2 * Copyright (C) 2007 Oracle. All rights reserved.
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.
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.
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.
19 #include <linux/version.h>
21 #include <linux/blkdev.h>
22 #include <linux/scatterlist.h>
23 #include <linux/swap.h>
24 #include <linux/radix-tree.h>
25 #include <linux/writeback.h>
26 #include <linux/buffer_head.h> // for block_sync_page
27 #include <linux/workqueue.h>
28 #include <linux/kthread.h>
29 #include <linux/freezer.h>
34 #include "transaction.h"
35 #include "btrfs_inode.h"
37 #include "print-tree.h"
38 #include "async-thread.h"
40 #include "ref-cache.h"
44 static int check_tree_block(struct btrfs_root
*root
, struct extent_buffer
*buf
)
46 if (extent_buffer_blocknr(buf
) != btrfs_header_blocknr(buf
)) {
47 printk(KERN_CRIT
"buf blocknr(buf) is %llu, header is %llu\n",
48 (unsigned long long)extent_buffer_blocknr(buf
),
49 (unsigned long long)btrfs_header_blocknr(buf
));
56 static struct extent_io_ops btree_extent_io_ops
;
57 static void end_workqueue_fn(struct btrfs_work
*work
);
60 * end_io_wq structs are used to do processing in task context when an IO is
61 * complete. This is used during reads to verify checksums, and it is used
62 * by writes to insert metadata for new file extents after IO is complete.
68 struct btrfs_fs_info
*info
;
71 struct list_head list
;
72 struct btrfs_work work
;
76 * async submit bios are used to offload expensive checksumming
77 * onto the worker threads. They checksum file and metadata bios
78 * just before they are sent down the IO stack.
80 struct async_submit_bio
{
83 struct list_head list
;
84 extent_submit_bio_hook_t
*submit_bio_start
;
85 extent_submit_bio_hook_t
*submit_bio_done
;
88 unsigned long bio_flags
;
89 struct btrfs_work work
;
93 * extents on the btree inode are pretty simple, there's one extent
94 * that covers the entire device
96 static struct extent_map
*btree_get_extent(struct inode
*inode
,
97 struct page
*page
, size_t page_offset
, u64 start
, u64 len
,
100 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
101 struct extent_map
*em
;
104 spin_lock(&em_tree
->lock
);
105 em
= lookup_extent_mapping(em_tree
, start
, len
);
108 BTRFS_I(inode
)->root
->fs_info
->fs_devices
->latest_bdev
;
109 spin_unlock(&em_tree
->lock
);
112 spin_unlock(&em_tree
->lock
);
114 em
= alloc_extent_map(GFP_NOFS
);
116 em
= ERR_PTR(-ENOMEM
);
121 em
->block_len
= (u64
)-1;
123 em
->bdev
= BTRFS_I(inode
)->root
->fs_info
->fs_devices
->latest_bdev
;
125 spin_lock(&em_tree
->lock
);
126 ret
= add_extent_mapping(em_tree
, em
);
127 if (ret
== -EEXIST
) {
128 u64 failed_start
= em
->start
;
129 u64 failed_len
= em
->len
;
131 printk("failed to insert %Lu %Lu -> %Lu into tree\n",
132 em
->start
, em
->len
, em
->block_start
);
134 em
= lookup_extent_mapping(em_tree
, start
, len
);
136 printk("after failing, found %Lu %Lu %Lu\n",
137 em
->start
, em
->len
, em
->block_start
);
140 em
= lookup_extent_mapping(em_tree
, failed_start
,
143 printk("double failure lookup gives us "
144 "%Lu %Lu -> %Lu\n", em
->start
,
145 em
->len
, em
->block_start
);
154 spin_unlock(&em_tree
->lock
);
162 u32
btrfs_csum_data(struct btrfs_root
*root
, char *data
, u32 seed
, size_t len
)
164 return btrfs_crc32c(seed
, data
, len
);
167 void btrfs_csum_final(u32 crc
, char *result
)
169 *(__le32
*)result
= ~cpu_to_le32(crc
);
173 * compute the csum for a btree block, and either verify it or write it
174 * into the csum field of the block.
176 static int csum_tree_block(struct btrfs_root
*root
, struct extent_buffer
*buf
,
179 char result
[BTRFS_CRC32_SIZE
];
181 unsigned long cur_len
;
182 unsigned long offset
= BTRFS_CSUM_SIZE
;
183 char *map_token
= NULL
;
185 unsigned long map_start
;
186 unsigned long map_len
;
190 len
= buf
->len
- offset
;
192 err
= map_private_extent_buffer(buf
, offset
, 32,
194 &map_start
, &map_len
, KM_USER0
);
196 printk("failed to map extent buffer! %lu\n",
200 cur_len
= min(len
, map_len
- (offset
- map_start
));
201 crc
= btrfs_csum_data(root
, kaddr
+ offset
- map_start
,
205 unmap_extent_buffer(buf
, map_token
, KM_USER0
);
207 btrfs_csum_final(crc
, result
);
210 /* FIXME, this is not good */
211 if (memcmp_extent_buffer(buf
, result
, 0, BTRFS_CRC32_SIZE
)) {
214 memcpy(&found
, result
, BTRFS_CRC32_SIZE
);
216 read_extent_buffer(buf
, &val
, 0, BTRFS_CRC32_SIZE
);
217 printk("btrfs: %s checksum verify failed on %llu "
218 "wanted %X found %X level %d\n",
219 root
->fs_info
->sb
->s_id
,
220 buf
->start
, val
, found
, btrfs_header_level(buf
));
224 write_extent_buffer(buf
, result
, 0, BTRFS_CRC32_SIZE
);
230 * we can't consider a given block up to date unless the transid of the
231 * block matches the transid in the parent node's pointer. This is how we
232 * detect blocks that either didn't get written at all or got written
233 * in the wrong place.
235 static int verify_parent_transid(struct extent_io_tree
*io_tree
,
236 struct extent_buffer
*eb
, u64 parent_transid
)
240 if (!parent_transid
|| btrfs_header_generation(eb
) == parent_transid
)
243 lock_extent(io_tree
, eb
->start
, eb
->start
+ eb
->len
- 1, GFP_NOFS
);
244 if (extent_buffer_uptodate(io_tree
, eb
) &&
245 btrfs_header_generation(eb
) == parent_transid
) {
249 printk("parent transid verify failed on %llu wanted %llu found %llu\n",
250 (unsigned long long)eb
->start
,
251 (unsigned long long)parent_transid
,
252 (unsigned long long)btrfs_header_generation(eb
));
254 clear_extent_buffer_uptodate(io_tree
, eb
);
256 unlock_extent(io_tree
, eb
->start
, eb
->start
+ eb
->len
- 1,
262 * helper to read a given tree block, doing retries as required when
263 * the checksums don't match and we have alternate mirrors to try.
265 static int btree_read_extent_buffer_pages(struct btrfs_root
*root
,
266 struct extent_buffer
*eb
,
267 u64 start
, u64 parent_transid
)
269 struct extent_io_tree
*io_tree
;
274 io_tree
= &BTRFS_I(root
->fs_info
->btree_inode
)->io_tree
;
276 ret
= read_extent_buffer_pages(io_tree
, eb
, start
, 1,
277 btree_get_extent
, mirror_num
);
279 !verify_parent_transid(io_tree
, eb
, parent_transid
))
281 printk("read extent buffer pages failed with ret %d mirror no %d\n", ret
, mirror_num
);
282 num_copies
= btrfs_num_copies(&root
->fs_info
->mapping_tree
,
288 if (mirror_num
> num_copies
)
295 * checksum a dirty tree block before IO. This has extra checks to make
296 * sure we only fill in the checksum field in the first page of a multi-page block
298 static int csum_dirty_buffer(struct btrfs_root
*root
, struct page
*page
)
300 struct extent_io_tree
*tree
;
301 u64 start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
305 struct extent_buffer
*eb
;
308 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
310 if (page
->private == EXTENT_PAGE_PRIVATE
)
314 len
= page
->private >> 2;
318 eb
= alloc_extent_buffer(tree
, start
, len
, page
, GFP_NOFS
);
319 ret
= btree_read_extent_buffer_pages(root
, eb
, start
+ PAGE_CACHE_SIZE
,
320 btrfs_header_generation(eb
));
322 found_start
= btrfs_header_bytenr(eb
);
323 if (found_start
!= start
) {
324 printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
325 start
, found_start
, len
);
329 if (eb
->first_page
!= page
) {
330 printk("bad first page %lu %lu\n", eb
->first_page
->index
,
335 if (!PageUptodate(page
)) {
336 printk("csum not up to date page %lu\n", page
->index
);
340 found_level
= btrfs_header_level(eb
);
342 csum_tree_block(root
, eb
, 0);
344 free_extent_buffer(eb
);
349 static int check_tree_block_fsid(struct btrfs_root
*root
,
350 struct extent_buffer
*eb
)
352 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
353 u8 fsid
[BTRFS_UUID_SIZE
];
356 read_extent_buffer(eb
, fsid
, (unsigned long)btrfs_header_fsid(eb
),
359 if (!memcmp(fsid
, fs_devices
->fsid
, BTRFS_FSID_SIZE
)) {
363 fs_devices
= fs_devices
->seed
;
368 static int btree_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
369 struct extent_state
*state
)
371 struct extent_io_tree
*tree
;
375 struct extent_buffer
*eb
;
376 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
379 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
380 if (page
->private == EXTENT_PAGE_PRIVATE
)
384 len
= page
->private >> 2;
388 eb
= alloc_extent_buffer(tree
, start
, len
, page
, GFP_NOFS
);
390 found_start
= btrfs_header_bytenr(eb
);
391 if (found_start
!= start
) {
392 printk("bad tree block start %llu %llu\n",
393 (unsigned long long)found_start
,
394 (unsigned long long)eb
->start
);
398 if (eb
->first_page
!= page
) {
399 printk("bad first page %lu %lu\n", eb
->first_page
->index
,
405 if (check_tree_block_fsid(root
, eb
)) {
406 printk("bad fsid on block %Lu\n", eb
->start
);
410 found_level
= btrfs_header_level(eb
);
412 ret
= csum_tree_block(root
, eb
, 1);
416 end
= min_t(u64
, eb
->len
, PAGE_CACHE_SIZE
);
417 end
= eb
->start
+ end
- 1;
419 free_extent_buffer(eb
);
424 static void end_workqueue_bio(struct bio
*bio
, int err
)
426 struct end_io_wq
*end_io_wq
= bio
->bi_private
;
427 struct btrfs_fs_info
*fs_info
;
429 fs_info
= end_io_wq
->info
;
430 end_io_wq
->error
= err
;
431 end_io_wq
->work
.func
= end_workqueue_fn
;
432 end_io_wq
->work
.flags
= 0;
433 if (bio
->bi_rw
& (1 << BIO_RW
))
434 btrfs_queue_worker(&fs_info
->endio_write_workers
,
437 btrfs_queue_worker(&fs_info
->endio_workers
, &end_io_wq
->work
);
440 int btrfs_bio_wq_end_io(struct btrfs_fs_info
*info
, struct bio
*bio
,
443 struct end_io_wq
*end_io_wq
;
444 end_io_wq
= kmalloc(sizeof(*end_io_wq
), GFP_NOFS
);
448 end_io_wq
->private = bio
->bi_private
;
449 end_io_wq
->end_io
= bio
->bi_end_io
;
450 end_io_wq
->info
= info
;
451 end_io_wq
->error
= 0;
452 end_io_wq
->bio
= bio
;
453 end_io_wq
->metadata
= metadata
;
455 bio
->bi_private
= end_io_wq
;
456 bio
->bi_end_io
= end_workqueue_bio
;
460 unsigned long btrfs_async_submit_limit(struct btrfs_fs_info
*info
)
462 unsigned long limit
= min_t(unsigned long,
463 info
->workers
.max_workers
,
464 info
->fs_devices
->open_devices
);
468 int btrfs_congested_async(struct btrfs_fs_info
*info
, int iodone
)
470 return atomic_read(&info
->nr_async_bios
) >
471 btrfs_async_submit_limit(info
);
474 static void run_one_async_start(struct btrfs_work
*work
)
476 struct btrfs_fs_info
*fs_info
;
477 struct async_submit_bio
*async
;
479 async
= container_of(work
, struct async_submit_bio
, work
);
480 fs_info
= BTRFS_I(async
->inode
)->root
->fs_info
;
481 async
->submit_bio_start(async
->inode
, async
->rw
, async
->bio
,
482 async
->mirror_num
, async
->bio_flags
);
485 static void run_one_async_done(struct btrfs_work
*work
)
487 struct btrfs_fs_info
*fs_info
;
488 struct async_submit_bio
*async
;
491 async
= container_of(work
, struct async_submit_bio
, work
);
492 fs_info
= BTRFS_I(async
->inode
)->root
->fs_info
;
494 limit
= btrfs_async_submit_limit(fs_info
);
495 limit
= limit
* 2 / 3;
497 atomic_dec(&fs_info
->nr_async_submits
);
499 if (atomic_read(&fs_info
->nr_async_submits
) < limit
&&
500 waitqueue_active(&fs_info
->async_submit_wait
))
501 wake_up(&fs_info
->async_submit_wait
);
503 async
->submit_bio_done(async
->inode
, async
->rw
, async
->bio
,
504 async
->mirror_num
, async
->bio_flags
);
507 static void run_one_async_free(struct btrfs_work
*work
)
509 struct async_submit_bio
*async
;
511 async
= container_of(work
, struct async_submit_bio
, work
);
515 int btrfs_wq_submit_bio(struct btrfs_fs_info
*fs_info
, struct inode
*inode
,
516 int rw
, struct bio
*bio
, int mirror_num
,
517 unsigned long bio_flags
,
518 extent_submit_bio_hook_t
*submit_bio_start
,
519 extent_submit_bio_hook_t
*submit_bio_done
)
521 struct async_submit_bio
*async
;
523 async
= kmalloc(sizeof(*async
), GFP_NOFS
);
527 async
->inode
= inode
;
530 async
->mirror_num
= mirror_num
;
531 async
->submit_bio_start
= submit_bio_start
;
532 async
->submit_bio_done
= submit_bio_done
;
534 async
->work
.func
= run_one_async_start
;
535 async
->work
.ordered_func
= run_one_async_done
;
536 async
->work
.ordered_free
= run_one_async_free
;
538 async
->work
.flags
= 0;
539 async
->bio_flags
= bio_flags
;
541 atomic_inc(&fs_info
->nr_async_submits
);
542 btrfs_queue_worker(&fs_info
->workers
, &async
->work
);
544 int limit
= btrfs_async_submit_limit(fs_info
);
545 if (atomic_read(&fs_info
->nr_async_submits
) > limit
) {
546 wait_event_timeout(fs_info
->async_submit_wait
,
547 (atomic_read(&fs_info
->nr_async_submits
) < limit
),
550 wait_event_timeout(fs_info
->async_submit_wait
,
551 (atomic_read(&fs_info
->nr_async_bios
) < limit
),
555 while(atomic_read(&fs_info
->async_submit_draining
) &&
556 atomic_read(&fs_info
->nr_async_submits
)) {
557 wait_event(fs_info
->async_submit_wait
,
558 (atomic_read(&fs_info
->nr_async_submits
) == 0));
564 static int btree_csum_one_bio(struct bio
*bio
)
566 struct bio_vec
*bvec
= bio
->bi_io_vec
;
568 struct btrfs_root
*root
;
570 WARN_ON(bio
->bi_vcnt
<= 0);
571 while(bio_index
< bio
->bi_vcnt
) {
572 root
= BTRFS_I(bvec
->bv_page
->mapping
->host
)->root
;
573 csum_dirty_buffer(root
, bvec
->bv_page
);
580 static int __btree_submit_bio_start(struct inode
*inode
, int rw
,
581 struct bio
*bio
, int mirror_num
,
582 unsigned long bio_flags
)
585 * when we're called for a write, we're already in the async
586 * submission context. Just jump into btrfs_map_bio
588 btree_csum_one_bio(bio
);
592 static int __btree_submit_bio_done(struct inode
*inode
, int rw
, struct bio
*bio
,
593 int mirror_num
, unsigned long bio_flags
)
596 * when we're called for a write, we're already in the async
597 * submission context. Just jump into btrfs_map_bio
599 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
, mirror_num
, 1);
602 static int btree_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
603 int mirror_num
, unsigned long bio_flags
)
606 * kthread helpers are used to submit writes so that checksumming
607 * can happen in parallel across all CPUs
609 if (!(rw
& (1 << BIO_RW
))) {
612 * called for a read, do the setup so that checksum validation
613 * can happen in the async kernel threads
615 ret
= btrfs_bio_wq_end_io(BTRFS_I(inode
)->root
->fs_info
,
619 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
,
622 return btrfs_wq_submit_bio(BTRFS_I(inode
)->root
->fs_info
,
623 inode
, rw
, bio
, mirror_num
, 0,
624 __btree_submit_bio_start
,
625 __btree_submit_bio_done
);
628 static int btree_writepage(struct page
*page
, struct writeback_control
*wbc
)
630 struct extent_io_tree
*tree
;
631 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
633 if (current
->flags
& PF_MEMALLOC
) {
634 redirty_page_for_writepage(wbc
, page
);
638 return extent_write_full_page(tree
, page
, btree_get_extent
, wbc
);
641 static int btree_writepages(struct address_space
*mapping
,
642 struct writeback_control
*wbc
)
644 struct extent_io_tree
*tree
;
645 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
646 if (wbc
->sync_mode
== WB_SYNC_NONE
) {
649 unsigned long thresh
= 32 * 1024 * 1024;
651 if (wbc
->for_kupdate
)
654 num_dirty
= count_range_bits(tree
, &start
, (u64
)-1,
655 thresh
, EXTENT_DIRTY
);
656 if (num_dirty
< thresh
) {
660 return extent_writepages(tree
, mapping
, btree_get_extent
, wbc
);
663 static int btree_readpage(struct file
*file
, struct page
*page
)
665 struct extent_io_tree
*tree
;
666 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
667 return extent_read_full_page(tree
, page
, btree_get_extent
);
670 static int btree_releasepage(struct page
*page
, gfp_t gfp_flags
)
672 struct extent_io_tree
*tree
;
673 struct extent_map_tree
*map
;
676 if (PageWriteback(page
) || PageDirty(page
))
679 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
680 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
682 ret
= try_release_extent_state(map
, tree
, page
, gfp_flags
);
687 ret
= try_release_extent_buffer(tree
, page
);
689 ClearPagePrivate(page
);
690 set_page_private(page
, 0);
691 page_cache_release(page
);
697 static void btree_invalidatepage(struct page
*page
, unsigned long offset
)
699 struct extent_io_tree
*tree
;
700 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
701 extent_invalidatepage(tree
, page
, offset
);
702 btree_releasepage(page
, GFP_NOFS
);
703 if (PagePrivate(page
)) {
704 printk("warning page private not zero on page %Lu\n",
706 ClearPagePrivate(page
);
707 set_page_private(page
, 0);
708 page_cache_release(page
);
713 static int btree_writepage(struct page
*page
, struct writeback_control
*wbc
)
715 struct buffer_head
*bh
;
716 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
717 struct buffer_head
*head
;
718 if (!page_has_buffers(page
)) {
719 create_empty_buffers(page
, root
->fs_info
->sb
->s_blocksize
,
720 (1 << BH_Dirty
)|(1 << BH_Uptodate
));
722 head
= page_buffers(page
);
725 if (buffer_dirty(bh
))
726 csum_tree_block(root
, bh
, 0);
727 bh
= bh
->b_this_page
;
728 } while (bh
!= head
);
729 return block_write_full_page(page
, btree_get_block
, wbc
);
733 static struct address_space_operations btree_aops
= {
734 .readpage
= btree_readpage
,
735 .writepage
= btree_writepage
,
736 .writepages
= btree_writepages
,
737 .releasepage
= btree_releasepage
,
738 .invalidatepage
= btree_invalidatepage
,
739 .sync_page
= block_sync_page
,
742 int readahead_tree_block(struct btrfs_root
*root
, u64 bytenr
, u32 blocksize
,
745 struct extent_buffer
*buf
= NULL
;
746 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
749 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
752 read_extent_buffer_pages(&BTRFS_I(btree_inode
)->io_tree
,
753 buf
, 0, 0, btree_get_extent
, 0);
754 free_extent_buffer(buf
);
758 struct extent_buffer
*btrfs_find_tree_block(struct btrfs_root
*root
,
759 u64 bytenr
, u32 blocksize
)
761 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
762 struct extent_buffer
*eb
;
763 eb
= find_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
764 bytenr
, blocksize
, GFP_NOFS
);
768 struct extent_buffer
*btrfs_find_create_tree_block(struct btrfs_root
*root
,
769 u64 bytenr
, u32 blocksize
)
771 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
772 struct extent_buffer
*eb
;
774 eb
= alloc_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
775 bytenr
, blocksize
, NULL
, GFP_NOFS
);
780 int btrfs_write_tree_block(struct extent_buffer
*buf
)
782 return btrfs_fdatawrite_range(buf
->first_page
->mapping
, buf
->start
,
783 buf
->start
+ buf
->len
- 1, WB_SYNC_ALL
);
786 int btrfs_wait_tree_block_writeback(struct extent_buffer
*buf
)
788 return btrfs_wait_on_page_writeback_range(buf
->first_page
->mapping
,
789 buf
->start
, buf
->start
+ buf
->len
-1);
792 struct extent_buffer
*read_tree_block(struct btrfs_root
*root
, u64 bytenr
,
793 u32 blocksize
, u64 parent_transid
)
795 struct extent_buffer
*buf
= NULL
;
796 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
797 struct extent_io_tree
*io_tree
;
800 io_tree
= &BTRFS_I(btree_inode
)->io_tree
;
802 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
806 ret
= btree_read_extent_buffer_pages(root
, buf
, 0, parent_transid
);
809 buf
->flags
|= EXTENT_UPTODATE
;
817 int clean_tree_block(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
818 struct extent_buffer
*buf
)
820 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
821 if (btrfs_header_generation(buf
) ==
822 root
->fs_info
->running_transaction
->transid
) {
823 WARN_ON(!btrfs_tree_locked(buf
));
824 clear_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
,
830 static int __setup_root(u32 nodesize
, u32 leafsize
, u32 sectorsize
,
831 u32 stripesize
, struct btrfs_root
*root
,
832 struct btrfs_fs_info
*fs_info
,
836 root
->commit_root
= NULL
;
837 root
->ref_tree
= NULL
;
838 root
->sectorsize
= sectorsize
;
839 root
->nodesize
= nodesize
;
840 root
->leafsize
= leafsize
;
841 root
->stripesize
= stripesize
;
843 root
->track_dirty
= 0;
845 root
->fs_info
= fs_info
;
846 root
->objectid
= objectid
;
847 root
->last_trans
= 0;
848 root
->highest_inode
= 0;
849 root
->last_inode_alloc
= 0;
853 INIT_LIST_HEAD(&root
->dirty_list
);
854 INIT_LIST_HEAD(&root
->orphan_list
);
855 INIT_LIST_HEAD(&root
->dead_list
);
856 spin_lock_init(&root
->node_lock
);
857 spin_lock_init(&root
->list_lock
);
858 mutex_init(&root
->objectid_mutex
);
859 mutex_init(&root
->log_mutex
);
860 extent_io_tree_init(&root
->dirty_log_pages
,
861 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
863 btrfs_leaf_ref_tree_init(&root
->ref_tree_struct
);
864 root
->ref_tree
= &root
->ref_tree_struct
;
866 memset(&root
->root_key
, 0, sizeof(root
->root_key
));
867 memset(&root
->root_item
, 0, sizeof(root
->root_item
));
868 memset(&root
->defrag_progress
, 0, sizeof(root
->defrag_progress
));
869 memset(&root
->root_kobj
, 0, sizeof(root
->root_kobj
));
870 root
->defrag_trans_start
= fs_info
->generation
;
871 init_completion(&root
->kobj_unregister
);
872 root
->defrag_running
= 0;
873 root
->defrag_level
= 0;
874 root
->root_key
.objectid
= objectid
;
875 root
->anon_super
.s_root
= NULL
;
876 root
->anon_super
.s_dev
= 0;
877 INIT_LIST_HEAD(&root
->anon_super
.s_list
);
878 INIT_LIST_HEAD(&root
->anon_super
.s_instances
);
879 init_rwsem(&root
->anon_super
.s_umount
);
884 static int find_and_setup_root(struct btrfs_root
*tree_root
,
885 struct btrfs_fs_info
*fs_info
,
887 struct btrfs_root
*root
)
893 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
894 tree_root
->sectorsize
, tree_root
->stripesize
,
895 root
, fs_info
, objectid
);
896 ret
= btrfs_find_last_root(tree_root
, objectid
,
897 &root
->root_item
, &root
->root_key
);
900 generation
= btrfs_root_generation(&root
->root_item
);
901 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
902 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
903 blocksize
, generation
);
908 int btrfs_free_log_root_tree(struct btrfs_trans_handle
*trans
,
909 struct btrfs_fs_info
*fs_info
)
911 struct extent_buffer
*eb
;
912 struct btrfs_root
*log_root_tree
= fs_info
->log_root_tree
;
921 ret
= find_first_extent_bit(&log_root_tree
->dirty_log_pages
,
922 0, &start
, &end
, EXTENT_DIRTY
);
926 clear_extent_dirty(&log_root_tree
->dirty_log_pages
,
927 start
, end
, GFP_NOFS
);
929 eb
= fs_info
->log_root_tree
->node
;
931 WARN_ON(btrfs_header_level(eb
) != 0);
932 WARN_ON(btrfs_header_nritems(eb
) != 0);
934 ret
= btrfs_free_reserved_extent(fs_info
->tree_root
,
938 free_extent_buffer(eb
);
939 kfree(fs_info
->log_root_tree
);
940 fs_info
->log_root_tree
= NULL
;
944 int btrfs_init_log_root_tree(struct btrfs_trans_handle
*trans
,
945 struct btrfs_fs_info
*fs_info
)
947 struct btrfs_root
*root
;
948 struct btrfs_root
*tree_root
= fs_info
->tree_root
;
950 root
= kzalloc(sizeof(*root
), GFP_NOFS
);
954 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
955 tree_root
->sectorsize
, tree_root
->stripesize
,
956 root
, fs_info
, BTRFS_TREE_LOG_OBJECTID
);
958 root
->root_key
.objectid
= BTRFS_TREE_LOG_OBJECTID
;
959 root
->root_key
.type
= BTRFS_ROOT_ITEM_KEY
;
960 root
->root_key
.offset
= BTRFS_TREE_LOG_OBJECTID
;
963 root
->node
= btrfs_alloc_free_block(trans
, root
, root
->leafsize
,
964 0, BTRFS_TREE_LOG_OBJECTID
,
965 trans
->transid
, 0, 0, 0);
967 btrfs_set_header_nritems(root
->node
, 0);
968 btrfs_set_header_level(root
->node
, 0);
969 btrfs_set_header_bytenr(root
->node
, root
->node
->start
);
970 btrfs_set_header_generation(root
->node
, trans
->transid
);
971 btrfs_set_header_owner(root
->node
, BTRFS_TREE_LOG_OBJECTID
);
973 write_extent_buffer(root
->node
, root
->fs_info
->fsid
,
974 (unsigned long)btrfs_header_fsid(root
->node
),
976 btrfs_mark_buffer_dirty(root
->node
);
977 btrfs_tree_unlock(root
->node
);
978 fs_info
->log_root_tree
= root
;
982 struct btrfs_root
*btrfs_read_fs_root_no_radix(struct btrfs_root
*tree_root
,
983 struct btrfs_key
*location
)
985 struct btrfs_root
*root
;
986 struct btrfs_fs_info
*fs_info
= tree_root
->fs_info
;
987 struct btrfs_path
*path
;
988 struct extent_buffer
*l
;
994 root
= kzalloc(sizeof(*root
), GFP_NOFS
);
996 return ERR_PTR(-ENOMEM
);
997 if (location
->offset
== (u64
)-1) {
998 ret
= find_and_setup_root(tree_root
, fs_info
,
999 location
->objectid
, root
);
1002 return ERR_PTR(ret
);
1007 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
1008 tree_root
->sectorsize
, tree_root
->stripesize
,
1009 root
, fs_info
, location
->objectid
);
1011 path
= btrfs_alloc_path();
1013 ret
= btrfs_search_slot(NULL
, tree_root
, location
, path
, 0, 0);
1020 read_extent_buffer(l
, &root
->root_item
,
1021 btrfs_item_ptr_offset(l
, path
->slots
[0]),
1022 sizeof(root
->root_item
));
1023 memcpy(&root
->root_key
, location
, sizeof(*location
));
1026 btrfs_release_path(root
, path
);
1027 btrfs_free_path(path
);
1030 return ERR_PTR(ret
);
1032 generation
= btrfs_root_generation(&root
->root_item
);
1033 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
1034 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
1035 blocksize
, generation
);
1036 BUG_ON(!root
->node
);
1038 if (location
->objectid
!= BTRFS_TREE_LOG_OBJECTID
) {
1040 ret
= btrfs_find_highest_inode(root
, &highest_inode
);
1042 root
->highest_inode
= highest_inode
;
1043 root
->last_inode_alloc
= highest_inode
;
1049 struct btrfs_root
*btrfs_lookup_fs_root(struct btrfs_fs_info
*fs_info
,
1052 struct btrfs_root
*root
;
1054 if (root_objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1055 return fs_info
->tree_root
;
1056 if (root_objectid
== BTRFS_EXTENT_TREE_OBJECTID
)
1057 return fs_info
->extent_root
;
1059 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
1060 (unsigned long)root_objectid
);
1064 struct btrfs_root
*btrfs_read_fs_root_no_name(struct btrfs_fs_info
*fs_info
,
1065 struct btrfs_key
*location
)
1067 struct btrfs_root
*root
;
1070 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1071 return fs_info
->tree_root
;
1072 if (location
->objectid
== BTRFS_EXTENT_TREE_OBJECTID
)
1073 return fs_info
->extent_root
;
1074 if (location
->objectid
== BTRFS_CHUNK_TREE_OBJECTID
)
1075 return fs_info
->chunk_root
;
1076 if (location
->objectid
== BTRFS_DEV_TREE_OBJECTID
)
1077 return fs_info
->dev_root
;
1079 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
1080 (unsigned long)location
->objectid
);
1084 root
= btrfs_read_fs_root_no_radix(fs_info
->tree_root
, location
);
1088 set_anon_super(&root
->anon_super
, NULL
);
1090 ret
= radix_tree_insert(&fs_info
->fs_roots_radix
,
1091 (unsigned long)root
->root_key
.objectid
,
1094 free_extent_buffer(root
->node
);
1096 return ERR_PTR(ret
);
1098 if (!(fs_info
->sb
->s_flags
& MS_RDONLY
)) {
1099 ret
= btrfs_find_dead_roots(fs_info
->tree_root
,
1100 root
->root_key
.objectid
, root
);
1102 btrfs_orphan_cleanup(root
);
1107 struct btrfs_root
*btrfs_read_fs_root(struct btrfs_fs_info
*fs_info
,
1108 struct btrfs_key
*location
,
1109 const char *name
, int namelen
)
1111 struct btrfs_root
*root
;
1114 root
= btrfs_read_fs_root_no_name(fs_info
, location
);
1121 ret
= btrfs_set_root_name(root
, name
, namelen
);
1123 free_extent_buffer(root
->node
);
1125 return ERR_PTR(ret
);
1128 ret
= btrfs_sysfs_add_root(root
);
1130 free_extent_buffer(root
->node
);
1133 return ERR_PTR(ret
);
1140 static int add_hasher(struct btrfs_fs_info
*info
, char *type
) {
1141 struct btrfs_hasher
*hasher
;
1143 hasher
= kmalloc(sizeof(*hasher
), GFP_NOFS
);
1146 hasher
->hash_tfm
= crypto_alloc_hash(type
, 0, CRYPTO_ALG_ASYNC
);
1147 if (!hasher
->hash_tfm
) {
1151 spin_lock(&info
->hash_lock
);
1152 list_add(&hasher
->list
, &info
->hashers
);
1153 spin_unlock(&info
->hash_lock
);
1158 static int btrfs_congested_fn(void *congested_data
, int bdi_bits
)
1160 struct btrfs_fs_info
*info
= (struct btrfs_fs_info
*)congested_data
;
1162 struct list_head
*cur
;
1163 struct btrfs_device
*device
;
1164 struct backing_dev_info
*bdi
;
1166 if ((bdi_bits
& (1 << BDI_write_congested
)) &&
1167 btrfs_congested_async(info
, 0))
1170 list_for_each(cur
, &info
->fs_devices
->devices
) {
1171 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
1174 bdi
= blk_get_backing_dev_info(device
->bdev
);
1175 if (bdi
&& bdi_congested(bdi
, bdi_bits
)) {
1184 * this unplugs every device on the box, and it is only used when page
1187 static void __unplug_io_fn(struct backing_dev_info
*bdi
, struct page
*page
)
1189 struct list_head
*cur
;
1190 struct btrfs_device
*device
;
1191 struct btrfs_fs_info
*info
;
1193 info
= (struct btrfs_fs_info
*)bdi
->unplug_io_data
;
1194 list_for_each(cur
, &info
->fs_devices
->devices
) {
1195 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
1196 bdi
= blk_get_backing_dev_info(device
->bdev
);
1197 if (bdi
->unplug_io_fn
) {
1198 bdi
->unplug_io_fn(bdi
, page
);
1203 static void btrfs_unplug_io_fn(struct backing_dev_info
*bdi
, struct page
*page
)
1205 struct inode
*inode
;
1206 struct extent_map_tree
*em_tree
;
1207 struct extent_map
*em
;
1208 struct address_space
*mapping
;
1211 /* the generic O_DIRECT read code does this */
1213 __unplug_io_fn(bdi
, page
);
1218 * page->mapping may change at any time. Get a consistent copy
1219 * and use that for everything below
1222 mapping
= page
->mapping
;
1226 inode
= mapping
->host
;
1229 * don't do the expensive searching for a small number of
1232 if (BTRFS_I(inode
)->root
->fs_info
->fs_devices
->open_devices
<= 2) {
1233 __unplug_io_fn(bdi
, page
);
1237 offset
= page_offset(page
);
1239 em_tree
= &BTRFS_I(inode
)->extent_tree
;
1240 spin_lock(&em_tree
->lock
);
1241 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_CACHE_SIZE
);
1242 spin_unlock(&em_tree
->lock
);
1244 __unplug_io_fn(bdi
, page
);
1248 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
) {
1249 free_extent_map(em
);
1250 __unplug_io_fn(bdi
, page
);
1253 offset
= offset
- em
->start
;
1254 btrfs_unplug_page(&BTRFS_I(inode
)->root
->fs_info
->mapping_tree
,
1255 em
->block_start
+ offset
, page
);
1256 free_extent_map(em
);
1259 static int setup_bdi(struct btrfs_fs_info
*info
, struct backing_dev_info
*bdi
)
1262 bdi
->ra_pages
= default_backing_dev_info
.ra_pages
;
1264 bdi
->capabilities
= default_backing_dev_info
.capabilities
;
1265 bdi
->unplug_io_fn
= btrfs_unplug_io_fn
;
1266 bdi
->unplug_io_data
= info
;
1267 bdi
->congested_fn
= btrfs_congested_fn
;
1268 bdi
->congested_data
= info
;
1272 static int bio_ready_for_csum(struct bio
*bio
)
1278 struct extent_io_tree
*io_tree
= NULL
;
1279 struct btrfs_fs_info
*info
= NULL
;
1280 struct bio_vec
*bvec
;
1284 bio_for_each_segment(bvec
, bio
, i
) {
1285 page
= bvec
->bv_page
;
1286 if (page
->private == EXTENT_PAGE_PRIVATE
) {
1287 length
+= bvec
->bv_len
;
1290 if (!page
->private) {
1291 length
+= bvec
->bv_len
;
1294 length
= bvec
->bv_len
;
1295 buf_len
= page
->private >> 2;
1296 start
= page_offset(page
) + bvec
->bv_offset
;
1297 io_tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
1298 info
= BTRFS_I(page
->mapping
->host
)->root
->fs_info
;
1300 /* are we fully contained in this bio? */
1301 if (buf_len
<= length
)
1304 ret
= extent_range_uptodate(io_tree
, start
+ length
,
1305 start
+ buf_len
- 1);
1312 * called by the kthread helper functions to finally call the bio end_io
1313 * functions. This is where read checksum verification actually happens
1315 static void end_workqueue_fn(struct btrfs_work
*work
)
1318 struct end_io_wq
*end_io_wq
;
1319 struct btrfs_fs_info
*fs_info
;
1322 end_io_wq
= container_of(work
, struct end_io_wq
, work
);
1323 bio
= end_io_wq
->bio
;
1324 fs_info
= end_io_wq
->info
;
1326 /* metadata bios are special because the whole tree block must
1327 * be checksummed at once. This makes sure the entire block is in
1328 * ram and up to date before trying to verify things. For
1329 * blocksize <= pagesize, it is basically a noop
1331 if (end_io_wq
->metadata
&& !bio_ready_for_csum(bio
)) {
1332 btrfs_queue_worker(&fs_info
->endio_workers
,
1336 error
= end_io_wq
->error
;
1337 bio
->bi_private
= end_io_wq
->private;
1338 bio
->bi_end_io
= end_io_wq
->end_io
;
1340 bio_endio(bio
, error
);
1343 static int cleaner_kthread(void *arg
)
1345 struct btrfs_root
*root
= arg
;
1349 if (root
->fs_info
->closing
)
1352 vfs_check_frozen(root
->fs_info
->sb
, SB_FREEZE_WRITE
);
1353 mutex_lock(&root
->fs_info
->cleaner_mutex
);
1354 btrfs_clean_old_snapshots(root
);
1355 mutex_unlock(&root
->fs_info
->cleaner_mutex
);
1357 if (freezing(current
)) {
1361 if (root
->fs_info
->closing
)
1363 set_current_state(TASK_INTERRUPTIBLE
);
1365 __set_current_state(TASK_RUNNING
);
1367 } while (!kthread_should_stop());
1371 static int transaction_kthread(void *arg
)
1373 struct btrfs_root
*root
= arg
;
1374 struct btrfs_trans_handle
*trans
;
1375 struct btrfs_transaction
*cur
;
1377 unsigned long delay
;
1382 if (root
->fs_info
->closing
)
1386 vfs_check_frozen(root
->fs_info
->sb
, SB_FREEZE_WRITE
);
1387 mutex_lock(&root
->fs_info
->transaction_kthread_mutex
);
1389 if (root
->fs_info
->total_ref_cache_size
> 20 * 1024 * 1024) {
1390 printk("btrfs: total reference cache size %Lu\n",
1391 root
->fs_info
->total_ref_cache_size
);
1394 mutex_lock(&root
->fs_info
->trans_mutex
);
1395 cur
= root
->fs_info
->running_transaction
;
1397 mutex_unlock(&root
->fs_info
->trans_mutex
);
1401 now
= get_seconds();
1402 if (now
< cur
->start_time
|| now
- cur
->start_time
< 30) {
1403 mutex_unlock(&root
->fs_info
->trans_mutex
);
1407 mutex_unlock(&root
->fs_info
->trans_mutex
);
1408 trans
= btrfs_start_transaction(root
, 1);
1409 ret
= btrfs_commit_transaction(trans
, root
);
1411 wake_up_process(root
->fs_info
->cleaner_kthread
);
1412 mutex_unlock(&root
->fs_info
->transaction_kthread_mutex
);
1414 if (freezing(current
)) {
1417 if (root
->fs_info
->closing
)
1419 set_current_state(TASK_INTERRUPTIBLE
);
1420 schedule_timeout(delay
);
1421 __set_current_state(TASK_RUNNING
);
1423 } while (!kthread_should_stop());
1427 struct btrfs_root
*open_ctree(struct super_block
*sb
,
1428 struct btrfs_fs_devices
*fs_devices
,
1438 struct btrfs_key location
;
1439 struct buffer_head
*bh
;
1440 struct btrfs_root
*extent_root
= kzalloc(sizeof(struct btrfs_root
),
1442 struct btrfs_root
*tree_root
= kzalloc(sizeof(struct btrfs_root
),
1444 struct btrfs_fs_info
*fs_info
= kzalloc(sizeof(*fs_info
),
1446 struct btrfs_root
*chunk_root
= kzalloc(sizeof(struct btrfs_root
),
1448 struct btrfs_root
*dev_root
= kzalloc(sizeof(struct btrfs_root
),
1450 struct btrfs_root
*log_tree_root
;
1455 struct btrfs_super_block
*disk_super
;
1457 if (!extent_root
|| !tree_root
|| !fs_info
||
1458 !chunk_root
|| !dev_root
) {
1462 INIT_RADIX_TREE(&fs_info
->fs_roots_radix
, GFP_NOFS
);
1463 INIT_LIST_HEAD(&fs_info
->trans_list
);
1464 INIT_LIST_HEAD(&fs_info
->dead_roots
);
1465 INIT_LIST_HEAD(&fs_info
->hashers
);
1466 INIT_LIST_HEAD(&fs_info
->delalloc_inodes
);
1467 spin_lock_init(&fs_info
->hash_lock
);
1468 spin_lock_init(&fs_info
->delalloc_lock
);
1469 spin_lock_init(&fs_info
->new_trans_lock
);
1470 spin_lock_init(&fs_info
->ref_cache_lock
);
1472 init_completion(&fs_info
->kobj_unregister
);
1473 fs_info
->tree_root
= tree_root
;
1474 fs_info
->extent_root
= extent_root
;
1475 fs_info
->chunk_root
= chunk_root
;
1476 fs_info
->dev_root
= dev_root
;
1477 fs_info
->fs_devices
= fs_devices
;
1478 INIT_LIST_HEAD(&fs_info
->dirty_cowonly_roots
);
1479 INIT_LIST_HEAD(&fs_info
->space_info
);
1480 btrfs_mapping_init(&fs_info
->mapping_tree
);
1481 atomic_set(&fs_info
->nr_async_submits
, 0);
1482 atomic_set(&fs_info
->async_delalloc_pages
, 0);
1483 atomic_set(&fs_info
->async_submit_draining
, 0);
1484 atomic_set(&fs_info
->nr_async_bios
, 0);
1485 atomic_set(&fs_info
->throttles
, 0);
1486 atomic_set(&fs_info
->throttle_gen
, 0);
1488 fs_info
->max_extent
= (u64
)-1;
1489 fs_info
->max_inline
= 8192 * 1024;
1490 setup_bdi(fs_info
, &fs_info
->bdi
);
1491 fs_info
->btree_inode
= new_inode(sb
);
1492 fs_info
->btree_inode
->i_ino
= 1;
1493 fs_info
->btree_inode
->i_nlink
= 1;
1495 fs_info
->thread_pool_size
= min(num_online_cpus() + 2, 8);
1497 INIT_LIST_HEAD(&fs_info
->ordered_extents
);
1498 spin_lock_init(&fs_info
->ordered_extent_lock
);
1500 sb
->s_blocksize
= 4096;
1501 sb
->s_blocksize_bits
= blksize_bits(4096);
1504 * we set the i_size on the btree inode to the max possible int.
1505 * the real end of the address space is determined by all of
1506 * the devices in the system
1508 fs_info
->btree_inode
->i_size
= OFFSET_MAX
;
1509 fs_info
->btree_inode
->i_mapping
->a_ops
= &btree_aops
;
1510 fs_info
->btree_inode
->i_mapping
->backing_dev_info
= &fs_info
->bdi
;
1512 extent_io_tree_init(&BTRFS_I(fs_info
->btree_inode
)->io_tree
,
1513 fs_info
->btree_inode
->i_mapping
,
1515 extent_map_tree_init(&BTRFS_I(fs_info
->btree_inode
)->extent_tree
,
1518 BTRFS_I(fs_info
->btree_inode
)->io_tree
.ops
= &btree_extent_io_ops
;
1520 spin_lock_init(&fs_info
->block_group_cache_lock
);
1521 fs_info
->block_group_cache_tree
.rb_node
= NULL
;
1523 extent_io_tree_init(&fs_info
->pinned_extents
,
1524 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1525 extent_io_tree_init(&fs_info
->pending_del
,
1526 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1527 extent_io_tree_init(&fs_info
->extent_ins
,
1528 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1529 fs_info
->do_barriers
= 1;
1531 INIT_LIST_HEAD(&fs_info
->dead_reloc_roots
);
1532 btrfs_leaf_ref_tree_init(&fs_info
->reloc_ref_tree
);
1533 btrfs_leaf_ref_tree_init(&fs_info
->shared_ref_tree
);
1535 BTRFS_I(fs_info
->btree_inode
)->root
= tree_root
;
1536 memset(&BTRFS_I(fs_info
->btree_inode
)->location
, 0,
1537 sizeof(struct btrfs_key
));
1538 insert_inode_hash(fs_info
->btree_inode
);
1540 mutex_init(&fs_info
->trans_mutex
);
1541 mutex_init(&fs_info
->tree_log_mutex
);
1542 mutex_init(&fs_info
->drop_mutex
);
1543 mutex_init(&fs_info
->extent_ins_mutex
);
1544 mutex_init(&fs_info
->pinned_mutex
);
1545 mutex_init(&fs_info
->chunk_mutex
);
1546 mutex_init(&fs_info
->transaction_kthread_mutex
);
1547 mutex_init(&fs_info
->cleaner_mutex
);
1548 mutex_init(&fs_info
->volume_mutex
);
1549 mutex_init(&fs_info
->tree_reloc_mutex
);
1550 init_waitqueue_head(&fs_info
->transaction_throttle
);
1551 init_waitqueue_head(&fs_info
->transaction_wait
);
1552 init_waitqueue_head(&fs_info
->async_submit_wait
);
1553 init_waitqueue_head(&fs_info
->tree_log_wait
);
1554 atomic_set(&fs_info
->tree_log_commit
, 0);
1555 atomic_set(&fs_info
->tree_log_writers
, 0);
1556 fs_info
->tree_log_transid
= 0;
1559 ret
= add_hasher(fs_info
, "crc32c");
1561 printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
1566 __setup_root(4096, 4096, 4096, 4096, tree_root
,
1567 fs_info
, BTRFS_ROOT_TREE_OBJECTID
);
1570 bh
= __bread(fs_devices
->latest_bdev
,
1571 BTRFS_SUPER_INFO_OFFSET
/ 4096, 4096);
1575 memcpy(&fs_info
->super_copy
, bh
->b_data
, sizeof(fs_info
->super_copy
));
1578 memcpy(fs_info
->fsid
, fs_info
->super_copy
.fsid
, BTRFS_FSID_SIZE
);
1580 disk_super
= &fs_info
->super_copy
;
1581 if (!btrfs_super_root(disk_super
))
1582 goto fail_sb_buffer
;
1584 ret
= btrfs_parse_options(tree_root
, options
);
1587 goto fail_sb_buffer
;
1590 features
= btrfs_super_incompat_flags(disk_super
) &
1591 ~BTRFS_FEATURE_INCOMPAT_SUPP
;
1593 printk(KERN_ERR
"BTRFS: couldn't mount because of "
1594 "unsupported optional features (%Lx).\n",
1597 goto fail_sb_buffer
;
1600 features
= btrfs_super_compat_ro_flags(disk_super
) &
1601 ~BTRFS_FEATURE_COMPAT_RO_SUPP
;
1602 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
1603 printk(KERN_ERR
"BTRFS: couldn't mount RDWR because of "
1604 "unsupported option features (%Lx).\n",
1607 goto fail_sb_buffer
;
1611 * we need to start all the end_io workers up front because the
1612 * queue work function gets called at interrupt time, and so it
1613 * cannot dynamically grow.
1615 btrfs_init_workers(&fs_info
->workers
, "worker",
1616 fs_info
->thread_pool_size
);
1618 btrfs_init_workers(&fs_info
->delalloc_workers
, "delalloc",
1619 fs_info
->thread_pool_size
);
1621 btrfs_init_workers(&fs_info
->submit_workers
, "submit",
1622 min_t(u64
, fs_devices
->num_devices
,
1623 fs_info
->thread_pool_size
));
1625 /* a higher idle thresh on the submit workers makes it much more
1626 * likely that bios will be send down in a sane order to the
1629 fs_info
->submit_workers
.idle_thresh
= 64;
1631 fs_info
->workers
.idle_thresh
= 16;
1632 fs_info
->workers
.ordered
= 1;
1634 fs_info
->delalloc_workers
.idle_thresh
= 2;
1635 fs_info
->delalloc_workers
.ordered
= 1;
1637 btrfs_init_workers(&fs_info
->fixup_workers
, "fixup", 1);
1638 btrfs_init_workers(&fs_info
->endio_workers
, "endio",
1639 fs_info
->thread_pool_size
);
1640 btrfs_init_workers(&fs_info
->endio_write_workers
, "endio-write",
1641 fs_info
->thread_pool_size
);
1644 * endios are largely parallel and should have a very
1647 fs_info
->endio_workers
.idle_thresh
= 4;
1648 fs_info
->endio_write_workers
.idle_thresh
= 64;
1650 btrfs_start_workers(&fs_info
->workers
, 1);
1651 btrfs_start_workers(&fs_info
->submit_workers
, 1);
1652 btrfs_start_workers(&fs_info
->delalloc_workers
, 1);
1653 btrfs_start_workers(&fs_info
->fixup_workers
, 1);
1654 btrfs_start_workers(&fs_info
->endio_workers
, fs_info
->thread_pool_size
);
1655 btrfs_start_workers(&fs_info
->endio_write_workers
,
1656 fs_info
->thread_pool_size
);
1658 fs_info
->bdi
.ra_pages
*= btrfs_super_num_devices(disk_super
);
1659 fs_info
->bdi
.ra_pages
= max(fs_info
->bdi
.ra_pages
,
1660 4 * 1024 * 1024 / PAGE_CACHE_SIZE
);
1662 nodesize
= btrfs_super_nodesize(disk_super
);
1663 leafsize
= btrfs_super_leafsize(disk_super
);
1664 sectorsize
= btrfs_super_sectorsize(disk_super
);
1665 stripesize
= btrfs_super_stripesize(disk_super
);
1666 tree_root
->nodesize
= nodesize
;
1667 tree_root
->leafsize
= leafsize
;
1668 tree_root
->sectorsize
= sectorsize
;
1669 tree_root
->stripesize
= stripesize
;
1671 sb
->s_blocksize
= sectorsize
;
1672 sb
->s_blocksize_bits
= blksize_bits(sectorsize
);
1674 if (strncmp((char *)(&disk_super
->magic
), BTRFS_MAGIC
,
1675 sizeof(disk_super
->magic
))) {
1676 printk("btrfs: valid FS not found on %s\n", sb
->s_id
);
1677 goto fail_sb_buffer
;
1680 mutex_lock(&fs_info
->chunk_mutex
);
1681 ret
= btrfs_read_sys_array(tree_root
);
1682 mutex_unlock(&fs_info
->chunk_mutex
);
1684 printk("btrfs: failed to read the system array on %s\n",
1686 goto fail_sys_array
;
1689 blocksize
= btrfs_level_size(tree_root
,
1690 btrfs_super_chunk_root_level(disk_super
));
1691 generation
= btrfs_super_chunk_root_generation(disk_super
);
1693 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
1694 chunk_root
, fs_info
, BTRFS_CHUNK_TREE_OBJECTID
);
1696 chunk_root
->node
= read_tree_block(chunk_root
,
1697 btrfs_super_chunk_root(disk_super
),
1698 blocksize
, generation
);
1699 BUG_ON(!chunk_root
->node
);
1701 read_extent_buffer(chunk_root
->node
, fs_info
->chunk_tree_uuid
,
1702 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root
->node
),
1705 mutex_lock(&fs_info
->chunk_mutex
);
1706 ret
= btrfs_read_chunk_tree(chunk_root
);
1707 mutex_unlock(&fs_info
->chunk_mutex
);
1709 printk("btrfs: failed to read chunk tree on %s\n", sb
->s_id
);
1710 goto fail_chunk_root
;
1713 btrfs_close_extra_devices(fs_devices
);
1715 blocksize
= btrfs_level_size(tree_root
,
1716 btrfs_super_root_level(disk_super
));
1717 generation
= btrfs_super_generation(disk_super
);
1719 tree_root
->node
= read_tree_block(tree_root
,
1720 btrfs_super_root(disk_super
),
1721 blocksize
, generation
);
1722 if (!tree_root
->node
)
1723 goto fail_chunk_root
;
1726 ret
= find_and_setup_root(tree_root
, fs_info
,
1727 BTRFS_EXTENT_TREE_OBJECTID
, extent_root
);
1729 goto fail_tree_root
;
1730 extent_root
->track_dirty
= 1;
1732 ret
= find_and_setup_root(tree_root
, fs_info
,
1733 BTRFS_DEV_TREE_OBJECTID
, dev_root
);
1734 dev_root
->track_dirty
= 1;
1737 goto fail_extent_root
;
1739 btrfs_read_block_groups(extent_root
);
1741 fs_info
->generation
= generation
+ 1;
1742 fs_info
->last_trans_committed
= generation
;
1743 fs_info
->data_alloc_profile
= (u64
)-1;
1744 fs_info
->metadata_alloc_profile
= (u64
)-1;
1745 fs_info
->system_alloc_profile
= fs_info
->metadata_alloc_profile
;
1746 fs_info
->cleaner_kthread
= kthread_run(cleaner_kthread
, tree_root
,
1748 if (!fs_info
->cleaner_kthread
)
1749 goto fail_extent_root
;
1751 fs_info
->transaction_kthread
= kthread_run(transaction_kthread
,
1753 "btrfs-transaction");
1754 if (!fs_info
->transaction_kthread
)
1757 if (btrfs_super_log_root(disk_super
) != 0) {
1758 u64 bytenr
= btrfs_super_log_root(disk_super
);
1760 if (fs_devices
->rw_devices
== 0) {
1761 printk("Btrfs log replay required on RO media\n");
1763 goto fail_trans_kthread
;
1766 btrfs_level_size(tree_root
,
1767 btrfs_super_log_root_level(disk_super
));
1769 log_tree_root
= kzalloc(sizeof(struct btrfs_root
),
1772 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
1773 log_tree_root
, fs_info
, BTRFS_TREE_LOG_OBJECTID
);
1775 log_tree_root
->node
= read_tree_block(tree_root
, bytenr
,
1778 ret
= btrfs_recover_log_trees(log_tree_root
);
1781 if (sb
->s_flags
& MS_RDONLY
) {
1782 ret
= btrfs_commit_super(tree_root
);
1787 if (!(sb
->s_flags
& MS_RDONLY
)) {
1788 ret
= btrfs_cleanup_reloc_trees(tree_root
);
1792 location
.objectid
= BTRFS_FS_TREE_OBJECTID
;
1793 location
.type
= BTRFS_ROOT_ITEM_KEY
;
1794 location
.offset
= (u64
)-1;
1796 fs_info
->fs_root
= btrfs_read_fs_root_no_name(fs_info
, &location
);
1797 if (!fs_info
->fs_root
)
1798 goto fail_trans_kthread
;
1802 kthread_stop(fs_info
->transaction_kthread
);
1804 kthread_stop(fs_info
->cleaner_kthread
);
1807 * make sure we're done with the btree inode before we stop our
1810 filemap_write_and_wait(fs_info
->btree_inode
->i_mapping
);
1811 invalidate_inode_pages2(fs_info
->btree_inode
->i_mapping
);
1814 free_extent_buffer(extent_root
->node
);
1816 free_extent_buffer(tree_root
->node
);
1818 free_extent_buffer(chunk_root
->node
);
1820 free_extent_buffer(dev_root
->node
);
1822 btrfs_stop_workers(&fs_info
->fixup_workers
);
1823 btrfs_stop_workers(&fs_info
->delalloc_workers
);
1824 btrfs_stop_workers(&fs_info
->workers
);
1825 btrfs_stop_workers(&fs_info
->endio_workers
);
1826 btrfs_stop_workers(&fs_info
->endio_write_workers
);
1827 btrfs_stop_workers(&fs_info
->submit_workers
);
1829 invalidate_inode_pages2(fs_info
->btree_inode
->i_mapping
);
1830 iput(fs_info
->btree_inode
);
1832 btrfs_close_devices(fs_info
->fs_devices
);
1833 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
1837 bdi_destroy(&fs_info
->bdi
);
1841 return ERR_PTR(err
);
1844 static void btrfs_end_buffer_write_sync(struct buffer_head
*bh
, int uptodate
)
1846 char b
[BDEVNAME_SIZE
];
1849 set_buffer_uptodate(bh
);
1851 if (!buffer_eopnotsupp(bh
) && printk_ratelimit()) {
1852 printk(KERN_WARNING
"lost page write due to "
1853 "I/O error on %s\n",
1854 bdevname(bh
->b_bdev
, b
));
1856 /* note, we dont' set_buffer_write_io_error because we have
1857 * our own ways of dealing with the IO errors
1859 clear_buffer_uptodate(bh
);
1865 static int write_all_supers(struct btrfs_root
*root
)
1867 struct list_head
*cur
;
1868 struct list_head
*head
= &root
->fs_info
->fs_devices
->devices
;
1869 struct btrfs_device
*dev
;
1870 struct btrfs_super_block
*sb
;
1871 struct btrfs_dev_item
*dev_item
;
1872 struct buffer_head
*bh
;
1876 int total_errors
= 0;
1880 max_errors
= btrfs_super_num_devices(&root
->fs_info
->super_copy
) - 1;
1881 do_barriers
= !btrfs_test_opt(root
, NOBARRIER
);
1883 sb
= &root
->fs_info
->super_for_commit
;
1884 dev_item
= &sb
->dev_item
;
1885 list_for_each(cur
, head
) {
1886 dev
= list_entry(cur
, struct btrfs_device
, dev_list
);
1891 if (!dev
->in_fs_metadata
|| !dev
->writeable
)
1894 btrfs_set_stack_device_generation(dev_item
, 0);
1895 btrfs_set_stack_device_type(dev_item
, dev
->type
);
1896 btrfs_set_stack_device_id(dev_item
, dev
->devid
);
1897 btrfs_set_stack_device_total_bytes(dev_item
, dev
->total_bytes
);
1898 btrfs_set_stack_device_bytes_used(dev_item
, dev
->bytes_used
);
1899 btrfs_set_stack_device_io_align(dev_item
, dev
->io_align
);
1900 btrfs_set_stack_device_io_width(dev_item
, dev
->io_width
);
1901 btrfs_set_stack_device_sector_size(dev_item
, dev
->sector_size
);
1902 memcpy(dev_item
->uuid
, dev
->uuid
, BTRFS_UUID_SIZE
);
1903 memcpy(dev_item
->fsid
, dev
->fs_devices
->fsid
, BTRFS_UUID_SIZE
);
1904 flags
= btrfs_super_flags(sb
);
1905 btrfs_set_super_flags(sb
, flags
| BTRFS_HEADER_FLAG_WRITTEN
);
1909 crc
= btrfs_csum_data(root
, (char *)sb
+ BTRFS_CSUM_SIZE
, crc
,
1910 BTRFS_SUPER_INFO_SIZE
- BTRFS_CSUM_SIZE
);
1911 btrfs_csum_final(crc
, sb
->csum
);
1913 bh
= __getblk(dev
->bdev
, BTRFS_SUPER_INFO_OFFSET
/ 4096,
1914 BTRFS_SUPER_INFO_SIZE
);
1916 memcpy(bh
->b_data
, sb
, BTRFS_SUPER_INFO_SIZE
);
1917 dev
->pending_io
= bh
;
1920 set_buffer_uptodate(bh
);
1922 bh
->b_end_io
= btrfs_end_buffer_write_sync
;
1924 if (do_barriers
&& dev
->barriers
) {
1925 ret
= submit_bh(WRITE_BARRIER
, bh
);
1926 if (ret
== -EOPNOTSUPP
) {
1927 printk("btrfs: disabling barriers on dev %s\n",
1929 set_buffer_uptodate(bh
);
1933 ret
= submit_bh(WRITE
, bh
);
1936 ret
= submit_bh(WRITE
, bh
);
1941 if (total_errors
> max_errors
) {
1942 printk("btrfs: %d errors while writing supers\n", total_errors
);
1947 list_for_each(cur
, head
) {
1948 dev
= list_entry(cur
, struct btrfs_device
, dev_list
);
1951 if (!dev
->in_fs_metadata
|| !dev
->writeable
)
1954 BUG_ON(!dev
->pending_io
);
1955 bh
= dev
->pending_io
;
1957 if (!buffer_uptodate(dev
->pending_io
)) {
1958 if (do_barriers
&& dev
->barriers
) {
1959 printk("btrfs: disabling barriers on dev %s\n",
1961 set_buffer_uptodate(bh
);
1965 ret
= submit_bh(WRITE
, bh
);
1968 if (!buffer_uptodate(bh
))
1975 dev
->pending_io
= NULL
;
1978 if (total_errors
> max_errors
) {
1979 printk("btrfs: %d errors while writing supers\n", total_errors
);
1985 int write_ctree_super(struct btrfs_trans_handle
*trans
, struct btrfs_root
1990 ret
= write_all_supers(root
);
1994 int btrfs_free_fs_root(struct btrfs_fs_info
*fs_info
, struct btrfs_root
*root
)
1996 radix_tree_delete(&fs_info
->fs_roots_radix
,
1997 (unsigned long)root
->root_key
.objectid
);
1998 if (root
->anon_super
.s_dev
) {
1999 down_write(&root
->anon_super
.s_umount
);
2000 kill_anon_super(&root
->anon_super
);
2004 btrfs_sysfs_del_root(root
);
2007 free_extent_buffer(root
->node
);
2008 if (root
->commit_root
)
2009 free_extent_buffer(root
->commit_root
);
2016 static int del_fs_roots(struct btrfs_fs_info
*fs_info
)
2019 struct btrfs_root
*gang
[8];
2023 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
2028 for (i
= 0; i
< ret
; i
++)
2029 btrfs_free_fs_root(fs_info
, gang
[i
]);
2034 int btrfs_cleanup_fs_roots(struct btrfs_fs_info
*fs_info
)
2036 u64 root_objectid
= 0;
2037 struct btrfs_root
*gang
[8];
2042 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
2043 (void **)gang
, root_objectid
,
2047 for (i
= 0; i
< ret
; i
++) {
2048 root_objectid
= gang
[i
]->root_key
.objectid
;
2049 ret
= btrfs_find_dead_roots(fs_info
->tree_root
,
2050 root_objectid
, gang
[i
]);
2052 btrfs_orphan_cleanup(gang
[i
]);
2059 int btrfs_commit_super(struct btrfs_root
*root
)
2061 struct btrfs_trans_handle
*trans
;
2064 mutex_lock(&root
->fs_info
->cleaner_mutex
);
2065 btrfs_clean_old_snapshots(root
);
2066 mutex_unlock(&root
->fs_info
->cleaner_mutex
);
2067 trans
= btrfs_start_transaction(root
, 1);
2068 ret
= btrfs_commit_transaction(trans
, root
);
2070 /* run commit again to drop the original snapshot */
2071 trans
= btrfs_start_transaction(root
, 1);
2072 btrfs_commit_transaction(trans
, root
);
2073 ret
= btrfs_write_and_wait_transaction(NULL
, root
);
2076 ret
= write_ctree_super(NULL
, root
);
2080 int close_ctree(struct btrfs_root
*root
)
2082 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
2085 fs_info
->closing
= 1;
2088 kthread_stop(root
->fs_info
->transaction_kthread
);
2089 kthread_stop(root
->fs_info
->cleaner_kthread
);
2091 if (!(fs_info
->sb
->s_flags
& MS_RDONLY
)) {
2092 ret
= btrfs_commit_super(root
);
2094 printk("btrfs: commit super returns %d\n", ret
);
2098 if (fs_info
->delalloc_bytes
) {
2099 printk("btrfs: at unmount delalloc count %Lu\n",
2100 fs_info
->delalloc_bytes
);
2102 if (fs_info
->total_ref_cache_size
) {
2103 printk("btrfs: at umount reference cache size %Lu\n",
2104 fs_info
->total_ref_cache_size
);
2107 if (fs_info
->extent_root
->node
)
2108 free_extent_buffer(fs_info
->extent_root
->node
);
2110 if (fs_info
->tree_root
->node
)
2111 free_extent_buffer(fs_info
->tree_root
->node
);
2113 if (root
->fs_info
->chunk_root
->node
);
2114 free_extent_buffer(root
->fs_info
->chunk_root
->node
);
2116 if (root
->fs_info
->dev_root
->node
);
2117 free_extent_buffer(root
->fs_info
->dev_root
->node
);
2119 btrfs_free_block_groups(root
->fs_info
);
2121 del_fs_roots(fs_info
);
2123 iput(fs_info
->btree_inode
);
2125 btrfs_stop_workers(&fs_info
->fixup_workers
);
2126 btrfs_stop_workers(&fs_info
->delalloc_workers
);
2127 btrfs_stop_workers(&fs_info
->workers
);
2128 btrfs_stop_workers(&fs_info
->endio_workers
);
2129 btrfs_stop_workers(&fs_info
->endio_write_workers
);
2130 btrfs_stop_workers(&fs_info
->submit_workers
);
2133 while(!list_empty(&fs_info
->hashers
)) {
2134 struct btrfs_hasher
*hasher
;
2135 hasher
= list_entry(fs_info
->hashers
.next
, struct btrfs_hasher
,
2137 list_del(&hasher
->hashers
);
2138 crypto_free_hash(&fs_info
->hash_tfm
);
2142 btrfs_close_devices(fs_info
->fs_devices
);
2143 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
2145 bdi_destroy(&fs_info
->bdi
);
2147 kfree(fs_info
->extent_root
);
2148 kfree(fs_info
->tree_root
);
2149 kfree(fs_info
->chunk_root
);
2150 kfree(fs_info
->dev_root
);
2154 int btrfs_buffer_uptodate(struct extent_buffer
*buf
, u64 parent_transid
)
2157 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
2159 ret
= extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
, buf
);
2163 ret
= verify_parent_transid(&BTRFS_I(btree_inode
)->io_tree
, buf
,
2168 int btrfs_set_buffer_uptodate(struct extent_buffer
*buf
)
2170 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
2171 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
,
2175 void btrfs_mark_buffer_dirty(struct extent_buffer
*buf
)
2177 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
2178 u64 transid
= btrfs_header_generation(buf
);
2179 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
2181 WARN_ON(!btrfs_tree_locked(buf
));
2182 if (transid
!= root
->fs_info
->generation
) {
2183 printk(KERN_CRIT
"transid mismatch buffer %llu, found %Lu running %Lu\n",
2184 (unsigned long long)buf
->start
,
2185 transid
, root
->fs_info
->generation
);
2188 set_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
, buf
);
2191 void btrfs_btree_balance_dirty(struct btrfs_root
*root
, unsigned long nr
)
2194 * looks as though older kernels can get into trouble with
2195 * this code, they end up stuck in balance_dirty_pages forever
2197 struct extent_io_tree
*tree
;
2200 unsigned long thresh
= 32 * 1024 * 1024;
2201 tree
= &BTRFS_I(root
->fs_info
->btree_inode
)->io_tree
;
2203 if (current_is_pdflush() || current
->flags
& PF_MEMALLOC
)
2206 num_dirty
= count_range_bits(tree
, &start
, (u64
)-1,
2207 thresh
, EXTENT_DIRTY
);
2208 if (num_dirty
> thresh
) {
2209 balance_dirty_pages_ratelimited_nr(
2210 root
->fs_info
->btree_inode
->i_mapping
, 1);
2215 int btrfs_read_buffer(struct extent_buffer
*buf
, u64 parent_transid
)
2217 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
2219 ret
= btree_read_extent_buffer_pages(root
, buf
, 0, parent_transid
);
2221 buf
->flags
|= EXTENT_UPTODATE
;
2226 int btree_lock_page_hook(struct page
*page
)
2228 struct inode
*inode
= page
->mapping
->host
;
2229 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2230 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2231 struct extent_buffer
*eb
;
2233 u64 bytenr
= page_offset(page
);
2235 if (page
->private == EXTENT_PAGE_PRIVATE
)
2238 len
= page
->private >> 2;
2239 eb
= find_extent_buffer(io_tree
, bytenr
, len
, GFP_NOFS
);
2243 btrfs_tree_lock(eb
);
2244 spin_lock(&root
->fs_info
->hash_lock
);
2245 btrfs_set_header_flag(eb
, BTRFS_HEADER_FLAG_WRITTEN
);
2246 spin_unlock(&root
->fs_info
->hash_lock
);
2247 btrfs_tree_unlock(eb
);
2248 free_extent_buffer(eb
);
2254 static struct extent_io_ops btree_extent_io_ops
= {
2255 .write_cache_pages_lock_hook
= btree_lock_page_hook
,
2256 .readpage_end_io_hook
= btree_readpage_end_io_hook
,
2257 .submit_bio_hook
= btree_submit_bio_hook
,
2258 /* note we're sharing with inode.c for the merge bio hook */
2259 .merge_bio_hook
= btrfs_merge_bio_hook
,