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 CM |
28 | #include <linux/kthread.h> |
29 | #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20) | |
30 | # include <linux/freezer.h> | |
31 | #else | |
32 | # include <linux/sched.h> | |
33 | #endif | |
a5eb62e3 | 34 | #include "crc32c.h" |
eb60ceac CM |
35 | #include "ctree.h" |
36 | #include "disk-io.h" | |
e089f05c | 37 | #include "transaction.h" |
0f7d52f4 | 38 | #include "btrfs_inode.h" |
0b86a832 | 39 | #include "volumes.h" |
db94535d | 40 | #include "print-tree.h" |
8b712842 | 41 | #include "async-thread.h" |
925baedd | 42 | #include "locking.h" |
017e5369 | 43 | #include "ref-cache.h" |
eb60ceac | 44 | |
5f39d397 CM |
45 | #if 0 |
46 | static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) | |
7eccb903 | 47 | { |
5f39d397 CM |
48 | if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) { |
49 | printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n", | |
50 | (unsigned long long)extent_buffer_blocknr(buf), | |
51 | (unsigned long long)btrfs_header_blocknr(buf)); | |
39279cc3 | 52 | return 1; |
d98237b3 | 53 | } |
9a8dd150 | 54 | return 0; |
eb60ceac | 55 | } |
5f39d397 | 56 | #endif |
eb60ceac | 57 | |
d1310b2e | 58 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 59 | static void end_workqueue_fn(struct btrfs_work *work); |
ce9adaa5 CM |
60 | |
61 | struct end_io_wq { | |
62 | struct bio *bio; | |
63 | bio_end_io_t *end_io; | |
64 | void *private; | |
65 | struct btrfs_fs_info *info; | |
66 | int error; | |
22c59948 | 67 | int metadata; |
ce9adaa5 | 68 | struct list_head list; |
8b712842 | 69 | struct btrfs_work work; |
ce9adaa5 | 70 | }; |
0da5468f | 71 | |
44b8bd7e CM |
72 | struct async_submit_bio { |
73 | struct inode *inode; | |
74 | struct bio *bio; | |
75 | struct list_head list; | |
76 | extent_submit_bio_hook_t *submit_bio_hook; | |
77 | int rw; | |
78 | int mirror_num; | |
8b712842 | 79 | struct btrfs_work work; |
44b8bd7e CM |
80 | }; |
81 | ||
5f39d397 | 82 | struct extent_map *btree_get_extent(struct inode *inode, struct page *page, |
d1310b2e | 83 | size_t page_offset, u64 start, u64 len, |
5f39d397 | 84 | int create) |
7eccb903 | 85 | { |
5f39d397 CM |
86 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
87 | struct extent_map *em; | |
88 | int ret; | |
89 | ||
d1310b2e CM |
90 | spin_lock(&em_tree->lock); |
91 | em = lookup_extent_mapping(em_tree, start, len); | |
a061fc8d CM |
92 | if (em) { |
93 | em->bdev = | |
94 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
95 | spin_unlock(&em_tree->lock); | |
5f39d397 | 96 | goto out; |
a061fc8d CM |
97 | } |
98 | spin_unlock(&em_tree->lock); | |
7b13b7b1 | 99 | |
5f39d397 CM |
100 | em = alloc_extent_map(GFP_NOFS); |
101 | if (!em) { | |
102 | em = ERR_PTR(-ENOMEM); | |
103 | goto out; | |
104 | } | |
105 | em->start = 0; | |
0afbaf8c | 106 | em->len = (u64)-1; |
5f39d397 | 107 | em->block_start = 0; |
a061fc8d | 108 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e CM |
109 | |
110 | spin_lock(&em_tree->lock); | |
5f39d397 CM |
111 | ret = add_extent_mapping(em_tree, em); |
112 | if (ret == -EEXIST) { | |
0afbaf8c CM |
113 | u64 failed_start = em->start; |
114 | u64 failed_len = em->len; | |
115 | ||
116 | printk("failed to insert %Lu %Lu -> %Lu into tree\n", | |
117 | em->start, em->len, em->block_start); | |
5f39d397 | 118 | free_extent_map(em); |
7b13b7b1 | 119 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c CM |
120 | if (em) { |
121 | printk("after failing, found %Lu %Lu %Lu\n", | |
122 | em->start, em->len, em->block_start); | |
7b13b7b1 | 123 | ret = 0; |
0afbaf8c CM |
124 | } else { |
125 | em = lookup_extent_mapping(em_tree, failed_start, | |
126 | failed_len); | |
127 | if (em) { | |
128 | printk("double failure lookup gives us " | |
129 | "%Lu %Lu -> %Lu\n", em->start, | |
130 | em->len, em->block_start); | |
131 | free_extent_map(em); | |
132 | } | |
7b13b7b1 | 133 | ret = -EIO; |
0afbaf8c | 134 | } |
5f39d397 | 135 | } else if (ret) { |
7b13b7b1 CM |
136 | free_extent_map(em); |
137 | em = NULL; | |
5f39d397 | 138 | } |
7b13b7b1 CM |
139 | spin_unlock(&em_tree->lock); |
140 | ||
141 | if (ret) | |
142 | em = ERR_PTR(ret); | |
5f39d397 CM |
143 | out: |
144 | return em; | |
7eccb903 CM |
145 | } |
146 | ||
19c00ddc CM |
147 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
148 | { | |
a5eb62e3 | 149 | return btrfs_crc32c(seed, data, len); |
19c00ddc CM |
150 | } |
151 | ||
152 | void btrfs_csum_final(u32 crc, char *result) | |
153 | { | |
154 | *(__le32 *)result = ~cpu_to_le32(crc); | |
155 | } | |
156 | ||
157 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, | |
158 | int verify) | |
159 | { | |
160 | char result[BTRFS_CRC32_SIZE]; | |
161 | unsigned long len; | |
162 | unsigned long cur_len; | |
163 | unsigned long offset = BTRFS_CSUM_SIZE; | |
164 | char *map_token = NULL; | |
165 | char *kaddr; | |
166 | unsigned long map_start; | |
167 | unsigned long map_len; | |
168 | int err; | |
169 | u32 crc = ~(u32)0; | |
170 | ||
171 | len = buf->len - offset; | |
172 | while(len > 0) { | |
173 | err = map_private_extent_buffer(buf, offset, 32, | |
174 | &map_token, &kaddr, | |
175 | &map_start, &map_len, KM_USER0); | |
176 | if (err) { | |
177 | printk("failed to map extent buffer! %lu\n", | |
178 | offset); | |
179 | return 1; | |
180 | } | |
181 | cur_len = min(len, map_len - (offset - map_start)); | |
182 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
183 | crc, cur_len); | |
184 | len -= cur_len; | |
185 | offset += cur_len; | |
186 | unmap_extent_buffer(buf, map_token, KM_USER0); | |
187 | } | |
188 | btrfs_csum_final(crc, result); | |
189 | ||
190 | if (verify) { | |
e4204ded CM |
191 | int from_this_trans = 0; |
192 | ||
193 | if (root->fs_info->running_transaction && | |
194 | btrfs_header_generation(buf) == | |
195 | root->fs_info->running_transaction->transid) | |
196 | from_this_trans = 1; | |
197 | ||
198 | /* FIXME, this is not good */ | |
63b10fc4 | 199 | if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) { |
e4204ded CM |
200 | u32 val; |
201 | u32 found = 0; | |
202 | memcpy(&found, result, BTRFS_CRC32_SIZE); | |
203 | ||
204 | read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE); | |
205 | printk("btrfs: %s checksum verify failed on %llu " | |
63b10fc4 CM |
206 | "wanted %X found %X from_this_trans %d " |
207 | "level %d\n", | |
19c00ddc | 208 | root->fs_info->sb->s_id, |
63b10fc4 CM |
209 | buf->start, val, found, from_this_trans, |
210 | btrfs_header_level(buf)); | |
19c00ddc CM |
211 | return 1; |
212 | } | |
213 | } else { | |
214 | write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE); | |
215 | } | |
216 | return 0; | |
217 | } | |
218 | ||
1259ab75 CM |
219 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
220 | struct extent_buffer *eb, u64 parent_transid) | |
221 | { | |
222 | int ret; | |
223 | ||
224 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
225 | return 0; | |
226 | ||
227 | lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); | |
228 | if (extent_buffer_uptodate(io_tree, eb) && | |
229 | btrfs_header_generation(eb) == parent_transid) { | |
230 | ret = 0; | |
231 | goto out; | |
232 | } | |
233 | printk("parent transid verify failed on %llu wanted %llu found %llu\n", | |
234 | (unsigned long long)eb->start, | |
235 | (unsigned long long)parent_transid, | |
236 | (unsigned long long)btrfs_header_generation(eb)); | |
237 | ret = 1; | |
1259ab75 | 238 | clear_extent_buffer_uptodate(io_tree, eb); |
33958dc6 | 239 | out: |
1259ab75 CM |
240 | unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, |
241 | GFP_NOFS); | |
242 | return ret; | |
243 | ||
244 | } | |
245 | ||
f188591e CM |
246 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
247 | struct extent_buffer *eb, | |
ca7a79ad | 248 | u64 start, u64 parent_transid) |
f188591e CM |
249 | { |
250 | struct extent_io_tree *io_tree; | |
251 | int ret; | |
252 | int num_copies = 0; | |
253 | int mirror_num = 0; | |
254 | ||
255 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
256 | while (1) { | |
257 | ret = read_extent_buffer_pages(io_tree, eb, start, 1, | |
258 | btree_get_extent, mirror_num); | |
1259ab75 CM |
259 | if (!ret && |
260 | !verify_parent_transid(io_tree, eb, parent_transid)) | |
f188591e | 261 | return ret; |
4235298e | 262 | |
f188591e CM |
263 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
264 | eb->start, eb->len); | |
4235298e | 265 | if (num_copies == 1) |
f188591e | 266 | return ret; |
4235298e | 267 | |
f188591e | 268 | mirror_num++; |
4235298e | 269 | if (mirror_num > num_copies) |
f188591e | 270 | return ret; |
f188591e | 271 | } |
f188591e CM |
272 | return -EIO; |
273 | } | |
19c00ddc CM |
274 | |
275 | int csum_dirty_buffer(struct btrfs_root *root, struct page *page) | |
276 | { | |
d1310b2e | 277 | struct extent_io_tree *tree; |
35ebb934 | 278 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc CM |
279 | u64 found_start; |
280 | int found_level; | |
281 | unsigned long len; | |
282 | struct extent_buffer *eb; | |
f188591e CM |
283 | int ret; |
284 | ||
d1310b2e | 285 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc CM |
286 | |
287 | if (page->private == EXTENT_PAGE_PRIVATE) | |
288 | goto out; | |
289 | if (!page->private) | |
290 | goto out; | |
291 | len = page->private >> 2; | |
292 | if (len == 0) { | |
293 | WARN_ON(1); | |
294 | } | |
295 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); | |
ca7a79ad CM |
296 | ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, |
297 | btrfs_header_generation(eb)); | |
f188591e | 298 | BUG_ON(ret); |
19c00ddc CM |
299 | found_start = btrfs_header_bytenr(eb); |
300 | if (found_start != start) { | |
301 | printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n", | |
302 | start, found_start, len); | |
55c69072 CM |
303 | WARN_ON(1); |
304 | goto err; | |
305 | } | |
306 | if (eb->first_page != page) { | |
307 | printk("bad first page %lu %lu\n", eb->first_page->index, | |
308 | page->index); | |
309 | WARN_ON(1); | |
310 | goto err; | |
311 | } | |
312 | if (!PageUptodate(page)) { | |
313 | printk("csum not up to date page %lu\n", page->index); | |
314 | WARN_ON(1); | |
315 | goto err; | |
19c00ddc CM |
316 | } |
317 | found_level = btrfs_header_level(eb); | |
63b10fc4 CM |
318 | spin_lock(&root->fs_info->hash_lock); |
319 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); | |
320 | spin_unlock(&root->fs_info->hash_lock); | |
19c00ddc | 321 | csum_tree_block(root, eb, 0); |
55c69072 | 322 | err: |
19c00ddc CM |
323 | free_extent_buffer(eb); |
324 | out: | |
325 | return 0; | |
326 | } | |
327 | ||
0da5468f | 328 | static int btree_writepage_io_hook(struct page *page, u64 start, u64 end) |
d98237b3 | 329 | { |
19c00ddc | 330 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
19c00ddc CM |
331 | |
332 | csum_dirty_buffer(root, page); | |
0da5468f CM |
333 | return 0; |
334 | } | |
335 | ||
ce9adaa5 CM |
336 | int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
337 | struct extent_state *state) | |
338 | { | |
339 | struct extent_io_tree *tree; | |
340 | u64 found_start; | |
341 | int found_level; | |
342 | unsigned long len; | |
343 | struct extent_buffer *eb; | |
344 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 345 | int ret = 0; |
ce9adaa5 CM |
346 | |
347 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
348 | if (page->private == EXTENT_PAGE_PRIVATE) | |
349 | goto out; | |
350 | if (!page->private) | |
351 | goto out; | |
352 | len = page->private >> 2; | |
353 | if (len == 0) { | |
354 | WARN_ON(1); | |
355 | } | |
356 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); | |
f188591e | 357 | |
ce9adaa5 CM |
358 | found_start = btrfs_header_bytenr(eb); |
359 | if (found_start != start) { | |
f188591e | 360 | ret = -EIO; |
ce9adaa5 CM |
361 | goto err; |
362 | } | |
363 | if (eb->first_page != page) { | |
364 | printk("bad first page %lu %lu\n", eb->first_page->index, | |
365 | page->index); | |
366 | WARN_ON(1); | |
f188591e | 367 | ret = -EIO; |
ce9adaa5 CM |
368 | goto err; |
369 | } | |
1259ab75 CM |
370 | if (memcmp_extent_buffer(eb, root->fs_info->fsid, |
371 | (unsigned long)btrfs_header_fsid(eb), | |
372 | BTRFS_FSID_SIZE)) { | |
373 | printk("bad fsid on block %Lu\n", eb->start); | |
374 | ret = -EIO; | |
375 | goto err; | |
376 | } | |
ce9adaa5 CM |
377 | found_level = btrfs_header_level(eb); |
378 | ||
379 | ret = csum_tree_block(root, eb, 1); | |
f188591e CM |
380 | if (ret) |
381 | ret = -EIO; | |
ce9adaa5 CM |
382 | |
383 | end = min_t(u64, eb->len, PAGE_CACHE_SIZE); | |
384 | end = eb->start + end - 1; | |
ce9adaa5 CM |
385 | err: |
386 | free_extent_buffer(eb); | |
387 | out: | |
f188591e | 388 | return ret; |
ce9adaa5 CM |
389 | } |
390 | ||
391 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) | |
392 | static void end_workqueue_bio(struct bio *bio, int err) | |
393 | #else | |
394 | static int end_workqueue_bio(struct bio *bio, | |
395 | unsigned int bytes_done, int err) | |
396 | #endif | |
397 | { | |
398 | struct end_io_wq *end_io_wq = bio->bi_private; | |
399 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 CM |
400 | |
401 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
402 | if (bio->bi_size) | |
403 | return 1; | |
404 | #endif | |
405 | ||
406 | fs_info = end_io_wq->info; | |
ce9adaa5 | 407 | end_io_wq->error = err; |
8b712842 CM |
408 | end_io_wq->work.func = end_workqueue_fn; |
409 | end_io_wq->work.flags = 0; | |
e6dcd2dc CM |
410 | if (bio->bi_rw & (1 << BIO_RW)) |
411 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
412 | &end_io_wq->work); | |
413 | else | |
414 | btrfs_queue_worker(&fs_info->endio_workers, &end_io_wq->work); | |
ce9adaa5 CM |
415 | |
416 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
417 | return 0; | |
418 | #endif | |
419 | } | |
420 | ||
22c59948 CM |
421 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
422 | int metadata) | |
0b86a832 | 423 | { |
ce9adaa5 | 424 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
425 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
426 | if (!end_io_wq) | |
427 | return -ENOMEM; | |
428 | ||
429 | end_io_wq->private = bio->bi_private; | |
430 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 431 | end_io_wq->info = info; |
ce9adaa5 CM |
432 | end_io_wq->error = 0; |
433 | end_io_wq->bio = bio; | |
22c59948 | 434 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
435 | |
436 | bio->bi_private = end_io_wq; | |
437 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
438 | return 0; |
439 | } | |
440 | ||
8b712842 CM |
441 | static void run_one_async_submit(struct btrfs_work *work) |
442 | { | |
443 | struct btrfs_fs_info *fs_info; | |
444 | struct async_submit_bio *async; | |
445 | ||
446 | async = container_of(work, struct async_submit_bio, work); | |
447 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
448 | atomic_dec(&fs_info->nr_async_submits); | |
449 | async->submit_bio_hook(async->inode, async->rw, async->bio, | |
450 | async->mirror_num); | |
451 | kfree(async); | |
452 | } | |
453 | ||
44b8bd7e CM |
454 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
455 | int rw, struct bio *bio, int mirror_num, | |
456 | extent_submit_bio_hook_t *submit_bio_hook) | |
457 | { | |
458 | struct async_submit_bio *async; | |
459 | ||
460 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
461 | if (!async) | |
462 | return -ENOMEM; | |
463 | ||
464 | async->inode = inode; | |
465 | async->rw = rw; | |
466 | async->bio = bio; | |
467 | async->mirror_num = mirror_num; | |
468 | async->submit_bio_hook = submit_bio_hook; | |
8b712842 CM |
469 | async->work.func = run_one_async_submit; |
470 | async->work.flags = 0; | |
cb03c743 | 471 | atomic_inc(&fs_info->nr_async_submits); |
8b712842 | 472 | btrfs_queue_worker(&fs_info->workers, &async->work); |
44b8bd7e CM |
473 | return 0; |
474 | } | |
475 | ||
476 | static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, | |
f188591e | 477 | int mirror_num) |
22c59948 CM |
478 | { |
479 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
480 | u64 offset; | |
481 | int ret; | |
482 | ||
483 | offset = bio->bi_sector << 9; | |
484 | ||
8b712842 CM |
485 | /* |
486 | * when we're called for a write, we're already in the async | |
487 | * submission context. Just jump ingo btrfs_map_bio | |
488 | */ | |
22c59948 | 489 | if (rw & (1 << BIO_RW)) { |
8b712842 CM |
490 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, |
491 | mirror_num, 0); | |
22c59948 CM |
492 | } |
493 | ||
8b712842 CM |
494 | /* |
495 | * called for a read, do the setup so that checksum validation | |
496 | * can happen in the async kernel threads | |
497 | */ | |
22c59948 CM |
498 | ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1); |
499 | BUG_ON(ret); | |
ce9adaa5 | 500 | |
8b712842 | 501 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
502 | } |
503 | ||
44b8bd7e CM |
504 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
505 | int mirror_num) | |
506 | { | |
8b712842 CM |
507 | /* |
508 | * kthread helpers are used to submit writes so that checksumming | |
509 | * can happen in parallel across all CPUs | |
510 | */ | |
44b8bd7e CM |
511 | if (!(rw & (1 << BIO_RW))) { |
512 | return __btree_submit_bio_hook(inode, rw, bio, mirror_num); | |
513 | } | |
514 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, | |
515 | inode, rw, bio, mirror_num, | |
516 | __btree_submit_bio_hook); | |
517 | } | |
518 | ||
0da5468f CM |
519 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
520 | { | |
d1310b2e CM |
521 | struct extent_io_tree *tree; |
522 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
523 | return extent_write_full_page(tree, page, btree_get_extent, wbc); |
524 | } | |
0da5468f CM |
525 | |
526 | static int btree_writepages(struct address_space *mapping, | |
527 | struct writeback_control *wbc) | |
528 | { | |
d1310b2e CM |
529 | struct extent_io_tree *tree; |
530 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 531 | if (wbc->sync_mode == WB_SYNC_NONE) { |
793955bc CM |
532 | u64 num_dirty; |
533 | u64 start = 0; | |
534 | unsigned long thresh = 96 * 1024 * 1024; | |
448d640b CM |
535 | |
536 | if (wbc->for_kupdate) | |
537 | return 0; | |
538 | ||
ca664626 CM |
539 | if (current_is_pdflush()) { |
540 | thresh = 96 * 1024 * 1024; | |
541 | } else { | |
542 | thresh = 8 * 1024 * 1024; | |
543 | } | |
1832a6d5 CM |
544 | num_dirty = count_range_bits(tree, &start, (u64)-1, |
545 | thresh, EXTENT_DIRTY); | |
793955bc CM |
546 | if (num_dirty < thresh) { |
547 | return 0; | |
548 | } | |
549 | } | |
0da5468f CM |
550 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
551 | } | |
552 | ||
5f39d397 CM |
553 | int btree_readpage(struct file *file, struct page *page) |
554 | { | |
d1310b2e CM |
555 | struct extent_io_tree *tree; |
556 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
557 | return extent_read_full_page(tree, page, btree_get_extent); |
558 | } | |
22b0ebda | 559 | |
70dec807 | 560 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 561 | { |
d1310b2e CM |
562 | struct extent_io_tree *tree; |
563 | struct extent_map_tree *map; | |
5f39d397 | 564 | int ret; |
d98237b3 | 565 | |
d1310b2e CM |
566 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
567 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
6af118ce | 568 | |
7b13b7b1 | 569 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
6af118ce CM |
570 | if (!ret) { |
571 | return 0; | |
572 | } | |
573 | ||
574 | ret = try_release_extent_buffer(tree, page); | |
5f39d397 CM |
575 | if (ret == 1) { |
576 | ClearPagePrivate(page); | |
577 | set_page_private(page, 0); | |
578 | page_cache_release(page); | |
579 | } | |
6af118ce | 580 | |
d98237b3 CM |
581 | return ret; |
582 | } | |
583 | ||
5f39d397 | 584 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 585 | { |
d1310b2e CM |
586 | struct extent_io_tree *tree; |
587 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
588 | extent_invalidatepage(tree, page, offset); |
589 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 590 | if (PagePrivate(page)) { |
6af118ce CM |
591 | printk("warning page private not zero on page %Lu\n", |
592 | page_offset(page)); | |
9ad6b7bc CM |
593 | ClearPagePrivate(page); |
594 | set_page_private(page, 0); | |
595 | page_cache_release(page); | |
596 | } | |
d98237b3 CM |
597 | } |
598 | ||
5f39d397 | 599 | #if 0 |
d98237b3 | 600 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
ed2ff2cb | 601 | { |
87cbda5c | 602 | struct buffer_head *bh; |
0f7d52f4 | 603 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
87cbda5c | 604 | struct buffer_head *head; |
87cbda5c CM |
605 | if (!page_has_buffers(page)) { |
606 | create_empty_buffers(page, root->fs_info->sb->s_blocksize, | |
607 | (1 << BH_Dirty)|(1 << BH_Uptodate)); | |
608 | } | |
609 | head = page_buffers(page); | |
610 | bh = head; | |
611 | do { | |
612 | if (buffer_dirty(bh)) | |
613 | csum_tree_block(root, bh, 0); | |
614 | bh = bh->b_this_page; | |
615 | } while (bh != head); | |
d98237b3 | 616 | return block_write_full_page(page, btree_get_block, wbc); |
ed2ff2cb | 617 | } |
5f39d397 | 618 | #endif |
eb60ceac | 619 | |
d98237b3 CM |
620 | static struct address_space_operations btree_aops = { |
621 | .readpage = btree_readpage, | |
622 | .writepage = btree_writepage, | |
0da5468f | 623 | .writepages = btree_writepages, |
5f39d397 CM |
624 | .releasepage = btree_releasepage, |
625 | .invalidatepage = btree_invalidatepage, | |
d98237b3 CM |
626 | .sync_page = block_sync_page, |
627 | }; | |
628 | ||
ca7a79ad CM |
629 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
630 | u64 parent_transid) | |
090d1875 | 631 | { |
5f39d397 CM |
632 | struct extent_buffer *buf = NULL; |
633 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 634 | int ret = 0; |
090d1875 | 635 | |
db94535d | 636 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 637 | if (!buf) |
090d1875 | 638 | return 0; |
d1310b2e | 639 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
f188591e | 640 | buf, 0, 0, btree_get_extent, 0); |
5f39d397 | 641 | free_extent_buffer(buf); |
de428b63 | 642 | return ret; |
090d1875 CM |
643 | } |
644 | ||
0999df54 CM |
645 | struct extent_buffer *btrfs_find_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 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
651 | bytenr, blocksize, GFP_NOFS); | |
652 | return eb; | |
653 | } | |
654 | ||
655 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
656 | u64 bytenr, u32 blocksize) | |
657 | { | |
658 | struct inode *btree_inode = root->fs_info->btree_inode; | |
659 | struct extent_buffer *eb; | |
660 | ||
661 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
662 | bytenr, blocksize, NULL, GFP_NOFS); | |
663 | return eb; | |
664 | } | |
665 | ||
666 | ||
667 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, | |
ca7a79ad | 668 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
669 | { |
670 | struct extent_buffer *buf = NULL; | |
671 | struct inode *btree_inode = root->fs_info->btree_inode; | |
672 | struct extent_io_tree *io_tree; | |
673 | int ret; | |
674 | ||
675 | io_tree = &BTRFS_I(btree_inode)->io_tree; | |
676 | ||
677 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
678 | if (!buf) | |
679 | return NULL; | |
0999df54 | 680 | |
ca7a79ad | 681 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 CM |
682 | |
683 | if (ret == 0) { | |
684 | buf->flags |= EXTENT_UPTODATE; | |
685 | } | |
5f39d397 | 686 | return buf; |
ce9adaa5 | 687 | |
eb60ceac CM |
688 | } |
689 | ||
e089f05c | 690 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 691 | struct extent_buffer *buf) |
ed2ff2cb | 692 | { |
5f39d397 | 693 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 | 694 | if (btrfs_header_generation(buf) == |
925baedd CM |
695 | root->fs_info->running_transaction->transid) { |
696 | WARN_ON(!btrfs_tree_locked(buf)); | |
d1310b2e | 697 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 698 | buf); |
925baedd | 699 | } |
5f39d397 CM |
700 | return 0; |
701 | } | |
702 | ||
703 | int wait_on_tree_block_writeback(struct btrfs_root *root, | |
704 | struct extent_buffer *buf) | |
705 | { | |
706 | struct inode *btree_inode = root->fs_info->btree_inode; | |
d1310b2e | 707 | wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
708 | buf); |
709 | return 0; | |
710 | } | |
711 | ||
db94535d | 712 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 713 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 714 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 715 | u64 objectid) |
d97e63b6 | 716 | { |
cfaa7295 | 717 | root->node = NULL; |
0f7d52f4 | 718 | root->inode = NULL; |
a28ec197 | 719 | root->commit_root = NULL; |
31153d81 | 720 | root->ref_tree = NULL; |
db94535d CM |
721 | root->sectorsize = sectorsize; |
722 | root->nodesize = nodesize; | |
723 | root->leafsize = leafsize; | |
87ee04eb | 724 | root->stripesize = stripesize; |
123abc88 | 725 | root->ref_cows = 0; |
0b86a832 CM |
726 | root->track_dirty = 0; |
727 | ||
9f5fae2f | 728 | root->fs_info = fs_info; |
0f7d52f4 CM |
729 | root->objectid = objectid; |
730 | root->last_trans = 0; | |
1b05da2e CM |
731 | root->highest_inode = 0; |
732 | root->last_inode_alloc = 0; | |
58176a96 | 733 | root->name = NULL; |
4313b399 | 734 | root->in_sysfs = 0; |
0b86a832 CM |
735 | |
736 | INIT_LIST_HEAD(&root->dirty_list); | |
7b128766 | 737 | INIT_LIST_HEAD(&root->orphan_list); |
bcc63abb | 738 | INIT_LIST_HEAD(&root->dead_list); |
925baedd | 739 | spin_lock_init(&root->node_lock); |
bcc63abb | 740 | spin_lock_init(&root->list_lock); |
a2135011 | 741 | mutex_init(&root->objectid_mutex); |
017e5369 CM |
742 | |
743 | btrfs_leaf_ref_tree_init(&root->ref_tree_struct); | |
744 | root->ref_tree = &root->ref_tree_struct; | |
745 | ||
3768f368 CM |
746 | memset(&root->root_key, 0, sizeof(root->root_key)); |
747 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 748 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 749 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 750 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 751 | init_completion(&root->kobj_unregister); |
6702ed49 CM |
752 | root->defrag_running = 0; |
753 | root->defrag_level = 0; | |
4d775673 | 754 | root->root_key.objectid = objectid; |
3768f368 CM |
755 | return 0; |
756 | } | |
757 | ||
db94535d | 758 | static int find_and_setup_root(struct btrfs_root *tree_root, |
9f5fae2f CM |
759 | struct btrfs_fs_info *fs_info, |
760 | u64 objectid, | |
e20d96d6 | 761 | struct btrfs_root *root) |
3768f368 CM |
762 | { |
763 | int ret; | |
db94535d | 764 | u32 blocksize; |
3768f368 | 765 | |
db94535d | 766 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
767 | tree_root->sectorsize, tree_root->stripesize, |
768 | root, fs_info, objectid); | |
3768f368 CM |
769 | ret = btrfs_find_last_root(tree_root, objectid, |
770 | &root->root_item, &root->root_key); | |
771 | BUG_ON(ret); | |
772 | ||
db94535d CM |
773 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
774 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
ca7a79ad | 775 | blocksize, 0); |
3768f368 | 776 | BUG_ON(!root->node); |
d97e63b6 CM |
777 | return 0; |
778 | } | |
779 | ||
5eda7b5e CM |
780 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info, |
781 | struct btrfs_key *location) | |
0f7d52f4 CM |
782 | { |
783 | struct btrfs_root *root; | |
784 | struct btrfs_root *tree_root = fs_info->tree_root; | |
785 | struct btrfs_path *path; | |
5f39d397 | 786 | struct extent_buffer *l; |
1b05da2e | 787 | u64 highest_inode; |
db94535d | 788 | u32 blocksize; |
0f7d52f4 CM |
789 | int ret = 0; |
790 | ||
5eda7b5e | 791 | root = kzalloc(sizeof(*root), GFP_NOFS); |
0cf6c620 | 792 | if (!root) |
0f7d52f4 | 793 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 794 | if (location->offset == (u64)-1) { |
db94535d | 795 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
796 | location->objectid, root); |
797 | if (ret) { | |
0f7d52f4 CM |
798 | kfree(root); |
799 | return ERR_PTR(ret); | |
800 | } | |
801 | goto insert; | |
802 | } | |
803 | ||
db94535d | 804 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
805 | tree_root->sectorsize, tree_root->stripesize, |
806 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
807 | |
808 | path = btrfs_alloc_path(); | |
809 | BUG_ON(!path); | |
810 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); | |
811 | if (ret != 0) { | |
0f7d52f4 CM |
812 | if (ret > 0) |
813 | ret = -ENOENT; | |
814 | goto out; | |
815 | } | |
5f39d397 CM |
816 | l = path->nodes[0]; |
817 | read_extent_buffer(l, &root->root_item, | |
818 | btrfs_item_ptr_offset(l, path->slots[0]), | |
0f7d52f4 | 819 | sizeof(root->root_item)); |
44b36eb2 | 820 | memcpy(&root->root_key, location, sizeof(*location)); |
0f7d52f4 CM |
821 | ret = 0; |
822 | out: | |
823 | btrfs_release_path(root, path); | |
824 | btrfs_free_path(path); | |
825 | if (ret) { | |
826 | kfree(root); | |
827 | return ERR_PTR(ret); | |
828 | } | |
db94535d CM |
829 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
830 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
ca7a79ad | 831 | blocksize, 0); |
0f7d52f4 CM |
832 | BUG_ON(!root->node); |
833 | insert: | |
0f7d52f4 | 834 | root->ref_cows = 1; |
5eda7b5e CM |
835 | ret = btrfs_find_highest_inode(root, &highest_inode); |
836 | if (ret == 0) { | |
837 | root->highest_inode = highest_inode; | |
838 | root->last_inode_alloc = highest_inode; | |
839 | } | |
840 | return root; | |
841 | } | |
842 | ||
dc17ff8f CM |
843 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
844 | u64 root_objectid) | |
845 | { | |
846 | struct btrfs_root *root; | |
847 | ||
848 | if (root_objectid == BTRFS_ROOT_TREE_OBJECTID) | |
849 | return fs_info->tree_root; | |
850 | if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
851 | return fs_info->extent_root; | |
852 | ||
853 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
854 | (unsigned long)root_objectid); | |
855 | return root; | |
856 | } | |
857 | ||
edbd8d4e CM |
858 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
859 | struct btrfs_key *location) | |
5eda7b5e CM |
860 | { |
861 | struct btrfs_root *root; | |
862 | int ret; | |
863 | ||
edbd8d4e CM |
864 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
865 | return fs_info->tree_root; | |
866 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
867 | return fs_info->extent_root; | |
8f18cf13 CM |
868 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
869 | return fs_info->chunk_root; | |
870 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
871 | return fs_info->dev_root; | |
edbd8d4e | 872 | |
5eda7b5e CM |
873 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
874 | (unsigned long)location->objectid); | |
875 | if (root) | |
876 | return root; | |
877 | ||
878 | root = btrfs_read_fs_root_no_radix(fs_info, location); | |
879 | if (IS_ERR(root)) | |
880 | return root; | |
2619ba1f CM |
881 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
882 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 CM |
883 | root); |
884 | if (ret) { | |
5f39d397 | 885 | free_extent_buffer(root->node); |
0f7d52f4 CM |
886 | kfree(root); |
887 | return ERR_PTR(ret); | |
888 | } | |
edbd8d4e CM |
889 | ret = btrfs_find_dead_roots(fs_info->tree_root, |
890 | root->root_key.objectid, root); | |
891 | BUG_ON(ret); | |
892 | ||
893 | return root; | |
894 | } | |
895 | ||
896 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, | |
897 | struct btrfs_key *location, | |
898 | const char *name, int namelen) | |
899 | { | |
900 | struct btrfs_root *root; | |
901 | int ret; | |
902 | ||
903 | root = btrfs_read_fs_root_no_name(fs_info, location); | |
904 | if (!root) | |
905 | return NULL; | |
58176a96 | 906 | |
4313b399 CM |
907 | if (root->in_sysfs) |
908 | return root; | |
909 | ||
58176a96 JB |
910 | ret = btrfs_set_root_name(root, name, namelen); |
911 | if (ret) { | |
5f39d397 | 912 | free_extent_buffer(root->node); |
58176a96 JB |
913 | kfree(root); |
914 | return ERR_PTR(ret); | |
915 | } | |
916 | ||
917 | ret = btrfs_sysfs_add_root(root); | |
918 | if (ret) { | |
5f39d397 | 919 | free_extent_buffer(root->node); |
58176a96 JB |
920 | kfree(root->name); |
921 | kfree(root); | |
922 | return ERR_PTR(ret); | |
923 | } | |
4313b399 | 924 | root->in_sysfs = 1; |
0f7d52f4 CM |
925 | return root; |
926 | } | |
19c00ddc CM |
927 | #if 0 |
928 | static int add_hasher(struct btrfs_fs_info *info, char *type) { | |
929 | struct btrfs_hasher *hasher; | |
930 | ||
931 | hasher = kmalloc(sizeof(*hasher), GFP_NOFS); | |
932 | if (!hasher) | |
933 | return -ENOMEM; | |
934 | hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC); | |
935 | if (!hasher->hash_tfm) { | |
936 | kfree(hasher); | |
937 | return -EINVAL; | |
938 | } | |
939 | spin_lock(&info->hash_lock); | |
940 | list_add(&hasher->list, &info->hashers); | |
941 | spin_unlock(&info->hash_lock); | |
942 | return 0; | |
943 | } | |
944 | #endif | |
04160088 CM |
945 | |
946 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) | |
947 | { | |
948 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
949 | int ret = 0; | |
cb03c743 | 950 | int limit = 256 * info->fs_devices->open_devices; |
04160088 CM |
951 | struct list_head *cur; |
952 | struct btrfs_device *device; | |
953 | struct backing_dev_info *bdi; | |
954 | ||
cb03c743 CM |
955 | if ((bdi_bits & (1 << BDI_write_congested)) && |
956 | atomic_read(&info->nr_async_submits) > limit) { | |
957 | return 1; | |
958 | } | |
959 | ||
04160088 CM |
960 | list_for_each(cur, &info->fs_devices->devices) { |
961 | device = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
962 | if (!device->bdev) |
963 | continue; | |
04160088 CM |
964 | bdi = blk_get_backing_dev_info(device->bdev); |
965 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
966 | ret = 1; | |
967 | break; | |
968 | } | |
969 | } | |
970 | return ret; | |
971 | } | |
972 | ||
38b66988 CM |
973 | /* |
974 | * this unplugs every device on the box, and it is only used when page | |
975 | * is null | |
976 | */ | |
977 | static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) | |
978 | { | |
979 | struct list_head *cur; | |
980 | struct btrfs_device *device; | |
981 | struct btrfs_fs_info *info; | |
982 | ||
983 | info = (struct btrfs_fs_info *)bdi->unplug_io_data; | |
984 | list_for_each(cur, &info->fs_devices->devices) { | |
985 | device = list_entry(cur, struct btrfs_device, dev_list); | |
986 | bdi = blk_get_backing_dev_info(device->bdev); | |
987 | if (bdi->unplug_io_fn) { | |
988 | bdi->unplug_io_fn(bdi, page); | |
989 | } | |
990 | } | |
991 | } | |
992 | ||
04160088 CM |
993 | void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) |
994 | { | |
38b66988 | 995 | struct inode *inode; |
f2d8d74d CM |
996 | struct extent_map_tree *em_tree; |
997 | struct extent_map *em; | |
bcbfce8a | 998 | struct address_space *mapping; |
38b66988 CM |
999 | u64 offset; |
1000 | ||
bcbfce8a | 1001 | /* the generic O_DIRECT read code does this */ |
38b66988 CM |
1002 | if (!page) { |
1003 | __unplug_io_fn(bdi, page); | |
1004 | return; | |
1005 | } | |
1006 | ||
bcbfce8a CM |
1007 | /* |
1008 | * page->mapping may change at any time. Get a consistent copy | |
1009 | * and use that for everything below | |
1010 | */ | |
1011 | smp_mb(); | |
1012 | mapping = page->mapping; | |
1013 | if (!mapping) | |
1014 | return; | |
1015 | ||
1016 | inode = mapping->host; | |
38b66988 | 1017 | offset = page_offset(page); |
04160088 | 1018 | |
f2d8d74d CM |
1019 | em_tree = &BTRFS_I(inode)->extent_tree; |
1020 | spin_lock(&em_tree->lock); | |
1021 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); | |
1022 | spin_unlock(&em_tree->lock); | |
89642229 CM |
1023 | if (!em) { |
1024 | __unplug_io_fn(bdi, page); | |
f2d8d74d | 1025 | return; |
89642229 | 1026 | } |
f2d8d74d | 1027 | |
89642229 CM |
1028 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { |
1029 | free_extent_map(em); | |
1030 | __unplug_io_fn(bdi, page); | |
1031 | return; | |
1032 | } | |
f2d8d74d CM |
1033 | offset = offset - em->start; |
1034 | btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree, | |
1035 | em->block_start + offset, page); | |
1036 | free_extent_map(em); | |
04160088 CM |
1037 | } |
1038 | ||
1039 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) | |
1040 | { | |
51ebc0d3 | 1041 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) |
04160088 | 1042 | bdi_init(bdi); |
b248a415 | 1043 | #endif |
4575c9cc | 1044 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1045 | bdi->state = 0; |
1046 | bdi->capabilities = default_backing_dev_info.capabilities; | |
1047 | bdi->unplug_io_fn = btrfs_unplug_io_fn; | |
1048 | bdi->unplug_io_data = info; | |
1049 | bdi->congested_fn = btrfs_congested_fn; | |
1050 | bdi->congested_data = info; | |
1051 | return 0; | |
1052 | } | |
1053 | ||
ce9adaa5 CM |
1054 | static int bio_ready_for_csum(struct bio *bio) |
1055 | { | |
1056 | u64 length = 0; | |
1057 | u64 buf_len = 0; | |
1058 | u64 start = 0; | |
1059 | struct page *page; | |
1060 | struct extent_io_tree *io_tree = NULL; | |
1061 | struct btrfs_fs_info *info = NULL; | |
1062 | struct bio_vec *bvec; | |
1063 | int i; | |
1064 | int ret; | |
1065 | ||
1066 | bio_for_each_segment(bvec, bio, i) { | |
1067 | page = bvec->bv_page; | |
1068 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
1069 | length += bvec->bv_len; | |
1070 | continue; | |
1071 | } | |
1072 | if (!page->private) { | |
1073 | length += bvec->bv_len; | |
1074 | continue; | |
1075 | } | |
1076 | length = bvec->bv_len; | |
1077 | buf_len = page->private >> 2; | |
1078 | start = page_offset(page) + bvec->bv_offset; | |
1079 | io_tree = &BTRFS_I(page->mapping->host)->io_tree; | |
1080 | info = BTRFS_I(page->mapping->host)->root->fs_info; | |
1081 | } | |
1082 | /* are we fully contained in this bio? */ | |
1083 | if (buf_len <= length) | |
1084 | return 1; | |
1085 | ||
1086 | ret = extent_range_uptodate(io_tree, start + length, | |
1087 | start + buf_len - 1); | |
1088 | if (ret == 1) | |
1089 | return ret; | |
1090 | return ret; | |
1091 | } | |
1092 | ||
8b712842 CM |
1093 | /* |
1094 | * called by the kthread helper functions to finally call the bio end_io | |
1095 | * functions. This is where read checksum verification actually happens | |
1096 | */ | |
1097 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1098 | { |
ce9adaa5 | 1099 | struct bio *bio; |
8b712842 CM |
1100 | struct end_io_wq *end_io_wq; |
1101 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1102 | int error; |
ce9adaa5 | 1103 | |
8b712842 CM |
1104 | end_io_wq = container_of(work, struct end_io_wq, work); |
1105 | bio = end_io_wq->bio; | |
1106 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1107 | |
8b712842 CM |
1108 | /* metadata bios are special because the whole tree block must |
1109 | * be checksummed at once. This makes sure the entire block is in | |
1110 | * ram and up to date before trying to verify things. For | |
1111 | * blocksize <= pagesize, it is basically a noop | |
1112 | */ | |
1113 | if (end_io_wq->metadata && !bio_ready_for_csum(bio)) { | |
1114 | btrfs_queue_worker(&fs_info->endio_workers, | |
1115 | &end_io_wq->work); | |
1116 | return; | |
1117 | } | |
1118 | error = end_io_wq->error; | |
1119 | bio->bi_private = end_io_wq->private; | |
1120 | bio->bi_end_io = end_io_wq->end_io; | |
1121 | kfree(end_io_wq); | |
1122 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
1123 | bio_endio(bio, bio->bi_size, error); | |
44b8bd7e | 1124 | #else |
8b712842 | 1125 | bio_endio(bio, error); |
44b8bd7e | 1126 | #endif |
44b8bd7e CM |
1127 | } |
1128 | ||
a74a4b97 CM |
1129 | static int cleaner_kthread(void *arg) |
1130 | { | |
1131 | struct btrfs_root *root = arg; | |
1132 | ||
1133 | do { | |
1134 | smp_mb(); | |
1135 | if (root->fs_info->closing) | |
1136 | break; | |
1137 | ||
1138 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1139 | mutex_lock(&root->fs_info->cleaner_mutex); | |
a74a4b97 | 1140 | btrfs_clean_old_snapshots(root); |
a74a4b97 CM |
1141 | mutex_unlock(&root->fs_info->cleaner_mutex); |
1142 | ||
1143 | if (freezing(current)) { | |
1144 | refrigerator(); | |
1145 | } else { | |
1146 | smp_mb(); | |
1147 | if (root->fs_info->closing) | |
1148 | break; | |
1149 | set_current_state(TASK_INTERRUPTIBLE); | |
1150 | schedule(); | |
1151 | __set_current_state(TASK_RUNNING); | |
1152 | } | |
1153 | } while (!kthread_should_stop()); | |
1154 | return 0; | |
1155 | } | |
1156 | ||
1157 | static int transaction_kthread(void *arg) | |
1158 | { | |
1159 | struct btrfs_root *root = arg; | |
1160 | struct btrfs_trans_handle *trans; | |
1161 | struct btrfs_transaction *cur; | |
1162 | unsigned long now; | |
1163 | unsigned long delay; | |
1164 | int ret; | |
1165 | ||
1166 | do { | |
1167 | smp_mb(); | |
1168 | if (root->fs_info->closing) | |
1169 | break; | |
1170 | ||
1171 | delay = HZ * 30; | |
1172 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1173 | mutex_lock(&root->fs_info->transaction_kthread_mutex); | |
1174 | ||
ab78c84d CM |
1175 | if (root->fs_info->total_ref_cache_size > 20 * 1024 * 1024) { |
1176 | printk("btrfs: total reference cache size %Lu\n", | |
1177 | root->fs_info->total_ref_cache_size); | |
1178 | } | |
31153d81 | 1179 | |
a74a4b97 CM |
1180 | mutex_lock(&root->fs_info->trans_mutex); |
1181 | cur = root->fs_info->running_transaction; | |
1182 | if (!cur) { | |
1183 | mutex_unlock(&root->fs_info->trans_mutex); | |
1184 | goto sleep; | |
1185 | } | |
31153d81 | 1186 | |
a74a4b97 CM |
1187 | now = get_seconds(); |
1188 | if (now < cur->start_time || now - cur->start_time < 30) { | |
1189 | mutex_unlock(&root->fs_info->trans_mutex); | |
1190 | delay = HZ * 5; | |
1191 | goto sleep; | |
1192 | } | |
1193 | mutex_unlock(&root->fs_info->trans_mutex); | |
a74a4b97 CM |
1194 | trans = btrfs_start_transaction(root, 1); |
1195 | ret = btrfs_commit_transaction(trans, root); | |
1196 | sleep: | |
1197 | wake_up_process(root->fs_info->cleaner_kthread); | |
1198 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1199 | ||
1200 | if (freezing(current)) { | |
1201 | refrigerator(); | |
1202 | } else { | |
1203 | if (root->fs_info->closing) | |
1204 | break; | |
1205 | set_current_state(TASK_INTERRUPTIBLE); | |
1206 | schedule_timeout(delay); | |
1207 | __set_current_state(TASK_RUNNING); | |
1208 | } | |
1209 | } while (!kthread_should_stop()); | |
1210 | return 0; | |
1211 | } | |
1212 | ||
8a4b83cc | 1213 | struct btrfs_root *open_ctree(struct super_block *sb, |
dfe25020 CM |
1214 | struct btrfs_fs_devices *fs_devices, |
1215 | char *options) | |
2e635a27 | 1216 | { |
db94535d CM |
1217 | u32 sectorsize; |
1218 | u32 nodesize; | |
1219 | u32 leafsize; | |
1220 | u32 blocksize; | |
87ee04eb | 1221 | u32 stripesize; |
a061fc8d | 1222 | struct buffer_head *bh; |
e20d96d6 CM |
1223 | struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root), |
1224 | GFP_NOFS); | |
1225 | struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root), | |
1226 | GFP_NOFS); | |
8790d502 | 1227 | struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), |
e20d96d6 | 1228 | GFP_NOFS); |
0b86a832 CM |
1229 | struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root), |
1230 | GFP_NOFS); | |
1231 | struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root), | |
1232 | GFP_NOFS); | |
eb60ceac | 1233 | int ret; |
e58ca020 | 1234 | int err = -EINVAL; |
4543df7e | 1235 | |
2c90e5d6 | 1236 | struct btrfs_super_block *disk_super; |
8790d502 | 1237 | |
39279cc3 CM |
1238 | if (!extent_root || !tree_root || !fs_info) { |
1239 | err = -ENOMEM; | |
1240 | goto fail; | |
1241 | } | |
0f7d52f4 | 1242 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS); |
8fd17795 | 1243 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1244 | INIT_LIST_HEAD(&fs_info->dead_roots); |
19c00ddc CM |
1245 | INIT_LIST_HEAD(&fs_info->hashers); |
1246 | spin_lock_init(&fs_info->hash_lock); | |
1832a6d5 | 1247 | spin_lock_init(&fs_info->delalloc_lock); |
cee36a03 | 1248 | spin_lock_init(&fs_info->new_trans_lock); |
31153d81 | 1249 | spin_lock_init(&fs_info->ref_cache_lock); |
19c00ddc | 1250 | |
58176a96 | 1251 | init_completion(&fs_info->kobj_unregister); |
9f5fae2f CM |
1252 | fs_info->tree_root = tree_root; |
1253 | fs_info->extent_root = extent_root; | |
0b86a832 CM |
1254 | fs_info->chunk_root = chunk_root; |
1255 | fs_info->dev_root = dev_root; | |
8a4b83cc | 1256 | fs_info->fs_devices = fs_devices; |
0b86a832 | 1257 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1258 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1259 | btrfs_mapping_init(&fs_info->mapping_tree); |
cb03c743 | 1260 | atomic_set(&fs_info->nr_async_submits, 0); |
a2135011 | 1261 | atomic_set(&fs_info->throttles, 0); |
ab78c84d | 1262 | atomic_set(&fs_info->throttle_gen, 0); |
e20d96d6 | 1263 | fs_info->sb = sb; |
c59f8951 | 1264 | fs_info->max_extent = (u64)-1; |
6f568d35 | 1265 | fs_info->max_inline = 8192 * 1024; |
04160088 | 1266 | setup_bdi(fs_info, &fs_info->bdi); |
d98237b3 CM |
1267 | fs_info->btree_inode = new_inode(sb); |
1268 | fs_info->btree_inode->i_ino = 1; | |
2c90e5d6 | 1269 | fs_info->btree_inode->i_nlink = 1; |
4543df7e | 1270 | fs_info->thread_pool_size = min(num_online_cpus() + 2, 8); |
0afbaf8c | 1271 | |
3eaa2885 CM |
1272 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
1273 | spin_lock_init(&fs_info->ordered_extent_lock); | |
1274 | ||
a061fc8d CM |
1275 | sb->s_blocksize = 4096; |
1276 | sb->s_blocksize_bits = blksize_bits(4096); | |
1277 | ||
0afbaf8c CM |
1278 | /* |
1279 | * we set the i_size on the btree inode to the max possible int. | |
1280 | * the real end of the address space is determined by all of | |
1281 | * the devices in the system | |
1282 | */ | |
1283 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 1284 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
1285 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
1286 | ||
d1310b2e | 1287 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
5f39d397 CM |
1288 | fs_info->btree_inode->i_mapping, |
1289 | GFP_NOFS); | |
d1310b2e CM |
1290 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, |
1291 | GFP_NOFS); | |
1292 | ||
1293 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 1294 | |
d1310b2e | 1295 | extent_io_tree_init(&fs_info->free_space_cache, |
f510cfec | 1296 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1297 | extent_io_tree_init(&fs_info->block_group_cache, |
96b5179d | 1298 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1299 | extent_io_tree_init(&fs_info->pinned_extents, |
1a5bc167 | 1300 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1301 | extent_io_tree_init(&fs_info->pending_del, |
1a5bc167 | 1302 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1303 | extent_io_tree_init(&fs_info->extent_ins, |
1a5bc167 | 1304 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
e66f709b | 1305 | fs_info->do_barriers = 1; |
e18e4809 | 1306 | |
0f7d52f4 CM |
1307 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
1308 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
1309 | sizeof(struct btrfs_key)); | |
22b0ebda | 1310 | insert_inode_hash(fs_info->btree_inode); |
d98237b3 | 1311 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
39279cc3 | 1312 | |
79154b1b | 1313 | mutex_init(&fs_info->trans_mutex); |
a2135011 | 1314 | mutex_init(&fs_info->drop_mutex); |
925baedd CM |
1315 | mutex_init(&fs_info->alloc_mutex); |
1316 | mutex_init(&fs_info->chunk_mutex); | |
a74a4b97 CM |
1317 | mutex_init(&fs_info->transaction_kthread_mutex); |
1318 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 1319 | mutex_init(&fs_info->volume_mutex); |
e6dcd2dc | 1320 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 1321 | init_waitqueue_head(&fs_info->transaction_wait); |
3768f368 | 1322 | |
19c00ddc CM |
1323 | #if 0 |
1324 | ret = add_hasher(fs_info, "crc32c"); | |
1325 | if (ret) { | |
1326 | printk("btrfs: failed hash setup, modprobe cryptomgr?\n"); | |
1327 | err = -ENOMEM; | |
1328 | goto fail_iput; | |
1329 | } | |
1330 | #endif | |
0b86a832 | 1331 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 1332 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 1333 | |
d98237b3 | 1334 | |
a061fc8d CM |
1335 | bh = __bread(fs_devices->latest_bdev, |
1336 | BTRFS_SUPER_INFO_OFFSET / 4096, 4096); | |
1337 | if (!bh) | |
39279cc3 | 1338 | goto fail_iput; |
39279cc3 | 1339 | |
a061fc8d CM |
1340 | memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy)); |
1341 | brelse(bh); | |
5f39d397 | 1342 | |
a061fc8d | 1343 | memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE); |
0b86a832 | 1344 | |
5f39d397 | 1345 | disk_super = &fs_info->super_copy; |
0f7d52f4 | 1346 | if (!btrfs_super_root(disk_super)) |
39279cc3 | 1347 | goto fail_sb_buffer; |
0f7d52f4 | 1348 | |
edf24abe CH |
1349 | err = btrfs_parse_options(tree_root, options); |
1350 | if (err) | |
1351 | goto fail_sb_buffer; | |
dfe25020 | 1352 | |
4543df7e CM |
1353 | /* |
1354 | * we need to start all the end_io workers up front because the | |
1355 | * queue work function gets called at interrupt time, and so it | |
1356 | * cannot dynamically grow. | |
1357 | */ | |
1358 | btrfs_init_workers(&fs_info->workers, fs_info->thread_pool_size); | |
1cc127b5 | 1359 | btrfs_init_workers(&fs_info->submit_workers, fs_info->thread_pool_size); |
61b49440 CM |
1360 | |
1361 | /* a higher idle thresh on the submit workers makes it much more | |
1362 | * likely that bios will be send down in a sane order to the | |
1363 | * devices | |
1364 | */ | |
1365 | fs_info->submit_workers.idle_thresh = 64; | |
1366 | ||
247e743c | 1367 | btrfs_init_workers(&fs_info->fixup_workers, 1); |
4543df7e | 1368 | btrfs_init_workers(&fs_info->endio_workers, fs_info->thread_pool_size); |
e6dcd2dc CM |
1369 | btrfs_init_workers(&fs_info->endio_write_workers, |
1370 | fs_info->thread_pool_size); | |
61b49440 CM |
1371 | |
1372 | /* | |
1373 | * endios are largely parallel and should have a very | |
1374 | * low idle thresh | |
1375 | */ | |
1376 | fs_info->endio_workers.idle_thresh = 4; | |
1377 | fs_info->endio_write_workers.idle_thresh = 4; | |
1378 | ||
4543df7e | 1379 | btrfs_start_workers(&fs_info->workers, 1); |
1cc127b5 | 1380 | btrfs_start_workers(&fs_info->submit_workers, 1); |
247e743c | 1381 | btrfs_start_workers(&fs_info->fixup_workers, 1); |
4543df7e | 1382 | btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size); |
e6dcd2dc CM |
1383 | btrfs_start_workers(&fs_info->endio_write_workers, |
1384 | fs_info->thread_pool_size); | |
4543df7e | 1385 | |
edf24abe | 1386 | err = -EINVAL; |
a0af469b | 1387 | if (btrfs_super_num_devices(disk_super) > fs_devices->open_devices) { |
8a4b83cc CM |
1388 | printk("Btrfs: wanted %llu devices, but found %llu\n", |
1389 | (unsigned long long)btrfs_super_num_devices(disk_super), | |
a0af469b | 1390 | (unsigned long long)fs_devices->open_devices); |
dfe25020 CM |
1391 | if (btrfs_test_opt(tree_root, DEGRADED)) |
1392 | printk("continuing in degraded mode\n"); | |
1393 | else { | |
1394 | goto fail_sb_buffer; | |
1395 | } | |
8a4b83cc | 1396 | } |
dfe25020 | 1397 | |
4575c9cc CM |
1398 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
1399 | ||
db94535d CM |
1400 | nodesize = btrfs_super_nodesize(disk_super); |
1401 | leafsize = btrfs_super_leafsize(disk_super); | |
1402 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 1403 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
1404 | tree_root->nodesize = nodesize; |
1405 | tree_root->leafsize = leafsize; | |
1406 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 1407 | tree_root->stripesize = stripesize; |
a061fc8d CM |
1408 | |
1409 | sb->s_blocksize = sectorsize; | |
1410 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 1411 | |
39279cc3 CM |
1412 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
1413 | sizeof(disk_super->magic))) { | |
1414 | printk("btrfs: valid FS not found on %s\n", sb->s_id); | |
1415 | goto fail_sb_buffer; | |
1416 | } | |
19c00ddc | 1417 | |
925baedd | 1418 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1419 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 1420 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f CM |
1421 | if (ret) { |
1422 | printk("btrfs: failed to read the system array on %s\n", | |
1423 | sb->s_id); | |
1424 | goto fail_sys_array; | |
1425 | } | |
0b86a832 CM |
1426 | |
1427 | blocksize = btrfs_level_size(tree_root, | |
1428 | btrfs_super_chunk_root_level(disk_super)); | |
1429 | ||
1430 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1431 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
1432 | ||
1433 | chunk_root->node = read_tree_block(chunk_root, | |
1434 | btrfs_super_chunk_root(disk_super), | |
ca7a79ad | 1435 | blocksize, 0); |
0b86a832 CM |
1436 | BUG_ON(!chunk_root->node); |
1437 | ||
e17cade2 CM |
1438 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
1439 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), | |
1440 | BTRFS_UUID_SIZE); | |
1441 | ||
925baedd | 1442 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1443 | ret = btrfs_read_chunk_tree(chunk_root); |
925baedd | 1444 | mutex_unlock(&fs_info->chunk_mutex); |
0b86a832 CM |
1445 | BUG_ON(ret); |
1446 | ||
dfe25020 CM |
1447 | btrfs_close_extra_devices(fs_devices); |
1448 | ||
db94535d CM |
1449 | blocksize = btrfs_level_size(tree_root, |
1450 | btrfs_super_root_level(disk_super)); | |
19c00ddc | 1451 | |
0b86a832 | 1452 | |
e20d96d6 | 1453 | tree_root->node = read_tree_block(tree_root, |
db94535d | 1454 | btrfs_super_root(disk_super), |
ca7a79ad | 1455 | blocksize, 0); |
39279cc3 CM |
1456 | if (!tree_root->node) |
1457 | goto fail_sb_buffer; | |
3768f368 | 1458 | |
db94535d CM |
1459 | |
1460 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 1461 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 1462 | if (ret) |
39279cc3 | 1463 | goto fail_tree_root; |
0b86a832 CM |
1464 | extent_root->track_dirty = 1; |
1465 | ||
1466 | ret = find_and_setup_root(tree_root, fs_info, | |
1467 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
1468 | dev_root->track_dirty = 1; | |
1469 | ||
1470 | if (ret) | |
1471 | goto fail_extent_root; | |
3768f368 | 1472 | |
9078a3e1 CM |
1473 | btrfs_read_block_groups(extent_root); |
1474 | ||
0f7d52f4 | 1475 | fs_info->generation = btrfs_super_generation(disk_super) + 1; |
d18a2c44 CM |
1476 | fs_info->data_alloc_profile = (u64)-1; |
1477 | fs_info->metadata_alloc_profile = (u64)-1; | |
1478 | fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; | |
a74a4b97 CM |
1479 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
1480 | "btrfs-cleaner"); | |
1481 | if (!fs_info->cleaner_kthread) | |
1482 | goto fail_extent_root; | |
1483 | ||
1484 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
1485 | tree_root, | |
1486 | "btrfs-transaction"); | |
1487 | if (!fs_info->transaction_kthread) | |
3f157a2f | 1488 | goto fail_cleaner; |
a74a4b97 | 1489 | |
d18a2c44 | 1490 | |
0f7d52f4 | 1491 | return tree_root; |
39279cc3 | 1492 | |
3f157a2f | 1493 | fail_cleaner: |
a74a4b97 | 1494 | kthread_stop(fs_info->cleaner_kthread); |
0b86a832 CM |
1495 | fail_extent_root: |
1496 | free_extent_buffer(extent_root->node); | |
39279cc3 | 1497 | fail_tree_root: |
5f39d397 | 1498 | free_extent_buffer(tree_root->node); |
84eed90f | 1499 | fail_sys_array: |
39279cc3 | 1500 | fail_sb_buffer: |
247e743c | 1501 | btrfs_stop_workers(&fs_info->fixup_workers); |
8b712842 CM |
1502 | btrfs_stop_workers(&fs_info->workers); |
1503 | btrfs_stop_workers(&fs_info->endio_workers); | |
e6dcd2dc | 1504 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 1505 | btrfs_stop_workers(&fs_info->submit_workers); |
4543df7e CM |
1506 | fail_iput: |
1507 | iput(fs_info->btree_inode); | |
39279cc3 | 1508 | fail: |
dfe25020 | 1509 | btrfs_close_devices(fs_info->fs_devices); |
84eed90f CM |
1510 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
1511 | ||
39279cc3 CM |
1512 | kfree(extent_root); |
1513 | kfree(tree_root); | |
51ebc0d3 | 1514 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) |
2d2ae547 | 1515 | bdi_destroy(&fs_info->bdi); |
b248a415 | 1516 | #endif |
39279cc3 CM |
1517 | kfree(fs_info); |
1518 | return ERR_PTR(err); | |
eb60ceac CM |
1519 | } |
1520 | ||
f2984462 CM |
1521 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
1522 | { | |
1523 | char b[BDEVNAME_SIZE]; | |
1524 | ||
1525 | if (uptodate) { | |
1526 | set_buffer_uptodate(bh); | |
1527 | } else { | |
1528 | if (!buffer_eopnotsupp(bh) && printk_ratelimit()) { | |
1529 | printk(KERN_WARNING "lost page write due to " | |
1530 | "I/O error on %s\n", | |
1531 | bdevname(bh->b_bdev, b)); | |
1532 | } | |
1259ab75 CM |
1533 | /* note, we dont' set_buffer_write_io_error because we have |
1534 | * our own ways of dealing with the IO errors | |
1535 | */ | |
f2984462 CM |
1536 | clear_buffer_uptodate(bh); |
1537 | } | |
1538 | unlock_buffer(bh); | |
1539 | put_bh(bh); | |
1540 | } | |
1541 | ||
1542 | int write_all_supers(struct btrfs_root *root) | |
1543 | { | |
1544 | struct list_head *cur; | |
1545 | struct list_head *head = &root->fs_info->fs_devices->devices; | |
1546 | struct btrfs_device *dev; | |
a061fc8d | 1547 | struct btrfs_super_block *sb; |
f2984462 CM |
1548 | struct btrfs_dev_item *dev_item; |
1549 | struct buffer_head *bh; | |
1550 | int ret; | |
1551 | int do_barriers; | |
a236aed1 CM |
1552 | int max_errors; |
1553 | int total_errors = 0; | |
a061fc8d CM |
1554 | u32 crc; |
1555 | u64 flags; | |
f2984462 | 1556 | |
a236aed1 | 1557 | max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1; |
f2984462 CM |
1558 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
1559 | ||
a061fc8d CM |
1560 | sb = &root->fs_info->super_for_commit; |
1561 | dev_item = &sb->dev_item; | |
f2984462 CM |
1562 | list_for_each(cur, head) { |
1563 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1564 | if (!dev->bdev) { |
1565 | total_errors++; | |
1566 | continue; | |
1567 | } | |
1568 | if (!dev->in_fs_metadata) | |
1569 | continue; | |
1570 | ||
a061fc8d CM |
1571 | btrfs_set_stack_device_type(dev_item, dev->type); |
1572 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
1573 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
1574 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
1575 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
1576 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
1577 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
1578 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
1579 | flags = btrfs_super_flags(sb); | |
1580 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
1581 | ||
1582 | ||
1583 | crc = ~(u32)0; | |
1584 | crc = btrfs_csum_data(root, (char *)sb + BTRFS_CSUM_SIZE, crc, | |
1585 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
1586 | btrfs_csum_final(crc, sb->csum); | |
1587 | ||
1588 | bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET / 4096, | |
f2984462 CM |
1589 | BTRFS_SUPER_INFO_SIZE); |
1590 | ||
a061fc8d | 1591 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
f2984462 CM |
1592 | dev->pending_io = bh; |
1593 | ||
1594 | get_bh(bh); | |
1595 | set_buffer_uptodate(bh); | |
1596 | lock_buffer(bh); | |
1597 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
1598 | ||
1599 | if (do_barriers && dev->barriers) { | |
1600 | ret = submit_bh(WRITE_BARRIER, bh); | |
1601 | if (ret == -EOPNOTSUPP) { | |
1602 | printk("btrfs: disabling barriers on dev %s\n", | |
1603 | dev->name); | |
1604 | set_buffer_uptodate(bh); | |
1605 | dev->barriers = 0; | |
1606 | get_bh(bh); | |
1607 | lock_buffer(bh); | |
1608 | ret = submit_bh(WRITE, bh); | |
1609 | } | |
1610 | } else { | |
1611 | ret = submit_bh(WRITE, bh); | |
1612 | } | |
a236aed1 CM |
1613 | if (ret) |
1614 | total_errors++; | |
f2984462 | 1615 | } |
a236aed1 CM |
1616 | if (total_errors > max_errors) { |
1617 | printk("btrfs: %d errors while writing supers\n", total_errors); | |
1618 | BUG(); | |
1619 | } | |
1620 | total_errors = 0; | |
f2984462 CM |
1621 | |
1622 | list_for_each(cur, head) { | |
1623 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1624 | if (!dev->bdev) |
1625 | continue; | |
1626 | if (!dev->in_fs_metadata) | |
1627 | continue; | |
1628 | ||
f2984462 CM |
1629 | BUG_ON(!dev->pending_io); |
1630 | bh = dev->pending_io; | |
1631 | wait_on_buffer(bh); | |
1632 | if (!buffer_uptodate(dev->pending_io)) { | |
1633 | if (do_barriers && dev->barriers) { | |
1634 | printk("btrfs: disabling barriers on dev %s\n", | |
1635 | dev->name); | |
1636 | set_buffer_uptodate(bh); | |
1637 | get_bh(bh); | |
1638 | lock_buffer(bh); | |
1639 | dev->barriers = 0; | |
1640 | ret = submit_bh(WRITE, bh); | |
1641 | BUG_ON(ret); | |
1642 | wait_on_buffer(bh); | |
1259ab75 CM |
1643 | if (!buffer_uptodate(bh)) |
1644 | total_errors++; | |
f2984462 | 1645 | } else { |
a236aed1 | 1646 | total_errors++; |
f2984462 CM |
1647 | } |
1648 | ||
1649 | } | |
1650 | dev->pending_io = NULL; | |
1651 | brelse(bh); | |
1652 | } | |
a236aed1 CM |
1653 | if (total_errors > max_errors) { |
1654 | printk("btrfs: %d errors while writing supers\n", total_errors); | |
1655 | BUG(); | |
1656 | } | |
f2984462 CM |
1657 | return 0; |
1658 | } | |
1659 | ||
e089f05c | 1660 | int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root |
79154b1b | 1661 | *root) |
eb60ceac | 1662 | { |
e66f709b | 1663 | int ret; |
5f39d397 | 1664 | |
f2984462 | 1665 | ret = write_all_supers(root); |
5f39d397 | 1666 | return ret; |
cfaa7295 CM |
1667 | } |
1668 | ||
5eda7b5e | 1669 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f CM |
1670 | { |
1671 | radix_tree_delete(&fs_info->fs_roots_radix, | |
1672 | (unsigned long)root->root_key.objectid); | |
b99aa6cb CM |
1673 | if (root->in_sysfs) |
1674 | btrfs_sysfs_del_root(root); | |
2619ba1f CM |
1675 | if (root->inode) |
1676 | iput(root->inode); | |
1677 | if (root->node) | |
5f39d397 | 1678 | free_extent_buffer(root->node); |
2619ba1f | 1679 | if (root->commit_root) |
5f39d397 | 1680 | free_extent_buffer(root->commit_root); |
58176a96 JB |
1681 | if (root->name) |
1682 | kfree(root->name); | |
2619ba1f CM |
1683 | kfree(root); |
1684 | return 0; | |
1685 | } | |
1686 | ||
35b7e476 | 1687 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
1688 | { |
1689 | int ret; | |
1690 | struct btrfs_root *gang[8]; | |
1691 | int i; | |
1692 | ||
1693 | while(1) { | |
1694 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
1695 | (void **)gang, 0, | |
1696 | ARRAY_SIZE(gang)); | |
1697 | if (!ret) | |
1698 | break; | |
2619ba1f | 1699 | for (i = 0; i < ret; i++) |
5eda7b5e | 1700 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
1701 | } |
1702 | return 0; | |
1703 | } | |
b4100d64 | 1704 | |
e20d96d6 | 1705 | int close_ctree(struct btrfs_root *root) |
cfaa7295 | 1706 | { |
3768f368 | 1707 | int ret; |
e089f05c | 1708 | struct btrfs_trans_handle *trans; |
0f7d52f4 | 1709 | struct btrfs_fs_info *fs_info = root->fs_info; |
e089f05c | 1710 | |
facda1e7 | 1711 | fs_info->closing = 1; |
a2135011 CM |
1712 | smp_mb(); |
1713 | ||
a74a4b97 CM |
1714 | kthread_stop(root->fs_info->transaction_kthread); |
1715 | kthread_stop(root->fs_info->cleaner_kthread); | |
1716 | ||
a74a4b97 | 1717 | btrfs_clean_old_snapshots(root); |
79154b1b | 1718 | trans = btrfs_start_transaction(root, 1); |
54aa1f4d | 1719 | ret = btrfs_commit_transaction(trans, root); |
79154b1b CM |
1720 | /* run commit again to drop the original snapshot */ |
1721 | trans = btrfs_start_transaction(root, 1); | |
1722 | btrfs_commit_transaction(trans, root); | |
1723 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 1724 | BUG_ON(ret); |
d6bfde87 | 1725 | |
79154b1b | 1726 | write_ctree_super(NULL, root); |
0f7d52f4 | 1727 | |
b0c68f8b CM |
1728 | if (fs_info->delalloc_bytes) { |
1729 | printk("btrfs: at unmount delalloc count %Lu\n", | |
1730 | fs_info->delalloc_bytes); | |
1731 | } | |
31153d81 YZ |
1732 | if (fs_info->total_ref_cache_size) { |
1733 | printk("btrfs: at umount reference cache size %Lu\n", | |
1734 | fs_info->total_ref_cache_size); | |
1735 | } | |
bcc63abb | 1736 | |
0f7d52f4 | 1737 | if (fs_info->extent_root->node) |
5f39d397 | 1738 | free_extent_buffer(fs_info->extent_root->node); |
f510cfec | 1739 | |
0f7d52f4 | 1740 | if (fs_info->tree_root->node) |
5f39d397 | 1741 | free_extent_buffer(fs_info->tree_root->node); |
f510cfec | 1742 | |
0b86a832 CM |
1743 | if (root->fs_info->chunk_root->node); |
1744 | free_extent_buffer(root->fs_info->chunk_root->node); | |
1745 | ||
1746 | if (root->fs_info->dev_root->node); | |
1747 | free_extent_buffer(root->fs_info->dev_root->node); | |
1748 | ||
9078a3e1 | 1749 | btrfs_free_block_groups(root->fs_info); |
0f7d52f4 | 1750 | del_fs_roots(fs_info); |
d10c5f31 CM |
1751 | |
1752 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
1753 | ||
db94535d | 1754 | truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); |
9ad6b7bc | 1755 | |
247e743c | 1756 | btrfs_stop_workers(&fs_info->fixup_workers); |
8b712842 CM |
1757 | btrfs_stop_workers(&fs_info->workers); |
1758 | btrfs_stop_workers(&fs_info->endio_workers); | |
e6dcd2dc | 1759 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 1760 | btrfs_stop_workers(&fs_info->submit_workers); |
d6bfde87 | 1761 | |
db94535d | 1762 | iput(fs_info->btree_inode); |
19c00ddc CM |
1763 | #if 0 |
1764 | while(!list_empty(&fs_info->hashers)) { | |
1765 | struct btrfs_hasher *hasher; | |
1766 | hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher, | |
1767 | hashers); | |
1768 | list_del(&hasher->hashers); | |
1769 | crypto_free_hash(&fs_info->hash_tfm); | |
1770 | kfree(hasher); | |
1771 | } | |
1772 | #endif | |
dfe25020 | 1773 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 1774 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 1775 | |
51ebc0d3 | 1776 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) |
04160088 | 1777 | bdi_destroy(&fs_info->bdi); |
b248a415 | 1778 | #endif |
0b86a832 | 1779 | |
0f7d52f4 | 1780 | kfree(fs_info->extent_root); |
0f7d52f4 | 1781 | kfree(fs_info->tree_root); |
0b86a832 CM |
1782 | kfree(fs_info->chunk_root); |
1783 | kfree(fs_info->dev_root); | |
eb60ceac CM |
1784 | return 0; |
1785 | } | |
1786 | ||
1259ab75 | 1787 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 1788 | { |
1259ab75 | 1789 | int ret; |
810191ff | 1790 | struct inode *btree_inode = buf->first_page->mapping->host; |
1259ab75 CM |
1791 | |
1792 | ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); | |
1793 | if (!ret) | |
1794 | return ret; | |
1795 | ||
1796 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
1797 | parent_transid); | |
1798 | return !ret; | |
5f39d397 CM |
1799 | } |
1800 | ||
1801 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 1802 | { |
810191ff | 1803 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 1804 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
1805 | buf); |
1806 | } | |
6702ed49 | 1807 | |
5f39d397 CM |
1808 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
1809 | { | |
810191ff | 1810 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
5f39d397 CM |
1811 | u64 transid = btrfs_header_generation(buf); |
1812 | struct inode *btree_inode = root->fs_info->btree_inode; | |
6702ed49 | 1813 | |
925baedd | 1814 | WARN_ON(!btrfs_tree_locked(buf)); |
ccd467d6 CM |
1815 | if (transid != root->fs_info->generation) { |
1816 | printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n", | |
db94535d | 1817 | (unsigned long long)buf->start, |
ccd467d6 CM |
1818 | transid, root->fs_info->generation); |
1819 | WARN_ON(1); | |
1820 | } | |
d1310b2e | 1821 | set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf); |
eb60ceac CM |
1822 | } |
1823 | ||
d3c2fdcf | 1824 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
35b7e476 | 1825 | { |
188de649 CM |
1826 | /* |
1827 | * looks as though older kernels can get into trouble with | |
1828 | * this code, they end up stuck in balance_dirty_pages forever | |
1829 | */ | |
d6bfde87 CM |
1830 | struct extent_io_tree *tree; |
1831 | u64 num_dirty; | |
1832 | u64 start = 0; | |
1833 | unsigned long thresh = 16 * 1024 * 1024; | |
1834 | tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
1835 | ||
1836 | if (current_is_pdflush()) | |
1837 | return; | |
1838 | ||
1839 | num_dirty = count_range_bits(tree, &start, (u64)-1, | |
1840 | thresh, EXTENT_DIRTY); | |
1841 | if (num_dirty > thresh) { | |
1842 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 1843 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 1844 | } |
188de649 | 1845 | return; |
35b7e476 | 1846 | } |
6b80053d | 1847 | |
ca7a79ad | 1848 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 1849 | { |
810191ff | 1850 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
ce9adaa5 | 1851 | int ret; |
ca7a79ad | 1852 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 CM |
1853 | if (ret == 0) { |
1854 | buf->flags |= EXTENT_UPTODATE; | |
1855 | } | |
1856 | return ret; | |
6b80053d | 1857 | } |
0da5468f | 1858 | |
d1310b2e | 1859 | static struct extent_io_ops btree_extent_io_ops = { |
0da5468f | 1860 | .writepage_io_hook = btree_writepage_io_hook, |
ce9adaa5 | 1861 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
0b86a832 | 1862 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
1863 | /* note we're sharing with inode.c for the merge bio hook */ |
1864 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 1865 | }; |