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 void 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
;
104 * Lockdep class keys for extent_buffer->lock's in this root. For a given
105 * eb, the lockdep key is determined by the btrfs_root it belongs to and
106 * the level the eb occupies in the tree.
108 * Different roots are used for different purposes and may nest inside each
109 * other and they require separate keysets. As lockdep keys should be
110 * static, assign keysets according to the purpose of the root as indicated
111 * by btrfs_root->objectid. This ensures that all special purpose roots
112 * have separate keysets.
114 * Lock-nesting across peer nodes is always done with the immediate parent
115 * node locked thus preventing deadlock. As lockdep doesn't know this, use
116 * subclass to avoid triggering lockdep warning in such cases.
118 * The key is set by the readpage_end_io_hook after the buffer has passed
119 * csum validation but before the pages are unlocked. It is also set by
120 * btrfs_init_new_buffer on freshly allocated blocks.
122 * We also add a check to make sure the highest level of the tree is the
123 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
124 * needs update as well.
126 #ifdef CONFIG_DEBUG_LOCK_ALLOC
127 # if BTRFS_MAX_LEVEL != 8
131 static struct btrfs_lockdep_keyset
{
132 u64 id
; /* root objectid */
133 const char *name_stem
; /* lock name stem */
134 char names
[BTRFS_MAX_LEVEL
+ 1][20];
135 struct lock_class_key keys
[BTRFS_MAX_LEVEL
+ 1];
136 } btrfs_lockdep_keysets
[] = {
137 { .id
= BTRFS_ROOT_TREE_OBJECTID
, .name_stem
= "root" },
138 { .id
= BTRFS_EXTENT_TREE_OBJECTID
, .name_stem
= "extent" },
139 { .id
= BTRFS_CHUNK_TREE_OBJECTID
, .name_stem
= "chunk" },
140 { .id
= BTRFS_DEV_TREE_OBJECTID
, .name_stem
= "dev" },
141 { .id
= BTRFS_FS_TREE_OBJECTID
, .name_stem
= "fs" },
142 { .id
= BTRFS_CSUM_TREE_OBJECTID
, .name_stem
= "csum" },
143 { .id
= BTRFS_ORPHAN_OBJECTID
, .name_stem
= "orphan" },
144 { .id
= BTRFS_TREE_LOG_OBJECTID
, .name_stem
= "log" },
145 { .id
= BTRFS_TREE_RELOC_OBJECTID
, .name_stem
= "treloc" },
146 { .id
= BTRFS_DATA_RELOC_TREE_OBJECTID
, .name_stem
= "dreloc" },
147 { .id
= 0, .name_stem
= "tree" },
150 void __init
btrfs_init_lockdep(void)
154 /* initialize lockdep class names */
155 for (i
= 0; i
< ARRAY_SIZE(btrfs_lockdep_keysets
); i
++) {
156 struct btrfs_lockdep_keyset
*ks
= &btrfs_lockdep_keysets
[i
];
158 for (j
= 0; j
< ARRAY_SIZE(ks
->names
); j
++)
159 snprintf(ks
->names
[j
], sizeof(ks
->names
[j
]),
160 "btrfs-%s-%02d", ks
->name_stem
, j
);
164 void btrfs_set_buffer_lockdep_class(u64 objectid
, struct extent_buffer
*eb
,
167 struct btrfs_lockdep_keyset
*ks
;
169 BUG_ON(level
>= ARRAY_SIZE(ks
->keys
));
171 /* find the matching keyset, id 0 is the default entry */
172 for (ks
= btrfs_lockdep_keysets
; ks
->id
; ks
++)
173 if (ks
->id
== objectid
)
176 lockdep_set_class_and_name(&eb
->lock
,
177 &ks
->keys
[level
], ks
->names
[level
]);
183 * extents on the btree inode are pretty simple, there's one extent
184 * that covers the entire device
186 static struct extent_map
*btree_get_extent(struct inode
*inode
,
187 struct page
*page
, size_t pg_offset
, u64 start
, u64 len
,
190 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
191 struct extent_map
*em
;
194 read_lock(&em_tree
->lock
);
195 em
= lookup_extent_mapping(em_tree
, start
, len
);
198 BTRFS_I(inode
)->root
->fs_info
->fs_devices
->latest_bdev
;
199 read_unlock(&em_tree
->lock
);
202 read_unlock(&em_tree
->lock
);
204 em
= alloc_extent_map();
206 em
= ERR_PTR(-ENOMEM
);
211 em
->block_len
= (u64
)-1;
213 em
->bdev
= BTRFS_I(inode
)->root
->fs_info
->fs_devices
->latest_bdev
;
215 write_lock(&em_tree
->lock
);
216 ret
= add_extent_mapping(em_tree
, em
);
217 if (ret
== -EEXIST
) {
218 u64 failed_start
= em
->start
;
219 u64 failed_len
= em
->len
;
222 em
= lookup_extent_mapping(em_tree
, start
, len
);
226 em
= lookup_extent_mapping(em_tree
, failed_start
,
234 write_unlock(&em_tree
->lock
);
242 u32
btrfs_csum_data(struct btrfs_root
*root
, char *data
, u32 seed
, size_t len
)
244 return crc32c(seed
, data
, len
);
247 void btrfs_csum_final(u32 crc
, char *result
)
249 put_unaligned_le32(~crc
, result
);
253 * compute the csum for a btree block, and either verify it or write it
254 * into the csum field of the block.
256 static int csum_tree_block(struct btrfs_root
*root
, struct extent_buffer
*buf
,
259 u16 csum_size
= btrfs_super_csum_size(root
->fs_info
->super_copy
);
262 unsigned long cur_len
;
263 unsigned long offset
= BTRFS_CSUM_SIZE
;
265 unsigned long map_start
;
266 unsigned long map_len
;
269 unsigned long inline_result
;
271 len
= buf
->len
- offset
;
273 err
= map_private_extent_buffer(buf
, offset
, 32,
274 &kaddr
, &map_start
, &map_len
);
277 cur_len
= min(len
, map_len
- (offset
- map_start
));
278 crc
= btrfs_csum_data(root
, kaddr
+ offset
- map_start
,
283 if (csum_size
> sizeof(inline_result
)) {
284 result
= kzalloc(csum_size
* sizeof(char), GFP_NOFS
);
288 result
= (char *)&inline_result
;
291 btrfs_csum_final(crc
, result
);
294 if (memcmp_extent_buffer(buf
, result
, 0, csum_size
)) {
297 memcpy(&found
, result
, csum_size
);
299 read_extent_buffer(buf
, &val
, 0, csum_size
);
300 printk_ratelimited(KERN_INFO
"btrfs: %s checksum verify "
301 "failed on %llu wanted %X found %X "
303 root
->fs_info
->sb
->s_id
,
304 (unsigned long long)buf
->start
, val
, found
,
305 btrfs_header_level(buf
));
306 if (result
!= (char *)&inline_result
)
311 write_extent_buffer(buf
, result
, 0, csum_size
);
313 if (result
!= (char *)&inline_result
)
319 * we can't consider a given block up to date unless the transid of the
320 * block matches the transid in the parent node's pointer. This is how we
321 * detect blocks that either didn't get written at all or got written
322 * in the wrong place.
324 static int verify_parent_transid(struct extent_io_tree
*io_tree
,
325 struct extent_buffer
*eb
, u64 parent_transid
)
327 struct extent_state
*cached_state
= NULL
;
330 if (!parent_transid
|| btrfs_header_generation(eb
) == parent_transid
)
333 lock_extent_bits(io_tree
, eb
->start
, eb
->start
+ eb
->len
- 1,
335 if (extent_buffer_uptodate(io_tree
, eb
, cached_state
) &&
336 btrfs_header_generation(eb
) == parent_transid
) {
340 printk_ratelimited("parent transid verify failed on %llu wanted %llu "
342 (unsigned long long)eb
->start
,
343 (unsigned long long)parent_transid
,
344 (unsigned long long)btrfs_header_generation(eb
));
346 clear_extent_buffer_uptodate(io_tree
, eb
, &cached_state
);
348 unlock_extent_cached(io_tree
, eb
->start
, eb
->start
+ eb
->len
- 1,
349 &cached_state
, GFP_NOFS
);
354 * helper to read a given tree block, doing retries as required when
355 * the checksums don't match and we have alternate mirrors to try.
357 static int btree_read_extent_buffer_pages(struct btrfs_root
*root
,
358 struct extent_buffer
*eb
,
359 u64 start
, u64 parent_transid
)
361 struct extent_io_tree
*io_tree
;
366 clear_bit(EXTENT_BUFFER_CORRUPT
, &eb
->bflags
);
367 io_tree
= &BTRFS_I(root
->fs_info
->btree_inode
)->io_tree
;
369 ret
= read_extent_buffer_pages(io_tree
, eb
, start
,
371 btree_get_extent
, mirror_num
);
373 !verify_parent_transid(io_tree
, eb
, parent_transid
))
377 * This buffer's crc is fine, but its contents are corrupted, so
378 * there is no reason to read the other copies, they won't be
381 if (test_bit(EXTENT_BUFFER_CORRUPT
, &eb
->bflags
))
384 num_copies
= btrfs_num_copies(&root
->fs_info
->mapping_tree
,
390 if (mirror_num
> num_copies
)
397 * checksum a dirty tree block before IO. This has extra checks to make sure
398 * we only fill in the checksum field in the first page of a multi-page block
401 static int csum_dirty_buffer(struct btrfs_root
*root
, struct page
*page
)
403 struct extent_io_tree
*tree
;
404 u64 start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
407 struct extent_buffer
*eb
;
410 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
412 if (page
->private == EXTENT_PAGE_PRIVATE
) {
416 if (!page
->private) {
420 len
= page
->private >> 2;
423 eb
= alloc_extent_buffer(tree
, start
, len
, page
);
428 ret
= btree_read_extent_buffer_pages(root
, eb
, start
+ PAGE_CACHE_SIZE
,
429 btrfs_header_generation(eb
));
431 WARN_ON(!btrfs_header_flag(eb
, BTRFS_HEADER_FLAG_WRITTEN
));
433 found_start
= btrfs_header_bytenr(eb
);
434 if (found_start
!= start
) {
438 if (eb
->first_page
!= page
) {
442 if (!PageUptodate(page
)) {
446 csum_tree_block(root
, eb
, 0);
448 free_extent_buffer(eb
);
453 static int check_tree_block_fsid(struct btrfs_root
*root
,
454 struct extent_buffer
*eb
)
456 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
457 u8 fsid
[BTRFS_UUID_SIZE
];
460 read_extent_buffer(eb
, fsid
, (unsigned long)btrfs_header_fsid(eb
),
463 if (!memcmp(fsid
, fs_devices
->fsid
, BTRFS_FSID_SIZE
)) {
467 fs_devices
= fs_devices
->seed
;
472 #define CORRUPT(reason, eb, root, slot) \
473 printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \
474 "root=%llu, slot=%d\n", reason, \
475 (unsigned long long)btrfs_header_bytenr(eb), \
476 (unsigned long long)root->objectid, slot)
478 static noinline
int check_leaf(struct btrfs_root
*root
,
479 struct extent_buffer
*leaf
)
481 struct btrfs_key key
;
482 struct btrfs_key leaf_key
;
483 u32 nritems
= btrfs_header_nritems(leaf
);
489 /* Check the 0 item */
490 if (btrfs_item_offset_nr(leaf
, 0) + btrfs_item_size_nr(leaf
, 0) !=
491 BTRFS_LEAF_DATA_SIZE(root
)) {
492 CORRUPT("invalid item offset size pair", leaf
, root
, 0);
497 * Check to make sure each items keys are in the correct order and their
498 * offsets make sense. We only have to loop through nritems-1 because
499 * we check the current slot against the next slot, which verifies the
500 * next slot's offset+size makes sense and that the current's slot
503 for (slot
= 0; slot
< nritems
- 1; slot
++) {
504 btrfs_item_key_to_cpu(leaf
, &leaf_key
, slot
);
505 btrfs_item_key_to_cpu(leaf
, &key
, slot
+ 1);
507 /* Make sure the keys are in the right order */
508 if (btrfs_comp_cpu_keys(&leaf_key
, &key
) >= 0) {
509 CORRUPT("bad key order", leaf
, root
, slot
);
514 * Make sure the offset and ends are right, remember that the
515 * item data starts at the end of the leaf and grows towards the
518 if (btrfs_item_offset_nr(leaf
, slot
) !=
519 btrfs_item_end_nr(leaf
, slot
+ 1)) {
520 CORRUPT("slot offset bad", leaf
, root
, slot
);
525 * Check to make sure that we don't point outside of the leaf,
526 * just incase all the items are consistent to eachother, but
527 * all point outside of the leaf.
529 if (btrfs_item_end_nr(leaf
, slot
) >
530 BTRFS_LEAF_DATA_SIZE(root
)) {
531 CORRUPT("slot end outside of leaf", leaf
, root
, slot
);
539 static int btree_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
540 struct extent_state
*state
)
542 struct extent_io_tree
*tree
;
546 struct extent_buffer
*eb
;
547 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
550 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
551 if (page
->private == EXTENT_PAGE_PRIVATE
)
556 len
= page
->private >> 2;
559 eb
= alloc_extent_buffer(tree
, start
, len
, page
);
565 found_start
= btrfs_header_bytenr(eb
);
566 if (found_start
!= start
) {
567 printk_ratelimited(KERN_INFO
"btrfs bad tree block start "
569 (unsigned long long)found_start
,
570 (unsigned long long)eb
->start
);
574 if (eb
->first_page
!= page
) {
575 printk(KERN_INFO
"btrfs bad first page %lu %lu\n",
576 eb
->first_page
->index
, page
->index
);
581 if (check_tree_block_fsid(root
, eb
)) {
582 printk_ratelimited(KERN_INFO
"btrfs bad fsid on block %llu\n",
583 (unsigned long long)eb
->start
);
587 found_level
= btrfs_header_level(eb
);
589 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb
),
592 ret
= csum_tree_block(root
, eb
, 1);
599 * If this is a leaf block and it is corrupt, set the corrupt bit so
600 * that we don't try and read the other copies of this block, just
603 if (found_level
== 0 && check_leaf(root
, eb
)) {
604 set_bit(EXTENT_BUFFER_CORRUPT
, &eb
->bflags
);
608 end
= min_t(u64
, eb
->len
, PAGE_CACHE_SIZE
);
609 end
= eb
->start
+ end
- 1;
611 if (test_bit(EXTENT_BUFFER_READAHEAD
, &eb
->bflags
)) {
612 clear_bit(EXTENT_BUFFER_READAHEAD
, &eb
->bflags
);
613 btree_readahead_hook(root
, eb
, eb
->start
, ret
);
616 free_extent_buffer(eb
);
621 static int btree_io_failed_hook(struct bio
*failed_bio
,
622 struct page
*page
, u64 start
, u64 end
,
623 int mirror_num
, struct extent_state
*state
)
625 struct extent_io_tree
*tree
;
627 struct extent_buffer
*eb
;
628 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
630 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
631 if (page
->private == EXTENT_PAGE_PRIVATE
)
636 len
= page
->private >> 2;
639 eb
= alloc_extent_buffer(tree
, start
, len
, page
);
643 if (test_bit(EXTENT_BUFFER_READAHEAD
, &eb
->bflags
)) {
644 clear_bit(EXTENT_BUFFER_READAHEAD
, &eb
->bflags
);
645 btree_readahead_hook(root
, eb
, eb
->start
, -EIO
);
647 free_extent_buffer(eb
);
650 return -EIO
; /* we fixed nothing */
653 static void end_workqueue_bio(struct bio
*bio
, int err
)
655 struct end_io_wq
*end_io_wq
= bio
->bi_private
;
656 struct btrfs_fs_info
*fs_info
;
658 fs_info
= end_io_wq
->info
;
659 end_io_wq
->error
= err
;
660 end_io_wq
->work
.func
= end_workqueue_fn
;
661 end_io_wq
->work
.flags
= 0;
663 if (bio
->bi_rw
& REQ_WRITE
) {
664 if (end_io_wq
->metadata
== 1)
665 btrfs_queue_worker(&fs_info
->endio_meta_write_workers
,
667 else if (end_io_wq
->metadata
== 2)
668 btrfs_queue_worker(&fs_info
->endio_freespace_worker
,
671 btrfs_queue_worker(&fs_info
->endio_write_workers
,
674 if (end_io_wq
->metadata
)
675 btrfs_queue_worker(&fs_info
->endio_meta_workers
,
678 btrfs_queue_worker(&fs_info
->endio_workers
,
684 * For the metadata arg you want
687 * 1 - if normal metadta
688 * 2 - if writing to the free space cache area
690 int btrfs_bio_wq_end_io(struct btrfs_fs_info
*info
, struct bio
*bio
,
693 struct end_io_wq
*end_io_wq
;
694 end_io_wq
= kmalloc(sizeof(*end_io_wq
), GFP_NOFS
);
698 end_io_wq
->private = bio
->bi_private
;
699 end_io_wq
->end_io
= bio
->bi_end_io
;
700 end_io_wq
->info
= info
;
701 end_io_wq
->error
= 0;
702 end_io_wq
->bio
= bio
;
703 end_io_wq
->metadata
= metadata
;
705 bio
->bi_private
= end_io_wq
;
706 bio
->bi_end_io
= end_workqueue_bio
;
710 unsigned long btrfs_async_submit_limit(struct btrfs_fs_info
*info
)
712 unsigned long limit
= min_t(unsigned long,
713 info
->workers
.max_workers
,
714 info
->fs_devices
->open_devices
);
718 static void run_one_async_start(struct btrfs_work
*work
)
720 struct async_submit_bio
*async
;
722 async
= container_of(work
, struct async_submit_bio
, work
);
723 async
->submit_bio_start(async
->inode
, async
->rw
, async
->bio
,
724 async
->mirror_num
, async
->bio_flags
,
728 static void run_one_async_done(struct btrfs_work
*work
)
730 struct btrfs_fs_info
*fs_info
;
731 struct async_submit_bio
*async
;
734 async
= container_of(work
, struct async_submit_bio
, work
);
735 fs_info
= BTRFS_I(async
->inode
)->root
->fs_info
;
737 limit
= btrfs_async_submit_limit(fs_info
);
738 limit
= limit
* 2 / 3;
740 atomic_dec(&fs_info
->nr_async_submits
);
742 if (atomic_read(&fs_info
->nr_async_submits
) < limit
&&
743 waitqueue_active(&fs_info
->async_submit_wait
))
744 wake_up(&fs_info
->async_submit_wait
);
746 async
->submit_bio_done(async
->inode
, async
->rw
, async
->bio
,
747 async
->mirror_num
, async
->bio_flags
,
751 static void run_one_async_free(struct btrfs_work
*work
)
753 struct async_submit_bio
*async
;
755 async
= container_of(work
, struct async_submit_bio
, work
);
759 int btrfs_wq_submit_bio(struct btrfs_fs_info
*fs_info
, struct inode
*inode
,
760 int rw
, struct bio
*bio
, int mirror_num
,
761 unsigned long bio_flags
,
763 extent_submit_bio_hook_t
*submit_bio_start
,
764 extent_submit_bio_hook_t
*submit_bio_done
)
766 struct async_submit_bio
*async
;
768 async
= kmalloc(sizeof(*async
), GFP_NOFS
);
772 async
->inode
= inode
;
775 async
->mirror_num
= mirror_num
;
776 async
->submit_bio_start
= submit_bio_start
;
777 async
->submit_bio_done
= submit_bio_done
;
779 async
->work
.func
= run_one_async_start
;
780 async
->work
.ordered_func
= run_one_async_done
;
781 async
->work
.ordered_free
= run_one_async_free
;
783 async
->work
.flags
= 0;
784 async
->bio_flags
= bio_flags
;
785 async
->bio_offset
= bio_offset
;
787 atomic_inc(&fs_info
->nr_async_submits
);
790 btrfs_set_work_high_prio(&async
->work
);
792 btrfs_queue_worker(&fs_info
->workers
, &async
->work
);
794 while (atomic_read(&fs_info
->async_submit_draining
) &&
795 atomic_read(&fs_info
->nr_async_submits
)) {
796 wait_event(fs_info
->async_submit_wait
,
797 (atomic_read(&fs_info
->nr_async_submits
) == 0));
803 static int btree_csum_one_bio(struct bio
*bio
)
805 struct bio_vec
*bvec
= bio
->bi_io_vec
;
807 struct btrfs_root
*root
;
809 WARN_ON(bio
->bi_vcnt
<= 0);
810 while (bio_index
< bio
->bi_vcnt
) {
811 root
= BTRFS_I(bvec
->bv_page
->mapping
->host
)->root
;
812 csum_dirty_buffer(root
, bvec
->bv_page
);
819 static int __btree_submit_bio_start(struct inode
*inode
, int rw
,
820 struct bio
*bio
, int mirror_num
,
821 unsigned long bio_flags
,
825 * when we're called for a write, we're already in the async
826 * submission context. Just jump into btrfs_map_bio
828 btree_csum_one_bio(bio
);
832 static int __btree_submit_bio_done(struct inode
*inode
, int rw
, struct bio
*bio
,
833 int mirror_num
, unsigned long bio_flags
,
837 * when we're called for a write, we're already in the async
838 * submission context. Just jump into btrfs_map_bio
840 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
, mirror_num
, 1);
843 static int btree_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
844 int mirror_num
, unsigned long bio_flags
,
849 ret
= btrfs_bio_wq_end_io(BTRFS_I(inode
)->root
->fs_info
, bio
, 1);
853 if (!(rw
& REQ_WRITE
)) {
855 * called for a read, do the setup so that checksum validation
856 * can happen in the async kernel threads
858 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
,
863 * kthread helpers are used to submit writes so that checksumming
864 * can happen in parallel across all CPUs
866 return btrfs_wq_submit_bio(BTRFS_I(inode
)->root
->fs_info
,
867 inode
, rw
, bio
, mirror_num
, 0,
869 __btree_submit_bio_start
,
870 __btree_submit_bio_done
);
873 #ifdef CONFIG_MIGRATION
874 static int btree_migratepage(struct address_space
*mapping
,
875 struct page
*newpage
, struct page
*page
,
876 enum migrate_mode mode
)
879 * we can't safely write a btree page from here,
880 * we haven't done the locking hook
885 * Buffers may be managed in a filesystem specific way.
886 * We must have no buffers or drop them.
888 if (page_has_private(page
) &&
889 !try_to_release_page(page
, GFP_KERNEL
))
891 return migrate_page(mapping
, newpage
, page
, mode
);
895 static int btree_writepage(struct page
*page
, struct writeback_control
*wbc
)
897 struct extent_io_tree
*tree
;
898 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
899 struct extent_buffer
*eb
;
902 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
903 if (!(current
->flags
& PF_MEMALLOC
)) {
904 return extent_write_full_page(tree
, page
,
905 btree_get_extent
, wbc
);
908 redirty_page_for_writepage(wbc
, page
);
909 eb
= btrfs_find_tree_block(root
, page_offset(page
), PAGE_CACHE_SIZE
);
912 was_dirty
= test_and_set_bit(EXTENT_BUFFER_DIRTY
, &eb
->bflags
);
914 spin_lock(&root
->fs_info
->delalloc_lock
);
915 root
->fs_info
->dirty_metadata_bytes
+= PAGE_CACHE_SIZE
;
916 spin_unlock(&root
->fs_info
->delalloc_lock
);
918 free_extent_buffer(eb
);
924 static int btree_writepages(struct address_space
*mapping
,
925 struct writeback_control
*wbc
)
927 struct extent_io_tree
*tree
;
928 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
929 if (wbc
->sync_mode
== WB_SYNC_NONE
) {
930 struct btrfs_root
*root
= BTRFS_I(mapping
->host
)->root
;
932 unsigned long thresh
= 32 * 1024 * 1024;
934 if (wbc
->for_kupdate
)
937 /* this is a bit racy, but that's ok */
938 num_dirty
= root
->fs_info
->dirty_metadata_bytes
;
939 if (num_dirty
< thresh
)
942 return extent_writepages(tree
, mapping
, btree_get_extent
, wbc
);
945 static int btree_readpage(struct file
*file
, struct page
*page
)
947 struct extent_io_tree
*tree
;
948 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
949 return extent_read_full_page(tree
, page
, btree_get_extent
, 0);
952 static int btree_releasepage(struct page
*page
, gfp_t gfp_flags
)
954 struct extent_io_tree
*tree
;
955 struct extent_map_tree
*map
;
958 if (PageWriteback(page
) || PageDirty(page
))
961 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
962 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
965 * We need to mask out eg. __GFP_HIGHMEM and __GFP_DMA32 as we're doing
966 * slab allocation from alloc_extent_state down the callchain where
967 * it'd hit a BUG_ON as those flags are not allowed.
969 gfp_flags
&= ~GFP_SLAB_BUG_MASK
;
971 ret
= try_release_extent_state(map
, tree
, page
, gfp_flags
);
975 ret
= try_release_extent_buffer(tree
, page
);
977 ClearPagePrivate(page
);
978 set_page_private(page
, 0);
979 page_cache_release(page
);
985 static void btree_invalidatepage(struct page
*page
, unsigned long offset
)
987 struct extent_io_tree
*tree
;
988 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
989 extent_invalidatepage(tree
, page
, offset
);
990 btree_releasepage(page
, GFP_NOFS
);
991 if (PagePrivate(page
)) {
992 printk(KERN_WARNING
"btrfs warning page private not zero "
993 "on page %llu\n", (unsigned long long)page_offset(page
));
994 ClearPagePrivate(page
);
995 set_page_private(page
, 0);
996 page_cache_release(page
);
1000 static const struct address_space_operations btree_aops
= {
1001 .readpage
= btree_readpage
,
1002 .writepage
= btree_writepage
,
1003 .writepages
= btree_writepages
,
1004 .releasepage
= btree_releasepage
,
1005 .invalidatepage
= btree_invalidatepage
,
1006 #ifdef CONFIG_MIGRATION
1007 .migratepage
= btree_migratepage
,
1011 int readahead_tree_block(struct btrfs_root
*root
, u64 bytenr
, u32 blocksize
,
1014 struct extent_buffer
*buf
= NULL
;
1015 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1018 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
1021 read_extent_buffer_pages(&BTRFS_I(btree_inode
)->io_tree
,
1022 buf
, 0, WAIT_NONE
, btree_get_extent
, 0);
1023 free_extent_buffer(buf
);
1027 int reada_tree_block_flagged(struct btrfs_root
*root
, u64 bytenr
, u32 blocksize
,
1028 int mirror_num
, struct extent_buffer
**eb
)
1030 struct extent_buffer
*buf
= NULL
;
1031 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1032 struct extent_io_tree
*io_tree
= &BTRFS_I(btree_inode
)->io_tree
;
1035 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
1039 set_bit(EXTENT_BUFFER_READAHEAD
, &buf
->bflags
);
1041 ret
= read_extent_buffer_pages(io_tree
, buf
, 0, WAIT_PAGE_LOCK
,
1042 btree_get_extent
, mirror_num
);
1044 free_extent_buffer(buf
);
1048 if (test_bit(EXTENT_BUFFER_CORRUPT
, &buf
->bflags
)) {
1049 free_extent_buffer(buf
);
1051 } else if (extent_buffer_uptodate(io_tree
, buf
, NULL
)) {
1054 free_extent_buffer(buf
);
1059 struct extent_buffer
*btrfs_find_tree_block(struct btrfs_root
*root
,
1060 u64 bytenr
, u32 blocksize
)
1062 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1063 struct extent_buffer
*eb
;
1064 eb
= find_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
1069 struct extent_buffer
*btrfs_find_create_tree_block(struct btrfs_root
*root
,
1070 u64 bytenr
, u32 blocksize
)
1072 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1073 struct extent_buffer
*eb
;
1075 eb
= alloc_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
1076 bytenr
, blocksize
, NULL
);
1081 int btrfs_write_tree_block(struct extent_buffer
*buf
)
1083 return filemap_fdatawrite_range(buf
->first_page
->mapping
, buf
->start
,
1084 buf
->start
+ buf
->len
- 1);
1087 int btrfs_wait_tree_block_writeback(struct extent_buffer
*buf
)
1089 return filemap_fdatawait_range(buf
->first_page
->mapping
,
1090 buf
->start
, buf
->start
+ buf
->len
- 1);
1093 struct extent_buffer
*read_tree_block(struct btrfs_root
*root
, u64 bytenr
,
1094 u32 blocksize
, u64 parent_transid
)
1096 struct extent_buffer
*buf
= NULL
;
1099 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
1103 ret
= btree_read_extent_buffer_pages(root
, buf
, 0, parent_transid
);
1106 set_bit(EXTENT_BUFFER_UPTODATE
, &buf
->bflags
);
1111 void clean_tree_block(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
1112 struct extent_buffer
*buf
)
1114 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1115 if (btrfs_header_generation(buf
) ==
1116 root
->fs_info
->running_transaction
->transid
) {
1117 btrfs_assert_tree_locked(buf
);
1119 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY
, &buf
->bflags
)) {
1120 spin_lock(&root
->fs_info
->delalloc_lock
);
1121 if (root
->fs_info
->dirty_metadata_bytes
>= buf
->len
)
1122 root
->fs_info
->dirty_metadata_bytes
-= buf
->len
;
1124 spin_unlock(&root
->fs_info
->delalloc_lock
);
1125 btrfs_panic(root
->fs_info
, -EOVERFLOW
,
1126 "Can't clear %lu bytes from "
1127 " dirty_mdatadata_bytes (%lu)",
1129 root
->fs_info
->dirty_metadata_bytes
);
1131 spin_unlock(&root
->fs_info
->delalloc_lock
);
1134 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1135 btrfs_set_lock_blocking(buf
);
1136 clear_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
,
1141 static void __setup_root(u32 nodesize
, u32 leafsize
, u32 sectorsize
,
1142 u32 stripesize
, struct btrfs_root
*root
,
1143 struct btrfs_fs_info
*fs_info
,
1147 root
->commit_root
= NULL
;
1148 root
->sectorsize
= sectorsize
;
1149 root
->nodesize
= nodesize
;
1150 root
->leafsize
= leafsize
;
1151 root
->stripesize
= stripesize
;
1153 root
->track_dirty
= 0;
1155 root
->orphan_item_inserted
= 0;
1156 root
->orphan_cleanup_state
= 0;
1158 root
->objectid
= objectid
;
1159 root
->last_trans
= 0;
1160 root
->highest_objectid
= 0;
1162 root
->inode_tree
= RB_ROOT
;
1163 INIT_RADIX_TREE(&root
->delayed_nodes_tree
, GFP_ATOMIC
);
1164 root
->block_rsv
= NULL
;
1165 root
->orphan_block_rsv
= NULL
;
1167 INIT_LIST_HEAD(&root
->dirty_list
);
1168 INIT_LIST_HEAD(&root
->orphan_list
);
1169 INIT_LIST_HEAD(&root
->root_list
);
1170 spin_lock_init(&root
->orphan_lock
);
1171 spin_lock_init(&root
->inode_lock
);
1172 spin_lock_init(&root
->accounting_lock
);
1173 mutex_init(&root
->objectid_mutex
);
1174 mutex_init(&root
->log_mutex
);
1175 init_waitqueue_head(&root
->log_writer_wait
);
1176 init_waitqueue_head(&root
->log_commit_wait
[0]);
1177 init_waitqueue_head(&root
->log_commit_wait
[1]);
1178 atomic_set(&root
->log_commit
[0], 0);
1179 atomic_set(&root
->log_commit
[1], 0);
1180 atomic_set(&root
->log_writers
, 0);
1181 root
->log_batch
= 0;
1182 root
->log_transid
= 0;
1183 root
->last_log_commit
= 0;
1184 extent_io_tree_init(&root
->dirty_log_pages
,
1185 fs_info
->btree_inode
->i_mapping
);
1187 memset(&root
->root_key
, 0, sizeof(root
->root_key
));
1188 memset(&root
->root_item
, 0, sizeof(root
->root_item
));
1189 memset(&root
->defrag_progress
, 0, sizeof(root
->defrag_progress
));
1190 memset(&root
->root_kobj
, 0, sizeof(root
->root_kobj
));
1191 root
->defrag_trans_start
= fs_info
->generation
;
1192 init_completion(&root
->kobj_unregister
);
1193 root
->defrag_running
= 0;
1194 root
->root_key
.objectid
= objectid
;
1198 static int __must_check
find_and_setup_root(struct btrfs_root
*tree_root
,
1199 struct btrfs_fs_info
*fs_info
,
1201 struct btrfs_root
*root
)
1207 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
1208 tree_root
->sectorsize
, tree_root
->stripesize
,
1209 root
, fs_info
, objectid
);
1210 ret
= btrfs_find_last_root(tree_root
, objectid
,
1211 &root
->root_item
, &root
->root_key
);
1217 generation
= btrfs_root_generation(&root
->root_item
);
1218 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
1219 root
->commit_root
= NULL
;
1220 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
1221 blocksize
, generation
);
1222 if (!root
->node
|| !btrfs_buffer_uptodate(root
->node
, generation
)) {
1223 free_extent_buffer(root
->node
);
1227 root
->commit_root
= btrfs_root_node(root
);
1231 static struct btrfs_root
*btrfs_alloc_root(struct btrfs_fs_info
*fs_info
)
1233 struct btrfs_root
*root
= kzalloc(sizeof(*root
), GFP_NOFS
);
1235 root
->fs_info
= fs_info
;
1239 static struct btrfs_root
*alloc_log_tree(struct btrfs_trans_handle
*trans
,
1240 struct btrfs_fs_info
*fs_info
)
1242 struct btrfs_root
*root
;
1243 struct btrfs_root
*tree_root
= fs_info
->tree_root
;
1244 struct extent_buffer
*leaf
;
1246 root
= btrfs_alloc_root(fs_info
);
1248 return ERR_PTR(-ENOMEM
);
1250 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
1251 tree_root
->sectorsize
, tree_root
->stripesize
,
1252 root
, fs_info
, BTRFS_TREE_LOG_OBJECTID
);
1254 root
->root_key
.objectid
= BTRFS_TREE_LOG_OBJECTID
;
1255 root
->root_key
.type
= BTRFS_ROOT_ITEM_KEY
;
1256 root
->root_key
.offset
= BTRFS_TREE_LOG_OBJECTID
;
1258 * log trees do not get reference counted because they go away
1259 * before a real commit is actually done. They do store pointers
1260 * to file data extents, and those reference counts still get
1261 * updated (along with back refs to the log tree).
1265 leaf
= btrfs_alloc_free_block(trans
, root
, root
->leafsize
, 0,
1266 BTRFS_TREE_LOG_OBJECTID
, NULL
,
1270 return ERR_CAST(leaf
);
1273 memset_extent_buffer(leaf
, 0, 0, sizeof(struct btrfs_header
));
1274 btrfs_set_header_bytenr(leaf
, leaf
->start
);
1275 btrfs_set_header_generation(leaf
, trans
->transid
);
1276 btrfs_set_header_backref_rev(leaf
, BTRFS_MIXED_BACKREF_REV
);
1277 btrfs_set_header_owner(leaf
, BTRFS_TREE_LOG_OBJECTID
);
1280 write_extent_buffer(root
->node
, root
->fs_info
->fsid
,
1281 (unsigned long)btrfs_header_fsid(root
->node
),
1283 btrfs_mark_buffer_dirty(root
->node
);
1284 btrfs_tree_unlock(root
->node
);
1288 int btrfs_init_log_root_tree(struct btrfs_trans_handle
*trans
,
1289 struct btrfs_fs_info
*fs_info
)
1291 struct btrfs_root
*log_root
;
1293 log_root
= alloc_log_tree(trans
, fs_info
);
1294 if (IS_ERR(log_root
))
1295 return PTR_ERR(log_root
);
1296 WARN_ON(fs_info
->log_root_tree
);
1297 fs_info
->log_root_tree
= log_root
;
1301 int btrfs_add_log_tree(struct btrfs_trans_handle
*trans
,
1302 struct btrfs_root
*root
)
1304 struct btrfs_root
*log_root
;
1305 struct btrfs_inode_item
*inode_item
;
1307 log_root
= alloc_log_tree(trans
, root
->fs_info
);
1308 if (IS_ERR(log_root
))
1309 return PTR_ERR(log_root
);
1311 log_root
->last_trans
= trans
->transid
;
1312 log_root
->root_key
.offset
= root
->root_key
.objectid
;
1314 inode_item
= &log_root
->root_item
.inode
;
1315 inode_item
->generation
= cpu_to_le64(1);
1316 inode_item
->size
= cpu_to_le64(3);
1317 inode_item
->nlink
= cpu_to_le32(1);
1318 inode_item
->nbytes
= cpu_to_le64(root
->leafsize
);
1319 inode_item
->mode
= cpu_to_le32(S_IFDIR
| 0755);
1321 btrfs_set_root_node(&log_root
->root_item
, log_root
->node
);
1323 WARN_ON(root
->log_root
);
1324 root
->log_root
= log_root
;
1325 root
->log_transid
= 0;
1326 root
->last_log_commit
= 0;
1330 struct btrfs_root
*btrfs_read_fs_root_no_radix(struct btrfs_root
*tree_root
,
1331 struct btrfs_key
*location
)
1333 struct btrfs_root
*root
;
1334 struct btrfs_fs_info
*fs_info
= tree_root
->fs_info
;
1335 struct btrfs_path
*path
;
1336 struct extent_buffer
*l
;
1341 root
= btrfs_alloc_root(fs_info
);
1343 return ERR_PTR(-ENOMEM
);
1344 if (location
->offset
== (u64
)-1) {
1345 ret
= find_and_setup_root(tree_root
, fs_info
,
1346 location
->objectid
, root
);
1349 return ERR_PTR(ret
);
1354 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
1355 tree_root
->sectorsize
, tree_root
->stripesize
,
1356 root
, fs_info
, location
->objectid
);
1358 path
= btrfs_alloc_path();
1361 return ERR_PTR(-ENOMEM
);
1363 ret
= btrfs_search_slot(NULL
, tree_root
, location
, path
, 0, 0);
1366 read_extent_buffer(l
, &root
->root_item
,
1367 btrfs_item_ptr_offset(l
, path
->slots
[0]),
1368 sizeof(root
->root_item
));
1369 memcpy(&root
->root_key
, location
, sizeof(*location
));
1371 btrfs_free_path(path
);
1376 return ERR_PTR(ret
);
1379 generation
= btrfs_root_generation(&root
->root_item
);
1380 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
1381 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
1382 blocksize
, generation
);
1383 root
->commit_root
= btrfs_root_node(root
);
1384 BUG_ON(!root
->node
);
1386 if (location
->objectid
!= BTRFS_TREE_LOG_OBJECTID
) {
1388 btrfs_check_and_init_root_item(&root
->root_item
);
1394 struct btrfs_root
*btrfs_read_fs_root_no_name(struct btrfs_fs_info
*fs_info
,
1395 struct btrfs_key
*location
)
1397 struct btrfs_root
*root
;
1400 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1401 return fs_info
->tree_root
;
1402 if (location
->objectid
== BTRFS_EXTENT_TREE_OBJECTID
)
1403 return fs_info
->extent_root
;
1404 if (location
->objectid
== BTRFS_CHUNK_TREE_OBJECTID
)
1405 return fs_info
->chunk_root
;
1406 if (location
->objectid
== BTRFS_DEV_TREE_OBJECTID
)
1407 return fs_info
->dev_root
;
1408 if (location
->objectid
== BTRFS_CSUM_TREE_OBJECTID
)
1409 return fs_info
->csum_root
;
1411 spin_lock(&fs_info
->fs_roots_radix_lock
);
1412 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
1413 (unsigned long)location
->objectid
);
1414 spin_unlock(&fs_info
->fs_roots_radix_lock
);
1418 root
= btrfs_read_fs_root_no_radix(fs_info
->tree_root
, location
);
1422 root
->free_ino_ctl
= kzalloc(sizeof(*root
->free_ino_ctl
), GFP_NOFS
);
1423 root
->free_ino_pinned
= kzalloc(sizeof(*root
->free_ino_pinned
),
1425 if (!root
->free_ino_pinned
|| !root
->free_ino_ctl
) {
1430 btrfs_init_free_ino_ctl(root
);
1431 mutex_init(&root
->fs_commit_mutex
);
1432 spin_lock_init(&root
->cache_lock
);
1433 init_waitqueue_head(&root
->cache_wait
);
1435 ret
= get_anon_bdev(&root
->anon_dev
);
1439 if (btrfs_root_refs(&root
->root_item
) == 0) {
1444 ret
= btrfs_find_orphan_item(fs_info
->tree_root
, location
->objectid
);
1448 root
->orphan_item_inserted
= 1;
1450 ret
= radix_tree_preload(GFP_NOFS
& ~__GFP_HIGHMEM
);
1454 spin_lock(&fs_info
->fs_roots_radix_lock
);
1455 ret
= radix_tree_insert(&fs_info
->fs_roots_radix
,
1456 (unsigned long)root
->root_key
.objectid
,
1461 spin_unlock(&fs_info
->fs_roots_radix_lock
);
1462 radix_tree_preload_end();
1464 if (ret
== -EEXIST
) {
1471 ret
= btrfs_find_dead_roots(fs_info
->tree_root
,
1472 root
->root_key
.objectid
);
1477 return ERR_PTR(ret
);
1480 static int btrfs_congested_fn(void *congested_data
, int bdi_bits
)
1482 struct btrfs_fs_info
*info
= (struct btrfs_fs_info
*)congested_data
;
1484 struct btrfs_device
*device
;
1485 struct backing_dev_info
*bdi
;
1488 list_for_each_entry_rcu(device
, &info
->fs_devices
->devices
, dev_list
) {
1491 bdi
= blk_get_backing_dev_info(device
->bdev
);
1492 if (bdi
&& bdi_congested(bdi
, bdi_bits
)) {
1502 * If this fails, caller must call bdi_destroy() to get rid of the
1505 static int setup_bdi(struct btrfs_fs_info
*info
, struct backing_dev_info
*bdi
)
1509 bdi
->capabilities
= BDI_CAP_MAP_COPY
;
1510 err
= bdi_setup_and_register(bdi
, "btrfs", BDI_CAP_MAP_COPY
);
1514 bdi
->ra_pages
= default_backing_dev_info
.ra_pages
;
1515 bdi
->congested_fn
= btrfs_congested_fn
;
1516 bdi
->congested_data
= info
;
1520 static int bio_ready_for_csum(struct bio
*bio
)
1526 struct extent_io_tree
*io_tree
= NULL
;
1527 struct bio_vec
*bvec
;
1531 bio_for_each_segment(bvec
, bio
, i
) {
1532 page
= bvec
->bv_page
;
1533 if (page
->private == EXTENT_PAGE_PRIVATE
) {
1534 length
+= bvec
->bv_len
;
1537 if (!page
->private) {
1538 length
+= bvec
->bv_len
;
1541 length
= bvec
->bv_len
;
1542 buf_len
= page
->private >> 2;
1543 start
= page_offset(page
) + bvec
->bv_offset
;
1544 io_tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
1546 /* are we fully contained in this bio? */
1547 if (buf_len
<= length
)
1550 ret
= extent_range_uptodate(io_tree
, start
+ length
,
1551 start
+ buf_len
- 1);
1556 * called by the kthread helper functions to finally call the bio end_io
1557 * functions. This is where read checksum verification actually happens
1559 static void end_workqueue_fn(struct btrfs_work
*work
)
1562 struct end_io_wq
*end_io_wq
;
1563 struct btrfs_fs_info
*fs_info
;
1566 end_io_wq
= container_of(work
, struct end_io_wq
, work
);
1567 bio
= end_io_wq
->bio
;
1568 fs_info
= end_io_wq
->info
;
1570 /* metadata bio reads are special because the whole tree block must
1571 * be checksummed at once. This makes sure the entire block is in
1572 * ram and up to date before trying to verify things. For
1573 * blocksize <= pagesize, it is basically a noop
1575 if (!(bio
->bi_rw
& REQ_WRITE
) && end_io_wq
->metadata
&&
1576 !bio_ready_for_csum(bio
)) {
1577 btrfs_queue_worker(&fs_info
->endio_meta_workers
,
1581 error
= end_io_wq
->error
;
1582 bio
->bi_private
= end_io_wq
->private;
1583 bio
->bi_end_io
= end_io_wq
->end_io
;
1585 bio_endio(bio
, error
);
1588 static int cleaner_kthread(void *arg
)
1590 struct btrfs_root
*root
= arg
;
1593 vfs_check_frozen(root
->fs_info
->sb
, SB_FREEZE_WRITE
);
1595 if (!(root
->fs_info
->sb
->s_flags
& MS_RDONLY
) &&
1596 mutex_trylock(&root
->fs_info
->cleaner_mutex
)) {
1597 btrfs_run_delayed_iputs(root
);
1598 btrfs_clean_old_snapshots(root
);
1599 mutex_unlock(&root
->fs_info
->cleaner_mutex
);
1600 btrfs_run_defrag_inodes(root
->fs_info
);
1603 if (!try_to_freeze()) {
1604 set_current_state(TASK_INTERRUPTIBLE
);
1605 if (!kthread_should_stop())
1607 __set_current_state(TASK_RUNNING
);
1609 } while (!kthread_should_stop());
1613 static int transaction_kthread(void *arg
)
1615 struct btrfs_root
*root
= arg
;
1616 struct btrfs_trans_handle
*trans
;
1617 struct btrfs_transaction
*cur
;
1620 unsigned long delay
;
1625 vfs_check_frozen(root
->fs_info
->sb
, SB_FREEZE_WRITE
);
1626 mutex_lock(&root
->fs_info
->transaction_kthread_mutex
);
1628 spin_lock(&root
->fs_info
->trans_lock
);
1629 cur
= root
->fs_info
->running_transaction
;
1631 spin_unlock(&root
->fs_info
->trans_lock
);
1635 now
= get_seconds();
1636 if (!cur
->blocked
&&
1637 (now
< cur
->start_time
|| now
- cur
->start_time
< 30)) {
1638 spin_unlock(&root
->fs_info
->trans_lock
);
1642 transid
= cur
->transid
;
1643 spin_unlock(&root
->fs_info
->trans_lock
);
1645 trans
= btrfs_join_transaction(root
);
1646 BUG_ON(IS_ERR(trans
));
1647 if (transid
== trans
->transid
) {
1648 ret
= btrfs_commit_transaction(trans
, root
);
1651 btrfs_end_transaction(trans
, root
);
1654 wake_up_process(root
->fs_info
->cleaner_kthread
);
1655 mutex_unlock(&root
->fs_info
->transaction_kthread_mutex
);
1657 if (!try_to_freeze()) {
1658 set_current_state(TASK_INTERRUPTIBLE
);
1659 if (!kthread_should_stop() &&
1660 !btrfs_transaction_blocked(root
->fs_info
))
1661 schedule_timeout(delay
);
1662 __set_current_state(TASK_RUNNING
);
1664 } while (!kthread_should_stop());
1669 * this will find the highest generation in the array of
1670 * root backups. The index of the highest array is returned,
1671 * or -1 if we can't find anything.
1673 * We check to make sure the array is valid by comparing the
1674 * generation of the latest root in the array with the generation
1675 * in the super block. If they don't match we pitch it.
1677 static int find_newest_super_backup(struct btrfs_fs_info
*info
, u64 newest_gen
)
1680 int newest_index
= -1;
1681 struct btrfs_root_backup
*root_backup
;
1684 for (i
= 0; i
< BTRFS_NUM_BACKUP_ROOTS
; i
++) {
1685 root_backup
= info
->super_copy
->super_roots
+ i
;
1686 cur
= btrfs_backup_tree_root_gen(root_backup
);
1687 if (cur
== newest_gen
)
1691 /* check to see if we actually wrapped around */
1692 if (newest_index
== BTRFS_NUM_BACKUP_ROOTS
- 1) {
1693 root_backup
= info
->super_copy
->super_roots
;
1694 cur
= btrfs_backup_tree_root_gen(root_backup
);
1695 if (cur
== newest_gen
)
1698 return newest_index
;
1703 * find the oldest backup so we know where to store new entries
1704 * in the backup array. This will set the backup_root_index
1705 * field in the fs_info struct
1707 static void find_oldest_super_backup(struct btrfs_fs_info
*info
,
1710 int newest_index
= -1;
1712 newest_index
= find_newest_super_backup(info
, newest_gen
);
1713 /* if there was garbage in there, just move along */
1714 if (newest_index
== -1) {
1715 info
->backup_root_index
= 0;
1717 info
->backup_root_index
= (newest_index
+ 1) % BTRFS_NUM_BACKUP_ROOTS
;
1722 * copy all the root pointers into the super backup array.
1723 * this will bump the backup pointer by one when it is
1726 static void backup_super_roots(struct btrfs_fs_info
*info
)
1729 struct btrfs_root_backup
*root_backup
;
1732 next_backup
= info
->backup_root_index
;
1733 last_backup
= (next_backup
+ BTRFS_NUM_BACKUP_ROOTS
- 1) %
1734 BTRFS_NUM_BACKUP_ROOTS
;
1737 * just overwrite the last backup if we're at the same generation
1738 * this happens only at umount
1740 root_backup
= info
->super_for_commit
->super_roots
+ last_backup
;
1741 if (btrfs_backup_tree_root_gen(root_backup
) ==
1742 btrfs_header_generation(info
->tree_root
->node
))
1743 next_backup
= last_backup
;
1745 root_backup
= info
->super_for_commit
->super_roots
+ next_backup
;
1748 * make sure all of our padding and empty slots get zero filled
1749 * regardless of which ones we use today
1751 memset(root_backup
, 0, sizeof(*root_backup
));
1753 info
->backup_root_index
= (next_backup
+ 1) % BTRFS_NUM_BACKUP_ROOTS
;
1755 btrfs_set_backup_tree_root(root_backup
, info
->tree_root
->node
->start
);
1756 btrfs_set_backup_tree_root_gen(root_backup
,
1757 btrfs_header_generation(info
->tree_root
->node
));
1759 btrfs_set_backup_tree_root_level(root_backup
,
1760 btrfs_header_level(info
->tree_root
->node
));
1762 btrfs_set_backup_chunk_root(root_backup
, info
->chunk_root
->node
->start
);
1763 btrfs_set_backup_chunk_root_gen(root_backup
,
1764 btrfs_header_generation(info
->chunk_root
->node
));
1765 btrfs_set_backup_chunk_root_level(root_backup
,
1766 btrfs_header_level(info
->chunk_root
->node
));
1768 btrfs_set_backup_extent_root(root_backup
, info
->extent_root
->node
->start
);
1769 btrfs_set_backup_extent_root_gen(root_backup
,
1770 btrfs_header_generation(info
->extent_root
->node
));
1771 btrfs_set_backup_extent_root_level(root_backup
,
1772 btrfs_header_level(info
->extent_root
->node
));
1775 * we might commit during log recovery, which happens before we set
1776 * the fs_root. Make sure it is valid before we fill it in.
1778 if (info
->fs_root
&& info
->fs_root
->node
) {
1779 btrfs_set_backup_fs_root(root_backup
,
1780 info
->fs_root
->node
->start
);
1781 btrfs_set_backup_fs_root_gen(root_backup
,
1782 btrfs_header_generation(info
->fs_root
->node
));
1783 btrfs_set_backup_fs_root_level(root_backup
,
1784 btrfs_header_level(info
->fs_root
->node
));
1787 btrfs_set_backup_dev_root(root_backup
, info
->dev_root
->node
->start
);
1788 btrfs_set_backup_dev_root_gen(root_backup
,
1789 btrfs_header_generation(info
->dev_root
->node
));
1790 btrfs_set_backup_dev_root_level(root_backup
,
1791 btrfs_header_level(info
->dev_root
->node
));
1793 btrfs_set_backup_csum_root(root_backup
, info
->csum_root
->node
->start
);
1794 btrfs_set_backup_csum_root_gen(root_backup
,
1795 btrfs_header_generation(info
->csum_root
->node
));
1796 btrfs_set_backup_csum_root_level(root_backup
,
1797 btrfs_header_level(info
->csum_root
->node
));
1799 btrfs_set_backup_total_bytes(root_backup
,
1800 btrfs_super_total_bytes(info
->super_copy
));
1801 btrfs_set_backup_bytes_used(root_backup
,
1802 btrfs_super_bytes_used(info
->super_copy
));
1803 btrfs_set_backup_num_devices(root_backup
,
1804 btrfs_super_num_devices(info
->super_copy
));
1807 * if we don't copy this out to the super_copy, it won't get remembered
1808 * for the next commit
1810 memcpy(&info
->super_copy
->super_roots
,
1811 &info
->super_for_commit
->super_roots
,
1812 sizeof(*root_backup
) * BTRFS_NUM_BACKUP_ROOTS
);
1816 * this copies info out of the root backup array and back into
1817 * the in-memory super block. It is meant to help iterate through
1818 * the array, so you send it the number of backups you've already
1819 * tried and the last backup index you used.
1821 * this returns -1 when it has tried all the backups
1823 static noinline
int next_root_backup(struct btrfs_fs_info
*info
,
1824 struct btrfs_super_block
*super
,
1825 int *num_backups_tried
, int *backup_index
)
1827 struct btrfs_root_backup
*root_backup
;
1828 int newest
= *backup_index
;
1830 if (*num_backups_tried
== 0) {
1831 u64 gen
= btrfs_super_generation(super
);
1833 newest
= find_newest_super_backup(info
, gen
);
1837 *backup_index
= newest
;
1838 *num_backups_tried
= 1;
1839 } else if (*num_backups_tried
== BTRFS_NUM_BACKUP_ROOTS
) {
1840 /* we've tried all the backups, all done */
1843 /* jump to the next oldest backup */
1844 newest
= (*backup_index
+ BTRFS_NUM_BACKUP_ROOTS
- 1) %
1845 BTRFS_NUM_BACKUP_ROOTS
;
1846 *backup_index
= newest
;
1847 *num_backups_tried
+= 1;
1849 root_backup
= super
->super_roots
+ newest
;
1851 btrfs_set_super_generation(super
,
1852 btrfs_backup_tree_root_gen(root_backup
));
1853 btrfs_set_super_root(super
, btrfs_backup_tree_root(root_backup
));
1854 btrfs_set_super_root_level(super
,
1855 btrfs_backup_tree_root_level(root_backup
));
1856 btrfs_set_super_bytes_used(super
, btrfs_backup_bytes_used(root_backup
));
1859 * fixme: the total bytes and num_devices need to match or we should
1862 btrfs_set_super_total_bytes(super
, btrfs_backup_total_bytes(root_backup
));
1863 btrfs_set_super_num_devices(super
, btrfs_backup_num_devices(root_backup
));
1867 /* helper to cleanup tree roots */
1868 static void free_root_pointers(struct btrfs_fs_info
*info
, int chunk_root
)
1870 free_extent_buffer(info
->tree_root
->node
);
1871 free_extent_buffer(info
->tree_root
->commit_root
);
1872 free_extent_buffer(info
->dev_root
->node
);
1873 free_extent_buffer(info
->dev_root
->commit_root
);
1874 free_extent_buffer(info
->extent_root
->node
);
1875 free_extent_buffer(info
->extent_root
->commit_root
);
1876 free_extent_buffer(info
->csum_root
->node
);
1877 free_extent_buffer(info
->csum_root
->commit_root
);
1879 info
->tree_root
->node
= NULL
;
1880 info
->tree_root
->commit_root
= NULL
;
1881 info
->dev_root
->node
= NULL
;
1882 info
->dev_root
->commit_root
= NULL
;
1883 info
->extent_root
->node
= NULL
;
1884 info
->extent_root
->commit_root
= NULL
;
1885 info
->csum_root
->node
= NULL
;
1886 info
->csum_root
->commit_root
= NULL
;
1889 free_extent_buffer(info
->chunk_root
->node
);
1890 free_extent_buffer(info
->chunk_root
->commit_root
);
1891 info
->chunk_root
->node
= NULL
;
1892 info
->chunk_root
->commit_root
= NULL
;
1897 int open_ctree(struct super_block
*sb
,
1898 struct btrfs_fs_devices
*fs_devices
,
1908 struct btrfs_key location
;
1909 struct buffer_head
*bh
;
1910 struct btrfs_super_block
*disk_super
;
1911 struct btrfs_fs_info
*fs_info
= btrfs_sb(sb
);
1912 struct btrfs_root
*tree_root
;
1913 struct btrfs_root
*extent_root
;
1914 struct btrfs_root
*csum_root
;
1915 struct btrfs_root
*chunk_root
;
1916 struct btrfs_root
*dev_root
;
1917 struct btrfs_root
*log_tree_root
;
1920 int num_backups_tried
= 0;
1921 int backup_index
= 0;
1923 tree_root
= fs_info
->tree_root
= btrfs_alloc_root(fs_info
);
1924 extent_root
= fs_info
->extent_root
= btrfs_alloc_root(fs_info
);
1925 csum_root
= fs_info
->csum_root
= btrfs_alloc_root(fs_info
);
1926 chunk_root
= fs_info
->chunk_root
= btrfs_alloc_root(fs_info
);
1927 dev_root
= fs_info
->dev_root
= btrfs_alloc_root(fs_info
);
1929 if (!tree_root
|| !extent_root
|| !csum_root
||
1930 !chunk_root
|| !dev_root
) {
1935 ret
= init_srcu_struct(&fs_info
->subvol_srcu
);
1941 ret
= setup_bdi(fs_info
, &fs_info
->bdi
);
1947 fs_info
->btree_inode
= new_inode(sb
);
1948 if (!fs_info
->btree_inode
) {
1953 mapping_set_gfp_mask(fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1955 INIT_RADIX_TREE(&fs_info
->fs_roots_radix
, GFP_ATOMIC
);
1956 INIT_LIST_HEAD(&fs_info
->trans_list
);
1957 INIT_LIST_HEAD(&fs_info
->dead_roots
);
1958 INIT_LIST_HEAD(&fs_info
->delayed_iputs
);
1959 INIT_LIST_HEAD(&fs_info
->hashers
);
1960 INIT_LIST_HEAD(&fs_info
->delalloc_inodes
);
1961 INIT_LIST_HEAD(&fs_info
->ordered_operations
);
1962 INIT_LIST_HEAD(&fs_info
->caching_block_groups
);
1963 spin_lock_init(&fs_info
->delalloc_lock
);
1964 spin_lock_init(&fs_info
->trans_lock
);
1965 spin_lock_init(&fs_info
->ref_cache_lock
);
1966 spin_lock_init(&fs_info
->fs_roots_radix_lock
);
1967 spin_lock_init(&fs_info
->delayed_iput_lock
);
1968 spin_lock_init(&fs_info
->defrag_inodes_lock
);
1969 spin_lock_init(&fs_info
->free_chunk_lock
);
1970 mutex_init(&fs_info
->reloc_mutex
);
1972 init_completion(&fs_info
->kobj_unregister
);
1973 INIT_LIST_HEAD(&fs_info
->dirty_cowonly_roots
);
1974 INIT_LIST_HEAD(&fs_info
->space_info
);
1975 btrfs_mapping_init(&fs_info
->mapping_tree
);
1976 btrfs_init_block_rsv(&fs_info
->global_block_rsv
);
1977 btrfs_init_block_rsv(&fs_info
->delalloc_block_rsv
);
1978 btrfs_init_block_rsv(&fs_info
->trans_block_rsv
);
1979 btrfs_init_block_rsv(&fs_info
->chunk_block_rsv
);
1980 btrfs_init_block_rsv(&fs_info
->empty_block_rsv
);
1981 btrfs_init_block_rsv(&fs_info
->delayed_block_rsv
);
1982 atomic_set(&fs_info
->nr_async_submits
, 0);
1983 atomic_set(&fs_info
->async_delalloc_pages
, 0);
1984 atomic_set(&fs_info
->async_submit_draining
, 0);
1985 atomic_set(&fs_info
->nr_async_bios
, 0);
1986 atomic_set(&fs_info
->defrag_running
, 0);
1988 fs_info
->max_inline
= 8192 * 1024;
1989 fs_info
->metadata_ratio
= 0;
1990 fs_info
->defrag_inodes
= RB_ROOT
;
1991 fs_info
->trans_no_join
= 0;
1992 fs_info
->free_chunk_space
= 0;
1994 /* readahead state */
1995 INIT_RADIX_TREE(&fs_info
->reada_tree
, GFP_NOFS
& ~__GFP_WAIT
);
1996 spin_lock_init(&fs_info
->reada_lock
);
1998 fs_info
->thread_pool_size
= min_t(unsigned long,
1999 num_online_cpus() + 2, 8);
2001 INIT_LIST_HEAD(&fs_info
->ordered_extents
);
2002 spin_lock_init(&fs_info
->ordered_extent_lock
);
2003 fs_info
->delayed_root
= kmalloc(sizeof(struct btrfs_delayed_root
),
2005 if (!fs_info
->delayed_root
) {
2009 btrfs_init_delayed_root(fs_info
->delayed_root
);
2011 mutex_init(&fs_info
->scrub_lock
);
2012 atomic_set(&fs_info
->scrubs_running
, 0);
2013 atomic_set(&fs_info
->scrub_pause_req
, 0);
2014 atomic_set(&fs_info
->scrubs_paused
, 0);
2015 atomic_set(&fs_info
->scrub_cancel_req
, 0);
2016 init_waitqueue_head(&fs_info
->scrub_pause_wait
);
2017 init_rwsem(&fs_info
->scrub_super_lock
);
2018 fs_info
->scrub_workers_refcnt
= 0;
2019 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2020 fs_info
->check_integrity_print_mask
= 0;
2023 spin_lock_init(&fs_info
->balance_lock
);
2024 mutex_init(&fs_info
->balance_mutex
);
2025 atomic_set(&fs_info
->balance_running
, 0);
2026 atomic_set(&fs_info
->balance_pause_req
, 0);
2027 atomic_set(&fs_info
->balance_cancel_req
, 0);
2028 fs_info
->balance_ctl
= NULL
;
2029 init_waitqueue_head(&fs_info
->balance_wait_q
);
2031 sb
->s_blocksize
= 4096;
2032 sb
->s_blocksize_bits
= blksize_bits(4096);
2033 sb
->s_bdi
= &fs_info
->bdi
;
2035 fs_info
->btree_inode
->i_ino
= BTRFS_BTREE_INODE_OBJECTID
;
2036 set_nlink(fs_info
->btree_inode
, 1);
2038 * we set the i_size on the btree inode to the max possible int.
2039 * the real end of the address space is determined by all of
2040 * the devices in the system
2042 fs_info
->btree_inode
->i_size
= OFFSET_MAX
;
2043 fs_info
->btree_inode
->i_mapping
->a_ops
= &btree_aops
;
2044 fs_info
->btree_inode
->i_mapping
->backing_dev_info
= &fs_info
->bdi
;
2046 RB_CLEAR_NODE(&BTRFS_I(fs_info
->btree_inode
)->rb_node
);
2047 extent_io_tree_init(&BTRFS_I(fs_info
->btree_inode
)->io_tree
,
2048 fs_info
->btree_inode
->i_mapping
);
2049 extent_map_tree_init(&BTRFS_I(fs_info
->btree_inode
)->extent_tree
);
2051 BTRFS_I(fs_info
->btree_inode
)->io_tree
.ops
= &btree_extent_io_ops
;
2053 BTRFS_I(fs_info
->btree_inode
)->root
= tree_root
;
2054 memset(&BTRFS_I(fs_info
->btree_inode
)->location
, 0,
2055 sizeof(struct btrfs_key
));
2056 BTRFS_I(fs_info
->btree_inode
)->dummy_inode
= 1;
2057 insert_inode_hash(fs_info
->btree_inode
);
2059 spin_lock_init(&fs_info
->block_group_cache_lock
);
2060 fs_info
->block_group_cache_tree
= RB_ROOT
;
2062 extent_io_tree_init(&fs_info
->freed_extents
[0],
2063 fs_info
->btree_inode
->i_mapping
);
2064 extent_io_tree_init(&fs_info
->freed_extents
[1],
2065 fs_info
->btree_inode
->i_mapping
);
2066 fs_info
->pinned_extents
= &fs_info
->freed_extents
[0];
2067 fs_info
->do_barriers
= 1;
2070 mutex_init(&fs_info
->ordered_operations_mutex
);
2071 mutex_init(&fs_info
->tree_log_mutex
);
2072 mutex_init(&fs_info
->chunk_mutex
);
2073 mutex_init(&fs_info
->transaction_kthread_mutex
);
2074 mutex_init(&fs_info
->cleaner_mutex
);
2075 mutex_init(&fs_info
->volume_mutex
);
2076 init_rwsem(&fs_info
->extent_commit_sem
);
2077 init_rwsem(&fs_info
->cleanup_work_sem
);
2078 init_rwsem(&fs_info
->subvol_sem
);
2080 btrfs_init_free_cluster(&fs_info
->meta_alloc_cluster
);
2081 btrfs_init_free_cluster(&fs_info
->data_alloc_cluster
);
2083 init_waitqueue_head(&fs_info
->transaction_throttle
);
2084 init_waitqueue_head(&fs_info
->transaction_wait
);
2085 init_waitqueue_head(&fs_info
->transaction_blocked_wait
);
2086 init_waitqueue_head(&fs_info
->async_submit_wait
);
2088 __setup_root(4096, 4096, 4096, 4096, tree_root
,
2089 fs_info
, BTRFS_ROOT_TREE_OBJECTID
);
2091 bh
= btrfs_read_dev_super(fs_devices
->latest_bdev
);
2097 memcpy(fs_info
->super_copy
, bh
->b_data
, sizeof(*fs_info
->super_copy
));
2098 memcpy(fs_info
->super_for_commit
, fs_info
->super_copy
,
2099 sizeof(*fs_info
->super_for_commit
));
2102 memcpy(fs_info
->fsid
, fs_info
->super_copy
->fsid
, BTRFS_FSID_SIZE
);
2104 disk_super
= fs_info
->super_copy
;
2105 if (!btrfs_super_root(disk_super
))
2108 /* check FS state, whether FS is broken. */
2109 fs_info
->fs_state
|= btrfs_super_flags(disk_super
);
2111 btrfs_check_super_valid(fs_info
, sb
->s_flags
& MS_RDONLY
);
2114 * run through our array of backup supers and setup
2115 * our ring pointer to the oldest one
2117 generation
= btrfs_super_generation(disk_super
);
2118 find_oldest_super_backup(fs_info
, generation
);
2121 * In the long term, we'll store the compression type in the super
2122 * block, and it'll be used for per file compression control.
2124 fs_info
->compress_type
= BTRFS_COMPRESS_ZLIB
;
2126 ret
= btrfs_parse_options(tree_root
, options
);
2132 features
= btrfs_super_incompat_flags(disk_super
) &
2133 ~BTRFS_FEATURE_INCOMPAT_SUPP
;
2135 printk(KERN_ERR
"BTRFS: couldn't mount because of "
2136 "unsupported optional features (%Lx).\n",
2137 (unsigned long long)features
);
2142 features
= btrfs_super_incompat_flags(disk_super
);
2143 features
|= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF
;
2144 if (tree_root
->fs_info
->compress_type
& BTRFS_COMPRESS_LZO
)
2145 features
|= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO
;
2146 btrfs_set_super_incompat_flags(disk_super
, features
);
2148 features
= btrfs_super_compat_ro_flags(disk_super
) &
2149 ~BTRFS_FEATURE_COMPAT_RO_SUPP
;
2150 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
2151 printk(KERN_ERR
"BTRFS: couldn't mount RDWR because of "
2152 "unsupported option features (%Lx).\n",
2153 (unsigned long long)features
);
2158 btrfs_init_workers(&fs_info
->generic_worker
,
2159 "genwork", 1, NULL
);
2161 btrfs_init_workers(&fs_info
->workers
, "worker",
2162 fs_info
->thread_pool_size
,
2163 &fs_info
->generic_worker
);
2165 btrfs_init_workers(&fs_info
->delalloc_workers
, "delalloc",
2166 fs_info
->thread_pool_size
,
2167 &fs_info
->generic_worker
);
2169 btrfs_init_workers(&fs_info
->submit_workers
, "submit",
2170 min_t(u64
, fs_devices
->num_devices
,
2171 fs_info
->thread_pool_size
),
2172 &fs_info
->generic_worker
);
2174 btrfs_init_workers(&fs_info
->caching_workers
, "cache",
2175 2, &fs_info
->generic_worker
);
2177 /* a higher idle thresh on the submit workers makes it much more
2178 * likely that bios will be send down in a sane order to the
2181 fs_info
->submit_workers
.idle_thresh
= 64;
2183 fs_info
->workers
.idle_thresh
= 16;
2184 fs_info
->workers
.ordered
= 1;
2186 fs_info
->delalloc_workers
.idle_thresh
= 2;
2187 fs_info
->delalloc_workers
.ordered
= 1;
2189 btrfs_init_workers(&fs_info
->fixup_workers
, "fixup", 1,
2190 &fs_info
->generic_worker
);
2191 btrfs_init_workers(&fs_info
->endio_workers
, "endio",
2192 fs_info
->thread_pool_size
,
2193 &fs_info
->generic_worker
);
2194 btrfs_init_workers(&fs_info
->endio_meta_workers
, "endio-meta",
2195 fs_info
->thread_pool_size
,
2196 &fs_info
->generic_worker
);
2197 btrfs_init_workers(&fs_info
->endio_meta_write_workers
,
2198 "endio-meta-write", fs_info
->thread_pool_size
,
2199 &fs_info
->generic_worker
);
2200 btrfs_init_workers(&fs_info
->endio_write_workers
, "endio-write",
2201 fs_info
->thread_pool_size
,
2202 &fs_info
->generic_worker
);
2203 btrfs_init_workers(&fs_info
->endio_freespace_worker
, "freespace-write",
2204 1, &fs_info
->generic_worker
);
2205 btrfs_init_workers(&fs_info
->delayed_workers
, "delayed-meta",
2206 fs_info
->thread_pool_size
,
2207 &fs_info
->generic_worker
);
2208 btrfs_init_workers(&fs_info
->readahead_workers
, "readahead",
2209 fs_info
->thread_pool_size
,
2210 &fs_info
->generic_worker
);
2213 * endios are largely parallel and should have a very
2216 fs_info
->endio_workers
.idle_thresh
= 4;
2217 fs_info
->endio_meta_workers
.idle_thresh
= 4;
2219 fs_info
->endio_write_workers
.idle_thresh
= 2;
2220 fs_info
->endio_meta_write_workers
.idle_thresh
= 2;
2221 fs_info
->readahead_workers
.idle_thresh
= 2;
2224 * btrfs_start_workers can really only fail because of ENOMEM so just
2225 * return -ENOMEM if any of these fail.
2227 ret
= btrfs_start_workers(&fs_info
->workers
);
2228 ret
|= btrfs_start_workers(&fs_info
->generic_worker
);
2229 ret
|= btrfs_start_workers(&fs_info
->submit_workers
);
2230 ret
|= btrfs_start_workers(&fs_info
->delalloc_workers
);
2231 ret
|= btrfs_start_workers(&fs_info
->fixup_workers
);
2232 ret
|= btrfs_start_workers(&fs_info
->endio_workers
);
2233 ret
|= btrfs_start_workers(&fs_info
->endio_meta_workers
);
2234 ret
|= btrfs_start_workers(&fs_info
->endio_meta_write_workers
);
2235 ret
|= btrfs_start_workers(&fs_info
->endio_write_workers
);
2236 ret
|= btrfs_start_workers(&fs_info
->endio_freespace_worker
);
2237 ret
|= btrfs_start_workers(&fs_info
->delayed_workers
);
2238 ret
|= btrfs_start_workers(&fs_info
->caching_workers
);
2239 ret
|= btrfs_start_workers(&fs_info
->readahead_workers
);
2242 goto fail_sb_buffer
;
2245 fs_info
->bdi
.ra_pages
*= btrfs_super_num_devices(disk_super
);
2246 fs_info
->bdi
.ra_pages
= max(fs_info
->bdi
.ra_pages
,
2247 4 * 1024 * 1024 / PAGE_CACHE_SIZE
);
2249 nodesize
= btrfs_super_nodesize(disk_super
);
2250 leafsize
= btrfs_super_leafsize(disk_super
);
2251 sectorsize
= btrfs_super_sectorsize(disk_super
);
2252 stripesize
= btrfs_super_stripesize(disk_super
);
2253 tree_root
->nodesize
= nodesize
;
2254 tree_root
->leafsize
= leafsize
;
2255 tree_root
->sectorsize
= sectorsize
;
2256 tree_root
->stripesize
= stripesize
;
2258 sb
->s_blocksize
= sectorsize
;
2259 sb
->s_blocksize_bits
= blksize_bits(sectorsize
);
2261 if (strncmp((char *)(&disk_super
->magic
), BTRFS_MAGIC
,
2262 sizeof(disk_super
->magic
))) {
2263 printk(KERN_INFO
"btrfs: valid FS not found on %s\n", sb
->s_id
);
2264 goto fail_sb_buffer
;
2267 if (sectorsize
< PAGE_SIZE
) {
2268 printk(KERN_WARNING
"btrfs: Incompatible sector size "
2269 "found on %s\n", sb
->s_id
);
2270 goto fail_sb_buffer
;
2273 mutex_lock(&fs_info
->chunk_mutex
);
2274 ret
= btrfs_read_sys_array(tree_root
);
2275 mutex_unlock(&fs_info
->chunk_mutex
);
2277 printk(KERN_WARNING
"btrfs: failed to read the system "
2278 "array on %s\n", sb
->s_id
);
2279 goto fail_sb_buffer
;
2282 blocksize
= btrfs_level_size(tree_root
,
2283 btrfs_super_chunk_root_level(disk_super
));
2284 generation
= btrfs_super_chunk_root_generation(disk_super
);
2286 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
2287 chunk_root
, fs_info
, BTRFS_CHUNK_TREE_OBJECTID
);
2289 chunk_root
->node
= read_tree_block(chunk_root
,
2290 btrfs_super_chunk_root(disk_super
),
2291 blocksize
, generation
);
2292 BUG_ON(!chunk_root
->node
);
2293 if (!test_bit(EXTENT_BUFFER_UPTODATE
, &chunk_root
->node
->bflags
)) {
2294 printk(KERN_WARNING
"btrfs: failed to read chunk root on %s\n",
2296 goto fail_tree_roots
;
2298 btrfs_set_root_node(&chunk_root
->root_item
, chunk_root
->node
);
2299 chunk_root
->commit_root
= btrfs_root_node(chunk_root
);
2301 read_extent_buffer(chunk_root
->node
, fs_info
->chunk_tree_uuid
,
2302 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root
->node
),
2305 ret
= btrfs_read_chunk_tree(chunk_root
);
2307 printk(KERN_WARNING
"btrfs: failed to read chunk tree on %s\n",
2309 goto fail_tree_roots
;
2312 btrfs_close_extra_devices(fs_devices
);
2314 if (!fs_devices
->latest_bdev
) {
2315 printk(KERN_CRIT
"btrfs: failed to read devices on %s\n",
2317 goto fail_tree_roots
;
2321 blocksize
= btrfs_level_size(tree_root
,
2322 btrfs_super_root_level(disk_super
));
2323 generation
= btrfs_super_generation(disk_super
);
2325 tree_root
->node
= read_tree_block(tree_root
,
2326 btrfs_super_root(disk_super
),
2327 blocksize
, generation
);
2328 if (!tree_root
->node
||
2329 !test_bit(EXTENT_BUFFER_UPTODATE
, &tree_root
->node
->bflags
)) {
2330 printk(KERN_WARNING
"btrfs: failed to read tree root on %s\n",
2333 goto recovery_tree_root
;
2336 btrfs_set_root_node(&tree_root
->root_item
, tree_root
->node
);
2337 tree_root
->commit_root
= btrfs_root_node(tree_root
);
2339 ret
= find_and_setup_root(tree_root
, fs_info
,
2340 BTRFS_EXTENT_TREE_OBJECTID
, extent_root
);
2342 goto recovery_tree_root
;
2343 extent_root
->track_dirty
= 1;
2345 ret
= find_and_setup_root(tree_root
, fs_info
,
2346 BTRFS_DEV_TREE_OBJECTID
, dev_root
);
2348 goto recovery_tree_root
;
2349 dev_root
->track_dirty
= 1;
2351 ret
= find_and_setup_root(tree_root
, fs_info
,
2352 BTRFS_CSUM_TREE_OBJECTID
, csum_root
);
2354 goto recovery_tree_root
;
2356 csum_root
->track_dirty
= 1;
2358 fs_info
->generation
= generation
;
2359 fs_info
->last_trans_committed
= generation
;
2361 ret
= btrfs_init_space_info(fs_info
);
2363 printk(KERN_ERR
"Failed to initial space info: %d\n", ret
);
2364 goto fail_block_groups
;
2367 ret
= btrfs_read_block_groups(extent_root
);
2369 printk(KERN_ERR
"Failed to read block groups: %d\n", ret
);
2370 goto fail_block_groups
;
2373 fs_info
->cleaner_kthread
= kthread_run(cleaner_kthread
, tree_root
,
2375 if (IS_ERR(fs_info
->cleaner_kthread
))
2376 goto fail_block_groups
;
2378 fs_info
->transaction_kthread
= kthread_run(transaction_kthread
,
2380 "btrfs-transaction");
2381 if (IS_ERR(fs_info
->transaction_kthread
))
2384 if (!btrfs_test_opt(tree_root
, SSD
) &&
2385 !btrfs_test_opt(tree_root
, NOSSD
) &&
2386 !fs_info
->fs_devices
->rotating
) {
2387 printk(KERN_INFO
"Btrfs detected SSD devices, enabling SSD "
2389 btrfs_set_opt(fs_info
->mount_opt
, SSD
);
2392 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2393 if (btrfs_test_opt(tree_root
, CHECK_INTEGRITY
)) {
2394 ret
= btrfsic_mount(tree_root
, fs_devices
,
2395 btrfs_test_opt(tree_root
,
2396 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA
) ?
2398 fs_info
->check_integrity_print_mask
);
2400 printk(KERN_WARNING
"btrfs: failed to initialize"
2401 " integrity check module %s\n", sb
->s_id
);
2405 /* do not make disk changes in broken FS */
2406 if (btrfs_super_log_root(disk_super
) != 0 &&
2407 !(fs_info
->fs_state
& BTRFS_SUPER_FLAG_ERROR
)) {
2408 u64 bytenr
= btrfs_super_log_root(disk_super
);
2410 if (fs_devices
->rw_devices
== 0) {
2411 printk(KERN_WARNING
"Btrfs log replay required "
2414 goto fail_trans_kthread
;
2417 btrfs_level_size(tree_root
,
2418 btrfs_super_log_root_level(disk_super
));
2420 log_tree_root
= btrfs_alloc_root(fs_info
);
2421 if (!log_tree_root
) {
2423 goto fail_trans_kthread
;
2426 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
2427 log_tree_root
, fs_info
, BTRFS_TREE_LOG_OBJECTID
);
2429 log_tree_root
->node
= read_tree_block(tree_root
, bytenr
,
2432 ret
= btrfs_recover_log_trees(log_tree_root
);
2435 if (sb
->s_flags
& MS_RDONLY
) {
2436 ret
= btrfs_commit_super(tree_root
);
2441 ret
= btrfs_find_orphan_roots(tree_root
);
2444 if (!(sb
->s_flags
& MS_RDONLY
)) {
2445 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",
2870 list_for_each_entry_rcu(dev
, head
, dev_list
) {
2873 if (!dev
->in_fs_metadata
|| !dev
->writeable
)
2876 ret
= write_dev_supers(dev
, sb
, do_barriers
, 1, max_mirrors
);
2880 mutex_unlock(&root
->fs_info
->fs_devices
->device_list_mutex
);
2881 if (total_errors
> max_errors
) {
2882 printk(KERN_ERR
"btrfs: %d errors while writing supers\n",
2889 int write_ctree_super(struct btrfs_trans_handle
*trans
,
2890 struct btrfs_root
*root
, int max_mirrors
)
2894 ret
= write_all_supers(root
, max_mirrors
);
2898 /* Kill all outstanding I/O */
2899 void btrfs_abort_devices(struct btrfs_root
*root
)
2901 struct list_head
*head
;
2902 struct btrfs_device
*dev
;
2903 mutex_lock(&root
->fs_info
->fs_devices
->device_list_mutex
);
2904 head
= &root
->fs_info
->fs_devices
->devices
;
2905 list_for_each_entry_rcu(dev
, head
, dev_list
) {
2906 blk_abort_queue(dev
->bdev
->bd_disk
->queue
);
2908 mutex_unlock(&root
->fs_info
->fs_devices
->device_list_mutex
);
2911 void btrfs_free_fs_root(struct btrfs_fs_info
*fs_info
, struct btrfs_root
*root
)
2913 spin_lock(&fs_info
->fs_roots_radix_lock
);
2914 radix_tree_delete(&fs_info
->fs_roots_radix
,
2915 (unsigned long)root
->root_key
.objectid
);
2916 spin_unlock(&fs_info
->fs_roots_radix_lock
);
2918 if (btrfs_root_refs(&root
->root_item
) == 0)
2919 synchronize_srcu(&fs_info
->subvol_srcu
);
2921 __btrfs_remove_free_space_cache(root
->free_ino_pinned
);
2922 __btrfs_remove_free_space_cache(root
->free_ino_ctl
);
2926 static void free_fs_root(struct btrfs_root
*root
)
2928 iput(root
->cache_inode
);
2929 WARN_ON(!RB_EMPTY_ROOT(&root
->inode_tree
));
2931 free_anon_bdev(root
->anon_dev
);
2932 free_extent_buffer(root
->node
);
2933 free_extent_buffer(root
->commit_root
);
2934 kfree(root
->free_ino_ctl
);
2935 kfree(root
->free_ino_pinned
);
2940 static void del_fs_roots(struct btrfs_fs_info
*fs_info
)
2943 struct btrfs_root
*gang
[8];
2946 while (!list_empty(&fs_info
->dead_roots
)) {
2947 gang
[0] = list_entry(fs_info
->dead_roots
.next
,
2948 struct btrfs_root
, root_list
);
2949 list_del(&gang
[0]->root_list
);
2951 if (gang
[0]->in_radix
) {
2952 btrfs_free_fs_root(fs_info
, gang
[0]);
2954 free_extent_buffer(gang
[0]->node
);
2955 free_extent_buffer(gang
[0]->commit_root
);
2961 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
2966 for (i
= 0; i
< ret
; i
++)
2967 btrfs_free_fs_root(fs_info
, gang
[i
]);
2971 int btrfs_cleanup_fs_roots(struct btrfs_fs_info
*fs_info
)
2973 u64 root_objectid
= 0;
2974 struct btrfs_root
*gang
[8];
2979 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
2980 (void **)gang
, root_objectid
,
2985 root_objectid
= gang
[ret
- 1]->root_key
.objectid
+ 1;
2986 for (i
= 0; i
< ret
; i
++) {
2989 root_objectid
= gang
[i
]->root_key
.objectid
;
2990 err
= btrfs_orphan_cleanup(gang
[i
]);
2999 int btrfs_commit_super(struct btrfs_root
*root
)
3001 struct btrfs_trans_handle
*trans
;
3004 mutex_lock(&root
->fs_info
->cleaner_mutex
);
3005 btrfs_run_delayed_iputs(root
);
3006 btrfs_clean_old_snapshots(root
);
3007 mutex_unlock(&root
->fs_info
->cleaner_mutex
);
3009 /* wait until ongoing cleanup work done */
3010 down_write(&root
->fs_info
->cleanup_work_sem
);
3011 up_write(&root
->fs_info
->cleanup_work_sem
);
3013 trans
= btrfs_join_transaction(root
);
3015 return PTR_ERR(trans
);
3016 ret
= btrfs_commit_transaction(trans
, root
);
3018 /* run commit again to drop the original snapshot */
3019 trans
= btrfs_join_transaction(root
);
3021 return PTR_ERR(trans
);
3022 btrfs_commit_transaction(trans
, root
);
3023 ret
= btrfs_write_and_wait_transaction(NULL
, root
);
3026 ret
= write_ctree_super(NULL
, root
, 0);
3030 int close_ctree(struct btrfs_root
*root
)
3032 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
3035 fs_info
->closing
= 1;
3038 /* pause restriper - we want to resume on mount */
3039 btrfs_pause_balance(root
->fs_info
);
3041 btrfs_scrub_cancel(root
);
3043 /* wait for any defraggers to finish */
3044 wait_event(fs_info
->transaction_wait
,
3045 (atomic_read(&fs_info
->defrag_running
) == 0));
3047 /* clear out the rbtree of defraggable inodes */
3048 btrfs_run_defrag_inodes(fs_info
);
3051 * Here come 2 situations when btrfs is broken to flip readonly:
3053 * 1. when btrfs flips readonly somewhere else before
3054 * btrfs_commit_super, sb->s_flags has MS_RDONLY flag,
3055 * and btrfs will skip to write sb directly to keep
3056 * ERROR state on disk.
3058 * 2. when btrfs flips readonly just in btrfs_commit_super,
3059 * and in such case, btrfs cannot write sb via btrfs_commit_super,
3060 * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag,
3061 * btrfs will cleanup all FS resources first and write sb then.
3063 if (!(fs_info
->sb
->s_flags
& MS_RDONLY
)) {
3064 ret
= btrfs_commit_super(root
);
3066 printk(KERN_ERR
"btrfs: commit super ret %d\n", ret
);
3069 if (fs_info
->fs_state
& BTRFS_SUPER_FLAG_ERROR
) {
3070 ret
= btrfs_error_commit_super(root
);
3072 printk(KERN_ERR
"btrfs: commit super ret %d\n", ret
);
3075 btrfs_put_block_group_cache(fs_info
);
3077 kthread_stop(fs_info
->transaction_kthread
);
3078 kthread_stop(fs_info
->cleaner_kthread
);
3080 fs_info
->closing
= 2;
3083 if (fs_info
->delalloc_bytes
) {
3084 printk(KERN_INFO
"btrfs: at unmount delalloc count %llu\n",
3085 (unsigned long long)fs_info
->delalloc_bytes
);
3087 if (fs_info
->total_ref_cache_size
) {
3088 printk(KERN_INFO
"btrfs: at umount reference cache size %llu\n",
3089 (unsigned long long)fs_info
->total_ref_cache_size
);
3092 free_extent_buffer(fs_info
->extent_root
->node
);
3093 free_extent_buffer(fs_info
->extent_root
->commit_root
);
3094 free_extent_buffer(fs_info
->tree_root
->node
);
3095 free_extent_buffer(fs_info
->tree_root
->commit_root
);
3096 free_extent_buffer(fs_info
->chunk_root
->node
);
3097 free_extent_buffer(fs_info
->chunk_root
->commit_root
);
3098 free_extent_buffer(fs_info
->dev_root
->node
);
3099 free_extent_buffer(fs_info
->dev_root
->commit_root
);
3100 free_extent_buffer(fs_info
->csum_root
->node
);
3101 free_extent_buffer(fs_info
->csum_root
->commit_root
);
3103 btrfs_free_block_groups(fs_info
);
3105 del_fs_roots(fs_info
);
3107 iput(fs_info
->btree_inode
);
3109 btrfs_stop_workers(&fs_info
->generic_worker
);
3110 btrfs_stop_workers(&fs_info
->fixup_workers
);
3111 btrfs_stop_workers(&fs_info
->delalloc_workers
);
3112 btrfs_stop_workers(&fs_info
->workers
);
3113 btrfs_stop_workers(&fs_info
->endio_workers
);
3114 btrfs_stop_workers(&fs_info
->endio_meta_workers
);
3115 btrfs_stop_workers(&fs_info
->endio_meta_write_workers
);
3116 btrfs_stop_workers(&fs_info
->endio_write_workers
);
3117 btrfs_stop_workers(&fs_info
->endio_freespace_worker
);
3118 btrfs_stop_workers(&fs_info
->submit_workers
);
3119 btrfs_stop_workers(&fs_info
->delayed_workers
);
3120 btrfs_stop_workers(&fs_info
->caching_workers
);
3121 btrfs_stop_workers(&fs_info
->readahead_workers
);
3123 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3124 if (btrfs_test_opt(root
, CHECK_INTEGRITY
))
3125 btrfsic_unmount(root
, fs_info
->fs_devices
);
3128 btrfs_close_devices(fs_info
->fs_devices
);
3129 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
3131 bdi_destroy(&fs_info
->bdi
);
3132 cleanup_srcu_struct(&fs_info
->subvol_srcu
);
3137 int btrfs_buffer_uptodate(struct extent_buffer
*buf
, u64 parent_transid
)
3140 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
3142 ret
= extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
, buf
,
3147 ret
= verify_parent_transid(&BTRFS_I(btree_inode
)->io_tree
, buf
,
3152 int btrfs_set_buffer_uptodate(struct extent_buffer
*buf
)
3154 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
3155 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
,
3159 void btrfs_mark_buffer_dirty(struct extent_buffer
*buf
)
3161 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
3162 u64 transid
= btrfs_header_generation(buf
);
3163 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
3166 btrfs_assert_tree_locked(buf
);
3167 if (transid
!= root
->fs_info
->generation
) {
3168 printk(KERN_CRIT
"btrfs transid mismatch buffer %llu, "
3169 "found %llu running %llu\n",
3170 (unsigned long long)buf
->start
,
3171 (unsigned long long)transid
,
3172 (unsigned long long)root
->fs_info
->generation
);
3175 was_dirty
= set_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
,
3178 spin_lock(&root
->fs_info
->delalloc_lock
);
3179 root
->fs_info
->dirty_metadata_bytes
+= buf
->len
;
3180 spin_unlock(&root
->fs_info
->delalloc_lock
);
3184 void btrfs_btree_balance_dirty(struct btrfs_root
*root
, unsigned long nr
)
3187 * looks as though older kernels can get into trouble with
3188 * this code, they end up stuck in balance_dirty_pages forever
3191 unsigned long thresh
= 32 * 1024 * 1024;
3193 if (current
->flags
& PF_MEMALLOC
)
3196 btrfs_balance_delayed_items(root
);
3198 num_dirty
= root
->fs_info
->dirty_metadata_bytes
;
3200 if (num_dirty
> thresh
) {
3201 balance_dirty_pages_ratelimited_nr(
3202 root
->fs_info
->btree_inode
->i_mapping
, 1);
3207 void __btrfs_btree_balance_dirty(struct btrfs_root
*root
, unsigned long nr
)
3210 * looks as though older kernels can get into trouble with
3211 * this code, they end up stuck in balance_dirty_pages forever
3214 unsigned long thresh
= 32 * 1024 * 1024;
3216 if (current
->flags
& PF_MEMALLOC
)
3219 num_dirty
= root
->fs_info
->dirty_metadata_bytes
;
3221 if (num_dirty
> thresh
) {
3222 balance_dirty_pages_ratelimited_nr(
3223 root
->fs_info
->btree_inode
->i_mapping
, 1);
3228 int btrfs_read_buffer(struct extent_buffer
*buf
, u64 parent_transid
)
3230 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
3232 ret
= btree_read_extent_buffer_pages(root
, buf
, 0, parent_transid
);
3234 set_bit(EXTENT_BUFFER_UPTODATE
, &buf
->bflags
);
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_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
3244 struct extent_buffer
*eb
;
3246 u64 bytenr
= page_offset(page
);
3248 if (page
->private == EXTENT_PAGE_PRIVATE
)
3251 len
= page
->private >> 2;
3252 eb
= find_extent_buffer(io_tree
, bytenr
, len
);
3256 if (!btrfs_try_tree_write_lock(eb
)) {
3258 btrfs_tree_lock(eb
);
3260 btrfs_set_header_flag(eb
, BTRFS_HEADER_FLAG_WRITTEN
);
3262 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY
, &eb
->bflags
)) {
3263 spin_lock(&root
->fs_info
->delalloc_lock
);
3264 if (root
->fs_info
->dirty_metadata_bytes
>= eb
->len
)
3265 root
->fs_info
->dirty_metadata_bytes
-= eb
->len
;
3268 spin_unlock(&root
->fs_info
->delalloc_lock
);
3271 btrfs_tree_unlock(eb
);
3272 free_extent_buffer(eb
);
3274 if (!trylock_page(page
)) {
3281 static void btrfs_check_super_valid(struct btrfs_fs_info
*fs_info
,
3287 if (fs_info
->fs_state
& BTRFS_SUPER_FLAG_ERROR
)
3288 printk(KERN_WARNING
"warning: mount fs with errors, "
3289 "running btrfsck is recommended\n");
3292 int btrfs_error_commit_super(struct btrfs_root
*root
)
3296 mutex_lock(&root
->fs_info
->cleaner_mutex
);
3297 btrfs_run_delayed_iputs(root
);
3298 mutex_unlock(&root
->fs_info
->cleaner_mutex
);
3300 down_write(&root
->fs_info
->cleanup_work_sem
);
3301 up_write(&root
->fs_info
->cleanup_work_sem
);
3303 /* cleanup FS via transaction */
3304 btrfs_cleanup_transaction(root
);
3306 ret
= write_ctree_super(NULL
, root
, 0);
3311 static void btrfs_destroy_ordered_operations(struct btrfs_root
*root
)
3313 struct btrfs_inode
*btrfs_inode
;
3314 struct list_head splice
;
3316 INIT_LIST_HEAD(&splice
);
3318 mutex_lock(&root
->fs_info
->ordered_operations_mutex
);
3319 spin_lock(&root
->fs_info
->ordered_extent_lock
);
3321 list_splice_init(&root
->fs_info
->ordered_operations
, &splice
);
3322 while (!list_empty(&splice
)) {
3323 btrfs_inode
= list_entry(splice
.next
, struct btrfs_inode
,
3324 ordered_operations
);
3326 list_del_init(&btrfs_inode
->ordered_operations
);
3328 btrfs_invalidate_inodes(btrfs_inode
->root
);
3331 spin_unlock(&root
->fs_info
->ordered_extent_lock
);
3332 mutex_unlock(&root
->fs_info
->ordered_operations_mutex
);
3335 static void btrfs_destroy_ordered_extents(struct btrfs_root
*root
)
3337 struct list_head splice
;
3338 struct btrfs_ordered_extent
*ordered
;
3339 struct inode
*inode
;
3341 INIT_LIST_HEAD(&splice
);
3343 spin_lock(&root
->fs_info
->ordered_extent_lock
);
3345 list_splice_init(&root
->fs_info
->ordered_extents
, &splice
);
3346 while (!list_empty(&splice
)) {
3347 ordered
= list_entry(splice
.next
, struct btrfs_ordered_extent
,
3350 list_del_init(&ordered
->root_extent_list
);
3351 atomic_inc(&ordered
->refs
);
3353 /* the inode may be getting freed (in sys_unlink path). */
3354 inode
= igrab(ordered
->inode
);
3356 spin_unlock(&root
->fs_info
->ordered_extent_lock
);
3360 atomic_set(&ordered
->refs
, 1);
3361 btrfs_put_ordered_extent(ordered
);
3363 spin_lock(&root
->fs_info
->ordered_extent_lock
);
3366 spin_unlock(&root
->fs_info
->ordered_extent_lock
);
3369 static int btrfs_destroy_delayed_refs(struct btrfs_transaction
*trans
,
3370 struct btrfs_root
*root
)
3372 struct rb_node
*node
;
3373 struct btrfs_delayed_ref_root
*delayed_refs
;
3374 struct btrfs_delayed_ref_node
*ref
;
3377 delayed_refs
= &trans
->delayed_refs
;
3379 spin_lock(&delayed_refs
->lock
);
3380 if (delayed_refs
->num_entries
== 0) {
3381 spin_unlock(&delayed_refs
->lock
);
3382 printk(KERN_INFO
"delayed_refs has NO entry\n");
3386 node
= rb_first(&delayed_refs
->root
);
3388 ref
= rb_entry(node
, struct btrfs_delayed_ref_node
, rb_node
);
3389 node
= rb_next(node
);
3392 rb_erase(&ref
->rb_node
, &delayed_refs
->root
);
3393 delayed_refs
->num_entries
--;
3395 atomic_set(&ref
->refs
, 1);
3396 if (btrfs_delayed_ref_is_head(ref
)) {
3397 struct btrfs_delayed_ref_head
*head
;
3399 head
= btrfs_delayed_node_to_head(ref
);
3400 mutex_lock(&head
->mutex
);
3401 kfree(head
->extent_op
);
3402 delayed_refs
->num_heads
--;
3403 if (list_empty(&head
->cluster
))
3404 delayed_refs
->num_heads_ready
--;
3405 list_del_init(&head
->cluster
);
3406 mutex_unlock(&head
->mutex
);
3409 spin_unlock(&delayed_refs
->lock
);
3410 btrfs_put_delayed_ref(ref
);
3413 spin_lock(&delayed_refs
->lock
);
3416 spin_unlock(&delayed_refs
->lock
);
3421 static void btrfs_destroy_pending_snapshots(struct btrfs_transaction
*t
)
3423 struct btrfs_pending_snapshot
*snapshot
;
3424 struct list_head splice
;
3426 INIT_LIST_HEAD(&splice
);
3428 list_splice_init(&t
->pending_snapshots
, &splice
);
3430 while (!list_empty(&splice
)) {
3431 snapshot
= list_entry(splice
.next
,
3432 struct btrfs_pending_snapshot
,
3435 list_del_init(&snapshot
->list
);
3441 static void btrfs_destroy_delalloc_inodes(struct btrfs_root
*root
)
3443 struct btrfs_inode
*btrfs_inode
;
3444 struct list_head splice
;
3446 INIT_LIST_HEAD(&splice
);
3448 spin_lock(&root
->fs_info
->delalloc_lock
);
3449 list_splice_init(&root
->fs_info
->delalloc_inodes
, &splice
);
3451 while (!list_empty(&splice
)) {
3452 btrfs_inode
= list_entry(splice
.next
, struct btrfs_inode
,
3455 list_del_init(&btrfs_inode
->delalloc_inodes
);
3457 btrfs_invalidate_inodes(btrfs_inode
->root
);
3460 spin_unlock(&root
->fs_info
->delalloc_lock
);
3463 static int btrfs_destroy_marked_extents(struct btrfs_root
*root
,
3464 struct extent_io_tree
*dirty_pages
,
3469 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
3470 struct extent_buffer
*eb
;
3474 unsigned long index
;
3477 ret
= find_first_extent_bit(dirty_pages
, start
, &start
, &end
,
3482 clear_extent_bits(dirty_pages
, start
, end
, mark
, GFP_NOFS
);
3483 while (start
<= end
) {
3484 index
= start
>> PAGE_CACHE_SHIFT
;
3485 start
= (u64
)(index
+ 1) << PAGE_CACHE_SHIFT
;
3486 page
= find_get_page(btree_inode
->i_mapping
, index
);
3489 offset
= page_offset(page
);
3491 spin_lock(&dirty_pages
->buffer_lock
);
3492 eb
= radix_tree_lookup(
3493 &(&BTRFS_I(page
->mapping
->host
)->io_tree
)->buffer
,
3494 offset
>> PAGE_CACHE_SHIFT
);
3495 spin_unlock(&dirty_pages
->buffer_lock
);
3497 ret
= test_and_clear_bit(EXTENT_BUFFER_DIRTY
,
3499 atomic_set(&eb
->refs
, 1);
3501 if (PageWriteback(page
))
3502 end_page_writeback(page
);
3505 if (PageDirty(page
)) {
3506 clear_page_dirty_for_io(page
);
3507 spin_lock_irq(&page
->mapping
->tree_lock
);
3508 radix_tree_tag_clear(&page
->mapping
->page_tree
,
3510 PAGECACHE_TAG_DIRTY
);
3511 spin_unlock_irq(&page
->mapping
->tree_lock
);
3514 page
->mapping
->a_ops
->invalidatepage(page
, 0);
3522 static int btrfs_destroy_pinned_extent(struct btrfs_root
*root
,
3523 struct extent_io_tree
*pinned_extents
)
3525 struct extent_io_tree
*unpin
;
3530 unpin
= pinned_extents
;
3532 ret
= find_first_extent_bit(unpin
, 0, &start
, &end
,
3538 if (btrfs_test_opt(root
, DISCARD
))
3539 ret
= btrfs_error_discard_extent(root
, start
,
3543 clear_extent_dirty(unpin
, start
, end
, GFP_NOFS
);
3544 btrfs_error_unpin_extent_range(root
, start
, end
);
3551 void btrfs_cleanup_one_transaction(struct btrfs_transaction
*cur_trans
,
3552 struct btrfs_root
*root
)
3554 btrfs_destroy_delayed_refs(cur_trans
, root
);
3555 btrfs_block_rsv_release(root
, &root
->fs_info
->trans_block_rsv
,
3556 cur_trans
->dirty_pages
.dirty_bytes
);
3558 /* FIXME: cleanup wait for commit */
3559 cur_trans
->in_commit
= 1;
3560 cur_trans
->blocked
= 1;
3561 if (waitqueue_active(&root
->fs_info
->transaction_blocked_wait
))
3562 wake_up(&root
->fs_info
->transaction_blocked_wait
);
3564 cur_trans
->blocked
= 0;
3565 if (waitqueue_active(&root
->fs_info
->transaction_wait
))
3566 wake_up(&root
->fs_info
->transaction_wait
);
3568 cur_trans
->commit_done
= 1;
3569 if (waitqueue_active(&cur_trans
->commit_wait
))
3570 wake_up(&cur_trans
->commit_wait
);
3572 btrfs_destroy_pending_snapshots(cur_trans
);
3574 btrfs_destroy_marked_extents(root
, &cur_trans
->dirty_pages
,
3578 memset(cur_trans, 0, sizeof(*cur_trans));
3579 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
3583 int btrfs_cleanup_transaction(struct btrfs_root
*root
)
3585 struct btrfs_transaction
*t
;
3588 mutex_lock(&root
->fs_info
->transaction_kthread_mutex
);
3590 spin_lock(&root
->fs_info
->trans_lock
);
3591 list_splice_init(&root
->fs_info
->trans_list
, &list
);
3592 root
->fs_info
->trans_no_join
= 1;
3593 spin_unlock(&root
->fs_info
->trans_lock
);
3595 while (!list_empty(&list
)) {
3596 t
= list_entry(list
.next
, struct btrfs_transaction
, list
);
3600 btrfs_destroy_ordered_operations(root
);
3602 btrfs_destroy_ordered_extents(root
);
3604 btrfs_destroy_delayed_refs(t
, root
);
3606 btrfs_block_rsv_release(root
,
3607 &root
->fs_info
->trans_block_rsv
,
3608 t
->dirty_pages
.dirty_bytes
);
3610 /* FIXME: cleanup wait for commit */
3613 if (waitqueue_active(&root
->fs_info
->transaction_blocked_wait
))
3614 wake_up(&root
->fs_info
->transaction_blocked_wait
);
3617 if (waitqueue_active(&root
->fs_info
->transaction_wait
))
3618 wake_up(&root
->fs_info
->transaction_wait
);
3621 if (waitqueue_active(&t
->commit_wait
))
3622 wake_up(&t
->commit_wait
);
3624 btrfs_destroy_pending_snapshots(t
);
3626 btrfs_destroy_delalloc_inodes(root
);
3628 spin_lock(&root
->fs_info
->trans_lock
);
3629 root
->fs_info
->running_transaction
= NULL
;
3630 spin_unlock(&root
->fs_info
->trans_lock
);
3632 btrfs_destroy_marked_extents(root
, &t
->dirty_pages
,
3635 btrfs_destroy_pinned_extent(root
,
3636 root
->fs_info
->pinned_extents
);
3638 atomic_set(&t
->use_count
, 0);
3639 list_del_init(&t
->list
);
3640 memset(t
, 0, sizeof(*t
));
3641 kmem_cache_free(btrfs_transaction_cachep
, t
);
3644 spin_lock(&root
->fs_info
->trans_lock
);
3645 root
->fs_info
->trans_no_join
= 0;
3646 spin_unlock(&root
->fs_info
->trans_lock
);
3647 mutex_unlock(&root
->fs_info
->transaction_kthread_mutex
);
3652 static struct extent_io_ops btree_extent_io_ops
= {
3653 .write_cache_pages_lock_hook
= btree_lock_page_hook
,
3654 .readpage_end_io_hook
= btree_readpage_end_io_hook
,
3655 .readpage_io_failed_hook
= btree_io_failed_hook
,
3656 .submit_bio_hook
= btree_submit_bio_hook
,
3657 /* note we're sharing with inode.c for the merge bio hook */
3658 .merge_bio_hook
= btrfs_merge_bio_hook
,