]>
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 | ||
a74a4b97 | 19 | #include <linux/version.h> |
e20d96d6 | 20 | #include <linux/fs.h> |
d98237b3 | 21 | #include <linux/blkdev.h> |
87cbda5c | 22 | #include <linux/scatterlist.h> |
22b0ebda | 23 | #include <linux/swap.h> |
0f7d52f4 | 24 | #include <linux/radix-tree.h> |
35b7e476 | 25 | #include <linux/writeback.h> |
5f39d397 | 26 | #include <linux/buffer_head.h> // for block_sync_page |
ce9adaa5 | 27 | #include <linux/workqueue.h> |
a74a4b97 | 28 | #include <linux/kthread.h> |
4b4e25f2 CM |
29 | #include <linux/freezer.h> |
30 | #include "compat.h" | |
a5eb62e3 | 31 | #include "crc32c.h" |
eb60ceac CM |
32 | #include "ctree.h" |
33 | #include "disk-io.h" | |
e089f05c | 34 | #include "transaction.h" |
0f7d52f4 | 35 | #include "btrfs_inode.h" |
0b86a832 | 36 | #include "volumes.h" |
db94535d | 37 | #include "print-tree.h" |
8b712842 | 38 | #include "async-thread.h" |
925baedd | 39 | #include "locking.h" |
017e5369 | 40 | #include "ref-cache.h" |
e02119d5 | 41 | #include "tree-log.h" |
eb60ceac | 42 | |
5f39d397 CM |
43 | #if 0 |
44 | static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) | |
7eccb903 | 45 | { |
5f39d397 CM |
46 | if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) { |
47 | printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n", | |
48 | (unsigned long long)extent_buffer_blocknr(buf), | |
49 | (unsigned long long)btrfs_header_blocknr(buf)); | |
39279cc3 | 50 | return 1; |
d98237b3 | 51 | } |
9a8dd150 | 52 | return 0; |
eb60ceac | 53 | } |
5f39d397 | 54 | #endif |
eb60ceac | 55 | |
d1310b2e | 56 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 57 | static void end_workqueue_fn(struct btrfs_work *work); |
ce9adaa5 | 58 | |
d352ac68 CM |
59 | /* |
60 | * end_io_wq structs are used to do processing in task context when an IO is | |
61 | * complete. This is used during reads to verify checksums, and it is used | |
62 | * by writes to insert metadata for new file extents after IO is complete. | |
63 | */ | |
ce9adaa5 CM |
64 | struct end_io_wq { |
65 | struct bio *bio; | |
66 | bio_end_io_t *end_io; | |
67 | void *private; | |
68 | struct btrfs_fs_info *info; | |
69 | int error; | |
22c59948 | 70 | int metadata; |
ce9adaa5 | 71 | struct list_head list; |
8b712842 | 72 | struct btrfs_work work; |
ce9adaa5 | 73 | }; |
0da5468f | 74 | |
d352ac68 CM |
75 | /* |
76 | * async submit bios are used to offload expensive checksumming | |
77 | * onto the worker threads. They checksum file and metadata bios | |
78 | * just before they are sent down the IO stack. | |
79 | */ | |
44b8bd7e CM |
80 | struct async_submit_bio { |
81 | struct inode *inode; | |
82 | struct bio *bio; | |
83 | struct list_head list; | |
4a69a410 CM |
84 | extent_submit_bio_hook_t *submit_bio_start; |
85 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
86 | int rw; |
87 | int mirror_num; | |
c8b97818 | 88 | unsigned long bio_flags; |
8b712842 | 89 | struct btrfs_work work; |
44b8bd7e CM |
90 | }; |
91 | ||
d352ac68 CM |
92 | /* |
93 | * extents on the btree inode are pretty simple, there's one extent | |
94 | * that covers the entire device | |
95 | */ | |
b2950863 CH |
96 | static struct extent_map *btree_get_extent(struct inode *inode, |
97 | struct page *page, size_t page_offset, u64 start, u64 len, | |
98 | int create) | |
7eccb903 | 99 | { |
5f39d397 CM |
100 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
101 | struct extent_map *em; | |
102 | int ret; | |
103 | ||
d1310b2e CM |
104 | spin_lock(&em_tree->lock); |
105 | em = lookup_extent_mapping(em_tree, start, len); | |
a061fc8d CM |
106 | if (em) { |
107 | em->bdev = | |
108 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
109 | spin_unlock(&em_tree->lock); | |
5f39d397 | 110 | goto out; |
a061fc8d CM |
111 | } |
112 | spin_unlock(&em_tree->lock); | |
7b13b7b1 | 113 | |
5f39d397 CM |
114 | em = alloc_extent_map(GFP_NOFS); |
115 | if (!em) { | |
116 | em = ERR_PTR(-ENOMEM); | |
117 | goto out; | |
118 | } | |
119 | em->start = 0; | |
0afbaf8c | 120 | em->len = (u64)-1; |
c8b97818 | 121 | em->block_len = (u64)-1; |
5f39d397 | 122 | em->block_start = 0; |
a061fc8d | 123 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e CM |
124 | |
125 | spin_lock(&em_tree->lock); | |
5f39d397 CM |
126 | ret = add_extent_mapping(em_tree, em); |
127 | if (ret == -EEXIST) { | |
0afbaf8c CM |
128 | u64 failed_start = em->start; |
129 | u64 failed_len = em->len; | |
130 | ||
131 | printk("failed to insert %Lu %Lu -> %Lu into tree\n", | |
132 | em->start, em->len, em->block_start); | |
5f39d397 | 133 | free_extent_map(em); |
7b13b7b1 | 134 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c CM |
135 | if (em) { |
136 | printk("after failing, found %Lu %Lu %Lu\n", | |
137 | em->start, em->len, em->block_start); | |
7b13b7b1 | 138 | ret = 0; |
0afbaf8c CM |
139 | } else { |
140 | em = lookup_extent_mapping(em_tree, failed_start, | |
141 | failed_len); | |
142 | if (em) { | |
143 | printk("double failure lookup gives us " | |
144 | "%Lu %Lu -> %Lu\n", em->start, | |
145 | em->len, em->block_start); | |
146 | free_extent_map(em); | |
147 | } | |
7b13b7b1 | 148 | ret = -EIO; |
0afbaf8c | 149 | } |
5f39d397 | 150 | } else if (ret) { |
7b13b7b1 CM |
151 | free_extent_map(em); |
152 | em = NULL; | |
5f39d397 | 153 | } |
7b13b7b1 CM |
154 | spin_unlock(&em_tree->lock); |
155 | ||
156 | if (ret) | |
157 | em = ERR_PTR(ret); | |
5f39d397 CM |
158 | out: |
159 | return em; | |
7eccb903 CM |
160 | } |
161 | ||
19c00ddc CM |
162 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
163 | { | |
a5eb62e3 | 164 | return btrfs_crc32c(seed, data, len); |
19c00ddc CM |
165 | } |
166 | ||
167 | void btrfs_csum_final(u32 crc, char *result) | |
168 | { | |
169 | *(__le32 *)result = ~cpu_to_le32(crc); | |
170 | } | |
171 | ||
d352ac68 CM |
172 | /* |
173 | * compute the csum for a btree block, and either verify it or write it | |
174 | * into the csum field of the block. | |
175 | */ | |
19c00ddc CM |
176 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
177 | int verify) | |
178 | { | |
607d432d JB |
179 | u16 csum_size = |
180 | btrfs_super_csum_size(&root->fs_info->super_copy); | |
181 | char *result = NULL; | |
19c00ddc CM |
182 | unsigned long len; |
183 | unsigned long cur_len; | |
184 | unsigned long offset = BTRFS_CSUM_SIZE; | |
185 | char *map_token = NULL; | |
186 | char *kaddr; | |
187 | unsigned long map_start; | |
188 | unsigned long map_len; | |
189 | int err; | |
190 | u32 crc = ~(u32)0; | |
607d432d | 191 | unsigned long inline_result; |
19c00ddc CM |
192 | |
193 | len = buf->len - offset; | |
194 | while(len > 0) { | |
195 | err = map_private_extent_buffer(buf, offset, 32, | |
196 | &map_token, &kaddr, | |
197 | &map_start, &map_len, KM_USER0); | |
198 | if (err) { | |
199 | printk("failed to map extent buffer! %lu\n", | |
200 | offset); | |
201 | return 1; | |
202 | } | |
203 | cur_len = min(len, map_len - (offset - map_start)); | |
204 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
205 | crc, cur_len); | |
206 | len -= cur_len; | |
207 | offset += cur_len; | |
208 | unmap_extent_buffer(buf, map_token, KM_USER0); | |
209 | } | |
607d432d JB |
210 | if (csum_size > sizeof(inline_result)) { |
211 | result = kzalloc(csum_size * sizeof(char), GFP_NOFS); | |
212 | if (!result) | |
213 | return 1; | |
214 | } else { | |
215 | result = (char *)&inline_result; | |
216 | } | |
217 | ||
19c00ddc CM |
218 | btrfs_csum_final(crc, result); |
219 | ||
220 | if (verify) { | |
e4204ded | 221 | /* FIXME, this is not good */ |
607d432d | 222 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
223 | u32 val; |
224 | u32 found = 0; | |
607d432d | 225 | memcpy(&found, result, csum_size); |
e4204ded | 226 | |
607d432d | 227 | read_extent_buffer(buf, &val, 0, csum_size); |
e4204ded | 228 | printk("btrfs: %s checksum verify failed on %llu " |
2dd3e67b | 229 | "wanted %X found %X level %d\n", |
19c00ddc | 230 | root->fs_info->sb->s_id, |
2dd3e67b | 231 | buf->start, val, found, btrfs_header_level(buf)); |
607d432d JB |
232 | if (result != (char *)&inline_result) |
233 | kfree(result); | |
19c00ddc CM |
234 | return 1; |
235 | } | |
236 | } else { | |
607d432d | 237 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 238 | } |
607d432d JB |
239 | if (result != (char *)&inline_result) |
240 | kfree(result); | |
19c00ddc CM |
241 | return 0; |
242 | } | |
243 | ||
d352ac68 CM |
244 | /* |
245 | * we can't consider a given block up to date unless the transid of the | |
246 | * block matches the transid in the parent node's pointer. This is how we | |
247 | * detect blocks that either didn't get written at all or got written | |
248 | * in the wrong place. | |
249 | */ | |
1259ab75 CM |
250 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
251 | struct extent_buffer *eb, u64 parent_transid) | |
252 | { | |
253 | int ret; | |
254 | ||
255 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
256 | return 0; | |
257 | ||
258 | lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); | |
259 | if (extent_buffer_uptodate(io_tree, eb) && | |
260 | btrfs_header_generation(eb) == parent_transid) { | |
261 | ret = 0; | |
262 | goto out; | |
263 | } | |
264 | printk("parent transid verify failed on %llu wanted %llu found %llu\n", | |
265 | (unsigned long long)eb->start, | |
266 | (unsigned long long)parent_transid, | |
267 | (unsigned long long)btrfs_header_generation(eb)); | |
268 | ret = 1; | |
1259ab75 | 269 | clear_extent_buffer_uptodate(io_tree, eb); |
33958dc6 | 270 | out: |
1259ab75 CM |
271 | unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, |
272 | GFP_NOFS); | |
273 | return ret; | |
1259ab75 CM |
274 | } |
275 | ||
d352ac68 CM |
276 | /* |
277 | * helper to read a given tree block, doing retries as required when | |
278 | * the checksums don't match and we have alternate mirrors to try. | |
279 | */ | |
f188591e CM |
280 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
281 | struct extent_buffer *eb, | |
ca7a79ad | 282 | u64 start, u64 parent_transid) |
f188591e CM |
283 | { |
284 | struct extent_io_tree *io_tree; | |
285 | int ret; | |
286 | int num_copies = 0; | |
287 | int mirror_num = 0; | |
288 | ||
289 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
290 | while (1) { | |
291 | ret = read_extent_buffer_pages(io_tree, eb, start, 1, | |
292 | btree_get_extent, mirror_num); | |
1259ab75 CM |
293 | if (!ret && |
294 | !verify_parent_transid(io_tree, eb, parent_transid)) | |
f188591e | 295 | return ret; |
a1b32a59 | 296 | printk("read extent buffer pages failed with ret %d mirror no %d\n", ret, mirror_num); |
f188591e CM |
297 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
298 | eb->start, eb->len); | |
4235298e | 299 | if (num_copies == 1) |
f188591e | 300 | return ret; |
4235298e | 301 | |
f188591e | 302 | mirror_num++; |
4235298e | 303 | if (mirror_num > num_copies) |
f188591e | 304 | return ret; |
f188591e | 305 | } |
f188591e CM |
306 | return -EIO; |
307 | } | |
19c00ddc | 308 | |
d352ac68 CM |
309 | /* |
310 | * checksum a dirty tree block before IO. This has extra checks to make | |
311 | * sure we only fill in the checksum field in the first page of a multi-page block | |
312 | */ | |
b2950863 | 313 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 314 | { |
d1310b2e | 315 | struct extent_io_tree *tree; |
35ebb934 | 316 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc CM |
317 | u64 found_start; |
318 | int found_level; | |
319 | unsigned long len; | |
320 | struct extent_buffer *eb; | |
f188591e CM |
321 | int ret; |
322 | ||
d1310b2e | 323 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc CM |
324 | |
325 | if (page->private == EXTENT_PAGE_PRIVATE) | |
326 | goto out; | |
327 | if (!page->private) | |
328 | goto out; | |
329 | len = page->private >> 2; | |
330 | if (len == 0) { | |
331 | WARN_ON(1); | |
332 | } | |
333 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); | |
ca7a79ad CM |
334 | ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, |
335 | btrfs_header_generation(eb)); | |
f188591e | 336 | BUG_ON(ret); |
19c00ddc CM |
337 | found_start = btrfs_header_bytenr(eb); |
338 | if (found_start != start) { | |
339 | printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n", | |
340 | start, found_start, len); | |
55c69072 CM |
341 | WARN_ON(1); |
342 | goto err; | |
343 | } | |
344 | if (eb->first_page != page) { | |
345 | printk("bad first page %lu %lu\n", eb->first_page->index, | |
346 | page->index); | |
347 | WARN_ON(1); | |
348 | goto err; | |
349 | } | |
350 | if (!PageUptodate(page)) { | |
351 | printk("csum not up to date page %lu\n", page->index); | |
352 | WARN_ON(1); | |
353 | goto err; | |
19c00ddc CM |
354 | } |
355 | found_level = btrfs_header_level(eb); | |
4bef0848 | 356 | |
19c00ddc | 357 | csum_tree_block(root, eb, 0); |
55c69072 | 358 | err: |
19c00ddc CM |
359 | free_extent_buffer(eb); |
360 | out: | |
361 | return 0; | |
362 | } | |
363 | ||
2b82032c YZ |
364 | static int check_tree_block_fsid(struct btrfs_root *root, |
365 | struct extent_buffer *eb) | |
366 | { | |
367 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
368 | u8 fsid[BTRFS_UUID_SIZE]; | |
369 | int ret = 1; | |
370 | ||
371 | read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), | |
372 | BTRFS_FSID_SIZE); | |
373 | while (fs_devices) { | |
374 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
375 | ret = 0; | |
376 | break; | |
377 | } | |
378 | fs_devices = fs_devices->seed; | |
379 | } | |
380 | return ret; | |
381 | } | |
382 | ||
b2950863 | 383 | static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
ce9adaa5 CM |
384 | struct extent_state *state) |
385 | { | |
386 | struct extent_io_tree *tree; | |
387 | u64 found_start; | |
388 | int found_level; | |
389 | unsigned long len; | |
390 | struct extent_buffer *eb; | |
391 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 392 | int ret = 0; |
ce9adaa5 CM |
393 | |
394 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
395 | if (page->private == EXTENT_PAGE_PRIVATE) | |
396 | goto out; | |
397 | if (!page->private) | |
398 | goto out; | |
399 | len = page->private >> 2; | |
400 | if (len == 0) { | |
401 | WARN_ON(1); | |
402 | } | |
403 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); | |
f188591e | 404 | |
ce9adaa5 | 405 | found_start = btrfs_header_bytenr(eb); |
23a07867 | 406 | if (found_start != start) { |
a1b32a59 CM |
407 | printk("bad tree block start %llu %llu\n", |
408 | (unsigned long long)found_start, | |
409 | (unsigned long long)eb->start); | |
f188591e | 410 | ret = -EIO; |
ce9adaa5 CM |
411 | goto err; |
412 | } | |
413 | if (eb->first_page != page) { | |
414 | printk("bad first page %lu %lu\n", eb->first_page->index, | |
415 | page->index); | |
416 | WARN_ON(1); | |
f188591e | 417 | ret = -EIO; |
ce9adaa5 CM |
418 | goto err; |
419 | } | |
2b82032c | 420 | if (check_tree_block_fsid(root, eb)) { |
1259ab75 CM |
421 | printk("bad fsid on block %Lu\n", eb->start); |
422 | ret = -EIO; | |
423 | goto err; | |
424 | } | |
ce9adaa5 CM |
425 | found_level = btrfs_header_level(eb); |
426 | ||
427 | ret = csum_tree_block(root, eb, 1); | |
f188591e CM |
428 | if (ret) |
429 | ret = -EIO; | |
ce9adaa5 CM |
430 | |
431 | end = min_t(u64, eb->len, PAGE_CACHE_SIZE); | |
432 | end = eb->start + end - 1; | |
ce9adaa5 CM |
433 | err: |
434 | free_extent_buffer(eb); | |
435 | out: | |
f188591e | 436 | return ret; |
ce9adaa5 CM |
437 | } |
438 | ||
ce9adaa5 | 439 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
440 | { |
441 | struct end_io_wq *end_io_wq = bio->bi_private; | |
442 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 443 | |
ce9adaa5 | 444 | fs_info = end_io_wq->info; |
ce9adaa5 | 445 | end_io_wq->error = err; |
8b712842 CM |
446 | end_io_wq->work.func = end_workqueue_fn; |
447 | end_io_wq->work.flags = 0; | |
d20f7043 CM |
448 | |
449 | if (bio->bi_rw & (1 << BIO_RW)) { | |
e6dcd2dc CM |
450 | btrfs_queue_worker(&fs_info->endio_write_workers, |
451 | &end_io_wq->work); | |
d20f7043 CM |
452 | } else { |
453 | if (end_io_wq->metadata) | |
454 | btrfs_queue_worker(&fs_info->endio_meta_workers, | |
455 | &end_io_wq->work); | |
456 | else | |
457 | btrfs_queue_worker(&fs_info->endio_workers, | |
458 | &end_io_wq->work); | |
459 | } | |
ce9adaa5 CM |
460 | } |
461 | ||
22c59948 CM |
462 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
463 | int metadata) | |
0b86a832 | 464 | { |
ce9adaa5 | 465 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
466 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
467 | if (!end_io_wq) | |
468 | return -ENOMEM; | |
469 | ||
470 | end_io_wq->private = bio->bi_private; | |
471 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 472 | end_io_wq->info = info; |
ce9adaa5 CM |
473 | end_io_wq->error = 0; |
474 | end_io_wq->bio = bio; | |
22c59948 | 475 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
476 | |
477 | bio->bi_private = end_io_wq; | |
478 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
479 | return 0; |
480 | } | |
481 | ||
b64a2851 | 482 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 483 | { |
4854ddd0 CM |
484 | unsigned long limit = min_t(unsigned long, |
485 | info->workers.max_workers, | |
486 | info->fs_devices->open_devices); | |
487 | return 256 * limit; | |
488 | } | |
0986fe9e | 489 | |
4854ddd0 CM |
490 | int btrfs_congested_async(struct btrfs_fs_info *info, int iodone) |
491 | { | |
b64a2851 CM |
492 | return atomic_read(&info->nr_async_bios) > |
493 | btrfs_async_submit_limit(info); | |
0986fe9e CM |
494 | } |
495 | ||
4a69a410 CM |
496 | static void run_one_async_start(struct btrfs_work *work) |
497 | { | |
498 | struct btrfs_fs_info *fs_info; | |
499 | struct async_submit_bio *async; | |
500 | ||
501 | async = container_of(work, struct async_submit_bio, work); | |
502 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
503 | async->submit_bio_start(async->inode, async->rw, async->bio, | |
504 | async->mirror_num, async->bio_flags); | |
505 | } | |
506 | ||
507 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
508 | { |
509 | struct btrfs_fs_info *fs_info; | |
510 | struct async_submit_bio *async; | |
4854ddd0 | 511 | int limit; |
8b712842 CM |
512 | |
513 | async = container_of(work, struct async_submit_bio, work); | |
514 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 515 | |
b64a2851 | 516 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
517 | limit = limit * 2 / 3; |
518 | ||
8b712842 | 519 | atomic_dec(&fs_info->nr_async_submits); |
0986fe9e | 520 | |
b64a2851 CM |
521 | if (atomic_read(&fs_info->nr_async_submits) < limit && |
522 | waitqueue_active(&fs_info->async_submit_wait)) | |
4854ddd0 CM |
523 | wake_up(&fs_info->async_submit_wait); |
524 | ||
4a69a410 | 525 | async->submit_bio_done(async->inode, async->rw, async->bio, |
c8b97818 | 526 | async->mirror_num, async->bio_flags); |
4a69a410 CM |
527 | } |
528 | ||
529 | static void run_one_async_free(struct btrfs_work *work) | |
530 | { | |
531 | struct async_submit_bio *async; | |
532 | ||
533 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
534 | kfree(async); |
535 | } | |
536 | ||
44b8bd7e CM |
537 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
538 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 539 | unsigned long bio_flags, |
4a69a410 CM |
540 | extent_submit_bio_hook_t *submit_bio_start, |
541 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
542 | { |
543 | struct async_submit_bio *async; | |
544 | ||
545 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
546 | if (!async) | |
547 | return -ENOMEM; | |
548 | ||
549 | async->inode = inode; | |
550 | async->rw = rw; | |
551 | async->bio = bio; | |
552 | async->mirror_num = mirror_num; | |
4a69a410 CM |
553 | async->submit_bio_start = submit_bio_start; |
554 | async->submit_bio_done = submit_bio_done; | |
555 | ||
556 | async->work.func = run_one_async_start; | |
557 | async->work.ordered_func = run_one_async_done; | |
558 | async->work.ordered_free = run_one_async_free; | |
559 | ||
8b712842 | 560 | async->work.flags = 0; |
c8b97818 | 561 | async->bio_flags = bio_flags; |
8c8bee1d | 562 | |
cb03c743 | 563 | atomic_inc(&fs_info->nr_async_submits); |
8b712842 | 564 | btrfs_queue_worker(&fs_info->workers, &async->work); |
d2c3f4f6 | 565 | #if 0 |
7c2ca468 | 566 | int limit = btrfs_async_submit_limit(fs_info); |
9473f16c CM |
567 | if (atomic_read(&fs_info->nr_async_submits) > limit) { |
568 | wait_event_timeout(fs_info->async_submit_wait, | |
4854ddd0 CM |
569 | (atomic_read(&fs_info->nr_async_submits) < limit), |
570 | HZ/10); | |
9473f16c CM |
571 | |
572 | wait_event_timeout(fs_info->async_submit_wait, | |
573 | (atomic_read(&fs_info->nr_async_bios) < limit), | |
574 | HZ/10); | |
575 | } | |
d2c3f4f6 | 576 | #endif |
771ed689 CM |
577 | while(atomic_read(&fs_info->async_submit_draining) && |
578 | atomic_read(&fs_info->nr_async_submits)) { | |
579 | wait_event(fs_info->async_submit_wait, | |
580 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
581 | } | |
582 | ||
44b8bd7e CM |
583 | return 0; |
584 | } | |
585 | ||
ce3ed71a CM |
586 | static int btree_csum_one_bio(struct bio *bio) |
587 | { | |
588 | struct bio_vec *bvec = bio->bi_io_vec; | |
589 | int bio_index = 0; | |
590 | struct btrfs_root *root; | |
591 | ||
592 | WARN_ON(bio->bi_vcnt <= 0); | |
593 | while(bio_index < bio->bi_vcnt) { | |
594 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; | |
595 | csum_dirty_buffer(root, bvec->bv_page); | |
596 | bio_index++; | |
597 | bvec++; | |
598 | } | |
599 | return 0; | |
600 | } | |
601 | ||
4a69a410 CM |
602 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
603 | struct bio *bio, int mirror_num, | |
604 | unsigned long bio_flags) | |
22c59948 | 605 | { |
8b712842 CM |
606 | /* |
607 | * when we're called for a write, we're already in the async | |
5443be45 | 608 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 609 | */ |
4a69a410 CM |
610 | btree_csum_one_bio(bio); |
611 | return 0; | |
612 | } | |
22c59948 | 613 | |
4a69a410 CM |
614 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
615 | int mirror_num, unsigned long bio_flags) | |
616 | { | |
8b712842 | 617 | /* |
4a69a410 CM |
618 | * when we're called for a write, we're already in the async |
619 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 620 | */ |
8b712842 | 621 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
622 | } |
623 | ||
44b8bd7e | 624 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
c8b97818 | 625 | int mirror_num, unsigned long bio_flags) |
44b8bd7e | 626 | { |
8b712842 CM |
627 | /* |
628 | * kthread helpers are used to submit writes so that checksumming | |
629 | * can happen in parallel across all CPUs | |
630 | */ | |
44b8bd7e | 631 | if (!(rw & (1 << BIO_RW))) { |
4a69a410 CM |
632 | int ret; |
633 | /* | |
634 | * called for a read, do the setup so that checksum validation | |
635 | * can happen in the async kernel threads | |
636 | */ | |
637 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, | |
638 | bio, 1); | |
639 | BUG_ON(ret); | |
640 | ||
641 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
6f3577bd | 642 | mirror_num, 0); |
44b8bd7e CM |
643 | } |
644 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, | |
c8b97818 | 645 | inode, rw, bio, mirror_num, 0, |
4a69a410 CM |
646 | __btree_submit_bio_start, |
647 | __btree_submit_bio_done); | |
44b8bd7e CM |
648 | } |
649 | ||
0da5468f CM |
650 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
651 | { | |
d1310b2e CM |
652 | struct extent_io_tree *tree; |
653 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5443be45 CM |
654 | |
655 | if (current->flags & PF_MEMALLOC) { | |
656 | redirty_page_for_writepage(wbc, page); | |
657 | unlock_page(page); | |
658 | return 0; | |
659 | } | |
5f39d397 CM |
660 | return extent_write_full_page(tree, page, btree_get_extent, wbc); |
661 | } | |
0da5468f CM |
662 | |
663 | static int btree_writepages(struct address_space *mapping, | |
664 | struct writeback_control *wbc) | |
665 | { | |
d1310b2e CM |
666 | struct extent_io_tree *tree; |
667 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 668 | if (wbc->sync_mode == WB_SYNC_NONE) { |
793955bc CM |
669 | u64 num_dirty; |
670 | u64 start = 0; | |
24ab9cd8 | 671 | unsigned long thresh = 32 * 1024 * 1024; |
448d640b CM |
672 | |
673 | if (wbc->for_kupdate) | |
674 | return 0; | |
675 | ||
1832a6d5 CM |
676 | num_dirty = count_range_bits(tree, &start, (u64)-1, |
677 | thresh, EXTENT_DIRTY); | |
793955bc CM |
678 | if (num_dirty < thresh) { |
679 | return 0; | |
680 | } | |
681 | } | |
0da5468f CM |
682 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
683 | } | |
684 | ||
b2950863 | 685 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 686 | { |
d1310b2e CM |
687 | struct extent_io_tree *tree; |
688 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
689 | return extent_read_full_page(tree, page, btree_get_extent); |
690 | } | |
22b0ebda | 691 | |
70dec807 | 692 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 693 | { |
d1310b2e CM |
694 | struct extent_io_tree *tree; |
695 | struct extent_map_tree *map; | |
5f39d397 | 696 | int ret; |
d98237b3 | 697 | |
98509cfc CM |
698 | if (PageWriteback(page) || PageDirty(page)) |
699 | return 0; | |
700 | ||
d1310b2e CM |
701 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
702 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
6af118ce | 703 | |
7b13b7b1 | 704 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
6af118ce CM |
705 | if (!ret) { |
706 | return 0; | |
707 | } | |
708 | ||
709 | ret = try_release_extent_buffer(tree, page); | |
5f39d397 CM |
710 | if (ret == 1) { |
711 | ClearPagePrivate(page); | |
712 | set_page_private(page, 0); | |
713 | page_cache_release(page); | |
714 | } | |
6af118ce | 715 | |
d98237b3 CM |
716 | return ret; |
717 | } | |
718 | ||
5f39d397 | 719 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 720 | { |
d1310b2e CM |
721 | struct extent_io_tree *tree; |
722 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
723 | extent_invalidatepage(tree, page, offset); |
724 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 725 | if (PagePrivate(page)) { |
6af118ce CM |
726 | printk("warning page private not zero on page %Lu\n", |
727 | page_offset(page)); | |
9ad6b7bc CM |
728 | ClearPagePrivate(page); |
729 | set_page_private(page, 0); | |
730 | page_cache_release(page); | |
731 | } | |
d98237b3 CM |
732 | } |
733 | ||
5f39d397 | 734 | #if 0 |
d98237b3 | 735 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
ed2ff2cb | 736 | { |
87cbda5c | 737 | struct buffer_head *bh; |
0f7d52f4 | 738 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
87cbda5c | 739 | struct buffer_head *head; |
87cbda5c CM |
740 | if (!page_has_buffers(page)) { |
741 | create_empty_buffers(page, root->fs_info->sb->s_blocksize, | |
742 | (1 << BH_Dirty)|(1 << BH_Uptodate)); | |
743 | } | |
744 | head = page_buffers(page); | |
745 | bh = head; | |
746 | do { | |
747 | if (buffer_dirty(bh)) | |
748 | csum_tree_block(root, bh, 0); | |
749 | bh = bh->b_this_page; | |
750 | } while (bh != head); | |
d98237b3 | 751 | return block_write_full_page(page, btree_get_block, wbc); |
ed2ff2cb | 752 | } |
5f39d397 | 753 | #endif |
eb60ceac | 754 | |
d98237b3 CM |
755 | static struct address_space_operations btree_aops = { |
756 | .readpage = btree_readpage, | |
757 | .writepage = btree_writepage, | |
0da5468f | 758 | .writepages = btree_writepages, |
5f39d397 CM |
759 | .releasepage = btree_releasepage, |
760 | .invalidatepage = btree_invalidatepage, | |
d98237b3 CM |
761 | .sync_page = block_sync_page, |
762 | }; | |
763 | ||
ca7a79ad CM |
764 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
765 | u64 parent_transid) | |
090d1875 | 766 | { |
5f39d397 CM |
767 | struct extent_buffer *buf = NULL; |
768 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 769 | int ret = 0; |
090d1875 | 770 | |
db94535d | 771 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 772 | if (!buf) |
090d1875 | 773 | return 0; |
d1310b2e | 774 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
f188591e | 775 | buf, 0, 0, btree_get_extent, 0); |
5f39d397 | 776 | free_extent_buffer(buf); |
de428b63 | 777 | return ret; |
090d1875 CM |
778 | } |
779 | ||
0999df54 CM |
780 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
781 | u64 bytenr, u32 blocksize) | |
782 | { | |
783 | struct inode *btree_inode = root->fs_info->btree_inode; | |
784 | struct extent_buffer *eb; | |
785 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
786 | bytenr, blocksize, GFP_NOFS); | |
787 | return eb; | |
788 | } | |
789 | ||
790 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
791 | u64 bytenr, u32 blocksize) | |
792 | { | |
793 | struct inode *btree_inode = root->fs_info->btree_inode; | |
794 | struct extent_buffer *eb; | |
795 | ||
796 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
797 | bytenr, blocksize, NULL, GFP_NOFS); | |
798 | return eb; | |
799 | } | |
800 | ||
801 | ||
e02119d5 CM |
802 | int btrfs_write_tree_block(struct extent_buffer *buf) |
803 | { | |
804 | return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start, | |
24ab9cd8 | 805 | buf->start + buf->len - 1, WB_SYNC_ALL); |
e02119d5 CM |
806 | } |
807 | ||
808 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
809 | { | |
810 | return btrfs_wait_on_page_writeback_range(buf->first_page->mapping, | |
811 | buf->start, buf->start + buf->len -1); | |
812 | } | |
813 | ||
0999df54 | 814 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 815 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
816 | { |
817 | struct extent_buffer *buf = NULL; | |
818 | struct inode *btree_inode = root->fs_info->btree_inode; | |
819 | struct extent_io_tree *io_tree; | |
820 | int ret; | |
821 | ||
822 | io_tree = &BTRFS_I(btree_inode)->io_tree; | |
823 | ||
824 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
825 | if (!buf) | |
826 | return NULL; | |
0999df54 | 827 | |
ca7a79ad | 828 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 CM |
829 | |
830 | if (ret == 0) { | |
831 | buf->flags |= EXTENT_UPTODATE; | |
a1b32a59 CM |
832 | } else { |
833 | WARN_ON(1); | |
ce9adaa5 | 834 | } |
5f39d397 | 835 | return buf; |
ce9adaa5 | 836 | |
eb60ceac CM |
837 | } |
838 | ||
e089f05c | 839 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 840 | struct extent_buffer *buf) |
ed2ff2cb | 841 | { |
5f39d397 | 842 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 | 843 | if (btrfs_header_generation(buf) == |
925baedd CM |
844 | root->fs_info->running_transaction->transid) { |
845 | WARN_ON(!btrfs_tree_locked(buf)); | |
d1310b2e | 846 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 847 | buf); |
925baedd | 848 | } |
5f39d397 CM |
849 | return 0; |
850 | } | |
851 | ||
db94535d | 852 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 853 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 854 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 855 | u64 objectid) |
d97e63b6 | 856 | { |
cfaa7295 | 857 | root->node = NULL; |
a28ec197 | 858 | root->commit_root = NULL; |
31153d81 | 859 | root->ref_tree = NULL; |
db94535d CM |
860 | root->sectorsize = sectorsize; |
861 | root->nodesize = nodesize; | |
862 | root->leafsize = leafsize; | |
87ee04eb | 863 | root->stripesize = stripesize; |
123abc88 | 864 | root->ref_cows = 0; |
0b86a832 CM |
865 | root->track_dirty = 0; |
866 | ||
9f5fae2f | 867 | root->fs_info = fs_info; |
0f7d52f4 CM |
868 | root->objectid = objectid; |
869 | root->last_trans = 0; | |
1b05da2e CM |
870 | root->highest_inode = 0; |
871 | root->last_inode_alloc = 0; | |
58176a96 | 872 | root->name = NULL; |
4313b399 | 873 | root->in_sysfs = 0; |
0b86a832 CM |
874 | |
875 | INIT_LIST_HEAD(&root->dirty_list); | |
7b128766 | 876 | INIT_LIST_HEAD(&root->orphan_list); |
bcc63abb | 877 | INIT_LIST_HEAD(&root->dead_list); |
925baedd | 878 | spin_lock_init(&root->node_lock); |
bcc63abb | 879 | spin_lock_init(&root->list_lock); |
a2135011 | 880 | mutex_init(&root->objectid_mutex); |
e02119d5 | 881 | mutex_init(&root->log_mutex); |
d0c803c4 CM |
882 | extent_io_tree_init(&root->dirty_log_pages, |
883 | fs_info->btree_inode->i_mapping, GFP_NOFS); | |
017e5369 CM |
884 | |
885 | btrfs_leaf_ref_tree_init(&root->ref_tree_struct); | |
886 | root->ref_tree = &root->ref_tree_struct; | |
887 | ||
3768f368 CM |
888 | memset(&root->root_key, 0, sizeof(root->root_key)); |
889 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 890 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 891 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 892 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 893 | init_completion(&root->kobj_unregister); |
6702ed49 CM |
894 | root->defrag_running = 0; |
895 | root->defrag_level = 0; | |
4d775673 | 896 | root->root_key.objectid = objectid; |
3394e160 CM |
897 | root->anon_super.s_root = NULL; |
898 | root->anon_super.s_dev = 0; | |
899 | INIT_LIST_HEAD(&root->anon_super.s_list); | |
900 | INIT_LIST_HEAD(&root->anon_super.s_instances); | |
901 | init_rwsem(&root->anon_super.s_umount); | |
902 | ||
3768f368 CM |
903 | return 0; |
904 | } | |
905 | ||
db94535d | 906 | static int find_and_setup_root(struct btrfs_root *tree_root, |
9f5fae2f CM |
907 | struct btrfs_fs_info *fs_info, |
908 | u64 objectid, | |
e20d96d6 | 909 | struct btrfs_root *root) |
3768f368 CM |
910 | { |
911 | int ret; | |
db94535d | 912 | u32 blocksize; |
84234f3a | 913 | u64 generation; |
3768f368 | 914 | |
db94535d | 915 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
916 | tree_root->sectorsize, tree_root->stripesize, |
917 | root, fs_info, objectid); | |
3768f368 CM |
918 | ret = btrfs_find_last_root(tree_root, objectid, |
919 | &root->root_item, &root->root_key); | |
920 | BUG_ON(ret); | |
921 | ||
84234f3a | 922 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
923 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
924 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 925 | blocksize, generation); |
3768f368 | 926 | BUG_ON(!root->node); |
d97e63b6 CM |
927 | return 0; |
928 | } | |
929 | ||
e02119d5 CM |
930 | int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans, |
931 | struct btrfs_fs_info *fs_info) | |
932 | { | |
933 | struct extent_buffer *eb; | |
d0c803c4 CM |
934 | struct btrfs_root *log_root_tree = fs_info->log_root_tree; |
935 | u64 start = 0; | |
936 | u64 end = 0; | |
e02119d5 CM |
937 | int ret; |
938 | ||
d0c803c4 | 939 | if (!log_root_tree) |
e02119d5 CM |
940 | return 0; |
941 | ||
d0c803c4 CM |
942 | while(1) { |
943 | ret = find_first_extent_bit(&log_root_tree->dirty_log_pages, | |
944 | 0, &start, &end, EXTENT_DIRTY); | |
945 | if (ret) | |
946 | break; | |
947 | ||
948 | clear_extent_dirty(&log_root_tree->dirty_log_pages, | |
949 | start, end, GFP_NOFS); | |
950 | } | |
e02119d5 CM |
951 | eb = fs_info->log_root_tree->node; |
952 | ||
953 | WARN_ON(btrfs_header_level(eb) != 0); | |
954 | WARN_ON(btrfs_header_nritems(eb) != 0); | |
955 | ||
d00aff00 CM |
956 | ret = btrfs_free_reserved_extent(fs_info->tree_root, |
957 | eb->start, eb->len); | |
e02119d5 CM |
958 | BUG_ON(ret); |
959 | ||
960 | free_extent_buffer(eb); | |
961 | kfree(fs_info->log_root_tree); | |
962 | fs_info->log_root_tree = NULL; | |
963 | return 0; | |
964 | } | |
965 | ||
966 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
967 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
968 | { |
969 | struct btrfs_root *root; | |
970 | struct btrfs_root *tree_root = fs_info->tree_root; | |
e02119d5 CM |
971 | |
972 | root = kzalloc(sizeof(*root), GFP_NOFS); | |
973 | if (!root) | |
974 | return -ENOMEM; | |
975 | ||
976 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
977 | tree_root->sectorsize, tree_root->stripesize, | |
978 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
979 | ||
980 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
981 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
982 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
983 | root->ref_cows = 0; | |
984 | ||
985 | root->node = btrfs_alloc_free_block(trans, root, root->leafsize, | |
31840ae1 ZY |
986 | 0, BTRFS_TREE_LOG_OBJECTID, |
987 | trans->transid, 0, 0, 0); | |
e02119d5 CM |
988 | |
989 | btrfs_set_header_nritems(root->node, 0); | |
990 | btrfs_set_header_level(root->node, 0); | |
991 | btrfs_set_header_bytenr(root->node, root->node->start); | |
992 | btrfs_set_header_generation(root->node, trans->transid); | |
993 | btrfs_set_header_owner(root->node, BTRFS_TREE_LOG_OBJECTID); | |
994 | ||
995 | write_extent_buffer(root->node, root->fs_info->fsid, | |
996 | (unsigned long)btrfs_header_fsid(root->node), | |
997 | BTRFS_FSID_SIZE); | |
998 | btrfs_mark_buffer_dirty(root->node); | |
999 | btrfs_tree_unlock(root->node); | |
1000 | fs_info->log_root_tree = root; | |
1001 | return 0; | |
1002 | } | |
1003 | ||
1004 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
1005 | struct btrfs_key *location) | |
1006 | { | |
1007 | struct btrfs_root *root; | |
1008 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1009 | struct btrfs_path *path; |
5f39d397 | 1010 | struct extent_buffer *l; |
1b05da2e | 1011 | u64 highest_inode; |
84234f3a | 1012 | u64 generation; |
db94535d | 1013 | u32 blocksize; |
0f7d52f4 CM |
1014 | int ret = 0; |
1015 | ||
5eda7b5e | 1016 | root = kzalloc(sizeof(*root), GFP_NOFS); |
0cf6c620 | 1017 | if (!root) |
0f7d52f4 | 1018 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 1019 | if (location->offset == (u64)-1) { |
db94535d | 1020 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
1021 | location->objectid, root); |
1022 | if (ret) { | |
0f7d52f4 CM |
1023 | kfree(root); |
1024 | return ERR_PTR(ret); | |
1025 | } | |
1026 | goto insert; | |
1027 | } | |
1028 | ||
db94535d | 1029 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1030 | tree_root->sectorsize, tree_root->stripesize, |
1031 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
1032 | |
1033 | path = btrfs_alloc_path(); | |
1034 | BUG_ON(!path); | |
1035 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); | |
1036 | if (ret != 0) { | |
0f7d52f4 CM |
1037 | if (ret > 0) |
1038 | ret = -ENOENT; | |
1039 | goto out; | |
1040 | } | |
5f39d397 CM |
1041 | l = path->nodes[0]; |
1042 | read_extent_buffer(l, &root->root_item, | |
1043 | btrfs_item_ptr_offset(l, path->slots[0]), | |
0f7d52f4 | 1044 | sizeof(root->root_item)); |
44b36eb2 | 1045 | memcpy(&root->root_key, location, sizeof(*location)); |
0f7d52f4 CM |
1046 | ret = 0; |
1047 | out: | |
1048 | btrfs_release_path(root, path); | |
1049 | btrfs_free_path(path); | |
1050 | if (ret) { | |
1051 | kfree(root); | |
1052 | return ERR_PTR(ret); | |
1053 | } | |
84234f3a | 1054 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1055 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1056 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1057 | blocksize, generation); |
0f7d52f4 CM |
1058 | BUG_ON(!root->node); |
1059 | insert: | |
e02119d5 CM |
1060 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { |
1061 | root->ref_cows = 1; | |
1062 | ret = btrfs_find_highest_inode(root, &highest_inode); | |
1063 | if (ret == 0) { | |
1064 | root->highest_inode = highest_inode; | |
1065 | root->last_inode_alloc = highest_inode; | |
1066 | } | |
5eda7b5e CM |
1067 | } |
1068 | return root; | |
1069 | } | |
1070 | ||
dc17ff8f CM |
1071 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1072 | u64 root_objectid) | |
1073 | { | |
1074 | struct btrfs_root *root; | |
1075 | ||
1076 | if (root_objectid == BTRFS_ROOT_TREE_OBJECTID) | |
1077 | return fs_info->tree_root; | |
1078 | if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1079 | return fs_info->extent_root; | |
1080 | ||
1081 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1082 | (unsigned long)root_objectid); | |
1083 | return root; | |
1084 | } | |
1085 | ||
edbd8d4e CM |
1086 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
1087 | struct btrfs_key *location) | |
5eda7b5e CM |
1088 | { |
1089 | struct btrfs_root *root; | |
1090 | int ret; | |
1091 | ||
edbd8d4e CM |
1092 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1093 | return fs_info->tree_root; | |
1094 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1095 | return fs_info->extent_root; | |
8f18cf13 CM |
1096 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1097 | return fs_info->chunk_root; | |
1098 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1099 | return fs_info->dev_root; | |
edbd8d4e | 1100 | |
5eda7b5e CM |
1101 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
1102 | (unsigned long)location->objectid); | |
1103 | if (root) | |
1104 | return root; | |
1105 | ||
e02119d5 | 1106 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
1107 | if (IS_ERR(root)) |
1108 | return root; | |
3394e160 CM |
1109 | |
1110 | set_anon_super(&root->anon_super, NULL); | |
1111 | ||
2619ba1f CM |
1112 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
1113 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 CM |
1114 | root); |
1115 | if (ret) { | |
5f39d397 | 1116 | free_extent_buffer(root->node); |
0f7d52f4 CM |
1117 | kfree(root); |
1118 | return ERR_PTR(ret); | |
1119 | } | |
c146afad YZ |
1120 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
1121 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
1122 | root->root_key.objectid, root); | |
1123 | BUG_ON(ret); | |
1124 | btrfs_orphan_cleanup(root); | |
1125 | } | |
edbd8d4e CM |
1126 | return root; |
1127 | } | |
1128 | ||
1129 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, | |
1130 | struct btrfs_key *location, | |
1131 | const char *name, int namelen) | |
1132 | { | |
1133 | struct btrfs_root *root; | |
1134 | int ret; | |
1135 | ||
1136 | root = btrfs_read_fs_root_no_name(fs_info, location); | |
1137 | if (!root) | |
1138 | return NULL; | |
58176a96 | 1139 | |
4313b399 CM |
1140 | if (root->in_sysfs) |
1141 | return root; | |
1142 | ||
58176a96 JB |
1143 | ret = btrfs_set_root_name(root, name, namelen); |
1144 | if (ret) { | |
5f39d397 | 1145 | free_extent_buffer(root->node); |
58176a96 JB |
1146 | kfree(root); |
1147 | return ERR_PTR(ret); | |
1148 | } | |
ea9e8b11 | 1149 | #if 0 |
58176a96 JB |
1150 | ret = btrfs_sysfs_add_root(root); |
1151 | if (ret) { | |
5f39d397 | 1152 | free_extent_buffer(root->node); |
58176a96 JB |
1153 | kfree(root->name); |
1154 | kfree(root); | |
1155 | return ERR_PTR(ret); | |
1156 | } | |
ea9e8b11 | 1157 | #endif |
4313b399 | 1158 | root->in_sysfs = 1; |
0f7d52f4 CM |
1159 | return root; |
1160 | } | |
19c00ddc CM |
1161 | #if 0 |
1162 | static int add_hasher(struct btrfs_fs_info *info, char *type) { | |
1163 | struct btrfs_hasher *hasher; | |
1164 | ||
1165 | hasher = kmalloc(sizeof(*hasher), GFP_NOFS); | |
1166 | if (!hasher) | |
1167 | return -ENOMEM; | |
1168 | hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC); | |
1169 | if (!hasher->hash_tfm) { | |
1170 | kfree(hasher); | |
1171 | return -EINVAL; | |
1172 | } | |
1173 | spin_lock(&info->hash_lock); | |
1174 | list_add(&hasher->list, &info->hashers); | |
1175 | spin_unlock(&info->hash_lock); | |
1176 | return 0; | |
1177 | } | |
1178 | #endif | |
04160088 CM |
1179 | |
1180 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) | |
1181 | { | |
1182 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1183 | int ret = 0; | |
1184 | struct list_head *cur; | |
1185 | struct btrfs_device *device; | |
1186 | struct backing_dev_info *bdi; | |
6f3577bd | 1187 | #if 0 |
cb03c743 | 1188 | if ((bdi_bits & (1 << BDI_write_congested)) && |
777e6bd7 | 1189 | btrfs_congested_async(info, 0)) |
cb03c743 | 1190 | return 1; |
6f3577bd | 1191 | #endif |
04160088 CM |
1192 | list_for_each(cur, &info->fs_devices->devices) { |
1193 | device = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1194 | if (!device->bdev) |
1195 | continue; | |
04160088 CM |
1196 | bdi = blk_get_backing_dev_info(device->bdev); |
1197 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1198 | ret = 1; | |
1199 | break; | |
1200 | } | |
1201 | } | |
1202 | return ret; | |
1203 | } | |
1204 | ||
38b66988 CM |
1205 | /* |
1206 | * this unplugs every device on the box, and it is only used when page | |
1207 | * is null | |
1208 | */ | |
1209 | static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) | |
1210 | { | |
1211 | struct list_head *cur; | |
1212 | struct btrfs_device *device; | |
1213 | struct btrfs_fs_info *info; | |
1214 | ||
1215 | info = (struct btrfs_fs_info *)bdi->unplug_io_data; | |
1216 | list_for_each(cur, &info->fs_devices->devices) { | |
1217 | device = list_entry(cur, struct btrfs_device, dev_list); | |
d20f7043 CM |
1218 | if (!device->bdev) |
1219 | continue; | |
1220 | ||
38b66988 CM |
1221 | bdi = blk_get_backing_dev_info(device->bdev); |
1222 | if (bdi->unplug_io_fn) { | |
1223 | bdi->unplug_io_fn(bdi, page); | |
1224 | } | |
1225 | } | |
1226 | } | |
1227 | ||
b2950863 | 1228 | static void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) |
04160088 | 1229 | { |
38b66988 | 1230 | struct inode *inode; |
f2d8d74d CM |
1231 | struct extent_map_tree *em_tree; |
1232 | struct extent_map *em; | |
bcbfce8a | 1233 | struct address_space *mapping; |
38b66988 CM |
1234 | u64 offset; |
1235 | ||
bcbfce8a | 1236 | /* the generic O_DIRECT read code does this */ |
9f0ba5bd | 1237 | if (1 || !page) { |
38b66988 CM |
1238 | __unplug_io_fn(bdi, page); |
1239 | return; | |
1240 | } | |
1241 | ||
bcbfce8a CM |
1242 | /* |
1243 | * page->mapping may change at any time. Get a consistent copy | |
1244 | * and use that for everything below | |
1245 | */ | |
1246 | smp_mb(); | |
1247 | mapping = page->mapping; | |
1248 | if (!mapping) | |
1249 | return; | |
1250 | ||
1251 | inode = mapping->host; | |
240d5d48 CM |
1252 | |
1253 | /* | |
1254 | * don't do the expensive searching for a small number of | |
1255 | * devices | |
1256 | */ | |
1257 | if (BTRFS_I(inode)->root->fs_info->fs_devices->open_devices <= 2) { | |
1258 | __unplug_io_fn(bdi, page); | |
1259 | return; | |
1260 | } | |
1261 | ||
38b66988 | 1262 | offset = page_offset(page); |
04160088 | 1263 | |
f2d8d74d CM |
1264 | em_tree = &BTRFS_I(inode)->extent_tree; |
1265 | spin_lock(&em_tree->lock); | |
1266 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); | |
1267 | spin_unlock(&em_tree->lock); | |
89642229 CM |
1268 | if (!em) { |
1269 | __unplug_io_fn(bdi, page); | |
f2d8d74d | 1270 | return; |
89642229 | 1271 | } |
f2d8d74d | 1272 | |
89642229 CM |
1273 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { |
1274 | free_extent_map(em); | |
1275 | __unplug_io_fn(bdi, page); | |
1276 | return; | |
1277 | } | |
f2d8d74d CM |
1278 | offset = offset - em->start; |
1279 | btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree, | |
1280 | em->block_start + offset, page); | |
1281 | free_extent_map(em); | |
04160088 CM |
1282 | } |
1283 | ||
1284 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) | |
1285 | { | |
1286 | bdi_init(bdi); | |
4575c9cc | 1287 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1288 | bdi->state = 0; |
1289 | bdi->capabilities = default_backing_dev_info.capabilities; | |
1290 | bdi->unplug_io_fn = btrfs_unplug_io_fn; | |
1291 | bdi->unplug_io_data = info; | |
1292 | bdi->congested_fn = btrfs_congested_fn; | |
1293 | bdi->congested_data = info; | |
1294 | return 0; | |
1295 | } | |
1296 | ||
ce9adaa5 CM |
1297 | static int bio_ready_for_csum(struct bio *bio) |
1298 | { | |
1299 | u64 length = 0; | |
1300 | u64 buf_len = 0; | |
1301 | u64 start = 0; | |
1302 | struct page *page; | |
1303 | struct extent_io_tree *io_tree = NULL; | |
1304 | struct btrfs_fs_info *info = NULL; | |
1305 | struct bio_vec *bvec; | |
1306 | int i; | |
1307 | int ret; | |
1308 | ||
1309 | bio_for_each_segment(bvec, bio, i) { | |
1310 | page = bvec->bv_page; | |
1311 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
1312 | length += bvec->bv_len; | |
1313 | continue; | |
1314 | } | |
1315 | if (!page->private) { | |
1316 | length += bvec->bv_len; | |
1317 | continue; | |
1318 | } | |
1319 | length = bvec->bv_len; | |
1320 | buf_len = page->private >> 2; | |
1321 | start = page_offset(page) + bvec->bv_offset; | |
1322 | io_tree = &BTRFS_I(page->mapping->host)->io_tree; | |
1323 | info = BTRFS_I(page->mapping->host)->root->fs_info; | |
1324 | } | |
1325 | /* are we fully contained in this bio? */ | |
1326 | if (buf_len <= length) | |
1327 | return 1; | |
1328 | ||
1329 | ret = extent_range_uptodate(io_tree, start + length, | |
1330 | start + buf_len - 1); | |
1331 | if (ret == 1) | |
1332 | return ret; | |
1333 | return ret; | |
1334 | } | |
1335 | ||
8b712842 CM |
1336 | /* |
1337 | * called by the kthread helper functions to finally call the bio end_io | |
1338 | * functions. This is where read checksum verification actually happens | |
1339 | */ | |
1340 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1341 | { |
ce9adaa5 | 1342 | struct bio *bio; |
8b712842 CM |
1343 | struct end_io_wq *end_io_wq; |
1344 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1345 | int error; |
ce9adaa5 | 1346 | |
8b712842 CM |
1347 | end_io_wq = container_of(work, struct end_io_wq, work); |
1348 | bio = end_io_wq->bio; | |
1349 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1350 | |
8b712842 CM |
1351 | /* metadata bios are special because the whole tree block must |
1352 | * be checksummed at once. This makes sure the entire block is in | |
1353 | * ram and up to date before trying to verify things. For | |
1354 | * blocksize <= pagesize, it is basically a noop | |
1355 | */ | |
1356 | if (end_io_wq->metadata && !bio_ready_for_csum(bio)) { | |
d20f7043 | 1357 | btrfs_queue_worker(&fs_info->endio_meta_workers, |
8b712842 CM |
1358 | &end_io_wq->work); |
1359 | return; | |
1360 | } | |
1361 | error = end_io_wq->error; | |
1362 | bio->bi_private = end_io_wq->private; | |
1363 | bio->bi_end_io = end_io_wq->end_io; | |
1364 | kfree(end_io_wq); | |
8b712842 | 1365 | bio_endio(bio, error); |
44b8bd7e CM |
1366 | } |
1367 | ||
a74a4b97 CM |
1368 | static int cleaner_kthread(void *arg) |
1369 | { | |
1370 | struct btrfs_root *root = arg; | |
1371 | ||
1372 | do { | |
1373 | smp_mb(); | |
1374 | if (root->fs_info->closing) | |
1375 | break; | |
1376 | ||
1377 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1378 | mutex_lock(&root->fs_info->cleaner_mutex); | |
a74a4b97 | 1379 | btrfs_clean_old_snapshots(root); |
a74a4b97 CM |
1380 | mutex_unlock(&root->fs_info->cleaner_mutex); |
1381 | ||
1382 | if (freezing(current)) { | |
1383 | refrigerator(); | |
1384 | } else { | |
1385 | smp_mb(); | |
1386 | if (root->fs_info->closing) | |
1387 | break; | |
1388 | set_current_state(TASK_INTERRUPTIBLE); | |
1389 | schedule(); | |
1390 | __set_current_state(TASK_RUNNING); | |
1391 | } | |
1392 | } while (!kthread_should_stop()); | |
1393 | return 0; | |
1394 | } | |
1395 | ||
1396 | static int transaction_kthread(void *arg) | |
1397 | { | |
1398 | struct btrfs_root *root = arg; | |
1399 | struct btrfs_trans_handle *trans; | |
1400 | struct btrfs_transaction *cur; | |
1401 | unsigned long now; | |
1402 | unsigned long delay; | |
1403 | int ret; | |
1404 | ||
1405 | do { | |
1406 | smp_mb(); | |
1407 | if (root->fs_info->closing) | |
1408 | break; | |
1409 | ||
1410 | delay = HZ * 30; | |
1411 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1412 | mutex_lock(&root->fs_info->transaction_kthread_mutex); | |
1413 | ||
ab78c84d CM |
1414 | if (root->fs_info->total_ref_cache_size > 20 * 1024 * 1024) { |
1415 | printk("btrfs: total reference cache size %Lu\n", | |
1416 | root->fs_info->total_ref_cache_size); | |
1417 | } | |
31153d81 | 1418 | |
a74a4b97 CM |
1419 | mutex_lock(&root->fs_info->trans_mutex); |
1420 | cur = root->fs_info->running_transaction; | |
1421 | if (!cur) { | |
1422 | mutex_unlock(&root->fs_info->trans_mutex); | |
1423 | goto sleep; | |
1424 | } | |
31153d81 | 1425 | |
a74a4b97 CM |
1426 | now = get_seconds(); |
1427 | if (now < cur->start_time || now - cur->start_time < 30) { | |
1428 | mutex_unlock(&root->fs_info->trans_mutex); | |
1429 | delay = HZ * 5; | |
1430 | goto sleep; | |
1431 | } | |
1432 | mutex_unlock(&root->fs_info->trans_mutex); | |
a74a4b97 CM |
1433 | trans = btrfs_start_transaction(root, 1); |
1434 | ret = btrfs_commit_transaction(trans, root); | |
1435 | sleep: | |
1436 | wake_up_process(root->fs_info->cleaner_kthread); | |
1437 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1438 | ||
1439 | if (freezing(current)) { | |
1440 | refrigerator(); | |
1441 | } else { | |
1442 | if (root->fs_info->closing) | |
1443 | break; | |
1444 | set_current_state(TASK_INTERRUPTIBLE); | |
1445 | schedule_timeout(delay); | |
1446 | __set_current_state(TASK_RUNNING); | |
1447 | } | |
1448 | } while (!kthread_should_stop()); | |
1449 | return 0; | |
1450 | } | |
1451 | ||
8a4b83cc | 1452 | struct btrfs_root *open_ctree(struct super_block *sb, |
dfe25020 CM |
1453 | struct btrfs_fs_devices *fs_devices, |
1454 | char *options) | |
2e635a27 | 1455 | { |
db94535d CM |
1456 | u32 sectorsize; |
1457 | u32 nodesize; | |
1458 | u32 leafsize; | |
1459 | u32 blocksize; | |
87ee04eb | 1460 | u32 stripesize; |
84234f3a | 1461 | u64 generation; |
f2b636e8 | 1462 | u64 features; |
3de4586c | 1463 | struct btrfs_key location; |
a061fc8d | 1464 | struct buffer_head *bh; |
e02119d5 | 1465 | struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1466 | GFP_NOFS); |
d20f7043 CM |
1467 | struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root), |
1468 | GFP_NOFS); | |
e02119d5 | 1469 | struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1470 | GFP_NOFS); |
8790d502 | 1471 | struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), |
e20d96d6 | 1472 | GFP_NOFS); |
e02119d5 | 1473 | struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1474 | GFP_NOFS); |
e02119d5 | 1475 | struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1476 | GFP_NOFS); |
e02119d5 CM |
1477 | struct btrfs_root *log_tree_root; |
1478 | ||
eb60ceac | 1479 | int ret; |
e58ca020 | 1480 | int err = -EINVAL; |
4543df7e | 1481 | |
2c90e5d6 | 1482 | struct btrfs_super_block *disk_super; |
8790d502 | 1483 | |
0463bb4e | 1484 | if (!extent_root || !tree_root || !fs_info || |
d20f7043 | 1485 | !chunk_root || !dev_root || !csum_root) { |
39279cc3 CM |
1486 | err = -ENOMEM; |
1487 | goto fail; | |
1488 | } | |
0f7d52f4 | 1489 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS); |
8fd17795 | 1490 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1491 | INIT_LIST_HEAD(&fs_info->dead_roots); |
19c00ddc | 1492 | INIT_LIST_HEAD(&fs_info->hashers); |
ea8c2819 | 1493 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
19c00ddc | 1494 | spin_lock_init(&fs_info->hash_lock); |
1832a6d5 | 1495 | spin_lock_init(&fs_info->delalloc_lock); |
cee36a03 | 1496 | spin_lock_init(&fs_info->new_trans_lock); |
31153d81 | 1497 | spin_lock_init(&fs_info->ref_cache_lock); |
19c00ddc | 1498 | |
58176a96 | 1499 | init_completion(&fs_info->kobj_unregister); |
9f5fae2f CM |
1500 | fs_info->tree_root = tree_root; |
1501 | fs_info->extent_root = extent_root; | |
d20f7043 | 1502 | fs_info->csum_root = csum_root; |
0b86a832 CM |
1503 | fs_info->chunk_root = chunk_root; |
1504 | fs_info->dev_root = dev_root; | |
8a4b83cc | 1505 | fs_info->fs_devices = fs_devices; |
0b86a832 | 1506 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1507 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1508 | btrfs_mapping_init(&fs_info->mapping_tree); |
cb03c743 | 1509 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 1510 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 1511 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 1512 | atomic_set(&fs_info->nr_async_bios, 0); |
a2135011 | 1513 | atomic_set(&fs_info->throttles, 0); |
ab78c84d | 1514 | atomic_set(&fs_info->throttle_gen, 0); |
e20d96d6 | 1515 | fs_info->sb = sb; |
c59f8951 | 1516 | fs_info->max_extent = (u64)-1; |
6f568d35 | 1517 | fs_info->max_inline = 8192 * 1024; |
04160088 | 1518 | setup_bdi(fs_info, &fs_info->bdi); |
d98237b3 CM |
1519 | fs_info->btree_inode = new_inode(sb); |
1520 | fs_info->btree_inode->i_ino = 1; | |
2c90e5d6 | 1521 | fs_info->btree_inode->i_nlink = 1; |
c8b97818 | 1522 | |
4543df7e | 1523 | fs_info->thread_pool_size = min(num_online_cpus() + 2, 8); |
0afbaf8c | 1524 | |
3eaa2885 CM |
1525 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
1526 | spin_lock_init(&fs_info->ordered_extent_lock); | |
1527 | ||
a061fc8d CM |
1528 | sb->s_blocksize = 4096; |
1529 | sb->s_blocksize_bits = blksize_bits(4096); | |
1530 | ||
0afbaf8c CM |
1531 | /* |
1532 | * we set the i_size on the btree inode to the max possible int. | |
1533 | * the real end of the address space is determined by all of | |
1534 | * the devices in the system | |
1535 | */ | |
1536 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 1537 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
1538 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
1539 | ||
d1310b2e | 1540 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
5f39d397 CM |
1541 | fs_info->btree_inode->i_mapping, |
1542 | GFP_NOFS); | |
d1310b2e CM |
1543 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, |
1544 | GFP_NOFS); | |
1545 | ||
1546 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 1547 | |
0f9dd46c JB |
1548 | spin_lock_init(&fs_info->block_group_cache_lock); |
1549 | fs_info->block_group_cache_tree.rb_node = NULL; | |
1550 | ||
d1310b2e | 1551 | extent_io_tree_init(&fs_info->pinned_extents, |
1a5bc167 | 1552 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1553 | extent_io_tree_init(&fs_info->pending_del, |
1a5bc167 | 1554 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1555 | extent_io_tree_init(&fs_info->extent_ins, |
1a5bc167 | 1556 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
e66f709b | 1557 | fs_info->do_barriers = 1; |
e18e4809 | 1558 | |
1a40e23b ZY |
1559 | INIT_LIST_HEAD(&fs_info->dead_reloc_roots); |
1560 | btrfs_leaf_ref_tree_init(&fs_info->reloc_ref_tree); | |
e4657689 ZY |
1561 | btrfs_leaf_ref_tree_init(&fs_info->shared_ref_tree); |
1562 | ||
0f7d52f4 CM |
1563 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
1564 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
1565 | sizeof(struct btrfs_key)); | |
22b0ebda | 1566 | insert_inode_hash(fs_info->btree_inode); |
39279cc3 | 1567 | |
79154b1b | 1568 | mutex_init(&fs_info->trans_mutex); |
e02119d5 | 1569 | mutex_init(&fs_info->tree_log_mutex); |
a2135011 | 1570 | mutex_init(&fs_info->drop_mutex); |
25179201 JB |
1571 | mutex_init(&fs_info->extent_ins_mutex); |
1572 | mutex_init(&fs_info->pinned_mutex); | |
925baedd | 1573 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
1574 | mutex_init(&fs_info->transaction_kthread_mutex); |
1575 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 1576 | mutex_init(&fs_info->volume_mutex); |
1a40e23b | 1577 | mutex_init(&fs_info->tree_reloc_mutex); |
e6dcd2dc | 1578 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 1579 | init_waitqueue_head(&fs_info->transaction_wait); |
4854ddd0 | 1580 | init_waitqueue_head(&fs_info->async_submit_wait); |
e02119d5 CM |
1581 | init_waitqueue_head(&fs_info->tree_log_wait); |
1582 | atomic_set(&fs_info->tree_log_commit, 0); | |
1583 | atomic_set(&fs_info->tree_log_writers, 0); | |
1584 | fs_info->tree_log_transid = 0; | |
3768f368 | 1585 | |
19c00ddc CM |
1586 | #if 0 |
1587 | ret = add_hasher(fs_info, "crc32c"); | |
1588 | if (ret) { | |
1589 | printk("btrfs: failed hash setup, modprobe cryptomgr?\n"); | |
1590 | err = -ENOMEM; | |
1591 | goto fail_iput; | |
1592 | } | |
1593 | #endif | |
0b86a832 | 1594 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 1595 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 1596 | |
d98237b3 | 1597 | |
a061fc8d CM |
1598 | bh = __bread(fs_devices->latest_bdev, |
1599 | BTRFS_SUPER_INFO_OFFSET / 4096, 4096); | |
1600 | if (!bh) | |
39279cc3 | 1601 | goto fail_iput; |
39279cc3 | 1602 | |
a061fc8d CM |
1603 | memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy)); |
1604 | brelse(bh); | |
5f39d397 | 1605 | |
a061fc8d | 1606 | memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE); |
0b86a832 | 1607 | |
5f39d397 | 1608 | disk_super = &fs_info->super_copy; |
0f7d52f4 | 1609 | if (!btrfs_super_root(disk_super)) |
c6e2bac1 | 1610 | goto fail_iput; |
0f7d52f4 | 1611 | |
2b82032c YZ |
1612 | ret = btrfs_parse_options(tree_root, options); |
1613 | if (ret) { | |
1614 | err = ret; | |
c6e2bac1 | 1615 | goto fail_iput; |
2b82032c | 1616 | } |
dfe25020 | 1617 | |
f2b636e8 JB |
1618 | features = btrfs_super_incompat_flags(disk_super) & |
1619 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
1620 | if (features) { | |
1621 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
1622 | "unsupported optional features (%Lx).\n", | |
1623 | features); | |
1624 | err = -EINVAL; | |
c6e2bac1 | 1625 | goto fail_iput; |
f2b636e8 JB |
1626 | } |
1627 | ||
1628 | features = btrfs_super_compat_ro_flags(disk_super) & | |
1629 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
1630 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
1631 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
1632 | "unsupported option features (%Lx).\n", | |
1633 | features); | |
1634 | err = -EINVAL; | |
c6e2bac1 | 1635 | goto fail_iput; |
f2b636e8 JB |
1636 | } |
1637 | ||
4543df7e CM |
1638 | /* |
1639 | * we need to start all the end_io workers up front because the | |
1640 | * queue work function gets called at interrupt time, and so it | |
1641 | * cannot dynamically grow. | |
1642 | */ | |
5443be45 CM |
1643 | btrfs_init_workers(&fs_info->workers, "worker", |
1644 | fs_info->thread_pool_size); | |
c8b97818 | 1645 | |
771ed689 CM |
1646 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
1647 | fs_info->thread_pool_size); | |
1648 | ||
5443be45 | 1649 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 CM |
1650 | min_t(u64, fs_devices->num_devices, |
1651 | fs_info->thread_pool_size)); | |
61b49440 CM |
1652 | |
1653 | /* a higher idle thresh on the submit workers makes it much more | |
1654 | * likely that bios will be send down in a sane order to the | |
1655 | * devices | |
1656 | */ | |
1657 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 1658 | |
771ed689 | 1659 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 1660 | fs_info->workers.ordered = 1; |
61b49440 | 1661 | |
771ed689 CM |
1662 | fs_info->delalloc_workers.idle_thresh = 2; |
1663 | fs_info->delalloc_workers.ordered = 1; | |
1664 | ||
5443be45 CM |
1665 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1); |
1666 | btrfs_init_workers(&fs_info->endio_workers, "endio", | |
1667 | fs_info->thread_pool_size); | |
d20f7043 CM |
1668 | btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", |
1669 | fs_info->thread_pool_size); | |
5443be45 | 1670 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", |
e6dcd2dc | 1671 | fs_info->thread_pool_size); |
61b49440 CM |
1672 | |
1673 | /* | |
1674 | * endios are largely parallel and should have a very | |
1675 | * low idle thresh | |
1676 | */ | |
1677 | fs_info->endio_workers.idle_thresh = 4; | |
3a5f1d45 | 1678 | fs_info->endio_write_workers.idle_thresh = 64; |
61b49440 | 1679 | |
4543df7e | 1680 | btrfs_start_workers(&fs_info->workers, 1); |
1cc127b5 | 1681 | btrfs_start_workers(&fs_info->submit_workers, 1); |
771ed689 | 1682 | btrfs_start_workers(&fs_info->delalloc_workers, 1); |
247e743c | 1683 | btrfs_start_workers(&fs_info->fixup_workers, 1); |
4543df7e | 1684 | btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size); |
d20f7043 CM |
1685 | btrfs_start_workers(&fs_info->endio_meta_workers, |
1686 | fs_info->thread_pool_size); | |
e6dcd2dc CM |
1687 | btrfs_start_workers(&fs_info->endio_write_workers, |
1688 | fs_info->thread_pool_size); | |
4543df7e | 1689 | |
4575c9cc | 1690 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
1691 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
1692 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 1693 | |
db94535d CM |
1694 | nodesize = btrfs_super_nodesize(disk_super); |
1695 | leafsize = btrfs_super_leafsize(disk_super); | |
1696 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 1697 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
1698 | tree_root->nodesize = nodesize; |
1699 | tree_root->leafsize = leafsize; | |
1700 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 1701 | tree_root->stripesize = stripesize; |
a061fc8d CM |
1702 | |
1703 | sb->s_blocksize = sectorsize; | |
1704 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 1705 | |
39279cc3 CM |
1706 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
1707 | sizeof(disk_super->magic))) { | |
1708 | printk("btrfs: valid FS not found on %s\n", sb->s_id); | |
1709 | goto fail_sb_buffer; | |
1710 | } | |
19c00ddc | 1711 | |
925baedd | 1712 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1713 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 1714 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f CM |
1715 | if (ret) { |
1716 | printk("btrfs: failed to read the system array on %s\n", | |
1717 | sb->s_id); | |
1718 | goto fail_sys_array; | |
1719 | } | |
0b86a832 CM |
1720 | |
1721 | blocksize = btrfs_level_size(tree_root, | |
1722 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 1723 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
1724 | |
1725 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1726 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
1727 | ||
1728 | chunk_root->node = read_tree_block(chunk_root, | |
1729 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 1730 | blocksize, generation); |
0b86a832 CM |
1731 | BUG_ON(!chunk_root->node); |
1732 | ||
e17cade2 CM |
1733 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
1734 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), | |
1735 | BTRFS_UUID_SIZE); | |
1736 | ||
925baedd | 1737 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1738 | ret = btrfs_read_chunk_tree(chunk_root); |
925baedd | 1739 | mutex_unlock(&fs_info->chunk_mutex); |
2b82032c YZ |
1740 | if (ret) { |
1741 | printk("btrfs: failed to read chunk tree on %s\n", sb->s_id); | |
1742 | goto fail_chunk_root; | |
1743 | } | |
0b86a832 | 1744 | |
dfe25020 CM |
1745 | btrfs_close_extra_devices(fs_devices); |
1746 | ||
db94535d CM |
1747 | blocksize = btrfs_level_size(tree_root, |
1748 | btrfs_super_root_level(disk_super)); | |
84234f3a | 1749 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 1750 | |
e20d96d6 | 1751 | tree_root->node = read_tree_block(tree_root, |
db94535d | 1752 | btrfs_super_root(disk_super), |
84234f3a | 1753 | blocksize, generation); |
39279cc3 | 1754 | if (!tree_root->node) |
2b82032c | 1755 | goto fail_chunk_root; |
3768f368 | 1756 | |
db94535d CM |
1757 | |
1758 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 1759 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 1760 | if (ret) |
39279cc3 | 1761 | goto fail_tree_root; |
0b86a832 CM |
1762 | extent_root->track_dirty = 1; |
1763 | ||
1764 | ret = find_and_setup_root(tree_root, fs_info, | |
1765 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
1766 | dev_root->track_dirty = 1; | |
1767 | ||
1768 | if (ret) | |
1769 | goto fail_extent_root; | |
3768f368 | 1770 | |
d20f7043 CM |
1771 | ret = find_and_setup_root(tree_root, fs_info, |
1772 | BTRFS_CSUM_TREE_OBJECTID, csum_root); | |
1773 | if (ret) | |
1774 | goto fail_extent_root; | |
1775 | ||
1776 | csum_root->track_dirty = 1; | |
1777 | ||
9078a3e1 CM |
1778 | btrfs_read_block_groups(extent_root); |
1779 | ||
c146afad YZ |
1780 | fs_info->generation = generation + 1; |
1781 | fs_info->last_trans_committed = generation; | |
d18a2c44 CM |
1782 | fs_info->data_alloc_profile = (u64)-1; |
1783 | fs_info->metadata_alloc_profile = (u64)-1; | |
1784 | fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; | |
a74a4b97 CM |
1785 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
1786 | "btrfs-cleaner"); | |
1787 | if (!fs_info->cleaner_kthread) | |
d20f7043 | 1788 | goto fail_csum_root; |
a74a4b97 CM |
1789 | |
1790 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
1791 | tree_root, | |
1792 | "btrfs-transaction"); | |
1793 | if (!fs_info->transaction_kthread) | |
3f157a2f | 1794 | goto fail_cleaner; |
a74a4b97 | 1795 | |
e02119d5 | 1796 | if (btrfs_super_log_root(disk_super) != 0) { |
e02119d5 CM |
1797 | u64 bytenr = btrfs_super_log_root(disk_super); |
1798 | ||
7c2ca468 CM |
1799 | if (fs_devices->rw_devices == 0) { |
1800 | printk("Btrfs log replay required on RO media\n"); | |
1801 | err = -EIO; | |
1802 | goto fail_trans_kthread; | |
1803 | } | |
e02119d5 CM |
1804 | blocksize = |
1805 | btrfs_level_size(tree_root, | |
1806 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 1807 | |
e02119d5 CM |
1808 | log_tree_root = kzalloc(sizeof(struct btrfs_root), |
1809 | GFP_NOFS); | |
1810 | ||
1811 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1812 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1813 | ||
1814 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
1815 | blocksize, |
1816 | generation + 1); | |
e02119d5 CM |
1817 | ret = btrfs_recover_log_trees(log_tree_root); |
1818 | BUG_ON(ret); | |
e556ce2c YZ |
1819 | |
1820 | if (sb->s_flags & MS_RDONLY) { | |
1821 | ret = btrfs_commit_super(tree_root); | |
1822 | BUG_ON(ret); | |
1823 | } | |
e02119d5 | 1824 | } |
1a40e23b | 1825 | |
7c2ca468 CM |
1826 | if (!(sb->s_flags & MS_RDONLY)) { |
1827 | ret = btrfs_cleanup_reloc_trees(tree_root); | |
1828 | BUG_ON(ret); | |
1829 | } | |
1a40e23b | 1830 | |
3de4586c CM |
1831 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
1832 | location.type = BTRFS_ROOT_ITEM_KEY; | |
1833 | location.offset = (u64)-1; | |
1834 | ||
3de4586c CM |
1835 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
1836 | if (!fs_info->fs_root) | |
7c2ca468 | 1837 | goto fail_trans_kthread; |
0f7d52f4 | 1838 | return tree_root; |
39279cc3 | 1839 | |
7c2ca468 CM |
1840 | fail_trans_kthread: |
1841 | kthread_stop(fs_info->transaction_kthread); | |
3f157a2f | 1842 | fail_cleaner: |
a74a4b97 | 1843 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
1844 | |
1845 | /* | |
1846 | * make sure we're done with the btree inode before we stop our | |
1847 | * kthreads | |
1848 | */ | |
1849 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
1850 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
1851 | ||
d20f7043 CM |
1852 | fail_csum_root: |
1853 | free_extent_buffer(csum_root->node); | |
0b86a832 CM |
1854 | fail_extent_root: |
1855 | free_extent_buffer(extent_root->node); | |
39279cc3 | 1856 | fail_tree_root: |
5f39d397 | 1857 | free_extent_buffer(tree_root->node); |
2b82032c YZ |
1858 | fail_chunk_root: |
1859 | free_extent_buffer(chunk_root->node); | |
84eed90f | 1860 | fail_sys_array: |
7c2ca468 | 1861 | free_extent_buffer(dev_root->node); |
39279cc3 | 1862 | fail_sb_buffer: |
247e743c | 1863 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 1864 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
1865 | btrfs_stop_workers(&fs_info->workers); |
1866 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 1867 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
e6dcd2dc | 1868 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 1869 | btrfs_stop_workers(&fs_info->submit_workers); |
4543df7e | 1870 | fail_iput: |
7c2ca468 | 1871 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
4543df7e | 1872 | iput(fs_info->btree_inode); |
39279cc3 | 1873 | fail: |
dfe25020 | 1874 | btrfs_close_devices(fs_info->fs_devices); |
84eed90f CM |
1875 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
1876 | ||
39279cc3 CM |
1877 | kfree(extent_root); |
1878 | kfree(tree_root); | |
2d2ae547 | 1879 | bdi_destroy(&fs_info->bdi); |
39279cc3 | 1880 | kfree(fs_info); |
83afeac4 JM |
1881 | kfree(chunk_root); |
1882 | kfree(dev_root); | |
d20f7043 | 1883 | kfree(csum_root); |
39279cc3 | 1884 | return ERR_PTR(err); |
eb60ceac CM |
1885 | } |
1886 | ||
f2984462 CM |
1887 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
1888 | { | |
1889 | char b[BDEVNAME_SIZE]; | |
1890 | ||
1891 | if (uptodate) { | |
1892 | set_buffer_uptodate(bh); | |
1893 | } else { | |
1894 | if (!buffer_eopnotsupp(bh) && printk_ratelimit()) { | |
1895 | printk(KERN_WARNING "lost page write due to " | |
1896 | "I/O error on %s\n", | |
1897 | bdevname(bh->b_bdev, b)); | |
1898 | } | |
1259ab75 CM |
1899 | /* note, we dont' set_buffer_write_io_error because we have |
1900 | * our own ways of dealing with the IO errors | |
1901 | */ | |
f2984462 CM |
1902 | clear_buffer_uptodate(bh); |
1903 | } | |
1904 | unlock_buffer(bh); | |
1905 | put_bh(bh); | |
1906 | } | |
1907 | ||
b2950863 | 1908 | static int write_all_supers(struct btrfs_root *root) |
f2984462 CM |
1909 | { |
1910 | struct list_head *cur; | |
1911 | struct list_head *head = &root->fs_info->fs_devices->devices; | |
1912 | struct btrfs_device *dev; | |
a061fc8d | 1913 | struct btrfs_super_block *sb; |
f2984462 CM |
1914 | struct btrfs_dev_item *dev_item; |
1915 | struct buffer_head *bh; | |
1916 | int ret; | |
1917 | int do_barriers; | |
a236aed1 CM |
1918 | int max_errors; |
1919 | int total_errors = 0; | |
a061fc8d CM |
1920 | u32 crc; |
1921 | u64 flags; | |
f2984462 | 1922 | |
a236aed1 | 1923 | max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1; |
f2984462 CM |
1924 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
1925 | ||
a061fc8d CM |
1926 | sb = &root->fs_info->super_for_commit; |
1927 | dev_item = &sb->dev_item; | |
f2984462 CM |
1928 | list_for_each(cur, head) { |
1929 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1930 | if (!dev->bdev) { |
1931 | total_errors++; | |
1932 | continue; | |
1933 | } | |
2b82032c | 1934 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
1935 | continue; |
1936 | ||
2b82032c | 1937 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
1938 | btrfs_set_stack_device_type(dev_item, dev->type); |
1939 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
1940 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
1941 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
1942 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
1943 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
1944 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
1945 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 1946 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a061fc8d CM |
1947 | flags = btrfs_super_flags(sb); |
1948 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
1949 | ||
1950 | ||
1951 | crc = ~(u32)0; | |
1952 | crc = btrfs_csum_data(root, (char *)sb + BTRFS_CSUM_SIZE, crc, | |
1953 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
1954 | btrfs_csum_final(crc, sb->csum); | |
1955 | ||
1956 | bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET / 4096, | |
f2984462 CM |
1957 | BTRFS_SUPER_INFO_SIZE); |
1958 | ||
a061fc8d | 1959 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
f2984462 CM |
1960 | dev->pending_io = bh; |
1961 | ||
1962 | get_bh(bh); | |
1963 | set_buffer_uptodate(bh); | |
1964 | lock_buffer(bh); | |
1965 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
1966 | ||
1967 | if (do_barriers && dev->barriers) { | |
1968 | ret = submit_bh(WRITE_BARRIER, bh); | |
1969 | if (ret == -EOPNOTSUPP) { | |
1970 | printk("btrfs: disabling barriers on dev %s\n", | |
1971 | dev->name); | |
1972 | set_buffer_uptodate(bh); | |
1973 | dev->barriers = 0; | |
1974 | get_bh(bh); | |
1975 | lock_buffer(bh); | |
1976 | ret = submit_bh(WRITE, bh); | |
1977 | } | |
1978 | } else { | |
1979 | ret = submit_bh(WRITE, bh); | |
1980 | } | |
a236aed1 CM |
1981 | if (ret) |
1982 | total_errors++; | |
f2984462 | 1983 | } |
a236aed1 CM |
1984 | if (total_errors > max_errors) { |
1985 | printk("btrfs: %d errors while writing supers\n", total_errors); | |
1986 | BUG(); | |
1987 | } | |
1988 | total_errors = 0; | |
f2984462 CM |
1989 | |
1990 | list_for_each(cur, head) { | |
1991 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1992 | if (!dev->bdev) |
1993 | continue; | |
2b82032c | 1994 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
1995 | continue; |
1996 | ||
f2984462 CM |
1997 | BUG_ON(!dev->pending_io); |
1998 | bh = dev->pending_io; | |
1999 | wait_on_buffer(bh); | |
2000 | if (!buffer_uptodate(dev->pending_io)) { | |
2001 | if (do_barriers && dev->barriers) { | |
2002 | printk("btrfs: disabling barriers on dev %s\n", | |
2003 | dev->name); | |
2004 | set_buffer_uptodate(bh); | |
2005 | get_bh(bh); | |
2006 | lock_buffer(bh); | |
2007 | dev->barriers = 0; | |
2008 | ret = submit_bh(WRITE, bh); | |
2009 | BUG_ON(ret); | |
2010 | wait_on_buffer(bh); | |
1259ab75 CM |
2011 | if (!buffer_uptodate(bh)) |
2012 | total_errors++; | |
f2984462 | 2013 | } else { |
a236aed1 | 2014 | total_errors++; |
f2984462 CM |
2015 | } |
2016 | ||
2017 | } | |
2018 | dev->pending_io = NULL; | |
2019 | brelse(bh); | |
2020 | } | |
a236aed1 CM |
2021 | if (total_errors > max_errors) { |
2022 | printk("btrfs: %d errors while writing supers\n", total_errors); | |
2023 | BUG(); | |
2024 | } | |
f2984462 CM |
2025 | return 0; |
2026 | } | |
2027 | ||
e089f05c | 2028 | int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root |
79154b1b | 2029 | *root) |
eb60ceac | 2030 | { |
e66f709b | 2031 | int ret; |
5f39d397 | 2032 | |
f2984462 | 2033 | ret = write_all_supers(root); |
5f39d397 | 2034 | return ret; |
cfaa7295 CM |
2035 | } |
2036 | ||
5eda7b5e | 2037 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f CM |
2038 | { |
2039 | radix_tree_delete(&fs_info->fs_roots_radix, | |
2040 | (unsigned long)root->root_key.objectid); | |
3394e160 CM |
2041 | if (root->anon_super.s_dev) { |
2042 | down_write(&root->anon_super.s_umount); | |
2043 | kill_anon_super(&root->anon_super); | |
2044 | } | |
ea9e8b11 | 2045 | #if 0 |
b99aa6cb CM |
2046 | if (root->in_sysfs) |
2047 | btrfs_sysfs_del_root(root); | |
ea9e8b11 | 2048 | #endif |
2619ba1f | 2049 | if (root->node) |
5f39d397 | 2050 | free_extent_buffer(root->node); |
2619ba1f | 2051 | if (root->commit_root) |
5f39d397 | 2052 | free_extent_buffer(root->commit_root); |
58176a96 JB |
2053 | if (root->name) |
2054 | kfree(root->name); | |
2619ba1f CM |
2055 | kfree(root); |
2056 | return 0; | |
2057 | } | |
2058 | ||
35b7e476 | 2059 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
2060 | { |
2061 | int ret; | |
2062 | struct btrfs_root *gang[8]; | |
2063 | int i; | |
2064 | ||
2065 | while(1) { | |
2066 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2067 | (void **)gang, 0, | |
2068 | ARRAY_SIZE(gang)); | |
2069 | if (!ret) | |
2070 | break; | |
2619ba1f | 2071 | for (i = 0; i < ret; i++) |
5eda7b5e | 2072 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
2073 | } |
2074 | return 0; | |
2075 | } | |
b4100d64 | 2076 | |
c146afad | 2077 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 2078 | { |
c146afad YZ |
2079 | u64 root_objectid = 0; |
2080 | struct btrfs_root *gang[8]; | |
2081 | int i; | |
3768f368 | 2082 | int ret; |
e089f05c | 2083 | |
c146afad YZ |
2084 | while (1) { |
2085 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2086 | (void **)gang, root_objectid, | |
2087 | ARRAY_SIZE(gang)); | |
2088 | if (!ret) | |
2089 | break; | |
2090 | for (i = 0; i < ret; i++) { | |
2091 | root_objectid = gang[i]->root_key.objectid; | |
2092 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
2093 | root_objectid, gang[i]); | |
2094 | BUG_ON(ret); | |
2095 | btrfs_orphan_cleanup(gang[i]); | |
2096 | } | |
2097 | root_objectid++; | |
2098 | } | |
2099 | return 0; | |
2100 | } | |
a2135011 | 2101 | |
c146afad YZ |
2102 | int btrfs_commit_super(struct btrfs_root *root) |
2103 | { | |
2104 | struct btrfs_trans_handle *trans; | |
2105 | int ret; | |
a74a4b97 | 2106 | |
c146afad | 2107 | mutex_lock(&root->fs_info->cleaner_mutex); |
a74a4b97 | 2108 | btrfs_clean_old_snapshots(root); |
c146afad | 2109 | mutex_unlock(&root->fs_info->cleaner_mutex); |
79154b1b | 2110 | trans = btrfs_start_transaction(root, 1); |
54aa1f4d | 2111 | ret = btrfs_commit_transaction(trans, root); |
c146afad YZ |
2112 | BUG_ON(ret); |
2113 | /* run commit again to drop the original snapshot */ | |
79154b1b CM |
2114 | trans = btrfs_start_transaction(root, 1); |
2115 | btrfs_commit_transaction(trans, root); | |
2116 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 2117 | BUG_ON(ret); |
d6bfde87 | 2118 | |
c146afad YZ |
2119 | ret = write_ctree_super(NULL, root); |
2120 | return ret; | |
2121 | } | |
2122 | ||
2123 | int close_ctree(struct btrfs_root *root) | |
2124 | { | |
2125 | struct btrfs_fs_info *fs_info = root->fs_info; | |
2126 | int ret; | |
2127 | ||
2128 | fs_info->closing = 1; | |
2129 | smp_mb(); | |
2130 | ||
2131 | kthread_stop(root->fs_info->transaction_kthread); | |
2132 | kthread_stop(root->fs_info->cleaner_kthread); | |
2133 | ||
2134 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { | |
2135 | ret = btrfs_commit_super(root); | |
2136 | if (ret) { | |
2137 | printk("btrfs: commit super returns %d\n", ret); | |
2138 | } | |
2139 | } | |
0f7d52f4 | 2140 | |
b0c68f8b CM |
2141 | if (fs_info->delalloc_bytes) { |
2142 | printk("btrfs: at unmount delalloc count %Lu\n", | |
2143 | fs_info->delalloc_bytes); | |
2144 | } | |
31153d81 YZ |
2145 | if (fs_info->total_ref_cache_size) { |
2146 | printk("btrfs: at umount reference cache size %Lu\n", | |
2147 | fs_info->total_ref_cache_size); | |
2148 | } | |
bcc63abb | 2149 | |
0f7d52f4 | 2150 | if (fs_info->extent_root->node) |
5f39d397 | 2151 | free_extent_buffer(fs_info->extent_root->node); |
f510cfec | 2152 | |
0f7d52f4 | 2153 | if (fs_info->tree_root->node) |
5f39d397 | 2154 | free_extent_buffer(fs_info->tree_root->node); |
f510cfec | 2155 | |
0b86a832 CM |
2156 | if (root->fs_info->chunk_root->node); |
2157 | free_extent_buffer(root->fs_info->chunk_root->node); | |
2158 | ||
2159 | if (root->fs_info->dev_root->node); | |
2160 | free_extent_buffer(root->fs_info->dev_root->node); | |
2161 | ||
d20f7043 CM |
2162 | if (root->fs_info->csum_root->node); |
2163 | free_extent_buffer(root->fs_info->csum_root->node); | |
2164 | ||
9078a3e1 | 2165 | btrfs_free_block_groups(root->fs_info); |
d10c5f31 | 2166 | |
c146afad | 2167 | del_fs_roots(fs_info); |
d10c5f31 | 2168 | |
c146afad | 2169 | iput(fs_info->btree_inode); |
9ad6b7bc | 2170 | |
247e743c | 2171 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2172 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2173 | btrfs_stop_workers(&fs_info->workers); |
2174 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 2175 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
e6dcd2dc | 2176 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 2177 | btrfs_stop_workers(&fs_info->submit_workers); |
d6bfde87 | 2178 | |
19c00ddc CM |
2179 | #if 0 |
2180 | while(!list_empty(&fs_info->hashers)) { | |
2181 | struct btrfs_hasher *hasher; | |
2182 | hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher, | |
2183 | hashers); | |
2184 | list_del(&hasher->hashers); | |
2185 | crypto_free_hash(&fs_info->hash_tfm); | |
2186 | kfree(hasher); | |
2187 | } | |
2188 | #endif | |
dfe25020 | 2189 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 2190 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 2191 | |
04160088 | 2192 | bdi_destroy(&fs_info->bdi); |
0b86a832 | 2193 | |
0f7d52f4 | 2194 | kfree(fs_info->extent_root); |
0f7d52f4 | 2195 | kfree(fs_info->tree_root); |
0b86a832 CM |
2196 | kfree(fs_info->chunk_root); |
2197 | kfree(fs_info->dev_root); | |
d20f7043 | 2198 | kfree(fs_info->csum_root); |
eb60ceac CM |
2199 | return 0; |
2200 | } | |
2201 | ||
1259ab75 | 2202 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 2203 | { |
1259ab75 | 2204 | int ret; |
810191ff | 2205 | struct inode *btree_inode = buf->first_page->mapping->host; |
1259ab75 CM |
2206 | |
2207 | ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); | |
2208 | if (!ret) | |
2209 | return ret; | |
2210 | ||
2211 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
2212 | parent_transid); | |
2213 | return !ret; | |
5f39d397 CM |
2214 | } |
2215 | ||
2216 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 2217 | { |
810191ff | 2218 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 2219 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
2220 | buf); |
2221 | } | |
6702ed49 | 2222 | |
5f39d397 CM |
2223 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
2224 | { | |
810191ff | 2225 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
5f39d397 CM |
2226 | u64 transid = btrfs_header_generation(buf); |
2227 | struct inode *btree_inode = root->fs_info->btree_inode; | |
6702ed49 | 2228 | |
925baedd | 2229 | WARN_ON(!btrfs_tree_locked(buf)); |
ccd467d6 CM |
2230 | if (transid != root->fs_info->generation) { |
2231 | printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n", | |
db94535d | 2232 | (unsigned long long)buf->start, |
ccd467d6 CM |
2233 | transid, root->fs_info->generation); |
2234 | WARN_ON(1); | |
2235 | } | |
d1310b2e | 2236 | set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf); |
eb60ceac CM |
2237 | } |
2238 | ||
d3c2fdcf | 2239 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
35b7e476 | 2240 | { |
188de649 CM |
2241 | /* |
2242 | * looks as though older kernels can get into trouble with | |
2243 | * this code, they end up stuck in balance_dirty_pages forever | |
2244 | */ | |
d6bfde87 CM |
2245 | struct extent_io_tree *tree; |
2246 | u64 num_dirty; | |
2247 | u64 start = 0; | |
771ed689 | 2248 | unsigned long thresh = 32 * 1024 * 1024; |
d6bfde87 CM |
2249 | tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
2250 | ||
b64a2851 | 2251 | if (current_is_pdflush() || current->flags & PF_MEMALLOC) |
d6bfde87 CM |
2252 | return; |
2253 | ||
2254 | num_dirty = count_range_bits(tree, &start, (u64)-1, | |
2255 | thresh, EXTENT_DIRTY); | |
2256 | if (num_dirty > thresh) { | |
2257 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 2258 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 2259 | } |
188de649 | 2260 | return; |
35b7e476 | 2261 | } |
6b80053d | 2262 | |
ca7a79ad | 2263 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 2264 | { |
810191ff | 2265 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
ce9adaa5 | 2266 | int ret; |
ca7a79ad | 2267 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 CM |
2268 | if (ret == 0) { |
2269 | buf->flags |= EXTENT_UPTODATE; | |
2270 | } | |
2271 | return ret; | |
6b80053d | 2272 | } |
0da5468f | 2273 | |
4bef0848 CM |
2274 | int btree_lock_page_hook(struct page *page) |
2275 | { | |
2276 | struct inode *inode = page->mapping->host; | |
2277 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
2278 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
2279 | struct extent_buffer *eb; | |
2280 | unsigned long len; | |
2281 | u64 bytenr = page_offset(page); | |
2282 | ||
2283 | if (page->private == EXTENT_PAGE_PRIVATE) | |
2284 | goto out; | |
2285 | ||
2286 | len = page->private >> 2; | |
2287 | eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS); | |
2288 | if (!eb) | |
2289 | goto out; | |
2290 | ||
2291 | btrfs_tree_lock(eb); | |
2292 | spin_lock(&root->fs_info->hash_lock); | |
2293 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); | |
2294 | spin_unlock(&root->fs_info->hash_lock); | |
2295 | btrfs_tree_unlock(eb); | |
2296 | free_extent_buffer(eb); | |
2297 | out: | |
2298 | lock_page(page); | |
2299 | return 0; | |
2300 | } | |
2301 | ||
d1310b2e | 2302 | static struct extent_io_ops btree_extent_io_ops = { |
4bef0848 | 2303 | .write_cache_pages_lock_hook = btree_lock_page_hook, |
ce9adaa5 | 2304 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
0b86a832 | 2305 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
2306 | /* note we're sharing with inode.c for the merge bio hook */ |
2307 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 2308 | }; |