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> |
4b4e25f2 | 30 | #include "compat.h" |
eb60ceac CM |
31 | #include "ctree.h" |
32 | #include "disk-io.h" | |
e089f05c | 33 | #include "transaction.h" |
0f7d52f4 | 34 | #include "btrfs_inode.h" |
0b86a832 | 35 | #include "volumes.h" |
db94535d | 36 | #include "print-tree.h" |
8b712842 | 37 | #include "async-thread.h" |
925baedd | 38 | #include "locking.h" |
e02119d5 | 39 | #include "tree-log.h" |
fa9c0d79 | 40 | #include "free-space-cache.h" |
eb60ceac | 41 | |
d1310b2e | 42 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 43 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 44 | static void free_fs_root(struct btrfs_root *root); |
ce9adaa5 | 45 | |
ad081f14 JA |
46 | static atomic_t btrfs_bdi_num = ATOMIC_INIT(0); |
47 | ||
d352ac68 CM |
48 | /* |
49 | * end_io_wq structs are used to do processing in task context when an IO is | |
50 | * complete. This is used during reads to verify checksums, and it is used | |
51 | * by writes to insert metadata for new file extents after IO is complete. | |
52 | */ | |
ce9adaa5 CM |
53 | struct end_io_wq { |
54 | struct bio *bio; | |
55 | bio_end_io_t *end_io; | |
56 | void *private; | |
57 | struct btrfs_fs_info *info; | |
58 | int error; | |
22c59948 | 59 | int metadata; |
ce9adaa5 | 60 | struct list_head list; |
8b712842 | 61 | struct btrfs_work work; |
ce9adaa5 | 62 | }; |
0da5468f | 63 | |
d352ac68 CM |
64 | /* |
65 | * async submit bios are used to offload expensive checksumming | |
66 | * onto the worker threads. They checksum file and metadata bios | |
67 | * just before they are sent down the IO stack. | |
68 | */ | |
44b8bd7e CM |
69 | struct async_submit_bio { |
70 | struct inode *inode; | |
71 | struct bio *bio; | |
72 | struct list_head list; | |
4a69a410 CM |
73 | extent_submit_bio_hook_t *submit_bio_start; |
74 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
75 | int rw; |
76 | int mirror_num; | |
c8b97818 | 77 | unsigned long bio_flags; |
8b712842 | 78 | struct btrfs_work work; |
44b8bd7e CM |
79 | }; |
80 | ||
4008c04a CM |
81 | /* These are used to set the lockdep class on the extent buffer locks. |
82 | * The class is set by the readpage_end_io_hook after the buffer has | |
83 | * passed csum validation but before the pages are unlocked. | |
84 | * | |
85 | * The lockdep class is also set by btrfs_init_new_buffer on freshly | |
86 | * allocated blocks. | |
87 | * | |
88 | * The class is based on the level in the tree block, which allows lockdep | |
89 | * to know that lower nodes nest inside the locks of higher nodes. | |
90 | * | |
91 | * We also add a check to make sure the highest level of the tree is | |
92 | * the same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this | |
93 | * code needs update as well. | |
94 | */ | |
95 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
96 | # if BTRFS_MAX_LEVEL != 8 | |
97 | # error | |
98 | # endif | |
99 | static struct lock_class_key btrfs_eb_class[BTRFS_MAX_LEVEL + 1]; | |
100 | static const char *btrfs_eb_name[BTRFS_MAX_LEVEL + 1] = { | |
101 | /* leaf */ | |
102 | "btrfs-extent-00", | |
103 | "btrfs-extent-01", | |
104 | "btrfs-extent-02", | |
105 | "btrfs-extent-03", | |
106 | "btrfs-extent-04", | |
107 | "btrfs-extent-05", | |
108 | "btrfs-extent-06", | |
109 | "btrfs-extent-07", | |
110 | /* highest possible level */ | |
111 | "btrfs-extent-08", | |
112 | }; | |
113 | #endif | |
114 | ||
d352ac68 CM |
115 | /* |
116 | * extents on the btree inode are pretty simple, there's one extent | |
117 | * that covers the entire device | |
118 | */ | |
b2950863 CH |
119 | static struct extent_map *btree_get_extent(struct inode *inode, |
120 | struct page *page, size_t page_offset, u64 start, u64 len, | |
121 | int create) | |
7eccb903 | 122 | { |
5f39d397 CM |
123 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
124 | struct extent_map *em; | |
125 | int ret; | |
126 | ||
890871be | 127 | read_lock(&em_tree->lock); |
d1310b2e | 128 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d CM |
129 | if (em) { |
130 | em->bdev = | |
131 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
890871be | 132 | read_unlock(&em_tree->lock); |
5f39d397 | 133 | goto out; |
a061fc8d | 134 | } |
890871be | 135 | read_unlock(&em_tree->lock); |
7b13b7b1 | 136 | |
5f39d397 CM |
137 | em = alloc_extent_map(GFP_NOFS); |
138 | if (!em) { | |
139 | em = ERR_PTR(-ENOMEM); | |
140 | goto out; | |
141 | } | |
142 | em->start = 0; | |
0afbaf8c | 143 | em->len = (u64)-1; |
c8b97818 | 144 | em->block_len = (u64)-1; |
5f39d397 | 145 | em->block_start = 0; |
a061fc8d | 146 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e | 147 | |
890871be | 148 | write_lock(&em_tree->lock); |
5f39d397 CM |
149 | ret = add_extent_mapping(em_tree, em); |
150 | if (ret == -EEXIST) { | |
0afbaf8c CM |
151 | u64 failed_start = em->start; |
152 | u64 failed_len = em->len; | |
153 | ||
5f39d397 | 154 | free_extent_map(em); |
7b13b7b1 | 155 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c | 156 | if (em) { |
7b13b7b1 | 157 | ret = 0; |
0afbaf8c CM |
158 | } else { |
159 | em = lookup_extent_mapping(em_tree, failed_start, | |
160 | failed_len); | |
7b13b7b1 | 161 | ret = -EIO; |
0afbaf8c | 162 | } |
5f39d397 | 163 | } else if (ret) { |
7b13b7b1 CM |
164 | free_extent_map(em); |
165 | em = NULL; | |
5f39d397 | 166 | } |
890871be | 167 | write_unlock(&em_tree->lock); |
7b13b7b1 CM |
168 | |
169 | if (ret) | |
170 | em = ERR_PTR(ret); | |
5f39d397 CM |
171 | out: |
172 | return em; | |
7eccb903 CM |
173 | } |
174 | ||
19c00ddc CM |
175 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
176 | { | |
163e783e | 177 | return crc32c(seed, data, len); |
19c00ddc CM |
178 | } |
179 | ||
180 | void btrfs_csum_final(u32 crc, char *result) | |
181 | { | |
182 | *(__le32 *)result = ~cpu_to_le32(crc); | |
183 | } | |
184 | ||
d352ac68 CM |
185 | /* |
186 | * compute the csum for a btree block, and either verify it or write it | |
187 | * into the csum field of the block. | |
188 | */ | |
19c00ddc CM |
189 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
190 | int verify) | |
191 | { | |
607d432d JB |
192 | u16 csum_size = |
193 | btrfs_super_csum_size(&root->fs_info->super_copy); | |
194 | char *result = NULL; | |
19c00ddc CM |
195 | unsigned long len; |
196 | unsigned long cur_len; | |
197 | unsigned long offset = BTRFS_CSUM_SIZE; | |
198 | char *map_token = NULL; | |
199 | char *kaddr; | |
200 | unsigned long map_start; | |
201 | unsigned long map_len; | |
202 | int err; | |
203 | u32 crc = ~(u32)0; | |
607d432d | 204 | unsigned long inline_result; |
19c00ddc CM |
205 | |
206 | len = buf->len - offset; | |
d397712b | 207 | while (len > 0) { |
19c00ddc CM |
208 | err = map_private_extent_buffer(buf, offset, 32, |
209 | &map_token, &kaddr, | |
210 | &map_start, &map_len, KM_USER0); | |
d397712b | 211 | if (err) |
19c00ddc | 212 | return 1; |
19c00ddc CM |
213 | cur_len = min(len, map_len - (offset - map_start)); |
214 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
215 | crc, cur_len); | |
216 | len -= cur_len; | |
217 | offset += cur_len; | |
218 | unmap_extent_buffer(buf, map_token, KM_USER0); | |
219 | } | |
607d432d JB |
220 | if (csum_size > sizeof(inline_result)) { |
221 | result = kzalloc(csum_size * sizeof(char), GFP_NOFS); | |
222 | if (!result) | |
223 | return 1; | |
224 | } else { | |
225 | result = (char *)&inline_result; | |
226 | } | |
227 | ||
19c00ddc CM |
228 | btrfs_csum_final(crc, result); |
229 | ||
230 | if (verify) { | |
607d432d | 231 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
232 | u32 val; |
233 | u32 found = 0; | |
607d432d | 234 | memcpy(&found, result, csum_size); |
e4204ded | 235 | |
607d432d | 236 | read_extent_buffer(buf, &val, 0, csum_size); |
193f284d CM |
237 | if (printk_ratelimit()) { |
238 | printk(KERN_INFO "btrfs: %s checksum verify " | |
239 | "failed on %llu wanted %X found %X " | |
240 | "level %d\n", | |
241 | root->fs_info->sb->s_id, | |
242 | (unsigned long long)buf->start, val, found, | |
243 | btrfs_header_level(buf)); | |
244 | } | |
607d432d JB |
245 | if (result != (char *)&inline_result) |
246 | kfree(result); | |
19c00ddc CM |
247 | return 1; |
248 | } | |
249 | } else { | |
607d432d | 250 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 251 | } |
607d432d JB |
252 | if (result != (char *)&inline_result) |
253 | kfree(result); | |
19c00ddc CM |
254 | return 0; |
255 | } | |
256 | ||
d352ac68 CM |
257 | /* |
258 | * we can't consider a given block up to date unless the transid of the | |
259 | * block matches the transid in the parent node's pointer. This is how we | |
260 | * detect blocks that either didn't get written at all or got written | |
261 | * in the wrong place. | |
262 | */ | |
1259ab75 CM |
263 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
264 | struct extent_buffer *eb, u64 parent_transid) | |
265 | { | |
266 | int ret; | |
267 | ||
268 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
269 | return 0; | |
270 | ||
271 | lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); | |
272 | if (extent_buffer_uptodate(io_tree, eb) && | |
273 | btrfs_header_generation(eb) == parent_transid) { | |
274 | ret = 0; | |
275 | goto out; | |
276 | } | |
193f284d CM |
277 | if (printk_ratelimit()) { |
278 | printk("parent transid verify failed on %llu wanted %llu " | |
279 | "found %llu\n", | |
280 | (unsigned long long)eb->start, | |
281 | (unsigned long long)parent_transid, | |
282 | (unsigned long long)btrfs_header_generation(eb)); | |
283 | } | |
1259ab75 | 284 | ret = 1; |
1259ab75 | 285 | clear_extent_buffer_uptodate(io_tree, eb); |
33958dc6 | 286 | out: |
1259ab75 CM |
287 | unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, |
288 | GFP_NOFS); | |
289 | return ret; | |
1259ab75 CM |
290 | } |
291 | ||
d352ac68 CM |
292 | /* |
293 | * helper to read a given tree block, doing retries as required when | |
294 | * the checksums don't match and we have alternate mirrors to try. | |
295 | */ | |
f188591e CM |
296 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
297 | struct extent_buffer *eb, | |
ca7a79ad | 298 | u64 start, u64 parent_transid) |
f188591e CM |
299 | { |
300 | struct extent_io_tree *io_tree; | |
301 | int ret; | |
302 | int num_copies = 0; | |
303 | int mirror_num = 0; | |
304 | ||
305 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
306 | while (1) { | |
307 | ret = read_extent_buffer_pages(io_tree, eb, start, 1, | |
308 | btree_get_extent, mirror_num); | |
1259ab75 CM |
309 | if (!ret && |
310 | !verify_parent_transid(io_tree, eb, parent_transid)) | |
f188591e | 311 | return ret; |
d397712b | 312 | |
f188591e CM |
313 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
314 | eb->start, eb->len); | |
4235298e | 315 | if (num_copies == 1) |
f188591e | 316 | return ret; |
4235298e | 317 | |
f188591e | 318 | mirror_num++; |
4235298e | 319 | if (mirror_num > num_copies) |
f188591e | 320 | return ret; |
f188591e | 321 | } |
f188591e CM |
322 | return -EIO; |
323 | } | |
19c00ddc | 324 | |
d352ac68 | 325 | /* |
d397712b CM |
326 | * checksum a dirty tree block before IO. This has extra checks to make sure |
327 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 328 | */ |
d397712b | 329 | |
b2950863 | 330 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 331 | { |
d1310b2e | 332 | struct extent_io_tree *tree; |
35ebb934 | 333 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc CM |
334 | u64 found_start; |
335 | int found_level; | |
336 | unsigned long len; | |
337 | struct extent_buffer *eb; | |
f188591e CM |
338 | int ret; |
339 | ||
d1310b2e | 340 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc CM |
341 | |
342 | if (page->private == EXTENT_PAGE_PRIVATE) | |
343 | goto out; | |
344 | if (!page->private) | |
345 | goto out; | |
346 | len = page->private >> 2; | |
d397712b CM |
347 | WARN_ON(len == 0); |
348 | ||
19c00ddc | 349 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); |
ca7a79ad CM |
350 | ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, |
351 | btrfs_header_generation(eb)); | |
f188591e | 352 | BUG_ON(ret); |
19c00ddc CM |
353 | found_start = btrfs_header_bytenr(eb); |
354 | if (found_start != start) { | |
55c69072 CM |
355 | WARN_ON(1); |
356 | goto err; | |
357 | } | |
358 | if (eb->first_page != page) { | |
55c69072 CM |
359 | WARN_ON(1); |
360 | goto err; | |
361 | } | |
362 | if (!PageUptodate(page)) { | |
55c69072 CM |
363 | WARN_ON(1); |
364 | goto err; | |
19c00ddc CM |
365 | } |
366 | found_level = btrfs_header_level(eb); | |
4bef0848 | 367 | |
19c00ddc | 368 | csum_tree_block(root, eb, 0); |
55c69072 | 369 | err: |
19c00ddc CM |
370 | free_extent_buffer(eb); |
371 | out: | |
372 | return 0; | |
373 | } | |
374 | ||
2b82032c YZ |
375 | static int check_tree_block_fsid(struct btrfs_root *root, |
376 | struct extent_buffer *eb) | |
377 | { | |
378 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
379 | u8 fsid[BTRFS_UUID_SIZE]; | |
380 | int ret = 1; | |
381 | ||
382 | read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), | |
383 | BTRFS_FSID_SIZE); | |
384 | while (fs_devices) { | |
385 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
386 | ret = 0; | |
387 | break; | |
388 | } | |
389 | fs_devices = fs_devices->seed; | |
390 | } | |
391 | return ret; | |
392 | } | |
393 | ||
4008c04a CM |
394 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
395 | void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, int level) | |
396 | { | |
397 | lockdep_set_class_and_name(&eb->lock, | |
398 | &btrfs_eb_class[level], | |
399 | btrfs_eb_name[level]); | |
400 | } | |
401 | #endif | |
402 | ||
b2950863 | 403 | static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
ce9adaa5 CM |
404 | struct extent_state *state) |
405 | { | |
406 | struct extent_io_tree *tree; | |
407 | u64 found_start; | |
408 | int found_level; | |
409 | unsigned long len; | |
410 | struct extent_buffer *eb; | |
411 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 412 | int ret = 0; |
ce9adaa5 CM |
413 | |
414 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
415 | if (page->private == EXTENT_PAGE_PRIVATE) | |
416 | goto out; | |
417 | if (!page->private) | |
418 | goto out; | |
d397712b | 419 | |
ce9adaa5 | 420 | len = page->private >> 2; |
d397712b CM |
421 | WARN_ON(len == 0); |
422 | ||
ce9adaa5 | 423 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); |
f188591e | 424 | |
ce9adaa5 | 425 | found_start = btrfs_header_bytenr(eb); |
23a07867 | 426 | if (found_start != start) { |
193f284d CM |
427 | if (printk_ratelimit()) { |
428 | printk(KERN_INFO "btrfs bad tree block start " | |
429 | "%llu %llu\n", | |
430 | (unsigned long long)found_start, | |
431 | (unsigned long long)eb->start); | |
432 | } | |
f188591e | 433 | ret = -EIO; |
ce9adaa5 CM |
434 | goto err; |
435 | } | |
436 | if (eb->first_page != page) { | |
d397712b CM |
437 | printk(KERN_INFO "btrfs bad first page %lu %lu\n", |
438 | eb->first_page->index, page->index); | |
ce9adaa5 | 439 | WARN_ON(1); |
f188591e | 440 | ret = -EIO; |
ce9adaa5 CM |
441 | goto err; |
442 | } | |
2b82032c | 443 | if (check_tree_block_fsid(root, eb)) { |
193f284d CM |
444 | if (printk_ratelimit()) { |
445 | printk(KERN_INFO "btrfs bad fsid on block %llu\n", | |
446 | (unsigned long long)eb->start); | |
447 | } | |
1259ab75 CM |
448 | ret = -EIO; |
449 | goto err; | |
450 | } | |
ce9adaa5 CM |
451 | found_level = btrfs_header_level(eb); |
452 | ||
4008c04a CM |
453 | btrfs_set_buffer_lockdep_class(eb, found_level); |
454 | ||
ce9adaa5 | 455 | ret = csum_tree_block(root, eb, 1); |
f188591e CM |
456 | if (ret) |
457 | ret = -EIO; | |
ce9adaa5 CM |
458 | |
459 | end = min_t(u64, eb->len, PAGE_CACHE_SIZE); | |
460 | end = eb->start + end - 1; | |
ce9adaa5 CM |
461 | err: |
462 | free_extent_buffer(eb); | |
463 | out: | |
f188591e | 464 | return ret; |
ce9adaa5 CM |
465 | } |
466 | ||
ce9adaa5 | 467 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
468 | { |
469 | struct end_io_wq *end_io_wq = bio->bi_private; | |
470 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 471 | |
ce9adaa5 | 472 | fs_info = end_io_wq->info; |
ce9adaa5 | 473 | end_io_wq->error = err; |
8b712842 CM |
474 | end_io_wq->work.func = end_workqueue_fn; |
475 | end_io_wq->work.flags = 0; | |
d20f7043 CM |
476 | |
477 | if (bio->bi_rw & (1 << BIO_RW)) { | |
cad321ad CM |
478 | if (end_io_wq->metadata) |
479 | btrfs_queue_worker(&fs_info->endio_meta_write_workers, | |
480 | &end_io_wq->work); | |
481 | else | |
482 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
483 | &end_io_wq->work); | |
d20f7043 CM |
484 | } else { |
485 | if (end_io_wq->metadata) | |
486 | btrfs_queue_worker(&fs_info->endio_meta_workers, | |
487 | &end_io_wq->work); | |
488 | else | |
489 | btrfs_queue_worker(&fs_info->endio_workers, | |
490 | &end_io_wq->work); | |
491 | } | |
ce9adaa5 CM |
492 | } |
493 | ||
22c59948 CM |
494 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
495 | int metadata) | |
0b86a832 | 496 | { |
ce9adaa5 | 497 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
498 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
499 | if (!end_io_wq) | |
500 | return -ENOMEM; | |
501 | ||
502 | end_io_wq->private = bio->bi_private; | |
503 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 504 | end_io_wq->info = info; |
ce9adaa5 CM |
505 | end_io_wq->error = 0; |
506 | end_io_wq->bio = bio; | |
22c59948 | 507 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
508 | |
509 | bio->bi_private = end_io_wq; | |
510 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
511 | return 0; |
512 | } | |
513 | ||
b64a2851 | 514 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 515 | { |
4854ddd0 CM |
516 | unsigned long limit = min_t(unsigned long, |
517 | info->workers.max_workers, | |
518 | info->fs_devices->open_devices); | |
519 | return 256 * limit; | |
520 | } | |
0986fe9e | 521 | |
4854ddd0 CM |
522 | int btrfs_congested_async(struct btrfs_fs_info *info, int iodone) |
523 | { | |
b64a2851 CM |
524 | return atomic_read(&info->nr_async_bios) > |
525 | btrfs_async_submit_limit(info); | |
0986fe9e CM |
526 | } |
527 | ||
4a69a410 CM |
528 | static void run_one_async_start(struct btrfs_work *work) |
529 | { | |
530 | struct btrfs_fs_info *fs_info; | |
531 | struct async_submit_bio *async; | |
532 | ||
533 | async = container_of(work, struct async_submit_bio, work); | |
534 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
535 | async->submit_bio_start(async->inode, async->rw, async->bio, | |
536 | async->mirror_num, async->bio_flags); | |
537 | } | |
538 | ||
539 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
540 | { |
541 | struct btrfs_fs_info *fs_info; | |
542 | struct async_submit_bio *async; | |
4854ddd0 | 543 | int limit; |
8b712842 CM |
544 | |
545 | async = container_of(work, struct async_submit_bio, work); | |
546 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 547 | |
b64a2851 | 548 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
549 | limit = limit * 2 / 3; |
550 | ||
8b712842 | 551 | atomic_dec(&fs_info->nr_async_submits); |
0986fe9e | 552 | |
b64a2851 CM |
553 | if (atomic_read(&fs_info->nr_async_submits) < limit && |
554 | waitqueue_active(&fs_info->async_submit_wait)) | |
4854ddd0 CM |
555 | wake_up(&fs_info->async_submit_wait); |
556 | ||
4a69a410 | 557 | async->submit_bio_done(async->inode, async->rw, async->bio, |
c8b97818 | 558 | async->mirror_num, async->bio_flags); |
4a69a410 CM |
559 | } |
560 | ||
561 | static void run_one_async_free(struct btrfs_work *work) | |
562 | { | |
563 | struct async_submit_bio *async; | |
564 | ||
565 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
566 | kfree(async); |
567 | } | |
568 | ||
44b8bd7e CM |
569 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
570 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 571 | unsigned long bio_flags, |
4a69a410 CM |
572 | extent_submit_bio_hook_t *submit_bio_start, |
573 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
574 | { |
575 | struct async_submit_bio *async; | |
576 | ||
577 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
578 | if (!async) | |
579 | return -ENOMEM; | |
580 | ||
581 | async->inode = inode; | |
582 | async->rw = rw; | |
583 | async->bio = bio; | |
584 | async->mirror_num = mirror_num; | |
4a69a410 CM |
585 | async->submit_bio_start = submit_bio_start; |
586 | async->submit_bio_done = submit_bio_done; | |
587 | ||
588 | async->work.func = run_one_async_start; | |
589 | async->work.ordered_func = run_one_async_done; | |
590 | async->work.ordered_free = run_one_async_free; | |
591 | ||
8b712842 | 592 | async->work.flags = 0; |
c8b97818 | 593 | async->bio_flags = bio_flags; |
8c8bee1d | 594 | |
cb03c743 | 595 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 CM |
596 | |
597 | if (rw & (1 << BIO_RW_SYNCIO)) | |
598 | btrfs_set_work_high_prio(&async->work); | |
599 | ||
8b712842 | 600 | btrfs_queue_worker(&fs_info->workers, &async->work); |
9473f16c | 601 | |
d397712b | 602 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
603 | atomic_read(&fs_info->nr_async_submits)) { |
604 | wait_event(fs_info->async_submit_wait, | |
605 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
606 | } | |
607 | ||
44b8bd7e CM |
608 | return 0; |
609 | } | |
610 | ||
ce3ed71a CM |
611 | static int btree_csum_one_bio(struct bio *bio) |
612 | { | |
613 | struct bio_vec *bvec = bio->bi_io_vec; | |
614 | int bio_index = 0; | |
615 | struct btrfs_root *root; | |
616 | ||
617 | WARN_ON(bio->bi_vcnt <= 0); | |
d397712b | 618 | while (bio_index < bio->bi_vcnt) { |
ce3ed71a CM |
619 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
620 | csum_dirty_buffer(root, bvec->bv_page); | |
621 | bio_index++; | |
622 | bvec++; | |
623 | } | |
624 | return 0; | |
625 | } | |
626 | ||
4a69a410 CM |
627 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
628 | struct bio *bio, int mirror_num, | |
629 | unsigned long bio_flags) | |
22c59948 | 630 | { |
8b712842 CM |
631 | /* |
632 | * when we're called for a write, we're already in the async | |
5443be45 | 633 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 634 | */ |
4a69a410 CM |
635 | btree_csum_one_bio(bio); |
636 | return 0; | |
637 | } | |
22c59948 | 638 | |
4a69a410 CM |
639 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
640 | int mirror_num, unsigned long bio_flags) | |
641 | { | |
8b712842 | 642 | /* |
4a69a410 CM |
643 | * when we're called for a write, we're already in the async |
644 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 645 | */ |
8b712842 | 646 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
647 | } |
648 | ||
44b8bd7e | 649 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
c8b97818 | 650 | int mirror_num, unsigned long bio_flags) |
44b8bd7e | 651 | { |
cad321ad CM |
652 | int ret; |
653 | ||
654 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, | |
655 | bio, 1); | |
656 | BUG_ON(ret); | |
657 | ||
44b8bd7e | 658 | if (!(rw & (1 << BIO_RW))) { |
4a69a410 CM |
659 | /* |
660 | * called for a read, do the setup so that checksum validation | |
661 | * can happen in the async kernel threads | |
662 | */ | |
4a69a410 | 663 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, |
6f3577bd | 664 | mirror_num, 0); |
44b8bd7e | 665 | } |
d313d7a3 | 666 | |
cad321ad CM |
667 | /* |
668 | * kthread helpers are used to submit writes so that checksumming | |
669 | * can happen in parallel across all CPUs | |
670 | */ | |
44b8bd7e | 671 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, |
c8b97818 | 672 | inode, rw, bio, mirror_num, 0, |
4a69a410 CM |
673 | __btree_submit_bio_start, |
674 | __btree_submit_bio_done); | |
44b8bd7e CM |
675 | } |
676 | ||
0da5468f CM |
677 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
678 | { | |
d1310b2e | 679 | struct extent_io_tree *tree; |
b9473439 CM |
680 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
681 | struct extent_buffer *eb; | |
682 | int was_dirty; | |
683 | ||
d1310b2e | 684 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
b9473439 CM |
685 | if (!(current->flags & PF_MEMALLOC)) { |
686 | return extent_write_full_page(tree, page, | |
687 | btree_get_extent, wbc); | |
688 | } | |
5443be45 | 689 | |
b9473439 CM |
690 | redirty_page_for_writepage(wbc, page); |
691 | eb = btrfs_find_tree_block(root, page_offset(page), | |
692 | PAGE_CACHE_SIZE); | |
693 | WARN_ON(!eb); | |
694 | ||
695 | was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags); | |
696 | if (!was_dirty) { | |
697 | spin_lock(&root->fs_info->delalloc_lock); | |
698 | root->fs_info->dirty_metadata_bytes += PAGE_CACHE_SIZE; | |
699 | spin_unlock(&root->fs_info->delalloc_lock); | |
5443be45 | 700 | } |
b9473439 CM |
701 | free_extent_buffer(eb); |
702 | ||
703 | unlock_page(page); | |
704 | return 0; | |
5f39d397 | 705 | } |
0da5468f CM |
706 | |
707 | static int btree_writepages(struct address_space *mapping, | |
708 | struct writeback_control *wbc) | |
709 | { | |
d1310b2e CM |
710 | struct extent_io_tree *tree; |
711 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 712 | if (wbc->sync_mode == WB_SYNC_NONE) { |
b9473439 | 713 | struct btrfs_root *root = BTRFS_I(mapping->host)->root; |
793955bc | 714 | u64 num_dirty; |
24ab9cd8 | 715 | unsigned long thresh = 32 * 1024 * 1024; |
448d640b CM |
716 | |
717 | if (wbc->for_kupdate) | |
718 | return 0; | |
719 | ||
b9473439 CM |
720 | /* this is a bit racy, but that's ok */ |
721 | num_dirty = root->fs_info->dirty_metadata_bytes; | |
d397712b | 722 | if (num_dirty < thresh) |
793955bc | 723 | return 0; |
793955bc | 724 | } |
0da5468f CM |
725 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
726 | } | |
727 | ||
b2950863 | 728 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 729 | { |
d1310b2e CM |
730 | struct extent_io_tree *tree; |
731 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
732 | return extent_read_full_page(tree, page, btree_get_extent); |
733 | } | |
22b0ebda | 734 | |
70dec807 | 735 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 736 | { |
d1310b2e CM |
737 | struct extent_io_tree *tree; |
738 | struct extent_map_tree *map; | |
5f39d397 | 739 | int ret; |
d98237b3 | 740 | |
98509cfc | 741 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 742 | return 0; |
98509cfc | 743 | |
d1310b2e CM |
744 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
745 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
6af118ce | 746 | |
7b13b7b1 | 747 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
d397712b | 748 | if (!ret) |
6af118ce | 749 | return 0; |
6af118ce CM |
750 | |
751 | ret = try_release_extent_buffer(tree, page); | |
5f39d397 CM |
752 | if (ret == 1) { |
753 | ClearPagePrivate(page); | |
754 | set_page_private(page, 0); | |
755 | page_cache_release(page); | |
756 | } | |
6af118ce | 757 | |
d98237b3 CM |
758 | return ret; |
759 | } | |
760 | ||
5f39d397 | 761 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 762 | { |
d1310b2e CM |
763 | struct extent_io_tree *tree; |
764 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
765 | extent_invalidatepage(tree, page, offset); |
766 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 767 | if (PagePrivate(page)) { |
d397712b CM |
768 | printk(KERN_WARNING "btrfs warning page private not zero " |
769 | "on page %llu\n", (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
770 | ClearPagePrivate(page); |
771 | set_page_private(page, 0); | |
772 | page_cache_release(page); | |
773 | } | |
d98237b3 CM |
774 | } |
775 | ||
d98237b3 CM |
776 | static struct address_space_operations btree_aops = { |
777 | .readpage = btree_readpage, | |
778 | .writepage = btree_writepage, | |
0da5468f | 779 | .writepages = btree_writepages, |
5f39d397 CM |
780 | .releasepage = btree_releasepage, |
781 | .invalidatepage = btree_invalidatepage, | |
d98237b3 CM |
782 | .sync_page = block_sync_page, |
783 | }; | |
784 | ||
ca7a79ad CM |
785 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
786 | u64 parent_transid) | |
090d1875 | 787 | { |
5f39d397 CM |
788 | struct extent_buffer *buf = NULL; |
789 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 790 | int ret = 0; |
090d1875 | 791 | |
db94535d | 792 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 793 | if (!buf) |
090d1875 | 794 | return 0; |
d1310b2e | 795 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
f188591e | 796 | buf, 0, 0, btree_get_extent, 0); |
5f39d397 | 797 | free_extent_buffer(buf); |
de428b63 | 798 | return ret; |
090d1875 CM |
799 | } |
800 | ||
0999df54 CM |
801 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
802 | u64 bytenr, u32 blocksize) | |
803 | { | |
804 | struct inode *btree_inode = root->fs_info->btree_inode; | |
805 | struct extent_buffer *eb; | |
806 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
807 | bytenr, blocksize, GFP_NOFS); | |
808 | return eb; | |
809 | } | |
810 | ||
811 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
812 | u64 bytenr, u32 blocksize) | |
813 | { | |
814 | struct inode *btree_inode = root->fs_info->btree_inode; | |
815 | struct extent_buffer *eb; | |
816 | ||
817 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
818 | bytenr, blocksize, NULL, GFP_NOFS); | |
819 | return eb; | |
820 | } | |
821 | ||
822 | ||
e02119d5 CM |
823 | int btrfs_write_tree_block(struct extent_buffer *buf) |
824 | { | |
825 | return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start, | |
24ab9cd8 | 826 | buf->start + buf->len - 1, WB_SYNC_ALL); |
e02119d5 CM |
827 | } |
828 | ||
829 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
830 | { | |
831 | return btrfs_wait_on_page_writeback_range(buf->first_page->mapping, | |
d397712b | 832 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
833 | } |
834 | ||
0999df54 | 835 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 836 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
837 | { |
838 | struct extent_buffer *buf = NULL; | |
839 | struct inode *btree_inode = root->fs_info->btree_inode; | |
840 | struct extent_io_tree *io_tree; | |
841 | int ret; | |
842 | ||
843 | io_tree = &BTRFS_I(btree_inode)->io_tree; | |
844 | ||
845 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
846 | if (!buf) | |
847 | return NULL; | |
0999df54 | 848 | |
ca7a79ad | 849 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 | 850 | |
d397712b | 851 | if (ret == 0) |
b4ce94de | 852 | set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags); |
5f39d397 | 853 | return buf; |
ce9adaa5 | 854 | |
eb60ceac CM |
855 | } |
856 | ||
e089f05c | 857 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 858 | struct extent_buffer *buf) |
ed2ff2cb | 859 | { |
5f39d397 | 860 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 | 861 | if (btrfs_header_generation(buf) == |
925baedd | 862 | root->fs_info->running_transaction->transid) { |
b9447ef8 | 863 | btrfs_assert_tree_locked(buf); |
b4ce94de | 864 | |
b9473439 CM |
865 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
866 | spin_lock(&root->fs_info->delalloc_lock); | |
867 | if (root->fs_info->dirty_metadata_bytes >= buf->len) | |
868 | root->fs_info->dirty_metadata_bytes -= buf->len; | |
869 | else | |
870 | WARN_ON(1); | |
871 | spin_unlock(&root->fs_info->delalloc_lock); | |
872 | } | |
b4ce94de | 873 | |
b9473439 CM |
874 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
875 | btrfs_set_lock_blocking(buf); | |
d1310b2e | 876 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 877 | buf); |
925baedd | 878 | } |
5f39d397 CM |
879 | return 0; |
880 | } | |
881 | ||
db94535d | 882 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 883 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 884 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 885 | u64 objectid) |
d97e63b6 | 886 | { |
cfaa7295 | 887 | root->node = NULL; |
a28ec197 | 888 | root->commit_root = NULL; |
db94535d CM |
889 | root->sectorsize = sectorsize; |
890 | root->nodesize = nodesize; | |
891 | root->leafsize = leafsize; | |
87ee04eb | 892 | root->stripesize = stripesize; |
123abc88 | 893 | root->ref_cows = 0; |
0b86a832 CM |
894 | root->track_dirty = 0; |
895 | ||
9f5fae2f | 896 | root->fs_info = fs_info; |
0f7d52f4 CM |
897 | root->objectid = objectid; |
898 | root->last_trans = 0; | |
13a8a7c8 | 899 | root->highest_objectid = 0; |
58176a96 | 900 | root->name = NULL; |
4313b399 | 901 | root->in_sysfs = 0; |
5d4f98a2 | 902 | root->inode_tree.rb_node = NULL; |
0b86a832 CM |
903 | |
904 | INIT_LIST_HEAD(&root->dirty_list); | |
7b128766 | 905 | INIT_LIST_HEAD(&root->orphan_list); |
5d4f98a2 | 906 | INIT_LIST_HEAD(&root->root_list); |
925baedd | 907 | spin_lock_init(&root->node_lock); |
bcc63abb | 908 | spin_lock_init(&root->list_lock); |
5d4f98a2 | 909 | spin_lock_init(&root->inode_lock); |
a2135011 | 910 | mutex_init(&root->objectid_mutex); |
e02119d5 | 911 | mutex_init(&root->log_mutex); |
7237f183 YZ |
912 | init_waitqueue_head(&root->log_writer_wait); |
913 | init_waitqueue_head(&root->log_commit_wait[0]); | |
914 | init_waitqueue_head(&root->log_commit_wait[1]); | |
915 | atomic_set(&root->log_commit[0], 0); | |
916 | atomic_set(&root->log_commit[1], 0); | |
917 | atomic_set(&root->log_writers, 0); | |
918 | root->log_batch = 0; | |
919 | root->log_transid = 0; | |
d0c803c4 CM |
920 | extent_io_tree_init(&root->dirty_log_pages, |
921 | fs_info->btree_inode->i_mapping, GFP_NOFS); | |
017e5369 | 922 | |
3768f368 CM |
923 | memset(&root->root_key, 0, sizeof(root->root_key)); |
924 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 925 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 926 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 927 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 928 | init_completion(&root->kobj_unregister); |
6702ed49 CM |
929 | root->defrag_running = 0; |
930 | root->defrag_level = 0; | |
4d775673 | 931 | root->root_key.objectid = objectid; |
3394e160 CM |
932 | root->anon_super.s_root = NULL; |
933 | root->anon_super.s_dev = 0; | |
934 | INIT_LIST_HEAD(&root->anon_super.s_list); | |
935 | INIT_LIST_HEAD(&root->anon_super.s_instances); | |
936 | init_rwsem(&root->anon_super.s_umount); | |
937 | ||
3768f368 CM |
938 | return 0; |
939 | } | |
940 | ||
db94535d | 941 | static int find_and_setup_root(struct btrfs_root *tree_root, |
9f5fae2f CM |
942 | struct btrfs_fs_info *fs_info, |
943 | u64 objectid, | |
e20d96d6 | 944 | struct btrfs_root *root) |
3768f368 CM |
945 | { |
946 | int ret; | |
db94535d | 947 | u32 blocksize; |
84234f3a | 948 | u64 generation; |
3768f368 | 949 | |
db94535d | 950 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
951 | tree_root->sectorsize, tree_root->stripesize, |
952 | root, fs_info, objectid); | |
3768f368 CM |
953 | ret = btrfs_find_last_root(tree_root, objectid, |
954 | &root->root_item, &root->root_key); | |
4df27c4d YZ |
955 | if (ret > 0) |
956 | return -ENOENT; | |
3768f368 CM |
957 | BUG_ON(ret); |
958 | ||
84234f3a | 959 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
960 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
961 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 962 | blocksize, generation); |
3768f368 | 963 | BUG_ON(!root->node); |
4df27c4d | 964 | root->commit_root = btrfs_root_node(root); |
d97e63b6 CM |
965 | return 0; |
966 | } | |
967 | ||
e02119d5 CM |
968 | int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans, |
969 | struct btrfs_fs_info *fs_info) | |
970 | { | |
971 | struct extent_buffer *eb; | |
d0c803c4 CM |
972 | struct btrfs_root *log_root_tree = fs_info->log_root_tree; |
973 | u64 start = 0; | |
974 | u64 end = 0; | |
e02119d5 CM |
975 | int ret; |
976 | ||
d0c803c4 | 977 | if (!log_root_tree) |
e02119d5 CM |
978 | return 0; |
979 | ||
d397712b | 980 | while (1) { |
d0c803c4 CM |
981 | ret = find_first_extent_bit(&log_root_tree->dirty_log_pages, |
982 | 0, &start, &end, EXTENT_DIRTY); | |
983 | if (ret) | |
984 | break; | |
985 | ||
986 | clear_extent_dirty(&log_root_tree->dirty_log_pages, | |
987 | start, end, GFP_NOFS); | |
988 | } | |
e02119d5 CM |
989 | eb = fs_info->log_root_tree->node; |
990 | ||
991 | WARN_ON(btrfs_header_level(eb) != 0); | |
992 | WARN_ON(btrfs_header_nritems(eb) != 0); | |
993 | ||
d00aff00 CM |
994 | ret = btrfs_free_reserved_extent(fs_info->tree_root, |
995 | eb->start, eb->len); | |
e02119d5 CM |
996 | BUG_ON(ret); |
997 | ||
998 | free_extent_buffer(eb); | |
999 | kfree(fs_info->log_root_tree); | |
1000 | fs_info->log_root_tree = NULL; | |
1001 | return 0; | |
1002 | } | |
1003 | ||
7237f183 YZ |
1004 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1005 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1006 | { |
1007 | struct btrfs_root *root; | |
1008 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1009 | struct extent_buffer *leaf; |
e02119d5 CM |
1010 | |
1011 | root = kzalloc(sizeof(*root), GFP_NOFS); | |
1012 | if (!root) | |
7237f183 | 1013 | return ERR_PTR(-ENOMEM); |
e02119d5 CM |
1014 | |
1015 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1016 | tree_root->sectorsize, tree_root->stripesize, | |
1017 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1018 | ||
1019 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1020 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1021 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
7237f183 YZ |
1022 | /* |
1023 | * log trees do not get reference counted because they go away | |
1024 | * before a real commit is actually done. They do store pointers | |
1025 | * to file data extents, and those reference counts still get | |
1026 | * updated (along with back refs to the log tree). | |
1027 | */ | |
e02119d5 CM |
1028 | root->ref_cows = 0; |
1029 | ||
5d4f98a2 YZ |
1030 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, |
1031 | BTRFS_TREE_LOG_OBJECTID, NULL, 0, 0, 0); | |
7237f183 YZ |
1032 | if (IS_ERR(leaf)) { |
1033 | kfree(root); | |
1034 | return ERR_CAST(leaf); | |
1035 | } | |
e02119d5 | 1036 | |
5d4f98a2 YZ |
1037 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1038 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1039 | btrfs_set_header_generation(leaf, trans->transid); | |
1040 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1041 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1042 | root->node = leaf; |
e02119d5 CM |
1043 | |
1044 | write_extent_buffer(root->node, root->fs_info->fsid, | |
1045 | (unsigned long)btrfs_header_fsid(root->node), | |
1046 | BTRFS_FSID_SIZE); | |
1047 | btrfs_mark_buffer_dirty(root->node); | |
1048 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1049 | return root; |
1050 | } | |
1051 | ||
1052 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1053 | struct btrfs_fs_info *fs_info) | |
1054 | { | |
1055 | struct btrfs_root *log_root; | |
1056 | ||
1057 | log_root = alloc_log_tree(trans, fs_info); | |
1058 | if (IS_ERR(log_root)) | |
1059 | return PTR_ERR(log_root); | |
1060 | WARN_ON(fs_info->log_root_tree); | |
1061 | fs_info->log_root_tree = log_root; | |
1062 | return 0; | |
1063 | } | |
1064 | ||
1065 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1066 | struct btrfs_root *root) | |
1067 | { | |
1068 | struct btrfs_root *log_root; | |
1069 | struct btrfs_inode_item *inode_item; | |
1070 | ||
1071 | log_root = alloc_log_tree(trans, root->fs_info); | |
1072 | if (IS_ERR(log_root)) | |
1073 | return PTR_ERR(log_root); | |
1074 | ||
1075 | log_root->last_trans = trans->transid; | |
1076 | log_root->root_key.offset = root->root_key.objectid; | |
1077 | ||
1078 | inode_item = &log_root->root_item.inode; | |
1079 | inode_item->generation = cpu_to_le64(1); | |
1080 | inode_item->size = cpu_to_le64(3); | |
1081 | inode_item->nlink = cpu_to_le32(1); | |
1082 | inode_item->nbytes = cpu_to_le64(root->leafsize); | |
1083 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); | |
1084 | ||
5d4f98a2 | 1085 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1086 | |
1087 | WARN_ON(root->log_root); | |
1088 | root->log_root = log_root; | |
1089 | root->log_transid = 0; | |
e02119d5 CM |
1090 | return 0; |
1091 | } | |
1092 | ||
1093 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
1094 | struct btrfs_key *location) | |
1095 | { | |
1096 | struct btrfs_root *root; | |
1097 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1098 | struct btrfs_path *path; |
5f39d397 | 1099 | struct extent_buffer *l; |
84234f3a | 1100 | u64 generation; |
db94535d | 1101 | u32 blocksize; |
0f7d52f4 CM |
1102 | int ret = 0; |
1103 | ||
5eda7b5e | 1104 | root = kzalloc(sizeof(*root), GFP_NOFS); |
0cf6c620 | 1105 | if (!root) |
0f7d52f4 | 1106 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 1107 | if (location->offset == (u64)-1) { |
db94535d | 1108 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
1109 | location->objectid, root); |
1110 | if (ret) { | |
0f7d52f4 CM |
1111 | kfree(root); |
1112 | return ERR_PTR(ret); | |
1113 | } | |
13a8a7c8 | 1114 | goto out; |
0f7d52f4 CM |
1115 | } |
1116 | ||
db94535d | 1117 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1118 | tree_root->sectorsize, tree_root->stripesize, |
1119 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
1120 | |
1121 | path = btrfs_alloc_path(); | |
1122 | BUG_ON(!path); | |
1123 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); | |
13a8a7c8 YZ |
1124 | if (ret == 0) { |
1125 | l = path->nodes[0]; | |
1126 | read_extent_buffer(l, &root->root_item, | |
1127 | btrfs_item_ptr_offset(l, path->slots[0]), | |
1128 | sizeof(root->root_item)); | |
1129 | memcpy(&root->root_key, location, sizeof(*location)); | |
0f7d52f4 | 1130 | } |
0f7d52f4 CM |
1131 | btrfs_free_path(path); |
1132 | if (ret) { | |
13a8a7c8 YZ |
1133 | if (ret > 0) |
1134 | ret = -ENOENT; | |
0f7d52f4 CM |
1135 | return ERR_PTR(ret); |
1136 | } | |
13a8a7c8 | 1137 | |
84234f3a | 1138 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1139 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1140 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1141 | blocksize, generation); |
5d4f98a2 | 1142 | root->commit_root = btrfs_root_node(root); |
0f7d52f4 | 1143 | BUG_ON(!root->node); |
13a8a7c8 YZ |
1144 | out: |
1145 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) | |
e02119d5 | 1146 | root->ref_cows = 1; |
13a8a7c8 | 1147 | |
5eda7b5e CM |
1148 | return root; |
1149 | } | |
1150 | ||
dc17ff8f CM |
1151 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1152 | u64 root_objectid) | |
1153 | { | |
1154 | struct btrfs_root *root; | |
1155 | ||
1156 | if (root_objectid == BTRFS_ROOT_TREE_OBJECTID) | |
1157 | return fs_info->tree_root; | |
1158 | if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1159 | return fs_info->extent_root; | |
1160 | ||
1161 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1162 | (unsigned long)root_objectid); | |
1163 | return root; | |
1164 | } | |
1165 | ||
edbd8d4e CM |
1166 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
1167 | struct btrfs_key *location) | |
5eda7b5e CM |
1168 | { |
1169 | struct btrfs_root *root; | |
1170 | int ret; | |
1171 | ||
edbd8d4e CM |
1172 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1173 | return fs_info->tree_root; | |
1174 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1175 | return fs_info->extent_root; | |
8f18cf13 CM |
1176 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1177 | return fs_info->chunk_root; | |
1178 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1179 | return fs_info->dev_root; | |
0403e47e YZ |
1180 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1181 | return fs_info->csum_root; | |
4df27c4d YZ |
1182 | again: |
1183 | spin_lock(&fs_info->fs_roots_radix_lock); | |
5eda7b5e CM |
1184 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
1185 | (unsigned long)location->objectid); | |
4df27c4d | 1186 | spin_unlock(&fs_info->fs_roots_radix_lock); |
5eda7b5e CM |
1187 | if (root) |
1188 | return root; | |
1189 | ||
4df27c4d YZ |
1190 | ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid); |
1191 | if (ret == 0) | |
1192 | ret = -ENOENT; | |
1193 | if (ret < 0) | |
1194 | return ERR_PTR(ret); | |
1195 | ||
e02119d5 | 1196 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
1197 | if (IS_ERR(root)) |
1198 | return root; | |
3394e160 | 1199 | |
4df27c4d | 1200 | WARN_ON(btrfs_root_refs(&root->root_item) == 0); |
3394e160 CM |
1201 | set_anon_super(&root->anon_super, NULL); |
1202 | ||
4df27c4d YZ |
1203 | ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); |
1204 | if (ret) | |
1205 | goto fail; | |
1206 | ||
1207 | spin_lock(&fs_info->fs_roots_radix_lock); | |
2619ba1f CM |
1208 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
1209 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 | 1210 | root); |
4df27c4d YZ |
1211 | if (ret == 0) |
1212 | root->in_radix = 1; | |
1213 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1214 | radix_tree_preload_end(); | |
0f7d52f4 | 1215 | if (ret) { |
4df27c4d YZ |
1216 | if (ret == -EEXIST) { |
1217 | free_fs_root(root); | |
1218 | goto again; | |
1219 | } | |
1220 | goto fail; | |
0f7d52f4 | 1221 | } |
4df27c4d YZ |
1222 | |
1223 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
1224 | root->root_key.objectid); | |
1225 | WARN_ON(ret); | |
1226 | ||
1227 | if (!(fs_info->sb->s_flags & MS_RDONLY)) | |
c146afad | 1228 | btrfs_orphan_cleanup(root); |
4df27c4d | 1229 | |
edbd8d4e | 1230 | return root; |
4df27c4d YZ |
1231 | fail: |
1232 | free_fs_root(root); | |
1233 | return ERR_PTR(ret); | |
edbd8d4e CM |
1234 | } |
1235 | ||
1236 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, | |
1237 | struct btrfs_key *location, | |
1238 | const char *name, int namelen) | |
1239 | { | |
4df27c4d YZ |
1240 | return btrfs_read_fs_root_no_name(fs_info, location); |
1241 | #if 0 | |
edbd8d4e CM |
1242 | struct btrfs_root *root; |
1243 | int ret; | |
1244 | ||
1245 | root = btrfs_read_fs_root_no_name(fs_info, location); | |
1246 | if (!root) | |
1247 | return NULL; | |
58176a96 | 1248 | |
4313b399 CM |
1249 | if (root->in_sysfs) |
1250 | return root; | |
1251 | ||
58176a96 JB |
1252 | ret = btrfs_set_root_name(root, name, namelen); |
1253 | if (ret) { | |
5f39d397 | 1254 | free_extent_buffer(root->node); |
58176a96 JB |
1255 | kfree(root); |
1256 | return ERR_PTR(ret); | |
1257 | } | |
4df27c4d | 1258 | |
58176a96 JB |
1259 | ret = btrfs_sysfs_add_root(root); |
1260 | if (ret) { | |
5f39d397 | 1261 | free_extent_buffer(root->node); |
58176a96 JB |
1262 | kfree(root->name); |
1263 | kfree(root); | |
1264 | return ERR_PTR(ret); | |
1265 | } | |
4313b399 | 1266 | root->in_sysfs = 1; |
0f7d52f4 | 1267 | return root; |
4df27c4d | 1268 | #endif |
0f7d52f4 | 1269 | } |
04160088 CM |
1270 | |
1271 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) | |
1272 | { | |
1273 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1274 | int ret = 0; | |
04160088 CM |
1275 | struct btrfs_device *device; |
1276 | struct backing_dev_info *bdi; | |
b7967db7 | 1277 | |
c6e30871 | 1278 | list_for_each_entry(device, &info->fs_devices->devices, dev_list) { |
dfe25020 CM |
1279 | if (!device->bdev) |
1280 | continue; | |
04160088 CM |
1281 | bdi = blk_get_backing_dev_info(device->bdev); |
1282 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1283 | ret = 1; | |
1284 | break; | |
1285 | } | |
1286 | } | |
1287 | return ret; | |
1288 | } | |
1289 | ||
38b66988 CM |
1290 | /* |
1291 | * this unplugs every device on the box, and it is only used when page | |
1292 | * is null | |
1293 | */ | |
1294 | static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) | |
1295 | { | |
38b66988 CM |
1296 | struct btrfs_device *device; |
1297 | struct btrfs_fs_info *info; | |
1298 | ||
1299 | info = (struct btrfs_fs_info *)bdi->unplug_io_data; | |
c6e30871 | 1300 | list_for_each_entry(device, &info->fs_devices->devices, dev_list) { |
d20f7043 CM |
1301 | if (!device->bdev) |
1302 | continue; | |
1303 | ||
38b66988 | 1304 | bdi = blk_get_backing_dev_info(device->bdev); |
d397712b | 1305 | if (bdi->unplug_io_fn) |
38b66988 | 1306 | bdi->unplug_io_fn(bdi, page); |
38b66988 CM |
1307 | } |
1308 | } | |
1309 | ||
b2950863 | 1310 | static void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) |
04160088 | 1311 | { |
38b66988 | 1312 | struct inode *inode; |
f2d8d74d CM |
1313 | struct extent_map_tree *em_tree; |
1314 | struct extent_map *em; | |
bcbfce8a | 1315 | struct address_space *mapping; |
38b66988 CM |
1316 | u64 offset; |
1317 | ||
bcbfce8a | 1318 | /* the generic O_DIRECT read code does this */ |
9f0ba5bd | 1319 | if (1 || !page) { |
38b66988 CM |
1320 | __unplug_io_fn(bdi, page); |
1321 | return; | |
1322 | } | |
1323 | ||
bcbfce8a CM |
1324 | /* |
1325 | * page->mapping may change at any time. Get a consistent copy | |
1326 | * and use that for everything below | |
1327 | */ | |
1328 | smp_mb(); | |
1329 | mapping = page->mapping; | |
1330 | if (!mapping) | |
1331 | return; | |
1332 | ||
1333 | inode = mapping->host; | |
240d5d48 CM |
1334 | |
1335 | /* | |
1336 | * don't do the expensive searching for a small number of | |
1337 | * devices | |
1338 | */ | |
1339 | if (BTRFS_I(inode)->root->fs_info->fs_devices->open_devices <= 2) { | |
1340 | __unplug_io_fn(bdi, page); | |
1341 | return; | |
1342 | } | |
1343 | ||
38b66988 | 1344 | offset = page_offset(page); |
04160088 | 1345 | |
f2d8d74d | 1346 | em_tree = &BTRFS_I(inode)->extent_tree; |
890871be | 1347 | read_lock(&em_tree->lock); |
f2d8d74d | 1348 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); |
890871be | 1349 | read_unlock(&em_tree->lock); |
89642229 CM |
1350 | if (!em) { |
1351 | __unplug_io_fn(bdi, page); | |
f2d8d74d | 1352 | return; |
89642229 | 1353 | } |
f2d8d74d | 1354 | |
89642229 CM |
1355 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { |
1356 | free_extent_map(em); | |
1357 | __unplug_io_fn(bdi, page); | |
1358 | return; | |
1359 | } | |
f2d8d74d CM |
1360 | offset = offset - em->start; |
1361 | btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree, | |
1362 | em->block_start + offset, page); | |
1363 | free_extent_map(em); | |
04160088 CM |
1364 | } |
1365 | ||
ad081f14 JA |
1366 | /* |
1367 | * If this fails, caller must call bdi_destroy() to get rid of the | |
1368 | * bdi again. | |
1369 | */ | |
04160088 CM |
1370 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1371 | { | |
ad081f14 JA |
1372 | int err; |
1373 | ||
1374 | bdi->capabilities = BDI_CAP_MAP_COPY; | |
1375 | err = bdi_init(bdi); | |
1376 | if (err) | |
1377 | return err; | |
1378 | ||
1379 | err = bdi_register(bdi, NULL, "btrfs-%d", | |
1380 | atomic_inc_return(&btrfs_bdi_num)); | |
76dda93c YZ |
1381 | if (err) { |
1382 | bdi_destroy(bdi); | |
ad081f14 | 1383 | return err; |
76dda93c | 1384 | } |
ad081f14 | 1385 | |
4575c9cc | 1386 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1387 | bdi->unplug_io_fn = btrfs_unplug_io_fn; |
1388 | bdi->unplug_io_data = info; | |
1389 | bdi->congested_fn = btrfs_congested_fn; | |
1390 | bdi->congested_data = info; | |
1391 | return 0; | |
1392 | } | |
1393 | ||
ce9adaa5 CM |
1394 | static int bio_ready_for_csum(struct bio *bio) |
1395 | { | |
1396 | u64 length = 0; | |
1397 | u64 buf_len = 0; | |
1398 | u64 start = 0; | |
1399 | struct page *page; | |
1400 | struct extent_io_tree *io_tree = NULL; | |
1401 | struct btrfs_fs_info *info = NULL; | |
1402 | struct bio_vec *bvec; | |
1403 | int i; | |
1404 | int ret; | |
1405 | ||
1406 | bio_for_each_segment(bvec, bio, i) { | |
1407 | page = bvec->bv_page; | |
1408 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
1409 | length += bvec->bv_len; | |
1410 | continue; | |
1411 | } | |
1412 | if (!page->private) { | |
1413 | length += bvec->bv_len; | |
1414 | continue; | |
1415 | } | |
1416 | length = bvec->bv_len; | |
1417 | buf_len = page->private >> 2; | |
1418 | start = page_offset(page) + bvec->bv_offset; | |
1419 | io_tree = &BTRFS_I(page->mapping->host)->io_tree; | |
1420 | info = BTRFS_I(page->mapping->host)->root->fs_info; | |
1421 | } | |
1422 | /* are we fully contained in this bio? */ | |
1423 | if (buf_len <= length) | |
1424 | return 1; | |
1425 | ||
1426 | ret = extent_range_uptodate(io_tree, start + length, | |
1427 | start + buf_len - 1); | |
ce9adaa5 CM |
1428 | return ret; |
1429 | } | |
1430 | ||
8b712842 CM |
1431 | /* |
1432 | * called by the kthread helper functions to finally call the bio end_io | |
1433 | * functions. This is where read checksum verification actually happens | |
1434 | */ | |
1435 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1436 | { |
ce9adaa5 | 1437 | struct bio *bio; |
8b712842 CM |
1438 | struct end_io_wq *end_io_wq; |
1439 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1440 | int error; |
ce9adaa5 | 1441 | |
8b712842 CM |
1442 | end_io_wq = container_of(work, struct end_io_wq, work); |
1443 | bio = end_io_wq->bio; | |
1444 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1445 | |
cad321ad | 1446 | /* metadata bio reads are special because the whole tree block must |
8b712842 CM |
1447 | * be checksummed at once. This makes sure the entire block is in |
1448 | * ram and up to date before trying to verify things. For | |
1449 | * blocksize <= pagesize, it is basically a noop | |
1450 | */ | |
cad321ad CM |
1451 | if (!(bio->bi_rw & (1 << BIO_RW)) && end_io_wq->metadata && |
1452 | !bio_ready_for_csum(bio)) { | |
d20f7043 | 1453 | btrfs_queue_worker(&fs_info->endio_meta_workers, |
8b712842 CM |
1454 | &end_io_wq->work); |
1455 | return; | |
1456 | } | |
1457 | error = end_io_wq->error; | |
1458 | bio->bi_private = end_io_wq->private; | |
1459 | bio->bi_end_io = end_io_wq->end_io; | |
1460 | kfree(end_io_wq); | |
8b712842 | 1461 | bio_endio(bio, error); |
44b8bd7e CM |
1462 | } |
1463 | ||
a74a4b97 CM |
1464 | static int cleaner_kthread(void *arg) |
1465 | { | |
1466 | struct btrfs_root *root = arg; | |
1467 | ||
1468 | do { | |
1469 | smp_mb(); | |
1470 | if (root->fs_info->closing) | |
1471 | break; | |
1472 | ||
1473 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
76dda93c YZ |
1474 | |
1475 | if (!(root->fs_info->sb->s_flags & MS_RDONLY) && | |
1476 | mutex_trylock(&root->fs_info->cleaner_mutex)) { | |
1477 | btrfs_clean_old_snapshots(root); | |
1478 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
1479 | } | |
a74a4b97 CM |
1480 | |
1481 | if (freezing(current)) { | |
1482 | refrigerator(); | |
1483 | } else { | |
1484 | smp_mb(); | |
1485 | if (root->fs_info->closing) | |
1486 | break; | |
1487 | set_current_state(TASK_INTERRUPTIBLE); | |
1488 | schedule(); | |
1489 | __set_current_state(TASK_RUNNING); | |
1490 | } | |
1491 | } while (!kthread_should_stop()); | |
1492 | return 0; | |
1493 | } | |
1494 | ||
1495 | static int transaction_kthread(void *arg) | |
1496 | { | |
1497 | struct btrfs_root *root = arg; | |
1498 | struct btrfs_trans_handle *trans; | |
1499 | struct btrfs_transaction *cur; | |
1500 | unsigned long now; | |
1501 | unsigned long delay; | |
1502 | int ret; | |
1503 | ||
1504 | do { | |
1505 | smp_mb(); | |
1506 | if (root->fs_info->closing) | |
1507 | break; | |
1508 | ||
1509 | delay = HZ * 30; | |
1510 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1511 | mutex_lock(&root->fs_info->transaction_kthread_mutex); | |
1512 | ||
1513 | mutex_lock(&root->fs_info->trans_mutex); | |
1514 | cur = root->fs_info->running_transaction; | |
1515 | if (!cur) { | |
1516 | mutex_unlock(&root->fs_info->trans_mutex); | |
1517 | goto sleep; | |
1518 | } | |
31153d81 | 1519 | |
a74a4b97 CM |
1520 | now = get_seconds(); |
1521 | if (now < cur->start_time || now - cur->start_time < 30) { | |
1522 | mutex_unlock(&root->fs_info->trans_mutex); | |
1523 | delay = HZ * 5; | |
1524 | goto sleep; | |
1525 | } | |
1526 | mutex_unlock(&root->fs_info->trans_mutex); | |
a74a4b97 CM |
1527 | trans = btrfs_start_transaction(root, 1); |
1528 | ret = btrfs_commit_transaction(trans, root); | |
56bec294 | 1529 | |
a74a4b97 CM |
1530 | sleep: |
1531 | wake_up_process(root->fs_info->cleaner_kthread); | |
1532 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1533 | ||
1534 | if (freezing(current)) { | |
1535 | refrigerator(); | |
1536 | } else { | |
1537 | if (root->fs_info->closing) | |
1538 | break; | |
1539 | set_current_state(TASK_INTERRUPTIBLE); | |
1540 | schedule_timeout(delay); | |
1541 | __set_current_state(TASK_RUNNING); | |
1542 | } | |
1543 | } while (!kthread_should_stop()); | |
1544 | return 0; | |
1545 | } | |
1546 | ||
8a4b83cc | 1547 | struct btrfs_root *open_ctree(struct super_block *sb, |
dfe25020 CM |
1548 | struct btrfs_fs_devices *fs_devices, |
1549 | char *options) | |
2e635a27 | 1550 | { |
db94535d CM |
1551 | u32 sectorsize; |
1552 | u32 nodesize; | |
1553 | u32 leafsize; | |
1554 | u32 blocksize; | |
87ee04eb | 1555 | u32 stripesize; |
84234f3a | 1556 | u64 generation; |
f2b636e8 | 1557 | u64 features; |
3de4586c | 1558 | struct btrfs_key location; |
a061fc8d | 1559 | struct buffer_head *bh; |
e02119d5 | 1560 | struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1561 | GFP_NOFS); |
d20f7043 CM |
1562 | struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root), |
1563 | GFP_NOFS); | |
e02119d5 | 1564 | struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1565 | GFP_NOFS); |
8790d502 | 1566 | struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), |
e20d96d6 | 1567 | GFP_NOFS); |
e02119d5 | 1568 | struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1569 | GFP_NOFS); |
e02119d5 | 1570 | struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1571 | GFP_NOFS); |
e02119d5 CM |
1572 | struct btrfs_root *log_tree_root; |
1573 | ||
eb60ceac | 1574 | int ret; |
e58ca020 | 1575 | int err = -EINVAL; |
4543df7e | 1576 | |
2c90e5d6 | 1577 | struct btrfs_super_block *disk_super; |
8790d502 | 1578 | |
0463bb4e | 1579 | if (!extent_root || !tree_root || !fs_info || |
d20f7043 | 1580 | !chunk_root || !dev_root || !csum_root) { |
39279cc3 CM |
1581 | err = -ENOMEM; |
1582 | goto fail; | |
1583 | } | |
76dda93c YZ |
1584 | |
1585 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
1586 | if (ret) { | |
1587 | err = ret; | |
1588 | goto fail; | |
1589 | } | |
1590 | ||
1591 | ret = setup_bdi(fs_info, &fs_info->bdi); | |
1592 | if (ret) { | |
1593 | err = ret; | |
1594 | goto fail_srcu; | |
1595 | } | |
1596 | ||
1597 | fs_info->btree_inode = new_inode(sb); | |
1598 | if (!fs_info->btree_inode) { | |
1599 | err = -ENOMEM; | |
1600 | goto fail_bdi; | |
1601 | } | |
1602 | ||
1603 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); | |
8fd17795 | 1604 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1605 | INIT_LIST_HEAD(&fs_info->dead_roots); |
19c00ddc | 1606 | INIT_LIST_HEAD(&fs_info->hashers); |
ea8c2819 | 1607 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
5a3f23d5 | 1608 | INIT_LIST_HEAD(&fs_info->ordered_operations); |
11833d66 | 1609 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
1832a6d5 | 1610 | spin_lock_init(&fs_info->delalloc_lock); |
cee36a03 | 1611 | spin_lock_init(&fs_info->new_trans_lock); |
31153d81 | 1612 | spin_lock_init(&fs_info->ref_cache_lock); |
76dda93c | 1613 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
19c00ddc | 1614 | |
58176a96 | 1615 | init_completion(&fs_info->kobj_unregister); |
9f5fae2f CM |
1616 | fs_info->tree_root = tree_root; |
1617 | fs_info->extent_root = extent_root; | |
d20f7043 | 1618 | fs_info->csum_root = csum_root; |
0b86a832 CM |
1619 | fs_info->chunk_root = chunk_root; |
1620 | fs_info->dev_root = dev_root; | |
8a4b83cc | 1621 | fs_info->fs_devices = fs_devices; |
0b86a832 | 1622 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1623 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1624 | btrfs_mapping_init(&fs_info->mapping_tree); |
cb03c743 | 1625 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 1626 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 1627 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 1628 | atomic_set(&fs_info->nr_async_bios, 0); |
e20d96d6 | 1629 | fs_info->sb = sb; |
c59f8951 | 1630 | fs_info->max_extent = (u64)-1; |
6f568d35 | 1631 | fs_info->max_inline = 8192 * 1024; |
97e728d4 | 1632 | fs_info->metadata_ratio = 8; |
c8b97818 | 1633 | |
b34b086c CM |
1634 | fs_info->thread_pool_size = min_t(unsigned long, |
1635 | num_online_cpus() + 2, 8); | |
0afbaf8c | 1636 | |
3eaa2885 CM |
1637 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
1638 | spin_lock_init(&fs_info->ordered_extent_lock); | |
1639 | ||
a061fc8d CM |
1640 | sb->s_blocksize = 4096; |
1641 | sb->s_blocksize_bits = blksize_bits(4096); | |
1642 | ||
76dda93c YZ |
1643 | fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; |
1644 | fs_info->btree_inode->i_nlink = 1; | |
0afbaf8c CM |
1645 | /* |
1646 | * we set the i_size on the btree inode to the max possible int. | |
1647 | * the real end of the address space is determined by all of | |
1648 | * the devices in the system | |
1649 | */ | |
1650 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 1651 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
1652 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
1653 | ||
5d4f98a2 | 1654 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); |
d1310b2e | 1655 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
5f39d397 CM |
1656 | fs_info->btree_inode->i_mapping, |
1657 | GFP_NOFS); | |
d1310b2e CM |
1658 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, |
1659 | GFP_NOFS); | |
1660 | ||
1661 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 1662 | |
76dda93c YZ |
1663 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
1664 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
1665 | sizeof(struct btrfs_key)); | |
1666 | BTRFS_I(fs_info->btree_inode)->dummy_inode = 1; | |
1667 | ||
0f9dd46c JB |
1668 | spin_lock_init(&fs_info->block_group_cache_lock); |
1669 | fs_info->block_group_cache_tree.rb_node = NULL; | |
1670 | ||
11833d66 | 1671 | extent_io_tree_init(&fs_info->freed_extents[0], |
1a5bc167 | 1672 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
11833d66 YZ |
1673 | extent_io_tree_init(&fs_info->freed_extents[1], |
1674 | fs_info->btree_inode->i_mapping, GFP_NOFS); | |
1675 | fs_info->pinned_extents = &fs_info->freed_extents[0]; | |
e66f709b | 1676 | fs_info->do_barriers = 1; |
e18e4809 | 1677 | |
39279cc3 | 1678 | |
79154b1b | 1679 | mutex_init(&fs_info->trans_mutex); |
5a3f23d5 | 1680 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 1681 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 1682 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
1683 | mutex_init(&fs_info->transaction_kthread_mutex); |
1684 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 1685 | mutex_init(&fs_info->volume_mutex); |
276e680d | 1686 | init_rwsem(&fs_info->extent_commit_sem); |
76dda93c | 1687 | init_rwsem(&fs_info->subvol_sem); |
fa9c0d79 CM |
1688 | |
1689 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); | |
1690 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
1691 | ||
e6dcd2dc | 1692 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 1693 | init_waitqueue_head(&fs_info->transaction_wait); |
4854ddd0 | 1694 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 1695 | |
0b86a832 | 1696 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 1697 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 1698 | |
d98237b3 | 1699 | |
a512bbf8 | 1700 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
a061fc8d | 1701 | if (!bh) |
39279cc3 | 1702 | goto fail_iput; |
39279cc3 | 1703 | |
a061fc8d | 1704 | memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy)); |
2d69a0f8 YZ |
1705 | memcpy(&fs_info->super_for_commit, &fs_info->super_copy, |
1706 | sizeof(fs_info->super_for_commit)); | |
a061fc8d | 1707 | brelse(bh); |
5f39d397 | 1708 | |
a061fc8d | 1709 | memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE); |
0b86a832 | 1710 | |
5f39d397 | 1711 | disk_super = &fs_info->super_copy; |
0f7d52f4 | 1712 | if (!btrfs_super_root(disk_super)) |
c6e2bac1 | 1713 | goto fail_iput; |
0f7d52f4 | 1714 | |
2b82032c YZ |
1715 | ret = btrfs_parse_options(tree_root, options); |
1716 | if (ret) { | |
1717 | err = ret; | |
c6e2bac1 | 1718 | goto fail_iput; |
2b82032c | 1719 | } |
dfe25020 | 1720 | |
f2b636e8 JB |
1721 | features = btrfs_super_incompat_flags(disk_super) & |
1722 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
1723 | if (features) { | |
1724 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
1725 | "unsupported optional features (%Lx).\n", | |
21380931 | 1726 | (unsigned long long)features); |
f2b636e8 | 1727 | err = -EINVAL; |
c6e2bac1 | 1728 | goto fail_iput; |
f2b636e8 JB |
1729 | } |
1730 | ||
5d4f98a2 YZ |
1731 | features = btrfs_super_incompat_flags(disk_super); |
1732 | if (!(features & BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF)) { | |
1733 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; | |
1734 | btrfs_set_super_incompat_flags(disk_super, features); | |
1735 | } | |
1736 | ||
f2b636e8 JB |
1737 | features = btrfs_super_compat_ro_flags(disk_super) & |
1738 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
1739 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
1740 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
1741 | "unsupported option features (%Lx).\n", | |
21380931 | 1742 | (unsigned long long)features); |
f2b636e8 | 1743 | err = -EINVAL; |
c6e2bac1 | 1744 | goto fail_iput; |
f2b636e8 | 1745 | } |
9042846b | 1746 | printk("thread pool is %d\n", fs_info->thread_pool_size); |
4543df7e CM |
1747 | /* |
1748 | * we need to start all the end_io workers up front because the | |
1749 | * queue work function gets called at interrupt time, and so it | |
1750 | * cannot dynamically grow. | |
1751 | */ | |
5443be45 CM |
1752 | btrfs_init_workers(&fs_info->workers, "worker", |
1753 | fs_info->thread_pool_size); | |
c8b97818 | 1754 | |
771ed689 CM |
1755 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
1756 | fs_info->thread_pool_size); | |
1757 | ||
5443be45 | 1758 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 CM |
1759 | min_t(u64, fs_devices->num_devices, |
1760 | fs_info->thread_pool_size)); | |
61b49440 CM |
1761 | |
1762 | /* a higher idle thresh on the submit workers makes it much more | |
1763 | * likely that bios will be send down in a sane order to the | |
1764 | * devices | |
1765 | */ | |
1766 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 1767 | |
771ed689 | 1768 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 1769 | fs_info->workers.ordered = 1; |
61b49440 | 1770 | |
771ed689 CM |
1771 | fs_info->delalloc_workers.idle_thresh = 2; |
1772 | fs_info->delalloc_workers.ordered = 1; | |
1773 | ||
5443be45 CM |
1774 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1); |
1775 | btrfs_init_workers(&fs_info->endio_workers, "endio", | |
1776 | fs_info->thread_pool_size); | |
d20f7043 CM |
1777 | btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", |
1778 | fs_info->thread_pool_size); | |
cad321ad CM |
1779 | btrfs_init_workers(&fs_info->endio_meta_write_workers, |
1780 | "endio-meta-write", fs_info->thread_pool_size); | |
5443be45 | 1781 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", |
e6dcd2dc | 1782 | fs_info->thread_pool_size); |
61b49440 CM |
1783 | |
1784 | /* | |
1785 | * endios are largely parallel and should have a very | |
1786 | * low idle thresh | |
1787 | */ | |
1788 | fs_info->endio_workers.idle_thresh = 4; | |
b51912c9 CM |
1789 | fs_info->endio_meta_workers.idle_thresh = 4; |
1790 | ||
9042846b CM |
1791 | fs_info->endio_write_workers.idle_thresh = 2; |
1792 | fs_info->endio_meta_write_workers.idle_thresh = 2; | |
1793 | ||
1794 | fs_info->endio_workers.atomic_worker_start = 1; | |
1795 | fs_info->endio_meta_workers.atomic_worker_start = 1; | |
1796 | fs_info->endio_write_workers.atomic_worker_start = 1; | |
1797 | fs_info->endio_meta_write_workers.atomic_worker_start = 1; | |
61b49440 | 1798 | |
4543df7e | 1799 | btrfs_start_workers(&fs_info->workers, 1); |
1cc127b5 | 1800 | btrfs_start_workers(&fs_info->submit_workers, 1); |
771ed689 | 1801 | btrfs_start_workers(&fs_info->delalloc_workers, 1); |
247e743c | 1802 | btrfs_start_workers(&fs_info->fixup_workers, 1); |
9042846b CM |
1803 | btrfs_start_workers(&fs_info->endio_workers, 1); |
1804 | btrfs_start_workers(&fs_info->endio_meta_workers, 1); | |
1805 | btrfs_start_workers(&fs_info->endio_meta_write_workers, 1); | |
1806 | btrfs_start_workers(&fs_info->endio_write_workers, 1); | |
4543df7e | 1807 | |
4575c9cc | 1808 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
1809 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
1810 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 1811 | |
db94535d CM |
1812 | nodesize = btrfs_super_nodesize(disk_super); |
1813 | leafsize = btrfs_super_leafsize(disk_super); | |
1814 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 1815 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
1816 | tree_root->nodesize = nodesize; |
1817 | tree_root->leafsize = leafsize; | |
1818 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 1819 | tree_root->stripesize = stripesize; |
a061fc8d CM |
1820 | |
1821 | sb->s_blocksize = sectorsize; | |
1822 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 1823 | |
39279cc3 CM |
1824 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
1825 | sizeof(disk_super->magic))) { | |
d397712b | 1826 | printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
1827 | goto fail_sb_buffer; |
1828 | } | |
19c00ddc | 1829 | |
925baedd | 1830 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 1831 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 1832 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 1833 | if (ret) { |
d397712b CM |
1834 | printk(KERN_WARNING "btrfs: failed to read the system " |
1835 | "array on %s\n", sb->s_id); | |
5d4f98a2 | 1836 | goto fail_sb_buffer; |
84eed90f | 1837 | } |
0b86a832 CM |
1838 | |
1839 | blocksize = btrfs_level_size(tree_root, | |
1840 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 1841 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
1842 | |
1843 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1844 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
1845 | ||
1846 | chunk_root->node = read_tree_block(chunk_root, | |
1847 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 1848 | blocksize, generation); |
0b86a832 | 1849 | BUG_ON(!chunk_root->node); |
83121942 DW |
1850 | if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { |
1851 | printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n", | |
1852 | sb->s_id); | |
1853 | goto fail_chunk_root; | |
1854 | } | |
5d4f98a2 YZ |
1855 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
1856 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 1857 | |
e17cade2 | 1858 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
d397712b CM |
1859 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), |
1860 | BTRFS_UUID_SIZE); | |
e17cade2 | 1861 | |
925baedd | 1862 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1863 | ret = btrfs_read_chunk_tree(chunk_root); |
925baedd | 1864 | mutex_unlock(&fs_info->chunk_mutex); |
2b82032c | 1865 | if (ret) { |
d397712b CM |
1866 | printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", |
1867 | sb->s_id); | |
2b82032c YZ |
1868 | goto fail_chunk_root; |
1869 | } | |
0b86a832 | 1870 | |
dfe25020 CM |
1871 | btrfs_close_extra_devices(fs_devices); |
1872 | ||
db94535d CM |
1873 | blocksize = btrfs_level_size(tree_root, |
1874 | btrfs_super_root_level(disk_super)); | |
84234f3a | 1875 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 1876 | |
e20d96d6 | 1877 | tree_root->node = read_tree_block(tree_root, |
db94535d | 1878 | btrfs_super_root(disk_super), |
84234f3a | 1879 | blocksize, generation); |
39279cc3 | 1880 | if (!tree_root->node) |
2b82032c | 1881 | goto fail_chunk_root; |
83121942 DW |
1882 | if (!test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { |
1883 | printk(KERN_WARNING "btrfs: failed to read tree root on %s\n", | |
1884 | sb->s_id); | |
1885 | goto fail_tree_root; | |
1886 | } | |
5d4f98a2 YZ |
1887 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
1888 | tree_root->commit_root = btrfs_root_node(tree_root); | |
db94535d CM |
1889 | |
1890 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 1891 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 1892 | if (ret) |
39279cc3 | 1893 | goto fail_tree_root; |
0b86a832 CM |
1894 | extent_root->track_dirty = 1; |
1895 | ||
1896 | ret = find_and_setup_root(tree_root, fs_info, | |
1897 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
0b86a832 CM |
1898 | if (ret) |
1899 | goto fail_extent_root; | |
5d4f98a2 | 1900 | dev_root->track_dirty = 1; |
3768f368 | 1901 | |
d20f7043 CM |
1902 | ret = find_and_setup_root(tree_root, fs_info, |
1903 | BTRFS_CSUM_TREE_OBJECTID, csum_root); | |
1904 | if (ret) | |
5d4f98a2 | 1905 | goto fail_dev_root; |
d20f7043 CM |
1906 | |
1907 | csum_root->track_dirty = 1; | |
1908 | ||
9078a3e1 CM |
1909 | btrfs_read_block_groups(extent_root); |
1910 | ||
2d69a0f8 | 1911 | fs_info->generation = generation; |
c146afad | 1912 | fs_info->last_trans_committed = generation; |
d18a2c44 CM |
1913 | fs_info->data_alloc_profile = (u64)-1; |
1914 | fs_info->metadata_alloc_profile = (u64)-1; | |
1915 | fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; | |
a74a4b97 CM |
1916 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
1917 | "btrfs-cleaner"); | |
57506d50 | 1918 | if (IS_ERR(fs_info->cleaner_kthread)) |
d20f7043 | 1919 | goto fail_csum_root; |
a74a4b97 CM |
1920 | |
1921 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
1922 | tree_root, | |
1923 | "btrfs-transaction"); | |
57506d50 | 1924 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 1925 | goto fail_cleaner; |
a74a4b97 | 1926 | |
c289811c CM |
1927 | if (!btrfs_test_opt(tree_root, SSD) && |
1928 | !btrfs_test_opt(tree_root, NOSSD) && | |
1929 | !fs_info->fs_devices->rotating) { | |
1930 | printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD " | |
1931 | "mode\n"); | |
1932 | btrfs_set_opt(fs_info->mount_opt, SSD); | |
1933 | } | |
1934 | ||
e02119d5 | 1935 | if (btrfs_super_log_root(disk_super) != 0) { |
e02119d5 CM |
1936 | u64 bytenr = btrfs_super_log_root(disk_super); |
1937 | ||
7c2ca468 | 1938 | if (fs_devices->rw_devices == 0) { |
d397712b CM |
1939 | printk(KERN_WARNING "Btrfs log replay required " |
1940 | "on RO media\n"); | |
7c2ca468 CM |
1941 | err = -EIO; |
1942 | goto fail_trans_kthread; | |
1943 | } | |
e02119d5 CM |
1944 | blocksize = |
1945 | btrfs_level_size(tree_root, | |
1946 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 1947 | |
e02119d5 CM |
1948 | log_tree_root = kzalloc(sizeof(struct btrfs_root), |
1949 | GFP_NOFS); | |
1950 | ||
1951 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1952 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1953 | ||
1954 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
1955 | blocksize, |
1956 | generation + 1); | |
e02119d5 CM |
1957 | ret = btrfs_recover_log_trees(log_tree_root); |
1958 | BUG_ON(ret); | |
e556ce2c YZ |
1959 | |
1960 | if (sb->s_flags & MS_RDONLY) { | |
1961 | ret = btrfs_commit_super(tree_root); | |
1962 | BUG_ON(ret); | |
1963 | } | |
e02119d5 | 1964 | } |
1a40e23b | 1965 | |
76dda93c YZ |
1966 | ret = btrfs_find_orphan_roots(tree_root); |
1967 | BUG_ON(ret); | |
1968 | ||
7c2ca468 | 1969 | if (!(sb->s_flags & MS_RDONLY)) { |
5d4f98a2 | 1970 | ret = btrfs_recover_relocation(tree_root); |
7c2ca468 CM |
1971 | BUG_ON(ret); |
1972 | } | |
1a40e23b | 1973 | |
3de4586c CM |
1974 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
1975 | location.type = BTRFS_ROOT_ITEM_KEY; | |
1976 | location.offset = (u64)-1; | |
1977 | ||
3de4586c CM |
1978 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
1979 | if (!fs_info->fs_root) | |
7c2ca468 | 1980 | goto fail_trans_kthread; |
c289811c | 1981 | |
0f7d52f4 | 1982 | return tree_root; |
39279cc3 | 1983 | |
7c2ca468 CM |
1984 | fail_trans_kthread: |
1985 | kthread_stop(fs_info->transaction_kthread); | |
3f157a2f | 1986 | fail_cleaner: |
a74a4b97 | 1987 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
1988 | |
1989 | /* | |
1990 | * make sure we're done with the btree inode before we stop our | |
1991 | * kthreads | |
1992 | */ | |
1993 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
1994 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
1995 | ||
d20f7043 CM |
1996 | fail_csum_root: |
1997 | free_extent_buffer(csum_root->node); | |
5d4f98a2 YZ |
1998 | free_extent_buffer(csum_root->commit_root); |
1999 | fail_dev_root: | |
2000 | free_extent_buffer(dev_root->node); | |
2001 | free_extent_buffer(dev_root->commit_root); | |
0b86a832 CM |
2002 | fail_extent_root: |
2003 | free_extent_buffer(extent_root->node); | |
5d4f98a2 | 2004 | free_extent_buffer(extent_root->commit_root); |
39279cc3 | 2005 | fail_tree_root: |
5f39d397 | 2006 | free_extent_buffer(tree_root->node); |
5d4f98a2 | 2007 | free_extent_buffer(tree_root->commit_root); |
2b82032c YZ |
2008 | fail_chunk_root: |
2009 | free_extent_buffer(chunk_root->node); | |
5d4f98a2 | 2010 | free_extent_buffer(chunk_root->commit_root); |
39279cc3 | 2011 | fail_sb_buffer: |
247e743c | 2012 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2013 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2014 | btrfs_stop_workers(&fs_info->workers); |
2015 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 2016 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 2017 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 2018 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 2019 | btrfs_stop_workers(&fs_info->submit_workers); |
4543df7e | 2020 | fail_iput: |
7c2ca468 | 2021 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
4543df7e | 2022 | iput(fs_info->btree_inode); |
7e662854 | 2023 | |
dfe25020 | 2024 | btrfs_close_devices(fs_info->fs_devices); |
84eed90f | 2025 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
ad081f14 | 2026 | fail_bdi: |
7e662854 | 2027 | bdi_destroy(&fs_info->bdi); |
76dda93c YZ |
2028 | fail_srcu: |
2029 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 2030 | fail: |
39279cc3 CM |
2031 | kfree(extent_root); |
2032 | kfree(tree_root); | |
2033 | kfree(fs_info); | |
83afeac4 JM |
2034 | kfree(chunk_root); |
2035 | kfree(dev_root); | |
d20f7043 | 2036 | kfree(csum_root); |
39279cc3 | 2037 | return ERR_PTR(err); |
eb60ceac CM |
2038 | } |
2039 | ||
f2984462 CM |
2040 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
2041 | { | |
2042 | char b[BDEVNAME_SIZE]; | |
2043 | ||
2044 | if (uptodate) { | |
2045 | set_buffer_uptodate(bh); | |
2046 | } else { | |
2047 | if (!buffer_eopnotsupp(bh) && printk_ratelimit()) { | |
2048 | printk(KERN_WARNING "lost page write due to " | |
2049 | "I/O error on %s\n", | |
2050 | bdevname(bh->b_bdev, b)); | |
2051 | } | |
1259ab75 CM |
2052 | /* note, we dont' set_buffer_write_io_error because we have |
2053 | * our own ways of dealing with the IO errors | |
2054 | */ | |
f2984462 CM |
2055 | clear_buffer_uptodate(bh); |
2056 | } | |
2057 | unlock_buffer(bh); | |
2058 | put_bh(bh); | |
2059 | } | |
2060 | ||
a512bbf8 YZ |
2061 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
2062 | { | |
2063 | struct buffer_head *bh; | |
2064 | struct buffer_head *latest = NULL; | |
2065 | struct btrfs_super_block *super; | |
2066 | int i; | |
2067 | u64 transid = 0; | |
2068 | u64 bytenr; | |
2069 | ||
2070 | /* we would like to check all the supers, but that would make | |
2071 | * a btrfs mount succeed after a mkfs from a different FS. | |
2072 | * So, we need to add a special mount option to scan for | |
2073 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
2074 | */ | |
2075 | for (i = 0; i < 1; i++) { | |
2076 | bytenr = btrfs_sb_offset(i); | |
2077 | if (bytenr + 4096 >= i_size_read(bdev->bd_inode)) | |
2078 | break; | |
2079 | bh = __bread(bdev, bytenr / 4096, 4096); | |
2080 | if (!bh) | |
2081 | continue; | |
2082 | ||
2083 | super = (struct btrfs_super_block *)bh->b_data; | |
2084 | if (btrfs_super_bytenr(super) != bytenr || | |
2085 | strncmp((char *)(&super->magic), BTRFS_MAGIC, | |
2086 | sizeof(super->magic))) { | |
2087 | brelse(bh); | |
2088 | continue; | |
2089 | } | |
2090 | ||
2091 | if (!latest || btrfs_super_generation(super) > transid) { | |
2092 | brelse(latest); | |
2093 | latest = bh; | |
2094 | transid = btrfs_super_generation(super); | |
2095 | } else { | |
2096 | brelse(bh); | |
2097 | } | |
2098 | } | |
2099 | return latest; | |
2100 | } | |
2101 | ||
4eedeb75 HH |
2102 | /* |
2103 | * this should be called twice, once with wait == 0 and | |
2104 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
2105 | * we write are pinned. | |
2106 | * | |
2107 | * They are released when wait == 1 is done. | |
2108 | * max_mirrors must be the same for both runs, and it indicates how | |
2109 | * many supers on this one device should be written. | |
2110 | * | |
2111 | * max_mirrors == 0 means to write them all. | |
2112 | */ | |
a512bbf8 YZ |
2113 | static int write_dev_supers(struct btrfs_device *device, |
2114 | struct btrfs_super_block *sb, | |
2115 | int do_barriers, int wait, int max_mirrors) | |
2116 | { | |
2117 | struct buffer_head *bh; | |
2118 | int i; | |
2119 | int ret; | |
2120 | int errors = 0; | |
2121 | u32 crc; | |
2122 | u64 bytenr; | |
2123 | int last_barrier = 0; | |
2124 | ||
2125 | if (max_mirrors == 0) | |
2126 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
2127 | ||
2128 | /* make sure only the last submit_bh does a barrier */ | |
2129 | if (do_barriers) { | |
2130 | for (i = 0; i < max_mirrors; i++) { | |
2131 | bytenr = btrfs_sb_offset(i); | |
2132 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= | |
2133 | device->total_bytes) | |
2134 | break; | |
2135 | last_barrier = i; | |
2136 | } | |
2137 | } | |
2138 | ||
2139 | for (i = 0; i < max_mirrors; i++) { | |
2140 | bytenr = btrfs_sb_offset(i); | |
2141 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) | |
2142 | break; | |
2143 | ||
2144 | if (wait) { | |
2145 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
2146 | BTRFS_SUPER_INFO_SIZE); | |
2147 | BUG_ON(!bh); | |
a512bbf8 | 2148 | wait_on_buffer(bh); |
4eedeb75 HH |
2149 | if (!buffer_uptodate(bh)) |
2150 | errors++; | |
2151 | ||
2152 | /* drop our reference */ | |
2153 | brelse(bh); | |
2154 | ||
2155 | /* drop the reference from the wait == 0 run */ | |
2156 | brelse(bh); | |
2157 | continue; | |
a512bbf8 YZ |
2158 | } else { |
2159 | btrfs_set_super_bytenr(sb, bytenr); | |
2160 | ||
2161 | crc = ~(u32)0; | |
2162 | crc = btrfs_csum_data(NULL, (char *)sb + | |
2163 | BTRFS_CSUM_SIZE, crc, | |
2164 | BTRFS_SUPER_INFO_SIZE - | |
2165 | BTRFS_CSUM_SIZE); | |
2166 | btrfs_csum_final(crc, sb->csum); | |
2167 | ||
4eedeb75 HH |
2168 | /* |
2169 | * one reference for us, and we leave it for the | |
2170 | * caller | |
2171 | */ | |
a512bbf8 YZ |
2172 | bh = __getblk(device->bdev, bytenr / 4096, |
2173 | BTRFS_SUPER_INFO_SIZE); | |
2174 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); | |
2175 | ||
4eedeb75 | 2176 | /* one reference for submit_bh */ |
a512bbf8 | 2177 | get_bh(bh); |
4eedeb75 HH |
2178 | |
2179 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
2180 | lock_buffer(bh); |
2181 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
2182 | } | |
2183 | ||
2184 | if (i == last_barrier && do_barriers && device->barriers) { | |
2185 | ret = submit_bh(WRITE_BARRIER, bh); | |
2186 | if (ret == -EOPNOTSUPP) { | |
2187 | printk("btrfs: disabling barriers on dev %s\n", | |
2188 | device->name); | |
2189 | set_buffer_uptodate(bh); | |
2190 | device->barriers = 0; | |
4eedeb75 | 2191 | /* one reference for submit_bh */ |
a512bbf8 YZ |
2192 | get_bh(bh); |
2193 | lock_buffer(bh); | |
ffbd517d | 2194 | ret = submit_bh(WRITE_SYNC, bh); |
a512bbf8 YZ |
2195 | } |
2196 | } else { | |
ffbd517d | 2197 | ret = submit_bh(WRITE_SYNC, bh); |
a512bbf8 YZ |
2198 | } |
2199 | ||
4eedeb75 | 2200 | if (ret) |
a512bbf8 | 2201 | errors++; |
a512bbf8 YZ |
2202 | } |
2203 | return errors < i ? 0 : -1; | |
2204 | } | |
2205 | ||
2206 | int write_all_supers(struct btrfs_root *root, int max_mirrors) | |
f2984462 | 2207 | { |
e5e9a520 | 2208 | struct list_head *head; |
f2984462 | 2209 | struct btrfs_device *dev; |
a061fc8d | 2210 | struct btrfs_super_block *sb; |
f2984462 | 2211 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
2212 | int ret; |
2213 | int do_barriers; | |
a236aed1 CM |
2214 | int max_errors; |
2215 | int total_errors = 0; | |
a061fc8d | 2216 | u64 flags; |
f2984462 | 2217 | |
a236aed1 | 2218 | max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1; |
f2984462 CM |
2219 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
2220 | ||
a061fc8d CM |
2221 | sb = &root->fs_info->super_for_commit; |
2222 | dev_item = &sb->dev_item; | |
e5e9a520 CM |
2223 | |
2224 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); | |
2225 | head = &root->fs_info->fs_devices->devices; | |
c6e30871 | 2226 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
2227 | if (!dev->bdev) { |
2228 | total_errors++; | |
2229 | continue; | |
2230 | } | |
2b82032c | 2231 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2232 | continue; |
2233 | ||
2b82032c | 2234 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
2235 | btrfs_set_stack_device_type(dev_item, dev->type); |
2236 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
2237 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
2238 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
2239 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
2240 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
2241 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
2242 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 2243 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 2244 | |
a061fc8d CM |
2245 | flags = btrfs_super_flags(sb); |
2246 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
2247 | ||
a512bbf8 | 2248 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
2249 | if (ret) |
2250 | total_errors++; | |
f2984462 | 2251 | } |
a236aed1 | 2252 | if (total_errors > max_errors) { |
d397712b CM |
2253 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2254 | total_errors); | |
a236aed1 CM |
2255 | BUG(); |
2256 | } | |
f2984462 | 2257 | |
a512bbf8 | 2258 | total_errors = 0; |
c6e30871 | 2259 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
2260 | if (!dev->bdev) |
2261 | continue; | |
2b82032c | 2262 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2263 | continue; |
2264 | ||
a512bbf8 YZ |
2265 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
2266 | if (ret) | |
2267 | total_errors++; | |
f2984462 | 2268 | } |
e5e9a520 | 2269 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 2270 | if (total_errors > max_errors) { |
d397712b CM |
2271 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2272 | total_errors); | |
a236aed1 CM |
2273 | BUG(); |
2274 | } | |
f2984462 CM |
2275 | return 0; |
2276 | } | |
2277 | ||
a512bbf8 YZ |
2278 | int write_ctree_super(struct btrfs_trans_handle *trans, |
2279 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 2280 | { |
e66f709b | 2281 | int ret; |
5f39d397 | 2282 | |
a512bbf8 | 2283 | ret = write_all_supers(root, max_mirrors); |
5f39d397 | 2284 | return ret; |
cfaa7295 CM |
2285 | } |
2286 | ||
5eda7b5e | 2287 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f | 2288 | { |
4df27c4d | 2289 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
2290 | radix_tree_delete(&fs_info->fs_roots_radix, |
2291 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 2292 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
2293 | |
2294 | if (btrfs_root_refs(&root->root_item) == 0) | |
2295 | synchronize_srcu(&fs_info->subvol_srcu); | |
2296 | ||
4df27c4d YZ |
2297 | free_fs_root(root); |
2298 | return 0; | |
2299 | } | |
2300 | ||
2301 | static void free_fs_root(struct btrfs_root *root) | |
2302 | { | |
2303 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); | |
3394e160 CM |
2304 | if (root->anon_super.s_dev) { |
2305 | down_write(&root->anon_super.s_umount); | |
2306 | kill_anon_super(&root->anon_super); | |
2307 | } | |
4df27c4d YZ |
2308 | free_extent_buffer(root->node); |
2309 | free_extent_buffer(root->commit_root); | |
d397712b | 2310 | kfree(root->name); |
2619ba1f | 2311 | kfree(root); |
2619ba1f CM |
2312 | } |
2313 | ||
35b7e476 | 2314 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
2315 | { |
2316 | int ret; | |
2317 | struct btrfs_root *gang[8]; | |
2318 | int i; | |
2319 | ||
76dda93c YZ |
2320 | while (!list_empty(&fs_info->dead_roots)) { |
2321 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2322 | struct btrfs_root, root_list); | |
2323 | list_del(&gang[0]->root_list); | |
2324 | ||
2325 | if (gang[0]->in_radix) { | |
2326 | btrfs_free_fs_root(fs_info, gang[0]); | |
2327 | } else { | |
2328 | free_extent_buffer(gang[0]->node); | |
2329 | free_extent_buffer(gang[0]->commit_root); | |
2330 | kfree(gang[0]); | |
2331 | } | |
2332 | } | |
2333 | ||
d397712b | 2334 | while (1) { |
0f7d52f4 CM |
2335 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
2336 | (void **)gang, 0, | |
2337 | ARRAY_SIZE(gang)); | |
2338 | if (!ret) | |
2339 | break; | |
2619ba1f | 2340 | for (i = 0; i < ret; i++) |
5eda7b5e | 2341 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
2342 | } |
2343 | return 0; | |
2344 | } | |
b4100d64 | 2345 | |
c146afad | 2346 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 2347 | { |
c146afad YZ |
2348 | u64 root_objectid = 0; |
2349 | struct btrfs_root *gang[8]; | |
2350 | int i; | |
3768f368 | 2351 | int ret; |
e089f05c | 2352 | |
c146afad YZ |
2353 | while (1) { |
2354 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2355 | (void **)gang, root_objectid, | |
2356 | ARRAY_SIZE(gang)); | |
2357 | if (!ret) | |
2358 | break; | |
5d4f98a2 YZ |
2359 | |
2360 | root_objectid = gang[ret - 1]->root_key.objectid + 1; | |
c146afad YZ |
2361 | for (i = 0; i < ret; i++) { |
2362 | root_objectid = gang[i]->root_key.objectid; | |
c146afad YZ |
2363 | btrfs_orphan_cleanup(gang[i]); |
2364 | } | |
2365 | root_objectid++; | |
2366 | } | |
2367 | return 0; | |
2368 | } | |
a2135011 | 2369 | |
c146afad YZ |
2370 | int btrfs_commit_super(struct btrfs_root *root) |
2371 | { | |
2372 | struct btrfs_trans_handle *trans; | |
2373 | int ret; | |
a74a4b97 | 2374 | |
c146afad | 2375 | mutex_lock(&root->fs_info->cleaner_mutex); |
a74a4b97 | 2376 | btrfs_clean_old_snapshots(root); |
c146afad | 2377 | mutex_unlock(&root->fs_info->cleaner_mutex); |
79154b1b | 2378 | trans = btrfs_start_transaction(root, 1); |
54aa1f4d | 2379 | ret = btrfs_commit_transaction(trans, root); |
c146afad YZ |
2380 | BUG_ON(ret); |
2381 | /* run commit again to drop the original snapshot */ | |
79154b1b CM |
2382 | trans = btrfs_start_transaction(root, 1); |
2383 | btrfs_commit_transaction(trans, root); | |
2384 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 2385 | BUG_ON(ret); |
d6bfde87 | 2386 | |
a512bbf8 | 2387 | ret = write_ctree_super(NULL, root, 0); |
c146afad YZ |
2388 | return ret; |
2389 | } | |
2390 | ||
2391 | int close_ctree(struct btrfs_root *root) | |
2392 | { | |
2393 | struct btrfs_fs_info *fs_info = root->fs_info; | |
2394 | int ret; | |
2395 | ||
2396 | fs_info->closing = 1; | |
2397 | smp_mb(); | |
2398 | ||
2399 | kthread_stop(root->fs_info->transaction_kthread); | |
2400 | kthread_stop(root->fs_info->cleaner_kthread); | |
2401 | ||
2402 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { | |
2403 | ret = btrfs_commit_super(root); | |
d397712b CM |
2404 | if (ret) |
2405 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
c146afad | 2406 | } |
0f7d52f4 | 2407 | |
f25784b3 YZ |
2408 | fs_info->closing = 2; |
2409 | smp_mb(); | |
2410 | ||
b0c68f8b | 2411 | if (fs_info->delalloc_bytes) { |
d397712b | 2412 | printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n", |
21380931 | 2413 | (unsigned long long)fs_info->delalloc_bytes); |
b0c68f8b | 2414 | } |
31153d81 | 2415 | if (fs_info->total_ref_cache_size) { |
d397712b CM |
2416 | printk(KERN_INFO "btrfs: at umount reference cache size %llu\n", |
2417 | (unsigned long long)fs_info->total_ref_cache_size); | |
31153d81 | 2418 | } |
bcc63abb | 2419 | |
5d4f98a2 YZ |
2420 | free_extent_buffer(fs_info->extent_root->node); |
2421 | free_extent_buffer(fs_info->extent_root->commit_root); | |
2422 | free_extent_buffer(fs_info->tree_root->node); | |
2423 | free_extent_buffer(fs_info->tree_root->commit_root); | |
2424 | free_extent_buffer(root->fs_info->chunk_root->node); | |
2425 | free_extent_buffer(root->fs_info->chunk_root->commit_root); | |
2426 | free_extent_buffer(root->fs_info->dev_root->node); | |
2427 | free_extent_buffer(root->fs_info->dev_root->commit_root); | |
2428 | free_extent_buffer(root->fs_info->csum_root->node); | |
2429 | free_extent_buffer(root->fs_info->csum_root->commit_root); | |
d20f7043 | 2430 | |
9078a3e1 | 2431 | btrfs_free_block_groups(root->fs_info); |
d10c5f31 | 2432 | |
c146afad | 2433 | del_fs_roots(fs_info); |
d10c5f31 | 2434 | |
c146afad | 2435 | iput(fs_info->btree_inode); |
9ad6b7bc | 2436 | |
247e743c | 2437 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2438 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2439 | btrfs_stop_workers(&fs_info->workers); |
2440 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 2441 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 2442 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 2443 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 2444 | btrfs_stop_workers(&fs_info->submit_workers); |
d6bfde87 | 2445 | |
dfe25020 | 2446 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 2447 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 2448 | |
04160088 | 2449 | bdi_destroy(&fs_info->bdi); |
76dda93c | 2450 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 2451 | |
0f7d52f4 | 2452 | kfree(fs_info->extent_root); |
0f7d52f4 | 2453 | kfree(fs_info->tree_root); |
0b86a832 CM |
2454 | kfree(fs_info->chunk_root); |
2455 | kfree(fs_info->dev_root); | |
d20f7043 | 2456 | kfree(fs_info->csum_root); |
eb60ceac CM |
2457 | return 0; |
2458 | } | |
2459 | ||
1259ab75 | 2460 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 2461 | { |
1259ab75 | 2462 | int ret; |
810191ff | 2463 | struct inode *btree_inode = buf->first_page->mapping->host; |
1259ab75 CM |
2464 | |
2465 | ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); | |
2466 | if (!ret) | |
2467 | return ret; | |
2468 | ||
2469 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
2470 | parent_transid); | |
2471 | return !ret; | |
5f39d397 CM |
2472 | } |
2473 | ||
2474 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 2475 | { |
810191ff | 2476 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 2477 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
2478 | buf); |
2479 | } | |
6702ed49 | 2480 | |
5f39d397 CM |
2481 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
2482 | { | |
810191ff | 2483 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
5f39d397 CM |
2484 | u64 transid = btrfs_header_generation(buf); |
2485 | struct inode *btree_inode = root->fs_info->btree_inode; | |
b9473439 | 2486 | int was_dirty; |
b4ce94de | 2487 | |
b9447ef8 | 2488 | btrfs_assert_tree_locked(buf); |
ccd467d6 | 2489 | if (transid != root->fs_info->generation) { |
d397712b CM |
2490 | printk(KERN_CRIT "btrfs transid mismatch buffer %llu, " |
2491 | "found %llu running %llu\n", | |
db94535d | 2492 | (unsigned long long)buf->start, |
d397712b CM |
2493 | (unsigned long long)transid, |
2494 | (unsigned long long)root->fs_info->generation); | |
ccd467d6 CM |
2495 | WARN_ON(1); |
2496 | } | |
b9473439 CM |
2497 | was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
2498 | buf); | |
2499 | if (!was_dirty) { | |
2500 | spin_lock(&root->fs_info->delalloc_lock); | |
2501 | root->fs_info->dirty_metadata_bytes += buf->len; | |
2502 | spin_unlock(&root->fs_info->delalloc_lock); | |
2503 | } | |
eb60ceac CM |
2504 | } |
2505 | ||
d3c2fdcf | 2506 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
35b7e476 | 2507 | { |
188de649 CM |
2508 | /* |
2509 | * looks as though older kernels can get into trouble with | |
2510 | * this code, they end up stuck in balance_dirty_pages forever | |
2511 | */ | |
d6bfde87 | 2512 | u64 num_dirty; |
771ed689 | 2513 | unsigned long thresh = 32 * 1024 * 1024; |
d6bfde87 | 2514 | |
6933c02e | 2515 | if (current->flags & PF_MEMALLOC) |
d6bfde87 CM |
2516 | return; |
2517 | ||
585ad2c3 CM |
2518 | num_dirty = root->fs_info->dirty_metadata_bytes; |
2519 | ||
d6bfde87 CM |
2520 | if (num_dirty > thresh) { |
2521 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 2522 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 2523 | } |
188de649 | 2524 | return; |
35b7e476 | 2525 | } |
6b80053d | 2526 | |
ca7a79ad | 2527 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 2528 | { |
810191ff | 2529 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
ce9adaa5 | 2530 | int ret; |
ca7a79ad | 2531 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
d397712b | 2532 | if (ret == 0) |
b4ce94de | 2533 | set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags); |
ce9adaa5 | 2534 | return ret; |
6b80053d | 2535 | } |
0da5468f | 2536 | |
4bef0848 CM |
2537 | int btree_lock_page_hook(struct page *page) |
2538 | { | |
2539 | struct inode *inode = page->mapping->host; | |
b9473439 | 2540 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4bef0848 CM |
2541 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
2542 | struct extent_buffer *eb; | |
2543 | unsigned long len; | |
2544 | u64 bytenr = page_offset(page); | |
2545 | ||
2546 | if (page->private == EXTENT_PAGE_PRIVATE) | |
2547 | goto out; | |
2548 | ||
2549 | len = page->private >> 2; | |
2550 | eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS); | |
2551 | if (!eb) | |
2552 | goto out; | |
2553 | ||
2554 | btrfs_tree_lock(eb); | |
4bef0848 | 2555 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); |
b9473439 CM |
2556 | |
2557 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) { | |
2558 | spin_lock(&root->fs_info->delalloc_lock); | |
2559 | if (root->fs_info->dirty_metadata_bytes >= eb->len) | |
2560 | root->fs_info->dirty_metadata_bytes -= eb->len; | |
2561 | else | |
2562 | WARN_ON(1); | |
2563 | spin_unlock(&root->fs_info->delalloc_lock); | |
2564 | } | |
2565 | ||
4bef0848 CM |
2566 | btrfs_tree_unlock(eb); |
2567 | free_extent_buffer(eb); | |
2568 | out: | |
2569 | lock_page(page); | |
2570 | return 0; | |
2571 | } | |
2572 | ||
d1310b2e | 2573 | static struct extent_io_ops btree_extent_io_ops = { |
4bef0848 | 2574 | .write_cache_pages_lock_hook = btree_lock_page_hook, |
ce9adaa5 | 2575 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
0b86a832 | 2576 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
2577 | /* note we're sharing with inode.c for the merge bio hook */ |
2578 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 2579 | }; |