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 | ||
a74a4b97 | 19 | #include <linux/version.h> |
e20d96d6 | 20 | #include <linux/fs.h> |
d98237b3 | 21 | #include <linux/blkdev.h> |
87cbda5c | 22 | #include <linux/scatterlist.h> |
22b0ebda | 23 | #include <linux/swap.h> |
0f7d52f4 | 24 | #include <linux/radix-tree.h> |
35b7e476 | 25 | #include <linux/writeback.h> |
d397712b | 26 | #include <linux/buffer_head.h> |
ce9adaa5 | 27 | #include <linux/workqueue.h> |
a74a4b97 | 28 | #include <linux/kthread.h> |
4b4e25f2 CM |
29 | #include <linux/freezer.h> |
30 | #include "compat.h" | |
a5eb62e3 | 31 | #include "crc32c.h" |
eb60ceac CM |
32 | #include "ctree.h" |
33 | #include "disk-io.h" | |
e089f05c | 34 | #include "transaction.h" |
0f7d52f4 | 35 | #include "btrfs_inode.h" |
0b86a832 | 36 | #include "volumes.h" |
db94535d | 37 | #include "print-tree.h" |
8b712842 | 38 | #include "async-thread.h" |
925baedd | 39 | #include "locking.h" |
017e5369 | 40 | #include "ref-cache.h" |
e02119d5 | 41 | #include "tree-log.h" |
eb60ceac | 42 | |
d1310b2e | 43 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 44 | static void end_workqueue_fn(struct btrfs_work *work); |
ce9adaa5 | 45 | |
d352ac68 CM |
46 | /* |
47 | * end_io_wq structs are used to do processing in task context when an IO is | |
48 | * complete. This is used during reads to verify checksums, and it is used | |
49 | * by writes to insert metadata for new file extents after IO is complete. | |
50 | */ | |
ce9adaa5 CM |
51 | struct end_io_wq { |
52 | struct bio *bio; | |
53 | bio_end_io_t *end_io; | |
54 | void *private; | |
55 | struct btrfs_fs_info *info; | |
56 | int error; | |
22c59948 | 57 | int metadata; |
ce9adaa5 | 58 | struct list_head list; |
8b712842 | 59 | struct btrfs_work work; |
ce9adaa5 | 60 | }; |
0da5468f | 61 | |
d352ac68 CM |
62 | /* |
63 | * async submit bios are used to offload expensive checksumming | |
64 | * onto the worker threads. They checksum file and metadata bios | |
65 | * just before they are sent down the IO stack. | |
66 | */ | |
44b8bd7e CM |
67 | struct async_submit_bio { |
68 | struct inode *inode; | |
69 | struct bio *bio; | |
70 | struct list_head list; | |
4a69a410 CM |
71 | extent_submit_bio_hook_t *submit_bio_start; |
72 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
73 | int rw; |
74 | int mirror_num; | |
c8b97818 | 75 | unsigned long bio_flags; |
8b712842 | 76 | struct btrfs_work work; |
44b8bd7e CM |
77 | }; |
78 | ||
d352ac68 CM |
79 | /* |
80 | * extents on the btree inode are pretty simple, there's one extent | |
81 | * that covers the entire device | |
82 | */ | |
b2950863 CH |
83 | static struct extent_map *btree_get_extent(struct inode *inode, |
84 | struct page *page, size_t page_offset, u64 start, u64 len, | |
85 | int create) | |
7eccb903 | 86 | { |
5f39d397 CM |
87 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
88 | struct extent_map *em; | |
89 | int ret; | |
90 | ||
d1310b2e CM |
91 | spin_lock(&em_tree->lock); |
92 | em = lookup_extent_mapping(em_tree, start, len); | |
a061fc8d CM |
93 | if (em) { |
94 | em->bdev = | |
95 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
96 | spin_unlock(&em_tree->lock); | |
5f39d397 | 97 | goto out; |
a061fc8d CM |
98 | } |
99 | spin_unlock(&em_tree->lock); | |
7b13b7b1 | 100 | |
5f39d397 CM |
101 | em = alloc_extent_map(GFP_NOFS); |
102 | if (!em) { | |
103 | em = ERR_PTR(-ENOMEM); | |
104 | goto out; | |
105 | } | |
106 | em->start = 0; | |
0afbaf8c | 107 | em->len = (u64)-1; |
c8b97818 | 108 | em->block_len = (u64)-1; |
5f39d397 | 109 | em->block_start = 0; |
a061fc8d | 110 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e CM |
111 | |
112 | spin_lock(&em_tree->lock); | |
5f39d397 CM |
113 | ret = add_extent_mapping(em_tree, em); |
114 | if (ret == -EEXIST) { | |
0afbaf8c CM |
115 | u64 failed_start = em->start; |
116 | u64 failed_len = em->len; | |
117 | ||
5f39d397 | 118 | free_extent_map(em); |
7b13b7b1 | 119 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c | 120 | if (em) { |
7b13b7b1 | 121 | ret = 0; |
0afbaf8c CM |
122 | } else { |
123 | em = lookup_extent_mapping(em_tree, failed_start, | |
124 | failed_len); | |
7b13b7b1 | 125 | ret = -EIO; |
0afbaf8c | 126 | } |
5f39d397 | 127 | } else if (ret) { |
7b13b7b1 CM |
128 | free_extent_map(em); |
129 | em = NULL; | |
5f39d397 | 130 | } |
7b13b7b1 CM |
131 | spin_unlock(&em_tree->lock); |
132 | ||
133 | if (ret) | |
134 | em = ERR_PTR(ret); | |
5f39d397 CM |
135 | out: |
136 | return em; | |
7eccb903 CM |
137 | } |
138 | ||
19c00ddc CM |
139 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
140 | { | |
a5eb62e3 | 141 | return btrfs_crc32c(seed, data, len); |
19c00ddc CM |
142 | } |
143 | ||
144 | void btrfs_csum_final(u32 crc, char *result) | |
145 | { | |
146 | *(__le32 *)result = ~cpu_to_le32(crc); | |
147 | } | |
148 | ||
d352ac68 CM |
149 | /* |
150 | * compute the csum for a btree block, and either verify it or write it | |
151 | * into the csum field of the block. | |
152 | */ | |
19c00ddc CM |
153 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
154 | int verify) | |
155 | { | |
607d432d JB |
156 | u16 csum_size = |
157 | btrfs_super_csum_size(&root->fs_info->super_copy); | |
158 | char *result = NULL; | |
19c00ddc CM |
159 | unsigned long len; |
160 | unsigned long cur_len; | |
161 | unsigned long offset = BTRFS_CSUM_SIZE; | |
162 | char *map_token = NULL; | |
163 | char *kaddr; | |
164 | unsigned long map_start; | |
165 | unsigned long map_len; | |
166 | int err; | |
167 | u32 crc = ~(u32)0; | |
607d432d | 168 | unsigned long inline_result; |
19c00ddc CM |
169 | |
170 | len = buf->len - offset; | |
d397712b | 171 | while (len > 0) { |
19c00ddc CM |
172 | err = map_private_extent_buffer(buf, offset, 32, |
173 | &map_token, &kaddr, | |
174 | &map_start, &map_len, KM_USER0); | |
d397712b | 175 | if (err) |
19c00ddc | 176 | return 1; |
19c00ddc CM |
177 | cur_len = min(len, map_len - (offset - map_start)); |
178 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
179 | crc, cur_len); | |
180 | len -= cur_len; | |
181 | offset += cur_len; | |
182 | unmap_extent_buffer(buf, map_token, KM_USER0); | |
183 | } | |
607d432d JB |
184 | if (csum_size > sizeof(inline_result)) { |
185 | result = kzalloc(csum_size * sizeof(char), GFP_NOFS); | |
186 | if (!result) | |
187 | return 1; | |
188 | } else { | |
189 | result = (char *)&inline_result; | |
190 | } | |
191 | ||
19c00ddc CM |
192 | btrfs_csum_final(crc, result); |
193 | ||
194 | if (verify) { | |
607d432d | 195 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
196 | u32 val; |
197 | u32 found = 0; | |
607d432d | 198 | memcpy(&found, result, csum_size); |
e4204ded | 199 | |
607d432d | 200 | read_extent_buffer(buf, &val, 0, csum_size); |
d397712b CM |
201 | printk(KERN_INFO "btrfs: %s checksum verify failed " |
202 | "on %llu wanted %X found %X level %d\n", | |
19c00ddc | 203 | root->fs_info->sb->s_id, |
2dd3e67b | 204 | buf->start, val, found, btrfs_header_level(buf)); |
607d432d JB |
205 | if (result != (char *)&inline_result) |
206 | kfree(result); | |
19c00ddc CM |
207 | return 1; |
208 | } | |
209 | } else { | |
607d432d | 210 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 211 | } |
607d432d JB |
212 | if (result != (char *)&inline_result) |
213 | kfree(result); | |
19c00ddc CM |
214 | return 0; |
215 | } | |
216 | ||
d352ac68 CM |
217 | /* |
218 | * we can't consider a given block up to date unless the transid of the | |
219 | * block matches the transid in the parent node's pointer. This is how we | |
220 | * detect blocks that either didn't get written at all or got written | |
221 | * in the wrong place. | |
222 | */ | |
1259ab75 CM |
223 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
224 | struct extent_buffer *eb, u64 parent_transid) | |
225 | { | |
226 | int ret; | |
227 | ||
228 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
229 | return 0; | |
230 | ||
231 | lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); | |
232 | if (extent_buffer_uptodate(io_tree, eb) && | |
233 | btrfs_header_generation(eb) == parent_transid) { | |
234 | ret = 0; | |
235 | goto out; | |
236 | } | |
237 | printk("parent transid verify failed on %llu wanted %llu found %llu\n", | |
238 | (unsigned long long)eb->start, | |
239 | (unsigned long long)parent_transid, | |
240 | (unsigned long long)btrfs_header_generation(eb)); | |
241 | ret = 1; | |
1259ab75 | 242 | clear_extent_buffer_uptodate(io_tree, eb); |
33958dc6 | 243 | out: |
1259ab75 CM |
244 | unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, |
245 | GFP_NOFS); | |
246 | return ret; | |
1259ab75 CM |
247 | } |
248 | ||
d352ac68 CM |
249 | /* |
250 | * helper to read a given tree block, doing retries as required when | |
251 | * the checksums don't match and we have alternate mirrors to try. | |
252 | */ | |
f188591e CM |
253 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
254 | struct extent_buffer *eb, | |
ca7a79ad | 255 | u64 start, u64 parent_transid) |
f188591e CM |
256 | { |
257 | struct extent_io_tree *io_tree; | |
258 | int ret; | |
259 | int num_copies = 0; | |
260 | int mirror_num = 0; | |
261 | ||
262 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
263 | while (1) { | |
264 | ret = read_extent_buffer_pages(io_tree, eb, start, 1, | |
265 | btree_get_extent, mirror_num); | |
1259ab75 CM |
266 | if (!ret && |
267 | !verify_parent_transid(io_tree, eb, parent_transid)) | |
f188591e | 268 | return ret; |
d397712b | 269 | |
f188591e CM |
270 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
271 | eb->start, eb->len); | |
4235298e | 272 | if (num_copies == 1) |
f188591e | 273 | return ret; |
4235298e | 274 | |
f188591e | 275 | mirror_num++; |
4235298e | 276 | if (mirror_num > num_copies) |
f188591e | 277 | return ret; |
f188591e | 278 | } |
f188591e CM |
279 | return -EIO; |
280 | } | |
19c00ddc | 281 | |
d352ac68 | 282 | /* |
d397712b CM |
283 | * checksum a dirty tree block before IO. This has extra checks to make sure |
284 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 285 | */ |
d397712b | 286 | |
b2950863 | 287 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 288 | { |
d1310b2e | 289 | struct extent_io_tree *tree; |
35ebb934 | 290 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc CM |
291 | u64 found_start; |
292 | int found_level; | |
293 | unsigned long len; | |
294 | struct extent_buffer *eb; | |
f188591e CM |
295 | int ret; |
296 | ||
d1310b2e | 297 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc CM |
298 | |
299 | if (page->private == EXTENT_PAGE_PRIVATE) | |
300 | goto out; | |
301 | if (!page->private) | |
302 | goto out; | |
303 | len = page->private >> 2; | |
d397712b CM |
304 | WARN_ON(len == 0); |
305 | ||
19c00ddc | 306 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); |
ca7a79ad CM |
307 | ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, |
308 | btrfs_header_generation(eb)); | |
f188591e | 309 | BUG_ON(ret); |
19c00ddc CM |
310 | found_start = btrfs_header_bytenr(eb); |
311 | if (found_start != start) { | |
55c69072 CM |
312 | WARN_ON(1); |
313 | goto err; | |
314 | } | |
315 | if (eb->first_page != page) { | |
55c69072 CM |
316 | WARN_ON(1); |
317 | goto err; | |
318 | } | |
319 | if (!PageUptodate(page)) { | |
55c69072 CM |
320 | WARN_ON(1); |
321 | goto err; | |
19c00ddc CM |
322 | } |
323 | found_level = btrfs_header_level(eb); | |
4bef0848 | 324 | |
19c00ddc | 325 | csum_tree_block(root, eb, 0); |
55c69072 | 326 | err: |
19c00ddc CM |
327 | free_extent_buffer(eb); |
328 | out: | |
329 | return 0; | |
330 | } | |
331 | ||
2b82032c YZ |
332 | static int check_tree_block_fsid(struct btrfs_root *root, |
333 | struct extent_buffer *eb) | |
334 | { | |
335 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
336 | u8 fsid[BTRFS_UUID_SIZE]; | |
337 | int ret = 1; | |
338 | ||
339 | read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), | |
340 | BTRFS_FSID_SIZE); | |
341 | while (fs_devices) { | |
342 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
343 | ret = 0; | |
344 | break; | |
345 | } | |
346 | fs_devices = fs_devices->seed; | |
347 | } | |
348 | return ret; | |
349 | } | |
350 | ||
b2950863 | 351 | static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
ce9adaa5 CM |
352 | struct extent_state *state) |
353 | { | |
354 | struct extent_io_tree *tree; | |
355 | u64 found_start; | |
356 | int found_level; | |
357 | unsigned long len; | |
358 | struct extent_buffer *eb; | |
359 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 360 | int ret = 0; |
ce9adaa5 CM |
361 | |
362 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
363 | if (page->private == EXTENT_PAGE_PRIVATE) | |
364 | goto out; | |
365 | if (!page->private) | |
366 | goto out; | |
d397712b | 367 | |
ce9adaa5 | 368 | len = page->private >> 2; |
d397712b CM |
369 | WARN_ON(len == 0); |
370 | ||
ce9adaa5 | 371 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); |
f188591e | 372 | |
ce9adaa5 | 373 | found_start = btrfs_header_bytenr(eb); |
23a07867 | 374 | if (found_start != start) { |
d397712b | 375 | printk(KERN_INFO "btrfs bad tree block start %llu %llu\n", |
a1b32a59 CM |
376 | (unsigned long long)found_start, |
377 | (unsigned long long)eb->start); | |
f188591e | 378 | ret = -EIO; |
ce9adaa5 CM |
379 | goto err; |
380 | } | |
381 | if (eb->first_page != page) { | |
d397712b CM |
382 | printk(KERN_INFO "btrfs bad first page %lu %lu\n", |
383 | eb->first_page->index, page->index); | |
ce9adaa5 | 384 | WARN_ON(1); |
f188591e | 385 | ret = -EIO; |
ce9adaa5 CM |
386 | goto err; |
387 | } | |
2b82032c | 388 | if (check_tree_block_fsid(root, eb)) { |
d397712b CM |
389 | printk(KERN_INFO "btrfs bad fsid on block %llu\n", |
390 | (unsigned long long)eb->start); | |
1259ab75 CM |
391 | ret = -EIO; |
392 | goto err; | |
393 | } | |
ce9adaa5 CM |
394 | found_level = btrfs_header_level(eb); |
395 | ||
396 | ret = csum_tree_block(root, eb, 1); | |
f188591e CM |
397 | if (ret) |
398 | ret = -EIO; | |
ce9adaa5 CM |
399 | |
400 | end = min_t(u64, eb->len, PAGE_CACHE_SIZE); | |
401 | end = eb->start + end - 1; | |
ce9adaa5 CM |
402 | err: |
403 | free_extent_buffer(eb); | |
404 | out: | |
f188591e | 405 | return ret; |
ce9adaa5 CM |
406 | } |
407 | ||
ce9adaa5 | 408 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
409 | { |
410 | struct end_io_wq *end_io_wq = bio->bi_private; | |
411 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 412 | |
ce9adaa5 | 413 | fs_info = end_io_wq->info; |
ce9adaa5 | 414 | end_io_wq->error = err; |
8b712842 CM |
415 | end_io_wq->work.func = end_workqueue_fn; |
416 | end_io_wq->work.flags = 0; | |
d20f7043 CM |
417 | |
418 | if (bio->bi_rw & (1 << BIO_RW)) { | |
cad321ad CM |
419 | if (end_io_wq->metadata) |
420 | btrfs_queue_worker(&fs_info->endio_meta_write_workers, | |
421 | &end_io_wq->work); | |
422 | else | |
423 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
424 | &end_io_wq->work); | |
d20f7043 CM |
425 | } else { |
426 | if (end_io_wq->metadata) | |
427 | btrfs_queue_worker(&fs_info->endio_meta_workers, | |
428 | &end_io_wq->work); | |
429 | else | |
430 | btrfs_queue_worker(&fs_info->endio_workers, | |
431 | &end_io_wq->work); | |
432 | } | |
ce9adaa5 CM |
433 | } |
434 | ||
22c59948 CM |
435 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
436 | int metadata) | |
0b86a832 | 437 | { |
ce9adaa5 | 438 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
439 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
440 | if (!end_io_wq) | |
441 | return -ENOMEM; | |
442 | ||
443 | end_io_wq->private = bio->bi_private; | |
444 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 445 | end_io_wq->info = info; |
ce9adaa5 CM |
446 | end_io_wq->error = 0; |
447 | end_io_wq->bio = bio; | |
22c59948 | 448 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
449 | |
450 | bio->bi_private = end_io_wq; | |
451 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
452 | return 0; |
453 | } | |
454 | ||
b64a2851 | 455 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 456 | { |
4854ddd0 CM |
457 | unsigned long limit = min_t(unsigned long, |
458 | info->workers.max_workers, | |
459 | info->fs_devices->open_devices); | |
460 | return 256 * limit; | |
461 | } | |
0986fe9e | 462 | |
4854ddd0 CM |
463 | int btrfs_congested_async(struct btrfs_fs_info *info, int iodone) |
464 | { | |
b64a2851 CM |
465 | return atomic_read(&info->nr_async_bios) > |
466 | btrfs_async_submit_limit(info); | |
0986fe9e CM |
467 | } |
468 | ||
4a69a410 CM |
469 | static void run_one_async_start(struct btrfs_work *work) |
470 | { | |
471 | struct btrfs_fs_info *fs_info; | |
472 | struct async_submit_bio *async; | |
473 | ||
474 | async = container_of(work, struct async_submit_bio, work); | |
475 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
476 | async->submit_bio_start(async->inode, async->rw, async->bio, | |
477 | async->mirror_num, async->bio_flags); | |
478 | } | |
479 | ||
480 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
481 | { |
482 | struct btrfs_fs_info *fs_info; | |
483 | struct async_submit_bio *async; | |
4854ddd0 | 484 | int limit; |
8b712842 CM |
485 | |
486 | async = container_of(work, struct async_submit_bio, work); | |
487 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 488 | |
b64a2851 | 489 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
490 | limit = limit * 2 / 3; |
491 | ||
8b712842 | 492 | atomic_dec(&fs_info->nr_async_submits); |
0986fe9e | 493 | |
b64a2851 CM |
494 | if (atomic_read(&fs_info->nr_async_submits) < limit && |
495 | waitqueue_active(&fs_info->async_submit_wait)) | |
4854ddd0 CM |
496 | wake_up(&fs_info->async_submit_wait); |
497 | ||
4a69a410 | 498 | async->submit_bio_done(async->inode, async->rw, async->bio, |
c8b97818 | 499 | async->mirror_num, async->bio_flags); |
4a69a410 CM |
500 | } |
501 | ||
502 | static void run_one_async_free(struct btrfs_work *work) | |
503 | { | |
504 | struct async_submit_bio *async; | |
505 | ||
506 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
507 | kfree(async); |
508 | } | |
509 | ||
44b8bd7e CM |
510 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
511 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 512 | unsigned long bio_flags, |
4a69a410 CM |
513 | extent_submit_bio_hook_t *submit_bio_start, |
514 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
515 | { |
516 | struct async_submit_bio *async; | |
517 | ||
518 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
519 | if (!async) | |
520 | return -ENOMEM; | |
521 | ||
522 | async->inode = inode; | |
523 | async->rw = rw; | |
524 | async->bio = bio; | |
525 | async->mirror_num = mirror_num; | |
4a69a410 CM |
526 | async->submit_bio_start = submit_bio_start; |
527 | async->submit_bio_done = submit_bio_done; | |
528 | ||
529 | async->work.func = run_one_async_start; | |
530 | async->work.ordered_func = run_one_async_done; | |
531 | async->work.ordered_free = run_one_async_free; | |
532 | ||
8b712842 | 533 | async->work.flags = 0; |
c8b97818 | 534 | async->bio_flags = bio_flags; |
8c8bee1d | 535 | |
cb03c743 | 536 | atomic_inc(&fs_info->nr_async_submits); |
8b712842 | 537 | btrfs_queue_worker(&fs_info->workers, &async->work); |
d2c3f4f6 | 538 | #if 0 |
7c2ca468 | 539 | int limit = btrfs_async_submit_limit(fs_info); |
9473f16c CM |
540 | if (atomic_read(&fs_info->nr_async_submits) > limit) { |
541 | wait_event_timeout(fs_info->async_submit_wait, | |
4854ddd0 CM |
542 | (atomic_read(&fs_info->nr_async_submits) < limit), |
543 | HZ/10); | |
9473f16c CM |
544 | |
545 | wait_event_timeout(fs_info->async_submit_wait, | |
546 | (atomic_read(&fs_info->nr_async_bios) < limit), | |
547 | HZ/10); | |
548 | } | |
d2c3f4f6 | 549 | #endif |
d397712b | 550 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
551 | atomic_read(&fs_info->nr_async_submits)) { |
552 | wait_event(fs_info->async_submit_wait, | |
553 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
554 | } | |
555 | ||
44b8bd7e CM |
556 | return 0; |
557 | } | |
558 | ||
ce3ed71a CM |
559 | static int btree_csum_one_bio(struct bio *bio) |
560 | { | |
561 | struct bio_vec *bvec = bio->bi_io_vec; | |
562 | int bio_index = 0; | |
563 | struct btrfs_root *root; | |
564 | ||
565 | WARN_ON(bio->bi_vcnt <= 0); | |
d397712b | 566 | while (bio_index < bio->bi_vcnt) { |
ce3ed71a CM |
567 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
568 | csum_dirty_buffer(root, bvec->bv_page); | |
569 | bio_index++; | |
570 | bvec++; | |
571 | } | |
572 | return 0; | |
573 | } | |
574 | ||
4a69a410 CM |
575 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
576 | struct bio *bio, int mirror_num, | |
577 | unsigned long bio_flags) | |
22c59948 | 578 | { |
8b712842 CM |
579 | /* |
580 | * when we're called for a write, we're already in the async | |
5443be45 | 581 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 582 | */ |
4a69a410 CM |
583 | btree_csum_one_bio(bio); |
584 | return 0; | |
585 | } | |
22c59948 | 586 | |
4a69a410 CM |
587 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
588 | int mirror_num, unsigned long bio_flags) | |
589 | { | |
8b712842 | 590 | /* |
4a69a410 CM |
591 | * when we're called for a write, we're already in the async |
592 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 593 | */ |
8b712842 | 594 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
595 | } |
596 | ||
44b8bd7e | 597 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
c8b97818 | 598 | int mirror_num, unsigned long bio_flags) |
44b8bd7e | 599 | { |
cad321ad CM |
600 | int ret; |
601 | ||
602 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, | |
603 | bio, 1); | |
604 | BUG_ON(ret); | |
605 | ||
44b8bd7e | 606 | if (!(rw & (1 << BIO_RW))) { |
4a69a410 CM |
607 | /* |
608 | * called for a read, do the setup so that checksum validation | |
609 | * can happen in the async kernel threads | |
610 | */ | |
4a69a410 | 611 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, |
6f3577bd | 612 | mirror_num, 0); |
44b8bd7e | 613 | } |
cad321ad CM |
614 | /* |
615 | * kthread helpers are used to submit writes so that checksumming | |
616 | * can happen in parallel across all CPUs | |
617 | */ | |
44b8bd7e | 618 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, |
c8b97818 | 619 | inode, rw, bio, mirror_num, 0, |
4a69a410 CM |
620 | __btree_submit_bio_start, |
621 | __btree_submit_bio_done); | |
44b8bd7e CM |
622 | } |
623 | ||
0da5468f CM |
624 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
625 | { | |
d1310b2e CM |
626 | struct extent_io_tree *tree; |
627 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5443be45 CM |
628 | |
629 | if (current->flags & PF_MEMALLOC) { | |
630 | redirty_page_for_writepage(wbc, page); | |
631 | unlock_page(page); | |
632 | return 0; | |
633 | } | |
5f39d397 CM |
634 | return extent_write_full_page(tree, page, btree_get_extent, wbc); |
635 | } | |
0da5468f CM |
636 | |
637 | static int btree_writepages(struct address_space *mapping, | |
638 | struct writeback_control *wbc) | |
639 | { | |
d1310b2e CM |
640 | struct extent_io_tree *tree; |
641 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 642 | if (wbc->sync_mode == WB_SYNC_NONE) { |
793955bc CM |
643 | u64 num_dirty; |
644 | u64 start = 0; | |
24ab9cd8 | 645 | unsigned long thresh = 32 * 1024 * 1024; |
448d640b CM |
646 | |
647 | if (wbc->for_kupdate) | |
648 | return 0; | |
649 | ||
1832a6d5 CM |
650 | num_dirty = count_range_bits(tree, &start, (u64)-1, |
651 | thresh, EXTENT_DIRTY); | |
d397712b | 652 | if (num_dirty < thresh) |
793955bc | 653 | return 0; |
793955bc | 654 | } |
0da5468f CM |
655 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
656 | } | |
657 | ||
b2950863 | 658 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 659 | { |
d1310b2e CM |
660 | struct extent_io_tree *tree; |
661 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
662 | return extent_read_full_page(tree, page, btree_get_extent); |
663 | } | |
22b0ebda | 664 | |
70dec807 | 665 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 666 | { |
d1310b2e CM |
667 | struct extent_io_tree *tree; |
668 | struct extent_map_tree *map; | |
5f39d397 | 669 | int ret; |
d98237b3 | 670 | |
98509cfc | 671 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 672 | return 0; |
98509cfc | 673 | |
d1310b2e CM |
674 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
675 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
6af118ce | 676 | |
7b13b7b1 | 677 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
d397712b | 678 | if (!ret) |
6af118ce | 679 | return 0; |
6af118ce CM |
680 | |
681 | ret = try_release_extent_buffer(tree, page); | |
5f39d397 CM |
682 | if (ret == 1) { |
683 | ClearPagePrivate(page); | |
684 | set_page_private(page, 0); | |
685 | page_cache_release(page); | |
686 | } | |
6af118ce | 687 | |
d98237b3 CM |
688 | return ret; |
689 | } | |
690 | ||
5f39d397 | 691 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 692 | { |
d1310b2e CM |
693 | struct extent_io_tree *tree; |
694 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
695 | extent_invalidatepage(tree, page, offset); |
696 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 697 | if (PagePrivate(page)) { |
d397712b CM |
698 | printk(KERN_WARNING "btrfs warning page private not zero " |
699 | "on page %llu\n", (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
700 | ClearPagePrivate(page); |
701 | set_page_private(page, 0); | |
702 | page_cache_release(page); | |
703 | } | |
d98237b3 CM |
704 | } |
705 | ||
5f39d397 | 706 | #if 0 |
d98237b3 | 707 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
ed2ff2cb | 708 | { |
87cbda5c | 709 | struct buffer_head *bh; |
0f7d52f4 | 710 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
87cbda5c | 711 | struct buffer_head *head; |
87cbda5c CM |
712 | if (!page_has_buffers(page)) { |
713 | create_empty_buffers(page, root->fs_info->sb->s_blocksize, | |
714 | (1 << BH_Dirty)|(1 << BH_Uptodate)); | |
715 | } | |
716 | head = page_buffers(page); | |
717 | bh = head; | |
718 | do { | |
719 | if (buffer_dirty(bh)) | |
720 | csum_tree_block(root, bh, 0); | |
721 | bh = bh->b_this_page; | |
722 | } while (bh != head); | |
d98237b3 | 723 | return block_write_full_page(page, btree_get_block, wbc); |
ed2ff2cb | 724 | } |
5f39d397 | 725 | #endif |
eb60ceac | 726 | |
d98237b3 CM |
727 | static struct address_space_operations btree_aops = { |
728 | .readpage = btree_readpage, | |
729 | .writepage = btree_writepage, | |
0da5468f | 730 | .writepages = btree_writepages, |
5f39d397 CM |
731 | .releasepage = btree_releasepage, |
732 | .invalidatepage = btree_invalidatepage, | |
d98237b3 CM |
733 | .sync_page = block_sync_page, |
734 | }; | |
735 | ||
ca7a79ad CM |
736 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
737 | u64 parent_transid) | |
090d1875 | 738 | { |
5f39d397 CM |
739 | struct extent_buffer *buf = NULL; |
740 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 741 | int ret = 0; |
090d1875 | 742 | |
db94535d | 743 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 744 | if (!buf) |
090d1875 | 745 | return 0; |
d1310b2e | 746 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
f188591e | 747 | buf, 0, 0, btree_get_extent, 0); |
5f39d397 | 748 | free_extent_buffer(buf); |
de428b63 | 749 | return ret; |
090d1875 CM |
750 | } |
751 | ||
0999df54 CM |
752 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
753 | u64 bytenr, u32 blocksize) | |
754 | { | |
755 | struct inode *btree_inode = root->fs_info->btree_inode; | |
756 | struct extent_buffer *eb; | |
757 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
758 | bytenr, blocksize, GFP_NOFS); | |
759 | return eb; | |
760 | } | |
761 | ||
762 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
763 | u64 bytenr, u32 blocksize) | |
764 | { | |
765 | struct inode *btree_inode = root->fs_info->btree_inode; | |
766 | struct extent_buffer *eb; | |
767 | ||
768 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
769 | bytenr, blocksize, NULL, GFP_NOFS); | |
770 | return eb; | |
771 | } | |
772 | ||
773 | ||
e02119d5 CM |
774 | int btrfs_write_tree_block(struct extent_buffer *buf) |
775 | { | |
776 | return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start, | |
24ab9cd8 | 777 | buf->start + buf->len - 1, WB_SYNC_ALL); |
e02119d5 CM |
778 | } |
779 | ||
780 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
781 | { | |
782 | return btrfs_wait_on_page_writeback_range(buf->first_page->mapping, | |
d397712b | 783 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
784 | } |
785 | ||
0999df54 | 786 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 787 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
788 | { |
789 | struct extent_buffer *buf = NULL; | |
790 | struct inode *btree_inode = root->fs_info->btree_inode; | |
791 | struct extent_io_tree *io_tree; | |
792 | int ret; | |
793 | ||
794 | io_tree = &BTRFS_I(btree_inode)->io_tree; | |
795 | ||
796 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
797 | if (!buf) | |
798 | return NULL; | |
0999df54 | 799 | |
ca7a79ad | 800 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 | 801 | |
d397712b | 802 | if (ret == 0) |
ce9adaa5 | 803 | buf->flags |= EXTENT_UPTODATE; |
d397712b | 804 | else |
a1b32a59 | 805 | WARN_ON(1); |
5f39d397 | 806 | return buf; |
ce9adaa5 | 807 | |
eb60ceac CM |
808 | } |
809 | ||
e089f05c | 810 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 811 | struct extent_buffer *buf) |
ed2ff2cb | 812 | { |
5f39d397 | 813 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 | 814 | if (btrfs_header_generation(buf) == |
925baedd CM |
815 | root->fs_info->running_transaction->transid) { |
816 | WARN_ON(!btrfs_tree_locked(buf)); | |
d1310b2e | 817 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 818 | buf); |
925baedd | 819 | } |
5f39d397 CM |
820 | return 0; |
821 | } | |
822 | ||
db94535d | 823 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 824 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 825 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 826 | u64 objectid) |
d97e63b6 | 827 | { |
cfaa7295 | 828 | root->node = NULL; |
a28ec197 | 829 | root->commit_root = NULL; |
31153d81 | 830 | root->ref_tree = NULL; |
db94535d CM |
831 | root->sectorsize = sectorsize; |
832 | root->nodesize = nodesize; | |
833 | root->leafsize = leafsize; | |
87ee04eb | 834 | root->stripesize = stripesize; |
123abc88 | 835 | root->ref_cows = 0; |
0b86a832 CM |
836 | root->track_dirty = 0; |
837 | ||
9f5fae2f | 838 | root->fs_info = fs_info; |
0f7d52f4 CM |
839 | root->objectid = objectid; |
840 | root->last_trans = 0; | |
1b05da2e CM |
841 | root->highest_inode = 0; |
842 | root->last_inode_alloc = 0; | |
58176a96 | 843 | root->name = NULL; |
4313b399 | 844 | root->in_sysfs = 0; |
0b86a832 CM |
845 | |
846 | INIT_LIST_HEAD(&root->dirty_list); | |
7b128766 | 847 | INIT_LIST_HEAD(&root->orphan_list); |
bcc63abb | 848 | INIT_LIST_HEAD(&root->dead_list); |
925baedd | 849 | spin_lock_init(&root->node_lock); |
bcc63abb | 850 | spin_lock_init(&root->list_lock); |
a2135011 | 851 | mutex_init(&root->objectid_mutex); |
e02119d5 | 852 | mutex_init(&root->log_mutex); |
d0c803c4 CM |
853 | extent_io_tree_init(&root->dirty_log_pages, |
854 | fs_info->btree_inode->i_mapping, GFP_NOFS); | |
017e5369 CM |
855 | |
856 | btrfs_leaf_ref_tree_init(&root->ref_tree_struct); | |
857 | root->ref_tree = &root->ref_tree_struct; | |
858 | ||
3768f368 CM |
859 | memset(&root->root_key, 0, sizeof(root->root_key)); |
860 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 861 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 862 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 863 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 864 | init_completion(&root->kobj_unregister); |
6702ed49 CM |
865 | root->defrag_running = 0; |
866 | root->defrag_level = 0; | |
4d775673 | 867 | root->root_key.objectid = objectid; |
3394e160 CM |
868 | root->anon_super.s_root = NULL; |
869 | root->anon_super.s_dev = 0; | |
870 | INIT_LIST_HEAD(&root->anon_super.s_list); | |
871 | INIT_LIST_HEAD(&root->anon_super.s_instances); | |
872 | init_rwsem(&root->anon_super.s_umount); | |
873 | ||
3768f368 CM |
874 | return 0; |
875 | } | |
876 | ||
db94535d | 877 | static int find_and_setup_root(struct btrfs_root *tree_root, |
9f5fae2f CM |
878 | struct btrfs_fs_info *fs_info, |
879 | u64 objectid, | |
e20d96d6 | 880 | struct btrfs_root *root) |
3768f368 CM |
881 | { |
882 | int ret; | |
db94535d | 883 | u32 blocksize; |
84234f3a | 884 | u64 generation; |
3768f368 | 885 | |
db94535d | 886 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
887 | tree_root->sectorsize, tree_root->stripesize, |
888 | root, fs_info, objectid); | |
3768f368 CM |
889 | ret = btrfs_find_last_root(tree_root, objectid, |
890 | &root->root_item, &root->root_key); | |
891 | BUG_ON(ret); | |
892 | ||
84234f3a | 893 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
894 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
895 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 896 | blocksize, generation); |
3768f368 | 897 | BUG_ON(!root->node); |
d97e63b6 CM |
898 | return 0; |
899 | } | |
900 | ||
e02119d5 CM |
901 | int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans, |
902 | struct btrfs_fs_info *fs_info) | |
903 | { | |
904 | struct extent_buffer *eb; | |
d0c803c4 CM |
905 | struct btrfs_root *log_root_tree = fs_info->log_root_tree; |
906 | u64 start = 0; | |
907 | u64 end = 0; | |
e02119d5 CM |
908 | int ret; |
909 | ||
d0c803c4 | 910 | if (!log_root_tree) |
e02119d5 CM |
911 | return 0; |
912 | ||
d397712b | 913 | while (1) { |
d0c803c4 CM |
914 | ret = find_first_extent_bit(&log_root_tree->dirty_log_pages, |
915 | 0, &start, &end, EXTENT_DIRTY); | |
916 | if (ret) | |
917 | break; | |
918 | ||
919 | clear_extent_dirty(&log_root_tree->dirty_log_pages, | |
920 | start, end, GFP_NOFS); | |
921 | } | |
e02119d5 CM |
922 | eb = fs_info->log_root_tree->node; |
923 | ||
924 | WARN_ON(btrfs_header_level(eb) != 0); | |
925 | WARN_ON(btrfs_header_nritems(eb) != 0); | |
926 | ||
d00aff00 CM |
927 | ret = btrfs_free_reserved_extent(fs_info->tree_root, |
928 | eb->start, eb->len); | |
e02119d5 CM |
929 | BUG_ON(ret); |
930 | ||
931 | free_extent_buffer(eb); | |
932 | kfree(fs_info->log_root_tree); | |
933 | fs_info->log_root_tree = NULL; | |
934 | return 0; | |
935 | } | |
936 | ||
937 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
938 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
939 | { |
940 | struct btrfs_root *root; | |
941 | struct btrfs_root *tree_root = fs_info->tree_root; | |
e02119d5 CM |
942 | |
943 | root = kzalloc(sizeof(*root), GFP_NOFS); | |
944 | if (!root) | |
945 | return -ENOMEM; | |
946 | ||
947 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
948 | tree_root->sectorsize, tree_root->stripesize, | |
949 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
950 | ||
951 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
952 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
953 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
954 | root->ref_cows = 0; | |
955 | ||
956 | root->node = btrfs_alloc_free_block(trans, root, root->leafsize, | |
31840ae1 ZY |
957 | 0, BTRFS_TREE_LOG_OBJECTID, |
958 | trans->transid, 0, 0, 0); | |
e02119d5 CM |
959 | |
960 | btrfs_set_header_nritems(root->node, 0); | |
961 | btrfs_set_header_level(root->node, 0); | |
962 | btrfs_set_header_bytenr(root->node, root->node->start); | |
963 | btrfs_set_header_generation(root->node, trans->transid); | |
964 | btrfs_set_header_owner(root->node, BTRFS_TREE_LOG_OBJECTID); | |
965 | ||
966 | write_extent_buffer(root->node, root->fs_info->fsid, | |
967 | (unsigned long)btrfs_header_fsid(root->node), | |
968 | BTRFS_FSID_SIZE); | |
969 | btrfs_mark_buffer_dirty(root->node); | |
970 | btrfs_tree_unlock(root->node); | |
971 | fs_info->log_root_tree = root; | |
972 | return 0; | |
973 | } | |
974 | ||
975 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
976 | struct btrfs_key *location) | |
977 | { | |
978 | struct btrfs_root *root; | |
979 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 980 | struct btrfs_path *path; |
5f39d397 | 981 | struct extent_buffer *l; |
1b05da2e | 982 | u64 highest_inode; |
84234f3a | 983 | u64 generation; |
db94535d | 984 | u32 blocksize; |
0f7d52f4 CM |
985 | int ret = 0; |
986 | ||
5eda7b5e | 987 | root = kzalloc(sizeof(*root), GFP_NOFS); |
0cf6c620 | 988 | if (!root) |
0f7d52f4 | 989 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 990 | if (location->offset == (u64)-1) { |
db94535d | 991 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
992 | location->objectid, root); |
993 | if (ret) { | |
0f7d52f4 CM |
994 | kfree(root); |
995 | return ERR_PTR(ret); | |
996 | } | |
997 | goto insert; | |
998 | } | |
999 | ||
db94535d | 1000 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1001 | tree_root->sectorsize, tree_root->stripesize, |
1002 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
1003 | |
1004 | path = btrfs_alloc_path(); | |
1005 | BUG_ON(!path); | |
1006 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); | |
1007 | if (ret != 0) { | |
0f7d52f4 CM |
1008 | if (ret > 0) |
1009 | ret = -ENOENT; | |
1010 | goto out; | |
1011 | } | |
5f39d397 CM |
1012 | l = path->nodes[0]; |
1013 | read_extent_buffer(l, &root->root_item, | |
1014 | btrfs_item_ptr_offset(l, path->slots[0]), | |
0f7d52f4 | 1015 | sizeof(root->root_item)); |
44b36eb2 | 1016 | memcpy(&root->root_key, location, sizeof(*location)); |
0f7d52f4 CM |
1017 | ret = 0; |
1018 | out: | |
1019 | btrfs_release_path(root, path); | |
1020 | btrfs_free_path(path); | |
1021 | if (ret) { | |
1022 | kfree(root); | |
1023 | return ERR_PTR(ret); | |
1024 | } | |
84234f3a | 1025 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1026 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1027 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1028 | blocksize, generation); |
0f7d52f4 CM |
1029 | BUG_ON(!root->node); |
1030 | insert: | |
e02119d5 CM |
1031 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { |
1032 | root->ref_cows = 1; | |
1033 | ret = btrfs_find_highest_inode(root, &highest_inode); | |
1034 | if (ret == 0) { | |
1035 | root->highest_inode = highest_inode; | |
1036 | root->last_inode_alloc = highest_inode; | |
1037 | } | |
5eda7b5e CM |
1038 | } |
1039 | return root; | |
1040 | } | |
1041 | ||
dc17ff8f CM |
1042 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1043 | u64 root_objectid) | |
1044 | { | |
1045 | struct btrfs_root *root; | |
1046 | ||
1047 | if (root_objectid == BTRFS_ROOT_TREE_OBJECTID) | |
1048 | return fs_info->tree_root; | |
1049 | if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1050 | return fs_info->extent_root; | |
1051 | ||
1052 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1053 | (unsigned long)root_objectid); | |
1054 | return root; | |
1055 | } | |
1056 | ||
edbd8d4e CM |
1057 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
1058 | struct btrfs_key *location) | |
5eda7b5e CM |
1059 | { |
1060 | struct btrfs_root *root; | |
1061 | int ret; | |
1062 | ||
edbd8d4e CM |
1063 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1064 | return fs_info->tree_root; | |
1065 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1066 | return fs_info->extent_root; | |
8f18cf13 CM |
1067 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1068 | return fs_info->chunk_root; | |
1069 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1070 | return fs_info->dev_root; | |
0403e47e YZ |
1071 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1072 | return fs_info->csum_root; | |
edbd8d4e | 1073 | |
5eda7b5e CM |
1074 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
1075 | (unsigned long)location->objectid); | |
1076 | if (root) | |
1077 | return root; | |
1078 | ||
e02119d5 | 1079 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
1080 | if (IS_ERR(root)) |
1081 | return root; | |
3394e160 CM |
1082 | |
1083 | set_anon_super(&root->anon_super, NULL); | |
1084 | ||
2619ba1f CM |
1085 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
1086 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 CM |
1087 | root); |
1088 | if (ret) { | |
5f39d397 | 1089 | free_extent_buffer(root->node); |
0f7d52f4 CM |
1090 | kfree(root); |
1091 | return ERR_PTR(ret); | |
1092 | } | |
c146afad YZ |
1093 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
1094 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
1095 | root->root_key.objectid, root); | |
1096 | BUG_ON(ret); | |
1097 | btrfs_orphan_cleanup(root); | |
1098 | } | |
edbd8d4e CM |
1099 | return root; |
1100 | } | |
1101 | ||
1102 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, | |
1103 | struct btrfs_key *location, | |
1104 | const char *name, int namelen) | |
1105 | { | |
1106 | struct btrfs_root *root; | |
1107 | int ret; | |
1108 | ||
1109 | root = btrfs_read_fs_root_no_name(fs_info, location); | |
1110 | if (!root) | |
1111 | return NULL; | |
58176a96 | 1112 | |
4313b399 CM |
1113 | if (root->in_sysfs) |
1114 | return root; | |
1115 | ||
58176a96 JB |
1116 | ret = btrfs_set_root_name(root, name, namelen); |
1117 | if (ret) { | |
5f39d397 | 1118 | free_extent_buffer(root->node); |
58176a96 JB |
1119 | kfree(root); |
1120 | return ERR_PTR(ret); | |
1121 | } | |
ea9e8b11 | 1122 | #if 0 |
58176a96 JB |
1123 | ret = btrfs_sysfs_add_root(root); |
1124 | if (ret) { | |
5f39d397 | 1125 | free_extent_buffer(root->node); |
58176a96 JB |
1126 | kfree(root->name); |
1127 | kfree(root); | |
1128 | return ERR_PTR(ret); | |
1129 | } | |
ea9e8b11 | 1130 | #endif |
4313b399 | 1131 | root->in_sysfs = 1; |
0f7d52f4 CM |
1132 | return root; |
1133 | } | |
04160088 CM |
1134 | |
1135 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) | |
1136 | { | |
1137 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1138 | int ret = 0; | |
1139 | struct list_head *cur; | |
1140 | struct btrfs_device *device; | |
1141 | struct backing_dev_info *bdi; | |
6f3577bd | 1142 | #if 0 |
cb03c743 | 1143 | if ((bdi_bits & (1 << BDI_write_congested)) && |
777e6bd7 | 1144 | btrfs_congested_async(info, 0)) |
cb03c743 | 1145 | return 1; |
6f3577bd | 1146 | #endif |
04160088 CM |
1147 | list_for_each(cur, &info->fs_devices->devices) { |
1148 | device = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1149 | if (!device->bdev) |
1150 | continue; | |
04160088 CM |
1151 | bdi = blk_get_backing_dev_info(device->bdev); |
1152 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1153 | ret = 1; | |
1154 | break; | |
1155 | } | |
1156 | } | |
1157 | return ret; | |
1158 | } | |
1159 | ||
38b66988 CM |
1160 | /* |
1161 | * this unplugs every device on the box, and it is only used when page | |
1162 | * is null | |
1163 | */ | |
1164 | static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) | |
1165 | { | |
1166 | struct list_head *cur; | |
1167 | struct btrfs_device *device; | |
1168 | struct btrfs_fs_info *info; | |
1169 | ||
1170 | info = (struct btrfs_fs_info *)bdi->unplug_io_data; | |
1171 | list_for_each(cur, &info->fs_devices->devices) { | |
1172 | device = list_entry(cur, struct btrfs_device, dev_list); | |
d20f7043 CM |
1173 | if (!device->bdev) |
1174 | continue; | |
1175 | ||
38b66988 | 1176 | bdi = blk_get_backing_dev_info(device->bdev); |
d397712b | 1177 | if (bdi->unplug_io_fn) |
38b66988 | 1178 | bdi->unplug_io_fn(bdi, page); |
38b66988 CM |
1179 | } |
1180 | } | |
1181 | ||
b2950863 | 1182 | static void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) |
04160088 | 1183 | { |
38b66988 | 1184 | struct inode *inode; |
f2d8d74d CM |
1185 | struct extent_map_tree *em_tree; |
1186 | struct extent_map *em; | |
bcbfce8a | 1187 | struct address_space *mapping; |
38b66988 CM |
1188 | u64 offset; |
1189 | ||
bcbfce8a | 1190 | /* the generic O_DIRECT read code does this */ |
9f0ba5bd | 1191 | if (1 || !page) { |
38b66988 CM |
1192 | __unplug_io_fn(bdi, page); |
1193 | return; | |
1194 | } | |
1195 | ||
bcbfce8a CM |
1196 | /* |
1197 | * page->mapping may change at any time. Get a consistent copy | |
1198 | * and use that for everything below | |
1199 | */ | |
1200 | smp_mb(); | |
1201 | mapping = page->mapping; | |
1202 | if (!mapping) | |
1203 | return; | |
1204 | ||
1205 | inode = mapping->host; | |
240d5d48 CM |
1206 | |
1207 | /* | |
1208 | * don't do the expensive searching for a small number of | |
1209 | * devices | |
1210 | */ | |
1211 | if (BTRFS_I(inode)->root->fs_info->fs_devices->open_devices <= 2) { | |
1212 | __unplug_io_fn(bdi, page); | |
1213 | return; | |
1214 | } | |
1215 | ||
38b66988 | 1216 | offset = page_offset(page); |
04160088 | 1217 | |
f2d8d74d CM |
1218 | em_tree = &BTRFS_I(inode)->extent_tree; |
1219 | spin_lock(&em_tree->lock); | |
1220 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); | |
1221 | spin_unlock(&em_tree->lock); | |
89642229 CM |
1222 | if (!em) { |
1223 | __unplug_io_fn(bdi, page); | |
f2d8d74d | 1224 | return; |
89642229 | 1225 | } |
f2d8d74d | 1226 | |
89642229 CM |
1227 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { |
1228 | free_extent_map(em); | |
1229 | __unplug_io_fn(bdi, page); | |
1230 | return; | |
1231 | } | |
f2d8d74d CM |
1232 | offset = offset - em->start; |
1233 | btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree, | |
1234 | em->block_start + offset, page); | |
1235 | free_extent_map(em); | |
04160088 CM |
1236 | } |
1237 | ||
1238 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) | |
1239 | { | |
1240 | bdi_init(bdi); | |
4575c9cc | 1241 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1242 | bdi->state = 0; |
1243 | bdi->capabilities = default_backing_dev_info.capabilities; | |
1244 | bdi->unplug_io_fn = btrfs_unplug_io_fn; | |
1245 | bdi->unplug_io_data = info; | |
1246 | bdi->congested_fn = btrfs_congested_fn; | |
1247 | bdi->congested_data = info; | |
1248 | return 0; | |
1249 | } | |
1250 | ||
ce9adaa5 CM |
1251 | static int bio_ready_for_csum(struct bio *bio) |
1252 | { | |
1253 | u64 length = 0; | |
1254 | u64 buf_len = 0; | |
1255 | u64 start = 0; | |
1256 | struct page *page; | |
1257 | struct extent_io_tree *io_tree = NULL; | |
1258 | struct btrfs_fs_info *info = NULL; | |
1259 | struct bio_vec *bvec; | |
1260 | int i; | |
1261 | int ret; | |
1262 | ||
1263 | bio_for_each_segment(bvec, bio, i) { | |
1264 | page = bvec->bv_page; | |
1265 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
1266 | length += bvec->bv_len; | |
1267 | continue; | |
1268 | } | |
1269 | if (!page->private) { | |
1270 | length += bvec->bv_len; | |
1271 | continue; | |
1272 | } | |
1273 | length = bvec->bv_len; | |
1274 | buf_len = page->private >> 2; | |
1275 | start = page_offset(page) + bvec->bv_offset; | |
1276 | io_tree = &BTRFS_I(page->mapping->host)->io_tree; | |
1277 | info = BTRFS_I(page->mapping->host)->root->fs_info; | |
1278 | } | |
1279 | /* are we fully contained in this bio? */ | |
1280 | if (buf_len <= length) | |
1281 | return 1; | |
1282 | ||
1283 | ret = extent_range_uptodate(io_tree, start + length, | |
1284 | start + buf_len - 1); | |
1285 | if (ret == 1) | |
1286 | return ret; | |
1287 | return ret; | |
1288 | } | |
1289 | ||
8b712842 CM |
1290 | /* |
1291 | * called by the kthread helper functions to finally call the bio end_io | |
1292 | * functions. This is where read checksum verification actually happens | |
1293 | */ | |
1294 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1295 | { |
ce9adaa5 | 1296 | struct bio *bio; |
8b712842 CM |
1297 | struct end_io_wq *end_io_wq; |
1298 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1299 | int error; |
ce9adaa5 | 1300 | |
8b712842 CM |
1301 | end_io_wq = container_of(work, struct end_io_wq, work); |
1302 | bio = end_io_wq->bio; | |
1303 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1304 | |
cad321ad | 1305 | /* metadata bio reads are special because the whole tree block must |
8b712842 CM |
1306 | * be checksummed at once. This makes sure the entire block is in |
1307 | * ram and up to date before trying to verify things. For | |
1308 | * blocksize <= pagesize, it is basically a noop | |
1309 | */ | |
cad321ad CM |
1310 | if (!(bio->bi_rw & (1 << BIO_RW)) && end_io_wq->metadata && |
1311 | !bio_ready_for_csum(bio)) { | |
d20f7043 | 1312 | btrfs_queue_worker(&fs_info->endio_meta_workers, |
8b712842 CM |
1313 | &end_io_wq->work); |
1314 | return; | |
1315 | } | |
1316 | error = end_io_wq->error; | |
1317 | bio->bi_private = end_io_wq->private; | |
1318 | bio->bi_end_io = end_io_wq->end_io; | |
1319 | kfree(end_io_wq); | |
8b712842 | 1320 | bio_endio(bio, error); |
44b8bd7e CM |
1321 | } |
1322 | ||
a74a4b97 CM |
1323 | static int cleaner_kthread(void *arg) |
1324 | { | |
1325 | struct btrfs_root *root = arg; | |
1326 | ||
1327 | do { | |
1328 | smp_mb(); | |
1329 | if (root->fs_info->closing) | |
1330 | break; | |
1331 | ||
1332 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1333 | mutex_lock(&root->fs_info->cleaner_mutex); | |
a74a4b97 | 1334 | btrfs_clean_old_snapshots(root); |
a74a4b97 CM |
1335 | mutex_unlock(&root->fs_info->cleaner_mutex); |
1336 | ||
1337 | if (freezing(current)) { | |
1338 | refrigerator(); | |
1339 | } else { | |
1340 | smp_mb(); | |
1341 | if (root->fs_info->closing) | |
1342 | break; | |
1343 | set_current_state(TASK_INTERRUPTIBLE); | |
1344 | schedule(); | |
1345 | __set_current_state(TASK_RUNNING); | |
1346 | } | |
1347 | } while (!kthread_should_stop()); | |
1348 | return 0; | |
1349 | } | |
1350 | ||
1351 | static int transaction_kthread(void *arg) | |
1352 | { | |
1353 | struct btrfs_root *root = arg; | |
1354 | struct btrfs_trans_handle *trans; | |
1355 | struct btrfs_transaction *cur; | |
1356 | unsigned long now; | |
1357 | unsigned long delay; | |
1358 | int ret; | |
1359 | ||
1360 | do { | |
1361 | smp_mb(); | |
1362 | if (root->fs_info->closing) | |
1363 | break; | |
1364 | ||
1365 | delay = HZ * 30; | |
1366 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1367 | mutex_lock(&root->fs_info->transaction_kthread_mutex); | |
1368 | ||
ab78c84d | 1369 | if (root->fs_info->total_ref_cache_size > 20 * 1024 * 1024) { |
d397712b CM |
1370 | printk(KERN_INFO "btrfs: total reference cache " |
1371 | "size %llu\n", | |
1372 | root->fs_info->total_ref_cache_size); | |
ab78c84d | 1373 | } |
31153d81 | 1374 | |
a74a4b97 CM |
1375 | mutex_lock(&root->fs_info->trans_mutex); |
1376 | cur = root->fs_info->running_transaction; | |
1377 | if (!cur) { | |
1378 | mutex_unlock(&root->fs_info->trans_mutex); | |
1379 | goto sleep; | |
1380 | } | |
31153d81 | 1381 | |
a74a4b97 CM |
1382 | now = get_seconds(); |
1383 | if (now < cur->start_time || now - cur->start_time < 30) { | |
1384 | mutex_unlock(&root->fs_info->trans_mutex); | |
1385 | delay = HZ * 5; | |
1386 | goto sleep; | |
1387 | } | |
1388 | mutex_unlock(&root->fs_info->trans_mutex); | |
a74a4b97 CM |
1389 | trans = btrfs_start_transaction(root, 1); |
1390 | ret = btrfs_commit_transaction(trans, root); | |
1391 | sleep: | |
1392 | wake_up_process(root->fs_info->cleaner_kthread); | |
1393 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1394 | ||
1395 | if (freezing(current)) { | |
1396 | refrigerator(); | |
1397 | } else { | |
1398 | if (root->fs_info->closing) | |
1399 | break; | |
1400 | set_current_state(TASK_INTERRUPTIBLE); | |
1401 | schedule_timeout(delay); | |
1402 | __set_current_state(TASK_RUNNING); | |
1403 | } | |
1404 | } while (!kthread_should_stop()); | |
1405 | return 0; | |
1406 | } | |
1407 | ||
8a4b83cc | 1408 | struct btrfs_root *open_ctree(struct super_block *sb, |
dfe25020 CM |
1409 | struct btrfs_fs_devices *fs_devices, |
1410 | char *options) | |
2e635a27 | 1411 | { |
db94535d CM |
1412 | u32 sectorsize; |
1413 | u32 nodesize; | |
1414 | u32 leafsize; | |
1415 | u32 blocksize; | |
87ee04eb | 1416 | u32 stripesize; |
84234f3a | 1417 | u64 generation; |
f2b636e8 | 1418 | u64 features; |
3de4586c | 1419 | struct btrfs_key location; |
a061fc8d | 1420 | struct buffer_head *bh; |
e02119d5 | 1421 | struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1422 | GFP_NOFS); |
d20f7043 CM |
1423 | struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root), |
1424 | GFP_NOFS); | |
e02119d5 | 1425 | struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1426 | GFP_NOFS); |
8790d502 | 1427 | struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), |
e20d96d6 | 1428 | GFP_NOFS); |
e02119d5 | 1429 | struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1430 | GFP_NOFS); |
e02119d5 | 1431 | struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1432 | GFP_NOFS); |
e02119d5 CM |
1433 | struct btrfs_root *log_tree_root; |
1434 | ||
eb60ceac | 1435 | int ret; |
e58ca020 | 1436 | int err = -EINVAL; |
4543df7e | 1437 | |
2c90e5d6 | 1438 | struct btrfs_super_block *disk_super; |
8790d502 | 1439 | |
0463bb4e | 1440 | if (!extent_root || !tree_root || !fs_info || |
d20f7043 | 1441 | !chunk_root || !dev_root || !csum_root) { |
39279cc3 CM |
1442 | err = -ENOMEM; |
1443 | goto fail; | |
1444 | } | |
0f7d52f4 | 1445 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS); |
8fd17795 | 1446 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1447 | INIT_LIST_HEAD(&fs_info->dead_roots); |
19c00ddc | 1448 | INIT_LIST_HEAD(&fs_info->hashers); |
ea8c2819 | 1449 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
19c00ddc | 1450 | spin_lock_init(&fs_info->hash_lock); |
1832a6d5 | 1451 | spin_lock_init(&fs_info->delalloc_lock); |
cee36a03 | 1452 | spin_lock_init(&fs_info->new_trans_lock); |
31153d81 | 1453 | spin_lock_init(&fs_info->ref_cache_lock); |
19c00ddc | 1454 | |
58176a96 | 1455 | init_completion(&fs_info->kobj_unregister); |
9f5fae2f CM |
1456 | fs_info->tree_root = tree_root; |
1457 | fs_info->extent_root = extent_root; | |
d20f7043 | 1458 | fs_info->csum_root = csum_root; |
0b86a832 CM |
1459 | fs_info->chunk_root = chunk_root; |
1460 | fs_info->dev_root = dev_root; | |
8a4b83cc | 1461 | fs_info->fs_devices = fs_devices; |
0b86a832 | 1462 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1463 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1464 | btrfs_mapping_init(&fs_info->mapping_tree); |
cb03c743 | 1465 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 1466 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 1467 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 1468 | atomic_set(&fs_info->nr_async_bios, 0); |
a2135011 | 1469 | atomic_set(&fs_info->throttles, 0); |
ab78c84d | 1470 | atomic_set(&fs_info->throttle_gen, 0); |
e20d96d6 | 1471 | fs_info->sb = sb; |
c59f8951 | 1472 | fs_info->max_extent = (u64)-1; |
6f568d35 | 1473 | fs_info->max_inline = 8192 * 1024; |
04160088 | 1474 | setup_bdi(fs_info, &fs_info->bdi); |
d98237b3 CM |
1475 | fs_info->btree_inode = new_inode(sb); |
1476 | fs_info->btree_inode->i_ino = 1; | |
2c90e5d6 | 1477 | fs_info->btree_inode->i_nlink = 1; |
c8b97818 | 1478 | |
b34b086c CM |
1479 | fs_info->thread_pool_size = min_t(unsigned long, |
1480 | num_online_cpus() + 2, 8); | |
0afbaf8c | 1481 | |
3eaa2885 CM |
1482 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
1483 | spin_lock_init(&fs_info->ordered_extent_lock); | |
1484 | ||
a061fc8d CM |
1485 | sb->s_blocksize = 4096; |
1486 | sb->s_blocksize_bits = blksize_bits(4096); | |
1487 | ||
0afbaf8c CM |
1488 | /* |
1489 | * we set the i_size on the btree inode to the max possible int. | |
1490 | * the real end of the address space is determined by all of | |
1491 | * the devices in the system | |
1492 | */ | |
1493 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 1494 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
1495 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
1496 | ||
d1310b2e | 1497 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
5f39d397 CM |
1498 | fs_info->btree_inode->i_mapping, |
1499 | GFP_NOFS); | |
d1310b2e CM |
1500 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, |
1501 | GFP_NOFS); | |
1502 | ||
1503 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 1504 | |
0f9dd46c JB |
1505 | spin_lock_init(&fs_info->block_group_cache_lock); |
1506 | fs_info->block_group_cache_tree.rb_node = NULL; | |
1507 | ||
d1310b2e | 1508 | extent_io_tree_init(&fs_info->pinned_extents, |
1a5bc167 | 1509 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1510 | extent_io_tree_init(&fs_info->pending_del, |
1a5bc167 | 1511 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1512 | extent_io_tree_init(&fs_info->extent_ins, |
1a5bc167 | 1513 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
e66f709b | 1514 | fs_info->do_barriers = 1; |
e18e4809 | 1515 | |
1a40e23b ZY |
1516 | INIT_LIST_HEAD(&fs_info->dead_reloc_roots); |
1517 | btrfs_leaf_ref_tree_init(&fs_info->reloc_ref_tree); | |
e4657689 ZY |
1518 | btrfs_leaf_ref_tree_init(&fs_info->shared_ref_tree); |
1519 | ||
0f7d52f4 CM |
1520 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
1521 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
1522 | sizeof(struct btrfs_key)); | |
22b0ebda | 1523 | insert_inode_hash(fs_info->btree_inode); |
39279cc3 | 1524 | |
79154b1b | 1525 | mutex_init(&fs_info->trans_mutex); |
e02119d5 | 1526 | mutex_init(&fs_info->tree_log_mutex); |
a2135011 | 1527 | mutex_init(&fs_info->drop_mutex); |
25179201 JB |
1528 | mutex_init(&fs_info->extent_ins_mutex); |
1529 | mutex_init(&fs_info->pinned_mutex); | |
925baedd | 1530 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
1531 | mutex_init(&fs_info->transaction_kthread_mutex); |
1532 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 1533 | mutex_init(&fs_info->volume_mutex); |
1a40e23b | 1534 | mutex_init(&fs_info->tree_reloc_mutex); |
e6dcd2dc | 1535 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 1536 | init_waitqueue_head(&fs_info->transaction_wait); |
4854ddd0 | 1537 | init_waitqueue_head(&fs_info->async_submit_wait); |
e02119d5 CM |
1538 | init_waitqueue_head(&fs_info->tree_log_wait); |
1539 | atomic_set(&fs_info->tree_log_commit, 0); | |
1540 | atomic_set(&fs_info->tree_log_writers, 0); | |
1541 | fs_info->tree_log_transid = 0; | |
3768f368 | 1542 | |
0b86a832 | 1543 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 1544 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 1545 | |
d98237b3 | 1546 | |
a512bbf8 | 1547 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
a061fc8d | 1548 | if (!bh) |
39279cc3 | 1549 | goto fail_iput; |
39279cc3 | 1550 | |
a061fc8d | 1551 | memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy)); |
2d69a0f8 YZ |
1552 | memcpy(&fs_info->super_for_commit, &fs_info->super_copy, |
1553 | sizeof(fs_info->super_for_commit)); | |
a061fc8d | 1554 | brelse(bh); |
5f39d397 | 1555 | |
a061fc8d | 1556 | memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE); |
0b86a832 | 1557 | |
5f39d397 | 1558 | disk_super = &fs_info->super_copy; |
0f7d52f4 | 1559 | if (!btrfs_super_root(disk_super)) |
c6e2bac1 | 1560 | goto fail_iput; |
0f7d52f4 | 1561 | |
2b82032c YZ |
1562 | ret = btrfs_parse_options(tree_root, options); |
1563 | if (ret) { | |
1564 | err = ret; | |
c6e2bac1 | 1565 | goto fail_iput; |
2b82032c | 1566 | } |
dfe25020 | 1567 | |
f2b636e8 JB |
1568 | features = btrfs_super_incompat_flags(disk_super) & |
1569 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
1570 | if (features) { | |
1571 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
1572 | "unsupported optional features (%Lx).\n", | |
1573 | features); | |
1574 | err = -EINVAL; | |
c6e2bac1 | 1575 | goto fail_iput; |
f2b636e8 JB |
1576 | } |
1577 | ||
1578 | features = btrfs_super_compat_ro_flags(disk_super) & | |
1579 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
1580 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
1581 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
1582 | "unsupported option features (%Lx).\n", | |
1583 | features); | |
1584 | err = -EINVAL; | |
c6e2bac1 | 1585 | goto fail_iput; |
f2b636e8 JB |
1586 | } |
1587 | ||
4543df7e CM |
1588 | /* |
1589 | * we need to start all the end_io workers up front because the | |
1590 | * queue work function gets called at interrupt time, and so it | |
1591 | * cannot dynamically grow. | |
1592 | */ | |
5443be45 CM |
1593 | btrfs_init_workers(&fs_info->workers, "worker", |
1594 | fs_info->thread_pool_size); | |
c8b97818 | 1595 | |
771ed689 CM |
1596 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
1597 | fs_info->thread_pool_size); | |
1598 | ||
5443be45 | 1599 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 CM |
1600 | min_t(u64, fs_devices->num_devices, |
1601 | fs_info->thread_pool_size)); | |
61b49440 CM |
1602 | |
1603 | /* a higher idle thresh on the submit workers makes it much more | |
1604 | * likely that bios will be send down in a sane order to the | |
1605 | * devices | |
1606 | */ | |
1607 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 1608 | |
771ed689 | 1609 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 1610 | fs_info->workers.ordered = 1; |
61b49440 | 1611 | |
771ed689 CM |
1612 | fs_info->delalloc_workers.idle_thresh = 2; |
1613 | fs_info->delalloc_workers.ordered = 1; | |
1614 | ||
5443be45 CM |
1615 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1); |
1616 | btrfs_init_workers(&fs_info->endio_workers, "endio", | |
1617 | fs_info->thread_pool_size); | |
d20f7043 CM |
1618 | btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", |
1619 | fs_info->thread_pool_size); | |
cad321ad CM |
1620 | btrfs_init_workers(&fs_info->endio_meta_write_workers, |
1621 | "endio-meta-write", fs_info->thread_pool_size); | |
5443be45 | 1622 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", |
e6dcd2dc | 1623 | fs_info->thread_pool_size); |
61b49440 CM |
1624 | |
1625 | /* | |
1626 | * endios are largely parallel and should have a very | |
1627 | * low idle thresh | |
1628 | */ | |
1629 | fs_info->endio_workers.idle_thresh = 4; | |
3a5f1d45 | 1630 | fs_info->endio_write_workers.idle_thresh = 64; |
cad321ad | 1631 | fs_info->endio_meta_write_workers.idle_thresh = 64; |
61b49440 | 1632 | |
4543df7e | 1633 | btrfs_start_workers(&fs_info->workers, 1); |
1cc127b5 | 1634 | btrfs_start_workers(&fs_info->submit_workers, 1); |
771ed689 | 1635 | btrfs_start_workers(&fs_info->delalloc_workers, 1); |
247e743c | 1636 | btrfs_start_workers(&fs_info->fixup_workers, 1); |
4543df7e | 1637 | btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size); |
d20f7043 CM |
1638 | btrfs_start_workers(&fs_info->endio_meta_workers, |
1639 | fs_info->thread_pool_size); | |
cad321ad CM |
1640 | btrfs_start_workers(&fs_info->endio_meta_write_workers, |
1641 | fs_info->thread_pool_size); | |
e6dcd2dc CM |
1642 | btrfs_start_workers(&fs_info->endio_write_workers, |
1643 | fs_info->thread_pool_size); | |
4543df7e | 1644 | |
4575c9cc | 1645 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
1646 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
1647 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 1648 | |
db94535d CM |
1649 | nodesize = btrfs_super_nodesize(disk_super); |
1650 | leafsize = btrfs_super_leafsize(disk_super); | |
1651 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 1652 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
1653 | tree_root->nodesize = nodesize; |
1654 | tree_root->leafsize = leafsize; | |
1655 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 1656 | tree_root->stripesize = stripesize; |
a061fc8d CM |
1657 | |
1658 | sb->s_blocksize = sectorsize; | |
1659 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 1660 | |
39279cc3 CM |
1661 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
1662 | sizeof(disk_super->magic))) { | |
d397712b | 1663 | printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
1664 | goto fail_sb_buffer; |
1665 | } | |
19c00ddc | 1666 | |
925baedd | 1667 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 1668 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 1669 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 1670 | if (ret) { |
d397712b CM |
1671 | printk(KERN_WARNING "btrfs: failed to read the system " |
1672 | "array on %s\n", sb->s_id); | |
84eed90f CM |
1673 | goto fail_sys_array; |
1674 | } | |
0b86a832 CM |
1675 | |
1676 | blocksize = btrfs_level_size(tree_root, | |
1677 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 1678 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
1679 | |
1680 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1681 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
1682 | ||
1683 | chunk_root->node = read_tree_block(chunk_root, | |
1684 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 1685 | blocksize, generation); |
0b86a832 CM |
1686 | BUG_ON(!chunk_root->node); |
1687 | ||
e17cade2 | 1688 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
d397712b CM |
1689 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), |
1690 | BTRFS_UUID_SIZE); | |
e17cade2 | 1691 | |
925baedd | 1692 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1693 | ret = btrfs_read_chunk_tree(chunk_root); |
925baedd | 1694 | mutex_unlock(&fs_info->chunk_mutex); |
2b82032c | 1695 | if (ret) { |
d397712b CM |
1696 | printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", |
1697 | sb->s_id); | |
2b82032c YZ |
1698 | goto fail_chunk_root; |
1699 | } | |
0b86a832 | 1700 | |
dfe25020 CM |
1701 | btrfs_close_extra_devices(fs_devices); |
1702 | ||
db94535d CM |
1703 | blocksize = btrfs_level_size(tree_root, |
1704 | btrfs_super_root_level(disk_super)); | |
84234f3a | 1705 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 1706 | |
e20d96d6 | 1707 | tree_root->node = read_tree_block(tree_root, |
db94535d | 1708 | btrfs_super_root(disk_super), |
84234f3a | 1709 | blocksize, generation); |
39279cc3 | 1710 | if (!tree_root->node) |
2b82032c | 1711 | goto fail_chunk_root; |
3768f368 | 1712 | |
db94535d CM |
1713 | |
1714 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 1715 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 1716 | if (ret) |
39279cc3 | 1717 | goto fail_tree_root; |
0b86a832 CM |
1718 | extent_root->track_dirty = 1; |
1719 | ||
1720 | ret = find_and_setup_root(tree_root, fs_info, | |
1721 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
1722 | dev_root->track_dirty = 1; | |
1723 | ||
1724 | if (ret) | |
1725 | goto fail_extent_root; | |
3768f368 | 1726 | |
d20f7043 CM |
1727 | ret = find_and_setup_root(tree_root, fs_info, |
1728 | BTRFS_CSUM_TREE_OBJECTID, csum_root); | |
1729 | if (ret) | |
1730 | goto fail_extent_root; | |
1731 | ||
1732 | csum_root->track_dirty = 1; | |
1733 | ||
9078a3e1 CM |
1734 | btrfs_read_block_groups(extent_root); |
1735 | ||
2d69a0f8 | 1736 | fs_info->generation = generation; |
c146afad | 1737 | fs_info->last_trans_committed = generation; |
d18a2c44 CM |
1738 | fs_info->data_alloc_profile = (u64)-1; |
1739 | fs_info->metadata_alloc_profile = (u64)-1; | |
1740 | fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; | |
a74a4b97 CM |
1741 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
1742 | "btrfs-cleaner"); | |
1743 | if (!fs_info->cleaner_kthread) | |
d20f7043 | 1744 | goto fail_csum_root; |
a74a4b97 CM |
1745 | |
1746 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
1747 | tree_root, | |
1748 | "btrfs-transaction"); | |
1749 | if (!fs_info->transaction_kthread) | |
3f157a2f | 1750 | goto fail_cleaner; |
a74a4b97 | 1751 | |
e02119d5 | 1752 | if (btrfs_super_log_root(disk_super) != 0) { |
e02119d5 CM |
1753 | u64 bytenr = btrfs_super_log_root(disk_super); |
1754 | ||
7c2ca468 | 1755 | if (fs_devices->rw_devices == 0) { |
d397712b CM |
1756 | printk(KERN_WARNING "Btrfs log replay required " |
1757 | "on RO media\n"); | |
7c2ca468 CM |
1758 | err = -EIO; |
1759 | goto fail_trans_kthread; | |
1760 | } | |
e02119d5 CM |
1761 | blocksize = |
1762 | btrfs_level_size(tree_root, | |
1763 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 1764 | |
e02119d5 CM |
1765 | log_tree_root = kzalloc(sizeof(struct btrfs_root), |
1766 | GFP_NOFS); | |
1767 | ||
1768 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1769 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1770 | ||
1771 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
1772 | blocksize, |
1773 | generation + 1); | |
e02119d5 CM |
1774 | ret = btrfs_recover_log_trees(log_tree_root); |
1775 | BUG_ON(ret); | |
e556ce2c YZ |
1776 | |
1777 | if (sb->s_flags & MS_RDONLY) { | |
1778 | ret = btrfs_commit_super(tree_root); | |
1779 | BUG_ON(ret); | |
1780 | } | |
e02119d5 | 1781 | } |
1a40e23b | 1782 | |
7c2ca468 CM |
1783 | if (!(sb->s_flags & MS_RDONLY)) { |
1784 | ret = btrfs_cleanup_reloc_trees(tree_root); | |
1785 | BUG_ON(ret); | |
1786 | } | |
1a40e23b | 1787 | |
3de4586c CM |
1788 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
1789 | location.type = BTRFS_ROOT_ITEM_KEY; | |
1790 | location.offset = (u64)-1; | |
1791 | ||
3de4586c CM |
1792 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
1793 | if (!fs_info->fs_root) | |
7c2ca468 | 1794 | goto fail_trans_kthread; |
0f7d52f4 | 1795 | return tree_root; |
39279cc3 | 1796 | |
7c2ca468 CM |
1797 | fail_trans_kthread: |
1798 | kthread_stop(fs_info->transaction_kthread); | |
3f157a2f | 1799 | fail_cleaner: |
a74a4b97 | 1800 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
1801 | |
1802 | /* | |
1803 | * make sure we're done with the btree inode before we stop our | |
1804 | * kthreads | |
1805 | */ | |
1806 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
1807 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
1808 | ||
d20f7043 CM |
1809 | fail_csum_root: |
1810 | free_extent_buffer(csum_root->node); | |
0b86a832 CM |
1811 | fail_extent_root: |
1812 | free_extent_buffer(extent_root->node); | |
39279cc3 | 1813 | fail_tree_root: |
5f39d397 | 1814 | free_extent_buffer(tree_root->node); |
2b82032c YZ |
1815 | fail_chunk_root: |
1816 | free_extent_buffer(chunk_root->node); | |
84eed90f | 1817 | fail_sys_array: |
7c2ca468 | 1818 | free_extent_buffer(dev_root->node); |
39279cc3 | 1819 | fail_sb_buffer: |
247e743c | 1820 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 1821 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
1822 | btrfs_stop_workers(&fs_info->workers); |
1823 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 1824 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 1825 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 1826 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 1827 | btrfs_stop_workers(&fs_info->submit_workers); |
4543df7e | 1828 | fail_iput: |
7c2ca468 | 1829 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
4543df7e | 1830 | iput(fs_info->btree_inode); |
39279cc3 | 1831 | fail: |
dfe25020 | 1832 | btrfs_close_devices(fs_info->fs_devices); |
84eed90f CM |
1833 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
1834 | ||
39279cc3 CM |
1835 | kfree(extent_root); |
1836 | kfree(tree_root); | |
2d2ae547 | 1837 | bdi_destroy(&fs_info->bdi); |
39279cc3 | 1838 | kfree(fs_info); |
83afeac4 JM |
1839 | kfree(chunk_root); |
1840 | kfree(dev_root); | |
d20f7043 | 1841 | kfree(csum_root); |
39279cc3 | 1842 | return ERR_PTR(err); |
eb60ceac CM |
1843 | } |
1844 | ||
f2984462 CM |
1845 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
1846 | { | |
1847 | char b[BDEVNAME_SIZE]; | |
1848 | ||
1849 | if (uptodate) { | |
1850 | set_buffer_uptodate(bh); | |
1851 | } else { | |
1852 | if (!buffer_eopnotsupp(bh) && printk_ratelimit()) { | |
1853 | printk(KERN_WARNING "lost page write due to " | |
1854 | "I/O error on %s\n", | |
1855 | bdevname(bh->b_bdev, b)); | |
1856 | } | |
1259ab75 CM |
1857 | /* note, we dont' set_buffer_write_io_error because we have |
1858 | * our own ways of dealing with the IO errors | |
1859 | */ | |
f2984462 CM |
1860 | clear_buffer_uptodate(bh); |
1861 | } | |
1862 | unlock_buffer(bh); | |
1863 | put_bh(bh); | |
1864 | } | |
1865 | ||
a512bbf8 YZ |
1866 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
1867 | { | |
1868 | struct buffer_head *bh; | |
1869 | struct buffer_head *latest = NULL; | |
1870 | struct btrfs_super_block *super; | |
1871 | int i; | |
1872 | u64 transid = 0; | |
1873 | u64 bytenr; | |
1874 | ||
1875 | /* we would like to check all the supers, but that would make | |
1876 | * a btrfs mount succeed after a mkfs from a different FS. | |
1877 | * So, we need to add a special mount option to scan for | |
1878 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
1879 | */ | |
1880 | for (i = 0; i < 1; i++) { | |
1881 | bytenr = btrfs_sb_offset(i); | |
1882 | if (bytenr + 4096 >= i_size_read(bdev->bd_inode)) | |
1883 | break; | |
1884 | bh = __bread(bdev, bytenr / 4096, 4096); | |
1885 | if (!bh) | |
1886 | continue; | |
1887 | ||
1888 | super = (struct btrfs_super_block *)bh->b_data; | |
1889 | if (btrfs_super_bytenr(super) != bytenr || | |
1890 | strncmp((char *)(&super->magic), BTRFS_MAGIC, | |
1891 | sizeof(super->magic))) { | |
1892 | brelse(bh); | |
1893 | continue; | |
1894 | } | |
1895 | ||
1896 | if (!latest || btrfs_super_generation(super) > transid) { | |
1897 | brelse(latest); | |
1898 | latest = bh; | |
1899 | transid = btrfs_super_generation(super); | |
1900 | } else { | |
1901 | brelse(bh); | |
1902 | } | |
1903 | } | |
1904 | return latest; | |
1905 | } | |
1906 | ||
1907 | static int write_dev_supers(struct btrfs_device *device, | |
1908 | struct btrfs_super_block *sb, | |
1909 | int do_barriers, int wait, int max_mirrors) | |
1910 | { | |
1911 | struct buffer_head *bh; | |
1912 | int i; | |
1913 | int ret; | |
1914 | int errors = 0; | |
1915 | u32 crc; | |
1916 | u64 bytenr; | |
1917 | int last_barrier = 0; | |
1918 | ||
1919 | if (max_mirrors == 0) | |
1920 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
1921 | ||
1922 | /* make sure only the last submit_bh does a barrier */ | |
1923 | if (do_barriers) { | |
1924 | for (i = 0; i < max_mirrors; i++) { | |
1925 | bytenr = btrfs_sb_offset(i); | |
1926 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= | |
1927 | device->total_bytes) | |
1928 | break; | |
1929 | last_barrier = i; | |
1930 | } | |
1931 | } | |
1932 | ||
1933 | for (i = 0; i < max_mirrors; i++) { | |
1934 | bytenr = btrfs_sb_offset(i); | |
1935 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) | |
1936 | break; | |
1937 | ||
1938 | if (wait) { | |
1939 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
1940 | BTRFS_SUPER_INFO_SIZE); | |
1941 | BUG_ON(!bh); | |
1942 | brelse(bh); | |
1943 | wait_on_buffer(bh); | |
1944 | if (buffer_uptodate(bh)) { | |
1945 | brelse(bh); | |
1946 | continue; | |
1947 | } | |
1948 | } else { | |
1949 | btrfs_set_super_bytenr(sb, bytenr); | |
1950 | ||
1951 | crc = ~(u32)0; | |
1952 | crc = btrfs_csum_data(NULL, (char *)sb + | |
1953 | BTRFS_CSUM_SIZE, crc, | |
1954 | BTRFS_SUPER_INFO_SIZE - | |
1955 | BTRFS_CSUM_SIZE); | |
1956 | btrfs_csum_final(crc, sb->csum); | |
1957 | ||
1958 | bh = __getblk(device->bdev, bytenr / 4096, | |
1959 | BTRFS_SUPER_INFO_SIZE); | |
1960 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); | |
1961 | ||
1962 | set_buffer_uptodate(bh); | |
1963 | get_bh(bh); | |
1964 | lock_buffer(bh); | |
1965 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
1966 | } | |
1967 | ||
1968 | if (i == last_barrier && do_barriers && device->barriers) { | |
1969 | ret = submit_bh(WRITE_BARRIER, bh); | |
1970 | if (ret == -EOPNOTSUPP) { | |
1971 | printk("btrfs: disabling barriers on dev %s\n", | |
1972 | device->name); | |
1973 | set_buffer_uptodate(bh); | |
1974 | device->barriers = 0; | |
1975 | get_bh(bh); | |
1976 | lock_buffer(bh); | |
1977 | ret = submit_bh(WRITE, bh); | |
1978 | } | |
1979 | } else { | |
1980 | ret = submit_bh(WRITE, bh); | |
1981 | } | |
1982 | ||
1983 | if (!ret && wait) { | |
1984 | wait_on_buffer(bh); | |
1985 | if (!buffer_uptodate(bh)) | |
1986 | errors++; | |
1987 | } else if (ret) { | |
1988 | errors++; | |
1989 | } | |
1990 | if (wait) | |
1991 | brelse(bh); | |
1992 | } | |
1993 | return errors < i ? 0 : -1; | |
1994 | } | |
1995 | ||
1996 | int write_all_supers(struct btrfs_root *root, int max_mirrors) | |
f2984462 CM |
1997 | { |
1998 | struct list_head *cur; | |
1999 | struct list_head *head = &root->fs_info->fs_devices->devices; | |
2000 | struct btrfs_device *dev; | |
a061fc8d | 2001 | struct btrfs_super_block *sb; |
f2984462 | 2002 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
2003 | int ret; |
2004 | int do_barriers; | |
a236aed1 CM |
2005 | int max_errors; |
2006 | int total_errors = 0; | |
a061fc8d | 2007 | u64 flags; |
f2984462 | 2008 | |
a236aed1 | 2009 | max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1; |
f2984462 CM |
2010 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
2011 | ||
a061fc8d CM |
2012 | sb = &root->fs_info->super_for_commit; |
2013 | dev_item = &sb->dev_item; | |
f2984462 CM |
2014 | list_for_each(cur, head) { |
2015 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
2016 | if (!dev->bdev) { |
2017 | total_errors++; | |
2018 | continue; | |
2019 | } | |
2b82032c | 2020 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2021 | continue; |
2022 | ||
2b82032c | 2023 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
2024 | btrfs_set_stack_device_type(dev_item, dev->type); |
2025 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
2026 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
2027 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
2028 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
2029 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
2030 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
2031 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 2032 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 2033 | |
a061fc8d CM |
2034 | flags = btrfs_super_flags(sb); |
2035 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
2036 | ||
a512bbf8 | 2037 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
2038 | if (ret) |
2039 | total_errors++; | |
f2984462 | 2040 | } |
a236aed1 | 2041 | if (total_errors > max_errors) { |
d397712b CM |
2042 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2043 | total_errors); | |
a236aed1 CM |
2044 | BUG(); |
2045 | } | |
f2984462 | 2046 | |
a512bbf8 | 2047 | total_errors = 0; |
f2984462 CM |
2048 | list_for_each(cur, head) { |
2049 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
2050 | if (!dev->bdev) |
2051 | continue; | |
2b82032c | 2052 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2053 | continue; |
2054 | ||
a512bbf8 YZ |
2055 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
2056 | if (ret) | |
2057 | total_errors++; | |
f2984462 | 2058 | } |
a236aed1 | 2059 | if (total_errors > max_errors) { |
d397712b CM |
2060 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2061 | total_errors); | |
a236aed1 CM |
2062 | BUG(); |
2063 | } | |
f2984462 CM |
2064 | return 0; |
2065 | } | |
2066 | ||
a512bbf8 YZ |
2067 | int write_ctree_super(struct btrfs_trans_handle *trans, |
2068 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 2069 | { |
e66f709b | 2070 | int ret; |
5f39d397 | 2071 | |
a512bbf8 | 2072 | ret = write_all_supers(root, max_mirrors); |
5f39d397 | 2073 | return ret; |
cfaa7295 CM |
2074 | } |
2075 | ||
5eda7b5e | 2076 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f CM |
2077 | { |
2078 | radix_tree_delete(&fs_info->fs_roots_radix, | |
2079 | (unsigned long)root->root_key.objectid); | |
3394e160 CM |
2080 | if (root->anon_super.s_dev) { |
2081 | down_write(&root->anon_super.s_umount); | |
2082 | kill_anon_super(&root->anon_super); | |
2083 | } | |
2619ba1f | 2084 | if (root->node) |
5f39d397 | 2085 | free_extent_buffer(root->node); |
2619ba1f | 2086 | if (root->commit_root) |
5f39d397 | 2087 | free_extent_buffer(root->commit_root); |
d397712b | 2088 | kfree(root->name); |
2619ba1f CM |
2089 | kfree(root); |
2090 | return 0; | |
2091 | } | |
2092 | ||
35b7e476 | 2093 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
2094 | { |
2095 | int ret; | |
2096 | struct btrfs_root *gang[8]; | |
2097 | int i; | |
2098 | ||
d397712b | 2099 | while (1) { |
0f7d52f4 CM |
2100 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
2101 | (void **)gang, 0, | |
2102 | ARRAY_SIZE(gang)); | |
2103 | if (!ret) | |
2104 | break; | |
2619ba1f | 2105 | for (i = 0; i < ret; i++) |
5eda7b5e | 2106 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
2107 | } |
2108 | return 0; | |
2109 | } | |
b4100d64 | 2110 | |
c146afad | 2111 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 2112 | { |
c146afad YZ |
2113 | u64 root_objectid = 0; |
2114 | struct btrfs_root *gang[8]; | |
2115 | int i; | |
3768f368 | 2116 | int ret; |
e089f05c | 2117 | |
c146afad YZ |
2118 | while (1) { |
2119 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2120 | (void **)gang, root_objectid, | |
2121 | ARRAY_SIZE(gang)); | |
2122 | if (!ret) | |
2123 | break; | |
2124 | for (i = 0; i < ret; i++) { | |
2125 | root_objectid = gang[i]->root_key.objectid; | |
2126 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
2127 | root_objectid, gang[i]); | |
2128 | BUG_ON(ret); | |
2129 | btrfs_orphan_cleanup(gang[i]); | |
2130 | } | |
2131 | root_objectid++; | |
2132 | } | |
2133 | return 0; | |
2134 | } | |
a2135011 | 2135 | |
c146afad YZ |
2136 | int btrfs_commit_super(struct btrfs_root *root) |
2137 | { | |
2138 | struct btrfs_trans_handle *trans; | |
2139 | int ret; | |
a74a4b97 | 2140 | |
c146afad | 2141 | mutex_lock(&root->fs_info->cleaner_mutex); |
a74a4b97 | 2142 | btrfs_clean_old_snapshots(root); |
c146afad | 2143 | mutex_unlock(&root->fs_info->cleaner_mutex); |
79154b1b | 2144 | trans = btrfs_start_transaction(root, 1); |
54aa1f4d | 2145 | ret = btrfs_commit_transaction(trans, root); |
c146afad YZ |
2146 | BUG_ON(ret); |
2147 | /* run commit again to drop the original snapshot */ | |
79154b1b CM |
2148 | trans = btrfs_start_transaction(root, 1); |
2149 | btrfs_commit_transaction(trans, root); | |
2150 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 2151 | BUG_ON(ret); |
d6bfde87 | 2152 | |
a512bbf8 | 2153 | ret = write_ctree_super(NULL, root, 0); |
c146afad YZ |
2154 | return ret; |
2155 | } | |
2156 | ||
2157 | int close_ctree(struct btrfs_root *root) | |
2158 | { | |
2159 | struct btrfs_fs_info *fs_info = root->fs_info; | |
2160 | int ret; | |
2161 | ||
2162 | fs_info->closing = 1; | |
2163 | smp_mb(); | |
2164 | ||
2165 | kthread_stop(root->fs_info->transaction_kthread); | |
2166 | kthread_stop(root->fs_info->cleaner_kthread); | |
2167 | ||
2168 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { | |
2169 | ret = btrfs_commit_super(root); | |
d397712b CM |
2170 | if (ret) |
2171 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
c146afad | 2172 | } |
0f7d52f4 | 2173 | |
b0c68f8b | 2174 | if (fs_info->delalloc_bytes) { |
d397712b | 2175 | printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n", |
b0c68f8b CM |
2176 | fs_info->delalloc_bytes); |
2177 | } | |
31153d81 | 2178 | if (fs_info->total_ref_cache_size) { |
d397712b CM |
2179 | printk(KERN_INFO "btrfs: at umount reference cache size %llu\n", |
2180 | (unsigned long long)fs_info->total_ref_cache_size); | |
31153d81 | 2181 | } |
bcc63abb | 2182 | |
0f7d52f4 | 2183 | if (fs_info->extent_root->node) |
5f39d397 | 2184 | free_extent_buffer(fs_info->extent_root->node); |
f510cfec | 2185 | |
0f7d52f4 | 2186 | if (fs_info->tree_root->node) |
5f39d397 | 2187 | free_extent_buffer(fs_info->tree_root->node); |
f510cfec | 2188 | |
d397712b | 2189 | if (root->fs_info->chunk_root->node) |
0b86a832 CM |
2190 | free_extent_buffer(root->fs_info->chunk_root->node); |
2191 | ||
d397712b | 2192 | if (root->fs_info->dev_root->node) |
0b86a832 CM |
2193 | free_extent_buffer(root->fs_info->dev_root->node); |
2194 | ||
d397712b | 2195 | if (root->fs_info->csum_root->node) |
d20f7043 CM |
2196 | free_extent_buffer(root->fs_info->csum_root->node); |
2197 | ||
9078a3e1 | 2198 | btrfs_free_block_groups(root->fs_info); |
d10c5f31 | 2199 | |
c146afad | 2200 | del_fs_roots(fs_info); |
d10c5f31 | 2201 | |
c146afad | 2202 | iput(fs_info->btree_inode); |
9ad6b7bc | 2203 | |
247e743c | 2204 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2205 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2206 | btrfs_stop_workers(&fs_info->workers); |
2207 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 2208 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 2209 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 2210 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 2211 | btrfs_stop_workers(&fs_info->submit_workers); |
d6bfde87 | 2212 | |
19c00ddc | 2213 | #if 0 |
d397712b | 2214 | while (!list_empty(&fs_info->hashers)) { |
19c00ddc CM |
2215 | struct btrfs_hasher *hasher; |
2216 | hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher, | |
2217 | hashers); | |
2218 | list_del(&hasher->hashers); | |
2219 | crypto_free_hash(&fs_info->hash_tfm); | |
2220 | kfree(hasher); | |
2221 | } | |
2222 | #endif | |
dfe25020 | 2223 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 2224 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 2225 | |
04160088 | 2226 | bdi_destroy(&fs_info->bdi); |
0b86a832 | 2227 | |
0f7d52f4 | 2228 | kfree(fs_info->extent_root); |
0f7d52f4 | 2229 | kfree(fs_info->tree_root); |
0b86a832 CM |
2230 | kfree(fs_info->chunk_root); |
2231 | kfree(fs_info->dev_root); | |
d20f7043 | 2232 | kfree(fs_info->csum_root); |
eb60ceac CM |
2233 | return 0; |
2234 | } | |
2235 | ||
1259ab75 | 2236 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 2237 | { |
1259ab75 | 2238 | int ret; |
810191ff | 2239 | struct inode *btree_inode = buf->first_page->mapping->host; |
1259ab75 CM |
2240 | |
2241 | ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); | |
2242 | if (!ret) | |
2243 | return ret; | |
2244 | ||
2245 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
2246 | parent_transid); | |
2247 | return !ret; | |
5f39d397 CM |
2248 | } |
2249 | ||
2250 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 2251 | { |
810191ff | 2252 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 2253 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
2254 | buf); |
2255 | } | |
6702ed49 | 2256 | |
5f39d397 CM |
2257 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
2258 | { | |
810191ff | 2259 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
5f39d397 CM |
2260 | u64 transid = btrfs_header_generation(buf); |
2261 | struct inode *btree_inode = root->fs_info->btree_inode; | |
6702ed49 | 2262 | |
925baedd | 2263 | WARN_ON(!btrfs_tree_locked(buf)); |
ccd467d6 | 2264 | if (transid != root->fs_info->generation) { |
d397712b CM |
2265 | printk(KERN_CRIT "btrfs transid mismatch buffer %llu, " |
2266 | "found %llu running %llu\n", | |
db94535d | 2267 | (unsigned long long)buf->start, |
d397712b CM |
2268 | (unsigned long long)transid, |
2269 | (unsigned long long)root->fs_info->generation); | |
ccd467d6 CM |
2270 | WARN_ON(1); |
2271 | } | |
d1310b2e | 2272 | set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf); |
eb60ceac CM |
2273 | } |
2274 | ||
d3c2fdcf | 2275 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
35b7e476 | 2276 | { |
188de649 CM |
2277 | /* |
2278 | * looks as though older kernels can get into trouble with | |
2279 | * this code, they end up stuck in balance_dirty_pages forever | |
2280 | */ | |
d6bfde87 CM |
2281 | struct extent_io_tree *tree; |
2282 | u64 num_dirty; | |
2283 | u64 start = 0; | |
771ed689 | 2284 | unsigned long thresh = 32 * 1024 * 1024; |
d6bfde87 CM |
2285 | tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
2286 | ||
b64a2851 | 2287 | if (current_is_pdflush() || current->flags & PF_MEMALLOC) |
d6bfde87 CM |
2288 | return; |
2289 | ||
2290 | num_dirty = count_range_bits(tree, &start, (u64)-1, | |
2291 | thresh, EXTENT_DIRTY); | |
2292 | if (num_dirty > thresh) { | |
2293 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 2294 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 2295 | } |
188de649 | 2296 | return; |
35b7e476 | 2297 | } |
6b80053d | 2298 | |
ca7a79ad | 2299 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 2300 | { |
810191ff | 2301 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
ce9adaa5 | 2302 | int ret; |
ca7a79ad | 2303 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
d397712b | 2304 | if (ret == 0) |
ce9adaa5 | 2305 | buf->flags |= EXTENT_UPTODATE; |
ce9adaa5 | 2306 | return ret; |
6b80053d | 2307 | } |
0da5468f | 2308 | |
4bef0848 CM |
2309 | int btree_lock_page_hook(struct page *page) |
2310 | { | |
2311 | struct inode *inode = page->mapping->host; | |
2312 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
2313 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
2314 | struct extent_buffer *eb; | |
2315 | unsigned long len; | |
2316 | u64 bytenr = page_offset(page); | |
2317 | ||
2318 | if (page->private == EXTENT_PAGE_PRIVATE) | |
2319 | goto out; | |
2320 | ||
2321 | len = page->private >> 2; | |
2322 | eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS); | |
2323 | if (!eb) | |
2324 | goto out; | |
2325 | ||
2326 | btrfs_tree_lock(eb); | |
2327 | spin_lock(&root->fs_info->hash_lock); | |
2328 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); | |
2329 | spin_unlock(&root->fs_info->hash_lock); | |
2330 | btrfs_tree_unlock(eb); | |
2331 | free_extent_buffer(eb); | |
2332 | out: | |
2333 | lock_page(page); | |
2334 | return 0; | |
2335 | } | |
2336 | ||
d1310b2e | 2337 | static struct extent_io_ops btree_extent_io_ops = { |
4bef0848 | 2338 | .write_cache_pages_lock_hook = btree_lock_page_hook, |
ce9adaa5 | 2339 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
0b86a832 | 2340 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
2341 | /* note we're sharing with inode.c for the merge bio hook */ |
2342 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 2343 | }; |