]>
Commit | Line | Data |
---|---|---|
6cbd5570 CM |
1 | /* |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
e20d96d6 | 19 | #include <linux/fs.h> |
d98237b3 | 20 | #include <linux/blkdev.h> |
87cbda5c | 21 | #include <linux/scatterlist.h> |
22b0ebda | 22 | #include <linux/swap.h> |
0f7d52f4 | 23 | #include <linux/radix-tree.h> |
35b7e476 | 24 | #include <linux/writeback.h> |
d397712b | 25 | #include <linux/buffer_head.h> |
ce9adaa5 | 26 | #include <linux/workqueue.h> |
a74a4b97 | 27 | #include <linux/kthread.h> |
4b4e25f2 | 28 | #include <linux/freezer.h> |
163e783e | 29 | #include <linux/crc32c.h> |
5a0e3ad6 | 30 | #include <linux/slab.h> |
4b4e25f2 | 31 | #include "compat.h" |
eb60ceac CM |
32 | #include "ctree.h" |
33 | #include "disk-io.h" | |
e089f05c | 34 | #include "transaction.h" |
0f7d52f4 | 35 | #include "btrfs_inode.h" |
0b86a832 | 36 | #include "volumes.h" |
db94535d | 37 | #include "print-tree.h" |
8b712842 | 38 | #include "async-thread.h" |
925baedd | 39 | #include "locking.h" |
e02119d5 | 40 | #include "tree-log.h" |
fa9c0d79 | 41 | #include "free-space-cache.h" |
eb60ceac | 42 | |
d1310b2e | 43 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 44 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 45 | static void free_fs_root(struct btrfs_root *root); |
ce9adaa5 | 46 | |
d352ac68 CM |
47 | /* |
48 | * end_io_wq structs are used to do processing in task context when an IO is | |
49 | * complete. This is used during reads to verify checksums, and it is used | |
50 | * by writes to insert metadata for new file extents after IO is complete. | |
51 | */ | |
ce9adaa5 CM |
52 | struct end_io_wq { |
53 | struct bio *bio; | |
54 | bio_end_io_t *end_io; | |
55 | void *private; | |
56 | struct btrfs_fs_info *info; | |
57 | int error; | |
22c59948 | 58 | int metadata; |
ce9adaa5 | 59 | struct list_head list; |
8b712842 | 60 | struct btrfs_work work; |
ce9adaa5 | 61 | }; |
0da5468f | 62 | |
d352ac68 CM |
63 | /* |
64 | * async submit bios are used to offload expensive checksumming | |
65 | * onto the worker threads. They checksum file and metadata bios | |
66 | * just before they are sent down the IO stack. | |
67 | */ | |
44b8bd7e CM |
68 | struct async_submit_bio { |
69 | struct inode *inode; | |
70 | struct bio *bio; | |
71 | struct list_head list; | |
4a69a410 CM |
72 | extent_submit_bio_hook_t *submit_bio_start; |
73 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
74 | int rw; |
75 | int mirror_num; | |
c8b97818 | 76 | unsigned long bio_flags; |
eaf25d93 CM |
77 | /* |
78 | * bio_offset is optional, can be used if the pages in the bio | |
79 | * can't tell us where in the file the bio should go | |
80 | */ | |
81 | u64 bio_offset; | |
8b712842 | 82 | struct btrfs_work work; |
44b8bd7e CM |
83 | }; |
84 | ||
4008c04a CM |
85 | /* These are used to set the lockdep class on the extent buffer locks. |
86 | * The class is set by the readpage_end_io_hook after the buffer has | |
87 | * passed csum validation but before the pages are unlocked. | |
88 | * | |
89 | * The lockdep class is also set by btrfs_init_new_buffer on freshly | |
90 | * allocated blocks. | |
91 | * | |
92 | * The class is based on the level in the tree block, which allows lockdep | |
93 | * to know that lower nodes nest inside the locks of higher nodes. | |
94 | * | |
95 | * We also add a check to make sure the highest level of the tree is | |
96 | * the same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this | |
97 | * code needs update as well. | |
98 | */ | |
99 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
100 | # if BTRFS_MAX_LEVEL != 8 | |
101 | # error | |
102 | # endif | |
103 | static struct lock_class_key btrfs_eb_class[BTRFS_MAX_LEVEL + 1]; | |
104 | static const char *btrfs_eb_name[BTRFS_MAX_LEVEL + 1] = { | |
105 | /* leaf */ | |
106 | "btrfs-extent-00", | |
107 | "btrfs-extent-01", | |
108 | "btrfs-extent-02", | |
109 | "btrfs-extent-03", | |
110 | "btrfs-extent-04", | |
111 | "btrfs-extent-05", | |
112 | "btrfs-extent-06", | |
113 | "btrfs-extent-07", | |
114 | /* highest possible level */ | |
115 | "btrfs-extent-08", | |
116 | }; | |
117 | #endif | |
118 | ||
d352ac68 CM |
119 | /* |
120 | * extents on the btree inode are pretty simple, there's one extent | |
121 | * that covers the entire device | |
122 | */ | |
b2950863 CH |
123 | static struct extent_map *btree_get_extent(struct inode *inode, |
124 | struct page *page, size_t page_offset, u64 start, u64 len, | |
125 | int create) | |
7eccb903 | 126 | { |
5f39d397 CM |
127 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
128 | struct extent_map *em; | |
129 | int ret; | |
130 | ||
890871be | 131 | read_lock(&em_tree->lock); |
d1310b2e | 132 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d CM |
133 | if (em) { |
134 | em->bdev = | |
135 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
890871be | 136 | read_unlock(&em_tree->lock); |
5f39d397 | 137 | goto out; |
a061fc8d | 138 | } |
890871be | 139 | read_unlock(&em_tree->lock); |
7b13b7b1 | 140 | |
5f39d397 CM |
141 | em = alloc_extent_map(GFP_NOFS); |
142 | if (!em) { | |
143 | em = ERR_PTR(-ENOMEM); | |
144 | goto out; | |
145 | } | |
146 | em->start = 0; | |
0afbaf8c | 147 | em->len = (u64)-1; |
c8b97818 | 148 | em->block_len = (u64)-1; |
5f39d397 | 149 | em->block_start = 0; |
a061fc8d | 150 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e | 151 | |
890871be | 152 | write_lock(&em_tree->lock); |
5f39d397 CM |
153 | ret = add_extent_mapping(em_tree, em); |
154 | if (ret == -EEXIST) { | |
0afbaf8c CM |
155 | u64 failed_start = em->start; |
156 | u64 failed_len = em->len; | |
157 | ||
5f39d397 | 158 | free_extent_map(em); |
7b13b7b1 | 159 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c | 160 | if (em) { |
7b13b7b1 | 161 | ret = 0; |
0afbaf8c CM |
162 | } else { |
163 | em = lookup_extent_mapping(em_tree, failed_start, | |
164 | failed_len); | |
7b13b7b1 | 165 | ret = -EIO; |
0afbaf8c | 166 | } |
5f39d397 | 167 | } else if (ret) { |
7b13b7b1 CM |
168 | free_extent_map(em); |
169 | em = NULL; | |
5f39d397 | 170 | } |
890871be | 171 | write_unlock(&em_tree->lock); |
7b13b7b1 CM |
172 | |
173 | if (ret) | |
174 | em = ERR_PTR(ret); | |
5f39d397 CM |
175 | out: |
176 | return em; | |
7eccb903 CM |
177 | } |
178 | ||
19c00ddc CM |
179 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
180 | { | |
163e783e | 181 | return crc32c(seed, data, len); |
19c00ddc CM |
182 | } |
183 | ||
184 | void btrfs_csum_final(u32 crc, char *result) | |
185 | { | |
186 | *(__le32 *)result = ~cpu_to_le32(crc); | |
187 | } | |
188 | ||
d352ac68 CM |
189 | /* |
190 | * compute the csum for a btree block, and either verify it or write it | |
191 | * into the csum field of the block. | |
192 | */ | |
19c00ddc CM |
193 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
194 | int verify) | |
195 | { | |
607d432d JB |
196 | u16 csum_size = |
197 | btrfs_super_csum_size(&root->fs_info->super_copy); | |
198 | char *result = NULL; | |
19c00ddc CM |
199 | unsigned long len; |
200 | unsigned long cur_len; | |
201 | unsigned long offset = BTRFS_CSUM_SIZE; | |
202 | char *map_token = NULL; | |
203 | char *kaddr; | |
204 | unsigned long map_start; | |
205 | unsigned long map_len; | |
206 | int err; | |
207 | u32 crc = ~(u32)0; | |
607d432d | 208 | unsigned long inline_result; |
19c00ddc CM |
209 | |
210 | len = buf->len - offset; | |
d397712b | 211 | while (len > 0) { |
19c00ddc CM |
212 | err = map_private_extent_buffer(buf, offset, 32, |
213 | &map_token, &kaddr, | |
214 | &map_start, &map_len, KM_USER0); | |
d397712b | 215 | if (err) |
19c00ddc | 216 | return 1; |
19c00ddc CM |
217 | cur_len = min(len, map_len - (offset - map_start)); |
218 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
219 | crc, cur_len); | |
220 | len -= cur_len; | |
221 | offset += cur_len; | |
222 | unmap_extent_buffer(buf, map_token, KM_USER0); | |
223 | } | |
607d432d JB |
224 | if (csum_size > sizeof(inline_result)) { |
225 | result = kzalloc(csum_size * sizeof(char), GFP_NOFS); | |
226 | if (!result) | |
227 | return 1; | |
228 | } else { | |
229 | result = (char *)&inline_result; | |
230 | } | |
231 | ||
19c00ddc CM |
232 | btrfs_csum_final(crc, result); |
233 | ||
234 | if (verify) { | |
607d432d | 235 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
236 | u32 val; |
237 | u32 found = 0; | |
607d432d | 238 | memcpy(&found, result, csum_size); |
e4204ded | 239 | |
607d432d | 240 | read_extent_buffer(buf, &val, 0, csum_size); |
193f284d CM |
241 | if (printk_ratelimit()) { |
242 | printk(KERN_INFO "btrfs: %s checksum verify " | |
243 | "failed on %llu wanted %X found %X " | |
244 | "level %d\n", | |
245 | root->fs_info->sb->s_id, | |
246 | (unsigned long long)buf->start, val, found, | |
247 | btrfs_header_level(buf)); | |
248 | } | |
607d432d JB |
249 | if (result != (char *)&inline_result) |
250 | kfree(result); | |
19c00ddc CM |
251 | return 1; |
252 | } | |
253 | } else { | |
607d432d | 254 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 255 | } |
607d432d JB |
256 | if (result != (char *)&inline_result) |
257 | kfree(result); | |
19c00ddc CM |
258 | return 0; |
259 | } | |
260 | ||
d352ac68 CM |
261 | /* |
262 | * we can't consider a given block up to date unless the transid of the | |
263 | * block matches the transid in the parent node's pointer. This is how we | |
264 | * detect blocks that either didn't get written at all or got written | |
265 | * in the wrong place. | |
266 | */ | |
1259ab75 CM |
267 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
268 | struct extent_buffer *eb, u64 parent_transid) | |
269 | { | |
2ac55d41 | 270 | struct extent_state *cached_state = NULL; |
1259ab75 CM |
271 | int ret; |
272 | ||
273 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
274 | return 0; | |
275 | ||
2ac55d41 JB |
276 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
277 | 0, &cached_state, GFP_NOFS); | |
278 | if (extent_buffer_uptodate(io_tree, eb, cached_state) && | |
1259ab75 CM |
279 | btrfs_header_generation(eb) == parent_transid) { |
280 | ret = 0; | |
281 | goto out; | |
282 | } | |
193f284d CM |
283 | if (printk_ratelimit()) { |
284 | printk("parent transid verify failed on %llu wanted %llu " | |
285 | "found %llu\n", | |
286 | (unsigned long long)eb->start, | |
287 | (unsigned long long)parent_transid, | |
288 | (unsigned long long)btrfs_header_generation(eb)); | |
289 | } | |
1259ab75 | 290 | ret = 1; |
2ac55d41 | 291 | clear_extent_buffer_uptodate(io_tree, eb, &cached_state); |
33958dc6 | 292 | out: |
2ac55d41 JB |
293 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
294 | &cached_state, GFP_NOFS); | |
1259ab75 | 295 | return ret; |
1259ab75 CM |
296 | } |
297 | ||
d352ac68 CM |
298 | /* |
299 | * helper to read a given tree block, doing retries as required when | |
300 | * the checksums don't match and we have alternate mirrors to try. | |
301 | */ | |
f188591e CM |
302 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
303 | struct extent_buffer *eb, | |
ca7a79ad | 304 | u64 start, u64 parent_transid) |
f188591e CM |
305 | { |
306 | struct extent_io_tree *io_tree; | |
307 | int ret; | |
308 | int num_copies = 0; | |
309 | int mirror_num = 0; | |
310 | ||
311 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
312 | while (1) { | |
313 | ret = read_extent_buffer_pages(io_tree, eb, start, 1, | |
314 | btree_get_extent, mirror_num); | |
1259ab75 CM |
315 | if (!ret && |
316 | !verify_parent_transid(io_tree, eb, parent_transid)) | |
f188591e | 317 | return ret; |
d397712b | 318 | |
f188591e CM |
319 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
320 | eb->start, eb->len); | |
4235298e | 321 | if (num_copies == 1) |
f188591e | 322 | return ret; |
4235298e | 323 | |
f188591e | 324 | mirror_num++; |
4235298e | 325 | if (mirror_num > num_copies) |
f188591e | 326 | return ret; |
f188591e | 327 | } |
f188591e CM |
328 | return -EIO; |
329 | } | |
19c00ddc | 330 | |
d352ac68 | 331 | /* |
d397712b CM |
332 | * checksum a dirty tree block before IO. This has extra checks to make sure |
333 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 334 | */ |
d397712b | 335 | |
b2950863 | 336 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 337 | { |
d1310b2e | 338 | struct extent_io_tree *tree; |
35ebb934 | 339 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc | 340 | u64 found_start; |
19c00ddc CM |
341 | unsigned long len; |
342 | struct extent_buffer *eb; | |
f188591e CM |
343 | int ret; |
344 | ||
d1310b2e | 345 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc CM |
346 | |
347 | if (page->private == EXTENT_PAGE_PRIVATE) | |
348 | goto out; | |
349 | if (!page->private) | |
350 | goto out; | |
351 | len = page->private >> 2; | |
d397712b CM |
352 | WARN_ON(len == 0); |
353 | ||
19c00ddc | 354 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); |
ca7a79ad CM |
355 | ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, |
356 | btrfs_header_generation(eb)); | |
f188591e | 357 | BUG_ON(ret); |
19c00ddc CM |
358 | found_start = btrfs_header_bytenr(eb); |
359 | if (found_start != start) { | |
55c69072 CM |
360 | WARN_ON(1); |
361 | goto err; | |
362 | } | |
363 | if (eb->first_page != page) { | |
55c69072 CM |
364 | WARN_ON(1); |
365 | goto err; | |
366 | } | |
367 | if (!PageUptodate(page)) { | |
55c69072 CM |
368 | WARN_ON(1); |
369 | goto err; | |
19c00ddc | 370 | } |
19c00ddc | 371 | csum_tree_block(root, eb, 0); |
55c69072 | 372 | err: |
19c00ddc CM |
373 | free_extent_buffer(eb); |
374 | out: | |
375 | return 0; | |
376 | } | |
377 | ||
2b82032c YZ |
378 | static int check_tree_block_fsid(struct btrfs_root *root, |
379 | struct extent_buffer *eb) | |
380 | { | |
381 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
382 | u8 fsid[BTRFS_UUID_SIZE]; | |
383 | int ret = 1; | |
384 | ||
385 | read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), | |
386 | BTRFS_FSID_SIZE); | |
387 | while (fs_devices) { | |
388 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
389 | ret = 0; | |
390 | break; | |
391 | } | |
392 | fs_devices = fs_devices->seed; | |
393 | } | |
394 | return ret; | |
395 | } | |
396 | ||
4008c04a CM |
397 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
398 | void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, int level) | |
399 | { | |
400 | lockdep_set_class_and_name(&eb->lock, | |
401 | &btrfs_eb_class[level], | |
402 | btrfs_eb_name[level]); | |
403 | } | |
404 | #endif | |
405 | ||
b2950863 | 406 | static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
ce9adaa5 CM |
407 | struct extent_state *state) |
408 | { | |
409 | struct extent_io_tree *tree; | |
410 | u64 found_start; | |
411 | int found_level; | |
412 | unsigned long len; | |
413 | struct extent_buffer *eb; | |
414 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 415 | int ret = 0; |
ce9adaa5 CM |
416 | |
417 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
418 | if (page->private == EXTENT_PAGE_PRIVATE) | |
419 | goto out; | |
420 | if (!page->private) | |
421 | goto out; | |
d397712b | 422 | |
ce9adaa5 | 423 | len = page->private >> 2; |
d397712b CM |
424 | WARN_ON(len == 0); |
425 | ||
ce9adaa5 | 426 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); |
f188591e | 427 | |
ce9adaa5 | 428 | found_start = btrfs_header_bytenr(eb); |
23a07867 | 429 | if (found_start != start) { |
193f284d CM |
430 | if (printk_ratelimit()) { |
431 | printk(KERN_INFO "btrfs bad tree block start " | |
432 | "%llu %llu\n", | |
433 | (unsigned long long)found_start, | |
434 | (unsigned long long)eb->start); | |
435 | } | |
f188591e | 436 | ret = -EIO; |
ce9adaa5 CM |
437 | goto err; |
438 | } | |
439 | if (eb->first_page != page) { | |
d397712b CM |
440 | printk(KERN_INFO "btrfs bad first page %lu %lu\n", |
441 | eb->first_page->index, page->index); | |
ce9adaa5 | 442 | WARN_ON(1); |
f188591e | 443 | ret = -EIO; |
ce9adaa5 CM |
444 | goto err; |
445 | } | |
2b82032c | 446 | if (check_tree_block_fsid(root, eb)) { |
193f284d CM |
447 | if (printk_ratelimit()) { |
448 | printk(KERN_INFO "btrfs bad fsid on block %llu\n", | |
449 | (unsigned long long)eb->start); | |
450 | } | |
1259ab75 CM |
451 | ret = -EIO; |
452 | goto err; | |
453 | } | |
ce9adaa5 CM |
454 | found_level = btrfs_header_level(eb); |
455 | ||
4008c04a CM |
456 | btrfs_set_buffer_lockdep_class(eb, found_level); |
457 | ||
ce9adaa5 | 458 | ret = csum_tree_block(root, eb, 1); |
f188591e CM |
459 | if (ret) |
460 | ret = -EIO; | |
ce9adaa5 CM |
461 | |
462 | end = min_t(u64, eb->len, PAGE_CACHE_SIZE); | |
463 | end = eb->start + end - 1; | |
ce9adaa5 CM |
464 | err: |
465 | free_extent_buffer(eb); | |
466 | out: | |
f188591e | 467 | return ret; |
ce9adaa5 CM |
468 | } |
469 | ||
ce9adaa5 | 470 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
471 | { |
472 | struct end_io_wq *end_io_wq = bio->bi_private; | |
473 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 474 | |
ce9adaa5 | 475 | fs_info = end_io_wq->info; |
ce9adaa5 | 476 | end_io_wq->error = err; |
8b712842 CM |
477 | end_io_wq->work.func = end_workqueue_fn; |
478 | end_io_wq->work.flags = 0; | |
d20f7043 | 479 | |
7b6d91da | 480 | if (bio->bi_rw & REQ_WRITE) { |
0cb59c99 | 481 | if (end_io_wq->metadata == 1) |
cad321ad CM |
482 | btrfs_queue_worker(&fs_info->endio_meta_write_workers, |
483 | &end_io_wq->work); | |
0cb59c99 JB |
484 | else if (end_io_wq->metadata == 2) |
485 | btrfs_queue_worker(&fs_info->endio_freespace_worker, | |
486 | &end_io_wq->work); | |
cad321ad CM |
487 | else |
488 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
489 | &end_io_wq->work); | |
d20f7043 CM |
490 | } else { |
491 | if (end_io_wq->metadata) | |
492 | btrfs_queue_worker(&fs_info->endio_meta_workers, | |
493 | &end_io_wq->work); | |
494 | else | |
495 | btrfs_queue_worker(&fs_info->endio_workers, | |
496 | &end_io_wq->work); | |
497 | } | |
ce9adaa5 CM |
498 | } |
499 | ||
0cb59c99 JB |
500 | /* |
501 | * For the metadata arg you want | |
502 | * | |
503 | * 0 - if data | |
504 | * 1 - if normal metadta | |
505 | * 2 - if writing to the free space cache area | |
506 | */ | |
22c59948 CM |
507 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
508 | int metadata) | |
0b86a832 | 509 | { |
ce9adaa5 | 510 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
511 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
512 | if (!end_io_wq) | |
513 | return -ENOMEM; | |
514 | ||
515 | end_io_wq->private = bio->bi_private; | |
516 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 517 | end_io_wq->info = info; |
ce9adaa5 CM |
518 | end_io_wq->error = 0; |
519 | end_io_wq->bio = bio; | |
22c59948 | 520 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
521 | |
522 | bio->bi_private = end_io_wq; | |
523 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
524 | return 0; |
525 | } | |
526 | ||
b64a2851 | 527 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 528 | { |
4854ddd0 CM |
529 | unsigned long limit = min_t(unsigned long, |
530 | info->workers.max_workers, | |
531 | info->fs_devices->open_devices); | |
532 | return 256 * limit; | |
533 | } | |
0986fe9e | 534 | |
4854ddd0 CM |
535 | int btrfs_congested_async(struct btrfs_fs_info *info, int iodone) |
536 | { | |
b64a2851 CM |
537 | return atomic_read(&info->nr_async_bios) > |
538 | btrfs_async_submit_limit(info); | |
0986fe9e CM |
539 | } |
540 | ||
4a69a410 CM |
541 | static void run_one_async_start(struct btrfs_work *work) |
542 | { | |
4a69a410 CM |
543 | struct async_submit_bio *async; |
544 | ||
545 | async = container_of(work, struct async_submit_bio, work); | |
4a69a410 | 546 | async->submit_bio_start(async->inode, async->rw, async->bio, |
eaf25d93 CM |
547 | async->mirror_num, async->bio_flags, |
548 | async->bio_offset); | |
4a69a410 CM |
549 | } |
550 | ||
551 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
552 | { |
553 | struct btrfs_fs_info *fs_info; | |
554 | struct async_submit_bio *async; | |
4854ddd0 | 555 | int limit; |
8b712842 CM |
556 | |
557 | async = container_of(work, struct async_submit_bio, work); | |
558 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 559 | |
b64a2851 | 560 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
561 | limit = limit * 2 / 3; |
562 | ||
8b712842 | 563 | atomic_dec(&fs_info->nr_async_submits); |
0986fe9e | 564 | |
b64a2851 CM |
565 | if (atomic_read(&fs_info->nr_async_submits) < limit && |
566 | waitqueue_active(&fs_info->async_submit_wait)) | |
4854ddd0 CM |
567 | wake_up(&fs_info->async_submit_wait); |
568 | ||
4a69a410 | 569 | async->submit_bio_done(async->inode, async->rw, async->bio, |
eaf25d93 CM |
570 | async->mirror_num, async->bio_flags, |
571 | async->bio_offset); | |
4a69a410 CM |
572 | } |
573 | ||
574 | static void run_one_async_free(struct btrfs_work *work) | |
575 | { | |
576 | struct async_submit_bio *async; | |
577 | ||
578 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
579 | kfree(async); |
580 | } | |
581 | ||
44b8bd7e CM |
582 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
583 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 584 | unsigned long bio_flags, |
eaf25d93 | 585 | u64 bio_offset, |
4a69a410 CM |
586 | extent_submit_bio_hook_t *submit_bio_start, |
587 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
588 | { |
589 | struct async_submit_bio *async; | |
590 | ||
591 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
592 | if (!async) | |
593 | return -ENOMEM; | |
594 | ||
595 | async->inode = inode; | |
596 | async->rw = rw; | |
597 | async->bio = bio; | |
598 | async->mirror_num = mirror_num; | |
4a69a410 CM |
599 | async->submit_bio_start = submit_bio_start; |
600 | async->submit_bio_done = submit_bio_done; | |
601 | ||
602 | async->work.func = run_one_async_start; | |
603 | async->work.ordered_func = run_one_async_done; | |
604 | async->work.ordered_free = run_one_async_free; | |
605 | ||
8b712842 | 606 | async->work.flags = 0; |
c8b97818 | 607 | async->bio_flags = bio_flags; |
eaf25d93 | 608 | async->bio_offset = bio_offset; |
8c8bee1d | 609 | |
cb03c743 | 610 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 | 611 | |
7b6d91da | 612 | if (rw & REQ_SYNC) |
d313d7a3 CM |
613 | btrfs_set_work_high_prio(&async->work); |
614 | ||
8b712842 | 615 | btrfs_queue_worker(&fs_info->workers, &async->work); |
9473f16c | 616 | |
d397712b | 617 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
618 | atomic_read(&fs_info->nr_async_submits)) { |
619 | wait_event(fs_info->async_submit_wait, | |
620 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
621 | } | |
622 | ||
44b8bd7e CM |
623 | return 0; |
624 | } | |
625 | ||
ce3ed71a CM |
626 | static int btree_csum_one_bio(struct bio *bio) |
627 | { | |
628 | struct bio_vec *bvec = bio->bi_io_vec; | |
629 | int bio_index = 0; | |
630 | struct btrfs_root *root; | |
631 | ||
632 | WARN_ON(bio->bi_vcnt <= 0); | |
d397712b | 633 | while (bio_index < bio->bi_vcnt) { |
ce3ed71a CM |
634 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
635 | csum_dirty_buffer(root, bvec->bv_page); | |
636 | bio_index++; | |
637 | bvec++; | |
638 | } | |
639 | return 0; | |
640 | } | |
641 | ||
4a69a410 CM |
642 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
643 | struct bio *bio, int mirror_num, | |
eaf25d93 CM |
644 | unsigned long bio_flags, |
645 | u64 bio_offset) | |
22c59948 | 646 | { |
8b712842 CM |
647 | /* |
648 | * when we're called for a write, we're already in the async | |
5443be45 | 649 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 650 | */ |
4a69a410 CM |
651 | btree_csum_one_bio(bio); |
652 | return 0; | |
653 | } | |
22c59948 | 654 | |
4a69a410 | 655 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
656 | int mirror_num, unsigned long bio_flags, |
657 | u64 bio_offset) | |
4a69a410 | 658 | { |
8b712842 | 659 | /* |
4a69a410 CM |
660 | * when we're called for a write, we're already in the async |
661 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 662 | */ |
8b712842 | 663 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
664 | } |
665 | ||
44b8bd7e | 666 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
667 | int mirror_num, unsigned long bio_flags, |
668 | u64 bio_offset) | |
44b8bd7e | 669 | { |
cad321ad CM |
670 | int ret; |
671 | ||
672 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, | |
673 | bio, 1); | |
674 | BUG_ON(ret); | |
675 | ||
7b6d91da | 676 | if (!(rw & REQ_WRITE)) { |
4a69a410 CM |
677 | /* |
678 | * called for a read, do the setup so that checksum validation | |
679 | * can happen in the async kernel threads | |
680 | */ | |
4a69a410 | 681 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, |
6f3577bd | 682 | mirror_num, 0); |
44b8bd7e | 683 | } |
d313d7a3 | 684 | |
cad321ad CM |
685 | /* |
686 | * kthread helpers are used to submit writes so that checksumming | |
687 | * can happen in parallel across all CPUs | |
688 | */ | |
44b8bd7e | 689 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, |
c8b97818 | 690 | inode, rw, bio, mirror_num, 0, |
eaf25d93 | 691 | bio_offset, |
4a69a410 CM |
692 | __btree_submit_bio_start, |
693 | __btree_submit_bio_done); | |
44b8bd7e CM |
694 | } |
695 | ||
0da5468f CM |
696 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
697 | { | |
d1310b2e | 698 | struct extent_io_tree *tree; |
b9473439 CM |
699 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
700 | struct extent_buffer *eb; | |
701 | int was_dirty; | |
702 | ||
d1310b2e | 703 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
b9473439 CM |
704 | if (!(current->flags & PF_MEMALLOC)) { |
705 | return extent_write_full_page(tree, page, | |
706 | btree_get_extent, wbc); | |
707 | } | |
5443be45 | 708 | |
b9473439 CM |
709 | redirty_page_for_writepage(wbc, page); |
710 | eb = btrfs_find_tree_block(root, page_offset(page), | |
711 | PAGE_CACHE_SIZE); | |
712 | WARN_ON(!eb); | |
713 | ||
714 | was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags); | |
715 | if (!was_dirty) { | |
716 | spin_lock(&root->fs_info->delalloc_lock); | |
717 | root->fs_info->dirty_metadata_bytes += PAGE_CACHE_SIZE; | |
718 | spin_unlock(&root->fs_info->delalloc_lock); | |
5443be45 | 719 | } |
b9473439 CM |
720 | free_extent_buffer(eb); |
721 | ||
722 | unlock_page(page); | |
723 | return 0; | |
5f39d397 | 724 | } |
0da5468f CM |
725 | |
726 | static int btree_writepages(struct address_space *mapping, | |
727 | struct writeback_control *wbc) | |
728 | { | |
d1310b2e CM |
729 | struct extent_io_tree *tree; |
730 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 731 | if (wbc->sync_mode == WB_SYNC_NONE) { |
b9473439 | 732 | struct btrfs_root *root = BTRFS_I(mapping->host)->root; |
793955bc | 733 | u64 num_dirty; |
24ab9cd8 | 734 | unsigned long thresh = 32 * 1024 * 1024; |
448d640b CM |
735 | |
736 | if (wbc->for_kupdate) | |
737 | return 0; | |
738 | ||
b9473439 CM |
739 | /* this is a bit racy, but that's ok */ |
740 | num_dirty = root->fs_info->dirty_metadata_bytes; | |
d397712b | 741 | if (num_dirty < thresh) |
793955bc | 742 | return 0; |
793955bc | 743 | } |
0da5468f CM |
744 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
745 | } | |
746 | ||
b2950863 | 747 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 748 | { |
d1310b2e CM |
749 | struct extent_io_tree *tree; |
750 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
751 | return extent_read_full_page(tree, page, btree_get_extent); |
752 | } | |
22b0ebda | 753 | |
70dec807 | 754 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 755 | { |
d1310b2e CM |
756 | struct extent_io_tree *tree; |
757 | struct extent_map_tree *map; | |
5f39d397 | 758 | int ret; |
d98237b3 | 759 | |
98509cfc | 760 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 761 | return 0; |
98509cfc | 762 | |
d1310b2e CM |
763 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
764 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
6af118ce | 765 | |
7b13b7b1 | 766 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
d397712b | 767 | if (!ret) |
6af118ce | 768 | return 0; |
6af118ce CM |
769 | |
770 | ret = try_release_extent_buffer(tree, page); | |
5f39d397 CM |
771 | if (ret == 1) { |
772 | ClearPagePrivate(page); | |
773 | set_page_private(page, 0); | |
774 | page_cache_release(page); | |
775 | } | |
6af118ce | 776 | |
d98237b3 CM |
777 | return ret; |
778 | } | |
779 | ||
5f39d397 | 780 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 781 | { |
d1310b2e CM |
782 | struct extent_io_tree *tree; |
783 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
784 | extent_invalidatepage(tree, page, offset); |
785 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 786 | if (PagePrivate(page)) { |
d397712b CM |
787 | printk(KERN_WARNING "btrfs warning page private not zero " |
788 | "on page %llu\n", (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
789 | ClearPagePrivate(page); |
790 | set_page_private(page, 0); | |
791 | page_cache_release(page); | |
792 | } | |
d98237b3 CM |
793 | } |
794 | ||
7f09410b | 795 | static const struct address_space_operations btree_aops = { |
d98237b3 CM |
796 | .readpage = btree_readpage, |
797 | .writepage = btree_writepage, | |
0da5468f | 798 | .writepages = btree_writepages, |
5f39d397 CM |
799 | .releasepage = btree_releasepage, |
800 | .invalidatepage = btree_invalidatepage, | |
d98237b3 CM |
801 | .sync_page = block_sync_page, |
802 | }; | |
803 | ||
ca7a79ad CM |
804 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
805 | u64 parent_transid) | |
090d1875 | 806 | { |
5f39d397 CM |
807 | struct extent_buffer *buf = NULL; |
808 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 809 | int ret = 0; |
090d1875 | 810 | |
db94535d | 811 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 812 | if (!buf) |
090d1875 | 813 | return 0; |
d1310b2e | 814 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
f188591e | 815 | buf, 0, 0, btree_get_extent, 0); |
5f39d397 | 816 | free_extent_buffer(buf); |
de428b63 | 817 | return ret; |
090d1875 CM |
818 | } |
819 | ||
0999df54 CM |
820 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
821 | u64 bytenr, u32 blocksize) | |
822 | { | |
823 | struct inode *btree_inode = root->fs_info->btree_inode; | |
824 | struct extent_buffer *eb; | |
825 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
826 | bytenr, blocksize, GFP_NOFS); | |
827 | return eb; | |
828 | } | |
829 | ||
830 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
831 | u64 bytenr, u32 blocksize) | |
832 | { | |
833 | struct inode *btree_inode = root->fs_info->btree_inode; | |
834 | struct extent_buffer *eb; | |
835 | ||
836 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
837 | bytenr, blocksize, NULL, GFP_NOFS); | |
838 | return eb; | |
839 | } | |
840 | ||
841 | ||
e02119d5 CM |
842 | int btrfs_write_tree_block(struct extent_buffer *buf) |
843 | { | |
8aa38c31 CH |
844 | return filemap_fdatawrite_range(buf->first_page->mapping, buf->start, |
845 | buf->start + buf->len - 1); | |
e02119d5 CM |
846 | } |
847 | ||
848 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
849 | { | |
8aa38c31 CH |
850 | return filemap_fdatawait_range(buf->first_page->mapping, |
851 | buf->start, buf->start + buf->len - 1); | |
e02119d5 CM |
852 | } |
853 | ||
0999df54 | 854 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 855 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
856 | { |
857 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
858 | int ret; |
859 | ||
0999df54 CM |
860 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
861 | if (!buf) | |
862 | return NULL; | |
0999df54 | 863 | |
ca7a79ad | 864 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 | 865 | |
d397712b | 866 | if (ret == 0) |
b4ce94de | 867 | set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags); |
5f39d397 | 868 | return buf; |
ce9adaa5 | 869 | |
eb60ceac CM |
870 | } |
871 | ||
e089f05c | 872 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 873 | struct extent_buffer *buf) |
ed2ff2cb | 874 | { |
5f39d397 | 875 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 | 876 | if (btrfs_header_generation(buf) == |
925baedd | 877 | root->fs_info->running_transaction->transid) { |
b9447ef8 | 878 | btrfs_assert_tree_locked(buf); |
b4ce94de | 879 | |
b9473439 CM |
880 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
881 | spin_lock(&root->fs_info->delalloc_lock); | |
882 | if (root->fs_info->dirty_metadata_bytes >= buf->len) | |
883 | root->fs_info->dirty_metadata_bytes -= buf->len; | |
884 | else | |
885 | WARN_ON(1); | |
886 | spin_unlock(&root->fs_info->delalloc_lock); | |
887 | } | |
b4ce94de | 888 | |
b9473439 CM |
889 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
890 | btrfs_set_lock_blocking(buf); | |
d1310b2e | 891 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 892 | buf); |
925baedd | 893 | } |
5f39d397 CM |
894 | return 0; |
895 | } | |
896 | ||
db94535d | 897 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 898 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 899 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 900 | u64 objectid) |
d97e63b6 | 901 | { |
cfaa7295 | 902 | root->node = NULL; |
a28ec197 | 903 | root->commit_root = NULL; |
db94535d CM |
904 | root->sectorsize = sectorsize; |
905 | root->nodesize = nodesize; | |
906 | root->leafsize = leafsize; | |
87ee04eb | 907 | root->stripesize = stripesize; |
123abc88 | 908 | root->ref_cows = 0; |
0b86a832 | 909 | root->track_dirty = 0; |
c71bf099 | 910 | root->in_radix = 0; |
d68fc57b YZ |
911 | root->orphan_item_inserted = 0; |
912 | root->orphan_cleanup_state = 0; | |
0b86a832 | 913 | |
9f5fae2f | 914 | root->fs_info = fs_info; |
0f7d52f4 CM |
915 | root->objectid = objectid; |
916 | root->last_trans = 0; | |
13a8a7c8 | 917 | root->highest_objectid = 0; |
58176a96 | 918 | root->name = NULL; |
4313b399 | 919 | root->in_sysfs = 0; |
6bef4d31 | 920 | root->inode_tree = RB_ROOT; |
f0486c68 | 921 | root->block_rsv = NULL; |
d68fc57b | 922 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
923 | |
924 | INIT_LIST_HEAD(&root->dirty_list); | |
7b128766 | 925 | INIT_LIST_HEAD(&root->orphan_list); |
5d4f98a2 | 926 | INIT_LIST_HEAD(&root->root_list); |
925baedd | 927 | spin_lock_init(&root->node_lock); |
d68fc57b | 928 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 929 | spin_lock_init(&root->inode_lock); |
f0486c68 | 930 | spin_lock_init(&root->accounting_lock); |
a2135011 | 931 | mutex_init(&root->objectid_mutex); |
e02119d5 | 932 | mutex_init(&root->log_mutex); |
7237f183 YZ |
933 | init_waitqueue_head(&root->log_writer_wait); |
934 | init_waitqueue_head(&root->log_commit_wait[0]); | |
935 | init_waitqueue_head(&root->log_commit_wait[1]); | |
936 | atomic_set(&root->log_commit[0], 0); | |
937 | atomic_set(&root->log_commit[1], 0); | |
938 | atomic_set(&root->log_writers, 0); | |
939 | root->log_batch = 0; | |
940 | root->log_transid = 0; | |
257c62e1 | 941 | root->last_log_commit = 0; |
d0c803c4 CM |
942 | extent_io_tree_init(&root->dirty_log_pages, |
943 | fs_info->btree_inode->i_mapping, GFP_NOFS); | |
017e5369 | 944 | |
3768f368 CM |
945 | memset(&root->root_key, 0, sizeof(root->root_key)); |
946 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 947 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 948 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 949 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 950 | init_completion(&root->kobj_unregister); |
6702ed49 | 951 | root->defrag_running = 0; |
4d775673 | 952 | root->root_key.objectid = objectid; |
3394e160 CM |
953 | root->anon_super.s_root = NULL; |
954 | root->anon_super.s_dev = 0; | |
955 | INIT_LIST_HEAD(&root->anon_super.s_list); | |
956 | INIT_LIST_HEAD(&root->anon_super.s_instances); | |
957 | init_rwsem(&root->anon_super.s_umount); | |
958 | ||
3768f368 CM |
959 | return 0; |
960 | } | |
961 | ||
db94535d | 962 | static int find_and_setup_root(struct btrfs_root *tree_root, |
9f5fae2f CM |
963 | struct btrfs_fs_info *fs_info, |
964 | u64 objectid, | |
e20d96d6 | 965 | struct btrfs_root *root) |
3768f368 CM |
966 | { |
967 | int ret; | |
db94535d | 968 | u32 blocksize; |
84234f3a | 969 | u64 generation; |
3768f368 | 970 | |
db94535d | 971 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
972 | tree_root->sectorsize, tree_root->stripesize, |
973 | root, fs_info, objectid); | |
3768f368 CM |
974 | ret = btrfs_find_last_root(tree_root, objectid, |
975 | &root->root_item, &root->root_key); | |
4df27c4d YZ |
976 | if (ret > 0) |
977 | return -ENOENT; | |
3768f368 CM |
978 | BUG_ON(ret); |
979 | ||
84234f3a | 980 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
981 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
982 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 983 | blocksize, generation); |
3768f368 | 984 | BUG_ON(!root->node); |
4df27c4d | 985 | root->commit_root = btrfs_root_node(root); |
d97e63b6 CM |
986 | return 0; |
987 | } | |
988 | ||
7237f183 YZ |
989 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
990 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
991 | { |
992 | struct btrfs_root *root; | |
993 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 994 | struct extent_buffer *leaf; |
e02119d5 CM |
995 | |
996 | root = kzalloc(sizeof(*root), GFP_NOFS); | |
997 | if (!root) | |
7237f183 | 998 | return ERR_PTR(-ENOMEM); |
e02119d5 CM |
999 | |
1000 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1001 | tree_root->sectorsize, tree_root->stripesize, | |
1002 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1003 | ||
1004 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1005 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1006 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
7237f183 YZ |
1007 | /* |
1008 | * log trees do not get reference counted because they go away | |
1009 | * before a real commit is actually done. They do store pointers | |
1010 | * to file data extents, and those reference counts still get | |
1011 | * updated (along with back refs to the log tree). | |
1012 | */ | |
e02119d5 CM |
1013 | root->ref_cows = 0; |
1014 | ||
5d4f98a2 YZ |
1015 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, |
1016 | BTRFS_TREE_LOG_OBJECTID, NULL, 0, 0, 0); | |
7237f183 YZ |
1017 | if (IS_ERR(leaf)) { |
1018 | kfree(root); | |
1019 | return ERR_CAST(leaf); | |
1020 | } | |
e02119d5 | 1021 | |
5d4f98a2 YZ |
1022 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1023 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1024 | btrfs_set_header_generation(leaf, trans->transid); | |
1025 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1026 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1027 | root->node = leaf; |
e02119d5 CM |
1028 | |
1029 | write_extent_buffer(root->node, root->fs_info->fsid, | |
1030 | (unsigned long)btrfs_header_fsid(root->node), | |
1031 | BTRFS_FSID_SIZE); | |
1032 | btrfs_mark_buffer_dirty(root->node); | |
1033 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1034 | return root; |
1035 | } | |
1036 | ||
1037 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1038 | struct btrfs_fs_info *fs_info) | |
1039 | { | |
1040 | struct btrfs_root *log_root; | |
1041 | ||
1042 | log_root = alloc_log_tree(trans, fs_info); | |
1043 | if (IS_ERR(log_root)) | |
1044 | return PTR_ERR(log_root); | |
1045 | WARN_ON(fs_info->log_root_tree); | |
1046 | fs_info->log_root_tree = log_root; | |
1047 | return 0; | |
1048 | } | |
1049 | ||
1050 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1051 | struct btrfs_root *root) | |
1052 | { | |
1053 | struct btrfs_root *log_root; | |
1054 | struct btrfs_inode_item *inode_item; | |
1055 | ||
1056 | log_root = alloc_log_tree(trans, root->fs_info); | |
1057 | if (IS_ERR(log_root)) | |
1058 | return PTR_ERR(log_root); | |
1059 | ||
1060 | log_root->last_trans = trans->transid; | |
1061 | log_root->root_key.offset = root->root_key.objectid; | |
1062 | ||
1063 | inode_item = &log_root->root_item.inode; | |
1064 | inode_item->generation = cpu_to_le64(1); | |
1065 | inode_item->size = cpu_to_le64(3); | |
1066 | inode_item->nlink = cpu_to_le32(1); | |
1067 | inode_item->nbytes = cpu_to_le64(root->leafsize); | |
1068 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); | |
1069 | ||
5d4f98a2 | 1070 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1071 | |
1072 | WARN_ON(root->log_root); | |
1073 | root->log_root = log_root; | |
1074 | root->log_transid = 0; | |
257c62e1 | 1075 | root->last_log_commit = 0; |
e02119d5 CM |
1076 | return 0; |
1077 | } | |
1078 | ||
1079 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
1080 | struct btrfs_key *location) | |
1081 | { | |
1082 | struct btrfs_root *root; | |
1083 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1084 | struct btrfs_path *path; |
5f39d397 | 1085 | struct extent_buffer *l; |
84234f3a | 1086 | u64 generation; |
db94535d | 1087 | u32 blocksize; |
0f7d52f4 CM |
1088 | int ret = 0; |
1089 | ||
5eda7b5e | 1090 | root = kzalloc(sizeof(*root), GFP_NOFS); |
0cf6c620 | 1091 | if (!root) |
0f7d52f4 | 1092 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 1093 | if (location->offset == (u64)-1) { |
db94535d | 1094 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
1095 | location->objectid, root); |
1096 | if (ret) { | |
0f7d52f4 CM |
1097 | kfree(root); |
1098 | return ERR_PTR(ret); | |
1099 | } | |
13a8a7c8 | 1100 | goto out; |
0f7d52f4 CM |
1101 | } |
1102 | ||
db94535d | 1103 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1104 | tree_root->sectorsize, tree_root->stripesize, |
1105 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
1106 | |
1107 | path = btrfs_alloc_path(); | |
1108 | BUG_ON(!path); | |
1109 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); | |
13a8a7c8 YZ |
1110 | if (ret == 0) { |
1111 | l = path->nodes[0]; | |
1112 | read_extent_buffer(l, &root->root_item, | |
1113 | btrfs_item_ptr_offset(l, path->slots[0]), | |
1114 | sizeof(root->root_item)); | |
1115 | memcpy(&root->root_key, location, sizeof(*location)); | |
0f7d52f4 | 1116 | } |
0f7d52f4 CM |
1117 | btrfs_free_path(path); |
1118 | if (ret) { | |
13a8a7c8 YZ |
1119 | if (ret > 0) |
1120 | ret = -ENOENT; | |
0f7d52f4 CM |
1121 | return ERR_PTR(ret); |
1122 | } | |
13a8a7c8 | 1123 | |
84234f3a | 1124 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1125 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1126 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1127 | blocksize, generation); |
5d4f98a2 | 1128 | root->commit_root = btrfs_root_node(root); |
0f7d52f4 | 1129 | BUG_ON(!root->node); |
13a8a7c8 YZ |
1130 | out: |
1131 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) | |
e02119d5 | 1132 | root->ref_cows = 1; |
13a8a7c8 | 1133 | |
5eda7b5e CM |
1134 | return root; |
1135 | } | |
1136 | ||
dc17ff8f CM |
1137 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1138 | u64 root_objectid) | |
1139 | { | |
1140 | struct btrfs_root *root; | |
1141 | ||
1142 | if (root_objectid == BTRFS_ROOT_TREE_OBJECTID) | |
1143 | return fs_info->tree_root; | |
1144 | if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1145 | return fs_info->extent_root; | |
1146 | ||
1147 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1148 | (unsigned long)root_objectid); | |
1149 | return root; | |
1150 | } | |
1151 | ||
edbd8d4e CM |
1152 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
1153 | struct btrfs_key *location) | |
5eda7b5e CM |
1154 | { |
1155 | struct btrfs_root *root; | |
1156 | int ret; | |
1157 | ||
edbd8d4e CM |
1158 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1159 | return fs_info->tree_root; | |
1160 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1161 | return fs_info->extent_root; | |
8f18cf13 CM |
1162 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1163 | return fs_info->chunk_root; | |
1164 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1165 | return fs_info->dev_root; | |
0403e47e YZ |
1166 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1167 | return fs_info->csum_root; | |
4df27c4d YZ |
1168 | again: |
1169 | spin_lock(&fs_info->fs_roots_radix_lock); | |
5eda7b5e CM |
1170 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
1171 | (unsigned long)location->objectid); | |
4df27c4d | 1172 | spin_unlock(&fs_info->fs_roots_radix_lock); |
5eda7b5e CM |
1173 | if (root) |
1174 | return root; | |
1175 | ||
e02119d5 | 1176 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
1177 | if (IS_ERR(root)) |
1178 | return root; | |
3394e160 CM |
1179 | |
1180 | set_anon_super(&root->anon_super, NULL); | |
1181 | ||
d68fc57b YZ |
1182 | if (btrfs_root_refs(&root->root_item) == 0) { |
1183 | ret = -ENOENT; | |
1184 | goto fail; | |
1185 | } | |
1186 | ||
1187 | ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid); | |
1188 | if (ret < 0) | |
1189 | goto fail; | |
1190 | if (ret == 0) | |
1191 | root->orphan_item_inserted = 1; | |
1192 | ||
4df27c4d YZ |
1193 | ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); |
1194 | if (ret) | |
1195 | goto fail; | |
1196 | ||
1197 | spin_lock(&fs_info->fs_roots_radix_lock); | |
2619ba1f CM |
1198 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
1199 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 | 1200 | root); |
d68fc57b | 1201 | if (ret == 0) |
4df27c4d | 1202 | root->in_radix = 1; |
d68fc57b | 1203 | |
4df27c4d YZ |
1204 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1205 | radix_tree_preload_end(); | |
0f7d52f4 | 1206 | if (ret) { |
4df27c4d YZ |
1207 | if (ret == -EEXIST) { |
1208 | free_fs_root(root); | |
1209 | goto again; | |
1210 | } | |
1211 | goto fail; | |
0f7d52f4 | 1212 | } |
4df27c4d YZ |
1213 | |
1214 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
1215 | root->root_key.objectid); | |
1216 | WARN_ON(ret); | |
edbd8d4e | 1217 | return root; |
4df27c4d YZ |
1218 | fail: |
1219 | free_fs_root(root); | |
1220 | return ERR_PTR(ret); | |
edbd8d4e CM |
1221 | } |
1222 | ||
1223 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, | |
1224 | struct btrfs_key *location, | |
1225 | const char *name, int namelen) | |
1226 | { | |
4df27c4d YZ |
1227 | return btrfs_read_fs_root_no_name(fs_info, location); |
1228 | #if 0 | |
edbd8d4e CM |
1229 | struct btrfs_root *root; |
1230 | int ret; | |
1231 | ||
1232 | root = btrfs_read_fs_root_no_name(fs_info, location); | |
1233 | if (!root) | |
1234 | return NULL; | |
58176a96 | 1235 | |
4313b399 CM |
1236 | if (root->in_sysfs) |
1237 | return root; | |
1238 | ||
58176a96 JB |
1239 | ret = btrfs_set_root_name(root, name, namelen); |
1240 | if (ret) { | |
5f39d397 | 1241 | free_extent_buffer(root->node); |
58176a96 JB |
1242 | kfree(root); |
1243 | return ERR_PTR(ret); | |
1244 | } | |
4df27c4d | 1245 | |
58176a96 JB |
1246 | ret = btrfs_sysfs_add_root(root); |
1247 | if (ret) { | |
5f39d397 | 1248 | free_extent_buffer(root->node); |
58176a96 JB |
1249 | kfree(root->name); |
1250 | kfree(root); | |
1251 | return ERR_PTR(ret); | |
1252 | } | |
4313b399 | 1253 | root->in_sysfs = 1; |
0f7d52f4 | 1254 | return root; |
4df27c4d | 1255 | #endif |
0f7d52f4 | 1256 | } |
04160088 CM |
1257 | |
1258 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) | |
1259 | { | |
1260 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1261 | int ret = 0; | |
04160088 CM |
1262 | struct btrfs_device *device; |
1263 | struct backing_dev_info *bdi; | |
b7967db7 | 1264 | |
c6e30871 | 1265 | list_for_each_entry(device, &info->fs_devices->devices, dev_list) { |
dfe25020 CM |
1266 | if (!device->bdev) |
1267 | continue; | |
04160088 CM |
1268 | bdi = blk_get_backing_dev_info(device->bdev); |
1269 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1270 | ret = 1; | |
1271 | break; | |
1272 | } | |
1273 | } | |
1274 | return ret; | |
1275 | } | |
1276 | ||
38b66988 CM |
1277 | /* |
1278 | * this unplugs every device on the box, and it is only used when page | |
1279 | * is null | |
1280 | */ | |
1281 | static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) | |
1282 | { | |
38b66988 CM |
1283 | struct btrfs_device *device; |
1284 | struct btrfs_fs_info *info; | |
1285 | ||
1286 | info = (struct btrfs_fs_info *)bdi->unplug_io_data; | |
c6e30871 | 1287 | list_for_each_entry(device, &info->fs_devices->devices, dev_list) { |
d20f7043 CM |
1288 | if (!device->bdev) |
1289 | continue; | |
1290 | ||
38b66988 | 1291 | bdi = blk_get_backing_dev_info(device->bdev); |
d397712b | 1292 | if (bdi->unplug_io_fn) |
38b66988 | 1293 | bdi->unplug_io_fn(bdi, page); |
38b66988 CM |
1294 | } |
1295 | } | |
1296 | ||
b2950863 | 1297 | static void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) |
04160088 | 1298 | { |
38b66988 | 1299 | struct inode *inode; |
f2d8d74d CM |
1300 | struct extent_map_tree *em_tree; |
1301 | struct extent_map *em; | |
bcbfce8a | 1302 | struct address_space *mapping; |
38b66988 CM |
1303 | u64 offset; |
1304 | ||
bcbfce8a | 1305 | /* the generic O_DIRECT read code does this */ |
9f0ba5bd | 1306 | if (1 || !page) { |
38b66988 CM |
1307 | __unplug_io_fn(bdi, page); |
1308 | return; | |
1309 | } | |
1310 | ||
bcbfce8a CM |
1311 | /* |
1312 | * page->mapping may change at any time. Get a consistent copy | |
1313 | * and use that for everything below | |
1314 | */ | |
1315 | smp_mb(); | |
1316 | mapping = page->mapping; | |
1317 | if (!mapping) | |
1318 | return; | |
1319 | ||
1320 | inode = mapping->host; | |
240d5d48 CM |
1321 | |
1322 | /* | |
1323 | * don't do the expensive searching for a small number of | |
1324 | * devices | |
1325 | */ | |
1326 | if (BTRFS_I(inode)->root->fs_info->fs_devices->open_devices <= 2) { | |
1327 | __unplug_io_fn(bdi, page); | |
1328 | return; | |
1329 | } | |
1330 | ||
38b66988 | 1331 | offset = page_offset(page); |
04160088 | 1332 | |
f2d8d74d | 1333 | em_tree = &BTRFS_I(inode)->extent_tree; |
890871be | 1334 | read_lock(&em_tree->lock); |
f2d8d74d | 1335 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); |
890871be | 1336 | read_unlock(&em_tree->lock); |
89642229 CM |
1337 | if (!em) { |
1338 | __unplug_io_fn(bdi, page); | |
f2d8d74d | 1339 | return; |
89642229 | 1340 | } |
f2d8d74d | 1341 | |
89642229 CM |
1342 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { |
1343 | free_extent_map(em); | |
1344 | __unplug_io_fn(bdi, page); | |
1345 | return; | |
1346 | } | |
f2d8d74d CM |
1347 | offset = offset - em->start; |
1348 | btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree, | |
1349 | em->block_start + offset, page); | |
1350 | free_extent_map(em); | |
04160088 CM |
1351 | } |
1352 | ||
ad081f14 JA |
1353 | /* |
1354 | * If this fails, caller must call bdi_destroy() to get rid of the | |
1355 | * bdi again. | |
1356 | */ | |
04160088 CM |
1357 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1358 | { | |
ad081f14 JA |
1359 | int err; |
1360 | ||
1361 | bdi->capabilities = BDI_CAP_MAP_COPY; | |
e6d086d8 | 1362 | err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY); |
ad081f14 JA |
1363 | if (err) |
1364 | return err; | |
1365 | ||
4575c9cc | 1366 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1367 | bdi->unplug_io_fn = btrfs_unplug_io_fn; |
1368 | bdi->unplug_io_data = info; | |
1369 | bdi->congested_fn = btrfs_congested_fn; | |
1370 | bdi->congested_data = info; | |
1371 | return 0; | |
1372 | } | |
1373 | ||
ce9adaa5 CM |
1374 | static int bio_ready_for_csum(struct bio *bio) |
1375 | { | |
1376 | u64 length = 0; | |
1377 | u64 buf_len = 0; | |
1378 | u64 start = 0; | |
1379 | struct page *page; | |
1380 | struct extent_io_tree *io_tree = NULL; | |
ce9adaa5 CM |
1381 | struct bio_vec *bvec; |
1382 | int i; | |
1383 | int ret; | |
1384 | ||
1385 | bio_for_each_segment(bvec, bio, i) { | |
1386 | page = bvec->bv_page; | |
1387 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
1388 | length += bvec->bv_len; | |
1389 | continue; | |
1390 | } | |
1391 | if (!page->private) { | |
1392 | length += bvec->bv_len; | |
1393 | continue; | |
1394 | } | |
1395 | length = bvec->bv_len; | |
1396 | buf_len = page->private >> 2; | |
1397 | start = page_offset(page) + bvec->bv_offset; | |
1398 | io_tree = &BTRFS_I(page->mapping->host)->io_tree; | |
ce9adaa5 CM |
1399 | } |
1400 | /* are we fully contained in this bio? */ | |
1401 | if (buf_len <= length) | |
1402 | return 1; | |
1403 | ||
1404 | ret = extent_range_uptodate(io_tree, start + length, | |
1405 | start + buf_len - 1); | |
ce9adaa5 CM |
1406 | return ret; |
1407 | } | |
1408 | ||
8b712842 CM |
1409 | /* |
1410 | * called by the kthread helper functions to finally call the bio end_io | |
1411 | * functions. This is where read checksum verification actually happens | |
1412 | */ | |
1413 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1414 | { |
ce9adaa5 | 1415 | struct bio *bio; |
8b712842 CM |
1416 | struct end_io_wq *end_io_wq; |
1417 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1418 | int error; |
ce9adaa5 | 1419 | |
8b712842 CM |
1420 | end_io_wq = container_of(work, struct end_io_wq, work); |
1421 | bio = end_io_wq->bio; | |
1422 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1423 | |
cad321ad | 1424 | /* metadata bio reads are special because the whole tree block must |
8b712842 CM |
1425 | * be checksummed at once. This makes sure the entire block is in |
1426 | * ram and up to date before trying to verify things. For | |
1427 | * blocksize <= pagesize, it is basically a noop | |
1428 | */ | |
7b6d91da | 1429 | if (!(bio->bi_rw & REQ_WRITE) && end_io_wq->metadata && |
cad321ad | 1430 | !bio_ready_for_csum(bio)) { |
d20f7043 | 1431 | btrfs_queue_worker(&fs_info->endio_meta_workers, |
8b712842 CM |
1432 | &end_io_wq->work); |
1433 | return; | |
1434 | } | |
1435 | error = end_io_wq->error; | |
1436 | bio->bi_private = end_io_wq->private; | |
1437 | bio->bi_end_io = end_io_wq->end_io; | |
1438 | kfree(end_io_wq); | |
8b712842 | 1439 | bio_endio(bio, error); |
44b8bd7e CM |
1440 | } |
1441 | ||
a74a4b97 CM |
1442 | static int cleaner_kthread(void *arg) |
1443 | { | |
1444 | struct btrfs_root *root = arg; | |
1445 | ||
1446 | do { | |
a74a4b97 | 1447 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); |
76dda93c YZ |
1448 | |
1449 | if (!(root->fs_info->sb->s_flags & MS_RDONLY) && | |
1450 | mutex_trylock(&root->fs_info->cleaner_mutex)) { | |
24bbcf04 | 1451 | btrfs_run_delayed_iputs(root); |
76dda93c YZ |
1452 | btrfs_clean_old_snapshots(root); |
1453 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
1454 | } | |
a74a4b97 CM |
1455 | |
1456 | if (freezing(current)) { | |
1457 | refrigerator(); | |
1458 | } else { | |
a74a4b97 | 1459 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1460 | if (!kthread_should_stop()) |
1461 | schedule(); | |
a74a4b97 CM |
1462 | __set_current_state(TASK_RUNNING); |
1463 | } | |
1464 | } while (!kthread_should_stop()); | |
1465 | return 0; | |
1466 | } | |
1467 | ||
1468 | static int transaction_kthread(void *arg) | |
1469 | { | |
1470 | struct btrfs_root *root = arg; | |
1471 | struct btrfs_trans_handle *trans; | |
1472 | struct btrfs_transaction *cur; | |
8929ecfa | 1473 | u64 transid; |
a74a4b97 CM |
1474 | unsigned long now; |
1475 | unsigned long delay; | |
1476 | int ret; | |
1477 | ||
1478 | do { | |
a74a4b97 CM |
1479 | delay = HZ * 30; |
1480 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1481 | mutex_lock(&root->fs_info->transaction_kthread_mutex); | |
1482 | ||
8929ecfa | 1483 | spin_lock(&root->fs_info->new_trans_lock); |
a74a4b97 CM |
1484 | cur = root->fs_info->running_transaction; |
1485 | if (!cur) { | |
8929ecfa | 1486 | spin_unlock(&root->fs_info->new_trans_lock); |
a74a4b97 CM |
1487 | goto sleep; |
1488 | } | |
31153d81 | 1489 | |
a74a4b97 | 1490 | now = get_seconds(); |
8929ecfa YZ |
1491 | if (!cur->blocked && |
1492 | (now < cur->start_time || now - cur->start_time < 30)) { | |
1493 | spin_unlock(&root->fs_info->new_trans_lock); | |
a74a4b97 CM |
1494 | delay = HZ * 5; |
1495 | goto sleep; | |
1496 | } | |
8929ecfa YZ |
1497 | transid = cur->transid; |
1498 | spin_unlock(&root->fs_info->new_trans_lock); | |
56bec294 | 1499 | |
8929ecfa YZ |
1500 | trans = btrfs_join_transaction(root, 1); |
1501 | if (transid == trans->transid) { | |
1502 | ret = btrfs_commit_transaction(trans, root); | |
1503 | BUG_ON(ret); | |
1504 | } else { | |
1505 | btrfs_end_transaction(trans, root); | |
1506 | } | |
a74a4b97 CM |
1507 | sleep: |
1508 | wake_up_process(root->fs_info->cleaner_kthread); | |
1509 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1510 | ||
1511 | if (freezing(current)) { | |
1512 | refrigerator(); | |
1513 | } else { | |
a74a4b97 | 1514 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1515 | if (!kthread_should_stop() && |
1516 | !btrfs_transaction_blocked(root->fs_info)) | |
1517 | schedule_timeout(delay); | |
a74a4b97 CM |
1518 | __set_current_state(TASK_RUNNING); |
1519 | } | |
1520 | } while (!kthread_should_stop()); | |
1521 | return 0; | |
1522 | } | |
1523 | ||
8a4b83cc | 1524 | struct btrfs_root *open_ctree(struct super_block *sb, |
dfe25020 CM |
1525 | struct btrfs_fs_devices *fs_devices, |
1526 | char *options) | |
2e635a27 | 1527 | { |
db94535d CM |
1528 | u32 sectorsize; |
1529 | u32 nodesize; | |
1530 | u32 leafsize; | |
1531 | u32 blocksize; | |
87ee04eb | 1532 | u32 stripesize; |
84234f3a | 1533 | u64 generation; |
f2b636e8 | 1534 | u64 features; |
3de4586c | 1535 | struct btrfs_key location; |
a061fc8d | 1536 | struct buffer_head *bh; |
e02119d5 | 1537 | struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1538 | GFP_NOFS); |
d20f7043 CM |
1539 | struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root), |
1540 | GFP_NOFS); | |
e02119d5 | 1541 | struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1542 | GFP_NOFS); |
8790d502 | 1543 | struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), |
e20d96d6 | 1544 | GFP_NOFS); |
e02119d5 | 1545 | struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1546 | GFP_NOFS); |
e02119d5 | 1547 | struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1548 | GFP_NOFS); |
e02119d5 CM |
1549 | struct btrfs_root *log_tree_root; |
1550 | ||
eb60ceac | 1551 | int ret; |
e58ca020 | 1552 | int err = -EINVAL; |
4543df7e | 1553 | |
2c90e5d6 | 1554 | struct btrfs_super_block *disk_super; |
8790d502 | 1555 | |
0463bb4e | 1556 | if (!extent_root || !tree_root || !fs_info || |
d20f7043 | 1557 | !chunk_root || !dev_root || !csum_root) { |
39279cc3 CM |
1558 | err = -ENOMEM; |
1559 | goto fail; | |
1560 | } | |
76dda93c YZ |
1561 | |
1562 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
1563 | if (ret) { | |
1564 | err = ret; | |
1565 | goto fail; | |
1566 | } | |
1567 | ||
1568 | ret = setup_bdi(fs_info, &fs_info->bdi); | |
1569 | if (ret) { | |
1570 | err = ret; | |
1571 | goto fail_srcu; | |
1572 | } | |
1573 | ||
1574 | fs_info->btree_inode = new_inode(sb); | |
1575 | if (!fs_info->btree_inode) { | |
1576 | err = -ENOMEM; | |
1577 | goto fail_bdi; | |
1578 | } | |
1579 | ||
1580 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); | |
8fd17795 | 1581 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1582 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 1583 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
19c00ddc | 1584 | INIT_LIST_HEAD(&fs_info->hashers); |
ea8c2819 | 1585 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
5a3f23d5 | 1586 | INIT_LIST_HEAD(&fs_info->ordered_operations); |
11833d66 | 1587 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
1832a6d5 | 1588 | spin_lock_init(&fs_info->delalloc_lock); |
cee36a03 | 1589 | spin_lock_init(&fs_info->new_trans_lock); |
31153d81 | 1590 | spin_lock_init(&fs_info->ref_cache_lock); |
76dda93c | 1591 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 1592 | spin_lock_init(&fs_info->delayed_iput_lock); |
19c00ddc | 1593 | |
58176a96 | 1594 | init_completion(&fs_info->kobj_unregister); |
9f5fae2f CM |
1595 | fs_info->tree_root = tree_root; |
1596 | fs_info->extent_root = extent_root; | |
d20f7043 | 1597 | fs_info->csum_root = csum_root; |
0b86a832 CM |
1598 | fs_info->chunk_root = chunk_root; |
1599 | fs_info->dev_root = dev_root; | |
8a4b83cc | 1600 | fs_info->fs_devices = fs_devices; |
0b86a832 | 1601 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1602 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1603 | btrfs_mapping_init(&fs_info->mapping_tree); |
f0486c68 YZ |
1604 | btrfs_init_block_rsv(&fs_info->global_block_rsv); |
1605 | btrfs_init_block_rsv(&fs_info->delalloc_block_rsv); | |
1606 | btrfs_init_block_rsv(&fs_info->trans_block_rsv); | |
1607 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv); | |
1608 | btrfs_init_block_rsv(&fs_info->empty_block_rsv); | |
1609 | INIT_LIST_HEAD(&fs_info->durable_block_rsv_list); | |
1610 | mutex_init(&fs_info->durable_block_rsv_mutex); | |
cb03c743 | 1611 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 1612 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 1613 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 1614 | atomic_set(&fs_info->nr_async_bios, 0); |
e20d96d6 | 1615 | fs_info->sb = sb; |
6f568d35 | 1616 | fs_info->max_inline = 8192 * 1024; |
9ed74f2d | 1617 | fs_info->metadata_ratio = 0; |
c8b97818 | 1618 | |
b34b086c CM |
1619 | fs_info->thread_pool_size = min_t(unsigned long, |
1620 | num_online_cpus() + 2, 8); | |
0afbaf8c | 1621 | |
3eaa2885 CM |
1622 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
1623 | spin_lock_init(&fs_info->ordered_extent_lock); | |
1624 | ||
a061fc8d CM |
1625 | sb->s_blocksize = 4096; |
1626 | sb->s_blocksize_bits = blksize_bits(4096); | |
32a88aa1 | 1627 | sb->s_bdi = &fs_info->bdi; |
a061fc8d | 1628 | |
76dda93c YZ |
1629 | fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; |
1630 | fs_info->btree_inode->i_nlink = 1; | |
0afbaf8c CM |
1631 | /* |
1632 | * we set the i_size on the btree inode to the max possible int. | |
1633 | * the real end of the address space is determined by all of | |
1634 | * the devices in the system | |
1635 | */ | |
1636 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 1637 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
1638 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
1639 | ||
5d4f98a2 | 1640 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); |
d1310b2e | 1641 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
5f39d397 CM |
1642 | fs_info->btree_inode->i_mapping, |
1643 | GFP_NOFS); | |
d1310b2e CM |
1644 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, |
1645 | GFP_NOFS); | |
1646 | ||
1647 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 1648 | |
76dda93c YZ |
1649 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
1650 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
1651 | sizeof(struct btrfs_key)); | |
1652 | BTRFS_I(fs_info->btree_inode)->dummy_inode = 1; | |
c65ddb52 | 1653 | insert_inode_hash(fs_info->btree_inode); |
76dda93c | 1654 | |
0f9dd46c | 1655 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 1656 | fs_info->block_group_cache_tree = RB_ROOT; |
0f9dd46c | 1657 | |
11833d66 | 1658 | extent_io_tree_init(&fs_info->freed_extents[0], |
1a5bc167 | 1659 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
11833d66 YZ |
1660 | extent_io_tree_init(&fs_info->freed_extents[1], |
1661 | fs_info->btree_inode->i_mapping, GFP_NOFS); | |
1662 | fs_info->pinned_extents = &fs_info->freed_extents[0]; | |
e66f709b | 1663 | fs_info->do_barriers = 1; |
e18e4809 | 1664 | |
39279cc3 | 1665 | |
79154b1b | 1666 | mutex_init(&fs_info->trans_mutex); |
5a3f23d5 | 1667 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 1668 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 1669 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
1670 | mutex_init(&fs_info->transaction_kthread_mutex); |
1671 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 1672 | mutex_init(&fs_info->volume_mutex); |
276e680d | 1673 | init_rwsem(&fs_info->extent_commit_sem); |
c71bf099 | 1674 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 1675 | init_rwsem(&fs_info->subvol_sem); |
fa9c0d79 CM |
1676 | |
1677 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); | |
1678 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
1679 | ||
e6dcd2dc | 1680 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 1681 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 1682 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 1683 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 1684 | |
0b86a832 | 1685 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 1686 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 1687 | |
a512bbf8 | 1688 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
a061fc8d | 1689 | if (!bh) |
39279cc3 | 1690 | goto fail_iput; |
39279cc3 | 1691 | |
a061fc8d | 1692 | memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy)); |
2d69a0f8 YZ |
1693 | memcpy(&fs_info->super_for_commit, &fs_info->super_copy, |
1694 | sizeof(fs_info->super_for_commit)); | |
a061fc8d | 1695 | brelse(bh); |
5f39d397 | 1696 | |
a061fc8d | 1697 | memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE); |
0b86a832 | 1698 | |
5f39d397 | 1699 | disk_super = &fs_info->super_copy; |
0f7d52f4 | 1700 | if (!btrfs_super_root(disk_super)) |
c6e2bac1 | 1701 | goto fail_iput; |
0f7d52f4 | 1702 | |
2b82032c YZ |
1703 | ret = btrfs_parse_options(tree_root, options); |
1704 | if (ret) { | |
1705 | err = ret; | |
c6e2bac1 | 1706 | goto fail_iput; |
2b82032c | 1707 | } |
dfe25020 | 1708 | |
f2b636e8 JB |
1709 | features = btrfs_super_incompat_flags(disk_super) & |
1710 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
1711 | if (features) { | |
1712 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
1713 | "unsupported optional features (%Lx).\n", | |
21380931 | 1714 | (unsigned long long)features); |
f2b636e8 | 1715 | err = -EINVAL; |
c6e2bac1 | 1716 | goto fail_iput; |
f2b636e8 JB |
1717 | } |
1718 | ||
5d4f98a2 YZ |
1719 | features = btrfs_super_incompat_flags(disk_super); |
1720 | if (!(features & BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF)) { | |
1721 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; | |
1722 | btrfs_set_super_incompat_flags(disk_super, features); | |
1723 | } | |
1724 | ||
f2b636e8 JB |
1725 | features = btrfs_super_compat_ro_flags(disk_super) & |
1726 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
1727 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
1728 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
1729 | "unsupported option features (%Lx).\n", | |
21380931 | 1730 | (unsigned long long)features); |
f2b636e8 | 1731 | err = -EINVAL; |
c6e2bac1 | 1732 | goto fail_iput; |
f2b636e8 | 1733 | } |
61d92c32 CM |
1734 | |
1735 | btrfs_init_workers(&fs_info->generic_worker, | |
1736 | "genwork", 1, NULL); | |
1737 | ||
5443be45 | 1738 | btrfs_init_workers(&fs_info->workers, "worker", |
61d92c32 CM |
1739 | fs_info->thread_pool_size, |
1740 | &fs_info->generic_worker); | |
c8b97818 | 1741 | |
771ed689 | 1742 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
61d92c32 CM |
1743 | fs_info->thread_pool_size, |
1744 | &fs_info->generic_worker); | |
771ed689 | 1745 | |
5443be45 | 1746 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 | 1747 | min_t(u64, fs_devices->num_devices, |
61d92c32 CM |
1748 | fs_info->thread_pool_size), |
1749 | &fs_info->generic_worker); | |
61b49440 CM |
1750 | |
1751 | /* a higher idle thresh on the submit workers makes it much more | |
1752 | * likely that bios will be send down in a sane order to the | |
1753 | * devices | |
1754 | */ | |
1755 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 1756 | |
771ed689 | 1757 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 1758 | fs_info->workers.ordered = 1; |
61b49440 | 1759 | |
771ed689 CM |
1760 | fs_info->delalloc_workers.idle_thresh = 2; |
1761 | fs_info->delalloc_workers.ordered = 1; | |
1762 | ||
61d92c32 CM |
1763 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1, |
1764 | &fs_info->generic_worker); | |
5443be45 | 1765 | btrfs_init_workers(&fs_info->endio_workers, "endio", |
61d92c32 CM |
1766 | fs_info->thread_pool_size, |
1767 | &fs_info->generic_worker); | |
d20f7043 | 1768 | btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", |
61d92c32 CM |
1769 | fs_info->thread_pool_size, |
1770 | &fs_info->generic_worker); | |
cad321ad | 1771 | btrfs_init_workers(&fs_info->endio_meta_write_workers, |
61d92c32 CM |
1772 | "endio-meta-write", fs_info->thread_pool_size, |
1773 | &fs_info->generic_worker); | |
5443be45 | 1774 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", |
61d92c32 CM |
1775 | fs_info->thread_pool_size, |
1776 | &fs_info->generic_worker); | |
0cb59c99 JB |
1777 | btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write", |
1778 | 1, &fs_info->generic_worker); | |
61b49440 CM |
1779 | |
1780 | /* | |
1781 | * endios are largely parallel and should have a very | |
1782 | * low idle thresh | |
1783 | */ | |
1784 | fs_info->endio_workers.idle_thresh = 4; | |
b51912c9 CM |
1785 | fs_info->endio_meta_workers.idle_thresh = 4; |
1786 | ||
9042846b CM |
1787 | fs_info->endio_write_workers.idle_thresh = 2; |
1788 | fs_info->endio_meta_write_workers.idle_thresh = 2; | |
1789 | ||
4543df7e | 1790 | btrfs_start_workers(&fs_info->workers, 1); |
61d92c32 | 1791 | btrfs_start_workers(&fs_info->generic_worker, 1); |
1cc127b5 | 1792 | btrfs_start_workers(&fs_info->submit_workers, 1); |
771ed689 | 1793 | btrfs_start_workers(&fs_info->delalloc_workers, 1); |
247e743c | 1794 | btrfs_start_workers(&fs_info->fixup_workers, 1); |
9042846b CM |
1795 | btrfs_start_workers(&fs_info->endio_workers, 1); |
1796 | btrfs_start_workers(&fs_info->endio_meta_workers, 1); | |
1797 | btrfs_start_workers(&fs_info->endio_meta_write_workers, 1); | |
1798 | btrfs_start_workers(&fs_info->endio_write_workers, 1); | |
0cb59c99 | 1799 | btrfs_start_workers(&fs_info->endio_freespace_worker, 1); |
4543df7e | 1800 | |
4575c9cc | 1801 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
1802 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
1803 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 1804 | |
db94535d CM |
1805 | nodesize = btrfs_super_nodesize(disk_super); |
1806 | leafsize = btrfs_super_leafsize(disk_super); | |
1807 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 1808 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
1809 | tree_root->nodesize = nodesize; |
1810 | tree_root->leafsize = leafsize; | |
1811 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 1812 | tree_root->stripesize = stripesize; |
a061fc8d CM |
1813 | |
1814 | sb->s_blocksize = sectorsize; | |
1815 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 1816 | |
39279cc3 CM |
1817 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
1818 | sizeof(disk_super->magic))) { | |
d397712b | 1819 | printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
1820 | goto fail_sb_buffer; |
1821 | } | |
19c00ddc | 1822 | |
925baedd | 1823 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 1824 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 1825 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 1826 | if (ret) { |
d397712b CM |
1827 | printk(KERN_WARNING "btrfs: failed to read the system " |
1828 | "array on %s\n", sb->s_id); | |
5d4f98a2 | 1829 | goto fail_sb_buffer; |
84eed90f | 1830 | } |
0b86a832 CM |
1831 | |
1832 | blocksize = btrfs_level_size(tree_root, | |
1833 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 1834 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
1835 | |
1836 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1837 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
1838 | ||
1839 | chunk_root->node = read_tree_block(chunk_root, | |
1840 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 1841 | blocksize, generation); |
0b86a832 | 1842 | BUG_ON(!chunk_root->node); |
83121942 DW |
1843 | if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { |
1844 | printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n", | |
1845 | sb->s_id); | |
1846 | goto fail_chunk_root; | |
1847 | } | |
5d4f98a2 YZ |
1848 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
1849 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 1850 | |
e17cade2 | 1851 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
d397712b CM |
1852 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), |
1853 | BTRFS_UUID_SIZE); | |
e17cade2 | 1854 | |
925baedd | 1855 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1856 | ret = btrfs_read_chunk_tree(chunk_root); |
925baedd | 1857 | mutex_unlock(&fs_info->chunk_mutex); |
2b82032c | 1858 | if (ret) { |
d397712b CM |
1859 | printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", |
1860 | sb->s_id); | |
2b82032c YZ |
1861 | goto fail_chunk_root; |
1862 | } | |
0b86a832 | 1863 | |
dfe25020 CM |
1864 | btrfs_close_extra_devices(fs_devices); |
1865 | ||
db94535d CM |
1866 | blocksize = btrfs_level_size(tree_root, |
1867 | btrfs_super_root_level(disk_super)); | |
84234f3a | 1868 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 1869 | |
e20d96d6 | 1870 | tree_root->node = read_tree_block(tree_root, |
db94535d | 1871 | btrfs_super_root(disk_super), |
84234f3a | 1872 | blocksize, generation); |
39279cc3 | 1873 | if (!tree_root->node) |
2b82032c | 1874 | goto fail_chunk_root; |
83121942 DW |
1875 | if (!test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { |
1876 | printk(KERN_WARNING "btrfs: failed to read tree root on %s\n", | |
1877 | sb->s_id); | |
1878 | goto fail_tree_root; | |
1879 | } | |
5d4f98a2 YZ |
1880 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
1881 | tree_root->commit_root = btrfs_root_node(tree_root); | |
db94535d CM |
1882 | |
1883 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 1884 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 1885 | if (ret) |
39279cc3 | 1886 | goto fail_tree_root; |
0b86a832 CM |
1887 | extent_root->track_dirty = 1; |
1888 | ||
1889 | ret = find_and_setup_root(tree_root, fs_info, | |
1890 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
0b86a832 CM |
1891 | if (ret) |
1892 | goto fail_extent_root; | |
5d4f98a2 | 1893 | dev_root->track_dirty = 1; |
3768f368 | 1894 | |
d20f7043 CM |
1895 | ret = find_and_setup_root(tree_root, fs_info, |
1896 | BTRFS_CSUM_TREE_OBJECTID, csum_root); | |
1897 | if (ret) | |
5d4f98a2 | 1898 | goto fail_dev_root; |
d20f7043 CM |
1899 | |
1900 | csum_root->track_dirty = 1; | |
1901 | ||
8929ecfa YZ |
1902 | fs_info->generation = generation; |
1903 | fs_info->last_trans_committed = generation; | |
1904 | fs_info->data_alloc_profile = (u64)-1; | |
1905 | fs_info->metadata_alloc_profile = (u64)-1; | |
1906 | fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; | |
1907 | ||
1b1d1f66 JB |
1908 | ret = btrfs_read_block_groups(extent_root); |
1909 | if (ret) { | |
1910 | printk(KERN_ERR "Failed to read block groups: %d\n", ret); | |
1911 | goto fail_block_groups; | |
1912 | } | |
9078a3e1 | 1913 | |
a74a4b97 CM |
1914 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
1915 | "btrfs-cleaner"); | |
57506d50 | 1916 | if (IS_ERR(fs_info->cleaner_kthread)) |
1b1d1f66 | 1917 | goto fail_block_groups; |
a74a4b97 CM |
1918 | |
1919 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
1920 | tree_root, | |
1921 | "btrfs-transaction"); | |
57506d50 | 1922 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 1923 | goto fail_cleaner; |
a74a4b97 | 1924 | |
c289811c CM |
1925 | if (!btrfs_test_opt(tree_root, SSD) && |
1926 | !btrfs_test_opt(tree_root, NOSSD) && | |
1927 | !fs_info->fs_devices->rotating) { | |
1928 | printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD " | |
1929 | "mode\n"); | |
1930 | btrfs_set_opt(fs_info->mount_opt, SSD); | |
1931 | } | |
1932 | ||
e02119d5 | 1933 | if (btrfs_super_log_root(disk_super) != 0) { |
e02119d5 CM |
1934 | u64 bytenr = btrfs_super_log_root(disk_super); |
1935 | ||
7c2ca468 | 1936 | if (fs_devices->rw_devices == 0) { |
d397712b CM |
1937 | printk(KERN_WARNING "Btrfs log replay required " |
1938 | "on RO media\n"); | |
7c2ca468 CM |
1939 | err = -EIO; |
1940 | goto fail_trans_kthread; | |
1941 | } | |
e02119d5 CM |
1942 | blocksize = |
1943 | btrfs_level_size(tree_root, | |
1944 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 1945 | |
676e4c86 DC |
1946 | log_tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS); |
1947 | if (!log_tree_root) { | |
1948 | err = -ENOMEM; | |
1949 | goto fail_trans_kthread; | |
1950 | } | |
e02119d5 CM |
1951 | |
1952 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1953 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1954 | ||
1955 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
1956 | blocksize, |
1957 | generation + 1); | |
e02119d5 CM |
1958 | ret = btrfs_recover_log_trees(log_tree_root); |
1959 | BUG_ON(ret); | |
e556ce2c YZ |
1960 | |
1961 | if (sb->s_flags & MS_RDONLY) { | |
1962 | ret = btrfs_commit_super(tree_root); | |
1963 | BUG_ON(ret); | |
1964 | } | |
e02119d5 | 1965 | } |
1a40e23b | 1966 | |
76dda93c YZ |
1967 | ret = btrfs_find_orphan_roots(tree_root); |
1968 | BUG_ON(ret); | |
1969 | ||
7c2ca468 | 1970 | if (!(sb->s_flags & MS_RDONLY)) { |
d68fc57b YZ |
1971 | ret = btrfs_cleanup_fs_roots(fs_info); |
1972 | BUG_ON(ret); | |
1973 | ||
5d4f98a2 | 1974 | ret = btrfs_recover_relocation(tree_root); |
d7ce5843 MX |
1975 | if (ret < 0) { |
1976 | printk(KERN_WARNING | |
1977 | "btrfs: failed to recover relocation\n"); | |
1978 | err = -EINVAL; | |
1979 | goto fail_trans_kthread; | |
1980 | } | |
7c2ca468 | 1981 | } |
1a40e23b | 1982 | |
3de4586c CM |
1983 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
1984 | location.type = BTRFS_ROOT_ITEM_KEY; | |
1985 | location.offset = (u64)-1; | |
1986 | ||
3de4586c CM |
1987 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
1988 | if (!fs_info->fs_root) | |
7c2ca468 | 1989 | goto fail_trans_kthread; |
3140c9a3 DC |
1990 | if (IS_ERR(fs_info->fs_root)) { |
1991 | err = PTR_ERR(fs_info->fs_root); | |
1992 | goto fail_trans_kthread; | |
1993 | } | |
c289811c | 1994 | |
e3acc2a6 JB |
1995 | if (!(sb->s_flags & MS_RDONLY)) { |
1996 | down_read(&fs_info->cleanup_work_sem); | |
1997 | btrfs_orphan_cleanup(fs_info->fs_root); | |
0af3d00b | 1998 | btrfs_orphan_cleanup(fs_info->tree_root); |
e3acc2a6 JB |
1999 | up_read(&fs_info->cleanup_work_sem); |
2000 | } | |
2001 | ||
0f7d52f4 | 2002 | return tree_root; |
39279cc3 | 2003 | |
7c2ca468 CM |
2004 | fail_trans_kthread: |
2005 | kthread_stop(fs_info->transaction_kthread); | |
3f157a2f | 2006 | fail_cleaner: |
a74a4b97 | 2007 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
2008 | |
2009 | /* | |
2010 | * make sure we're done with the btree inode before we stop our | |
2011 | * kthreads | |
2012 | */ | |
2013 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
2014 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
2015 | ||
1b1d1f66 JB |
2016 | fail_block_groups: |
2017 | btrfs_free_block_groups(fs_info); | |
d20f7043 | 2018 | free_extent_buffer(csum_root->node); |
5d4f98a2 YZ |
2019 | free_extent_buffer(csum_root->commit_root); |
2020 | fail_dev_root: | |
2021 | free_extent_buffer(dev_root->node); | |
2022 | free_extent_buffer(dev_root->commit_root); | |
0b86a832 CM |
2023 | fail_extent_root: |
2024 | free_extent_buffer(extent_root->node); | |
5d4f98a2 | 2025 | free_extent_buffer(extent_root->commit_root); |
39279cc3 | 2026 | fail_tree_root: |
5f39d397 | 2027 | free_extent_buffer(tree_root->node); |
5d4f98a2 | 2028 | free_extent_buffer(tree_root->commit_root); |
2b82032c YZ |
2029 | fail_chunk_root: |
2030 | free_extent_buffer(chunk_root->node); | |
5d4f98a2 | 2031 | free_extent_buffer(chunk_root->commit_root); |
39279cc3 | 2032 | fail_sb_buffer: |
61d92c32 | 2033 | btrfs_stop_workers(&fs_info->generic_worker); |
247e743c | 2034 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2035 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2036 | btrfs_stop_workers(&fs_info->workers); |
2037 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 2038 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 2039 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 2040 | btrfs_stop_workers(&fs_info->endio_write_workers); |
0cb59c99 | 2041 | btrfs_stop_workers(&fs_info->endio_freespace_worker); |
1cc127b5 | 2042 | btrfs_stop_workers(&fs_info->submit_workers); |
4543df7e | 2043 | fail_iput: |
7c2ca468 | 2044 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
4543df7e | 2045 | iput(fs_info->btree_inode); |
7e662854 | 2046 | |
dfe25020 | 2047 | btrfs_close_devices(fs_info->fs_devices); |
84eed90f | 2048 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
ad081f14 | 2049 | fail_bdi: |
7e662854 | 2050 | bdi_destroy(&fs_info->bdi); |
76dda93c YZ |
2051 | fail_srcu: |
2052 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 2053 | fail: |
39279cc3 CM |
2054 | kfree(extent_root); |
2055 | kfree(tree_root); | |
2056 | kfree(fs_info); | |
83afeac4 JM |
2057 | kfree(chunk_root); |
2058 | kfree(dev_root); | |
d20f7043 | 2059 | kfree(csum_root); |
39279cc3 | 2060 | return ERR_PTR(err); |
eb60ceac CM |
2061 | } |
2062 | ||
f2984462 CM |
2063 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
2064 | { | |
2065 | char b[BDEVNAME_SIZE]; | |
2066 | ||
2067 | if (uptodate) { | |
2068 | set_buffer_uptodate(bh); | |
2069 | } else { | |
2070 | if (!buffer_eopnotsupp(bh) && printk_ratelimit()) { | |
2071 | printk(KERN_WARNING "lost page write due to " | |
2072 | "I/O error on %s\n", | |
2073 | bdevname(bh->b_bdev, b)); | |
2074 | } | |
1259ab75 CM |
2075 | /* note, we dont' set_buffer_write_io_error because we have |
2076 | * our own ways of dealing with the IO errors | |
2077 | */ | |
f2984462 CM |
2078 | clear_buffer_uptodate(bh); |
2079 | } | |
2080 | unlock_buffer(bh); | |
2081 | put_bh(bh); | |
2082 | } | |
2083 | ||
a512bbf8 YZ |
2084 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
2085 | { | |
2086 | struct buffer_head *bh; | |
2087 | struct buffer_head *latest = NULL; | |
2088 | struct btrfs_super_block *super; | |
2089 | int i; | |
2090 | u64 transid = 0; | |
2091 | u64 bytenr; | |
2092 | ||
2093 | /* we would like to check all the supers, but that would make | |
2094 | * a btrfs mount succeed after a mkfs from a different FS. | |
2095 | * So, we need to add a special mount option to scan for | |
2096 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
2097 | */ | |
2098 | for (i = 0; i < 1; i++) { | |
2099 | bytenr = btrfs_sb_offset(i); | |
2100 | if (bytenr + 4096 >= i_size_read(bdev->bd_inode)) | |
2101 | break; | |
2102 | bh = __bread(bdev, bytenr / 4096, 4096); | |
2103 | if (!bh) | |
2104 | continue; | |
2105 | ||
2106 | super = (struct btrfs_super_block *)bh->b_data; | |
2107 | if (btrfs_super_bytenr(super) != bytenr || | |
2108 | strncmp((char *)(&super->magic), BTRFS_MAGIC, | |
2109 | sizeof(super->magic))) { | |
2110 | brelse(bh); | |
2111 | continue; | |
2112 | } | |
2113 | ||
2114 | if (!latest || btrfs_super_generation(super) > transid) { | |
2115 | brelse(latest); | |
2116 | latest = bh; | |
2117 | transid = btrfs_super_generation(super); | |
2118 | } else { | |
2119 | brelse(bh); | |
2120 | } | |
2121 | } | |
2122 | return latest; | |
2123 | } | |
2124 | ||
4eedeb75 HH |
2125 | /* |
2126 | * this should be called twice, once with wait == 0 and | |
2127 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
2128 | * we write are pinned. | |
2129 | * | |
2130 | * They are released when wait == 1 is done. | |
2131 | * max_mirrors must be the same for both runs, and it indicates how | |
2132 | * many supers on this one device should be written. | |
2133 | * | |
2134 | * max_mirrors == 0 means to write them all. | |
2135 | */ | |
a512bbf8 YZ |
2136 | static int write_dev_supers(struct btrfs_device *device, |
2137 | struct btrfs_super_block *sb, | |
2138 | int do_barriers, int wait, int max_mirrors) | |
2139 | { | |
2140 | struct buffer_head *bh; | |
2141 | int i; | |
2142 | int ret; | |
2143 | int errors = 0; | |
2144 | u32 crc; | |
2145 | u64 bytenr; | |
2146 | int last_barrier = 0; | |
2147 | ||
2148 | if (max_mirrors == 0) | |
2149 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
2150 | ||
2151 | /* make sure only the last submit_bh does a barrier */ | |
2152 | if (do_barriers) { | |
2153 | for (i = 0; i < max_mirrors; i++) { | |
2154 | bytenr = btrfs_sb_offset(i); | |
2155 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= | |
2156 | device->total_bytes) | |
2157 | break; | |
2158 | last_barrier = i; | |
2159 | } | |
2160 | } | |
2161 | ||
2162 | for (i = 0; i < max_mirrors; i++) { | |
2163 | bytenr = btrfs_sb_offset(i); | |
2164 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) | |
2165 | break; | |
2166 | ||
2167 | if (wait) { | |
2168 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
2169 | BTRFS_SUPER_INFO_SIZE); | |
2170 | BUG_ON(!bh); | |
a512bbf8 | 2171 | wait_on_buffer(bh); |
4eedeb75 HH |
2172 | if (!buffer_uptodate(bh)) |
2173 | errors++; | |
2174 | ||
2175 | /* drop our reference */ | |
2176 | brelse(bh); | |
2177 | ||
2178 | /* drop the reference from the wait == 0 run */ | |
2179 | brelse(bh); | |
2180 | continue; | |
a512bbf8 YZ |
2181 | } else { |
2182 | btrfs_set_super_bytenr(sb, bytenr); | |
2183 | ||
2184 | crc = ~(u32)0; | |
2185 | crc = btrfs_csum_data(NULL, (char *)sb + | |
2186 | BTRFS_CSUM_SIZE, crc, | |
2187 | BTRFS_SUPER_INFO_SIZE - | |
2188 | BTRFS_CSUM_SIZE); | |
2189 | btrfs_csum_final(crc, sb->csum); | |
2190 | ||
4eedeb75 HH |
2191 | /* |
2192 | * one reference for us, and we leave it for the | |
2193 | * caller | |
2194 | */ | |
a512bbf8 YZ |
2195 | bh = __getblk(device->bdev, bytenr / 4096, |
2196 | BTRFS_SUPER_INFO_SIZE); | |
2197 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); | |
2198 | ||
4eedeb75 | 2199 | /* one reference for submit_bh */ |
a512bbf8 | 2200 | get_bh(bh); |
4eedeb75 HH |
2201 | |
2202 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
2203 | lock_buffer(bh); |
2204 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
2205 | } | |
2206 | ||
2207 | if (i == last_barrier && do_barriers && device->barriers) { | |
2208 | ret = submit_bh(WRITE_BARRIER, bh); | |
2209 | if (ret == -EOPNOTSUPP) { | |
2210 | printk("btrfs: disabling barriers on dev %s\n", | |
2211 | device->name); | |
2212 | set_buffer_uptodate(bh); | |
2213 | device->barriers = 0; | |
4eedeb75 | 2214 | /* one reference for submit_bh */ |
a512bbf8 YZ |
2215 | get_bh(bh); |
2216 | lock_buffer(bh); | |
ffbd517d | 2217 | ret = submit_bh(WRITE_SYNC, bh); |
a512bbf8 YZ |
2218 | } |
2219 | } else { | |
ffbd517d | 2220 | ret = submit_bh(WRITE_SYNC, bh); |
a512bbf8 YZ |
2221 | } |
2222 | ||
4eedeb75 | 2223 | if (ret) |
a512bbf8 | 2224 | errors++; |
a512bbf8 YZ |
2225 | } |
2226 | return errors < i ? 0 : -1; | |
2227 | } | |
2228 | ||
2229 | int write_all_supers(struct btrfs_root *root, int max_mirrors) | |
f2984462 | 2230 | { |
e5e9a520 | 2231 | struct list_head *head; |
f2984462 | 2232 | struct btrfs_device *dev; |
a061fc8d | 2233 | struct btrfs_super_block *sb; |
f2984462 | 2234 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
2235 | int ret; |
2236 | int do_barriers; | |
a236aed1 CM |
2237 | int max_errors; |
2238 | int total_errors = 0; | |
a061fc8d | 2239 | u64 flags; |
f2984462 | 2240 | |
a236aed1 | 2241 | max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1; |
f2984462 CM |
2242 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
2243 | ||
a061fc8d CM |
2244 | sb = &root->fs_info->super_for_commit; |
2245 | dev_item = &sb->dev_item; | |
e5e9a520 CM |
2246 | |
2247 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); | |
2248 | head = &root->fs_info->fs_devices->devices; | |
c6e30871 | 2249 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
2250 | if (!dev->bdev) { |
2251 | total_errors++; | |
2252 | continue; | |
2253 | } | |
2b82032c | 2254 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2255 | continue; |
2256 | ||
2b82032c | 2257 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
2258 | btrfs_set_stack_device_type(dev_item, dev->type); |
2259 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
2260 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
2261 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
2262 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
2263 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
2264 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
2265 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 2266 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 2267 | |
a061fc8d CM |
2268 | flags = btrfs_super_flags(sb); |
2269 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
2270 | ||
a512bbf8 | 2271 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
2272 | if (ret) |
2273 | total_errors++; | |
f2984462 | 2274 | } |
a236aed1 | 2275 | if (total_errors > max_errors) { |
d397712b CM |
2276 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2277 | total_errors); | |
a236aed1 CM |
2278 | BUG(); |
2279 | } | |
f2984462 | 2280 | |
a512bbf8 | 2281 | total_errors = 0; |
c6e30871 | 2282 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
2283 | if (!dev->bdev) |
2284 | continue; | |
2b82032c | 2285 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2286 | continue; |
2287 | ||
a512bbf8 YZ |
2288 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
2289 | if (ret) | |
2290 | total_errors++; | |
f2984462 | 2291 | } |
e5e9a520 | 2292 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 2293 | if (total_errors > max_errors) { |
d397712b CM |
2294 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2295 | total_errors); | |
a236aed1 CM |
2296 | BUG(); |
2297 | } | |
f2984462 CM |
2298 | return 0; |
2299 | } | |
2300 | ||
a512bbf8 YZ |
2301 | int write_ctree_super(struct btrfs_trans_handle *trans, |
2302 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 2303 | { |
e66f709b | 2304 | int ret; |
5f39d397 | 2305 | |
a512bbf8 | 2306 | ret = write_all_supers(root, max_mirrors); |
5f39d397 | 2307 | return ret; |
cfaa7295 CM |
2308 | } |
2309 | ||
5eda7b5e | 2310 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f | 2311 | { |
4df27c4d | 2312 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
2313 | radix_tree_delete(&fs_info->fs_roots_radix, |
2314 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 2315 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
2316 | |
2317 | if (btrfs_root_refs(&root->root_item) == 0) | |
2318 | synchronize_srcu(&fs_info->subvol_srcu); | |
2319 | ||
4df27c4d YZ |
2320 | free_fs_root(root); |
2321 | return 0; | |
2322 | } | |
2323 | ||
2324 | static void free_fs_root(struct btrfs_root *root) | |
2325 | { | |
2326 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); | |
3394e160 CM |
2327 | if (root->anon_super.s_dev) { |
2328 | down_write(&root->anon_super.s_umount); | |
2329 | kill_anon_super(&root->anon_super); | |
2330 | } | |
4df27c4d YZ |
2331 | free_extent_buffer(root->node); |
2332 | free_extent_buffer(root->commit_root); | |
d397712b | 2333 | kfree(root->name); |
2619ba1f | 2334 | kfree(root); |
2619ba1f CM |
2335 | } |
2336 | ||
35b7e476 | 2337 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
2338 | { |
2339 | int ret; | |
2340 | struct btrfs_root *gang[8]; | |
2341 | int i; | |
2342 | ||
76dda93c YZ |
2343 | while (!list_empty(&fs_info->dead_roots)) { |
2344 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2345 | struct btrfs_root, root_list); | |
2346 | list_del(&gang[0]->root_list); | |
2347 | ||
2348 | if (gang[0]->in_radix) { | |
2349 | btrfs_free_fs_root(fs_info, gang[0]); | |
2350 | } else { | |
2351 | free_extent_buffer(gang[0]->node); | |
2352 | free_extent_buffer(gang[0]->commit_root); | |
2353 | kfree(gang[0]); | |
2354 | } | |
2355 | } | |
2356 | ||
d397712b | 2357 | while (1) { |
0f7d52f4 CM |
2358 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
2359 | (void **)gang, 0, | |
2360 | ARRAY_SIZE(gang)); | |
2361 | if (!ret) | |
2362 | break; | |
2619ba1f | 2363 | for (i = 0; i < ret; i++) |
5eda7b5e | 2364 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
2365 | } |
2366 | return 0; | |
2367 | } | |
b4100d64 | 2368 | |
c146afad | 2369 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 2370 | { |
c146afad YZ |
2371 | u64 root_objectid = 0; |
2372 | struct btrfs_root *gang[8]; | |
2373 | int i; | |
3768f368 | 2374 | int ret; |
e089f05c | 2375 | |
c146afad YZ |
2376 | while (1) { |
2377 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2378 | (void **)gang, root_objectid, | |
2379 | ARRAY_SIZE(gang)); | |
2380 | if (!ret) | |
2381 | break; | |
5d4f98a2 YZ |
2382 | |
2383 | root_objectid = gang[ret - 1]->root_key.objectid + 1; | |
c146afad YZ |
2384 | for (i = 0; i < ret; i++) { |
2385 | root_objectid = gang[i]->root_key.objectid; | |
c146afad YZ |
2386 | btrfs_orphan_cleanup(gang[i]); |
2387 | } | |
2388 | root_objectid++; | |
2389 | } | |
2390 | return 0; | |
2391 | } | |
a2135011 | 2392 | |
c146afad YZ |
2393 | int btrfs_commit_super(struct btrfs_root *root) |
2394 | { | |
2395 | struct btrfs_trans_handle *trans; | |
2396 | int ret; | |
a74a4b97 | 2397 | |
c146afad | 2398 | mutex_lock(&root->fs_info->cleaner_mutex); |
24bbcf04 | 2399 | btrfs_run_delayed_iputs(root); |
a74a4b97 | 2400 | btrfs_clean_old_snapshots(root); |
c146afad | 2401 | mutex_unlock(&root->fs_info->cleaner_mutex); |
c71bf099 YZ |
2402 | |
2403 | /* wait until ongoing cleanup work done */ | |
2404 | down_write(&root->fs_info->cleanup_work_sem); | |
2405 | up_write(&root->fs_info->cleanup_work_sem); | |
2406 | ||
a22285a6 | 2407 | trans = btrfs_join_transaction(root, 1); |
54aa1f4d | 2408 | ret = btrfs_commit_transaction(trans, root); |
c146afad YZ |
2409 | BUG_ON(ret); |
2410 | /* run commit again to drop the original snapshot */ | |
a22285a6 | 2411 | trans = btrfs_join_transaction(root, 1); |
79154b1b CM |
2412 | btrfs_commit_transaction(trans, root); |
2413 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 2414 | BUG_ON(ret); |
d6bfde87 | 2415 | |
a512bbf8 | 2416 | ret = write_ctree_super(NULL, root, 0); |
c146afad YZ |
2417 | return ret; |
2418 | } | |
2419 | ||
2420 | int close_ctree(struct btrfs_root *root) | |
2421 | { | |
2422 | struct btrfs_fs_info *fs_info = root->fs_info; | |
2423 | int ret; | |
2424 | ||
2425 | fs_info->closing = 1; | |
2426 | smp_mb(); | |
2427 | ||
0af3d00b | 2428 | btrfs_put_block_group_cache(fs_info); |
c146afad YZ |
2429 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
2430 | ret = btrfs_commit_super(root); | |
d397712b CM |
2431 | if (ret) |
2432 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
c146afad | 2433 | } |
0f7d52f4 | 2434 | |
8929ecfa YZ |
2435 | kthread_stop(root->fs_info->transaction_kthread); |
2436 | kthread_stop(root->fs_info->cleaner_kthread); | |
2437 | ||
f25784b3 YZ |
2438 | fs_info->closing = 2; |
2439 | smp_mb(); | |
2440 | ||
b0c68f8b | 2441 | if (fs_info->delalloc_bytes) { |
d397712b | 2442 | printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n", |
21380931 | 2443 | (unsigned long long)fs_info->delalloc_bytes); |
b0c68f8b | 2444 | } |
31153d81 | 2445 | if (fs_info->total_ref_cache_size) { |
d397712b CM |
2446 | printk(KERN_INFO "btrfs: at umount reference cache size %llu\n", |
2447 | (unsigned long long)fs_info->total_ref_cache_size); | |
31153d81 | 2448 | } |
bcc63abb | 2449 | |
5d4f98a2 YZ |
2450 | free_extent_buffer(fs_info->extent_root->node); |
2451 | free_extent_buffer(fs_info->extent_root->commit_root); | |
2452 | free_extent_buffer(fs_info->tree_root->node); | |
2453 | free_extent_buffer(fs_info->tree_root->commit_root); | |
2454 | free_extent_buffer(root->fs_info->chunk_root->node); | |
2455 | free_extent_buffer(root->fs_info->chunk_root->commit_root); | |
2456 | free_extent_buffer(root->fs_info->dev_root->node); | |
2457 | free_extent_buffer(root->fs_info->dev_root->commit_root); | |
2458 | free_extent_buffer(root->fs_info->csum_root->node); | |
2459 | free_extent_buffer(root->fs_info->csum_root->commit_root); | |
d20f7043 | 2460 | |
9078a3e1 | 2461 | btrfs_free_block_groups(root->fs_info); |
d10c5f31 | 2462 | |
c146afad | 2463 | del_fs_roots(fs_info); |
d10c5f31 | 2464 | |
c146afad | 2465 | iput(fs_info->btree_inode); |
9ad6b7bc | 2466 | |
61d92c32 | 2467 | btrfs_stop_workers(&fs_info->generic_worker); |
247e743c | 2468 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2469 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2470 | btrfs_stop_workers(&fs_info->workers); |
2471 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 2472 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 2473 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 2474 | btrfs_stop_workers(&fs_info->endio_write_workers); |
0cb59c99 | 2475 | btrfs_stop_workers(&fs_info->endio_freespace_worker); |
1cc127b5 | 2476 | btrfs_stop_workers(&fs_info->submit_workers); |
d6bfde87 | 2477 | |
dfe25020 | 2478 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 2479 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 2480 | |
04160088 | 2481 | bdi_destroy(&fs_info->bdi); |
76dda93c | 2482 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 2483 | |
0f7d52f4 | 2484 | kfree(fs_info->extent_root); |
0f7d52f4 | 2485 | kfree(fs_info->tree_root); |
0b86a832 CM |
2486 | kfree(fs_info->chunk_root); |
2487 | kfree(fs_info->dev_root); | |
d20f7043 | 2488 | kfree(fs_info->csum_root); |
eb60ceac CM |
2489 | return 0; |
2490 | } | |
2491 | ||
1259ab75 | 2492 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 2493 | { |
1259ab75 | 2494 | int ret; |
810191ff | 2495 | struct inode *btree_inode = buf->first_page->mapping->host; |
1259ab75 | 2496 | |
2ac55d41 JB |
2497 | ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf, |
2498 | NULL); | |
1259ab75 CM |
2499 | if (!ret) |
2500 | return ret; | |
2501 | ||
2502 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
2503 | parent_transid); | |
2504 | return !ret; | |
5f39d397 CM |
2505 | } |
2506 | ||
2507 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 2508 | { |
810191ff | 2509 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 2510 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
2511 | buf); |
2512 | } | |
6702ed49 | 2513 | |
5f39d397 CM |
2514 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
2515 | { | |
810191ff | 2516 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
5f39d397 CM |
2517 | u64 transid = btrfs_header_generation(buf); |
2518 | struct inode *btree_inode = root->fs_info->btree_inode; | |
b9473439 | 2519 | int was_dirty; |
b4ce94de | 2520 | |
b9447ef8 | 2521 | btrfs_assert_tree_locked(buf); |
ccd467d6 | 2522 | if (transid != root->fs_info->generation) { |
d397712b CM |
2523 | printk(KERN_CRIT "btrfs transid mismatch buffer %llu, " |
2524 | "found %llu running %llu\n", | |
db94535d | 2525 | (unsigned long long)buf->start, |
d397712b CM |
2526 | (unsigned long long)transid, |
2527 | (unsigned long long)root->fs_info->generation); | |
ccd467d6 CM |
2528 | WARN_ON(1); |
2529 | } | |
b9473439 CM |
2530 | was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
2531 | buf); | |
2532 | if (!was_dirty) { | |
2533 | spin_lock(&root->fs_info->delalloc_lock); | |
2534 | root->fs_info->dirty_metadata_bytes += buf->len; | |
2535 | spin_unlock(&root->fs_info->delalloc_lock); | |
2536 | } | |
eb60ceac CM |
2537 | } |
2538 | ||
d3c2fdcf | 2539 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
35b7e476 | 2540 | { |
188de649 CM |
2541 | /* |
2542 | * looks as though older kernels can get into trouble with | |
2543 | * this code, they end up stuck in balance_dirty_pages forever | |
2544 | */ | |
d6bfde87 | 2545 | u64 num_dirty; |
771ed689 | 2546 | unsigned long thresh = 32 * 1024 * 1024; |
d6bfde87 | 2547 | |
6933c02e | 2548 | if (current->flags & PF_MEMALLOC) |
d6bfde87 CM |
2549 | return; |
2550 | ||
585ad2c3 CM |
2551 | num_dirty = root->fs_info->dirty_metadata_bytes; |
2552 | ||
d6bfde87 CM |
2553 | if (num_dirty > thresh) { |
2554 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 2555 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 2556 | } |
188de649 | 2557 | return; |
35b7e476 | 2558 | } |
6b80053d | 2559 | |
ca7a79ad | 2560 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 2561 | { |
810191ff | 2562 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
ce9adaa5 | 2563 | int ret; |
ca7a79ad | 2564 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
d397712b | 2565 | if (ret == 0) |
b4ce94de | 2566 | set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags); |
ce9adaa5 | 2567 | return ret; |
6b80053d | 2568 | } |
0da5468f | 2569 | |
4bef0848 CM |
2570 | int btree_lock_page_hook(struct page *page) |
2571 | { | |
2572 | struct inode *inode = page->mapping->host; | |
b9473439 | 2573 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4bef0848 CM |
2574 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
2575 | struct extent_buffer *eb; | |
2576 | unsigned long len; | |
2577 | u64 bytenr = page_offset(page); | |
2578 | ||
2579 | if (page->private == EXTENT_PAGE_PRIVATE) | |
2580 | goto out; | |
2581 | ||
2582 | len = page->private >> 2; | |
2583 | eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS); | |
2584 | if (!eb) | |
2585 | goto out; | |
2586 | ||
2587 | btrfs_tree_lock(eb); | |
4bef0848 | 2588 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); |
b9473439 CM |
2589 | |
2590 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) { | |
2591 | spin_lock(&root->fs_info->delalloc_lock); | |
2592 | if (root->fs_info->dirty_metadata_bytes >= eb->len) | |
2593 | root->fs_info->dirty_metadata_bytes -= eb->len; | |
2594 | else | |
2595 | WARN_ON(1); | |
2596 | spin_unlock(&root->fs_info->delalloc_lock); | |
2597 | } | |
2598 | ||
4bef0848 CM |
2599 | btrfs_tree_unlock(eb); |
2600 | free_extent_buffer(eb); | |
2601 | out: | |
2602 | lock_page(page); | |
2603 | return 0; | |
2604 | } | |
2605 | ||
d1310b2e | 2606 | static struct extent_io_ops btree_extent_io_ops = { |
4bef0848 | 2607 | .write_cache_pages_lock_hook = btree_lock_page_hook, |
ce9adaa5 | 2608 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
0b86a832 | 2609 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
2610 | /* note we're sharing with inode.c for the merge bio hook */ |
2611 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 2612 | }; |