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.
20 #include <linux/blkdev.h>
21 #include <linux/scatterlist.h>
22 #include <linux/swap.h>
23 #include <linux/radix-tree.h>
24 #include <linux/writeback.h>
25 #include <linux/buffer_head.h>
26 #include <linux/workqueue.h>
27 #include <linux/kthread.h>
28 #include <linux/freezer.h>
29 #include <linux/crc32c.h>
30 #include <linux/slab.h>
31 #include <linux/migrate.h>
32 #include <asm/unaligned.h>
36 #include "transaction.h"
37 #include "btrfs_inode.h"
39 #include "print-tree.h"
40 #include "async-thread.h"
43 #include "free-space-cache.h"
45 static struct extent_io_ops btree_extent_io_ops
;
46 static void end_workqueue_fn(struct btrfs_work
*work
);
47 static void free_fs_root(struct btrfs_root
*root
);
48 static void btrfs_check_super_valid(struct btrfs_fs_info
*fs_info
,
50 static int btrfs_destroy_ordered_operations(struct btrfs_root
*root
);
51 static int btrfs_destroy_ordered_extents(struct btrfs_root
*root
);
52 static int btrfs_destroy_delayed_refs(struct btrfs_transaction
*trans
,
53 struct btrfs_root
*root
);
54 static int btrfs_destroy_pending_snapshots(struct btrfs_transaction
*t
);
55 static int btrfs_destroy_delalloc_inodes(struct btrfs_root
*root
);
56 static int btrfs_destroy_marked_extents(struct btrfs_root
*root
,
57 struct extent_io_tree
*dirty_pages
,
59 static int btrfs_destroy_pinned_extent(struct btrfs_root
*root
,
60 struct extent_io_tree
*pinned_extents
);
61 static int btrfs_cleanup_transaction(struct btrfs_root
*root
);
64 * end_io_wq structs are used to do processing in task context when an IO is
65 * complete. This is used during reads to verify checksums, and it is used
66 * by writes to insert metadata for new file extents after IO is complete.
72 struct btrfs_fs_info
*info
;
75 struct list_head list
;
76 struct btrfs_work work
;
80 * async submit bios are used to offload expensive checksumming
81 * onto the worker threads. They checksum file and metadata bios
82 * just before they are sent down the IO stack.
84 struct async_submit_bio
{
87 struct list_head list
;
88 extent_submit_bio_hook_t
*submit_bio_start
;
89 extent_submit_bio_hook_t
*submit_bio_done
;
92 unsigned long bio_flags
;
94 * bio_offset is optional, can be used if the pages in the bio
95 * can't tell us where in the file the bio should go
98 struct btrfs_work work
;
101 /* These are used to set the lockdep class on the extent buffer locks.
102 * The class is set by the readpage_end_io_hook after the buffer has
103 * passed csum validation but before the pages are unlocked.
105 * The lockdep class is also set by btrfs_init_new_buffer on freshly
108 * The class is based on the level in the tree block, which allows lockdep
109 * to know that lower nodes nest inside the locks of higher nodes.
111 * We also add a check to make sure the highest level of the tree is
112 * the same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this
113 * code needs update as well.
115 #ifdef CONFIG_DEBUG_LOCK_ALLOC
116 # if BTRFS_MAX_LEVEL != 8
119 static struct lock_class_key btrfs_eb_class
[BTRFS_MAX_LEVEL
+ 1];
120 static const char *btrfs_eb_name
[BTRFS_MAX_LEVEL
+ 1] = {
130 /* highest possible level */
136 * extents on the btree inode are pretty simple, there's one extent
137 * that covers the entire device
139 static struct extent_map
*btree_get_extent(struct inode
*inode
,
140 struct page
*page
, size_t pg_offset
, u64 start
, u64 len
,
143 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
144 struct extent_map
*em
;
147 read_lock(&em_tree
->lock
);
148 em
= lookup_extent_mapping(em_tree
, start
, len
);
151 BTRFS_I(inode
)->root
->fs_info
->fs_devices
->latest_bdev
;
152 read_unlock(&em_tree
->lock
);
155 read_unlock(&em_tree
->lock
);
157 em
= alloc_extent_map();
159 em
= ERR_PTR(-ENOMEM
);
164 em
->block_len
= (u64
)-1;
166 em
->bdev
= BTRFS_I(inode
)->root
->fs_info
->fs_devices
->latest_bdev
;
168 write_lock(&em_tree
->lock
);
169 ret
= add_extent_mapping(em_tree
, em
);
170 if (ret
== -EEXIST
) {
171 u64 failed_start
= em
->start
;
172 u64 failed_len
= em
->len
;
175 em
= lookup_extent_mapping(em_tree
, start
, len
);
179 em
= lookup_extent_mapping(em_tree
, failed_start
,
187 write_unlock(&em_tree
->lock
);
195 u32
btrfs_csum_data(struct btrfs_root
*root
, char *data
, u32 seed
, size_t len
)
197 return crc32c(seed
, data
, len
);
200 void btrfs_csum_final(u32 crc
, char *result
)
202 put_unaligned_le32(~crc
, result
);
206 * compute the csum for a btree block, and either verify it or write it
207 * into the csum field of the block.
209 static int csum_tree_block(struct btrfs_root
*root
, struct extent_buffer
*buf
,
213 btrfs_super_csum_size(&root
->fs_info
->super_copy
);
216 unsigned long cur_len
;
217 unsigned long offset
= BTRFS_CSUM_SIZE
;
218 char *map_token
= NULL
;
220 unsigned long map_start
;
221 unsigned long map_len
;
224 unsigned long inline_result
;
226 len
= buf
->len
- offset
;
228 err
= map_private_extent_buffer(buf
, offset
, 32,
230 &map_start
, &map_len
, KM_USER0
);
233 cur_len
= min(len
, map_len
- (offset
- map_start
));
234 crc
= btrfs_csum_data(root
, kaddr
+ offset
- map_start
,
238 unmap_extent_buffer(buf
, map_token
, KM_USER0
);
240 if (csum_size
> sizeof(inline_result
)) {
241 result
= kzalloc(csum_size
* sizeof(char), GFP_NOFS
);
245 result
= (char *)&inline_result
;
248 btrfs_csum_final(crc
, result
);
251 if (memcmp_extent_buffer(buf
, result
, 0, csum_size
)) {
254 memcpy(&found
, result
, csum_size
);
256 read_extent_buffer(buf
, &val
, 0, csum_size
);
257 if (printk_ratelimit()) {
258 printk(KERN_INFO
"btrfs: %s checksum verify "
259 "failed on %llu wanted %X found %X "
261 root
->fs_info
->sb
->s_id
,
262 (unsigned long long)buf
->start
, val
, found
,
263 btrfs_header_level(buf
));
265 if (result
!= (char *)&inline_result
)
270 write_extent_buffer(buf
, result
, 0, csum_size
);
272 if (result
!= (char *)&inline_result
)
278 * we can't consider a given block up to date unless the transid of the
279 * block matches the transid in the parent node's pointer. This is how we
280 * detect blocks that either didn't get written at all or got written
281 * in the wrong place.
283 static int verify_parent_transid(struct extent_io_tree
*io_tree
,
284 struct extent_buffer
*eb
, u64 parent_transid
)
286 struct extent_state
*cached_state
= NULL
;
289 if (!parent_transid
|| btrfs_header_generation(eb
) == parent_transid
)
292 lock_extent_bits(io_tree
, eb
->start
, eb
->start
+ eb
->len
- 1,
293 0, &cached_state
, GFP_NOFS
);
294 if (extent_buffer_uptodate(io_tree
, eb
, cached_state
) &&
295 btrfs_header_generation(eb
) == parent_transid
) {
299 if (printk_ratelimit()) {
300 printk("parent transid verify failed on %llu wanted %llu "
302 (unsigned long long)eb
->start
,
303 (unsigned long long)parent_transid
,
304 (unsigned long long)btrfs_header_generation(eb
));
307 clear_extent_buffer_uptodate(io_tree
, eb
, &cached_state
);
309 unlock_extent_cached(io_tree
, eb
->start
, eb
->start
+ eb
->len
- 1,
310 &cached_state
, GFP_NOFS
);
315 * helper to read a given tree block, doing retries as required when
316 * the checksums don't match and we have alternate mirrors to try.
318 static int btree_read_extent_buffer_pages(struct btrfs_root
*root
,
319 struct extent_buffer
*eb
,
320 u64 start
, u64 parent_transid
)
322 struct extent_io_tree
*io_tree
;
327 clear_bit(EXTENT_BUFFER_CORRUPT
, &eb
->bflags
);
328 io_tree
= &BTRFS_I(root
->fs_info
->btree_inode
)->io_tree
;
330 ret
= read_extent_buffer_pages(io_tree
, eb
, start
, 1,
331 btree_get_extent
, mirror_num
);
333 !verify_parent_transid(io_tree
, eb
, parent_transid
))
337 * This buffer's crc is fine, but its contents are corrupted, so
338 * there is no reason to read the other copies, they won't be
341 if (test_bit(EXTENT_BUFFER_CORRUPT
, &eb
->bflags
))
344 num_copies
= btrfs_num_copies(&root
->fs_info
->mapping_tree
,
350 if (mirror_num
> num_copies
)
357 * checksum a dirty tree block before IO. This has extra checks to make sure
358 * we only fill in the checksum field in the first page of a multi-page block
361 static int csum_dirty_buffer(struct btrfs_root
*root
, struct page
*page
)
363 struct extent_io_tree
*tree
;
364 u64 start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
367 struct extent_buffer
*eb
;
370 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
372 if (page
->private == EXTENT_PAGE_PRIVATE
) {
376 if (!page
->private) {
380 len
= page
->private >> 2;
383 eb
= alloc_extent_buffer(tree
, start
, len
, page
);
388 ret
= btree_read_extent_buffer_pages(root
, eb
, start
+ PAGE_CACHE_SIZE
,
389 btrfs_header_generation(eb
));
391 WARN_ON(!btrfs_header_flag(eb
, BTRFS_HEADER_FLAG_WRITTEN
));
393 found_start
= btrfs_header_bytenr(eb
);
394 if (found_start
!= start
) {
398 if (eb
->first_page
!= page
) {
402 if (!PageUptodate(page
)) {
406 csum_tree_block(root
, eb
, 0);
408 free_extent_buffer(eb
);
413 static int check_tree_block_fsid(struct btrfs_root
*root
,
414 struct extent_buffer
*eb
)
416 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
417 u8 fsid
[BTRFS_UUID_SIZE
];
420 read_extent_buffer(eb
, fsid
, (unsigned long)btrfs_header_fsid(eb
),
423 if (!memcmp(fsid
, fs_devices
->fsid
, BTRFS_FSID_SIZE
)) {
427 fs_devices
= fs_devices
->seed
;
432 #define CORRUPT(reason, eb, root, slot) \
433 printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \
434 "root=%llu, slot=%d\n", reason, \
435 (unsigned long long)btrfs_header_bytenr(eb), \
436 (unsigned long long)root->objectid, slot)
438 static noinline
int check_leaf(struct btrfs_root
*root
,
439 struct extent_buffer
*leaf
)
441 struct btrfs_key key
;
442 struct btrfs_key leaf_key
;
443 u32 nritems
= btrfs_header_nritems(leaf
);
449 /* Check the 0 item */
450 if (btrfs_item_offset_nr(leaf
, 0) + btrfs_item_size_nr(leaf
, 0) !=
451 BTRFS_LEAF_DATA_SIZE(root
)) {
452 CORRUPT("invalid item offset size pair", leaf
, root
, 0);
457 * Check to make sure each items keys are in the correct order and their
458 * offsets make sense. We only have to loop through nritems-1 because
459 * we check the current slot against the next slot, which verifies the
460 * next slot's offset+size makes sense and that the current's slot
463 for (slot
= 0; slot
< nritems
- 1; slot
++) {
464 btrfs_item_key_to_cpu(leaf
, &leaf_key
, slot
);
465 btrfs_item_key_to_cpu(leaf
, &key
, slot
+ 1);
467 /* Make sure the keys are in the right order */
468 if (btrfs_comp_cpu_keys(&leaf_key
, &key
) >= 0) {
469 CORRUPT("bad key order", leaf
, root
, slot
);
474 * Make sure the offset and ends are right, remember that the
475 * item data starts at the end of the leaf and grows towards the
478 if (btrfs_item_offset_nr(leaf
, slot
) !=
479 btrfs_item_end_nr(leaf
, slot
+ 1)) {
480 CORRUPT("slot offset bad", leaf
, root
, slot
);
485 * Check to make sure that we don't point outside of the leaf,
486 * just incase all the items are consistent to eachother, but
487 * all point outside of the leaf.
489 if (btrfs_item_end_nr(leaf
, slot
) >
490 BTRFS_LEAF_DATA_SIZE(root
)) {
491 CORRUPT("slot end outside of leaf", leaf
, root
, slot
);
499 #ifdef CONFIG_DEBUG_LOCK_ALLOC
500 void btrfs_set_buffer_lockdep_class(struct extent_buffer
*eb
, int level
)
502 lockdep_set_class_and_name(&eb
->lock
,
503 &btrfs_eb_class
[level
],
504 btrfs_eb_name
[level
]);
508 static int btree_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
509 struct extent_state
*state
)
511 struct extent_io_tree
*tree
;
515 struct extent_buffer
*eb
;
516 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
519 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
520 if (page
->private == EXTENT_PAGE_PRIVATE
)
525 len
= page
->private >> 2;
528 eb
= alloc_extent_buffer(tree
, start
, len
, page
);
534 found_start
= btrfs_header_bytenr(eb
);
535 if (found_start
!= start
) {
536 if (printk_ratelimit()) {
537 printk(KERN_INFO
"btrfs bad tree block start "
539 (unsigned long long)found_start
,
540 (unsigned long long)eb
->start
);
545 if (eb
->first_page
!= page
) {
546 printk(KERN_INFO
"btrfs bad first page %lu %lu\n",
547 eb
->first_page
->index
, page
->index
);
552 if (check_tree_block_fsid(root
, eb
)) {
553 if (printk_ratelimit()) {
554 printk(KERN_INFO
"btrfs bad fsid on block %llu\n",
555 (unsigned long long)eb
->start
);
560 found_level
= btrfs_header_level(eb
);
562 btrfs_set_buffer_lockdep_class(eb
, found_level
);
564 ret
= csum_tree_block(root
, eb
, 1);
571 * If this is a leaf block and it is corrupt, set the corrupt bit so
572 * that we don't try and read the other copies of this block, just
575 if (found_level
== 0 && check_leaf(root
, eb
)) {
576 set_bit(EXTENT_BUFFER_CORRUPT
, &eb
->bflags
);
580 end
= min_t(u64
, eb
->len
, PAGE_CACHE_SIZE
);
581 end
= eb
->start
+ end
- 1;
583 free_extent_buffer(eb
);
588 static void end_workqueue_bio(struct bio
*bio
, int err
)
590 struct end_io_wq
*end_io_wq
= bio
->bi_private
;
591 struct btrfs_fs_info
*fs_info
;
593 fs_info
= end_io_wq
->info
;
594 end_io_wq
->error
= err
;
595 end_io_wq
->work
.func
= end_workqueue_fn
;
596 end_io_wq
->work
.flags
= 0;
598 if (bio
->bi_rw
& REQ_WRITE
) {
599 if (end_io_wq
->metadata
== 1)
600 btrfs_queue_worker(&fs_info
->endio_meta_write_workers
,
602 else if (end_io_wq
->metadata
== 2)
603 btrfs_queue_worker(&fs_info
->endio_freespace_worker
,
606 btrfs_queue_worker(&fs_info
->endio_write_workers
,
609 if (end_io_wq
->metadata
)
610 btrfs_queue_worker(&fs_info
->endio_meta_workers
,
613 btrfs_queue_worker(&fs_info
->endio_workers
,
619 * For the metadata arg you want
622 * 1 - if normal metadta
623 * 2 - if writing to the free space cache area
625 int btrfs_bio_wq_end_io(struct btrfs_fs_info
*info
, struct bio
*bio
,
628 struct end_io_wq
*end_io_wq
;
629 end_io_wq
= kmalloc(sizeof(*end_io_wq
), GFP_NOFS
);
633 end_io_wq
->private = bio
->bi_private
;
634 end_io_wq
->end_io
= bio
->bi_end_io
;
635 end_io_wq
->info
= info
;
636 end_io_wq
->error
= 0;
637 end_io_wq
->bio
= bio
;
638 end_io_wq
->metadata
= metadata
;
640 bio
->bi_private
= end_io_wq
;
641 bio
->bi_end_io
= end_workqueue_bio
;
645 unsigned long btrfs_async_submit_limit(struct btrfs_fs_info
*info
)
647 unsigned long limit
= min_t(unsigned long,
648 info
->workers
.max_workers
,
649 info
->fs_devices
->open_devices
);
653 static void run_one_async_start(struct btrfs_work
*work
)
655 struct async_submit_bio
*async
;
657 async
= container_of(work
, struct async_submit_bio
, work
);
658 async
->submit_bio_start(async
->inode
, async
->rw
, async
->bio
,
659 async
->mirror_num
, async
->bio_flags
,
663 static void run_one_async_done(struct btrfs_work
*work
)
665 struct btrfs_fs_info
*fs_info
;
666 struct async_submit_bio
*async
;
669 async
= container_of(work
, struct async_submit_bio
, work
);
670 fs_info
= BTRFS_I(async
->inode
)->root
->fs_info
;
672 limit
= btrfs_async_submit_limit(fs_info
);
673 limit
= limit
* 2 / 3;
675 atomic_dec(&fs_info
->nr_async_submits
);
677 if (atomic_read(&fs_info
->nr_async_submits
) < limit
&&
678 waitqueue_active(&fs_info
->async_submit_wait
))
679 wake_up(&fs_info
->async_submit_wait
);
681 async
->submit_bio_done(async
->inode
, async
->rw
, async
->bio
,
682 async
->mirror_num
, async
->bio_flags
,
686 static void run_one_async_free(struct btrfs_work
*work
)
688 struct async_submit_bio
*async
;
690 async
= container_of(work
, struct async_submit_bio
, work
);
694 int btrfs_wq_submit_bio(struct btrfs_fs_info
*fs_info
, struct inode
*inode
,
695 int rw
, struct bio
*bio
, int mirror_num
,
696 unsigned long bio_flags
,
698 extent_submit_bio_hook_t
*submit_bio_start
,
699 extent_submit_bio_hook_t
*submit_bio_done
)
701 struct async_submit_bio
*async
;
703 async
= kmalloc(sizeof(*async
), GFP_NOFS
);
707 async
->inode
= inode
;
710 async
->mirror_num
= mirror_num
;
711 async
->submit_bio_start
= submit_bio_start
;
712 async
->submit_bio_done
= submit_bio_done
;
714 async
->work
.func
= run_one_async_start
;
715 async
->work
.ordered_func
= run_one_async_done
;
716 async
->work
.ordered_free
= run_one_async_free
;
718 async
->work
.flags
= 0;
719 async
->bio_flags
= bio_flags
;
720 async
->bio_offset
= bio_offset
;
722 atomic_inc(&fs_info
->nr_async_submits
);
725 btrfs_set_work_high_prio(&async
->work
);
727 btrfs_queue_worker(&fs_info
->workers
, &async
->work
);
729 while (atomic_read(&fs_info
->async_submit_draining
) &&
730 atomic_read(&fs_info
->nr_async_submits
)) {
731 wait_event(fs_info
->async_submit_wait
,
732 (atomic_read(&fs_info
->nr_async_submits
) == 0));
738 static int btree_csum_one_bio(struct bio
*bio
)
740 struct bio_vec
*bvec
= bio
->bi_io_vec
;
742 struct btrfs_root
*root
;
744 WARN_ON(bio
->bi_vcnt
<= 0);
745 while (bio_index
< bio
->bi_vcnt
) {
746 root
= BTRFS_I(bvec
->bv_page
->mapping
->host
)->root
;
747 csum_dirty_buffer(root
, bvec
->bv_page
);
754 static int __btree_submit_bio_start(struct inode
*inode
, int rw
,
755 struct bio
*bio
, int mirror_num
,
756 unsigned long bio_flags
,
760 * when we're called for a write, we're already in the async
761 * submission context. Just jump into btrfs_map_bio
763 btree_csum_one_bio(bio
);
767 static int __btree_submit_bio_done(struct inode
*inode
, int rw
, struct bio
*bio
,
768 int mirror_num
, unsigned long bio_flags
,
772 * when we're called for a write, we're already in the async
773 * submission context. Just jump into btrfs_map_bio
775 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
, mirror_num
, 1);
778 static int btree_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
779 int mirror_num
, unsigned long bio_flags
,
784 ret
= btrfs_bio_wq_end_io(BTRFS_I(inode
)->root
->fs_info
,
788 if (!(rw
& REQ_WRITE
)) {
790 * called for a read, do the setup so that checksum validation
791 * can happen in the async kernel threads
793 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
,
798 * kthread helpers are used to submit writes so that checksumming
799 * can happen in parallel across all CPUs
801 return btrfs_wq_submit_bio(BTRFS_I(inode
)->root
->fs_info
,
802 inode
, rw
, bio
, mirror_num
, 0,
804 __btree_submit_bio_start
,
805 __btree_submit_bio_done
);
808 #ifdef CONFIG_MIGRATION
809 static int btree_migratepage(struct address_space
*mapping
,
810 struct page
*newpage
, struct page
*page
)
813 * we can't safely write a btree page from here,
814 * we haven't done the locking hook
819 * Buffers may be managed in a filesystem specific way.
820 * We must have no buffers or drop them.
822 if (page_has_private(page
) &&
823 !try_to_release_page(page
, GFP_KERNEL
))
825 return migrate_page(mapping
, newpage
, page
);
829 static int btree_writepage(struct page
*page
, struct writeback_control
*wbc
)
831 struct extent_io_tree
*tree
;
832 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
833 struct extent_buffer
*eb
;
836 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
837 if (!(current
->flags
& PF_MEMALLOC
)) {
838 return extent_write_full_page(tree
, page
,
839 btree_get_extent
, wbc
);
842 redirty_page_for_writepage(wbc
, page
);
843 eb
= btrfs_find_tree_block(root
, page_offset(page
), PAGE_CACHE_SIZE
);
846 was_dirty
= test_and_set_bit(EXTENT_BUFFER_DIRTY
, &eb
->bflags
);
848 spin_lock(&root
->fs_info
->delalloc_lock
);
849 root
->fs_info
->dirty_metadata_bytes
+= PAGE_CACHE_SIZE
;
850 spin_unlock(&root
->fs_info
->delalloc_lock
);
852 free_extent_buffer(eb
);
858 static int btree_writepages(struct address_space
*mapping
,
859 struct writeback_control
*wbc
)
861 struct extent_io_tree
*tree
;
862 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
863 if (wbc
->sync_mode
== WB_SYNC_NONE
) {
864 struct btrfs_root
*root
= BTRFS_I(mapping
->host
)->root
;
866 unsigned long thresh
= 32 * 1024 * 1024;
868 if (wbc
->for_kupdate
)
871 /* this is a bit racy, but that's ok */
872 num_dirty
= root
->fs_info
->dirty_metadata_bytes
;
873 if (num_dirty
< thresh
)
876 return extent_writepages(tree
, mapping
, btree_get_extent
, wbc
);
879 static int btree_readpage(struct file
*file
, struct page
*page
)
881 struct extent_io_tree
*tree
;
882 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
883 return extent_read_full_page(tree
, page
, btree_get_extent
);
886 static int btree_releasepage(struct page
*page
, gfp_t gfp_flags
)
888 struct extent_io_tree
*tree
;
889 struct extent_map_tree
*map
;
892 if (PageWriteback(page
) || PageDirty(page
))
895 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
896 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
898 ret
= try_release_extent_state(map
, tree
, page
, gfp_flags
);
902 ret
= try_release_extent_buffer(tree
, page
);
904 ClearPagePrivate(page
);
905 set_page_private(page
, 0);
906 page_cache_release(page
);
912 static void btree_invalidatepage(struct page
*page
, unsigned long offset
)
914 struct extent_io_tree
*tree
;
915 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
916 extent_invalidatepage(tree
, page
, offset
);
917 btree_releasepage(page
, GFP_NOFS
);
918 if (PagePrivate(page
)) {
919 printk(KERN_WARNING
"btrfs warning page private not zero "
920 "on page %llu\n", (unsigned long long)page_offset(page
));
921 ClearPagePrivate(page
);
922 set_page_private(page
, 0);
923 page_cache_release(page
);
927 static const struct address_space_operations btree_aops
= {
928 .readpage
= btree_readpage
,
929 .writepage
= btree_writepage
,
930 .writepages
= btree_writepages
,
931 .releasepage
= btree_releasepage
,
932 .invalidatepage
= btree_invalidatepage
,
933 #ifdef CONFIG_MIGRATION
934 .migratepage
= btree_migratepage
,
938 int readahead_tree_block(struct btrfs_root
*root
, u64 bytenr
, u32 blocksize
,
941 struct extent_buffer
*buf
= NULL
;
942 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
945 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
948 read_extent_buffer_pages(&BTRFS_I(btree_inode
)->io_tree
,
949 buf
, 0, 0, btree_get_extent
, 0);
950 free_extent_buffer(buf
);
954 struct extent_buffer
*btrfs_find_tree_block(struct btrfs_root
*root
,
955 u64 bytenr
, u32 blocksize
)
957 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
958 struct extent_buffer
*eb
;
959 eb
= find_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
964 struct extent_buffer
*btrfs_find_create_tree_block(struct btrfs_root
*root
,
965 u64 bytenr
, u32 blocksize
)
967 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
968 struct extent_buffer
*eb
;
970 eb
= alloc_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
971 bytenr
, blocksize
, NULL
);
976 int btrfs_write_tree_block(struct extent_buffer
*buf
)
978 return filemap_fdatawrite_range(buf
->first_page
->mapping
, buf
->start
,
979 buf
->start
+ buf
->len
- 1);
982 int btrfs_wait_tree_block_writeback(struct extent_buffer
*buf
)
984 return filemap_fdatawait_range(buf
->first_page
->mapping
,
985 buf
->start
, buf
->start
+ buf
->len
- 1);
988 struct extent_buffer
*read_tree_block(struct btrfs_root
*root
, u64 bytenr
,
989 u32 blocksize
, u64 parent_transid
)
991 struct extent_buffer
*buf
= NULL
;
994 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
998 ret
= btree_read_extent_buffer_pages(root
, buf
, 0, parent_transid
);
1001 set_bit(EXTENT_BUFFER_UPTODATE
, &buf
->bflags
);
1006 int clean_tree_block(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
1007 struct extent_buffer
*buf
)
1009 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1010 if (btrfs_header_generation(buf
) ==
1011 root
->fs_info
->running_transaction
->transid
) {
1012 btrfs_assert_tree_locked(buf
);
1014 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY
, &buf
->bflags
)) {
1015 spin_lock(&root
->fs_info
->delalloc_lock
);
1016 if (root
->fs_info
->dirty_metadata_bytes
>= buf
->len
)
1017 root
->fs_info
->dirty_metadata_bytes
-= buf
->len
;
1020 spin_unlock(&root
->fs_info
->delalloc_lock
);
1023 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1024 btrfs_set_lock_blocking(buf
);
1025 clear_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
,
1031 static int __setup_root(u32 nodesize
, u32 leafsize
, u32 sectorsize
,
1032 u32 stripesize
, struct btrfs_root
*root
,
1033 struct btrfs_fs_info
*fs_info
,
1037 root
->commit_root
= NULL
;
1038 root
->sectorsize
= sectorsize
;
1039 root
->nodesize
= nodesize
;
1040 root
->leafsize
= leafsize
;
1041 root
->stripesize
= stripesize
;
1043 root
->track_dirty
= 0;
1045 root
->orphan_item_inserted
= 0;
1046 root
->orphan_cleanup_state
= 0;
1048 root
->fs_info
= fs_info
;
1049 root
->objectid
= objectid
;
1050 root
->last_trans
= 0;
1051 root
->highest_objectid
= 0;
1054 root
->inode_tree
= RB_ROOT
;
1055 root
->block_rsv
= NULL
;
1056 root
->orphan_block_rsv
= NULL
;
1058 INIT_LIST_HEAD(&root
->dirty_list
);
1059 INIT_LIST_HEAD(&root
->orphan_list
);
1060 INIT_LIST_HEAD(&root
->root_list
);
1061 spin_lock_init(&root
->node_lock
);
1062 spin_lock_init(&root
->orphan_lock
);
1063 spin_lock_init(&root
->inode_lock
);
1064 spin_lock_init(&root
->accounting_lock
);
1065 mutex_init(&root
->objectid_mutex
);
1066 mutex_init(&root
->log_mutex
);
1067 init_waitqueue_head(&root
->log_writer_wait
);
1068 init_waitqueue_head(&root
->log_commit_wait
[0]);
1069 init_waitqueue_head(&root
->log_commit_wait
[1]);
1070 atomic_set(&root
->log_commit
[0], 0);
1071 atomic_set(&root
->log_commit
[1], 0);
1072 atomic_set(&root
->log_writers
, 0);
1073 root
->log_batch
= 0;
1074 root
->log_transid
= 0;
1075 root
->last_log_commit
= 0;
1076 extent_io_tree_init(&root
->dirty_log_pages
,
1077 fs_info
->btree_inode
->i_mapping
);
1079 memset(&root
->root_key
, 0, sizeof(root
->root_key
));
1080 memset(&root
->root_item
, 0, sizeof(root
->root_item
));
1081 memset(&root
->defrag_progress
, 0, sizeof(root
->defrag_progress
));
1082 memset(&root
->root_kobj
, 0, sizeof(root
->root_kobj
));
1083 root
->defrag_trans_start
= fs_info
->generation
;
1084 init_completion(&root
->kobj_unregister
);
1085 root
->defrag_running
= 0;
1086 root
->root_key
.objectid
= objectid
;
1087 root
->anon_super
.s_root
= NULL
;
1088 root
->anon_super
.s_dev
= 0;
1089 INIT_LIST_HEAD(&root
->anon_super
.s_list
);
1090 INIT_LIST_HEAD(&root
->anon_super
.s_instances
);
1091 init_rwsem(&root
->anon_super
.s_umount
);
1096 static int find_and_setup_root(struct btrfs_root
*tree_root
,
1097 struct btrfs_fs_info
*fs_info
,
1099 struct btrfs_root
*root
)
1105 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
1106 tree_root
->sectorsize
, tree_root
->stripesize
,
1107 root
, fs_info
, objectid
);
1108 ret
= btrfs_find_last_root(tree_root
, objectid
,
1109 &root
->root_item
, &root
->root_key
);
1114 generation
= btrfs_root_generation(&root
->root_item
);
1115 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
1116 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
1117 blocksize
, generation
);
1118 if (!root
->node
|| !btrfs_buffer_uptodate(root
->node
, generation
)) {
1119 free_extent_buffer(root
->node
);
1122 root
->commit_root
= btrfs_root_node(root
);
1126 static struct btrfs_root
*alloc_log_tree(struct btrfs_trans_handle
*trans
,
1127 struct btrfs_fs_info
*fs_info
)
1129 struct btrfs_root
*root
;
1130 struct btrfs_root
*tree_root
= fs_info
->tree_root
;
1131 struct extent_buffer
*leaf
;
1133 root
= kzalloc(sizeof(*root
), GFP_NOFS
);
1135 return ERR_PTR(-ENOMEM
);
1137 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
1138 tree_root
->sectorsize
, tree_root
->stripesize
,
1139 root
, fs_info
, BTRFS_TREE_LOG_OBJECTID
);
1141 root
->root_key
.objectid
= BTRFS_TREE_LOG_OBJECTID
;
1142 root
->root_key
.type
= BTRFS_ROOT_ITEM_KEY
;
1143 root
->root_key
.offset
= BTRFS_TREE_LOG_OBJECTID
;
1145 * log trees do not get reference counted because they go away
1146 * before a real commit is actually done. They do store pointers
1147 * to file data extents, and those reference counts still get
1148 * updated (along with back refs to the log tree).
1152 leaf
= btrfs_alloc_free_block(trans
, root
, root
->leafsize
, 0,
1153 BTRFS_TREE_LOG_OBJECTID
, NULL
, 0, 0, 0);
1156 return ERR_CAST(leaf
);
1159 memset_extent_buffer(leaf
, 0, 0, sizeof(struct btrfs_header
));
1160 btrfs_set_header_bytenr(leaf
, leaf
->start
);
1161 btrfs_set_header_generation(leaf
, trans
->transid
);
1162 btrfs_set_header_backref_rev(leaf
, BTRFS_MIXED_BACKREF_REV
);
1163 btrfs_set_header_owner(leaf
, BTRFS_TREE_LOG_OBJECTID
);
1166 write_extent_buffer(root
->node
, root
->fs_info
->fsid
,
1167 (unsigned long)btrfs_header_fsid(root
->node
),
1169 btrfs_mark_buffer_dirty(root
->node
);
1170 btrfs_tree_unlock(root
->node
);
1174 int btrfs_init_log_root_tree(struct btrfs_trans_handle
*trans
,
1175 struct btrfs_fs_info
*fs_info
)
1177 struct btrfs_root
*log_root
;
1179 log_root
= alloc_log_tree(trans
, fs_info
);
1180 if (IS_ERR(log_root
))
1181 return PTR_ERR(log_root
);
1182 WARN_ON(fs_info
->log_root_tree
);
1183 fs_info
->log_root_tree
= log_root
;
1187 int btrfs_add_log_tree(struct btrfs_trans_handle
*trans
,
1188 struct btrfs_root
*root
)
1190 struct btrfs_root
*log_root
;
1191 struct btrfs_inode_item
*inode_item
;
1193 log_root
= alloc_log_tree(trans
, root
->fs_info
);
1194 if (IS_ERR(log_root
))
1195 return PTR_ERR(log_root
);
1197 log_root
->last_trans
= trans
->transid
;
1198 log_root
->root_key
.offset
= root
->root_key
.objectid
;
1200 inode_item
= &log_root
->root_item
.inode
;
1201 inode_item
->generation
= cpu_to_le64(1);
1202 inode_item
->size
= cpu_to_le64(3);
1203 inode_item
->nlink
= cpu_to_le32(1);
1204 inode_item
->nbytes
= cpu_to_le64(root
->leafsize
);
1205 inode_item
->mode
= cpu_to_le32(S_IFDIR
| 0755);
1207 btrfs_set_root_node(&log_root
->root_item
, log_root
->node
);
1209 WARN_ON(root
->log_root
);
1210 root
->log_root
= log_root
;
1211 root
->log_transid
= 0;
1212 root
->last_log_commit
= 0;
1216 struct btrfs_root
*btrfs_read_fs_root_no_radix(struct btrfs_root
*tree_root
,
1217 struct btrfs_key
*location
)
1219 struct btrfs_root
*root
;
1220 struct btrfs_fs_info
*fs_info
= tree_root
->fs_info
;
1221 struct btrfs_path
*path
;
1222 struct extent_buffer
*l
;
1227 root
= kzalloc(sizeof(*root
), GFP_NOFS
);
1229 return ERR_PTR(-ENOMEM
);
1230 if (location
->offset
== (u64
)-1) {
1231 ret
= find_and_setup_root(tree_root
, fs_info
,
1232 location
->objectid
, root
);
1235 return ERR_PTR(ret
);
1240 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
1241 tree_root
->sectorsize
, tree_root
->stripesize
,
1242 root
, fs_info
, location
->objectid
);
1244 path
= btrfs_alloc_path();
1247 return ERR_PTR(-ENOMEM
);
1249 ret
= btrfs_search_slot(NULL
, tree_root
, location
, path
, 0, 0);
1252 read_extent_buffer(l
, &root
->root_item
,
1253 btrfs_item_ptr_offset(l
, path
->slots
[0]),
1254 sizeof(root
->root_item
));
1255 memcpy(&root
->root_key
, location
, sizeof(*location
));
1257 btrfs_free_path(path
);
1262 return ERR_PTR(ret
);
1265 generation
= btrfs_root_generation(&root
->root_item
);
1266 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
1267 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
1268 blocksize
, generation
);
1269 root
->commit_root
= btrfs_root_node(root
);
1270 BUG_ON(!root
->node
);
1272 if (location
->objectid
!= BTRFS_TREE_LOG_OBJECTID
) {
1274 btrfs_check_and_init_root_item(&root
->root_item
);
1280 struct btrfs_root
*btrfs_read_fs_root_no_name(struct btrfs_fs_info
*fs_info
,
1281 struct btrfs_key
*location
)
1283 struct btrfs_root
*root
;
1286 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1287 return fs_info
->tree_root
;
1288 if (location
->objectid
== BTRFS_EXTENT_TREE_OBJECTID
)
1289 return fs_info
->extent_root
;
1290 if (location
->objectid
== BTRFS_CHUNK_TREE_OBJECTID
)
1291 return fs_info
->chunk_root
;
1292 if (location
->objectid
== BTRFS_DEV_TREE_OBJECTID
)
1293 return fs_info
->dev_root
;
1294 if (location
->objectid
== BTRFS_CSUM_TREE_OBJECTID
)
1295 return fs_info
->csum_root
;
1297 spin_lock(&fs_info
->fs_roots_radix_lock
);
1298 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
1299 (unsigned long)location
->objectid
);
1300 spin_unlock(&fs_info
->fs_roots_radix_lock
);
1304 root
= btrfs_read_fs_root_no_radix(fs_info
->tree_root
, location
);
1308 set_anon_super(&root
->anon_super
, NULL
);
1310 if (btrfs_root_refs(&root
->root_item
) == 0) {
1315 ret
= btrfs_find_orphan_item(fs_info
->tree_root
, location
->objectid
);
1319 root
->orphan_item_inserted
= 1;
1321 ret
= radix_tree_preload(GFP_NOFS
& ~__GFP_HIGHMEM
);
1325 spin_lock(&fs_info
->fs_roots_radix_lock
);
1326 ret
= radix_tree_insert(&fs_info
->fs_roots_radix
,
1327 (unsigned long)root
->root_key
.objectid
,
1332 spin_unlock(&fs_info
->fs_roots_radix_lock
);
1333 radix_tree_preload_end();
1335 if (ret
== -EEXIST
) {
1342 ret
= btrfs_find_dead_roots(fs_info
->tree_root
,
1343 root
->root_key
.objectid
);
1348 return ERR_PTR(ret
);
1351 static int btrfs_congested_fn(void *congested_data
, int bdi_bits
)
1353 struct btrfs_fs_info
*info
= (struct btrfs_fs_info
*)congested_data
;
1355 struct btrfs_device
*device
;
1356 struct backing_dev_info
*bdi
;
1358 list_for_each_entry(device
, &info
->fs_devices
->devices
, dev_list
) {
1361 bdi
= blk_get_backing_dev_info(device
->bdev
);
1362 if (bdi
&& bdi_congested(bdi
, bdi_bits
)) {
1371 * If this fails, caller must call bdi_destroy() to get rid of the
1374 static int setup_bdi(struct btrfs_fs_info
*info
, struct backing_dev_info
*bdi
)
1378 bdi
->capabilities
= BDI_CAP_MAP_COPY
;
1379 err
= bdi_setup_and_register(bdi
, "btrfs", BDI_CAP_MAP_COPY
);
1383 bdi
->ra_pages
= default_backing_dev_info
.ra_pages
;
1384 bdi
->congested_fn
= btrfs_congested_fn
;
1385 bdi
->congested_data
= info
;
1389 static int bio_ready_for_csum(struct bio
*bio
)
1395 struct extent_io_tree
*io_tree
= NULL
;
1396 struct bio_vec
*bvec
;
1400 bio_for_each_segment(bvec
, bio
, i
) {
1401 page
= bvec
->bv_page
;
1402 if (page
->private == EXTENT_PAGE_PRIVATE
) {
1403 length
+= bvec
->bv_len
;
1406 if (!page
->private) {
1407 length
+= bvec
->bv_len
;
1410 length
= bvec
->bv_len
;
1411 buf_len
= page
->private >> 2;
1412 start
= page_offset(page
) + bvec
->bv_offset
;
1413 io_tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
1415 /* are we fully contained in this bio? */
1416 if (buf_len
<= length
)
1419 ret
= extent_range_uptodate(io_tree
, start
+ length
,
1420 start
+ buf_len
- 1);
1425 * called by the kthread helper functions to finally call the bio end_io
1426 * functions. This is where read checksum verification actually happens
1428 static void end_workqueue_fn(struct btrfs_work
*work
)
1431 struct end_io_wq
*end_io_wq
;
1432 struct btrfs_fs_info
*fs_info
;
1435 end_io_wq
= container_of(work
, struct end_io_wq
, work
);
1436 bio
= end_io_wq
->bio
;
1437 fs_info
= end_io_wq
->info
;
1439 /* metadata bio reads are special because the whole tree block must
1440 * be checksummed at once. This makes sure the entire block is in
1441 * ram and up to date before trying to verify things. For
1442 * blocksize <= pagesize, it is basically a noop
1444 if (!(bio
->bi_rw
& REQ_WRITE
) && end_io_wq
->metadata
&&
1445 !bio_ready_for_csum(bio
)) {
1446 btrfs_queue_worker(&fs_info
->endio_meta_workers
,
1450 error
= end_io_wq
->error
;
1451 bio
->bi_private
= end_io_wq
->private;
1452 bio
->bi_end_io
= end_io_wq
->end_io
;
1454 bio_endio(bio
, error
);
1457 static int cleaner_kthread(void *arg
)
1459 struct btrfs_root
*root
= arg
;
1462 vfs_check_frozen(root
->fs_info
->sb
, SB_FREEZE_WRITE
);
1464 if (!(root
->fs_info
->sb
->s_flags
& MS_RDONLY
) &&
1465 mutex_trylock(&root
->fs_info
->cleaner_mutex
)) {
1466 btrfs_run_delayed_iputs(root
);
1467 btrfs_clean_old_snapshots(root
);
1468 mutex_unlock(&root
->fs_info
->cleaner_mutex
);
1471 if (freezing(current
)) {
1474 set_current_state(TASK_INTERRUPTIBLE
);
1475 if (!kthread_should_stop())
1477 __set_current_state(TASK_RUNNING
);
1479 } while (!kthread_should_stop());
1483 static int transaction_kthread(void *arg
)
1485 struct btrfs_root
*root
= arg
;
1486 struct btrfs_trans_handle
*trans
;
1487 struct btrfs_transaction
*cur
;
1490 unsigned long delay
;
1495 vfs_check_frozen(root
->fs_info
->sb
, SB_FREEZE_WRITE
);
1496 mutex_lock(&root
->fs_info
->transaction_kthread_mutex
);
1498 spin_lock(&root
->fs_info
->new_trans_lock
);
1499 cur
= root
->fs_info
->running_transaction
;
1501 spin_unlock(&root
->fs_info
->new_trans_lock
);
1505 now
= get_seconds();
1506 if (!cur
->blocked
&&
1507 (now
< cur
->start_time
|| now
- cur
->start_time
< 30)) {
1508 spin_unlock(&root
->fs_info
->new_trans_lock
);
1512 transid
= cur
->transid
;
1513 spin_unlock(&root
->fs_info
->new_trans_lock
);
1515 trans
= btrfs_join_transaction(root
, 1);
1516 BUG_ON(IS_ERR(trans
));
1517 if (transid
== trans
->transid
) {
1518 ret
= btrfs_commit_transaction(trans
, root
);
1521 btrfs_end_transaction(trans
, root
);
1524 wake_up_process(root
->fs_info
->cleaner_kthread
);
1525 mutex_unlock(&root
->fs_info
->transaction_kthread_mutex
);
1527 if (freezing(current
)) {
1530 set_current_state(TASK_INTERRUPTIBLE
);
1531 if (!kthread_should_stop() &&
1532 !btrfs_transaction_blocked(root
->fs_info
))
1533 schedule_timeout(delay
);
1534 __set_current_state(TASK_RUNNING
);
1536 } while (!kthread_should_stop());
1540 struct btrfs_root
*open_ctree(struct super_block
*sb
,
1541 struct btrfs_fs_devices
*fs_devices
,
1551 struct btrfs_key location
;
1552 struct buffer_head
*bh
;
1553 struct btrfs_root
*extent_root
= kzalloc(sizeof(struct btrfs_root
),
1555 struct btrfs_root
*csum_root
= kzalloc(sizeof(struct btrfs_root
),
1557 struct btrfs_root
*tree_root
= btrfs_sb(sb
);
1558 struct btrfs_fs_info
*fs_info
= NULL
;
1559 struct btrfs_root
*chunk_root
= kzalloc(sizeof(struct btrfs_root
),
1561 struct btrfs_root
*dev_root
= kzalloc(sizeof(struct btrfs_root
),
1563 struct btrfs_root
*log_tree_root
;
1568 struct btrfs_super_block
*disk_super
;
1570 if (!extent_root
|| !tree_root
|| !tree_root
->fs_info
||
1571 !chunk_root
|| !dev_root
|| !csum_root
) {
1575 fs_info
= tree_root
->fs_info
;
1577 ret
= init_srcu_struct(&fs_info
->subvol_srcu
);
1583 ret
= setup_bdi(fs_info
, &fs_info
->bdi
);
1589 fs_info
->btree_inode
= new_inode(sb
);
1590 if (!fs_info
->btree_inode
) {
1595 fs_info
->btree_inode
->i_mapping
->flags
&= ~__GFP_FS
;
1597 INIT_RADIX_TREE(&fs_info
->fs_roots_radix
, GFP_ATOMIC
);
1598 INIT_LIST_HEAD(&fs_info
->trans_list
);
1599 INIT_LIST_HEAD(&fs_info
->dead_roots
);
1600 INIT_LIST_HEAD(&fs_info
->delayed_iputs
);
1601 INIT_LIST_HEAD(&fs_info
->hashers
);
1602 INIT_LIST_HEAD(&fs_info
->delalloc_inodes
);
1603 INIT_LIST_HEAD(&fs_info
->ordered_operations
);
1604 INIT_LIST_HEAD(&fs_info
->caching_block_groups
);
1605 spin_lock_init(&fs_info
->delalloc_lock
);
1606 spin_lock_init(&fs_info
->new_trans_lock
);
1607 spin_lock_init(&fs_info
->ref_cache_lock
);
1608 spin_lock_init(&fs_info
->fs_roots_radix_lock
);
1609 spin_lock_init(&fs_info
->delayed_iput_lock
);
1611 init_completion(&fs_info
->kobj_unregister
);
1612 fs_info
->tree_root
= tree_root
;
1613 fs_info
->extent_root
= extent_root
;
1614 fs_info
->csum_root
= csum_root
;
1615 fs_info
->chunk_root
= chunk_root
;
1616 fs_info
->dev_root
= dev_root
;
1617 fs_info
->fs_devices
= fs_devices
;
1618 INIT_LIST_HEAD(&fs_info
->dirty_cowonly_roots
);
1619 INIT_LIST_HEAD(&fs_info
->space_info
);
1620 btrfs_mapping_init(&fs_info
->mapping_tree
);
1621 btrfs_init_block_rsv(&fs_info
->global_block_rsv
);
1622 btrfs_init_block_rsv(&fs_info
->delalloc_block_rsv
);
1623 btrfs_init_block_rsv(&fs_info
->trans_block_rsv
);
1624 btrfs_init_block_rsv(&fs_info
->chunk_block_rsv
);
1625 btrfs_init_block_rsv(&fs_info
->empty_block_rsv
);
1626 INIT_LIST_HEAD(&fs_info
->durable_block_rsv_list
);
1627 mutex_init(&fs_info
->durable_block_rsv_mutex
);
1628 atomic_set(&fs_info
->nr_async_submits
, 0);
1629 atomic_set(&fs_info
->async_delalloc_pages
, 0);
1630 atomic_set(&fs_info
->async_submit_draining
, 0);
1631 atomic_set(&fs_info
->nr_async_bios
, 0);
1633 fs_info
->max_inline
= 8192 * 1024;
1634 fs_info
->metadata_ratio
= 0;
1636 fs_info
->thread_pool_size
= min_t(unsigned long,
1637 num_online_cpus() + 2, 8);
1639 INIT_LIST_HEAD(&fs_info
->ordered_extents
);
1640 spin_lock_init(&fs_info
->ordered_extent_lock
);
1642 sb
->s_blocksize
= 4096;
1643 sb
->s_blocksize_bits
= blksize_bits(4096);
1644 sb
->s_bdi
= &fs_info
->bdi
;
1646 fs_info
->btree_inode
->i_ino
= BTRFS_BTREE_INODE_OBJECTID
;
1647 fs_info
->btree_inode
->i_nlink
= 1;
1649 * we set the i_size on the btree inode to the max possible int.
1650 * the real end of the address space is determined by all of
1651 * the devices in the system
1653 fs_info
->btree_inode
->i_size
= OFFSET_MAX
;
1654 fs_info
->btree_inode
->i_mapping
->a_ops
= &btree_aops
;
1655 fs_info
->btree_inode
->i_mapping
->backing_dev_info
= &fs_info
->bdi
;
1657 RB_CLEAR_NODE(&BTRFS_I(fs_info
->btree_inode
)->rb_node
);
1658 extent_io_tree_init(&BTRFS_I(fs_info
->btree_inode
)->io_tree
,
1659 fs_info
->btree_inode
->i_mapping
);
1660 extent_map_tree_init(&BTRFS_I(fs_info
->btree_inode
)->extent_tree
);
1662 BTRFS_I(fs_info
->btree_inode
)->io_tree
.ops
= &btree_extent_io_ops
;
1664 BTRFS_I(fs_info
->btree_inode
)->root
= tree_root
;
1665 memset(&BTRFS_I(fs_info
->btree_inode
)->location
, 0,
1666 sizeof(struct btrfs_key
));
1667 BTRFS_I(fs_info
->btree_inode
)->dummy_inode
= 1;
1668 insert_inode_hash(fs_info
->btree_inode
);
1670 spin_lock_init(&fs_info
->block_group_cache_lock
);
1671 fs_info
->block_group_cache_tree
= RB_ROOT
;
1673 extent_io_tree_init(&fs_info
->freed_extents
[0],
1674 fs_info
->btree_inode
->i_mapping
);
1675 extent_io_tree_init(&fs_info
->freed_extents
[1],
1676 fs_info
->btree_inode
->i_mapping
);
1677 fs_info
->pinned_extents
= &fs_info
->freed_extents
[0];
1678 fs_info
->do_barriers
= 1;
1681 mutex_init(&fs_info
->trans_mutex
);
1682 mutex_init(&fs_info
->ordered_operations_mutex
);
1683 mutex_init(&fs_info
->tree_log_mutex
);
1684 mutex_init(&fs_info
->chunk_mutex
);
1685 mutex_init(&fs_info
->transaction_kthread_mutex
);
1686 mutex_init(&fs_info
->cleaner_mutex
);
1687 mutex_init(&fs_info
->volume_mutex
);
1688 init_rwsem(&fs_info
->extent_commit_sem
);
1689 init_rwsem(&fs_info
->cleanup_work_sem
);
1690 init_rwsem(&fs_info
->subvol_sem
);
1692 btrfs_init_free_cluster(&fs_info
->meta_alloc_cluster
);
1693 btrfs_init_free_cluster(&fs_info
->data_alloc_cluster
);
1695 init_waitqueue_head(&fs_info
->transaction_throttle
);
1696 init_waitqueue_head(&fs_info
->transaction_wait
);
1697 init_waitqueue_head(&fs_info
->transaction_blocked_wait
);
1698 init_waitqueue_head(&fs_info
->async_submit_wait
);
1700 __setup_root(4096, 4096, 4096, 4096, tree_root
,
1701 fs_info
, BTRFS_ROOT_TREE_OBJECTID
);
1703 bh
= btrfs_read_dev_super(fs_devices
->latest_bdev
);
1709 memcpy(&fs_info
->super_copy
, bh
->b_data
, sizeof(fs_info
->super_copy
));
1710 memcpy(&fs_info
->super_for_commit
, &fs_info
->super_copy
,
1711 sizeof(fs_info
->super_for_commit
));
1714 memcpy(fs_info
->fsid
, fs_info
->super_copy
.fsid
, BTRFS_FSID_SIZE
);
1716 disk_super
= &fs_info
->super_copy
;
1717 if (!btrfs_super_root(disk_super
))
1720 /* check FS state, whether FS is broken. */
1721 fs_info
->fs_state
|= btrfs_super_flags(disk_super
);
1723 btrfs_check_super_valid(fs_info
, sb
->s_flags
& MS_RDONLY
);
1726 * In the long term, we'll store the compression type in the super
1727 * block, and it'll be used for per file compression control.
1729 fs_info
->compress_type
= BTRFS_COMPRESS_ZLIB
;
1731 ret
= btrfs_parse_options(tree_root
, options
);
1737 features
= btrfs_super_incompat_flags(disk_super
) &
1738 ~BTRFS_FEATURE_INCOMPAT_SUPP
;
1740 printk(KERN_ERR
"BTRFS: couldn't mount because of "
1741 "unsupported optional features (%Lx).\n",
1742 (unsigned long long)features
);
1747 features
= btrfs_super_incompat_flags(disk_super
);
1748 features
|= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF
;
1749 if (tree_root
->fs_info
->compress_type
& BTRFS_COMPRESS_LZO
)
1750 features
|= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO
;
1751 btrfs_set_super_incompat_flags(disk_super
, features
);
1753 features
= btrfs_super_compat_ro_flags(disk_super
) &
1754 ~BTRFS_FEATURE_COMPAT_RO_SUPP
;
1755 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
1756 printk(KERN_ERR
"BTRFS: couldn't mount RDWR because of "
1757 "unsupported option features (%Lx).\n",
1758 (unsigned long long)features
);
1763 btrfs_init_workers(&fs_info
->generic_worker
,
1764 "genwork", 1, NULL
);
1766 btrfs_init_workers(&fs_info
->workers
, "worker",
1767 fs_info
->thread_pool_size
,
1768 &fs_info
->generic_worker
);
1770 btrfs_init_workers(&fs_info
->delalloc_workers
, "delalloc",
1771 fs_info
->thread_pool_size
,
1772 &fs_info
->generic_worker
);
1774 btrfs_init_workers(&fs_info
->submit_workers
, "submit",
1775 min_t(u64
, fs_devices
->num_devices
,
1776 fs_info
->thread_pool_size
),
1777 &fs_info
->generic_worker
);
1779 /* a higher idle thresh on the submit workers makes it much more
1780 * likely that bios will be send down in a sane order to the
1783 fs_info
->submit_workers
.idle_thresh
= 64;
1785 fs_info
->workers
.idle_thresh
= 16;
1786 fs_info
->workers
.ordered
= 1;
1788 fs_info
->delalloc_workers
.idle_thresh
= 2;
1789 fs_info
->delalloc_workers
.ordered
= 1;
1791 btrfs_init_workers(&fs_info
->fixup_workers
, "fixup", 1,
1792 &fs_info
->generic_worker
);
1793 btrfs_init_workers(&fs_info
->endio_workers
, "endio",
1794 fs_info
->thread_pool_size
,
1795 &fs_info
->generic_worker
);
1796 btrfs_init_workers(&fs_info
->endio_meta_workers
, "endio-meta",
1797 fs_info
->thread_pool_size
,
1798 &fs_info
->generic_worker
);
1799 btrfs_init_workers(&fs_info
->endio_meta_write_workers
,
1800 "endio-meta-write", fs_info
->thread_pool_size
,
1801 &fs_info
->generic_worker
);
1802 btrfs_init_workers(&fs_info
->endio_write_workers
, "endio-write",
1803 fs_info
->thread_pool_size
,
1804 &fs_info
->generic_worker
);
1805 btrfs_init_workers(&fs_info
->endio_freespace_worker
, "freespace-write",
1806 1, &fs_info
->generic_worker
);
1809 * endios are largely parallel and should have a very
1812 fs_info
->endio_workers
.idle_thresh
= 4;
1813 fs_info
->endio_meta_workers
.idle_thresh
= 4;
1815 fs_info
->endio_write_workers
.idle_thresh
= 2;
1816 fs_info
->endio_meta_write_workers
.idle_thresh
= 2;
1818 btrfs_start_workers(&fs_info
->workers
, 1);
1819 btrfs_start_workers(&fs_info
->generic_worker
, 1);
1820 btrfs_start_workers(&fs_info
->submit_workers
, 1);
1821 btrfs_start_workers(&fs_info
->delalloc_workers
, 1);
1822 btrfs_start_workers(&fs_info
->fixup_workers
, 1);
1823 btrfs_start_workers(&fs_info
->endio_workers
, 1);
1824 btrfs_start_workers(&fs_info
->endio_meta_workers
, 1);
1825 btrfs_start_workers(&fs_info
->endio_meta_write_workers
, 1);
1826 btrfs_start_workers(&fs_info
->endio_write_workers
, 1);
1827 btrfs_start_workers(&fs_info
->endio_freespace_worker
, 1);
1829 fs_info
->bdi
.ra_pages
*= btrfs_super_num_devices(disk_super
);
1830 fs_info
->bdi
.ra_pages
= max(fs_info
->bdi
.ra_pages
,
1831 4 * 1024 * 1024 / PAGE_CACHE_SIZE
);
1833 nodesize
= btrfs_super_nodesize(disk_super
);
1834 leafsize
= btrfs_super_leafsize(disk_super
);
1835 sectorsize
= btrfs_super_sectorsize(disk_super
);
1836 stripesize
= btrfs_super_stripesize(disk_super
);
1837 tree_root
->nodesize
= nodesize
;
1838 tree_root
->leafsize
= leafsize
;
1839 tree_root
->sectorsize
= sectorsize
;
1840 tree_root
->stripesize
= stripesize
;
1842 sb
->s_blocksize
= sectorsize
;
1843 sb
->s_blocksize_bits
= blksize_bits(sectorsize
);
1845 if (strncmp((char *)(&disk_super
->magic
), BTRFS_MAGIC
,
1846 sizeof(disk_super
->magic
))) {
1847 printk(KERN_INFO
"btrfs: valid FS not found on %s\n", sb
->s_id
);
1848 goto fail_sb_buffer
;
1851 mutex_lock(&fs_info
->chunk_mutex
);
1852 ret
= btrfs_read_sys_array(tree_root
);
1853 mutex_unlock(&fs_info
->chunk_mutex
);
1855 printk(KERN_WARNING
"btrfs: failed to read the system "
1856 "array on %s\n", sb
->s_id
);
1857 goto fail_sb_buffer
;
1860 blocksize
= btrfs_level_size(tree_root
,
1861 btrfs_super_chunk_root_level(disk_super
));
1862 generation
= btrfs_super_chunk_root_generation(disk_super
);
1864 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
1865 chunk_root
, fs_info
, BTRFS_CHUNK_TREE_OBJECTID
);
1867 chunk_root
->node
= read_tree_block(chunk_root
,
1868 btrfs_super_chunk_root(disk_super
),
1869 blocksize
, generation
);
1870 BUG_ON(!chunk_root
->node
);
1871 if (!test_bit(EXTENT_BUFFER_UPTODATE
, &chunk_root
->node
->bflags
)) {
1872 printk(KERN_WARNING
"btrfs: failed to read chunk root on %s\n",
1874 goto fail_chunk_root
;
1876 btrfs_set_root_node(&chunk_root
->root_item
, chunk_root
->node
);
1877 chunk_root
->commit_root
= btrfs_root_node(chunk_root
);
1879 read_extent_buffer(chunk_root
->node
, fs_info
->chunk_tree_uuid
,
1880 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root
->node
),
1883 mutex_lock(&fs_info
->chunk_mutex
);
1884 ret
= btrfs_read_chunk_tree(chunk_root
);
1885 mutex_unlock(&fs_info
->chunk_mutex
);
1887 printk(KERN_WARNING
"btrfs: failed to read chunk tree on %s\n",
1889 goto fail_chunk_root
;
1892 btrfs_close_extra_devices(fs_devices
);
1894 blocksize
= btrfs_level_size(tree_root
,
1895 btrfs_super_root_level(disk_super
));
1896 generation
= btrfs_super_generation(disk_super
);
1898 tree_root
->node
= read_tree_block(tree_root
,
1899 btrfs_super_root(disk_super
),
1900 blocksize
, generation
);
1901 if (!tree_root
->node
)
1902 goto fail_chunk_root
;
1903 if (!test_bit(EXTENT_BUFFER_UPTODATE
, &tree_root
->node
->bflags
)) {
1904 printk(KERN_WARNING
"btrfs: failed to read tree root on %s\n",
1906 goto fail_tree_root
;
1908 btrfs_set_root_node(&tree_root
->root_item
, tree_root
->node
);
1909 tree_root
->commit_root
= btrfs_root_node(tree_root
);
1911 ret
= find_and_setup_root(tree_root
, fs_info
,
1912 BTRFS_EXTENT_TREE_OBJECTID
, extent_root
);
1914 goto fail_tree_root
;
1915 extent_root
->track_dirty
= 1;
1917 ret
= find_and_setup_root(tree_root
, fs_info
,
1918 BTRFS_DEV_TREE_OBJECTID
, dev_root
);
1920 goto fail_extent_root
;
1921 dev_root
->track_dirty
= 1;
1923 ret
= find_and_setup_root(tree_root
, fs_info
,
1924 BTRFS_CSUM_TREE_OBJECTID
, csum_root
);
1928 csum_root
->track_dirty
= 1;
1930 fs_info
->generation
= generation
;
1931 fs_info
->last_trans_committed
= generation
;
1932 fs_info
->data_alloc_profile
= (u64
)-1;
1933 fs_info
->metadata_alloc_profile
= (u64
)-1;
1934 fs_info
->system_alloc_profile
= fs_info
->metadata_alloc_profile
;
1936 ret
= btrfs_init_space_info(fs_info
);
1938 printk(KERN_ERR
"Failed to initial space info: %d\n", ret
);
1939 goto fail_block_groups
;
1942 ret
= btrfs_read_block_groups(extent_root
);
1944 printk(KERN_ERR
"Failed to read block groups: %d\n", ret
);
1945 goto fail_block_groups
;
1948 fs_info
->cleaner_kthread
= kthread_run(cleaner_kthread
, tree_root
,
1950 if (IS_ERR(fs_info
->cleaner_kthread
))
1951 goto fail_block_groups
;
1953 fs_info
->transaction_kthread
= kthread_run(transaction_kthread
,
1955 "btrfs-transaction");
1956 if (IS_ERR(fs_info
->transaction_kthread
))
1959 if (!btrfs_test_opt(tree_root
, SSD
) &&
1960 !btrfs_test_opt(tree_root
, NOSSD
) &&
1961 !fs_info
->fs_devices
->rotating
) {
1962 printk(KERN_INFO
"Btrfs detected SSD devices, enabling SSD "
1964 btrfs_set_opt(fs_info
->mount_opt
, SSD
);
1967 /* do not make disk changes in broken FS */
1968 if (btrfs_super_log_root(disk_super
) != 0 &&
1969 !(fs_info
->fs_state
& BTRFS_SUPER_FLAG_ERROR
)) {
1970 u64 bytenr
= btrfs_super_log_root(disk_super
);
1972 if (fs_devices
->rw_devices
== 0) {
1973 printk(KERN_WARNING
"Btrfs log replay required "
1976 goto fail_trans_kthread
;
1979 btrfs_level_size(tree_root
,
1980 btrfs_super_log_root_level(disk_super
));
1982 log_tree_root
= kzalloc(sizeof(struct btrfs_root
), GFP_NOFS
);
1983 if (!log_tree_root
) {
1985 goto fail_trans_kthread
;
1988 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
1989 log_tree_root
, fs_info
, BTRFS_TREE_LOG_OBJECTID
);
1991 log_tree_root
->node
= read_tree_block(tree_root
, bytenr
,
1994 ret
= btrfs_recover_log_trees(log_tree_root
);
1997 if (sb
->s_flags
& MS_RDONLY
) {
1998 ret
= btrfs_commit_super(tree_root
);
2003 ret
= btrfs_find_orphan_roots(tree_root
);
2006 if (!(sb
->s_flags
& MS_RDONLY
)) {
2007 ret
= btrfs_cleanup_fs_roots(fs_info
);
2010 ret
= btrfs_recover_relocation(tree_root
);
2013 "btrfs: failed to recover relocation\n");
2015 goto fail_trans_kthread
;
2019 location
.objectid
= BTRFS_FS_TREE_OBJECTID
;
2020 location
.type
= BTRFS_ROOT_ITEM_KEY
;
2021 location
.offset
= (u64
)-1;
2023 fs_info
->fs_root
= btrfs_read_fs_root_no_name(fs_info
, &location
);
2024 if (!fs_info
->fs_root
)
2025 goto fail_trans_kthread
;
2026 if (IS_ERR(fs_info
->fs_root
)) {
2027 err
= PTR_ERR(fs_info
->fs_root
);
2028 goto fail_trans_kthread
;
2031 if (!(sb
->s_flags
& MS_RDONLY
)) {
2032 down_read(&fs_info
->cleanup_work_sem
);
2033 err
= btrfs_orphan_cleanup(fs_info
->fs_root
);
2035 err
= btrfs_orphan_cleanup(fs_info
->tree_root
);
2036 up_read(&fs_info
->cleanup_work_sem
);
2038 close_ctree(tree_root
);
2039 return ERR_PTR(err
);
2046 kthread_stop(fs_info
->transaction_kthread
);
2048 kthread_stop(fs_info
->cleaner_kthread
);
2051 * make sure we're done with the btree inode before we stop our
2054 filemap_write_and_wait(fs_info
->btree_inode
->i_mapping
);
2055 invalidate_inode_pages2(fs_info
->btree_inode
->i_mapping
);
2058 btrfs_free_block_groups(fs_info
);
2059 free_extent_buffer(csum_root
->node
);
2060 free_extent_buffer(csum_root
->commit_root
);
2062 free_extent_buffer(dev_root
->node
);
2063 free_extent_buffer(dev_root
->commit_root
);
2065 free_extent_buffer(extent_root
->node
);
2066 free_extent_buffer(extent_root
->commit_root
);
2068 free_extent_buffer(tree_root
->node
);
2069 free_extent_buffer(tree_root
->commit_root
);
2071 free_extent_buffer(chunk_root
->node
);
2072 free_extent_buffer(chunk_root
->commit_root
);
2074 btrfs_stop_workers(&fs_info
->generic_worker
);
2075 btrfs_stop_workers(&fs_info
->fixup_workers
);
2076 btrfs_stop_workers(&fs_info
->delalloc_workers
);
2077 btrfs_stop_workers(&fs_info
->workers
);
2078 btrfs_stop_workers(&fs_info
->endio_workers
);
2079 btrfs_stop_workers(&fs_info
->endio_meta_workers
);
2080 btrfs_stop_workers(&fs_info
->endio_meta_write_workers
);
2081 btrfs_stop_workers(&fs_info
->endio_write_workers
);
2082 btrfs_stop_workers(&fs_info
->endio_freespace_worker
);
2083 btrfs_stop_workers(&fs_info
->submit_workers
);
2085 invalidate_inode_pages2(fs_info
->btree_inode
->i_mapping
);
2086 iput(fs_info
->btree_inode
);
2088 btrfs_close_devices(fs_info
->fs_devices
);
2089 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
2091 bdi_destroy(&fs_info
->bdi
);
2093 cleanup_srcu_struct(&fs_info
->subvol_srcu
);
2101 return ERR_PTR(err
);
2104 static void btrfs_end_buffer_write_sync(struct buffer_head
*bh
, int uptodate
)
2106 char b
[BDEVNAME_SIZE
];
2109 set_buffer_uptodate(bh
);
2111 if (printk_ratelimit()) {
2112 printk(KERN_WARNING
"lost page write due to "
2113 "I/O error on %s\n",
2114 bdevname(bh
->b_bdev
, b
));
2116 /* note, we dont' set_buffer_write_io_error because we have
2117 * our own ways of dealing with the IO errors
2119 clear_buffer_uptodate(bh
);
2125 struct buffer_head
*btrfs_read_dev_super(struct block_device
*bdev
)
2127 struct buffer_head
*bh
;
2128 struct buffer_head
*latest
= NULL
;
2129 struct btrfs_super_block
*super
;
2134 /* we would like to check all the supers, but that would make
2135 * a btrfs mount succeed after a mkfs from a different FS.
2136 * So, we need to add a special mount option to scan for
2137 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
2139 for (i
= 0; i
< 1; i
++) {
2140 bytenr
= btrfs_sb_offset(i
);
2141 if (bytenr
+ 4096 >= i_size_read(bdev
->bd_inode
))
2143 bh
= __bread(bdev
, bytenr
/ 4096, 4096);
2147 super
= (struct btrfs_super_block
*)bh
->b_data
;
2148 if (btrfs_super_bytenr(super
) != bytenr
||
2149 strncmp((char *)(&super
->magic
), BTRFS_MAGIC
,
2150 sizeof(super
->magic
))) {
2155 if (!latest
|| btrfs_super_generation(super
) > transid
) {
2158 transid
= btrfs_super_generation(super
);
2167 * this should be called twice, once with wait == 0 and
2168 * once with wait == 1. When wait == 0 is done, all the buffer heads
2169 * we write are pinned.
2171 * They are released when wait == 1 is done.
2172 * max_mirrors must be the same for both runs, and it indicates how
2173 * many supers on this one device should be written.
2175 * max_mirrors == 0 means to write them all.
2177 static int write_dev_supers(struct btrfs_device
*device
,
2178 struct btrfs_super_block
*sb
,
2179 int do_barriers
, int wait
, int max_mirrors
)
2181 struct buffer_head
*bh
;
2187 int last_barrier
= 0;
2189 if (max_mirrors
== 0)
2190 max_mirrors
= BTRFS_SUPER_MIRROR_MAX
;
2192 /* make sure only the last submit_bh does a barrier */
2194 for (i
= 0; i
< max_mirrors
; i
++) {
2195 bytenr
= btrfs_sb_offset(i
);
2196 if (bytenr
+ BTRFS_SUPER_INFO_SIZE
>=
2197 device
->total_bytes
)
2203 for (i
= 0; i
< max_mirrors
; i
++) {
2204 bytenr
= btrfs_sb_offset(i
);
2205 if (bytenr
+ BTRFS_SUPER_INFO_SIZE
>= device
->total_bytes
)
2209 bh
= __find_get_block(device
->bdev
, bytenr
/ 4096,
2210 BTRFS_SUPER_INFO_SIZE
);
2213 if (!buffer_uptodate(bh
))
2216 /* drop our reference */
2219 /* drop the reference from the wait == 0 run */
2223 btrfs_set_super_bytenr(sb
, bytenr
);
2226 crc
= btrfs_csum_data(NULL
, (char *)sb
+
2227 BTRFS_CSUM_SIZE
, crc
,
2228 BTRFS_SUPER_INFO_SIZE
-
2230 btrfs_csum_final(crc
, sb
->csum
);
2233 * one reference for us, and we leave it for the
2236 bh
= __getblk(device
->bdev
, bytenr
/ 4096,
2237 BTRFS_SUPER_INFO_SIZE
);
2238 memcpy(bh
->b_data
, sb
, BTRFS_SUPER_INFO_SIZE
);
2240 /* one reference for submit_bh */
2243 set_buffer_uptodate(bh
);
2245 bh
->b_end_io
= btrfs_end_buffer_write_sync
;
2248 if (i
== last_barrier
&& do_barriers
)
2249 ret
= submit_bh(WRITE_FLUSH_FUA
, bh
);
2251 ret
= submit_bh(WRITE_SYNC
, bh
);
2256 return errors
< i
? 0 : -1;
2259 int write_all_supers(struct btrfs_root
*root
, int max_mirrors
)
2261 struct list_head
*head
;
2262 struct btrfs_device
*dev
;
2263 struct btrfs_super_block
*sb
;
2264 struct btrfs_dev_item
*dev_item
;
2268 int total_errors
= 0;
2271 max_errors
= btrfs_super_num_devices(&root
->fs_info
->super_copy
) - 1;
2272 do_barriers
= !btrfs_test_opt(root
, NOBARRIER
);
2274 sb
= &root
->fs_info
->super_for_commit
;
2275 dev_item
= &sb
->dev_item
;
2277 mutex_lock(&root
->fs_info
->fs_devices
->device_list_mutex
);
2278 head
= &root
->fs_info
->fs_devices
->devices
;
2279 list_for_each_entry(dev
, head
, dev_list
) {
2284 if (!dev
->in_fs_metadata
|| !dev
->writeable
)
2287 btrfs_set_stack_device_generation(dev_item
, 0);
2288 btrfs_set_stack_device_type(dev_item
, dev
->type
);
2289 btrfs_set_stack_device_id(dev_item
, dev
->devid
);
2290 btrfs_set_stack_device_total_bytes(dev_item
, dev
->total_bytes
);
2291 btrfs_set_stack_device_bytes_used(dev_item
, dev
->bytes_used
);
2292 btrfs_set_stack_device_io_align(dev_item
, dev
->io_align
);
2293 btrfs_set_stack_device_io_width(dev_item
, dev
->io_width
);
2294 btrfs_set_stack_device_sector_size(dev_item
, dev
->sector_size
);
2295 memcpy(dev_item
->uuid
, dev
->uuid
, BTRFS_UUID_SIZE
);
2296 memcpy(dev_item
->fsid
, dev
->fs_devices
->fsid
, BTRFS_UUID_SIZE
);
2298 flags
= btrfs_super_flags(sb
);
2299 btrfs_set_super_flags(sb
, flags
| BTRFS_HEADER_FLAG_WRITTEN
);
2301 ret
= write_dev_supers(dev
, sb
, do_barriers
, 0, max_mirrors
);
2305 if (total_errors
> max_errors
) {
2306 printk(KERN_ERR
"btrfs: %d errors while writing supers\n",
2312 list_for_each_entry(dev
, head
, dev_list
) {
2315 if (!dev
->in_fs_metadata
|| !dev
->writeable
)
2318 ret
= write_dev_supers(dev
, sb
, do_barriers
, 1, max_mirrors
);
2322 mutex_unlock(&root
->fs_info
->fs_devices
->device_list_mutex
);
2323 if (total_errors
> max_errors
) {
2324 printk(KERN_ERR
"btrfs: %d errors while writing supers\n",
2331 int write_ctree_super(struct btrfs_trans_handle
*trans
,
2332 struct btrfs_root
*root
, int max_mirrors
)
2336 ret
= write_all_supers(root
, max_mirrors
);
2340 int btrfs_free_fs_root(struct btrfs_fs_info
*fs_info
, struct btrfs_root
*root
)
2342 spin_lock(&fs_info
->fs_roots_radix_lock
);
2343 radix_tree_delete(&fs_info
->fs_roots_radix
,
2344 (unsigned long)root
->root_key
.objectid
);
2345 spin_unlock(&fs_info
->fs_roots_radix_lock
);
2347 if (btrfs_root_refs(&root
->root_item
) == 0)
2348 synchronize_srcu(&fs_info
->subvol_srcu
);
2354 static void free_fs_root(struct btrfs_root
*root
)
2356 WARN_ON(!RB_EMPTY_ROOT(&root
->inode_tree
));
2357 if (root
->anon_super
.s_dev
) {
2358 down_write(&root
->anon_super
.s_umount
);
2359 kill_anon_super(&root
->anon_super
);
2361 free_extent_buffer(root
->node
);
2362 free_extent_buffer(root
->commit_root
);
2367 static int del_fs_roots(struct btrfs_fs_info
*fs_info
)
2370 struct btrfs_root
*gang
[8];
2373 while (!list_empty(&fs_info
->dead_roots
)) {
2374 gang
[0] = list_entry(fs_info
->dead_roots
.next
,
2375 struct btrfs_root
, root_list
);
2376 list_del(&gang
[0]->root_list
);
2378 if (gang
[0]->in_radix
) {
2379 btrfs_free_fs_root(fs_info
, gang
[0]);
2381 free_extent_buffer(gang
[0]->node
);
2382 free_extent_buffer(gang
[0]->commit_root
);
2388 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
2393 for (i
= 0; i
< ret
; i
++)
2394 btrfs_free_fs_root(fs_info
, gang
[i
]);
2399 int btrfs_cleanup_fs_roots(struct btrfs_fs_info
*fs_info
)
2401 u64 root_objectid
= 0;
2402 struct btrfs_root
*gang
[8];
2407 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
2408 (void **)gang
, root_objectid
,
2413 root_objectid
= gang
[ret
- 1]->root_key
.objectid
+ 1;
2414 for (i
= 0; i
< ret
; i
++) {
2417 root_objectid
= gang
[i
]->root_key
.objectid
;
2418 err
= btrfs_orphan_cleanup(gang
[i
]);
2427 int btrfs_commit_super(struct btrfs_root
*root
)
2429 struct btrfs_trans_handle
*trans
;
2432 mutex_lock(&root
->fs_info
->cleaner_mutex
);
2433 btrfs_run_delayed_iputs(root
);
2434 btrfs_clean_old_snapshots(root
);
2435 mutex_unlock(&root
->fs_info
->cleaner_mutex
);
2437 /* wait until ongoing cleanup work done */
2438 down_write(&root
->fs_info
->cleanup_work_sem
);
2439 up_write(&root
->fs_info
->cleanup_work_sem
);
2441 trans
= btrfs_join_transaction(root
, 1);
2443 return PTR_ERR(trans
);
2444 ret
= btrfs_commit_transaction(trans
, root
);
2446 /* run commit again to drop the original snapshot */
2447 trans
= btrfs_join_transaction(root
, 1);
2449 return PTR_ERR(trans
);
2450 btrfs_commit_transaction(trans
, root
);
2451 ret
= btrfs_write_and_wait_transaction(NULL
, root
);
2454 ret
= write_ctree_super(NULL
, root
, 0);
2458 int close_ctree(struct btrfs_root
*root
)
2460 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
2463 fs_info
->closing
= 1;
2466 btrfs_put_block_group_cache(fs_info
);
2469 * Here come 2 situations when btrfs is broken to flip readonly:
2471 * 1. when btrfs flips readonly somewhere else before
2472 * btrfs_commit_super, sb->s_flags has MS_RDONLY flag,
2473 * and btrfs will skip to write sb directly to keep
2474 * ERROR state on disk.
2476 * 2. when btrfs flips readonly just in btrfs_commit_super,
2477 * and in such case, btrfs cannot write sb via btrfs_commit_super,
2478 * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag,
2479 * btrfs will cleanup all FS resources first and write sb then.
2481 if (!(fs_info
->sb
->s_flags
& MS_RDONLY
)) {
2482 ret
= btrfs_commit_super(root
);
2484 printk(KERN_ERR
"btrfs: commit super ret %d\n", ret
);
2487 if (fs_info
->fs_state
& BTRFS_SUPER_FLAG_ERROR
) {
2488 ret
= btrfs_error_commit_super(root
);
2490 printk(KERN_ERR
"btrfs: commit super ret %d\n", ret
);
2493 kthread_stop(root
->fs_info
->transaction_kthread
);
2494 kthread_stop(root
->fs_info
->cleaner_kthread
);
2496 fs_info
->closing
= 2;
2499 if (fs_info
->delalloc_bytes
) {
2500 printk(KERN_INFO
"btrfs: at unmount delalloc count %llu\n",
2501 (unsigned long long)fs_info
->delalloc_bytes
);
2503 if (fs_info
->total_ref_cache_size
) {
2504 printk(KERN_INFO
"btrfs: at umount reference cache size %llu\n",
2505 (unsigned long long)fs_info
->total_ref_cache_size
);
2508 free_extent_buffer(fs_info
->extent_root
->node
);
2509 free_extent_buffer(fs_info
->extent_root
->commit_root
);
2510 free_extent_buffer(fs_info
->tree_root
->node
);
2511 free_extent_buffer(fs_info
->tree_root
->commit_root
);
2512 free_extent_buffer(root
->fs_info
->chunk_root
->node
);
2513 free_extent_buffer(root
->fs_info
->chunk_root
->commit_root
);
2514 free_extent_buffer(root
->fs_info
->dev_root
->node
);
2515 free_extent_buffer(root
->fs_info
->dev_root
->commit_root
);
2516 free_extent_buffer(root
->fs_info
->csum_root
->node
);
2517 free_extent_buffer(root
->fs_info
->csum_root
->commit_root
);
2519 btrfs_free_block_groups(root
->fs_info
);
2521 del_fs_roots(fs_info
);
2523 iput(fs_info
->btree_inode
);
2525 btrfs_stop_workers(&fs_info
->generic_worker
);
2526 btrfs_stop_workers(&fs_info
->fixup_workers
);
2527 btrfs_stop_workers(&fs_info
->delalloc_workers
);
2528 btrfs_stop_workers(&fs_info
->workers
);
2529 btrfs_stop_workers(&fs_info
->endio_workers
);
2530 btrfs_stop_workers(&fs_info
->endio_meta_workers
);
2531 btrfs_stop_workers(&fs_info
->endio_meta_write_workers
);
2532 btrfs_stop_workers(&fs_info
->endio_write_workers
);
2533 btrfs_stop_workers(&fs_info
->endio_freespace_worker
);
2534 btrfs_stop_workers(&fs_info
->submit_workers
);
2536 btrfs_close_devices(fs_info
->fs_devices
);
2537 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
2539 bdi_destroy(&fs_info
->bdi
);
2540 cleanup_srcu_struct(&fs_info
->subvol_srcu
);
2542 kfree(fs_info
->extent_root
);
2543 kfree(fs_info
->tree_root
);
2544 kfree(fs_info
->chunk_root
);
2545 kfree(fs_info
->dev_root
);
2546 kfree(fs_info
->csum_root
);
2552 int btrfs_buffer_uptodate(struct extent_buffer
*buf
, u64 parent_transid
)
2555 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
2557 ret
= extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
, buf
,
2562 ret
= verify_parent_transid(&BTRFS_I(btree_inode
)->io_tree
, buf
,
2567 int btrfs_set_buffer_uptodate(struct extent_buffer
*buf
)
2569 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
2570 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
,
2574 void btrfs_mark_buffer_dirty(struct extent_buffer
*buf
)
2576 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
2577 u64 transid
= btrfs_header_generation(buf
);
2578 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
2581 btrfs_assert_tree_locked(buf
);
2582 if (transid
!= root
->fs_info
->generation
) {
2583 printk(KERN_CRIT
"btrfs transid mismatch buffer %llu, "
2584 "found %llu running %llu\n",
2585 (unsigned long long)buf
->start
,
2586 (unsigned long long)transid
,
2587 (unsigned long long)root
->fs_info
->generation
);
2590 was_dirty
= set_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
,
2593 spin_lock(&root
->fs_info
->delalloc_lock
);
2594 root
->fs_info
->dirty_metadata_bytes
+= buf
->len
;
2595 spin_unlock(&root
->fs_info
->delalloc_lock
);
2599 void btrfs_btree_balance_dirty(struct btrfs_root
*root
, unsigned long nr
)
2602 * looks as though older kernels can get into trouble with
2603 * this code, they end up stuck in balance_dirty_pages forever
2606 unsigned long thresh
= 32 * 1024 * 1024;
2608 if (current
->flags
& PF_MEMALLOC
)
2611 num_dirty
= root
->fs_info
->dirty_metadata_bytes
;
2613 if (num_dirty
> thresh
) {
2614 balance_dirty_pages_ratelimited_nr(
2615 root
->fs_info
->btree_inode
->i_mapping
, 1);
2620 int btrfs_read_buffer(struct extent_buffer
*buf
, u64 parent_transid
)
2622 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
2624 ret
= btree_read_extent_buffer_pages(root
, buf
, 0, parent_transid
);
2626 set_bit(EXTENT_BUFFER_UPTODATE
, &buf
->bflags
);
2630 int btree_lock_page_hook(struct page
*page
)
2632 struct inode
*inode
= page
->mapping
->host
;
2633 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2634 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2635 struct extent_buffer
*eb
;
2637 u64 bytenr
= page_offset(page
);
2639 if (page
->private == EXTENT_PAGE_PRIVATE
)
2642 len
= page
->private >> 2;
2643 eb
= find_extent_buffer(io_tree
, bytenr
, len
);
2647 btrfs_tree_lock(eb
);
2648 btrfs_set_header_flag(eb
, BTRFS_HEADER_FLAG_WRITTEN
);
2650 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY
, &eb
->bflags
)) {
2651 spin_lock(&root
->fs_info
->delalloc_lock
);
2652 if (root
->fs_info
->dirty_metadata_bytes
>= eb
->len
)
2653 root
->fs_info
->dirty_metadata_bytes
-= eb
->len
;
2656 spin_unlock(&root
->fs_info
->delalloc_lock
);
2659 btrfs_tree_unlock(eb
);
2660 free_extent_buffer(eb
);
2666 static void btrfs_check_super_valid(struct btrfs_fs_info
*fs_info
,
2672 if (fs_info
->fs_state
& BTRFS_SUPER_FLAG_ERROR
)
2673 printk(KERN_WARNING
"warning: mount fs with errors, "
2674 "running btrfsck is recommended\n");
2677 int btrfs_error_commit_super(struct btrfs_root
*root
)
2681 mutex_lock(&root
->fs_info
->cleaner_mutex
);
2682 btrfs_run_delayed_iputs(root
);
2683 mutex_unlock(&root
->fs_info
->cleaner_mutex
);
2685 down_write(&root
->fs_info
->cleanup_work_sem
);
2686 up_write(&root
->fs_info
->cleanup_work_sem
);
2688 /* cleanup FS via transaction */
2689 btrfs_cleanup_transaction(root
);
2691 ret
= write_ctree_super(NULL
, root
, 0);
2696 static int btrfs_destroy_ordered_operations(struct btrfs_root
*root
)
2698 struct btrfs_inode
*btrfs_inode
;
2699 struct list_head splice
;
2701 INIT_LIST_HEAD(&splice
);
2703 mutex_lock(&root
->fs_info
->ordered_operations_mutex
);
2704 spin_lock(&root
->fs_info
->ordered_extent_lock
);
2706 list_splice_init(&root
->fs_info
->ordered_operations
, &splice
);
2707 while (!list_empty(&splice
)) {
2708 btrfs_inode
= list_entry(splice
.next
, struct btrfs_inode
,
2709 ordered_operations
);
2711 list_del_init(&btrfs_inode
->ordered_operations
);
2713 btrfs_invalidate_inodes(btrfs_inode
->root
);
2716 spin_unlock(&root
->fs_info
->ordered_extent_lock
);
2717 mutex_unlock(&root
->fs_info
->ordered_operations_mutex
);
2722 static int btrfs_destroy_ordered_extents(struct btrfs_root
*root
)
2724 struct list_head splice
;
2725 struct btrfs_ordered_extent
*ordered
;
2726 struct inode
*inode
;
2728 INIT_LIST_HEAD(&splice
);
2730 spin_lock(&root
->fs_info
->ordered_extent_lock
);
2732 list_splice_init(&root
->fs_info
->ordered_extents
, &splice
);
2733 while (!list_empty(&splice
)) {
2734 ordered
= list_entry(splice
.next
, struct btrfs_ordered_extent
,
2737 list_del_init(&ordered
->root_extent_list
);
2738 atomic_inc(&ordered
->refs
);
2740 /* the inode may be getting freed (in sys_unlink path). */
2741 inode
= igrab(ordered
->inode
);
2743 spin_unlock(&root
->fs_info
->ordered_extent_lock
);
2747 atomic_set(&ordered
->refs
, 1);
2748 btrfs_put_ordered_extent(ordered
);
2750 spin_lock(&root
->fs_info
->ordered_extent_lock
);
2753 spin_unlock(&root
->fs_info
->ordered_extent_lock
);
2758 static int btrfs_destroy_delayed_refs(struct btrfs_transaction
*trans
,
2759 struct btrfs_root
*root
)
2761 struct rb_node
*node
;
2762 struct btrfs_delayed_ref_root
*delayed_refs
;
2763 struct btrfs_delayed_ref_node
*ref
;
2766 delayed_refs
= &trans
->delayed_refs
;
2768 spin_lock(&delayed_refs
->lock
);
2769 if (delayed_refs
->num_entries
== 0) {
2770 spin_unlock(&delayed_refs
->lock
);
2771 printk(KERN_INFO
"delayed_refs has NO entry\n");
2775 node
= rb_first(&delayed_refs
->root
);
2777 ref
= rb_entry(node
, struct btrfs_delayed_ref_node
, rb_node
);
2778 node
= rb_next(node
);
2781 rb_erase(&ref
->rb_node
, &delayed_refs
->root
);
2782 delayed_refs
->num_entries
--;
2784 atomic_set(&ref
->refs
, 1);
2785 if (btrfs_delayed_ref_is_head(ref
)) {
2786 struct btrfs_delayed_ref_head
*head
;
2788 head
= btrfs_delayed_node_to_head(ref
);
2789 mutex_lock(&head
->mutex
);
2790 kfree(head
->extent_op
);
2791 delayed_refs
->num_heads
--;
2792 if (list_empty(&head
->cluster
))
2793 delayed_refs
->num_heads_ready
--;
2794 list_del_init(&head
->cluster
);
2795 mutex_unlock(&head
->mutex
);
2798 spin_unlock(&delayed_refs
->lock
);
2799 btrfs_put_delayed_ref(ref
);
2802 spin_lock(&delayed_refs
->lock
);
2805 spin_unlock(&delayed_refs
->lock
);
2810 static int btrfs_destroy_pending_snapshots(struct btrfs_transaction
*t
)
2812 struct btrfs_pending_snapshot
*snapshot
;
2813 struct list_head splice
;
2815 INIT_LIST_HEAD(&splice
);
2817 list_splice_init(&t
->pending_snapshots
, &splice
);
2819 while (!list_empty(&splice
)) {
2820 snapshot
= list_entry(splice
.next
,
2821 struct btrfs_pending_snapshot
,
2824 list_del_init(&snapshot
->list
);
2832 static int btrfs_destroy_delalloc_inodes(struct btrfs_root
*root
)
2834 struct btrfs_inode
*btrfs_inode
;
2835 struct list_head splice
;
2837 INIT_LIST_HEAD(&splice
);
2839 list_splice_init(&root
->fs_info
->delalloc_inodes
, &splice
);
2841 spin_lock(&root
->fs_info
->delalloc_lock
);
2843 while (!list_empty(&splice
)) {
2844 btrfs_inode
= list_entry(splice
.next
, struct btrfs_inode
,
2847 list_del_init(&btrfs_inode
->delalloc_inodes
);
2849 btrfs_invalidate_inodes(btrfs_inode
->root
);
2852 spin_unlock(&root
->fs_info
->delalloc_lock
);
2857 static int btrfs_destroy_marked_extents(struct btrfs_root
*root
,
2858 struct extent_io_tree
*dirty_pages
,
2863 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
2864 struct extent_buffer
*eb
;
2868 unsigned long index
;
2871 ret
= find_first_extent_bit(dirty_pages
, start
, &start
, &end
,
2876 clear_extent_bits(dirty_pages
, start
, end
, mark
, GFP_NOFS
);
2877 while (start
<= end
) {
2878 index
= start
>> PAGE_CACHE_SHIFT
;
2879 start
= (u64
)(index
+ 1) << PAGE_CACHE_SHIFT
;
2880 page
= find_get_page(btree_inode
->i_mapping
, index
);
2883 offset
= page_offset(page
);
2885 spin_lock(&dirty_pages
->buffer_lock
);
2886 eb
= radix_tree_lookup(
2887 &(&BTRFS_I(page
->mapping
->host
)->io_tree
)->buffer
,
2888 offset
>> PAGE_CACHE_SHIFT
);
2889 spin_unlock(&dirty_pages
->buffer_lock
);
2891 ret
= test_and_clear_bit(EXTENT_BUFFER_DIRTY
,
2893 atomic_set(&eb
->refs
, 1);
2895 if (PageWriteback(page
))
2896 end_page_writeback(page
);
2899 if (PageDirty(page
)) {
2900 clear_page_dirty_for_io(page
);
2901 spin_lock_irq(&page
->mapping
->tree_lock
);
2902 radix_tree_tag_clear(&page
->mapping
->page_tree
,
2904 PAGECACHE_TAG_DIRTY
);
2905 spin_unlock_irq(&page
->mapping
->tree_lock
);
2908 page
->mapping
->a_ops
->invalidatepage(page
, 0);
2916 static int btrfs_destroy_pinned_extent(struct btrfs_root
*root
,
2917 struct extent_io_tree
*pinned_extents
)
2919 struct extent_io_tree
*unpin
;
2924 unpin
= pinned_extents
;
2926 ret
= find_first_extent_bit(unpin
, 0, &start
, &end
,
2932 if (btrfs_test_opt(root
, DISCARD
))
2933 ret
= btrfs_error_discard_extent(root
, start
,
2937 clear_extent_dirty(unpin
, start
, end
, GFP_NOFS
);
2938 btrfs_error_unpin_extent_range(root
, start
, end
);
2945 static int btrfs_cleanup_transaction(struct btrfs_root
*root
)
2947 struct btrfs_transaction
*t
;
2952 mutex_lock(&root
->fs_info
->trans_mutex
);
2953 mutex_lock(&root
->fs_info
->transaction_kthread_mutex
);
2955 list_splice_init(&root
->fs_info
->trans_list
, &list
);
2956 while (!list_empty(&list
)) {
2957 t
= list_entry(list
.next
, struct btrfs_transaction
, list
);
2961 btrfs_destroy_ordered_operations(root
);
2963 btrfs_destroy_ordered_extents(root
);
2965 btrfs_destroy_delayed_refs(t
, root
);
2967 btrfs_block_rsv_release(root
,
2968 &root
->fs_info
->trans_block_rsv
,
2969 t
->dirty_pages
.dirty_bytes
);
2971 /* FIXME: cleanup wait for commit */
2974 if (waitqueue_active(&root
->fs_info
->transaction_blocked_wait
))
2975 wake_up(&root
->fs_info
->transaction_blocked_wait
);
2978 if (waitqueue_active(&root
->fs_info
->transaction_wait
))
2979 wake_up(&root
->fs_info
->transaction_wait
);
2980 mutex_unlock(&root
->fs_info
->trans_mutex
);
2982 mutex_lock(&root
->fs_info
->trans_mutex
);
2984 if (waitqueue_active(&t
->commit_wait
))
2985 wake_up(&t
->commit_wait
);
2986 mutex_unlock(&root
->fs_info
->trans_mutex
);
2988 mutex_lock(&root
->fs_info
->trans_mutex
);
2990 btrfs_destroy_pending_snapshots(t
);
2992 btrfs_destroy_delalloc_inodes(root
);
2994 spin_lock(&root
->fs_info
->new_trans_lock
);
2995 root
->fs_info
->running_transaction
= NULL
;
2996 spin_unlock(&root
->fs_info
->new_trans_lock
);
2998 btrfs_destroy_marked_extents(root
, &t
->dirty_pages
,
3001 btrfs_destroy_pinned_extent(root
,
3002 root
->fs_info
->pinned_extents
);
3004 atomic_set(&t
->use_count
, 0);
3005 list_del_init(&t
->list
);
3006 memset(t
, 0, sizeof(*t
));
3007 kmem_cache_free(btrfs_transaction_cachep
, t
);
3010 mutex_unlock(&root
->fs_info
->transaction_kthread_mutex
);
3011 mutex_unlock(&root
->fs_info
->trans_mutex
);
3016 static struct extent_io_ops btree_extent_io_ops
= {
3017 .write_cache_pages_lock_hook
= btree_lock_page_hook
,
3018 .readpage_end_io_hook
= btree_readpage_end_io_hook
,
3019 .submit_bio_hook
= btree_submit_bio_hook
,
3020 /* note we're sharing with inode.c for the merge bio hook */
3021 .merge_bio_hook
= btrfs_merge_bio_hook
,