]>
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> |
5f39d397 | 25 | #include <linux/buffer_head.h> // for block_sync_page |
ce9adaa5 | 26 | #include <linux/workqueue.h> |
a5eb62e3 | 27 | #include "crc32c.h" |
eb60ceac CM |
28 | #include "ctree.h" |
29 | #include "disk-io.h" | |
e089f05c | 30 | #include "transaction.h" |
0f7d52f4 | 31 | #include "btrfs_inode.h" |
0b86a832 | 32 | #include "volumes.h" |
db94535d | 33 | #include "print-tree.h" |
8b712842 | 34 | #include "async-thread.h" |
925baedd | 35 | #include "locking.h" |
eb60ceac | 36 | |
5f39d397 CM |
37 | #if 0 |
38 | static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) | |
7eccb903 | 39 | { |
5f39d397 CM |
40 | if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) { |
41 | printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n", | |
42 | (unsigned long long)extent_buffer_blocknr(buf), | |
43 | (unsigned long long)btrfs_header_blocknr(buf)); | |
39279cc3 | 44 | return 1; |
d98237b3 | 45 | } |
9a8dd150 | 46 | return 0; |
eb60ceac | 47 | } |
5f39d397 | 48 | #endif |
eb60ceac | 49 | |
d1310b2e | 50 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 51 | static void end_workqueue_fn(struct btrfs_work *work); |
ce9adaa5 CM |
52 | |
53 | struct end_io_wq { | |
54 | struct bio *bio; | |
55 | bio_end_io_t *end_io; | |
56 | void *private; | |
57 | struct btrfs_fs_info *info; | |
58 | int error; | |
22c59948 | 59 | int metadata; |
ce9adaa5 | 60 | struct list_head list; |
8b712842 | 61 | struct btrfs_work work; |
ce9adaa5 | 62 | }; |
0da5468f | 63 | |
44b8bd7e CM |
64 | struct async_submit_bio { |
65 | struct inode *inode; | |
66 | struct bio *bio; | |
67 | struct list_head list; | |
68 | extent_submit_bio_hook_t *submit_bio_hook; | |
69 | int rw; | |
70 | int mirror_num; | |
8b712842 | 71 | struct btrfs_work work; |
44b8bd7e CM |
72 | }; |
73 | ||
5f39d397 | 74 | struct extent_map *btree_get_extent(struct inode *inode, struct page *page, |
d1310b2e | 75 | size_t page_offset, u64 start, u64 len, |
5f39d397 | 76 | int create) |
7eccb903 | 77 | { |
5f39d397 CM |
78 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
79 | struct extent_map *em; | |
80 | int ret; | |
81 | ||
d1310b2e CM |
82 | spin_lock(&em_tree->lock); |
83 | em = lookup_extent_mapping(em_tree, start, len); | |
a061fc8d CM |
84 | if (em) { |
85 | em->bdev = | |
86 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
87 | spin_unlock(&em_tree->lock); | |
5f39d397 | 88 | goto out; |
a061fc8d CM |
89 | } |
90 | spin_unlock(&em_tree->lock); | |
7b13b7b1 | 91 | |
5f39d397 CM |
92 | em = alloc_extent_map(GFP_NOFS); |
93 | if (!em) { | |
94 | em = ERR_PTR(-ENOMEM); | |
95 | goto out; | |
96 | } | |
97 | em->start = 0; | |
0afbaf8c | 98 | em->len = (u64)-1; |
5f39d397 | 99 | em->block_start = 0; |
a061fc8d | 100 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e CM |
101 | |
102 | spin_lock(&em_tree->lock); | |
5f39d397 CM |
103 | ret = add_extent_mapping(em_tree, em); |
104 | if (ret == -EEXIST) { | |
0afbaf8c CM |
105 | u64 failed_start = em->start; |
106 | u64 failed_len = em->len; | |
107 | ||
108 | printk("failed to insert %Lu %Lu -> %Lu into tree\n", | |
109 | em->start, em->len, em->block_start); | |
5f39d397 | 110 | free_extent_map(em); |
7b13b7b1 | 111 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c CM |
112 | if (em) { |
113 | printk("after failing, found %Lu %Lu %Lu\n", | |
114 | em->start, em->len, em->block_start); | |
7b13b7b1 | 115 | ret = 0; |
0afbaf8c CM |
116 | } else { |
117 | em = lookup_extent_mapping(em_tree, failed_start, | |
118 | failed_len); | |
119 | if (em) { | |
120 | printk("double failure lookup gives us " | |
121 | "%Lu %Lu -> %Lu\n", em->start, | |
122 | em->len, em->block_start); | |
123 | free_extent_map(em); | |
124 | } | |
7b13b7b1 | 125 | ret = -EIO; |
0afbaf8c | 126 | } |
5f39d397 | 127 | } else if (ret) { |
7b13b7b1 CM |
128 | free_extent_map(em); |
129 | em = NULL; | |
5f39d397 | 130 | } |
7b13b7b1 CM |
131 | spin_unlock(&em_tree->lock); |
132 | ||
133 | if (ret) | |
134 | em = ERR_PTR(ret); | |
5f39d397 CM |
135 | out: |
136 | return em; | |
7eccb903 CM |
137 | } |
138 | ||
19c00ddc CM |
139 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
140 | { | |
a5eb62e3 | 141 | return btrfs_crc32c(seed, data, len); |
19c00ddc CM |
142 | } |
143 | ||
144 | void btrfs_csum_final(u32 crc, char *result) | |
145 | { | |
146 | *(__le32 *)result = ~cpu_to_le32(crc); | |
147 | } | |
148 | ||
149 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, | |
150 | int verify) | |
151 | { | |
152 | char result[BTRFS_CRC32_SIZE]; | |
153 | unsigned long len; | |
154 | unsigned long cur_len; | |
155 | unsigned long offset = BTRFS_CSUM_SIZE; | |
156 | char *map_token = NULL; | |
157 | char *kaddr; | |
158 | unsigned long map_start; | |
159 | unsigned long map_len; | |
160 | int err; | |
161 | u32 crc = ~(u32)0; | |
162 | ||
163 | len = buf->len - offset; | |
164 | while(len > 0) { | |
165 | err = map_private_extent_buffer(buf, offset, 32, | |
166 | &map_token, &kaddr, | |
167 | &map_start, &map_len, KM_USER0); | |
168 | if (err) { | |
169 | printk("failed to map extent buffer! %lu\n", | |
170 | offset); | |
171 | return 1; | |
172 | } | |
173 | cur_len = min(len, map_len - (offset - map_start)); | |
174 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
175 | crc, cur_len); | |
176 | len -= cur_len; | |
177 | offset += cur_len; | |
178 | unmap_extent_buffer(buf, map_token, KM_USER0); | |
179 | } | |
180 | btrfs_csum_final(crc, result); | |
181 | ||
182 | if (verify) { | |
e4204ded CM |
183 | int from_this_trans = 0; |
184 | ||
185 | if (root->fs_info->running_transaction && | |
186 | btrfs_header_generation(buf) == | |
187 | root->fs_info->running_transaction->transid) | |
188 | from_this_trans = 1; | |
189 | ||
190 | /* FIXME, this is not good */ | |
63b10fc4 | 191 | if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) { |
e4204ded CM |
192 | u32 val; |
193 | u32 found = 0; | |
194 | memcpy(&found, result, BTRFS_CRC32_SIZE); | |
195 | ||
196 | read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE); | |
197 | printk("btrfs: %s checksum verify failed on %llu " | |
63b10fc4 CM |
198 | "wanted %X found %X from_this_trans %d " |
199 | "level %d\n", | |
19c00ddc | 200 | root->fs_info->sb->s_id, |
63b10fc4 CM |
201 | buf->start, val, found, from_this_trans, |
202 | btrfs_header_level(buf)); | |
19c00ddc CM |
203 | return 1; |
204 | } | |
205 | } else { | |
206 | write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE); | |
207 | } | |
208 | return 0; | |
209 | } | |
210 | ||
1259ab75 CM |
211 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
212 | struct extent_buffer *eb, u64 parent_transid) | |
213 | { | |
214 | int ret; | |
215 | ||
216 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
217 | return 0; | |
218 | ||
219 | lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); | |
220 | if (extent_buffer_uptodate(io_tree, eb) && | |
221 | btrfs_header_generation(eb) == parent_transid) { | |
222 | ret = 0; | |
223 | goto out; | |
224 | } | |
225 | printk("parent transid verify failed on %llu wanted %llu found %llu\n", | |
226 | (unsigned long long)eb->start, | |
227 | (unsigned long long)parent_transid, | |
228 | (unsigned long long)btrfs_header_generation(eb)); | |
229 | ret = 1; | |
230 | out: | |
231 | clear_extent_buffer_uptodate(io_tree, eb); | |
232 | unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, | |
233 | GFP_NOFS); | |
234 | return ret; | |
235 | ||
236 | } | |
237 | ||
f188591e CM |
238 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
239 | struct extent_buffer *eb, | |
ca7a79ad | 240 | u64 start, u64 parent_transid) |
f188591e CM |
241 | { |
242 | struct extent_io_tree *io_tree; | |
243 | int ret; | |
244 | int num_copies = 0; | |
245 | int mirror_num = 0; | |
246 | ||
247 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
248 | while (1) { | |
249 | ret = read_extent_buffer_pages(io_tree, eb, start, 1, | |
250 | btree_get_extent, mirror_num); | |
1259ab75 CM |
251 | if (!ret && |
252 | !verify_parent_transid(io_tree, eb, parent_transid)) | |
f188591e | 253 | return ret; |
4235298e | 254 | |
f188591e CM |
255 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
256 | eb->start, eb->len); | |
4235298e | 257 | if (num_copies == 1) |
f188591e | 258 | return ret; |
4235298e | 259 | |
f188591e | 260 | mirror_num++; |
4235298e | 261 | if (mirror_num > num_copies) |
f188591e | 262 | return ret; |
f188591e | 263 | } |
f188591e CM |
264 | return -EIO; |
265 | } | |
19c00ddc CM |
266 | |
267 | int csum_dirty_buffer(struct btrfs_root *root, struct page *page) | |
268 | { | |
d1310b2e | 269 | struct extent_io_tree *tree; |
35ebb934 | 270 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc CM |
271 | u64 found_start; |
272 | int found_level; | |
273 | unsigned long len; | |
274 | struct extent_buffer *eb; | |
f188591e CM |
275 | int ret; |
276 | ||
d1310b2e | 277 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc CM |
278 | |
279 | if (page->private == EXTENT_PAGE_PRIVATE) | |
280 | goto out; | |
281 | if (!page->private) | |
282 | goto out; | |
283 | len = page->private >> 2; | |
284 | if (len == 0) { | |
285 | WARN_ON(1); | |
286 | } | |
287 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); | |
ca7a79ad CM |
288 | ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, |
289 | btrfs_header_generation(eb)); | |
f188591e | 290 | BUG_ON(ret); |
e18e4809 | 291 | btrfs_clear_buffer_defrag(eb); |
19c00ddc CM |
292 | found_start = btrfs_header_bytenr(eb); |
293 | if (found_start != start) { | |
294 | printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n", | |
295 | start, found_start, len); | |
55c69072 CM |
296 | WARN_ON(1); |
297 | goto err; | |
298 | } | |
299 | if (eb->first_page != page) { | |
300 | printk("bad first page %lu %lu\n", eb->first_page->index, | |
301 | page->index); | |
302 | WARN_ON(1); | |
303 | goto err; | |
304 | } | |
305 | if (!PageUptodate(page)) { | |
306 | printk("csum not up to date page %lu\n", page->index); | |
307 | WARN_ON(1); | |
308 | goto err; | |
19c00ddc CM |
309 | } |
310 | found_level = btrfs_header_level(eb); | |
63b10fc4 CM |
311 | spin_lock(&root->fs_info->hash_lock); |
312 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); | |
313 | spin_unlock(&root->fs_info->hash_lock); | |
19c00ddc | 314 | csum_tree_block(root, eb, 0); |
55c69072 | 315 | err: |
19c00ddc CM |
316 | free_extent_buffer(eb); |
317 | out: | |
318 | return 0; | |
319 | } | |
320 | ||
0da5468f | 321 | static int btree_writepage_io_hook(struct page *page, u64 start, u64 end) |
d98237b3 | 322 | { |
19c00ddc | 323 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
19c00ddc CM |
324 | |
325 | csum_dirty_buffer(root, page); | |
0da5468f CM |
326 | return 0; |
327 | } | |
328 | ||
ce9adaa5 CM |
329 | int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
330 | struct extent_state *state) | |
331 | { | |
332 | struct extent_io_tree *tree; | |
333 | u64 found_start; | |
334 | int found_level; | |
335 | unsigned long len; | |
336 | struct extent_buffer *eb; | |
337 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 338 | int ret = 0; |
ce9adaa5 CM |
339 | |
340 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
341 | if (page->private == EXTENT_PAGE_PRIVATE) | |
342 | goto out; | |
343 | if (!page->private) | |
344 | goto out; | |
345 | len = page->private >> 2; | |
346 | if (len == 0) { | |
347 | WARN_ON(1); | |
348 | } | |
349 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); | |
f188591e | 350 | |
ce9adaa5 CM |
351 | btrfs_clear_buffer_defrag(eb); |
352 | found_start = btrfs_header_bytenr(eb); | |
353 | if (found_start != start) { | |
f188591e | 354 | ret = -EIO; |
ce9adaa5 CM |
355 | goto err; |
356 | } | |
357 | if (eb->first_page != page) { | |
358 | printk("bad first page %lu %lu\n", eb->first_page->index, | |
359 | page->index); | |
360 | WARN_ON(1); | |
f188591e | 361 | ret = -EIO; |
ce9adaa5 CM |
362 | goto err; |
363 | } | |
1259ab75 CM |
364 | if (memcmp_extent_buffer(eb, root->fs_info->fsid, |
365 | (unsigned long)btrfs_header_fsid(eb), | |
366 | BTRFS_FSID_SIZE)) { | |
367 | printk("bad fsid on block %Lu\n", eb->start); | |
368 | ret = -EIO; | |
369 | goto err; | |
370 | } | |
ce9adaa5 CM |
371 | found_level = btrfs_header_level(eb); |
372 | ||
373 | ret = csum_tree_block(root, eb, 1); | |
f188591e CM |
374 | if (ret) |
375 | ret = -EIO; | |
ce9adaa5 CM |
376 | |
377 | end = min_t(u64, eb->len, PAGE_CACHE_SIZE); | |
378 | end = eb->start + end - 1; | |
379 | release_extent_buffer_tail_pages(eb); | |
380 | err: | |
381 | free_extent_buffer(eb); | |
382 | out: | |
f188591e | 383 | return ret; |
ce9adaa5 CM |
384 | } |
385 | ||
386 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) | |
387 | static void end_workqueue_bio(struct bio *bio, int err) | |
388 | #else | |
389 | static int end_workqueue_bio(struct bio *bio, | |
390 | unsigned int bytes_done, int err) | |
391 | #endif | |
392 | { | |
393 | struct end_io_wq *end_io_wq = bio->bi_private; | |
394 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 CM |
395 | |
396 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
397 | if (bio->bi_size) | |
398 | return 1; | |
399 | #endif | |
400 | ||
401 | fs_info = end_io_wq->info; | |
ce9adaa5 | 402 | end_io_wq->error = err; |
8b712842 CM |
403 | end_io_wq->work.func = end_workqueue_fn; |
404 | end_io_wq->work.flags = 0; | |
405 | btrfs_queue_worker(&fs_info->endio_workers, &end_io_wq->work); | |
ce9adaa5 CM |
406 | |
407 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
408 | return 0; | |
409 | #endif | |
410 | } | |
411 | ||
22c59948 CM |
412 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
413 | int metadata) | |
0b86a832 | 414 | { |
ce9adaa5 | 415 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
416 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
417 | if (!end_io_wq) | |
418 | return -ENOMEM; | |
419 | ||
420 | end_io_wq->private = bio->bi_private; | |
421 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 422 | end_io_wq->info = info; |
ce9adaa5 CM |
423 | end_io_wq->error = 0; |
424 | end_io_wq->bio = bio; | |
22c59948 | 425 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
426 | |
427 | bio->bi_private = end_io_wq; | |
428 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
429 | return 0; |
430 | } | |
431 | ||
8b712842 CM |
432 | static void run_one_async_submit(struct btrfs_work *work) |
433 | { | |
434 | struct btrfs_fs_info *fs_info; | |
435 | struct async_submit_bio *async; | |
436 | ||
437 | async = container_of(work, struct async_submit_bio, work); | |
438 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
439 | atomic_dec(&fs_info->nr_async_submits); | |
440 | async->submit_bio_hook(async->inode, async->rw, async->bio, | |
441 | async->mirror_num); | |
442 | kfree(async); | |
443 | } | |
444 | ||
44b8bd7e CM |
445 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
446 | int rw, struct bio *bio, int mirror_num, | |
447 | extent_submit_bio_hook_t *submit_bio_hook) | |
448 | { | |
449 | struct async_submit_bio *async; | |
450 | ||
451 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
452 | if (!async) | |
453 | return -ENOMEM; | |
454 | ||
455 | async->inode = inode; | |
456 | async->rw = rw; | |
457 | async->bio = bio; | |
458 | async->mirror_num = mirror_num; | |
459 | async->submit_bio_hook = submit_bio_hook; | |
8b712842 CM |
460 | async->work.func = run_one_async_submit; |
461 | async->work.flags = 0; | |
cb03c743 | 462 | atomic_inc(&fs_info->nr_async_submits); |
8b712842 | 463 | btrfs_queue_worker(&fs_info->workers, &async->work); |
44b8bd7e CM |
464 | return 0; |
465 | } | |
466 | ||
467 | static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, | |
f188591e | 468 | int mirror_num) |
22c59948 CM |
469 | { |
470 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
471 | u64 offset; | |
472 | int ret; | |
473 | ||
474 | offset = bio->bi_sector << 9; | |
475 | ||
8b712842 CM |
476 | /* |
477 | * when we're called for a write, we're already in the async | |
478 | * submission context. Just jump ingo btrfs_map_bio | |
479 | */ | |
22c59948 | 480 | if (rw & (1 << BIO_RW)) { |
8b712842 CM |
481 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, |
482 | mirror_num, 0); | |
22c59948 CM |
483 | } |
484 | ||
8b712842 CM |
485 | /* |
486 | * called for a read, do the setup so that checksum validation | |
487 | * can happen in the async kernel threads | |
488 | */ | |
22c59948 CM |
489 | ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1); |
490 | BUG_ON(ret); | |
ce9adaa5 | 491 | |
8b712842 | 492 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
493 | } |
494 | ||
44b8bd7e CM |
495 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
496 | int mirror_num) | |
497 | { | |
8b712842 CM |
498 | /* |
499 | * kthread helpers are used to submit writes so that checksumming | |
500 | * can happen in parallel across all CPUs | |
501 | */ | |
44b8bd7e CM |
502 | if (!(rw & (1 << BIO_RW))) { |
503 | return __btree_submit_bio_hook(inode, rw, bio, mirror_num); | |
504 | } | |
505 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, | |
506 | inode, rw, bio, mirror_num, | |
507 | __btree_submit_bio_hook); | |
508 | } | |
509 | ||
0da5468f CM |
510 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
511 | { | |
d1310b2e CM |
512 | struct extent_io_tree *tree; |
513 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
514 | return extent_write_full_page(tree, page, btree_get_extent, wbc); |
515 | } | |
0da5468f CM |
516 | |
517 | static int btree_writepages(struct address_space *mapping, | |
518 | struct writeback_control *wbc) | |
519 | { | |
d1310b2e CM |
520 | struct extent_io_tree *tree; |
521 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 522 | if (wbc->sync_mode == WB_SYNC_NONE) { |
793955bc CM |
523 | u64 num_dirty; |
524 | u64 start = 0; | |
525 | unsigned long thresh = 96 * 1024 * 1024; | |
448d640b CM |
526 | |
527 | if (wbc->for_kupdate) | |
528 | return 0; | |
529 | ||
ca664626 CM |
530 | if (current_is_pdflush()) { |
531 | thresh = 96 * 1024 * 1024; | |
532 | } else { | |
533 | thresh = 8 * 1024 * 1024; | |
534 | } | |
1832a6d5 CM |
535 | num_dirty = count_range_bits(tree, &start, (u64)-1, |
536 | thresh, EXTENT_DIRTY); | |
793955bc CM |
537 | if (num_dirty < thresh) { |
538 | return 0; | |
539 | } | |
540 | } | |
0da5468f CM |
541 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
542 | } | |
543 | ||
5f39d397 CM |
544 | int btree_readpage(struct file *file, struct page *page) |
545 | { | |
d1310b2e CM |
546 | struct extent_io_tree *tree; |
547 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
548 | return extent_read_full_page(tree, page, btree_get_extent); |
549 | } | |
22b0ebda | 550 | |
70dec807 | 551 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 552 | { |
d1310b2e CM |
553 | struct extent_io_tree *tree; |
554 | struct extent_map_tree *map; | |
5f39d397 | 555 | int ret; |
d98237b3 | 556 | |
3dd39914 CM |
557 | if (page_count(page) > 3) { |
558 | /* once for page->private, once for the caller, once | |
559 | * once for the page cache | |
560 | */ | |
561 | return 0; | |
562 | } | |
d1310b2e CM |
563 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
564 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
7b13b7b1 | 565 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
5f39d397 | 566 | if (ret == 1) { |
728131d8 | 567 | invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE); |
5f39d397 CM |
568 | ClearPagePrivate(page); |
569 | set_page_private(page, 0); | |
570 | page_cache_release(page); | |
571 | } | |
d98237b3 CM |
572 | return ret; |
573 | } | |
574 | ||
5f39d397 | 575 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 576 | { |
d1310b2e CM |
577 | struct extent_io_tree *tree; |
578 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
579 | extent_invalidatepage(tree, page, offset); |
580 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 581 | if (PagePrivate(page)) { |
4ef64eae | 582 | invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE); |
9ad6b7bc CM |
583 | ClearPagePrivate(page); |
584 | set_page_private(page, 0); | |
585 | page_cache_release(page); | |
586 | } | |
d98237b3 CM |
587 | } |
588 | ||
5f39d397 | 589 | #if 0 |
d98237b3 | 590 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
ed2ff2cb | 591 | { |
87cbda5c | 592 | struct buffer_head *bh; |
0f7d52f4 | 593 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
87cbda5c | 594 | struct buffer_head *head; |
87cbda5c CM |
595 | if (!page_has_buffers(page)) { |
596 | create_empty_buffers(page, root->fs_info->sb->s_blocksize, | |
597 | (1 << BH_Dirty)|(1 << BH_Uptodate)); | |
598 | } | |
599 | head = page_buffers(page); | |
600 | bh = head; | |
601 | do { | |
602 | if (buffer_dirty(bh)) | |
603 | csum_tree_block(root, bh, 0); | |
604 | bh = bh->b_this_page; | |
605 | } while (bh != head); | |
d98237b3 | 606 | return block_write_full_page(page, btree_get_block, wbc); |
ed2ff2cb | 607 | } |
5f39d397 | 608 | #endif |
eb60ceac | 609 | |
d98237b3 CM |
610 | static struct address_space_operations btree_aops = { |
611 | .readpage = btree_readpage, | |
612 | .writepage = btree_writepage, | |
0da5468f | 613 | .writepages = btree_writepages, |
5f39d397 CM |
614 | .releasepage = btree_releasepage, |
615 | .invalidatepage = btree_invalidatepage, | |
d98237b3 CM |
616 | .sync_page = block_sync_page, |
617 | }; | |
618 | ||
ca7a79ad CM |
619 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
620 | u64 parent_transid) | |
090d1875 | 621 | { |
5f39d397 CM |
622 | struct extent_buffer *buf = NULL; |
623 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 624 | int ret = 0; |
090d1875 | 625 | |
db94535d | 626 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 627 | if (!buf) |
090d1875 | 628 | return 0; |
d1310b2e | 629 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
f188591e | 630 | buf, 0, 0, btree_get_extent, 0); |
5f39d397 | 631 | free_extent_buffer(buf); |
de428b63 | 632 | return ret; |
090d1875 CM |
633 | } |
634 | ||
0999df54 CM |
635 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
636 | u64 bytenr, u32 blocksize) | |
637 | { | |
638 | struct inode *btree_inode = root->fs_info->btree_inode; | |
639 | struct extent_buffer *eb; | |
640 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
641 | bytenr, blocksize, GFP_NOFS); | |
642 | return eb; | |
643 | } | |
644 | ||
645 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
646 | u64 bytenr, u32 blocksize) | |
647 | { | |
648 | struct inode *btree_inode = root->fs_info->btree_inode; | |
649 | struct extent_buffer *eb; | |
650 | ||
651 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
652 | bytenr, blocksize, NULL, GFP_NOFS); | |
653 | return eb; | |
654 | } | |
655 | ||
656 | ||
657 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, | |
ca7a79ad | 658 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
659 | { |
660 | struct extent_buffer *buf = NULL; | |
661 | struct inode *btree_inode = root->fs_info->btree_inode; | |
662 | struct extent_io_tree *io_tree; | |
663 | int ret; | |
664 | ||
665 | io_tree = &BTRFS_I(btree_inode)->io_tree; | |
666 | ||
667 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
668 | if (!buf) | |
669 | return NULL; | |
0999df54 | 670 | |
ca7a79ad | 671 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 CM |
672 | |
673 | if (ret == 0) { | |
674 | buf->flags |= EXTENT_UPTODATE; | |
675 | } | |
5f39d397 | 676 | return buf; |
ce9adaa5 | 677 | |
eb60ceac CM |
678 | } |
679 | ||
e089f05c | 680 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 681 | struct extent_buffer *buf) |
ed2ff2cb | 682 | { |
5f39d397 | 683 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 | 684 | if (btrfs_header_generation(buf) == |
925baedd CM |
685 | root->fs_info->running_transaction->transid) { |
686 | WARN_ON(!btrfs_tree_locked(buf)); | |
d1310b2e | 687 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 688 | buf); |
925baedd | 689 | } |
5f39d397 CM |
690 | return 0; |
691 | } | |
692 | ||
693 | int wait_on_tree_block_writeback(struct btrfs_root *root, | |
694 | struct extent_buffer *buf) | |
695 | { | |
696 | struct inode *btree_inode = root->fs_info->btree_inode; | |
d1310b2e | 697 | wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
698 | buf); |
699 | return 0; | |
700 | } | |
701 | ||
db94535d | 702 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 703 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 704 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 705 | u64 objectid) |
d97e63b6 | 706 | { |
cfaa7295 | 707 | root->node = NULL; |
0f7d52f4 | 708 | root->inode = NULL; |
a28ec197 | 709 | root->commit_root = NULL; |
db94535d CM |
710 | root->sectorsize = sectorsize; |
711 | root->nodesize = nodesize; | |
712 | root->leafsize = leafsize; | |
87ee04eb | 713 | root->stripesize = stripesize; |
123abc88 | 714 | root->ref_cows = 0; |
0b86a832 CM |
715 | root->track_dirty = 0; |
716 | ||
9f5fae2f | 717 | root->fs_info = fs_info; |
0f7d52f4 CM |
718 | root->objectid = objectid; |
719 | root->last_trans = 0; | |
1b05da2e CM |
720 | root->highest_inode = 0; |
721 | root->last_inode_alloc = 0; | |
58176a96 | 722 | root->name = NULL; |
4313b399 | 723 | root->in_sysfs = 0; |
0b86a832 CM |
724 | |
725 | INIT_LIST_HEAD(&root->dirty_list); | |
925baedd | 726 | spin_lock_init(&root->node_lock); |
3768f368 CM |
727 | memset(&root->root_key, 0, sizeof(root->root_key)); |
728 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 729 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 JB |
730 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
731 | init_completion(&root->kobj_unregister); | |
6702ed49 CM |
732 | root->defrag_running = 0; |
733 | root->defrag_level = 0; | |
4d775673 | 734 | root->root_key.objectid = objectid; |
3768f368 CM |
735 | return 0; |
736 | } | |
737 | ||
db94535d | 738 | static int find_and_setup_root(struct btrfs_root *tree_root, |
9f5fae2f CM |
739 | struct btrfs_fs_info *fs_info, |
740 | u64 objectid, | |
e20d96d6 | 741 | struct btrfs_root *root) |
3768f368 CM |
742 | { |
743 | int ret; | |
db94535d | 744 | u32 blocksize; |
3768f368 | 745 | |
db94535d | 746 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
747 | tree_root->sectorsize, tree_root->stripesize, |
748 | root, fs_info, objectid); | |
3768f368 CM |
749 | ret = btrfs_find_last_root(tree_root, objectid, |
750 | &root->root_item, &root->root_key); | |
751 | BUG_ON(ret); | |
752 | ||
db94535d CM |
753 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
754 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
ca7a79ad | 755 | blocksize, 0); |
3768f368 | 756 | BUG_ON(!root->node); |
d97e63b6 CM |
757 | return 0; |
758 | } | |
759 | ||
5eda7b5e CM |
760 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info, |
761 | struct btrfs_key *location) | |
0f7d52f4 CM |
762 | { |
763 | struct btrfs_root *root; | |
764 | struct btrfs_root *tree_root = fs_info->tree_root; | |
765 | struct btrfs_path *path; | |
5f39d397 | 766 | struct extent_buffer *l; |
1b05da2e | 767 | u64 highest_inode; |
db94535d | 768 | u32 blocksize; |
0f7d52f4 CM |
769 | int ret = 0; |
770 | ||
5eda7b5e | 771 | root = kzalloc(sizeof(*root), GFP_NOFS); |
0cf6c620 | 772 | if (!root) |
0f7d52f4 | 773 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 774 | if (location->offset == (u64)-1) { |
db94535d | 775 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
776 | location->objectid, root); |
777 | if (ret) { | |
0f7d52f4 CM |
778 | kfree(root); |
779 | return ERR_PTR(ret); | |
780 | } | |
781 | goto insert; | |
782 | } | |
783 | ||
db94535d | 784 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
785 | tree_root->sectorsize, tree_root->stripesize, |
786 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
787 | |
788 | path = btrfs_alloc_path(); | |
789 | BUG_ON(!path); | |
790 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); | |
791 | if (ret != 0) { | |
0f7d52f4 CM |
792 | if (ret > 0) |
793 | ret = -ENOENT; | |
794 | goto out; | |
795 | } | |
5f39d397 CM |
796 | l = path->nodes[0]; |
797 | read_extent_buffer(l, &root->root_item, | |
798 | btrfs_item_ptr_offset(l, path->slots[0]), | |
0f7d52f4 | 799 | sizeof(root->root_item)); |
44b36eb2 | 800 | memcpy(&root->root_key, location, sizeof(*location)); |
0f7d52f4 CM |
801 | ret = 0; |
802 | out: | |
803 | btrfs_release_path(root, path); | |
804 | btrfs_free_path(path); | |
805 | if (ret) { | |
806 | kfree(root); | |
807 | return ERR_PTR(ret); | |
808 | } | |
db94535d CM |
809 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
810 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
ca7a79ad | 811 | blocksize, 0); |
0f7d52f4 CM |
812 | BUG_ON(!root->node); |
813 | insert: | |
0f7d52f4 | 814 | root->ref_cows = 1; |
5eda7b5e CM |
815 | ret = btrfs_find_highest_inode(root, &highest_inode); |
816 | if (ret == 0) { | |
817 | root->highest_inode = highest_inode; | |
818 | root->last_inode_alloc = highest_inode; | |
819 | } | |
820 | return root; | |
821 | } | |
822 | ||
dc17ff8f CM |
823 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
824 | u64 root_objectid) | |
825 | { | |
826 | struct btrfs_root *root; | |
827 | ||
828 | if (root_objectid == BTRFS_ROOT_TREE_OBJECTID) | |
829 | return fs_info->tree_root; | |
830 | if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
831 | return fs_info->extent_root; | |
832 | ||
833 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
834 | (unsigned long)root_objectid); | |
835 | return root; | |
836 | } | |
837 | ||
edbd8d4e CM |
838 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
839 | struct btrfs_key *location) | |
5eda7b5e CM |
840 | { |
841 | struct btrfs_root *root; | |
842 | int ret; | |
843 | ||
edbd8d4e CM |
844 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
845 | return fs_info->tree_root; | |
846 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
847 | return fs_info->extent_root; | |
8f18cf13 CM |
848 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
849 | return fs_info->chunk_root; | |
850 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
851 | return fs_info->dev_root; | |
edbd8d4e | 852 | |
5eda7b5e CM |
853 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
854 | (unsigned long)location->objectid); | |
855 | if (root) | |
856 | return root; | |
857 | ||
858 | root = btrfs_read_fs_root_no_radix(fs_info, location); | |
859 | if (IS_ERR(root)) | |
860 | return root; | |
2619ba1f CM |
861 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
862 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 CM |
863 | root); |
864 | if (ret) { | |
5f39d397 | 865 | free_extent_buffer(root->node); |
0f7d52f4 CM |
866 | kfree(root); |
867 | return ERR_PTR(ret); | |
868 | } | |
edbd8d4e CM |
869 | ret = btrfs_find_dead_roots(fs_info->tree_root, |
870 | root->root_key.objectid, root); | |
871 | BUG_ON(ret); | |
872 | ||
873 | return root; | |
874 | } | |
875 | ||
876 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, | |
877 | struct btrfs_key *location, | |
878 | const char *name, int namelen) | |
879 | { | |
880 | struct btrfs_root *root; | |
881 | int ret; | |
882 | ||
883 | root = btrfs_read_fs_root_no_name(fs_info, location); | |
884 | if (!root) | |
885 | return NULL; | |
58176a96 | 886 | |
4313b399 CM |
887 | if (root->in_sysfs) |
888 | return root; | |
889 | ||
58176a96 JB |
890 | ret = btrfs_set_root_name(root, name, namelen); |
891 | if (ret) { | |
5f39d397 | 892 | free_extent_buffer(root->node); |
58176a96 JB |
893 | kfree(root); |
894 | return ERR_PTR(ret); | |
895 | } | |
896 | ||
897 | ret = btrfs_sysfs_add_root(root); | |
898 | if (ret) { | |
5f39d397 | 899 | free_extent_buffer(root->node); |
58176a96 JB |
900 | kfree(root->name); |
901 | kfree(root); | |
902 | return ERR_PTR(ret); | |
903 | } | |
4313b399 | 904 | root->in_sysfs = 1; |
0f7d52f4 CM |
905 | return root; |
906 | } | |
19c00ddc CM |
907 | #if 0 |
908 | static int add_hasher(struct btrfs_fs_info *info, char *type) { | |
909 | struct btrfs_hasher *hasher; | |
910 | ||
911 | hasher = kmalloc(sizeof(*hasher), GFP_NOFS); | |
912 | if (!hasher) | |
913 | return -ENOMEM; | |
914 | hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC); | |
915 | if (!hasher->hash_tfm) { | |
916 | kfree(hasher); | |
917 | return -EINVAL; | |
918 | } | |
919 | spin_lock(&info->hash_lock); | |
920 | list_add(&hasher->list, &info->hashers); | |
921 | spin_unlock(&info->hash_lock); | |
922 | return 0; | |
923 | } | |
924 | #endif | |
04160088 CM |
925 | |
926 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) | |
927 | { | |
928 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
929 | int ret = 0; | |
cb03c743 | 930 | int limit = 256 * info->fs_devices->open_devices; |
04160088 CM |
931 | struct list_head *cur; |
932 | struct btrfs_device *device; | |
933 | struct backing_dev_info *bdi; | |
934 | ||
cb03c743 CM |
935 | if ((bdi_bits & (1 << BDI_write_congested)) && |
936 | atomic_read(&info->nr_async_submits) > limit) { | |
937 | return 1; | |
938 | } | |
939 | ||
04160088 CM |
940 | list_for_each(cur, &info->fs_devices->devices) { |
941 | device = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
942 | if (!device->bdev) |
943 | continue; | |
04160088 CM |
944 | bdi = blk_get_backing_dev_info(device->bdev); |
945 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
946 | ret = 1; | |
947 | break; | |
948 | } | |
949 | } | |
950 | return ret; | |
951 | } | |
952 | ||
38b66988 CM |
953 | /* |
954 | * this unplugs every device on the box, and it is only used when page | |
955 | * is null | |
956 | */ | |
957 | static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) | |
958 | { | |
959 | struct list_head *cur; | |
960 | struct btrfs_device *device; | |
961 | struct btrfs_fs_info *info; | |
962 | ||
963 | info = (struct btrfs_fs_info *)bdi->unplug_io_data; | |
964 | list_for_each(cur, &info->fs_devices->devices) { | |
965 | device = list_entry(cur, struct btrfs_device, dev_list); | |
966 | bdi = blk_get_backing_dev_info(device->bdev); | |
967 | if (bdi->unplug_io_fn) { | |
968 | bdi->unplug_io_fn(bdi, page); | |
969 | } | |
970 | } | |
971 | } | |
972 | ||
04160088 CM |
973 | void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) |
974 | { | |
38b66988 | 975 | struct inode *inode; |
f2d8d74d CM |
976 | struct extent_map_tree *em_tree; |
977 | struct extent_map *em; | |
bcbfce8a | 978 | struct address_space *mapping; |
38b66988 CM |
979 | u64 offset; |
980 | ||
bcbfce8a | 981 | /* the generic O_DIRECT read code does this */ |
38b66988 CM |
982 | if (!page) { |
983 | __unplug_io_fn(bdi, page); | |
984 | return; | |
985 | } | |
986 | ||
bcbfce8a CM |
987 | /* |
988 | * page->mapping may change at any time. Get a consistent copy | |
989 | * and use that for everything below | |
990 | */ | |
991 | smp_mb(); | |
992 | mapping = page->mapping; | |
993 | if (!mapping) | |
994 | return; | |
995 | ||
996 | inode = mapping->host; | |
38b66988 | 997 | offset = page_offset(page); |
04160088 | 998 | |
f2d8d74d CM |
999 | em_tree = &BTRFS_I(inode)->extent_tree; |
1000 | spin_lock(&em_tree->lock); | |
1001 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); | |
1002 | spin_unlock(&em_tree->lock); | |
1003 | if (!em) | |
1004 | return; | |
1005 | ||
1006 | offset = offset - em->start; | |
1007 | btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree, | |
1008 | em->block_start + offset, page); | |
1009 | free_extent_map(em); | |
04160088 CM |
1010 | } |
1011 | ||
1012 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) | |
1013 | { | |
51ebc0d3 | 1014 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) |
04160088 | 1015 | bdi_init(bdi); |
b248a415 | 1016 | #endif |
4575c9cc | 1017 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1018 | bdi->state = 0; |
1019 | bdi->capabilities = default_backing_dev_info.capabilities; | |
1020 | bdi->unplug_io_fn = btrfs_unplug_io_fn; | |
1021 | bdi->unplug_io_data = info; | |
1022 | bdi->congested_fn = btrfs_congested_fn; | |
1023 | bdi->congested_data = info; | |
1024 | return 0; | |
1025 | } | |
1026 | ||
ce9adaa5 CM |
1027 | static int bio_ready_for_csum(struct bio *bio) |
1028 | { | |
1029 | u64 length = 0; | |
1030 | u64 buf_len = 0; | |
1031 | u64 start = 0; | |
1032 | struct page *page; | |
1033 | struct extent_io_tree *io_tree = NULL; | |
1034 | struct btrfs_fs_info *info = NULL; | |
1035 | struct bio_vec *bvec; | |
1036 | int i; | |
1037 | int ret; | |
1038 | ||
1039 | bio_for_each_segment(bvec, bio, i) { | |
1040 | page = bvec->bv_page; | |
1041 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
1042 | length += bvec->bv_len; | |
1043 | continue; | |
1044 | } | |
1045 | if (!page->private) { | |
1046 | length += bvec->bv_len; | |
1047 | continue; | |
1048 | } | |
1049 | length = bvec->bv_len; | |
1050 | buf_len = page->private >> 2; | |
1051 | start = page_offset(page) + bvec->bv_offset; | |
1052 | io_tree = &BTRFS_I(page->mapping->host)->io_tree; | |
1053 | info = BTRFS_I(page->mapping->host)->root->fs_info; | |
1054 | } | |
1055 | /* are we fully contained in this bio? */ | |
1056 | if (buf_len <= length) | |
1057 | return 1; | |
1058 | ||
1059 | ret = extent_range_uptodate(io_tree, start + length, | |
1060 | start + buf_len - 1); | |
1061 | if (ret == 1) | |
1062 | return ret; | |
1063 | return ret; | |
1064 | } | |
1065 | ||
8b712842 CM |
1066 | /* |
1067 | * called by the kthread helper functions to finally call the bio end_io | |
1068 | * functions. This is where read checksum verification actually happens | |
1069 | */ | |
1070 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1071 | { |
ce9adaa5 | 1072 | struct bio *bio; |
8b712842 CM |
1073 | struct end_io_wq *end_io_wq; |
1074 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1075 | int error; |
ce9adaa5 | 1076 | |
8b712842 CM |
1077 | end_io_wq = container_of(work, struct end_io_wq, work); |
1078 | bio = end_io_wq->bio; | |
1079 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1080 | |
8b712842 CM |
1081 | /* metadata bios are special because the whole tree block must |
1082 | * be checksummed at once. This makes sure the entire block is in | |
1083 | * ram and up to date before trying to verify things. For | |
1084 | * blocksize <= pagesize, it is basically a noop | |
1085 | */ | |
1086 | if (end_io_wq->metadata && !bio_ready_for_csum(bio)) { | |
1087 | btrfs_queue_worker(&fs_info->endio_workers, | |
1088 | &end_io_wq->work); | |
1089 | return; | |
1090 | } | |
1091 | error = end_io_wq->error; | |
1092 | bio->bi_private = end_io_wq->private; | |
1093 | bio->bi_end_io = end_io_wq->end_io; | |
1094 | kfree(end_io_wq); | |
1095 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
1096 | bio_endio(bio, bio->bi_size, error); | |
44b8bd7e | 1097 | #else |
8b712842 | 1098 | bio_endio(bio, error); |
44b8bd7e | 1099 | #endif |
44b8bd7e CM |
1100 | } |
1101 | ||
8a4b83cc | 1102 | struct btrfs_root *open_ctree(struct super_block *sb, |
dfe25020 CM |
1103 | struct btrfs_fs_devices *fs_devices, |
1104 | char *options) | |
2e635a27 | 1105 | { |
db94535d CM |
1106 | u32 sectorsize; |
1107 | u32 nodesize; | |
1108 | u32 leafsize; | |
1109 | u32 blocksize; | |
87ee04eb | 1110 | u32 stripesize; |
a061fc8d | 1111 | struct buffer_head *bh; |
e20d96d6 CM |
1112 | struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root), |
1113 | GFP_NOFS); | |
1114 | struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root), | |
1115 | GFP_NOFS); | |
8790d502 | 1116 | struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), |
e20d96d6 | 1117 | GFP_NOFS); |
0b86a832 CM |
1118 | struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root), |
1119 | GFP_NOFS); | |
1120 | struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root), | |
1121 | GFP_NOFS); | |
eb60ceac | 1122 | int ret; |
e58ca020 | 1123 | int err = -EINVAL; |
4543df7e | 1124 | |
2c90e5d6 | 1125 | struct btrfs_super_block *disk_super; |
8790d502 | 1126 | |
39279cc3 CM |
1127 | if (!extent_root || !tree_root || !fs_info) { |
1128 | err = -ENOMEM; | |
1129 | goto fail; | |
1130 | } | |
0f7d52f4 | 1131 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS); |
8fd17795 | 1132 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1133 | INIT_LIST_HEAD(&fs_info->dead_roots); |
19c00ddc CM |
1134 | INIT_LIST_HEAD(&fs_info->hashers); |
1135 | spin_lock_init(&fs_info->hash_lock); | |
1832a6d5 | 1136 | spin_lock_init(&fs_info->delalloc_lock); |
cee36a03 | 1137 | spin_lock_init(&fs_info->new_trans_lock); |
19c00ddc | 1138 | |
58176a96 | 1139 | init_completion(&fs_info->kobj_unregister); |
9f5fae2f CM |
1140 | fs_info->tree_root = tree_root; |
1141 | fs_info->extent_root = extent_root; | |
0b86a832 CM |
1142 | fs_info->chunk_root = chunk_root; |
1143 | fs_info->dev_root = dev_root; | |
8a4b83cc | 1144 | fs_info->fs_devices = fs_devices; |
0b86a832 | 1145 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1146 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1147 | btrfs_mapping_init(&fs_info->mapping_tree); |
cb03c743 | 1148 | atomic_set(&fs_info->nr_async_submits, 0); |
e20d96d6 | 1149 | fs_info->sb = sb; |
c59f8951 | 1150 | fs_info->max_extent = (u64)-1; |
6f568d35 | 1151 | fs_info->max_inline = 8192 * 1024; |
04160088 | 1152 | setup_bdi(fs_info, &fs_info->bdi); |
d98237b3 CM |
1153 | fs_info->btree_inode = new_inode(sb); |
1154 | fs_info->btree_inode->i_ino = 1; | |
2c90e5d6 | 1155 | fs_info->btree_inode->i_nlink = 1; |
4543df7e | 1156 | fs_info->thread_pool_size = min(num_online_cpus() + 2, 8); |
0afbaf8c | 1157 | |
a061fc8d CM |
1158 | sb->s_blocksize = 4096; |
1159 | sb->s_blocksize_bits = blksize_bits(4096); | |
1160 | ||
0afbaf8c CM |
1161 | /* |
1162 | * we set the i_size on the btree inode to the max possible int. | |
1163 | * the real end of the address space is determined by all of | |
1164 | * the devices in the system | |
1165 | */ | |
1166 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 1167 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
1168 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
1169 | ||
d1310b2e | 1170 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
5f39d397 CM |
1171 | fs_info->btree_inode->i_mapping, |
1172 | GFP_NOFS); | |
d1310b2e CM |
1173 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, |
1174 | GFP_NOFS); | |
1175 | ||
1176 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 1177 | |
d1310b2e | 1178 | extent_io_tree_init(&fs_info->free_space_cache, |
f510cfec | 1179 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1180 | extent_io_tree_init(&fs_info->block_group_cache, |
96b5179d | 1181 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1182 | extent_io_tree_init(&fs_info->pinned_extents, |
1a5bc167 | 1183 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1184 | extent_io_tree_init(&fs_info->pending_del, |
1a5bc167 | 1185 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1186 | extent_io_tree_init(&fs_info->extent_ins, |
1a5bc167 | 1187 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
e66f709b | 1188 | fs_info->do_barriers = 1; |
e18e4809 | 1189 | |
6da6abae CM |
1190 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) |
1191 | INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info); | |
1192 | #else | |
08607c1b | 1193 | INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner); |
6da6abae | 1194 | #endif |
0f7d52f4 CM |
1195 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
1196 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
1197 | sizeof(struct btrfs_key)); | |
22b0ebda | 1198 | insert_inode_hash(fs_info->btree_inode); |
d98237b3 | 1199 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
39279cc3 | 1200 | |
79154b1b | 1201 | mutex_init(&fs_info->trans_mutex); |
d561c025 | 1202 | mutex_init(&fs_info->fs_mutex); |
925baedd CM |
1203 | mutex_init(&fs_info->alloc_mutex); |
1204 | mutex_init(&fs_info->chunk_mutex); | |
3768f368 | 1205 | |
19c00ddc CM |
1206 | #if 0 |
1207 | ret = add_hasher(fs_info, "crc32c"); | |
1208 | if (ret) { | |
1209 | printk("btrfs: failed hash setup, modprobe cryptomgr?\n"); | |
1210 | err = -ENOMEM; | |
1211 | goto fail_iput; | |
1212 | } | |
1213 | #endif | |
0b86a832 | 1214 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 1215 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 1216 | |
d98237b3 | 1217 | |
a061fc8d CM |
1218 | bh = __bread(fs_devices->latest_bdev, |
1219 | BTRFS_SUPER_INFO_OFFSET / 4096, 4096); | |
1220 | if (!bh) | |
39279cc3 | 1221 | goto fail_iput; |
39279cc3 | 1222 | |
a061fc8d CM |
1223 | memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy)); |
1224 | brelse(bh); | |
5f39d397 | 1225 | |
a061fc8d | 1226 | memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE); |
0b86a832 | 1227 | |
5f39d397 | 1228 | disk_super = &fs_info->super_copy; |
0f7d52f4 | 1229 | if (!btrfs_super_root(disk_super)) |
39279cc3 | 1230 | goto fail_sb_buffer; |
0f7d52f4 | 1231 | |
edf24abe CH |
1232 | err = btrfs_parse_options(tree_root, options); |
1233 | if (err) | |
1234 | goto fail_sb_buffer; | |
dfe25020 | 1235 | |
4543df7e CM |
1236 | /* |
1237 | * we need to start all the end_io workers up front because the | |
1238 | * queue work function gets called at interrupt time, and so it | |
1239 | * cannot dynamically grow. | |
1240 | */ | |
1241 | btrfs_init_workers(&fs_info->workers, fs_info->thread_pool_size); | |
1cc127b5 | 1242 | btrfs_init_workers(&fs_info->submit_workers, fs_info->thread_pool_size); |
4543df7e CM |
1243 | btrfs_init_workers(&fs_info->endio_workers, fs_info->thread_pool_size); |
1244 | btrfs_start_workers(&fs_info->workers, 1); | |
1cc127b5 | 1245 | btrfs_start_workers(&fs_info->submit_workers, 1); |
4543df7e CM |
1246 | btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size); |
1247 | ||
1248 | ||
edf24abe | 1249 | err = -EINVAL; |
a0af469b | 1250 | if (btrfs_super_num_devices(disk_super) > fs_devices->open_devices) { |
8a4b83cc CM |
1251 | printk("Btrfs: wanted %llu devices, but found %llu\n", |
1252 | (unsigned long long)btrfs_super_num_devices(disk_super), | |
a0af469b | 1253 | (unsigned long long)fs_devices->open_devices); |
dfe25020 CM |
1254 | if (btrfs_test_opt(tree_root, DEGRADED)) |
1255 | printk("continuing in degraded mode\n"); | |
1256 | else { | |
1257 | goto fail_sb_buffer; | |
1258 | } | |
8a4b83cc | 1259 | } |
dfe25020 | 1260 | |
4575c9cc CM |
1261 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
1262 | ||
db94535d CM |
1263 | nodesize = btrfs_super_nodesize(disk_super); |
1264 | leafsize = btrfs_super_leafsize(disk_super); | |
1265 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 1266 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
1267 | tree_root->nodesize = nodesize; |
1268 | tree_root->leafsize = leafsize; | |
1269 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 1270 | tree_root->stripesize = stripesize; |
a061fc8d CM |
1271 | |
1272 | sb->s_blocksize = sectorsize; | |
1273 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 1274 | |
39279cc3 CM |
1275 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
1276 | sizeof(disk_super->magic))) { | |
1277 | printk("btrfs: valid FS not found on %s\n", sb->s_id); | |
1278 | goto fail_sb_buffer; | |
1279 | } | |
19c00ddc | 1280 | |
0b86a832 | 1281 | mutex_lock(&fs_info->fs_mutex); |
0d81ba5d | 1282 | |
925baedd | 1283 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1284 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 1285 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f CM |
1286 | if (ret) { |
1287 | printk("btrfs: failed to read the system array on %s\n", | |
1288 | sb->s_id); | |
1289 | goto fail_sys_array; | |
1290 | } | |
0b86a832 CM |
1291 | |
1292 | blocksize = btrfs_level_size(tree_root, | |
1293 | btrfs_super_chunk_root_level(disk_super)); | |
1294 | ||
1295 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1296 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
1297 | ||
1298 | chunk_root->node = read_tree_block(chunk_root, | |
1299 | btrfs_super_chunk_root(disk_super), | |
ca7a79ad | 1300 | blocksize, 0); |
0b86a832 CM |
1301 | BUG_ON(!chunk_root->node); |
1302 | ||
e17cade2 CM |
1303 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
1304 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), | |
1305 | BTRFS_UUID_SIZE); | |
1306 | ||
925baedd | 1307 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1308 | ret = btrfs_read_chunk_tree(chunk_root); |
925baedd | 1309 | mutex_unlock(&fs_info->chunk_mutex); |
0b86a832 CM |
1310 | BUG_ON(ret); |
1311 | ||
dfe25020 CM |
1312 | btrfs_close_extra_devices(fs_devices); |
1313 | ||
db94535d CM |
1314 | blocksize = btrfs_level_size(tree_root, |
1315 | btrfs_super_root_level(disk_super)); | |
19c00ddc | 1316 | |
0b86a832 | 1317 | |
e20d96d6 | 1318 | tree_root->node = read_tree_block(tree_root, |
db94535d | 1319 | btrfs_super_root(disk_super), |
ca7a79ad | 1320 | blocksize, 0); |
39279cc3 CM |
1321 | if (!tree_root->node) |
1322 | goto fail_sb_buffer; | |
3768f368 | 1323 | |
db94535d CM |
1324 | |
1325 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 1326 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 1327 | if (ret) |
39279cc3 | 1328 | goto fail_tree_root; |
0b86a832 CM |
1329 | extent_root->track_dirty = 1; |
1330 | ||
1331 | ret = find_and_setup_root(tree_root, fs_info, | |
1332 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
1333 | dev_root->track_dirty = 1; | |
1334 | ||
1335 | if (ret) | |
1336 | goto fail_extent_root; | |
3768f368 | 1337 | |
9078a3e1 CM |
1338 | btrfs_read_block_groups(extent_root); |
1339 | ||
0f7d52f4 | 1340 | fs_info->generation = btrfs_super_generation(disk_super) + 1; |
d18a2c44 CM |
1341 | fs_info->data_alloc_profile = (u64)-1; |
1342 | fs_info->metadata_alloc_profile = (u64)-1; | |
1343 | fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; | |
1344 | ||
5be6f7f1 | 1345 | mutex_unlock(&fs_info->fs_mutex); |
0f7d52f4 | 1346 | return tree_root; |
39279cc3 | 1347 | |
0b86a832 CM |
1348 | fail_extent_root: |
1349 | free_extent_buffer(extent_root->node); | |
39279cc3 | 1350 | fail_tree_root: |
5f39d397 | 1351 | free_extent_buffer(tree_root->node); |
84eed90f CM |
1352 | fail_sys_array: |
1353 | mutex_unlock(&fs_info->fs_mutex); | |
39279cc3 | 1354 | fail_sb_buffer: |
2d2ae547 | 1355 | extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree); |
8b712842 CM |
1356 | btrfs_stop_workers(&fs_info->workers); |
1357 | btrfs_stop_workers(&fs_info->endio_workers); | |
1cc127b5 | 1358 | btrfs_stop_workers(&fs_info->submit_workers); |
4543df7e CM |
1359 | fail_iput: |
1360 | iput(fs_info->btree_inode); | |
39279cc3 | 1361 | fail: |
dfe25020 | 1362 | btrfs_close_devices(fs_info->fs_devices); |
84eed90f CM |
1363 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
1364 | ||
39279cc3 CM |
1365 | kfree(extent_root); |
1366 | kfree(tree_root); | |
51ebc0d3 | 1367 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) |
2d2ae547 | 1368 | bdi_destroy(&fs_info->bdi); |
b248a415 | 1369 | #endif |
39279cc3 CM |
1370 | kfree(fs_info); |
1371 | return ERR_PTR(err); | |
eb60ceac CM |
1372 | } |
1373 | ||
f2984462 CM |
1374 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
1375 | { | |
1376 | char b[BDEVNAME_SIZE]; | |
1377 | ||
1378 | if (uptodate) { | |
1379 | set_buffer_uptodate(bh); | |
1380 | } else { | |
1381 | if (!buffer_eopnotsupp(bh) && printk_ratelimit()) { | |
1382 | printk(KERN_WARNING "lost page write due to " | |
1383 | "I/O error on %s\n", | |
1384 | bdevname(bh->b_bdev, b)); | |
1385 | } | |
1259ab75 CM |
1386 | /* note, we dont' set_buffer_write_io_error because we have |
1387 | * our own ways of dealing with the IO errors | |
1388 | */ | |
f2984462 CM |
1389 | clear_buffer_uptodate(bh); |
1390 | } | |
1391 | unlock_buffer(bh); | |
1392 | put_bh(bh); | |
1393 | } | |
1394 | ||
1395 | int write_all_supers(struct btrfs_root *root) | |
1396 | { | |
1397 | struct list_head *cur; | |
1398 | struct list_head *head = &root->fs_info->fs_devices->devices; | |
1399 | struct btrfs_device *dev; | |
a061fc8d | 1400 | struct btrfs_super_block *sb; |
f2984462 CM |
1401 | struct btrfs_dev_item *dev_item; |
1402 | struct buffer_head *bh; | |
1403 | int ret; | |
1404 | int do_barriers; | |
a236aed1 CM |
1405 | int max_errors; |
1406 | int total_errors = 0; | |
a061fc8d CM |
1407 | u32 crc; |
1408 | u64 flags; | |
f2984462 | 1409 | |
a236aed1 | 1410 | max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1; |
f2984462 CM |
1411 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
1412 | ||
a061fc8d CM |
1413 | sb = &root->fs_info->super_for_commit; |
1414 | dev_item = &sb->dev_item; | |
f2984462 CM |
1415 | list_for_each(cur, head) { |
1416 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1417 | if (!dev->bdev) { |
1418 | total_errors++; | |
1419 | continue; | |
1420 | } | |
1421 | if (!dev->in_fs_metadata) | |
1422 | continue; | |
1423 | ||
a061fc8d CM |
1424 | btrfs_set_stack_device_type(dev_item, dev->type); |
1425 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
1426 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
1427 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
1428 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
1429 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
1430 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
1431 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
1432 | flags = btrfs_super_flags(sb); | |
1433 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
1434 | ||
1435 | ||
1436 | crc = ~(u32)0; | |
1437 | crc = btrfs_csum_data(root, (char *)sb + BTRFS_CSUM_SIZE, crc, | |
1438 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
1439 | btrfs_csum_final(crc, sb->csum); | |
1440 | ||
1441 | bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET / 4096, | |
f2984462 CM |
1442 | BTRFS_SUPER_INFO_SIZE); |
1443 | ||
a061fc8d | 1444 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
f2984462 CM |
1445 | dev->pending_io = bh; |
1446 | ||
1447 | get_bh(bh); | |
1448 | set_buffer_uptodate(bh); | |
1449 | lock_buffer(bh); | |
1450 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
1451 | ||
1452 | if (do_barriers && dev->barriers) { | |
1453 | ret = submit_bh(WRITE_BARRIER, bh); | |
1454 | if (ret == -EOPNOTSUPP) { | |
1455 | printk("btrfs: disabling barriers on dev %s\n", | |
1456 | dev->name); | |
1457 | set_buffer_uptodate(bh); | |
1458 | dev->barriers = 0; | |
1459 | get_bh(bh); | |
1460 | lock_buffer(bh); | |
1461 | ret = submit_bh(WRITE, bh); | |
1462 | } | |
1463 | } else { | |
1464 | ret = submit_bh(WRITE, bh); | |
1465 | } | |
a236aed1 CM |
1466 | if (ret) |
1467 | total_errors++; | |
f2984462 | 1468 | } |
a236aed1 CM |
1469 | if (total_errors > max_errors) { |
1470 | printk("btrfs: %d errors while writing supers\n", total_errors); | |
1471 | BUG(); | |
1472 | } | |
1473 | total_errors = 0; | |
f2984462 CM |
1474 | |
1475 | list_for_each(cur, head) { | |
1476 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1477 | if (!dev->bdev) |
1478 | continue; | |
1479 | if (!dev->in_fs_metadata) | |
1480 | continue; | |
1481 | ||
f2984462 CM |
1482 | BUG_ON(!dev->pending_io); |
1483 | bh = dev->pending_io; | |
1484 | wait_on_buffer(bh); | |
1485 | if (!buffer_uptodate(dev->pending_io)) { | |
1486 | if (do_barriers && dev->barriers) { | |
1487 | printk("btrfs: disabling barriers on dev %s\n", | |
1488 | dev->name); | |
1489 | set_buffer_uptodate(bh); | |
1490 | get_bh(bh); | |
1491 | lock_buffer(bh); | |
1492 | dev->barriers = 0; | |
1493 | ret = submit_bh(WRITE, bh); | |
1494 | BUG_ON(ret); | |
1495 | wait_on_buffer(bh); | |
1259ab75 CM |
1496 | if (!buffer_uptodate(bh)) |
1497 | total_errors++; | |
f2984462 | 1498 | } else { |
a236aed1 | 1499 | total_errors++; |
f2984462 CM |
1500 | } |
1501 | ||
1502 | } | |
1503 | dev->pending_io = NULL; | |
1504 | brelse(bh); | |
1505 | } | |
a236aed1 CM |
1506 | if (total_errors > max_errors) { |
1507 | printk("btrfs: %d errors while writing supers\n", total_errors); | |
1508 | BUG(); | |
1509 | } | |
f2984462 CM |
1510 | return 0; |
1511 | } | |
1512 | ||
e089f05c | 1513 | int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root |
79154b1b | 1514 | *root) |
eb60ceac | 1515 | { |
e66f709b | 1516 | int ret; |
5f39d397 | 1517 | |
f2984462 | 1518 | ret = write_all_supers(root); |
5f39d397 | 1519 | return ret; |
cfaa7295 CM |
1520 | } |
1521 | ||
5eda7b5e | 1522 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f CM |
1523 | { |
1524 | radix_tree_delete(&fs_info->fs_roots_radix, | |
1525 | (unsigned long)root->root_key.objectid); | |
b99aa6cb CM |
1526 | if (root->in_sysfs) |
1527 | btrfs_sysfs_del_root(root); | |
2619ba1f CM |
1528 | if (root->inode) |
1529 | iput(root->inode); | |
1530 | if (root->node) | |
5f39d397 | 1531 | free_extent_buffer(root->node); |
2619ba1f | 1532 | if (root->commit_root) |
5f39d397 | 1533 | free_extent_buffer(root->commit_root); |
58176a96 JB |
1534 | if (root->name) |
1535 | kfree(root->name); | |
2619ba1f CM |
1536 | kfree(root); |
1537 | return 0; | |
1538 | } | |
1539 | ||
35b7e476 | 1540 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
1541 | { |
1542 | int ret; | |
1543 | struct btrfs_root *gang[8]; | |
1544 | int i; | |
1545 | ||
1546 | while(1) { | |
1547 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
1548 | (void **)gang, 0, | |
1549 | ARRAY_SIZE(gang)); | |
1550 | if (!ret) | |
1551 | break; | |
2619ba1f | 1552 | for (i = 0; i < ret; i++) |
5eda7b5e | 1553 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
1554 | } |
1555 | return 0; | |
1556 | } | |
b4100d64 | 1557 | |
e20d96d6 | 1558 | int close_ctree(struct btrfs_root *root) |
cfaa7295 | 1559 | { |
3768f368 | 1560 | int ret; |
e089f05c | 1561 | struct btrfs_trans_handle *trans; |
0f7d52f4 | 1562 | struct btrfs_fs_info *fs_info = root->fs_info; |
e089f05c | 1563 | |
facda1e7 | 1564 | fs_info->closing = 1; |
08607c1b | 1565 | btrfs_transaction_flush_work(root); |
0f7d52f4 | 1566 | mutex_lock(&fs_info->fs_mutex); |
6702ed49 | 1567 | btrfs_defrag_dirty_roots(root->fs_info); |
79154b1b | 1568 | trans = btrfs_start_transaction(root, 1); |
54aa1f4d | 1569 | ret = btrfs_commit_transaction(trans, root); |
79154b1b CM |
1570 | /* run commit again to drop the original snapshot */ |
1571 | trans = btrfs_start_transaction(root, 1); | |
1572 | btrfs_commit_transaction(trans, root); | |
1573 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 1574 | BUG_ON(ret); |
d6bfde87 | 1575 | |
79154b1b | 1576 | write_ctree_super(NULL, root); |
0f7d52f4 CM |
1577 | mutex_unlock(&fs_info->fs_mutex); |
1578 | ||
9ad6b7bc CM |
1579 | btrfs_transaction_flush_work(root); |
1580 | ||
b0c68f8b CM |
1581 | if (fs_info->delalloc_bytes) { |
1582 | printk("btrfs: at unmount delalloc count %Lu\n", | |
1583 | fs_info->delalloc_bytes); | |
1584 | } | |
0f7d52f4 | 1585 | if (fs_info->extent_root->node) |
5f39d397 | 1586 | free_extent_buffer(fs_info->extent_root->node); |
f510cfec | 1587 | |
0f7d52f4 | 1588 | if (fs_info->tree_root->node) |
5f39d397 | 1589 | free_extent_buffer(fs_info->tree_root->node); |
f510cfec | 1590 | |
0b86a832 CM |
1591 | if (root->fs_info->chunk_root->node); |
1592 | free_extent_buffer(root->fs_info->chunk_root->node); | |
1593 | ||
1594 | if (root->fs_info->dev_root->node); | |
1595 | free_extent_buffer(root->fs_info->dev_root->node); | |
1596 | ||
9078a3e1 | 1597 | btrfs_free_block_groups(root->fs_info); |
0f7d52f4 | 1598 | del_fs_roots(fs_info); |
d10c5f31 CM |
1599 | |
1600 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
1601 | ||
d1310b2e CM |
1602 | extent_io_tree_empty_lru(&fs_info->free_space_cache); |
1603 | extent_io_tree_empty_lru(&fs_info->block_group_cache); | |
1604 | extent_io_tree_empty_lru(&fs_info->pinned_extents); | |
1605 | extent_io_tree_empty_lru(&fs_info->pending_del); | |
1606 | extent_io_tree_empty_lru(&fs_info->extent_ins); | |
1607 | extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree); | |
d10c5f31 | 1608 | |
db94535d | 1609 | truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); |
9ad6b7bc | 1610 | |
8b712842 CM |
1611 | btrfs_stop_workers(&fs_info->workers); |
1612 | btrfs_stop_workers(&fs_info->endio_workers); | |
1cc127b5 | 1613 | btrfs_stop_workers(&fs_info->submit_workers); |
d6bfde87 | 1614 | |
db94535d | 1615 | iput(fs_info->btree_inode); |
19c00ddc CM |
1616 | #if 0 |
1617 | while(!list_empty(&fs_info->hashers)) { | |
1618 | struct btrfs_hasher *hasher; | |
1619 | hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher, | |
1620 | hashers); | |
1621 | list_del(&hasher->hashers); | |
1622 | crypto_free_hash(&fs_info->hash_tfm); | |
1623 | kfree(hasher); | |
1624 | } | |
1625 | #endif | |
dfe25020 | 1626 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 1627 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 1628 | |
51ebc0d3 | 1629 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) |
04160088 | 1630 | bdi_destroy(&fs_info->bdi); |
b248a415 | 1631 | #endif |
0b86a832 | 1632 | |
0f7d52f4 | 1633 | kfree(fs_info->extent_root); |
0f7d52f4 | 1634 | kfree(fs_info->tree_root); |
0b86a832 CM |
1635 | kfree(fs_info->chunk_root); |
1636 | kfree(fs_info->dev_root); | |
eb60ceac CM |
1637 | return 0; |
1638 | } | |
1639 | ||
1259ab75 | 1640 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 1641 | { |
1259ab75 | 1642 | int ret; |
810191ff | 1643 | struct inode *btree_inode = buf->first_page->mapping->host; |
1259ab75 CM |
1644 | |
1645 | ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); | |
1646 | if (!ret) | |
1647 | return ret; | |
1648 | ||
1649 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
1650 | parent_transid); | |
1651 | return !ret; | |
5f39d397 CM |
1652 | } |
1653 | ||
1654 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 1655 | { |
810191ff | 1656 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 1657 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
1658 | buf); |
1659 | } | |
6702ed49 | 1660 | |
5f39d397 CM |
1661 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
1662 | { | |
810191ff | 1663 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
5f39d397 CM |
1664 | u64 transid = btrfs_header_generation(buf); |
1665 | struct inode *btree_inode = root->fs_info->btree_inode; | |
6702ed49 | 1666 | |
925baedd | 1667 | WARN_ON(!btrfs_tree_locked(buf)); |
ccd467d6 CM |
1668 | if (transid != root->fs_info->generation) { |
1669 | printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n", | |
db94535d | 1670 | (unsigned long long)buf->start, |
ccd467d6 CM |
1671 | transid, root->fs_info->generation); |
1672 | WARN_ON(1); | |
1673 | } | |
d1310b2e | 1674 | set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf); |
eb60ceac CM |
1675 | } |
1676 | ||
e2008b61 CM |
1677 | void btrfs_throttle(struct btrfs_root *root) |
1678 | { | |
55c69072 CM |
1679 | struct backing_dev_info *bdi; |
1680 | ||
a061fc8d | 1681 | bdi = &root->fs_info->bdi; |
04005cc7 CM |
1682 | if (root->fs_info->throttles && bdi_write_congested(bdi)) { |
1683 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18) | |
55c69072 | 1684 | congestion_wait(WRITE, HZ/20); |
04005cc7 CM |
1685 | #else |
1686 | blk_congestion_wait(WRITE, HZ/20); | |
1687 | #endif | |
1688 | } | |
e2008b61 CM |
1689 | } |
1690 | ||
d3c2fdcf | 1691 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
35b7e476 | 1692 | { |
188de649 CM |
1693 | /* |
1694 | * looks as though older kernels can get into trouble with | |
1695 | * this code, they end up stuck in balance_dirty_pages forever | |
1696 | */ | |
d6bfde87 CM |
1697 | struct extent_io_tree *tree; |
1698 | u64 num_dirty; | |
1699 | u64 start = 0; | |
1700 | unsigned long thresh = 16 * 1024 * 1024; | |
1701 | tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
1702 | ||
1703 | if (current_is_pdflush()) | |
1704 | return; | |
1705 | ||
1706 | num_dirty = count_range_bits(tree, &start, (u64)-1, | |
1707 | thresh, EXTENT_DIRTY); | |
1708 | if (num_dirty > thresh) { | |
1709 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 1710 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 1711 | } |
188de649 | 1712 | return; |
35b7e476 | 1713 | } |
6b80053d CM |
1714 | |
1715 | void btrfs_set_buffer_defrag(struct extent_buffer *buf) | |
1716 | { | |
810191ff | 1717 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1718 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1719 | set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start, |
6b80053d CM |
1720 | buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS); |
1721 | } | |
1722 | ||
1723 | void btrfs_set_buffer_defrag_done(struct extent_buffer *buf) | |
1724 | { | |
810191ff | 1725 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1726 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1727 | set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start, |
6b80053d CM |
1728 | buf->start + buf->len - 1, EXTENT_DEFRAG_DONE, |
1729 | GFP_NOFS); | |
1730 | } | |
1731 | ||
1732 | int btrfs_buffer_defrag(struct extent_buffer *buf) | |
1733 | { | |
810191ff | 1734 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1735 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1736 | return test_range_bit(&BTRFS_I(btree_inode)->io_tree, |
6b80053d CM |
1737 | buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, 0); |
1738 | } | |
1739 | ||
1740 | int btrfs_buffer_defrag_done(struct extent_buffer *buf) | |
1741 | { | |
810191ff | 1742 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1743 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1744 | return test_range_bit(&BTRFS_I(btree_inode)->io_tree, |
6b80053d CM |
1745 | buf->start, buf->start + buf->len - 1, |
1746 | EXTENT_DEFRAG_DONE, 0); | |
1747 | } | |
1748 | ||
1749 | int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf) | |
1750 | { | |
810191ff | 1751 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1752 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1753 | return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree, |
6b80053d CM |
1754 | buf->start, buf->start + buf->len - 1, |
1755 | EXTENT_DEFRAG_DONE, GFP_NOFS); | |
1756 | } | |
1757 | ||
1758 | int btrfs_clear_buffer_defrag(struct extent_buffer *buf) | |
1759 | { | |
810191ff | 1760 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1761 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1762 | return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree, |
6b80053d CM |
1763 | buf->start, buf->start + buf->len - 1, |
1764 | EXTENT_DEFRAG, GFP_NOFS); | |
1765 | } | |
1766 | ||
ca7a79ad | 1767 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 1768 | { |
810191ff | 1769 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
ce9adaa5 | 1770 | int ret; |
ca7a79ad | 1771 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 CM |
1772 | if (ret == 0) { |
1773 | buf->flags |= EXTENT_UPTODATE; | |
1774 | } | |
1775 | return ret; | |
6b80053d | 1776 | } |
0da5468f | 1777 | |
d1310b2e | 1778 | static struct extent_io_ops btree_extent_io_ops = { |
0da5468f | 1779 | .writepage_io_hook = btree_writepage_io_hook, |
ce9adaa5 | 1780 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
0b86a832 | 1781 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
1782 | /* note we're sharing with inode.c for the merge bio hook */ |
1783 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 1784 | }; |