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 <linux/ratelimit.h>
33 #include <asm/unaligned.h>
37 #include "transaction.h"
38 #include "btrfs_inode.h"
40 #include "print-tree.h"
41 #include "async-thread.h"
44 #include "free-space-cache.h"
45 #include "inode-map.h"
46 #include "check-integrity.h"
48 static struct extent_io_ops btree_extent_io_ops
;
49 static void end_workqueue_fn(struct btrfs_work
*work
);
50 static void free_fs_root(struct btrfs_root
*root
);
51 static int btrfs_check_super_valid(struct btrfs_fs_info
*fs_info
,
53 static void btrfs_destroy_ordered_operations(struct btrfs_root
*root
);
54 static void btrfs_destroy_ordered_extents(struct btrfs_root
*root
);
55 static int btrfs_destroy_delayed_refs(struct btrfs_transaction
*trans
,
56 struct btrfs_root
*root
);
57 static void btrfs_destroy_pending_snapshots(struct btrfs_transaction
*t
);
58 static void btrfs_destroy_delalloc_inodes(struct btrfs_root
*root
);
59 static int btrfs_destroy_marked_extents(struct btrfs_root
*root
,
60 struct extent_io_tree
*dirty_pages
,
62 static int btrfs_destroy_pinned_extent(struct btrfs_root
*root
,
63 struct extent_io_tree
*pinned_extents
);
66 * end_io_wq structs are used to do processing in task context when an IO is
67 * complete. This is used during reads to verify checksums, and it is used
68 * by writes to insert metadata for new file extents after IO is complete.
74 struct btrfs_fs_info
*info
;
77 struct list_head list
;
78 struct btrfs_work work
;
82 * async submit bios are used to offload expensive checksumming
83 * onto the worker threads. They checksum file and metadata bios
84 * just before they are sent down the IO stack.
86 struct async_submit_bio
{
89 struct list_head list
;
90 extent_submit_bio_hook_t
*submit_bio_start
;
91 extent_submit_bio_hook_t
*submit_bio_done
;
94 unsigned long bio_flags
;
96 * bio_offset is optional, can be used if the pages in the bio
97 * can't tell us where in the file the bio should go
100 struct btrfs_work work
;
105 * Lockdep class keys for extent_buffer->lock's in this root. For a given
106 * eb, the lockdep key is determined by the btrfs_root it belongs to and
107 * the level the eb occupies in the tree.
109 * Different roots are used for different purposes and may nest inside each
110 * other and they require separate keysets. As lockdep keys should be
111 * static, assign keysets according to the purpose of the root as indicated
112 * by btrfs_root->objectid. This ensures that all special purpose roots
113 * have separate keysets.
115 * Lock-nesting across peer nodes is always done with the immediate parent
116 * node locked thus preventing deadlock. As lockdep doesn't know this, use
117 * subclass to avoid triggering lockdep warning in such cases.
119 * The key is set by the readpage_end_io_hook after the buffer has passed
120 * csum validation but before the pages are unlocked. It is also set by
121 * btrfs_init_new_buffer on freshly allocated blocks.
123 * We also add a check to make sure the highest level of the tree is the
124 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
125 * needs update as well.
127 #ifdef CONFIG_DEBUG_LOCK_ALLOC
128 # if BTRFS_MAX_LEVEL != 8
132 static struct btrfs_lockdep_keyset
{
133 u64 id
; /* root objectid */
134 const char *name_stem
; /* lock name stem */
135 char names
[BTRFS_MAX_LEVEL
+ 1][20];
136 struct lock_class_key keys
[BTRFS_MAX_LEVEL
+ 1];
137 } btrfs_lockdep_keysets
[] = {
138 { .id
= BTRFS_ROOT_TREE_OBJECTID
, .name_stem
= "root" },
139 { .id
= BTRFS_EXTENT_TREE_OBJECTID
, .name_stem
= "extent" },
140 { .id
= BTRFS_CHUNK_TREE_OBJECTID
, .name_stem
= "chunk" },
141 { .id
= BTRFS_DEV_TREE_OBJECTID
, .name_stem
= "dev" },
142 { .id
= BTRFS_FS_TREE_OBJECTID
, .name_stem
= "fs" },
143 { .id
= BTRFS_CSUM_TREE_OBJECTID
, .name_stem
= "csum" },
144 { .id
= BTRFS_ORPHAN_OBJECTID
, .name_stem
= "orphan" },
145 { .id
= BTRFS_TREE_LOG_OBJECTID
, .name_stem
= "log" },
146 { .id
= BTRFS_TREE_RELOC_OBJECTID
, .name_stem
= "treloc" },
147 { .id
= BTRFS_DATA_RELOC_TREE_OBJECTID
, .name_stem
= "dreloc" },
148 { .id
= 0, .name_stem
= "tree" },
151 void __init
btrfs_init_lockdep(void)
155 /* initialize lockdep class names */
156 for (i
= 0; i
< ARRAY_SIZE(btrfs_lockdep_keysets
); i
++) {
157 struct btrfs_lockdep_keyset
*ks
= &btrfs_lockdep_keysets
[i
];
159 for (j
= 0; j
< ARRAY_SIZE(ks
->names
); j
++)
160 snprintf(ks
->names
[j
], sizeof(ks
->names
[j
]),
161 "btrfs-%s-%02d", ks
->name_stem
, j
);
165 void btrfs_set_buffer_lockdep_class(u64 objectid
, struct extent_buffer
*eb
,
168 struct btrfs_lockdep_keyset
*ks
;
170 BUG_ON(level
>= ARRAY_SIZE(ks
->keys
));
172 /* find the matching keyset, id 0 is the default entry */
173 for (ks
= btrfs_lockdep_keysets
; ks
->id
; ks
++)
174 if (ks
->id
== objectid
)
177 lockdep_set_class_and_name(&eb
->lock
,
178 &ks
->keys
[level
], ks
->names
[level
]);
184 * extents on the btree inode are pretty simple, there's one extent
185 * that covers the entire device
187 static struct extent_map
*btree_get_extent(struct inode
*inode
,
188 struct page
*page
, size_t pg_offset
, u64 start
, u64 len
,
191 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
192 struct extent_map
*em
;
195 read_lock(&em_tree
->lock
);
196 em
= lookup_extent_mapping(em_tree
, start
, len
);
199 BTRFS_I(inode
)->root
->fs_info
->fs_devices
->latest_bdev
;
200 read_unlock(&em_tree
->lock
);
203 read_unlock(&em_tree
->lock
);
205 em
= alloc_extent_map();
207 em
= ERR_PTR(-ENOMEM
);
212 em
->block_len
= (u64
)-1;
214 em
->bdev
= BTRFS_I(inode
)->root
->fs_info
->fs_devices
->latest_bdev
;
216 write_lock(&em_tree
->lock
);
217 ret
= add_extent_mapping(em_tree
, em
);
218 if (ret
== -EEXIST
) {
219 u64 failed_start
= em
->start
;
220 u64 failed_len
= em
->len
;
223 em
= lookup_extent_mapping(em_tree
, start
, len
);
227 em
= lookup_extent_mapping(em_tree
, failed_start
,
235 write_unlock(&em_tree
->lock
);
243 u32
btrfs_csum_data(struct btrfs_root
*root
, char *data
, u32 seed
, size_t len
)
245 return crc32c(seed
, data
, len
);
248 void btrfs_csum_final(u32 crc
, char *result
)
250 put_unaligned_le32(~crc
, result
);
254 * compute the csum for a btree block, and either verify it or write it
255 * into the csum field of the block.
257 static int csum_tree_block(struct btrfs_root
*root
, struct extent_buffer
*buf
,
260 u16 csum_size
= btrfs_super_csum_size(root
->fs_info
->super_copy
);
263 unsigned long cur_len
;
264 unsigned long offset
= BTRFS_CSUM_SIZE
;
266 unsigned long map_start
;
267 unsigned long map_len
;
270 unsigned long inline_result
;
272 len
= buf
->len
- offset
;
274 err
= map_private_extent_buffer(buf
, offset
, 32,
275 &kaddr
, &map_start
, &map_len
);
278 cur_len
= min(len
, map_len
- (offset
- map_start
));
279 crc
= btrfs_csum_data(root
, kaddr
+ offset
- map_start
,
284 if (csum_size
> sizeof(inline_result
)) {
285 result
= kzalloc(csum_size
* sizeof(char), GFP_NOFS
);
289 result
= (char *)&inline_result
;
292 btrfs_csum_final(crc
, result
);
295 if (memcmp_extent_buffer(buf
, result
, 0, csum_size
)) {
298 memcpy(&found
, result
, csum_size
);
300 read_extent_buffer(buf
, &val
, 0, csum_size
);
301 printk_ratelimited(KERN_INFO
"btrfs: %s checksum verify "
302 "failed on %llu wanted %X found %X "
304 root
->fs_info
->sb
->s_id
,
305 (unsigned long long)buf
->start
, val
, found
,
306 btrfs_header_level(buf
));
307 if (result
!= (char *)&inline_result
)
312 write_extent_buffer(buf
, result
, 0, csum_size
);
314 if (result
!= (char *)&inline_result
)
320 * we can't consider a given block up to date unless the transid of the
321 * block matches the transid in the parent node's pointer. This is how we
322 * detect blocks that either didn't get written at all or got written
323 * in the wrong place.
325 static int verify_parent_transid(struct extent_io_tree
*io_tree
,
326 struct extent_buffer
*eb
, u64 parent_transid
)
328 struct extent_state
*cached_state
= NULL
;
331 if (!parent_transid
|| btrfs_header_generation(eb
) == parent_transid
)
334 lock_extent_bits(io_tree
, eb
->start
, eb
->start
+ eb
->len
- 1,
336 if (extent_buffer_uptodate(eb
) &&
337 btrfs_header_generation(eb
) == parent_transid
) {
341 printk_ratelimited("parent transid verify failed on %llu wanted %llu "
343 (unsigned long long)eb
->start
,
344 (unsigned long long)parent_transid
,
345 (unsigned long long)btrfs_header_generation(eb
));
347 clear_extent_buffer_uptodate(eb
);
349 unlock_extent_cached(io_tree
, eb
->start
, eb
->start
+ eb
->len
- 1,
350 &cached_state
, GFP_NOFS
);
355 * helper to read a given tree block, doing retries as required when
356 * the checksums don't match and we have alternate mirrors to try.
358 static int btree_read_extent_buffer_pages(struct btrfs_root
*root
,
359 struct extent_buffer
*eb
,
360 u64 start
, u64 parent_transid
)
362 struct extent_io_tree
*io_tree
;
367 int failed_mirror
= 0;
369 clear_bit(EXTENT_BUFFER_CORRUPT
, &eb
->bflags
);
370 io_tree
= &BTRFS_I(root
->fs_info
->btree_inode
)->io_tree
;
372 ret
= read_extent_buffer_pages(io_tree
, eb
, start
,
374 btree_get_extent
, mirror_num
);
375 if (!ret
&& !verify_parent_transid(io_tree
, eb
, parent_transid
))
379 * This buffer's crc is fine, but its contents are corrupted, so
380 * there is no reason to read the other copies, they won't be
383 if (test_bit(EXTENT_BUFFER_CORRUPT
, &eb
->bflags
))
386 num_copies
= btrfs_num_copies(&root
->fs_info
->mapping_tree
,
391 if (!failed_mirror
) {
393 failed_mirror
= eb
->read_mirror
;
397 if (mirror_num
== failed_mirror
)
400 if (mirror_num
> num_copies
)
405 repair_eb_io_failure(root
, eb
, failed_mirror
);
411 * checksum a dirty tree block before IO. This has extra checks to make sure
412 * we only fill in the checksum field in the first page of a multi-page block
415 static int csum_dirty_buffer(struct btrfs_root
*root
, struct page
*page
)
417 struct extent_io_tree
*tree
;
418 u64 start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
420 struct extent_buffer
*eb
;
422 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
424 eb
= (struct extent_buffer
*)page
->private;
425 if (page
!= eb
->pages
[0])
427 found_start
= btrfs_header_bytenr(eb
);
428 if (found_start
!= start
) {
432 if (eb
->pages
[0] != page
) {
436 if (!PageUptodate(page
)) {
440 csum_tree_block(root
, eb
, 0);
444 static int check_tree_block_fsid(struct btrfs_root
*root
,
445 struct extent_buffer
*eb
)
447 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
448 u8 fsid
[BTRFS_UUID_SIZE
];
451 read_extent_buffer(eb
, fsid
, (unsigned long)btrfs_header_fsid(eb
),
454 if (!memcmp(fsid
, fs_devices
->fsid
, BTRFS_FSID_SIZE
)) {
458 fs_devices
= fs_devices
->seed
;
463 #define CORRUPT(reason, eb, root, slot) \
464 printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \
465 "root=%llu, slot=%d\n", reason, \
466 (unsigned long long)btrfs_header_bytenr(eb), \
467 (unsigned long long)root->objectid, slot)
469 static noinline
int check_leaf(struct btrfs_root
*root
,
470 struct extent_buffer
*leaf
)
472 struct btrfs_key key
;
473 struct btrfs_key leaf_key
;
474 u32 nritems
= btrfs_header_nritems(leaf
);
480 /* Check the 0 item */
481 if (btrfs_item_offset_nr(leaf
, 0) + btrfs_item_size_nr(leaf
, 0) !=
482 BTRFS_LEAF_DATA_SIZE(root
)) {
483 CORRUPT("invalid item offset size pair", leaf
, root
, 0);
488 * Check to make sure each items keys are in the correct order and their
489 * offsets make sense. We only have to loop through nritems-1 because
490 * we check the current slot against the next slot, which verifies the
491 * next slot's offset+size makes sense and that the current's slot
494 for (slot
= 0; slot
< nritems
- 1; slot
++) {
495 btrfs_item_key_to_cpu(leaf
, &leaf_key
, slot
);
496 btrfs_item_key_to_cpu(leaf
, &key
, slot
+ 1);
498 /* Make sure the keys are in the right order */
499 if (btrfs_comp_cpu_keys(&leaf_key
, &key
) >= 0) {
500 CORRUPT("bad key order", leaf
, root
, slot
);
505 * Make sure the offset and ends are right, remember that the
506 * item data starts at the end of the leaf and grows towards the
509 if (btrfs_item_offset_nr(leaf
, slot
) !=
510 btrfs_item_end_nr(leaf
, slot
+ 1)) {
511 CORRUPT("slot offset bad", leaf
, root
, slot
);
516 * Check to make sure that we don't point outside of the leaf,
517 * just incase all the items are consistent to eachother, but
518 * all point outside of the leaf.
520 if (btrfs_item_end_nr(leaf
, slot
) >
521 BTRFS_LEAF_DATA_SIZE(root
)) {
522 CORRUPT("slot end outside of leaf", leaf
, root
, slot
);
530 struct extent_buffer
*find_eb_for_page(struct extent_io_tree
*tree
,
531 struct page
*page
, int max_walk
)
533 struct extent_buffer
*eb
;
534 u64 start
= page_offset(page
);
538 if (start
< max_walk
)
541 min_start
= start
- max_walk
;
543 while (start
>= min_start
) {
544 eb
= find_extent_buffer(tree
, start
, 0);
547 * we found an extent buffer and it contains our page
550 if (eb
->start
<= target
&&
551 eb
->start
+ eb
->len
> target
)
554 /* we found an extent buffer that wasn't for us */
555 free_extent_buffer(eb
);
560 start
-= PAGE_CACHE_SIZE
;
565 static int btree_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
566 struct extent_state
*state
, int mirror
)
568 struct extent_io_tree
*tree
;
571 struct extent_buffer
*eb
;
572 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
579 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
580 eb
= (struct extent_buffer
*)page
->private;
582 /* the pending IO might have been the only thing that kept this buffer
583 * in memory. Make sure we have a ref for all this other checks
585 extent_buffer_get(eb
);
587 reads_done
= atomic_dec_and_test(&eb
->io_pages
);
591 eb
->read_mirror
= mirror
;
592 if (test_bit(EXTENT_BUFFER_IOERR
, &eb
->bflags
)) {
597 found_start
= btrfs_header_bytenr(eb
);
598 if (found_start
!= eb
->start
) {
599 printk_ratelimited(KERN_INFO
"btrfs bad tree block start "
601 (unsigned long long)found_start
,
602 (unsigned long long)eb
->start
);
606 if (check_tree_block_fsid(root
, eb
)) {
607 printk_ratelimited(KERN_INFO
"btrfs bad fsid on block %llu\n",
608 (unsigned long long)eb
->start
);
612 found_level
= btrfs_header_level(eb
);
614 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb
),
617 ret
= csum_tree_block(root
, eb
, 1);
624 * If this is a leaf block and it is corrupt, set the corrupt bit so
625 * that we don't try and read the other copies of this block, just
628 if (found_level
== 0 && check_leaf(root
, eb
)) {
629 set_bit(EXTENT_BUFFER_CORRUPT
, &eb
->bflags
);
634 set_extent_buffer_uptodate(eb
);
636 if (test_bit(EXTENT_BUFFER_READAHEAD
, &eb
->bflags
)) {
637 clear_bit(EXTENT_BUFFER_READAHEAD
, &eb
->bflags
);
638 btree_readahead_hook(root
, eb
, eb
->start
, ret
);
642 clear_extent_buffer_uptodate(eb
);
643 free_extent_buffer(eb
);
648 static int btree_io_failed_hook(struct page
*page
, int failed_mirror
)
650 struct extent_buffer
*eb
;
651 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
653 eb
= (struct extent_buffer
*)page
->private;
654 set_bit(EXTENT_BUFFER_IOERR
, &eb
->bflags
);
655 eb
->read_mirror
= failed_mirror
;
656 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD
, &eb
->bflags
))
657 btree_readahead_hook(root
, eb
, eb
->start
, -EIO
);
658 return -EIO
; /* we fixed nothing */
661 static void end_workqueue_bio(struct bio
*bio
, int err
)
663 struct end_io_wq
*end_io_wq
= bio
->bi_private
;
664 struct btrfs_fs_info
*fs_info
;
666 fs_info
= end_io_wq
->info
;
667 end_io_wq
->error
= err
;
668 end_io_wq
->work
.func
= end_workqueue_fn
;
669 end_io_wq
->work
.flags
= 0;
671 if (bio
->bi_rw
& REQ_WRITE
) {
672 if (end_io_wq
->metadata
== 1)
673 btrfs_queue_worker(&fs_info
->endio_meta_write_workers
,
675 else if (end_io_wq
->metadata
== 2)
676 btrfs_queue_worker(&fs_info
->endio_freespace_worker
,
679 btrfs_queue_worker(&fs_info
->endio_write_workers
,
682 if (end_io_wq
->metadata
)
683 btrfs_queue_worker(&fs_info
->endio_meta_workers
,
686 btrfs_queue_worker(&fs_info
->endio_workers
,
692 * For the metadata arg you want
695 * 1 - if normal metadta
696 * 2 - if writing to the free space cache area
698 int btrfs_bio_wq_end_io(struct btrfs_fs_info
*info
, struct bio
*bio
,
701 struct end_io_wq
*end_io_wq
;
702 end_io_wq
= kmalloc(sizeof(*end_io_wq
), GFP_NOFS
);
706 end_io_wq
->private = bio
->bi_private
;
707 end_io_wq
->end_io
= bio
->bi_end_io
;
708 end_io_wq
->info
= info
;
709 end_io_wq
->error
= 0;
710 end_io_wq
->bio
= bio
;
711 end_io_wq
->metadata
= metadata
;
713 bio
->bi_private
= end_io_wq
;
714 bio
->bi_end_io
= end_workqueue_bio
;
718 unsigned long btrfs_async_submit_limit(struct btrfs_fs_info
*info
)
720 unsigned long limit
= min_t(unsigned long,
721 info
->workers
.max_workers
,
722 info
->fs_devices
->open_devices
);
726 static void run_one_async_start(struct btrfs_work
*work
)
728 struct async_submit_bio
*async
;
731 async
= container_of(work
, struct async_submit_bio
, work
);
732 ret
= async
->submit_bio_start(async
->inode
, async
->rw
, async
->bio
,
733 async
->mirror_num
, async
->bio_flags
,
739 static void run_one_async_done(struct btrfs_work
*work
)
741 struct btrfs_fs_info
*fs_info
;
742 struct async_submit_bio
*async
;
745 async
= container_of(work
, struct async_submit_bio
, work
);
746 fs_info
= BTRFS_I(async
->inode
)->root
->fs_info
;
748 limit
= btrfs_async_submit_limit(fs_info
);
749 limit
= limit
* 2 / 3;
751 atomic_dec(&fs_info
->nr_async_submits
);
753 if (atomic_read(&fs_info
->nr_async_submits
) < limit
&&
754 waitqueue_active(&fs_info
->async_submit_wait
))
755 wake_up(&fs_info
->async_submit_wait
);
757 /* If an error occured we just want to clean up the bio and move on */
759 bio_endio(async
->bio
, async
->error
);
763 async
->submit_bio_done(async
->inode
, async
->rw
, async
->bio
,
764 async
->mirror_num
, async
->bio_flags
,
768 static void run_one_async_free(struct btrfs_work
*work
)
770 struct async_submit_bio
*async
;
772 async
= container_of(work
, struct async_submit_bio
, work
);
776 int btrfs_wq_submit_bio(struct btrfs_fs_info
*fs_info
, struct inode
*inode
,
777 int rw
, struct bio
*bio
, int mirror_num
,
778 unsigned long bio_flags
,
780 extent_submit_bio_hook_t
*submit_bio_start
,
781 extent_submit_bio_hook_t
*submit_bio_done
)
783 struct async_submit_bio
*async
;
785 async
= kmalloc(sizeof(*async
), GFP_NOFS
);
789 async
->inode
= inode
;
792 async
->mirror_num
= mirror_num
;
793 async
->submit_bio_start
= submit_bio_start
;
794 async
->submit_bio_done
= submit_bio_done
;
796 async
->work
.func
= run_one_async_start
;
797 async
->work
.ordered_func
= run_one_async_done
;
798 async
->work
.ordered_free
= run_one_async_free
;
800 async
->work
.flags
= 0;
801 async
->bio_flags
= bio_flags
;
802 async
->bio_offset
= bio_offset
;
806 atomic_inc(&fs_info
->nr_async_submits
);
809 btrfs_set_work_high_prio(&async
->work
);
811 btrfs_queue_worker(&fs_info
->workers
, &async
->work
);
813 while (atomic_read(&fs_info
->async_submit_draining
) &&
814 atomic_read(&fs_info
->nr_async_submits
)) {
815 wait_event(fs_info
->async_submit_wait
,
816 (atomic_read(&fs_info
->nr_async_submits
) == 0));
822 static int btree_csum_one_bio(struct bio
*bio
)
824 struct bio_vec
*bvec
= bio
->bi_io_vec
;
826 struct btrfs_root
*root
;
829 WARN_ON(bio
->bi_vcnt
<= 0);
830 while (bio_index
< bio
->bi_vcnt
) {
831 root
= BTRFS_I(bvec
->bv_page
->mapping
->host
)->root
;
832 ret
= csum_dirty_buffer(root
, bvec
->bv_page
);
841 static int __btree_submit_bio_start(struct inode
*inode
, int rw
,
842 struct bio
*bio
, int mirror_num
,
843 unsigned long bio_flags
,
847 * when we're called for a write, we're already in the async
848 * submission context. Just jump into btrfs_map_bio
850 return btree_csum_one_bio(bio
);
853 static int __btree_submit_bio_done(struct inode
*inode
, int rw
, struct bio
*bio
,
854 int mirror_num
, unsigned long bio_flags
,
858 * when we're called for a write, we're already in the async
859 * submission context. Just jump into btrfs_map_bio
861 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
, mirror_num
, 1);
864 static int btree_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
865 int mirror_num
, unsigned long bio_flags
,
870 if (!(rw
& REQ_WRITE
)) {
873 * called for a read, do the setup so that checksum validation
874 * can happen in the async kernel threads
876 ret
= btrfs_bio_wq_end_io(BTRFS_I(inode
)->root
->fs_info
,
880 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
,
885 * kthread helpers are used to submit writes so that checksumming
886 * can happen in parallel across all CPUs
888 return btrfs_wq_submit_bio(BTRFS_I(inode
)->root
->fs_info
,
889 inode
, rw
, bio
, mirror_num
, 0,
891 __btree_submit_bio_start
,
892 __btree_submit_bio_done
);
895 #ifdef CONFIG_MIGRATION
896 static int btree_migratepage(struct address_space
*mapping
,
897 struct page
*newpage
, struct page
*page
,
898 enum migrate_mode mode
)
901 * we can't safely write a btree page from here,
902 * we haven't done the locking hook
907 * Buffers may be managed in a filesystem specific way.
908 * We must have no buffers or drop them.
910 if (page_has_private(page
) &&
911 !try_to_release_page(page
, GFP_KERNEL
))
913 return migrate_page(mapping
, newpage
, page
, mode
);
918 static int btree_writepages(struct address_space
*mapping
,
919 struct writeback_control
*wbc
)
921 struct extent_io_tree
*tree
;
922 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
923 if (wbc
->sync_mode
== WB_SYNC_NONE
) {
924 struct btrfs_root
*root
= BTRFS_I(mapping
->host
)->root
;
926 unsigned long thresh
= 32 * 1024 * 1024;
928 if (wbc
->for_kupdate
)
931 /* this is a bit racy, but that's ok */
932 num_dirty
= root
->fs_info
->dirty_metadata_bytes
;
933 if (num_dirty
< thresh
)
936 return btree_write_cache_pages(mapping
, wbc
);
939 static int btree_readpage(struct file
*file
, struct page
*page
)
941 struct extent_io_tree
*tree
;
942 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
943 return extent_read_full_page(tree
, page
, btree_get_extent
, 0);
946 static int btree_releasepage(struct page
*page
, gfp_t gfp_flags
)
948 if (PageWriteback(page
) || PageDirty(page
))
951 * We need to mask out eg. __GFP_HIGHMEM and __GFP_DMA32 as we're doing
952 * slab allocation from alloc_extent_state down the callchain where
953 * it'd hit a BUG_ON as those flags are not allowed.
955 gfp_flags
&= ~GFP_SLAB_BUG_MASK
;
957 return try_release_extent_buffer(page
, gfp_flags
);
960 static void btree_invalidatepage(struct page
*page
, unsigned long offset
)
962 struct extent_io_tree
*tree
;
963 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
964 extent_invalidatepage(tree
, page
, offset
);
965 btree_releasepage(page
, GFP_NOFS
);
966 if (PagePrivate(page
)) {
967 printk(KERN_WARNING
"btrfs warning page private not zero "
968 "on page %llu\n", (unsigned long long)page_offset(page
));
969 ClearPagePrivate(page
);
970 set_page_private(page
, 0);
971 page_cache_release(page
);
975 static int btree_set_page_dirty(struct page
*page
)
977 struct extent_buffer
*eb
;
979 BUG_ON(!PagePrivate(page
));
980 eb
= (struct extent_buffer
*)page
->private;
982 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY
, &eb
->bflags
));
983 BUG_ON(!atomic_read(&eb
->refs
));
984 btrfs_assert_tree_locked(eb
);
985 return __set_page_dirty_nobuffers(page
);
988 static const struct address_space_operations btree_aops
= {
989 .readpage
= btree_readpage
,
990 .writepages
= btree_writepages
,
991 .releasepage
= btree_releasepage
,
992 .invalidatepage
= btree_invalidatepage
,
993 #ifdef CONFIG_MIGRATION
994 .migratepage
= btree_migratepage
,
996 .set_page_dirty
= btree_set_page_dirty
,
999 int readahead_tree_block(struct btrfs_root
*root
, u64 bytenr
, u32 blocksize
,
1002 struct extent_buffer
*buf
= NULL
;
1003 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1006 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
1009 read_extent_buffer_pages(&BTRFS_I(btree_inode
)->io_tree
,
1010 buf
, 0, WAIT_NONE
, btree_get_extent
, 0);
1011 free_extent_buffer(buf
);
1015 int reada_tree_block_flagged(struct btrfs_root
*root
, u64 bytenr
, u32 blocksize
,
1016 int mirror_num
, struct extent_buffer
**eb
)
1018 struct extent_buffer
*buf
= NULL
;
1019 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1020 struct extent_io_tree
*io_tree
= &BTRFS_I(btree_inode
)->io_tree
;
1023 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
1027 set_bit(EXTENT_BUFFER_READAHEAD
, &buf
->bflags
);
1029 ret
= read_extent_buffer_pages(io_tree
, buf
, 0, WAIT_PAGE_LOCK
,
1030 btree_get_extent
, mirror_num
);
1032 free_extent_buffer(buf
);
1036 if (test_bit(EXTENT_BUFFER_CORRUPT
, &buf
->bflags
)) {
1037 free_extent_buffer(buf
);
1039 } else if (extent_buffer_uptodate(buf
)) {
1042 free_extent_buffer(buf
);
1047 struct extent_buffer
*btrfs_find_tree_block(struct btrfs_root
*root
,
1048 u64 bytenr
, u32 blocksize
)
1050 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1051 struct extent_buffer
*eb
;
1052 eb
= find_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
1057 struct extent_buffer
*btrfs_find_create_tree_block(struct btrfs_root
*root
,
1058 u64 bytenr
, u32 blocksize
)
1060 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1061 struct extent_buffer
*eb
;
1063 eb
= alloc_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
1069 int btrfs_write_tree_block(struct extent_buffer
*buf
)
1071 return filemap_fdatawrite_range(buf
->pages
[0]->mapping
, buf
->start
,
1072 buf
->start
+ buf
->len
- 1);
1075 int btrfs_wait_tree_block_writeback(struct extent_buffer
*buf
)
1077 return filemap_fdatawait_range(buf
->pages
[0]->mapping
,
1078 buf
->start
, buf
->start
+ buf
->len
- 1);
1081 struct extent_buffer
*read_tree_block(struct btrfs_root
*root
, u64 bytenr
,
1082 u32 blocksize
, u64 parent_transid
)
1084 struct extent_buffer
*buf
= NULL
;
1087 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
1091 ret
= btree_read_extent_buffer_pages(root
, buf
, 0, parent_transid
);
1096 void clean_tree_block(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
1097 struct extent_buffer
*buf
)
1099 if (btrfs_header_generation(buf
) ==
1100 root
->fs_info
->running_transaction
->transid
) {
1101 btrfs_assert_tree_locked(buf
);
1103 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY
, &buf
->bflags
)) {
1104 spin_lock(&root
->fs_info
->delalloc_lock
);
1105 if (root
->fs_info
->dirty_metadata_bytes
>= buf
->len
)
1106 root
->fs_info
->dirty_metadata_bytes
-= buf
->len
;
1108 spin_unlock(&root
->fs_info
->delalloc_lock
);
1109 btrfs_panic(root
->fs_info
, -EOVERFLOW
,
1110 "Can't clear %lu bytes from "
1111 " dirty_mdatadata_bytes (%lu)",
1113 root
->fs_info
->dirty_metadata_bytes
);
1115 spin_unlock(&root
->fs_info
->delalloc_lock
);
1118 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1119 btrfs_set_lock_blocking(buf
);
1120 clear_extent_buffer_dirty(buf
);
1124 static void __setup_root(u32 nodesize
, u32 leafsize
, u32 sectorsize
,
1125 u32 stripesize
, struct btrfs_root
*root
,
1126 struct btrfs_fs_info
*fs_info
,
1130 root
->commit_root
= NULL
;
1131 root
->sectorsize
= sectorsize
;
1132 root
->nodesize
= nodesize
;
1133 root
->leafsize
= leafsize
;
1134 root
->stripesize
= stripesize
;
1136 root
->track_dirty
= 0;
1138 root
->orphan_item_inserted
= 0;
1139 root
->orphan_cleanup_state
= 0;
1141 root
->objectid
= objectid
;
1142 root
->last_trans
= 0;
1143 root
->highest_objectid
= 0;
1145 root
->inode_tree
= RB_ROOT
;
1146 INIT_RADIX_TREE(&root
->delayed_nodes_tree
, GFP_ATOMIC
);
1147 root
->block_rsv
= NULL
;
1148 root
->orphan_block_rsv
= NULL
;
1150 INIT_LIST_HEAD(&root
->dirty_list
);
1151 INIT_LIST_HEAD(&root
->orphan_list
);
1152 INIT_LIST_HEAD(&root
->root_list
);
1153 spin_lock_init(&root
->orphan_lock
);
1154 spin_lock_init(&root
->inode_lock
);
1155 spin_lock_init(&root
->accounting_lock
);
1156 mutex_init(&root
->objectid_mutex
);
1157 mutex_init(&root
->log_mutex
);
1158 init_waitqueue_head(&root
->log_writer_wait
);
1159 init_waitqueue_head(&root
->log_commit_wait
[0]);
1160 init_waitqueue_head(&root
->log_commit_wait
[1]);
1161 atomic_set(&root
->log_commit
[0], 0);
1162 atomic_set(&root
->log_commit
[1], 0);
1163 atomic_set(&root
->log_writers
, 0);
1164 root
->log_batch
= 0;
1165 root
->log_transid
= 0;
1166 root
->last_log_commit
= 0;
1167 extent_io_tree_init(&root
->dirty_log_pages
,
1168 fs_info
->btree_inode
->i_mapping
);
1170 memset(&root
->root_key
, 0, sizeof(root
->root_key
));
1171 memset(&root
->root_item
, 0, sizeof(root
->root_item
));
1172 memset(&root
->defrag_progress
, 0, sizeof(root
->defrag_progress
));
1173 memset(&root
->root_kobj
, 0, sizeof(root
->root_kobj
));
1174 root
->defrag_trans_start
= fs_info
->generation
;
1175 init_completion(&root
->kobj_unregister
);
1176 root
->defrag_running
= 0;
1177 root
->root_key
.objectid
= objectid
;
1181 static int __must_check
find_and_setup_root(struct btrfs_root
*tree_root
,
1182 struct btrfs_fs_info
*fs_info
,
1184 struct btrfs_root
*root
)
1190 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
1191 tree_root
->sectorsize
, tree_root
->stripesize
,
1192 root
, fs_info
, objectid
);
1193 ret
= btrfs_find_last_root(tree_root
, objectid
,
1194 &root
->root_item
, &root
->root_key
);
1200 generation
= btrfs_root_generation(&root
->root_item
);
1201 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
1202 root
->commit_root
= NULL
;
1203 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
1204 blocksize
, generation
);
1205 if (!root
->node
|| !btrfs_buffer_uptodate(root
->node
, generation
)) {
1206 free_extent_buffer(root
->node
);
1210 root
->commit_root
= btrfs_root_node(root
);
1214 static struct btrfs_root
*btrfs_alloc_root(struct btrfs_fs_info
*fs_info
)
1216 struct btrfs_root
*root
= kzalloc(sizeof(*root
), GFP_NOFS
);
1218 root
->fs_info
= fs_info
;
1222 static struct btrfs_root
*alloc_log_tree(struct btrfs_trans_handle
*trans
,
1223 struct btrfs_fs_info
*fs_info
)
1225 struct btrfs_root
*root
;
1226 struct btrfs_root
*tree_root
= fs_info
->tree_root
;
1227 struct extent_buffer
*leaf
;
1229 root
= btrfs_alloc_root(fs_info
);
1231 return ERR_PTR(-ENOMEM
);
1233 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
1234 tree_root
->sectorsize
, tree_root
->stripesize
,
1235 root
, fs_info
, BTRFS_TREE_LOG_OBJECTID
);
1237 root
->root_key
.objectid
= BTRFS_TREE_LOG_OBJECTID
;
1238 root
->root_key
.type
= BTRFS_ROOT_ITEM_KEY
;
1239 root
->root_key
.offset
= BTRFS_TREE_LOG_OBJECTID
;
1241 * log trees do not get reference counted because they go away
1242 * before a real commit is actually done. They do store pointers
1243 * to file data extents, and those reference counts still get
1244 * updated (along with back refs to the log tree).
1248 leaf
= btrfs_alloc_free_block(trans
, root
, root
->leafsize
, 0,
1249 BTRFS_TREE_LOG_OBJECTID
, NULL
,
1253 return ERR_CAST(leaf
);
1256 memset_extent_buffer(leaf
, 0, 0, sizeof(struct btrfs_header
));
1257 btrfs_set_header_bytenr(leaf
, leaf
->start
);
1258 btrfs_set_header_generation(leaf
, trans
->transid
);
1259 btrfs_set_header_backref_rev(leaf
, BTRFS_MIXED_BACKREF_REV
);
1260 btrfs_set_header_owner(leaf
, BTRFS_TREE_LOG_OBJECTID
);
1263 write_extent_buffer(root
->node
, root
->fs_info
->fsid
,
1264 (unsigned long)btrfs_header_fsid(root
->node
),
1266 btrfs_mark_buffer_dirty(root
->node
);
1267 btrfs_tree_unlock(root
->node
);
1271 int btrfs_init_log_root_tree(struct btrfs_trans_handle
*trans
,
1272 struct btrfs_fs_info
*fs_info
)
1274 struct btrfs_root
*log_root
;
1276 log_root
= alloc_log_tree(trans
, fs_info
);
1277 if (IS_ERR(log_root
))
1278 return PTR_ERR(log_root
);
1279 WARN_ON(fs_info
->log_root_tree
);
1280 fs_info
->log_root_tree
= log_root
;
1284 int btrfs_add_log_tree(struct btrfs_trans_handle
*trans
,
1285 struct btrfs_root
*root
)
1287 struct btrfs_root
*log_root
;
1288 struct btrfs_inode_item
*inode_item
;
1290 log_root
= alloc_log_tree(trans
, root
->fs_info
);
1291 if (IS_ERR(log_root
))
1292 return PTR_ERR(log_root
);
1294 log_root
->last_trans
= trans
->transid
;
1295 log_root
->root_key
.offset
= root
->root_key
.objectid
;
1297 inode_item
= &log_root
->root_item
.inode
;
1298 inode_item
->generation
= cpu_to_le64(1);
1299 inode_item
->size
= cpu_to_le64(3);
1300 inode_item
->nlink
= cpu_to_le32(1);
1301 inode_item
->nbytes
= cpu_to_le64(root
->leafsize
);
1302 inode_item
->mode
= cpu_to_le32(S_IFDIR
| 0755);
1304 btrfs_set_root_node(&log_root
->root_item
, log_root
->node
);
1306 WARN_ON(root
->log_root
);
1307 root
->log_root
= log_root
;
1308 root
->log_transid
= 0;
1309 root
->last_log_commit
= 0;
1313 struct btrfs_root
*btrfs_read_fs_root_no_radix(struct btrfs_root
*tree_root
,
1314 struct btrfs_key
*location
)
1316 struct btrfs_root
*root
;
1317 struct btrfs_fs_info
*fs_info
= tree_root
->fs_info
;
1318 struct btrfs_path
*path
;
1319 struct extent_buffer
*l
;
1324 root
= btrfs_alloc_root(fs_info
);
1326 return ERR_PTR(-ENOMEM
);
1327 if (location
->offset
== (u64
)-1) {
1328 ret
= find_and_setup_root(tree_root
, fs_info
,
1329 location
->objectid
, root
);
1332 return ERR_PTR(ret
);
1337 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
1338 tree_root
->sectorsize
, tree_root
->stripesize
,
1339 root
, fs_info
, location
->objectid
);
1341 path
= btrfs_alloc_path();
1344 return ERR_PTR(-ENOMEM
);
1346 ret
= btrfs_search_slot(NULL
, tree_root
, location
, path
, 0, 0);
1349 read_extent_buffer(l
, &root
->root_item
,
1350 btrfs_item_ptr_offset(l
, path
->slots
[0]),
1351 sizeof(root
->root_item
));
1352 memcpy(&root
->root_key
, location
, sizeof(*location
));
1354 btrfs_free_path(path
);
1359 return ERR_PTR(ret
);
1362 generation
= btrfs_root_generation(&root
->root_item
);
1363 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
1364 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
1365 blocksize
, generation
);
1366 root
->commit_root
= btrfs_root_node(root
);
1367 BUG_ON(!root
->node
); /* -ENOMEM */
1369 if (location
->objectid
!= BTRFS_TREE_LOG_OBJECTID
) {
1371 btrfs_check_and_init_root_item(&root
->root_item
);
1377 struct btrfs_root
*btrfs_read_fs_root_no_name(struct btrfs_fs_info
*fs_info
,
1378 struct btrfs_key
*location
)
1380 struct btrfs_root
*root
;
1383 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1384 return fs_info
->tree_root
;
1385 if (location
->objectid
== BTRFS_EXTENT_TREE_OBJECTID
)
1386 return fs_info
->extent_root
;
1387 if (location
->objectid
== BTRFS_CHUNK_TREE_OBJECTID
)
1388 return fs_info
->chunk_root
;
1389 if (location
->objectid
== BTRFS_DEV_TREE_OBJECTID
)
1390 return fs_info
->dev_root
;
1391 if (location
->objectid
== BTRFS_CSUM_TREE_OBJECTID
)
1392 return fs_info
->csum_root
;
1394 spin_lock(&fs_info
->fs_roots_radix_lock
);
1395 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
1396 (unsigned long)location
->objectid
);
1397 spin_unlock(&fs_info
->fs_roots_radix_lock
);
1401 root
= btrfs_read_fs_root_no_radix(fs_info
->tree_root
, location
);
1405 root
->free_ino_ctl
= kzalloc(sizeof(*root
->free_ino_ctl
), GFP_NOFS
);
1406 root
->free_ino_pinned
= kzalloc(sizeof(*root
->free_ino_pinned
),
1408 if (!root
->free_ino_pinned
|| !root
->free_ino_ctl
) {
1413 btrfs_init_free_ino_ctl(root
);
1414 mutex_init(&root
->fs_commit_mutex
);
1415 spin_lock_init(&root
->cache_lock
);
1416 init_waitqueue_head(&root
->cache_wait
);
1418 ret
= get_anon_bdev(&root
->anon_dev
);
1422 if (btrfs_root_refs(&root
->root_item
) == 0) {
1427 ret
= btrfs_find_orphan_item(fs_info
->tree_root
, location
->objectid
);
1431 root
->orphan_item_inserted
= 1;
1433 ret
= radix_tree_preload(GFP_NOFS
& ~__GFP_HIGHMEM
);
1437 spin_lock(&fs_info
->fs_roots_radix_lock
);
1438 ret
= radix_tree_insert(&fs_info
->fs_roots_radix
,
1439 (unsigned long)root
->root_key
.objectid
,
1444 spin_unlock(&fs_info
->fs_roots_radix_lock
);
1445 radix_tree_preload_end();
1447 if (ret
== -EEXIST
) {
1454 ret
= btrfs_find_dead_roots(fs_info
->tree_root
,
1455 root
->root_key
.objectid
);
1460 return ERR_PTR(ret
);
1463 static int btrfs_congested_fn(void *congested_data
, int bdi_bits
)
1465 struct btrfs_fs_info
*info
= (struct btrfs_fs_info
*)congested_data
;
1467 struct btrfs_device
*device
;
1468 struct backing_dev_info
*bdi
;
1471 list_for_each_entry_rcu(device
, &info
->fs_devices
->devices
, dev_list
) {
1474 bdi
= blk_get_backing_dev_info(device
->bdev
);
1475 if (bdi
&& bdi_congested(bdi
, bdi_bits
)) {
1485 * If this fails, caller must call bdi_destroy() to get rid of the
1488 static int setup_bdi(struct btrfs_fs_info
*info
, struct backing_dev_info
*bdi
)
1492 bdi
->capabilities
= BDI_CAP_MAP_COPY
;
1493 err
= bdi_setup_and_register(bdi
, "btrfs", BDI_CAP_MAP_COPY
);
1497 bdi
->ra_pages
= default_backing_dev_info
.ra_pages
;
1498 bdi
->congested_fn
= btrfs_congested_fn
;
1499 bdi
->congested_data
= info
;
1504 * called by the kthread helper functions to finally call the bio end_io
1505 * functions. This is where read checksum verification actually happens
1507 static void end_workqueue_fn(struct btrfs_work
*work
)
1510 struct end_io_wq
*end_io_wq
;
1511 struct btrfs_fs_info
*fs_info
;
1514 end_io_wq
= container_of(work
, struct end_io_wq
, work
);
1515 bio
= end_io_wq
->bio
;
1516 fs_info
= end_io_wq
->info
;
1518 error
= end_io_wq
->error
;
1519 bio
->bi_private
= end_io_wq
->private;
1520 bio
->bi_end_io
= end_io_wq
->end_io
;
1522 bio_endio(bio
, error
);
1525 static int cleaner_kthread(void *arg
)
1527 struct btrfs_root
*root
= arg
;
1530 vfs_check_frozen(root
->fs_info
->sb
, SB_FREEZE_WRITE
);
1532 if (!(root
->fs_info
->sb
->s_flags
& MS_RDONLY
) &&
1533 mutex_trylock(&root
->fs_info
->cleaner_mutex
)) {
1534 btrfs_run_delayed_iputs(root
);
1535 btrfs_clean_old_snapshots(root
);
1536 mutex_unlock(&root
->fs_info
->cleaner_mutex
);
1537 btrfs_run_defrag_inodes(root
->fs_info
);
1540 if (!try_to_freeze()) {
1541 set_current_state(TASK_INTERRUPTIBLE
);
1542 if (!kthread_should_stop())
1544 __set_current_state(TASK_RUNNING
);
1546 } while (!kthread_should_stop());
1550 static int transaction_kthread(void *arg
)
1552 struct btrfs_root
*root
= arg
;
1553 struct btrfs_trans_handle
*trans
;
1554 struct btrfs_transaction
*cur
;
1557 unsigned long delay
;
1561 cannot_commit
= false;
1563 vfs_check_frozen(root
->fs_info
->sb
, SB_FREEZE_WRITE
);
1564 mutex_lock(&root
->fs_info
->transaction_kthread_mutex
);
1566 spin_lock(&root
->fs_info
->trans_lock
);
1567 cur
= root
->fs_info
->running_transaction
;
1569 spin_unlock(&root
->fs_info
->trans_lock
);
1573 now
= get_seconds();
1574 if (!cur
->blocked
&&
1575 (now
< cur
->start_time
|| now
- cur
->start_time
< 30)) {
1576 spin_unlock(&root
->fs_info
->trans_lock
);
1580 transid
= cur
->transid
;
1581 spin_unlock(&root
->fs_info
->trans_lock
);
1583 /* If the file system is aborted, this will always fail. */
1584 trans
= btrfs_join_transaction(root
);
1585 if (IS_ERR(trans
)) {
1586 cannot_commit
= true;
1589 if (transid
== trans
->transid
) {
1590 btrfs_commit_transaction(trans
, root
);
1592 btrfs_end_transaction(trans
, root
);
1595 wake_up_process(root
->fs_info
->cleaner_kthread
);
1596 mutex_unlock(&root
->fs_info
->transaction_kthread_mutex
);
1598 if (!try_to_freeze()) {
1599 set_current_state(TASK_INTERRUPTIBLE
);
1600 if (!kthread_should_stop() &&
1601 (!btrfs_transaction_blocked(root
->fs_info
) ||
1603 schedule_timeout(delay
);
1604 __set_current_state(TASK_RUNNING
);
1606 } while (!kthread_should_stop());
1611 * this will find the highest generation in the array of
1612 * root backups. The index of the highest array is returned,
1613 * or -1 if we can't find anything.
1615 * We check to make sure the array is valid by comparing the
1616 * generation of the latest root in the array with the generation
1617 * in the super block. If they don't match we pitch it.
1619 static int find_newest_super_backup(struct btrfs_fs_info
*info
, u64 newest_gen
)
1622 int newest_index
= -1;
1623 struct btrfs_root_backup
*root_backup
;
1626 for (i
= 0; i
< BTRFS_NUM_BACKUP_ROOTS
; i
++) {
1627 root_backup
= info
->super_copy
->super_roots
+ i
;
1628 cur
= btrfs_backup_tree_root_gen(root_backup
);
1629 if (cur
== newest_gen
)
1633 /* check to see if we actually wrapped around */
1634 if (newest_index
== BTRFS_NUM_BACKUP_ROOTS
- 1) {
1635 root_backup
= info
->super_copy
->super_roots
;
1636 cur
= btrfs_backup_tree_root_gen(root_backup
);
1637 if (cur
== newest_gen
)
1640 return newest_index
;
1645 * find the oldest backup so we know where to store new entries
1646 * in the backup array. This will set the backup_root_index
1647 * field in the fs_info struct
1649 static void find_oldest_super_backup(struct btrfs_fs_info
*info
,
1652 int newest_index
= -1;
1654 newest_index
= find_newest_super_backup(info
, newest_gen
);
1655 /* if there was garbage in there, just move along */
1656 if (newest_index
== -1) {
1657 info
->backup_root_index
= 0;
1659 info
->backup_root_index
= (newest_index
+ 1) % BTRFS_NUM_BACKUP_ROOTS
;
1664 * copy all the root pointers into the super backup array.
1665 * this will bump the backup pointer by one when it is
1668 static void backup_super_roots(struct btrfs_fs_info
*info
)
1671 struct btrfs_root_backup
*root_backup
;
1674 next_backup
= info
->backup_root_index
;
1675 last_backup
= (next_backup
+ BTRFS_NUM_BACKUP_ROOTS
- 1) %
1676 BTRFS_NUM_BACKUP_ROOTS
;
1679 * just overwrite the last backup if we're at the same generation
1680 * this happens only at umount
1682 root_backup
= info
->super_for_commit
->super_roots
+ last_backup
;
1683 if (btrfs_backup_tree_root_gen(root_backup
) ==
1684 btrfs_header_generation(info
->tree_root
->node
))
1685 next_backup
= last_backup
;
1687 root_backup
= info
->super_for_commit
->super_roots
+ next_backup
;
1690 * make sure all of our padding and empty slots get zero filled
1691 * regardless of which ones we use today
1693 memset(root_backup
, 0, sizeof(*root_backup
));
1695 info
->backup_root_index
= (next_backup
+ 1) % BTRFS_NUM_BACKUP_ROOTS
;
1697 btrfs_set_backup_tree_root(root_backup
, info
->tree_root
->node
->start
);
1698 btrfs_set_backup_tree_root_gen(root_backup
,
1699 btrfs_header_generation(info
->tree_root
->node
));
1701 btrfs_set_backup_tree_root_level(root_backup
,
1702 btrfs_header_level(info
->tree_root
->node
));
1704 btrfs_set_backup_chunk_root(root_backup
, info
->chunk_root
->node
->start
);
1705 btrfs_set_backup_chunk_root_gen(root_backup
,
1706 btrfs_header_generation(info
->chunk_root
->node
));
1707 btrfs_set_backup_chunk_root_level(root_backup
,
1708 btrfs_header_level(info
->chunk_root
->node
));
1710 btrfs_set_backup_extent_root(root_backup
, info
->extent_root
->node
->start
);
1711 btrfs_set_backup_extent_root_gen(root_backup
,
1712 btrfs_header_generation(info
->extent_root
->node
));
1713 btrfs_set_backup_extent_root_level(root_backup
,
1714 btrfs_header_level(info
->extent_root
->node
));
1717 * we might commit during log recovery, which happens before we set
1718 * the fs_root. Make sure it is valid before we fill it in.
1720 if (info
->fs_root
&& info
->fs_root
->node
) {
1721 btrfs_set_backup_fs_root(root_backup
,
1722 info
->fs_root
->node
->start
);
1723 btrfs_set_backup_fs_root_gen(root_backup
,
1724 btrfs_header_generation(info
->fs_root
->node
));
1725 btrfs_set_backup_fs_root_level(root_backup
,
1726 btrfs_header_level(info
->fs_root
->node
));
1729 btrfs_set_backup_dev_root(root_backup
, info
->dev_root
->node
->start
);
1730 btrfs_set_backup_dev_root_gen(root_backup
,
1731 btrfs_header_generation(info
->dev_root
->node
));
1732 btrfs_set_backup_dev_root_level(root_backup
,
1733 btrfs_header_level(info
->dev_root
->node
));
1735 btrfs_set_backup_csum_root(root_backup
, info
->csum_root
->node
->start
);
1736 btrfs_set_backup_csum_root_gen(root_backup
,
1737 btrfs_header_generation(info
->csum_root
->node
));
1738 btrfs_set_backup_csum_root_level(root_backup
,
1739 btrfs_header_level(info
->csum_root
->node
));
1741 btrfs_set_backup_total_bytes(root_backup
,
1742 btrfs_super_total_bytes(info
->super_copy
));
1743 btrfs_set_backup_bytes_used(root_backup
,
1744 btrfs_super_bytes_used(info
->super_copy
));
1745 btrfs_set_backup_num_devices(root_backup
,
1746 btrfs_super_num_devices(info
->super_copy
));
1749 * if we don't copy this out to the super_copy, it won't get remembered
1750 * for the next commit
1752 memcpy(&info
->super_copy
->super_roots
,
1753 &info
->super_for_commit
->super_roots
,
1754 sizeof(*root_backup
) * BTRFS_NUM_BACKUP_ROOTS
);
1758 * this copies info out of the root backup array and back into
1759 * the in-memory super block. It is meant to help iterate through
1760 * the array, so you send it the number of backups you've already
1761 * tried and the last backup index you used.
1763 * this returns -1 when it has tried all the backups
1765 static noinline
int next_root_backup(struct btrfs_fs_info
*info
,
1766 struct btrfs_super_block
*super
,
1767 int *num_backups_tried
, int *backup_index
)
1769 struct btrfs_root_backup
*root_backup
;
1770 int newest
= *backup_index
;
1772 if (*num_backups_tried
== 0) {
1773 u64 gen
= btrfs_super_generation(super
);
1775 newest
= find_newest_super_backup(info
, gen
);
1779 *backup_index
= newest
;
1780 *num_backups_tried
= 1;
1781 } else if (*num_backups_tried
== BTRFS_NUM_BACKUP_ROOTS
) {
1782 /* we've tried all the backups, all done */
1785 /* jump to the next oldest backup */
1786 newest
= (*backup_index
+ BTRFS_NUM_BACKUP_ROOTS
- 1) %
1787 BTRFS_NUM_BACKUP_ROOTS
;
1788 *backup_index
= newest
;
1789 *num_backups_tried
+= 1;
1791 root_backup
= super
->super_roots
+ newest
;
1793 btrfs_set_super_generation(super
,
1794 btrfs_backup_tree_root_gen(root_backup
));
1795 btrfs_set_super_root(super
, btrfs_backup_tree_root(root_backup
));
1796 btrfs_set_super_root_level(super
,
1797 btrfs_backup_tree_root_level(root_backup
));
1798 btrfs_set_super_bytes_used(super
, btrfs_backup_bytes_used(root_backup
));
1801 * fixme: the total bytes and num_devices need to match or we should
1804 btrfs_set_super_total_bytes(super
, btrfs_backup_total_bytes(root_backup
));
1805 btrfs_set_super_num_devices(super
, btrfs_backup_num_devices(root_backup
));
1809 /* helper to cleanup tree roots */
1810 static void free_root_pointers(struct btrfs_fs_info
*info
, int chunk_root
)
1812 free_extent_buffer(info
->tree_root
->node
);
1813 free_extent_buffer(info
->tree_root
->commit_root
);
1814 free_extent_buffer(info
->dev_root
->node
);
1815 free_extent_buffer(info
->dev_root
->commit_root
);
1816 free_extent_buffer(info
->extent_root
->node
);
1817 free_extent_buffer(info
->extent_root
->commit_root
);
1818 free_extent_buffer(info
->csum_root
->node
);
1819 free_extent_buffer(info
->csum_root
->commit_root
);
1821 info
->tree_root
->node
= NULL
;
1822 info
->tree_root
->commit_root
= NULL
;
1823 info
->dev_root
->node
= NULL
;
1824 info
->dev_root
->commit_root
= NULL
;
1825 info
->extent_root
->node
= NULL
;
1826 info
->extent_root
->commit_root
= NULL
;
1827 info
->csum_root
->node
= NULL
;
1828 info
->csum_root
->commit_root
= NULL
;
1831 free_extent_buffer(info
->chunk_root
->node
);
1832 free_extent_buffer(info
->chunk_root
->commit_root
);
1833 info
->chunk_root
->node
= NULL
;
1834 info
->chunk_root
->commit_root
= NULL
;
1839 int open_ctree(struct super_block
*sb
,
1840 struct btrfs_fs_devices
*fs_devices
,
1850 struct btrfs_key location
;
1851 struct buffer_head
*bh
;
1852 struct btrfs_super_block
*disk_super
;
1853 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
1854 struct btrfs_root
*tree_root
;
1855 struct btrfs_root
*extent_root
;
1856 struct btrfs_root
*csum_root
;
1857 struct btrfs_root
*chunk_root
;
1858 struct btrfs_root
*dev_root
;
1859 struct btrfs_root
*log_tree_root
;
1862 int num_backups_tried
= 0;
1863 int backup_index
= 0;
1865 tree_root
= fs_info
->tree_root
= btrfs_alloc_root(fs_info
);
1866 extent_root
= fs_info
->extent_root
= btrfs_alloc_root(fs_info
);
1867 csum_root
= fs_info
->csum_root
= btrfs_alloc_root(fs_info
);
1868 chunk_root
= fs_info
->chunk_root
= btrfs_alloc_root(fs_info
);
1869 dev_root
= fs_info
->dev_root
= btrfs_alloc_root(fs_info
);
1871 if (!tree_root
|| !extent_root
|| !csum_root
||
1872 !chunk_root
|| !dev_root
) {
1877 ret
= init_srcu_struct(&fs_info
->subvol_srcu
);
1883 ret
= setup_bdi(fs_info
, &fs_info
->bdi
);
1889 fs_info
->btree_inode
= new_inode(sb
);
1890 if (!fs_info
->btree_inode
) {
1895 mapping_set_gfp_mask(fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1897 INIT_RADIX_TREE(&fs_info
->fs_roots_radix
, GFP_ATOMIC
);
1898 INIT_LIST_HEAD(&fs_info
->trans_list
);
1899 INIT_LIST_HEAD(&fs_info
->dead_roots
);
1900 INIT_LIST_HEAD(&fs_info
->delayed_iputs
);
1901 INIT_LIST_HEAD(&fs_info
->hashers
);
1902 INIT_LIST_HEAD(&fs_info
->delalloc_inodes
);
1903 INIT_LIST_HEAD(&fs_info
->ordered_operations
);
1904 INIT_LIST_HEAD(&fs_info
->caching_block_groups
);
1905 spin_lock_init(&fs_info
->delalloc_lock
);
1906 spin_lock_init(&fs_info
->trans_lock
);
1907 spin_lock_init(&fs_info
->ref_cache_lock
);
1908 spin_lock_init(&fs_info
->fs_roots_radix_lock
);
1909 spin_lock_init(&fs_info
->delayed_iput_lock
);
1910 spin_lock_init(&fs_info
->defrag_inodes_lock
);
1911 spin_lock_init(&fs_info
->free_chunk_lock
);
1912 mutex_init(&fs_info
->reloc_mutex
);
1914 init_completion(&fs_info
->kobj_unregister
);
1915 INIT_LIST_HEAD(&fs_info
->dirty_cowonly_roots
);
1916 INIT_LIST_HEAD(&fs_info
->space_info
);
1917 btrfs_mapping_init(&fs_info
->mapping_tree
);
1918 btrfs_init_block_rsv(&fs_info
->global_block_rsv
);
1919 btrfs_init_block_rsv(&fs_info
->delalloc_block_rsv
);
1920 btrfs_init_block_rsv(&fs_info
->trans_block_rsv
);
1921 btrfs_init_block_rsv(&fs_info
->chunk_block_rsv
);
1922 btrfs_init_block_rsv(&fs_info
->empty_block_rsv
);
1923 btrfs_init_block_rsv(&fs_info
->delayed_block_rsv
);
1924 atomic_set(&fs_info
->nr_async_submits
, 0);
1925 atomic_set(&fs_info
->async_delalloc_pages
, 0);
1926 atomic_set(&fs_info
->async_submit_draining
, 0);
1927 atomic_set(&fs_info
->nr_async_bios
, 0);
1928 atomic_set(&fs_info
->defrag_running
, 0);
1930 fs_info
->max_inline
= 8192 * 1024;
1931 fs_info
->metadata_ratio
= 0;
1932 fs_info
->defrag_inodes
= RB_ROOT
;
1933 fs_info
->trans_no_join
= 0;
1934 fs_info
->free_chunk_space
= 0;
1936 /* readahead state */
1937 INIT_RADIX_TREE(&fs_info
->reada_tree
, GFP_NOFS
& ~__GFP_WAIT
);
1938 spin_lock_init(&fs_info
->reada_lock
);
1940 fs_info
->thread_pool_size
= min_t(unsigned long,
1941 num_online_cpus() + 2, 8);
1943 INIT_LIST_HEAD(&fs_info
->ordered_extents
);
1944 spin_lock_init(&fs_info
->ordered_extent_lock
);
1945 fs_info
->delayed_root
= kmalloc(sizeof(struct btrfs_delayed_root
),
1947 if (!fs_info
->delayed_root
) {
1951 btrfs_init_delayed_root(fs_info
->delayed_root
);
1953 mutex_init(&fs_info
->scrub_lock
);
1954 atomic_set(&fs_info
->scrubs_running
, 0);
1955 atomic_set(&fs_info
->scrub_pause_req
, 0);
1956 atomic_set(&fs_info
->scrubs_paused
, 0);
1957 atomic_set(&fs_info
->scrub_cancel_req
, 0);
1958 init_waitqueue_head(&fs_info
->scrub_pause_wait
);
1959 init_rwsem(&fs_info
->scrub_super_lock
);
1960 fs_info
->scrub_workers_refcnt
= 0;
1961 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1962 fs_info
->check_integrity_print_mask
= 0;
1965 spin_lock_init(&fs_info
->balance_lock
);
1966 mutex_init(&fs_info
->balance_mutex
);
1967 atomic_set(&fs_info
->balance_running
, 0);
1968 atomic_set(&fs_info
->balance_pause_req
, 0);
1969 atomic_set(&fs_info
->balance_cancel_req
, 0);
1970 fs_info
->balance_ctl
= NULL
;
1971 init_waitqueue_head(&fs_info
->balance_wait_q
);
1973 sb
->s_blocksize
= 4096;
1974 sb
->s_blocksize_bits
= blksize_bits(4096);
1975 sb
->s_bdi
= &fs_info
->bdi
;
1977 fs_info
->btree_inode
->i_ino
= BTRFS_BTREE_INODE_OBJECTID
;
1978 set_nlink(fs_info
->btree_inode
, 1);
1980 * we set the i_size on the btree inode to the max possible int.
1981 * the real end of the address space is determined by all of
1982 * the devices in the system
1984 fs_info
->btree_inode
->i_size
= OFFSET_MAX
;
1985 fs_info
->btree_inode
->i_mapping
->a_ops
= &btree_aops
;
1986 fs_info
->btree_inode
->i_mapping
->backing_dev_info
= &fs_info
->bdi
;
1988 RB_CLEAR_NODE(&BTRFS_I(fs_info
->btree_inode
)->rb_node
);
1989 extent_io_tree_init(&BTRFS_I(fs_info
->btree_inode
)->io_tree
,
1990 fs_info
->btree_inode
->i_mapping
);
1991 BTRFS_I(fs_info
->btree_inode
)->io_tree
.track_uptodate
= 0;
1992 extent_map_tree_init(&BTRFS_I(fs_info
->btree_inode
)->extent_tree
);
1994 BTRFS_I(fs_info
->btree_inode
)->io_tree
.ops
= &btree_extent_io_ops
;
1996 BTRFS_I(fs_info
->btree_inode
)->root
= tree_root
;
1997 memset(&BTRFS_I(fs_info
->btree_inode
)->location
, 0,
1998 sizeof(struct btrfs_key
));
1999 BTRFS_I(fs_info
->btree_inode
)->dummy_inode
= 1;
2000 insert_inode_hash(fs_info
->btree_inode
);
2002 spin_lock_init(&fs_info
->block_group_cache_lock
);
2003 fs_info
->block_group_cache_tree
= RB_ROOT
;
2005 extent_io_tree_init(&fs_info
->freed_extents
[0],
2006 fs_info
->btree_inode
->i_mapping
);
2007 extent_io_tree_init(&fs_info
->freed_extents
[1],
2008 fs_info
->btree_inode
->i_mapping
);
2009 fs_info
->pinned_extents
= &fs_info
->freed_extents
[0];
2010 fs_info
->do_barriers
= 1;
2013 mutex_init(&fs_info
->ordered_operations_mutex
);
2014 mutex_init(&fs_info
->tree_log_mutex
);
2015 mutex_init(&fs_info
->chunk_mutex
);
2016 mutex_init(&fs_info
->transaction_kthread_mutex
);
2017 mutex_init(&fs_info
->cleaner_mutex
);
2018 mutex_init(&fs_info
->volume_mutex
);
2019 init_rwsem(&fs_info
->extent_commit_sem
);
2020 init_rwsem(&fs_info
->cleanup_work_sem
);
2021 init_rwsem(&fs_info
->subvol_sem
);
2023 btrfs_init_free_cluster(&fs_info
->meta_alloc_cluster
);
2024 btrfs_init_free_cluster(&fs_info
->data_alloc_cluster
);
2026 init_waitqueue_head(&fs_info
->transaction_throttle
);
2027 init_waitqueue_head(&fs_info
->transaction_wait
);
2028 init_waitqueue_head(&fs_info
->transaction_blocked_wait
);
2029 init_waitqueue_head(&fs_info
->async_submit_wait
);
2031 __setup_root(4096, 4096, 4096, 4096, tree_root
,
2032 fs_info
, BTRFS_ROOT_TREE_OBJECTID
);
2034 invalidate_bdev(fs_devices
->latest_bdev
);
2035 bh
= btrfs_read_dev_super(fs_devices
->latest_bdev
);
2041 memcpy(fs_info
->super_copy
, bh
->b_data
, sizeof(*fs_info
->super_copy
));
2042 memcpy(fs_info
->super_for_commit
, fs_info
->super_copy
,
2043 sizeof(*fs_info
->super_for_commit
));
2046 memcpy(fs_info
->fsid
, fs_info
->super_copy
->fsid
, BTRFS_FSID_SIZE
);
2048 disk_super
= fs_info
->super_copy
;
2049 if (!btrfs_super_root(disk_super
))
2052 /* check FS state, whether FS is broken. */
2053 fs_info
->fs_state
|= btrfs_super_flags(disk_super
);
2055 ret
= btrfs_check_super_valid(fs_info
, sb
->s_flags
& MS_RDONLY
);
2057 printk(KERN_ERR
"btrfs: superblock contains fatal errors\n");
2063 * run through our array of backup supers and setup
2064 * our ring pointer to the oldest one
2066 generation
= btrfs_super_generation(disk_super
);
2067 find_oldest_super_backup(fs_info
, generation
);
2070 * In the long term, we'll store the compression type in the super
2071 * block, and it'll be used for per file compression control.
2073 fs_info
->compress_type
= BTRFS_COMPRESS_ZLIB
;
2075 ret
= btrfs_parse_options(tree_root
, options
);
2081 features
= btrfs_super_incompat_flags(disk_super
) &
2082 ~BTRFS_FEATURE_INCOMPAT_SUPP
;
2084 printk(KERN_ERR
"BTRFS: couldn't mount because of "
2085 "unsupported optional features (%Lx).\n",
2086 (unsigned long long)features
);
2091 if (btrfs_super_leafsize(disk_super
) !=
2092 btrfs_super_nodesize(disk_super
)) {
2093 printk(KERN_ERR
"BTRFS: couldn't mount because metadata "
2094 "blocksizes don't match. node %d leaf %d\n",
2095 btrfs_super_nodesize(disk_super
),
2096 btrfs_super_leafsize(disk_super
));
2100 if (btrfs_super_leafsize(disk_super
) > BTRFS_MAX_METADATA_BLOCKSIZE
) {
2101 printk(KERN_ERR
"BTRFS: couldn't mount because metadata "
2102 "blocksize (%d) was too large\n",
2103 btrfs_super_leafsize(disk_super
));
2108 features
= btrfs_super_incompat_flags(disk_super
);
2109 features
|= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF
;
2110 if (tree_root
->fs_info
->compress_type
& BTRFS_COMPRESS_LZO
)
2111 features
|= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO
;
2114 * flag our filesystem as having big metadata blocks if
2115 * they are bigger than the page size
2117 if (btrfs_super_leafsize(disk_super
) > PAGE_CACHE_SIZE
) {
2118 if (!(features
& BTRFS_FEATURE_INCOMPAT_BIG_METADATA
))
2119 printk(KERN_INFO
"btrfs flagging fs with big metadata feature\n");
2120 features
|= BTRFS_FEATURE_INCOMPAT_BIG_METADATA
;
2123 nodesize
= btrfs_super_nodesize(disk_super
);
2124 leafsize
= btrfs_super_leafsize(disk_super
);
2125 sectorsize
= btrfs_super_sectorsize(disk_super
);
2126 stripesize
= btrfs_super_stripesize(disk_super
);
2129 * mixed block groups end up with duplicate but slightly offset
2130 * extent buffers for the same range. It leads to corruptions
2132 if ((features
& BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS
) &&
2133 (sectorsize
!= leafsize
)) {
2134 printk(KERN_WARNING
"btrfs: unequal leaf/node/sector sizes "
2135 "are not allowed for mixed block groups on %s\n",
2140 btrfs_set_super_incompat_flags(disk_super
, features
);
2142 features
= btrfs_super_compat_ro_flags(disk_super
) &
2143 ~BTRFS_FEATURE_COMPAT_RO_SUPP
;
2144 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
2145 printk(KERN_ERR
"BTRFS: couldn't mount RDWR because of "
2146 "unsupported option features (%Lx).\n",
2147 (unsigned long long)features
);
2152 btrfs_init_workers(&fs_info
->generic_worker
,
2153 "genwork", 1, NULL
);
2155 btrfs_init_workers(&fs_info
->workers
, "worker",
2156 fs_info
->thread_pool_size
,
2157 &fs_info
->generic_worker
);
2159 btrfs_init_workers(&fs_info
->delalloc_workers
, "delalloc",
2160 fs_info
->thread_pool_size
,
2161 &fs_info
->generic_worker
);
2163 btrfs_init_workers(&fs_info
->submit_workers
, "submit",
2164 min_t(u64
, fs_devices
->num_devices
,
2165 fs_info
->thread_pool_size
),
2166 &fs_info
->generic_worker
);
2168 btrfs_init_workers(&fs_info
->caching_workers
, "cache",
2169 2, &fs_info
->generic_worker
);
2171 /* a higher idle thresh on the submit workers makes it much more
2172 * likely that bios will be send down in a sane order to the
2175 fs_info
->submit_workers
.idle_thresh
= 64;
2177 fs_info
->workers
.idle_thresh
= 16;
2178 fs_info
->workers
.ordered
= 1;
2180 fs_info
->delalloc_workers
.idle_thresh
= 2;
2181 fs_info
->delalloc_workers
.ordered
= 1;
2183 btrfs_init_workers(&fs_info
->fixup_workers
, "fixup", 1,
2184 &fs_info
->generic_worker
);
2185 btrfs_init_workers(&fs_info
->endio_workers
, "endio",
2186 fs_info
->thread_pool_size
,
2187 &fs_info
->generic_worker
);
2188 btrfs_init_workers(&fs_info
->endio_meta_workers
, "endio-meta",
2189 fs_info
->thread_pool_size
,
2190 &fs_info
->generic_worker
);
2191 btrfs_init_workers(&fs_info
->endio_meta_write_workers
,
2192 "endio-meta-write", fs_info
->thread_pool_size
,
2193 &fs_info
->generic_worker
);
2194 btrfs_init_workers(&fs_info
->endio_write_workers
, "endio-write",
2195 fs_info
->thread_pool_size
,
2196 &fs_info
->generic_worker
);
2197 btrfs_init_workers(&fs_info
->endio_freespace_worker
, "freespace-write",
2198 1, &fs_info
->generic_worker
);
2199 btrfs_init_workers(&fs_info
->delayed_workers
, "delayed-meta",
2200 fs_info
->thread_pool_size
,
2201 &fs_info
->generic_worker
);
2202 btrfs_init_workers(&fs_info
->readahead_workers
, "readahead",
2203 fs_info
->thread_pool_size
,
2204 &fs_info
->generic_worker
);
2207 * endios are largely parallel and should have a very
2210 fs_info
->endio_workers
.idle_thresh
= 4;
2211 fs_info
->endio_meta_workers
.idle_thresh
= 4;
2213 fs_info
->endio_write_workers
.idle_thresh
= 2;
2214 fs_info
->endio_meta_write_workers
.idle_thresh
= 2;
2215 fs_info
->readahead_workers
.idle_thresh
= 2;
2218 * btrfs_start_workers can really only fail because of ENOMEM so just
2219 * return -ENOMEM if any of these fail.
2221 ret
= btrfs_start_workers(&fs_info
->workers
);
2222 ret
|= btrfs_start_workers(&fs_info
->generic_worker
);
2223 ret
|= btrfs_start_workers(&fs_info
->submit_workers
);
2224 ret
|= btrfs_start_workers(&fs_info
->delalloc_workers
);
2225 ret
|= btrfs_start_workers(&fs_info
->fixup_workers
);
2226 ret
|= btrfs_start_workers(&fs_info
->endio_workers
);
2227 ret
|= btrfs_start_workers(&fs_info
->endio_meta_workers
);
2228 ret
|= btrfs_start_workers(&fs_info
->endio_meta_write_workers
);
2229 ret
|= btrfs_start_workers(&fs_info
->endio_write_workers
);
2230 ret
|= btrfs_start_workers(&fs_info
->endio_freespace_worker
);
2231 ret
|= btrfs_start_workers(&fs_info
->delayed_workers
);
2232 ret
|= btrfs_start_workers(&fs_info
->caching_workers
);
2233 ret
|= btrfs_start_workers(&fs_info
->readahead_workers
);
2236 goto fail_sb_buffer
;
2239 fs_info
->bdi
.ra_pages
*= btrfs_super_num_devices(disk_super
);
2240 fs_info
->bdi
.ra_pages
= max(fs_info
->bdi
.ra_pages
,
2241 4 * 1024 * 1024 / PAGE_CACHE_SIZE
);
2243 tree_root
->nodesize
= nodesize
;
2244 tree_root
->leafsize
= leafsize
;
2245 tree_root
->sectorsize
= sectorsize
;
2246 tree_root
->stripesize
= stripesize
;
2248 sb
->s_blocksize
= sectorsize
;
2249 sb
->s_blocksize_bits
= blksize_bits(sectorsize
);
2251 if (strncmp((char *)(&disk_super
->magic
), BTRFS_MAGIC
,
2252 sizeof(disk_super
->magic
))) {
2253 printk(KERN_INFO
"btrfs: valid FS not found on %s\n", sb
->s_id
);
2254 goto fail_sb_buffer
;
2257 if (sectorsize
!= PAGE_SIZE
) {
2258 printk(KERN_WARNING
"btrfs: Incompatible sector size(%lu) "
2259 "found on %s\n", (unsigned long)sectorsize
, sb
->s_id
);
2260 goto fail_sb_buffer
;
2263 mutex_lock(&fs_info
->chunk_mutex
);
2264 ret
= btrfs_read_sys_array(tree_root
);
2265 mutex_unlock(&fs_info
->chunk_mutex
);
2267 printk(KERN_WARNING
"btrfs: failed to read the system "
2268 "array on %s\n", sb
->s_id
);
2269 goto fail_sb_buffer
;
2272 blocksize
= btrfs_level_size(tree_root
,
2273 btrfs_super_chunk_root_level(disk_super
));
2274 generation
= btrfs_super_chunk_root_generation(disk_super
);
2276 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
2277 chunk_root
, fs_info
, BTRFS_CHUNK_TREE_OBJECTID
);
2279 chunk_root
->node
= read_tree_block(chunk_root
,
2280 btrfs_super_chunk_root(disk_super
),
2281 blocksize
, generation
);
2282 BUG_ON(!chunk_root
->node
); /* -ENOMEM */
2283 if (!test_bit(EXTENT_BUFFER_UPTODATE
, &chunk_root
->node
->bflags
)) {
2284 printk(KERN_WARNING
"btrfs: failed to read chunk root on %s\n",
2286 goto fail_tree_roots
;
2288 btrfs_set_root_node(&chunk_root
->root_item
, chunk_root
->node
);
2289 chunk_root
->commit_root
= btrfs_root_node(chunk_root
);
2291 read_extent_buffer(chunk_root
->node
, fs_info
->chunk_tree_uuid
,
2292 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root
->node
),
2295 ret
= btrfs_read_chunk_tree(chunk_root
);
2297 printk(KERN_WARNING
"btrfs: failed to read chunk tree on %s\n",
2299 goto fail_tree_roots
;
2302 btrfs_close_extra_devices(fs_devices
);
2304 if (!fs_devices
->latest_bdev
) {
2305 printk(KERN_CRIT
"btrfs: failed to read devices on %s\n",
2307 goto fail_tree_roots
;
2311 blocksize
= btrfs_level_size(tree_root
,
2312 btrfs_super_root_level(disk_super
));
2313 generation
= btrfs_super_generation(disk_super
);
2315 tree_root
->node
= read_tree_block(tree_root
,
2316 btrfs_super_root(disk_super
),
2317 blocksize
, generation
);
2318 if (!tree_root
->node
||
2319 !test_bit(EXTENT_BUFFER_UPTODATE
, &tree_root
->node
->bflags
)) {
2320 printk(KERN_WARNING
"btrfs: failed to read tree root on %s\n",
2323 goto recovery_tree_root
;
2326 btrfs_set_root_node(&tree_root
->root_item
, tree_root
->node
);
2327 tree_root
->commit_root
= btrfs_root_node(tree_root
);
2329 ret
= find_and_setup_root(tree_root
, fs_info
,
2330 BTRFS_EXTENT_TREE_OBJECTID
, extent_root
);
2332 goto recovery_tree_root
;
2333 extent_root
->track_dirty
= 1;
2335 ret
= find_and_setup_root(tree_root
, fs_info
,
2336 BTRFS_DEV_TREE_OBJECTID
, dev_root
);
2338 goto recovery_tree_root
;
2339 dev_root
->track_dirty
= 1;
2341 ret
= find_and_setup_root(tree_root
, fs_info
,
2342 BTRFS_CSUM_TREE_OBJECTID
, csum_root
);
2344 goto recovery_tree_root
;
2346 csum_root
->track_dirty
= 1;
2348 fs_info
->generation
= generation
;
2349 fs_info
->last_trans_committed
= generation
;
2351 ret
= btrfs_init_space_info(fs_info
);
2353 printk(KERN_ERR
"Failed to initial space info: %d\n", ret
);
2354 goto fail_block_groups
;
2357 ret
= btrfs_read_block_groups(extent_root
);
2359 printk(KERN_ERR
"Failed to read block groups: %d\n", ret
);
2360 goto fail_block_groups
;
2363 fs_info
->cleaner_kthread
= kthread_run(cleaner_kthread
, tree_root
,
2365 if (IS_ERR(fs_info
->cleaner_kthread
))
2366 goto fail_block_groups
;
2368 fs_info
->transaction_kthread
= kthread_run(transaction_kthread
,
2370 "btrfs-transaction");
2371 if (IS_ERR(fs_info
->transaction_kthread
))
2374 if (!btrfs_test_opt(tree_root
, SSD
) &&
2375 !btrfs_test_opt(tree_root
, NOSSD
) &&
2376 !fs_info
->fs_devices
->rotating
) {
2377 printk(KERN_INFO
"Btrfs detected SSD devices, enabling SSD "
2379 btrfs_set_opt(fs_info
->mount_opt
, SSD
);
2382 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2383 if (btrfs_test_opt(tree_root
, CHECK_INTEGRITY
)) {
2384 ret
= btrfsic_mount(tree_root
, fs_devices
,
2385 btrfs_test_opt(tree_root
,
2386 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA
) ?
2388 fs_info
->check_integrity_print_mask
);
2390 printk(KERN_WARNING
"btrfs: failed to initialize"
2391 " integrity check module %s\n", sb
->s_id
);
2395 /* do not make disk changes in broken FS */
2396 if (btrfs_super_log_root(disk_super
) != 0 &&
2397 !(fs_info
->fs_state
& BTRFS_SUPER_FLAG_ERROR
)) {
2398 u64 bytenr
= btrfs_super_log_root(disk_super
);
2400 if (fs_devices
->rw_devices
== 0) {
2401 printk(KERN_WARNING
"Btrfs log replay required "
2404 goto fail_trans_kthread
;
2407 btrfs_level_size(tree_root
,
2408 btrfs_super_log_root_level(disk_super
));
2410 log_tree_root
= btrfs_alloc_root(fs_info
);
2411 if (!log_tree_root
) {
2413 goto fail_trans_kthread
;
2416 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
2417 log_tree_root
, fs_info
, BTRFS_TREE_LOG_OBJECTID
);
2419 log_tree_root
->node
= read_tree_block(tree_root
, bytenr
,
2422 /* returns with log_tree_root freed on success */
2423 ret
= btrfs_recover_log_trees(log_tree_root
);
2425 btrfs_error(tree_root
->fs_info
, ret
,
2426 "Failed to recover log tree");
2427 free_extent_buffer(log_tree_root
->node
);
2428 kfree(log_tree_root
);
2429 goto fail_trans_kthread
;
2432 if (sb
->s_flags
& MS_RDONLY
) {
2433 ret
= btrfs_commit_super(tree_root
);
2435 goto fail_trans_kthread
;
2439 ret
= btrfs_find_orphan_roots(tree_root
);
2441 goto fail_trans_kthread
;
2443 if (!(sb
->s_flags
& MS_RDONLY
)) {
2444 ret
= btrfs_cleanup_fs_roots(fs_info
);
2448 ret
= btrfs_recover_relocation(tree_root
);
2451 "btrfs: failed to recover relocation\n");
2453 goto fail_trans_kthread
;
2457 location
.objectid
= BTRFS_FS_TREE_OBJECTID
;
2458 location
.type
= BTRFS_ROOT_ITEM_KEY
;
2459 location
.offset
= (u64
)-1;
2461 fs_info
->fs_root
= btrfs_read_fs_root_no_name(fs_info
, &location
);
2462 if (!fs_info
->fs_root
)
2463 goto fail_trans_kthread
;
2464 if (IS_ERR(fs_info
->fs_root
)) {
2465 err
= PTR_ERR(fs_info
->fs_root
);
2466 goto fail_trans_kthread
;
2469 if (!(sb
->s_flags
& MS_RDONLY
)) {
2470 down_read(&fs_info
->cleanup_work_sem
);
2471 err
= btrfs_orphan_cleanup(fs_info
->fs_root
);
2473 err
= btrfs_orphan_cleanup(fs_info
->tree_root
);
2474 up_read(&fs_info
->cleanup_work_sem
);
2477 err
= btrfs_recover_balance(fs_info
->tree_root
);
2480 close_ctree(tree_root
);
2488 kthread_stop(fs_info
->transaction_kthread
);
2490 kthread_stop(fs_info
->cleaner_kthread
);
2493 * make sure we're done with the btree inode before we stop our
2496 filemap_write_and_wait(fs_info
->btree_inode
->i_mapping
);
2497 invalidate_inode_pages2(fs_info
->btree_inode
->i_mapping
);
2500 btrfs_free_block_groups(fs_info
);
2503 free_root_pointers(fs_info
, 1);
2506 btrfs_stop_workers(&fs_info
->generic_worker
);
2507 btrfs_stop_workers(&fs_info
->readahead_workers
);
2508 btrfs_stop_workers(&fs_info
->fixup_workers
);
2509 btrfs_stop_workers(&fs_info
->delalloc_workers
);
2510 btrfs_stop_workers(&fs_info
->workers
);
2511 btrfs_stop_workers(&fs_info
->endio_workers
);
2512 btrfs_stop_workers(&fs_info
->endio_meta_workers
);
2513 btrfs_stop_workers(&fs_info
->endio_meta_write_workers
);
2514 btrfs_stop_workers(&fs_info
->endio_write_workers
);
2515 btrfs_stop_workers(&fs_info
->endio_freespace_worker
);
2516 btrfs_stop_workers(&fs_info
->submit_workers
);
2517 btrfs_stop_workers(&fs_info
->delayed_workers
);
2518 btrfs_stop_workers(&fs_info
->caching_workers
);
2521 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
2523 invalidate_inode_pages2(fs_info
->btree_inode
->i_mapping
);
2524 iput(fs_info
->btree_inode
);
2526 bdi_destroy(&fs_info
->bdi
);
2528 cleanup_srcu_struct(&fs_info
->subvol_srcu
);
2530 btrfs_close_devices(fs_info
->fs_devices
);
2534 if (!btrfs_test_opt(tree_root
, RECOVERY
))
2535 goto fail_tree_roots
;
2537 free_root_pointers(fs_info
, 0);
2539 /* don't use the log in recovery mode, it won't be valid */
2540 btrfs_set_super_log_root(disk_super
, 0);
2542 /* we can't trust the free space cache either */
2543 btrfs_set_opt(fs_info
->mount_opt
, CLEAR_CACHE
);
2545 ret
= next_root_backup(fs_info
, fs_info
->super_copy
,
2546 &num_backups_tried
, &backup_index
);
2548 goto fail_block_groups
;
2549 goto retry_root_backup
;
2552 static void btrfs_end_buffer_write_sync(struct buffer_head
*bh
, int uptodate
)
2554 char b
[BDEVNAME_SIZE
];
2557 set_buffer_uptodate(bh
);
2559 printk_ratelimited(KERN_WARNING
"lost page write due to "
2560 "I/O error on %s\n",
2561 bdevname(bh
->b_bdev
, b
));
2562 /* note, we dont' set_buffer_write_io_error because we have
2563 * our own ways of dealing with the IO errors
2565 clear_buffer_uptodate(bh
);
2571 struct buffer_head
*btrfs_read_dev_super(struct block_device
*bdev
)
2573 struct buffer_head
*bh
;
2574 struct buffer_head
*latest
= NULL
;
2575 struct btrfs_super_block
*super
;
2580 /* we would like to check all the supers, but that would make
2581 * a btrfs mount succeed after a mkfs from a different FS.
2582 * So, we need to add a special mount option to scan for
2583 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
2585 for (i
= 0; i
< 1; i
++) {
2586 bytenr
= btrfs_sb_offset(i
);
2587 if (bytenr
+ 4096 >= i_size_read(bdev
->bd_inode
))
2589 bh
= __bread(bdev
, bytenr
/ 4096, 4096);
2593 super
= (struct btrfs_super_block
*)bh
->b_data
;
2594 if (btrfs_super_bytenr(super
) != bytenr
||
2595 strncmp((char *)(&super
->magic
), BTRFS_MAGIC
,
2596 sizeof(super
->magic
))) {
2601 if (!latest
|| btrfs_super_generation(super
) > transid
) {
2604 transid
= btrfs_super_generation(super
);
2613 * this should be called twice, once with wait == 0 and
2614 * once with wait == 1. When wait == 0 is done, all the buffer heads
2615 * we write are pinned.
2617 * They are released when wait == 1 is done.
2618 * max_mirrors must be the same for both runs, and it indicates how
2619 * many supers on this one device should be written.
2621 * max_mirrors == 0 means to write them all.
2623 static int write_dev_supers(struct btrfs_device
*device
,
2624 struct btrfs_super_block
*sb
,
2625 int do_barriers
, int wait
, int max_mirrors
)
2627 struct buffer_head
*bh
;
2634 if (max_mirrors
== 0)
2635 max_mirrors
= BTRFS_SUPER_MIRROR_MAX
;
2637 for (i
= 0; i
< max_mirrors
; i
++) {
2638 bytenr
= btrfs_sb_offset(i
);
2639 if (bytenr
+ BTRFS_SUPER_INFO_SIZE
>= device
->total_bytes
)
2643 bh
= __find_get_block(device
->bdev
, bytenr
/ 4096,
2644 BTRFS_SUPER_INFO_SIZE
);
2647 if (!buffer_uptodate(bh
))
2650 /* drop our reference */
2653 /* drop the reference from the wait == 0 run */
2657 btrfs_set_super_bytenr(sb
, bytenr
);
2660 crc
= btrfs_csum_data(NULL
, (char *)sb
+
2661 BTRFS_CSUM_SIZE
, crc
,
2662 BTRFS_SUPER_INFO_SIZE
-
2664 btrfs_csum_final(crc
, sb
->csum
);
2667 * one reference for us, and we leave it for the
2670 bh
= __getblk(device
->bdev
, bytenr
/ 4096,
2671 BTRFS_SUPER_INFO_SIZE
);
2672 memcpy(bh
->b_data
, sb
, BTRFS_SUPER_INFO_SIZE
);
2674 /* one reference for submit_bh */
2677 set_buffer_uptodate(bh
);
2679 bh
->b_end_io
= btrfs_end_buffer_write_sync
;
2683 * we fua the first super. The others we allow
2686 ret
= btrfsic_submit_bh(WRITE_FUA
, bh
);
2690 return errors
< i
? 0 : -1;
2694 * endio for the write_dev_flush, this will wake anyone waiting
2695 * for the barrier when it is done
2697 static void btrfs_end_empty_barrier(struct bio
*bio
, int err
)
2700 if (err
== -EOPNOTSUPP
)
2701 set_bit(BIO_EOPNOTSUPP
, &bio
->bi_flags
);
2702 clear_bit(BIO_UPTODATE
, &bio
->bi_flags
);
2704 if (bio
->bi_private
)
2705 complete(bio
->bi_private
);
2710 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
2711 * sent down. With wait == 1, it waits for the previous flush.
2713 * any device where the flush fails with eopnotsupp are flagged as not-barrier
2716 static int write_dev_flush(struct btrfs_device
*device
, int wait
)
2721 if (device
->nobarriers
)
2725 bio
= device
->flush_bio
;
2729 wait_for_completion(&device
->flush_wait
);
2731 if (bio_flagged(bio
, BIO_EOPNOTSUPP
)) {
2732 printk("btrfs: disabling barriers on dev %s\n",
2734 device
->nobarriers
= 1;
2736 if (!bio_flagged(bio
, BIO_UPTODATE
)) {
2740 /* drop the reference from the wait == 0 run */
2742 device
->flush_bio
= NULL
;
2748 * one reference for us, and we leave it for the
2751 device
->flush_bio
= NULL
;;
2752 bio
= bio_alloc(GFP_NOFS
, 0);
2756 bio
->bi_end_io
= btrfs_end_empty_barrier
;
2757 bio
->bi_bdev
= device
->bdev
;
2758 init_completion(&device
->flush_wait
);
2759 bio
->bi_private
= &device
->flush_wait
;
2760 device
->flush_bio
= bio
;
2763 btrfsic_submit_bio(WRITE_FLUSH
, bio
);
2769 * send an empty flush down to each device in parallel,
2770 * then wait for them
2772 static int barrier_all_devices(struct btrfs_fs_info
*info
)
2774 struct list_head
*head
;
2775 struct btrfs_device
*dev
;
2779 /* send down all the barriers */
2780 head
= &info
->fs_devices
->devices
;
2781 list_for_each_entry_rcu(dev
, head
, dev_list
) {
2786 if (!dev
->in_fs_metadata
|| !dev
->writeable
)
2789 ret
= write_dev_flush(dev
, 0);
2794 /* wait for all the barriers */
2795 list_for_each_entry_rcu(dev
, head
, dev_list
) {
2800 if (!dev
->in_fs_metadata
|| !dev
->writeable
)
2803 ret
= write_dev_flush(dev
, 1);
2812 int write_all_supers(struct btrfs_root
*root
, int max_mirrors
)
2814 struct list_head
*head
;
2815 struct btrfs_device
*dev
;
2816 struct btrfs_super_block
*sb
;
2817 struct btrfs_dev_item
*dev_item
;
2821 int total_errors
= 0;
2824 max_errors
= btrfs_super_num_devices(root
->fs_info
->super_copy
) - 1;
2825 do_barriers
= !btrfs_test_opt(root
, NOBARRIER
);
2826 backup_super_roots(root
->fs_info
);
2828 sb
= root
->fs_info
->super_for_commit
;
2829 dev_item
= &sb
->dev_item
;
2831 mutex_lock(&root
->fs_info
->fs_devices
->device_list_mutex
);
2832 head
= &root
->fs_info
->fs_devices
->devices
;
2835 barrier_all_devices(root
->fs_info
);
2837 list_for_each_entry_rcu(dev
, head
, dev_list
) {
2842 if (!dev
->in_fs_metadata
|| !dev
->writeable
)
2845 btrfs_set_stack_device_generation(dev_item
, 0);
2846 btrfs_set_stack_device_type(dev_item
, dev
->type
);
2847 btrfs_set_stack_device_id(dev_item
, dev
->devid
);
2848 btrfs_set_stack_device_total_bytes(dev_item
, dev
->total_bytes
);
2849 btrfs_set_stack_device_bytes_used(dev_item
, dev
->bytes_used
);
2850 btrfs_set_stack_device_io_align(dev_item
, dev
->io_align
);
2851 btrfs_set_stack_device_io_width(dev_item
, dev
->io_width
);
2852 btrfs_set_stack_device_sector_size(dev_item
, dev
->sector_size
);
2853 memcpy(dev_item
->uuid
, dev
->uuid
, BTRFS_UUID_SIZE
);
2854 memcpy(dev_item
->fsid
, dev
->fs_devices
->fsid
, BTRFS_UUID_SIZE
);
2856 flags
= btrfs_super_flags(sb
);
2857 btrfs_set_super_flags(sb
, flags
| BTRFS_HEADER_FLAG_WRITTEN
);
2859 ret
= write_dev_supers(dev
, sb
, do_barriers
, 0, max_mirrors
);
2863 if (total_errors
> max_errors
) {
2864 printk(KERN_ERR
"btrfs: %d errors while writing supers\n",
2867 /* This shouldn't happen. FUA is masked off if unsupported */
2872 list_for_each_entry_rcu(dev
, head
, dev_list
) {
2875 if (!dev
->in_fs_metadata
|| !dev
->writeable
)
2878 ret
= write_dev_supers(dev
, sb
, do_barriers
, 1, max_mirrors
);
2882 mutex_unlock(&root
->fs_info
->fs_devices
->device_list_mutex
);
2883 if (total_errors
> max_errors
) {
2884 btrfs_error(root
->fs_info
, -EIO
,
2885 "%d errors while writing supers", total_errors
);
2891 int write_ctree_super(struct btrfs_trans_handle
*trans
,
2892 struct btrfs_root
*root
, int max_mirrors
)
2896 ret
= write_all_supers(root
, max_mirrors
);
2900 /* Kill all outstanding I/O */
2901 void btrfs_abort_devices(struct btrfs_root
*root
)
2903 struct list_head
*head
;
2904 struct btrfs_device
*dev
;
2905 mutex_lock(&root
->fs_info
->fs_devices
->device_list_mutex
);
2906 head
= &root
->fs_info
->fs_devices
->devices
;
2907 list_for_each_entry_rcu(dev
, head
, dev_list
) {
2908 blk_abort_queue(dev
->bdev
->bd_disk
->queue
);
2910 mutex_unlock(&root
->fs_info
->fs_devices
->device_list_mutex
);
2913 void btrfs_free_fs_root(struct btrfs_fs_info
*fs_info
, struct btrfs_root
*root
)
2915 spin_lock(&fs_info
->fs_roots_radix_lock
);
2916 radix_tree_delete(&fs_info
->fs_roots_radix
,
2917 (unsigned long)root
->root_key
.objectid
);
2918 spin_unlock(&fs_info
->fs_roots_radix_lock
);
2920 if (btrfs_root_refs(&root
->root_item
) == 0)
2921 synchronize_srcu(&fs_info
->subvol_srcu
);
2923 __btrfs_remove_free_space_cache(root
->free_ino_pinned
);
2924 __btrfs_remove_free_space_cache(root
->free_ino_ctl
);
2928 static void free_fs_root(struct btrfs_root
*root
)
2930 iput(root
->cache_inode
);
2931 WARN_ON(!RB_EMPTY_ROOT(&root
->inode_tree
));
2933 free_anon_bdev(root
->anon_dev
);
2934 free_extent_buffer(root
->node
);
2935 free_extent_buffer(root
->commit_root
);
2936 kfree(root
->free_ino_ctl
);
2937 kfree(root
->free_ino_pinned
);
2942 static void del_fs_roots(struct btrfs_fs_info
*fs_info
)
2945 struct btrfs_root
*gang
[8];
2948 while (!list_empty(&fs_info
->dead_roots
)) {
2949 gang
[0] = list_entry(fs_info
->dead_roots
.next
,
2950 struct btrfs_root
, root_list
);
2951 list_del(&gang
[0]->root_list
);
2953 if (gang
[0]->in_radix
) {
2954 btrfs_free_fs_root(fs_info
, gang
[0]);
2956 free_extent_buffer(gang
[0]->node
);
2957 free_extent_buffer(gang
[0]->commit_root
);
2963 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
2968 for (i
= 0; i
< ret
; i
++)
2969 btrfs_free_fs_root(fs_info
, gang
[i
]);
2973 int btrfs_cleanup_fs_roots(struct btrfs_fs_info
*fs_info
)
2975 u64 root_objectid
= 0;
2976 struct btrfs_root
*gang
[8];
2981 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
2982 (void **)gang
, root_objectid
,
2987 root_objectid
= gang
[ret
- 1]->root_key
.objectid
+ 1;
2988 for (i
= 0; i
< ret
; i
++) {
2991 root_objectid
= gang
[i
]->root_key
.objectid
;
2992 err
= btrfs_orphan_cleanup(gang
[i
]);
3001 int btrfs_commit_super(struct btrfs_root
*root
)
3003 struct btrfs_trans_handle
*trans
;
3006 mutex_lock(&root
->fs_info
->cleaner_mutex
);
3007 btrfs_run_delayed_iputs(root
);
3008 btrfs_clean_old_snapshots(root
);
3009 mutex_unlock(&root
->fs_info
->cleaner_mutex
);
3011 /* wait until ongoing cleanup work done */
3012 down_write(&root
->fs_info
->cleanup_work_sem
);
3013 up_write(&root
->fs_info
->cleanup_work_sem
);
3015 trans
= btrfs_join_transaction(root
);
3017 return PTR_ERR(trans
);
3018 ret
= btrfs_commit_transaction(trans
, root
);
3021 /* run commit again to drop the original snapshot */
3022 trans
= btrfs_join_transaction(root
);
3024 return PTR_ERR(trans
);
3025 ret
= btrfs_commit_transaction(trans
, root
);
3028 ret
= btrfs_write_and_wait_transaction(NULL
, root
);
3030 btrfs_error(root
->fs_info
, ret
,
3031 "Failed to sync btree inode to disk.");
3035 ret
= write_ctree_super(NULL
, root
, 0);
3039 int close_ctree(struct btrfs_root
*root
)
3041 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
3044 fs_info
->closing
= 1;
3047 /* pause restriper - we want to resume on mount */
3048 btrfs_pause_balance(root
->fs_info
);
3050 btrfs_scrub_cancel(root
);
3052 /* wait for any defraggers to finish */
3053 wait_event(fs_info
->transaction_wait
,
3054 (atomic_read(&fs_info
->defrag_running
) == 0));
3056 /* clear out the rbtree of defraggable inodes */
3057 btrfs_run_defrag_inodes(fs_info
);
3060 * Here come 2 situations when btrfs is broken to flip readonly:
3062 * 1. when btrfs flips readonly somewhere else before
3063 * btrfs_commit_super, sb->s_flags has MS_RDONLY flag,
3064 * and btrfs will skip to write sb directly to keep
3065 * ERROR state on disk.
3067 * 2. when btrfs flips readonly just in btrfs_commit_super,
3068 * and in such case, btrfs cannot write sb via btrfs_commit_super,
3069 * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag,
3070 * btrfs will cleanup all FS resources first and write sb then.
3072 if (!(fs_info
->sb
->s_flags
& MS_RDONLY
)) {
3073 ret
= btrfs_commit_super(root
);
3075 printk(KERN_ERR
"btrfs: commit super ret %d\n", ret
);
3078 if (fs_info
->fs_state
& BTRFS_SUPER_FLAG_ERROR
) {
3079 ret
= btrfs_error_commit_super(root
);
3081 printk(KERN_ERR
"btrfs: commit super ret %d\n", ret
);
3084 btrfs_put_block_group_cache(fs_info
);
3086 kthread_stop(fs_info
->transaction_kthread
);
3087 kthread_stop(fs_info
->cleaner_kthread
);
3089 fs_info
->closing
= 2;
3092 if (fs_info
->delalloc_bytes
) {
3093 printk(KERN_INFO
"btrfs: at unmount delalloc count %llu\n",
3094 (unsigned long long)fs_info
->delalloc_bytes
);
3096 if (fs_info
->total_ref_cache_size
) {
3097 printk(KERN_INFO
"btrfs: at umount reference cache size %llu\n",
3098 (unsigned long long)fs_info
->total_ref_cache_size
);
3101 free_extent_buffer(fs_info
->extent_root
->node
);
3102 free_extent_buffer(fs_info
->extent_root
->commit_root
);
3103 free_extent_buffer(fs_info
->tree_root
->node
);
3104 free_extent_buffer(fs_info
->tree_root
->commit_root
);
3105 free_extent_buffer(fs_info
->chunk_root
->node
);
3106 free_extent_buffer(fs_info
->chunk_root
->commit_root
);
3107 free_extent_buffer(fs_info
->dev_root
->node
);
3108 free_extent_buffer(fs_info
->dev_root
->commit_root
);
3109 free_extent_buffer(fs_info
->csum_root
->node
);
3110 free_extent_buffer(fs_info
->csum_root
->commit_root
);
3112 btrfs_free_block_groups(fs_info
);
3114 del_fs_roots(fs_info
);
3116 iput(fs_info
->btree_inode
);
3118 btrfs_stop_workers(&fs_info
->generic_worker
);
3119 btrfs_stop_workers(&fs_info
->fixup_workers
);
3120 btrfs_stop_workers(&fs_info
->delalloc_workers
);
3121 btrfs_stop_workers(&fs_info
->workers
);
3122 btrfs_stop_workers(&fs_info
->endio_workers
);
3123 btrfs_stop_workers(&fs_info
->endio_meta_workers
);
3124 btrfs_stop_workers(&fs_info
->endio_meta_write_workers
);
3125 btrfs_stop_workers(&fs_info
->endio_write_workers
);
3126 btrfs_stop_workers(&fs_info
->endio_freespace_worker
);
3127 btrfs_stop_workers(&fs_info
->submit_workers
);
3128 btrfs_stop_workers(&fs_info
->delayed_workers
);
3129 btrfs_stop_workers(&fs_info
->caching_workers
);
3130 btrfs_stop_workers(&fs_info
->readahead_workers
);
3132 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3133 if (btrfs_test_opt(root
, CHECK_INTEGRITY
))
3134 btrfsic_unmount(root
, fs_info
->fs_devices
);
3137 btrfs_close_devices(fs_info
->fs_devices
);
3138 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
3140 bdi_destroy(&fs_info
->bdi
);
3141 cleanup_srcu_struct(&fs_info
->subvol_srcu
);
3146 int btrfs_buffer_uptodate(struct extent_buffer
*buf
, u64 parent_transid
)
3149 struct inode
*btree_inode
= buf
->pages
[0]->mapping
->host
;
3151 ret
= extent_buffer_uptodate(buf
);
3155 ret
= verify_parent_transid(&BTRFS_I(btree_inode
)->io_tree
, buf
,
3160 int btrfs_set_buffer_uptodate(struct extent_buffer
*buf
)
3162 return set_extent_buffer_uptodate(buf
);
3165 void btrfs_mark_buffer_dirty(struct extent_buffer
*buf
)
3167 struct btrfs_root
*root
= BTRFS_I(buf
->pages
[0]->mapping
->host
)->root
;
3168 u64 transid
= btrfs_header_generation(buf
);
3171 btrfs_assert_tree_locked(buf
);
3172 if (transid
!= root
->fs_info
->generation
) {
3173 printk(KERN_CRIT
"btrfs transid mismatch buffer %llu, "
3174 "found %llu running %llu\n",
3175 (unsigned long long)buf
->start
,
3176 (unsigned long long)transid
,
3177 (unsigned long long)root
->fs_info
->generation
);
3180 was_dirty
= set_extent_buffer_dirty(buf
);
3182 spin_lock(&root
->fs_info
->delalloc_lock
);
3183 root
->fs_info
->dirty_metadata_bytes
+= buf
->len
;
3184 spin_unlock(&root
->fs_info
->delalloc_lock
);
3188 void btrfs_btree_balance_dirty(struct btrfs_root
*root
, unsigned long nr
)
3191 * looks as though older kernels can get into trouble with
3192 * this code, they end up stuck in balance_dirty_pages forever
3195 unsigned long thresh
= 32 * 1024 * 1024;
3197 if (current
->flags
& PF_MEMALLOC
)
3200 btrfs_balance_delayed_items(root
);
3202 num_dirty
= root
->fs_info
->dirty_metadata_bytes
;
3204 if (num_dirty
> thresh
) {
3205 balance_dirty_pages_ratelimited_nr(
3206 root
->fs_info
->btree_inode
->i_mapping
, 1);
3211 void __btrfs_btree_balance_dirty(struct btrfs_root
*root
, unsigned long nr
)
3214 * looks as though older kernels can get into trouble with
3215 * this code, they end up stuck in balance_dirty_pages forever
3218 unsigned long thresh
= 32 * 1024 * 1024;
3220 if (current
->flags
& PF_MEMALLOC
)
3223 num_dirty
= root
->fs_info
->dirty_metadata_bytes
;
3225 if (num_dirty
> thresh
) {
3226 balance_dirty_pages_ratelimited_nr(
3227 root
->fs_info
->btree_inode
->i_mapping
, 1);
3232 int btrfs_read_buffer(struct extent_buffer
*buf
, u64 parent_transid
)
3234 struct btrfs_root
*root
= BTRFS_I(buf
->pages
[0]->mapping
->host
)->root
;
3235 return btree_read_extent_buffer_pages(root
, buf
, 0, parent_transid
);
3238 static int btree_lock_page_hook(struct page
*page
, void *data
,
3239 void (*flush_fn
)(void *))
3241 struct inode
*inode
= page
->mapping
->host
;
3242 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3243 struct extent_buffer
*eb
;
3246 * We culled this eb but the page is still hanging out on the mapping,
3249 if (!PagePrivate(page
))
3252 eb
= (struct extent_buffer
*)page
->private;
3257 if (page
!= eb
->pages
[0])
3260 if (!btrfs_try_tree_write_lock(eb
)) {
3262 btrfs_tree_lock(eb
);
3264 btrfs_set_header_flag(eb
, BTRFS_HEADER_FLAG_WRITTEN
);
3266 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY
, &eb
->bflags
)) {
3267 spin_lock(&root
->fs_info
->delalloc_lock
);
3268 if (root
->fs_info
->dirty_metadata_bytes
>= eb
->len
)
3269 root
->fs_info
->dirty_metadata_bytes
-= eb
->len
;
3272 spin_unlock(&root
->fs_info
->delalloc_lock
);
3275 btrfs_tree_unlock(eb
);
3277 if (!trylock_page(page
)) {
3284 static int btrfs_check_super_valid(struct btrfs_fs_info
*fs_info
,
3287 if (btrfs_super_csum_type(fs_info
->super_copy
) >= ARRAY_SIZE(btrfs_csum_sizes
)) {
3288 printk(KERN_ERR
"btrfs: unsupported checksum algorithm\n");
3295 if (fs_info
->fs_state
& BTRFS_SUPER_FLAG_ERROR
) {
3296 printk(KERN_WARNING
"warning: mount fs with errors, "
3297 "running btrfsck is recommended\n");
3303 int btrfs_error_commit_super(struct btrfs_root
*root
)
3307 mutex_lock(&root
->fs_info
->cleaner_mutex
);
3308 btrfs_run_delayed_iputs(root
);
3309 mutex_unlock(&root
->fs_info
->cleaner_mutex
);
3311 down_write(&root
->fs_info
->cleanup_work_sem
);
3312 up_write(&root
->fs_info
->cleanup_work_sem
);
3314 /* cleanup FS via transaction */
3315 btrfs_cleanup_transaction(root
);
3317 ret
= write_ctree_super(NULL
, root
, 0);
3322 static void btrfs_destroy_ordered_operations(struct btrfs_root
*root
)
3324 struct btrfs_inode
*btrfs_inode
;
3325 struct list_head splice
;
3327 INIT_LIST_HEAD(&splice
);
3329 mutex_lock(&root
->fs_info
->ordered_operations_mutex
);
3330 spin_lock(&root
->fs_info
->ordered_extent_lock
);
3332 list_splice_init(&root
->fs_info
->ordered_operations
, &splice
);
3333 while (!list_empty(&splice
)) {
3334 btrfs_inode
= list_entry(splice
.next
, struct btrfs_inode
,
3335 ordered_operations
);
3337 list_del_init(&btrfs_inode
->ordered_operations
);
3339 btrfs_invalidate_inodes(btrfs_inode
->root
);
3342 spin_unlock(&root
->fs_info
->ordered_extent_lock
);
3343 mutex_unlock(&root
->fs_info
->ordered_operations_mutex
);
3346 static void btrfs_destroy_ordered_extents(struct btrfs_root
*root
)
3348 struct list_head splice
;
3349 struct btrfs_ordered_extent
*ordered
;
3350 struct inode
*inode
;
3352 INIT_LIST_HEAD(&splice
);
3354 spin_lock(&root
->fs_info
->ordered_extent_lock
);
3356 list_splice_init(&root
->fs_info
->ordered_extents
, &splice
);
3357 while (!list_empty(&splice
)) {
3358 ordered
= list_entry(splice
.next
, struct btrfs_ordered_extent
,
3361 list_del_init(&ordered
->root_extent_list
);
3362 atomic_inc(&ordered
->refs
);
3364 /* the inode may be getting freed (in sys_unlink path). */
3365 inode
= igrab(ordered
->inode
);
3367 spin_unlock(&root
->fs_info
->ordered_extent_lock
);
3371 atomic_set(&ordered
->refs
, 1);
3372 btrfs_put_ordered_extent(ordered
);
3374 spin_lock(&root
->fs_info
->ordered_extent_lock
);
3377 spin_unlock(&root
->fs_info
->ordered_extent_lock
);
3380 int btrfs_destroy_delayed_refs(struct btrfs_transaction
*trans
,
3381 struct btrfs_root
*root
)
3383 struct rb_node
*node
;
3384 struct btrfs_delayed_ref_root
*delayed_refs
;
3385 struct btrfs_delayed_ref_node
*ref
;
3388 delayed_refs
= &trans
->delayed_refs
;
3391 spin_lock(&delayed_refs
->lock
);
3392 if (delayed_refs
->num_entries
== 0) {
3393 spin_unlock(&delayed_refs
->lock
);
3394 printk(KERN_INFO
"delayed_refs has NO entry\n");
3398 node
= rb_first(&delayed_refs
->root
);
3400 ref
= rb_entry(node
, struct btrfs_delayed_ref_node
, rb_node
);
3401 node
= rb_next(node
);
3404 rb_erase(&ref
->rb_node
, &delayed_refs
->root
);
3405 delayed_refs
->num_entries
--;
3407 atomic_set(&ref
->refs
, 1);
3408 if (btrfs_delayed_ref_is_head(ref
)) {
3409 struct btrfs_delayed_ref_head
*head
;
3411 head
= btrfs_delayed_node_to_head(ref
);
3412 spin_unlock(&delayed_refs
->lock
);
3413 mutex_lock(&head
->mutex
);
3414 kfree(head
->extent_op
);
3415 delayed_refs
->num_heads
--;
3416 if (list_empty(&head
->cluster
))
3417 delayed_refs
->num_heads_ready
--;
3418 list_del_init(&head
->cluster
);
3419 mutex_unlock(&head
->mutex
);
3420 btrfs_put_delayed_ref(ref
);
3423 spin_unlock(&delayed_refs
->lock
);
3424 btrfs_put_delayed_ref(ref
);
3427 spin_lock(&delayed_refs
->lock
);
3430 spin_unlock(&delayed_refs
->lock
);
3435 static void btrfs_destroy_pending_snapshots(struct btrfs_transaction
*t
)
3437 struct btrfs_pending_snapshot
*snapshot
;
3438 struct list_head splice
;
3440 INIT_LIST_HEAD(&splice
);
3442 list_splice_init(&t
->pending_snapshots
, &splice
);
3444 while (!list_empty(&splice
)) {
3445 snapshot
= list_entry(splice
.next
,
3446 struct btrfs_pending_snapshot
,
3449 list_del_init(&snapshot
->list
);
3455 static void btrfs_destroy_delalloc_inodes(struct btrfs_root
*root
)
3457 struct btrfs_inode
*btrfs_inode
;
3458 struct list_head splice
;
3460 INIT_LIST_HEAD(&splice
);
3462 spin_lock(&root
->fs_info
->delalloc_lock
);
3463 list_splice_init(&root
->fs_info
->delalloc_inodes
, &splice
);
3465 while (!list_empty(&splice
)) {
3466 btrfs_inode
= list_entry(splice
.next
, struct btrfs_inode
,
3469 list_del_init(&btrfs_inode
->delalloc_inodes
);
3471 btrfs_invalidate_inodes(btrfs_inode
->root
);
3474 spin_unlock(&root
->fs_info
->delalloc_lock
);
3477 static int btrfs_destroy_marked_extents(struct btrfs_root
*root
,
3478 struct extent_io_tree
*dirty_pages
,
3483 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
3484 struct extent_buffer
*eb
;
3488 unsigned long index
;
3491 ret
= find_first_extent_bit(dirty_pages
, start
, &start
, &end
,
3496 clear_extent_bits(dirty_pages
, start
, end
, mark
, GFP_NOFS
);
3497 while (start
<= end
) {
3498 index
= start
>> PAGE_CACHE_SHIFT
;
3499 start
= (u64
)(index
+ 1) << PAGE_CACHE_SHIFT
;
3500 page
= find_get_page(btree_inode
->i_mapping
, index
);
3503 offset
= page_offset(page
);
3505 spin_lock(&dirty_pages
->buffer_lock
);
3506 eb
= radix_tree_lookup(
3507 &(&BTRFS_I(page
->mapping
->host
)->io_tree
)->buffer
,
3508 offset
>> PAGE_CACHE_SHIFT
);
3509 spin_unlock(&dirty_pages
->buffer_lock
);
3511 ret
= test_and_clear_bit(EXTENT_BUFFER_DIRTY
,
3513 atomic_set(&eb
->refs
, 1);
3515 if (PageWriteback(page
))
3516 end_page_writeback(page
);
3519 if (PageDirty(page
)) {
3520 clear_page_dirty_for_io(page
);
3521 spin_lock_irq(&page
->mapping
->tree_lock
);
3522 radix_tree_tag_clear(&page
->mapping
->page_tree
,
3524 PAGECACHE_TAG_DIRTY
);
3525 spin_unlock_irq(&page
->mapping
->tree_lock
);
3528 page
->mapping
->a_ops
->invalidatepage(page
, 0);
3536 static int btrfs_destroy_pinned_extent(struct btrfs_root
*root
,
3537 struct extent_io_tree
*pinned_extents
)
3539 struct extent_io_tree
*unpin
;
3544 unpin
= pinned_extents
;
3546 ret
= find_first_extent_bit(unpin
, 0, &start
, &end
,
3552 if (btrfs_test_opt(root
, DISCARD
))
3553 ret
= btrfs_error_discard_extent(root
, start
,
3557 clear_extent_dirty(unpin
, start
, end
, GFP_NOFS
);
3558 btrfs_error_unpin_extent_range(root
, start
, end
);
3565 void btrfs_cleanup_one_transaction(struct btrfs_transaction
*cur_trans
,
3566 struct btrfs_root
*root
)
3568 btrfs_destroy_delayed_refs(cur_trans
, root
);
3569 btrfs_block_rsv_release(root
, &root
->fs_info
->trans_block_rsv
,
3570 cur_trans
->dirty_pages
.dirty_bytes
);
3572 /* FIXME: cleanup wait for commit */
3573 cur_trans
->in_commit
= 1;
3574 cur_trans
->blocked
= 1;
3575 if (waitqueue_active(&root
->fs_info
->transaction_blocked_wait
))
3576 wake_up(&root
->fs_info
->transaction_blocked_wait
);
3578 cur_trans
->blocked
= 0;
3579 if (waitqueue_active(&root
->fs_info
->transaction_wait
))
3580 wake_up(&root
->fs_info
->transaction_wait
);
3582 cur_trans
->commit_done
= 1;
3583 if (waitqueue_active(&cur_trans
->commit_wait
))
3584 wake_up(&cur_trans
->commit_wait
);
3586 btrfs_destroy_pending_snapshots(cur_trans
);
3588 btrfs_destroy_marked_extents(root
, &cur_trans
->dirty_pages
,
3592 memset(cur_trans, 0, sizeof(*cur_trans));
3593 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
3597 int btrfs_cleanup_transaction(struct btrfs_root
*root
)
3599 struct btrfs_transaction
*t
;
3602 mutex_lock(&root
->fs_info
->transaction_kthread_mutex
);
3604 spin_lock(&root
->fs_info
->trans_lock
);
3605 list_splice_init(&root
->fs_info
->trans_list
, &list
);
3606 root
->fs_info
->trans_no_join
= 1;
3607 spin_unlock(&root
->fs_info
->trans_lock
);
3609 while (!list_empty(&list
)) {
3610 t
= list_entry(list
.next
, struct btrfs_transaction
, list
);
3614 btrfs_destroy_ordered_operations(root
);
3616 btrfs_destroy_ordered_extents(root
);
3618 btrfs_destroy_delayed_refs(t
, root
);
3620 btrfs_block_rsv_release(root
,
3621 &root
->fs_info
->trans_block_rsv
,
3622 t
->dirty_pages
.dirty_bytes
);
3624 /* FIXME: cleanup wait for commit */
3627 if (waitqueue_active(&root
->fs_info
->transaction_blocked_wait
))
3628 wake_up(&root
->fs_info
->transaction_blocked_wait
);
3631 if (waitqueue_active(&root
->fs_info
->transaction_wait
))
3632 wake_up(&root
->fs_info
->transaction_wait
);
3635 if (waitqueue_active(&t
->commit_wait
))
3636 wake_up(&t
->commit_wait
);
3638 btrfs_destroy_pending_snapshots(t
);
3640 btrfs_destroy_delalloc_inodes(root
);
3642 spin_lock(&root
->fs_info
->trans_lock
);
3643 root
->fs_info
->running_transaction
= NULL
;
3644 spin_unlock(&root
->fs_info
->trans_lock
);
3646 btrfs_destroy_marked_extents(root
, &t
->dirty_pages
,
3649 btrfs_destroy_pinned_extent(root
,
3650 root
->fs_info
->pinned_extents
);
3652 atomic_set(&t
->use_count
, 0);
3653 list_del_init(&t
->list
);
3654 memset(t
, 0, sizeof(*t
));
3655 kmem_cache_free(btrfs_transaction_cachep
, t
);
3658 spin_lock(&root
->fs_info
->trans_lock
);
3659 root
->fs_info
->trans_no_join
= 0;
3660 spin_unlock(&root
->fs_info
->trans_lock
);
3661 mutex_unlock(&root
->fs_info
->transaction_kthread_mutex
);
3666 static int btree_writepage_io_failed_hook(struct bio
*bio
, struct page
*page
,
3668 struct extent_state
*state
)
3670 struct super_block
*sb
= page
->mapping
->host
->i_sb
;
3671 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
3672 btrfs_error(fs_info
, -EIO
,
3673 "Error occured while writing out btree at %llu", start
);
3677 static struct extent_io_ops btree_extent_io_ops
= {
3678 .write_cache_pages_lock_hook
= btree_lock_page_hook
,
3679 .readpage_end_io_hook
= btree_readpage_end_io_hook
,
3680 .readpage_io_failed_hook
= btree_io_failed_hook
,
3681 .submit_bio_hook
= btree_submit_bio_hook
,
3682 /* note we're sharing with inode.c for the merge bio hook */
3683 .merge_bio_hook
= btrfs_merge_bio_hook
,
3684 .writepage_io_failed_hook
= btree_writepage_io_failed_hook
,