]>
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); |
ce9adaa5 | 44 | |
ad081f14 JA |
45 | static atomic_t btrfs_bdi_num = ATOMIC_INIT(0); |
46 | ||
d352ac68 CM |
47 | /* |
48 | * end_io_wq structs are used to do processing in task context when an IO is | |
49 | * complete. This is used during reads to verify checksums, and it is used | |
50 | * by writes to insert metadata for new file extents after IO is complete. | |
51 | */ | |
ce9adaa5 CM |
52 | struct end_io_wq { |
53 | struct bio *bio; | |
54 | bio_end_io_t *end_io; | |
55 | void *private; | |
56 | struct btrfs_fs_info *info; | |
57 | int error; | |
22c59948 | 58 | int metadata; |
ce9adaa5 | 59 | struct list_head list; |
8b712842 | 60 | struct btrfs_work work; |
ce9adaa5 | 61 | }; |
0da5468f | 62 | |
d352ac68 CM |
63 | /* |
64 | * async submit bios are used to offload expensive checksumming | |
65 | * onto the worker threads. They checksum file and metadata bios | |
66 | * just before they are sent down the IO stack. | |
67 | */ | |
44b8bd7e CM |
68 | struct async_submit_bio { |
69 | struct inode *inode; | |
70 | struct bio *bio; | |
71 | struct list_head list; | |
4a69a410 CM |
72 | extent_submit_bio_hook_t *submit_bio_start; |
73 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
74 | int rw; |
75 | int mirror_num; | |
c8b97818 | 76 | unsigned long bio_flags; |
8b712842 | 77 | struct btrfs_work work; |
44b8bd7e CM |
78 | }; |
79 | ||
4008c04a CM |
80 | /* These are used to set the lockdep class on the extent buffer locks. |
81 | * The class is set by the readpage_end_io_hook after the buffer has | |
82 | * passed csum validation but before the pages are unlocked. | |
83 | * | |
84 | * The lockdep class is also set by btrfs_init_new_buffer on freshly | |
85 | * allocated blocks. | |
86 | * | |
87 | * The class is based on the level in the tree block, which allows lockdep | |
88 | * to know that lower nodes nest inside the locks of higher nodes. | |
89 | * | |
90 | * We also add a check to make sure the highest level of the tree is | |
91 | * the same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this | |
92 | * code needs update as well. | |
93 | */ | |
94 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
95 | # if BTRFS_MAX_LEVEL != 8 | |
96 | # error | |
97 | # endif | |
98 | static struct lock_class_key btrfs_eb_class[BTRFS_MAX_LEVEL + 1]; | |
99 | static const char *btrfs_eb_name[BTRFS_MAX_LEVEL + 1] = { | |
100 | /* leaf */ | |
101 | "btrfs-extent-00", | |
102 | "btrfs-extent-01", | |
103 | "btrfs-extent-02", | |
104 | "btrfs-extent-03", | |
105 | "btrfs-extent-04", | |
106 | "btrfs-extent-05", | |
107 | "btrfs-extent-06", | |
108 | "btrfs-extent-07", | |
109 | /* highest possible level */ | |
110 | "btrfs-extent-08", | |
111 | }; | |
112 | #endif | |
113 | ||
d352ac68 CM |
114 | /* |
115 | * extents on the btree inode are pretty simple, there's one extent | |
116 | * that covers the entire device | |
117 | */ | |
b2950863 CH |
118 | static struct extent_map *btree_get_extent(struct inode *inode, |
119 | struct page *page, size_t page_offset, u64 start, u64 len, | |
120 | int create) | |
7eccb903 | 121 | { |
5f39d397 CM |
122 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
123 | struct extent_map *em; | |
124 | int ret; | |
125 | ||
890871be | 126 | read_lock(&em_tree->lock); |
d1310b2e | 127 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d CM |
128 | if (em) { |
129 | em->bdev = | |
130 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
890871be | 131 | read_unlock(&em_tree->lock); |
5f39d397 | 132 | goto out; |
a061fc8d | 133 | } |
890871be | 134 | read_unlock(&em_tree->lock); |
7b13b7b1 | 135 | |
5f39d397 CM |
136 | em = alloc_extent_map(GFP_NOFS); |
137 | if (!em) { | |
138 | em = ERR_PTR(-ENOMEM); | |
139 | goto out; | |
140 | } | |
141 | em->start = 0; | |
0afbaf8c | 142 | em->len = (u64)-1; |
c8b97818 | 143 | em->block_len = (u64)-1; |
5f39d397 | 144 | em->block_start = 0; |
a061fc8d | 145 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e | 146 | |
890871be | 147 | write_lock(&em_tree->lock); |
5f39d397 CM |
148 | ret = add_extent_mapping(em_tree, em); |
149 | if (ret == -EEXIST) { | |
0afbaf8c CM |
150 | u64 failed_start = em->start; |
151 | u64 failed_len = em->len; | |
152 | ||
5f39d397 | 153 | free_extent_map(em); |
7b13b7b1 | 154 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c | 155 | if (em) { |
7b13b7b1 | 156 | ret = 0; |
0afbaf8c CM |
157 | } else { |
158 | em = lookup_extent_mapping(em_tree, failed_start, | |
159 | failed_len); | |
7b13b7b1 | 160 | ret = -EIO; |
0afbaf8c | 161 | } |
5f39d397 | 162 | } else if (ret) { |
7b13b7b1 CM |
163 | free_extent_map(em); |
164 | em = NULL; | |
5f39d397 | 165 | } |
890871be | 166 | write_unlock(&em_tree->lock); |
7b13b7b1 CM |
167 | |
168 | if (ret) | |
169 | em = ERR_PTR(ret); | |
5f39d397 CM |
170 | out: |
171 | return em; | |
7eccb903 CM |
172 | } |
173 | ||
19c00ddc CM |
174 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
175 | { | |
163e783e | 176 | return crc32c(seed, data, len); |
19c00ddc CM |
177 | } |
178 | ||
179 | void btrfs_csum_final(u32 crc, char *result) | |
180 | { | |
181 | *(__le32 *)result = ~cpu_to_le32(crc); | |
182 | } | |
183 | ||
d352ac68 CM |
184 | /* |
185 | * compute the csum for a btree block, and either verify it or write it | |
186 | * into the csum field of the block. | |
187 | */ | |
19c00ddc CM |
188 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
189 | int verify) | |
190 | { | |
607d432d JB |
191 | u16 csum_size = |
192 | btrfs_super_csum_size(&root->fs_info->super_copy); | |
193 | char *result = NULL; | |
19c00ddc CM |
194 | unsigned long len; |
195 | unsigned long cur_len; | |
196 | unsigned long offset = BTRFS_CSUM_SIZE; | |
197 | char *map_token = NULL; | |
198 | char *kaddr; | |
199 | unsigned long map_start; | |
200 | unsigned long map_len; | |
201 | int err; | |
202 | u32 crc = ~(u32)0; | |
607d432d | 203 | unsigned long inline_result; |
19c00ddc CM |
204 | |
205 | len = buf->len - offset; | |
d397712b | 206 | while (len > 0) { |
19c00ddc CM |
207 | err = map_private_extent_buffer(buf, offset, 32, |
208 | &map_token, &kaddr, | |
209 | &map_start, &map_len, KM_USER0); | |
d397712b | 210 | if (err) |
19c00ddc | 211 | return 1; |
19c00ddc CM |
212 | cur_len = min(len, map_len - (offset - map_start)); |
213 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
214 | crc, cur_len); | |
215 | len -= cur_len; | |
216 | offset += cur_len; | |
217 | unmap_extent_buffer(buf, map_token, KM_USER0); | |
218 | } | |
607d432d JB |
219 | if (csum_size > sizeof(inline_result)) { |
220 | result = kzalloc(csum_size * sizeof(char), GFP_NOFS); | |
221 | if (!result) | |
222 | return 1; | |
223 | } else { | |
224 | result = (char *)&inline_result; | |
225 | } | |
226 | ||
19c00ddc CM |
227 | btrfs_csum_final(crc, result); |
228 | ||
229 | if (verify) { | |
607d432d | 230 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
231 | u32 val; |
232 | u32 found = 0; | |
607d432d | 233 | memcpy(&found, result, csum_size); |
e4204ded | 234 | |
607d432d | 235 | read_extent_buffer(buf, &val, 0, csum_size); |
193f284d CM |
236 | if (printk_ratelimit()) { |
237 | printk(KERN_INFO "btrfs: %s checksum verify " | |
238 | "failed on %llu wanted %X found %X " | |
239 | "level %d\n", | |
240 | root->fs_info->sb->s_id, | |
241 | (unsigned long long)buf->start, val, found, | |
242 | btrfs_header_level(buf)); | |
243 | } | |
607d432d JB |
244 | if (result != (char *)&inline_result) |
245 | kfree(result); | |
19c00ddc CM |
246 | return 1; |
247 | } | |
248 | } else { | |
607d432d | 249 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 250 | } |
607d432d JB |
251 | if (result != (char *)&inline_result) |
252 | kfree(result); | |
19c00ddc CM |
253 | return 0; |
254 | } | |
255 | ||
d352ac68 CM |
256 | /* |
257 | * we can't consider a given block up to date unless the transid of the | |
258 | * block matches the transid in the parent node's pointer. This is how we | |
259 | * detect blocks that either didn't get written at all or got written | |
260 | * in the wrong place. | |
261 | */ | |
1259ab75 CM |
262 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
263 | struct extent_buffer *eb, u64 parent_transid) | |
264 | { | |
265 | int ret; | |
266 | ||
267 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
268 | return 0; | |
269 | ||
270 | lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); | |
271 | if (extent_buffer_uptodate(io_tree, eb) && | |
272 | btrfs_header_generation(eb) == parent_transid) { | |
273 | ret = 0; | |
274 | goto out; | |
275 | } | |
193f284d CM |
276 | if (printk_ratelimit()) { |
277 | printk("parent transid verify failed on %llu wanted %llu " | |
278 | "found %llu\n", | |
279 | (unsigned long long)eb->start, | |
280 | (unsigned long long)parent_transid, | |
281 | (unsigned long long)btrfs_header_generation(eb)); | |
282 | } | |
1259ab75 | 283 | ret = 1; |
1259ab75 | 284 | clear_extent_buffer_uptodate(io_tree, eb); |
33958dc6 | 285 | out: |
1259ab75 CM |
286 | unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, |
287 | GFP_NOFS); | |
288 | return ret; | |
1259ab75 CM |
289 | } |
290 | ||
d352ac68 CM |
291 | /* |
292 | * helper to read a given tree block, doing retries as required when | |
293 | * the checksums don't match and we have alternate mirrors to try. | |
294 | */ | |
f188591e CM |
295 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
296 | struct extent_buffer *eb, | |
ca7a79ad | 297 | u64 start, u64 parent_transid) |
f188591e CM |
298 | { |
299 | struct extent_io_tree *io_tree; | |
300 | int ret; | |
301 | int num_copies = 0; | |
302 | int mirror_num = 0; | |
303 | ||
304 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
305 | while (1) { | |
306 | ret = read_extent_buffer_pages(io_tree, eb, start, 1, | |
307 | btree_get_extent, mirror_num); | |
1259ab75 CM |
308 | if (!ret && |
309 | !verify_parent_transid(io_tree, eb, parent_transid)) | |
f188591e | 310 | return ret; |
d397712b | 311 | |
f188591e CM |
312 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
313 | eb->start, eb->len); | |
4235298e | 314 | if (num_copies == 1) |
f188591e | 315 | return ret; |
4235298e | 316 | |
f188591e | 317 | mirror_num++; |
4235298e | 318 | if (mirror_num > num_copies) |
f188591e | 319 | return ret; |
f188591e | 320 | } |
f188591e CM |
321 | return -EIO; |
322 | } | |
19c00ddc | 323 | |
d352ac68 | 324 | /* |
d397712b CM |
325 | * checksum a dirty tree block before IO. This has extra checks to make sure |
326 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 327 | */ |
d397712b | 328 | |
b2950863 | 329 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 330 | { |
d1310b2e | 331 | struct extent_io_tree *tree; |
35ebb934 | 332 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc CM |
333 | u64 found_start; |
334 | int found_level; | |
335 | unsigned long len; | |
336 | struct extent_buffer *eb; | |
f188591e CM |
337 | int ret; |
338 | ||
d1310b2e | 339 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc CM |
340 | |
341 | if (page->private == EXTENT_PAGE_PRIVATE) | |
342 | goto out; | |
343 | if (!page->private) | |
344 | goto out; | |
345 | len = page->private >> 2; | |
d397712b CM |
346 | WARN_ON(len == 0); |
347 | ||
19c00ddc | 348 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); |
ca7a79ad CM |
349 | ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, |
350 | btrfs_header_generation(eb)); | |
f188591e | 351 | BUG_ON(ret); |
19c00ddc CM |
352 | found_start = btrfs_header_bytenr(eb); |
353 | if (found_start != start) { | |
55c69072 CM |
354 | WARN_ON(1); |
355 | goto err; | |
356 | } | |
357 | if (eb->first_page != page) { | |
55c69072 CM |
358 | WARN_ON(1); |
359 | goto err; | |
360 | } | |
361 | if (!PageUptodate(page)) { | |
55c69072 CM |
362 | WARN_ON(1); |
363 | goto err; | |
19c00ddc CM |
364 | } |
365 | found_level = btrfs_header_level(eb); | |
4bef0848 | 366 | |
19c00ddc | 367 | csum_tree_block(root, eb, 0); |
55c69072 | 368 | err: |
19c00ddc CM |
369 | free_extent_buffer(eb); |
370 | out: | |
371 | return 0; | |
372 | } | |
373 | ||
2b82032c YZ |
374 | static int check_tree_block_fsid(struct btrfs_root *root, |
375 | struct extent_buffer *eb) | |
376 | { | |
377 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
378 | u8 fsid[BTRFS_UUID_SIZE]; | |
379 | int ret = 1; | |
380 | ||
381 | read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), | |
382 | BTRFS_FSID_SIZE); | |
383 | while (fs_devices) { | |
384 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
385 | ret = 0; | |
386 | break; | |
387 | } | |
388 | fs_devices = fs_devices->seed; | |
389 | } | |
390 | return ret; | |
391 | } | |
392 | ||
4008c04a CM |
393 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
394 | void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, int level) | |
395 | { | |
396 | lockdep_set_class_and_name(&eb->lock, | |
397 | &btrfs_eb_class[level], | |
398 | btrfs_eb_name[level]); | |
399 | } | |
400 | #endif | |
401 | ||
b2950863 | 402 | static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
ce9adaa5 CM |
403 | struct extent_state *state) |
404 | { | |
405 | struct extent_io_tree *tree; | |
406 | u64 found_start; | |
407 | int found_level; | |
408 | unsigned long len; | |
409 | struct extent_buffer *eb; | |
410 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 411 | int ret = 0; |
ce9adaa5 CM |
412 | |
413 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
414 | if (page->private == EXTENT_PAGE_PRIVATE) | |
415 | goto out; | |
416 | if (!page->private) | |
417 | goto out; | |
d397712b | 418 | |
ce9adaa5 | 419 | len = page->private >> 2; |
d397712b CM |
420 | WARN_ON(len == 0); |
421 | ||
ce9adaa5 | 422 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); |
f188591e | 423 | |
ce9adaa5 | 424 | found_start = btrfs_header_bytenr(eb); |
23a07867 | 425 | if (found_start != start) { |
193f284d CM |
426 | if (printk_ratelimit()) { |
427 | printk(KERN_INFO "btrfs bad tree block start " | |
428 | "%llu %llu\n", | |
429 | (unsigned long long)found_start, | |
430 | (unsigned long long)eb->start); | |
431 | } | |
f188591e | 432 | ret = -EIO; |
ce9adaa5 CM |
433 | goto err; |
434 | } | |
435 | if (eb->first_page != page) { | |
d397712b CM |
436 | printk(KERN_INFO "btrfs bad first page %lu %lu\n", |
437 | eb->first_page->index, page->index); | |
ce9adaa5 | 438 | WARN_ON(1); |
f188591e | 439 | ret = -EIO; |
ce9adaa5 CM |
440 | goto err; |
441 | } | |
2b82032c | 442 | if (check_tree_block_fsid(root, eb)) { |
193f284d CM |
443 | if (printk_ratelimit()) { |
444 | printk(KERN_INFO "btrfs bad fsid on block %llu\n", | |
445 | (unsigned long long)eb->start); | |
446 | } | |
1259ab75 CM |
447 | ret = -EIO; |
448 | goto err; | |
449 | } | |
ce9adaa5 CM |
450 | found_level = btrfs_header_level(eb); |
451 | ||
4008c04a CM |
452 | btrfs_set_buffer_lockdep_class(eb, found_level); |
453 | ||
ce9adaa5 | 454 | ret = csum_tree_block(root, eb, 1); |
f188591e CM |
455 | if (ret) |
456 | ret = -EIO; | |
ce9adaa5 CM |
457 | |
458 | end = min_t(u64, eb->len, PAGE_CACHE_SIZE); | |
459 | end = eb->start + end - 1; | |
ce9adaa5 CM |
460 | err: |
461 | free_extent_buffer(eb); | |
462 | out: | |
f188591e | 463 | return ret; |
ce9adaa5 CM |
464 | } |
465 | ||
ce9adaa5 | 466 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
467 | { |
468 | struct end_io_wq *end_io_wq = bio->bi_private; | |
469 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 470 | |
ce9adaa5 | 471 | fs_info = end_io_wq->info; |
ce9adaa5 | 472 | end_io_wq->error = err; |
8b712842 CM |
473 | end_io_wq->work.func = end_workqueue_fn; |
474 | end_io_wq->work.flags = 0; | |
d20f7043 CM |
475 | |
476 | if (bio->bi_rw & (1 << BIO_RW)) { | |
cad321ad CM |
477 | if (end_io_wq->metadata) |
478 | btrfs_queue_worker(&fs_info->endio_meta_write_workers, | |
479 | &end_io_wq->work); | |
480 | else | |
481 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
482 | &end_io_wq->work); | |
d20f7043 CM |
483 | } else { |
484 | if (end_io_wq->metadata) | |
485 | btrfs_queue_worker(&fs_info->endio_meta_workers, | |
486 | &end_io_wq->work); | |
487 | else | |
488 | btrfs_queue_worker(&fs_info->endio_workers, | |
489 | &end_io_wq->work); | |
490 | } | |
ce9adaa5 CM |
491 | } |
492 | ||
22c59948 CM |
493 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
494 | int metadata) | |
0b86a832 | 495 | { |
ce9adaa5 | 496 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
497 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
498 | if (!end_io_wq) | |
499 | return -ENOMEM; | |
500 | ||
501 | end_io_wq->private = bio->bi_private; | |
502 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 503 | end_io_wq->info = info; |
ce9adaa5 CM |
504 | end_io_wq->error = 0; |
505 | end_io_wq->bio = bio; | |
22c59948 | 506 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
507 | |
508 | bio->bi_private = end_io_wq; | |
509 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
510 | return 0; |
511 | } | |
512 | ||
b64a2851 | 513 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 514 | { |
4854ddd0 CM |
515 | unsigned long limit = min_t(unsigned long, |
516 | info->workers.max_workers, | |
517 | info->fs_devices->open_devices); | |
518 | return 256 * limit; | |
519 | } | |
0986fe9e | 520 | |
4854ddd0 CM |
521 | int btrfs_congested_async(struct btrfs_fs_info *info, int iodone) |
522 | { | |
b64a2851 CM |
523 | return atomic_read(&info->nr_async_bios) > |
524 | btrfs_async_submit_limit(info); | |
0986fe9e CM |
525 | } |
526 | ||
4a69a410 CM |
527 | static void run_one_async_start(struct btrfs_work *work) |
528 | { | |
529 | struct btrfs_fs_info *fs_info; | |
530 | struct async_submit_bio *async; | |
531 | ||
532 | async = container_of(work, struct async_submit_bio, work); | |
533 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
534 | async->submit_bio_start(async->inode, async->rw, async->bio, | |
535 | async->mirror_num, async->bio_flags); | |
536 | } | |
537 | ||
538 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
539 | { |
540 | struct btrfs_fs_info *fs_info; | |
541 | struct async_submit_bio *async; | |
4854ddd0 | 542 | int limit; |
8b712842 CM |
543 | |
544 | async = container_of(work, struct async_submit_bio, work); | |
545 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 546 | |
b64a2851 | 547 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
548 | limit = limit * 2 / 3; |
549 | ||
8b712842 | 550 | atomic_dec(&fs_info->nr_async_submits); |
0986fe9e | 551 | |
b64a2851 CM |
552 | if (atomic_read(&fs_info->nr_async_submits) < limit && |
553 | waitqueue_active(&fs_info->async_submit_wait)) | |
4854ddd0 CM |
554 | wake_up(&fs_info->async_submit_wait); |
555 | ||
4a69a410 | 556 | async->submit_bio_done(async->inode, async->rw, async->bio, |
c8b97818 | 557 | async->mirror_num, async->bio_flags); |
4a69a410 CM |
558 | } |
559 | ||
560 | static void run_one_async_free(struct btrfs_work *work) | |
561 | { | |
562 | struct async_submit_bio *async; | |
563 | ||
564 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
565 | kfree(async); |
566 | } | |
567 | ||
44b8bd7e CM |
568 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
569 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 570 | unsigned long bio_flags, |
4a69a410 CM |
571 | extent_submit_bio_hook_t *submit_bio_start, |
572 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
573 | { |
574 | struct async_submit_bio *async; | |
575 | ||
576 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
577 | if (!async) | |
578 | return -ENOMEM; | |
579 | ||
580 | async->inode = inode; | |
581 | async->rw = rw; | |
582 | async->bio = bio; | |
583 | async->mirror_num = mirror_num; | |
4a69a410 CM |
584 | async->submit_bio_start = submit_bio_start; |
585 | async->submit_bio_done = submit_bio_done; | |
586 | ||
587 | async->work.func = run_one_async_start; | |
588 | async->work.ordered_func = run_one_async_done; | |
589 | async->work.ordered_free = run_one_async_free; | |
590 | ||
8b712842 | 591 | async->work.flags = 0; |
c8b97818 | 592 | async->bio_flags = bio_flags; |
8c8bee1d | 593 | |
cb03c743 | 594 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 CM |
595 | |
596 | if (rw & (1 << BIO_RW_SYNCIO)) | |
597 | btrfs_set_work_high_prio(&async->work); | |
598 | ||
8b712842 | 599 | btrfs_queue_worker(&fs_info->workers, &async->work); |
9473f16c | 600 | |
d397712b | 601 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
602 | atomic_read(&fs_info->nr_async_submits)) { |
603 | wait_event(fs_info->async_submit_wait, | |
604 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
605 | } | |
606 | ||
44b8bd7e CM |
607 | return 0; |
608 | } | |
609 | ||
ce3ed71a CM |
610 | static int btree_csum_one_bio(struct bio *bio) |
611 | { | |
612 | struct bio_vec *bvec = bio->bi_io_vec; | |
613 | int bio_index = 0; | |
614 | struct btrfs_root *root; | |
615 | ||
616 | WARN_ON(bio->bi_vcnt <= 0); | |
d397712b | 617 | while (bio_index < bio->bi_vcnt) { |
ce3ed71a CM |
618 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
619 | csum_dirty_buffer(root, bvec->bv_page); | |
620 | bio_index++; | |
621 | bvec++; | |
622 | } | |
623 | return 0; | |
624 | } | |
625 | ||
4a69a410 CM |
626 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
627 | struct bio *bio, int mirror_num, | |
628 | unsigned long bio_flags) | |
22c59948 | 629 | { |
8b712842 CM |
630 | /* |
631 | * when we're called for a write, we're already in the async | |
5443be45 | 632 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 633 | */ |
4a69a410 CM |
634 | btree_csum_one_bio(bio); |
635 | return 0; | |
636 | } | |
22c59948 | 637 | |
4a69a410 CM |
638 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
639 | int mirror_num, unsigned long bio_flags) | |
640 | { | |
8b712842 | 641 | /* |
4a69a410 CM |
642 | * when we're called for a write, we're already in the async |
643 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 644 | */ |
8b712842 | 645 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
646 | } |
647 | ||
44b8bd7e | 648 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
c8b97818 | 649 | int mirror_num, unsigned long bio_flags) |
44b8bd7e | 650 | { |
cad321ad CM |
651 | int ret; |
652 | ||
653 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, | |
654 | bio, 1); | |
655 | BUG_ON(ret); | |
656 | ||
44b8bd7e | 657 | if (!(rw & (1 << BIO_RW))) { |
4a69a410 CM |
658 | /* |
659 | * called for a read, do the setup so that checksum validation | |
660 | * can happen in the async kernel threads | |
661 | */ | |
4a69a410 | 662 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, |
6f3577bd | 663 | mirror_num, 0); |
44b8bd7e | 664 | } |
d313d7a3 | 665 | |
cad321ad CM |
666 | /* |
667 | * kthread helpers are used to submit writes so that checksumming | |
668 | * can happen in parallel across all CPUs | |
669 | */ | |
44b8bd7e | 670 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, |
c8b97818 | 671 | inode, rw, bio, mirror_num, 0, |
4a69a410 CM |
672 | __btree_submit_bio_start, |
673 | __btree_submit_bio_done); | |
44b8bd7e CM |
674 | } |
675 | ||
0da5468f CM |
676 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
677 | { | |
d1310b2e | 678 | struct extent_io_tree *tree; |
b9473439 CM |
679 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
680 | struct extent_buffer *eb; | |
681 | int was_dirty; | |
682 | ||
d1310b2e | 683 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
b9473439 CM |
684 | if (!(current->flags & PF_MEMALLOC)) { |
685 | return extent_write_full_page(tree, page, | |
686 | btree_get_extent, wbc); | |
687 | } | |
5443be45 | 688 | |
b9473439 CM |
689 | redirty_page_for_writepage(wbc, page); |
690 | eb = btrfs_find_tree_block(root, page_offset(page), | |
691 | PAGE_CACHE_SIZE); | |
692 | WARN_ON(!eb); | |
693 | ||
694 | was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags); | |
695 | if (!was_dirty) { | |
696 | spin_lock(&root->fs_info->delalloc_lock); | |
697 | root->fs_info->dirty_metadata_bytes += PAGE_CACHE_SIZE; | |
698 | spin_unlock(&root->fs_info->delalloc_lock); | |
5443be45 | 699 | } |
b9473439 CM |
700 | free_extent_buffer(eb); |
701 | ||
702 | unlock_page(page); | |
703 | return 0; | |
5f39d397 | 704 | } |
0da5468f CM |
705 | |
706 | static int btree_writepages(struct address_space *mapping, | |
707 | struct writeback_control *wbc) | |
708 | { | |
d1310b2e CM |
709 | struct extent_io_tree *tree; |
710 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 711 | if (wbc->sync_mode == WB_SYNC_NONE) { |
b9473439 | 712 | struct btrfs_root *root = BTRFS_I(mapping->host)->root; |
793955bc | 713 | u64 num_dirty; |
24ab9cd8 | 714 | unsigned long thresh = 32 * 1024 * 1024; |
448d640b CM |
715 | |
716 | if (wbc->for_kupdate) | |
717 | return 0; | |
718 | ||
b9473439 CM |
719 | /* this is a bit racy, but that's ok */ |
720 | num_dirty = root->fs_info->dirty_metadata_bytes; | |
d397712b | 721 | if (num_dirty < thresh) |
793955bc | 722 | return 0; |
793955bc | 723 | } |
0da5468f CM |
724 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
725 | } | |
726 | ||
b2950863 | 727 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 728 | { |
d1310b2e CM |
729 | struct extent_io_tree *tree; |
730 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
731 | return extent_read_full_page(tree, page, btree_get_extent); |
732 | } | |
22b0ebda | 733 | |
70dec807 | 734 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 735 | { |
d1310b2e CM |
736 | struct extent_io_tree *tree; |
737 | struct extent_map_tree *map; | |
5f39d397 | 738 | int ret; |
d98237b3 | 739 | |
98509cfc | 740 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 741 | return 0; |
98509cfc | 742 | |
d1310b2e CM |
743 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
744 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
6af118ce | 745 | |
7b13b7b1 | 746 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
d397712b | 747 | if (!ret) |
6af118ce | 748 | return 0; |
6af118ce CM |
749 | |
750 | ret = try_release_extent_buffer(tree, page); | |
5f39d397 CM |
751 | if (ret == 1) { |
752 | ClearPagePrivate(page); | |
753 | set_page_private(page, 0); | |
754 | page_cache_release(page); | |
755 | } | |
6af118ce | 756 | |
d98237b3 CM |
757 | return ret; |
758 | } | |
759 | ||
5f39d397 | 760 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 761 | { |
d1310b2e CM |
762 | struct extent_io_tree *tree; |
763 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
764 | extent_invalidatepage(tree, page, offset); |
765 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 766 | if (PagePrivate(page)) { |
d397712b CM |
767 | printk(KERN_WARNING "btrfs warning page private not zero " |
768 | "on page %llu\n", (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
769 | ClearPagePrivate(page); |
770 | set_page_private(page, 0); | |
771 | page_cache_release(page); | |
772 | } | |
d98237b3 CM |
773 | } |
774 | ||
d98237b3 CM |
775 | static struct address_space_operations btree_aops = { |
776 | .readpage = btree_readpage, | |
777 | .writepage = btree_writepage, | |
0da5468f | 778 | .writepages = btree_writepages, |
5f39d397 CM |
779 | .releasepage = btree_releasepage, |
780 | .invalidatepage = btree_invalidatepage, | |
d98237b3 CM |
781 | .sync_page = block_sync_page, |
782 | }; | |
783 | ||
ca7a79ad CM |
784 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
785 | u64 parent_transid) | |
090d1875 | 786 | { |
5f39d397 CM |
787 | struct extent_buffer *buf = NULL; |
788 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 789 | int ret = 0; |
090d1875 | 790 | |
db94535d | 791 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 792 | if (!buf) |
090d1875 | 793 | return 0; |
d1310b2e | 794 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
f188591e | 795 | buf, 0, 0, btree_get_extent, 0); |
5f39d397 | 796 | free_extent_buffer(buf); |
de428b63 | 797 | return ret; |
090d1875 CM |
798 | } |
799 | ||
0999df54 CM |
800 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
801 | u64 bytenr, u32 blocksize) | |
802 | { | |
803 | struct inode *btree_inode = root->fs_info->btree_inode; | |
804 | struct extent_buffer *eb; | |
805 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
806 | bytenr, blocksize, GFP_NOFS); | |
807 | return eb; | |
808 | } | |
809 | ||
810 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
811 | u64 bytenr, u32 blocksize) | |
812 | { | |
813 | struct inode *btree_inode = root->fs_info->btree_inode; | |
814 | struct extent_buffer *eb; | |
815 | ||
816 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
817 | bytenr, blocksize, NULL, GFP_NOFS); | |
818 | return eb; | |
819 | } | |
820 | ||
821 | ||
e02119d5 CM |
822 | int btrfs_write_tree_block(struct extent_buffer *buf) |
823 | { | |
824 | return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start, | |
24ab9cd8 | 825 | buf->start + buf->len - 1, WB_SYNC_ALL); |
e02119d5 CM |
826 | } |
827 | ||
828 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
829 | { | |
830 | return btrfs_wait_on_page_writeback_range(buf->first_page->mapping, | |
d397712b | 831 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
832 | } |
833 | ||
0999df54 | 834 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 835 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
836 | { |
837 | struct extent_buffer *buf = NULL; | |
838 | struct inode *btree_inode = root->fs_info->btree_inode; | |
839 | struct extent_io_tree *io_tree; | |
840 | int ret; | |
841 | ||
842 | io_tree = &BTRFS_I(btree_inode)->io_tree; | |
843 | ||
844 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
845 | if (!buf) | |
846 | return NULL; | |
0999df54 | 847 | |
ca7a79ad | 848 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 | 849 | |
d397712b | 850 | if (ret == 0) |
b4ce94de | 851 | set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags); |
5f39d397 | 852 | return buf; |
ce9adaa5 | 853 | |
eb60ceac CM |
854 | } |
855 | ||
e089f05c | 856 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 857 | struct extent_buffer *buf) |
ed2ff2cb | 858 | { |
5f39d397 | 859 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 | 860 | if (btrfs_header_generation(buf) == |
925baedd | 861 | root->fs_info->running_transaction->transid) { |
b9447ef8 | 862 | btrfs_assert_tree_locked(buf); |
b4ce94de | 863 | |
b9473439 CM |
864 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
865 | spin_lock(&root->fs_info->delalloc_lock); | |
866 | if (root->fs_info->dirty_metadata_bytes >= buf->len) | |
867 | root->fs_info->dirty_metadata_bytes -= buf->len; | |
868 | else | |
869 | WARN_ON(1); | |
870 | spin_unlock(&root->fs_info->delalloc_lock); | |
871 | } | |
b4ce94de | 872 | |
b9473439 CM |
873 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
874 | btrfs_set_lock_blocking(buf); | |
d1310b2e | 875 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 876 | buf); |
925baedd | 877 | } |
5f39d397 CM |
878 | return 0; |
879 | } | |
880 | ||
db94535d | 881 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 882 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 883 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 884 | u64 objectid) |
d97e63b6 | 885 | { |
cfaa7295 | 886 | root->node = NULL; |
a28ec197 | 887 | root->commit_root = NULL; |
db94535d CM |
888 | root->sectorsize = sectorsize; |
889 | root->nodesize = nodesize; | |
890 | root->leafsize = leafsize; | |
87ee04eb | 891 | root->stripesize = stripesize; |
123abc88 | 892 | root->ref_cows = 0; |
0b86a832 CM |
893 | root->track_dirty = 0; |
894 | ||
9f5fae2f | 895 | root->fs_info = fs_info; |
0f7d52f4 CM |
896 | root->objectid = objectid; |
897 | root->last_trans = 0; | |
1b05da2e CM |
898 | root->highest_inode = 0; |
899 | root->last_inode_alloc = 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); | |
955 | BUG_ON(ret); | |
956 | ||
84234f3a | 957 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
958 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
959 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 960 | blocksize, generation); |
5d4f98a2 | 961 | root->commit_root = btrfs_root_node(root); |
3768f368 | 962 | BUG_ON(!root->node); |
d97e63b6 CM |
963 | return 0; |
964 | } | |
965 | ||
e02119d5 CM |
966 | int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans, |
967 | struct btrfs_fs_info *fs_info) | |
968 | { | |
969 | struct extent_buffer *eb; | |
d0c803c4 CM |
970 | struct btrfs_root *log_root_tree = fs_info->log_root_tree; |
971 | u64 start = 0; | |
972 | u64 end = 0; | |
e02119d5 CM |
973 | int ret; |
974 | ||
d0c803c4 | 975 | if (!log_root_tree) |
e02119d5 CM |
976 | return 0; |
977 | ||
d397712b | 978 | while (1) { |
d0c803c4 CM |
979 | ret = find_first_extent_bit(&log_root_tree->dirty_log_pages, |
980 | 0, &start, &end, EXTENT_DIRTY); | |
981 | if (ret) | |
982 | break; | |
983 | ||
984 | clear_extent_dirty(&log_root_tree->dirty_log_pages, | |
985 | start, end, GFP_NOFS); | |
986 | } | |
e02119d5 CM |
987 | eb = fs_info->log_root_tree->node; |
988 | ||
989 | WARN_ON(btrfs_header_level(eb) != 0); | |
990 | WARN_ON(btrfs_header_nritems(eb) != 0); | |
991 | ||
d00aff00 CM |
992 | ret = btrfs_free_reserved_extent(fs_info->tree_root, |
993 | eb->start, eb->len); | |
e02119d5 CM |
994 | BUG_ON(ret); |
995 | ||
996 | free_extent_buffer(eb); | |
997 | kfree(fs_info->log_root_tree); | |
998 | fs_info->log_root_tree = NULL; | |
999 | return 0; | |
1000 | } | |
1001 | ||
7237f183 YZ |
1002 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1003 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1004 | { |
1005 | struct btrfs_root *root; | |
1006 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1007 | struct extent_buffer *leaf; |
e02119d5 CM |
1008 | |
1009 | root = kzalloc(sizeof(*root), GFP_NOFS); | |
1010 | if (!root) | |
7237f183 | 1011 | return ERR_PTR(-ENOMEM); |
e02119d5 CM |
1012 | |
1013 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1014 | tree_root->sectorsize, tree_root->stripesize, | |
1015 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1016 | ||
1017 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1018 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1019 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
7237f183 YZ |
1020 | /* |
1021 | * log trees do not get reference counted because they go away | |
1022 | * before a real commit is actually done. They do store pointers | |
1023 | * to file data extents, and those reference counts still get | |
1024 | * updated (along with back refs to the log tree). | |
1025 | */ | |
e02119d5 CM |
1026 | root->ref_cows = 0; |
1027 | ||
5d4f98a2 YZ |
1028 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, |
1029 | BTRFS_TREE_LOG_OBJECTID, NULL, 0, 0, 0); | |
7237f183 YZ |
1030 | if (IS_ERR(leaf)) { |
1031 | kfree(root); | |
1032 | return ERR_CAST(leaf); | |
1033 | } | |
e02119d5 | 1034 | |
5d4f98a2 YZ |
1035 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1036 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1037 | btrfs_set_header_generation(leaf, trans->transid); | |
1038 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1039 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1040 | root->node = leaf; |
e02119d5 CM |
1041 | |
1042 | write_extent_buffer(root->node, root->fs_info->fsid, | |
1043 | (unsigned long)btrfs_header_fsid(root->node), | |
1044 | BTRFS_FSID_SIZE); | |
1045 | btrfs_mark_buffer_dirty(root->node); | |
1046 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1047 | return root; |
1048 | } | |
1049 | ||
1050 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1051 | struct btrfs_fs_info *fs_info) | |
1052 | { | |
1053 | struct btrfs_root *log_root; | |
1054 | ||
1055 | log_root = alloc_log_tree(trans, fs_info); | |
1056 | if (IS_ERR(log_root)) | |
1057 | return PTR_ERR(log_root); | |
1058 | WARN_ON(fs_info->log_root_tree); | |
1059 | fs_info->log_root_tree = log_root; | |
1060 | return 0; | |
1061 | } | |
1062 | ||
1063 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1064 | struct btrfs_root *root) | |
1065 | { | |
1066 | struct btrfs_root *log_root; | |
1067 | struct btrfs_inode_item *inode_item; | |
1068 | ||
1069 | log_root = alloc_log_tree(trans, root->fs_info); | |
1070 | if (IS_ERR(log_root)) | |
1071 | return PTR_ERR(log_root); | |
1072 | ||
1073 | log_root->last_trans = trans->transid; | |
1074 | log_root->root_key.offset = root->root_key.objectid; | |
1075 | ||
1076 | inode_item = &log_root->root_item.inode; | |
1077 | inode_item->generation = cpu_to_le64(1); | |
1078 | inode_item->size = cpu_to_le64(3); | |
1079 | inode_item->nlink = cpu_to_le32(1); | |
1080 | inode_item->nbytes = cpu_to_le64(root->leafsize); | |
1081 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); | |
1082 | ||
5d4f98a2 | 1083 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1084 | |
1085 | WARN_ON(root->log_root); | |
1086 | root->log_root = log_root; | |
1087 | root->log_transid = 0; | |
e02119d5 CM |
1088 | return 0; |
1089 | } | |
1090 | ||
1091 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
1092 | struct btrfs_key *location) | |
1093 | { | |
1094 | struct btrfs_root *root; | |
1095 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1096 | struct btrfs_path *path; |
5f39d397 | 1097 | struct extent_buffer *l; |
1b05da2e | 1098 | u64 highest_inode; |
84234f3a | 1099 | u64 generation; |
db94535d | 1100 | u32 blocksize; |
0f7d52f4 CM |
1101 | int ret = 0; |
1102 | ||
5eda7b5e | 1103 | root = kzalloc(sizeof(*root), GFP_NOFS); |
0cf6c620 | 1104 | if (!root) |
0f7d52f4 | 1105 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 1106 | if (location->offset == (u64)-1) { |
db94535d | 1107 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
1108 | location->objectid, root); |
1109 | if (ret) { | |
0f7d52f4 CM |
1110 | kfree(root); |
1111 | return ERR_PTR(ret); | |
1112 | } | |
1113 | goto insert; | |
1114 | } | |
1115 | ||
db94535d | 1116 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1117 | tree_root->sectorsize, tree_root->stripesize, |
1118 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
1119 | |
1120 | path = btrfs_alloc_path(); | |
1121 | BUG_ON(!path); | |
1122 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); | |
1123 | if (ret != 0) { | |
0f7d52f4 CM |
1124 | if (ret > 0) |
1125 | ret = -ENOENT; | |
1126 | goto out; | |
1127 | } | |
5f39d397 CM |
1128 | l = path->nodes[0]; |
1129 | read_extent_buffer(l, &root->root_item, | |
1130 | btrfs_item_ptr_offset(l, path->slots[0]), | |
0f7d52f4 | 1131 | sizeof(root->root_item)); |
44b36eb2 | 1132 | memcpy(&root->root_key, location, sizeof(*location)); |
0f7d52f4 CM |
1133 | ret = 0; |
1134 | out: | |
1135 | btrfs_release_path(root, path); | |
1136 | btrfs_free_path(path); | |
1137 | if (ret) { | |
1138 | kfree(root); | |
1139 | return ERR_PTR(ret); | |
1140 | } | |
84234f3a | 1141 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1142 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1143 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1144 | blocksize, generation); |
5d4f98a2 | 1145 | root->commit_root = btrfs_root_node(root); |
0f7d52f4 CM |
1146 | BUG_ON(!root->node); |
1147 | insert: | |
e02119d5 CM |
1148 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { |
1149 | root->ref_cows = 1; | |
1150 | ret = btrfs_find_highest_inode(root, &highest_inode); | |
1151 | if (ret == 0) { | |
1152 | root->highest_inode = highest_inode; | |
1153 | root->last_inode_alloc = highest_inode; | |
1154 | } | |
5eda7b5e CM |
1155 | } |
1156 | return root; | |
1157 | } | |
1158 | ||
dc17ff8f CM |
1159 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1160 | u64 root_objectid) | |
1161 | { | |
1162 | struct btrfs_root *root; | |
1163 | ||
1164 | if (root_objectid == BTRFS_ROOT_TREE_OBJECTID) | |
1165 | return fs_info->tree_root; | |
1166 | if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1167 | return fs_info->extent_root; | |
1168 | ||
1169 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1170 | (unsigned long)root_objectid); | |
1171 | return root; | |
1172 | } | |
1173 | ||
edbd8d4e CM |
1174 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
1175 | struct btrfs_key *location) | |
5eda7b5e CM |
1176 | { |
1177 | struct btrfs_root *root; | |
1178 | int ret; | |
1179 | ||
edbd8d4e CM |
1180 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1181 | return fs_info->tree_root; | |
1182 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1183 | return fs_info->extent_root; | |
8f18cf13 CM |
1184 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1185 | return fs_info->chunk_root; | |
1186 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1187 | return fs_info->dev_root; | |
0403e47e YZ |
1188 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1189 | return fs_info->csum_root; | |
edbd8d4e | 1190 | |
5eda7b5e CM |
1191 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
1192 | (unsigned long)location->objectid); | |
1193 | if (root) | |
1194 | return root; | |
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 CM |
1199 | |
1200 | set_anon_super(&root->anon_super, NULL); | |
1201 | ||
2619ba1f CM |
1202 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
1203 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 CM |
1204 | root); |
1205 | if (ret) { | |
5f39d397 | 1206 | free_extent_buffer(root->node); |
0f7d52f4 CM |
1207 | kfree(root); |
1208 | return ERR_PTR(ret); | |
1209 | } | |
c146afad YZ |
1210 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
1211 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
5d4f98a2 | 1212 | root->root_key.objectid); |
c146afad YZ |
1213 | BUG_ON(ret); |
1214 | btrfs_orphan_cleanup(root); | |
1215 | } | |
edbd8d4e CM |
1216 | return root; |
1217 | } | |
1218 | ||
1219 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, | |
1220 | struct btrfs_key *location, | |
1221 | const char *name, int namelen) | |
1222 | { | |
1223 | struct btrfs_root *root; | |
1224 | int ret; | |
1225 | ||
1226 | root = btrfs_read_fs_root_no_name(fs_info, location); | |
1227 | if (!root) | |
1228 | return NULL; | |
58176a96 | 1229 | |
4313b399 CM |
1230 | if (root->in_sysfs) |
1231 | return root; | |
1232 | ||
58176a96 JB |
1233 | ret = btrfs_set_root_name(root, name, namelen); |
1234 | if (ret) { | |
5f39d397 | 1235 | free_extent_buffer(root->node); |
58176a96 JB |
1236 | kfree(root); |
1237 | return ERR_PTR(ret); | |
1238 | } | |
ea9e8b11 | 1239 | #if 0 |
58176a96 JB |
1240 | ret = btrfs_sysfs_add_root(root); |
1241 | if (ret) { | |
5f39d397 | 1242 | free_extent_buffer(root->node); |
58176a96 JB |
1243 | kfree(root->name); |
1244 | kfree(root); | |
1245 | return ERR_PTR(ret); | |
1246 | } | |
ea9e8b11 | 1247 | #endif |
4313b399 | 1248 | root->in_sysfs = 1; |
0f7d52f4 CM |
1249 | return root; |
1250 | } | |
04160088 CM |
1251 | |
1252 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) | |
1253 | { | |
1254 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1255 | int ret = 0; | |
04160088 CM |
1256 | struct btrfs_device *device; |
1257 | struct backing_dev_info *bdi; | |
b7967db7 | 1258 | |
c6e30871 | 1259 | list_for_each_entry(device, &info->fs_devices->devices, dev_list) { |
dfe25020 CM |
1260 | if (!device->bdev) |
1261 | continue; | |
04160088 CM |
1262 | bdi = blk_get_backing_dev_info(device->bdev); |
1263 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1264 | ret = 1; | |
1265 | break; | |
1266 | } | |
1267 | } | |
1268 | return ret; | |
1269 | } | |
1270 | ||
38b66988 CM |
1271 | /* |
1272 | * this unplugs every device on the box, and it is only used when page | |
1273 | * is null | |
1274 | */ | |
1275 | static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) | |
1276 | { | |
38b66988 CM |
1277 | struct btrfs_device *device; |
1278 | struct btrfs_fs_info *info; | |
1279 | ||
1280 | info = (struct btrfs_fs_info *)bdi->unplug_io_data; | |
c6e30871 | 1281 | list_for_each_entry(device, &info->fs_devices->devices, dev_list) { |
d20f7043 CM |
1282 | if (!device->bdev) |
1283 | continue; | |
1284 | ||
38b66988 | 1285 | bdi = blk_get_backing_dev_info(device->bdev); |
d397712b | 1286 | if (bdi->unplug_io_fn) |
38b66988 | 1287 | bdi->unplug_io_fn(bdi, page); |
38b66988 CM |
1288 | } |
1289 | } | |
1290 | ||
b2950863 | 1291 | static void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) |
04160088 | 1292 | { |
38b66988 | 1293 | struct inode *inode; |
f2d8d74d CM |
1294 | struct extent_map_tree *em_tree; |
1295 | struct extent_map *em; | |
bcbfce8a | 1296 | struct address_space *mapping; |
38b66988 CM |
1297 | u64 offset; |
1298 | ||
bcbfce8a | 1299 | /* the generic O_DIRECT read code does this */ |
9f0ba5bd | 1300 | if (1 || !page) { |
38b66988 CM |
1301 | __unplug_io_fn(bdi, page); |
1302 | return; | |
1303 | } | |
1304 | ||
bcbfce8a CM |
1305 | /* |
1306 | * page->mapping may change at any time. Get a consistent copy | |
1307 | * and use that for everything below | |
1308 | */ | |
1309 | smp_mb(); | |
1310 | mapping = page->mapping; | |
1311 | if (!mapping) | |
1312 | return; | |
1313 | ||
1314 | inode = mapping->host; | |
240d5d48 CM |
1315 | |
1316 | /* | |
1317 | * don't do the expensive searching for a small number of | |
1318 | * devices | |
1319 | */ | |
1320 | if (BTRFS_I(inode)->root->fs_info->fs_devices->open_devices <= 2) { | |
1321 | __unplug_io_fn(bdi, page); | |
1322 | return; | |
1323 | } | |
1324 | ||
38b66988 | 1325 | offset = page_offset(page); |
04160088 | 1326 | |
f2d8d74d | 1327 | em_tree = &BTRFS_I(inode)->extent_tree; |
890871be | 1328 | read_lock(&em_tree->lock); |
f2d8d74d | 1329 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); |
890871be | 1330 | read_unlock(&em_tree->lock); |
89642229 CM |
1331 | if (!em) { |
1332 | __unplug_io_fn(bdi, page); | |
f2d8d74d | 1333 | return; |
89642229 | 1334 | } |
f2d8d74d | 1335 | |
89642229 CM |
1336 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { |
1337 | free_extent_map(em); | |
1338 | __unplug_io_fn(bdi, page); | |
1339 | return; | |
1340 | } | |
f2d8d74d CM |
1341 | offset = offset - em->start; |
1342 | btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree, | |
1343 | em->block_start + offset, page); | |
1344 | free_extent_map(em); | |
04160088 CM |
1345 | } |
1346 | ||
ad081f14 JA |
1347 | /* |
1348 | * If this fails, caller must call bdi_destroy() to get rid of the | |
1349 | * bdi again. | |
1350 | */ | |
04160088 CM |
1351 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1352 | { | |
ad081f14 JA |
1353 | int err; |
1354 | ||
1355 | bdi->capabilities = BDI_CAP_MAP_COPY; | |
1356 | err = bdi_init(bdi); | |
1357 | if (err) | |
1358 | return err; | |
1359 | ||
1360 | err = bdi_register(bdi, NULL, "btrfs-%d", | |
1361 | atomic_inc_return(&btrfs_bdi_num)); | |
1362 | if (err) | |
1363 | return err; | |
1364 | ||
4575c9cc | 1365 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1366 | bdi->unplug_io_fn = btrfs_unplug_io_fn; |
1367 | bdi->unplug_io_data = info; | |
1368 | bdi->congested_fn = btrfs_congested_fn; | |
1369 | bdi->congested_data = info; | |
1370 | return 0; | |
1371 | } | |
1372 | ||
ce9adaa5 CM |
1373 | static int bio_ready_for_csum(struct bio *bio) |
1374 | { | |
1375 | u64 length = 0; | |
1376 | u64 buf_len = 0; | |
1377 | u64 start = 0; | |
1378 | struct page *page; | |
1379 | struct extent_io_tree *io_tree = NULL; | |
1380 | struct btrfs_fs_info *info = NULL; | |
1381 | struct bio_vec *bvec; | |
1382 | int i; | |
1383 | int ret; | |
1384 | ||
1385 | bio_for_each_segment(bvec, bio, i) { | |
1386 | page = bvec->bv_page; | |
1387 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
1388 | length += bvec->bv_len; | |
1389 | continue; | |
1390 | } | |
1391 | if (!page->private) { | |
1392 | length += bvec->bv_len; | |
1393 | continue; | |
1394 | } | |
1395 | length = bvec->bv_len; | |
1396 | buf_len = page->private >> 2; | |
1397 | start = page_offset(page) + bvec->bv_offset; | |
1398 | io_tree = &BTRFS_I(page->mapping->host)->io_tree; | |
1399 | info = BTRFS_I(page->mapping->host)->root->fs_info; | |
1400 | } | |
1401 | /* are we fully contained in this bio? */ | |
1402 | if (buf_len <= length) | |
1403 | return 1; | |
1404 | ||
1405 | ret = extent_range_uptodate(io_tree, start + length, | |
1406 | start + buf_len - 1); | |
ce9adaa5 CM |
1407 | return ret; |
1408 | } | |
1409 | ||
8b712842 CM |
1410 | /* |
1411 | * called by the kthread helper functions to finally call the bio end_io | |
1412 | * functions. This is where read checksum verification actually happens | |
1413 | */ | |
1414 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1415 | { |
ce9adaa5 | 1416 | struct bio *bio; |
8b712842 CM |
1417 | struct end_io_wq *end_io_wq; |
1418 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1419 | int error; |
ce9adaa5 | 1420 | |
8b712842 CM |
1421 | end_io_wq = container_of(work, struct end_io_wq, work); |
1422 | bio = end_io_wq->bio; | |
1423 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1424 | |
cad321ad | 1425 | /* metadata bio reads are special because the whole tree block must |
8b712842 CM |
1426 | * be checksummed at once. This makes sure the entire block is in |
1427 | * ram and up to date before trying to verify things. For | |
1428 | * blocksize <= pagesize, it is basically a noop | |
1429 | */ | |
cad321ad CM |
1430 | if (!(bio->bi_rw & (1 << BIO_RW)) && end_io_wq->metadata && |
1431 | !bio_ready_for_csum(bio)) { | |
d20f7043 | 1432 | btrfs_queue_worker(&fs_info->endio_meta_workers, |
8b712842 CM |
1433 | &end_io_wq->work); |
1434 | return; | |
1435 | } | |
1436 | error = end_io_wq->error; | |
1437 | bio->bi_private = end_io_wq->private; | |
1438 | bio->bi_end_io = end_io_wq->end_io; | |
1439 | kfree(end_io_wq); | |
8b712842 | 1440 | bio_endio(bio, error); |
44b8bd7e CM |
1441 | } |
1442 | ||
a74a4b97 CM |
1443 | static int cleaner_kthread(void *arg) |
1444 | { | |
1445 | struct btrfs_root *root = arg; | |
1446 | ||
1447 | do { | |
1448 | smp_mb(); | |
1449 | if (root->fs_info->closing) | |
1450 | break; | |
1451 | ||
1452 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1453 | mutex_lock(&root->fs_info->cleaner_mutex); | |
a74a4b97 | 1454 | btrfs_clean_old_snapshots(root); |
a74a4b97 CM |
1455 | mutex_unlock(&root->fs_info->cleaner_mutex); |
1456 | ||
1457 | if (freezing(current)) { | |
1458 | refrigerator(); | |
1459 | } else { | |
1460 | smp_mb(); | |
1461 | if (root->fs_info->closing) | |
1462 | break; | |
1463 | set_current_state(TASK_INTERRUPTIBLE); | |
1464 | schedule(); | |
1465 | __set_current_state(TASK_RUNNING); | |
1466 | } | |
1467 | } while (!kthread_should_stop()); | |
1468 | return 0; | |
1469 | } | |
1470 | ||
1471 | static int transaction_kthread(void *arg) | |
1472 | { | |
1473 | struct btrfs_root *root = arg; | |
1474 | struct btrfs_trans_handle *trans; | |
1475 | struct btrfs_transaction *cur; | |
1476 | unsigned long now; | |
1477 | unsigned long delay; | |
1478 | int ret; | |
1479 | ||
1480 | do { | |
1481 | smp_mb(); | |
1482 | if (root->fs_info->closing) | |
1483 | break; | |
1484 | ||
1485 | delay = HZ * 30; | |
1486 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1487 | mutex_lock(&root->fs_info->transaction_kthread_mutex); | |
1488 | ||
1489 | mutex_lock(&root->fs_info->trans_mutex); | |
1490 | cur = root->fs_info->running_transaction; | |
1491 | if (!cur) { | |
1492 | mutex_unlock(&root->fs_info->trans_mutex); | |
1493 | goto sleep; | |
1494 | } | |
31153d81 | 1495 | |
a74a4b97 CM |
1496 | now = get_seconds(); |
1497 | if (now < cur->start_time || now - cur->start_time < 30) { | |
1498 | mutex_unlock(&root->fs_info->trans_mutex); | |
1499 | delay = HZ * 5; | |
1500 | goto sleep; | |
1501 | } | |
1502 | mutex_unlock(&root->fs_info->trans_mutex); | |
a74a4b97 CM |
1503 | trans = btrfs_start_transaction(root, 1); |
1504 | ret = btrfs_commit_transaction(trans, root); | |
56bec294 | 1505 | |
a74a4b97 CM |
1506 | sleep: |
1507 | wake_up_process(root->fs_info->cleaner_kthread); | |
1508 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1509 | ||
1510 | if (freezing(current)) { | |
1511 | refrigerator(); | |
1512 | } else { | |
1513 | if (root->fs_info->closing) | |
1514 | break; | |
1515 | set_current_state(TASK_INTERRUPTIBLE); | |
1516 | schedule_timeout(delay); | |
1517 | __set_current_state(TASK_RUNNING); | |
1518 | } | |
1519 | } while (!kthread_should_stop()); | |
1520 | return 0; | |
1521 | } | |
1522 | ||
8a4b83cc | 1523 | struct btrfs_root *open_ctree(struct super_block *sb, |
dfe25020 CM |
1524 | struct btrfs_fs_devices *fs_devices, |
1525 | char *options) | |
2e635a27 | 1526 | { |
db94535d CM |
1527 | u32 sectorsize; |
1528 | u32 nodesize; | |
1529 | u32 leafsize; | |
1530 | u32 blocksize; | |
87ee04eb | 1531 | u32 stripesize; |
84234f3a | 1532 | u64 generation; |
f2b636e8 | 1533 | u64 features; |
3de4586c | 1534 | struct btrfs_key location; |
a061fc8d | 1535 | struct buffer_head *bh; |
e02119d5 | 1536 | struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1537 | GFP_NOFS); |
d20f7043 CM |
1538 | struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root), |
1539 | GFP_NOFS); | |
e02119d5 | 1540 | struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1541 | GFP_NOFS); |
8790d502 | 1542 | struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), |
e20d96d6 | 1543 | GFP_NOFS); |
e02119d5 | 1544 | struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1545 | GFP_NOFS); |
e02119d5 | 1546 | struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1547 | GFP_NOFS); |
e02119d5 CM |
1548 | struct btrfs_root *log_tree_root; |
1549 | ||
eb60ceac | 1550 | int ret; |
e58ca020 | 1551 | int err = -EINVAL; |
4543df7e | 1552 | |
2c90e5d6 | 1553 | struct btrfs_super_block *disk_super; |
8790d502 | 1554 | |
0463bb4e | 1555 | if (!extent_root || !tree_root || !fs_info || |
d20f7043 | 1556 | !chunk_root || !dev_root || !csum_root) { |
39279cc3 CM |
1557 | err = -ENOMEM; |
1558 | goto fail; | |
1559 | } | |
0f7d52f4 | 1560 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS); |
8fd17795 | 1561 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1562 | INIT_LIST_HEAD(&fs_info->dead_roots); |
19c00ddc | 1563 | INIT_LIST_HEAD(&fs_info->hashers); |
ea8c2819 | 1564 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
5a3f23d5 | 1565 | INIT_LIST_HEAD(&fs_info->ordered_operations); |
11833d66 | 1566 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
1832a6d5 | 1567 | spin_lock_init(&fs_info->delalloc_lock); |
cee36a03 | 1568 | spin_lock_init(&fs_info->new_trans_lock); |
31153d81 | 1569 | spin_lock_init(&fs_info->ref_cache_lock); |
19c00ddc | 1570 | |
58176a96 | 1571 | init_completion(&fs_info->kobj_unregister); |
9f5fae2f CM |
1572 | fs_info->tree_root = tree_root; |
1573 | fs_info->extent_root = extent_root; | |
d20f7043 | 1574 | fs_info->csum_root = csum_root; |
0b86a832 CM |
1575 | fs_info->chunk_root = chunk_root; |
1576 | fs_info->dev_root = dev_root; | |
8a4b83cc | 1577 | fs_info->fs_devices = fs_devices; |
0b86a832 | 1578 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1579 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1580 | btrfs_mapping_init(&fs_info->mapping_tree); |
cb03c743 | 1581 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 1582 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 1583 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 1584 | atomic_set(&fs_info->nr_async_bios, 0); |
e20d96d6 | 1585 | fs_info->sb = sb; |
c59f8951 | 1586 | fs_info->max_extent = (u64)-1; |
6f568d35 | 1587 | fs_info->max_inline = 8192 * 1024; |
ad081f14 JA |
1588 | if (setup_bdi(fs_info, &fs_info->bdi)) |
1589 | goto fail_bdi; | |
d98237b3 CM |
1590 | fs_info->btree_inode = new_inode(sb); |
1591 | fs_info->btree_inode->i_ino = 1; | |
2c90e5d6 | 1592 | fs_info->btree_inode->i_nlink = 1; |
97e728d4 | 1593 | fs_info->metadata_ratio = 8; |
c8b97818 | 1594 | |
b34b086c CM |
1595 | fs_info->thread_pool_size = min_t(unsigned long, |
1596 | num_online_cpus() + 2, 8); | |
0afbaf8c | 1597 | |
3eaa2885 CM |
1598 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
1599 | spin_lock_init(&fs_info->ordered_extent_lock); | |
1600 | ||
a061fc8d CM |
1601 | sb->s_blocksize = 4096; |
1602 | sb->s_blocksize_bits = blksize_bits(4096); | |
1603 | ||
0afbaf8c CM |
1604 | /* |
1605 | * we set the i_size on the btree inode to the max possible int. | |
1606 | * the real end of the address space is determined by all of | |
1607 | * the devices in the system | |
1608 | */ | |
1609 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 1610 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
1611 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
1612 | ||
5d4f98a2 | 1613 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); |
d1310b2e | 1614 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
5f39d397 CM |
1615 | fs_info->btree_inode->i_mapping, |
1616 | GFP_NOFS); | |
d1310b2e CM |
1617 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, |
1618 | GFP_NOFS); | |
1619 | ||
1620 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 1621 | |
0f9dd46c JB |
1622 | spin_lock_init(&fs_info->block_group_cache_lock); |
1623 | fs_info->block_group_cache_tree.rb_node = NULL; | |
1624 | ||
11833d66 | 1625 | extent_io_tree_init(&fs_info->freed_extents[0], |
1a5bc167 | 1626 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
11833d66 YZ |
1627 | extent_io_tree_init(&fs_info->freed_extents[1], |
1628 | fs_info->btree_inode->i_mapping, GFP_NOFS); | |
1629 | fs_info->pinned_extents = &fs_info->freed_extents[0]; | |
e66f709b | 1630 | fs_info->do_barriers = 1; |
e18e4809 | 1631 | |
0f7d52f4 CM |
1632 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
1633 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
1634 | sizeof(struct btrfs_key)); | |
22b0ebda | 1635 | insert_inode_hash(fs_info->btree_inode); |
39279cc3 | 1636 | |
79154b1b | 1637 | mutex_init(&fs_info->trans_mutex); |
5a3f23d5 | 1638 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 1639 | mutex_init(&fs_info->tree_log_mutex); |
a2135011 | 1640 | mutex_init(&fs_info->drop_mutex); |
925baedd | 1641 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
1642 | mutex_init(&fs_info->transaction_kthread_mutex); |
1643 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 1644 | mutex_init(&fs_info->volume_mutex); |
1a40e23b | 1645 | mutex_init(&fs_info->tree_reloc_mutex); |
276e680d | 1646 | init_rwsem(&fs_info->extent_commit_sem); |
fa9c0d79 CM |
1647 | |
1648 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); | |
1649 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
1650 | ||
e6dcd2dc | 1651 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 1652 | init_waitqueue_head(&fs_info->transaction_wait); |
4854ddd0 | 1653 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 1654 | |
0b86a832 | 1655 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 1656 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 1657 | |
d98237b3 | 1658 | |
a512bbf8 | 1659 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
a061fc8d | 1660 | if (!bh) |
39279cc3 | 1661 | goto fail_iput; |
39279cc3 | 1662 | |
a061fc8d | 1663 | memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy)); |
2d69a0f8 YZ |
1664 | memcpy(&fs_info->super_for_commit, &fs_info->super_copy, |
1665 | sizeof(fs_info->super_for_commit)); | |
a061fc8d | 1666 | brelse(bh); |
5f39d397 | 1667 | |
a061fc8d | 1668 | memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE); |
0b86a832 | 1669 | |
5f39d397 | 1670 | disk_super = &fs_info->super_copy; |
0f7d52f4 | 1671 | if (!btrfs_super_root(disk_super)) |
c6e2bac1 | 1672 | goto fail_iput; |
0f7d52f4 | 1673 | |
2b82032c YZ |
1674 | ret = btrfs_parse_options(tree_root, options); |
1675 | if (ret) { | |
1676 | err = ret; | |
c6e2bac1 | 1677 | goto fail_iput; |
2b82032c | 1678 | } |
dfe25020 | 1679 | |
f2b636e8 JB |
1680 | features = btrfs_super_incompat_flags(disk_super) & |
1681 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
1682 | if (features) { | |
1683 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
1684 | "unsupported optional features (%Lx).\n", | |
21380931 | 1685 | (unsigned long long)features); |
f2b636e8 | 1686 | err = -EINVAL; |
c6e2bac1 | 1687 | goto fail_iput; |
f2b636e8 JB |
1688 | } |
1689 | ||
5d4f98a2 YZ |
1690 | features = btrfs_super_incompat_flags(disk_super); |
1691 | if (!(features & BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF)) { | |
1692 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; | |
1693 | btrfs_set_super_incompat_flags(disk_super, features); | |
1694 | } | |
1695 | ||
f2b636e8 JB |
1696 | features = btrfs_super_compat_ro_flags(disk_super) & |
1697 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
1698 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
1699 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
1700 | "unsupported option features (%Lx).\n", | |
21380931 | 1701 | (unsigned long long)features); |
f2b636e8 | 1702 | err = -EINVAL; |
c6e2bac1 | 1703 | goto fail_iput; |
f2b636e8 | 1704 | } |
9042846b | 1705 | printk("thread pool is %d\n", fs_info->thread_pool_size); |
4543df7e CM |
1706 | /* |
1707 | * we need to start all the end_io workers up front because the | |
1708 | * queue work function gets called at interrupt time, and so it | |
1709 | * cannot dynamically grow. | |
1710 | */ | |
5443be45 CM |
1711 | btrfs_init_workers(&fs_info->workers, "worker", |
1712 | fs_info->thread_pool_size); | |
c8b97818 | 1713 | |
771ed689 CM |
1714 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
1715 | fs_info->thread_pool_size); | |
1716 | ||
5443be45 | 1717 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 CM |
1718 | min_t(u64, fs_devices->num_devices, |
1719 | fs_info->thread_pool_size)); | |
61b49440 CM |
1720 | |
1721 | /* a higher idle thresh on the submit workers makes it much more | |
1722 | * likely that bios will be send down in a sane order to the | |
1723 | * devices | |
1724 | */ | |
1725 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 1726 | |
771ed689 | 1727 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 1728 | fs_info->workers.ordered = 1; |
61b49440 | 1729 | |
771ed689 CM |
1730 | fs_info->delalloc_workers.idle_thresh = 2; |
1731 | fs_info->delalloc_workers.ordered = 1; | |
1732 | ||
5443be45 CM |
1733 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1); |
1734 | btrfs_init_workers(&fs_info->endio_workers, "endio", | |
1735 | fs_info->thread_pool_size); | |
d20f7043 CM |
1736 | btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", |
1737 | fs_info->thread_pool_size); | |
cad321ad CM |
1738 | btrfs_init_workers(&fs_info->endio_meta_write_workers, |
1739 | "endio-meta-write", fs_info->thread_pool_size); | |
5443be45 | 1740 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", |
e6dcd2dc | 1741 | fs_info->thread_pool_size); |
61b49440 CM |
1742 | |
1743 | /* | |
1744 | * endios are largely parallel and should have a very | |
1745 | * low idle thresh | |
1746 | */ | |
1747 | fs_info->endio_workers.idle_thresh = 4; | |
b51912c9 CM |
1748 | fs_info->endio_meta_workers.idle_thresh = 4; |
1749 | ||
9042846b CM |
1750 | fs_info->endio_write_workers.idle_thresh = 2; |
1751 | fs_info->endio_meta_write_workers.idle_thresh = 2; | |
1752 | ||
1753 | fs_info->endio_workers.atomic_worker_start = 1; | |
1754 | fs_info->endio_meta_workers.atomic_worker_start = 1; | |
1755 | fs_info->endio_write_workers.atomic_worker_start = 1; | |
1756 | fs_info->endio_meta_write_workers.atomic_worker_start = 1; | |
61b49440 | 1757 | |
4543df7e | 1758 | btrfs_start_workers(&fs_info->workers, 1); |
1cc127b5 | 1759 | btrfs_start_workers(&fs_info->submit_workers, 1); |
771ed689 | 1760 | btrfs_start_workers(&fs_info->delalloc_workers, 1); |
247e743c | 1761 | btrfs_start_workers(&fs_info->fixup_workers, 1); |
9042846b CM |
1762 | btrfs_start_workers(&fs_info->endio_workers, 1); |
1763 | btrfs_start_workers(&fs_info->endio_meta_workers, 1); | |
1764 | btrfs_start_workers(&fs_info->endio_meta_write_workers, 1); | |
1765 | btrfs_start_workers(&fs_info->endio_write_workers, 1); | |
4543df7e | 1766 | |
4575c9cc | 1767 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
1768 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
1769 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 1770 | |
db94535d CM |
1771 | nodesize = btrfs_super_nodesize(disk_super); |
1772 | leafsize = btrfs_super_leafsize(disk_super); | |
1773 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 1774 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
1775 | tree_root->nodesize = nodesize; |
1776 | tree_root->leafsize = leafsize; | |
1777 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 1778 | tree_root->stripesize = stripesize; |
a061fc8d CM |
1779 | |
1780 | sb->s_blocksize = sectorsize; | |
1781 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 1782 | |
39279cc3 CM |
1783 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
1784 | sizeof(disk_super->magic))) { | |
d397712b | 1785 | printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
1786 | goto fail_sb_buffer; |
1787 | } | |
19c00ddc | 1788 | |
925baedd | 1789 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 1790 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 1791 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 1792 | if (ret) { |
d397712b CM |
1793 | printk(KERN_WARNING "btrfs: failed to read the system " |
1794 | "array on %s\n", sb->s_id); | |
5d4f98a2 | 1795 | goto fail_sb_buffer; |
84eed90f | 1796 | } |
0b86a832 CM |
1797 | |
1798 | blocksize = btrfs_level_size(tree_root, | |
1799 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 1800 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
1801 | |
1802 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1803 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
1804 | ||
1805 | chunk_root->node = read_tree_block(chunk_root, | |
1806 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 1807 | blocksize, generation); |
0b86a832 | 1808 | BUG_ON(!chunk_root->node); |
83121942 DW |
1809 | if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { |
1810 | printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n", | |
1811 | sb->s_id); | |
1812 | goto fail_chunk_root; | |
1813 | } | |
5d4f98a2 YZ |
1814 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
1815 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 1816 | |
e17cade2 | 1817 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
d397712b CM |
1818 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), |
1819 | BTRFS_UUID_SIZE); | |
e17cade2 | 1820 | |
925baedd | 1821 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1822 | ret = btrfs_read_chunk_tree(chunk_root); |
925baedd | 1823 | mutex_unlock(&fs_info->chunk_mutex); |
2b82032c | 1824 | if (ret) { |
d397712b CM |
1825 | printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", |
1826 | sb->s_id); | |
2b82032c YZ |
1827 | goto fail_chunk_root; |
1828 | } | |
0b86a832 | 1829 | |
dfe25020 CM |
1830 | btrfs_close_extra_devices(fs_devices); |
1831 | ||
db94535d CM |
1832 | blocksize = btrfs_level_size(tree_root, |
1833 | btrfs_super_root_level(disk_super)); | |
84234f3a | 1834 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 1835 | |
e20d96d6 | 1836 | tree_root->node = read_tree_block(tree_root, |
db94535d | 1837 | btrfs_super_root(disk_super), |
84234f3a | 1838 | blocksize, generation); |
39279cc3 | 1839 | if (!tree_root->node) |
2b82032c | 1840 | goto fail_chunk_root; |
83121942 DW |
1841 | if (!test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { |
1842 | printk(KERN_WARNING "btrfs: failed to read tree root on %s\n", | |
1843 | sb->s_id); | |
1844 | goto fail_tree_root; | |
1845 | } | |
5d4f98a2 YZ |
1846 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
1847 | tree_root->commit_root = btrfs_root_node(tree_root); | |
db94535d CM |
1848 | |
1849 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 1850 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 1851 | if (ret) |
39279cc3 | 1852 | goto fail_tree_root; |
0b86a832 CM |
1853 | extent_root->track_dirty = 1; |
1854 | ||
1855 | ret = find_and_setup_root(tree_root, fs_info, | |
1856 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
0b86a832 CM |
1857 | if (ret) |
1858 | goto fail_extent_root; | |
5d4f98a2 | 1859 | dev_root->track_dirty = 1; |
3768f368 | 1860 | |
d20f7043 CM |
1861 | ret = find_and_setup_root(tree_root, fs_info, |
1862 | BTRFS_CSUM_TREE_OBJECTID, csum_root); | |
1863 | if (ret) | |
5d4f98a2 | 1864 | goto fail_dev_root; |
d20f7043 CM |
1865 | |
1866 | csum_root->track_dirty = 1; | |
1867 | ||
9078a3e1 CM |
1868 | btrfs_read_block_groups(extent_root); |
1869 | ||
2d69a0f8 | 1870 | fs_info->generation = generation; |
c146afad | 1871 | fs_info->last_trans_committed = generation; |
d18a2c44 CM |
1872 | fs_info->data_alloc_profile = (u64)-1; |
1873 | fs_info->metadata_alloc_profile = (u64)-1; | |
1874 | fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; | |
a74a4b97 CM |
1875 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
1876 | "btrfs-cleaner"); | |
57506d50 | 1877 | if (IS_ERR(fs_info->cleaner_kthread)) |
d20f7043 | 1878 | goto fail_csum_root; |
a74a4b97 CM |
1879 | |
1880 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
1881 | tree_root, | |
1882 | "btrfs-transaction"); | |
57506d50 | 1883 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 1884 | goto fail_cleaner; |
a74a4b97 | 1885 | |
c289811c CM |
1886 | if (!btrfs_test_opt(tree_root, SSD) && |
1887 | !btrfs_test_opt(tree_root, NOSSD) && | |
1888 | !fs_info->fs_devices->rotating) { | |
1889 | printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD " | |
1890 | "mode\n"); | |
1891 | btrfs_set_opt(fs_info->mount_opt, SSD); | |
1892 | } | |
1893 | ||
e02119d5 | 1894 | if (btrfs_super_log_root(disk_super) != 0) { |
e02119d5 CM |
1895 | u64 bytenr = btrfs_super_log_root(disk_super); |
1896 | ||
7c2ca468 | 1897 | if (fs_devices->rw_devices == 0) { |
d397712b CM |
1898 | printk(KERN_WARNING "Btrfs log replay required " |
1899 | "on RO media\n"); | |
7c2ca468 CM |
1900 | err = -EIO; |
1901 | goto fail_trans_kthread; | |
1902 | } | |
e02119d5 CM |
1903 | blocksize = |
1904 | btrfs_level_size(tree_root, | |
1905 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 1906 | |
e02119d5 CM |
1907 | log_tree_root = kzalloc(sizeof(struct btrfs_root), |
1908 | GFP_NOFS); | |
1909 | ||
1910 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1911 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1912 | ||
1913 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
1914 | blocksize, |
1915 | generation + 1); | |
e02119d5 CM |
1916 | ret = btrfs_recover_log_trees(log_tree_root); |
1917 | BUG_ON(ret); | |
e556ce2c YZ |
1918 | |
1919 | if (sb->s_flags & MS_RDONLY) { | |
1920 | ret = btrfs_commit_super(tree_root); | |
1921 | BUG_ON(ret); | |
1922 | } | |
e02119d5 | 1923 | } |
1a40e23b | 1924 | |
7c2ca468 | 1925 | if (!(sb->s_flags & MS_RDONLY)) { |
5d4f98a2 | 1926 | ret = btrfs_recover_relocation(tree_root); |
7c2ca468 CM |
1927 | BUG_ON(ret); |
1928 | } | |
1a40e23b | 1929 | |
3de4586c CM |
1930 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
1931 | location.type = BTRFS_ROOT_ITEM_KEY; | |
1932 | location.offset = (u64)-1; | |
1933 | ||
3de4586c CM |
1934 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
1935 | if (!fs_info->fs_root) | |
7c2ca468 | 1936 | goto fail_trans_kthread; |
c289811c | 1937 | |
0f7d52f4 | 1938 | return tree_root; |
39279cc3 | 1939 | |
7c2ca468 CM |
1940 | fail_trans_kthread: |
1941 | kthread_stop(fs_info->transaction_kthread); | |
3f157a2f | 1942 | fail_cleaner: |
a74a4b97 | 1943 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
1944 | |
1945 | /* | |
1946 | * make sure we're done with the btree inode before we stop our | |
1947 | * kthreads | |
1948 | */ | |
1949 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
1950 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
1951 | ||
d20f7043 CM |
1952 | fail_csum_root: |
1953 | free_extent_buffer(csum_root->node); | |
5d4f98a2 YZ |
1954 | free_extent_buffer(csum_root->commit_root); |
1955 | fail_dev_root: | |
1956 | free_extent_buffer(dev_root->node); | |
1957 | free_extent_buffer(dev_root->commit_root); | |
0b86a832 CM |
1958 | fail_extent_root: |
1959 | free_extent_buffer(extent_root->node); | |
5d4f98a2 | 1960 | free_extent_buffer(extent_root->commit_root); |
39279cc3 | 1961 | fail_tree_root: |
5f39d397 | 1962 | free_extent_buffer(tree_root->node); |
5d4f98a2 | 1963 | free_extent_buffer(tree_root->commit_root); |
2b82032c YZ |
1964 | fail_chunk_root: |
1965 | free_extent_buffer(chunk_root->node); | |
5d4f98a2 | 1966 | free_extent_buffer(chunk_root->commit_root); |
39279cc3 | 1967 | fail_sb_buffer: |
247e743c | 1968 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 1969 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
1970 | btrfs_stop_workers(&fs_info->workers); |
1971 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 1972 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 1973 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 1974 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 1975 | btrfs_stop_workers(&fs_info->submit_workers); |
4543df7e | 1976 | fail_iput: |
7c2ca468 | 1977 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
4543df7e | 1978 | iput(fs_info->btree_inode); |
7e662854 | 1979 | |
dfe25020 | 1980 | btrfs_close_devices(fs_info->fs_devices); |
84eed90f | 1981 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
ad081f14 | 1982 | fail_bdi: |
7e662854 | 1983 | bdi_destroy(&fs_info->bdi); |
7e662854 | 1984 | fail: |
39279cc3 CM |
1985 | kfree(extent_root); |
1986 | kfree(tree_root); | |
1987 | kfree(fs_info); | |
83afeac4 JM |
1988 | kfree(chunk_root); |
1989 | kfree(dev_root); | |
d20f7043 | 1990 | kfree(csum_root); |
39279cc3 | 1991 | return ERR_PTR(err); |
eb60ceac CM |
1992 | } |
1993 | ||
f2984462 CM |
1994 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
1995 | { | |
1996 | char b[BDEVNAME_SIZE]; | |
1997 | ||
1998 | if (uptodate) { | |
1999 | set_buffer_uptodate(bh); | |
2000 | } else { | |
2001 | if (!buffer_eopnotsupp(bh) && printk_ratelimit()) { | |
2002 | printk(KERN_WARNING "lost page write due to " | |
2003 | "I/O error on %s\n", | |
2004 | bdevname(bh->b_bdev, b)); | |
2005 | } | |
1259ab75 CM |
2006 | /* note, we dont' set_buffer_write_io_error because we have |
2007 | * our own ways of dealing with the IO errors | |
2008 | */ | |
f2984462 CM |
2009 | clear_buffer_uptodate(bh); |
2010 | } | |
2011 | unlock_buffer(bh); | |
2012 | put_bh(bh); | |
2013 | } | |
2014 | ||
a512bbf8 YZ |
2015 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
2016 | { | |
2017 | struct buffer_head *bh; | |
2018 | struct buffer_head *latest = NULL; | |
2019 | struct btrfs_super_block *super; | |
2020 | int i; | |
2021 | u64 transid = 0; | |
2022 | u64 bytenr; | |
2023 | ||
2024 | /* we would like to check all the supers, but that would make | |
2025 | * a btrfs mount succeed after a mkfs from a different FS. | |
2026 | * So, we need to add a special mount option to scan for | |
2027 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
2028 | */ | |
2029 | for (i = 0; i < 1; i++) { | |
2030 | bytenr = btrfs_sb_offset(i); | |
2031 | if (bytenr + 4096 >= i_size_read(bdev->bd_inode)) | |
2032 | break; | |
2033 | bh = __bread(bdev, bytenr / 4096, 4096); | |
2034 | if (!bh) | |
2035 | continue; | |
2036 | ||
2037 | super = (struct btrfs_super_block *)bh->b_data; | |
2038 | if (btrfs_super_bytenr(super) != bytenr || | |
2039 | strncmp((char *)(&super->magic), BTRFS_MAGIC, | |
2040 | sizeof(super->magic))) { | |
2041 | brelse(bh); | |
2042 | continue; | |
2043 | } | |
2044 | ||
2045 | if (!latest || btrfs_super_generation(super) > transid) { | |
2046 | brelse(latest); | |
2047 | latest = bh; | |
2048 | transid = btrfs_super_generation(super); | |
2049 | } else { | |
2050 | brelse(bh); | |
2051 | } | |
2052 | } | |
2053 | return latest; | |
2054 | } | |
2055 | ||
4eedeb75 HH |
2056 | /* |
2057 | * this should be called twice, once with wait == 0 and | |
2058 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
2059 | * we write are pinned. | |
2060 | * | |
2061 | * They are released when wait == 1 is done. | |
2062 | * max_mirrors must be the same for both runs, and it indicates how | |
2063 | * many supers on this one device should be written. | |
2064 | * | |
2065 | * max_mirrors == 0 means to write them all. | |
2066 | */ | |
a512bbf8 YZ |
2067 | static int write_dev_supers(struct btrfs_device *device, |
2068 | struct btrfs_super_block *sb, | |
2069 | int do_barriers, int wait, int max_mirrors) | |
2070 | { | |
2071 | struct buffer_head *bh; | |
2072 | int i; | |
2073 | int ret; | |
2074 | int errors = 0; | |
2075 | u32 crc; | |
2076 | u64 bytenr; | |
2077 | int last_barrier = 0; | |
2078 | ||
2079 | if (max_mirrors == 0) | |
2080 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
2081 | ||
2082 | /* make sure only the last submit_bh does a barrier */ | |
2083 | if (do_barriers) { | |
2084 | for (i = 0; i < max_mirrors; i++) { | |
2085 | bytenr = btrfs_sb_offset(i); | |
2086 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= | |
2087 | device->total_bytes) | |
2088 | break; | |
2089 | last_barrier = i; | |
2090 | } | |
2091 | } | |
2092 | ||
2093 | for (i = 0; i < max_mirrors; i++) { | |
2094 | bytenr = btrfs_sb_offset(i); | |
2095 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) | |
2096 | break; | |
2097 | ||
2098 | if (wait) { | |
2099 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
2100 | BTRFS_SUPER_INFO_SIZE); | |
2101 | BUG_ON(!bh); | |
a512bbf8 | 2102 | wait_on_buffer(bh); |
4eedeb75 HH |
2103 | if (!buffer_uptodate(bh)) |
2104 | errors++; | |
2105 | ||
2106 | /* drop our reference */ | |
2107 | brelse(bh); | |
2108 | ||
2109 | /* drop the reference from the wait == 0 run */ | |
2110 | brelse(bh); | |
2111 | continue; | |
a512bbf8 YZ |
2112 | } else { |
2113 | btrfs_set_super_bytenr(sb, bytenr); | |
2114 | ||
2115 | crc = ~(u32)0; | |
2116 | crc = btrfs_csum_data(NULL, (char *)sb + | |
2117 | BTRFS_CSUM_SIZE, crc, | |
2118 | BTRFS_SUPER_INFO_SIZE - | |
2119 | BTRFS_CSUM_SIZE); | |
2120 | btrfs_csum_final(crc, sb->csum); | |
2121 | ||
4eedeb75 HH |
2122 | /* |
2123 | * one reference for us, and we leave it for the | |
2124 | * caller | |
2125 | */ | |
a512bbf8 YZ |
2126 | bh = __getblk(device->bdev, bytenr / 4096, |
2127 | BTRFS_SUPER_INFO_SIZE); | |
2128 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); | |
2129 | ||
4eedeb75 | 2130 | /* one reference for submit_bh */ |
a512bbf8 | 2131 | get_bh(bh); |
4eedeb75 HH |
2132 | |
2133 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
2134 | lock_buffer(bh); |
2135 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
2136 | } | |
2137 | ||
2138 | if (i == last_barrier && do_barriers && device->barriers) { | |
2139 | ret = submit_bh(WRITE_BARRIER, bh); | |
2140 | if (ret == -EOPNOTSUPP) { | |
2141 | printk("btrfs: disabling barriers on dev %s\n", | |
2142 | device->name); | |
2143 | set_buffer_uptodate(bh); | |
2144 | device->barriers = 0; | |
4eedeb75 | 2145 | /* one reference for submit_bh */ |
a512bbf8 YZ |
2146 | get_bh(bh); |
2147 | lock_buffer(bh); | |
ffbd517d | 2148 | ret = submit_bh(WRITE_SYNC, bh); |
a512bbf8 YZ |
2149 | } |
2150 | } else { | |
ffbd517d | 2151 | ret = submit_bh(WRITE_SYNC, bh); |
a512bbf8 YZ |
2152 | } |
2153 | ||
4eedeb75 | 2154 | if (ret) |
a512bbf8 | 2155 | errors++; |
a512bbf8 YZ |
2156 | } |
2157 | return errors < i ? 0 : -1; | |
2158 | } | |
2159 | ||
2160 | int write_all_supers(struct btrfs_root *root, int max_mirrors) | |
f2984462 | 2161 | { |
e5e9a520 | 2162 | struct list_head *head; |
f2984462 | 2163 | struct btrfs_device *dev; |
a061fc8d | 2164 | struct btrfs_super_block *sb; |
f2984462 | 2165 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
2166 | int ret; |
2167 | int do_barriers; | |
a236aed1 CM |
2168 | int max_errors; |
2169 | int total_errors = 0; | |
a061fc8d | 2170 | u64 flags; |
f2984462 | 2171 | |
a236aed1 | 2172 | max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1; |
f2984462 CM |
2173 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
2174 | ||
a061fc8d CM |
2175 | sb = &root->fs_info->super_for_commit; |
2176 | dev_item = &sb->dev_item; | |
e5e9a520 CM |
2177 | |
2178 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); | |
2179 | head = &root->fs_info->fs_devices->devices; | |
c6e30871 | 2180 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
2181 | if (!dev->bdev) { |
2182 | total_errors++; | |
2183 | continue; | |
2184 | } | |
2b82032c | 2185 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2186 | continue; |
2187 | ||
2b82032c | 2188 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
2189 | btrfs_set_stack_device_type(dev_item, dev->type); |
2190 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
2191 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
2192 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
2193 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
2194 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
2195 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
2196 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 2197 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 2198 | |
a061fc8d CM |
2199 | flags = btrfs_super_flags(sb); |
2200 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
2201 | ||
a512bbf8 | 2202 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
2203 | if (ret) |
2204 | total_errors++; | |
f2984462 | 2205 | } |
a236aed1 | 2206 | if (total_errors > max_errors) { |
d397712b CM |
2207 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2208 | total_errors); | |
a236aed1 CM |
2209 | BUG(); |
2210 | } | |
f2984462 | 2211 | |
a512bbf8 | 2212 | total_errors = 0; |
c6e30871 | 2213 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
2214 | if (!dev->bdev) |
2215 | continue; | |
2b82032c | 2216 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2217 | continue; |
2218 | ||
a512bbf8 YZ |
2219 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
2220 | if (ret) | |
2221 | total_errors++; | |
f2984462 | 2222 | } |
e5e9a520 | 2223 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 2224 | if (total_errors > max_errors) { |
d397712b CM |
2225 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2226 | total_errors); | |
a236aed1 CM |
2227 | BUG(); |
2228 | } | |
f2984462 CM |
2229 | return 0; |
2230 | } | |
2231 | ||
a512bbf8 YZ |
2232 | int write_ctree_super(struct btrfs_trans_handle *trans, |
2233 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 2234 | { |
e66f709b | 2235 | int ret; |
5f39d397 | 2236 | |
a512bbf8 | 2237 | ret = write_all_supers(root, max_mirrors); |
5f39d397 | 2238 | return ret; |
cfaa7295 CM |
2239 | } |
2240 | ||
5eda7b5e | 2241 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f | 2242 | { |
5d4f98a2 | 2243 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
2619ba1f CM |
2244 | radix_tree_delete(&fs_info->fs_roots_radix, |
2245 | (unsigned long)root->root_key.objectid); | |
3394e160 CM |
2246 | if (root->anon_super.s_dev) { |
2247 | down_write(&root->anon_super.s_umount); | |
2248 | kill_anon_super(&root->anon_super); | |
2249 | } | |
2619ba1f | 2250 | if (root->node) |
5f39d397 | 2251 | free_extent_buffer(root->node); |
2619ba1f | 2252 | if (root->commit_root) |
5f39d397 | 2253 | free_extent_buffer(root->commit_root); |
d397712b | 2254 | kfree(root->name); |
2619ba1f CM |
2255 | kfree(root); |
2256 | return 0; | |
2257 | } | |
2258 | ||
35b7e476 | 2259 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
2260 | { |
2261 | int ret; | |
2262 | struct btrfs_root *gang[8]; | |
2263 | int i; | |
2264 | ||
d397712b | 2265 | while (1) { |
0f7d52f4 CM |
2266 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
2267 | (void **)gang, 0, | |
2268 | ARRAY_SIZE(gang)); | |
2269 | if (!ret) | |
2270 | break; | |
2619ba1f | 2271 | for (i = 0; i < ret; i++) |
5eda7b5e | 2272 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
2273 | } |
2274 | return 0; | |
2275 | } | |
b4100d64 | 2276 | |
c146afad | 2277 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 2278 | { |
c146afad YZ |
2279 | u64 root_objectid = 0; |
2280 | struct btrfs_root *gang[8]; | |
2281 | int i; | |
3768f368 | 2282 | int ret; |
e089f05c | 2283 | |
c146afad YZ |
2284 | while (1) { |
2285 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2286 | (void **)gang, root_objectid, | |
2287 | ARRAY_SIZE(gang)); | |
2288 | if (!ret) | |
2289 | break; | |
5d4f98a2 YZ |
2290 | |
2291 | root_objectid = gang[ret - 1]->root_key.objectid + 1; | |
c146afad YZ |
2292 | for (i = 0; i < ret; i++) { |
2293 | root_objectid = gang[i]->root_key.objectid; | |
2294 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
5d4f98a2 | 2295 | root_objectid); |
c146afad YZ |
2296 | BUG_ON(ret); |
2297 | btrfs_orphan_cleanup(gang[i]); | |
2298 | } | |
2299 | root_objectid++; | |
2300 | } | |
2301 | return 0; | |
2302 | } | |
a2135011 | 2303 | |
c146afad YZ |
2304 | int btrfs_commit_super(struct btrfs_root *root) |
2305 | { | |
2306 | struct btrfs_trans_handle *trans; | |
2307 | int ret; | |
a74a4b97 | 2308 | |
c146afad | 2309 | mutex_lock(&root->fs_info->cleaner_mutex); |
a74a4b97 | 2310 | btrfs_clean_old_snapshots(root); |
c146afad | 2311 | mutex_unlock(&root->fs_info->cleaner_mutex); |
79154b1b | 2312 | trans = btrfs_start_transaction(root, 1); |
54aa1f4d | 2313 | ret = btrfs_commit_transaction(trans, root); |
c146afad YZ |
2314 | BUG_ON(ret); |
2315 | /* run commit again to drop the original snapshot */ | |
79154b1b CM |
2316 | trans = btrfs_start_transaction(root, 1); |
2317 | btrfs_commit_transaction(trans, root); | |
2318 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 2319 | BUG_ON(ret); |
d6bfde87 | 2320 | |
a512bbf8 | 2321 | ret = write_ctree_super(NULL, root, 0); |
c146afad YZ |
2322 | return ret; |
2323 | } | |
2324 | ||
2325 | int close_ctree(struct btrfs_root *root) | |
2326 | { | |
2327 | struct btrfs_fs_info *fs_info = root->fs_info; | |
2328 | int ret; | |
2329 | ||
2330 | fs_info->closing = 1; | |
2331 | smp_mb(); | |
2332 | ||
2333 | kthread_stop(root->fs_info->transaction_kthread); | |
2334 | kthread_stop(root->fs_info->cleaner_kthread); | |
2335 | ||
2336 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { | |
2337 | ret = btrfs_commit_super(root); | |
d397712b CM |
2338 | if (ret) |
2339 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
c146afad | 2340 | } |
0f7d52f4 | 2341 | |
f25784b3 YZ |
2342 | fs_info->closing = 2; |
2343 | smp_mb(); | |
2344 | ||
b0c68f8b | 2345 | if (fs_info->delalloc_bytes) { |
d397712b | 2346 | printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n", |
21380931 | 2347 | (unsigned long long)fs_info->delalloc_bytes); |
b0c68f8b | 2348 | } |
31153d81 | 2349 | if (fs_info->total_ref_cache_size) { |
d397712b CM |
2350 | printk(KERN_INFO "btrfs: at umount reference cache size %llu\n", |
2351 | (unsigned long long)fs_info->total_ref_cache_size); | |
31153d81 | 2352 | } |
bcc63abb | 2353 | |
5d4f98a2 YZ |
2354 | free_extent_buffer(fs_info->extent_root->node); |
2355 | free_extent_buffer(fs_info->extent_root->commit_root); | |
2356 | free_extent_buffer(fs_info->tree_root->node); | |
2357 | free_extent_buffer(fs_info->tree_root->commit_root); | |
2358 | free_extent_buffer(root->fs_info->chunk_root->node); | |
2359 | free_extent_buffer(root->fs_info->chunk_root->commit_root); | |
2360 | free_extent_buffer(root->fs_info->dev_root->node); | |
2361 | free_extent_buffer(root->fs_info->dev_root->commit_root); | |
2362 | free_extent_buffer(root->fs_info->csum_root->node); | |
2363 | free_extent_buffer(root->fs_info->csum_root->commit_root); | |
d20f7043 | 2364 | |
9078a3e1 | 2365 | btrfs_free_block_groups(root->fs_info); |
d10c5f31 | 2366 | |
c146afad | 2367 | del_fs_roots(fs_info); |
d10c5f31 | 2368 | |
c146afad | 2369 | iput(fs_info->btree_inode); |
9ad6b7bc | 2370 | |
247e743c | 2371 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2372 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2373 | btrfs_stop_workers(&fs_info->workers); |
2374 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 2375 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 2376 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 2377 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 2378 | btrfs_stop_workers(&fs_info->submit_workers); |
d6bfde87 | 2379 | |
dfe25020 | 2380 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 2381 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 2382 | |
04160088 | 2383 | bdi_destroy(&fs_info->bdi); |
0b86a832 | 2384 | |
0f7d52f4 | 2385 | kfree(fs_info->extent_root); |
0f7d52f4 | 2386 | kfree(fs_info->tree_root); |
0b86a832 CM |
2387 | kfree(fs_info->chunk_root); |
2388 | kfree(fs_info->dev_root); | |
d20f7043 | 2389 | kfree(fs_info->csum_root); |
eb60ceac CM |
2390 | return 0; |
2391 | } | |
2392 | ||
1259ab75 | 2393 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 2394 | { |
1259ab75 | 2395 | int ret; |
810191ff | 2396 | struct inode *btree_inode = buf->first_page->mapping->host; |
1259ab75 CM |
2397 | |
2398 | ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); | |
2399 | if (!ret) | |
2400 | return ret; | |
2401 | ||
2402 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
2403 | parent_transid); | |
2404 | return !ret; | |
5f39d397 CM |
2405 | } |
2406 | ||
2407 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 2408 | { |
810191ff | 2409 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 2410 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
2411 | buf); |
2412 | } | |
6702ed49 | 2413 | |
5f39d397 CM |
2414 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
2415 | { | |
810191ff | 2416 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
5f39d397 CM |
2417 | u64 transid = btrfs_header_generation(buf); |
2418 | struct inode *btree_inode = root->fs_info->btree_inode; | |
b9473439 | 2419 | int was_dirty; |
b4ce94de | 2420 | |
b9447ef8 | 2421 | btrfs_assert_tree_locked(buf); |
ccd467d6 | 2422 | if (transid != root->fs_info->generation) { |
d397712b CM |
2423 | printk(KERN_CRIT "btrfs transid mismatch buffer %llu, " |
2424 | "found %llu running %llu\n", | |
db94535d | 2425 | (unsigned long long)buf->start, |
d397712b CM |
2426 | (unsigned long long)transid, |
2427 | (unsigned long long)root->fs_info->generation); | |
ccd467d6 CM |
2428 | WARN_ON(1); |
2429 | } | |
b9473439 CM |
2430 | was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
2431 | buf); | |
2432 | if (!was_dirty) { | |
2433 | spin_lock(&root->fs_info->delalloc_lock); | |
2434 | root->fs_info->dirty_metadata_bytes += buf->len; | |
2435 | spin_unlock(&root->fs_info->delalloc_lock); | |
2436 | } | |
eb60ceac CM |
2437 | } |
2438 | ||
d3c2fdcf | 2439 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
35b7e476 | 2440 | { |
188de649 CM |
2441 | /* |
2442 | * looks as though older kernels can get into trouble with | |
2443 | * this code, they end up stuck in balance_dirty_pages forever | |
2444 | */ | |
d6bfde87 | 2445 | u64 num_dirty; |
771ed689 | 2446 | unsigned long thresh = 32 * 1024 * 1024; |
d6bfde87 | 2447 | |
6933c02e | 2448 | if (current->flags & PF_MEMALLOC) |
d6bfde87 CM |
2449 | return; |
2450 | ||
585ad2c3 CM |
2451 | num_dirty = root->fs_info->dirty_metadata_bytes; |
2452 | ||
d6bfde87 CM |
2453 | if (num_dirty > thresh) { |
2454 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 2455 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 2456 | } |
188de649 | 2457 | return; |
35b7e476 | 2458 | } |
6b80053d | 2459 | |
ca7a79ad | 2460 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 2461 | { |
810191ff | 2462 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
ce9adaa5 | 2463 | int ret; |
ca7a79ad | 2464 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
d397712b | 2465 | if (ret == 0) |
b4ce94de | 2466 | set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags); |
ce9adaa5 | 2467 | return ret; |
6b80053d | 2468 | } |
0da5468f | 2469 | |
4bef0848 CM |
2470 | int btree_lock_page_hook(struct page *page) |
2471 | { | |
2472 | struct inode *inode = page->mapping->host; | |
b9473439 | 2473 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4bef0848 CM |
2474 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
2475 | struct extent_buffer *eb; | |
2476 | unsigned long len; | |
2477 | u64 bytenr = page_offset(page); | |
2478 | ||
2479 | if (page->private == EXTENT_PAGE_PRIVATE) | |
2480 | goto out; | |
2481 | ||
2482 | len = page->private >> 2; | |
2483 | eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS); | |
2484 | if (!eb) | |
2485 | goto out; | |
2486 | ||
2487 | btrfs_tree_lock(eb); | |
4bef0848 | 2488 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); |
b9473439 CM |
2489 | |
2490 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) { | |
2491 | spin_lock(&root->fs_info->delalloc_lock); | |
2492 | if (root->fs_info->dirty_metadata_bytes >= eb->len) | |
2493 | root->fs_info->dirty_metadata_bytes -= eb->len; | |
2494 | else | |
2495 | WARN_ON(1); | |
2496 | spin_unlock(&root->fs_info->delalloc_lock); | |
2497 | } | |
2498 | ||
4bef0848 CM |
2499 | btrfs_tree_unlock(eb); |
2500 | free_extent_buffer(eb); | |
2501 | out: | |
2502 | lock_page(page); | |
2503 | return 0; | |
2504 | } | |
2505 | ||
d1310b2e | 2506 | static struct extent_io_ops btree_extent_io_ops = { |
4bef0848 | 2507 | .write_cache_pages_lock_hook = btree_lock_page_hook, |
ce9adaa5 | 2508 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
0b86a832 | 2509 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
2510 | /* note we're sharing with inode.c for the merge bio hook */ |
2511 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 2512 | }; |