2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/version.h>
21 #include <linux/blkdev.h>
22 #include <linux/scatterlist.h>
23 #include <linux/swap.h>
24 #include <linux/radix-tree.h>
25 #include <linux/writeback.h>
26 #include <linux/buffer_head.h> // for block_sync_page
27 #include <linux/workqueue.h>
28 #include <linux/kthread.h>
29 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20)
30 # include <linux/freezer.h>
32 # include <linux/sched.h>
37 #include "transaction.h"
38 #include "btrfs_inode.h"
40 #include "print-tree.h"
41 #include "async-thread.h"
43 #include "ref-cache.h"
46 static int check_tree_block(struct btrfs_root
*root
, struct extent_buffer
*buf
)
48 if (extent_buffer_blocknr(buf
) != btrfs_header_blocknr(buf
)) {
49 printk(KERN_CRIT
"buf blocknr(buf) is %llu, header is %llu\n",
50 (unsigned long long)extent_buffer_blocknr(buf
),
51 (unsigned long long)btrfs_header_blocknr(buf
));
58 static struct extent_io_ops btree_extent_io_ops
;
59 static void end_workqueue_fn(struct btrfs_work
*work
);
65 struct btrfs_fs_info
*info
;
68 struct list_head list
;
69 struct btrfs_work work
;
72 struct async_submit_bio
{
75 struct list_head list
;
76 extent_submit_bio_hook_t
*submit_bio_hook
;
79 struct btrfs_work work
;
82 struct extent_map
*btree_get_extent(struct inode
*inode
, struct page
*page
,
83 size_t page_offset
, u64 start
, u64 len
,
86 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
87 struct extent_map
*em
;
90 spin_lock(&em_tree
->lock
);
91 em
= lookup_extent_mapping(em_tree
, start
, len
);
94 BTRFS_I(inode
)->root
->fs_info
->fs_devices
->latest_bdev
;
95 spin_unlock(&em_tree
->lock
);
98 spin_unlock(&em_tree
->lock
);
100 em
= alloc_extent_map(GFP_NOFS
);
102 em
= ERR_PTR(-ENOMEM
);
108 em
->bdev
= BTRFS_I(inode
)->root
->fs_info
->fs_devices
->latest_bdev
;
110 spin_lock(&em_tree
->lock
);
111 ret
= add_extent_mapping(em_tree
, em
);
112 if (ret
== -EEXIST
) {
113 u64 failed_start
= em
->start
;
114 u64 failed_len
= em
->len
;
116 printk("failed to insert %Lu %Lu -> %Lu into tree\n",
117 em
->start
, em
->len
, em
->block_start
);
119 em
= lookup_extent_mapping(em_tree
, start
, len
);
121 printk("after failing, found %Lu %Lu %Lu\n",
122 em
->start
, em
->len
, em
->block_start
);
125 em
= lookup_extent_mapping(em_tree
, failed_start
,
128 printk("double failure lookup gives us "
129 "%Lu %Lu -> %Lu\n", em
->start
,
130 em
->len
, em
->block_start
);
139 spin_unlock(&em_tree
->lock
);
147 u32
btrfs_csum_data(struct btrfs_root
*root
, char *data
, u32 seed
, size_t len
)
149 return btrfs_crc32c(seed
, data
, len
);
152 void btrfs_csum_final(u32 crc
, char *result
)
154 *(__le32
*)result
= ~cpu_to_le32(crc
);
157 static int csum_tree_block(struct btrfs_root
*root
, struct extent_buffer
*buf
,
160 char result
[BTRFS_CRC32_SIZE
];
162 unsigned long cur_len
;
163 unsigned long offset
= BTRFS_CSUM_SIZE
;
164 char *map_token
= NULL
;
166 unsigned long map_start
;
167 unsigned long map_len
;
171 len
= buf
->len
- offset
;
173 err
= map_private_extent_buffer(buf
, offset
, 32,
175 &map_start
, &map_len
, KM_USER0
);
177 printk("failed to map extent buffer! %lu\n",
181 cur_len
= min(len
, map_len
- (offset
- map_start
));
182 crc
= btrfs_csum_data(root
, kaddr
+ offset
- map_start
,
186 unmap_extent_buffer(buf
, map_token
, KM_USER0
);
188 btrfs_csum_final(crc
, result
);
191 int from_this_trans
= 0;
193 if (root
->fs_info
->running_transaction
&&
194 btrfs_header_generation(buf
) ==
195 root
->fs_info
->running_transaction
->transid
)
198 /* FIXME, this is not good */
199 if (memcmp_extent_buffer(buf
, result
, 0, BTRFS_CRC32_SIZE
)) {
202 memcpy(&found
, result
, BTRFS_CRC32_SIZE
);
204 read_extent_buffer(buf
, &val
, 0, BTRFS_CRC32_SIZE
);
205 printk("btrfs: %s checksum verify failed on %llu "
206 "wanted %X found %X from_this_trans %d "
208 root
->fs_info
->sb
->s_id
,
209 buf
->start
, val
, found
, from_this_trans
,
210 btrfs_header_level(buf
));
214 write_extent_buffer(buf
, result
, 0, BTRFS_CRC32_SIZE
);
219 static int verify_parent_transid(struct extent_io_tree
*io_tree
,
220 struct extent_buffer
*eb
, u64 parent_transid
)
224 if (!parent_transid
|| btrfs_header_generation(eb
) == parent_transid
)
227 lock_extent(io_tree
, eb
->start
, eb
->start
+ eb
->len
- 1, GFP_NOFS
);
228 if (extent_buffer_uptodate(io_tree
, eb
) &&
229 btrfs_header_generation(eb
) == parent_transid
) {
233 printk("parent transid verify failed on %llu wanted %llu found %llu\n",
234 (unsigned long long)eb
->start
,
235 (unsigned long long)parent_transid
,
236 (unsigned long long)btrfs_header_generation(eb
));
238 clear_extent_buffer_uptodate(io_tree
, eb
);
240 unlock_extent(io_tree
, eb
->start
, eb
->start
+ eb
->len
- 1,
246 static int btree_read_extent_buffer_pages(struct btrfs_root
*root
,
247 struct extent_buffer
*eb
,
248 u64 start
, u64 parent_transid
)
250 struct extent_io_tree
*io_tree
;
255 io_tree
= &BTRFS_I(root
->fs_info
->btree_inode
)->io_tree
;
257 ret
= read_extent_buffer_pages(io_tree
, eb
, start
, 1,
258 btree_get_extent
, mirror_num
);
260 !verify_parent_transid(io_tree
, eb
, parent_transid
))
263 num_copies
= btrfs_num_copies(&root
->fs_info
->mapping_tree
,
269 if (mirror_num
> num_copies
)
275 int csum_dirty_buffer(struct btrfs_root
*root
, struct page
*page
)
277 struct extent_io_tree
*tree
;
278 u64 start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
282 struct extent_buffer
*eb
;
285 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
287 if (page
->private == EXTENT_PAGE_PRIVATE
)
291 len
= page
->private >> 2;
295 eb
= alloc_extent_buffer(tree
, start
, len
, page
, GFP_NOFS
);
296 ret
= btree_read_extent_buffer_pages(root
, eb
, start
+ PAGE_CACHE_SIZE
,
297 btrfs_header_generation(eb
));
299 found_start
= btrfs_header_bytenr(eb
);
300 if (found_start
!= start
) {
301 printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
302 start
, found_start
, len
);
306 if (eb
->first_page
!= page
) {
307 printk("bad first page %lu %lu\n", eb
->first_page
->index
,
312 if (!PageUptodate(page
)) {
313 printk("csum not up to date page %lu\n", page
->index
);
317 found_level
= btrfs_header_level(eb
);
318 spin_lock(&root
->fs_info
->hash_lock
);
319 btrfs_set_header_flag(eb
, BTRFS_HEADER_FLAG_WRITTEN
);
320 spin_unlock(&root
->fs_info
->hash_lock
);
321 csum_tree_block(root
, eb
, 0);
323 free_extent_buffer(eb
);
328 static int btree_writepage_io_hook(struct page
*page
, u64 start
, u64 end
)
330 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
332 csum_dirty_buffer(root
, page
);
336 int btree_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
337 struct extent_state
*state
)
339 struct extent_io_tree
*tree
;
343 struct extent_buffer
*eb
;
344 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
347 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
348 if (page
->private == EXTENT_PAGE_PRIVATE
)
352 len
= page
->private >> 2;
356 eb
= alloc_extent_buffer(tree
, start
, len
, page
, GFP_NOFS
);
358 found_start
= btrfs_header_bytenr(eb
);
359 if (found_start
!= start
) {
363 if (eb
->first_page
!= page
) {
364 printk("bad first page %lu %lu\n", eb
->first_page
->index
,
370 if (memcmp_extent_buffer(eb
, root
->fs_info
->fsid
,
371 (unsigned long)btrfs_header_fsid(eb
),
373 printk("bad fsid on block %Lu\n", eb
->start
);
377 found_level
= btrfs_header_level(eb
);
379 ret
= csum_tree_block(root
, eb
, 1);
383 end
= min_t(u64
, eb
->len
, PAGE_CACHE_SIZE
);
384 end
= eb
->start
+ end
- 1;
386 free_extent_buffer(eb
);
391 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
392 static void end_workqueue_bio(struct bio
*bio
, int err
)
394 static int end_workqueue_bio(struct bio
*bio
,
395 unsigned int bytes_done
, int err
)
398 struct end_io_wq
*end_io_wq
= bio
->bi_private
;
399 struct btrfs_fs_info
*fs_info
;
401 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
406 fs_info
= end_io_wq
->info
;
407 end_io_wq
->error
= err
;
408 end_io_wq
->work
.func
= end_workqueue_fn
;
409 end_io_wq
->work
.flags
= 0;
410 if (bio
->bi_rw
& (1 << BIO_RW
))
411 btrfs_queue_worker(&fs_info
->endio_write_workers
,
414 btrfs_queue_worker(&fs_info
->endio_workers
, &end_io_wq
->work
);
416 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
421 int btrfs_bio_wq_end_io(struct btrfs_fs_info
*info
, struct bio
*bio
,
424 struct end_io_wq
*end_io_wq
;
425 end_io_wq
= kmalloc(sizeof(*end_io_wq
), GFP_NOFS
);
429 end_io_wq
->private = bio
->bi_private
;
430 end_io_wq
->end_io
= bio
->bi_end_io
;
431 end_io_wq
->info
= info
;
432 end_io_wq
->error
= 0;
433 end_io_wq
->bio
= bio
;
434 end_io_wq
->metadata
= metadata
;
436 bio
->bi_private
= end_io_wq
;
437 bio
->bi_end_io
= end_workqueue_bio
;
441 static void run_one_async_submit(struct btrfs_work
*work
)
443 struct btrfs_fs_info
*fs_info
;
444 struct async_submit_bio
*async
;
446 async
= container_of(work
, struct async_submit_bio
, work
);
447 fs_info
= BTRFS_I(async
->inode
)->root
->fs_info
;
448 atomic_dec(&fs_info
->nr_async_submits
);
449 async
->submit_bio_hook(async
->inode
, async
->rw
, async
->bio
,
454 int btrfs_wq_submit_bio(struct btrfs_fs_info
*fs_info
, struct inode
*inode
,
455 int rw
, struct bio
*bio
, int mirror_num
,
456 extent_submit_bio_hook_t
*submit_bio_hook
)
458 struct async_submit_bio
*async
;
460 async
= kmalloc(sizeof(*async
), GFP_NOFS
);
464 async
->inode
= inode
;
467 async
->mirror_num
= mirror_num
;
468 async
->submit_bio_hook
= submit_bio_hook
;
469 async
->work
.func
= run_one_async_submit
;
470 async
->work
.flags
= 0;
471 atomic_inc(&fs_info
->nr_async_submits
);
472 btrfs_queue_worker(&fs_info
->workers
, &async
->work
);
476 static int __btree_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
479 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
483 offset
= bio
->bi_sector
<< 9;
486 * when we're called for a write, we're already in the async
487 * submission context. Just jump ingo btrfs_map_bio
489 if (rw
& (1 << BIO_RW
)) {
490 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
,
495 * called for a read, do the setup so that checksum validation
496 * can happen in the async kernel threads
498 ret
= btrfs_bio_wq_end_io(root
->fs_info
, bio
, 1);
501 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
, mirror_num
, 1);
504 static int btree_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
508 * kthread helpers are used to submit writes so that checksumming
509 * can happen in parallel across all CPUs
511 if (!(rw
& (1 << BIO_RW
))) {
512 return __btree_submit_bio_hook(inode
, rw
, bio
, mirror_num
);
514 return btrfs_wq_submit_bio(BTRFS_I(inode
)->root
->fs_info
,
515 inode
, rw
, bio
, mirror_num
,
516 __btree_submit_bio_hook
);
519 static int btree_writepage(struct page
*page
, struct writeback_control
*wbc
)
521 struct extent_io_tree
*tree
;
522 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
523 return extent_write_full_page(tree
, page
, btree_get_extent
, wbc
);
526 static int btree_writepages(struct address_space
*mapping
,
527 struct writeback_control
*wbc
)
529 struct extent_io_tree
*tree
;
530 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
531 if (wbc
->sync_mode
== WB_SYNC_NONE
) {
534 unsigned long thresh
= 96 * 1024 * 1024;
536 if (wbc
->for_kupdate
)
539 if (current_is_pdflush()) {
540 thresh
= 96 * 1024 * 1024;
542 thresh
= 8 * 1024 * 1024;
544 num_dirty
= count_range_bits(tree
, &start
, (u64
)-1,
545 thresh
, EXTENT_DIRTY
);
546 if (num_dirty
< thresh
) {
550 return extent_writepages(tree
, mapping
, btree_get_extent
, wbc
);
553 int btree_readpage(struct file
*file
, struct page
*page
)
555 struct extent_io_tree
*tree
;
556 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
557 return extent_read_full_page(tree
, page
, btree_get_extent
);
560 static int btree_releasepage(struct page
*page
, gfp_t gfp_flags
)
562 struct extent_io_tree
*tree
;
563 struct extent_map_tree
*map
;
566 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
567 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
569 ret
= try_release_extent_state(map
, tree
, page
, gfp_flags
);
574 ret
= try_release_extent_buffer(tree
, page
);
576 ClearPagePrivate(page
);
577 set_page_private(page
, 0);
578 page_cache_release(page
);
584 static void btree_invalidatepage(struct page
*page
, unsigned long offset
)
586 struct extent_io_tree
*tree
;
587 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
588 extent_invalidatepage(tree
, page
, offset
);
589 btree_releasepage(page
, GFP_NOFS
);
590 if (PagePrivate(page
)) {
591 printk("warning page private not zero on page %Lu\n",
593 ClearPagePrivate(page
);
594 set_page_private(page
, 0);
595 page_cache_release(page
);
600 static int btree_writepage(struct page
*page
, struct writeback_control
*wbc
)
602 struct buffer_head
*bh
;
603 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
604 struct buffer_head
*head
;
605 if (!page_has_buffers(page
)) {
606 create_empty_buffers(page
, root
->fs_info
->sb
->s_blocksize
,
607 (1 << BH_Dirty
)|(1 << BH_Uptodate
));
609 head
= page_buffers(page
);
612 if (buffer_dirty(bh
))
613 csum_tree_block(root
, bh
, 0);
614 bh
= bh
->b_this_page
;
615 } while (bh
!= head
);
616 return block_write_full_page(page
, btree_get_block
, wbc
);
620 static struct address_space_operations btree_aops
= {
621 .readpage
= btree_readpage
,
622 .writepage
= btree_writepage
,
623 .writepages
= btree_writepages
,
624 .releasepage
= btree_releasepage
,
625 .invalidatepage
= btree_invalidatepage
,
626 .sync_page
= block_sync_page
,
629 int readahead_tree_block(struct btrfs_root
*root
, u64 bytenr
, u32 blocksize
,
632 struct extent_buffer
*buf
= NULL
;
633 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
636 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
639 read_extent_buffer_pages(&BTRFS_I(btree_inode
)->io_tree
,
640 buf
, 0, 0, btree_get_extent
, 0);
641 free_extent_buffer(buf
);
645 struct extent_buffer
*btrfs_find_tree_block(struct btrfs_root
*root
,
646 u64 bytenr
, u32 blocksize
)
648 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
649 struct extent_buffer
*eb
;
650 eb
= find_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
651 bytenr
, blocksize
, GFP_NOFS
);
655 struct extent_buffer
*btrfs_find_create_tree_block(struct btrfs_root
*root
,
656 u64 bytenr
, u32 blocksize
)
658 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
659 struct extent_buffer
*eb
;
661 eb
= alloc_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
662 bytenr
, blocksize
, NULL
, GFP_NOFS
);
667 struct extent_buffer
*read_tree_block(struct btrfs_root
*root
, u64 bytenr
,
668 u32 blocksize
, u64 parent_transid
)
670 struct extent_buffer
*buf
= NULL
;
671 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
672 struct extent_io_tree
*io_tree
;
675 io_tree
= &BTRFS_I(btree_inode
)->io_tree
;
677 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
681 ret
= btree_read_extent_buffer_pages(root
, buf
, 0, parent_transid
);
684 buf
->flags
|= EXTENT_UPTODATE
;
690 int clean_tree_block(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
691 struct extent_buffer
*buf
)
693 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
694 if (btrfs_header_generation(buf
) ==
695 root
->fs_info
->running_transaction
->transid
) {
696 WARN_ON(!btrfs_tree_locked(buf
));
697 clear_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
,
703 int wait_on_tree_block_writeback(struct btrfs_root
*root
,
704 struct extent_buffer
*buf
)
706 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
707 wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode
)->io_tree
,
712 static int __setup_root(u32 nodesize
, u32 leafsize
, u32 sectorsize
,
713 u32 stripesize
, struct btrfs_root
*root
,
714 struct btrfs_fs_info
*fs_info
,
719 root
->commit_root
= NULL
;
720 root
->ref_tree
= NULL
;
721 root
->sectorsize
= sectorsize
;
722 root
->nodesize
= nodesize
;
723 root
->leafsize
= leafsize
;
724 root
->stripesize
= stripesize
;
726 root
->track_dirty
= 0;
728 root
->fs_info
= fs_info
;
729 root
->objectid
= objectid
;
730 root
->last_trans
= 0;
731 root
->highest_inode
= 0;
732 root
->last_inode_alloc
= 0;
736 INIT_LIST_HEAD(&root
->dirty_list
);
737 INIT_LIST_HEAD(&root
->orphan_list
);
738 INIT_LIST_HEAD(&root
->dead_list
);
739 spin_lock_init(&root
->node_lock
);
740 spin_lock_init(&root
->list_lock
);
741 mutex_init(&root
->objectid_mutex
);
743 btrfs_leaf_ref_tree_init(&root
->ref_tree_struct
);
744 root
->ref_tree
= &root
->ref_tree_struct
;
746 memset(&root
->root_key
, 0, sizeof(root
->root_key
));
747 memset(&root
->root_item
, 0, sizeof(root
->root_item
));
748 memset(&root
->defrag_progress
, 0, sizeof(root
->defrag_progress
));
749 memset(&root
->root_kobj
, 0, sizeof(root
->root_kobj
));
750 root
->defrag_trans_start
= fs_info
->generation
;
751 init_completion(&root
->kobj_unregister
);
752 root
->defrag_running
= 0;
753 root
->defrag_level
= 0;
754 root
->root_key
.objectid
= objectid
;
758 static int find_and_setup_root(struct btrfs_root
*tree_root
,
759 struct btrfs_fs_info
*fs_info
,
761 struct btrfs_root
*root
)
766 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
767 tree_root
->sectorsize
, tree_root
->stripesize
,
768 root
, fs_info
, objectid
);
769 ret
= btrfs_find_last_root(tree_root
, objectid
,
770 &root
->root_item
, &root
->root_key
);
773 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
774 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
780 struct btrfs_root
*btrfs_read_fs_root_no_radix(struct btrfs_fs_info
*fs_info
,
781 struct btrfs_key
*location
)
783 struct btrfs_root
*root
;
784 struct btrfs_root
*tree_root
= fs_info
->tree_root
;
785 struct btrfs_path
*path
;
786 struct extent_buffer
*l
;
791 root
= kzalloc(sizeof(*root
), GFP_NOFS
);
793 return ERR_PTR(-ENOMEM
);
794 if (location
->offset
== (u64
)-1) {
795 ret
= find_and_setup_root(tree_root
, fs_info
,
796 location
->objectid
, root
);
804 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
805 tree_root
->sectorsize
, tree_root
->stripesize
,
806 root
, fs_info
, location
->objectid
);
808 path
= btrfs_alloc_path();
810 ret
= btrfs_search_slot(NULL
, tree_root
, location
, path
, 0, 0);
817 read_extent_buffer(l
, &root
->root_item
,
818 btrfs_item_ptr_offset(l
, path
->slots
[0]),
819 sizeof(root
->root_item
));
820 memcpy(&root
->root_key
, location
, sizeof(*location
));
823 btrfs_release_path(root
, path
);
824 btrfs_free_path(path
);
829 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
830 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
835 ret
= btrfs_find_highest_inode(root
, &highest_inode
);
837 root
->highest_inode
= highest_inode
;
838 root
->last_inode_alloc
= highest_inode
;
843 struct btrfs_root
*btrfs_lookup_fs_root(struct btrfs_fs_info
*fs_info
,
846 struct btrfs_root
*root
;
848 if (root_objectid
== BTRFS_ROOT_TREE_OBJECTID
)
849 return fs_info
->tree_root
;
850 if (root_objectid
== BTRFS_EXTENT_TREE_OBJECTID
)
851 return fs_info
->extent_root
;
853 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
854 (unsigned long)root_objectid
);
858 struct btrfs_root
*btrfs_read_fs_root_no_name(struct btrfs_fs_info
*fs_info
,
859 struct btrfs_key
*location
)
861 struct btrfs_root
*root
;
864 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
865 return fs_info
->tree_root
;
866 if (location
->objectid
== BTRFS_EXTENT_TREE_OBJECTID
)
867 return fs_info
->extent_root
;
868 if (location
->objectid
== BTRFS_CHUNK_TREE_OBJECTID
)
869 return fs_info
->chunk_root
;
870 if (location
->objectid
== BTRFS_DEV_TREE_OBJECTID
)
871 return fs_info
->dev_root
;
873 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
874 (unsigned long)location
->objectid
);
878 root
= btrfs_read_fs_root_no_radix(fs_info
, location
);
881 ret
= radix_tree_insert(&fs_info
->fs_roots_radix
,
882 (unsigned long)root
->root_key
.objectid
,
885 free_extent_buffer(root
->node
);
889 ret
= btrfs_find_dead_roots(fs_info
->tree_root
,
890 root
->root_key
.objectid
, root
);
896 struct btrfs_root
*btrfs_read_fs_root(struct btrfs_fs_info
*fs_info
,
897 struct btrfs_key
*location
,
898 const char *name
, int namelen
)
900 struct btrfs_root
*root
;
903 root
= btrfs_read_fs_root_no_name(fs_info
, location
);
910 ret
= btrfs_set_root_name(root
, name
, namelen
);
912 free_extent_buffer(root
->node
);
917 ret
= btrfs_sysfs_add_root(root
);
919 free_extent_buffer(root
->node
);
928 static int add_hasher(struct btrfs_fs_info
*info
, char *type
) {
929 struct btrfs_hasher
*hasher
;
931 hasher
= kmalloc(sizeof(*hasher
), GFP_NOFS
);
934 hasher
->hash_tfm
= crypto_alloc_hash(type
, 0, CRYPTO_ALG_ASYNC
);
935 if (!hasher
->hash_tfm
) {
939 spin_lock(&info
->hash_lock
);
940 list_add(&hasher
->list
, &info
->hashers
);
941 spin_unlock(&info
->hash_lock
);
946 static int btrfs_congested_fn(void *congested_data
, int bdi_bits
)
948 struct btrfs_fs_info
*info
= (struct btrfs_fs_info
*)congested_data
;
950 int limit
= 256 * info
->fs_devices
->open_devices
;
951 struct list_head
*cur
;
952 struct btrfs_device
*device
;
953 struct backing_dev_info
*bdi
;
955 if ((bdi_bits
& (1 << BDI_write_congested
)) &&
956 atomic_read(&info
->nr_async_submits
) > limit
) {
960 list_for_each(cur
, &info
->fs_devices
->devices
) {
961 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
964 bdi
= blk_get_backing_dev_info(device
->bdev
);
965 if (bdi
&& bdi_congested(bdi
, bdi_bits
)) {
974 * this unplugs every device on the box, and it is only used when page
977 static void __unplug_io_fn(struct backing_dev_info
*bdi
, struct page
*page
)
979 struct list_head
*cur
;
980 struct btrfs_device
*device
;
981 struct btrfs_fs_info
*info
;
983 info
= (struct btrfs_fs_info
*)bdi
->unplug_io_data
;
984 list_for_each(cur
, &info
->fs_devices
->devices
) {
985 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
986 bdi
= blk_get_backing_dev_info(device
->bdev
);
987 if (bdi
->unplug_io_fn
) {
988 bdi
->unplug_io_fn(bdi
, page
);
993 void btrfs_unplug_io_fn(struct backing_dev_info
*bdi
, struct page
*page
)
996 struct extent_map_tree
*em_tree
;
997 struct extent_map
*em
;
998 struct address_space
*mapping
;
1001 /* the generic O_DIRECT read code does this */
1003 __unplug_io_fn(bdi
, page
);
1008 * page->mapping may change at any time. Get a consistent copy
1009 * and use that for everything below
1012 mapping
= page
->mapping
;
1016 inode
= mapping
->host
;
1017 offset
= page_offset(page
);
1019 em_tree
= &BTRFS_I(inode
)->extent_tree
;
1020 spin_lock(&em_tree
->lock
);
1021 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_CACHE_SIZE
);
1022 spin_unlock(&em_tree
->lock
);
1024 __unplug_io_fn(bdi
, page
);
1028 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
) {
1029 free_extent_map(em
);
1030 __unplug_io_fn(bdi
, page
);
1033 offset
= offset
- em
->start
;
1034 btrfs_unplug_page(&BTRFS_I(inode
)->root
->fs_info
->mapping_tree
,
1035 em
->block_start
+ offset
, page
);
1036 free_extent_map(em
);
1039 static int setup_bdi(struct btrfs_fs_info
*info
, struct backing_dev_info
*bdi
)
1041 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
1044 bdi
->ra_pages
= default_backing_dev_info
.ra_pages
;
1046 bdi
->capabilities
= default_backing_dev_info
.capabilities
;
1047 bdi
->unplug_io_fn
= btrfs_unplug_io_fn
;
1048 bdi
->unplug_io_data
= info
;
1049 bdi
->congested_fn
= btrfs_congested_fn
;
1050 bdi
->congested_data
= info
;
1054 static int bio_ready_for_csum(struct bio
*bio
)
1060 struct extent_io_tree
*io_tree
= NULL
;
1061 struct btrfs_fs_info
*info
= NULL
;
1062 struct bio_vec
*bvec
;
1066 bio_for_each_segment(bvec
, bio
, i
) {
1067 page
= bvec
->bv_page
;
1068 if (page
->private == EXTENT_PAGE_PRIVATE
) {
1069 length
+= bvec
->bv_len
;
1072 if (!page
->private) {
1073 length
+= bvec
->bv_len
;
1076 length
= bvec
->bv_len
;
1077 buf_len
= page
->private >> 2;
1078 start
= page_offset(page
) + bvec
->bv_offset
;
1079 io_tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
1080 info
= BTRFS_I(page
->mapping
->host
)->root
->fs_info
;
1082 /* are we fully contained in this bio? */
1083 if (buf_len
<= length
)
1086 ret
= extent_range_uptodate(io_tree
, start
+ length
,
1087 start
+ buf_len
- 1);
1094 * called by the kthread helper functions to finally call the bio end_io
1095 * functions. This is where read checksum verification actually happens
1097 static void end_workqueue_fn(struct btrfs_work
*work
)
1100 struct end_io_wq
*end_io_wq
;
1101 struct btrfs_fs_info
*fs_info
;
1104 end_io_wq
= container_of(work
, struct end_io_wq
, work
);
1105 bio
= end_io_wq
->bio
;
1106 fs_info
= end_io_wq
->info
;
1108 /* metadata bios are special because the whole tree block must
1109 * be checksummed at once. This makes sure the entire block is in
1110 * ram and up to date before trying to verify things. For
1111 * blocksize <= pagesize, it is basically a noop
1113 if (end_io_wq
->metadata
&& !bio_ready_for_csum(bio
)) {
1114 btrfs_queue_worker(&fs_info
->endio_workers
,
1118 error
= end_io_wq
->error
;
1119 bio
->bi_private
= end_io_wq
->private;
1120 bio
->bi_end_io
= end_io_wq
->end_io
;
1122 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
1123 bio_endio(bio
, bio
->bi_size
, error
);
1125 bio_endio(bio
, error
);
1129 static int cleaner_kthread(void *arg
)
1131 struct btrfs_root
*root
= arg
;
1135 if (root
->fs_info
->closing
)
1138 vfs_check_frozen(root
->fs_info
->sb
, SB_FREEZE_WRITE
);
1139 mutex_lock(&root
->fs_info
->cleaner_mutex
);
1140 btrfs_clean_old_snapshots(root
);
1141 mutex_unlock(&root
->fs_info
->cleaner_mutex
);
1143 if (freezing(current
)) {
1147 if (root
->fs_info
->closing
)
1149 set_current_state(TASK_INTERRUPTIBLE
);
1151 __set_current_state(TASK_RUNNING
);
1153 } while (!kthread_should_stop());
1157 static int transaction_kthread(void *arg
)
1159 struct btrfs_root
*root
= arg
;
1160 struct btrfs_trans_handle
*trans
;
1161 struct btrfs_transaction
*cur
;
1163 unsigned long delay
;
1168 if (root
->fs_info
->closing
)
1172 vfs_check_frozen(root
->fs_info
->sb
, SB_FREEZE_WRITE
);
1173 mutex_lock(&root
->fs_info
->transaction_kthread_mutex
);
1175 if (root
->fs_info
->total_ref_cache_size
> 20 * 1024 * 1024) {
1176 printk("btrfs: total reference cache size %Lu\n",
1177 root
->fs_info
->total_ref_cache_size
);
1180 mutex_lock(&root
->fs_info
->trans_mutex
);
1181 cur
= root
->fs_info
->running_transaction
;
1183 mutex_unlock(&root
->fs_info
->trans_mutex
);
1187 now
= get_seconds();
1188 if (now
< cur
->start_time
|| now
- cur
->start_time
< 30) {
1189 mutex_unlock(&root
->fs_info
->trans_mutex
);
1193 mutex_unlock(&root
->fs_info
->trans_mutex
);
1194 trans
= btrfs_start_transaction(root
, 1);
1195 ret
= btrfs_commit_transaction(trans
, root
);
1197 wake_up_process(root
->fs_info
->cleaner_kthread
);
1198 mutex_unlock(&root
->fs_info
->transaction_kthread_mutex
);
1200 if (freezing(current
)) {
1203 if (root
->fs_info
->closing
)
1205 set_current_state(TASK_INTERRUPTIBLE
);
1206 schedule_timeout(delay
);
1207 __set_current_state(TASK_RUNNING
);
1209 } while (!kthread_should_stop());
1213 struct btrfs_root
*open_ctree(struct super_block
*sb
,
1214 struct btrfs_fs_devices
*fs_devices
,
1222 struct buffer_head
*bh
;
1223 struct btrfs_root
*extent_root
= kmalloc(sizeof(struct btrfs_root
),
1225 struct btrfs_root
*tree_root
= kmalloc(sizeof(struct btrfs_root
),
1227 struct btrfs_fs_info
*fs_info
= kzalloc(sizeof(*fs_info
),
1229 struct btrfs_root
*chunk_root
= kmalloc(sizeof(struct btrfs_root
),
1231 struct btrfs_root
*dev_root
= kmalloc(sizeof(struct btrfs_root
),
1236 struct btrfs_super_block
*disk_super
;
1238 if (!extent_root
|| !tree_root
|| !fs_info
) {
1242 INIT_RADIX_TREE(&fs_info
->fs_roots_radix
, GFP_NOFS
);
1243 INIT_LIST_HEAD(&fs_info
->trans_list
);
1244 INIT_LIST_HEAD(&fs_info
->dead_roots
);
1245 INIT_LIST_HEAD(&fs_info
->hashers
);
1246 spin_lock_init(&fs_info
->hash_lock
);
1247 spin_lock_init(&fs_info
->delalloc_lock
);
1248 spin_lock_init(&fs_info
->new_trans_lock
);
1249 spin_lock_init(&fs_info
->ref_cache_lock
);
1251 init_completion(&fs_info
->kobj_unregister
);
1252 fs_info
->tree_root
= tree_root
;
1253 fs_info
->extent_root
= extent_root
;
1254 fs_info
->chunk_root
= chunk_root
;
1255 fs_info
->dev_root
= dev_root
;
1256 fs_info
->fs_devices
= fs_devices
;
1257 INIT_LIST_HEAD(&fs_info
->dirty_cowonly_roots
);
1258 INIT_LIST_HEAD(&fs_info
->space_info
);
1259 btrfs_mapping_init(&fs_info
->mapping_tree
);
1260 atomic_set(&fs_info
->nr_async_submits
, 0);
1261 atomic_set(&fs_info
->throttles
, 0);
1262 atomic_set(&fs_info
->throttle_gen
, 0);
1264 fs_info
->max_extent
= (u64
)-1;
1265 fs_info
->max_inline
= 8192 * 1024;
1266 setup_bdi(fs_info
, &fs_info
->bdi
);
1267 fs_info
->btree_inode
= new_inode(sb
);
1268 fs_info
->btree_inode
->i_ino
= 1;
1269 fs_info
->btree_inode
->i_nlink
= 1;
1270 fs_info
->thread_pool_size
= min(num_online_cpus() + 2, 8);
1272 INIT_LIST_HEAD(&fs_info
->ordered_extents
);
1273 spin_lock_init(&fs_info
->ordered_extent_lock
);
1275 sb
->s_blocksize
= 4096;
1276 sb
->s_blocksize_bits
= blksize_bits(4096);
1279 * we set the i_size on the btree inode to the max possible int.
1280 * the real end of the address space is determined by all of
1281 * the devices in the system
1283 fs_info
->btree_inode
->i_size
= OFFSET_MAX
;
1284 fs_info
->btree_inode
->i_mapping
->a_ops
= &btree_aops
;
1285 fs_info
->btree_inode
->i_mapping
->backing_dev_info
= &fs_info
->bdi
;
1287 extent_io_tree_init(&BTRFS_I(fs_info
->btree_inode
)->io_tree
,
1288 fs_info
->btree_inode
->i_mapping
,
1290 extent_map_tree_init(&BTRFS_I(fs_info
->btree_inode
)->extent_tree
,
1293 BTRFS_I(fs_info
->btree_inode
)->io_tree
.ops
= &btree_extent_io_ops
;
1295 extent_io_tree_init(&fs_info
->free_space_cache
,
1296 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1297 extent_io_tree_init(&fs_info
->block_group_cache
,
1298 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1299 extent_io_tree_init(&fs_info
->pinned_extents
,
1300 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1301 extent_io_tree_init(&fs_info
->pending_del
,
1302 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1303 extent_io_tree_init(&fs_info
->extent_ins
,
1304 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1305 fs_info
->do_barriers
= 1;
1307 BTRFS_I(fs_info
->btree_inode
)->root
= tree_root
;
1308 memset(&BTRFS_I(fs_info
->btree_inode
)->location
, 0,
1309 sizeof(struct btrfs_key
));
1310 insert_inode_hash(fs_info
->btree_inode
);
1311 mapping_set_gfp_mask(fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1313 mutex_init(&fs_info
->trans_mutex
);
1314 mutex_init(&fs_info
->drop_mutex
);
1315 mutex_init(&fs_info
->alloc_mutex
);
1316 mutex_init(&fs_info
->chunk_mutex
);
1317 mutex_init(&fs_info
->transaction_kthread_mutex
);
1318 mutex_init(&fs_info
->cleaner_mutex
);
1319 mutex_init(&fs_info
->volume_mutex
);
1320 init_waitqueue_head(&fs_info
->transaction_throttle
);
1321 init_waitqueue_head(&fs_info
->transaction_wait
);
1324 ret
= add_hasher(fs_info
, "crc32c");
1326 printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
1331 __setup_root(4096, 4096, 4096, 4096, tree_root
,
1332 fs_info
, BTRFS_ROOT_TREE_OBJECTID
);
1335 bh
= __bread(fs_devices
->latest_bdev
,
1336 BTRFS_SUPER_INFO_OFFSET
/ 4096, 4096);
1340 memcpy(&fs_info
->super_copy
, bh
->b_data
, sizeof(fs_info
->super_copy
));
1343 memcpy(fs_info
->fsid
, fs_info
->super_copy
.fsid
, BTRFS_FSID_SIZE
);
1345 disk_super
= &fs_info
->super_copy
;
1346 if (!btrfs_super_root(disk_super
))
1347 goto fail_sb_buffer
;
1349 err
= btrfs_parse_options(tree_root
, options
);
1351 goto fail_sb_buffer
;
1354 * we need to start all the end_io workers up front because the
1355 * queue work function gets called at interrupt time, and so it
1356 * cannot dynamically grow.
1358 btrfs_init_workers(&fs_info
->workers
, fs_info
->thread_pool_size
);
1359 btrfs_init_workers(&fs_info
->submit_workers
, fs_info
->thread_pool_size
);
1360 btrfs_init_workers(&fs_info
->fixup_workers
, 1);
1361 btrfs_init_workers(&fs_info
->endio_workers
, fs_info
->thread_pool_size
);
1362 btrfs_init_workers(&fs_info
->endio_write_workers
,
1363 fs_info
->thread_pool_size
);
1364 btrfs_start_workers(&fs_info
->workers
, 1);
1365 btrfs_start_workers(&fs_info
->submit_workers
, 1);
1366 btrfs_start_workers(&fs_info
->fixup_workers
, 1);
1367 btrfs_start_workers(&fs_info
->endio_workers
, fs_info
->thread_pool_size
);
1368 btrfs_start_workers(&fs_info
->endio_write_workers
,
1369 fs_info
->thread_pool_size
);
1372 if (btrfs_super_num_devices(disk_super
) > fs_devices
->open_devices
) {
1373 printk("Btrfs: wanted %llu devices, but found %llu\n",
1374 (unsigned long long)btrfs_super_num_devices(disk_super
),
1375 (unsigned long long)fs_devices
->open_devices
);
1376 if (btrfs_test_opt(tree_root
, DEGRADED
))
1377 printk("continuing in degraded mode\n");
1379 goto fail_sb_buffer
;
1383 fs_info
->bdi
.ra_pages
*= btrfs_super_num_devices(disk_super
);
1385 nodesize
= btrfs_super_nodesize(disk_super
);
1386 leafsize
= btrfs_super_leafsize(disk_super
);
1387 sectorsize
= btrfs_super_sectorsize(disk_super
);
1388 stripesize
= btrfs_super_stripesize(disk_super
);
1389 tree_root
->nodesize
= nodesize
;
1390 tree_root
->leafsize
= leafsize
;
1391 tree_root
->sectorsize
= sectorsize
;
1392 tree_root
->stripesize
= stripesize
;
1394 sb
->s_blocksize
= sectorsize
;
1395 sb
->s_blocksize_bits
= blksize_bits(sectorsize
);
1397 if (strncmp((char *)(&disk_super
->magic
), BTRFS_MAGIC
,
1398 sizeof(disk_super
->magic
))) {
1399 printk("btrfs: valid FS not found on %s\n", sb
->s_id
);
1400 goto fail_sb_buffer
;
1403 mutex_lock(&fs_info
->chunk_mutex
);
1404 ret
= btrfs_read_sys_array(tree_root
);
1405 mutex_unlock(&fs_info
->chunk_mutex
);
1407 printk("btrfs: failed to read the system array on %s\n",
1409 goto fail_sys_array
;
1412 blocksize
= btrfs_level_size(tree_root
,
1413 btrfs_super_chunk_root_level(disk_super
));
1415 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
1416 chunk_root
, fs_info
, BTRFS_CHUNK_TREE_OBJECTID
);
1418 chunk_root
->node
= read_tree_block(chunk_root
,
1419 btrfs_super_chunk_root(disk_super
),
1421 BUG_ON(!chunk_root
->node
);
1423 read_extent_buffer(chunk_root
->node
, fs_info
->chunk_tree_uuid
,
1424 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root
->node
),
1427 mutex_lock(&fs_info
->chunk_mutex
);
1428 ret
= btrfs_read_chunk_tree(chunk_root
);
1429 mutex_unlock(&fs_info
->chunk_mutex
);
1432 btrfs_close_extra_devices(fs_devices
);
1434 blocksize
= btrfs_level_size(tree_root
,
1435 btrfs_super_root_level(disk_super
));
1438 tree_root
->node
= read_tree_block(tree_root
,
1439 btrfs_super_root(disk_super
),
1441 if (!tree_root
->node
)
1442 goto fail_sb_buffer
;
1445 ret
= find_and_setup_root(tree_root
, fs_info
,
1446 BTRFS_EXTENT_TREE_OBJECTID
, extent_root
);
1448 goto fail_tree_root
;
1449 extent_root
->track_dirty
= 1;
1451 ret
= find_and_setup_root(tree_root
, fs_info
,
1452 BTRFS_DEV_TREE_OBJECTID
, dev_root
);
1453 dev_root
->track_dirty
= 1;
1456 goto fail_extent_root
;
1458 btrfs_read_block_groups(extent_root
);
1460 fs_info
->generation
= btrfs_super_generation(disk_super
) + 1;
1461 fs_info
->data_alloc_profile
= (u64
)-1;
1462 fs_info
->metadata_alloc_profile
= (u64
)-1;
1463 fs_info
->system_alloc_profile
= fs_info
->metadata_alloc_profile
;
1464 fs_info
->cleaner_kthread
= kthread_run(cleaner_kthread
, tree_root
,
1466 if (!fs_info
->cleaner_kthread
)
1467 goto fail_extent_root
;
1469 fs_info
->transaction_kthread
= kthread_run(transaction_kthread
,
1471 "btrfs-transaction");
1472 if (!fs_info
->transaction_kthread
)
1479 kthread_stop(fs_info
->cleaner_kthread
);
1481 free_extent_buffer(extent_root
->node
);
1483 free_extent_buffer(tree_root
->node
);
1486 btrfs_stop_workers(&fs_info
->fixup_workers
);
1487 btrfs_stop_workers(&fs_info
->workers
);
1488 btrfs_stop_workers(&fs_info
->endio_workers
);
1489 btrfs_stop_workers(&fs_info
->endio_write_workers
);
1490 btrfs_stop_workers(&fs_info
->submit_workers
);
1492 iput(fs_info
->btree_inode
);
1494 btrfs_close_devices(fs_info
->fs_devices
);
1495 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
1499 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
1500 bdi_destroy(&fs_info
->bdi
);
1503 return ERR_PTR(err
);
1506 static void btrfs_end_buffer_write_sync(struct buffer_head
*bh
, int uptodate
)
1508 char b
[BDEVNAME_SIZE
];
1511 set_buffer_uptodate(bh
);
1513 if (!buffer_eopnotsupp(bh
) && printk_ratelimit()) {
1514 printk(KERN_WARNING
"lost page write due to "
1515 "I/O error on %s\n",
1516 bdevname(bh
->b_bdev
, b
));
1518 /* note, we dont' set_buffer_write_io_error because we have
1519 * our own ways of dealing with the IO errors
1521 clear_buffer_uptodate(bh
);
1527 int write_all_supers(struct btrfs_root
*root
)
1529 struct list_head
*cur
;
1530 struct list_head
*head
= &root
->fs_info
->fs_devices
->devices
;
1531 struct btrfs_device
*dev
;
1532 struct btrfs_super_block
*sb
;
1533 struct btrfs_dev_item
*dev_item
;
1534 struct buffer_head
*bh
;
1538 int total_errors
= 0;
1542 max_errors
= btrfs_super_num_devices(&root
->fs_info
->super_copy
) - 1;
1543 do_barriers
= !btrfs_test_opt(root
, NOBARRIER
);
1545 sb
= &root
->fs_info
->super_for_commit
;
1546 dev_item
= &sb
->dev_item
;
1547 list_for_each(cur
, head
) {
1548 dev
= list_entry(cur
, struct btrfs_device
, dev_list
);
1553 if (!dev
->in_fs_metadata
)
1556 btrfs_set_stack_device_type(dev_item
, dev
->type
);
1557 btrfs_set_stack_device_id(dev_item
, dev
->devid
);
1558 btrfs_set_stack_device_total_bytes(dev_item
, dev
->total_bytes
);
1559 btrfs_set_stack_device_bytes_used(dev_item
, dev
->bytes_used
);
1560 btrfs_set_stack_device_io_align(dev_item
, dev
->io_align
);
1561 btrfs_set_stack_device_io_width(dev_item
, dev
->io_width
);
1562 btrfs_set_stack_device_sector_size(dev_item
, dev
->sector_size
);
1563 memcpy(dev_item
->uuid
, dev
->uuid
, BTRFS_UUID_SIZE
);
1564 flags
= btrfs_super_flags(sb
);
1565 btrfs_set_super_flags(sb
, flags
| BTRFS_HEADER_FLAG_WRITTEN
);
1569 crc
= btrfs_csum_data(root
, (char *)sb
+ BTRFS_CSUM_SIZE
, crc
,
1570 BTRFS_SUPER_INFO_SIZE
- BTRFS_CSUM_SIZE
);
1571 btrfs_csum_final(crc
, sb
->csum
);
1573 bh
= __getblk(dev
->bdev
, BTRFS_SUPER_INFO_OFFSET
/ 4096,
1574 BTRFS_SUPER_INFO_SIZE
);
1576 memcpy(bh
->b_data
, sb
, BTRFS_SUPER_INFO_SIZE
);
1577 dev
->pending_io
= bh
;
1580 set_buffer_uptodate(bh
);
1582 bh
->b_end_io
= btrfs_end_buffer_write_sync
;
1584 if (do_barriers
&& dev
->barriers
) {
1585 ret
= submit_bh(WRITE_BARRIER
, bh
);
1586 if (ret
== -EOPNOTSUPP
) {
1587 printk("btrfs: disabling barriers on dev %s\n",
1589 set_buffer_uptodate(bh
);
1593 ret
= submit_bh(WRITE
, bh
);
1596 ret
= submit_bh(WRITE
, bh
);
1601 if (total_errors
> max_errors
) {
1602 printk("btrfs: %d errors while writing supers\n", total_errors
);
1607 list_for_each(cur
, head
) {
1608 dev
= list_entry(cur
, struct btrfs_device
, dev_list
);
1611 if (!dev
->in_fs_metadata
)
1614 BUG_ON(!dev
->pending_io
);
1615 bh
= dev
->pending_io
;
1617 if (!buffer_uptodate(dev
->pending_io
)) {
1618 if (do_barriers
&& dev
->barriers
) {
1619 printk("btrfs: disabling barriers on dev %s\n",
1621 set_buffer_uptodate(bh
);
1625 ret
= submit_bh(WRITE
, bh
);
1628 if (!buffer_uptodate(bh
))
1635 dev
->pending_io
= NULL
;
1638 if (total_errors
> max_errors
) {
1639 printk("btrfs: %d errors while writing supers\n", total_errors
);
1645 int write_ctree_super(struct btrfs_trans_handle
*trans
, struct btrfs_root
1650 ret
= write_all_supers(root
);
1654 int btrfs_free_fs_root(struct btrfs_fs_info
*fs_info
, struct btrfs_root
*root
)
1656 radix_tree_delete(&fs_info
->fs_roots_radix
,
1657 (unsigned long)root
->root_key
.objectid
);
1659 btrfs_sysfs_del_root(root
);
1663 free_extent_buffer(root
->node
);
1664 if (root
->commit_root
)
1665 free_extent_buffer(root
->commit_root
);
1672 static int del_fs_roots(struct btrfs_fs_info
*fs_info
)
1675 struct btrfs_root
*gang
[8];
1679 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
1684 for (i
= 0; i
< ret
; i
++)
1685 btrfs_free_fs_root(fs_info
, gang
[i
]);
1690 int close_ctree(struct btrfs_root
*root
)
1693 struct btrfs_trans_handle
*trans
;
1694 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1696 fs_info
->closing
= 1;
1699 kthread_stop(root
->fs_info
->transaction_kthread
);
1700 kthread_stop(root
->fs_info
->cleaner_kthread
);
1702 btrfs_clean_old_snapshots(root
);
1703 trans
= btrfs_start_transaction(root
, 1);
1704 ret
= btrfs_commit_transaction(trans
, root
);
1705 /* run commit again to drop the original snapshot */
1706 trans
= btrfs_start_transaction(root
, 1);
1707 btrfs_commit_transaction(trans
, root
);
1708 ret
= btrfs_write_and_wait_transaction(NULL
, root
);
1711 write_ctree_super(NULL
, root
);
1713 if (fs_info
->delalloc_bytes
) {
1714 printk("btrfs: at unmount delalloc count %Lu\n",
1715 fs_info
->delalloc_bytes
);
1717 if (fs_info
->total_ref_cache_size
) {
1718 printk("btrfs: at umount reference cache size %Lu\n",
1719 fs_info
->total_ref_cache_size
);
1722 if (fs_info
->extent_root
->node
)
1723 free_extent_buffer(fs_info
->extent_root
->node
);
1725 if (fs_info
->tree_root
->node
)
1726 free_extent_buffer(fs_info
->tree_root
->node
);
1728 if (root
->fs_info
->chunk_root
->node
);
1729 free_extent_buffer(root
->fs_info
->chunk_root
->node
);
1731 if (root
->fs_info
->dev_root
->node
);
1732 free_extent_buffer(root
->fs_info
->dev_root
->node
);
1734 btrfs_free_block_groups(root
->fs_info
);
1735 del_fs_roots(fs_info
);
1737 filemap_write_and_wait(fs_info
->btree_inode
->i_mapping
);
1739 truncate_inode_pages(fs_info
->btree_inode
->i_mapping
, 0);
1741 btrfs_stop_workers(&fs_info
->fixup_workers
);
1742 btrfs_stop_workers(&fs_info
->workers
);
1743 btrfs_stop_workers(&fs_info
->endio_workers
);
1744 btrfs_stop_workers(&fs_info
->endio_write_workers
);
1745 btrfs_stop_workers(&fs_info
->submit_workers
);
1747 iput(fs_info
->btree_inode
);
1749 while(!list_empty(&fs_info
->hashers
)) {
1750 struct btrfs_hasher
*hasher
;
1751 hasher
= list_entry(fs_info
->hashers
.next
, struct btrfs_hasher
,
1753 list_del(&hasher
->hashers
);
1754 crypto_free_hash(&fs_info
->hash_tfm
);
1758 btrfs_close_devices(fs_info
->fs_devices
);
1759 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
1761 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
1762 bdi_destroy(&fs_info
->bdi
);
1765 kfree(fs_info
->extent_root
);
1766 kfree(fs_info
->tree_root
);
1767 kfree(fs_info
->chunk_root
);
1768 kfree(fs_info
->dev_root
);
1772 int btrfs_buffer_uptodate(struct extent_buffer
*buf
, u64 parent_transid
)
1775 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
1777 ret
= extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
, buf
);
1781 ret
= verify_parent_transid(&BTRFS_I(btree_inode
)->io_tree
, buf
,
1786 int btrfs_set_buffer_uptodate(struct extent_buffer
*buf
)
1788 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
1789 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
,
1793 void btrfs_mark_buffer_dirty(struct extent_buffer
*buf
)
1795 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1796 u64 transid
= btrfs_header_generation(buf
);
1797 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1799 WARN_ON(!btrfs_tree_locked(buf
));
1800 if (transid
!= root
->fs_info
->generation
) {
1801 printk(KERN_CRIT
"transid mismatch buffer %llu, found %Lu running %Lu\n",
1802 (unsigned long long)buf
->start
,
1803 transid
, root
->fs_info
->generation
);
1806 set_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
, buf
);
1809 void btrfs_btree_balance_dirty(struct btrfs_root
*root
, unsigned long nr
)
1812 * looks as though older kernels can get into trouble with
1813 * this code, they end up stuck in balance_dirty_pages forever
1815 struct extent_io_tree
*tree
;
1818 unsigned long thresh
= 16 * 1024 * 1024;
1819 tree
= &BTRFS_I(root
->fs_info
->btree_inode
)->io_tree
;
1821 if (current_is_pdflush())
1824 num_dirty
= count_range_bits(tree
, &start
, (u64
)-1,
1825 thresh
, EXTENT_DIRTY
);
1826 if (num_dirty
> thresh
) {
1827 balance_dirty_pages_ratelimited_nr(
1828 root
->fs_info
->btree_inode
->i_mapping
, 1);
1833 int btrfs_read_buffer(struct extent_buffer
*buf
, u64 parent_transid
)
1835 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1837 ret
= btree_read_extent_buffer_pages(root
, buf
, 0, parent_transid
);
1839 buf
->flags
|= EXTENT_UPTODATE
;
1844 static struct extent_io_ops btree_extent_io_ops
= {
1845 .writepage_io_hook
= btree_writepage_io_hook
,
1846 .readpage_end_io_hook
= btree_readpage_end_io_hook
,
1847 .submit_bio_hook
= btree_submit_bio_hook
,
1848 /* note we're sharing with inode.c for the merge bio hook */
1849 .merge_bio_hook
= btrfs_merge_bio_hook
,