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" |
eb60ceac | 41 | |
d1310b2e | 42 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 43 | static void end_workqueue_fn(struct btrfs_work *work); |
ce9adaa5 | 44 | |
d352ac68 CM |
45 | /* |
46 | * end_io_wq structs are used to do processing in task context when an IO is | |
47 | * complete. This is used during reads to verify checksums, and it is used | |
48 | * by writes to insert metadata for new file extents after IO is complete. | |
49 | */ | |
ce9adaa5 CM |
50 | struct end_io_wq { |
51 | struct bio *bio; | |
52 | bio_end_io_t *end_io; | |
53 | void *private; | |
54 | struct btrfs_fs_info *info; | |
55 | int error; | |
22c59948 | 56 | int metadata; |
ce9adaa5 | 57 | struct list_head list; |
8b712842 | 58 | struct btrfs_work work; |
ce9adaa5 | 59 | }; |
0da5468f | 60 | |
d352ac68 CM |
61 | /* |
62 | * async submit bios are used to offload expensive checksumming | |
63 | * onto the worker threads. They checksum file and metadata bios | |
64 | * just before they are sent down the IO stack. | |
65 | */ | |
44b8bd7e CM |
66 | struct async_submit_bio { |
67 | struct inode *inode; | |
68 | struct bio *bio; | |
69 | struct list_head list; | |
4a69a410 CM |
70 | extent_submit_bio_hook_t *submit_bio_start; |
71 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
72 | int rw; |
73 | int mirror_num; | |
c8b97818 | 74 | unsigned long bio_flags; |
8b712842 | 75 | struct btrfs_work work; |
44b8bd7e CM |
76 | }; |
77 | ||
4008c04a CM |
78 | /* These are used to set the lockdep class on the extent buffer locks. |
79 | * The class is set by the readpage_end_io_hook after the buffer has | |
80 | * passed csum validation but before the pages are unlocked. | |
81 | * | |
82 | * The lockdep class is also set by btrfs_init_new_buffer on freshly | |
83 | * allocated blocks. | |
84 | * | |
85 | * The class is based on the level in the tree block, which allows lockdep | |
86 | * to know that lower nodes nest inside the locks of higher nodes. | |
87 | * | |
88 | * We also add a check to make sure the highest level of the tree is | |
89 | * the same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this | |
90 | * code needs update as well. | |
91 | */ | |
92 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
93 | # if BTRFS_MAX_LEVEL != 8 | |
94 | # error | |
95 | # endif | |
96 | static struct lock_class_key btrfs_eb_class[BTRFS_MAX_LEVEL + 1]; | |
97 | static const char *btrfs_eb_name[BTRFS_MAX_LEVEL + 1] = { | |
98 | /* leaf */ | |
99 | "btrfs-extent-00", | |
100 | "btrfs-extent-01", | |
101 | "btrfs-extent-02", | |
102 | "btrfs-extent-03", | |
103 | "btrfs-extent-04", | |
104 | "btrfs-extent-05", | |
105 | "btrfs-extent-06", | |
106 | "btrfs-extent-07", | |
107 | /* highest possible level */ | |
108 | "btrfs-extent-08", | |
109 | }; | |
110 | #endif | |
111 | ||
d352ac68 CM |
112 | /* |
113 | * extents on the btree inode are pretty simple, there's one extent | |
114 | * that covers the entire device | |
115 | */ | |
b2950863 CH |
116 | static struct extent_map *btree_get_extent(struct inode *inode, |
117 | struct page *page, size_t page_offset, u64 start, u64 len, | |
118 | int create) | |
7eccb903 | 119 | { |
5f39d397 CM |
120 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
121 | struct extent_map *em; | |
122 | int ret; | |
123 | ||
d1310b2e CM |
124 | spin_lock(&em_tree->lock); |
125 | em = lookup_extent_mapping(em_tree, start, len); | |
a061fc8d CM |
126 | if (em) { |
127 | em->bdev = | |
128 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
129 | spin_unlock(&em_tree->lock); | |
5f39d397 | 130 | goto out; |
a061fc8d CM |
131 | } |
132 | spin_unlock(&em_tree->lock); | |
7b13b7b1 | 133 | |
5f39d397 CM |
134 | em = alloc_extent_map(GFP_NOFS); |
135 | if (!em) { | |
136 | em = ERR_PTR(-ENOMEM); | |
137 | goto out; | |
138 | } | |
139 | em->start = 0; | |
0afbaf8c | 140 | em->len = (u64)-1; |
c8b97818 | 141 | em->block_len = (u64)-1; |
5f39d397 | 142 | em->block_start = 0; |
a061fc8d | 143 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e CM |
144 | |
145 | spin_lock(&em_tree->lock); | |
5f39d397 CM |
146 | ret = add_extent_mapping(em_tree, em); |
147 | if (ret == -EEXIST) { | |
0afbaf8c CM |
148 | u64 failed_start = em->start; |
149 | u64 failed_len = em->len; | |
150 | ||
5f39d397 | 151 | free_extent_map(em); |
7b13b7b1 | 152 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c | 153 | if (em) { |
7b13b7b1 | 154 | ret = 0; |
0afbaf8c CM |
155 | } else { |
156 | em = lookup_extent_mapping(em_tree, failed_start, | |
157 | failed_len); | |
7b13b7b1 | 158 | ret = -EIO; |
0afbaf8c | 159 | } |
5f39d397 | 160 | } else if (ret) { |
7b13b7b1 CM |
161 | free_extent_map(em); |
162 | em = NULL; | |
5f39d397 | 163 | } |
7b13b7b1 CM |
164 | spin_unlock(&em_tree->lock); |
165 | ||
166 | if (ret) | |
167 | em = ERR_PTR(ret); | |
5f39d397 CM |
168 | out: |
169 | return em; | |
7eccb903 CM |
170 | } |
171 | ||
19c00ddc CM |
172 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
173 | { | |
a5eb62e3 | 174 | return btrfs_crc32c(seed, data, len); |
19c00ddc CM |
175 | } |
176 | ||
177 | void btrfs_csum_final(u32 crc, char *result) | |
178 | { | |
179 | *(__le32 *)result = ~cpu_to_le32(crc); | |
180 | } | |
181 | ||
d352ac68 CM |
182 | /* |
183 | * compute the csum for a btree block, and either verify it or write it | |
184 | * into the csum field of the block. | |
185 | */ | |
19c00ddc CM |
186 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
187 | int verify) | |
188 | { | |
607d432d JB |
189 | u16 csum_size = |
190 | btrfs_super_csum_size(&root->fs_info->super_copy); | |
191 | char *result = NULL; | |
19c00ddc CM |
192 | unsigned long len; |
193 | unsigned long cur_len; | |
194 | unsigned long offset = BTRFS_CSUM_SIZE; | |
195 | char *map_token = NULL; | |
196 | char *kaddr; | |
197 | unsigned long map_start; | |
198 | unsigned long map_len; | |
199 | int err; | |
200 | u32 crc = ~(u32)0; | |
607d432d | 201 | unsigned long inline_result; |
19c00ddc CM |
202 | |
203 | len = buf->len - offset; | |
d397712b | 204 | while (len > 0) { |
19c00ddc CM |
205 | err = map_private_extent_buffer(buf, offset, 32, |
206 | &map_token, &kaddr, | |
207 | &map_start, &map_len, KM_USER0); | |
d397712b | 208 | if (err) |
19c00ddc | 209 | return 1; |
19c00ddc CM |
210 | cur_len = min(len, map_len - (offset - map_start)); |
211 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
212 | crc, cur_len); | |
213 | len -= cur_len; | |
214 | offset += cur_len; | |
215 | unmap_extent_buffer(buf, map_token, KM_USER0); | |
216 | } | |
607d432d JB |
217 | if (csum_size > sizeof(inline_result)) { |
218 | result = kzalloc(csum_size * sizeof(char), GFP_NOFS); | |
219 | if (!result) | |
220 | return 1; | |
221 | } else { | |
222 | result = (char *)&inline_result; | |
223 | } | |
224 | ||
19c00ddc CM |
225 | btrfs_csum_final(crc, result); |
226 | ||
227 | if (verify) { | |
607d432d | 228 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
229 | u32 val; |
230 | u32 found = 0; | |
607d432d | 231 | memcpy(&found, result, csum_size); |
e4204ded | 232 | |
607d432d | 233 | read_extent_buffer(buf, &val, 0, csum_size); |
d397712b CM |
234 | printk(KERN_INFO "btrfs: %s checksum verify failed " |
235 | "on %llu wanted %X found %X level %d\n", | |
19c00ddc | 236 | root->fs_info->sb->s_id, |
2dd3e67b | 237 | buf->start, val, found, btrfs_header_level(buf)); |
607d432d JB |
238 | if (result != (char *)&inline_result) |
239 | kfree(result); | |
19c00ddc CM |
240 | return 1; |
241 | } | |
242 | } else { | |
607d432d | 243 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 244 | } |
607d432d JB |
245 | if (result != (char *)&inline_result) |
246 | kfree(result); | |
19c00ddc CM |
247 | return 0; |
248 | } | |
249 | ||
d352ac68 CM |
250 | /* |
251 | * we can't consider a given block up to date unless the transid of the | |
252 | * block matches the transid in the parent node's pointer. This is how we | |
253 | * detect blocks that either didn't get written at all or got written | |
254 | * in the wrong place. | |
255 | */ | |
1259ab75 CM |
256 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
257 | struct extent_buffer *eb, u64 parent_transid) | |
258 | { | |
259 | int ret; | |
260 | ||
261 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
262 | return 0; | |
263 | ||
264 | lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); | |
265 | if (extent_buffer_uptodate(io_tree, eb) && | |
266 | btrfs_header_generation(eb) == parent_transid) { | |
267 | ret = 0; | |
268 | goto out; | |
269 | } | |
270 | printk("parent transid verify failed on %llu wanted %llu found %llu\n", | |
271 | (unsigned long long)eb->start, | |
272 | (unsigned long long)parent_transid, | |
273 | (unsigned long long)btrfs_header_generation(eb)); | |
274 | ret = 1; | |
1259ab75 | 275 | clear_extent_buffer_uptodate(io_tree, eb); |
33958dc6 | 276 | out: |
1259ab75 CM |
277 | unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, |
278 | GFP_NOFS); | |
279 | return ret; | |
1259ab75 CM |
280 | } |
281 | ||
d352ac68 CM |
282 | /* |
283 | * helper to read a given tree block, doing retries as required when | |
284 | * the checksums don't match and we have alternate mirrors to try. | |
285 | */ | |
f188591e CM |
286 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
287 | struct extent_buffer *eb, | |
ca7a79ad | 288 | u64 start, u64 parent_transid) |
f188591e CM |
289 | { |
290 | struct extent_io_tree *io_tree; | |
291 | int ret; | |
292 | int num_copies = 0; | |
293 | int mirror_num = 0; | |
294 | ||
295 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
296 | while (1) { | |
297 | ret = read_extent_buffer_pages(io_tree, eb, start, 1, | |
298 | btree_get_extent, mirror_num); | |
1259ab75 CM |
299 | if (!ret && |
300 | !verify_parent_transid(io_tree, eb, parent_transid)) | |
f188591e | 301 | return ret; |
d397712b | 302 | |
f188591e CM |
303 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
304 | eb->start, eb->len); | |
4235298e | 305 | if (num_copies == 1) |
f188591e | 306 | return ret; |
4235298e | 307 | |
f188591e | 308 | mirror_num++; |
4235298e | 309 | if (mirror_num > num_copies) |
f188591e | 310 | return ret; |
f188591e | 311 | } |
f188591e CM |
312 | return -EIO; |
313 | } | |
19c00ddc | 314 | |
d352ac68 | 315 | /* |
d397712b CM |
316 | * checksum a dirty tree block before IO. This has extra checks to make sure |
317 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 318 | */ |
d397712b | 319 | |
b2950863 | 320 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 321 | { |
d1310b2e | 322 | struct extent_io_tree *tree; |
35ebb934 | 323 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc CM |
324 | u64 found_start; |
325 | int found_level; | |
326 | unsigned long len; | |
327 | struct extent_buffer *eb; | |
f188591e CM |
328 | int ret; |
329 | ||
d1310b2e | 330 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc CM |
331 | |
332 | if (page->private == EXTENT_PAGE_PRIVATE) | |
333 | goto out; | |
334 | if (!page->private) | |
335 | goto out; | |
336 | len = page->private >> 2; | |
d397712b CM |
337 | WARN_ON(len == 0); |
338 | ||
19c00ddc | 339 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); |
ca7a79ad CM |
340 | ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, |
341 | btrfs_header_generation(eb)); | |
f188591e | 342 | BUG_ON(ret); |
19c00ddc CM |
343 | found_start = btrfs_header_bytenr(eb); |
344 | if (found_start != start) { | |
55c69072 CM |
345 | WARN_ON(1); |
346 | goto err; | |
347 | } | |
348 | if (eb->first_page != page) { | |
55c69072 CM |
349 | WARN_ON(1); |
350 | goto err; | |
351 | } | |
352 | if (!PageUptodate(page)) { | |
55c69072 CM |
353 | WARN_ON(1); |
354 | goto err; | |
19c00ddc CM |
355 | } |
356 | found_level = btrfs_header_level(eb); | |
4bef0848 | 357 | |
19c00ddc | 358 | csum_tree_block(root, eb, 0); |
55c69072 | 359 | err: |
19c00ddc CM |
360 | free_extent_buffer(eb); |
361 | out: | |
362 | return 0; | |
363 | } | |
364 | ||
2b82032c YZ |
365 | static int check_tree_block_fsid(struct btrfs_root *root, |
366 | struct extent_buffer *eb) | |
367 | { | |
368 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
369 | u8 fsid[BTRFS_UUID_SIZE]; | |
370 | int ret = 1; | |
371 | ||
372 | read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), | |
373 | BTRFS_FSID_SIZE); | |
374 | while (fs_devices) { | |
375 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
376 | ret = 0; | |
377 | break; | |
378 | } | |
379 | fs_devices = fs_devices->seed; | |
380 | } | |
381 | return ret; | |
382 | } | |
383 | ||
4008c04a CM |
384 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
385 | void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, int level) | |
386 | { | |
387 | lockdep_set_class_and_name(&eb->lock, | |
388 | &btrfs_eb_class[level], | |
389 | btrfs_eb_name[level]); | |
390 | } | |
391 | #endif | |
392 | ||
b2950863 | 393 | static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
ce9adaa5 CM |
394 | struct extent_state *state) |
395 | { | |
396 | struct extent_io_tree *tree; | |
397 | u64 found_start; | |
398 | int found_level; | |
399 | unsigned long len; | |
400 | struct extent_buffer *eb; | |
401 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 402 | int ret = 0; |
ce9adaa5 CM |
403 | |
404 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
405 | if (page->private == EXTENT_PAGE_PRIVATE) | |
406 | goto out; | |
407 | if (!page->private) | |
408 | goto out; | |
d397712b | 409 | |
ce9adaa5 | 410 | len = page->private >> 2; |
d397712b CM |
411 | WARN_ON(len == 0); |
412 | ||
ce9adaa5 | 413 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); |
f188591e | 414 | |
ce9adaa5 | 415 | found_start = btrfs_header_bytenr(eb); |
23a07867 | 416 | if (found_start != start) { |
d397712b | 417 | printk(KERN_INFO "btrfs bad tree block start %llu %llu\n", |
a1b32a59 CM |
418 | (unsigned long long)found_start, |
419 | (unsigned long long)eb->start); | |
f188591e | 420 | ret = -EIO; |
ce9adaa5 CM |
421 | goto err; |
422 | } | |
423 | if (eb->first_page != page) { | |
d397712b CM |
424 | printk(KERN_INFO "btrfs bad first page %lu %lu\n", |
425 | eb->first_page->index, page->index); | |
ce9adaa5 | 426 | WARN_ON(1); |
f188591e | 427 | ret = -EIO; |
ce9adaa5 CM |
428 | goto err; |
429 | } | |
2b82032c | 430 | if (check_tree_block_fsid(root, eb)) { |
d397712b CM |
431 | printk(KERN_INFO "btrfs bad fsid on block %llu\n", |
432 | (unsigned long long)eb->start); | |
1259ab75 CM |
433 | ret = -EIO; |
434 | goto err; | |
435 | } | |
ce9adaa5 CM |
436 | found_level = btrfs_header_level(eb); |
437 | ||
4008c04a CM |
438 | btrfs_set_buffer_lockdep_class(eb, found_level); |
439 | ||
ce9adaa5 | 440 | ret = csum_tree_block(root, eb, 1); |
f188591e CM |
441 | if (ret) |
442 | ret = -EIO; | |
ce9adaa5 CM |
443 | |
444 | end = min_t(u64, eb->len, PAGE_CACHE_SIZE); | |
445 | end = eb->start + end - 1; | |
ce9adaa5 CM |
446 | err: |
447 | free_extent_buffer(eb); | |
448 | out: | |
f188591e | 449 | return ret; |
ce9adaa5 CM |
450 | } |
451 | ||
ce9adaa5 | 452 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
453 | { |
454 | struct end_io_wq *end_io_wq = bio->bi_private; | |
455 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 456 | |
ce9adaa5 | 457 | fs_info = end_io_wq->info; |
ce9adaa5 | 458 | end_io_wq->error = err; |
8b712842 CM |
459 | end_io_wq->work.func = end_workqueue_fn; |
460 | end_io_wq->work.flags = 0; | |
d20f7043 CM |
461 | |
462 | if (bio->bi_rw & (1 << BIO_RW)) { | |
cad321ad CM |
463 | if (end_io_wq->metadata) |
464 | btrfs_queue_worker(&fs_info->endio_meta_write_workers, | |
465 | &end_io_wq->work); | |
466 | else | |
467 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
468 | &end_io_wq->work); | |
d20f7043 CM |
469 | } else { |
470 | if (end_io_wq->metadata) | |
471 | btrfs_queue_worker(&fs_info->endio_meta_workers, | |
472 | &end_io_wq->work); | |
473 | else | |
474 | btrfs_queue_worker(&fs_info->endio_workers, | |
475 | &end_io_wq->work); | |
476 | } | |
ce9adaa5 CM |
477 | } |
478 | ||
22c59948 CM |
479 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
480 | int metadata) | |
0b86a832 | 481 | { |
ce9adaa5 | 482 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
483 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
484 | if (!end_io_wq) | |
485 | return -ENOMEM; | |
486 | ||
487 | end_io_wq->private = bio->bi_private; | |
488 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 489 | end_io_wq->info = info; |
ce9adaa5 CM |
490 | end_io_wq->error = 0; |
491 | end_io_wq->bio = bio; | |
22c59948 | 492 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
493 | |
494 | bio->bi_private = end_io_wq; | |
495 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
496 | return 0; |
497 | } | |
498 | ||
b64a2851 | 499 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 500 | { |
4854ddd0 CM |
501 | unsigned long limit = min_t(unsigned long, |
502 | info->workers.max_workers, | |
503 | info->fs_devices->open_devices); | |
504 | return 256 * limit; | |
505 | } | |
0986fe9e | 506 | |
4854ddd0 CM |
507 | int btrfs_congested_async(struct btrfs_fs_info *info, int iodone) |
508 | { | |
b64a2851 CM |
509 | return atomic_read(&info->nr_async_bios) > |
510 | btrfs_async_submit_limit(info); | |
0986fe9e CM |
511 | } |
512 | ||
4a69a410 CM |
513 | static void run_one_async_start(struct btrfs_work *work) |
514 | { | |
515 | struct btrfs_fs_info *fs_info; | |
516 | struct async_submit_bio *async; | |
517 | ||
518 | async = container_of(work, struct async_submit_bio, work); | |
519 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
520 | async->submit_bio_start(async->inode, async->rw, async->bio, | |
521 | async->mirror_num, async->bio_flags); | |
522 | } | |
523 | ||
524 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
525 | { |
526 | struct btrfs_fs_info *fs_info; | |
527 | struct async_submit_bio *async; | |
4854ddd0 | 528 | int limit; |
8b712842 CM |
529 | |
530 | async = container_of(work, struct async_submit_bio, work); | |
531 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 532 | |
b64a2851 | 533 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
534 | limit = limit * 2 / 3; |
535 | ||
8b712842 | 536 | atomic_dec(&fs_info->nr_async_submits); |
0986fe9e | 537 | |
b64a2851 CM |
538 | if (atomic_read(&fs_info->nr_async_submits) < limit && |
539 | waitqueue_active(&fs_info->async_submit_wait)) | |
4854ddd0 CM |
540 | wake_up(&fs_info->async_submit_wait); |
541 | ||
4a69a410 | 542 | async->submit_bio_done(async->inode, async->rw, async->bio, |
c8b97818 | 543 | async->mirror_num, async->bio_flags); |
4a69a410 CM |
544 | } |
545 | ||
546 | static void run_one_async_free(struct btrfs_work *work) | |
547 | { | |
548 | struct async_submit_bio *async; | |
549 | ||
550 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
551 | kfree(async); |
552 | } | |
553 | ||
44b8bd7e CM |
554 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
555 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 556 | unsigned long bio_flags, |
4a69a410 CM |
557 | extent_submit_bio_hook_t *submit_bio_start, |
558 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
559 | { |
560 | struct async_submit_bio *async; | |
561 | ||
562 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
563 | if (!async) | |
564 | return -ENOMEM; | |
565 | ||
566 | async->inode = inode; | |
567 | async->rw = rw; | |
568 | async->bio = bio; | |
569 | async->mirror_num = mirror_num; | |
4a69a410 CM |
570 | async->submit_bio_start = submit_bio_start; |
571 | async->submit_bio_done = submit_bio_done; | |
572 | ||
573 | async->work.func = run_one_async_start; | |
574 | async->work.ordered_func = run_one_async_done; | |
575 | async->work.ordered_free = run_one_async_free; | |
576 | ||
8b712842 | 577 | async->work.flags = 0; |
c8b97818 | 578 | async->bio_flags = bio_flags; |
8c8bee1d | 579 | |
cb03c743 | 580 | atomic_inc(&fs_info->nr_async_submits); |
8b712842 | 581 | btrfs_queue_worker(&fs_info->workers, &async->work); |
d2c3f4f6 | 582 | #if 0 |
7c2ca468 | 583 | int limit = btrfs_async_submit_limit(fs_info); |
9473f16c CM |
584 | if (atomic_read(&fs_info->nr_async_submits) > limit) { |
585 | wait_event_timeout(fs_info->async_submit_wait, | |
4854ddd0 CM |
586 | (atomic_read(&fs_info->nr_async_submits) < limit), |
587 | HZ/10); | |
9473f16c CM |
588 | |
589 | wait_event_timeout(fs_info->async_submit_wait, | |
590 | (atomic_read(&fs_info->nr_async_bios) < limit), | |
591 | HZ/10); | |
592 | } | |
d2c3f4f6 | 593 | #endif |
d397712b | 594 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
595 | atomic_read(&fs_info->nr_async_submits)) { |
596 | wait_event(fs_info->async_submit_wait, | |
597 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
598 | } | |
599 | ||
44b8bd7e CM |
600 | return 0; |
601 | } | |
602 | ||
ce3ed71a CM |
603 | static int btree_csum_one_bio(struct bio *bio) |
604 | { | |
605 | struct bio_vec *bvec = bio->bi_io_vec; | |
606 | int bio_index = 0; | |
607 | struct btrfs_root *root; | |
608 | ||
609 | WARN_ON(bio->bi_vcnt <= 0); | |
d397712b | 610 | while (bio_index < bio->bi_vcnt) { |
ce3ed71a CM |
611 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
612 | csum_dirty_buffer(root, bvec->bv_page); | |
613 | bio_index++; | |
614 | bvec++; | |
615 | } | |
616 | return 0; | |
617 | } | |
618 | ||
4a69a410 CM |
619 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
620 | struct bio *bio, int mirror_num, | |
621 | unsigned long bio_flags) | |
22c59948 | 622 | { |
8b712842 CM |
623 | /* |
624 | * when we're called for a write, we're already in the async | |
5443be45 | 625 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 626 | */ |
4a69a410 CM |
627 | btree_csum_one_bio(bio); |
628 | return 0; | |
629 | } | |
22c59948 | 630 | |
4a69a410 CM |
631 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
632 | int mirror_num, unsigned long bio_flags) | |
633 | { | |
8b712842 | 634 | /* |
4a69a410 CM |
635 | * when we're called for a write, we're already in the async |
636 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 637 | */ |
8b712842 | 638 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
639 | } |
640 | ||
44b8bd7e | 641 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
c8b97818 | 642 | int mirror_num, unsigned long bio_flags) |
44b8bd7e | 643 | { |
cad321ad CM |
644 | int ret; |
645 | ||
646 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, | |
647 | bio, 1); | |
648 | BUG_ON(ret); | |
649 | ||
44b8bd7e | 650 | if (!(rw & (1 << BIO_RW))) { |
4a69a410 CM |
651 | /* |
652 | * called for a read, do the setup so that checksum validation | |
653 | * can happen in the async kernel threads | |
654 | */ | |
4a69a410 | 655 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, |
6f3577bd | 656 | mirror_num, 0); |
44b8bd7e | 657 | } |
cad321ad CM |
658 | /* |
659 | * kthread helpers are used to submit writes so that checksumming | |
660 | * can happen in parallel across all CPUs | |
661 | */ | |
44b8bd7e | 662 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, |
c8b97818 | 663 | inode, rw, bio, mirror_num, 0, |
4a69a410 CM |
664 | __btree_submit_bio_start, |
665 | __btree_submit_bio_done); | |
44b8bd7e CM |
666 | } |
667 | ||
0da5468f CM |
668 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
669 | { | |
d1310b2e CM |
670 | struct extent_io_tree *tree; |
671 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5443be45 CM |
672 | |
673 | if (current->flags & PF_MEMALLOC) { | |
674 | redirty_page_for_writepage(wbc, page); | |
675 | unlock_page(page); | |
676 | return 0; | |
677 | } | |
5f39d397 CM |
678 | return extent_write_full_page(tree, page, btree_get_extent, wbc); |
679 | } | |
0da5468f CM |
680 | |
681 | static int btree_writepages(struct address_space *mapping, | |
682 | struct writeback_control *wbc) | |
683 | { | |
d1310b2e CM |
684 | struct extent_io_tree *tree; |
685 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 686 | if (wbc->sync_mode == WB_SYNC_NONE) { |
793955bc CM |
687 | u64 num_dirty; |
688 | u64 start = 0; | |
24ab9cd8 | 689 | unsigned long thresh = 32 * 1024 * 1024; |
448d640b CM |
690 | |
691 | if (wbc->for_kupdate) | |
692 | return 0; | |
693 | ||
1832a6d5 CM |
694 | num_dirty = count_range_bits(tree, &start, (u64)-1, |
695 | thresh, EXTENT_DIRTY); | |
d397712b | 696 | if (num_dirty < thresh) |
793955bc | 697 | return 0; |
793955bc | 698 | } |
0da5468f CM |
699 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
700 | } | |
701 | ||
b2950863 | 702 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 703 | { |
d1310b2e CM |
704 | struct extent_io_tree *tree; |
705 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
706 | return extent_read_full_page(tree, page, btree_get_extent); |
707 | } | |
22b0ebda | 708 | |
70dec807 | 709 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 710 | { |
d1310b2e CM |
711 | struct extent_io_tree *tree; |
712 | struct extent_map_tree *map; | |
5f39d397 | 713 | int ret; |
d98237b3 | 714 | |
98509cfc | 715 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 716 | return 0; |
98509cfc | 717 | |
d1310b2e CM |
718 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
719 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
6af118ce | 720 | |
7b13b7b1 | 721 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
d397712b | 722 | if (!ret) |
6af118ce | 723 | return 0; |
6af118ce CM |
724 | |
725 | ret = try_release_extent_buffer(tree, page); | |
5f39d397 CM |
726 | if (ret == 1) { |
727 | ClearPagePrivate(page); | |
728 | set_page_private(page, 0); | |
729 | page_cache_release(page); | |
730 | } | |
6af118ce | 731 | |
d98237b3 CM |
732 | return ret; |
733 | } | |
734 | ||
5f39d397 | 735 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 736 | { |
d1310b2e CM |
737 | struct extent_io_tree *tree; |
738 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
739 | extent_invalidatepage(tree, page, offset); |
740 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 741 | if (PagePrivate(page)) { |
d397712b CM |
742 | printk(KERN_WARNING "btrfs warning page private not zero " |
743 | "on page %llu\n", (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
744 | ClearPagePrivate(page); |
745 | set_page_private(page, 0); | |
746 | page_cache_release(page); | |
747 | } | |
d98237b3 CM |
748 | } |
749 | ||
5f39d397 | 750 | #if 0 |
d98237b3 | 751 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
ed2ff2cb | 752 | { |
87cbda5c | 753 | struct buffer_head *bh; |
0f7d52f4 | 754 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
87cbda5c | 755 | struct buffer_head *head; |
87cbda5c CM |
756 | if (!page_has_buffers(page)) { |
757 | create_empty_buffers(page, root->fs_info->sb->s_blocksize, | |
758 | (1 << BH_Dirty)|(1 << BH_Uptodate)); | |
759 | } | |
760 | head = page_buffers(page); | |
761 | bh = head; | |
762 | do { | |
763 | if (buffer_dirty(bh)) | |
764 | csum_tree_block(root, bh, 0); | |
765 | bh = bh->b_this_page; | |
766 | } while (bh != head); | |
d98237b3 | 767 | return block_write_full_page(page, btree_get_block, wbc); |
ed2ff2cb | 768 | } |
5f39d397 | 769 | #endif |
eb60ceac | 770 | |
d98237b3 CM |
771 | static struct address_space_operations btree_aops = { |
772 | .readpage = btree_readpage, | |
773 | .writepage = btree_writepage, | |
0da5468f | 774 | .writepages = btree_writepages, |
5f39d397 CM |
775 | .releasepage = btree_releasepage, |
776 | .invalidatepage = btree_invalidatepage, | |
d98237b3 CM |
777 | .sync_page = block_sync_page, |
778 | }; | |
779 | ||
ca7a79ad CM |
780 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
781 | u64 parent_transid) | |
090d1875 | 782 | { |
5f39d397 CM |
783 | struct extent_buffer *buf = NULL; |
784 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 785 | int ret = 0; |
090d1875 | 786 | |
db94535d | 787 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 788 | if (!buf) |
090d1875 | 789 | return 0; |
d1310b2e | 790 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
f188591e | 791 | buf, 0, 0, btree_get_extent, 0); |
5f39d397 | 792 | free_extent_buffer(buf); |
de428b63 | 793 | return ret; |
090d1875 CM |
794 | } |
795 | ||
0999df54 CM |
796 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
797 | u64 bytenr, u32 blocksize) | |
798 | { | |
799 | struct inode *btree_inode = root->fs_info->btree_inode; | |
800 | struct extent_buffer *eb; | |
801 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
802 | bytenr, blocksize, GFP_NOFS); | |
803 | return eb; | |
804 | } | |
805 | ||
806 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
807 | u64 bytenr, u32 blocksize) | |
808 | { | |
809 | struct inode *btree_inode = root->fs_info->btree_inode; | |
810 | struct extent_buffer *eb; | |
811 | ||
812 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
813 | bytenr, blocksize, NULL, GFP_NOFS); | |
814 | return eb; | |
815 | } | |
816 | ||
817 | ||
e02119d5 CM |
818 | int btrfs_write_tree_block(struct extent_buffer *buf) |
819 | { | |
820 | return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start, | |
24ab9cd8 | 821 | buf->start + buf->len - 1, WB_SYNC_ALL); |
e02119d5 CM |
822 | } |
823 | ||
824 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
825 | { | |
826 | return btrfs_wait_on_page_writeback_range(buf->first_page->mapping, | |
d397712b | 827 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
828 | } |
829 | ||
0999df54 | 830 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 831 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
832 | { |
833 | struct extent_buffer *buf = NULL; | |
834 | struct inode *btree_inode = root->fs_info->btree_inode; | |
835 | struct extent_io_tree *io_tree; | |
836 | int ret; | |
837 | ||
838 | io_tree = &BTRFS_I(btree_inode)->io_tree; | |
839 | ||
840 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
841 | if (!buf) | |
842 | return NULL; | |
0999df54 | 843 | |
ca7a79ad | 844 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 | 845 | |
d397712b | 846 | if (ret == 0) |
b4ce94de | 847 | set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags); |
d397712b | 848 | else |
a1b32a59 | 849 | WARN_ON(1); |
5f39d397 | 850 | return buf; |
ce9adaa5 | 851 | |
eb60ceac CM |
852 | } |
853 | ||
e089f05c | 854 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 855 | struct extent_buffer *buf) |
ed2ff2cb | 856 | { |
5f39d397 | 857 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 | 858 | if (btrfs_header_generation(buf) == |
925baedd CM |
859 | root->fs_info->running_transaction->transid) { |
860 | WARN_ON(!btrfs_tree_locked(buf)); | |
b4ce94de CM |
861 | |
862 | /* ugh, clear_extent_buffer_dirty can be expensive */ | |
863 | btrfs_set_lock_blocking(buf); | |
864 | ||
d1310b2e | 865 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 866 | buf); |
925baedd | 867 | } |
5f39d397 CM |
868 | return 0; |
869 | } | |
870 | ||
db94535d | 871 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 872 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 873 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 874 | u64 objectid) |
d97e63b6 | 875 | { |
cfaa7295 | 876 | root->node = NULL; |
a28ec197 | 877 | root->commit_root = NULL; |
31153d81 | 878 | root->ref_tree = NULL; |
db94535d CM |
879 | root->sectorsize = sectorsize; |
880 | root->nodesize = nodesize; | |
881 | root->leafsize = leafsize; | |
87ee04eb | 882 | root->stripesize = stripesize; |
123abc88 | 883 | root->ref_cows = 0; |
0b86a832 CM |
884 | root->track_dirty = 0; |
885 | ||
9f5fae2f | 886 | root->fs_info = fs_info; |
0f7d52f4 CM |
887 | root->objectid = objectid; |
888 | root->last_trans = 0; | |
1b05da2e CM |
889 | root->highest_inode = 0; |
890 | root->last_inode_alloc = 0; | |
58176a96 | 891 | root->name = NULL; |
4313b399 | 892 | root->in_sysfs = 0; |
0b86a832 CM |
893 | |
894 | INIT_LIST_HEAD(&root->dirty_list); | |
7b128766 | 895 | INIT_LIST_HEAD(&root->orphan_list); |
bcc63abb | 896 | INIT_LIST_HEAD(&root->dead_list); |
925baedd | 897 | spin_lock_init(&root->node_lock); |
bcc63abb | 898 | spin_lock_init(&root->list_lock); |
a2135011 | 899 | mutex_init(&root->objectid_mutex); |
e02119d5 | 900 | mutex_init(&root->log_mutex); |
7237f183 YZ |
901 | init_waitqueue_head(&root->log_writer_wait); |
902 | init_waitqueue_head(&root->log_commit_wait[0]); | |
903 | init_waitqueue_head(&root->log_commit_wait[1]); | |
904 | atomic_set(&root->log_commit[0], 0); | |
905 | atomic_set(&root->log_commit[1], 0); | |
906 | atomic_set(&root->log_writers, 0); | |
907 | root->log_batch = 0; | |
908 | root->log_transid = 0; | |
d0c803c4 CM |
909 | extent_io_tree_init(&root->dirty_log_pages, |
910 | fs_info->btree_inode->i_mapping, GFP_NOFS); | |
017e5369 CM |
911 | |
912 | btrfs_leaf_ref_tree_init(&root->ref_tree_struct); | |
913 | root->ref_tree = &root->ref_tree_struct; | |
914 | ||
3768f368 CM |
915 | memset(&root->root_key, 0, sizeof(root->root_key)); |
916 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 917 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 918 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 919 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 920 | init_completion(&root->kobj_unregister); |
6702ed49 CM |
921 | root->defrag_running = 0; |
922 | root->defrag_level = 0; | |
4d775673 | 923 | root->root_key.objectid = objectid; |
3394e160 CM |
924 | root->anon_super.s_root = NULL; |
925 | root->anon_super.s_dev = 0; | |
926 | INIT_LIST_HEAD(&root->anon_super.s_list); | |
927 | INIT_LIST_HEAD(&root->anon_super.s_instances); | |
928 | init_rwsem(&root->anon_super.s_umount); | |
929 | ||
3768f368 CM |
930 | return 0; |
931 | } | |
932 | ||
db94535d | 933 | static int find_and_setup_root(struct btrfs_root *tree_root, |
9f5fae2f CM |
934 | struct btrfs_fs_info *fs_info, |
935 | u64 objectid, | |
e20d96d6 | 936 | struct btrfs_root *root) |
3768f368 CM |
937 | { |
938 | int ret; | |
db94535d | 939 | u32 blocksize; |
84234f3a | 940 | u64 generation; |
3768f368 | 941 | |
db94535d | 942 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
943 | tree_root->sectorsize, tree_root->stripesize, |
944 | root, fs_info, objectid); | |
3768f368 CM |
945 | ret = btrfs_find_last_root(tree_root, objectid, |
946 | &root->root_item, &root->root_key); | |
947 | BUG_ON(ret); | |
948 | ||
84234f3a | 949 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
950 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
951 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 952 | blocksize, generation); |
3768f368 | 953 | BUG_ON(!root->node); |
d97e63b6 CM |
954 | return 0; |
955 | } | |
956 | ||
e02119d5 CM |
957 | int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans, |
958 | struct btrfs_fs_info *fs_info) | |
959 | { | |
960 | struct extent_buffer *eb; | |
d0c803c4 CM |
961 | struct btrfs_root *log_root_tree = fs_info->log_root_tree; |
962 | u64 start = 0; | |
963 | u64 end = 0; | |
e02119d5 CM |
964 | int ret; |
965 | ||
d0c803c4 | 966 | if (!log_root_tree) |
e02119d5 CM |
967 | return 0; |
968 | ||
d397712b | 969 | while (1) { |
d0c803c4 CM |
970 | ret = find_first_extent_bit(&log_root_tree->dirty_log_pages, |
971 | 0, &start, &end, EXTENT_DIRTY); | |
972 | if (ret) | |
973 | break; | |
974 | ||
975 | clear_extent_dirty(&log_root_tree->dirty_log_pages, | |
976 | start, end, GFP_NOFS); | |
977 | } | |
e02119d5 CM |
978 | eb = fs_info->log_root_tree->node; |
979 | ||
980 | WARN_ON(btrfs_header_level(eb) != 0); | |
981 | WARN_ON(btrfs_header_nritems(eb) != 0); | |
982 | ||
d00aff00 CM |
983 | ret = btrfs_free_reserved_extent(fs_info->tree_root, |
984 | eb->start, eb->len); | |
e02119d5 CM |
985 | BUG_ON(ret); |
986 | ||
987 | free_extent_buffer(eb); | |
988 | kfree(fs_info->log_root_tree); | |
989 | fs_info->log_root_tree = NULL; | |
990 | return 0; | |
991 | } | |
992 | ||
7237f183 YZ |
993 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
994 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
995 | { |
996 | struct btrfs_root *root; | |
997 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 998 | struct extent_buffer *leaf; |
e02119d5 CM |
999 | |
1000 | root = kzalloc(sizeof(*root), GFP_NOFS); | |
1001 | if (!root) | |
7237f183 | 1002 | return ERR_PTR(-ENOMEM); |
e02119d5 CM |
1003 | |
1004 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1005 | tree_root->sectorsize, tree_root->stripesize, | |
1006 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1007 | ||
1008 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1009 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1010 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
7237f183 YZ |
1011 | /* |
1012 | * log trees do not get reference counted because they go away | |
1013 | * before a real commit is actually done. They do store pointers | |
1014 | * to file data extents, and those reference counts still get | |
1015 | * updated (along with back refs to the log tree). | |
1016 | */ | |
e02119d5 CM |
1017 | root->ref_cows = 0; |
1018 | ||
7237f183 YZ |
1019 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, |
1020 | 0, BTRFS_TREE_LOG_OBJECTID, | |
1021 | trans->transid, 0, 0, 0); | |
1022 | if (IS_ERR(leaf)) { | |
1023 | kfree(root); | |
1024 | return ERR_CAST(leaf); | |
1025 | } | |
e02119d5 | 1026 | |
7237f183 | 1027 | root->node = leaf; |
e02119d5 CM |
1028 | btrfs_set_header_nritems(root->node, 0); |
1029 | btrfs_set_header_level(root->node, 0); | |
1030 | btrfs_set_header_bytenr(root->node, root->node->start); | |
1031 | btrfs_set_header_generation(root->node, trans->transid); | |
1032 | btrfs_set_header_owner(root->node, BTRFS_TREE_LOG_OBJECTID); | |
1033 | ||
1034 | write_extent_buffer(root->node, root->fs_info->fsid, | |
1035 | (unsigned long)btrfs_header_fsid(root->node), | |
1036 | BTRFS_FSID_SIZE); | |
1037 | btrfs_mark_buffer_dirty(root->node); | |
1038 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1039 | return root; |
1040 | } | |
1041 | ||
1042 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1043 | struct btrfs_fs_info *fs_info) | |
1044 | { | |
1045 | struct btrfs_root *log_root; | |
1046 | ||
1047 | log_root = alloc_log_tree(trans, fs_info); | |
1048 | if (IS_ERR(log_root)) | |
1049 | return PTR_ERR(log_root); | |
1050 | WARN_ON(fs_info->log_root_tree); | |
1051 | fs_info->log_root_tree = log_root; | |
1052 | return 0; | |
1053 | } | |
1054 | ||
1055 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1056 | struct btrfs_root *root) | |
1057 | { | |
1058 | struct btrfs_root *log_root; | |
1059 | struct btrfs_inode_item *inode_item; | |
1060 | ||
1061 | log_root = alloc_log_tree(trans, root->fs_info); | |
1062 | if (IS_ERR(log_root)) | |
1063 | return PTR_ERR(log_root); | |
1064 | ||
1065 | log_root->last_trans = trans->transid; | |
1066 | log_root->root_key.offset = root->root_key.objectid; | |
1067 | ||
1068 | inode_item = &log_root->root_item.inode; | |
1069 | inode_item->generation = cpu_to_le64(1); | |
1070 | inode_item->size = cpu_to_le64(3); | |
1071 | inode_item->nlink = cpu_to_le32(1); | |
1072 | inode_item->nbytes = cpu_to_le64(root->leafsize); | |
1073 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); | |
1074 | ||
1075 | btrfs_set_root_bytenr(&log_root->root_item, log_root->node->start); | |
1076 | btrfs_set_root_generation(&log_root->root_item, trans->transid); | |
1077 | ||
1078 | WARN_ON(root->log_root); | |
1079 | root->log_root = log_root; | |
1080 | root->log_transid = 0; | |
e02119d5 CM |
1081 | return 0; |
1082 | } | |
1083 | ||
1084 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
1085 | struct btrfs_key *location) | |
1086 | { | |
1087 | struct btrfs_root *root; | |
1088 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1089 | struct btrfs_path *path; |
5f39d397 | 1090 | struct extent_buffer *l; |
1b05da2e | 1091 | u64 highest_inode; |
84234f3a | 1092 | u64 generation; |
db94535d | 1093 | u32 blocksize; |
0f7d52f4 CM |
1094 | int ret = 0; |
1095 | ||
5eda7b5e | 1096 | root = kzalloc(sizeof(*root), GFP_NOFS); |
0cf6c620 | 1097 | if (!root) |
0f7d52f4 | 1098 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 1099 | if (location->offset == (u64)-1) { |
db94535d | 1100 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
1101 | location->objectid, root); |
1102 | if (ret) { | |
0f7d52f4 CM |
1103 | kfree(root); |
1104 | return ERR_PTR(ret); | |
1105 | } | |
1106 | goto insert; | |
1107 | } | |
1108 | ||
db94535d | 1109 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1110 | tree_root->sectorsize, tree_root->stripesize, |
1111 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
1112 | |
1113 | path = btrfs_alloc_path(); | |
1114 | BUG_ON(!path); | |
1115 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); | |
1116 | if (ret != 0) { | |
0f7d52f4 CM |
1117 | if (ret > 0) |
1118 | ret = -ENOENT; | |
1119 | goto out; | |
1120 | } | |
5f39d397 CM |
1121 | l = path->nodes[0]; |
1122 | read_extent_buffer(l, &root->root_item, | |
1123 | btrfs_item_ptr_offset(l, path->slots[0]), | |
0f7d52f4 | 1124 | sizeof(root->root_item)); |
44b36eb2 | 1125 | memcpy(&root->root_key, location, sizeof(*location)); |
0f7d52f4 CM |
1126 | ret = 0; |
1127 | out: | |
1128 | btrfs_release_path(root, path); | |
1129 | btrfs_free_path(path); | |
1130 | if (ret) { | |
1131 | kfree(root); | |
1132 | return ERR_PTR(ret); | |
1133 | } | |
84234f3a | 1134 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1135 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1136 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1137 | blocksize, generation); |
0f7d52f4 CM |
1138 | BUG_ON(!root->node); |
1139 | insert: | |
e02119d5 CM |
1140 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { |
1141 | root->ref_cows = 1; | |
1142 | ret = btrfs_find_highest_inode(root, &highest_inode); | |
1143 | if (ret == 0) { | |
1144 | root->highest_inode = highest_inode; | |
1145 | root->last_inode_alloc = highest_inode; | |
1146 | } | |
5eda7b5e CM |
1147 | } |
1148 | return root; | |
1149 | } | |
1150 | ||
dc17ff8f CM |
1151 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1152 | u64 root_objectid) | |
1153 | { | |
1154 | struct btrfs_root *root; | |
1155 | ||
1156 | if (root_objectid == BTRFS_ROOT_TREE_OBJECTID) | |
1157 | return fs_info->tree_root; | |
1158 | if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1159 | return fs_info->extent_root; | |
1160 | ||
1161 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1162 | (unsigned long)root_objectid); | |
1163 | return root; | |
1164 | } | |
1165 | ||
edbd8d4e CM |
1166 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
1167 | struct btrfs_key *location) | |
5eda7b5e CM |
1168 | { |
1169 | struct btrfs_root *root; | |
1170 | int ret; | |
1171 | ||
edbd8d4e CM |
1172 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1173 | return fs_info->tree_root; | |
1174 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1175 | return fs_info->extent_root; | |
8f18cf13 CM |
1176 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1177 | return fs_info->chunk_root; | |
1178 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1179 | return fs_info->dev_root; | |
0403e47e YZ |
1180 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1181 | return fs_info->csum_root; | |
edbd8d4e | 1182 | |
5eda7b5e CM |
1183 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
1184 | (unsigned long)location->objectid); | |
1185 | if (root) | |
1186 | return root; | |
1187 | ||
e02119d5 | 1188 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
1189 | if (IS_ERR(root)) |
1190 | return root; | |
3394e160 CM |
1191 | |
1192 | set_anon_super(&root->anon_super, NULL); | |
1193 | ||
2619ba1f CM |
1194 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
1195 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 CM |
1196 | root); |
1197 | if (ret) { | |
5f39d397 | 1198 | free_extent_buffer(root->node); |
0f7d52f4 CM |
1199 | kfree(root); |
1200 | return ERR_PTR(ret); | |
1201 | } | |
c146afad YZ |
1202 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
1203 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
1204 | root->root_key.objectid, root); | |
1205 | BUG_ON(ret); | |
1206 | btrfs_orphan_cleanup(root); | |
1207 | } | |
edbd8d4e CM |
1208 | return root; |
1209 | } | |
1210 | ||
1211 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, | |
1212 | struct btrfs_key *location, | |
1213 | const char *name, int namelen) | |
1214 | { | |
1215 | struct btrfs_root *root; | |
1216 | int ret; | |
1217 | ||
1218 | root = btrfs_read_fs_root_no_name(fs_info, location); | |
1219 | if (!root) | |
1220 | return NULL; | |
58176a96 | 1221 | |
4313b399 CM |
1222 | if (root->in_sysfs) |
1223 | return root; | |
1224 | ||
58176a96 JB |
1225 | ret = btrfs_set_root_name(root, name, namelen); |
1226 | if (ret) { | |
5f39d397 | 1227 | free_extent_buffer(root->node); |
58176a96 JB |
1228 | kfree(root); |
1229 | return ERR_PTR(ret); | |
1230 | } | |
ea9e8b11 | 1231 | #if 0 |
58176a96 JB |
1232 | ret = btrfs_sysfs_add_root(root); |
1233 | if (ret) { | |
5f39d397 | 1234 | free_extent_buffer(root->node); |
58176a96 JB |
1235 | kfree(root->name); |
1236 | kfree(root); | |
1237 | return ERR_PTR(ret); | |
1238 | } | |
ea9e8b11 | 1239 | #endif |
4313b399 | 1240 | root->in_sysfs = 1; |
0f7d52f4 CM |
1241 | return root; |
1242 | } | |
04160088 CM |
1243 | |
1244 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) | |
1245 | { | |
1246 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1247 | int ret = 0; | |
04160088 CM |
1248 | struct btrfs_device *device; |
1249 | struct backing_dev_info *bdi; | |
6f3577bd | 1250 | #if 0 |
cb03c743 | 1251 | if ((bdi_bits & (1 << BDI_write_congested)) && |
777e6bd7 | 1252 | btrfs_congested_async(info, 0)) |
cb03c743 | 1253 | return 1; |
6f3577bd | 1254 | #endif |
c6e30871 | 1255 | list_for_each_entry(device, &info->fs_devices->devices, dev_list) { |
dfe25020 CM |
1256 | if (!device->bdev) |
1257 | continue; | |
04160088 CM |
1258 | bdi = blk_get_backing_dev_info(device->bdev); |
1259 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1260 | ret = 1; | |
1261 | break; | |
1262 | } | |
1263 | } | |
1264 | return ret; | |
1265 | } | |
1266 | ||
38b66988 CM |
1267 | /* |
1268 | * this unplugs every device on the box, and it is only used when page | |
1269 | * is null | |
1270 | */ | |
1271 | static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) | |
1272 | { | |
38b66988 CM |
1273 | struct btrfs_device *device; |
1274 | struct btrfs_fs_info *info; | |
1275 | ||
1276 | info = (struct btrfs_fs_info *)bdi->unplug_io_data; | |
c6e30871 | 1277 | list_for_each_entry(device, &info->fs_devices->devices, dev_list) { |
d20f7043 CM |
1278 | if (!device->bdev) |
1279 | continue; | |
1280 | ||
38b66988 | 1281 | bdi = blk_get_backing_dev_info(device->bdev); |
d397712b | 1282 | if (bdi->unplug_io_fn) |
38b66988 | 1283 | bdi->unplug_io_fn(bdi, page); |
38b66988 CM |
1284 | } |
1285 | } | |
1286 | ||
b2950863 | 1287 | static void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) |
04160088 | 1288 | { |
38b66988 | 1289 | struct inode *inode; |
f2d8d74d CM |
1290 | struct extent_map_tree *em_tree; |
1291 | struct extent_map *em; | |
bcbfce8a | 1292 | struct address_space *mapping; |
38b66988 CM |
1293 | u64 offset; |
1294 | ||
bcbfce8a | 1295 | /* the generic O_DIRECT read code does this */ |
9f0ba5bd | 1296 | if (1 || !page) { |
38b66988 CM |
1297 | __unplug_io_fn(bdi, page); |
1298 | return; | |
1299 | } | |
1300 | ||
bcbfce8a CM |
1301 | /* |
1302 | * page->mapping may change at any time. Get a consistent copy | |
1303 | * and use that for everything below | |
1304 | */ | |
1305 | smp_mb(); | |
1306 | mapping = page->mapping; | |
1307 | if (!mapping) | |
1308 | return; | |
1309 | ||
1310 | inode = mapping->host; | |
240d5d48 CM |
1311 | |
1312 | /* | |
1313 | * don't do the expensive searching for a small number of | |
1314 | * devices | |
1315 | */ | |
1316 | if (BTRFS_I(inode)->root->fs_info->fs_devices->open_devices <= 2) { | |
1317 | __unplug_io_fn(bdi, page); | |
1318 | return; | |
1319 | } | |
1320 | ||
38b66988 | 1321 | offset = page_offset(page); |
04160088 | 1322 | |
f2d8d74d CM |
1323 | em_tree = &BTRFS_I(inode)->extent_tree; |
1324 | spin_lock(&em_tree->lock); | |
1325 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); | |
1326 | spin_unlock(&em_tree->lock); | |
89642229 CM |
1327 | if (!em) { |
1328 | __unplug_io_fn(bdi, page); | |
f2d8d74d | 1329 | return; |
89642229 | 1330 | } |
f2d8d74d | 1331 | |
89642229 CM |
1332 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { |
1333 | free_extent_map(em); | |
1334 | __unplug_io_fn(bdi, page); | |
1335 | return; | |
1336 | } | |
f2d8d74d CM |
1337 | offset = offset - em->start; |
1338 | btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree, | |
1339 | em->block_start + offset, page); | |
1340 | free_extent_map(em); | |
04160088 CM |
1341 | } |
1342 | ||
1343 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) | |
1344 | { | |
1345 | bdi_init(bdi); | |
4575c9cc | 1346 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1347 | bdi->state = 0; |
1348 | bdi->capabilities = default_backing_dev_info.capabilities; | |
1349 | bdi->unplug_io_fn = btrfs_unplug_io_fn; | |
1350 | bdi->unplug_io_data = info; | |
1351 | bdi->congested_fn = btrfs_congested_fn; | |
1352 | bdi->congested_data = info; | |
1353 | return 0; | |
1354 | } | |
1355 | ||
ce9adaa5 CM |
1356 | static int bio_ready_for_csum(struct bio *bio) |
1357 | { | |
1358 | u64 length = 0; | |
1359 | u64 buf_len = 0; | |
1360 | u64 start = 0; | |
1361 | struct page *page; | |
1362 | struct extent_io_tree *io_tree = NULL; | |
1363 | struct btrfs_fs_info *info = NULL; | |
1364 | struct bio_vec *bvec; | |
1365 | int i; | |
1366 | int ret; | |
1367 | ||
1368 | bio_for_each_segment(bvec, bio, i) { | |
1369 | page = bvec->bv_page; | |
1370 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
1371 | length += bvec->bv_len; | |
1372 | continue; | |
1373 | } | |
1374 | if (!page->private) { | |
1375 | length += bvec->bv_len; | |
1376 | continue; | |
1377 | } | |
1378 | length = bvec->bv_len; | |
1379 | buf_len = page->private >> 2; | |
1380 | start = page_offset(page) + bvec->bv_offset; | |
1381 | io_tree = &BTRFS_I(page->mapping->host)->io_tree; | |
1382 | info = BTRFS_I(page->mapping->host)->root->fs_info; | |
1383 | } | |
1384 | /* are we fully contained in this bio? */ | |
1385 | if (buf_len <= length) | |
1386 | return 1; | |
1387 | ||
1388 | ret = extent_range_uptodate(io_tree, start + length, | |
1389 | start + buf_len - 1); | |
1390 | if (ret == 1) | |
1391 | return ret; | |
1392 | return ret; | |
1393 | } | |
1394 | ||
8b712842 CM |
1395 | /* |
1396 | * called by the kthread helper functions to finally call the bio end_io | |
1397 | * functions. This is where read checksum verification actually happens | |
1398 | */ | |
1399 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1400 | { |
ce9adaa5 | 1401 | struct bio *bio; |
8b712842 CM |
1402 | struct end_io_wq *end_io_wq; |
1403 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1404 | int error; |
ce9adaa5 | 1405 | |
8b712842 CM |
1406 | end_io_wq = container_of(work, struct end_io_wq, work); |
1407 | bio = end_io_wq->bio; | |
1408 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1409 | |
cad321ad | 1410 | /* metadata bio reads are special because the whole tree block must |
8b712842 CM |
1411 | * be checksummed at once. This makes sure the entire block is in |
1412 | * ram and up to date before trying to verify things. For | |
1413 | * blocksize <= pagesize, it is basically a noop | |
1414 | */ | |
cad321ad CM |
1415 | if (!(bio->bi_rw & (1 << BIO_RW)) && end_io_wq->metadata && |
1416 | !bio_ready_for_csum(bio)) { | |
d20f7043 | 1417 | btrfs_queue_worker(&fs_info->endio_meta_workers, |
8b712842 CM |
1418 | &end_io_wq->work); |
1419 | return; | |
1420 | } | |
1421 | error = end_io_wq->error; | |
1422 | bio->bi_private = end_io_wq->private; | |
1423 | bio->bi_end_io = end_io_wq->end_io; | |
1424 | kfree(end_io_wq); | |
8b712842 | 1425 | bio_endio(bio, error); |
44b8bd7e CM |
1426 | } |
1427 | ||
a74a4b97 CM |
1428 | static int cleaner_kthread(void *arg) |
1429 | { | |
1430 | struct btrfs_root *root = arg; | |
1431 | ||
1432 | do { | |
1433 | smp_mb(); | |
1434 | if (root->fs_info->closing) | |
1435 | break; | |
1436 | ||
1437 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1438 | mutex_lock(&root->fs_info->cleaner_mutex); | |
a74a4b97 | 1439 | btrfs_clean_old_snapshots(root); |
a74a4b97 CM |
1440 | mutex_unlock(&root->fs_info->cleaner_mutex); |
1441 | ||
1442 | if (freezing(current)) { | |
1443 | refrigerator(); | |
1444 | } else { | |
1445 | smp_mb(); | |
1446 | if (root->fs_info->closing) | |
1447 | break; | |
1448 | set_current_state(TASK_INTERRUPTIBLE); | |
1449 | schedule(); | |
1450 | __set_current_state(TASK_RUNNING); | |
1451 | } | |
1452 | } while (!kthread_should_stop()); | |
1453 | return 0; | |
1454 | } | |
1455 | ||
1456 | static int transaction_kthread(void *arg) | |
1457 | { | |
1458 | struct btrfs_root *root = arg; | |
1459 | struct btrfs_trans_handle *trans; | |
1460 | struct btrfs_transaction *cur; | |
1461 | unsigned long now; | |
1462 | unsigned long delay; | |
1463 | int ret; | |
1464 | ||
1465 | do { | |
1466 | smp_mb(); | |
1467 | if (root->fs_info->closing) | |
1468 | break; | |
1469 | ||
1470 | delay = HZ * 30; | |
1471 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1472 | mutex_lock(&root->fs_info->transaction_kthread_mutex); | |
1473 | ||
ab78c84d | 1474 | if (root->fs_info->total_ref_cache_size > 20 * 1024 * 1024) { |
d397712b CM |
1475 | printk(KERN_INFO "btrfs: total reference cache " |
1476 | "size %llu\n", | |
1477 | root->fs_info->total_ref_cache_size); | |
ab78c84d | 1478 | } |
31153d81 | 1479 | |
a74a4b97 CM |
1480 | mutex_lock(&root->fs_info->trans_mutex); |
1481 | cur = root->fs_info->running_transaction; | |
1482 | if (!cur) { | |
1483 | mutex_unlock(&root->fs_info->trans_mutex); | |
1484 | goto sleep; | |
1485 | } | |
31153d81 | 1486 | |
a74a4b97 CM |
1487 | now = get_seconds(); |
1488 | if (now < cur->start_time || now - cur->start_time < 30) { | |
1489 | mutex_unlock(&root->fs_info->trans_mutex); | |
1490 | delay = HZ * 5; | |
1491 | goto sleep; | |
1492 | } | |
1493 | mutex_unlock(&root->fs_info->trans_mutex); | |
a74a4b97 CM |
1494 | trans = btrfs_start_transaction(root, 1); |
1495 | ret = btrfs_commit_transaction(trans, root); | |
1496 | sleep: | |
1497 | wake_up_process(root->fs_info->cleaner_kthread); | |
1498 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1499 | ||
1500 | if (freezing(current)) { | |
1501 | refrigerator(); | |
1502 | } else { | |
1503 | if (root->fs_info->closing) | |
1504 | break; | |
1505 | set_current_state(TASK_INTERRUPTIBLE); | |
1506 | schedule_timeout(delay); | |
1507 | __set_current_state(TASK_RUNNING); | |
1508 | } | |
1509 | } while (!kthread_should_stop()); | |
1510 | return 0; | |
1511 | } | |
1512 | ||
8a4b83cc | 1513 | struct btrfs_root *open_ctree(struct super_block *sb, |
dfe25020 CM |
1514 | struct btrfs_fs_devices *fs_devices, |
1515 | char *options) | |
2e635a27 | 1516 | { |
db94535d CM |
1517 | u32 sectorsize; |
1518 | u32 nodesize; | |
1519 | u32 leafsize; | |
1520 | u32 blocksize; | |
87ee04eb | 1521 | u32 stripesize; |
84234f3a | 1522 | u64 generation; |
f2b636e8 | 1523 | u64 features; |
3de4586c | 1524 | struct btrfs_key location; |
a061fc8d | 1525 | struct buffer_head *bh; |
e02119d5 | 1526 | struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1527 | GFP_NOFS); |
d20f7043 CM |
1528 | struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root), |
1529 | GFP_NOFS); | |
e02119d5 | 1530 | struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1531 | GFP_NOFS); |
8790d502 | 1532 | struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), |
e20d96d6 | 1533 | GFP_NOFS); |
e02119d5 | 1534 | struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1535 | GFP_NOFS); |
e02119d5 | 1536 | struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1537 | GFP_NOFS); |
e02119d5 CM |
1538 | struct btrfs_root *log_tree_root; |
1539 | ||
eb60ceac | 1540 | int ret; |
e58ca020 | 1541 | int err = -EINVAL; |
4543df7e | 1542 | |
2c90e5d6 | 1543 | struct btrfs_super_block *disk_super; |
8790d502 | 1544 | |
0463bb4e | 1545 | if (!extent_root || !tree_root || !fs_info || |
d20f7043 | 1546 | !chunk_root || !dev_root || !csum_root) { |
39279cc3 CM |
1547 | err = -ENOMEM; |
1548 | goto fail; | |
1549 | } | |
0f7d52f4 | 1550 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS); |
8fd17795 | 1551 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1552 | INIT_LIST_HEAD(&fs_info->dead_roots); |
19c00ddc | 1553 | INIT_LIST_HEAD(&fs_info->hashers); |
ea8c2819 | 1554 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
1832a6d5 | 1555 | spin_lock_init(&fs_info->delalloc_lock); |
cee36a03 | 1556 | spin_lock_init(&fs_info->new_trans_lock); |
31153d81 | 1557 | spin_lock_init(&fs_info->ref_cache_lock); |
19c00ddc | 1558 | |
58176a96 | 1559 | init_completion(&fs_info->kobj_unregister); |
9f5fae2f CM |
1560 | fs_info->tree_root = tree_root; |
1561 | fs_info->extent_root = extent_root; | |
d20f7043 | 1562 | fs_info->csum_root = csum_root; |
0b86a832 CM |
1563 | fs_info->chunk_root = chunk_root; |
1564 | fs_info->dev_root = dev_root; | |
8a4b83cc | 1565 | fs_info->fs_devices = fs_devices; |
0b86a832 | 1566 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1567 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1568 | btrfs_mapping_init(&fs_info->mapping_tree); |
cb03c743 | 1569 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 1570 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 1571 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 1572 | atomic_set(&fs_info->nr_async_bios, 0); |
a2135011 | 1573 | atomic_set(&fs_info->throttles, 0); |
ab78c84d | 1574 | atomic_set(&fs_info->throttle_gen, 0); |
e20d96d6 | 1575 | fs_info->sb = sb; |
c59f8951 | 1576 | fs_info->max_extent = (u64)-1; |
6f568d35 | 1577 | fs_info->max_inline = 8192 * 1024; |
04160088 | 1578 | setup_bdi(fs_info, &fs_info->bdi); |
d98237b3 CM |
1579 | fs_info->btree_inode = new_inode(sb); |
1580 | fs_info->btree_inode->i_ino = 1; | |
2c90e5d6 | 1581 | fs_info->btree_inode->i_nlink = 1; |
c8b97818 | 1582 | |
b34b086c CM |
1583 | fs_info->thread_pool_size = min_t(unsigned long, |
1584 | num_online_cpus() + 2, 8); | |
0afbaf8c | 1585 | |
3eaa2885 CM |
1586 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
1587 | spin_lock_init(&fs_info->ordered_extent_lock); | |
1588 | ||
a061fc8d CM |
1589 | sb->s_blocksize = 4096; |
1590 | sb->s_blocksize_bits = blksize_bits(4096); | |
1591 | ||
0afbaf8c CM |
1592 | /* |
1593 | * we set the i_size on the btree inode to the max possible int. | |
1594 | * the real end of the address space is determined by all of | |
1595 | * the devices in the system | |
1596 | */ | |
1597 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 1598 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
1599 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
1600 | ||
d1310b2e | 1601 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
5f39d397 CM |
1602 | fs_info->btree_inode->i_mapping, |
1603 | GFP_NOFS); | |
d1310b2e CM |
1604 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, |
1605 | GFP_NOFS); | |
1606 | ||
1607 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 1608 | |
0f9dd46c JB |
1609 | spin_lock_init(&fs_info->block_group_cache_lock); |
1610 | fs_info->block_group_cache_tree.rb_node = NULL; | |
1611 | ||
d1310b2e | 1612 | extent_io_tree_init(&fs_info->pinned_extents, |
1a5bc167 | 1613 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1614 | extent_io_tree_init(&fs_info->pending_del, |
1a5bc167 | 1615 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1616 | extent_io_tree_init(&fs_info->extent_ins, |
1a5bc167 | 1617 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
e66f709b | 1618 | fs_info->do_barriers = 1; |
e18e4809 | 1619 | |
1a40e23b ZY |
1620 | INIT_LIST_HEAD(&fs_info->dead_reloc_roots); |
1621 | btrfs_leaf_ref_tree_init(&fs_info->reloc_ref_tree); | |
e4657689 ZY |
1622 | btrfs_leaf_ref_tree_init(&fs_info->shared_ref_tree); |
1623 | ||
0f7d52f4 CM |
1624 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
1625 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
1626 | sizeof(struct btrfs_key)); | |
22b0ebda | 1627 | insert_inode_hash(fs_info->btree_inode); |
39279cc3 | 1628 | |
79154b1b | 1629 | mutex_init(&fs_info->trans_mutex); |
e02119d5 | 1630 | mutex_init(&fs_info->tree_log_mutex); |
a2135011 | 1631 | mutex_init(&fs_info->drop_mutex); |
25179201 JB |
1632 | mutex_init(&fs_info->extent_ins_mutex); |
1633 | mutex_init(&fs_info->pinned_mutex); | |
925baedd | 1634 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
1635 | mutex_init(&fs_info->transaction_kthread_mutex); |
1636 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 1637 | mutex_init(&fs_info->volume_mutex); |
1a40e23b | 1638 | mutex_init(&fs_info->tree_reloc_mutex); |
e6dcd2dc | 1639 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 1640 | init_waitqueue_head(&fs_info->transaction_wait); |
4854ddd0 | 1641 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 1642 | |
0b86a832 | 1643 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 1644 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 1645 | |
d98237b3 | 1646 | |
a512bbf8 | 1647 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
a061fc8d | 1648 | if (!bh) |
39279cc3 | 1649 | goto fail_iput; |
39279cc3 | 1650 | |
a061fc8d | 1651 | memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy)); |
2d69a0f8 YZ |
1652 | memcpy(&fs_info->super_for_commit, &fs_info->super_copy, |
1653 | sizeof(fs_info->super_for_commit)); | |
a061fc8d | 1654 | brelse(bh); |
5f39d397 | 1655 | |
a061fc8d | 1656 | memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE); |
0b86a832 | 1657 | |
5f39d397 | 1658 | disk_super = &fs_info->super_copy; |
0f7d52f4 | 1659 | if (!btrfs_super_root(disk_super)) |
c6e2bac1 | 1660 | goto fail_iput; |
0f7d52f4 | 1661 | |
2b82032c YZ |
1662 | ret = btrfs_parse_options(tree_root, options); |
1663 | if (ret) { | |
1664 | err = ret; | |
c6e2bac1 | 1665 | goto fail_iput; |
2b82032c | 1666 | } |
dfe25020 | 1667 | |
f2b636e8 JB |
1668 | features = btrfs_super_incompat_flags(disk_super) & |
1669 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
1670 | if (features) { | |
1671 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
1672 | "unsupported optional features (%Lx).\n", | |
1673 | features); | |
1674 | err = -EINVAL; | |
c6e2bac1 | 1675 | goto fail_iput; |
f2b636e8 JB |
1676 | } |
1677 | ||
1678 | features = btrfs_super_compat_ro_flags(disk_super) & | |
1679 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
1680 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
1681 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
1682 | "unsupported option features (%Lx).\n", | |
1683 | features); | |
1684 | err = -EINVAL; | |
c6e2bac1 | 1685 | goto fail_iput; |
f2b636e8 JB |
1686 | } |
1687 | ||
4543df7e CM |
1688 | /* |
1689 | * we need to start all the end_io workers up front because the | |
1690 | * queue work function gets called at interrupt time, and so it | |
1691 | * cannot dynamically grow. | |
1692 | */ | |
5443be45 CM |
1693 | btrfs_init_workers(&fs_info->workers, "worker", |
1694 | fs_info->thread_pool_size); | |
c8b97818 | 1695 | |
771ed689 CM |
1696 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
1697 | fs_info->thread_pool_size); | |
1698 | ||
5443be45 | 1699 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 CM |
1700 | min_t(u64, fs_devices->num_devices, |
1701 | fs_info->thread_pool_size)); | |
61b49440 CM |
1702 | |
1703 | /* a higher idle thresh on the submit workers makes it much more | |
1704 | * likely that bios will be send down in a sane order to the | |
1705 | * devices | |
1706 | */ | |
1707 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 1708 | |
771ed689 | 1709 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 1710 | fs_info->workers.ordered = 1; |
61b49440 | 1711 | |
771ed689 CM |
1712 | fs_info->delalloc_workers.idle_thresh = 2; |
1713 | fs_info->delalloc_workers.ordered = 1; | |
1714 | ||
5443be45 CM |
1715 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1); |
1716 | btrfs_init_workers(&fs_info->endio_workers, "endio", | |
1717 | fs_info->thread_pool_size); | |
d20f7043 CM |
1718 | btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", |
1719 | fs_info->thread_pool_size); | |
cad321ad CM |
1720 | btrfs_init_workers(&fs_info->endio_meta_write_workers, |
1721 | "endio-meta-write", fs_info->thread_pool_size); | |
5443be45 | 1722 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", |
e6dcd2dc | 1723 | fs_info->thread_pool_size); |
61b49440 CM |
1724 | |
1725 | /* | |
1726 | * endios are largely parallel and should have a very | |
1727 | * low idle thresh | |
1728 | */ | |
1729 | fs_info->endio_workers.idle_thresh = 4; | |
b51912c9 CM |
1730 | fs_info->endio_meta_workers.idle_thresh = 4; |
1731 | ||
3a5f1d45 | 1732 | fs_info->endio_write_workers.idle_thresh = 64; |
cad321ad | 1733 | fs_info->endio_meta_write_workers.idle_thresh = 64; |
61b49440 | 1734 | |
4543df7e | 1735 | btrfs_start_workers(&fs_info->workers, 1); |
1cc127b5 | 1736 | btrfs_start_workers(&fs_info->submit_workers, 1); |
771ed689 | 1737 | btrfs_start_workers(&fs_info->delalloc_workers, 1); |
247e743c | 1738 | btrfs_start_workers(&fs_info->fixup_workers, 1); |
4543df7e | 1739 | btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size); |
d20f7043 CM |
1740 | btrfs_start_workers(&fs_info->endio_meta_workers, |
1741 | fs_info->thread_pool_size); | |
cad321ad CM |
1742 | btrfs_start_workers(&fs_info->endio_meta_write_workers, |
1743 | fs_info->thread_pool_size); | |
e6dcd2dc CM |
1744 | btrfs_start_workers(&fs_info->endio_write_workers, |
1745 | fs_info->thread_pool_size); | |
4543df7e | 1746 | |
4575c9cc | 1747 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
1748 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
1749 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 1750 | |
db94535d CM |
1751 | nodesize = btrfs_super_nodesize(disk_super); |
1752 | leafsize = btrfs_super_leafsize(disk_super); | |
1753 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 1754 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
1755 | tree_root->nodesize = nodesize; |
1756 | tree_root->leafsize = leafsize; | |
1757 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 1758 | tree_root->stripesize = stripesize; |
a061fc8d CM |
1759 | |
1760 | sb->s_blocksize = sectorsize; | |
1761 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 1762 | |
39279cc3 CM |
1763 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
1764 | sizeof(disk_super->magic))) { | |
d397712b | 1765 | printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
1766 | goto fail_sb_buffer; |
1767 | } | |
19c00ddc | 1768 | |
925baedd | 1769 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 1770 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 1771 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 1772 | if (ret) { |
d397712b CM |
1773 | printk(KERN_WARNING "btrfs: failed to read the system " |
1774 | "array on %s\n", sb->s_id); | |
84eed90f CM |
1775 | goto fail_sys_array; |
1776 | } | |
0b86a832 CM |
1777 | |
1778 | blocksize = btrfs_level_size(tree_root, | |
1779 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 1780 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
1781 | |
1782 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1783 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
1784 | ||
1785 | chunk_root->node = read_tree_block(chunk_root, | |
1786 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 1787 | blocksize, generation); |
0b86a832 CM |
1788 | BUG_ON(!chunk_root->node); |
1789 | ||
e17cade2 | 1790 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
d397712b CM |
1791 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), |
1792 | BTRFS_UUID_SIZE); | |
e17cade2 | 1793 | |
925baedd | 1794 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1795 | ret = btrfs_read_chunk_tree(chunk_root); |
925baedd | 1796 | mutex_unlock(&fs_info->chunk_mutex); |
2b82032c | 1797 | if (ret) { |
d397712b CM |
1798 | printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", |
1799 | sb->s_id); | |
2b82032c YZ |
1800 | goto fail_chunk_root; |
1801 | } | |
0b86a832 | 1802 | |
dfe25020 CM |
1803 | btrfs_close_extra_devices(fs_devices); |
1804 | ||
db94535d CM |
1805 | blocksize = btrfs_level_size(tree_root, |
1806 | btrfs_super_root_level(disk_super)); | |
84234f3a | 1807 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 1808 | |
e20d96d6 | 1809 | tree_root->node = read_tree_block(tree_root, |
db94535d | 1810 | btrfs_super_root(disk_super), |
84234f3a | 1811 | blocksize, generation); |
39279cc3 | 1812 | if (!tree_root->node) |
2b82032c | 1813 | goto fail_chunk_root; |
3768f368 | 1814 | |
db94535d CM |
1815 | |
1816 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 1817 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 1818 | if (ret) |
39279cc3 | 1819 | goto fail_tree_root; |
0b86a832 CM |
1820 | extent_root->track_dirty = 1; |
1821 | ||
1822 | ret = find_and_setup_root(tree_root, fs_info, | |
1823 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
1824 | dev_root->track_dirty = 1; | |
0b86a832 CM |
1825 | if (ret) |
1826 | goto fail_extent_root; | |
3768f368 | 1827 | |
d20f7043 CM |
1828 | ret = find_and_setup_root(tree_root, fs_info, |
1829 | BTRFS_CSUM_TREE_OBJECTID, csum_root); | |
1830 | if (ret) | |
1831 | goto fail_extent_root; | |
1832 | ||
1833 | csum_root->track_dirty = 1; | |
1834 | ||
9078a3e1 CM |
1835 | btrfs_read_block_groups(extent_root); |
1836 | ||
2d69a0f8 | 1837 | fs_info->generation = generation; |
c146afad | 1838 | fs_info->last_trans_committed = generation; |
d18a2c44 CM |
1839 | fs_info->data_alloc_profile = (u64)-1; |
1840 | fs_info->metadata_alloc_profile = (u64)-1; | |
1841 | fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; | |
a74a4b97 CM |
1842 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
1843 | "btrfs-cleaner"); | |
57506d50 | 1844 | if (IS_ERR(fs_info->cleaner_kthread)) |
d20f7043 | 1845 | goto fail_csum_root; |
a74a4b97 CM |
1846 | |
1847 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
1848 | tree_root, | |
1849 | "btrfs-transaction"); | |
57506d50 | 1850 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 1851 | goto fail_cleaner; |
a74a4b97 | 1852 | |
e02119d5 | 1853 | if (btrfs_super_log_root(disk_super) != 0) { |
e02119d5 CM |
1854 | u64 bytenr = btrfs_super_log_root(disk_super); |
1855 | ||
7c2ca468 | 1856 | if (fs_devices->rw_devices == 0) { |
d397712b CM |
1857 | printk(KERN_WARNING "Btrfs log replay required " |
1858 | "on RO media\n"); | |
7c2ca468 CM |
1859 | err = -EIO; |
1860 | goto fail_trans_kthread; | |
1861 | } | |
e02119d5 CM |
1862 | blocksize = |
1863 | btrfs_level_size(tree_root, | |
1864 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 1865 | |
e02119d5 CM |
1866 | log_tree_root = kzalloc(sizeof(struct btrfs_root), |
1867 | GFP_NOFS); | |
1868 | ||
1869 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1870 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1871 | ||
1872 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
1873 | blocksize, |
1874 | generation + 1); | |
e02119d5 CM |
1875 | ret = btrfs_recover_log_trees(log_tree_root); |
1876 | BUG_ON(ret); | |
e556ce2c YZ |
1877 | |
1878 | if (sb->s_flags & MS_RDONLY) { | |
1879 | ret = btrfs_commit_super(tree_root); | |
1880 | BUG_ON(ret); | |
1881 | } | |
e02119d5 | 1882 | } |
1a40e23b | 1883 | |
7c2ca468 CM |
1884 | if (!(sb->s_flags & MS_RDONLY)) { |
1885 | ret = btrfs_cleanup_reloc_trees(tree_root); | |
1886 | BUG_ON(ret); | |
1887 | } | |
1a40e23b | 1888 | |
3de4586c CM |
1889 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
1890 | location.type = BTRFS_ROOT_ITEM_KEY; | |
1891 | location.offset = (u64)-1; | |
1892 | ||
3de4586c CM |
1893 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
1894 | if (!fs_info->fs_root) | |
7c2ca468 | 1895 | goto fail_trans_kthread; |
0f7d52f4 | 1896 | return tree_root; |
39279cc3 | 1897 | |
7c2ca468 CM |
1898 | fail_trans_kthread: |
1899 | kthread_stop(fs_info->transaction_kthread); | |
3f157a2f | 1900 | fail_cleaner: |
a74a4b97 | 1901 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
1902 | |
1903 | /* | |
1904 | * make sure we're done with the btree inode before we stop our | |
1905 | * kthreads | |
1906 | */ | |
1907 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
1908 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
1909 | ||
d20f7043 CM |
1910 | fail_csum_root: |
1911 | free_extent_buffer(csum_root->node); | |
0b86a832 CM |
1912 | fail_extent_root: |
1913 | free_extent_buffer(extent_root->node); | |
39279cc3 | 1914 | fail_tree_root: |
5f39d397 | 1915 | free_extent_buffer(tree_root->node); |
2b82032c YZ |
1916 | fail_chunk_root: |
1917 | free_extent_buffer(chunk_root->node); | |
84eed90f | 1918 | fail_sys_array: |
7c2ca468 | 1919 | free_extent_buffer(dev_root->node); |
39279cc3 | 1920 | fail_sb_buffer: |
247e743c | 1921 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 1922 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
1923 | btrfs_stop_workers(&fs_info->workers); |
1924 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 1925 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 1926 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 1927 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 1928 | btrfs_stop_workers(&fs_info->submit_workers); |
4543df7e | 1929 | fail_iput: |
7c2ca468 | 1930 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
4543df7e | 1931 | iput(fs_info->btree_inode); |
7e662854 | 1932 | |
dfe25020 | 1933 | btrfs_close_devices(fs_info->fs_devices); |
84eed90f | 1934 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
7e662854 | 1935 | bdi_destroy(&fs_info->bdi); |
84eed90f | 1936 | |
7e662854 | 1937 | fail: |
39279cc3 CM |
1938 | kfree(extent_root); |
1939 | kfree(tree_root); | |
1940 | kfree(fs_info); | |
83afeac4 JM |
1941 | kfree(chunk_root); |
1942 | kfree(dev_root); | |
d20f7043 | 1943 | kfree(csum_root); |
39279cc3 | 1944 | return ERR_PTR(err); |
eb60ceac CM |
1945 | } |
1946 | ||
f2984462 CM |
1947 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
1948 | { | |
1949 | char b[BDEVNAME_SIZE]; | |
1950 | ||
1951 | if (uptodate) { | |
1952 | set_buffer_uptodate(bh); | |
1953 | } else { | |
1954 | if (!buffer_eopnotsupp(bh) && printk_ratelimit()) { | |
1955 | printk(KERN_WARNING "lost page write due to " | |
1956 | "I/O error on %s\n", | |
1957 | bdevname(bh->b_bdev, b)); | |
1958 | } | |
1259ab75 CM |
1959 | /* note, we dont' set_buffer_write_io_error because we have |
1960 | * our own ways of dealing with the IO errors | |
1961 | */ | |
f2984462 CM |
1962 | clear_buffer_uptodate(bh); |
1963 | } | |
1964 | unlock_buffer(bh); | |
1965 | put_bh(bh); | |
1966 | } | |
1967 | ||
a512bbf8 YZ |
1968 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
1969 | { | |
1970 | struct buffer_head *bh; | |
1971 | struct buffer_head *latest = NULL; | |
1972 | struct btrfs_super_block *super; | |
1973 | int i; | |
1974 | u64 transid = 0; | |
1975 | u64 bytenr; | |
1976 | ||
1977 | /* we would like to check all the supers, but that would make | |
1978 | * a btrfs mount succeed after a mkfs from a different FS. | |
1979 | * So, we need to add a special mount option to scan for | |
1980 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
1981 | */ | |
1982 | for (i = 0; i < 1; i++) { | |
1983 | bytenr = btrfs_sb_offset(i); | |
1984 | if (bytenr + 4096 >= i_size_read(bdev->bd_inode)) | |
1985 | break; | |
1986 | bh = __bread(bdev, bytenr / 4096, 4096); | |
1987 | if (!bh) | |
1988 | continue; | |
1989 | ||
1990 | super = (struct btrfs_super_block *)bh->b_data; | |
1991 | if (btrfs_super_bytenr(super) != bytenr || | |
1992 | strncmp((char *)(&super->magic), BTRFS_MAGIC, | |
1993 | sizeof(super->magic))) { | |
1994 | brelse(bh); | |
1995 | continue; | |
1996 | } | |
1997 | ||
1998 | if (!latest || btrfs_super_generation(super) > transid) { | |
1999 | brelse(latest); | |
2000 | latest = bh; | |
2001 | transid = btrfs_super_generation(super); | |
2002 | } else { | |
2003 | brelse(bh); | |
2004 | } | |
2005 | } | |
2006 | return latest; | |
2007 | } | |
2008 | ||
2009 | static int write_dev_supers(struct btrfs_device *device, | |
2010 | struct btrfs_super_block *sb, | |
2011 | int do_barriers, int wait, int max_mirrors) | |
2012 | { | |
2013 | struct buffer_head *bh; | |
2014 | int i; | |
2015 | int ret; | |
2016 | int errors = 0; | |
2017 | u32 crc; | |
2018 | u64 bytenr; | |
2019 | int last_barrier = 0; | |
2020 | ||
2021 | if (max_mirrors == 0) | |
2022 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
2023 | ||
2024 | /* make sure only the last submit_bh does a barrier */ | |
2025 | if (do_barriers) { | |
2026 | for (i = 0; i < max_mirrors; i++) { | |
2027 | bytenr = btrfs_sb_offset(i); | |
2028 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= | |
2029 | device->total_bytes) | |
2030 | break; | |
2031 | last_barrier = i; | |
2032 | } | |
2033 | } | |
2034 | ||
2035 | for (i = 0; i < max_mirrors; i++) { | |
2036 | bytenr = btrfs_sb_offset(i); | |
2037 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) | |
2038 | break; | |
2039 | ||
2040 | if (wait) { | |
2041 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
2042 | BTRFS_SUPER_INFO_SIZE); | |
2043 | BUG_ON(!bh); | |
2044 | brelse(bh); | |
2045 | wait_on_buffer(bh); | |
2046 | if (buffer_uptodate(bh)) { | |
2047 | brelse(bh); | |
2048 | continue; | |
2049 | } | |
2050 | } else { | |
2051 | btrfs_set_super_bytenr(sb, bytenr); | |
2052 | ||
2053 | crc = ~(u32)0; | |
2054 | crc = btrfs_csum_data(NULL, (char *)sb + | |
2055 | BTRFS_CSUM_SIZE, crc, | |
2056 | BTRFS_SUPER_INFO_SIZE - | |
2057 | BTRFS_CSUM_SIZE); | |
2058 | btrfs_csum_final(crc, sb->csum); | |
2059 | ||
2060 | bh = __getblk(device->bdev, bytenr / 4096, | |
2061 | BTRFS_SUPER_INFO_SIZE); | |
2062 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); | |
2063 | ||
2064 | set_buffer_uptodate(bh); | |
2065 | get_bh(bh); | |
2066 | lock_buffer(bh); | |
2067 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
2068 | } | |
2069 | ||
2070 | if (i == last_barrier && do_barriers && device->barriers) { | |
2071 | ret = submit_bh(WRITE_BARRIER, bh); | |
2072 | if (ret == -EOPNOTSUPP) { | |
2073 | printk("btrfs: disabling barriers on dev %s\n", | |
2074 | device->name); | |
2075 | set_buffer_uptodate(bh); | |
2076 | device->barriers = 0; | |
2077 | get_bh(bh); | |
2078 | lock_buffer(bh); | |
2079 | ret = submit_bh(WRITE, bh); | |
2080 | } | |
2081 | } else { | |
2082 | ret = submit_bh(WRITE, bh); | |
2083 | } | |
2084 | ||
2085 | if (!ret && wait) { | |
2086 | wait_on_buffer(bh); | |
2087 | if (!buffer_uptodate(bh)) | |
2088 | errors++; | |
2089 | } else if (ret) { | |
2090 | errors++; | |
2091 | } | |
2092 | if (wait) | |
2093 | brelse(bh); | |
2094 | } | |
2095 | return errors < i ? 0 : -1; | |
2096 | } | |
2097 | ||
2098 | int write_all_supers(struct btrfs_root *root, int max_mirrors) | |
f2984462 | 2099 | { |
f2984462 CM |
2100 | struct list_head *head = &root->fs_info->fs_devices->devices; |
2101 | struct btrfs_device *dev; | |
a061fc8d | 2102 | struct btrfs_super_block *sb; |
f2984462 | 2103 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
2104 | int ret; |
2105 | int do_barriers; | |
a236aed1 CM |
2106 | int max_errors; |
2107 | int total_errors = 0; | |
a061fc8d | 2108 | u64 flags; |
f2984462 | 2109 | |
a236aed1 | 2110 | max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1; |
f2984462 CM |
2111 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
2112 | ||
a061fc8d CM |
2113 | sb = &root->fs_info->super_for_commit; |
2114 | dev_item = &sb->dev_item; | |
c6e30871 | 2115 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
2116 | if (!dev->bdev) { |
2117 | total_errors++; | |
2118 | continue; | |
2119 | } | |
2b82032c | 2120 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2121 | continue; |
2122 | ||
2b82032c | 2123 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
2124 | btrfs_set_stack_device_type(dev_item, dev->type); |
2125 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
2126 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
2127 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
2128 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
2129 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
2130 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
2131 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 2132 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 2133 | |
a061fc8d CM |
2134 | flags = btrfs_super_flags(sb); |
2135 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
2136 | ||
a512bbf8 | 2137 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
2138 | if (ret) |
2139 | total_errors++; | |
f2984462 | 2140 | } |
a236aed1 | 2141 | if (total_errors > max_errors) { |
d397712b CM |
2142 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2143 | total_errors); | |
a236aed1 CM |
2144 | BUG(); |
2145 | } | |
f2984462 | 2146 | |
a512bbf8 | 2147 | total_errors = 0; |
c6e30871 | 2148 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
2149 | if (!dev->bdev) |
2150 | continue; | |
2b82032c | 2151 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2152 | continue; |
2153 | ||
a512bbf8 YZ |
2154 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
2155 | if (ret) | |
2156 | total_errors++; | |
f2984462 | 2157 | } |
a236aed1 | 2158 | if (total_errors > max_errors) { |
d397712b CM |
2159 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2160 | total_errors); | |
a236aed1 CM |
2161 | BUG(); |
2162 | } | |
f2984462 CM |
2163 | return 0; |
2164 | } | |
2165 | ||
a512bbf8 YZ |
2166 | int write_ctree_super(struct btrfs_trans_handle *trans, |
2167 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 2168 | { |
e66f709b | 2169 | int ret; |
5f39d397 | 2170 | |
a512bbf8 | 2171 | ret = write_all_supers(root, max_mirrors); |
5f39d397 | 2172 | return ret; |
cfaa7295 CM |
2173 | } |
2174 | ||
5eda7b5e | 2175 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f CM |
2176 | { |
2177 | radix_tree_delete(&fs_info->fs_roots_radix, | |
2178 | (unsigned long)root->root_key.objectid); | |
3394e160 CM |
2179 | if (root->anon_super.s_dev) { |
2180 | down_write(&root->anon_super.s_umount); | |
2181 | kill_anon_super(&root->anon_super); | |
2182 | } | |
2619ba1f | 2183 | if (root->node) |
5f39d397 | 2184 | free_extent_buffer(root->node); |
2619ba1f | 2185 | if (root->commit_root) |
5f39d397 | 2186 | free_extent_buffer(root->commit_root); |
d397712b | 2187 | kfree(root->name); |
2619ba1f CM |
2188 | kfree(root); |
2189 | return 0; | |
2190 | } | |
2191 | ||
35b7e476 | 2192 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
2193 | { |
2194 | int ret; | |
2195 | struct btrfs_root *gang[8]; | |
2196 | int i; | |
2197 | ||
d397712b | 2198 | while (1) { |
0f7d52f4 CM |
2199 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
2200 | (void **)gang, 0, | |
2201 | ARRAY_SIZE(gang)); | |
2202 | if (!ret) | |
2203 | break; | |
2619ba1f | 2204 | for (i = 0; i < ret; i++) |
5eda7b5e | 2205 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
2206 | } |
2207 | return 0; | |
2208 | } | |
b4100d64 | 2209 | |
c146afad | 2210 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 2211 | { |
c146afad YZ |
2212 | u64 root_objectid = 0; |
2213 | struct btrfs_root *gang[8]; | |
2214 | int i; | |
3768f368 | 2215 | int ret; |
e089f05c | 2216 | |
c146afad YZ |
2217 | while (1) { |
2218 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2219 | (void **)gang, root_objectid, | |
2220 | ARRAY_SIZE(gang)); | |
2221 | if (!ret) | |
2222 | break; | |
2223 | for (i = 0; i < ret; i++) { | |
2224 | root_objectid = gang[i]->root_key.objectid; | |
2225 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
2226 | root_objectid, gang[i]); | |
2227 | BUG_ON(ret); | |
2228 | btrfs_orphan_cleanup(gang[i]); | |
2229 | } | |
2230 | root_objectid++; | |
2231 | } | |
2232 | return 0; | |
2233 | } | |
a2135011 | 2234 | |
c146afad YZ |
2235 | int btrfs_commit_super(struct btrfs_root *root) |
2236 | { | |
2237 | struct btrfs_trans_handle *trans; | |
2238 | int ret; | |
a74a4b97 | 2239 | |
c146afad | 2240 | mutex_lock(&root->fs_info->cleaner_mutex); |
a74a4b97 | 2241 | btrfs_clean_old_snapshots(root); |
c146afad | 2242 | mutex_unlock(&root->fs_info->cleaner_mutex); |
79154b1b | 2243 | trans = btrfs_start_transaction(root, 1); |
54aa1f4d | 2244 | ret = btrfs_commit_transaction(trans, root); |
c146afad YZ |
2245 | BUG_ON(ret); |
2246 | /* run commit again to drop the original snapshot */ | |
79154b1b CM |
2247 | trans = btrfs_start_transaction(root, 1); |
2248 | btrfs_commit_transaction(trans, root); | |
2249 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 2250 | BUG_ON(ret); |
d6bfde87 | 2251 | |
a512bbf8 | 2252 | ret = write_ctree_super(NULL, root, 0); |
c146afad YZ |
2253 | return ret; |
2254 | } | |
2255 | ||
2256 | int close_ctree(struct btrfs_root *root) | |
2257 | { | |
2258 | struct btrfs_fs_info *fs_info = root->fs_info; | |
2259 | int ret; | |
2260 | ||
2261 | fs_info->closing = 1; | |
2262 | smp_mb(); | |
2263 | ||
2264 | kthread_stop(root->fs_info->transaction_kthread); | |
2265 | kthread_stop(root->fs_info->cleaner_kthread); | |
2266 | ||
2267 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { | |
2268 | ret = btrfs_commit_super(root); | |
d397712b CM |
2269 | if (ret) |
2270 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
c146afad | 2271 | } |
0f7d52f4 | 2272 | |
b0c68f8b | 2273 | if (fs_info->delalloc_bytes) { |
d397712b | 2274 | printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n", |
b0c68f8b CM |
2275 | fs_info->delalloc_bytes); |
2276 | } | |
31153d81 | 2277 | if (fs_info->total_ref_cache_size) { |
d397712b CM |
2278 | printk(KERN_INFO "btrfs: at umount reference cache size %llu\n", |
2279 | (unsigned long long)fs_info->total_ref_cache_size); | |
31153d81 | 2280 | } |
bcc63abb | 2281 | |
0f7d52f4 | 2282 | if (fs_info->extent_root->node) |
5f39d397 | 2283 | free_extent_buffer(fs_info->extent_root->node); |
f510cfec | 2284 | |
0f7d52f4 | 2285 | if (fs_info->tree_root->node) |
5f39d397 | 2286 | free_extent_buffer(fs_info->tree_root->node); |
f510cfec | 2287 | |
d397712b | 2288 | if (root->fs_info->chunk_root->node) |
0b86a832 CM |
2289 | free_extent_buffer(root->fs_info->chunk_root->node); |
2290 | ||
d397712b | 2291 | if (root->fs_info->dev_root->node) |
0b86a832 CM |
2292 | free_extent_buffer(root->fs_info->dev_root->node); |
2293 | ||
d397712b | 2294 | if (root->fs_info->csum_root->node) |
d20f7043 CM |
2295 | free_extent_buffer(root->fs_info->csum_root->node); |
2296 | ||
9078a3e1 | 2297 | btrfs_free_block_groups(root->fs_info); |
d10c5f31 | 2298 | |
c146afad | 2299 | del_fs_roots(fs_info); |
d10c5f31 | 2300 | |
c146afad | 2301 | iput(fs_info->btree_inode); |
9ad6b7bc | 2302 | |
247e743c | 2303 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2304 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2305 | btrfs_stop_workers(&fs_info->workers); |
2306 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 2307 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 2308 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 2309 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 2310 | btrfs_stop_workers(&fs_info->submit_workers); |
d6bfde87 | 2311 | |
19c00ddc | 2312 | #if 0 |
d397712b | 2313 | while (!list_empty(&fs_info->hashers)) { |
19c00ddc CM |
2314 | struct btrfs_hasher *hasher; |
2315 | hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher, | |
2316 | hashers); | |
2317 | list_del(&hasher->hashers); | |
2318 | crypto_free_hash(&fs_info->hash_tfm); | |
2319 | kfree(hasher); | |
2320 | } | |
2321 | #endif | |
dfe25020 | 2322 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 2323 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 2324 | |
04160088 | 2325 | bdi_destroy(&fs_info->bdi); |
0b86a832 | 2326 | |
0f7d52f4 | 2327 | kfree(fs_info->extent_root); |
0f7d52f4 | 2328 | kfree(fs_info->tree_root); |
0b86a832 CM |
2329 | kfree(fs_info->chunk_root); |
2330 | kfree(fs_info->dev_root); | |
d20f7043 | 2331 | kfree(fs_info->csum_root); |
eb60ceac CM |
2332 | return 0; |
2333 | } | |
2334 | ||
1259ab75 | 2335 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 2336 | { |
1259ab75 | 2337 | int ret; |
810191ff | 2338 | struct inode *btree_inode = buf->first_page->mapping->host; |
1259ab75 CM |
2339 | |
2340 | ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); | |
2341 | if (!ret) | |
2342 | return ret; | |
2343 | ||
2344 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
2345 | parent_transid); | |
2346 | return !ret; | |
5f39d397 CM |
2347 | } |
2348 | ||
2349 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 2350 | { |
810191ff | 2351 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 2352 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
2353 | buf); |
2354 | } | |
6702ed49 | 2355 | |
5f39d397 CM |
2356 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
2357 | { | |
810191ff | 2358 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
5f39d397 CM |
2359 | u64 transid = btrfs_header_generation(buf); |
2360 | struct inode *btree_inode = root->fs_info->btree_inode; | |
6702ed49 | 2361 | |
b4ce94de CM |
2362 | btrfs_set_lock_blocking(buf); |
2363 | ||
925baedd | 2364 | WARN_ON(!btrfs_tree_locked(buf)); |
ccd467d6 | 2365 | if (transid != root->fs_info->generation) { |
d397712b CM |
2366 | printk(KERN_CRIT "btrfs transid mismatch buffer %llu, " |
2367 | "found %llu running %llu\n", | |
db94535d | 2368 | (unsigned long long)buf->start, |
d397712b CM |
2369 | (unsigned long long)transid, |
2370 | (unsigned long long)root->fs_info->generation); | |
ccd467d6 CM |
2371 | WARN_ON(1); |
2372 | } | |
d1310b2e | 2373 | set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf); |
eb60ceac CM |
2374 | } |
2375 | ||
d3c2fdcf | 2376 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
35b7e476 | 2377 | { |
188de649 CM |
2378 | /* |
2379 | * looks as though older kernels can get into trouble with | |
2380 | * this code, they end up stuck in balance_dirty_pages forever | |
2381 | */ | |
d6bfde87 CM |
2382 | struct extent_io_tree *tree; |
2383 | u64 num_dirty; | |
2384 | u64 start = 0; | |
771ed689 | 2385 | unsigned long thresh = 32 * 1024 * 1024; |
d6bfde87 CM |
2386 | tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
2387 | ||
b64a2851 | 2388 | if (current_is_pdflush() || current->flags & PF_MEMALLOC) |
d6bfde87 CM |
2389 | return; |
2390 | ||
2391 | num_dirty = count_range_bits(tree, &start, (u64)-1, | |
2392 | thresh, EXTENT_DIRTY); | |
2393 | if (num_dirty > thresh) { | |
2394 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 2395 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 2396 | } |
188de649 | 2397 | return; |
35b7e476 | 2398 | } |
6b80053d | 2399 | |
ca7a79ad | 2400 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 2401 | { |
810191ff | 2402 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
ce9adaa5 | 2403 | int ret; |
ca7a79ad | 2404 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
d397712b | 2405 | if (ret == 0) |
b4ce94de | 2406 | set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags); |
ce9adaa5 | 2407 | return ret; |
6b80053d | 2408 | } |
0da5468f | 2409 | |
4bef0848 CM |
2410 | int btree_lock_page_hook(struct page *page) |
2411 | { | |
2412 | struct inode *inode = page->mapping->host; | |
4bef0848 CM |
2413 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
2414 | struct extent_buffer *eb; | |
2415 | unsigned long len; | |
2416 | u64 bytenr = page_offset(page); | |
2417 | ||
2418 | if (page->private == EXTENT_PAGE_PRIVATE) | |
2419 | goto out; | |
2420 | ||
2421 | len = page->private >> 2; | |
2422 | eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS); | |
2423 | if (!eb) | |
2424 | goto out; | |
2425 | ||
2426 | btrfs_tree_lock(eb); | |
4bef0848 | 2427 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); |
4bef0848 CM |
2428 | btrfs_tree_unlock(eb); |
2429 | free_extent_buffer(eb); | |
2430 | out: | |
2431 | lock_page(page); | |
2432 | return 0; | |
2433 | } | |
2434 | ||
d1310b2e | 2435 | static struct extent_io_ops btree_extent_io_ops = { |
4bef0848 | 2436 | .write_cache_pages_lock_hook = btree_lock_page_hook, |
ce9adaa5 | 2437 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
0b86a832 | 2438 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
2439 | /* note we're sharing with inode.c for the merge bio hook */ |
2440 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 2441 | }; |