb82ab88 | 1135 | buf, 0, WAIT_NONE, btree_get_extent, 0); |
5f39d397 | 1136 | free_extent_buffer(buf); |
090d1875 CM |
1137 | } |
1138 | ||
c0dcaa4d | 1139 | int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, |
ab0fff03 AJ |
1140 | int mirror_num, struct extent_buffer **eb) |
1141 | { | |
1142 | struct extent_buffer *buf = NULL; | |
1143 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1144 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; | |
1145 | int ret; | |
1146 | ||
a83fffb7 | 1147 | buf = btrfs_find_create_tree_block(root, bytenr); |
c871b0f2 | 1148 | if (IS_ERR(buf)) |
ab0fff03 AJ |
1149 | return 0; |
1150 | ||
1151 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1152 | ||
1153 | ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, | |
1154 | btree_get_extent, mirror_num); | |
1155 | if (ret) { | |
1156 | free_extent_buffer(buf); | |
1157 | return ret; | |
1158 | } | |
1159 | ||
1160 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1161 | free_extent_buffer(buf); | |
1162 | return -EIO; | |
0b32f4bb | 1163 | } else if (extent_buffer_uptodate(buf)) { |
ab0fff03 AJ |
1164 | *eb = buf; |
1165 | } else { | |
1166 | free_extent_buffer(buf); | |
1167 | } | |
1168 | return 0; | |
1169 | } | |
1170 | ||
01d58472 | 1171 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_fs_info *fs_info, |
0308af44 | 1172 | u64 bytenr) |
0999df54 | 1173 | { |
01d58472 | 1174 | return find_extent_buffer(fs_info, bytenr); |
0999df54 CM |
1175 | } |
1176 | ||
1177 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
a83fffb7 | 1178 | u64 bytenr) |
0999df54 | 1179 | { |
f5ee5c9a | 1180 | if (btrfs_is_testing(root->fs_info)) |
b9ef22de FX |
1181 | return alloc_test_extent_buffer(root->fs_info, bytenr, |
1182 | root->nodesize); | |
ce3e6984 | 1183 | return alloc_extent_buffer(root->fs_info, bytenr); |
0999df54 CM |
1184 | } |
1185 | ||
1186 | ||
e02119d5 CM |
1187 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1188 | { | |
727011e0 | 1189 | return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, |
8aa38c31 | 1190 | buf->start + buf->len - 1); |
e02119d5 CM |
1191 | } |
1192 | ||
1193 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
1194 | { | |
727011e0 | 1195 | return filemap_fdatawait_range(buf->pages[0]->mapping, |
8aa38c31 | 1196 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
1197 | } |
1198 | ||
0999df54 | 1199 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ce86cd59 | 1200 | u64 parent_transid) |
0999df54 CM |
1201 | { |
1202 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1203 | int ret; |
1204 | ||
a83fffb7 | 1205 | buf = btrfs_find_create_tree_block(root, bytenr); |
c871b0f2 LB |
1206 | if (IS_ERR(buf)) |
1207 | return buf; | |
0999df54 | 1208 | |
ca7a79ad | 1209 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
0f0fe8f7 FDBM |
1210 | if (ret) { |
1211 | free_extent_buffer(buf); | |
64c043de | 1212 | return ERR_PTR(ret); |
0f0fe8f7 | 1213 | } |
5f39d397 | 1214 | return buf; |
ce9adaa5 | 1215 | |
eb60ceac CM |
1216 | } |
1217 | ||
01d58472 DD |
1218 | void clean_tree_block(struct btrfs_trans_handle *trans, |
1219 | struct btrfs_fs_info *fs_info, | |
d5c13f92 | 1220 | struct extent_buffer *buf) |
ed2ff2cb | 1221 | { |
55c69072 | 1222 | if (btrfs_header_generation(buf) == |
e2d84521 | 1223 | fs_info->running_transaction->transid) { |
b9447ef8 | 1224 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1225 | |
b9473439 | 1226 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
e2d84521 MX |
1227 | __percpu_counter_add(&fs_info->dirty_metadata_bytes, |
1228 | -buf->len, | |
1229 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1230 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1231 | btrfs_set_lock_blocking(buf); | |
1232 | clear_extent_buffer_dirty(buf); | |
1233 | } | |
925baedd | 1234 | } |
5f39d397 CM |
1235 | } |
1236 | ||
8257b2dc MX |
1237 | static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) |
1238 | { | |
1239 | struct btrfs_subvolume_writers *writers; | |
1240 | int ret; | |
1241 | ||
1242 | writers = kmalloc(sizeof(*writers), GFP_NOFS); | |
1243 | if (!writers) | |
1244 | return ERR_PTR(-ENOMEM); | |
1245 | ||
908c7f19 | 1246 | ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL); |
8257b2dc MX |
1247 | if (ret < 0) { |
1248 | kfree(writers); | |
1249 | return ERR_PTR(ret); | |
1250 | } | |
1251 | ||
1252 | init_waitqueue_head(&writers->wait); | |
1253 | return writers; | |
1254 | } | |
1255 | ||
1256 | static void | |
1257 | btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) | |
1258 | { | |
1259 | percpu_counter_destroy(&writers->counter); | |
1260 | kfree(writers); | |
1261 | } | |
1262 | ||
707e8a07 DS |
1263 | static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize, |
1264 | struct btrfs_root *root, struct btrfs_fs_info *fs_info, | |
143bede5 | 1265 | u64 objectid) |
d97e63b6 | 1266 | { |
7c0260ee | 1267 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
cfaa7295 | 1268 | root->node = NULL; |
a28ec197 | 1269 | root->commit_root = NULL; |
db94535d CM |
1270 | root->sectorsize = sectorsize; |
1271 | root->nodesize = nodesize; | |
87ee04eb | 1272 | root->stripesize = stripesize; |
27cdeb70 | 1273 | root->state = 0; |
d68fc57b | 1274 | root->orphan_cleanup_state = 0; |
0b86a832 | 1275 | |
0f7d52f4 CM |
1276 | root->objectid = objectid; |
1277 | root->last_trans = 0; | |
13a8a7c8 | 1278 | root->highest_objectid = 0; |
eb73c1b7 | 1279 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1280 | root->nr_ordered_extents = 0; |
58176a96 | 1281 | root->name = NULL; |
6bef4d31 | 1282 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1283 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1284 | root->block_rsv = NULL; |
d68fc57b | 1285 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
1286 | |
1287 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1288 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1289 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1290 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1291 | INIT_LIST_HEAD(&root->ordered_extents); |
1292 | INIT_LIST_HEAD(&root->ordered_root); | |
2ab28f32 JB |
1293 | INIT_LIST_HEAD(&root->logged_list[0]); |
1294 | INIT_LIST_HEAD(&root->logged_list[1]); | |
d68fc57b | 1295 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 1296 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1297 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1298 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1299 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1300 | spin_lock_init(&root->log_extents_lock[0]); |
1301 | spin_lock_init(&root->log_extents_lock[1]); | |
a2135011 | 1302 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1303 | mutex_init(&root->log_mutex); |
31f3d255 | 1304 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1305 | mutex_init(&root->delalloc_mutex); |
7237f183 YZ |
1306 | init_waitqueue_head(&root->log_writer_wait); |
1307 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1308 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1309 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1310 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1311 | atomic_set(&root->log_commit[0], 0); |
1312 | atomic_set(&root->log_commit[1], 0); | |
1313 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1314 | atomic_set(&root->log_batch, 0); |
8a35d95f | 1315 | atomic_set(&root->orphan_inodes, 0); |
b0feb9d9 | 1316 | atomic_set(&root->refs, 1); |
8257b2dc | 1317 | atomic_set(&root->will_be_snapshoted, 0); |
55eeaf05 | 1318 | atomic_set(&root->qgroup_meta_rsv, 0); |
7237f183 | 1319 | root->log_transid = 0; |
d1433deb | 1320 | root->log_transid_committed = -1; |
257c62e1 | 1321 | root->last_log_commit = 0; |
7c0260ee | 1322 | if (!dummy) |
06ea65a3 JB |
1323 | extent_io_tree_init(&root->dirty_log_pages, |
1324 | fs_info->btree_inode->i_mapping); | |
017e5369 | 1325 | |
3768f368 CM |
1326 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1327 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1328 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
7c0260ee | 1329 | if (!dummy) |
06ea65a3 JB |
1330 | root->defrag_trans_start = fs_info->generation; |
1331 | else | |
1332 | root->defrag_trans_start = 0; | |
4d775673 | 1333 | root->root_key.objectid = objectid; |
0ee5dc67 | 1334 | root->anon_dev = 0; |
8ea05e3a | 1335 | |
5f3ab90a | 1336 | spin_lock_init(&root->root_item_lock); |
3768f368 CM |
1337 | } |
1338 | ||
74e4d827 DS |
1339 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
1340 | gfp_t flags) | |
6f07e42e | 1341 | { |
74e4d827 | 1342 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e AV |
1343 | if (root) |
1344 | root->fs_info = fs_info; | |
1345 | return root; | |
1346 | } | |
1347 | ||
06ea65a3 JB |
1348 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1349 | /* Should only be used by the testing infrastructure */ | |
7c0260ee JM |
1350 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info, |
1351 | u32 sectorsize, u32 nodesize) | |
06ea65a3 JB |
1352 | { |
1353 | struct btrfs_root *root; | |
1354 | ||
7c0260ee JM |
1355 | if (!fs_info) |
1356 | return ERR_PTR(-EINVAL); | |
1357 | ||
1358 | root = btrfs_alloc_root(fs_info, GFP_KERNEL); | |
06ea65a3 JB |
1359 | if (!root) |
1360 | return ERR_PTR(-ENOMEM); | |
b9ef22de | 1361 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
7c0260ee | 1362 | __setup_root(nodesize, sectorsize, sectorsize, root, fs_info, |
ef9f2db3 | 1363 | BTRFS_ROOT_TREE_OBJECTID); |
faa2dbf0 | 1364 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1365 | |
1366 | return root; | |
1367 | } | |
1368 | #endif | |
1369 | ||
20897f5c AJ |
1370 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
1371 | struct btrfs_fs_info *fs_info, | |
1372 | u64 objectid) | |
1373 | { | |
1374 | struct extent_buffer *leaf; | |
1375 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1376 | struct btrfs_root *root; | |
1377 | struct btrfs_key key; | |
1378 | int ret = 0; | |
6463fe58 | 1379 | uuid_le uuid; |
20897f5c | 1380 | |
74e4d827 | 1381 | root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
20897f5c AJ |
1382 | if (!root) |
1383 | return ERR_PTR(-ENOMEM); | |
1384 | ||
707e8a07 DS |
1385 | __setup_root(tree_root->nodesize, tree_root->sectorsize, |
1386 | tree_root->stripesize, root, fs_info, objectid); | |
20897f5c AJ |
1387 | root->root_key.objectid = objectid; |
1388 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1389 | root->root_key.offset = 0; | |
1390 | ||
4d75f8a9 | 1391 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); |
20897f5c AJ |
1392 | if (IS_ERR(leaf)) { |
1393 | ret = PTR_ERR(leaf); | |
1dd05682 | 1394 | leaf = NULL; |
20897f5c AJ |
1395 | goto fail; |
1396 | } | |
1397 | ||
20897f5c AJ |
1398 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1399 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1400 | btrfs_set_header_generation(leaf, trans->transid); | |
1401 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1402 | btrfs_set_header_owner(leaf, objectid); | |
1403 | root->node = leaf; | |
1404 | ||
0a4e5586 | 1405 | write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(), |
20897f5c AJ |
1406 | BTRFS_FSID_SIZE); |
1407 | write_extent_buffer(leaf, fs_info->chunk_tree_uuid, | |
b308bc2f | 1408 | btrfs_header_chunk_tree_uuid(leaf), |
20897f5c AJ |
1409 | BTRFS_UUID_SIZE); |
1410 | btrfs_mark_buffer_dirty(leaf); | |
1411 | ||
1412 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1413 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c AJ |
1414 | |
1415 | root->root_item.flags = 0; | |
1416 | root->root_item.byte_limit = 0; | |
1417 | btrfs_set_root_bytenr(&root->root_item, leaf->start); | |
1418 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1419 | btrfs_set_root_level(&root->root_item, 0); | |
1420 | btrfs_set_root_refs(&root->root_item, 1); | |
1421 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1422 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1423 | btrfs_set_root_dirid(&root->root_item, 0); | |
6463fe58 SB |
1424 | uuid_le_gen(&uuid); |
1425 | memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); | |
20897f5c AJ |
1426 | root->root_item.drop_level = 0; |
1427 | ||
1428 | key.objectid = objectid; | |
1429 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1430 | key.offset = 0; | |
1431 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1432 | if (ret) | |
1433 | goto fail; | |
1434 | ||
1435 | btrfs_tree_unlock(leaf); | |
1436 | ||
1dd05682 TI |
1437 | return root; |
1438 | ||
20897f5c | 1439 | fail: |
1dd05682 TI |
1440 | if (leaf) { |
1441 | btrfs_tree_unlock(leaf); | |
59885b39 | 1442 | free_extent_buffer(root->commit_root); |
1dd05682 TI |
1443 | free_extent_buffer(leaf); |
1444 | } | |
1445 | kfree(root); | |
20897f5c | 1446 | |
1dd05682 | 1447 | return ERR_PTR(ret); |
20897f5c AJ |
1448 | } |
1449 | ||
7237f183 YZ |
1450 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1451 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1452 | { |
1453 | struct btrfs_root *root; | |
1454 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1455 | struct extent_buffer *leaf; |
e02119d5 | 1456 | |
74e4d827 | 1457 | root = btrfs_alloc_root(fs_info, GFP_NOFS); |
e02119d5 | 1458 | if (!root) |
7237f183 | 1459 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1460 | |
707e8a07 DS |
1461 | __setup_root(tree_root->nodesize, tree_root->sectorsize, |
1462 | tree_root->stripesize, root, fs_info, | |
1463 | BTRFS_TREE_LOG_OBJECTID); | |
e02119d5 CM |
1464 | |
1465 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1466 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1467 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1468 | |
7237f183 | 1469 | /* |
27cdeb70 MX |
1470 | * DON'T set REF_COWS for log trees |
1471 | * | |
7237f183 YZ |
1472 | * log trees do not get reference counted because they go away |
1473 | * before a real commit is actually done. They do store pointers | |
1474 | * to file data extents, and those reference counts still get | |
1475 | * updated (along with back refs to the log tree). | |
1476 | */ | |
e02119d5 | 1477 | |
4d75f8a9 DS |
1478 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
1479 | NULL, 0, 0, 0); | |
7237f183 YZ |
1480 | if (IS_ERR(leaf)) { |
1481 | kfree(root); | |
1482 | return ERR_CAST(leaf); | |
1483 | } | |
e02119d5 | 1484 | |
5d4f98a2 YZ |
1485 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1486 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1487 | btrfs_set_header_generation(leaf, trans->transid); | |
1488 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1489 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1490 | root->node = leaf; |
e02119d5 CM |
1491 | |
1492 | write_extent_buffer(root->node, root->fs_info->fsid, | |
0a4e5586 | 1493 | btrfs_header_fsid(), BTRFS_FSID_SIZE); |
e02119d5 CM |
1494 | btrfs_mark_buffer_dirty(root->node); |
1495 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1496 | return root; |
1497 | } | |
1498 | ||
1499 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1500 | struct btrfs_fs_info *fs_info) | |
1501 | { | |
1502 | struct btrfs_root *log_root; | |
1503 | ||
1504 | log_root = alloc_log_tree(trans, fs_info); | |
1505 | if (IS_ERR(log_root)) | |
1506 | return PTR_ERR(log_root); | |
1507 | WARN_ON(fs_info->log_root_tree); | |
1508 | fs_info->log_root_tree = log_root; | |
1509 | return 0; | |
1510 | } | |
1511 | ||
1512 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1513 | struct btrfs_root *root) | |
1514 | { | |
1515 | struct btrfs_root *log_root; | |
1516 | struct btrfs_inode_item *inode_item; | |
1517 | ||
1518 | log_root = alloc_log_tree(trans, root->fs_info); | |
1519 | if (IS_ERR(log_root)) | |
1520 | return PTR_ERR(log_root); | |
1521 | ||
1522 | log_root->last_trans = trans->transid; | |
1523 | log_root->root_key.offset = root->root_key.objectid; | |
1524 | ||
1525 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1526 | btrfs_set_stack_inode_generation(inode_item, 1); |
1527 | btrfs_set_stack_inode_size(inode_item, 3); | |
1528 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
707e8a07 | 1529 | btrfs_set_stack_inode_nbytes(inode_item, root->nodesize); |
3cae210f | 1530 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1531 | |
5d4f98a2 | 1532 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1533 | |
1534 | WARN_ON(root->log_root); | |
1535 | root->log_root = log_root; | |
1536 | root->log_transid = 0; | |
d1433deb | 1537 | root->log_transid_committed = -1; |
257c62e1 | 1538 | root->last_log_commit = 0; |
e02119d5 CM |
1539 | return 0; |
1540 | } | |
1541 | ||
35a3621b SB |
1542 | static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, |
1543 | struct btrfs_key *key) | |
e02119d5 CM |
1544 | { |
1545 | struct btrfs_root *root; | |
1546 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1547 | struct btrfs_path *path; |
84234f3a | 1548 | u64 generation; |
cb517eab | 1549 | int ret; |
0f7d52f4 | 1550 | |
cb517eab MX |
1551 | path = btrfs_alloc_path(); |
1552 | if (!path) | |
0f7d52f4 | 1553 | return ERR_PTR(-ENOMEM); |
cb517eab | 1554 | |
74e4d827 | 1555 | root = btrfs_alloc_root(fs_info, GFP_NOFS); |
cb517eab MX |
1556 | if (!root) { |
1557 | ret = -ENOMEM; | |
1558 | goto alloc_fail; | |
0f7d52f4 CM |
1559 | } |
1560 | ||
707e8a07 DS |
1561 | __setup_root(tree_root->nodesize, tree_root->sectorsize, |
1562 | tree_root->stripesize, root, fs_info, key->objectid); | |
0f7d52f4 | 1563 | |
cb517eab MX |
1564 | ret = btrfs_find_root(tree_root, key, path, |
1565 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1566 | if (ret) { |
13a8a7c8 YZ |
1567 | if (ret > 0) |
1568 | ret = -ENOENT; | |
cb517eab | 1569 | goto find_fail; |
0f7d52f4 | 1570 | } |
13a8a7c8 | 1571 | |
84234f3a | 1572 | generation = btrfs_root_generation(&root->root_item); |
db94535d | 1573 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), |
ce86cd59 | 1574 | generation); |
64c043de LB |
1575 | if (IS_ERR(root->node)) { |
1576 | ret = PTR_ERR(root->node); | |
cb517eab MX |
1577 | goto find_fail; |
1578 | } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
1579 | ret = -EIO; | |
64c043de LB |
1580 | free_extent_buffer(root->node); |
1581 | goto find_fail; | |
416bc658 | 1582 | } |
5d4f98a2 | 1583 | root->commit_root = btrfs_root_node(root); |
13a8a7c8 | 1584 | out: |
cb517eab MX |
1585 | btrfs_free_path(path); |
1586 | return root; | |
1587 | ||
cb517eab MX |
1588 | find_fail: |
1589 | kfree(root); | |
1590 | alloc_fail: | |
1591 | root = ERR_PTR(ret); | |
1592 | goto out; | |
1593 | } | |
1594 | ||
1595 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, | |
1596 | struct btrfs_key *location) | |
1597 | { | |
1598 | struct btrfs_root *root; | |
1599 | ||
1600 | root = btrfs_read_tree_root(tree_root, location); | |
1601 | if (IS_ERR(root)) | |
1602 | return root; | |
1603 | ||
1604 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { | |
27cdeb70 | 1605 | set_bit(BTRFS_ROOT_REF_COWS, &root->state); |
08fe4db1 LZ |
1606 | btrfs_check_and_init_root_item(&root->root_item); |
1607 | } | |
13a8a7c8 | 1608 | |
5eda7b5e CM |
1609 | return root; |
1610 | } | |
1611 | ||
cb517eab MX |
1612 | int btrfs_init_fs_root(struct btrfs_root *root) |
1613 | { | |
1614 | int ret; | |
8257b2dc | 1615 | struct btrfs_subvolume_writers *writers; |
cb517eab MX |
1616 | |
1617 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); | |
1618 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), | |
1619 | GFP_NOFS); | |
1620 | if (!root->free_ino_pinned || !root->free_ino_ctl) { | |
1621 | ret = -ENOMEM; | |
1622 | goto fail; | |
1623 | } | |
1624 | ||
8257b2dc MX |
1625 | writers = btrfs_alloc_subvolume_writers(); |
1626 | if (IS_ERR(writers)) { | |
1627 | ret = PTR_ERR(writers); | |
1628 | goto fail; | |
1629 | } | |
1630 | root->subv_writers = writers; | |
1631 | ||
cb517eab | 1632 | btrfs_init_free_ino_ctl(root); |
57cdc8db DS |
1633 | spin_lock_init(&root->ino_cache_lock); |
1634 | init_waitqueue_head(&root->ino_cache_wait); | |
cb517eab MX |
1635 | |
1636 | ret = get_anon_bdev(&root->anon_dev); | |
1637 | if (ret) | |
876d2cf1 | 1638 | goto fail; |
f32e48e9 CR |
1639 | |
1640 | mutex_lock(&root->objectid_mutex); | |
1641 | ret = btrfs_find_highest_objectid(root, | |
1642 | &root->highest_objectid); | |
1643 | if (ret) { | |
1644 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1645 | goto fail; |
f32e48e9 CR |
1646 | } |
1647 | ||
1648 | ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
1649 | ||
1650 | mutex_unlock(&root->objectid_mutex); | |
1651 | ||
cb517eab MX |
1652 | return 0; |
1653 | fail: | |
876d2cf1 | 1654 | /* the caller is responsible to call free_fs_root */ |
cb517eab MX |
1655 | return ret; |
1656 | } | |
1657 | ||
35bbb97f JM |
1658 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1659 | u64 root_id) | |
cb517eab MX |
1660 | { |
1661 | struct btrfs_root *root; | |
1662 | ||
1663 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1664 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1665 | (unsigned long)root_id); | |
1666 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1667 | return root; | |
1668 | } | |
1669 | ||
1670 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, | |
1671 | struct btrfs_root *root) | |
1672 | { | |
1673 | int ret; | |
1674 | ||
e1860a77 | 1675 | ret = radix_tree_preload(GFP_NOFS); |
cb517eab MX |
1676 | if (ret) |
1677 | return ret; | |
1678 | ||
1679 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1680 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1681 | (unsigned long)root->root_key.objectid, | |
1682 | root); | |
1683 | if (ret == 0) | |
27cdeb70 | 1684 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
cb517eab MX |
1685 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1686 | radix_tree_preload_end(); | |
1687 | ||
1688 | return ret; | |
1689 | } | |
1690 | ||
c00869f1 MX |
1691 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, |
1692 | struct btrfs_key *location, | |
1693 | bool check_ref) | |
5eda7b5e CM |
1694 | { |
1695 | struct btrfs_root *root; | |
381cf658 | 1696 | struct btrfs_path *path; |
1d4c08e0 | 1697 | struct btrfs_key key; |
5eda7b5e CM |
1698 | int ret; |
1699 | ||
edbd8d4e CM |
1700 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1701 | return fs_info->tree_root; | |
1702 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1703 | return fs_info->extent_root; | |
8f18cf13 CM |
1704 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1705 | return fs_info->chunk_root; | |
1706 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1707 | return fs_info->dev_root; | |
0403e47e YZ |
1708 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1709 | return fs_info->csum_root; | |
bcef60f2 AJ |
1710 | if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1711 | return fs_info->quota_root ? fs_info->quota_root : | |
1712 | ERR_PTR(-ENOENT); | |
f7a81ea4 SB |
1713 | if (location->objectid == BTRFS_UUID_TREE_OBJECTID) |
1714 | return fs_info->uuid_root ? fs_info->uuid_root : | |
1715 | ERR_PTR(-ENOENT); | |
70f6d82e OS |
1716 | if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) |
1717 | return fs_info->free_space_root ? fs_info->free_space_root : | |
1718 | ERR_PTR(-ENOENT); | |
4df27c4d | 1719 | again: |
cb517eab | 1720 | root = btrfs_lookup_fs_root(fs_info, location->objectid); |
48475471 | 1721 | if (root) { |
c00869f1 | 1722 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) |
48475471 | 1723 | return ERR_PTR(-ENOENT); |
5eda7b5e | 1724 | return root; |
48475471 | 1725 | } |
5eda7b5e | 1726 | |
cb517eab | 1727 | root = btrfs_read_fs_root(fs_info->tree_root, location); |
5eda7b5e CM |
1728 | if (IS_ERR(root)) |
1729 | return root; | |
3394e160 | 1730 | |
c00869f1 | 1731 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1732 | ret = -ENOENT; |
581bb050 | 1733 | goto fail; |
35a30d7c | 1734 | } |
581bb050 | 1735 | |
cb517eab | 1736 | ret = btrfs_init_fs_root(root); |
ac08aedf CM |
1737 | if (ret) |
1738 | goto fail; | |
3394e160 | 1739 | |
381cf658 DS |
1740 | path = btrfs_alloc_path(); |
1741 | if (!path) { | |
1742 | ret = -ENOMEM; | |
1743 | goto fail; | |
1744 | } | |
1d4c08e0 DS |
1745 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1746 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
1747 | key.offset = location->objectid; | |
1748 | ||
1749 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1750 | btrfs_free_path(path); |
d68fc57b YZ |
1751 | if (ret < 0) |
1752 | goto fail; | |
1753 | if (ret == 0) | |
27cdeb70 | 1754 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1755 | |
cb517eab | 1756 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1757 | if (ret) { |
4df27c4d YZ |
1758 | if (ret == -EEXIST) { |
1759 | free_fs_root(root); | |
1760 | goto again; | |
1761 | } | |
1762 | goto fail; | |
0f7d52f4 | 1763 | } |
edbd8d4e | 1764 | return root; |
4df27c4d YZ |
1765 | fail: |
1766 | free_fs_root(root); | |
1767 | return ERR_PTR(ret); | |
edbd8d4e CM |
1768 | } |
1769 | ||
04160088 CM |
1770 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1771 | { | |
1772 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1773 | int ret = 0; | |
04160088 CM |
1774 | struct btrfs_device *device; |
1775 | struct backing_dev_info *bdi; | |
b7967db7 | 1776 | |
1f78160c XG |
1777 | rcu_read_lock(); |
1778 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1779 | if (!device->bdev) |
1780 | continue; | |
04160088 | 1781 | bdi = blk_get_backing_dev_info(device->bdev); |
ff9ea323 | 1782 | if (bdi_congested(bdi, bdi_bits)) { |
04160088 CM |
1783 | ret = 1; |
1784 | break; | |
1785 | } | |
1786 | } | |
1f78160c | 1787 | rcu_read_unlock(); |
04160088 CM |
1788 | return ret; |
1789 | } | |
1790 | ||
04160088 CM |
1791 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1792 | { | |
ad081f14 JA |
1793 | int err; |
1794 | ||
b4caecd4 | 1795 | err = bdi_setup_and_register(bdi, "btrfs"); |
ad081f14 JA |
1796 | if (err) |
1797 | return err; | |
1798 | ||
09cbfeaf | 1799 | bdi->ra_pages = VM_MAX_READAHEAD * 1024 / PAGE_SIZE; |
04160088 CM |
1800 | bdi->congested_fn = btrfs_congested_fn; |
1801 | bdi->congested_data = info; | |
da2f0f74 | 1802 | bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK; |
04160088 CM |
1803 | return 0; |
1804 | } | |
1805 | ||
8b712842 CM |
1806 | /* |
1807 | * called by the kthread helper functions to finally call the bio end_io | |
1808 | * functions. This is where read checksum verification actually happens | |
1809 | */ | |
1810 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1811 | { |
ce9adaa5 | 1812 | struct bio *bio; |
97eb6b69 | 1813 | struct btrfs_end_io_wq *end_io_wq; |
ce9adaa5 | 1814 | |
97eb6b69 | 1815 | end_io_wq = container_of(work, struct btrfs_end_io_wq, work); |
8b712842 | 1816 | bio = end_io_wq->bio; |
ce9adaa5 | 1817 | |
4246a0b6 | 1818 | bio->bi_error = end_io_wq->error; |
8b712842 CM |
1819 | bio->bi_private = end_io_wq->private; |
1820 | bio->bi_end_io = end_io_wq->end_io; | |
97eb6b69 | 1821 | kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); |
4246a0b6 | 1822 | bio_endio(bio); |
44b8bd7e CM |
1823 | } |
1824 | ||
a74a4b97 CM |
1825 | static int cleaner_kthread(void *arg) |
1826 | { | |
1827 | struct btrfs_root *root = arg; | |
d0278245 | 1828 | int again; |
da288d28 | 1829 | struct btrfs_trans_handle *trans; |
a74a4b97 CM |
1830 | |
1831 | do { | |
d0278245 | 1832 | again = 0; |
a74a4b97 | 1833 | |
d0278245 | 1834 | /* Make the cleaner go to sleep early. */ |
babbf170 | 1835 | if (btrfs_need_cleaner_sleep(root)) |
d0278245 MX |
1836 | goto sleep; |
1837 | ||
90c711ab ZB |
1838 | /* |
1839 | * Do not do anything if we might cause open_ctree() to block | |
1840 | * before we have finished mounting the filesystem. | |
1841 | */ | |
1842 | if (!root->fs_info->open) | |
1843 | goto sleep; | |
1844 | ||
d0278245 MX |
1845 | if (!mutex_trylock(&root->fs_info->cleaner_mutex)) |
1846 | goto sleep; | |
1847 | ||
dc7f370c MX |
1848 | /* |
1849 | * Avoid the problem that we change the status of the fs | |
1850 | * during the above check and trylock. | |
1851 | */ | |
babbf170 | 1852 | if (btrfs_need_cleaner_sleep(root)) { |
dc7f370c MX |
1853 | mutex_unlock(&root->fs_info->cleaner_mutex); |
1854 | goto sleep; | |
76dda93c | 1855 | } |
a74a4b97 | 1856 | |
c2d6cb16 | 1857 | mutex_lock(&root->fs_info->cleaner_delayed_iput_mutex); |
d0278245 | 1858 | btrfs_run_delayed_iputs(root); |
c2d6cb16 FM |
1859 | mutex_unlock(&root->fs_info->cleaner_delayed_iput_mutex); |
1860 | ||
d0278245 MX |
1861 | again = btrfs_clean_one_deleted_snapshot(root); |
1862 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
1863 | ||
1864 | /* | |
05323cd1 MX |
1865 | * The defragger has dealt with the R/O remount and umount, |
1866 | * needn't do anything special here. | |
d0278245 MX |
1867 | */ |
1868 | btrfs_run_defrag_inodes(root->fs_info); | |
67c5e7d4 FM |
1869 | |
1870 | /* | |
1871 | * Acquires fs_info->delete_unused_bgs_mutex to avoid racing | |
1872 | * with relocation (btrfs_relocate_chunk) and relocation | |
1873 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
1874 | * after acquiring fs_info->delete_unused_bgs_mutex. So we | |
1875 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting | |
1876 | * unused block groups. | |
1877 | */ | |
1878 | btrfs_delete_unused_bgs(root->fs_info); | |
d0278245 | 1879 | sleep: |
838fe188 | 1880 | if (!again) { |
a74a4b97 | 1881 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1882 | if (!kthread_should_stop()) |
1883 | schedule(); | |
a74a4b97 CM |
1884 | __set_current_state(TASK_RUNNING); |
1885 | } | |
1886 | } while (!kthread_should_stop()); | |
da288d28 FM |
1887 | |
1888 | /* | |
1889 | * Transaction kthread is stopped before us and wakes us up. | |
1890 | * However we might have started a new transaction and COWed some | |
1891 | * tree blocks when deleting unused block groups for example. So | |
1892 | * make sure we commit the transaction we started to have a clean | |
1893 | * shutdown when evicting the btree inode - if it has dirty pages | |
1894 | * when we do the final iput() on it, eviction will trigger a | |
1895 | * writeback for it which will fail with null pointer dereferences | |
1896 | * since work queues and other resources were already released and | |
1897 | * destroyed by the time the iput/eviction/writeback is made. | |
1898 | */ | |
1899 | trans = btrfs_attach_transaction(root); | |
1900 | if (IS_ERR(trans)) { | |
1901 | if (PTR_ERR(trans) != -ENOENT) | |
1902 | btrfs_err(root->fs_info, | |
1903 | "cleaner transaction attach returned %ld", | |
1904 | PTR_ERR(trans)); | |
1905 | } else { | |
1906 | int ret; | |
1907 | ||
1908 | ret = btrfs_commit_transaction(trans, root); | |
1909 | if (ret) | |
1910 | btrfs_err(root->fs_info, | |
1911 | "cleaner open transaction commit returned %d", | |
1912 | ret); | |
1913 | } | |
1914 | ||
a74a4b97 CM |
1915 | return 0; |
1916 | } | |
1917 | ||
1918 | static int transaction_kthread(void *arg) | |
1919 | { | |
1920 | struct btrfs_root *root = arg; | |
1921 | struct btrfs_trans_handle *trans; | |
1922 | struct btrfs_transaction *cur; | |
8929ecfa | 1923 | u64 transid; |
a74a4b97 CM |
1924 | unsigned long now; |
1925 | unsigned long delay; | |
914b2007 | 1926 | bool cannot_commit; |
a74a4b97 CM |
1927 | |
1928 | do { | |
914b2007 | 1929 | cannot_commit = false; |
8b87dc17 | 1930 | delay = HZ * root->fs_info->commit_interval; |
a74a4b97 CM |
1931 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
1932 | ||
a4abeea4 | 1933 | spin_lock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1934 | cur = root->fs_info->running_transaction; |
1935 | if (!cur) { | |
a4abeea4 | 1936 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1937 | goto sleep; |
1938 | } | |
31153d81 | 1939 | |
a74a4b97 | 1940 | now = get_seconds(); |
4a9d8bde | 1941 | if (cur->state < TRANS_STATE_BLOCKED && |
8b87dc17 DS |
1942 | (now < cur->start_time || |
1943 | now - cur->start_time < root->fs_info->commit_interval)) { | |
a4abeea4 | 1944 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1945 | delay = HZ * 5; |
1946 | goto sleep; | |
1947 | } | |
8929ecfa | 1948 | transid = cur->transid; |
a4abeea4 | 1949 | spin_unlock(&root->fs_info->trans_lock); |
56bec294 | 1950 | |
79787eaa | 1951 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1952 | trans = btrfs_attach_transaction(root); |
914b2007 | 1953 | if (IS_ERR(trans)) { |
354aa0fb MX |
1954 | if (PTR_ERR(trans) != -ENOENT) |
1955 | cannot_commit = true; | |
79787eaa | 1956 | goto sleep; |
914b2007 | 1957 | } |
8929ecfa | 1958 | if (transid == trans->transid) { |
79787eaa | 1959 | btrfs_commit_transaction(trans, root); |
8929ecfa YZ |
1960 | } else { |
1961 | btrfs_end_transaction(trans, root); | |
1962 | } | |
a74a4b97 CM |
1963 | sleep: |
1964 | wake_up_process(root->fs_info->cleaner_kthread); | |
1965 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1966 | ||
4e121c06 JB |
1967 | if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, |
1968 | &root->fs_info->fs_state))) | |
1969 | btrfs_cleanup_transaction(root); | |
ce63f891 JK |
1970 | set_current_state(TASK_INTERRUPTIBLE); |
1971 | if (!kthread_should_stop() && | |
1972 | (!btrfs_transaction_blocked(root->fs_info) || | |
1973 | cannot_commit)) | |
1974 | schedule_timeout(delay); | |
1975 | __set_current_state(TASK_RUNNING); | |
a74a4b97 CM |
1976 | } while (!kthread_should_stop()); |
1977 | return 0; | |
1978 | } | |
1979 | ||
af31f5e5 CM |
1980 | /* |
1981 | * this will find the highest generation in the array of | |
1982 | * root backups. The index of the highest array is returned, | |
1983 | * or -1 if we can't find anything. | |
1984 | * | |
1985 | * We check to make sure the array is valid by comparing the | |
1986 | * generation of the latest root in the array with the generation | |
1987 | * in the super block. If they don't match we pitch it. | |
1988 | */ | |
1989 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1990 | { | |
1991 | u64 cur; | |
1992 | int newest_index = -1; | |
1993 | struct btrfs_root_backup *root_backup; | |
1994 | int i; | |
1995 | ||
1996 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1997 | root_backup = info->super_copy->super_roots + i; | |
1998 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1999 | if (cur == newest_gen) | |
2000 | newest_index = i; | |
2001 | } | |
2002 | ||
2003 | /* check to see if we actually wrapped around */ | |
2004 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
2005 | root_backup = info->super_copy->super_roots; | |
2006 | cur = btrfs_backup_tree_root_gen(root_backup); | |
2007 | if (cur == newest_gen) | |
2008 | newest_index = 0; | |
2009 | } | |
2010 | return newest_index; | |
2011 | } | |
2012 | ||
2013 | ||
2014 | /* | |
2015 | * find the oldest backup so we know where to store new entries | |
2016 | * in the backup array. This will set the backup_root_index | |
2017 | * field in the fs_info struct | |
2018 | */ | |
2019 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
2020 | u64 newest_gen) | |
2021 | { | |
2022 | int newest_index = -1; | |
2023 | ||
2024 | newest_index = find_newest_super_backup(info, newest_gen); | |
2025 | /* if there was garbage in there, just move along */ | |
2026 | if (newest_index == -1) { | |
2027 | info->backup_root_index = 0; | |
2028 | } else { | |
2029 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
2030 | } | |
2031 | } | |
2032 | ||
2033 | /* | |
2034 | * copy all the root pointers into the super backup array. | |
2035 | * this will bump the backup pointer by one when it is | |
2036 | * done | |
2037 | */ | |
2038 | static void backup_super_roots(struct btrfs_fs_info *info) | |
2039 | { | |
2040 | int next_backup; | |
2041 | struct btrfs_root_backup *root_backup; | |
2042 | int last_backup; | |
2043 | ||
2044 | next_backup = info->backup_root_index; | |
2045 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
2046 | BTRFS_NUM_BACKUP_ROOTS; | |
2047 | ||
2048 | /* | |
2049 | * just overwrite the last backup if we're at the same generation | |
2050 | * this happens only at umount | |
2051 | */ | |
2052 | root_backup = info->super_for_commit->super_roots + last_backup; | |
2053 | if (btrfs_backup_tree_root_gen(root_backup) == | |
2054 | btrfs_header_generation(info->tree_root->node)) | |
2055 | next_backup = last_backup; | |
2056 | ||
2057 | root_backup = info->super_for_commit->super_roots + next_backup; | |
2058 | ||
2059 | /* | |
2060 | * make sure all of our padding and empty slots get zero filled | |
2061 | * regardless of which ones we use today | |
2062 | */ | |
2063 | memset(root_backup, 0, sizeof(*root_backup)); | |
2064 | ||
2065 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
2066 | ||
2067 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
2068 | btrfs_set_backup_tree_root_gen(root_backup, | |
2069 | btrfs_header_generation(info->tree_root->node)); | |
2070 | ||
2071 | btrfs_set_backup_tree_root_level(root_backup, | |
2072 | btrfs_header_level(info->tree_root->node)); | |
2073 | ||
2074 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
2075 | btrfs_set_backup_chunk_root_gen(root_backup, | |
2076 | btrfs_header_generation(info->chunk_root->node)); | |
2077 | btrfs_set_backup_chunk_root_level(root_backup, | |
2078 | btrfs_header_level(info->chunk_root->node)); | |
2079 | ||
2080 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
2081 | btrfs_set_backup_extent_root_gen(root_backup, | |
2082 | btrfs_header_generation(info->extent_root->node)); | |
2083 | btrfs_set_backup_extent_root_level(root_backup, | |
2084 | btrfs_header_level(info->extent_root->node)); | |
2085 | ||
7c7e82a7 CM |
2086 | /* |
2087 | * we might commit during log recovery, which happens before we set | |
2088 | * the fs_root. Make sure it is valid before we fill it in. | |
2089 | */ | |
2090 | if (info->fs_root && info->fs_root->node) { | |
2091 | btrfs_set_backup_fs_root(root_backup, | |
2092 | info->fs_root->node->start); | |
2093 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 2094 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 2095 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 2096 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 2097 | } |
af31f5e5 CM |
2098 | |
2099 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
2100 | btrfs_set_backup_dev_root_gen(root_backup, | |
2101 | btrfs_header_generation(info->dev_root->node)); | |
2102 | btrfs_set_backup_dev_root_level(root_backup, | |
2103 | btrfs_header_level(info->dev_root->node)); | |
2104 | ||
2105 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
2106 | btrfs_set_backup_csum_root_gen(root_backup, | |
2107 | btrfs_header_generation(info->csum_root->node)); | |
2108 | btrfs_set_backup_csum_root_level(root_backup, | |
2109 | btrfs_header_level(info->csum_root->node)); | |
2110 | ||
2111 | btrfs_set_backup_total_bytes(root_backup, | |
2112 | btrfs_super_total_bytes(info->super_copy)); | |
2113 | btrfs_set_backup_bytes_used(root_backup, | |
2114 | btrfs_super_bytes_used(info->super_copy)); | |
2115 | btrfs_set_backup_num_devices(root_backup, | |
2116 | btrfs_super_num_devices(info->super_copy)); | |
2117 | ||
2118 | /* | |
2119 | * if we don't copy this out to the super_copy, it won't get remembered | |
2120 | * for the next commit | |
2121 | */ | |
2122 | memcpy(&info->super_copy->super_roots, | |
2123 | &info->super_for_commit->super_roots, | |
2124 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
2125 | } | |
2126 | ||
2127 | /* | |
2128 | * this copies info out of the root backup array and back into | |
2129 | * the in-memory super block. It is meant to help iterate through | |
2130 | * the array, so you send it the number of backups you've already | |
2131 | * tried and the last backup index you used. | |
2132 | * | |
2133 | * this returns -1 when it has tried all the backups | |
2134 | */ | |
2135 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
2136 | struct btrfs_super_block *super, | |
2137 | int *num_backups_tried, int *backup_index) | |
2138 | { | |
2139 | struct btrfs_root_backup *root_backup; | |
2140 | int newest = *backup_index; | |
2141 | ||
2142 | if (*num_backups_tried == 0) { | |
2143 | u64 gen = btrfs_super_generation(super); | |
2144 | ||
2145 | newest = find_newest_super_backup(info, gen); | |
2146 | if (newest == -1) | |
2147 | return -1; | |
2148 | ||
2149 | *backup_index = newest; | |
2150 | *num_backups_tried = 1; | |
2151 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
2152 | /* we've tried all the backups, all done */ | |
2153 | return -1; | |
2154 | } else { | |
2155 | /* jump to the next oldest backup */ | |
2156 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
2157 | BTRFS_NUM_BACKUP_ROOTS; | |
2158 | *backup_index = newest; | |
2159 | *num_backups_tried += 1; | |
2160 | } | |
2161 | root_backup = super->super_roots + newest; | |
2162 | ||
2163 | btrfs_set_super_generation(super, | |
2164 | btrfs_backup_tree_root_gen(root_backup)); | |
2165 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
2166 | btrfs_set_super_root_level(super, | |
2167 | btrfs_backup_tree_root_level(root_backup)); | |
2168 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
2169 | ||
2170 | /* | |
2171 | * fixme: the total bytes and num_devices need to match or we should | |
2172 | * need a fsck | |
2173 | */ | |
2174 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
2175 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
2176 | return 0; | |
2177 | } | |
2178 | ||
7abadb64 LB |
2179 | /* helper to cleanup workers */ |
2180 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2181 | { | |
dc6e3209 | 2182 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2183 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 2184 | btrfs_destroy_workqueue(fs_info->workers); |
fccb5d86 QW |
2185 | btrfs_destroy_workqueue(fs_info->endio_workers); |
2186 | btrfs_destroy_workqueue(fs_info->endio_meta_workers); | |
2187 | btrfs_destroy_workqueue(fs_info->endio_raid56_workers); | |
8b110e39 | 2188 | btrfs_destroy_workqueue(fs_info->endio_repair_workers); |
d05a33ac | 2189 | btrfs_destroy_workqueue(fs_info->rmw_workers); |
fccb5d86 QW |
2190 | btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); |
2191 | btrfs_destroy_workqueue(fs_info->endio_write_workers); | |
2192 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
a8c93d4e | 2193 | btrfs_destroy_workqueue(fs_info->submit_workers); |
5b3bc44e | 2194 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2195 | btrfs_destroy_workqueue(fs_info->caching_workers); |
736cfa15 | 2196 | btrfs_destroy_workqueue(fs_info->readahead_workers); |
a44903ab | 2197 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2198 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
a79b7d4b | 2199 | btrfs_destroy_workqueue(fs_info->extent_workers); |
7abadb64 LB |
2200 | } |
2201 | ||
2e9f5954 R |
2202 | static void free_root_extent_buffers(struct btrfs_root *root) |
2203 | { | |
2204 | if (root) { | |
2205 | free_extent_buffer(root->node); | |
2206 | free_extent_buffer(root->commit_root); | |
2207 | root->node = NULL; | |
2208 | root->commit_root = NULL; | |
2209 | } | |
2210 | } | |
2211 | ||
af31f5e5 CM |
2212 | /* helper to cleanup tree roots */ |
2213 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
2214 | { | |
2e9f5954 | 2215 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2216 | |
2e9f5954 R |
2217 | free_root_extent_buffers(info->dev_root); |
2218 | free_root_extent_buffers(info->extent_root); | |
2219 | free_root_extent_buffers(info->csum_root); | |
2220 | free_root_extent_buffers(info->quota_root); | |
2221 | free_root_extent_buffers(info->uuid_root); | |
2222 | if (chunk_root) | |
2223 | free_root_extent_buffers(info->chunk_root); | |
70f6d82e | 2224 | free_root_extent_buffers(info->free_space_root); |
af31f5e5 CM |
2225 | } |
2226 | ||
faa2dbf0 | 2227 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 JB |
2228 | { |
2229 | int ret; | |
2230 | struct btrfs_root *gang[8]; | |
2231 | int i; | |
2232 | ||
2233 | while (!list_empty(&fs_info->dead_roots)) { | |
2234 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2235 | struct btrfs_root, root_list); | |
2236 | list_del(&gang[0]->root_list); | |
2237 | ||
27cdeb70 | 2238 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) { |
cb517eab | 2239 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); |
171f6537 JB |
2240 | } else { |
2241 | free_extent_buffer(gang[0]->node); | |
2242 | free_extent_buffer(gang[0]->commit_root); | |
b0feb9d9 | 2243 | btrfs_put_fs_root(gang[0]); |
171f6537 JB |
2244 | } |
2245 | } | |
2246 | ||
2247 | while (1) { | |
2248 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2249 | (void **)gang, 0, | |
2250 | ARRAY_SIZE(gang)); | |
2251 | if (!ret) | |
2252 | break; | |
2253 | for (i = 0; i < ret; i++) | |
cb517eab | 2254 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); |
171f6537 | 2255 | } |
1a4319cc LB |
2256 | |
2257 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { | |
2258 | btrfs_free_log_root_tree(NULL, fs_info); | |
2259 | btrfs_destroy_pinned_extent(fs_info->tree_root, | |
2260 | fs_info->pinned_extents); | |
2261 | } | |
171f6537 | 2262 | } |
af31f5e5 | 2263 | |
638aa7ed ES |
2264 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2265 | { | |
2266 | mutex_init(&fs_info->scrub_lock); | |
2267 | atomic_set(&fs_info->scrubs_running, 0); | |
2268 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2269 | atomic_set(&fs_info->scrubs_paused, 0); | |
2270 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2271 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
2272 | fs_info->scrub_workers_refcnt = 0; | |
2273 | } | |
2274 | ||
779a65a4 ES |
2275 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2276 | { | |
2277 | spin_lock_init(&fs_info->balance_lock); | |
2278 | mutex_init(&fs_info->balance_mutex); | |
2279 | atomic_set(&fs_info->balance_running, 0); | |
2280 | atomic_set(&fs_info->balance_pause_req, 0); | |
2281 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2282 | fs_info->balance_ctl = NULL; | |
2283 | init_waitqueue_head(&fs_info->balance_wait_q); | |
2284 | } | |
2285 | ||
f37938e0 ES |
2286 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info, |
2287 | struct btrfs_root *tree_root) | |
2288 | { | |
2289 | fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2290 | set_nlink(fs_info->btree_inode, 1); | |
2291 | /* | |
2292 | * we set the i_size on the btree inode to the max possible int. | |
2293 | * the real end of the address space is determined by all of | |
2294 | * the devices in the system | |
2295 | */ | |
2296 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
2297 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 ES |
2298 | |
2299 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); | |
2300 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, | |
2301 | fs_info->btree_inode->i_mapping); | |
2302 | BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; | |
2303 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); | |
2304 | ||
2305 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
2306 | ||
2307 | BTRFS_I(fs_info->btree_inode)->root = tree_root; | |
2308 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
2309 | sizeof(struct btrfs_key)); | |
2310 | set_bit(BTRFS_INODE_DUMMY, | |
2311 | &BTRFS_I(fs_info->btree_inode)->runtime_flags); | |
2312 | btrfs_insert_inode_hash(fs_info->btree_inode); | |
2313 | } | |
2314 | ||
ad618368 ES |
2315 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2316 | { | |
2317 | fs_info->dev_replace.lock_owner = 0; | |
2318 | atomic_set(&fs_info->dev_replace.nesting_level, 0); | |
2319 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); | |
73beece9 LB |
2320 | rwlock_init(&fs_info->dev_replace.lock); |
2321 | atomic_set(&fs_info->dev_replace.read_locks, 0); | |
2322 | atomic_set(&fs_info->dev_replace.blocking_readers, 0); | |
ad618368 | 2323 | init_waitqueue_head(&fs_info->replace_wait); |
73beece9 | 2324 | init_waitqueue_head(&fs_info->dev_replace.read_lock_wq); |
ad618368 ES |
2325 | } |
2326 | ||
f9e92e40 ES |
2327 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2328 | { | |
2329 | spin_lock_init(&fs_info->qgroup_lock); | |
2330 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2331 | fs_info->qgroup_tree = RB_ROOT; | |
2332 | fs_info->qgroup_op_tree = RB_ROOT; | |
2333 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); | |
2334 | fs_info->qgroup_seq = 1; | |
2335 | fs_info->quota_enabled = 0; | |
2336 | fs_info->pending_quota_state = 0; | |
2337 | fs_info->qgroup_ulist = NULL; | |
d2c609b8 | 2338 | fs_info->qgroup_rescan_running = false; |
f9e92e40 ES |
2339 | mutex_init(&fs_info->qgroup_rescan_lock); |
2340 | } | |
2341 | ||
2a458198 ES |
2342 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, |
2343 | struct btrfs_fs_devices *fs_devices) | |
2344 | { | |
2345 | int max_active = fs_info->thread_pool_size; | |
6f011058 | 2346 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2347 | |
2348 | fs_info->workers = | |
cb001095 JM |
2349 | btrfs_alloc_workqueue(fs_info, "worker", |
2350 | flags | WQ_HIGHPRI, max_active, 16); | |
2a458198 ES |
2351 | |
2352 | fs_info->delalloc_workers = | |
cb001095 JM |
2353 | btrfs_alloc_workqueue(fs_info, "delalloc", |
2354 | flags, max_active, 2); | |
2a458198 ES |
2355 | |
2356 | fs_info->flush_workers = | |
cb001095 JM |
2357 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
2358 | flags, max_active, 0); | |
2a458198 ES |
2359 | |
2360 | fs_info->caching_workers = | |
cb001095 | 2361 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 ES |
2362 | |
2363 | /* | |
2364 | * a higher idle thresh on the submit workers makes it much more | |
2365 | * likely that bios will be send down in a sane order to the | |
2366 | * devices | |
2367 | */ | |
2368 | fs_info->submit_workers = | |
cb001095 | 2369 | btrfs_alloc_workqueue(fs_info, "submit", flags, |
2a458198 ES |
2370 | min_t(u64, fs_devices->num_devices, |
2371 | max_active), 64); | |
2372 | ||
2373 | fs_info->fixup_workers = | |
cb001095 | 2374 | btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); |
2a458198 ES |
2375 | |
2376 | /* | |
2377 | * endios are largely parallel and should have a very | |
2378 | * low idle thresh | |
2379 | */ | |
2380 | fs_info->endio_workers = | |
cb001095 | 2381 | btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); |
2a458198 | 2382 | fs_info->endio_meta_workers = |
cb001095 JM |
2383 | btrfs_alloc_workqueue(fs_info, "endio-meta", flags, |
2384 | max_active, 4); | |
2a458198 | 2385 | fs_info->endio_meta_write_workers = |
cb001095 JM |
2386 | btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, |
2387 | max_active, 2); | |
2a458198 | 2388 | fs_info->endio_raid56_workers = |
cb001095 JM |
2389 | btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, |
2390 | max_active, 4); | |
2a458198 | 2391 | fs_info->endio_repair_workers = |
cb001095 | 2392 | btrfs_alloc_workqueue(fs_info, "endio-repair", flags, 1, 0); |
2a458198 | 2393 | fs_info->rmw_workers = |
cb001095 | 2394 | btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); |
2a458198 | 2395 | fs_info->endio_write_workers = |
cb001095 JM |
2396 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2397 | max_active, 2); | |
2a458198 | 2398 | fs_info->endio_freespace_worker = |
cb001095 JM |
2399 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2400 | max_active, 0); | |
2a458198 | 2401 | fs_info->delayed_workers = |
cb001095 JM |
2402 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2403 | max_active, 0); | |
2a458198 | 2404 | fs_info->readahead_workers = |
cb001095 JM |
2405 | btrfs_alloc_workqueue(fs_info, "readahead", flags, |
2406 | max_active, 2); | |
2a458198 | 2407 | fs_info->qgroup_rescan_workers = |
cb001095 | 2408 | btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); |
2a458198 | 2409 | fs_info->extent_workers = |
cb001095 | 2410 | btrfs_alloc_workqueue(fs_info, "extent-refs", flags, |
2a458198 ES |
2411 | min_t(u64, fs_devices->num_devices, |
2412 | max_active), 8); | |
2413 | ||
2414 | if (!(fs_info->workers && fs_info->delalloc_workers && | |
2415 | fs_info->submit_workers && fs_info->flush_workers && | |
2416 | fs_info->endio_workers && fs_info->endio_meta_workers && | |
2417 | fs_info->endio_meta_write_workers && | |
2418 | fs_info->endio_repair_workers && | |
2419 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && | |
2420 | fs_info->endio_freespace_worker && fs_info->rmw_workers && | |
2421 | fs_info->caching_workers && fs_info->readahead_workers && | |
2422 | fs_info->fixup_workers && fs_info->delayed_workers && | |
2423 | fs_info->extent_workers && | |
2424 | fs_info->qgroup_rescan_workers)) { | |
2425 | return -ENOMEM; | |
2426 | } | |
2427 | ||
2428 | return 0; | |
2429 | } | |
2430 | ||
63443bf5 ES |
2431 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2432 | struct btrfs_fs_devices *fs_devices) | |
2433 | { | |
2434 | int ret; | |
2435 | struct btrfs_root *tree_root = fs_info->tree_root; | |
2436 | struct btrfs_root *log_tree_root; | |
2437 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2438 | u64 bytenr = btrfs_super_log_root(disk_super); | |
2439 | ||
2440 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2441 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2442 | return -EIO; |
2443 | } | |
2444 | ||
74e4d827 | 2445 | log_tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
63443bf5 ES |
2446 | if (!log_tree_root) |
2447 | return -ENOMEM; | |
2448 | ||
2449 | __setup_root(tree_root->nodesize, tree_root->sectorsize, | |
2450 | tree_root->stripesize, log_tree_root, fs_info, | |
2451 | BTRFS_TREE_LOG_OBJECTID); | |
2452 | ||
2453 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
2454 | fs_info->generation + 1); | |
64c043de | 2455 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2456 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2457 | ret = PTR_ERR(log_tree_root->node); |
64c043de | 2458 | kfree(log_tree_root); |
0eeff236 | 2459 | return ret; |
64c043de | 2460 | } else if (!extent_buffer_uptodate(log_tree_root->node)) { |
f14d104d | 2461 | btrfs_err(fs_info, "failed to read log tree"); |
63443bf5 ES |
2462 | free_extent_buffer(log_tree_root->node); |
2463 | kfree(log_tree_root); | |
2464 | return -EIO; | |
2465 | } | |
2466 | /* returns with log_tree_root freed on success */ | |
2467 | ret = btrfs_recover_log_trees(log_tree_root); | |
2468 | if (ret) { | |
34d97007 | 2469 | btrfs_handle_fs_error(tree_root->fs_info, ret, |
63443bf5 ES |
2470 | "Failed to recover log tree"); |
2471 | free_extent_buffer(log_tree_root->node); | |
2472 | kfree(log_tree_root); | |
2473 | return ret; | |
2474 | } | |
2475 | ||
2476 | if (fs_info->sb->s_flags & MS_RDONLY) { | |
2477 | ret = btrfs_commit_super(tree_root); | |
2478 | if (ret) | |
2479 | return ret; | |
2480 | } | |
2481 | ||
2482 | return 0; | |
2483 | } | |
2484 | ||
4bbcaa64 ES |
2485 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info, |
2486 | struct btrfs_root *tree_root) | |
2487 | { | |
a4f3d2c4 | 2488 | struct btrfs_root *root; |
4bbcaa64 ES |
2489 | struct btrfs_key location; |
2490 | int ret; | |
2491 | ||
2492 | location.objectid = BTRFS_EXTENT_TREE_OBJECTID; | |
2493 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2494 | location.offset = 0; | |
2495 | ||
a4f3d2c4 DS |
2496 | root = btrfs_read_tree_root(tree_root, &location); |
2497 | if (IS_ERR(root)) | |
2498 | return PTR_ERR(root); | |
2499 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2500 | fs_info->extent_root = root; | |
4bbcaa64 ES |
2501 | |
2502 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 DS |
2503 | root = btrfs_read_tree_root(tree_root, &location); |
2504 | if (IS_ERR(root)) | |
2505 | return PTR_ERR(root); | |
2506 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2507 | fs_info->dev_root = root; | |
4bbcaa64 ES |
2508 | btrfs_init_devices_late(fs_info); |
2509 | ||
2510 | location.objectid = BTRFS_CSUM_TREE_OBJECTID; | |
a4f3d2c4 DS |
2511 | root = btrfs_read_tree_root(tree_root, &location); |
2512 | if (IS_ERR(root)) | |
2513 | return PTR_ERR(root); | |
2514 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2515 | fs_info->csum_root = root; | |
4bbcaa64 ES |
2516 | |
2517 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; | |
a4f3d2c4 DS |
2518 | root = btrfs_read_tree_root(tree_root, &location); |
2519 | if (!IS_ERR(root)) { | |
2520 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
4bbcaa64 ES |
2521 | fs_info->quota_enabled = 1; |
2522 | fs_info->pending_quota_state = 1; | |
a4f3d2c4 | 2523 | fs_info->quota_root = root; |
4bbcaa64 ES |
2524 | } |
2525 | ||
2526 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2527 | root = btrfs_read_tree_root(tree_root, &location); |
2528 | if (IS_ERR(root)) { | |
2529 | ret = PTR_ERR(root); | |
4bbcaa64 ES |
2530 | if (ret != -ENOENT) |
2531 | return ret; | |
2532 | } else { | |
a4f3d2c4 DS |
2533 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2534 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2535 | } |
2536 | ||
70f6d82e OS |
2537 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
2538 | location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; | |
2539 | root = btrfs_read_tree_root(tree_root, &location); | |
2540 | if (IS_ERR(root)) | |
2541 | return PTR_ERR(root); | |
2542 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2543 | fs_info->free_space_root = root; | |
2544 | } | |
2545 | ||
4bbcaa64 ES |
2546 | return 0; |
2547 | } | |
2548 | ||
ad2b2c80 AV |
2549 | int open_ctree(struct super_block *sb, |
2550 | struct btrfs_fs_devices *fs_devices, | |
2551 | char *options) | |
2e635a27 | 2552 | { |
db94535d CM |
2553 | u32 sectorsize; |
2554 | u32 nodesize; | |
87ee04eb | 2555 | u32 stripesize; |
84234f3a | 2556 | u64 generation; |
f2b636e8 | 2557 | u64 features; |
3de4586c | 2558 | struct btrfs_key location; |
a061fc8d | 2559 | struct buffer_head *bh; |
4d34b278 | 2560 | struct btrfs_super_block *disk_super; |
815745cf | 2561 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
f84a8bd6 | 2562 | struct btrfs_root *tree_root; |
4d34b278 | 2563 | struct btrfs_root *chunk_root; |
eb60ceac | 2564 | int ret; |
e58ca020 | 2565 | int err = -EINVAL; |
af31f5e5 CM |
2566 | int num_backups_tried = 0; |
2567 | int backup_index = 0; | |
5cdc7ad3 | 2568 | int max_active; |
4543df7e | 2569 | |
74e4d827 DS |
2570 | tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
2571 | chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info, GFP_KERNEL); | |
cb517eab | 2572 | if (!tree_root || !chunk_root) { |
39279cc3 CM |
2573 | err = -ENOMEM; |
2574 | goto fail; | |
2575 | } | |
76dda93c YZ |
2576 | |
2577 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
2578 | if (ret) { | |
2579 | err = ret; | |
2580 | goto fail; | |
2581 | } | |
2582 | ||
2583 | ret = setup_bdi(fs_info, &fs_info->bdi); | |
2584 | if (ret) { | |
2585 | err = ret; | |
2586 | goto fail_srcu; | |
2587 | } | |
2588 | ||
908c7f19 | 2589 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); |
e2d84521 MX |
2590 | if (ret) { |
2591 | err = ret; | |
2592 | goto fail_bdi; | |
2593 | } | |
09cbfeaf | 2594 | fs_info->dirty_metadata_batch = PAGE_SIZE * |
e2d84521 MX |
2595 | (1 + ilog2(nr_cpu_ids)); |
2596 | ||
908c7f19 | 2597 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); |
963d678b MX |
2598 | if (ret) { |
2599 | err = ret; | |
2600 | goto fail_dirty_metadata_bytes; | |
2601 | } | |
2602 | ||
908c7f19 | 2603 | ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL); |
c404e0dc MX |
2604 | if (ret) { |
2605 | err = ret; | |
2606 | goto fail_delalloc_bytes; | |
2607 | } | |
2608 | ||
76dda93c YZ |
2609 | fs_info->btree_inode = new_inode(sb); |
2610 | if (!fs_info->btree_inode) { | |
2611 | err = -ENOMEM; | |
c404e0dc | 2612 | goto fail_bio_counter; |
76dda93c YZ |
2613 | } |
2614 | ||
a6591715 | 2615 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
1561deda | 2616 | |
76dda93c | 2617 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
f28491e0 | 2618 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2619 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2620 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2621 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2622 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2623 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2624 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2625 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 2626 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2627 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2628 | spin_lock_init(&fs_info->defrag_inodes_lock); |
2bf64758 | 2629 | spin_lock_init(&fs_info->free_chunk_lock); |
f29021b2 | 2630 | spin_lock_init(&fs_info->tree_mod_seq_lock); |
ceda0864 | 2631 | spin_lock_init(&fs_info->super_lock); |
fcebe456 | 2632 | spin_lock_init(&fs_info->qgroup_op_lock); |
f28491e0 | 2633 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2634 | spin_lock_init(&fs_info->unused_bgs_lock); |
f29021b2 | 2635 | rwlock_init(&fs_info->tree_mod_log_lock); |
d7c15171 | 2636 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
67c5e7d4 | 2637 | mutex_init(&fs_info->delete_unused_bgs_mutex); |
7585717f | 2638 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2639 | mutex_init(&fs_info->delalloc_root_mutex); |
c2d6cb16 | 2640 | mutex_init(&fs_info->cleaner_delayed_iput_mutex); |
de98ced9 | 2641 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2642 | |
0b86a832 | 2643 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2644 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2645 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 2646 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
0b86a832 | 2647 | btrfs_mapping_init(&fs_info->mapping_tree); |
66d8f3dd MX |
2648 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2649 | BTRFS_BLOCK_RSV_GLOBAL); | |
2650 | btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, | |
2651 | BTRFS_BLOCK_RSV_DELALLOC); | |
2652 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); | |
2653 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2654 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2655 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2656 | BTRFS_BLOCK_RSV_DELOPS); | |
cb03c743 | 2657 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 2658 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 2659 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 2660 | atomic_set(&fs_info->nr_async_bios, 0); |
4cb5300b | 2661 | atomic_set(&fs_info->defrag_running, 0); |
fcebe456 | 2662 | atomic_set(&fs_info->qgroup_op_seq, 0); |
2fefd558 | 2663 | atomic_set(&fs_info->reada_works_cnt, 0); |
fc36ed7e | 2664 | atomic64_set(&fs_info->tree_mod_seq, 0); |
9e7cc91a | 2665 | fs_info->fs_frozen = 0; |
e20d96d6 | 2666 | fs_info->sb = sb; |
95ac567a | 2667 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 2668 | fs_info->metadata_ratio = 0; |
4cb5300b | 2669 | fs_info->defrag_inodes = RB_ROOT; |
2bf64758 | 2670 | fs_info->free_chunk_space = 0; |
f29021b2 | 2671 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2672 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 2673 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
90519d66 | 2674 | /* readahead state */ |
d0164adc | 2675 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); |
90519d66 | 2676 | spin_lock_init(&fs_info->reada_lock); |
c8b97818 | 2677 | |
b34b086c CM |
2678 | fs_info->thread_pool_size = min_t(unsigned long, |
2679 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2680 | |
199c2a9c MX |
2681 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2682 | spin_lock_init(&fs_info->ordered_root_lock); | |
16cdcec7 | 2683 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
74e4d827 | 2684 | GFP_KERNEL); |
16cdcec7 MX |
2685 | if (!fs_info->delayed_root) { |
2686 | err = -ENOMEM; | |
2687 | goto fail_iput; | |
2688 | } | |
2689 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 2690 | |
638aa7ed | 2691 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
2692 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2693 | fs_info->check_integrity_print_mask = 0; | |
2694 | #endif | |
779a65a4 | 2695 | btrfs_init_balance(fs_info); |
21c7e756 | 2696 | btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); |
a2de733c | 2697 | |
a061fc8d CM |
2698 | sb->s_blocksize = 4096; |
2699 | sb->s_blocksize_bits = blksize_bits(4096); | |
32a88aa1 | 2700 | sb->s_bdi = &fs_info->bdi; |
a061fc8d | 2701 | |
f37938e0 | 2702 | btrfs_init_btree_inode(fs_info, tree_root); |
76dda93c | 2703 | |
0f9dd46c | 2704 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2705 | fs_info->block_group_cache_tree = RB_ROOT; |
a1897fdd | 2706 | fs_info->first_logical_byte = (u64)-1; |
0f9dd46c | 2707 | |
11833d66 | 2708 | extent_io_tree_init(&fs_info->freed_extents[0], |
f993c883 | 2709 | fs_info->btree_inode->i_mapping); |
11833d66 | 2710 | extent_io_tree_init(&fs_info->freed_extents[1], |
f993c883 | 2711 | fs_info->btree_inode->i_mapping); |
11833d66 | 2712 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
e66f709b | 2713 | fs_info->do_barriers = 1; |
e18e4809 | 2714 | |
39279cc3 | 2715 | |
5a3f23d5 | 2716 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2717 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2718 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2719 | mutex_init(&fs_info->transaction_kthread_mutex); |
2720 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 2721 | mutex_init(&fs_info->volume_mutex); |
1bbc621e | 2722 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 2723 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 2724 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2725 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2726 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 2727 | |
ad618368 | 2728 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 2729 | btrfs_init_qgroup(fs_info); |
416ac51d | 2730 | |
fa9c0d79 CM |
2731 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2732 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2733 | ||
e6dcd2dc | 2734 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2735 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2736 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2737 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2738 | |
04216820 FM |
2739 | INIT_LIST_HEAD(&fs_info->pinned_chunks); |
2740 | ||
53b381b3 DW |
2741 | ret = btrfs_alloc_stripe_hash_table(fs_info); |
2742 | if (ret) { | |
83c8266a | 2743 | err = ret; |
53b381b3 DW |
2744 | goto fail_alloc; |
2745 | } | |
2746 | ||
707e8a07 | 2747 | __setup_root(4096, 4096, 4096, tree_root, |
2c90e5d6 | 2748 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2749 | |
3c4bb26b | 2750 | invalidate_bdev(fs_devices->latest_bdev); |
1104a885 DS |
2751 | |
2752 | /* | |
2753 | * Read super block and check the signature bytes only | |
2754 | */ | |
a512bbf8 | 2755 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
92fc03fb AJ |
2756 | if (IS_ERR(bh)) { |
2757 | err = PTR_ERR(bh); | |
16cdcec7 | 2758 | goto fail_alloc; |
20b45077 | 2759 | } |
39279cc3 | 2760 | |
1104a885 DS |
2761 | /* |
2762 | * We want to check superblock checksum, the type is stored inside. | |
2763 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
2764 | */ | |
2765 | if (btrfs_check_super_csum(bh->b_data)) { | |
05135f59 | 2766 | btrfs_err(fs_info, "superblock checksum mismatch"); |
1104a885 | 2767 | err = -EINVAL; |
b2acdddf | 2768 | brelse(bh); |
1104a885 DS |
2769 | goto fail_alloc; |
2770 | } | |
2771 | ||
2772 | /* | |
2773 | * super_copy is zeroed at allocation time and we never touch the | |
2774 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
2775 | * the whole block of INFO_SIZE | |
2776 | */ | |
6c41761f DS |
2777 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2778 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2779 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2780 | brelse(bh); |
5f39d397 | 2781 | |
6c41761f | 2782 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2783 | |
1104a885 DS |
2784 | ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); |
2785 | if (ret) { | |
05135f59 | 2786 | btrfs_err(fs_info, "superblock contains fatal errors"); |
1104a885 DS |
2787 | err = -EINVAL; |
2788 | goto fail_alloc; | |
2789 | } | |
2790 | ||
6c41761f | 2791 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2792 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2793 | goto fail_alloc; |
0f7d52f4 | 2794 | |
acce952b | 2795 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
2796 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
2797 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 2798 | |
af31f5e5 CM |
2799 | /* |
2800 | * run through our array of backup supers and setup | |
2801 | * our ring pointer to the oldest one | |
2802 | */ | |
2803 | generation = btrfs_super_generation(disk_super); | |
2804 | find_oldest_super_backup(fs_info, generation); | |
2805 | ||
75e7cb7f LB |
2806 | /* |
2807 | * In the long term, we'll store the compression type in the super | |
2808 | * block, and it'll be used for per file compression control. | |
2809 | */ | |
2810 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2811 | ||
96da0919 | 2812 | ret = btrfs_parse_options(tree_root, options, sb->s_flags); |
2b82032c YZ |
2813 | if (ret) { |
2814 | err = ret; | |
16cdcec7 | 2815 | goto fail_alloc; |
2b82032c | 2816 | } |
dfe25020 | 2817 | |
f2b636e8 JB |
2818 | features = btrfs_super_incompat_flags(disk_super) & |
2819 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2820 | if (features) { | |
05135f59 DS |
2821 | btrfs_err(fs_info, |
2822 | "cannot mount because of unsupported optional features (%llx)", | |
2823 | features); | |
f2b636e8 | 2824 | err = -EINVAL; |
16cdcec7 | 2825 | goto fail_alloc; |
f2b636e8 JB |
2826 | } |
2827 | ||
5d4f98a2 | 2828 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 2829 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
69e380d1 | 2830 | if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 2831 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
727011e0 | 2832 | |
3173a18f | 2833 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
05135f59 | 2834 | btrfs_info(fs_info, "has skinny extents"); |
3173a18f | 2835 | |
727011e0 CM |
2836 | /* |
2837 | * flag our filesystem as having big metadata blocks if | |
2838 | * they are bigger than the page size | |
2839 | */ | |
09cbfeaf | 2840 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { |
727011e0 | 2841 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) |
05135f59 DS |
2842 | btrfs_info(fs_info, |
2843 | "flagging fs with big metadata feature"); | |
727011e0 CM |
2844 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; |
2845 | } | |
2846 | ||
bc3f116f | 2847 | nodesize = btrfs_super_nodesize(disk_super); |
bc3f116f | 2848 | sectorsize = btrfs_super_sectorsize(disk_super); |
b7f67055 | 2849 | stripesize = sectorsize; |
707e8a07 | 2850 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); |
963d678b | 2851 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); |
bc3f116f CM |
2852 | |
2853 | /* | |
2854 | * mixed block groups end up with duplicate but slightly offset | |
2855 | * extent buffers for the same range. It leads to corruptions | |
2856 | */ | |
2857 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
707e8a07 | 2858 | (sectorsize != nodesize)) { |
05135f59 DS |
2859 | btrfs_err(fs_info, |
2860 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
2861 | nodesize, sectorsize); | |
bc3f116f CM |
2862 | goto fail_alloc; |
2863 | } | |
2864 | ||
ceda0864 MX |
2865 | /* |
2866 | * Needn't use the lock because there is no other task which will | |
2867 | * update the flag. | |
2868 | */ | |
a6fa6fae | 2869 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 2870 | |
f2b636e8 JB |
2871 | features = btrfs_super_compat_ro_flags(disk_super) & |
2872 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
2873 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
05135f59 DS |
2874 | btrfs_err(fs_info, |
2875 | "cannot mount read-write because of unsupported optional features (%llx)", | |
c1c9ff7c | 2876 | features); |
f2b636e8 | 2877 | err = -EINVAL; |
16cdcec7 | 2878 | goto fail_alloc; |
f2b636e8 | 2879 | } |
61d92c32 | 2880 | |
5cdc7ad3 | 2881 | max_active = fs_info->thread_pool_size; |
61d92c32 | 2882 | |
2a458198 ES |
2883 | ret = btrfs_init_workqueues(fs_info, fs_devices); |
2884 | if (ret) { | |
2885 | err = ret; | |
0dc3b84a JB |
2886 | goto fail_sb_buffer; |
2887 | } | |
4543df7e | 2888 | |
4575c9cc | 2889 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 | 2890 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
09cbfeaf | 2891 | SZ_4M / PAGE_SIZE); |
4575c9cc | 2892 | |
db94535d | 2893 | tree_root->nodesize = nodesize; |
db94535d | 2894 | tree_root->sectorsize = sectorsize; |
87ee04eb | 2895 | tree_root->stripesize = stripesize; |
a061fc8d CM |
2896 | |
2897 | sb->s_blocksize = sectorsize; | |
2898 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 2899 | |
925baedd | 2900 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 2901 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 2902 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2903 | if (ret) { |
05135f59 | 2904 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 2905 | goto fail_sb_buffer; |
84eed90f | 2906 | } |
0b86a832 | 2907 | |
84234f3a | 2908 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 | 2909 | |
707e8a07 DS |
2910 | __setup_root(nodesize, sectorsize, stripesize, chunk_root, |
2911 | fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
0b86a832 CM |
2912 | |
2913 | chunk_root->node = read_tree_block(chunk_root, | |
2914 | btrfs_super_chunk_root(disk_super), | |
ce86cd59 | 2915 | generation); |
64c043de LB |
2916 | if (IS_ERR(chunk_root->node) || |
2917 | !extent_buffer_uptodate(chunk_root->node)) { | |
05135f59 | 2918 | btrfs_err(fs_info, "failed to read chunk root"); |
e5fffbac | 2919 | if (!IS_ERR(chunk_root->node)) |
2920 | free_extent_buffer(chunk_root->node); | |
95ab1f64 | 2921 | chunk_root->node = NULL; |
af31f5e5 | 2922 | goto fail_tree_roots; |
83121942 | 2923 | } |
5d4f98a2 YZ |
2924 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
2925 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 2926 | |
e17cade2 | 2927 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
b308bc2f | 2928 | btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); |
e17cade2 | 2929 | |
0b86a832 | 2930 | ret = btrfs_read_chunk_tree(chunk_root); |
2b82032c | 2931 | if (ret) { |
05135f59 | 2932 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 2933 | goto fail_tree_roots; |
2b82032c | 2934 | } |
0b86a832 | 2935 | |
8dabb742 SB |
2936 | /* |
2937 | * keep the device that is marked to be the target device for the | |
2938 | * dev_replace procedure | |
2939 | */ | |
9eaed21e | 2940 | btrfs_close_extra_devices(fs_devices, 0); |
dfe25020 | 2941 | |
a6b0d5c8 | 2942 | if (!fs_devices->latest_bdev) { |
05135f59 | 2943 | btrfs_err(fs_info, "failed to read devices"); |
a6b0d5c8 CM |
2944 | goto fail_tree_roots; |
2945 | } | |
2946 | ||
af31f5e5 | 2947 | retry_root_backup: |
84234f3a | 2948 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 2949 | |
e20d96d6 | 2950 | tree_root->node = read_tree_block(tree_root, |
db94535d | 2951 | btrfs_super_root(disk_super), |
ce86cd59 | 2952 | generation); |
64c043de LB |
2953 | if (IS_ERR(tree_root->node) || |
2954 | !extent_buffer_uptodate(tree_root->node)) { | |
05135f59 | 2955 | btrfs_warn(fs_info, "failed to read tree root"); |
e5fffbac | 2956 | if (!IS_ERR(tree_root->node)) |
2957 | free_extent_buffer(tree_root->node); | |
95ab1f64 | 2958 | tree_root->node = NULL; |
af31f5e5 | 2959 | goto recovery_tree_root; |
83121942 | 2960 | } |
af31f5e5 | 2961 | |
5d4f98a2 YZ |
2962 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
2963 | tree_root->commit_root = btrfs_root_node(tree_root); | |
69e9c6c6 | 2964 | btrfs_set_root_refs(&tree_root->root_item, 1); |
db94535d | 2965 | |
f32e48e9 CR |
2966 | mutex_lock(&tree_root->objectid_mutex); |
2967 | ret = btrfs_find_highest_objectid(tree_root, | |
2968 | &tree_root->highest_objectid); | |
2969 | if (ret) { | |
2970 | mutex_unlock(&tree_root->objectid_mutex); | |
2971 | goto recovery_tree_root; | |
2972 | } | |
2973 | ||
2974 | ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
2975 | ||
2976 | mutex_unlock(&tree_root->objectid_mutex); | |
2977 | ||
4bbcaa64 ES |
2978 | ret = btrfs_read_roots(fs_info, tree_root); |
2979 | if (ret) | |
af31f5e5 | 2980 | goto recovery_tree_root; |
f7a81ea4 | 2981 | |
8929ecfa YZ |
2982 | fs_info->generation = generation; |
2983 | fs_info->last_trans_committed = generation; | |
8929ecfa | 2984 | |
68310a5e ID |
2985 | ret = btrfs_recover_balance(fs_info); |
2986 | if (ret) { | |
05135f59 | 2987 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
2988 | goto fail_block_groups; |
2989 | } | |
2990 | ||
733f4fbb SB |
2991 | ret = btrfs_init_dev_stats(fs_info); |
2992 | if (ret) { | |
05135f59 | 2993 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
2994 | goto fail_block_groups; |
2995 | } | |
2996 | ||
8dabb742 SB |
2997 | ret = btrfs_init_dev_replace(fs_info); |
2998 | if (ret) { | |
05135f59 | 2999 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3000 | goto fail_block_groups; |
3001 | } | |
3002 | ||
9eaed21e | 3003 | btrfs_close_extra_devices(fs_devices, 1); |
8dabb742 | 3004 | |
b7c35e81 AJ |
3005 | ret = btrfs_sysfs_add_fsid(fs_devices, NULL); |
3006 | if (ret) { | |
05135f59 DS |
3007 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3008 | ret); | |
b7c35e81 AJ |
3009 | goto fail_block_groups; |
3010 | } | |
3011 | ||
3012 | ret = btrfs_sysfs_add_device(fs_devices); | |
3013 | if (ret) { | |
05135f59 DS |
3014 | btrfs_err(fs_info, "failed to init sysfs device interface: %d", |
3015 | ret); | |
b7c35e81 AJ |
3016 | goto fail_fsdev_sysfs; |
3017 | } | |
3018 | ||
96f3136e | 3019 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3020 | if (ret) { |
05135f59 | 3021 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3022 | goto fail_fsdev_sysfs; |
c59021f8 | 3023 | } |
3024 | ||
c59021f8 | 3025 | ret = btrfs_init_space_info(fs_info); |
3026 | if (ret) { | |
05135f59 | 3027 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3028 | goto fail_sysfs; |
c59021f8 | 3029 | } |
3030 | ||
4bbcaa64 | 3031 | ret = btrfs_read_block_groups(fs_info->extent_root); |
1b1d1f66 | 3032 | if (ret) { |
05135f59 | 3033 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3034 | goto fail_sysfs; |
1b1d1f66 | 3035 | } |
5af3e8cc SB |
3036 | fs_info->num_tolerated_disk_barrier_failures = |
3037 | btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); | |
292fd7fc SB |
3038 | if (fs_info->fs_devices->missing_devices > |
3039 | fs_info->num_tolerated_disk_barrier_failures && | |
3040 | !(sb->s_flags & MS_RDONLY)) { | |
05135f59 DS |
3041 | btrfs_warn(fs_info, |
3042 | "missing devices (%llu) exceeds the limit (%d), writeable mount is not allowed", | |
78fa1770 ZL |
3043 | fs_info->fs_devices->missing_devices, |
3044 | fs_info->num_tolerated_disk_barrier_failures); | |
2365dd3c | 3045 | goto fail_sysfs; |
292fd7fc | 3046 | } |
9078a3e1 | 3047 | |
a74a4b97 CM |
3048 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
3049 | "btrfs-cleaner"); | |
57506d50 | 3050 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 3051 | goto fail_sysfs; |
a74a4b97 CM |
3052 | |
3053 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3054 | tree_root, | |
3055 | "btrfs-transaction"); | |
57506d50 | 3056 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 3057 | goto fail_cleaner; |
a74a4b97 | 3058 | |
3cdde224 JM |
3059 | if (!btrfs_test_opt(tree_root->fs_info, SSD) && |
3060 | !btrfs_test_opt(tree_root->fs_info, NOSSD) && | |
c289811c | 3061 | !fs_info->fs_devices->rotating) { |
05135f59 | 3062 | btrfs_info(fs_info, "detected SSD devices, enabling SSD mode"); |
c289811c CM |
3063 | btrfs_set_opt(fs_info->mount_opt, SSD); |
3064 | } | |
3065 | ||
572d9ab7 | 3066 | /* |
01327610 | 3067 | * Mount does not set all options immediately, we can do it now and do |
572d9ab7 DS |
3068 | * not have to wait for transaction commit |
3069 | */ | |
3070 | btrfs_apply_pending_changes(fs_info); | |
3818aea2 | 3071 | |
21adbd5c | 3072 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3cdde224 | 3073 | if (btrfs_test_opt(tree_root->fs_info, CHECK_INTEGRITY)) { |
21adbd5c | 3074 | ret = btrfsic_mount(tree_root, fs_devices, |
3cdde224 | 3075 | btrfs_test_opt(tree_root->fs_info, |
21adbd5c SB |
3076 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? |
3077 | 1 : 0, | |
3078 | fs_info->check_integrity_print_mask); | |
3079 | if (ret) | |
05135f59 DS |
3080 | btrfs_warn(fs_info, |
3081 | "failed to initialize integrity check module: %d", | |
3082 | ret); | |
21adbd5c SB |
3083 | } |
3084 | #endif | |
bcef60f2 AJ |
3085 | ret = btrfs_read_qgroup_config(fs_info); |
3086 | if (ret) | |
3087 | goto fail_trans_kthread; | |
21adbd5c | 3088 | |
96da0919 QW |
3089 | /* do not make disk changes in broken FS or nologreplay is given */ |
3090 | if (btrfs_super_log_root(disk_super) != 0 && | |
3cdde224 | 3091 | !btrfs_test_opt(tree_root->fs_info, NOLOGREPLAY)) { |
63443bf5 | 3092 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 3093 | if (ret) { |
63443bf5 | 3094 | err = ret; |
28c16cbb | 3095 | goto fail_qgroup; |
79787eaa | 3096 | } |
e02119d5 | 3097 | } |
1a40e23b | 3098 | |
76dda93c | 3099 | ret = btrfs_find_orphan_roots(tree_root); |
79787eaa | 3100 | if (ret) |
28c16cbb | 3101 | goto fail_qgroup; |
76dda93c | 3102 | |
7c2ca468 | 3103 | if (!(sb->s_flags & MS_RDONLY)) { |
d68fc57b | 3104 | ret = btrfs_cleanup_fs_roots(fs_info); |
44c44af2 | 3105 | if (ret) |
28c16cbb | 3106 | goto fail_qgroup; |
90c711ab ZB |
3107 | |
3108 | mutex_lock(&fs_info->cleaner_mutex); | |
5d4f98a2 | 3109 | ret = btrfs_recover_relocation(tree_root); |
90c711ab | 3110 | mutex_unlock(&fs_info->cleaner_mutex); |
d7ce5843 | 3111 | if (ret < 0) { |
05135f59 DS |
3112 | btrfs_warn(fs_info, "failed to recover relocation: %d", |
3113 | ret); | |
d7ce5843 | 3114 | err = -EINVAL; |
bcef60f2 | 3115 | goto fail_qgroup; |
d7ce5843 | 3116 | } |
7c2ca468 | 3117 | } |
1a40e23b | 3118 | |
3de4586c CM |
3119 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
3120 | location.type = BTRFS_ROOT_ITEM_KEY; | |
cb517eab | 3121 | location.offset = 0; |
3de4586c | 3122 | |
3de4586c | 3123 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
3140c9a3 DC |
3124 | if (IS_ERR(fs_info->fs_root)) { |
3125 | err = PTR_ERR(fs_info->fs_root); | |
bcef60f2 | 3126 | goto fail_qgroup; |
3140c9a3 | 3127 | } |
c289811c | 3128 | |
2b6ba629 ID |
3129 | if (sb->s_flags & MS_RDONLY) |
3130 | return 0; | |
59641015 | 3131 | |
3cdde224 | 3132 | if (btrfs_test_opt(tree_root->fs_info, FREE_SPACE_TREE) && |
511711af | 3133 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
05135f59 | 3134 | btrfs_info(fs_info, "creating free space tree"); |
511711af CM |
3135 | ret = btrfs_create_free_space_tree(fs_info); |
3136 | if (ret) { | |
05135f59 DS |
3137 | btrfs_warn(fs_info, |
3138 | "failed to create free space tree: %d", ret); | |
511711af CM |
3139 | close_ctree(tree_root); |
3140 | return ret; | |
3141 | } | |
3142 | } | |
3143 | ||
2b6ba629 ID |
3144 | down_read(&fs_info->cleanup_work_sem); |
3145 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3146 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
e3acc2a6 | 3147 | up_read(&fs_info->cleanup_work_sem); |
2b6ba629 ID |
3148 | close_ctree(tree_root); |
3149 | return ret; | |
3150 | } | |
3151 | up_read(&fs_info->cleanup_work_sem); | |
59641015 | 3152 | |
2b6ba629 ID |
3153 | ret = btrfs_resume_balance_async(fs_info); |
3154 | if (ret) { | |
05135f59 | 3155 | btrfs_warn(fs_info, "failed to resume balance: %d", ret); |
2b6ba629 ID |
3156 | close_ctree(tree_root); |
3157 | return ret; | |
e3acc2a6 JB |
3158 | } |
3159 | ||
8dabb742 SB |
3160 | ret = btrfs_resume_dev_replace_async(fs_info); |
3161 | if (ret) { | |
05135f59 | 3162 | btrfs_warn(fs_info, "failed to resume device replace: %d", ret); |
8dabb742 SB |
3163 | close_ctree(tree_root); |
3164 | return ret; | |
3165 | } | |
3166 | ||
b382a324 JS |
3167 | btrfs_qgroup_rescan_resume(fs_info); |
3168 | ||
3cdde224 | 3169 | if (btrfs_test_opt(tree_root->fs_info, CLEAR_CACHE) && |
70f6d82e | 3170 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
05135f59 | 3171 | btrfs_info(fs_info, "clearing free space tree"); |
70f6d82e OS |
3172 | ret = btrfs_clear_free_space_tree(fs_info); |
3173 | if (ret) { | |
05135f59 DS |
3174 | btrfs_warn(fs_info, |
3175 | "failed to clear free space tree: %d", ret); | |
70f6d82e OS |
3176 | close_ctree(tree_root); |
3177 | return ret; | |
3178 | } | |
3179 | } | |
3180 | ||
4bbcaa64 | 3181 | if (!fs_info->uuid_root) { |
05135f59 | 3182 | btrfs_info(fs_info, "creating UUID tree"); |
f7a81ea4 SB |
3183 | ret = btrfs_create_uuid_tree(fs_info); |
3184 | if (ret) { | |
05135f59 DS |
3185 | btrfs_warn(fs_info, |
3186 | "failed to create the UUID tree: %d", ret); | |
f7a81ea4 SB |
3187 | close_ctree(tree_root); |
3188 | return ret; | |
3189 | } | |
3cdde224 | 3190 | } else if (btrfs_test_opt(tree_root->fs_info, RESCAN_UUID_TREE) || |
4bbcaa64 ES |
3191 | fs_info->generation != |
3192 | btrfs_super_uuid_tree_generation(disk_super)) { | |
05135f59 | 3193 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3194 | ret = btrfs_check_uuid_tree(fs_info); |
3195 | if (ret) { | |
05135f59 DS |
3196 | btrfs_warn(fs_info, |
3197 | "failed to check the UUID tree: %d", ret); | |
70f80175 SB |
3198 | close_ctree(tree_root); |
3199 | return ret; | |
3200 | } | |
3201 | } else { | |
3202 | fs_info->update_uuid_tree_gen = 1; | |
f7a81ea4 SB |
3203 | } |
3204 | ||
47ab2a6c JB |
3205 | fs_info->open = 1; |
3206 | ||
8dcddfa0 QW |
3207 | /* |
3208 | * backuproot only affect mount behavior, and if open_ctree succeeded, | |
3209 | * no need to keep the flag | |
3210 | */ | |
3211 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
3212 | ||
ad2b2c80 | 3213 | return 0; |
39279cc3 | 3214 | |
bcef60f2 AJ |
3215 | fail_qgroup: |
3216 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3217 | fail_trans_kthread: |
3218 | kthread_stop(fs_info->transaction_kthread); | |
54067ae9 | 3219 | btrfs_cleanup_transaction(fs_info->tree_root); |
faa2dbf0 | 3220 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3221 | fail_cleaner: |
a74a4b97 | 3222 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3223 | |
3224 | /* | |
3225 | * make sure we're done with the btree inode before we stop our | |
3226 | * kthreads | |
3227 | */ | |
3228 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3229 | |
2365dd3c | 3230 | fail_sysfs: |
6618a59b | 3231 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3232 | |
b7c35e81 AJ |
3233 | fail_fsdev_sysfs: |
3234 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3235 | ||
1b1d1f66 | 3236 | fail_block_groups: |
54067ae9 | 3237 | btrfs_put_block_group_cache(fs_info); |
1b1d1f66 | 3238 | btrfs_free_block_groups(fs_info); |
af31f5e5 CM |
3239 | |
3240 | fail_tree_roots: | |
3241 | free_root_pointers(fs_info, 1); | |
2b8195bb | 3242 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3243 | |
39279cc3 | 3244 | fail_sb_buffer: |
7abadb64 | 3245 | btrfs_stop_all_workers(fs_info); |
16cdcec7 | 3246 | fail_alloc: |
4543df7e | 3247 | fail_iput: |
586e46e2 ID |
3248 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3249 | ||
4543df7e | 3250 | iput(fs_info->btree_inode); |
c404e0dc MX |
3251 | fail_bio_counter: |
3252 | percpu_counter_destroy(&fs_info->bio_counter); | |
963d678b MX |
3253 | fail_delalloc_bytes: |
3254 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
e2d84521 MX |
3255 | fail_dirty_metadata_bytes: |
3256 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); | |
ad081f14 | 3257 | fail_bdi: |
7e662854 | 3258 | bdi_destroy(&fs_info->bdi); |
76dda93c YZ |
3259 | fail_srcu: |
3260 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 3261 | fail: |
53b381b3 | 3262 | btrfs_free_stripe_hash_table(fs_info); |
586e46e2 | 3263 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3264 | return err; |
af31f5e5 CM |
3265 | |
3266 | recovery_tree_root: | |
3cdde224 | 3267 | if (!btrfs_test_opt(tree_root->fs_info, USEBACKUPROOT)) |
af31f5e5 CM |
3268 | goto fail_tree_roots; |
3269 | ||
3270 | free_root_pointers(fs_info, 0); | |
3271 | ||
3272 | /* don't use the log in recovery mode, it won't be valid */ | |
3273 | btrfs_set_super_log_root(disk_super, 0); | |
3274 | ||
3275 | /* we can't trust the free space cache either */ | |
3276 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
3277 | ||
3278 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
3279 | &num_backups_tried, &backup_index); | |
3280 | if (ret == -1) | |
3281 | goto fail_block_groups; | |
3282 | goto retry_root_backup; | |
eb60ceac CM |
3283 | } |
3284 | ||
f2984462 CM |
3285 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
3286 | { | |
f2984462 CM |
3287 | if (uptodate) { |
3288 | set_buffer_uptodate(bh); | |
3289 | } else { | |
442a4f63 SB |
3290 | struct btrfs_device *device = (struct btrfs_device *) |
3291 | bh->b_private; | |
3292 | ||
b14af3b4 DS |
3293 | btrfs_warn_rl_in_rcu(device->dev_root->fs_info, |
3294 | "lost page write due to IO error on %s", | |
606686ee | 3295 | rcu_str_deref(device->name)); |
01327610 | 3296 | /* note, we don't set_buffer_write_io_error because we have |
1259ab75 CM |
3297 | * our own ways of dealing with the IO errors |
3298 | */ | |
f2984462 | 3299 | clear_buffer_uptodate(bh); |
442a4f63 | 3300 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); |
f2984462 CM |
3301 | } |
3302 | unlock_buffer(bh); | |
3303 | put_bh(bh); | |
3304 | } | |
3305 | ||
29c36d72 AJ |
3306 | int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num, |
3307 | struct buffer_head **bh_ret) | |
3308 | { | |
3309 | struct buffer_head *bh; | |
3310 | struct btrfs_super_block *super; | |
3311 | u64 bytenr; | |
3312 | ||
3313 | bytenr = btrfs_sb_offset(copy_num); | |
3314 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) | |
3315 | return -EINVAL; | |
3316 | ||
3317 | bh = __bread(bdev, bytenr / 4096, BTRFS_SUPER_INFO_SIZE); | |
3318 | /* | |
3319 | * If we fail to read from the underlying devices, as of now | |
3320 | * the best option we have is to mark it EIO. | |
3321 | */ | |
3322 | if (!bh) | |
3323 | return -EIO; | |
3324 | ||
3325 | super = (struct btrfs_super_block *)bh->b_data; | |
3326 | if (btrfs_super_bytenr(super) != bytenr || | |
3327 | btrfs_super_magic(super) != BTRFS_MAGIC) { | |
3328 | brelse(bh); | |
3329 | return -EINVAL; | |
3330 | } | |
3331 | ||
3332 | *bh_ret = bh; | |
3333 | return 0; | |
3334 | } | |
3335 | ||
3336 | ||
a512bbf8 YZ |
3337 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
3338 | { | |
3339 | struct buffer_head *bh; | |
3340 | struct buffer_head *latest = NULL; | |
3341 | struct btrfs_super_block *super; | |
3342 | int i; | |
3343 | u64 transid = 0; | |
92fc03fb | 3344 | int ret = -EINVAL; |
a512bbf8 YZ |
3345 | |
3346 | /* we would like to check all the supers, but that would make | |
3347 | * a btrfs mount succeed after a mkfs from a different FS. | |
3348 | * So, we need to add a special mount option to scan for | |
3349 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3350 | */ | |
3351 | for (i = 0; i < 1; i++) { | |
29c36d72 AJ |
3352 | ret = btrfs_read_dev_one_super(bdev, i, &bh); |
3353 | if (ret) | |
a512bbf8 YZ |
3354 | continue; |
3355 | ||
3356 | super = (struct btrfs_super_block *)bh->b_data; | |
a512bbf8 YZ |
3357 | |
3358 | if (!latest || btrfs_super_generation(super) > transid) { | |
3359 | brelse(latest); | |
3360 | latest = bh; | |
3361 | transid = btrfs_super_generation(super); | |
3362 | } else { | |
3363 | brelse(bh); | |
3364 | } | |
3365 | } | |
92fc03fb AJ |
3366 | |
3367 | if (!latest) | |
3368 | return ERR_PTR(ret); | |
3369 | ||
a512bbf8 YZ |
3370 | return latest; |
3371 | } | |
3372 | ||
4eedeb75 HH |
3373 | /* |
3374 | * this should be called twice, once with wait == 0 and | |
3375 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
3376 | * we write are pinned. | |
3377 | * | |
3378 | * They are released when wait == 1 is done. | |
3379 | * max_mirrors must be the same for both runs, and it indicates how | |
3380 | * many supers on this one device should be written. | |
3381 | * | |
3382 | * max_mirrors == 0 means to write them all. | |
3383 | */ | |
a512bbf8 YZ |
3384 | static int write_dev_supers(struct btrfs_device *device, |
3385 | struct btrfs_super_block *sb, | |
3386 | int do_barriers, int wait, int max_mirrors) | |
3387 | { | |
3388 | struct buffer_head *bh; | |
3389 | int i; | |
3390 | int ret; | |
3391 | int errors = 0; | |
3392 | u32 crc; | |
3393 | u64 bytenr; | |
a512bbf8 YZ |
3394 | |
3395 | if (max_mirrors == 0) | |
3396 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3397 | ||
a512bbf8 YZ |
3398 | for (i = 0; i < max_mirrors; i++) { |
3399 | bytenr = btrfs_sb_offset(i); | |
935e5cc9 MX |
3400 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3401 | device->commit_total_bytes) | |
a512bbf8 YZ |
3402 | break; |
3403 | ||
3404 | if (wait) { | |
3405 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
3406 | BTRFS_SUPER_INFO_SIZE); | |
634554dc JB |
3407 | if (!bh) { |
3408 | errors++; | |
3409 | continue; | |
3410 | } | |
a512bbf8 | 3411 | wait_on_buffer(bh); |
4eedeb75 HH |
3412 | if (!buffer_uptodate(bh)) |
3413 | errors++; | |
3414 | ||
3415 | /* drop our reference */ | |
3416 | brelse(bh); | |
3417 | ||
3418 | /* drop the reference from the wait == 0 run */ | |
3419 | brelse(bh); | |
3420 | continue; | |
a512bbf8 YZ |
3421 | } else { |
3422 | btrfs_set_super_bytenr(sb, bytenr); | |
3423 | ||
3424 | crc = ~(u32)0; | |
b0496686 | 3425 | crc = btrfs_csum_data((char *)sb + |
a512bbf8 YZ |
3426 | BTRFS_CSUM_SIZE, crc, |
3427 | BTRFS_SUPER_INFO_SIZE - | |
3428 | BTRFS_CSUM_SIZE); | |
3429 | btrfs_csum_final(crc, sb->csum); | |
3430 | ||
4eedeb75 HH |
3431 | /* |
3432 | * one reference for us, and we leave it for the | |
3433 | * caller | |
3434 | */ | |
a512bbf8 YZ |
3435 | bh = __getblk(device->bdev, bytenr / 4096, |
3436 | BTRFS_SUPER_INFO_SIZE); | |
634554dc | 3437 | if (!bh) { |
f14d104d DS |
3438 | btrfs_err(device->dev_root->fs_info, |
3439 | "couldn't get super buffer head for bytenr %llu", | |
3440 | bytenr); | |
634554dc JB |
3441 | errors++; |
3442 | continue; | |
3443 | } | |
3444 | ||
a512bbf8 YZ |
3445 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
3446 | ||
4eedeb75 | 3447 | /* one reference for submit_bh */ |
a512bbf8 | 3448 | get_bh(bh); |
4eedeb75 HH |
3449 | |
3450 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
3451 | lock_buffer(bh); |
3452 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
442a4f63 | 3453 | bh->b_private = device; |
a512bbf8 YZ |
3454 | } |
3455 | ||
387125fc CM |
3456 | /* |
3457 | * we fua the first super. The others we allow | |
3458 | * to go down lazy. | |
3459 | */ | |
e8117c26 | 3460 | if (i == 0) |
2a222ca9 | 3461 | ret = btrfsic_submit_bh(REQ_OP_WRITE, WRITE_FUA, bh); |
e8117c26 | 3462 | else |
2a222ca9 | 3463 | ret = btrfsic_submit_bh(REQ_OP_WRITE, WRITE_SYNC, bh); |
4eedeb75 | 3464 | if (ret) |
a512bbf8 | 3465 | errors++; |
a512bbf8 YZ |
3466 | } |
3467 | return errors < i ? 0 : -1; | |
3468 | } | |
3469 | ||
387125fc CM |
3470 | /* |
3471 | * endio for the write_dev_flush, this will wake anyone waiting | |
3472 | * for the barrier when it is done | |
3473 | */ | |
4246a0b6 | 3474 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 3475 | { |
387125fc CM |
3476 | if (bio->bi_private) |
3477 | complete(bio->bi_private); | |
3478 | bio_put(bio); | |
3479 | } | |
3480 | ||
3481 | /* | |
3482 | * trigger flushes for one the devices. If you pass wait == 0, the flushes are | |
3483 | * sent down. With wait == 1, it waits for the previous flush. | |
3484 | * | |
3485 | * any device where the flush fails with eopnotsupp are flagged as not-barrier | |
3486 | * capable | |
3487 | */ | |
3488 | static int write_dev_flush(struct btrfs_device *device, int wait) | |
3489 | { | |
3490 | struct bio *bio; | |
3491 | int ret = 0; | |
3492 | ||
3493 | if (device->nobarriers) | |
3494 | return 0; | |
3495 | ||
3496 | if (wait) { | |
3497 | bio = device->flush_bio; | |
3498 | if (!bio) | |
3499 | return 0; | |
3500 | ||
3501 | wait_for_completion(&device->flush_wait); | |
3502 | ||
4246a0b6 CH |
3503 | if (bio->bi_error) { |
3504 | ret = bio->bi_error; | |
5af3e8cc SB |
3505 | btrfs_dev_stat_inc_and_print(device, |
3506 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
387125fc CM |
3507 | } |
3508 | ||
3509 | /* drop the reference from the wait == 0 run */ | |
3510 | bio_put(bio); | |
3511 | device->flush_bio = NULL; | |
3512 | ||
3513 | return ret; | |
3514 | } | |
3515 | ||
3516 | /* | |
3517 | * one reference for us, and we leave it for the | |
3518 | * caller | |
3519 | */ | |
9c017abc | 3520 | device->flush_bio = NULL; |
9be3395b | 3521 | bio = btrfs_io_bio_alloc(GFP_NOFS, 0); |
387125fc CM |
3522 | if (!bio) |
3523 | return -ENOMEM; | |
3524 | ||
3525 | bio->bi_end_io = btrfs_end_empty_barrier; | |
3526 | bio->bi_bdev = device->bdev; | |
37226b21 | 3527 | bio_set_op_attrs(bio, REQ_OP_WRITE, WRITE_FLUSH); |
387125fc CM |
3528 | init_completion(&device->flush_wait); |
3529 | bio->bi_private = &device->flush_wait; | |
3530 | device->flush_bio = bio; | |
3531 | ||
3532 | bio_get(bio); | |
4e49ea4a | 3533 | btrfsic_submit_bio(bio); |
387125fc CM |
3534 | |
3535 | return 0; | |
3536 | } | |
3537 | ||
3538 | /* | |
3539 | * send an empty flush down to each device in parallel, | |
3540 | * then wait for them | |
3541 | */ | |
3542 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
3543 | { | |
3544 | struct list_head *head; | |
3545 | struct btrfs_device *dev; | |
5af3e8cc SB |
3546 | int errors_send = 0; |
3547 | int errors_wait = 0; | |
387125fc CM |
3548 | int ret; |
3549 | ||
3550 | /* send down all the barriers */ | |
3551 | head = &info->fs_devices->devices; | |
3552 | list_for_each_entry_rcu(dev, head, dev_list) { | |
f88ba6a2 HS |
3553 | if (dev->missing) |
3554 | continue; | |
387125fc | 3555 | if (!dev->bdev) { |
5af3e8cc | 3556 | errors_send++; |
387125fc CM |
3557 | continue; |
3558 | } | |
3559 | if (!dev->in_fs_metadata || !dev->writeable) | |
3560 | continue; | |
3561 | ||
3562 | ret = write_dev_flush(dev, 0); | |
3563 | if (ret) | |
5af3e8cc | 3564 | errors_send++; |
387125fc CM |
3565 | } |
3566 | ||
3567 | /* wait for all the barriers */ | |
3568 | list_for_each_entry_rcu(dev, head, dev_list) { | |
f88ba6a2 HS |
3569 | if (dev->missing) |
3570 | continue; | |
387125fc | 3571 | if (!dev->bdev) { |
5af3e8cc | 3572 | errors_wait++; |
387125fc CM |
3573 | continue; |
3574 | } | |
3575 | if (!dev->in_fs_metadata || !dev->writeable) | |
3576 | continue; | |
3577 | ||
3578 | ret = write_dev_flush(dev, 1); | |
3579 | if (ret) | |
5af3e8cc | 3580 | errors_wait++; |
387125fc | 3581 | } |
5af3e8cc SB |
3582 | if (errors_send > info->num_tolerated_disk_barrier_failures || |
3583 | errors_wait > info->num_tolerated_disk_barrier_failures) | |
387125fc CM |
3584 | return -EIO; |
3585 | return 0; | |
3586 | } | |
3587 | ||
943c6e99 ZL |
3588 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
3589 | { | |
8789f4fe ZL |
3590 | int raid_type; |
3591 | int min_tolerated = INT_MAX; | |
943c6e99 | 3592 | |
8789f4fe ZL |
3593 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
3594 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
3595 | min_tolerated = min(min_tolerated, | |
3596 | btrfs_raid_array[BTRFS_RAID_SINGLE]. | |
3597 | tolerated_failures); | |
943c6e99 | 3598 | |
8789f4fe ZL |
3599 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
3600 | if (raid_type == BTRFS_RAID_SINGLE) | |
3601 | continue; | |
3602 | if (!(flags & btrfs_raid_group[raid_type])) | |
3603 | continue; | |
3604 | min_tolerated = min(min_tolerated, | |
3605 | btrfs_raid_array[raid_type]. | |
3606 | tolerated_failures); | |
3607 | } | |
943c6e99 | 3608 | |
8789f4fe ZL |
3609 | if (min_tolerated == INT_MAX) { |
3610 | pr_warn("BTRFS: unknown raid flag: %llu\n", flags); | |
3611 | min_tolerated = 0; | |
3612 | } | |
3613 | ||
3614 | return min_tolerated; | |
943c6e99 ZL |
3615 | } |
3616 | ||
5af3e8cc SB |
3617 | int btrfs_calc_num_tolerated_disk_barrier_failures( |
3618 | struct btrfs_fs_info *fs_info) | |
3619 | { | |
3620 | struct btrfs_ioctl_space_info space; | |
3621 | struct btrfs_space_info *sinfo; | |
3622 | u64 types[] = {BTRFS_BLOCK_GROUP_DATA, | |
3623 | BTRFS_BLOCK_GROUP_SYSTEM, | |
3624 | BTRFS_BLOCK_GROUP_METADATA, | |
3625 | BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; | |
5af3e8cc SB |
3626 | int i; |
3627 | int c; | |
3628 | int num_tolerated_disk_barrier_failures = | |
3629 | (int)fs_info->fs_devices->num_devices; | |
3630 | ||
2c458045 | 3631 | for (i = 0; i < ARRAY_SIZE(types); i++) { |
5af3e8cc SB |
3632 | struct btrfs_space_info *tmp; |
3633 | ||
3634 | sinfo = NULL; | |
3635 | rcu_read_lock(); | |
3636 | list_for_each_entry_rcu(tmp, &fs_info->space_info, list) { | |
3637 | if (tmp->flags == types[i]) { | |
3638 | sinfo = tmp; | |
3639 | break; | |
3640 | } | |
3641 | } | |
3642 | rcu_read_unlock(); | |
3643 | ||
3644 | if (!sinfo) | |
3645 | continue; | |
3646 | ||
3647 | down_read(&sinfo->groups_sem); | |
3648 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
2c458045 ZL |
3649 | u64 flags; |
3650 | ||
3651 | if (list_empty(&sinfo->block_groups[c])) | |
3652 | continue; | |
3653 | ||
3654 | btrfs_get_block_group_info(&sinfo->block_groups[c], | |
3655 | &space); | |
3656 | if (space.total_bytes == 0 || space.used_bytes == 0) | |
3657 | continue; | |
3658 | flags = space.flags; | |
943c6e99 ZL |
3659 | |
3660 | num_tolerated_disk_barrier_failures = min( | |
3661 | num_tolerated_disk_barrier_failures, | |
3662 | btrfs_get_num_tolerated_disk_barrier_failures( | |
3663 | flags)); | |
5af3e8cc SB |
3664 | } |
3665 | up_read(&sinfo->groups_sem); | |
3666 | } | |
3667 | ||
3668 | return num_tolerated_disk_barrier_failures; | |
3669 | } | |
3670 | ||
48a3b636 | 3671 | static int write_all_supers(struct btrfs_root *root, int max_mirrors) |
f2984462 | 3672 | { |
e5e9a520 | 3673 | struct list_head *head; |
f2984462 | 3674 | struct btrfs_device *dev; |
a061fc8d | 3675 | struct btrfs_super_block *sb; |
f2984462 | 3676 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
3677 | int ret; |
3678 | int do_barriers; | |
a236aed1 CM |
3679 | int max_errors; |
3680 | int total_errors = 0; | |
a061fc8d | 3681 | u64 flags; |
f2984462 | 3682 | |
3cdde224 | 3683 | do_barriers = !btrfs_test_opt(root->fs_info, NOBARRIER); |
af31f5e5 | 3684 | backup_super_roots(root->fs_info); |
f2984462 | 3685 | |
6c41761f | 3686 | sb = root->fs_info->super_for_commit; |
a061fc8d | 3687 | dev_item = &sb->dev_item; |
e5e9a520 | 3688 | |
174ba509 | 3689 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); |
e5e9a520 | 3690 | head = &root->fs_info->fs_devices->devices; |
d7306801 | 3691 | max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; |
387125fc | 3692 | |
5af3e8cc SB |
3693 | if (do_barriers) { |
3694 | ret = barrier_all_devices(root->fs_info); | |
3695 | if (ret) { | |
3696 | mutex_unlock( | |
3697 | &root->fs_info->fs_devices->device_list_mutex); | |
34d97007 | 3698 | btrfs_handle_fs_error(root->fs_info, ret, |
5af3e8cc SB |
3699 | "errors while submitting device barriers."); |
3700 | return ret; | |
3701 | } | |
3702 | } | |
387125fc | 3703 | |
1f78160c | 3704 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3705 | if (!dev->bdev) { |
3706 | total_errors++; | |
3707 | continue; | |
3708 | } | |
2b82032c | 3709 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3710 | continue; |
3711 | ||
2b82032c | 3712 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
3713 | btrfs_set_stack_device_type(dev_item, dev->type); |
3714 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 3715 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 3716 | dev->commit_total_bytes); |
ce7213c7 MX |
3717 | btrfs_set_stack_device_bytes_used(dev_item, |
3718 | dev->commit_bytes_used); | |
a061fc8d CM |
3719 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
3720 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
3721 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
3722 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 3723 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 3724 | |
a061fc8d CM |
3725 | flags = btrfs_super_flags(sb); |
3726 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
3727 | ||
a512bbf8 | 3728 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
3729 | if (ret) |
3730 | total_errors++; | |
f2984462 | 3731 | } |
a236aed1 | 3732 | if (total_errors > max_errors) { |
efe120a0 | 3733 | btrfs_err(root->fs_info, "%d errors while writing supers", |
d397712b | 3734 | total_errors); |
a724b436 | 3735 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
79787eaa | 3736 | |
9d565ba4 | 3737 | /* FUA is masked off if unsupported and can't be the reason */ |
34d97007 | 3738 | btrfs_handle_fs_error(root->fs_info, -EIO, |
9d565ba4 SB |
3739 | "%d errors while writing supers", total_errors); |
3740 | return -EIO; | |
a236aed1 | 3741 | } |
f2984462 | 3742 | |
a512bbf8 | 3743 | total_errors = 0; |
1f78160c | 3744 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3745 | if (!dev->bdev) |
3746 | continue; | |
2b82032c | 3747 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3748 | continue; |
3749 | ||
a512bbf8 YZ |
3750 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
3751 | if (ret) | |
3752 | total_errors++; | |
f2984462 | 3753 | } |
174ba509 | 3754 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 3755 | if (total_errors > max_errors) { |
34d97007 | 3756 | btrfs_handle_fs_error(root->fs_info, -EIO, |
79787eaa JM |
3757 | "%d errors while writing supers", total_errors); |
3758 | return -EIO; | |
a236aed1 | 3759 | } |
f2984462 CM |
3760 | return 0; |
3761 | } | |
3762 | ||
a512bbf8 YZ |
3763 | int write_ctree_super(struct btrfs_trans_handle *trans, |
3764 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 3765 | { |
f570e757 | 3766 | return write_all_supers(root, max_mirrors); |
cfaa7295 CM |
3767 | } |
3768 | ||
cb517eab MX |
3769 | /* Drop a fs root from the radix tree and free it. */ |
3770 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
3771 | struct btrfs_root *root) | |
2619ba1f | 3772 | { |
4df27c4d | 3773 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
3774 | radix_tree_delete(&fs_info->fs_roots_radix, |
3775 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 3776 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
3777 | |
3778 | if (btrfs_root_refs(&root->root_item) == 0) | |
3779 | synchronize_srcu(&fs_info->subvol_srcu); | |
3780 | ||
1c1ea4f7 | 3781 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { |
3321719e | 3782 | btrfs_free_log(NULL, root); |
1c1ea4f7 LB |
3783 | if (root->reloc_root) { |
3784 | free_extent_buffer(root->reloc_root->node); | |
3785 | free_extent_buffer(root->reloc_root->commit_root); | |
3786 | btrfs_put_fs_root(root->reloc_root); | |
3787 | root->reloc_root = NULL; | |
3788 | } | |
3789 | } | |
3321719e | 3790 | |
faa2dbf0 JB |
3791 | if (root->free_ino_pinned) |
3792 | __btrfs_remove_free_space_cache(root->free_ino_pinned); | |
3793 | if (root->free_ino_ctl) | |
3794 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d | 3795 | free_fs_root(root); |
4df27c4d YZ |
3796 | } |
3797 | ||
3798 | static void free_fs_root(struct btrfs_root *root) | |
3799 | { | |
57cdc8db | 3800 | iput(root->ino_cache_inode); |
4df27c4d | 3801 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
1cb048f5 FDBM |
3802 | btrfs_free_block_rsv(root, root->orphan_block_rsv); |
3803 | root->orphan_block_rsv = NULL; | |
0ee5dc67 AV |
3804 | if (root->anon_dev) |
3805 | free_anon_bdev(root->anon_dev); | |
8257b2dc MX |
3806 | if (root->subv_writers) |
3807 | btrfs_free_subvolume_writers(root->subv_writers); | |
4df27c4d YZ |
3808 | free_extent_buffer(root->node); |
3809 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
3810 | kfree(root->free_ino_ctl); |
3811 | kfree(root->free_ino_pinned); | |
d397712b | 3812 | kfree(root->name); |
b0feb9d9 | 3813 | btrfs_put_fs_root(root); |
2619ba1f CM |
3814 | } |
3815 | ||
cb517eab MX |
3816 | void btrfs_free_fs_root(struct btrfs_root *root) |
3817 | { | |
3818 | free_fs_root(root); | |
2619ba1f CM |
3819 | } |
3820 | ||
c146afad | 3821 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 3822 | { |
c146afad YZ |
3823 | u64 root_objectid = 0; |
3824 | struct btrfs_root *gang[8]; | |
65d33fd7 QW |
3825 | int i = 0; |
3826 | int err = 0; | |
3827 | unsigned int ret = 0; | |
3828 | int index; | |
e089f05c | 3829 | |
c146afad | 3830 | while (1) { |
65d33fd7 | 3831 | index = srcu_read_lock(&fs_info->subvol_srcu); |
c146afad YZ |
3832 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
3833 | (void **)gang, root_objectid, | |
3834 | ARRAY_SIZE(gang)); | |
65d33fd7 QW |
3835 | if (!ret) { |
3836 | srcu_read_unlock(&fs_info->subvol_srcu, index); | |
c146afad | 3837 | break; |
65d33fd7 | 3838 | } |
5d4f98a2 | 3839 | root_objectid = gang[ret - 1]->root_key.objectid + 1; |
65d33fd7 | 3840 | |
c146afad | 3841 | for (i = 0; i < ret; i++) { |
65d33fd7 QW |
3842 | /* Avoid to grab roots in dead_roots */ |
3843 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
3844 | gang[i] = NULL; | |
3845 | continue; | |
3846 | } | |
3847 | /* grab all the search result for later use */ | |
3848 | gang[i] = btrfs_grab_fs_root(gang[i]); | |
3849 | } | |
3850 | srcu_read_unlock(&fs_info->subvol_srcu, index); | |
66b4ffd1 | 3851 | |
65d33fd7 QW |
3852 | for (i = 0; i < ret; i++) { |
3853 | if (!gang[i]) | |
3854 | continue; | |
c146afad | 3855 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
3856 | err = btrfs_orphan_cleanup(gang[i]); |
3857 | if (err) | |
65d33fd7 QW |
3858 | break; |
3859 | btrfs_put_fs_root(gang[i]); | |
c146afad YZ |
3860 | } |
3861 | root_objectid++; | |
3862 | } | |
65d33fd7 QW |
3863 | |
3864 | /* release the uncleaned roots due to error */ | |
3865 | for (; i < ret; i++) { | |
3866 | if (gang[i]) | |
3867 | btrfs_put_fs_root(gang[i]); | |
3868 | } | |
3869 | return err; | |
c146afad | 3870 | } |
a2135011 | 3871 | |
c146afad YZ |
3872 | int btrfs_commit_super(struct btrfs_root *root) |
3873 | { | |
3874 | struct btrfs_trans_handle *trans; | |
a74a4b97 | 3875 | |
c146afad | 3876 | mutex_lock(&root->fs_info->cleaner_mutex); |
24bbcf04 | 3877 | btrfs_run_delayed_iputs(root); |
c146afad | 3878 | mutex_unlock(&root->fs_info->cleaner_mutex); |
9d1a2a3a | 3879 | wake_up_process(root->fs_info->cleaner_kthread); |
c71bf099 YZ |
3880 | |
3881 | /* wait until ongoing cleanup work done */ | |
3882 | down_write(&root->fs_info->cleanup_work_sem); | |
3883 | up_write(&root->fs_info->cleanup_work_sem); | |
3884 | ||
7a7eaa40 | 3885 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3886 | if (IS_ERR(trans)) |
3887 | return PTR_ERR(trans); | |
d52c1bcc | 3888 | return btrfs_commit_transaction(trans, root); |
c146afad YZ |
3889 | } |
3890 | ||
3abdbd78 | 3891 | void close_ctree(struct btrfs_root *root) |
c146afad YZ |
3892 | { |
3893 | struct btrfs_fs_info *fs_info = root->fs_info; | |
3894 | int ret; | |
3895 | ||
3896 | fs_info->closing = 1; | |
3897 | smp_mb(); | |
3898 | ||
7343dd61 | 3899 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 3900 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 3901 | |
803b2f54 SB |
3902 | /* wait for the uuid_scan task to finish */ |
3903 | down(&fs_info->uuid_tree_rescan_sem); | |
3904 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
3905 | up(&fs_info->uuid_tree_rescan_sem); | |
3906 | ||
837d5b6e | 3907 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 3908 | btrfs_pause_balance(fs_info); |
837d5b6e | 3909 | |
8dabb742 SB |
3910 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
3911 | ||
aa1b8cd4 | 3912 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
3913 | |
3914 | /* wait for any defraggers to finish */ | |
3915 | wait_event(fs_info->transaction_wait, | |
3916 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3917 | ||
3918 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 3919 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 3920 | |
21c7e756 MX |
3921 | cancel_work_sync(&fs_info->async_reclaim_work); |
3922 | ||
c146afad | 3923 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
e44163e1 JM |
3924 | /* |
3925 | * If the cleaner thread is stopped and there are | |
3926 | * block groups queued for removal, the deletion will be | |
3927 | * skipped when we quit the cleaner thread. | |
3928 | */ | |
e44163e1 | 3929 | btrfs_delete_unused_bgs(root->fs_info); |
e44163e1 | 3930 | |
acce952b | 3931 | ret = btrfs_commit_super(root); |
3932 | if (ret) | |
04892340 | 3933 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 3934 | } |
3935 | ||
87533c47 | 3936 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) |
68ce9682 | 3937 | btrfs_error_commit_super(root); |
0f7d52f4 | 3938 | |
e3029d9f AV |
3939 | kthread_stop(fs_info->transaction_kthread); |
3940 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 3941 | |
f25784b3 YZ |
3942 | fs_info->closing = 2; |
3943 | smp_mb(); | |
3944 | ||
04892340 | 3945 | btrfs_free_qgroup_config(fs_info); |
bcef60f2 | 3946 | |
963d678b | 3947 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 3948 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 3949 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 3950 | } |
bcc63abb | 3951 | |
6618a59b | 3952 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 3953 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 3954 | |
faa2dbf0 | 3955 | btrfs_free_fs_roots(fs_info); |
d10c5f31 | 3956 | |
1a4319cc LB |
3957 | btrfs_put_block_group_cache(fs_info); |
3958 | ||
2b1360da JB |
3959 | btrfs_free_block_groups(fs_info); |
3960 | ||
de348ee0 WS |
3961 | /* |
3962 | * we must make sure there is not any read request to | |
3963 | * submit after we stopping all workers. | |
3964 | */ | |
3965 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
3966 | btrfs_stop_all_workers(fs_info); |
3967 | ||
47ab2a6c | 3968 | fs_info->open = 0; |
13e6c37b | 3969 | free_root_pointers(fs_info, 1); |
9ad6b7bc | 3970 | |
13e6c37b | 3971 | iput(fs_info->btree_inode); |
d6bfde87 | 3972 | |
21adbd5c | 3973 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3cdde224 | 3974 | if (btrfs_test_opt(root->fs_info, CHECK_INTEGRITY)) |
21adbd5c SB |
3975 | btrfsic_unmount(root, fs_info->fs_devices); |
3976 | #endif | |
3977 | ||
dfe25020 | 3978 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 3979 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 3980 | |
e2d84521 | 3981 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
963d678b | 3982 | percpu_counter_destroy(&fs_info->delalloc_bytes); |
c404e0dc | 3983 | percpu_counter_destroy(&fs_info->bio_counter); |
04160088 | 3984 | bdi_destroy(&fs_info->bdi); |
76dda93c | 3985 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 3986 | |
53b381b3 DW |
3987 | btrfs_free_stripe_hash_table(fs_info); |
3988 | ||
cdfb080e | 3989 | __btrfs_free_block_rsv(root->orphan_block_rsv); |
1cb048f5 | 3990 | root->orphan_block_rsv = NULL; |
04216820 FM |
3991 | |
3992 | lock_chunks(root); | |
3993 | while (!list_empty(&fs_info->pinned_chunks)) { | |
3994 | struct extent_map *em; | |
3995 | ||
3996 | em = list_first_entry(&fs_info->pinned_chunks, | |
3997 | struct extent_map, list); | |
3998 | list_del_init(&em->list); | |
3999 | free_extent_map(em); | |
4000 | } | |
4001 | unlock_chunks(root); | |
eb60ceac CM |
4002 | } |
4003 | ||
b9fab919 CM |
4004 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
4005 | int atomic) | |
5f39d397 | 4006 | { |
1259ab75 | 4007 | int ret; |
727011e0 | 4008 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 4009 | |
0b32f4bb | 4010 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
4011 | if (!ret) |
4012 | return ret; | |
4013 | ||
4014 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
4015 | parent_transid, atomic); |
4016 | if (ret == -EAGAIN) | |
4017 | return ret; | |
1259ab75 | 4018 | return !ret; |
5f39d397 CM |
4019 | } |
4020 | ||
5f39d397 CM |
4021 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
4022 | { | |
06ea65a3 | 4023 | struct btrfs_root *root; |
5f39d397 | 4024 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 4025 | int was_dirty; |
b4ce94de | 4026 | |
06ea65a3 JB |
4027 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4028 | /* | |
4029 | * This is a fast path so only do this check if we have sanity tests | |
4030 | * enabled. Normal people shouldn't be marking dummy buffers as dirty | |
4031 | * outside of the sanity tests. | |
4032 | */ | |
4033 | if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) | |
4034 | return; | |
4035 | #endif | |
4036 | root = BTRFS_I(buf->pages[0]->mapping->host)->root; | |
b9447ef8 | 4037 | btrfs_assert_tree_locked(buf); |
31b1a2bd JL |
4038 | if (transid != root->fs_info->generation) |
4039 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, " | |
d397712b | 4040 | "found %llu running %llu\n", |
c1c9ff7c | 4041 | buf->start, transid, root->fs_info->generation); |
0b32f4bb | 4042 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 MX |
4043 | if (!was_dirty) |
4044 | __percpu_counter_add(&root->fs_info->dirty_metadata_bytes, | |
4045 | buf->len, | |
4046 | root->fs_info->dirty_metadata_batch); | |
1f21ef0a FM |
4047 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
4048 | if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) { | |
4049 | btrfs_print_leaf(root, buf); | |
4050 | ASSERT(0); | |
4051 | } | |
4052 | #endif | |
eb60ceac CM |
4053 | } |
4054 | ||
b53d3f5d LB |
4055 | static void __btrfs_btree_balance_dirty(struct btrfs_root *root, |
4056 | int flush_delayed) | |
16cdcec7 MX |
4057 | { |
4058 | /* | |
4059 | * looks as though older kernels can get into trouble with | |
4060 | * this code, they end up stuck in balance_dirty_pages forever | |
4061 | */ | |
e2d84521 | 4062 | int ret; |
16cdcec7 MX |
4063 | |
4064 | if (current->flags & PF_MEMALLOC) | |
4065 | return; | |
4066 | ||
b53d3f5d LB |
4067 | if (flush_delayed) |
4068 | btrfs_balance_delayed_items(root); | |
16cdcec7 | 4069 | |
e2d84521 MX |
4070 | ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes, |
4071 | BTRFS_DIRTY_METADATA_THRESH); | |
4072 | if (ret > 0) { | |
d0e1d66b NJ |
4073 | balance_dirty_pages_ratelimited( |
4074 | root->fs_info->btree_inode->i_mapping); | |
16cdcec7 | 4075 | } |
16cdcec7 MX |
4076 | } |
4077 | ||
b53d3f5d | 4078 | void btrfs_btree_balance_dirty(struct btrfs_root *root) |
35b7e476 | 4079 | { |
b53d3f5d LB |
4080 | __btrfs_btree_balance_dirty(root, 1); |
4081 | } | |
585ad2c3 | 4082 | |
b53d3f5d LB |
4083 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root) |
4084 | { | |
4085 | __btrfs_btree_balance_dirty(root, 0); | |
35b7e476 | 4086 | } |
6b80053d | 4087 | |
ca7a79ad | 4088 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 4089 | { |
727011e0 | 4090 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
0b32f4bb | 4091 | return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
6b80053d | 4092 | } |
0da5468f | 4093 | |
fcd1f065 | 4094 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 4095 | int read_only) |
4096 | { | |
c926093e | 4097 | struct btrfs_super_block *sb = fs_info->super_copy; |
319e4d06 QW |
4098 | u64 nodesize = btrfs_super_nodesize(sb); |
4099 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
c926093e DS |
4100 | int ret = 0; |
4101 | ||
319e4d06 QW |
4102 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { |
4103 | printk(KERN_ERR "BTRFS: no valid FS found\n"); | |
4104 | ret = -EINVAL; | |
4105 | } | |
4106 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) | |
4107 | printk(KERN_WARNING "BTRFS: unrecognized super flag: %llu\n", | |
4108 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
21e7626b DS |
4109 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { |
4110 | printk(KERN_ERR "BTRFS: tree_root level too big: %d >= %d\n", | |
4111 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
c926093e DS |
4112 | ret = -EINVAL; |
4113 | } | |
21e7626b DS |
4114 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { |
4115 | printk(KERN_ERR "BTRFS: chunk_root level too big: %d >= %d\n", | |
4116 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
c926093e DS |
4117 | ret = -EINVAL; |
4118 | } | |
21e7626b DS |
4119 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { |
4120 | printk(KERN_ERR "BTRFS: log_root level too big: %d >= %d\n", | |
4121 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
c926093e DS |
4122 | ret = -EINVAL; |
4123 | } | |
4124 | ||
1104a885 | 4125 | /* |
319e4d06 QW |
4126 | * Check sectorsize and nodesize first, other check will need it. |
4127 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
1104a885 | 4128 | */ |
319e4d06 QW |
4129 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || |
4130 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
4131 | printk(KERN_ERR "BTRFS: invalid sectorsize %llu\n", sectorsize); | |
4132 | ret = -EINVAL; | |
4133 | } | |
4134 | /* Only PAGE SIZE is supported yet */ | |
09cbfeaf | 4135 | if (sectorsize != PAGE_SIZE) { |
319e4d06 | 4136 | printk(KERN_ERR "BTRFS: sectorsize %llu not supported yet, only support %lu\n", |
09cbfeaf | 4137 | sectorsize, PAGE_SIZE); |
319e4d06 QW |
4138 | ret = -EINVAL; |
4139 | } | |
4140 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || | |
4141 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
4142 | printk(KERN_ERR "BTRFS: invalid nodesize %llu\n", nodesize); | |
4143 | ret = -EINVAL; | |
4144 | } | |
4145 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
4146 | printk(KERN_ERR "BTRFS: invalid leafsize %u, should be %llu\n", | |
4147 | le32_to_cpu(sb->__unused_leafsize), | |
4148 | nodesize); | |
4149 | ret = -EINVAL; | |
4150 | } | |
4151 | ||
4152 | /* Root alignment check */ | |
4153 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
c926093e | 4154 | printk(KERN_WARNING "BTRFS: tree_root block unaligned: %llu\n", |
cd743fac | 4155 | btrfs_super_root(sb)); |
319e4d06 QW |
4156 | ret = -EINVAL; |
4157 | } | |
4158 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
cd743fac DS |
4159 | printk(KERN_WARNING "BTRFS: chunk_root block unaligned: %llu\n", |
4160 | btrfs_super_chunk_root(sb)); | |
75d6ad38 DS |
4161 | ret = -EINVAL; |
4162 | } | |
319e4d06 QW |
4163 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { |
4164 | printk(KERN_WARNING "BTRFS: log_root block unaligned: %llu\n", | |
4165 | btrfs_super_log_root(sb)); | |
75d6ad38 DS |
4166 | ret = -EINVAL; |
4167 | } | |
4168 | ||
c926093e DS |
4169 | if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) { |
4170 | printk(KERN_ERR "BTRFS: dev_item UUID does not match fsid: %pU != %pU\n", | |
4171 | fs_info->fsid, sb->dev_item.fsid); | |
4172 | ret = -EINVAL; | |
4173 | } | |
4174 | ||
4175 | /* | |
4176 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
4177 | * done later | |
4178 | */ | |
99e3ecfc LB |
4179 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { |
4180 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
4181 | btrfs_super_bytes_used(sb)); | |
4182 | ret = -EINVAL; | |
4183 | } | |
b7f67055 | 4184 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { |
99e3ecfc LB |
4185 | btrfs_err(fs_info, "invalid stripesize %u", |
4186 | btrfs_super_stripesize(sb)); | |
4187 | ret = -EINVAL; | |
4188 | } | |
21e7626b | 4189 | if (btrfs_super_num_devices(sb) > (1UL << 31)) |
c926093e | 4190 | printk(KERN_WARNING "BTRFS: suspicious number of devices: %llu\n", |
21e7626b | 4191 | btrfs_super_num_devices(sb)); |
75d6ad38 DS |
4192 | if (btrfs_super_num_devices(sb) == 0) { |
4193 | printk(KERN_ERR "BTRFS: number of devices is 0\n"); | |
4194 | ret = -EINVAL; | |
4195 | } | |
c926093e | 4196 | |
21e7626b | 4197 | if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) { |
c926093e | 4198 | printk(KERN_ERR "BTRFS: super offset mismatch %llu != %u\n", |
21e7626b | 4199 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); |
c926093e DS |
4200 | ret = -EINVAL; |
4201 | } | |
4202 | ||
ce7fca5f DS |
4203 | /* |
4204 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
4205 | * and one chunk | |
4206 | */ | |
4207 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
4208 | printk(KERN_ERR "BTRFS: system chunk array too big %u > %u\n", | |
4209 | btrfs_super_sys_array_size(sb), | |
4210 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
4211 | ret = -EINVAL; | |
4212 | } | |
4213 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
4214 | + sizeof(struct btrfs_chunk)) { | |
d2207129 | 4215 | printk(KERN_ERR "BTRFS: system chunk array too small %u < %zu\n", |
ce7fca5f DS |
4216 | btrfs_super_sys_array_size(sb), |
4217 | sizeof(struct btrfs_disk_key) | |
4218 | + sizeof(struct btrfs_chunk)); | |
4219 | ret = -EINVAL; | |
4220 | } | |
4221 | ||
c926093e DS |
4222 | /* |
4223 | * The generation is a global counter, we'll trust it more than the others | |
4224 | * but it's still possible that it's the one that's wrong. | |
4225 | */ | |
21e7626b | 4226 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) |
c926093e DS |
4227 | printk(KERN_WARNING |
4228 | "BTRFS: suspicious: generation < chunk_root_generation: %llu < %llu\n", | |
21e7626b DS |
4229 | btrfs_super_generation(sb), btrfs_super_chunk_root_generation(sb)); |
4230 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
4231 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
c926093e DS |
4232 | printk(KERN_WARNING |
4233 | "BTRFS: suspicious: generation < cache_generation: %llu < %llu\n", | |
21e7626b | 4234 | btrfs_super_generation(sb), btrfs_super_cache_generation(sb)); |
c926093e DS |
4235 | |
4236 | return ret; | |
acce952b | 4237 | } |
4238 | ||
48a3b636 | 4239 | static void btrfs_error_commit_super(struct btrfs_root *root) |
acce952b | 4240 | { |
acce952b | 4241 | mutex_lock(&root->fs_info->cleaner_mutex); |
4242 | btrfs_run_delayed_iputs(root); | |
4243 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
4244 | ||
4245 | down_write(&root->fs_info->cleanup_work_sem); | |
4246 | up_write(&root->fs_info->cleanup_work_sem); | |
4247 | ||
4248 | /* cleanup FS via transaction */ | |
4249 | btrfs_cleanup_transaction(root); | |
acce952b | 4250 | } |
4251 | ||
143bede5 | 4252 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4253 | { |
acce952b | 4254 | struct btrfs_ordered_extent *ordered; |
acce952b | 4255 | |
199c2a9c | 4256 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4257 | /* |
4258 | * This will just short circuit the ordered completion stuff which will | |
4259 | * make sure the ordered extent gets properly cleaned up. | |
4260 | */ | |
199c2a9c | 4261 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4262 | root_extent_list) |
4263 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4264 | spin_unlock(&root->ordered_extent_lock); |
4265 | } | |
4266 | ||
4267 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4268 | { | |
4269 | struct btrfs_root *root; | |
4270 | struct list_head splice; | |
4271 | ||
4272 | INIT_LIST_HEAD(&splice); | |
4273 | ||
4274 | spin_lock(&fs_info->ordered_root_lock); | |
4275 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4276 | while (!list_empty(&splice)) { | |
4277 | root = list_first_entry(&splice, struct btrfs_root, | |
4278 | ordered_root); | |
1de2cfde JB |
4279 | list_move_tail(&root->ordered_root, |
4280 | &fs_info->ordered_roots); | |
199c2a9c | 4281 | |
2a85d9ca | 4282 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4283 | btrfs_destroy_ordered_extents(root); |
4284 | ||
2a85d9ca LB |
4285 | cond_resched(); |
4286 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4287 | } |
4288 | spin_unlock(&fs_info->ordered_root_lock); | |
acce952b | 4289 | } |
4290 | ||
35a3621b SB |
4291 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
4292 | struct btrfs_root *root) | |
acce952b | 4293 | { |
4294 | struct rb_node *node; | |
4295 | struct btrfs_delayed_ref_root *delayed_refs; | |
4296 | struct btrfs_delayed_ref_node *ref; | |
4297 | int ret = 0; | |
4298 | ||
4299 | delayed_refs = &trans->delayed_refs; | |
4300 | ||
4301 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4302 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4303 | spin_unlock(&delayed_refs->lock); |
efe120a0 | 4304 | btrfs_info(root->fs_info, "delayed_refs has NO entry"); |
acce952b | 4305 | return ret; |
4306 | } | |
4307 | ||
d7df2c79 JB |
4308 | while ((node = rb_first(&delayed_refs->href_root)) != NULL) { |
4309 | struct btrfs_delayed_ref_head *head; | |
c6fc2454 | 4310 | struct btrfs_delayed_ref_node *tmp; |
e78417d1 | 4311 | bool pin_bytes = false; |
acce952b | 4312 | |
d7df2c79 JB |
4313 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4314 | href_node); | |
4315 | if (!mutex_trylock(&head->mutex)) { | |
4316 | atomic_inc(&head->node.refs); | |
4317 | spin_unlock(&delayed_refs->lock); | |
eb12db69 | 4318 | |
d7df2c79 | 4319 | mutex_lock(&head->mutex); |
e78417d1 | 4320 | mutex_unlock(&head->mutex); |
d7df2c79 JB |
4321 | btrfs_put_delayed_ref(&head->node); |
4322 | spin_lock(&delayed_refs->lock); | |
4323 | continue; | |
4324 | } | |
4325 | spin_lock(&head->lock); | |
c6fc2454 QW |
4326 | list_for_each_entry_safe_reverse(ref, tmp, &head->ref_list, |
4327 | list) { | |
d7df2c79 | 4328 | ref->in_tree = 0; |
c6fc2454 | 4329 | list_del(&ref->list); |
d7df2c79 JB |
4330 | atomic_dec(&delayed_refs->num_entries); |
4331 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 4332 | } |
d7df2c79 JB |
4333 | if (head->must_insert_reserved) |
4334 | pin_bytes = true; | |
4335 | btrfs_free_delayed_extent_op(head->extent_op); | |
4336 | delayed_refs->num_heads--; | |
4337 | if (head->processing == 0) | |
4338 | delayed_refs->num_heads_ready--; | |
4339 | atomic_dec(&delayed_refs->num_entries); | |
4340 | head->node.in_tree = 0; | |
4341 | rb_erase(&head->href_node, &delayed_refs->href_root); | |
4342 | spin_unlock(&head->lock); | |
4343 | spin_unlock(&delayed_refs->lock); | |
4344 | mutex_unlock(&head->mutex); | |
acce952b | 4345 | |
d7df2c79 JB |
4346 | if (pin_bytes) |
4347 | btrfs_pin_extent(root, head->node.bytenr, | |
4348 | head->node.num_bytes, 1); | |
4349 | btrfs_put_delayed_ref(&head->node); | |
acce952b | 4350 | cond_resched(); |
4351 | spin_lock(&delayed_refs->lock); | |
4352 | } | |
4353 | ||
4354 | spin_unlock(&delayed_refs->lock); | |
4355 | ||
4356 | return ret; | |
4357 | } | |
4358 | ||
143bede5 | 4359 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4360 | { |
4361 | struct btrfs_inode *btrfs_inode; | |
4362 | struct list_head splice; | |
4363 | ||
4364 | INIT_LIST_HEAD(&splice); | |
4365 | ||
eb73c1b7 MX |
4366 | spin_lock(&root->delalloc_lock); |
4367 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4368 | |
4369 | while (!list_empty(&splice)) { | |
eb73c1b7 MX |
4370 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4371 | delalloc_inodes); | |
acce952b | 4372 | |
4373 | list_del_init(&btrfs_inode->delalloc_inodes); | |
df0af1a5 MX |
4374 | clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, |
4375 | &btrfs_inode->runtime_flags); | |
eb73c1b7 | 4376 | spin_unlock(&root->delalloc_lock); |
acce952b | 4377 | |
4378 | btrfs_invalidate_inodes(btrfs_inode->root); | |
b216cbfb | 4379 | |
eb73c1b7 | 4380 | spin_lock(&root->delalloc_lock); |
acce952b | 4381 | } |
4382 | ||
eb73c1b7 MX |
4383 | spin_unlock(&root->delalloc_lock); |
4384 | } | |
4385 | ||
4386 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4387 | { | |
4388 | struct btrfs_root *root; | |
4389 | struct list_head splice; | |
4390 | ||
4391 | INIT_LIST_HEAD(&splice); | |
4392 | ||
4393 | spin_lock(&fs_info->delalloc_root_lock); | |
4394 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4395 | while (!list_empty(&splice)) { | |
4396 | root = list_first_entry(&splice, struct btrfs_root, | |
4397 | delalloc_root); | |
4398 | list_del_init(&root->delalloc_root); | |
4399 | root = btrfs_grab_fs_root(root); | |
4400 | BUG_ON(!root); | |
4401 | spin_unlock(&fs_info->delalloc_root_lock); | |
4402 | ||
4403 | btrfs_destroy_delalloc_inodes(root); | |
4404 | btrfs_put_fs_root(root); | |
4405 | ||
4406 | spin_lock(&fs_info->delalloc_root_lock); | |
4407 | } | |
4408 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4409 | } |
4410 | ||
4411 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, | |
4412 | struct extent_io_tree *dirty_pages, | |
4413 | int mark) | |
4414 | { | |
4415 | int ret; | |
acce952b | 4416 | struct extent_buffer *eb; |
4417 | u64 start = 0; | |
4418 | u64 end; | |
acce952b | 4419 | |
4420 | while (1) { | |
4421 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 4422 | mark, NULL); |
acce952b | 4423 | if (ret) |
4424 | break; | |
4425 | ||
91166212 | 4426 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 4427 | while (start <= end) { |
01d58472 | 4428 | eb = btrfs_find_tree_block(root->fs_info, start); |
707e8a07 | 4429 | start += root->nodesize; |
fd8b2b61 | 4430 | if (!eb) |
acce952b | 4431 | continue; |
fd8b2b61 | 4432 | wait_on_extent_buffer_writeback(eb); |
acce952b | 4433 | |
fd8b2b61 JB |
4434 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
4435 | &eb->bflags)) | |
4436 | clear_extent_buffer_dirty(eb); | |
4437 | free_extent_buffer_stale(eb); | |
acce952b | 4438 | } |
4439 | } | |
4440 | ||
4441 | return ret; | |
4442 | } | |
4443 | ||
4444 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
4445 | struct extent_io_tree *pinned_extents) | |
4446 | { | |
4447 | struct extent_io_tree *unpin; | |
4448 | u64 start; | |
4449 | u64 end; | |
4450 | int ret; | |
ed0eaa14 | 4451 | bool loop = true; |
acce952b | 4452 | |
4453 | unpin = pinned_extents; | |
ed0eaa14 | 4454 | again: |
acce952b | 4455 | while (1) { |
4456 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
e6138876 | 4457 | EXTENT_DIRTY, NULL); |
acce952b | 4458 | if (ret) |
4459 | break; | |
4460 | ||
af6f8f60 | 4461 | clear_extent_dirty(unpin, start, end); |
acce952b | 4462 | btrfs_error_unpin_extent_range(root, start, end); |
4463 | cond_resched(); | |
4464 | } | |
4465 | ||
ed0eaa14 LB |
4466 | if (loop) { |
4467 | if (unpin == &root->fs_info->freed_extents[0]) | |
4468 | unpin = &root->fs_info->freed_extents[1]; | |
4469 | else | |
4470 | unpin = &root->fs_info->freed_extents[0]; | |
4471 | loop = false; | |
4472 | goto again; | |
4473 | } | |
4474 | ||
acce952b | 4475 | return 0; |
4476 | } | |
4477 | ||
49b25e05 JM |
4478 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
4479 | struct btrfs_root *root) | |
4480 | { | |
4481 | btrfs_destroy_delayed_refs(cur_trans, root); | |
49b25e05 | 4482 | |
4a9d8bde | 4483 | cur_trans->state = TRANS_STATE_COMMIT_START; |
d7096fc3 | 4484 | wake_up(&root->fs_info->transaction_blocked_wait); |
49b25e05 | 4485 | |
4a9d8bde | 4486 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
d7096fc3 | 4487 | wake_up(&root->fs_info->transaction_wait); |
49b25e05 | 4488 | |
67cde344 MX |
4489 | btrfs_destroy_delayed_inodes(root); |
4490 | btrfs_assert_delayed_root_empty(root); | |
49b25e05 | 4491 | |
49b25e05 JM |
4492 | btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, |
4493 | EXTENT_DIRTY); | |
6e841e32 LB |
4494 | btrfs_destroy_pinned_extent(root, |
4495 | root->fs_info->pinned_extents); | |
49b25e05 | 4496 | |
4a9d8bde MX |
4497 | cur_trans->state =TRANS_STATE_COMPLETED; |
4498 | wake_up(&cur_trans->commit_wait); | |
4499 | ||
49b25e05 JM |
4500 | /* |
4501 | memset(cur_trans, 0, sizeof(*cur_trans)); | |
4502 | kmem_cache_free(btrfs_transaction_cachep, cur_trans); | |
4503 | */ | |
4504 | } | |
4505 | ||
48a3b636 | 4506 | static int btrfs_cleanup_transaction(struct btrfs_root *root) |
acce952b | 4507 | { |
4508 | struct btrfs_transaction *t; | |
acce952b | 4509 | |
acce952b | 4510 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
4511 | ||
a4abeea4 | 4512 | spin_lock(&root->fs_info->trans_lock); |
724e2315 JB |
4513 | while (!list_empty(&root->fs_info->trans_list)) { |
4514 | t = list_first_entry(&root->fs_info->trans_list, | |
4515 | struct btrfs_transaction, list); | |
4516 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
4517 | atomic_inc(&t->use_count); | |
4518 | spin_unlock(&root->fs_info->trans_lock); | |
4519 | btrfs_wait_for_commit(root, t->transid); | |
4520 | btrfs_put_transaction(t); | |
4521 | spin_lock(&root->fs_info->trans_lock); | |
4522 | continue; | |
4523 | } | |
4524 | if (t == root->fs_info->running_transaction) { | |
4525 | t->state = TRANS_STATE_COMMIT_DOING; | |
4526 | spin_unlock(&root->fs_info->trans_lock); | |
4527 | /* | |
4528 | * We wait for 0 num_writers since we don't hold a trans | |
4529 | * handle open currently for this transaction. | |
4530 | */ | |
4531 | wait_event(t->writer_wait, | |
4532 | atomic_read(&t->num_writers) == 0); | |
4533 | } else { | |
4534 | spin_unlock(&root->fs_info->trans_lock); | |
4535 | } | |
4536 | btrfs_cleanup_one_transaction(t, root); | |
4a9d8bde | 4537 | |
724e2315 JB |
4538 | spin_lock(&root->fs_info->trans_lock); |
4539 | if (t == root->fs_info->running_transaction) | |
4540 | root->fs_info->running_transaction = NULL; | |
acce952b | 4541 | list_del_init(&t->list); |
724e2315 | 4542 | spin_unlock(&root->fs_info->trans_lock); |
acce952b | 4543 | |
724e2315 JB |
4544 | btrfs_put_transaction(t); |
4545 | trace_btrfs_transaction_commit(root); | |
4546 | spin_lock(&root->fs_info->trans_lock); | |
4547 | } | |
4548 | spin_unlock(&root->fs_info->trans_lock); | |
4549 | btrfs_destroy_all_ordered_extents(root->fs_info); | |
4550 | btrfs_destroy_delayed_inodes(root); | |
4551 | btrfs_assert_delayed_root_empty(root); | |
4552 | btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents); | |
4553 | btrfs_destroy_all_delalloc_inodes(root->fs_info); | |
acce952b | 4554 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); |
acce952b | 4555 | |
4556 | return 0; | |
4557 | } | |
4558 | ||
e8c9f186 | 4559 | static const struct extent_io_ops btree_extent_io_ops = { |
ce9adaa5 | 4560 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
4bb31e92 | 4561 | .readpage_io_failed_hook = btree_io_failed_hook, |
0b86a832 | 4562 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
4563 | /* note we're sharing with inode.c for the merge bio hook */ |
4564 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 4565 | }; |