Commit | Line | Data |
---|---|---|
6cbd5570 CM |
1 | /* |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
e20d96d6 | 19 | #include <linux/fs.h> |
d98237b3 | 20 | #include <linux/blkdev.h> |
87cbda5c | 21 | #include <linux/scatterlist.h> |
22b0ebda | 22 | #include <linux/swap.h> |
0f7d52f4 | 23 | #include <linux/radix-tree.h> |
35b7e476 | 24 | #include <linux/writeback.h> |
d397712b | 25 | #include <linux/buffer_head.h> |
ce9adaa5 | 26 | #include <linux/workqueue.h> |
a74a4b97 | 27 | #include <linux/kthread.h> |
4b4e25f2 CM |
28 | #include <linux/freezer.h> |
29 | #include "compat.h" | |
a5eb62e3 | 30 | #include "crc32c.h" |
eb60ceac CM |
31 | #include "ctree.h" |
32 | #include "disk-io.h" | |
e089f05c | 33 | #include "transaction.h" |
0f7d52f4 | 34 | #include "btrfs_inode.h" |
0b86a832 | 35 | #include "volumes.h" |
db94535d | 36 | #include "print-tree.h" |
8b712842 | 37 | #include "async-thread.h" |
925baedd | 38 | #include "locking.h" |
017e5369 | 39 | #include "ref-cache.h" |
e02119d5 | 40 | #include "tree-log.h" |
fa9c0d79 | 41 | #include "free-space-cache.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 | ||
4008c04a CM |
79 | /* These are used to set the lockdep class on the extent buffer locks. |
80 | * The class is set by the readpage_end_io_hook after the buffer has | |
81 | * passed csum validation but before the pages are unlocked. | |
82 | * | |
83 | * The lockdep class is also set by btrfs_init_new_buffer on freshly | |
84 | * allocated blocks. | |
85 | * | |
86 | * The class is based on the level in the tree block, which allows lockdep | |
87 | * to know that lower nodes nest inside the locks of higher nodes. | |
88 | * | |
89 | * We also add a check to make sure the highest level of the tree is | |
90 | * the same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this | |
91 | * code needs update as well. | |
92 | */ | |
93 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
94 | # if BTRFS_MAX_LEVEL != 8 | |
95 | # error | |
96 | # endif | |
97 | static struct lock_class_key btrfs_eb_class[BTRFS_MAX_LEVEL + 1]; | |
98 | static const char *btrfs_eb_name[BTRFS_MAX_LEVEL + 1] = { | |
99 | /* leaf */ | |
100 | "btrfs-extent-00", | |
101 | "btrfs-extent-01", | |
102 | "btrfs-extent-02", | |
103 | "btrfs-extent-03", | |
104 | "btrfs-extent-04", | |
105 | "btrfs-extent-05", | |
106 | "btrfs-extent-06", | |
107 | "btrfs-extent-07", | |
108 | /* highest possible level */ | |
109 | "btrfs-extent-08", | |
110 | }; | |
111 | #endif | |
112 | ||
d352ac68 CM |
113 | /* |
114 | * extents on the btree inode are pretty simple, there's one extent | |
115 | * that covers the entire device | |
116 | */ | |
b2950863 CH |
117 | static struct extent_map *btree_get_extent(struct inode *inode, |
118 | struct page *page, size_t page_offset, u64 start, u64 len, | |
119 | int create) | |
7eccb903 | 120 | { |
5f39d397 CM |
121 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
122 | struct extent_map *em; | |
123 | int ret; | |
124 | ||
d1310b2e CM |
125 | spin_lock(&em_tree->lock); |
126 | em = lookup_extent_mapping(em_tree, start, len); | |
a061fc8d CM |
127 | if (em) { |
128 | em->bdev = | |
129 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
130 | spin_unlock(&em_tree->lock); | |
5f39d397 | 131 | goto out; |
a061fc8d CM |
132 | } |
133 | spin_unlock(&em_tree->lock); | |
7b13b7b1 | 134 | |
5f39d397 CM |
135 | em = alloc_extent_map(GFP_NOFS); |
136 | if (!em) { | |
137 | em = ERR_PTR(-ENOMEM); | |
138 | goto out; | |
139 | } | |
140 | em->start = 0; | |
0afbaf8c | 141 | em->len = (u64)-1; |
c8b97818 | 142 | em->block_len = (u64)-1; |
5f39d397 | 143 | em->block_start = 0; |
a061fc8d | 144 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e CM |
145 | |
146 | spin_lock(&em_tree->lock); | |
5f39d397 CM |
147 | ret = add_extent_mapping(em_tree, em); |
148 | if (ret == -EEXIST) { | |
0afbaf8c CM |
149 | u64 failed_start = em->start; |
150 | u64 failed_len = em->len; | |
151 | ||
5f39d397 | 152 | free_extent_map(em); |
7b13b7b1 | 153 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c | 154 | if (em) { |
7b13b7b1 | 155 | ret = 0; |
0afbaf8c CM |
156 | } else { |
157 | em = lookup_extent_mapping(em_tree, failed_start, | |
158 | failed_len); | |
7b13b7b1 | 159 | ret = -EIO; |
0afbaf8c | 160 | } |
5f39d397 | 161 | } else if (ret) { |
7b13b7b1 CM |
162 | free_extent_map(em); |
163 | em = NULL; | |
5f39d397 | 164 | } |
7b13b7b1 CM |
165 | spin_unlock(&em_tree->lock); |
166 | ||
167 | if (ret) | |
168 | em = ERR_PTR(ret); | |
5f39d397 CM |
169 | out: |
170 | return em; | |
7eccb903 CM |
171 | } |
172 | ||
19c00ddc CM |
173 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
174 | { | |
a5eb62e3 | 175 | return btrfs_crc32c(seed, data, len); |
19c00ddc CM |
176 | } |
177 | ||
178 | void btrfs_csum_final(u32 crc, char *result) | |
179 | { | |
180 | *(__le32 *)result = ~cpu_to_le32(crc); | |
181 | } | |
182 | ||
d352ac68 CM |
183 | /* |
184 | * compute the csum for a btree block, and either verify it or write it | |
185 | * into the csum field of the block. | |
186 | */ | |
19c00ddc CM |
187 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
188 | int verify) | |
189 | { | |
607d432d JB |
190 | u16 csum_size = |
191 | btrfs_super_csum_size(&root->fs_info->super_copy); | |
192 | char *result = NULL; | |
19c00ddc CM |
193 | unsigned long len; |
194 | unsigned long cur_len; | |
195 | unsigned long offset = BTRFS_CSUM_SIZE; | |
196 | char *map_token = NULL; | |
197 | char *kaddr; | |
198 | unsigned long map_start; | |
199 | unsigned long map_len; | |
200 | int err; | |
201 | u32 crc = ~(u32)0; | |
607d432d | 202 | unsigned long inline_result; |
19c00ddc CM |
203 | |
204 | len = buf->len - offset; | |
d397712b | 205 | while (len > 0) { |
19c00ddc CM |
206 | err = map_private_extent_buffer(buf, offset, 32, |
207 | &map_token, &kaddr, | |
208 | &map_start, &map_len, KM_USER0); | |
d397712b | 209 | if (err) |
19c00ddc | 210 | return 1; |
19c00ddc CM |
211 | cur_len = min(len, map_len - (offset - map_start)); |
212 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
213 | crc, cur_len); | |
214 | len -= cur_len; | |
215 | offset += cur_len; | |
216 | unmap_extent_buffer(buf, map_token, KM_USER0); | |
217 | } | |
607d432d JB |
218 | if (csum_size > sizeof(inline_result)) { |
219 | result = kzalloc(csum_size * sizeof(char), GFP_NOFS); | |
220 | if (!result) | |
221 | return 1; | |
222 | } else { | |
223 | result = (char *)&inline_result; | |
224 | } | |
225 | ||
19c00ddc CM |
226 | btrfs_csum_final(crc, result); |
227 | ||
228 | if (verify) { | |
607d432d | 229 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
230 | u32 val; |
231 | u32 found = 0; | |
607d432d | 232 | memcpy(&found, result, csum_size); |
e4204ded | 233 | |
607d432d | 234 | read_extent_buffer(buf, &val, 0, csum_size); |
d397712b CM |
235 | printk(KERN_INFO "btrfs: %s checksum verify failed " |
236 | "on %llu wanted %X found %X level %d\n", | |
19c00ddc | 237 | root->fs_info->sb->s_id, |
2dd3e67b | 238 | buf->start, val, found, btrfs_header_level(buf)); |
607d432d JB |
239 | if (result != (char *)&inline_result) |
240 | kfree(result); | |
19c00ddc CM |
241 | return 1; |
242 | } | |
243 | } else { | |
607d432d | 244 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 245 | } |
607d432d JB |
246 | if (result != (char *)&inline_result) |
247 | kfree(result); | |
19c00ddc CM |
248 | return 0; |
249 | } | |
250 | ||
d352ac68 CM |
251 | /* |
252 | * we can't consider a given block up to date unless the transid of the | |
253 | * block matches the transid in the parent node's pointer. This is how we | |
254 | * detect blocks that either didn't get written at all or got written | |
255 | * in the wrong place. | |
256 | */ | |
1259ab75 CM |
257 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
258 | struct extent_buffer *eb, u64 parent_transid) | |
259 | { | |
260 | int ret; | |
261 | ||
262 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
263 | return 0; | |
264 | ||
265 | lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); | |
266 | if (extent_buffer_uptodate(io_tree, eb) && | |
267 | btrfs_header_generation(eb) == parent_transid) { | |
268 | ret = 0; | |
269 | goto out; | |
270 | } | |
271 | printk("parent transid verify failed on %llu wanted %llu found %llu\n", | |
272 | (unsigned long long)eb->start, | |
273 | (unsigned long long)parent_transid, | |
274 | (unsigned long long)btrfs_header_generation(eb)); | |
275 | ret = 1; | |
1259ab75 | 276 | clear_extent_buffer_uptodate(io_tree, eb); |
33958dc6 | 277 | out: |
1259ab75 CM |
278 | unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, |
279 | GFP_NOFS); | |
280 | return ret; | |
1259ab75 CM |
281 | } |
282 | ||
d352ac68 CM |
283 | /* |
284 | * helper to read a given tree block, doing retries as required when | |
285 | * the checksums don't match and we have alternate mirrors to try. | |
286 | */ | |
f188591e CM |
287 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
288 | struct extent_buffer *eb, | |
ca7a79ad | 289 | u64 start, u64 parent_transid) |
f188591e CM |
290 | { |
291 | struct extent_io_tree *io_tree; | |
292 | int ret; | |
293 | int num_copies = 0; | |
294 | int mirror_num = 0; | |
295 | ||
296 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
297 | while (1) { | |
298 | ret = read_extent_buffer_pages(io_tree, eb, start, 1, | |
299 | btree_get_extent, mirror_num); | |
1259ab75 CM |
300 | if (!ret && |
301 | !verify_parent_transid(io_tree, eb, parent_transid)) | |
f188591e | 302 | return ret; |
d397712b | 303 | |
f188591e CM |
304 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
305 | eb->start, eb->len); | |
4235298e | 306 | if (num_copies == 1) |
f188591e | 307 | return ret; |
4235298e | 308 | |
f188591e | 309 | mirror_num++; |
4235298e | 310 | if (mirror_num > num_copies) |
f188591e | 311 | return ret; |
f188591e | 312 | } |
f188591e CM |
313 | return -EIO; |
314 | } | |
19c00ddc | 315 | |
d352ac68 | 316 | /* |
d397712b CM |
317 | * checksum a dirty tree block before IO. This has extra checks to make sure |
318 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 319 | */ |
d397712b | 320 | |
b2950863 | 321 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 322 | { |
d1310b2e | 323 | struct extent_io_tree *tree; |
35ebb934 | 324 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc CM |
325 | u64 found_start; |
326 | int found_level; | |
327 | unsigned long len; | |
328 | struct extent_buffer *eb; | |
f188591e CM |
329 | int ret; |
330 | ||
d1310b2e | 331 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc CM |
332 | |
333 | if (page->private == EXTENT_PAGE_PRIVATE) | |
334 | goto out; | |
335 | if (!page->private) | |
336 | goto out; | |
337 | len = page->private >> 2; | |
d397712b CM |
338 | WARN_ON(len == 0); |
339 | ||
19c00ddc | 340 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); |
ca7a79ad CM |
341 | ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, |
342 | btrfs_header_generation(eb)); | |
f188591e | 343 | BUG_ON(ret); |
19c00ddc CM |
344 | found_start = btrfs_header_bytenr(eb); |
345 | if (found_start != start) { | |
55c69072 CM |
346 | WARN_ON(1); |
347 | goto err; | |
348 | } | |
349 | if (eb->first_page != page) { | |
55c69072 CM |
350 | WARN_ON(1); |
351 | goto err; | |
352 | } | |
353 | if (!PageUptodate(page)) { | |
55c69072 CM |
354 | WARN_ON(1); |
355 | goto err; | |
19c00ddc CM |
356 | } |
357 | found_level = btrfs_header_level(eb); | |
4bef0848 | 358 | |
19c00ddc | 359 | csum_tree_block(root, eb, 0); |
55c69072 | 360 | err: |
19c00ddc CM |
361 | free_extent_buffer(eb); |
362 | out: | |
363 | return 0; | |
364 | } | |
365 | ||
2b82032c YZ |
366 | static int check_tree_block_fsid(struct btrfs_root *root, |
367 | struct extent_buffer *eb) | |
368 | { | |
369 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
370 | u8 fsid[BTRFS_UUID_SIZE]; | |
371 | int ret = 1; | |
372 | ||
373 | read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), | |
374 | BTRFS_FSID_SIZE); | |
375 | while (fs_devices) { | |
376 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
377 | ret = 0; | |
378 | break; | |
379 | } | |
380 | fs_devices = fs_devices->seed; | |
381 | } | |
382 | return ret; | |
383 | } | |
384 | ||
4008c04a CM |
385 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
386 | void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, int level) | |
387 | { | |
388 | lockdep_set_class_and_name(&eb->lock, | |
389 | &btrfs_eb_class[level], | |
390 | btrfs_eb_name[level]); | |
391 | } | |
392 | #endif | |
393 | ||
b2950863 | 394 | static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
ce9adaa5 CM |
395 | struct extent_state *state) |
396 | { | |
397 | struct extent_io_tree *tree; | |
398 | u64 found_start; | |
399 | int found_level; | |
400 | unsigned long len; | |
401 | struct extent_buffer *eb; | |
402 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 403 | int ret = 0; |
ce9adaa5 CM |
404 | |
405 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
406 | if (page->private == EXTENT_PAGE_PRIVATE) | |
407 | goto out; | |
408 | if (!page->private) | |
409 | goto out; | |
d397712b | 410 | |
ce9adaa5 | 411 | len = page->private >> 2; |
d397712b CM |
412 | WARN_ON(len == 0); |
413 | ||
ce9adaa5 | 414 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); |
f188591e | 415 | |
ce9adaa5 | 416 | found_start = btrfs_header_bytenr(eb); |
23a07867 | 417 | if (found_start != start) { |
d397712b | 418 | printk(KERN_INFO "btrfs bad tree block start %llu %llu\n", |
a1b32a59 CM |
419 | (unsigned long long)found_start, |
420 | (unsigned long long)eb->start); | |
f188591e | 421 | ret = -EIO; |
ce9adaa5 CM |
422 | goto err; |
423 | } | |
424 | if (eb->first_page != page) { | |
d397712b CM |
425 | printk(KERN_INFO "btrfs bad first page %lu %lu\n", |
426 | eb->first_page->index, page->index); | |
ce9adaa5 | 427 | WARN_ON(1); |
f188591e | 428 | ret = -EIO; |
ce9adaa5 CM |
429 | goto err; |
430 | } | |
2b82032c | 431 | if (check_tree_block_fsid(root, eb)) { |
d397712b CM |
432 | printk(KERN_INFO "btrfs bad fsid on block %llu\n", |
433 | (unsigned long long)eb->start); | |
1259ab75 CM |
434 | ret = -EIO; |
435 | goto err; | |
436 | } | |
ce9adaa5 CM |
437 | found_level = btrfs_header_level(eb); |
438 | ||
4008c04a CM |
439 | btrfs_set_buffer_lockdep_class(eb, found_level); |
440 | ||
ce9adaa5 | 441 | ret = csum_tree_block(root, eb, 1); |
f188591e CM |
442 | if (ret) |
443 | ret = -EIO; | |
ce9adaa5 CM |
444 | |
445 | end = min_t(u64, eb->len, PAGE_CACHE_SIZE); | |
446 | end = eb->start + end - 1; | |
ce9adaa5 CM |
447 | err: |
448 | free_extent_buffer(eb); | |
449 | out: | |
f188591e | 450 | return ret; |
ce9adaa5 CM |
451 | } |
452 | ||
ce9adaa5 | 453 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
454 | { |
455 | struct end_io_wq *end_io_wq = bio->bi_private; | |
456 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 457 | |
ce9adaa5 | 458 | fs_info = end_io_wq->info; |
ce9adaa5 | 459 | end_io_wq->error = err; |
8b712842 CM |
460 | end_io_wq->work.func = end_workqueue_fn; |
461 | end_io_wq->work.flags = 0; | |
d20f7043 CM |
462 | |
463 | if (bio->bi_rw & (1 << BIO_RW)) { | |
cad321ad CM |
464 | if (end_io_wq->metadata) |
465 | btrfs_queue_worker(&fs_info->endio_meta_write_workers, | |
466 | &end_io_wq->work); | |
467 | else | |
468 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
469 | &end_io_wq->work); | |
d20f7043 CM |
470 | } else { |
471 | if (end_io_wq->metadata) | |
472 | btrfs_queue_worker(&fs_info->endio_meta_workers, | |
473 | &end_io_wq->work); | |
474 | else | |
475 | btrfs_queue_worker(&fs_info->endio_workers, | |
476 | &end_io_wq->work); | |
477 | } | |
ce9adaa5 CM |
478 | } |
479 | ||
22c59948 CM |
480 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
481 | int metadata) | |
0b86a832 | 482 | { |
ce9adaa5 | 483 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
484 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
485 | if (!end_io_wq) | |
486 | return -ENOMEM; | |
487 | ||
488 | end_io_wq->private = bio->bi_private; | |
489 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 490 | end_io_wq->info = info; |
ce9adaa5 CM |
491 | end_io_wq->error = 0; |
492 | end_io_wq->bio = bio; | |
22c59948 | 493 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
494 | |
495 | bio->bi_private = end_io_wq; | |
496 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
497 | return 0; |
498 | } | |
499 | ||
b64a2851 | 500 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 501 | { |
4854ddd0 CM |
502 | unsigned long limit = min_t(unsigned long, |
503 | info->workers.max_workers, | |
504 | info->fs_devices->open_devices); | |
505 | return 256 * limit; | |
506 | } | |
0986fe9e | 507 | |
4854ddd0 CM |
508 | int btrfs_congested_async(struct btrfs_fs_info *info, int iodone) |
509 | { | |
b64a2851 CM |
510 | return atomic_read(&info->nr_async_bios) > |
511 | btrfs_async_submit_limit(info); | |
0986fe9e CM |
512 | } |
513 | ||
4a69a410 CM |
514 | static void run_one_async_start(struct btrfs_work *work) |
515 | { | |
516 | struct btrfs_fs_info *fs_info; | |
517 | struct async_submit_bio *async; | |
518 | ||
519 | async = container_of(work, struct async_submit_bio, work); | |
520 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
521 | async->submit_bio_start(async->inode, async->rw, async->bio, | |
522 | async->mirror_num, async->bio_flags); | |
523 | } | |
524 | ||
525 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
526 | { |
527 | struct btrfs_fs_info *fs_info; | |
528 | struct async_submit_bio *async; | |
4854ddd0 | 529 | int limit; |
8b712842 CM |
530 | |
531 | async = container_of(work, struct async_submit_bio, work); | |
532 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 533 | |
b64a2851 | 534 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
535 | limit = limit * 2 / 3; |
536 | ||
8b712842 | 537 | atomic_dec(&fs_info->nr_async_submits); |
0986fe9e | 538 | |
b64a2851 CM |
539 | if (atomic_read(&fs_info->nr_async_submits) < limit && |
540 | waitqueue_active(&fs_info->async_submit_wait)) | |
4854ddd0 CM |
541 | wake_up(&fs_info->async_submit_wait); |
542 | ||
4a69a410 | 543 | async->submit_bio_done(async->inode, async->rw, async->bio, |
c8b97818 | 544 | async->mirror_num, async->bio_flags); |
4a69a410 CM |
545 | } |
546 | ||
547 | static void run_one_async_free(struct btrfs_work *work) | |
548 | { | |
549 | struct async_submit_bio *async; | |
550 | ||
551 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
552 | kfree(async); |
553 | } | |
554 | ||
44b8bd7e CM |
555 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
556 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 557 | unsigned long bio_flags, |
4a69a410 CM |
558 | extent_submit_bio_hook_t *submit_bio_start, |
559 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
560 | { |
561 | struct async_submit_bio *async; | |
562 | ||
563 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
564 | if (!async) | |
565 | return -ENOMEM; | |
566 | ||
567 | async->inode = inode; | |
568 | async->rw = rw; | |
569 | async->bio = bio; | |
570 | async->mirror_num = mirror_num; | |
4a69a410 CM |
571 | async->submit_bio_start = submit_bio_start; |
572 | async->submit_bio_done = submit_bio_done; | |
573 | ||
574 | async->work.func = run_one_async_start; | |
575 | async->work.ordered_func = run_one_async_done; | |
576 | async->work.ordered_free = run_one_async_free; | |
577 | ||
8b712842 | 578 | async->work.flags = 0; |
c8b97818 | 579 | async->bio_flags = bio_flags; |
8c8bee1d | 580 | |
cb03c743 | 581 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 CM |
582 | |
583 | if (rw & (1 << BIO_RW_SYNCIO)) | |
584 | btrfs_set_work_high_prio(&async->work); | |
585 | ||
8b712842 | 586 | btrfs_queue_worker(&fs_info->workers, &async->work); |
d2c3f4f6 | 587 | #if 0 |
7c2ca468 | 588 | int limit = btrfs_async_submit_limit(fs_info); |
9473f16c CM |
589 | if (atomic_read(&fs_info->nr_async_submits) > limit) { |
590 | wait_event_timeout(fs_info->async_submit_wait, | |
4854ddd0 CM |
591 | (atomic_read(&fs_info->nr_async_submits) < limit), |
592 | HZ/10); | |
9473f16c CM |
593 | |
594 | wait_event_timeout(fs_info->async_submit_wait, | |
595 | (atomic_read(&fs_info->nr_async_bios) < limit), | |
596 | HZ/10); | |
597 | } | |
d2c3f4f6 | 598 | #endif |
d397712b | 599 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
600 | atomic_read(&fs_info->nr_async_submits)) { |
601 | wait_event(fs_info->async_submit_wait, | |
602 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
603 | } | |
604 | ||
44b8bd7e CM |
605 | return 0; |
606 | } | |
607 | ||
ce3ed71a CM |
608 | static int btree_csum_one_bio(struct bio *bio) |
609 | { | |
610 | struct bio_vec *bvec = bio->bi_io_vec; | |
611 | int bio_index = 0; | |
612 | struct btrfs_root *root; | |
613 | ||
614 | WARN_ON(bio->bi_vcnt <= 0); | |
d397712b | 615 | while (bio_index < bio->bi_vcnt) { |
ce3ed71a CM |
616 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
617 | csum_dirty_buffer(root, bvec->bv_page); | |
618 | bio_index++; | |
619 | bvec++; | |
620 | } | |
621 | return 0; | |
622 | } | |
623 | ||
4a69a410 CM |
624 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
625 | struct bio *bio, int mirror_num, | |
626 | unsigned long bio_flags) | |
22c59948 | 627 | { |
8b712842 CM |
628 | /* |
629 | * when we're called for a write, we're already in the async | |
5443be45 | 630 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 631 | */ |
4a69a410 CM |
632 | btree_csum_one_bio(bio); |
633 | return 0; | |
634 | } | |
22c59948 | 635 | |
4a69a410 CM |
636 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
637 | int mirror_num, unsigned long bio_flags) | |
638 | { | |
8b712842 | 639 | /* |
4a69a410 CM |
640 | * when we're called for a write, we're already in the async |
641 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 642 | */ |
8b712842 | 643 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
644 | } |
645 | ||
44b8bd7e | 646 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
c8b97818 | 647 | int mirror_num, unsigned long bio_flags) |
44b8bd7e | 648 | { |
cad321ad CM |
649 | int ret; |
650 | ||
651 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, | |
652 | bio, 1); | |
653 | BUG_ON(ret); | |
654 | ||
44b8bd7e | 655 | if (!(rw & (1 << BIO_RW))) { |
4a69a410 CM |
656 | /* |
657 | * called for a read, do the setup so that checksum validation | |
658 | * can happen in the async kernel threads | |
659 | */ | |
4a69a410 | 660 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, |
6f3577bd | 661 | mirror_num, 0); |
44b8bd7e | 662 | } |
d313d7a3 | 663 | |
cad321ad CM |
664 | /* |
665 | * kthread helpers are used to submit writes so that checksumming | |
666 | * can happen in parallel across all CPUs | |
667 | */ | |
44b8bd7e | 668 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, |
c8b97818 | 669 | inode, rw, bio, mirror_num, 0, |
4a69a410 CM |
670 | __btree_submit_bio_start, |
671 | __btree_submit_bio_done); | |
44b8bd7e CM |
672 | } |
673 | ||
0da5468f CM |
674 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
675 | { | |
d1310b2e | 676 | struct extent_io_tree *tree; |
b9473439 CM |
677 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
678 | struct extent_buffer *eb; | |
679 | int was_dirty; | |
680 | ||
d1310b2e | 681 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
b9473439 CM |
682 | if (!(current->flags & PF_MEMALLOC)) { |
683 | return extent_write_full_page(tree, page, | |
684 | btree_get_extent, wbc); | |
685 | } | |
5443be45 | 686 | |
b9473439 CM |
687 | redirty_page_for_writepage(wbc, page); |
688 | eb = btrfs_find_tree_block(root, page_offset(page), | |
689 | PAGE_CACHE_SIZE); | |
690 | WARN_ON(!eb); | |
691 | ||
692 | was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags); | |
693 | if (!was_dirty) { | |
694 | spin_lock(&root->fs_info->delalloc_lock); | |
695 | root->fs_info->dirty_metadata_bytes += PAGE_CACHE_SIZE; | |
696 | spin_unlock(&root->fs_info->delalloc_lock); | |
5443be45 | 697 | } |
b9473439 CM |
698 | free_extent_buffer(eb); |
699 | ||
700 | unlock_page(page); | |
701 | return 0; | |
5f39d397 | 702 | } |
0da5468f CM |
703 | |
704 | static int btree_writepages(struct address_space *mapping, | |
705 | struct writeback_control *wbc) | |
706 | { | |
d1310b2e CM |
707 | struct extent_io_tree *tree; |
708 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 709 | if (wbc->sync_mode == WB_SYNC_NONE) { |
b9473439 | 710 | struct btrfs_root *root = BTRFS_I(mapping->host)->root; |
793955bc | 711 | u64 num_dirty; |
24ab9cd8 | 712 | unsigned long thresh = 32 * 1024 * 1024; |
448d640b CM |
713 | |
714 | if (wbc->for_kupdate) | |
715 | return 0; | |
716 | ||
b9473439 CM |
717 | /* this is a bit racy, but that's ok */ |
718 | num_dirty = root->fs_info->dirty_metadata_bytes; | |
d397712b | 719 | if (num_dirty < thresh) |
793955bc | 720 | return 0; |
793955bc | 721 | } |
0da5468f CM |
722 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
723 | } | |
724 | ||
b2950863 | 725 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 726 | { |
d1310b2e CM |
727 | struct extent_io_tree *tree; |
728 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
729 | return extent_read_full_page(tree, page, btree_get_extent); |
730 | } | |
22b0ebda | 731 | |
70dec807 | 732 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 733 | { |
d1310b2e CM |
734 | struct extent_io_tree *tree; |
735 | struct extent_map_tree *map; | |
5f39d397 | 736 | int ret; |
d98237b3 | 737 | |
98509cfc | 738 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 739 | return 0; |
98509cfc | 740 | |
d1310b2e CM |
741 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
742 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
6af118ce | 743 | |
7b13b7b1 | 744 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
d397712b | 745 | if (!ret) |
6af118ce | 746 | return 0; |
6af118ce CM |
747 | |
748 | ret = try_release_extent_buffer(tree, page); | |
5f39d397 CM |
749 | if (ret == 1) { |
750 | ClearPagePrivate(page); | |
751 | set_page_private(page, 0); | |
752 | page_cache_release(page); | |
753 | } | |
6af118ce | 754 | |
d98237b3 CM |
755 | return ret; |
756 | } | |
757 | ||
5f39d397 | 758 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 759 | { |
d1310b2e CM |
760 | struct extent_io_tree *tree; |
761 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
762 | extent_invalidatepage(tree, page, offset); |
763 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 764 | if (PagePrivate(page)) { |
d397712b CM |
765 | printk(KERN_WARNING "btrfs warning page private not zero " |
766 | "on page %llu\n", (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
767 | ClearPagePrivate(page); |
768 | set_page_private(page, 0); | |
769 | page_cache_release(page); | |
770 | } | |
d98237b3 CM |
771 | } |
772 | ||
5f39d397 | 773 | #if 0 |
d98237b3 | 774 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
ed2ff2cb | 775 | { |
87cbda5c | 776 | struct buffer_head *bh; |
0f7d52f4 | 777 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
87cbda5c | 778 | struct buffer_head *head; |
87cbda5c CM |
779 | if (!page_has_buffers(page)) { |
780 | create_empty_buffers(page, root->fs_info->sb->s_blocksize, | |
781 | (1 << BH_Dirty)|(1 << BH_Uptodate)); | |
782 | } | |
783 | head = page_buffers(page); | |
784 | bh = head; | |
785 | do { | |
786 | if (buffer_dirty(bh)) | |
787 | csum_tree_block(root, bh, 0); | |
788 | bh = bh->b_this_page; | |
789 | } while (bh != head); | |
d98237b3 | 790 | return block_write_full_page(page, btree_get_block, wbc); |
ed2ff2cb | 791 | } |
5f39d397 | 792 | #endif |
eb60ceac | 793 | |
d98237b3 CM |
794 | static struct address_space_operations btree_aops = { |
795 | .readpage = btree_readpage, | |
796 | .writepage = btree_writepage, | |
0da5468f | 797 | .writepages = btree_writepages, |
5f39d397 CM |
798 | .releasepage = btree_releasepage, |
799 | .invalidatepage = btree_invalidatepage, | |
d98237b3 CM |
800 | .sync_page = block_sync_page, |
801 | }; | |
802 | ||
ca7a79ad CM |
803 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
804 | u64 parent_transid) | |
090d1875 | 805 | { |
5f39d397 CM |
806 | struct extent_buffer *buf = NULL; |
807 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 808 | int ret = 0; |
090d1875 | 809 | |
db94535d | 810 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 811 | if (!buf) |
090d1875 | 812 | return 0; |
d1310b2e | 813 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
f188591e | 814 | buf, 0, 0, btree_get_extent, 0); |
5f39d397 | 815 | free_extent_buffer(buf); |
de428b63 | 816 | return ret; |
090d1875 CM |
817 | } |
818 | ||
0999df54 CM |
819 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
820 | u64 bytenr, u32 blocksize) | |
821 | { | |
822 | struct inode *btree_inode = root->fs_info->btree_inode; | |
823 | struct extent_buffer *eb; | |
824 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
825 | bytenr, blocksize, GFP_NOFS); | |
826 | return eb; | |
827 | } | |
828 | ||
829 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
830 | u64 bytenr, u32 blocksize) | |
831 | { | |
832 | struct inode *btree_inode = root->fs_info->btree_inode; | |
833 | struct extent_buffer *eb; | |
834 | ||
835 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
836 | bytenr, blocksize, NULL, GFP_NOFS); | |
837 | return eb; | |
838 | } | |
839 | ||
840 | ||
e02119d5 CM |
841 | int btrfs_write_tree_block(struct extent_buffer *buf) |
842 | { | |
843 | return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start, | |
24ab9cd8 | 844 | buf->start + buf->len - 1, WB_SYNC_ALL); |
e02119d5 CM |
845 | } |
846 | ||
847 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
848 | { | |
849 | return btrfs_wait_on_page_writeback_range(buf->first_page->mapping, | |
d397712b | 850 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
851 | } |
852 | ||
0999df54 | 853 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 854 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
855 | { |
856 | struct extent_buffer *buf = NULL; | |
857 | struct inode *btree_inode = root->fs_info->btree_inode; | |
858 | struct extent_io_tree *io_tree; | |
859 | int ret; | |
860 | ||
861 | io_tree = &BTRFS_I(btree_inode)->io_tree; | |
862 | ||
863 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
864 | if (!buf) | |
865 | return NULL; | |
0999df54 | 866 | |
ca7a79ad | 867 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 | 868 | |
d397712b | 869 | if (ret == 0) |
b4ce94de | 870 | set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags); |
d397712b | 871 | else |
a1b32a59 | 872 | WARN_ON(1); |
5f39d397 | 873 | return buf; |
ce9adaa5 | 874 | |
eb60ceac CM |
875 | } |
876 | ||
e089f05c | 877 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 878 | struct extent_buffer *buf) |
ed2ff2cb | 879 | { |
5f39d397 | 880 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 | 881 | if (btrfs_header_generation(buf) == |
925baedd | 882 | root->fs_info->running_transaction->transid) { |
b9447ef8 | 883 | btrfs_assert_tree_locked(buf); |
b4ce94de | 884 | |
b9473439 CM |
885 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
886 | spin_lock(&root->fs_info->delalloc_lock); | |
887 | if (root->fs_info->dirty_metadata_bytes >= buf->len) | |
888 | root->fs_info->dirty_metadata_bytes -= buf->len; | |
889 | else | |
890 | WARN_ON(1); | |
891 | spin_unlock(&root->fs_info->delalloc_lock); | |
892 | } | |
b4ce94de | 893 | |
b9473439 CM |
894 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
895 | btrfs_set_lock_blocking(buf); | |
d1310b2e | 896 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 897 | buf); |
925baedd | 898 | } |
5f39d397 CM |
899 | return 0; |
900 | } | |
901 | ||
db94535d | 902 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 903 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 904 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 905 | u64 objectid) |
d97e63b6 | 906 | { |
cfaa7295 | 907 | root->node = NULL; |
a28ec197 | 908 | root->commit_root = NULL; |
31153d81 | 909 | root->ref_tree = NULL; |
db94535d CM |
910 | root->sectorsize = sectorsize; |
911 | root->nodesize = nodesize; | |
912 | root->leafsize = leafsize; | |
87ee04eb | 913 | root->stripesize = stripesize; |
123abc88 | 914 | root->ref_cows = 0; |
0b86a832 CM |
915 | root->track_dirty = 0; |
916 | ||
9f5fae2f | 917 | root->fs_info = fs_info; |
0f7d52f4 CM |
918 | root->objectid = objectid; |
919 | root->last_trans = 0; | |
1b05da2e CM |
920 | root->highest_inode = 0; |
921 | root->last_inode_alloc = 0; | |
58176a96 | 922 | root->name = NULL; |
4313b399 | 923 | root->in_sysfs = 0; |
0b86a832 CM |
924 | |
925 | INIT_LIST_HEAD(&root->dirty_list); | |
7b128766 | 926 | INIT_LIST_HEAD(&root->orphan_list); |
bcc63abb | 927 | INIT_LIST_HEAD(&root->dead_list); |
925baedd | 928 | spin_lock_init(&root->node_lock); |
bcc63abb | 929 | spin_lock_init(&root->list_lock); |
a2135011 | 930 | mutex_init(&root->objectid_mutex); |
e02119d5 | 931 | mutex_init(&root->log_mutex); |
7237f183 YZ |
932 | init_waitqueue_head(&root->log_writer_wait); |
933 | init_waitqueue_head(&root->log_commit_wait[0]); | |
934 | init_waitqueue_head(&root->log_commit_wait[1]); | |
935 | atomic_set(&root->log_commit[0], 0); | |
936 | atomic_set(&root->log_commit[1], 0); | |
937 | atomic_set(&root->log_writers, 0); | |
938 | root->log_batch = 0; | |
939 | root->log_transid = 0; | |
d0c803c4 CM |
940 | extent_io_tree_init(&root->dirty_log_pages, |
941 | fs_info->btree_inode->i_mapping, GFP_NOFS); | |
017e5369 CM |
942 | |
943 | btrfs_leaf_ref_tree_init(&root->ref_tree_struct); | |
944 | root->ref_tree = &root->ref_tree_struct; | |
945 | ||
3768f368 CM |
946 | memset(&root->root_key, 0, sizeof(root->root_key)); |
947 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 948 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 949 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 950 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 951 | init_completion(&root->kobj_unregister); |
6702ed49 CM |
952 | root->defrag_running = 0; |
953 | root->defrag_level = 0; | |
4d775673 | 954 | root->root_key.objectid = objectid; |
3394e160 CM |
955 | root->anon_super.s_root = NULL; |
956 | root->anon_super.s_dev = 0; | |
957 | INIT_LIST_HEAD(&root->anon_super.s_list); | |
958 | INIT_LIST_HEAD(&root->anon_super.s_instances); | |
959 | init_rwsem(&root->anon_super.s_umount); | |
960 | ||
3768f368 CM |
961 | return 0; |
962 | } | |
963 | ||
db94535d | 964 | static int find_and_setup_root(struct btrfs_root *tree_root, |
9f5fae2f CM |
965 | struct btrfs_fs_info *fs_info, |
966 | u64 objectid, | |
e20d96d6 | 967 | struct btrfs_root *root) |
3768f368 CM |
968 | { |
969 | int ret; | |
db94535d | 970 | u32 blocksize; |
84234f3a | 971 | u64 generation; |
3768f368 | 972 | |
db94535d | 973 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
974 | tree_root->sectorsize, tree_root->stripesize, |
975 | root, fs_info, objectid); | |
3768f368 CM |
976 | ret = btrfs_find_last_root(tree_root, objectid, |
977 | &root->root_item, &root->root_key); | |
978 | BUG_ON(ret); | |
979 | ||
84234f3a | 980 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
981 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
982 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 983 | blocksize, generation); |
3768f368 | 984 | BUG_ON(!root->node); |
d97e63b6 CM |
985 | return 0; |
986 | } | |
987 | ||
e02119d5 CM |
988 | int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans, |
989 | struct btrfs_fs_info *fs_info) | |
990 | { | |
991 | struct extent_buffer *eb; | |
d0c803c4 CM |
992 | struct btrfs_root *log_root_tree = fs_info->log_root_tree; |
993 | u64 start = 0; | |
994 | u64 end = 0; | |
e02119d5 CM |
995 | int ret; |
996 | ||
d0c803c4 | 997 | if (!log_root_tree) |
e02119d5 CM |
998 | return 0; |
999 | ||
d397712b | 1000 | while (1) { |
d0c803c4 CM |
1001 | ret = find_first_extent_bit(&log_root_tree->dirty_log_pages, |
1002 | 0, &start, &end, EXTENT_DIRTY); | |
1003 | if (ret) | |
1004 | break; | |
1005 | ||
1006 | clear_extent_dirty(&log_root_tree->dirty_log_pages, | |
1007 | start, end, GFP_NOFS); | |
1008 | } | |
e02119d5 CM |
1009 | eb = fs_info->log_root_tree->node; |
1010 | ||
1011 | WARN_ON(btrfs_header_level(eb) != 0); | |
1012 | WARN_ON(btrfs_header_nritems(eb) != 0); | |
1013 | ||
d00aff00 CM |
1014 | ret = btrfs_free_reserved_extent(fs_info->tree_root, |
1015 | eb->start, eb->len); | |
e02119d5 CM |
1016 | BUG_ON(ret); |
1017 | ||
1018 | free_extent_buffer(eb); | |
1019 | kfree(fs_info->log_root_tree); | |
1020 | fs_info->log_root_tree = NULL; | |
1021 | return 0; | |
1022 | } | |
1023 | ||
7237f183 YZ |
1024 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1025 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1026 | { |
1027 | struct btrfs_root *root; | |
1028 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1029 | struct extent_buffer *leaf; |
e02119d5 CM |
1030 | |
1031 | root = kzalloc(sizeof(*root), GFP_NOFS); | |
1032 | if (!root) | |
7237f183 | 1033 | return ERR_PTR(-ENOMEM); |
e02119d5 CM |
1034 | |
1035 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1036 | tree_root->sectorsize, tree_root->stripesize, | |
1037 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1038 | ||
1039 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1040 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1041 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
7237f183 YZ |
1042 | /* |
1043 | * log trees do not get reference counted because they go away | |
1044 | * before a real commit is actually done. They do store pointers | |
1045 | * to file data extents, and those reference counts still get | |
1046 | * updated (along with back refs to the log tree). | |
1047 | */ | |
e02119d5 CM |
1048 | root->ref_cows = 0; |
1049 | ||
7237f183 YZ |
1050 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, |
1051 | 0, BTRFS_TREE_LOG_OBJECTID, | |
1052 | trans->transid, 0, 0, 0); | |
1053 | if (IS_ERR(leaf)) { | |
1054 | kfree(root); | |
1055 | return ERR_CAST(leaf); | |
1056 | } | |
e02119d5 | 1057 | |
7237f183 | 1058 | root->node = leaf; |
e02119d5 CM |
1059 | btrfs_set_header_nritems(root->node, 0); |
1060 | btrfs_set_header_level(root->node, 0); | |
1061 | btrfs_set_header_bytenr(root->node, root->node->start); | |
1062 | btrfs_set_header_generation(root->node, trans->transid); | |
1063 | btrfs_set_header_owner(root->node, BTRFS_TREE_LOG_OBJECTID); | |
1064 | ||
1065 | write_extent_buffer(root->node, root->fs_info->fsid, | |
1066 | (unsigned long)btrfs_header_fsid(root->node), | |
1067 | BTRFS_FSID_SIZE); | |
1068 | btrfs_mark_buffer_dirty(root->node); | |
1069 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1070 | return root; |
1071 | } | |
1072 | ||
1073 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1074 | struct btrfs_fs_info *fs_info) | |
1075 | { | |
1076 | struct btrfs_root *log_root; | |
1077 | ||
1078 | log_root = alloc_log_tree(trans, fs_info); | |
1079 | if (IS_ERR(log_root)) | |
1080 | return PTR_ERR(log_root); | |
1081 | WARN_ON(fs_info->log_root_tree); | |
1082 | fs_info->log_root_tree = log_root; | |
1083 | return 0; | |
1084 | } | |
1085 | ||
1086 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1087 | struct btrfs_root *root) | |
1088 | { | |
1089 | struct btrfs_root *log_root; | |
1090 | struct btrfs_inode_item *inode_item; | |
1091 | ||
1092 | log_root = alloc_log_tree(trans, root->fs_info); | |
1093 | if (IS_ERR(log_root)) | |
1094 | return PTR_ERR(log_root); | |
1095 | ||
1096 | log_root->last_trans = trans->transid; | |
1097 | log_root->root_key.offset = root->root_key.objectid; | |
1098 | ||
1099 | inode_item = &log_root->root_item.inode; | |
1100 | inode_item->generation = cpu_to_le64(1); | |
1101 | inode_item->size = cpu_to_le64(3); | |
1102 | inode_item->nlink = cpu_to_le32(1); | |
1103 | inode_item->nbytes = cpu_to_le64(root->leafsize); | |
1104 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); | |
1105 | ||
1106 | btrfs_set_root_bytenr(&log_root->root_item, log_root->node->start); | |
1107 | btrfs_set_root_generation(&log_root->root_item, trans->transid); | |
1108 | ||
1109 | WARN_ON(root->log_root); | |
1110 | root->log_root = log_root; | |
1111 | root->log_transid = 0; | |
e02119d5 CM |
1112 | return 0; |
1113 | } | |
1114 | ||
1115 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
1116 | struct btrfs_key *location) | |
1117 | { | |
1118 | struct btrfs_root *root; | |
1119 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1120 | struct btrfs_path *path; |
5f39d397 | 1121 | struct extent_buffer *l; |
1b05da2e | 1122 | u64 highest_inode; |
84234f3a | 1123 | u64 generation; |
db94535d | 1124 | u32 blocksize; |
0f7d52f4 CM |
1125 | int ret = 0; |
1126 | ||
5eda7b5e | 1127 | root = kzalloc(sizeof(*root), GFP_NOFS); |
0cf6c620 | 1128 | if (!root) |
0f7d52f4 | 1129 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 1130 | if (location->offset == (u64)-1) { |
db94535d | 1131 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
1132 | location->objectid, root); |
1133 | if (ret) { | |
0f7d52f4 CM |
1134 | kfree(root); |
1135 | return ERR_PTR(ret); | |
1136 | } | |
1137 | goto insert; | |
1138 | } | |
1139 | ||
db94535d | 1140 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1141 | tree_root->sectorsize, tree_root->stripesize, |
1142 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
1143 | |
1144 | path = btrfs_alloc_path(); | |
1145 | BUG_ON(!path); | |
1146 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); | |
1147 | if (ret != 0) { | |
0f7d52f4 CM |
1148 | if (ret > 0) |
1149 | ret = -ENOENT; | |
1150 | goto out; | |
1151 | } | |
5f39d397 CM |
1152 | l = path->nodes[0]; |
1153 | read_extent_buffer(l, &root->root_item, | |
1154 | btrfs_item_ptr_offset(l, path->slots[0]), | |
0f7d52f4 | 1155 | sizeof(root->root_item)); |
44b36eb2 | 1156 | memcpy(&root->root_key, location, sizeof(*location)); |
0f7d52f4 CM |
1157 | ret = 0; |
1158 | out: | |
1159 | btrfs_release_path(root, path); | |
1160 | btrfs_free_path(path); | |
1161 | if (ret) { | |
1162 | kfree(root); | |
1163 | return ERR_PTR(ret); | |
1164 | } | |
84234f3a | 1165 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1166 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1167 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1168 | blocksize, generation); |
0f7d52f4 CM |
1169 | BUG_ON(!root->node); |
1170 | insert: | |
e02119d5 CM |
1171 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { |
1172 | root->ref_cows = 1; | |
1173 | ret = btrfs_find_highest_inode(root, &highest_inode); | |
1174 | if (ret == 0) { | |
1175 | root->highest_inode = highest_inode; | |
1176 | root->last_inode_alloc = highest_inode; | |
1177 | } | |
5eda7b5e CM |
1178 | } |
1179 | return root; | |
1180 | } | |
1181 | ||
dc17ff8f CM |
1182 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1183 | u64 root_objectid) | |
1184 | { | |
1185 | struct btrfs_root *root; | |
1186 | ||
1187 | if (root_objectid == BTRFS_ROOT_TREE_OBJECTID) | |
1188 | return fs_info->tree_root; | |
1189 | if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1190 | return fs_info->extent_root; | |
1191 | ||
1192 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1193 | (unsigned long)root_objectid); | |
1194 | return root; | |
1195 | } | |
1196 | ||
edbd8d4e CM |
1197 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
1198 | struct btrfs_key *location) | |
5eda7b5e CM |
1199 | { |
1200 | struct btrfs_root *root; | |
1201 | int ret; | |
1202 | ||
edbd8d4e CM |
1203 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1204 | return fs_info->tree_root; | |
1205 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1206 | return fs_info->extent_root; | |
8f18cf13 CM |
1207 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1208 | return fs_info->chunk_root; | |
1209 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1210 | return fs_info->dev_root; | |
0403e47e YZ |
1211 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1212 | return fs_info->csum_root; | |
edbd8d4e | 1213 | |
5eda7b5e CM |
1214 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
1215 | (unsigned long)location->objectid); | |
1216 | if (root) | |
1217 | return root; | |
1218 | ||
e02119d5 | 1219 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
1220 | if (IS_ERR(root)) |
1221 | return root; | |
3394e160 CM |
1222 | |
1223 | set_anon_super(&root->anon_super, NULL); | |
1224 | ||
2619ba1f CM |
1225 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
1226 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 CM |
1227 | root); |
1228 | if (ret) { | |
5f39d397 | 1229 | free_extent_buffer(root->node); |
0f7d52f4 CM |
1230 | kfree(root); |
1231 | return ERR_PTR(ret); | |
1232 | } | |
c146afad YZ |
1233 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
1234 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
1235 | root->root_key.objectid, root); | |
1236 | BUG_ON(ret); | |
1237 | btrfs_orphan_cleanup(root); | |
1238 | } | |
edbd8d4e CM |
1239 | return root; |
1240 | } | |
1241 | ||
1242 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, | |
1243 | struct btrfs_key *location, | |
1244 | const char *name, int namelen) | |
1245 | { | |
1246 | struct btrfs_root *root; | |
1247 | int ret; | |
1248 | ||
1249 | root = btrfs_read_fs_root_no_name(fs_info, location); | |
1250 | if (!root) | |
1251 | return NULL; | |
58176a96 | 1252 | |
4313b399 CM |
1253 | if (root->in_sysfs) |
1254 | return root; | |
1255 | ||
58176a96 JB |
1256 | ret = btrfs_set_root_name(root, name, namelen); |
1257 | if (ret) { | |
5f39d397 | 1258 | free_extent_buffer(root->node); |
58176a96 JB |
1259 | kfree(root); |
1260 | return ERR_PTR(ret); | |
1261 | } | |
ea9e8b11 | 1262 | #if 0 |
58176a96 JB |
1263 | ret = btrfs_sysfs_add_root(root); |
1264 | if (ret) { | |
5f39d397 | 1265 | free_extent_buffer(root->node); |
58176a96 JB |
1266 | kfree(root->name); |
1267 | kfree(root); | |
1268 | return ERR_PTR(ret); | |
1269 | } | |
ea9e8b11 | 1270 | #endif |
4313b399 | 1271 | root->in_sysfs = 1; |
0f7d52f4 CM |
1272 | return root; |
1273 | } | |
04160088 CM |
1274 | |
1275 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) | |
1276 | { | |
1277 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1278 | int ret = 0; | |
04160088 CM |
1279 | struct btrfs_device *device; |
1280 | struct backing_dev_info *bdi; | |
6f3577bd | 1281 | #if 0 |
cb03c743 | 1282 | if ((bdi_bits & (1 << BDI_write_congested)) && |
777e6bd7 | 1283 | btrfs_congested_async(info, 0)) |
cb03c743 | 1284 | return 1; |
6f3577bd | 1285 | #endif |
c6e30871 | 1286 | list_for_each_entry(device, &info->fs_devices->devices, dev_list) { |
dfe25020 CM |
1287 | if (!device->bdev) |
1288 | continue; | |
04160088 CM |
1289 | bdi = blk_get_backing_dev_info(device->bdev); |
1290 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1291 | ret = 1; | |
1292 | break; | |
1293 | } | |
1294 | } | |
1295 | return ret; | |
1296 | } | |
1297 | ||
38b66988 CM |
1298 | /* |
1299 | * this unplugs every device on the box, and it is only used when page | |
1300 | * is null | |
1301 | */ | |
1302 | static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) | |
1303 | { | |
38b66988 CM |
1304 | struct btrfs_device *device; |
1305 | struct btrfs_fs_info *info; | |
1306 | ||
1307 | info = (struct btrfs_fs_info *)bdi->unplug_io_data; | |
c6e30871 | 1308 | list_for_each_entry(device, &info->fs_devices->devices, dev_list) { |
d20f7043 CM |
1309 | if (!device->bdev) |
1310 | continue; | |
1311 | ||
38b66988 | 1312 | bdi = blk_get_backing_dev_info(device->bdev); |
d397712b | 1313 | if (bdi->unplug_io_fn) |
38b66988 | 1314 | bdi->unplug_io_fn(bdi, page); |
38b66988 CM |
1315 | } |
1316 | } | |
1317 | ||
b2950863 | 1318 | static void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) |
04160088 | 1319 | { |
38b66988 | 1320 | struct inode *inode; |
f2d8d74d CM |
1321 | struct extent_map_tree *em_tree; |
1322 | struct extent_map *em; | |
bcbfce8a | 1323 | struct address_space *mapping; |
38b66988 CM |
1324 | u64 offset; |
1325 | ||
bcbfce8a | 1326 | /* the generic O_DIRECT read code does this */ |
9f0ba5bd | 1327 | if (1 || !page) { |
38b66988 CM |
1328 | __unplug_io_fn(bdi, page); |
1329 | return; | |
1330 | } | |
1331 | ||
bcbfce8a CM |
1332 | /* |
1333 | * page->mapping may change at any time. Get a consistent copy | |
1334 | * and use that for everything below | |
1335 | */ | |
1336 | smp_mb(); | |
1337 | mapping = page->mapping; | |
1338 | if (!mapping) | |
1339 | return; | |
1340 | ||
1341 | inode = mapping->host; | |
240d5d48 CM |
1342 | |
1343 | /* | |
1344 | * don't do the expensive searching for a small number of | |
1345 | * devices | |
1346 | */ | |
1347 | if (BTRFS_I(inode)->root->fs_info->fs_devices->open_devices <= 2) { | |
1348 | __unplug_io_fn(bdi, page); | |
1349 | return; | |
1350 | } | |
1351 | ||
38b66988 | 1352 | offset = page_offset(page); |
04160088 | 1353 | |
f2d8d74d CM |
1354 | em_tree = &BTRFS_I(inode)->extent_tree; |
1355 | spin_lock(&em_tree->lock); | |
1356 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); | |
1357 | spin_unlock(&em_tree->lock); | |
89642229 CM |
1358 | if (!em) { |
1359 | __unplug_io_fn(bdi, page); | |
f2d8d74d | 1360 | return; |
89642229 | 1361 | } |
f2d8d74d | 1362 | |
89642229 CM |
1363 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { |
1364 | free_extent_map(em); | |
1365 | __unplug_io_fn(bdi, page); | |
1366 | return; | |
1367 | } | |
f2d8d74d CM |
1368 | offset = offset - em->start; |
1369 | btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree, | |
1370 | em->block_start + offset, page); | |
1371 | free_extent_map(em); | |
04160088 CM |
1372 | } |
1373 | ||
1374 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) | |
1375 | { | |
1376 | bdi_init(bdi); | |
4575c9cc | 1377 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1378 | bdi->state = 0; |
1379 | bdi->capabilities = default_backing_dev_info.capabilities; | |
1380 | bdi->unplug_io_fn = btrfs_unplug_io_fn; | |
1381 | bdi->unplug_io_data = info; | |
1382 | bdi->congested_fn = btrfs_congested_fn; | |
1383 | bdi->congested_data = info; | |
1384 | return 0; | |
1385 | } | |
1386 | ||
ce9adaa5 CM |
1387 | static int bio_ready_for_csum(struct bio *bio) |
1388 | { | |
1389 | u64 length = 0; | |
1390 | u64 buf_len = 0; | |
1391 | u64 start = 0; | |
1392 | struct page *page; | |
1393 | struct extent_io_tree *io_tree = NULL; | |
1394 | struct btrfs_fs_info *info = NULL; | |
1395 | struct bio_vec *bvec; | |
1396 | int i; | |
1397 | int ret; | |
1398 | ||
1399 | bio_for_each_segment(bvec, bio, i) { | |
1400 | page = bvec->bv_page; | |
1401 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
1402 | length += bvec->bv_len; | |
1403 | continue; | |
1404 | } | |
1405 | if (!page->private) { | |
1406 | length += bvec->bv_len; | |
1407 | continue; | |
1408 | } | |
1409 | length = bvec->bv_len; | |
1410 | buf_len = page->private >> 2; | |
1411 | start = page_offset(page) + bvec->bv_offset; | |
1412 | io_tree = &BTRFS_I(page->mapping->host)->io_tree; | |
1413 | info = BTRFS_I(page->mapping->host)->root->fs_info; | |
1414 | } | |
1415 | /* are we fully contained in this bio? */ | |
1416 | if (buf_len <= length) | |
1417 | return 1; | |
1418 | ||
1419 | ret = extent_range_uptodate(io_tree, start + length, | |
1420 | start + buf_len - 1); | |
ce9adaa5 CM |
1421 | return ret; |
1422 | } | |
1423 | ||
8b712842 CM |
1424 | /* |
1425 | * called by the kthread helper functions to finally call the bio end_io | |
1426 | * functions. This is where read checksum verification actually happens | |
1427 | */ | |
1428 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1429 | { |
ce9adaa5 | 1430 | struct bio *bio; |
8b712842 CM |
1431 | struct end_io_wq *end_io_wq; |
1432 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1433 | int error; |
ce9adaa5 | 1434 | |
8b712842 CM |
1435 | end_io_wq = container_of(work, struct end_io_wq, work); |
1436 | bio = end_io_wq->bio; | |
1437 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1438 | |
cad321ad | 1439 | /* metadata bio reads are special because the whole tree block must |
8b712842 CM |
1440 | * be checksummed at once. This makes sure the entire block is in |
1441 | * ram and up to date before trying to verify things. For | |
1442 | * blocksize <= pagesize, it is basically a noop | |
1443 | */ | |
cad321ad CM |
1444 | if (!(bio->bi_rw & (1 << BIO_RW)) && end_io_wq->metadata && |
1445 | !bio_ready_for_csum(bio)) { | |
d20f7043 | 1446 | btrfs_queue_worker(&fs_info->endio_meta_workers, |
8b712842 CM |
1447 | &end_io_wq->work); |
1448 | return; | |
1449 | } | |
1450 | error = end_io_wq->error; | |
1451 | bio->bi_private = end_io_wq->private; | |
1452 | bio->bi_end_io = end_io_wq->end_io; | |
1453 | kfree(end_io_wq); | |
8b712842 | 1454 | bio_endio(bio, error); |
44b8bd7e CM |
1455 | } |
1456 | ||
a74a4b97 CM |
1457 | static int cleaner_kthread(void *arg) |
1458 | { | |
1459 | struct btrfs_root *root = arg; | |
1460 | ||
1461 | do { | |
1462 | smp_mb(); | |
1463 | if (root->fs_info->closing) | |
1464 | break; | |
1465 | ||
1466 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1467 | mutex_lock(&root->fs_info->cleaner_mutex); | |
a74a4b97 | 1468 | btrfs_clean_old_snapshots(root); |
a74a4b97 CM |
1469 | mutex_unlock(&root->fs_info->cleaner_mutex); |
1470 | ||
1471 | if (freezing(current)) { | |
1472 | refrigerator(); | |
1473 | } else { | |
1474 | smp_mb(); | |
1475 | if (root->fs_info->closing) | |
1476 | break; | |
1477 | set_current_state(TASK_INTERRUPTIBLE); | |
1478 | schedule(); | |
1479 | __set_current_state(TASK_RUNNING); | |
1480 | } | |
1481 | } while (!kthread_should_stop()); | |
1482 | return 0; | |
1483 | } | |
1484 | ||
1485 | static int transaction_kthread(void *arg) | |
1486 | { | |
1487 | struct btrfs_root *root = arg; | |
1488 | struct btrfs_trans_handle *trans; | |
1489 | struct btrfs_transaction *cur; | |
1490 | unsigned long now; | |
1491 | unsigned long delay; | |
1492 | int ret; | |
1493 | ||
1494 | do { | |
1495 | smp_mb(); | |
1496 | if (root->fs_info->closing) | |
1497 | break; | |
1498 | ||
1499 | delay = HZ * 30; | |
1500 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1501 | mutex_lock(&root->fs_info->transaction_kthread_mutex); | |
1502 | ||
1503 | mutex_lock(&root->fs_info->trans_mutex); | |
1504 | cur = root->fs_info->running_transaction; | |
1505 | if (!cur) { | |
1506 | mutex_unlock(&root->fs_info->trans_mutex); | |
1507 | goto sleep; | |
1508 | } | |
31153d81 | 1509 | |
a74a4b97 CM |
1510 | now = get_seconds(); |
1511 | if (now < cur->start_time || now - cur->start_time < 30) { | |
1512 | mutex_unlock(&root->fs_info->trans_mutex); | |
1513 | delay = HZ * 5; | |
1514 | goto sleep; | |
1515 | } | |
1516 | mutex_unlock(&root->fs_info->trans_mutex); | |
a74a4b97 CM |
1517 | trans = btrfs_start_transaction(root, 1); |
1518 | ret = btrfs_commit_transaction(trans, root); | |
56bec294 | 1519 | |
a74a4b97 CM |
1520 | sleep: |
1521 | wake_up_process(root->fs_info->cleaner_kthread); | |
1522 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1523 | ||
1524 | if (freezing(current)) { | |
1525 | refrigerator(); | |
1526 | } else { | |
1527 | if (root->fs_info->closing) | |
1528 | break; | |
1529 | set_current_state(TASK_INTERRUPTIBLE); | |
1530 | schedule_timeout(delay); | |
1531 | __set_current_state(TASK_RUNNING); | |
1532 | } | |
1533 | } while (!kthread_should_stop()); | |
1534 | return 0; | |
1535 | } | |
1536 | ||
8a4b83cc | 1537 | struct btrfs_root *open_ctree(struct super_block *sb, |
dfe25020 CM |
1538 | struct btrfs_fs_devices *fs_devices, |
1539 | char *options) | |
2e635a27 | 1540 | { |
db94535d CM |
1541 | u32 sectorsize; |
1542 | u32 nodesize; | |
1543 | u32 leafsize; | |
1544 | u32 blocksize; | |
87ee04eb | 1545 | u32 stripesize; |
84234f3a | 1546 | u64 generation; |
f2b636e8 | 1547 | u64 features; |
3de4586c | 1548 | struct btrfs_key location; |
a061fc8d | 1549 | struct buffer_head *bh; |
e02119d5 | 1550 | struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1551 | GFP_NOFS); |
d20f7043 CM |
1552 | struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root), |
1553 | GFP_NOFS); | |
e02119d5 | 1554 | struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1555 | GFP_NOFS); |
8790d502 | 1556 | struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), |
e20d96d6 | 1557 | GFP_NOFS); |
e02119d5 | 1558 | struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1559 | GFP_NOFS); |
e02119d5 | 1560 | struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1561 | GFP_NOFS); |
e02119d5 CM |
1562 | struct btrfs_root *log_tree_root; |
1563 | ||
eb60ceac | 1564 | int ret; |
e58ca020 | 1565 | int err = -EINVAL; |
4543df7e | 1566 | |
2c90e5d6 | 1567 | struct btrfs_super_block *disk_super; |
8790d502 | 1568 | |
0463bb4e | 1569 | if (!extent_root || !tree_root || !fs_info || |
d20f7043 | 1570 | !chunk_root || !dev_root || !csum_root) { |
39279cc3 CM |
1571 | err = -ENOMEM; |
1572 | goto fail; | |
1573 | } | |
0f7d52f4 | 1574 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS); |
8fd17795 | 1575 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1576 | INIT_LIST_HEAD(&fs_info->dead_roots); |
19c00ddc | 1577 | INIT_LIST_HEAD(&fs_info->hashers); |
ea8c2819 | 1578 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
5a3f23d5 | 1579 | INIT_LIST_HEAD(&fs_info->ordered_operations); |
1832a6d5 | 1580 | spin_lock_init(&fs_info->delalloc_lock); |
cee36a03 | 1581 | spin_lock_init(&fs_info->new_trans_lock); |
31153d81 | 1582 | spin_lock_init(&fs_info->ref_cache_lock); |
19c00ddc | 1583 | |
58176a96 | 1584 | init_completion(&fs_info->kobj_unregister); |
9f5fae2f CM |
1585 | fs_info->tree_root = tree_root; |
1586 | fs_info->extent_root = extent_root; | |
d20f7043 | 1587 | fs_info->csum_root = csum_root; |
0b86a832 CM |
1588 | fs_info->chunk_root = chunk_root; |
1589 | fs_info->dev_root = dev_root; | |
8a4b83cc | 1590 | fs_info->fs_devices = fs_devices; |
0b86a832 | 1591 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1592 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1593 | btrfs_mapping_init(&fs_info->mapping_tree); |
cb03c743 | 1594 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 1595 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 1596 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 1597 | atomic_set(&fs_info->nr_async_bios, 0); |
a2135011 | 1598 | atomic_set(&fs_info->throttles, 0); |
ab78c84d | 1599 | atomic_set(&fs_info->throttle_gen, 0); |
e20d96d6 | 1600 | fs_info->sb = sb; |
c59f8951 | 1601 | fs_info->max_extent = (u64)-1; |
6f568d35 | 1602 | fs_info->max_inline = 8192 * 1024; |
04160088 | 1603 | setup_bdi(fs_info, &fs_info->bdi); |
d98237b3 CM |
1604 | fs_info->btree_inode = new_inode(sb); |
1605 | fs_info->btree_inode->i_ino = 1; | |
2c90e5d6 | 1606 | fs_info->btree_inode->i_nlink = 1; |
97e728d4 | 1607 | fs_info->metadata_ratio = 8; |
c8b97818 | 1608 | |
b34b086c CM |
1609 | fs_info->thread_pool_size = min_t(unsigned long, |
1610 | num_online_cpus() + 2, 8); | |
0afbaf8c | 1611 | |
3eaa2885 CM |
1612 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
1613 | spin_lock_init(&fs_info->ordered_extent_lock); | |
1614 | ||
a061fc8d CM |
1615 | sb->s_blocksize = 4096; |
1616 | sb->s_blocksize_bits = blksize_bits(4096); | |
1617 | ||
0afbaf8c CM |
1618 | /* |
1619 | * we set the i_size on the btree inode to the max possible int. | |
1620 | * the real end of the address space is determined by all of | |
1621 | * the devices in the system | |
1622 | */ | |
1623 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 1624 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
1625 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
1626 | ||
d1310b2e | 1627 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
5f39d397 CM |
1628 | fs_info->btree_inode->i_mapping, |
1629 | GFP_NOFS); | |
d1310b2e CM |
1630 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, |
1631 | GFP_NOFS); | |
1632 | ||
1633 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 1634 | |
0f9dd46c JB |
1635 | spin_lock_init(&fs_info->block_group_cache_lock); |
1636 | fs_info->block_group_cache_tree.rb_node = NULL; | |
1637 | ||
d1310b2e | 1638 | extent_io_tree_init(&fs_info->pinned_extents, |
1a5bc167 | 1639 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
e66f709b | 1640 | fs_info->do_barriers = 1; |
e18e4809 | 1641 | |
1a40e23b ZY |
1642 | INIT_LIST_HEAD(&fs_info->dead_reloc_roots); |
1643 | btrfs_leaf_ref_tree_init(&fs_info->reloc_ref_tree); | |
e4657689 ZY |
1644 | btrfs_leaf_ref_tree_init(&fs_info->shared_ref_tree); |
1645 | ||
0f7d52f4 CM |
1646 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
1647 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
1648 | sizeof(struct btrfs_key)); | |
22b0ebda | 1649 | insert_inode_hash(fs_info->btree_inode); |
39279cc3 | 1650 | |
79154b1b | 1651 | mutex_init(&fs_info->trans_mutex); |
5a3f23d5 | 1652 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 1653 | mutex_init(&fs_info->tree_log_mutex); |
a2135011 | 1654 | mutex_init(&fs_info->drop_mutex); |
925baedd | 1655 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
1656 | mutex_init(&fs_info->transaction_kthread_mutex); |
1657 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 1658 | mutex_init(&fs_info->volume_mutex); |
1a40e23b | 1659 | mutex_init(&fs_info->tree_reloc_mutex); |
fa9c0d79 CM |
1660 | |
1661 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); | |
1662 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
1663 | ||
e6dcd2dc | 1664 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 1665 | init_waitqueue_head(&fs_info->transaction_wait); |
4854ddd0 | 1666 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 1667 | |
0b86a832 | 1668 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 1669 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 1670 | |
d98237b3 | 1671 | |
a512bbf8 | 1672 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
a061fc8d | 1673 | if (!bh) |
39279cc3 | 1674 | goto fail_iput; |
39279cc3 | 1675 | |
a061fc8d | 1676 | memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy)); |
2d69a0f8 YZ |
1677 | memcpy(&fs_info->super_for_commit, &fs_info->super_copy, |
1678 | sizeof(fs_info->super_for_commit)); | |
a061fc8d | 1679 | brelse(bh); |
5f39d397 | 1680 | |
a061fc8d | 1681 | memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE); |
0b86a832 | 1682 | |
5f39d397 | 1683 | disk_super = &fs_info->super_copy; |
0f7d52f4 | 1684 | if (!btrfs_super_root(disk_super)) |
c6e2bac1 | 1685 | goto fail_iput; |
0f7d52f4 | 1686 | |
2b82032c YZ |
1687 | ret = btrfs_parse_options(tree_root, options); |
1688 | if (ret) { | |
1689 | err = ret; | |
c6e2bac1 | 1690 | goto fail_iput; |
2b82032c | 1691 | } |
dfe25020 | 1692 | |
f2b636e8 JB |
1693 | features = btrfs_super_incompat_flags(disk_super) & |
1694 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
1695 | if (features) { | |
1696 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
1697 | "unsupported optional features (%Lx).\n", | |
1698 | features); | |
1699 | err = -EINVAL; | |
c6e2bac1 | 1700 | goto fail_iput; |
f2b636e8 JB |
1701 | } |
1702 | ||
1703 | features = btrfs_super_compat_ro_flags(disk_super) & | |
1704 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
1705 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
1706 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
1707 | "unsupported option features (%Lx).\n", | |
1708 | features); | |
1709 | err = -EINVAL; | |
c6e2bac1 | 1710 | goto fail_iput; |
f2b636e8 JB |
1711 | } |
1712 | ||
4543df7e CM |
1713 | /* |
1714 | * we need to start all the end_io workers up front because the | |
1715 | * queue work function gets called at interrupt time, and so it | |
1716 | * cannot dynamically grow. | |
1717 | */ | |
5443be45 CM |
1718 | btrfs_init_workers(&fs_info->workers, "worker", |
1719 | fs_info->thread_pool_size); | |
c8b97818 | 1720 | |
771ed689 CM |
1721 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
1722 | fs_info->thread_pool_size); | |
1723 | ||
5443be45 | 1724 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 CM |
1725 | min_t(u64, fs_devices->num_devices, |
1726 | fs_info->thread_pool_size)); | |
61b49440 CM |
1727 | |
1728 | /* a higher idle thresh on the submit workers makes it much more | |
1729 | * likely that bios will be send down in a sane order to the | |
1730 | * devices | |
1731 | */ | |
1732 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 1733 | |
771ed689 | 1734 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 1735 | fs_info->workers.ordered = 1; |
61b49440 | 1736 | |
771ed689 CM |
1737 | fs_info->delalloc_workers.idle_thresh = 2; |
1738 | fs_info->delalloc_workers.ordered = 1; | |
1739 | ||
5443be45 CM |
1740 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1); |
1741 | btrfs_init_workers(&fs_info->endio_workers, "endio", | |
1742 | fs_info->thread_pool_size); | |
d20f7043 CM |
1743 | btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", |
1744 | fs_info->thread_pool_size); | |
cad321ad CM |
1745 | btrfs_init_workers(&fs_info->endio_meta_write_workers, |
1746 | "endio-meta-write", fs_info->thread_pool_size); | |
5443be45 | 1747 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", |
e6dcd2dc | 1748 | fs_info->thread_pool_size); |
61b49440 CM |
1749 | |
1750 | /* | |
1751 | * endios are largely parallel and should have a very | |
1752 | * low idle thresh | |
1753 | */ | |
1754 | fs_info->endio_workers.idle_thresh = 4; | |
b51912c9 CM |
1755 | fs_info->endio_meta_workers.idle_thresh = 4; |
1756 | ||
3a5f1d45 | 1757 | fs_info->endio_write_workers.idle_thresh = 64; |
cad321ad | 1758 | fs_info->endio_meta_write_workers.idle_thresh = 64; |
61b49440 | 1759 | |
4543df7e | 1760 | btrfs_start_workers(&fs_info->workers, 1); |
1cc127b5 | 1761 | btrfs_start_workers(&fs_info->submit_workers, 1); |
771ed689 | 1762 | btrfs_start_workers(&fs_info->delalloc_workers, 1); |
247e743c | 1763 | btrfs_start_workers(&fs_info->fixup_workers, 1); |
4543df7e | 1764 | btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size); |
d20f7043 CM |
1765 | btrfs_start_workers(&fs_info->endio_meta_workers, |
1766 | fs_info->thread_pool_size); | |
cad321ad CM |
1767 | btrfs_start_workers(&fs_info->endio_meta_write_workers, |
1768 | fs_info->thread_pool_size); | |
e6dcd2dc CM |
1769 | btrfs_start_workers(&fs_info->endio_write_workers, |
1770 | fs_info->thread_pool_size); | |
4543df7e | 1771 | |
4575c9cc | 1772 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
1773 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
1774 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 1775 | |
db94535d CM |
1776 | nodesize = btrfs_super_nodesize(disk_super); |
1777 | leafsize = btrfs_super_leafsize(disk_super); | |
1778 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 1779 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
1780 | tree_root->nodesize = nodesize; |
1781 | tree_root->leafsize = leafsize; | |
1782 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 1783 | tree_root->stripesize = stripesize; |
a061fc8d CM |
1784 | |
1785 | sb->s_blocksize = sectorsize; | |
1786 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 1787 | |
39279cc3 CM |
1788 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
1789 | sizeof(disk_super->magic))) { | |
d397712b | 1790 | printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
1791 | goto fail_sb_buffer; |
1792 | } | |
19c00ddc | 1793 | |
925baedd | 1794 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 1795 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 1796 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 1797 | if (ret) { |
d397712b CM |
1798 | printk(KERN_WARNING "btrfs: failed to read the system " |
1799 | "array on %s\n", sb->s_id); | |
84eed90f CM |
1800 | goto fail_sys_array; |
1801 | } | |
0b86a832 CM |
1802 | |
1803 | blocksize = btrfs_level_size(tree_root, | |
1804 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 1805 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
1806 | |
1807 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1808 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
1809 | ||
1810 | chunk_root->node = read_tree_block(chunk_root, | |
1811 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 1812 | blocksize, generation); |
0b86a832 CM |
1813 | BUG_ON(!chunk_root->node); |
1814 | ||
e17cade2 | 1815 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
d397712b CM |
1816 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), |
1817 | BTRFS_UUID_SIZE); | |
e17cade2 | 1818 | |
925baedd | 1819 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1820 | ret = btrfs_read_chunk_tree(chunk_root); |
925baedd | 1821 | mutex_unlock(&fs_info->chunk_mutex); |
2b82032c | 1822 | if (ret) { |
d397712b CM |
1823 | printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", |
1824 | sb->s_id); | |
2b82032c YZ |
1825 | goto fail_chunk_root; |
1826 | } | |
0b86a832 | 1827 | |
dfe25020 CM |
1828 | btrfs_close_extra_devices(fs_devices); |
1829 | ||
db94535d CM |
1830 | blocksize = btrfs_level_size(tree_root, |
1831 | btrfs_super_root_level(disk_super)); | |
84234f3a | 1832 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 1833 | |
e20d96d6 | 1834 | tree_root->node = read_tree_block(tree_root, |
db94535d | 1835 | btrfs_super_root(disk_super), |
84234f3a | 1836 | blocksize, generation); |
39279cc3 | 1837 | if (!tree_root->node) |
2b82032c | 1838 | goto fail_chunk_root; |
3768f368 | 1839 | |
db94535d CM |
1840 | |
1841 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 1842 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 1843 | if (ret) |
39279cc3 | 1844 | goto fail_tree_root; |
0b86a832 CM |
1845 | extent_root->track_dirty = 1; |
1846 | ||
1847 | ret = find_and_setup_root(tree_root, fs_info, | |
1848 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
1849 | dev_root->track_dirty = 1; | |
0b86a832 CM |
1850 | if (ret) |
1851 | goto fail_extent_root; | |
3768f368 | 1852 | |
d20f7043 CM |
1853 | ret = find_and_setup_root(tree_root, fs_info, |
1854 | BTRFS_CSUM_TREE_OBJECTID, csum_root); | |
1855 | if (ret) | |
1856 | goto fail_extent_root; | |
1857 | ||
1858 | csum_root->track_dirty = 1; | |
1859 | ||
9078a3e1 CM |
1860 | btrfs_read_block_groups(extent_root); |
1861 | ||
2d69a0f8 | 1862 | fs_info->generation = generation; |
c146afad | 1863 | fs_info->last_trans_committed = generation; |
d18a2c44 CM |
1864 | fs_info->data_alloc_profile = (u64)-1; |
1865 | fs_info->metadata_alloc_profile = (u64)-1; | |
1866 | fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; | |
a74a4b97 CM |
1867 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
1868 | "btrfs-cleaner"); | |
57506d50 | 1869 | if (IS_ERR(fs_info->cleaner_kthread)) |
d20f7043 | 1870 | goto fail_csum_root; |
a74a4b97 CM |
1871 | |
1872 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
1873 | tree_root, | |
1874 | "btrfs-transaction"); | |
57506d50 | 1875 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 1876 | goto fail_cleaner; |
a74a4b97 | 1877 | |
e02119d5 | 1878 | if (btrfs_super_log_root(disk_super) != 0) { |
e02119d5 CM |
1879 | u64 bytenr = btrfs_super_log_root(disk_super); |
1880 | ||
7c2ca468 | 1881 | if (fs_devices->rw_devices == 0) { |
d397712b CM |
1882 | printk(KERN_WARNING "Btrfs log replay required " |
1883 | "on RO media\n"); | |
7c2ca468 CM |
1884 | err = -EIO; |
1885 | goto fail_trans_kthread; | |
1886 | } | |
e02119d5 CM |
1887 | blocksize = |
1888 | btrfs_level_size(tree_root, | |
1889 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 1890 | |
e02119d5 CM |
1891 | log_tree_root = kzalloc(sizeof(struct btrfs_root), |
1892 | GFP_NOFS); | |
1893 | ||
1894 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1895 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1896 | ||
1897 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
1898 | blocksize, |
1899 | generation + 1); | |
e02119d5 CM |
1900 | ret = btrfs_recover_log_trees(log_tree_root); |
1901 | BUG_ON(ret); | |
e556ce2c YZ |
1902 | |
1903 | if (sb->s_flags & MS_RDONLY) { | |
1904 | ret = btrfs_commit_super(tree_root); | |
1905 | BUG_ON(ret); | |
1906 | } | |
e02119d5 | 1907 | } |
1a40e23b | 1908 | |
7c2ca468 CM |
1909 | if (!(sb->s_flags & MS_RDONLY)) { |
1910 | ret = btrfs_cleanup_reloc_trees(tree_root); | |
1911 | BUG_ON(ret); | |
1912 | } | |
1a40e23b | 1913 | |
3de4586c CM |
1914 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
1915 | location.type = BTRFS_ROOT_ITEM_KEY; | |
1916 | location.offset = (u64)-1; | |
1917 | ||
3de4586c CM |
1918 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
1919 | if (!fs_info->fs_root) | |
7c2ca468 | 1920 | goto fail_trans_kthread; |
0f7d52f4 | 1921 | return tree_root; |
39279cc3 | 1922 | |
7c2ca468 CM |
1923 | fail_trans_kthread: |
1924 | kthread_stop(fs_info->transaction_kthread); | |
3f157a2f | 1925 | fail_cleaner: |
a74a4b97 | 1926 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
1927 | |
1928 | /* | |
1929 | * make sure we're done with the btree inode before we stop our | |
1930 | * kthreads | |
1931 | */ | |
1932 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
1933 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
1934 | ||
d20f7043 CM |
1935 | fail_csum_root: |
1936 | free_extent_buffer(csum_root->node); | |
0b86a832 CM |
1937 | fail_extent_root: |
1938 | free_extent_buffer(extent_root->node); | |
39279cc3 | 1939 | fail_tree_root: |
5f39d397 | 1940 | free_extent_buffer(tree_root->node); |
2b82032c YZ |
1941 | fail_chunk_root: |
1942 | free_extent_buffer(chunk_root->node); | |
84eed90f | 1943 | fail_sys_array: |
7c2ca468 | 1944 | free_extent_buffer(dev_root->node); |
39279cc3 | 1945 | fail_sb_buffer: |
247e743c | 1946 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 1947 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
1948 | btrfs_stop_workers(&fs_info->workers); |
1949 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 1950 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 1951 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 1952 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 1953 | btrfs_stop_workers(&fs_info->submit_workers); |
4543df7e | 1954 | fail_iput: |
7c2ca468 | 1955 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
4543df7e | 1956 | iput(fs_info->btree_inode); |
7e662854 | 1957 | |
dfe25020 | 1958 | btrfs_close_devices(fs_info->fs_devices); |
84eed90f | 1959 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
7e662854 | 1960 | bdi_destroy(&fs_info->bdi); |
84eed90f | 1961 | |
7e662854 | 1962 | fail: |
39279cc3 CM |
1963 | kfree(extent_root); |
1964 | kfree(tree_root); | |
1965 | kfree(fs_info); | |
83afeac4 JM |
1966 | kfree(chunk_root); |
1967 | kfree(dev_root); | |
d20f7043 | 1968 | kfree(csum_root); |
39279cc3 | 1969 | return ERR_PTR(err); |
eb60ceac CM |
1970 | } |
1971 | ||
f2984462 CM |
1972 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
1973 | { | |
1974 | char b[BDEVNAME_SIZE]; | |
1975 | ||
1976 | if (uptodate) { | |
1977 | set_buffer_uptodate(bh); | |
1978 | } else { | |
1979 | if (!buffer_eopnotsupp(bh) && printk_ratelimit()) { | |
1980 | printk(KERN_WARNING "lost page write due to " | |
1981 | "I/O error on %s\n", | |
1982 | bdevname(bh->b_bdev, b)); | |
1983 | } | |
1259ab75 CM |
1984 | /* note, we dont' set_buffer_write_io_error because we have |
1985 | * our own ways of dealing with the IO errors | |
1986 | */ | |
f2984462 CM |
1987 | clear_buffer_uptodate(bh); |
1988 | } | |
1989 | unlock_buffer(bh); | |
1990 | put_bh(bh); | |
1991 | } | |
1992 | ||
a512bbf8 YZ |
1993 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
1994 | { | |
1995 | struct buffer_head *bh; | |
1996 | struct buffer_head *latest = NULL; | |
1997 | struct btrfs_super_block *super; | |
1998 | int i; | |
1999 | u64 transid = 0; | |
2000 | u64 bytenr; | |
2001 | ||
2002 | /* we would like to check all the supers, but that would make | |
2003 | * a btrfs mount succeed after a mkfs from a different FS. | |
2004 | * So, we need to add a special mount option to scan for | |
2005 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
2006 | */ | |
2007 | for (i = 0; i < 1; i++) { | |
2008 | bytenr = btrfs_sb_offset(i); | |
2009 | if (bytenr + 4096 >= i_size_read(bdev->bd_inode)) | |
2010 | break; | |
2011 | bh = __bread(bdev, bytenr / 4096, 4096); | |
2012 | if (!bh) | |
2013 | continue; | |
2014 | ||
2015 | super = (struct btrfs_super_block *)bh->b_data; | |
2016 | if (btrfs_super_bytenr(super) != bytenr || | |
2017 | strncmp((char *)(&super->magic), BTRFS_MAGIC, | |
2018 | sizeof(super->magic))) { | |
2019 | brelse(bh); | |
2020 | continue; | |
2021 | } | |
2022 | ||
2023 | if (!latest || btrfs_super_generation(super) > transid) { | |
2024 | brelse(latest); | |
2025 | latest = bh; | |
2026 | transid = btrfs_super_generation(super); | |
2027 | } else { | |
2028 | brelse(bh); | |
2029 | } | |
2030 | } | |
2031 | return latest; | |
2032 | } | |
2033 | ||
2034 | static int write_dev_supers(struct btrfs_device *device, | |
2035 | struct btrfs_super_block *sb, | |
2036 | int do_barriers, int wait, int max_mirrors) | |
2037 | { | |
2038 | struct buffer_head *bh; | |
2039 | int i; | |
2040 | int ret; | |
2041 | int errors = 0; | |
2042 | u32 crc; | |
2043 | u64 bytenr; | |
2044 | int last_barrier = 0; | |
2045 | ||
2046 | if (max_mirrors == 0) | |
2047 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
2048 | ||
2049 | /* make sure only the last submit_bh does a barrier */ | |
2050 | if (do_barriers) { | |
2051 | for (i = 0; i < max_mirrors; i++) { | |
2052 | bytenr = btrfs_sb_offset(i); | |
2053 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= | |
2054 | device->total_bytes) | |
2055 | break; | |
2056 | last_barrier = i; | |
2057 | } | |
2058 | } | |
2059 | ||
2060 | for (i = 0; i < max_mirrors; i++) { | |
2061 | bytenr = btrfs_sb_offset(i); | |
2062 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) | |
2063 | break; | |
2064 | ||
2065 | if (wait) { | |
2066 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
2067 | BTRFS_SUPER_INFO_SIZE); | |
2068 | BUG_ON(!bh); | |
2069 | brelse(bh); | |
2070 | wait_on_buffer(bh); | |
2071 | if (buffer_uptodate(bh)) { | |
2072 | brelse(bh); | |
2073 | continue; | |
2074 | } | |
2075 | } else { | |
2076 | btrfs_set_super_bytenr(sb, bytenr); | |
2077 | ||
2078 | crc = ~(u32)0; | |
2079 | crc = btrfs_csum_data(NULL, (char *)sb + | |
2080 | BTRFS_CSUM_SIZE, crc, | |
2081 | BTRFS_SUPER_INFO_SIZE - | |
2082 | BTRFS_CSUM_SIZE); | |
2083 | btrfs_csum_final(crc, sb->csum); | |
2084 | ||
2085 | bh = __getblk(device->bdev, bytenr / 4096, | |
2086 | BTRFS_SUPER_INFO_SIZE); | |
2087 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); | |
2088 | ||
2089 | set_buffer_uptodate(bh); | |
2090 | get_bh(bh); | |
2091 | lock_buffer(bh); | |
2092 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
2093 | } | |
2094 | ||
2095 | if (i == last_barrier && do_barriers && device->barriers) { | |
2096 | ret = submit_bh(WRITE_BARRIER, bh); | |
2097 | if (ret == -EOPNOTSUPP) { | |
2098 | printk("btrfs: disabling barriers on dev %s\n", | |
2099 | device->name); | |
2100 | set_buffer_uptodate(bh); | |
2101 | device->barriers = 0; | |
2102 | get_bh(bh); | |
2103 | lock_buffer(bh); | |
ffbd517d | 2104 | ret = submit_bh(WRITE_SYNC, bh); |
a512bbf8 YZ |
2105 | } |
2106 | } else { | |
ffbd517d | 2107 | ret = submit_bh(WRITE_SYNC, bh); |
a512bbf8 YZ |
2108 | } |
2109 | ||
2110 | if (!ret && wait) { | |
2111 | wait_on_buffer(bh); | |
2112 | if (!buffer_uptodate(bh)) | |
2113 | errors++; | |
2114 | } else if (ret) { | |
2115 | errors++; | |
2116 | } | |
2117 | if (wait) | |
2118 | brelse(bh); | |
2119 | } | |
2120 | return errors < i ? 0 : -1; | |
2121 | } | |
2122 | ||
2123 | int write_all_supers(struct btrfs_root *root, int max_mirrors) | |
f2984462 | 2124 | { |
f2984462 CM |
2125 | struct list_head *head = &root->fs_info->fs_devices->devices; |
2126 | struct btrfs_device *dev; | |
a061fc8d | 2127 | struct btrfs_super_block *sb; |
f2984462 | 2128 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
2129 | int ret; |
2130 | int do_barriers; | |
a236aed1 CM |
2131 | int max_errors; |
2132 | int total_errors = 0; | |
a061fc8d | 2133 | u64 flags; |
f2984462 | 2134 | |
a236aed1 | 2135 | max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1; |
f2984462 CM |
2136 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
2137 | ||
a061fc8d CM |
2138 | sb = &root->fs_info->super_for_commit; |
2139 | dev_item = &sb->dev_item; | |
c6e30871 | 2140 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
2141 | if (!dev->bdev) { |
2142 | total_errors++; | |
2143 | continue; | |
2144 | } | |
2b82032c | 2145 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2146 | continue; |
2147 | ||
2b82032c | 2148 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
2149 | btrfs_set_stack_device_type(dev_item, dev->type); |
2150 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
2151 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
2152 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
2153 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
2154 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
2155 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
2156 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 2157 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 2158 | |
a061fc8d CM |
2159 | flags = btrfs_super_flags(sb); |
2160 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
2161 | ||
a512bbf8 | 2162 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
2163 | if (ret) |
2164 | total_errors++; | |
f2984462 | 2165 | } |
a236aed1 | 2166 | if (total_errors > max_errors) { |
d397712b CM |
2167 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2168 | total_errors); | |
a236aed1 CM |
2169 | BUG(); |
2170 | } | |
f2984462 | 2171 | |
a512bbf8 | 2172 | total_errors = 0; |
c6e30871 | 2173 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
2174 | if (!dev->bdev) |
2175 | continue; | |
2b82032c | 2176 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2177 | continue; |
2178 | ||
a512bbf8 YZ |
2179 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
2180 | if (ret) | |
2181 | total_errors++; | |
f2984462 | 2182 | } |
a236aed1 | 2183 | if (total_errors > max_errors) { |
d397712b CM |
2184 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2185 | total_errors); | |
a236aed1 CM |
2186 | BUG(); |
2187 | } | |
f2984462 CM |
2188 | return 0; |
2189 | } | |
2190 | ||
a512bbf8 YZ |
2191 | int write_ctree_super(struct btrfs_trans_handle *trans, |
2192 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 2193 | { |
e66f709b | 2194 | int ret; |
5f39d397 | 2195 | |
a512bbf8 | 2196 | ret = write_all_supers(root, max_mirrors); |
5f39d397 | 2197 | return ret; |
cfaa7295 CM |
2198 | } |
2199 | ||
5eda7b5e | 2200 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f CM |
2201 | { |
2202 | radix_tree_delete(&fs_info->fs_roots_radix, | |
2203 | (unsigned long)root->root_key.objectid); | |
3394e160 CM |
2204 | if (root->anon_super.s_dev) { |
2205 | down_write(&root->anon_super.s_umount); | |
2206 | kill_anon_super(&root->anon_super); | |
2207 | } | |
2619ba1f | 2208 | if (root->node) |
5f39d397 | 2209 | free_extent_buffer(root->node); |
2619ba1f | 2210 | if (root->commit_root) |
5f39d397 | 2211 | free_extent_buffer(root->commit_root); |
d397712b | 2212 | kfree(root->name); |
2619ba1f CM |
2213 | kfree(root); |
2214 | return 0; | |
2215 | } | |
2216 | ||
35b7e476 | 2217 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
2218 | { |
2219 | int ret; | |
2220 | struct btrfs_root *gang[8]; | |
2221 | int i; | |
2222 | ||
d397712b | 2223 | while (1) { |
0f7d52f4 CM |
2224 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
2225 | (void **)gang, 0, | |
2226 | ARRAY_SIZE(gang)); | |
2227 | if (!ret) | |
2228 | break; | |
2619ba1f | 2229 | for (i = 0; i < ret; i++) |
5eda7b5e | 2230 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
2231 | } |
2232 | return 0; | |
2233 | } | |
b4100d64 | 2234 | |
c146afad | 2235 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 2236 | { |
c146afad YZ |
2237 | u64 root_objectid = 0; |
2238 | struct btrfs_root *gang[8]; | |
2239 | int i; | |
3768f368 | 2240 | int ret; |
e089f05c | 2241 | |
c146afad YZ |
2242 | while (1) { |
2243 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2244 | (void **)gang, root_objectid, | |
2245 | ARRAY_SIZE(gang)); | |
2246 | if (!ret) | |
2247 | break; | |
2248 | for (i = 0; i < ret; i++) { | |
2249 | root_objectid = gang[i]->root_key.objectid; | |
2250 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
2251 | root_objectid, gang[i]); | |
2252 | BUG_ON(ret); | |
2253 | btrfs_orphan_cleanup(gang[i]); | |
2254 | } | |
2255 | root_objectid++; | |
2256 | } | |
2257 | return 0; | |
2258 | } | |
a2135011 | 2259 | |
c146afad YZ |
2260 | int btrfs_commit_super(struct btrfs_root *root) |
2261 | { | |
2262 | struct btrfs_trans_handle *trans; | |
2263 | int ret; | |
a74a4b97 | 2264 | |
c146afad | 2265 | mutex_lock(&root->fs_info->cleaner_mutex); |
a74a4b97 | 2266 | btrfs_clean_old_snapshots(root); |
c146afad | 2267 | mutex_unlock(&root->fs_info->cleaner_mutex); |
79154b1b | 2268 | trans = btrfs_start_transaction(root, 1); |
54aa1f4d | 2269 | ret = btrfs_commit_transaction(trans, root); |
c146afad YZ |
2270 | BUG_ON(ret); |
2271 | /* run commit again to drop the original snapshot */ | |
79154b1b CM |
2272 | trans = btrfs_start_transaction(root, 1); |
2273 | btrfs_commit_transaction(trans, root); | |
2274 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 2275 | BUG_ON(ret); |
d6bfde87 | 2276 | |
a512bbf8 | 2277 | ret = write_ctree_super(NULL, root, 0); |
c146afad YZ |
2278 | return ret; |
2279 | } | |
2280 | ||
2281 | int close_ctree(struct btrfs_root *root) | |
2282 | { | |
2283 | struct btrfs_fs_info *fs_info = root->fs_info; | |
2284 | int ret; | |
2285 | ||
2286 | fs_info->closing = 1; | |
2287 | smp_mb(); | |
2288 | ||
2289 | kthread_stop(root->fs_info->transaction_kthread); | |
2290 | kthread_stop(root->fs_info->cleaner_kthread); | |
2291 | ||
2292 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { | |
2293 | ret = btrfs_commit_super(root); | |
d397712b CM |
2294 | if (ret) |
2295 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
c146afad | 2296 | } |
0f7d52f4 | 2297 | |
b0c68f8b | 2298 | if (fs_info->delalloc_bytes) { |
d397712b | 2299 | printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n", |
b0c68f8b CM |
2300 | fs_info->delalloc_bytes); |
2301 | } | |
31153d81 | 2302 | if (fs_info->total_ref_cache_size) { |
d397712b CM |
2303 | printk(KERN_INFO "btrfs: at umount reference cache size %llu\n", |
2304 | (unsigned long long)fs_info->total_ref_cache_size); | |
31153d81 | 2305 | } |
bcc63abb | 2306 | |
0f7d52f4 | 2307 | if (fs_info->extent_root->node) |
5f39d397 | 2308 | free_extent_buffer(fs_info->extent_root->node); |
f510cfec | 2309 | |
0f7d52f4 | 2310 | if (fs_info->tree_root->node) |
5f39d397 | 2311 | free_extent_buffer(fs_info->tree_root->node); |
f510cfec | 2312 | |
d397712b | 2313 | if (root->fs_info->chunk_root->node) |
0b86a832 CM |
2314 | free_extent_buffer(root->fs_info->chunk_root->node); |
2315 | ||
d397712b | 2316 | if (root->fs_info->dev_root->node) |
0b86a832 CM |
2317 | free_extent_buffer(root->fs_info->dev_root->node); |
2318 | ||
d397712b | 2319 | if (root->fs_info->csum_root->node) |
d20f7043 CM |
2320 | free_extent_buffer(root->fs_info->csum_root->node); |
2321 | ||
9078a3e1 | 2322 | btrfs_free_block_groups(root->fs_info); |
d10c5f31 | 2323 | |
c146afad | 2324 | del_fs_roots(fs_info); |
d10c5f31 | 2325 | |
c146afad | 2326 | iput(fs_info->btree_inode); |
9ad6b7bc | 2327 | |
247e743c | 2328 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2329 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2330 | btrfs_stop_workers(&fs_info->workers); |
2331 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 2332 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 2333 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 2334 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 2335 | btrfs_stop_workers(&fs_info->submit_workers); |
d6bfde87 | 2336 | |
19c00ddc | 2337 | #if 0 |
d397712b | 2338 | while (!list_empty(&fs_info->hashers)) { |
19c00ddc CM |
2339 | struct btrfs_hasher *hasher; |
2340 | hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher, | |
2341 | hashers); | |
2342 | list_del(&hasher->hashers); | |
2343 | crypto_free_hash(&fs_info->hash_tfm); | |
2344 | kfree(hasher); | |
2345 | } | |
2346 | #endif | |
dfe25020 | 2347 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 2348 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 2349 | |
04160088 | 2350 | bdi_destroy(&fs_info->bdi); |
0b86a832 | 2351 | |
0f7d52f4 | 2352 | kfree(fs_info->extent_root); |
0f7d52f4 | 2353 | kfree(fs_info->tree_root); |
0b86a832 CM |
2354 | kfree(fs_info->chunk_root); |
2355 | kfree(fs_info->dev_root); | |
d20f7043 | 2356 | kfree(fs_info->csum_root); |
eb60ceac CM |
2357 | return 0; |
2358 | } | |
2359 | ||
1259ab75 | 2360 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 2361 | { |
1259ab75 | 2362 | int ret; |
810191ff | 2363 | struct inode *btree_inode = buf->first_page->mapping->host; |
1259ab75 CM |
2364 | |
2365 | ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); | |
2366 | if (!ret) | |
2367 | return ret; | |
2368 | ||
2369 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
2370 | parent_transid); | |
2371 | return !ret; | |
5f39d397 CM |
2372 | } |
2373 | ||
2374 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 2375 | { |
810191ff | 2376 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 2377 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
2378 | buf); |
2379 | } | |
6702ed49 | 2380 | |
5f39d397 CM |
2381 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
2382 | { | |
810191ff | 2383 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
5f39d397 CM |
2384 | u64 transid = btrfs_header_generation(buf); |
2385 | struct inode *btree_inode = root->fs_info->btree_inode; | |
b9473439 | 2386 | int was_dirty; |
b4ce94de | 2387 | |
b9447ef8 | 2388 | btrfs_assert_tree_locked(buf); |
ccd467d6 | 2389 | if (transid != root->fs_info->generation) { |
d397712b CM |
2390 | printk(KERN_CRIT "btrfs transid mismatch buffer %llu, " |
2391 | "found %llu running %llu\n", | |
db94535d | 2392 | (unsigned long long)buf->start, |
d397712b CM |
2393 | (unsigned long long)transid, |
2394 | (unsigned long long)root->fs_info->generation); | |
ccd467d6 CM |
2395 | WARN_ON(1); |
2396 | } | |
b9473439 CM |
2397 | was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
2398 | buf); | |
2399 | if (!was_dirty) { | |
2400 | spin_lock(&root->fs_info->delalloc_lock); | |
2401 | root->fs_info->dirty_metadata_bytes += buf->len; | |
2402 | spin_unlock(&root->fs_info->delalloc_lock); | |
2403 | } | |
eb60ceac CM |
2404 | } |
2405 | ||
d3c2fdcf | 2406 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
35b7e476 | 2407 | { |
188de649 CM |
2408 | /* |
2409 | * looks as though older kernels can get into trouble with | |
2410 | * this code, they end up stuck in balance_dirty_pages forever | |
2411 | */ | |
d6bfde87 CM |
2412 | struct extent_io_tree *tree; |
2413 | u64 num_dirty; | |
2414 | u64 start = 0; | |
771ed689 | 2415 | unsigned long thresh = 32 * 1024 * 1024; |
d6bfde87 CM |
2416 | tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
2417 | ||
6933c02e | 2418 | if (current->flags & PF_MEMALLOC) |
d6bfde87 CM |
2419 | return; |
2420 | ||
2421 | num_dirty = count_range_bits(tree, &start, (u64)-1, | |
2422 | thresh, EXTENT_DIRTY); | |
2423 | if (num_dirty > thresh) { | |
2424 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 2425 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 2426 | } |
188de649 | 2427 | return; |
35b7e476 | 2428 | } |
6b80053d | 2429 | |
ca7a79ad | 2430 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 2431 | { |
810191ff | 2432 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
ce9adaa5 | 2433 | int ret; |
ca7a79ad | 2434 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
d397712b | 2435 | if (ret == 0) |
b4ce94de | 2436 | set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags); |
ce9adaa5 | 2437 | return ret; |
6b80053d | 2438 | } |
0da5468f | 2439 | |
4bef0848 CM |
2440 | int btree_lock_page_hook(struct page *page) |
2441 | { | |
2442 | struct inode *inode = page->mapping->host; | |
b9473439 | 2443 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4bef0848 CM |
2444 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
2445 | struct extent_buffer *eb; | |
2446 | unsigned long len; | |
2447 | u64 bytenr = page_offset(page); | |
2448 | ||
2449 | if (page->private == EXTENT_PAGE_PRIVATE) | |
2450 | goto out; | |
2451 | ||
2452 | len = page->private >> 2; | |
2453 | eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS); | |
2454 | if (!eb) | |
2455 | goto out; | |
2456 | ||
2457 | btrfs_tree_lock(eb); | |
4bef0848 | 2458 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); |
b9473439 CM |
2459 | |
2460 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) { | |
2461 | spin_lock(&root->fs_info->delalloc_lock); | |
2462 | if (root->fs_info->dirty_metadata_bytes >= eb->len) | |
2463 | root->fs_info->dirty_metadata_bytes -= eb->len; | |
2464 | else | |
2465 | WARN_ON(1); | |
2466 | spin_unlock(&root->fs_info->delalloc_lock); | |
2467 | } | |
2468 | ||
4bef0848 CM |
2469 | btrfs_tree_unlock(eb); |
2470 | free_extent_buffer(eb); | |
2471 | out: | |
2472 | lock_page(page); | |
2473 | return 0; | |
2474 | } | |
2475 | ||
d1310b2e | 2476 | static struct extent_io_ops btree_extent_io_ops = { |
4bef0848 | 2477 | .write_cache_pages_lock_hook = btree_lock_page_hook, |
ce9adaa5 | 2478 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
0b86a832 | 2479 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
2480 | /* note we're sharing with inode.c for the merge bio hook */ |
2481 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 2482 | }; |