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>
35 #include "transaction.h"
36 #include "btrfs_inode.h"
38 #include "print-tree.h"
39 #include "async-thread.h"
42 #include "free-space-cache.h"
44 static struct extent_io_ops btree_extent_io_ops
;
45 static void end_workqueue_fn(struct btrfs_work
*work
);
46 static void free_fs_root(struct btrfs_root
*root
);
47 static void btrfs_check_super_valid(struct btrfs_fs_info
*fs_info
,
49 static int btrfs_destroy_ordered_operations(struct btrfs_root
*root
);
50 static int btrfs_destroy_ordered_extents(struct btrfs_root
*root
);
51 static int btrfs_destroy_delayed_refs(struct btrfs_transaction
*trans
,
52 struct btrfs_root
*root
);
53 static int btrfs_destroy_pending_snapshots(struct btrfs_transaction
*t
);
54 static int btrfs_destroy_delalloc_inodes(struct btrfs_root
*root
);
55 static int btrfs_destroy_marked_extents(struct btrfs_root
*root
,
56 struct extent_io_tree
*dirty_pages
,
58 static int btrfs_destroy_pinned_extent(struct btrfs_root
*root
,
59 struct extent_io_tree
*pinned_extents
);
60 static int btrfs_cleanup_transaction(struct btrfs_root
*root
);
63 * end_io_wq structs are used to do processing in task context when an IO is
64 * complete. This is used during reads to verify checksums, and it is used
65 * by writes to insert metadata for new file extents after IO is complete.
71 struct btrfs_fs_info
*info
;
74 struct list_head list
;
75 struct btrfs_work work
;
79 * async submit bios are used to offload expensive checksumming
80 * onto the worker threads. They checksum file and metadata bios
81 * just before they are sent down the IO stack.
83 struct async_submit_bio
{
86 struct list_head list
;
87 extent_submit_bio_hook_t
*submit_bio_start
;
88 extent_submit_bio_hook_t
*submit_bio_done
;
91 unsigned long bio_flags
;
93 * bio_offset is optional, can be used if the pages in the bio
94 * can't tell us where in the file the bio should go
97 struct btrfs_work work
;
100 /* These are used to set the lockdep class on the extent buffer locks.
101 * The class is set by the readpage_end_io_hook after the buffer has
102 * passed csum validation but before the pages are unlocked.
104 * The lockdep class is also set by btrfs_init_new_buffer on freshly
107 * The class is based on the level in the tree block, which allows lockdep
108 * to know that lower nodes nest inside the locks of higher nodes.
110 * We also add a check to make sure the highest level of the tree is
111 * the same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this
112 * code needs update as well.
114 #ifdef CONFIG_DEBUG_LOCK_ALLOC
115 # if BTRFS_MAX_LEVEL != 8
118 static struct lock_class_key btrfs_eb_class
[BTRFS_MAX_LEVEL
+ 1];
119 static const char *btrfs_eb_name
[BTRFS_MAX_LEVEL
+ 1] = {
129 /* highest possible level */
135 * extents on the btree inode are pretty simple, there's one extent
136 * that covers the entire device
138 static struct extent_map
*btree_get_extent(struct inode
*inode
,
139 struct page
*page
, size_t page_offset
, u64 start
, u64 len
,
142 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
143 struct extent_map
*em
;
146 read_lock(&em_tree
->lock
);
147 em
= lookup_extent_mapping(em_tree
, start
, len
);
150 BTRFS_I(inode
)->root
->fs_info
->fs_devices
->latest_bdev
;
151 read_unlock(&em_tree
->lock
);
154 read_unlock(&em_tree
->lock
);
156 em
= alloc_extent_map(GFP_NOFS
);
158 em
= ERR_PTR(-ENOMEM
);
163 em
->block_len
= (u64
)-1;
165 em
->bdev
= BTRFS_I(inode
)->root
->fs_info
->fs_devices
->latest_bdev
;
167 write_lock(&em_tree
->lock
);
168 ret
= add_extent_mapping(em_tree
, em
);
169 if (ret
== -EEXIST
) {
170 u64 failed_start
= em
->start
;
171 u64 failed_len
= em
->len
;
174 em
= lookup_extent_mapping(em_tree
, start
, len
);
178 em
= lookup_extent_mapping(em_tree
, failed_start
,
186 write_unlock(&em_tree
->lock
);
194 u32
btrfs_csum_data(struct btrfs_root
*root
, char *data
, u32 seed
, size_t len
)
196 return crc32c(seed
, data
, len
);
199 void btrfs_csum_final(u32 crc
, char *result
)
201 *(__le32
*)result
= ~cpu_to_le32(crc
);
205 * compute the csum for a btree block, and either verify it or write it
206 * into the csum field of the block.
208 static int csum_tree_block(struct btrfs_root
*root
, struct extent_buffer
*buf
,
212 btrfs_super_csum_size(&root
->fs_info
->super_copy
);
215 unsigned long cur_len
;
216 unsigned long offset
= BTRFS_CSUM_SIZE
;
217 char *map_token
= NULL
;
219 unsigned long map_start
;
220 unsigned long map_len
;
223 unsigned long inline_result
;
225 len
= buf
->len
- offset
;
227 err
= map_private_extent_buffer(buf
, offset
, 32,
229 &map_start
, &map_len
, KM_USER0
);
232 cur_len
= min(len
, map_len
- (offset
- map_start
));
233 crc
= btrfs_csum_data(root
, kaddr
+ offset
- map_start
,
237 unmap_extent_buffer(buf
, map_token
, KM_USER0
);
239 if (csum_size
> sizeof(inline_result
)) {
240 result
= kzalloc(csum_size
* sizeof(char), GFP_NOFS
);
244 result
= (char *)&inline_result
;
247 btrfs_csum_final(crc
, result
);
250 if (memcmp_extent_buffer(buf
, result
, 0, csum_size
)) {
253 memcpy(&found
, result
, csum_size
);
255 read_extent_buffer(buf
, &val
, 0, csum_size
);
256 if (printk_ratelimit()) {
257 printk(KERN_INFO
"btrfs: %s checksum verify "
258 "failed on %llu wanted %X found %X "
260 root
->fs_info
->sb
->s_id
,
261 (unsigned long long)buf
->start
, val
, found
,
262 btrfs_header_level(buf
));
264 if (result
!= (char *)&inline_result
)
269 write_extent_buffer(buf
, result
, 0, csum_size
);
271 if (result
!= (char *)&inline_result
)
277 * we can't consider a given block up to date unless the transid of the
278 * block matches the transid in the parent node's pointer. This is how we
279 * detect blocks that either didn't get written at all or got written
280 * in the wrong place.
282 static int verify_parent_transid(struct extent_io_tree
*io_tree
,
283 struct extent_buffer
*eb
, u64 parent_transid
)
285 struct extent_state
*cached_state
= NULL
;
288 if (!parent_transid
|| btrfs_header_generation(eb
) == parent_transid
)
291 lock_extent_bits(io_tree
, eb
->start
, eb
->start
+ eb
->len
- 1,
292 0, &cached_state
, GFP_NOFS
);
293 if (extent_buffer_uptodate(io_tree
, eb
, cached_state
) &&
294 btrfs_header_generation(eb
) == parent_transid
) {
298 if (printk_ratelimit()) {
299 printk("parent transid verify failed on %llu wanted %llu "
301 (unsigned long long)eb
->start
,
302 (unsigned long long)parent_transid
,
303 (unsigned long long)btrfs_header_generation(eb
));
306 clear_extent_buffer_uptodate(io_tree
, eb
, &cached_state
);
308 unlock_extent_cached(io_tree
, eb
->start
, eb
->start
+ eb
->len
- 1,
309 &cached_state
, GFP_NOFS
);
314 * helper to read a given tree block, doing retries as required when
315 * the checksums don't match and we have alternate mirrors to try.
317 static int btree_read_extent_buffer_pages(struct btrfs_root
*root
,
318 struct extent_buffer
*eb
,
319 u64 start
, u64 parent_transid
)
321 struct extent_io_tree
*io_tree
;
326 io_tree
= &BTRFS_I(root
->fs_info
->btree_inode
)->io_tree
;
328 ret
= read_extent_buffer_pages(io_tree
, eb
, start
, 1,
329 btree_get_extent
, mirror_num
);
331 !verify_parent_transid(io_tree
, eb
, parent_transid
))
334 num_copies
= btrfs_num_copies(&root
->fs_info
->mapping_tree
,
340 if (mirror_num
> num_copies
)
347 * checksum a dirty tree block before IO. This has extra checks to make sure
348 * we only fill in the checksum field in the first page of a multi-page block
351 static int csum_dirty_buffer(struct btrfs_root
*root
, struct page
*page
)
353 struct extent_io_tree
*tree
;
354 u64 start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
357 struct extent_buffer
*eb
;
360 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
362 if (page
->private == EXTENT_PAGE_PRIVATE
)
366 len
= page
->private >> 2;
369 eb
= alloc_extent_buffer(tree
, start
, len
, page
, GFP_NOFS
);
374 ret
= btree_read_extent_buffer_pages(root
, eb
, start
+ PAGE_CACHE_SIZE
,
375 btrfs_header_generation(eb
));
377 WARN_ON(!btrfs_header_flag(eb
, BTRFS_HEADER_FLAG_WRITTEN
));
379 found_start
= btrfs_header_bytenr(eb
);
380 if (found_start
!= start
) {
384 if (eb
->first_page
!= page
) {
388 if (!PageUptodate(page
)) {
392 csum_tree_block(root
, eb
, 0);
394 free_extent_buffer(eb
);
399 static int check_tree_block_fsid(struct btrfs_root
*root
,
400 struct extent_buffer
*eb
)
402 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
403 u8 fsid
[BTRFS_UUID_SIZE
];
406 read_extent_buffer(eb
, fsid
, (unsigned long)btrfs_header_fsid(eb
),
409 if (!memcmp(fsid
, fs_devices
->fsid
, BTRFS_FSID_SIZE
)) {
413 fs_devices
= fs_devices
->seed
;
418 #ifdef CONFIG_DEBUG_LOCK_ALLOC
419 void btrfs_set_buffer_lockdep_class(struct extent_buffer
*eb
, int level
)
421 lockdep_set_class_and_name(&eb
->lock
,
422 &btrfs_eb_class
[level
],
423 btrfs_eb_name
[level
]);
427 static int btree_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
428 struct extent_state
*state
)
430 struct extent_io_tree
*tree
;
434 struct extent_buffer
*eb
;
435 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
438 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
439 if (page
->private == EXTENT_PAGE_PRIVATE
)
444 len
= page
->private >> 2;
447 eb
= alloc_extent_buffer(tree
, start
, len
, page
, GFP_NOFS
);
453 found_start
= btrfs_header_bytenr(eb
);
454 if (found_start
!= start
) {
455 if (printk_ratelimit()) {
456 printk(KERN_INFO
"btrfs bad tree block start "
458 (unsigned long long)found_start
,
459 (unsigned long long)eb
->start
);
464 if (eb
->first_page
!= page
) {
465 printk(KERN_INFO
"btrfs bad first page %lu %lu\n",
466 eb
->first_page
->index
, page
->index
);
471 if (check_tree_block_fsid(root
, eb
)) {
472 if (printk_ratelimit()) {
473 printk(KERN_INFO
"btrfs bad fsid on block %llu\n",
474 (unsigned long long)eb
->start
);
479 found_level
= btrfs_header_level(eb
);
481 btrfs_set_buffer_lockdep_class(eb
, found_level
);
483 ret
= csum_tree_block(root
, eb
, 1);
487 end
= min_t(u64
, eb
->len
, PAGE_CACHE_SIZE
);
488 end
= eb
->start
+ end
- 1;
490 free_extent_buffer(eb
);
495 static void end_workqueue_bio(struct bio
*bio
, int err
)
497 struct end_io_wq
*end_io_wq
= bio
->bi_private
;
498 struct btrfs_fs_info
*fs_info
;
500 fs_info
= end_io_wq
->info
;
501 end_io_wq
->error
= err
;
502 end_io_wq
->work
.func
= end_workqueue_fn
;
503 end_io_wq
->work
.flags
= 0;
505 if (bio
->bi_rw
& REQ_WRITE
) {
506 if (end_io_wq
->metadata
== 1)
507 btrfs_queue_worker(&fs_info
->endio_meta_write_workers
,
509 else if (end_io_wq
->metadata
== 2)
510 btrfs_queue_worker(&fs_info
->endio_freespace_worker
,
513 btrfs_queue_worker(&fs_info
->endio_write_workers
,
516 if (end_io_wq
->metadata
)
517 btrfs_queue_worker(&fs_info
->endio_meta_workers
,
520 btrfs_queue_worker(&fs_info
->endio_workers
,
526 * For the metadata arg you want
529 * 1 - if normal metadta
530 * 2 - if writing to the free space cache area
532 int btrfs_bio_wq_end_io(struct btrfs_fs_info
*info
, struct bio
*bio
,
535 struct end_io_wq
*end_io_wq
;
536 end_io_wq
= kmalloc(sizeof(*end_io_wq
), GFP_NOFS
);
540 end_io_wq
->private = bio
->bi_private
;
541 end_io_wq
->end_io
= bio
->bi_end_io
;
542 end_io_wq
->info
= info
;
543 end_io_wq
->error
= 0;
544 end_io_wq
->bio
= bio
;
545 end_io_wq
->metadata
= metadata
;
547 bio
->bi_private
= end_io_wq
;
548 bio
->bi_end_io
= end_workqueue_bio
;
552 unsigned long btrfs_async_submit_limit(struct btrfs_fs_info
*info
)
554 unsigned long limit
= min_t(unsigned long,
555 info
->workers
.max_workers
,
556 info
->fs_devices
->open_devices
);
560 int btrfs_congested_async(struct btrfs_fs_info
*info
, int iodone
)
562 return atomic_read(&info
->nr_async_bios
) >
563 btrfs_async_submit_limit(info
);
566 static void run_one_async_start(struct btrfs_work
*work
)
568 struct async_submit_bio
*async
;
570 async
= container_of(work
, struct async_submit_bio
, work
);
571 async
->submit_bio_start(async
->inode
, async
->rw
, async
->bio
,
572 async
->mirror_num
, async
->bio_flags
,
576 static void run_one_async_done(struct btrfs_work
*work
)
578 struct btrfs_fs_info
*fs_info
;
579 struct async_submit_bio
*async
;
582 async
= container_of(work
, struct async_submit_bio
, work
);
583 fs_info
= BTRFS_I(async
->inode
)->root
->fs_info
;
585 limit
= btrfs_async_submit_limit(fs_info
);
586 limit
= limit
* 2 / 3;
588 atomic_dec(&fs_info
->nr_async_submits
);
590 if (atomic_read(&fs_info
->nr_async_submits
) < limit
&&
591 waitqueue_active(&fs_info
->async_submit_wait
))
592 wake_up(&fs_info
->async_submit_wait
);
594 async
->submit_bio_done(async
->inode
, async
->rw
, async
->bio
,
595 async
->mirror_num
, async
->bio_flags
,
599 static void run_one_async_free(struct btrfs_work
*work
)
601 struct async_submit_bio
*async
;
603 async
= container_of(work
, struct async_submit_bio
, work
);
607 int btrfs_wq_submit_bio(struct btrfs_fs_info
*fs_info
, struct inode
*inode
,
608 int rw
, struct bio
*bio
, int mirror_num
,
609 unsigned long bio_flags
,
611 extent_submit_bio_hook_t
*submit_bio_start
,
612 extent_submit_bio_hook_t
*submit_bio_done
)
614 struct async_submit_bio
*async
;
616 async
= kmalloc(sizeof(*async
), GFP_NOFS
);
620 async
->inode
= inode
;
623 async
->mirror_num
= mirror_num
;
624 async
->submit_bio_start
= submit_bio_start
;
625 async
->submit_bio_done
= submit_bio_done
;
627 async
->work
.func
= run_one_async_start
;
628 async
->work
.ordered_func
= run_one_async_done
;
629 async
->work
.ordered_free
= run_one_async_free
;
631 async
->work
.flags
= 0;
632 async
->bio_flags
= bio_flags
;
633 async
->bio_offset
= bio_offset
;
635 atomic_inc(&fs_info
->nr_async_submits
);
638 btrfs_set_work_high_prio(&async
->work
);
640 btrfs_queue_worker(&fs_info
->workers
, &async
->work
);
642 while (atomic_read(&fs_info
->async_submit_draining
) &&
643 atomic_read(&fs_info
->nr_async_submits
)) {
644 wait_event(fs_info
->async_submit_wait
,
645 (atomic_read(&fs_info
->nr_async_submits
) == 0));
651 static int btree_csum_one_bio(struct bio
*bio
)
653 struct bio_vec
*bvec
= bio
->bi_io_vec
;
655 struct btrfs_root
*root
;
657 WARN_ON(bio
->bi_vcnt
<= 0);
658 while (bio_index
< bio
->bi_vcnt
) {
659 root
= BTRFS_I(bvec
->bv_page
->mapping
->host
)->root
;
660 csum_dirty_buffer(root
, bvec
->bv_page
);
667 static int __btree_submit_bio_start(struct inode
*inode
, int rw
,
668 struct bio
*bio
, int mirror_num
,
669 unsigned long bio_flags
,
673 * when we're called for a write, we're already in the async
674 * submission context. Just jump into btrfs_map_bio
676 btree_csum_one_bio(bio
);
680 static int __btree_submit_bio_done(struct inode
*inode
, int rw
, struct bio
*bio
,
681 int mirror_num
, unsigned long bio_flags
,
685 * when we're called for a write, we're already in the async
686 * submission context. Just jump into btrfs_map_bio
688 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
, mirror_num
, 1);
691 static int btree_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
692 int mirror_num
, unsigned long bio_flags
,
697 ret
= btrfs_bio_wq_end_io(BTRFS_I(inode
)->root
->fs_info
,
701 if (!(rw
& REQ_WRITE
)) {
703 * called for a read, do the setup so that checksum validation
704 * can happen in the async kernel threads
706 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
,
711 * kthread helpers are used to submit writes so that checksumming
712 * can happen in parallel across all CPUs
714 return btrfs_wq_submit_bio(BTRFS_I(inode
)->root
->fs_info
,
715 inode
, rw
, bio
, mirror_num
, 0,
717 __btree_submit_bio_start
,
718 __btree_submit_bio_done
);
721 #ifdef CONFIG_MIGRATION
722 static int btree_migratepage(struct address_space
*mapping
,
723 struct page
*newpage
, struct page
*page
)
726 * we can't safely write a btree page from here,
727 * we haven't done the locking hook
732 * Buffers may be managed in a filesystem specific way.
733 * We must have no buffers or drop them.
735 if (page_has_private(page
) &&
736 !try_to_release_page(page
, GFP_KERNEL
))
738 return migrate_page(mapping
, newpage
, page
);
742 static int btree_writepage(struct page
*page
, struct writeback_control
*wbc
)
744 struct extent_io_tree
*tree
;
745 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
746 struct extent_buffer
*eb
;
749 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
750 if (!(current
->flags
& PF_MEMALLOC
)) {
751 return extent_write_full_page(tree
, page
,
752 btree_get_extent
, wbc
);
755 redirty_page_for_writepage(wbc
, page
);
756 eb
= btrfs_find_tree_block(root
, page_offset(page
), PAGE_CACHE_SIZE
);
759 was_dirty
= test_and_set_bit(EXTENT_BUFFER_DIRTY
, &eb
->bflags
);
761 spin_lock(&root
->fs_info
->delalloc_lock
);
762 root
->fs_info
->dirty_metadata_bytes
+= PAGE_CACHE_SIZE
;
763 spin_unlock(&root
->fs_info
->delalloc_lock
);
765 free_extent_buffer(eb
);
771 static int btree_writepages(struct address_space
*mapping
,
772 struct writeback_control
*wbc
)
774 struct extent_io_tree
*tree
;
775 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
776 if (wbc
->sync_mode
== WB_SYNC_NONE
) {
777 struct btrfs_root
*root
= BTRFS_I(mapping
->host
)->root
;
779 unsigned long thresh
= 32 * 1024 * 1024;
781 if (wbc
->for_kupdate
)
784 /* this is a bit racy, but that's ok */
785 num_dirty
= root
->fs_info
->dirty_metadata_bytes
;
786 if (num_dirty
< thresh
)
789 return extent_writepages(tree
, mapping
, btree_get_extent
, wbc
);
792 static int btree_readpage(struct file
*file
, struct page
*page
)
794 struct extent_io_tree
*tree
;
795 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
796 return extent_read_full_page(tree
, page
, btree_get_extent
);
799 static int btree_releasepage(struct page
*page
, gfp_t gfp_flags
)
801 struct extent_io_tree
*tree
;
802 struct extent_map_tree
*map
;
805 if (PageWriteback(page
) || PageDirty(page
))
808 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
809 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
811 ret
= try_release_extent_state(map
, tree
, page
, gfp_flags
);
815 ret
= try_release_extent_buffer(tree
, page
);
817 ClearPagePrivate(page
);
818 set_page_private(page
, 0);
819 page_cache_release(page
);
825 static void btree_invalidatepage(struct page
*page
, unsigned long offset
)
827 struct extent_io_tree
*tree
;
828 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
829 extent_invalidatepage(tree
, page
, offset
);
830 btree_releasepage(page
, GFP_NOFS
);
831 if (PagePrivate(page
)) {
832 printk(KERN_WARNING
"btrfs warning page private not zero "
833 "on page %llu\n", (unsigned long long)page_offset(page
));
834 ClearPagePrivate(page
);
835 set_page_private(page
, 0);
836 page_cache_release(page
);
840 static const struct address_space_operations btree_aops
= {
841 .readpage
= btree_readpage
,
842 .writepage
= btree_writepage
,
843 .writepages
= btree_writepages
,
844 .releasepage
= btree_releasepage
,
845 .invalidatepage
= btree_invalidatepage
,
846 .sync_page
= block_sync_page
,
847 #ifdef CONFIG_MIGRATION
848 .migratepage
= btree_migratepage
,
852 int readahead_tree_block(struct btrfs_root
*root
, u64 bytenr
, u32 blocksize
,
855 struct extent_buffer
*buf
= NULL
;
856 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
859 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
862 read_extent_buffer_pages(&BTRFS_I(btree_inode
)->io_tree
,
863 buf
, 0, 0, btree_get_extent
, 0);
864 free_extent_buffer(buf
);
868 struct extent_buffer
*btrfs_find_tree_block(struct btrfs_root
*root
,
869 u64 bytenr
, u32 blocksize
)
871 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
872 struct extent_buffer
*eb
;
873 eb
= find_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
874 bytenr
, blocksize
, GFP_NOFS
);
878 struct extent_buffer
*btrfs_find_create_tree_block(struct btrfs_root
*root
,
879 u64 bytenr
, u32 blocksize
)
881 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
882 struct extent_buffer
*eb
;
884 eb
= alloc_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
885 bytenr
, blocksize
, NULL
, GFP_NOFS
);
890 int btrfs_write_tree_block(struct extent_buffer
*buf
)
892 return filemap_fdatawrite_range(buf
->first_page
->mapping
, buf
->start
,
893 buf
->start
+ buf
->len
- 1);
896 int btrfs_wait_tree_block_writeback(struct extent_buffer
*buf
)
898 return filemap_fdatawait_range(buf
->first_page
->mapping
,
899 buf
->start
, buf
->start
+ buf
->len
- 1);
902 struct extent_buffer
*read_tree_block(struct btrfs_root
*root
, u64 bytenr
,
903 u32 blocksize
, u64 parent_transid
)
905 struct extent_buffer
*buf
= NULL
;
908 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
912 ret
= btree_read_extent_buffer_pages(root
, buf
, 0, parent_transid
);
915 set_bit(EXTENT_BUFFER_UPTODATE
, &buf
->bflags
);
920 int clean_tree_block(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
921 struct extent_buffer
*buf
)
923 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
924 if (btrfs_header_generation(buf
) ==
925 root
->fs_info
->running_transaction
->transid
) {
926 btrfs_assert_tree_locked(buf
);
928 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY
, &buf
->bflags
)) {
929 spin_lock(&root
->fs_info
->delalloc_lock
);
930 if (root
->fs_info
->dirty_metadata_bytes
>= buf
->len
)
931 root
->fs_info
->dirty_metadata_bytes
-= buf
->len
;
934 spin_unlock(&root
->fs_info
->delalloc_lock
);
937 /* ugh, clear_extent_buffer_dirty needs to lock the page */
938 btrfs_set_lock_blocking(buf
);
939 clear_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
,
945 static int __setup_root(u32 nodesize
, u32 leafsize
, u32 sectorsize
,
946 u32 stripesize
, struct btrfs_root
*root
,
947 struct btrfs_fs_info
*fs_info
,
951 root
->commit_root
= NULL
;
952 root
->sectorsize
= sectorsize
;
953 root
->nodesize
= nodesize
;
954 root
->leafsize
= leafsize
;
955 root
->stripesize
= stripesize
;
957 root
->track_dirty
= 0;
959 root
->orphan_item_inserted
= 0;
960 root
->orphan_cleanup_state
= 0;
962 root
->fs_info
= fs_info
;
963 root
->objectid
= objectid
;
964 root
->last_trans
= 0;
965 root
->highest_objectid
= 0;
968 root
->inode_tree
= RB_ROOT
;
969 root
->block_rsv
= NULL
;
970 root
->orphan_block_rsv
= NULL
;
972 INIT_LIST_HEAD(&root
->dirty_list
);
973 INIT_LIST_HEAD(&root
->orphan_list
);
974 INIT_LIST_HEAD(&root
->root_list
);
975 spin_lock_init(&root
->node_lock
);
976 spin_lock_init(&root
->orphan_lock
);
977 spin_lock_init(&root
->inode_lock
);
978 spin_lock_init(&root
->accounting_lock
);
979 mutex_init(&root
->objectid_mutex
);
980 mutex_init(&root
->log_mutex
);
981 init_waitqueue_head(&root
->log_writer_wait
);
982 init_waitqueue_head(&root
->log_commit_wait
[0]);
983 init_waitqueue_head(&root
->log_commit_wait
[1]);
984 atomic_set(&root
->log_commit
[0], 0);
985 atomic_set(&root
->log_commit
[1], 0);
986 atomic_set(&root
->log_writers
, 0);
988 root
->log_transid
= 0;
989 root
->last_log_commit
= 0;
990 extent_io_tree_init(&root
->dirty_log_pages
,
991 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
993 memset(&root
->root_key
, 0, sizeof(root
->root_key
));
994 memset(&root
->root_item
, 0, sizeof(root
->root_item
));
995 memset(&root
->defrag_progress
, 0, sizeof(root
->defrag_progress
));
996 memset(&root
->root_kobj
, 0, sizeof(root
->root_kobj
));
997 root
->defrag_trans_start
= fs_info
->generation
;
998 init_completion(&root
->kobj_unregister
);
999 root
->defrag_running
= 0;
1000 root
->root_key
.objectid
= objectid
;
1001 root
->anon_super
.s_root
= NULL
;
1002 root
->anon_super
.s_dev
= 0;
1003 INIT_LIST_HEAD(&root
->anon_super
.s_list
);
1004 INIT_LIST_HEAD(&root
->anon_super
.s_instances
);
1005 init_rwsem(&root
->anon_super
.s_umount
);
1010 static int find_and_setup_root(struct btrfs_root
*tree_root
,
1011 struct btrfs_fs_info
*fs_info
,
1013 struct btrfs_root
*root
)
1019 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
1020 tree_root
->sectorsize
, tree_root
->stripesize
,
1021 root
, fs_info
, objectid
);
1022 ret
= btrfs_find_last_root(tree_root
, objectid
,
1023 &root
->root_item
, &root
->root_key
);
1028 generation
= btrfs_root_generation(&root
->root_item
);
1029 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
1030 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
1031 blocksize
, generation
);
1032 if (!root
->node
|| !btrfs_buffer_uptodate(root
->node
, generation
)) {
1033 free_extent_buffer(root
->node
);
1036 root
->commit_root
= btrfs_root_node(root
);
1040 static struct btrfs_root
*alloc_log_tree(struct btrfs_trans_handle
*trans
,
1041 struct btrfs_fs_info
*fs_info
)
1043 struct btrfs_root
*root
;
1044 struct btrfs_root
*tree_root
= fs_info
->tree_root
;
1045 struct extent_buffer
*leaf
;
1047 root
= kzalloc(sizeof(*root
), GFP_NOFS
);
1049 return ERR_PTR(-ENOMEM
);
1051 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
1052 tree_root
->sectorsize
, tree_root
->stripesize
,
1053 root
, fs_info
, BTRFS_TREE_LOG_OBJECTID
);
1055 root
->root_key
.objectid
= BTRFS_TREE_LOG_OBJECTID
;
1056 root
->root_key
.type
= BTRFS_ROOT_ITEM_KEY
;
1057 root
->root_key
.offset
= BTRFS_TREE_LOG_OBJECTID
;
1059 * log trees do not get reference counted because they go away
1060 * before a real commit is actually done. They do store pointers
1061 * to file data extents, and those reference counts still get
1062 * updated (along with back refs to the log tree).
1066 leaf
= btrfs_alloc_free_block(trans
, root
, root
->leafsize
, 0,
1067 BTRFS_TREE_LOG_OBJECTID
, NULL
, 0, 0, 0);
1070 return ERR_CAST(leaf
);
1073 memset_extent_buffer(leaf
, 0, 0, sizeof(struct btrfs_header
));
1074 btrfs_set_header_bytenr(leaf
, leaf
->start
);
1075 btrfs_set_header_generation(leaf
, trans
->transid
);
1076 btrfs_set_header_backref_rev(leaf
, BTRFS_MIXED_BACKREF_REV
);
1077 btrfs_set_header_owner(leaf
, BTRFS_TREE_LOG_OBJECTID
);
1080 write_extent_buffer(root
->node
, root
->fs_info
->fsid
,
1081 (unsigned long)btrfs_header_fsid(root
->node
),
1083 btrfs_mark_buffer_dirty(root
->node
);
1084 btrfs_tree_unlock(root
->node
);
1088 int btrfs_init_log_root_tree(struct btrfs_trans_handle
*trans
,
1089 struct btrfs_fs_info
*fs_info
)
1091 struct btrfs_root
*log_root
;
1093 log_root
= alloc_log_tree(trans
, fs_info
);
1094 if (IS_ERR(log_root
))
1095 return PTR_ERR(log_root
);
1096 WARN_ON(fs_info
->log_root_tree
);
1097 fs_info
->log_root_tree
= log_root
;
1101 int btrfs_add_log_tree(struct btrfs_trans_handle
*trans
,
1102 struct btrfs_root
*root
)
1104 struct btrfs_root
*log_root
;
1105 struct btrfs_inode_item
*inode_item
;
1107 log_root
= alloc_log_tree(trans
, root
->fs_info
);
1108 if (IS_ERR(log_root
))
1109 return PTR_ERR(log_root
);
1111 log_root
->last_trans
= trans
->transid
;
1112 log_root
->root_key
.offset
= root
->root_key
.objectid
;
1114 inode_item
= &log_root
->root_item
.inode
;
1115 inode_item
->generation
= cpu_to_le64(1);
1116 inode_item
->size
= cpu_to_le64(3);
1117 inode_item
->nlink
= cpu_to_le32(1);
1118 inode_item
->nbytes
= cpu_to_le64(root
->leafsize
);
1119 inode_item
->mode
= cpu_to_le32(S_IFDIR
| 0755);
1121 btrfs_set_root_node(&log_root
->root_item
, log_root
->node
);
1123 WARN_ON(root
->log_root
);
1124 root
->log_root
= log_root
;
1125 root
->log_transid
= 0;
1126 root
->last_log_commit
= 0;
1130 struct btrfs_root
*btrfs_read_fs_root_no_radix(struct btrfs_root
*tree_root
,
1131 struct btrfs_key
*location
)
1133 struct btrfs_root
*root
;
1134 struct btrfs_fs_info
*fs_info
= tree_root
->fs_info
;
1135 struct btrfs_path
*path
;
1136 struct extent_buffer
*l
;
1141 root
= kzalloc(sizeof(*root
), GFP_NOFS
);
1143 return ERR_PTR(-ENOMEM
);
1144 if (location
->offset
== (u64
)-1) {
1145 ret
= find_and_setup_root(tree_root
, fs_info
,
1146 location
->objectid
, root
);
1149 return ERR_PTR(ret
);
1154 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
1155 tree_root
->sectorsize
, tree_root
->stripesize
,
1156 root
, fs_info
, location
->objectid
);
1158 path
= btrfs_alloc_path();
1160 ret
= btrfs_search_slot(NULL
, tree_root
, location
, path
, 0, 0);
1163 read_extent_buffer(l
, &root
->root_item
,
1164 btrfs_item_ptr_offset(l
, path
->slots
[0]),
1165 sizeof(root
->root_item
));
1166 memcpy(&root
->root_key
, location
, sizeof(*location
));
1168 btrfs_free_path(path
);
1173 return ERR_PTR(ret
);
1176 generation
= btrfs_root_generation(&root
->root_item
);
1177 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
1178 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
1179 blocksize
, generation
);
1180 root
->commit_root
= btrfs_root_node(root
);
1181 BUG_ON(!root
->node
);
1183 if (location
->objectid
!= BTRFS_TREE_LOG_OBJECTID
)
1189 struct btrfs_root
*btrfs_lookup_fs_root(struct btrfs_fs_info
*fs_info
,
1192 struct btrfs_root
*root
;
1194 if (root_objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1195 return fs_info
->tree_root
;
1196 if (root_objectid
== BTRFS_EXTENT_TREE_OBJECTID
)
1197 return fs_info
->extent_root
;
1199 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
1200 (unsigned long)root_objectid
);
1204 struct btrfs_root
*btrfs_read_fs_root_no_name(struct btrfs_fs_info
*fs_info
,
1205 struct btrfs_key
*location
)
1207 struct btrfs_root
*root
;
1210 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1211 return fs_info
->tree_root
;
1212 if (location
->objectid
== BTRFS_EXTENT_TREE_OBJECTID
)
1213 return fs_info
->extent_root
;
1214 if (location
->objectid
== BTRFS_CHUNK_TREE_OBJECTID
)
1215 return fs_info
->chunk_root
;
1216 if (location
->objectid
== BTRFS_DEV_TREE_OBJECTID
)
1217 return fs_info
->dev_root
;
1218 if (location
->objectid
== BTRFS_CSUM_TREE_OBJECTID
)
1219 return fs_info
->csum_root
;
1221 spin_lock(&fs_info
->fs_roots_radix_lock
);
1222 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
1223 (unsigned long)location
->objectid
);
1224 spin_unlock(&fs_info
->fs_roots_radix_lock
);
1228 root
= btrfs_read_fs_root_no_radix(fs_info
->tree_root
, location
);
1232 set_anon_super(&root
->anon_super
, NULL
);
1234 if (btrfs_root_refs(&root
->root_item
) == 0) {
1239 ret
= btrfs_find_orphan_item(fs_info
->tree_root
, location
->objectid
);
1243 root
->orphan_item_inserted
= 1;
1245 ret
= radix_tree_preload(GFP_NOFS
& ~__GFP_HIGHMEM
);
1249 spin_lock(&fs_info
->fs_roots_radix_lock
);
1250 ret
= radix_tree_insert(&fs_info
->fs_roots_radix
,
1251 (unsigned long)root
->root_key
.objectid
,
1256 spin_unlock(&fs_info
->fs_roots_radix_lock
);
1257 radix_tree_preload_end();
1259 if (ret
== -EEXIST
) {
1266 ret
= btrfs_find_dead_roots(fs_info
->tree_root
,
1267 root
->root_key
.objectid
);
1272 return ERR_PTR(ret
);
1275 struct btrfs_root
*btrfs_read_fs_root(struct btrfs_fs_info
*fs_info
,
1276 struct btrfs_key
*location
,
1277 const char *name
, int namelen
)
1279 return btrfs_read_fs_root_no_name(fs_info
, location
);
1281 struct btrfs_root
*root
;
1284 root
= btrfs_read_fs_root_no_name(fs_info
, location
);
1291 ret
= btrfs_set_root_name(root
, name
, namelen
);
1293 free_extent_buffer(root
->node
);
1295 return ERR_PTR(ret
);
1298 ret
= btrfs_sysfs_add_root(root
);
1300 free_extent_buffer(root
->node
);
1303 return ERR_PTR(ret
);
1310 static int btrfs_congested_fn(void *congested_data
, int bdi_bits
)
1312 struct btrfs_fs_info
*info
= (struct btrfs_fs_info
*)congested_data
;
1314 struct btrfs_device
*device
;
1315 struct backing_dev_info
*bdi
;
1317 list_for_each_entry(device
, &info
->fs_devices
->devices
, dev_list
) {
1320 bdi
= blk_get_backing_dev_info(device
->bdev
);
1321 if (bdi
&& bdi_congested(bdi
, bdi_bits
)) {
1330 * this unplugs every device on the box, and it is only used when page
1333 static void __unplug_io_fn(struct backing_dev_info
*bdi
, struct page
*page
)
1335 struct btrfs_device
*device
;
1336 struct btrfs_fs_info
*info
;
1338 info
= (struct btrfs_fs_info
*)bdi
->unplug_io_data
;
1339 list_for_each_entry(device
, &info
->fs_devices
->devices
, dev_list
) {
1343 bdi
= blk_get_backing_dev_info(device
->bdev
);
1344 if (bdi
->unplug_io_fn
)
1345 bdi
->unplug_io_fn(bdi
, page
);
1349 static void btrfs_unplug_io_fn(struct backing_dev_info
*bdi
, struct page
*page
)
1351 struct inode
*inode
;
1352 struct extent_map_tree
*em_tree
;
1353 struct extent_map
*em
;
1354 struct address_space
*mapping
;
1357 /* the generic O_DIRECT read code does this */
1359 __unplug_io_fn(bdi
, page
);
1364 * page->mapping may change at any time. Get a consistent copy
1365 * and use that for everything below
1368 mapping
= page
->mapping
;
1372 inode
= mapping
->host
;
1375 * don't do the expensive searching for a small number of
1378 if (BTRFS_I(inode
)->root
->fs_info
->fs_devices
->open_devices
<= 2) {
1379 __unplug_io_fn(bdi
, page
);
1383 offset
= page_offset(page
);
1385 em_tree
= &BTRFS_I(inode
)->extent_tree
;
1386 read_lock(&em_tree
->lock
);
1387 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_CACHE_SIZE
);
1388 read_unlock(&em_tree
->lock
);
1390 __unplug_io_fn(bdi
, page
);
1394 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
) {
1395 free_extent_map(em
);
1396 __unplug_io_fn(bdi
, page
);
1399 offset
= offset
- em
->start
;
1400 btrfs_unplug_page(&BTRFS_I(inode
)->root
->fs_info
->mapping_tree
,
1401 em
->block_start
+ offset
, page
);
1402 free_extent_map(em
);
1406 * If this fails, caller must call bdi_destroy() to get rid of the
1409 static int setup_bdi(struct btrfs_fs_info
*info
, struct backing_dev_info
*bdi
)
1413 bdi
->capabilities
= BDI_CAP_MAP_COPY
;
1414 err
= bdi_setup_and_register(bdi
, "btrfs", BDI_CAP_MAP_COPY
);
1418 bdi
->ra_pages
= default_backing_dev_info
.ra_pages
;
1419 bdi
->unplug_io_fn
= btrfs_unplug_io_fn
;
1420 bdi
->unplug_io_data
= info
;
1421 bdi
->congested_fn
= btrfs_congested_fn
;
1422 bdi
->congested_data
= info
;
1426 static int bio_ready_for_csum(struct bio
*bio
)
1432 struct extent_io_tree
*io_tree
= NULL
;
1433 struct bio_vec
*bvec
;
1437 bio_for_each_segment(bvec
, bio
, i
) {
1438 page
= bvec
->bv_page
;
1439 if (page
->private == EXTENT_PAGE_PRIVATE
) {
1440 length
+= bvec
->bv_len
;
1443 if (!page
->private) {
1444 length
+= bvec
->bv_len
;
1447 length
= bvec
->bv_len
;
1448 buf_len
= page
->private >> 2;
1449 start
= page_offset(page
) + bvec
->bv_offset
;
1450 io_tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
1452 /* are we fully contained in this bio? */
1453 if (buf_len
<= length
)
1456 ret
= extent_range_uptodate(io_tree
, start
+ length
,
1457 start
+ buf_len
- 1);
1462 * called by the kthread helper functions to finally call the bio end_io
1463 * functions. This is where read checksum verification actually happens
1465 static void end_workqueue_fn(struct btrfs_work
*work
)
1468 struct end_io_wq
*end_io_wq
;
1469 struct btrfs_fs_info
*fs_info
;
1472 end_io_wq
= container_of(work
, struct end_io_wq
, work
);
1473 bio
= end_io_wq
->bio
;
1474 fs_info
= end_io_wq
->info
;
1476 /* metadata bio reads are special because the whole tree block must
1477 * be checksummed at once. This makes sure the entire block is in
1478 * ram and up to date before trying to verify things. For
1479 * blocksize <= pagesize, it is basically a noop
1481 if (!(bio
->bi_rw
& REQ_WRITE
) && end_io_wq
->metadata
&&
1482 !bio_ready_for_csum(bio
)) {
1483 btrfs_queue_worker(&fs_info
->endio_meta_workers
,
1487 error
= end_io_wq
->error
;
1488 bio
->bi_private
= end_io_wq
->private;
1489 bio
->bi_end_io
= end_io_wq
->end_io
;
1491 bio_endio(bio
, error
);
1494 static int cleaner_kthread(void *arg
)
1496 struct btrfs_root
*root
= arg
;
1499 vfs_check_frozen(root
->fs_info
->sb
, SB_FREEZE_WRITE
);
1501 if (!(root
->fs_info
->sb
->s_flags
& MS_RDONLY
) &&
1502 mutex_trylock(&root
->fs_info
->cleaner_mutex
)) {
1503 btrfs_run_delayed_iputs(root
);
1504 btrfs_clean_old_snapshots(root
);
1505 mutex_unlock(&root
->fs_info
->cleaner_mutex
);
1508 if (freezing(current
)) {
1511 set_current_state(TASK_INTERRUPTIBLE
);
1512 if (!kthread_should_stop())
1514 __set_current_state(TASK_RUNNING
);
1516 } while (!kthread_should_stop());
1520 static int transaction_kthread(void *arg
)
1522 struct btrfs_root
*root
= arg
;
1523 struct btrfs_trans_handle
*trans
;
1524 struct btrfs_transaction
*cur
;
1527 unsigned long delay
;
1532 vfs_check_frozen(root
->fs_info
->sb
, SB_FREEZE_WRITE
);
1533 mutex_lock(&root
->fs_info
->transaction_kthread_mutex
);
1535 spin_lock(&root
->fs_info
->new_trans_lock
);
1536 cur
= root
->fs_info
->running_transaction
;
1538 spin_unlock(&root
->fs_info
->new_trans_lock
);
1542 now
= get_seconds();
1543 if (!cur
->blocked
&&
1544 (now
< cur
->start_time
|| now
- cur
->start_time
< 30)) {
1545 spin_unlock(&root
->fs_info
->new_trans_lock
);
1549 transid
= cur
->transid
;
1550 spin_unlock(&root
->fs_info
->new_trans_lock
);
1552 trans
= btrfs_join_transaction(root
, 1);
1553 BUG_ON(IS_ERR(trans
));
1554 if (transid
== trans
->transid
) {
1555 ret
= btrfs_commit_transaction(trans
, root
);
1558 btrfs_end_transaction(trans
, root
);
1561 wake_up_process(root
->fs_info
->cleaner_kthread
);
1562 mutex_unlock(&root
->fs_info
->transaction_kthread_mutex
);
1564 if (freezing(current
)) {
1567 set_current_state(TASK_INTERRUPTIBLE
);
1568 if (!kthread_should_stop() &&
1569 !btrfs_transaction_blocked(root
->fs_info
))
1570 schedule_timeout(delay
);
1571 __set_current_state(TASK_RUNNING
);
1573 } while (!kthread_should_stop());
1577 struct btrfs_root
*open_ctree(struct super_block
*sb
,
1578 struct btrfs_fs_devices
*fs_devices
,
1588 struct btrfs_key location
;
1589 struct buffer_head
*bh
;
1590 struct btrfs_root
*extent_root
= kzalloc(sizeof(struct btrfs_root
),
1592 struct btrfs_root
*csum_root
= kzalloc(sizeof(struct btrfs_root
),
1594 struct btrfs_root
*tree_root
= btrfs_sb(sb
);
1595 struct btrfs_fs_info
*fs_info
= tree_root
->fs_info
;
1596 struct btrfs_root
*chunk_root
= kzalloc(sizeof(struct btrfs_root
),
1598 struct btrfs_root
*dev_root
= kzalloc(sizeof(struct btrfs_root
),
1600 struct btrfs_root
*log_tree_root
;
1605 struct btrfs_super_block
*disk_super
;
1607 if (!extent_root
|| !tree_root
|| !fs_info
||
1608 !chunk_root
|| !dev_root
|| !csum_root
) {
1613 ret
= init_srcu_struct(&fs_info
->subvol_srcu
);
1619 ret
= setup_bdi(fs_info
, &fs_info
->bdi
);
1625 fs_info
->btree_inode
= new_inode(sb
);
1626 if (!fs_info
->btree_inode
) {
1631 INIT_RADIX_TREE(&fs_info
->fs_roots_radix
, GFP_ATOMIC
);
1632 INIT_LIST_HEAD(&fs_info
->trans_list
);
1633 INIT_LIST_HEAD(&fs_info
->dead_roots
);
1634 INIT_LIST_HEAD(&fs_info
->delayed_iputs
);
1635 INIT_LIST_HEAD(&fs_info
->hashers
);
1636 INIT_LIST_HEAD(&fs_info
->delalloc_inodes
);
1637 INIT_LIST_HEAD(&fs_info
->ordered_operations
);
1638 INIT_LIST_HEAD(&fs_info
->caching_block_groups
);
1639 spin_lock_init(&fs_info
->delalloc_lock
);
1640 spin_lock_init(&fs_info
->new_trans_lock
);
1641 spin_lock_init(&fs_info
->ref_cache_lock
);
1642 spin_lock_init(&fs_info
->fs_roots_radix_lock
);
1643 spin_lock_init(&fs_info
->delayed_iput_lock
);
1645 init_completion(&fs_info
->kobj_unregister
);
1646 fs_info
->tree_root
= tree_root
;
1647 fs_info
->extent_root
= extent_root
;
1648 fs_info
->csum_root
= csum_root
;
1649 fs_info
->chunk_root
= chunk_root
;
1650 fs_info
->dev_root
= dev_root
;
1651 fs_info
->fs_devices
= fs_devices
;
1652 INIT_LIST_HEAD(&fs_info
->dirty_cowonly_roots
);
1653 INIT_LIST_HEAD(&fs_info
->space_info
);
1654 btrfs_mapping_init(&fs_info
->mapping_tree
);
1655 btrfs_init_block_rsv(&fs_info
->global_block_rsv
);
1656 btrfs_init_block_rsv(&fs_info
->delalloc_block_rsv
);
1657 btrfs_init_block_rsv(&fs_info
->trans_block_rsv
);
1658 btrfs_init_block_rsv(&fs_info
->chunk_block_rsv
);
1659 btrfs_init_block_rsv(&fs_info
->empty_block_rsv
);
1660 INIT_LIST_HEAD(&fs_info
->durable_block_rsv_list
);
1661 mutex_init(&fs_info
->durable_block_rsv_mutex
);
1662 atomic_set(&fs_info
->nr_async_submits
, 0);
1663 atomic_set(&fs_info
->async_delalloc_pages
, 0);
1664 atomic_set(&fs_info
->async_submit_draining
, 0);
1665 atomic_set(&fs_info
->nr_async_bios
, 0);
1667 fs_info
->max_inline
= 8192 * 1024;
1668 fs_info
->metadata_ratio
= 0;
1670 fs_info
->thread_pool_size
= min_t(unsigned long,
1671 num_online_cpus() + 2, 8);
1673 INIT_LIST_HEAD(&fs_info
->ordered_extents
);
1674 spin_lock_init(&fs_info
->ordered_extent_lock
);
1676 sb
->s_blocksize
= 4096;
1677 sb
->s_blocksize_bits
= blksize_bits(4096);
1678 sb
->s_bdi
= &fs_info
->bdi
;
1680 fs_info
->btree_inode
->i_ino
= BTRFS_BTREE_INODE_OBJECTID
;
1681 fs_info
->btree_inode
->i_nlink
= 1;
1683 * we set the i_size on the btree inode to the max possible int.
1684 * the real end of the address space is determined by all of
1685 * the devices in the system
1687 fs_info
->btree_inode
->i_size
= OFFSET_MAX
;
1688 fs_info
->btree_inode
->i_mapping
->a_ops
= &btree_aops
;
1689 fs_info
->btree_inode
->i_mapping
->backing_dev_info
= &fs_info
->bdi
;
1691 RB_CLEAR_NODE(&BTRFS_I(fs_info
->btree_inode
)->rb_node
);
1692 extent_io_tree_init(&BTRFS_I(fs_info
->btree_inode
)->io_tree
,
1693 fs_info
->btree_inode
->i_mapping
,
1695 extent_map_tree_init(&BTRFS_I(fs_info
->btree_inode
)->extent_tree
,
1698 BTRFS_I(fs_info
->btree_inode
)->io_tree
.ops
= &btree_extent_io_ops
;
1700 BTRFS_I(fs_info
->btree_inode
)->root
= tree_root
;
1701 memset(&BTRFS_I(fs_info
->btree_inode
)->location
, 0,
1702 sizeof(struct btrfs_key
));
1703 BTRFS_I(fs_info
->btree_inode
)->dummy_inode
= 1;
1704 insert_inode_hash(fs_info
->btree_inode
);
1706 spin_lock_init(&fs_info
->block_group_cache_lock
);
1707 fs_info
->block_group_cache_tree
= RB_ROOT
;
1709 extent_io_tree_init(&fs_info
->freed_extents
[0],
1710 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1711 extent_io_tree_init(&fs_info
->freed_extents
[1],
1712 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1713 fs_info
->pinned_extents
= &fs_info
->freed_extents
[0];
1714 fs_info
->do_barriers
= 1;
1717 mutex_init(&fs_info
->trans_mutex
);
1718 mutex_init(&fs_info
->ordered_operations_mutex
);
1719 mutex_init(&fs_info
->tree_log_mutex
);
1720 mutex_init(&fs_info
->chunk_mutex
);
1721 mutex_init(&fs_info
->transaction_kthread_mutex
);
1722 mutex_init(&fs_info
->cleaner_mutex
);
1723 mutex_init(&fs_info
->volume_mutex
);
1724 init_rwsem(&fs_info
->extent_commit_sem
);
1725 init_rwsem(&fs_info
->cleanup_work_sem
);
1726 init_rwsem(&fs_info
->subvol_sem
);
1728 btrfs_init_free_cluster(&fs_info
->meta_alloc_cluster
);
1729 btrfs_init_free_cluster(&fs_info
->data_alloc_cluster
);
1731 init_waitqueue_head(&fs_info
->transaction_throttle
);
1732 init_waitqueue_head(&fs_info
->transaction_wait
);
1733 init_waitqueue_head(&fs_info
->transaction_blocked_wait
);
1734 init_waitqueue_head(&fs_info
->async_submit_wait
);
1736 __setup_root(4096, 4096, 4096, 4096, tree_root
,
1737 fs_info
, BTRFS_ROOT_TREE_OBJECTID
);
1739 bh
= btrfs_read_dev_super(fs_devices
->latest_bdev
);
1745 memcpy(&fs_info
->super_copy
, bh
->b_data
, sizeof(fs_info
->super_copy
));
1746 memcpy(&fs_info
->super_for_commit
, &fs_info
->super_copy
,
1747 sizeof(fs_info
->super_for_commit
));
1750 memcpy(fs_info
->fsid
, fs_info
->super_copy
.fsid
, BTRFS_FSID_SIZE
);
1752 disk_super
= &fs_info
->super_copy
;
1753 if (!btrfs_super_root(disk_super
))
1756 /* check FS state, whether FS is broken. */
1757 fs_info
->fs_state
|= btrfs_super_flags(disk_super
);
1759 btrfs_check_super_valid(fs_info
, sb
->s_flags
& MS_RDONLY
);
1761 ret
= btrfs_parse_options(tree_root
, options
);
1767 features
= btrfs_super_incompat_flags(disk_super
) &
1768 ~BTRFS_FEATURE_INCOMPAT_SUPP
;
1770 printk(KERN_ERR
"BTRFS: couldn't mount because of "
1771 "unsupported optional features (%Lx).\n",
1772 (unsigned long long)features
);
1777 features
= btrfs_super_incompat_flags(disk_super
);
1778 features
|= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF
;
1779 if (tree_root
->fs_info
->compress_type
& BTRFS_COMPRESS_LZO
)
1780 features
|= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO
;
1781 btrfs_set_super_incompat_flags(disk_super
, features
);
1783 features
= btrfs_super_compat_ro_flags(disk_super
) &
1784 ~BTRFS_FEATURE_COMPAT_RO_SUPP
;
1785 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
1786 printk(KERN_ERR
"BTRFS: couldn't mount RDWR because of "
1787 "unsupported option features (%Lx).\n",
1788 (unsigned long long)features
);
1793 btrfs_init_workers(&fs_info
->generic_worker
,
1794 "genwork", 1, NULL
);
1796 btrfs_init_workers(&fs_info
->workers
, "worker",
1797 fs_info
->thread_pool_size
,
1798 &fs_info
->generic_worker
);
1800 btrfs_init_workers(&fs_info
->delalloc_workers
, "delalloc",
1801 fs_info
->thread_pool_size
,
1802 &fs_info
->generic_worker
);
1804 btrfs_init_workers(&fs_info
->submit_workers
, "submit",
1805 min_t(u64
, fs_devices
->num_devices
,
1806 fs_info
->thread_pool_size
),
1807 &fs_info
->generic_worker
);
1809 /* a higher idle thresh on the submit workers makes it much more
1810 * likely that bios will be send down in a sane order to the
1813 fs_info
->submit_workers
.idle_thresh
= 64;
1815 fs_info
->workers
.idle_thresh
= 16;
1816 fs_info
->workers
.ordered
= 1;
1818 fs_info
->delalloc_workers
.idle_thresh
= 2;
1819 fs_info
->delalloc_workers
.ordered
= 1;
1821 btrfs_init_workers(&fs_info
->fixup_workers
, "fixup", 1,
1822 &fs_info
->generic_worker
);
1823 btrfs_init_workers(&fs_info
->endio_workers
, "endio",
1824 fs_info
->thread_pool_size
,
1825 &fs_info
->generic_worker
);
1826 btrfs_init_workers(&fs_info
->endio_meta_workers
, "endio-meta",
1827 fs_info
->thread_pool_size
,
1828 &fs_info
->generic_worker
);
1829 btrfs_init_workers(&fs_info
->endio_meta_write_workers
,
1830 "endio-meta-write", fs_info
->thread_pool_size
,
1831 &fs_info
->generic_worker
);
1832 btrfs_init_workers(&fs_info
->endio_write_workers
, "endio-write",
1833 fs_info
->thread_pool_size
,
1834 &fs_info
->generic_worker
);
1835 btrfs_init_workers(&fs_info
->endio_freespace_worker
, "freespace-write",
1836 1, &fs_info
->generic_worker
);
1839 * endios are largely parallel and should have a very
1842 fs_info
->endio_workers
.idle_thresh
= 4;
1843 fs_info
->endio_meta_workers
.idle_thresh
= 4;
1845 fs_info
->endio_write_workers
.idle_thresh
= 2;
1846 fs_info
->endio_meta_write_workers
.idle_thresh
= 2;
1848 btrfs_start_workers(&fs_info
->workers
, 1);
1849 btrfs_start_workers(&fs_info
->generic_worker
, 1);
1850 btrfs_start_workers(&fs_info
->submit_workers
, 1);
1851 btrfs_start_workers(&fs_info
->delalloc_workers
, 1);
1852 btrfs_start_workers(&fs_info
->fixup_workers
, 1);
1853 btrfs_start_workers(&fs_info
->endio_workers
, 1);
1854 btrfs_start_workers(&fs_info
->endio_meta_workers
, 1);
1855 btrfs_start_workers(&fs_info
->endio_meta_write_workers
, 1);
1856 btrfs_start_workers(&fs_info
->endio_write_workers
, 1);
1857 btrfs_start_workers(&fs_info
->endio_freespace_worker
, 1);
1859 fs_info
->bdi
.ra_pages
*= btrfs_super_num_devices(disk_super
);
1860 fs_info
->bdi
.ra_pages
= max(fs_info
->bdi
.ra_pages
,
1861 4 * 1024 * 1024 / PAGE_CACHE_SIZE
);
1863 nodesize
= btrfs_super_nodesize(disk_super
);
1864 leafsize
= btrfs_super_leafsize(disk_super
);
1865 sectorsize
= btrfs_super_sectorsize(disk_super
);
1866 stripesize
= btrfs_super_stripesize(disk_super
);
1867 tree_root
->nodesize
= nodesize
;
1868 tree_root
->leafsize
= leafsize
;
1869 tree_root
->sectorsize
= sectorsize
;
1870 tree_root
->stripesize
= stripesize
;
1872 sb
->s_blocksize
= sectorsize
;
1873 sb
->s_blocksize_bits
= blksize_bits(sectorsize
);
1875 if (strncmp((char *)(&disk_super
->magic
), BTRFS_MAGIC
,
1876 sizeof(disk_super
->magic
))) {
1877 printk(KERN_INFO
"btrfs: valid FS not found on %s\n", sb
->s_id
);
1878 goto fail_sb_buffer
;
1881 mutex_lock(&fs_info
->chunk_mutex
);
1882 ret
= btrfs_read_sys_array(tree_root
);
1883 mutex_unlock(&fs_info
->chunk_mutex
);
1885 printk(KERN_WARNING
"btrfs: failed to read the system "
1886 "array on %s\n", sb
->s_id
);
1887 goto fail_sb_buffer
;
1890 blocksize
= btrfs_level_size(tree_root
,
1891 btrfs_super_chunk_root_level(disk_super
));
1892 generation
= btrfs_super_chunk_root_generation(disk_super
);
1894 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
1895 chunk_root
, fs_info
, BTRFS_CHUNK_TREE_OBJECTID
);
1897 chunk_root
->node
= read_tree_block(chunk_root
,
1898 btrfs_super_chunk_root(disk_super
),
1899 blocksize
, generation
);
1900 BUG_ON(!chunk_root
->node
);
1901 if (!test_bit(EXTENT_BUFFER_UPTODATE
, &chunk_root
->node
->bflags
)) {
1902 printk(KERN_WARNING
"btrfs: failed to read chunk root on %s\n",
1904 goto fail_chunk_root
;
1906 btrfs_set_root_node(&chunk_root
->root_item
, chunk_root
->node
);
1907 chunk_root
->commit_root
= btrfs_root_node(chunk_root
);
1909 read_extent_buffer(chunk_root
->node
, fs_info
->chunk_tree_uuid
,
1910 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root
->node
),
1913 mutex_lock(&fs_info
->chunk_mutex
);
1914 ret
= btrfs_read_chunk_tree(chunk_root
);
1915 mutex_unlock(&fs_info
->chunk_mutex
);
1917 printk(KERN_WARNING
"btrfs: failed to read chunk tree on %s\n",
1919 goto fail_chunk_root
;
1922 btrfs_close_extra_devices(fs_devices
);
1924 blocksize
= btrfs_level_size(tree_root
,
1925 btrfs_super_root_level(disk_super
));
1926 generation
= btrfs_super_generation(disk_super
);
1928 tree_root
->node
= read_tree_block(tree_root
,
1929 btrfs_super_root(disk_super
),
1930 blocksize
, generation
);
1931 if (!tree_root
->node
)
1932 goto fail_chunk_root
;
1933 if (!test_bit(EXTENT_BUFFER_UPTODATE
, &tree_root
->node
->bflags
)) {
1934 printk(KERN_WARNING
"btrfs: failed to read tree root on %s\n",
1936 goto fail_tree_root
;
1938 btrfs_set_root_node(&tree_root
->root_item
, tree_root
->node
);
1939 tree_root
->commit_root
= btrfs_root_node(tree_root
);
1941 ret
= find_and_setup_root(tree_root
, fs_info
,
1942 BTRFS_EXTENT_TREE_OBJECTID
, extent_root
);
1944 goto fail_tree_root
;
1945 extent_root
->track_dirty
= 1;
1947 ret
= find_and_setup_root(tree_root
, fs_info
,
1948 BTRFS_DEV_TREE_OBJECTID
, dev_root
);
1950 goto fail_extent_root
;
1951 dev_root
->track_dirty
= 1;
1953 ret
= find_and_setup_root(tree_root
, fs_info
,
1954 BTRFS_CSUM_TREE_OBJECTID
, csum_root
);
1958 csum_root
->track_dirty
= 1;
1960 fs_info
->generation
= generation
;
1961 fs_info
->last_trans_committed
= generation
;
1962 fs_info
->data_alloc_profile
= (u64
)-1;
1963 fs_info
->metadata_alloc_profile
= (u64
)-1;
1964 fs_info
->system_alloc_profile
= fs_info
->metadata_alloc_profile
;
1966 ret
= btrfs_read_block_groups(extent_root
);
1968 printk(KERN_ERR
"Failed to read block groups: %d\n", ret
);
1969 goto fail_block_groups
;
1972 fs_info
->cleaner_kthread
= kthread_run(cleaner_kthread
, tree_root
,
1974 if (IS_ERR(fs_info
->cleaner_kthread
))
1975 goto fail_block_groups
;
1977 fs_info
->transaction_kthread
= kthread_run(transaction_kthread
,
1979 "btrfs-transaction");
1980 if (IS_ERR(fs_info
->transaction_kthread
))
1983 if (!btrfs_test_opt(tree_root
, SSD
) &&
1984 !btrfs_test_opt(tree_root
, NOSSD
) &&
1985 !fs_info
->fs_devices
->rotating
) {
1986 printk(KERN_INFO
"Btrfs detected SSD devices, enabling SSD "
1988 btrfs_set_opt(fs_info
->mount_opt
, SSD
);
1991 /* do not make disk changes in broken FS */
1992 if (btrfs_super_log_root(disk_super
) != 0 &&
1993 !(fs_info
->fs_state
& BTRFS_SUPER_FLAG_ERROR
)) {
1994 u64 bytenr
= btrfs_super_log_root(disk_super
);
1996 if (fs_devices
->rw_devices
== 0) {
1997 printk(KERN_WARNING
"Btrfs log replay required "
2000 goto fail_trans_kthread
;
2003 btrfs_level_size(tree_root
,
2004 btrfs_super_log_root_level(disk_super
));
2006 log_tree_root
= kzalloc(sizeof(struct btrfs_root
), GFP_NOFS
);
2007 if (!log_tree_root
) {
2009 goto fail_trans_kthread
;
2012 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
2013 log_tree_root
, fs_info
, BTRFS_TREE_LOG_OBJECTID
);
2015 log_tree_root
->node
= read_tree_block(tree_root
, bytenr
,
2018 ret
= btrfs_recover_log_trees(log_tree_root
);
2021 if (sb
->s_flags
& MS_RDONLY
) {
2022 ret
= btrfs_commit_super(tree_root
);
2027 ret
= btrfs_find_orphan_roots(tree_root
);
2030 if (!(sb
->s_flags
& MS_RDONLY
)) {
2031 ret
= btrfs_cleanup_fs_roots(fs_info
);
2034 ret
= btrfs_recover_relocation(tree_root
);
2037 "btrfs: failed to recover relocation\n");
2039 goto fail_trans_kthread
;
2043 location
.objectid
= BTRFS_FS_TREE_OBJECTID
;
2044 location
.type
= BTRFS_ROOT_ITEM_KEY
;
2045 location
.offset
= (u64
)-1;
2047 fs_info
->fs_root
= btrfs_read_fs_root_no_name(fs_info
, &location
);
2048 if (!fs_info
->fs_root
)
2049 goto fail_trans_kthread
;
2050 if (IS_ERR(fs_info
->fs_root
)) {
2051 err
= PTR_ERR(fs_info
->fs_root
);
2052 goto fail_trans_kthread
;
2055 if (!(sb
->s_flags
& MS_RDONLY
)) {
2056 down_read(&fs_info
->cleanup_work_sem
);
2057 btrfs_orphan_cleanup(fs_info
->fs_root
);
2058 btrfs_orphan_cleanup(fs_info
->tree_root
);
2059 up_read(&fs_info
->cleanup_work_sem
);
2065 kthread_stop(fs_info
->transaction_kthread
);
2067 kthread_stop(fs_info
->cleaner_kthread
);
2070 * make sure we're done with the btree inode before we stop our
2073 filemap_write_and_wait(fs_info
->btree_inode
->i_mapping
);
2074 invalidate_inode_pages2(fs_info
->btree_inode
->i_mapping
);
2077 btrfs_free_block_groups(fs_info
);
2078 free_extent_buffer(csum_root
->node
);
2079 free_extent_buffer(csum_root
->commit_root
);
2081 free_extent_buffer(dev_root
->node
);
2082 free_extent_buffer(dev_root
->commit_root
);
2084 free_extent_buffer(extent_root
->node
);
2085 free_extent_buffer(extent_root
->commit_root
);
2087 free_extent_buffer(tree_root
->node
);
2088 free_extent_buffer(tree_root
->commit_root
);
2090 free_extent_buffer(chunk_root
->node
);
2091 free_extent_buffer(chunk_root
->commit_root
);
2093 btrfs_stop_workers(&fs_info
->generic_worker
);
2094 btrfs_stop_workers(&fs_info
->fixup_workers
);
2095 btrfs_stop_workers(&fs_info
->delalloc_workers
);
2096 btrfs_stop_workers(&fs_info
->workers
);
2097 btrfs_stop_workers(&fs_info
->endio_workers
);
2098 btrfs_stop_workers(&fs_info
->endio_meta_workers
);
2099 btrfs_stop_workers(&fs_info
->endio_meta_write_workers
);
2100 btrfs_stop_workers(&fs_info
->endio_write_workers
);
2101 btrfs_stop_workers(&fs_info
->endio_freespace_worker
);
2102 btrfs_stop_workers(&fs_info
->submit_workers
);
2104 invalidate_inode_pages2(fs_info
->btree_inode
->i_mapping
);
2105 iput(fs_info
->btree_inode
);
2107 btrfs_close_devices(fs_info
->fs_devices
);
2108 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
2110 bdi_destroy(&fs_info
->bdi
);
2112 cleanup_srcu_struct(&fs_info
->subvol_srcu
);
2120 return ERR_PTR(err
);
2123 static void btrfs_end_buffer_write_sync(struct buffer_head
*bh
, int uptodate
)
2125 char b
[BDEVNAME_SIZE
];
2128 set_buffer_uptodate(bh
);
2130 if (!buffer_eopnotsupp(bh
) && printk_ratelimit()) {
2131 printk(KERN_WARNING
"lost page write due to "
2132 "I/O error on %s\n",
2133 bdevname(bh
->b_bdev
, b
));
2135 /* note, we dont' set_buffer_write_io_error because we have
2136 * our own ways of dealing with the IO errors
2138 clear_buffer_uptodate(bh
);
2144 struct buffer_head
*btrfs_read_dev_super(struct block_device
*bdev
)
2146 struct buffer_head
*bh
;
2147 struct buffer_head
*latest
= NULL
;
2148 struct btrfs_super_block
*super
;
2153 /* we would like to check all the supers, but that would make
2154 * a btrfs mount succeed after a mkfs from a different FS.
2155 * So, we need to add a special mount option to scan for
2156 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
2158 for (i
= 0; i
< 1; i
++) {
2159 bytenr
= btrfs_sb_offset(i
);
2160 if (bytenr
+ 4096 >= i_size_read(bdev
->bd_inode
))
2162 bh
= __bread(bdev
, bytenr
/ 4096, 4096);
2166 super
= (struct btrfs_super_block
*)bh
->b_data
;
2167 if (btrfs_super_bytenr(super
) != bytenr
||
2168 strncmp((char *)(&super
->magic
), BTRFS_MAGIC
,
2169 sizeof(super
->magic
))) {
2174 if (!latest
|| btrfs_super_generation(super
) > transid
) {
2177 transid
= btrfs_super_generation(super
);
2186 * this should be called twice, once with wait == 0 and
2187 * once with wait == 1. When wait == 0 is done, all the buffer heads
2188 * we write are pinned.
2190 * They are released when wait == 1 is done.
2191 * max_mirrors must be the same for both runs, and it indicates how
2192 * many supers on this one device should be written.
2194 * max_mirrors == 0 means to write them all.
2196 static int write_dev_supers(struct btrfs_device
*device
,
2197 struct btrfs_super_block
*sb
,
2198 int do_barriers
, int wait
, int max_mirrors
)
2200 struct buffer_head
*bh
;
2206 int last_barrier
= 0;
2208 if (max_mirrors
== 0)
2209 max_mirrors
= BTRFS_SUPER_MIRROR_MAX
;
2211 /* make sure only the last submit_bh does a barrier */
2213 for (i
= 0; i
< max_mirrors
; i
++) {
2214 bytenr
= btrfs_sb_offset(i
);
2215 if (bytenr
+ BTRFS_SUPER_INFO_SIZE
>=
2216 device
->total_bytes
)
2222 for (i
= 0; i
< max_mirrors
; i
++) {
2223 bytenr
= btrfs_sb_offset(i
);
2224 if (bytenr
+ BTRFS_SUPER_INFO_SIZE
>= device
->total_bytes
)
2228 bh
= __find_get_block(device
->bdev
, bytenr
/ 4096,
2229 BTRFS_SUPER_INFO_SIZE
);
2232 if (!buffer_uptodate(bh
))
2235 /* drop our reference */
2238 /* drop the reference from the wait == 0 run */
2242 btrfs_set_super_bytenr(sb
, bytenr
);
2245 crc
= btrfs_csum_data(NULL
, (char *)sb
+
2246 BTRFS_CSUM_SIZE
, crc
,
2247 BTRFS_SUPER_INFO_SIZE
-
2249 btrfs_csum_final(crc
, sb
->csum
);
2252 * one reference for us, and we leave it for the
2255 bh
= __getblk(device
->bdev
, bytenr
/ 4096,
2256 BTRFS_SUPER_INFO_SIZE
);
2257 memcpy(bh
->b_data
, sb
, BTRFS_SUPER_INFO_SIZE
);
2259 /* one reference for submit_bh */
2262 set_buffer_uptodate(bh
);
2264 bh
->b_end_io
= btrfs_end_buffer_write_sync
;
2267 if (i
== last_barrier
&& do_barriers
&& device
->barriers
) {
2268 ret
= submit_bh(WRITE_BARRIER
, bh
);
2269 if (ret
== -EOPNOTSUPP
) {
2270 printk("btrfs: disabling barriers on dev %s\n",
2272 set_buffer_uptodate(bh
);
2273 device
->barriers
= 0;
2274 /* one reference for submit_bh */
2277 ret
= submit_bh(WRITE_SYNC
, bh
);
2280 ret
= submit_bh(WRITE_SYNC
, bh
);
2286 return errors
< i
? 0 : -1;
2289 int write_all_supers(struct btrfs_root
*root
, int max_mirrors
)
2291 struct list_head
*head
;
2292 struct btrfs_device
*dev
;
2293 struct btrfs_super_block
*sb
;
2294 struct btrfs_dev_item
*dev_item
;
2298 int total_errors
= 0;
2301 max_errors
= btrfs_super_num_devices(&root
->fs_info
->super_copy
) - 1;
2302 do_barriers
= !btrfs_test_opt(root
, NOBARRIER
);
2304 sb
= &root
->fs_info
->super_for_commit
;
2305 dev_item
= &sb
->dev_item
;
2307 mutex_lock(&root
->fs_info
->fs_devices
->device_list_mutex
);
2308 head
= &root
->fs_info
->fs_devices
->devices
;
2309 list_for_each_entry(dev
, head
, dev_list
) {
2314 if (!dev
->in_fs_metadata
|| !dev
->writeable
)
2317 btrfs_set_stack_device_generation(dev_item
, 0);
2318 btrfs_set_stack_device_type(dev_item
, dev
->type
);
2319 btrfs_set_stack_device_id(dev_item
, dev
->devid
);
2320 btrfs_set_stack_device_total_bytes(dev_item
, dev
->total_bytes
);
2321 btrfs_set_stack_device_bytes_used(dev_item
, dev
->bytes_used
);
2322 btrfs_set_stack_device_io_align(dev_item
, dev
->io_align
);
2323 btrfs_set_stack_device_io_width(dev_item
, dev
->io_width
);
2324 btrfs_set_stack_device_sector_size(dev_item
, dev
->sector_size
);
2325 memcpy(dev_item
->uuid
, dev
->uuid
, BTRFS_UUID_SIZE
);
2326 memcpy(dev_item
->fsid
, dev
->fs_devices
->fsid
, BTRFS_UUID_SIZE
);
2328 flags
= btrfs_super_flags(sb
);
2329 btrfs_set_super_flags(sb
, flags
| BTRFS_HEADER_FLAG_WRITTEN
);
2331 ret
= write_dev_supers(dev
, sb
, do_barriers
, 0, max_mirrors
);
2335 if (total_errors
> max_errors
) {
2336 printk(KERN_ERR
"btrfs: %d errors while writing supers\n",
2342 list_for_each_entry(dev
, head
, dev_list
) {
2345 if (!dev
->in_fs_metadata
|| !dev
->writeable
)
2348 ret
= write_dev_supers(dev
, sb
, do_barriers
, 1, max_mirrors
);
2352 mutex_unlock(&root
->fs_info
->fs_devices
->device_list_mutex
);
2353 if (total_errors
> max_errors
) {
2354 printk(KERN_ERR
"btrfs: %d errors while writing supers\n",
2361 int write_ctree_super(struct btrfs_trans_handle
*trans
,
2362 struct btrfs_root
*root
, int max_mirrors
)
2366 ret
= write_all_supers(root
, max_mirrors
);
2370 int btrfs_free_fs_root(struct btrfs_fs_info
*fs_info
, struct btrfs_root
*root
)
2372 spin_lock(&fs_info
->fs_roots_radix_lock
);
2373 radix_tree_delete(&fs_info
->fs_roots_radix
,
2374 (unsigned long)root
->root_key
.objectid
);
2375 spin_unlock(&fs_info
->fs_roots_radix_lock
);
2377 if (btrfs_root_refs(&root
->root_item
) == 0)
2378 synchronize_srcu(&fs_info
->subvol_srcu
);
2384 static void free_fs_root(struct btrfs_root
*root
)
2386 WARN_ON(!RB_EMPTY_ROOT(&root
->inode_tree
));
2387 if (root
->anon_super
.s_dev
) {
2388 down_write(&root
->anon_super
.s_umount
);
2389 kill_anon_super(&root
->anon_super
);
2391 free_extent_buffer(root
->node
);
2392 free_extent_buffer(root
->commit_root
);
2397 static int del_fs_roots(struct btrfs_fs_info
*fs_info
)
2400 struct btrfs_root
*gang
[8];
2403 while (!list_empty(&fs_info
->dead_roots
)) {
2404 gang
[0] = list_entry(fs_info
->dead_roots
.next
,
2405 struct btrfs_root
, root_list
);
2406 list_del(&gang
[0]->root_list
);
2408 if (gang
[0]->in_radix
) {
2409 btrfs_free_fs_root(fs_info
, gang
[0]);
2411 free_extent_buffer(gang
[0]->node
);
2412 free_extent_buffer(gang
[0]->commit_root
);
2418 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
2423 for (i
= 0; i
< ret
; i
++)
2424 btrfs_free_fs_root(fs_info
, gang
[i
]);
2429 int btrfs_cleanup_fs_roots(struct btrfs_fs_info
*fs_info
)
2431 u64 root_objectid
= 0;
2432 struct btrfs_root
*gang
[8];
2437 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
2438 (void **)gang
, root_objectid
,
2443 root_objectid
= gang
[ret
- 1]->root_key
.objectid
+ 1;
2444 for (i
= 0; i
< ret
; i
++) {
2445 root_objectid
= gang
[i
]->root_key
.objectid
;
2446 btrfs_orphan_cleanup(gang
[i
]);
2453 int btrfs_commit_super(struct btrfs_root
*root
)
2455 struct btrfs_trans_handle
*trans
;
2458 mutex_lock(&root
->fs_info
->cleaner_mutex
);
2459 btrfs_run_delayed_iputs(root
);
2460 btrfs_clean_old_snapshots(root
);
2461 mutex_unlock(&root
->fs_info
->cleaner_mutex
);
2463 /* wait until ongoing cleanup work done */
2464 down_write(&root
->fs_info
->cleanup_work_sem
);
2465 up_write(&root
->fs_info
->cleanup_work_sem
);
2467 trans
= btrfs_join_transaction(root
, 1);
2469 return PTR_ERR(trans
);
2470 ret
= btrfs_commit_transaction(trans
, root
);
2472 /* run commit again to drop the original snapshot */
2473 trans
= btrfs_join_transaction(root
, 1);
2475 return PTR_ERR(trans
);
2476 btrfs_commit_transaction(trans
, root
);
2477 ret
= btrfs_write_and_wait_transaction(NULL
, root
);
2480 ret
= write_ctree_super(NULL
, root
, 0);
2484 int close_ctree(struct btrfs_root
*root
)
2486 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
2489 fs_info
->closing
= 1;
2492 btrfs_put_block_group_cache(fs_info
);
2495 * Here come 2 situations when btrfs is broken to flip readonly:
2497 * 1. when btrfs flips readonly somewhere else before
2498 * btrfs_commit_super, sb->s_flags has MS_RDONLY flag,
2499 * and btrfs will skip to write sb directly to keep
2500 * ERROR state on disk.
2502 * 2. when btrfs flips readonly just in btrfs_commit_super,
2503 * and in such case, btrfs cannnot write sb via btrfs_commit_super,
2504 * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag,
2505 * btrfs will cleanup all FS resources first and write sb then.
2507 if (!(fs_info
->sb
->s_flags
& MS_RDONLY
)) {
2508 ret
= btrfs_commit_super(root
);
2510 printk(KERN_ERR
"btrfs: commit super ret %d\n", ret
);
2513 if (fs_info
->fs_state
& BTRFS_SUPER_FLAG_ERROR
) {
2514 ret
= btrfs_error_commit_super(root
);
2516 printk(KERN_ERR
"btrfs: commit super ret %d\n", ret
);
2519 kthread_stop(root
->fs_info
->transaction_kthread
);
2520 kthread_stop(root
->fs_info
->cleaner_kthread
);
2522 fs_info
->closing
= 2;
2525 if (fs_info
->delalloc_bytes
) {
2526 printk(KERN_INFO
"btrfs: at unmount delalloc count %llu\n",
2527 (unsigned long long)fs_info
->delalloc_bytes
);
2529 if (fs_info
->total_ref_cache_size
) {
2530 printk(KERN_INFO
"btrfs: at umount reference cache size %llu\n",
2531 (unsigned long long)fs_info
->total_ref_cache_size
);
2534 free_extent_buffer(fs_info
->extent_root
->node
);
2535 free_extent_buffer(fs_info
->extent_root
->commit_root
);
2536 free_extent_buffer(fs_info
->tree_root
->node
);
2537 free_extent_buffer(fs_info
->tree_root
->commit_root
);
2538 free_extent_buffer(root
->fs_info
->chunk_root
->node
);
2539 free_extent_buffer(root
->fs_info
->chunk_root
->commit_root
);
2540 free_extent_buffer(root
->fs_info
->dev_root
->node
);
2541 free_extent_buffer(root
->fs_info
->dev_root
->commit_root
);
2542 free_extent_buffer(root
->fs_info
->csum_root
->node
);
2543 free_extent_buffer(root
->fs_info
->csum_root
->commit_root
);
2545 btrfs_free_block_groups(root
->fs_info
);
2547 del_fs_roots(fs_info
);
2549 iput(fs_info
->btree_inode
);
2551 btrfs_stop_workers(&fs_info
->generic_worker
);
2552 btrfs_stop_workers(&fs_info
->fixup_workers
);
2553 btrfs_stop_workers(&fs_info
->delalloc_workers
);
2554 btrfs_stop_workers(&fs_info
->workers
);
2555 btrfs_stop_workers(&fs_info
->endio_workers
);
2556 btrfs_stop_workers(&fs_info
->endio_meta_workers
);
2557 btrfs_stop_workers(&fs_info
->endio_meta_write_workers
);
2558 btrfs_stop_workers(&fs_info
->endio_write_workers
);
2559 btrfs_stop_workers(&fs_info
->endio_freespace_worker
);
2560 btrfs_stop_workers(&fs_info
->submit_workers
);
2562 btrfs_close_devices(fs_info
->fs_devices
);
2563 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
2565 bdi_destroy(&fs_info
->bdi
);
2566 cleanup_srcu_struct(&fs_info
->subvol_srcu
);
2568 kfree(fs_info
->extent_root
);
2569 kfree(fs_info
->tree_root
);
2570 kfree(fs_info
->chunk_root
);
2571 kfree(fs_info
->dev_root
);
2572 kfree(fs_info
->csum_root
);
2578 int btrfs_buffer_uptodate(struct extent_buffer
*buf
, u64 parent_transid
)
2581 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
2583 ret
= extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
, buf
,
2588 ret
= verify_parent_transid(&BTRFS_I(btree_inode
)->io_tree
, buf
,
2593 int btrfs_set_buffer_uptodate(struct extent_buffer
*buf
)
2595 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
2596 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
,
2600 void btrfs_mark_buffer_dirty(struct extent_buffer
*buf
)
2602 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
2603 u64 transid
= btrfs_header_generation(buf
);
2604 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
2607 btrfs_assert_tree_locked(buf
);
2608 if (transid
!= root
->fs_info
->generation
) {
2609 printk(KERN_CRIT
"btrfs transid mismatch buffer %llu, "
2610 "found %llu running %llu\n",
2611 (unsigned long long)buf
->start
,
2612 (unsigned long long)transid
,
2613 (unsigned long long)root
->fs_info
->generation
);
2616 was_dirty
= set_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
,
2619 spin_lock(&root
->fs_info
->delalloc_lock
);
2620 root
->fs_info
->dirty_metadata_bytes
+= buf
->len
;
2621 spin_unlock(&root
->fs_info
->delalloc_lock
);
2625 void btrfs_btree_balance_dirty(struct btrfs_root
*root
, unsigned long nr
)
2628 * looks as though older kernels can get into trouble with
2629 * this code, they end up stuck in balance_dirty_pages forever
2632 unsigned long thresh
= 32 * 1024 * 1024;
2634 if (current
->flags
& PF_MEMALLOC
)
2637 num_dirty
= root
->fs_info
->dirty_metadata_bytes
;
2639 if (num_dirty
> thresh
) {
2640 balance_dirty_pages_ratelimited_nr(
2641 root
->fs_info
->btree_inode
->i_mapping
, 1);
2646 int btrfs_read_buffer(struct extent_buffer
*buf
, u64 parent_transid
)
2648 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
2650 ret
= btree_read_extent_buffer_pages(root
, buf
, 0, parent_transid
);
2652 set_bit(EXTENT_BUFFER_UPTODATE
, &buf
->bflags
);
2656 int btree_lock_page_hook(struct page
*page
)
2658 struct inode
*inode
= page
->mapping
->host
;
2659 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2660 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2661 struct extent_buffer
*eb
;
2663 u64 bytenr
= page_offset(page
);
2665 if (page
->private == EXTENT_PAGE_PRIVATE
)
2668 len
= page
->private >> 2;
2669 eb
= find_extent_buffer(io_tree
, bytenr
, len
, GFP_NOFS
);
2673 btrfs_tree_lock(eb
);
2674 btrfs_set_header_flag(eb
, BTRFS_HEADER_FLAG_WRITTEN
);
2676 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY
, &eb
->bflags
)) {
2677 spin_lock(&root
->fs_info
->delalloc_lock
);
2678 if (root
->fs_info
->dirty_metadata_bytes
>= eb
->len
)
2679 root
->fs_info
->dirty_metadata_bytes
-= eb
->len
;
2682 spin_unlock(&root
->fs_info
->delalloc_lock
);
2685 btrfs_tree_unlock(eb
);
2686 free_extent_buffer(eb
);
2692 static void btrfs_check_super_valid(struct btrfs_fs_info
*fs_info
,
2698 if (fs_info
->fs_state
& BTRFS_SUPER_FLAG_ERROR
)
2699 printk(KERN_WARNING
"warning: mount fs with errors, "
2700 "running btrfsck is recommended\n");
2703 int btrfs_error_commit_super(struct btrfs_root
*root
)
2707 mutex_lock(&root
->fs_info
->cleaner_mutex
);
2708 btrfs_run_delayed_iputs(root
);
2709 mutex_unlock(&root
->fs_info
->cleaner_mutex
);
2711 down_write(&root
->fs_info
->cleanup_work_sem
);
2712 up_write(&root
->fs_info
->cleanup_work_sem
);
2714 /* cleanup FS via transaction */
2715 btrfs_cleanup_transaction(root
);
2717 ret
= write_ctree_super(NULL
, root
, 0);
2722 static int btrfs_destroy_ordered_operations(struct btrfs_root
*root
)
2724 struct btrfs_inode
*btrfs_inode
;
2725 struct list_head splice
;
2727 INIT_LIST_HEAD(&splice
);
2729 mutex_lock(&root
->fs_info
->ordered_operations_mutex
);
2730 spin_lock(&root
->fs_info
->ordered_extent_lock
);
2732 list_splice_init(&root
->fs_info
->ordered_operations
, &splice
);
2733 while (!list_empty(&splice
)) {
2734 btrfs_inode
= list_entry(splice
.next
, struct btrfs_inode
,
2735 ordered_operations
);
2737 list_del_init(&btrfs_inode
->ordered_operations
);
2739 btrfs_invalidate_inodes(btrfs_inode
->root
);
2742 spin_unlock(&root
->fs_info
->ordered_extent_lock
);
2743 mutex_unlock(&root
->fs_info
->ordered_operations_mutex
);
2748 static int btrfs_destroy_ordered_extents(struct btrfs_root
*root
)
2750 struct list_head splice
;
2751 struct btrfs_ordered_extent
*ordered
;
2752 struct inode
*inode
;
2754 INIT_LIST_HEAD(&splice
);
2756 spin_lock(&root
->fs_info
->ordered_extent_lock
);
2758 list_splice_init(&root
->fs_info
->ordered_extents
, &splice
);
2759 while (!list_empty(&splice
)) {
2760 ordered
= list_entry(splice
.next
, struct btrfs_ordered_extent
,
2763 list_del_init(&ordered
->root_extent_list
);
2764 atomic_inc(&ordered
->refs
);
2766 /* the inode may be getting freed (in sys_unlink path). */
2767 inode
= igrab(ordered
->inode
);
2769 spin_unlock(&root
->fs_info
->ordered_extent_lock
);
2773 atomic_set(&ordered
->refs
, 1);
2774 btrfs_put_ordered_extent(ordered
);
2776 spin_lock(&root
->fs_info
->ordered_extent_lock
);
2779 spin_unlock(&root
->fs_info
->ordered_extent_lock
);
2784 static int btrfs_destroy_delayed_refs(struct btrfs_transaction
*trans
,
2785 struct btrfs_root
*root
)
2787 struct rb_node
*node
;
2788 struct btrfs_delayed_ref_root
*delayed_refs
;
2789 struct btrfs_delayed_ref_node
*ref
;
2792 delayed_refs
= &trans
->delayed_refs
;
2794 spin_lock(&delayed_refs
->lock
);
2795 if (delayed_refs
->num_entries
== 0) {
2796 printk(KERN_INFO
"delayed_refs has NO entry\n");
2800 node
= rb_first(&delayed_refs
->root
);
2802 ref
= rb_entry(node
, struct btrfs_delayed_ref_node
, rb_node
);
2803 node
= rb_next(node
);
2806 rb_erase(&ref
->rb_node
, &delayed_refs
->root
);
2807 delayed_refs
->num_entries
--;
2809 atomic_set(&ref
->refs
, 1);
2810 if (btrfs_delayed_ref_is_head(ref
)) {
2811 struct btrfs_delayed_ref_head
*head
;
2813 head
= btrfs_delayed_node_to_head(ref
);
2814 mutex_lock(&head
->mutex
);
2815 kfree(head
->extent_op
);
2816 delayed_refs
->num_heads
--;
2817 if (list_empty(&head
->cluster
))
2818 delayed_refs
->num_heads_ready
--;
2819 list_del_init(&head
->cluster
);
2820 mutex_unlock(&head
->mutex
);
2823 spin_unlock(&delayed_refs
->lock
);
2824 btrfs_put_delayed_ref(ref
);
2827 spin_lock(&delayed_refs
->lock
);
2830 spin_unlock(&delayed_refs
->lock
);
2835 static int btrfs_destroy_pending_snapshots(struct btrfs_transaction
*t
)
2837 struct btrfs_pending_snapshot
*snapshot
;
2838 struct list_head splice
;
2840 INIT_LIST_HEAD(&splice
);
2842 list_splice_init(&t
->pending_snapshots
, &splice
);
2844 while (!list_empty(&splice
)) {
2845 snapshot
= list_entry(splice
.next
,
2846 struct btrfs_pending_snapshot
,
2849 list_del_init(&snapshot
->list
);
2857 static int btrfs_destroy_delalloc_inodes(struct btrfs_root
*root
)
2859 struct btrfs_inode
*btrfs_inode
;
2860 struct list_head splice
;
2862 INIT_LIST_HEAD(&splice
);
2864 list_splice_init(&root
->fs_info
->delalloc_inodes
, &splice
);
2866 spin_lock(&root
->fs_info
->delalloc_lock
);
2868 while (!list_empty(&splice
)) {
2869 btrfs_inode
= list_entry(splice
.next
, struct btrfs_inode
,
2872 list_del_init(&btrfs_inode
->delalloc_inodes
);
2874 btrfs_invalidate_inodes(btrfs_inode
->root
);
2877 spin_unlock(&root
->fs_info
->delalloc_lock
);
2882 static int btrfs_destroy_marked_extents(struct btrfs_root
*root
,
2883 struct extent_io_tree
*dirty_pages
,
2888 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
2889 struct extent_buffer
*eb
;
2893 unsigned long index
;
2896 ret
= find_first_extent_bit(dirty_pages
, start
, &start
, &end
,
2901 clear_extent_bits(dirty_pages
, start
, end
, mark
, GFP_NOFS
);
2902 while (start
<= end
) {
2903 index
= start
>> PAGE_CACHE_SHIFT
;
2904 start
= (u64
)(index
+ 1) << PAGE_CACHE_SHIFT
;
2905 page
= find_get_page(btree_inode
->i_mapping
, index
);
2908 offset
= page_offset(page
);
2910 spin_lock(&dirty_pages
->buffer_lock
);
2911 eb
= radix_tree_lookup(
2912 &(&BTRFS_I(page
->mapping
->host
)->io_tree
)->buffer
,
2913 offset
>> PAGE_CACHE_SHIFT
);
2914 spin_unlock(&dirty_pages
->buffer_lock
);
2916 ret
= test_and_clear_bit(EXTENT_BUFFER_DIRTY
,
2918 atomic_set(&eb
->refs
, 1);
2920 if (PageWriteback(page
))
2921 end_page_writeback(page
);
2924 if (PageDirty(page
)) {
2925 clear_page_dirty_for_io(page
);
2926 spin_lock_irq(&page
->mapping
->tree_lock
);
2927 radix_tree_tag_clear(&page
->mapping
->page_tree
,
2929 PAGECACHE_TAG_DIRTY
);
2930 spin_unlock_irq(&page
->mapping
->tree_lock
);
2933 page
->mapping
->a_ops
->invalidatepage(page
, 0);
2941 static int btrfs_destroy_pinned_extent(struct btrfs_root
*root
,
2942 struct extent_io_tree
*pinned_extents
)
2944 struct extent_io_tree
*unpin
;
2949 unpin
= pinned_extents
;
2951 ret
= find_first_extent_bit(unpin
, 0, &start
, &end
,
2957 ret
= btrfs_error_discard_extent(root
, start
, end
+ 1 - start
);
2959 clear_extent_dirty(unpin
, start
, end
, GFP_NOFS
);
2960 btrfs_error_unpin_extent_range(root
, start
, end
);
2967 static int btrfs_cleanup_transaction(struct btrfs_root
*root
)
2969 struct btrfs_transaction
*t
;
2974 mutex_lock(&root
->fs_info
->trans_mutex
);
2975 mutex_lock(&root
->fs_info
->transaction_kthread_mutex
);
2977 list_splice_init(&root
->fs_info
->trans_list
, &list
);
2978 while (!list_empty(&list
)) {
2979 t
= list_entry(list
.next
, struct btrfs_transaction
, list
);
2983 btrfs_destroy_ordered_operations(root
);
2985 btrfs_destroy_ordered_extents(root
);
2987 btrfs_destroy_delayed_refs(t
, root
);
2989 btrfs_block_rsv_release(root
,
2990 &root
->fs_info
->trans_block_rsv
,
2991 t
->dirty_pages
.dirty_bytes
);
2993 /* FIXME: cleanup wait for commit */
2996 if (waitqueue_active(&root
->fs_info
->transaction_blocked_wait
))
2997 wake_up(&root
->fs_info
->transaction_blocked_wait
);
3000 if (waitqueue_active(&root
->fs_info
->transaction_wait
))
3001 wake_up(&root
->fs_info
->transaction_wait
);
3002 mutex_unlock(&root
->fs_info
->trans_mutex
);
3004 mutex_lock(&root
->fs_info
->trans_mutex
);
3006 if (waitqueue_active(&t
->commit_wait
))
3007 wake_up(&t
->commit_wait
);
3008 mutex_unlock(&root
->fs_info
->trans_mutex
);
3010 mutex_lock(&root
->fs_info
->trans_mutex
);
3012 btrfs_destroy_pending_snapshots(t
);
3014 btrfs_destroy_delalloc_inodes(root
);
3016 spin_lock(&root
->fs_info
->new_trans_lock
);
3017 root
->fs_info
->running_transaction
= NULL
;
3018 spin_unlock(&root
->fs_info
->new_trans_lock
);
3020 btrfs_destroy_marked_extents(root
, &t
->dirty_pages
,
3023 btrfs_destroy_pinned_extent(root
,
3024 root
->fs_info
->pinned_extents
);
3027 list_del_init(&t
->list
);
3028 memset(t
, 0, sizeof(*t
));
3029 kmem_cache_free(btrfs_transaction_cachep
, t
);
3032 mutex_unlock(&root
->fs_info
->transaction_kthread_mutex
);
3033 mutex_unlock(&root
->fs_info
->trans_mutex
);
3038 static struct extent_io_ops btree_extent_io_ops
= {
3039 .write_cache_pages_lock_hook
= btree_lock_page_hook
,
3040 .readpage_end_io_hook
= btree_readpage_end_io_hook
,
3041 .submit_bio_hook
= btree_submit_bio_hook
,
3042 /* note we're sharing with inode.c for the merge bio hook */
3043 .merge_bio_hook
= btrfs_merge_bio_hook
,