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> // for block_sync_page
26 #include <linux/workqueue.h>
30 #include "transaction.h"
31 #include "btrfs_inode.h"
33 #include "print-tree.h"
36 static int check_tree_block(struct btrfs_root
*root
, struct extent_buffer
*buf
)
38 if (extent_buffer_blocknr(buf
) != btrfs_header_blocknr(buf
)) {
39 printk(KERN_CRIT
"buf blocknr(buf) is %llu, header is %llu\n",
40 (unsigned long long)extent_buffer_blocknr(buf
),
41 (unsigned long long)btrfs_header_blocknr(buf
));
48 static struct extent_io_ops btree_extent_io_ops
;
49 static struct workqueue_struct
*end_io_workqueue
;
50 static struct workqueue_struct
*async_submit_workqueue
;
56 struct btrfs_fs_info
*info
;
59 struct list_head list
;
62 struct async_submit_bio
{
65 struct list_head list
;
66 extent_submit_bio_hook_t
*submit_bio_hook
;
71 struct extent_map
*btree_get_extent(struct inode
*inode
, struct page
*page
,
72 size_t page_offset
, u64 start
, u64 len
,
75 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
76 struct extent_map
*em
;
79 spin_lock(&em_tree
->lock
);
80 em
= lookup_extent_mapping(em_tree
, start
, len
);
81 spin_unlock(&em_tree
->lock
);
85 em
= alloc_extent_map(GFP_NOFS
);
87 em
= ERR_PTR(-ENOMEM
);
93 em
->bdev
= inode
->i_sb
->s_bdev
;
95 spin_lock(&em_tree
->lock
);
96 ret
= add_extent_mapping(em_tree
, em
);
98 u64 failed_start
= em
->start
;
99 u64 failed_len
= em
->len
;
101 printk("failed to insert %Lu %Lu -> %Lu into tree\n",
102 em
->start
, em
->len
, em
->block_start
);
104 em
= lookup_extent_mapping(em_tree
, start
, len
);
106 printk("after failing, found %Lu %Lu %Lu\n",
107 em
->start
, em
->len
, em
->block_start
);
110 em
= lookup_extent_mapping(em_tree
, failed_start
,
113 printk("double failure lookup gives us "
114 "%Lu %Lu -> %Lu\n", em
->start
,
115 em
->len
, em
->block_start
);
124 spin_unlock(&em_tree
->lock
);
132 u32
btrfs_csum_data(struct btrfs_root
*root
, char *data
, u32 seed
, size_t len
)
134 return btrfs_crc32c(seed
, data
, len
);
137 void btrfs_csum_final(u32 crc
, char *result
)
139 *(__le32
*)result
= ~cpu_to_le32(crc
);
142 static int csum_tree_block(struct btrfs_root
*root
, struct extent_buffer
*buf
,
145 char result
[BTRFS_CRC32_SIZE
];
147 unsigned long cur_len
;
148 unsigned long offset
= BTRFS_CSUM_SIZE
;
149 char *map_token
= NULL
;
151 unsigned long map_start
;
152 unsigned long map_len
;
156 len
= buf
->len
- offset
;
158 err
= map_private_extent_buffer(buf
, offset
, 32,
160 &map_start
, &map_len
, KM_USER0
);
162 printk("failed to map extent buffer! %lu\n",
166 cur_len
= min(len
, map_len
- (offset
- map_start
));
167 crc
= btrfs_csum_data(root
, kaddr
+ offset
- map_start
,
171 unmap_extent_buffer(buf
, map_token
, KM_USER0
);
173 btrfs_csum_final(crc
, result
);
176 int from_this_trans
= 0;
178 if (root
->fs_info
->running_transaction
&&
179 btrfs_header_generation(buf
) ==
180 root
->fs_info
->running_transaction
->transid
)
183 /* FIXME, this is not good */
184 if (memcmp_extent_buffer(buf
, result
, 0, BTRFS_CRC32_SIZE
)) {
187 memcpy(&found
, result
, BTRFS_CRC32_SIZE
);
189 read_extent_buffer(buf
, &val
, 0, BTRFS_CRC32_SIZE
);
190 printk("btrfs: %s checksum verify failed on %llu "
191 "wanted %X found %X from_this_trans %d "
193 root
->fs_info
->sb
->s_id
,
194 buf
->start
, val
, found
, from_this_trans
,
195 btrfs_header_level(buf
));
199 write_extent_buffer(buf
, result
, 0, BTRFS_CRC32_SIZE
);
204 static int btree_read_extent_buffer_pages(struct btrfs_root
*root
,
205 struct extent_buffer
*eb
,
208 struct extent_io_tree
*io_tree
;
213 io_tree
= &BTRFS_I(root
->fs_info
->btree_inode
)->io_tree
;
215 ret
= read_extent_buffer_pages(io_tree
, eb
, start
, 1,
216 btree_get_extent
, mirror_num
);
219 printk("good read %Lu mirror %d total %d\n", eb
->start
, mirror_num
, num_copies
);
222 num_copies
= btrfs_num_copies(&root
->fs_info
->mapping_tree
,
224 printk("failed to read %Lu mirror %d total %d\n", eb
->start
, mirror_num
, num_copies
);
225 if (num_copies
== 1) {
226 printk("reading %Lu failed only one copy\n", eb
->start
);
230 if (mirror_num
> num_copies
) {
231 printk("bailing at mirror %d of %d\n", mirror_num
, num_copies
);
235 printk("read extent buffer page last\n");
239 int csum_dirty_buffer(struct btrfs_root
*root
, struct page
*page
)
241 struct extent_io_tree
*tree
;
242 u64 start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
246 struct extent_buffer
*eb
;
249 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
251 if (page
->private == EXTENT_PAGE_PRIVATE
)
255 len
= page
->private >> 2;
259 eb
= alloc_extent_buffer(tree
, start
, len
, page
, GFP_NOFS
);
260 ret
= btree_read_extent_buffer_pages(root
, eb
, start
+ PAGE_CACHE_SIZE
);
262 btrfs_clear_buffer_defrag(eb
);
263 found_start
= btrfs_header_bytenr(eb
);
264 if (found_start
!= start
) {
265 printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
266 start
, found_start
, len
);
270 if (eb
->first_page
!= page
) {
271 printk("bad first page %lu %lu\n", eb
->first_page
->index
,
276 if (!PageUptodate(page
)) {
277 printk("csum not up to date page %lu\n", page
->index
);
281 found_level
= btrfs_header_level(eb
);
282 spin_lock(&root
->fs_info
->hash_lock
);
283 btrfs_set_header_flag(eb
, BTRFS_HEADER_FLAG_WRITTEN
);
284 spin_unlock(&root
->fs_info
->hash_lock
);
285 csum_tree_block(root
, eb
, 0);
287 free_extent_buffer(eb
);
292 static int btree_writepage_io_hook(struct page
*page
, u64 start
, u64 end
)
294 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
296 csum_dirty_buffer(root
, page
);
300 int btree_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
301 struct extent_state
*state
)
303 struct extent_io_tree
*tree
;
307 struct extent_buffer
*eb
;
308 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
311 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
312 if (page
->private == EXTENT_PAGE_PRIVATE
)
316 len
= page
->private >> 2;
320 eb
= alloc_extent_buffer(tree
, start
, len
, page
, GFP_NOFS
);
322 btrfs_clear_buffer_defrag(eb
);
323 found_start
= btrfs_header_bytenr(eb
);
324 if (found_start
!= start
) {
325 printk("bad start on %Lu found %Lu\n", eb
->start
, found_start
);
329 if (eb
->first_page
!= page
) {
330 printk("bad first page %lu %lu\n", eb
->first_page
->index
,
336 found_level
= btrfs_header_level(eb
);
338 ret
= csum_tree_block(root
, eb
, 1);
342 end
= min_t(u64
, eb
->len
, PAGE_CACHE_SIZE
);
343 end
= eb
->start
+ end
- 1;
344 release_extent_buffer_tail_pages(eb
);
346 free_extent_buffer(eb
);
351 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
352 static void end_workqueue_bio(struct bio
*bio
, int err
)
354 static int end_workqueue_bio(struct bio
*bio
,
355 unsigned int bytes_done
, int err
)
358 struct end_io_wq
*end_io_wq
= bio
->bi_private
;
359 struct btrfs_fs_info
*fs_info
;
362 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
367 fs_info
= end_io_wq
->info
;
368 spin_lock_irqsave(&fs_info
->end_io_work_lock
, flags
);
369 end_io_wq
->error
= err
;
370 list_add_tail(&end_io_wq
->list
, &fs_info
->end_io_work_list
);
371 spin_unlock_irqrestore(&fs_info
->end_io_work_lock
, flags
);
372 queue_work(end_io_workqueue
, &fs_info
->end_io_work
);
374 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
379 int btrfs_bio_wq_end_io(struct btrfs_fs_info
*info
, struct bio
*bio
,
382 struct end_io_wq
*end_io_wq
;
383 end_io_wq
= kmalloc(sizeof(*end_io_wq
), GFP_NOFS
);
387 end_io_wq
->private = bio
->bi_private
;
388 end_io_wq
->end_io
= bio
->bi_end_io
;
389 end_io_wq
->info
= info
;
390 end_io_wq
->error
= 0;
391 end_io_wq
->bio
= bio
;
392 end_io_wq
->metadata
= metadata
;
394 bio
->bi_private
= end_io_wq
;
395 bio
->bi_end_io
= end_workqueue_bio
;
399 int btrfs_wq_submit_bio(struct btrfs_fs_info
*fs_info
, struct inode
*inode
,
400 int rw
, struct bio
*bio
, int mirror_num
,
401 extent_submit_bio_hook_t
*submit_bio_hook
)
403 struct async_submit_bio
*async
;
406 * inline writerback should stay inline, only hop to the async
407 * queue if we're pdflush
409 if (!current_is_pdflush())
410 return submit_bio_hook(inode
, rw
, bio
, mirror_num
);
412 async
= kmalloc(sizeof(*async
), GFP_NOFS
);
416 async
->inode
= inode
;
419 async
->mirror_num
= mirror_num
;
420 async
->submit_bio_hook
= submit_bio_hook
;
422 spin_lock(&fs_info
->async_submit_work_lock
);
423 list_add_tail(&async
->list
, &fs_info
->async_submit_work_list
);
424 spin_unlock(&fs_info
->async_submit_work_lock
);
426 queue_work(async_submit_workqueue
, &fs_info
->async_submit_work
);
430 static int __btree_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
433 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
437 offset
= bio
->bi_sector
<< 9;
439 if (rw
& (1 << BIO_RW
)) {
440 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
, mirror_num
);
443 ret
= btrfs_bio_wq_end_io(root
->fs_info
, bio
, 1);
446 if (offset
== BTRFS_SUPER_INFO_OFFSET
) {
447 bio
->bi_bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
451 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
, mirror_num
);
454 static int btree_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
457 if (!(rw
& (1 << BIO_RW
))) {
458 return __btree_submit_bio_hook(inode
, rw
, bio
, mirror_num
);
460 return btrfs_wq_submit_bio(BTRFS_I(inode
)->root
->fs_info
,
461 inode
, rw
, bio
, mirror_num
,
462 __btree_submit_bio_hook
);
465 static int btree_writepage(struct page
*page
, struct writeback_control
*wbc
)
467 struct extent_io_tree
*tree
;
468 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
469 return extent_write_full_page(tree
, page
, btree_get_extent
, wbc
);
472 static int btree_writepages(struct address_space
*mapping
,
473 struct writeback_control
*wbc
)
475 struct extent_io_tree
*tree
;
476 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
477 if (wbc
->sync_mode
== WB_SYNC_NONE
) {
480 unsigned long thresh
= 96 * 1024 * 1024;
482 if (wbc
->for_kupdate
)
485 if (current_is_pdflush()) {
486 thresh
= 96 * 1024 * 1024;
488 thresh
= 8 * 1024 * 1024;
490 num_dirty
= count_range_bits(tree
, &start
, (u64
)-1,
491 thresh
, EXTENT_DIRTY
);
492 if (num_dirty
< thresh
) {
496 return extent_writepages(tree
, mapping
, btree_get_extent
, wbc
);
499 int btree_readpage(struct file
*file
, struct page
*page
)
501 struct extent_io_tree
*tree
;
502 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
503 return extent_read_full_page(tree
, page
, btree_get_extent
);
506 static int btree_releasepage(struct page
*page
, gfp_t gfp_flags
)
508 struct extent_io_tree
*tree
;
509 struct extent_map_tree
*map
;
512 if (page_count(page
) > 3) {
513 /* once for page->private, once for the caller, once
514 * once for the page cache
518 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
519 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
520 ret
= try_release_extent_state(map
, tree
, page
, gfp_flags
);
522 invalidate_extent_lru(tree
, page_offset(page
), PAGE_CACHE_SIZE
);
523 ClearPagePrivate(page
);
524 set_page_private(page
, 0);
525 page_cache_release(page
);
530 static void btree_invalidatepage(struct page
*page
, unsigned long offset
)
532 struct extent_io_tree
*tree
;
533 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
534 extent_invalidatepage(tree
, page
, offset
);
535 btree_releasepage(page
, GFP_NOFS
);
536 if (PagePrivate(page
)) {
537 invalidate_extent_lru(tree
, page_offset(page
), PAGE_CACHE_SIZE
);
538 ClearPagePrivate(page
);
539 set_page_private(page
, 0);
540 page_cache_release(page
);
545 static int btree_writepage(struct page
*page
, struct writeback_control
*wbc
)
547 struct buffer_head
*bh
;
548 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
549 struct buffer_head
*head
;
550 if (!page_has_buffers(page
)) {
551 create_empty_buffers(page
, root
->fs_info
->sb
->s_blocksize
,
552 (1 << BH_Dirty
)|(1 << BH_Uptodate
));
554 head
= page_buffers(page
);
557 if (buffer_dirty(bh
))
558 csum_tree_block(root
, bh
, 0);
559 bh
= bh
->b_this_page
;
560 } while (bh
!= head
);
561 return block_write_full_page(page
, btree_get_block
, wbc
);
565 static struct address_space_operations btree_aops
= {
566 .readpage
= btree_readpage
,
567 .writepage
= btree_writepage
,
568 .writepages
= btree_writepages
,
569 .releasepage
= btree_releasepage
,
570 .invalidatepage
= btree_invalidatepage
,
571 .sync_page
= block_sync_page
,
574 int readahead_tree_block(struct btrfs_root
*root
, u64 bytenr
, u32 blocksize
)
576 struct extent_buffer
*buf
= NULL
;
577 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
580 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
583 read_extent_buffer_pages(&BTRFS_I(btree_inode
)->io_tree
,
584 buf
, 0, 0, btree_get_extent
, 0);
585 free_extent_buffer(buf
);
589 static int close_all_devices(struct btrfs_fs_info
*fs_info
)
591 struct list_head
*list
;
592 struct list_head
*next
;
593 struct btrfs_device
*device
;
595 list
= &fs_info
->fs_devices
->devices
;
596 list_for_each(next
, list
) {
597 device
= list_entry(next
, struct btrfs_device
, dev_list
);
598 if (device
->bdev
&& device
->bdev
!= fs_info
->sb
->s_bdev
)
599 close_bdev_excl(device
->bdev
);
605 int btrfs_verify_block_csum(struct btrfs_root
*root
,
606 struct extent_buffer
*buf
)
608 return btrfs_buffer_uptodate(buf
);
611 struct extent_buffer
*btrfs_find_tree_block(struct btrfs_root
*root
,
612 u64 bytenr
, u32 blocksize
)
614 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
615 struct extent_buffer
*eb
;
616 eb
= find_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
617 bytenr
, blocksize
, GFP_NOFS
);
621 struct extent_buffer
*btrfs_find_create_tree_block(struct btrfs_root
*root
,
622 u64 bytenr
, u32 blocksize
)
624 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
625 struct extent_buffer
*eb
;
627 eb
= alloc_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
628 bytenr
, blocksize
, NULL
, GFP_NOFS
);
633 struct extent_buffer
*read_tree_block(struct btrfs_root
*root
, u64 bytenr
,
636 struct extent_buffer
*buf
= NULL
;
637 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
638 struct extent_io_tree
*io_tree
;
641 io_tree
= &BTRFS_I(btree_inode
)->io_tree
;
643 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
647 ret
= btree_read_extent_buffer_pages(root
, buf
, 0);
650 buf
->flags
|= EXTENT_UPTODATE
;
656 int clean_tree_block(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
657 struct extent_buffer
*buf
)
659 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
660 if (btrfs_header_generation(buf
) ==
661 root
->fs_info
->running_transaction
->transid
)
662 clear_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
,
667 int wait_on_tree_block_writeback(struct btrfs_root
*root
,
668 struct extent_buffer
*buf
)
670 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
671 wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode
)->io_tree
,
676 static int __setup_root(u32 nodesize
, u32 leafsize
, u32 sectorsize
,
677 u32 stripesize
, struct btrfs_root
*root
,
678 struct btrfs_fs_info
*fs_info
,
683 root
->commit_root
= NULL
;
684 root
->sectorsize
= sectorsize
;
685 root
->nodesize
= nodesize
;
686 root
->leafsize
= leafsize
;
687 root
->stripesize
= stripesize
;
689 root
->track_dirty
= 0;
691 root
->fs_info
= fs_info
;
692 root
->objectid
= objectid
;
693 root
->last_trans
= 0;
694 root
->highest_inode
= 0;
695 root
->last_inode_alloc
= 0;
699 INIT_LIST_HEAD(&root
->dirty_list
);
700 memset(&root
->root_key
, 0, sizeof(root
->root_key
));
701 memset(&root
->root_item
, 0, sizeof(root
->root_item
));
702 memset(&root
->defrag_progress
, 0, sizeof(root
->defrag_progress
));
703 memset(&root
->root_kobj
, 0, sizeof(root
->root_kobj
));
704 init_completion(&root
->kobj_unregister
);
705 root
->defrag_running
= 0;
706 root
->defrag_level
= 0;
707 root
->root_key
.objectid
= objectid
;
711 static int find_and_setup_root(struct btrfs_root
*tree_root
,
712 struct btrfs_fs_info
*fs_info
,
714 struct btrfs_root
*root
)
719 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
720 tree_root
->sectorsize
, tree_root
->stripesize
,
721 root
, fs_info
, objectid
);
722 ret
= btrfs_find_last_root(tree_root
, objectid
,
723 &root
->root_item
, &root
->root_key
);
726 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
727 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
733 struct btrfs_root
*btrfs_read_fs_root_no_radix(struct btrfs_fs_info
*fs_info
,
734 struct btrfs_key
*location
)
736 struct btrfs_root
*root
;
737 struct btrfs_root
*tree_root
= fs_info
->tree_root
;
738 struct btrfs_path
*path
;
739 struct extent_buffer
*l
;
744 root
= kzalloc(sizeof(*root
), GFP_NOFS
);
746 return ERR_PTR(-ENOMEM
);
747 if (location
->offset
== (u64
)-1) {
748 ret
= find_and_setup_root(tree_root
, fs_info
,
749 location
->objectid
, root
);
757 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
758 tree_root
->sectorsize
, tree_root
->stripesize
,
759 root
, fs_info
, location
->objectid
);
761 path
= btrfs_alloc_path();
763 ret
= btrfs_search_slot(NULL
, tree_root
, location
, path
, 0, 0);
770 read_extent_buffer(l
, &root
->root_item
,
771 btrfs_item_ptr_offset(l
, path
->slots
[0]),
772 sizeof(root
->root_item
));
773 memcpy(&root
->root_key
, location
, sizeof(*location
));
776 btrfs_release_path(root
, path
);
777 btrfs_free_path(path
);
782 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
783 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
788 ret
= btrfs_find_highest_inode(root
, &highest_inode
);
790 root
->highest_inode
= highest_inode
;
791 root
->last_inode_alloc
= highest_inode
;
796 struct btrfs_root
*btrfs_lookup_fs_root(struct btrfs_fs_info
*fs_info
,
799 struct btrfs_root
*root
;
801 if (root_objectid
== BTRFS_ROOT_TREE_OBJECTID
)
802 return fs_info
->tree_root
;
803 if (root_objectid
== BTRFS_EXTENT_TREE_OBJECTID
)
804 return fs_info
->extent_root
;
806 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
807 (unsigned long)root_objectid
);
811 struct btrfs_root
*btrfs_read_fs_root_no_name(struct btrfs_fs_info
*fs_info
,
812 struct btrfs_key
*location
)
814 struct btrfs_root
*root
;
817 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
818 return fs_info
->tree_root
;
819 if (location
->objectid
== BTRFS_EXTENT_TREE_OBJECTID
)
820 return fs_info
->extent_root
;
822 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
823 (unsigned long)location
->objectid
);
827 root
= btrfs_read_fs_root_no_radix(fs_info
, location
);
830 ret
= radix_tree_insert(&fs_info
->fs_roots_radix
,
831 (unsigned long)root
->root_key
.objectid
,
834 free_extent_buffer(root
->node
);
838 ret
= btrfs_find_dead_roots(fs_info
->tree_root
,
839 root
->root_key
.objectid
, root
);
845 struct btrfs_root
*btrfs_read_fs_root(struct btrfs_fs_info
*fs_info
,
846 struct btrfs_key
*location
,
847 const char *name
, int namelen
)
849 struct btrfs_root
*root
;
852 root
= btrfs_read_fs_root_no_name(fs_info
, location
);
859 ret
= btrfs_set_root_name(root
, name
, namelen
);
861 free_extent_buffer(root
->node
);
866 ret
= btrfs_sysfs_add_root(root
);
868 free_extent_buffer(root
->node
);
877 static int add_hasher(struct btrfs_fs_info
*info
, char *type
) {
878 struct btrfs_hasher
*hasher
;
880 hasher
= kmalloc(sizeof(*hasher
), GFP_NOFS
);
883 hasher
->hash_tfm
= crypto_alloc_hash(type
, 0, CRYPTO_ALG_ASYNC
);
884 if (!hasher
->hash_tfm
) {
888 spin_lock(&info
->hash_lock
);
889 list_add(&hasher
->list
, &info
->hashers
);
890 spin_unlock(&info
->hash_lock
);
895 static int btrfs_congested_fn(void *congested_data
, int bdi_bits
)
897 struct btrfs_fs_info
*info
= (struct btrfs_fs_info
*)congested_data
;
899 struct list_head
*cur
;
900 struct btrfs_device
*device
;
901 struct backing_dev_info
*bdi
;
903 list_for_each(cur
, &info
->fs_devices
->devices
) {
904 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
905 bdi
= blk_get_backing_dev_info(device
->bdev
);
906 if (bdi
&& bdi_congested(bdi
, bdi_bits
)) {
915 * this unplugs every device on the box, and it is only used when page
918 static void __unplug_io_fn(struct backing_dev_info
*bdi
, struct page
*page
)
920 struct list_head
*cur
;
921 struct btrfs_device
*device
;
922 struct btrfs_fs_info
*info
;
924 info
= (struct btrfs_fs_info
*)bdi
->unplug_io_data
;
925 list_for_each(cur
, &info
->fs_devices
->devices
) {
926 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
927 bdi
= blk_get_backing_dev_info(device
->bdev
);
928 if (bdi
->unplug_io_fn
) {
929 bdi
->unplug_io_fn(bdi
, page
);
934 void btrfs_unplug_io_fn(struct backing_dev_info
*bdi
, struct page
*page
)
937 struct extent_map_tree
*em_tree
;
938 struct extent_map
*em
;
939 struct address_space
*mapping
;
942 /* the generic O_DIRECT read code does this */
944 __unplug_io_fn(bdi
, page
);
949 * page->mapping may change at any time. Get a consistent copy
950 * and use that for everything below
953 mapping
= page
->mapping
;
957 inode
= mapping
->host
;
958 offset
= page_offset(page
);
960 em_tree
= &BTRFS_I(inode
)->extent_tree
;
961 spin_lock(&em_tree
->lock
);
962 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_CACHE_SIZE
);
963 spin_unlock(&em_tree
->lock
);
967 offset
= offset
- em
->start
;
968 btrfs_unplug_page(&BTRFS_I(inode
)->root
->fs_info
->mapping_tree
,
969 em
->block_start
+ offset
, page
);
973 static int setup_bdi(struct btrfs_fs_info
*info
, struct backing_dev_info
*bdi
)
975 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
978 bdi
->ra_pages
= default_backing_dev_info
.ra_pages
;
980 bdi
->capabilities
= default_backing_dev_info
.capabilities
;
981 bdi
->unplug_io_fn
= btrfs_unplug_io_fn
;
982 bdi
->unplug_io_data
= info
;
983 bdi
->congested_fn
= btrfs_congested_fn
;
984 bdi
->congested_data
= info
;
988 static int bio_ready_for_csum(struct bio
*bio
)
994 struct extent_io_tree
*io_tree
= NULL
;
995 struct btrfs_fs_info
*info
= NULL
;
996 struct bio_vec
*bvec
;
1000 bio_for_each_segment(bvec
, bio
, i
) {
1001 page
= bvec
->bv_page
;
1002 if (page
->private == EXTENT_PAGE_PRIVATE
) {
1003 length
+= bvec
->bv_len
;
1006 if (!page
->private) {
1007 length
+= bvec
->bv_len
;
1010 length
= bvec
->bv_len
;
1011 buf_len
= page
->private >> 2;
1012 start
= page_offset(page
) + bvec
->bv_offset
;
1013 io_tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
1014 info
= BTRFS_I(page
->mapping
->host
)->root
->fs_info
;
1016 /* are we fully contained in this bio? */
1017 if (buf_len
<= length
)
1020 ret
= extent_range_uptodate(io_tree
, start
+ length
,
1021 start
+ buf_len
- 1);
1027 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1028 static void btrfs_end_io_csum(void *p
)
1030 static void btrfs_end_io_csum(struct work_struct
*work
)
1033 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1034 struct btrfs_fs_info
*fs_info
= p
;
1036 struct btrfs_fs_info
*fs_info
= container_of(work
,
1037 struct btrfs_fs_info
,
1040 unsigned long flags
;
1041 struct end_io_wq
*end_io_wq
;
1043 struct list_head
*next
;
1048 spin_lock_irqsave(&fs_info
->end_io_work_lock
, flags
);
1049 if (list_empty(&fs_info
->end_io_work_list
)) {
1050 spin_unlock_irqrestore(&fs_info
->end_io_work_lock
,
1054 next
= fs_info
->end_io_work_list
.next
;
1056 spin_unlock_irqrestore(&fs_info
->end_io_work_lock
, flags
);
1058 end_io_wq
= list_entry(next
, struct end_io_wq
, list
);
1060 bio
= end_io_wq
->bio
;
1061 if (end_io_wq
->metadata
&& !bio_ready_for_csum(bio
)) {
1062 spin_lock_irqsave(&fs_info
->end_io_work_lock
, flags
);
1063 was_empty
= list_empty(&fs_info
->end_io_work_list
);
1064 list_add_tail(&end_io_wq
->list
,
1065 &fs_info
->end_io_work_list
);
1066 spin_unlock_irqrestore(&fs_info
->end_io_work_lock
,
1072 error
= end_io_wq
->error
;
1073 bio
->bi_private
= end_io_wq
->private;
1074 bio
->bi_end_io
= end_io_wq
->end_io
;
1076 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
1077 bio_endio(bio
, bio
->bi_size
, error
);
1079 bio_endio(bio
, error
);
1084 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1085 static void btrfs_async_submit_work(void *p
)
1087 static void btrfs_async_submit_work(struct work_struct
*work
)
1090 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1091 struct btrfs_fs_info
*fs_info
= p
;
1093 struct btrfs_fs_info
*fs_info
= container_of(work
,
1094 struct btrfs_fs_info
,
1097 struct async_submit_bio
*async
;
1098 struct list_head
*next
;
1101 spin_lock(&fs_info
->async_submit_work_lock
);
1102 if (list_empty(&fs_info
->async_submit_work_list
)) {
1103 spin_unlock(&fs_info
->async_submit_work_lock
);
1106 next
= fs_info
->async_submit_work_list
.next
;
1108 spin_unlock(&fs_info
->async_submit_work_lock
);
1110 async
= list_entry(next
, struct async_submit_bio
, list
);
1111 async
->submit_bio_hook(async
->inode
, async
->rw
, async
->bio
,
1117 struct btrfs_root
*open_ctree(struct super_block
*sb
,
1118 struct btrfs_fs_devices
*fs_devices
)
1125 struct btrfs_root
*extent_root
= kmalloc(sizeof(struct btrfs_root
),
1127 struct btrfs_root
*tree_root
= kmalloc(sizeof(struct btrfs_root
),
1129 struct btrfs_fs_info
*fs_info
= kzalloc(sizeof(*fs_info
),
1131 struct btrfs_root
*chunk_root
= kmalloc(sizeof(struct btrfs_root
),
1133 struct btrfs_root
*dev_root
= kmalloc(sizeof(struct btrfs_root
),
1137 struct btrfs_super_block
*disk_super
;
1139 if (!extent_root
|| !tree_root
|| !fs_info
) {
1143 end_io_workqueue
= create_workqueue("btrfs-end-io");
1144 BUG_ON(!end_io_workqueue
);
1145 async_submit_workqueue
= create_workqueue("btrfs-async-submit");
1147 INIT_RADIX_TREE(&fs_info
->fs_roots_radix
, GFP_NOFS
);
1148 INIT_LIST_HEAD(&fs_info
->trans_list
);
1149 INIT_LIST_HEAD(&fs_info
->dead_roots
);
1150 INIT_LIST_HEAD(&fs_info
->hashers
);
1151 INIT_LIST_HEAD(&fs_info
->end_io_work_list
);
1152 INIT_LIST_HEAD(&fs_info
->async_submit_work_list
);
1153 spin_lock_init(&fs_info
->hash_lock
);
1154 spin_lock_init(&fs_info
->end_io_work_lock
);
1155 spin_lock_init(&fs_info
->async_submit_work_lock
);
1156 spin_lock_init(&fs_info
->delalloc_lock
);
1157 spin_lock_init(&fs_info
->new_trans_lock
);
1159 init_completion(&fs_info
->kobj_unregister
);
1160 sb_set_blocksize(sb
, BTRFS_SUPER_INFO_SIZE
);
1161 fs_info
->tree_root
= tree_root
;
1162 fs_info
->extent_root
= extent_root
;
1163 fs_info
->chunk_root
= chunk_root
;
1164 fs_info
->dev_root
= dev_root
;
1165 fs_info
->fs_devices
= fs_devices
;
1166 INIT_LIST_HEAD(&fs_info
->dirty_cowonly_roots
);
1167 INIT_LIST_HEAD(&fs_info
->space_info
);
1168 btrfs_mapping_init(&fs_info
->mapping_tree
);
1170 fs_info
->max_extent
= (u64
)-1;
1171 fs_info
->max_inline
= 8192 * 1024;
1172 setup_bdi(fs_info
, &fs_info
->bdi
);
1173 fs_info
->btree_inode
= new_inode(sb
);
1174 fs_info
->btree_inode
->i_ino
= 1;
1175 fs_info
->btree_inode
->i_nlink
= 1;
1178 * we set the i_size on the btree inode to the max possible int.
1179 * the real end of the address space is determined by all of
1180 * the devices in the system
1182 fs_info
->btree_inode
->i_size
= OFFSET_MAX
;
1183 fs_info
->btree_inode
->i_mapping
->a_ops
= &btree_aops
;
1184 fs_info
->btree_inode
->i_mapping
->backing_dev_info
= &fs_info
->bdi
;
1186 extent_io_tree_init(&BTRFS_I(fs_info
->btree_inode
)->io_tree
,
1187 fs_info
->btree_inode
->i_mapping
,
1189 extent_map_tree_init(&BTRFS_I(fs_info
->btree_inode
)->extent_tree
,
1192 BTRFS_I(fs_info
->btree_inode
)->io_tree
.ops
= &btree_extent_io_ops
;
1194 extent_io_tree_init(&fs_info
->free_space_cache
,
1195 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1196 extent_io_tree_init(&fs_info
->block_group_cache
,
1197 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1198 extent_io_tree_init(&fs_info
->pinned_extents
,
1199 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1200 extent_io_tree_init(&fs_info
->pending_del
,
1201 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1202 extent_io_tree_init(&fs_info
->extent_ins
,
1203 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1204 fs_info
->do_barriers
= 1;
1206 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1207 INIT_WORK(&fs_info
->end_io_work
, btrfs_end_io_csum
, fs_info
);
1208 INIT_WORK(&fs_info
->async_submit_work
, btrfs_async_submit_work
,
1210 INIT_WORK(&fs_info
->trans_work
, btrfs_transaction_cleaner
, fs_info
);
1212 INIT_WORK(&fs_info
->end_io_work
, btrfs_end_io_csum
);
1213 INIT_WORK(&fs_info
->async_submit_work
, btrfs_async_submit_work
);
1214 INIT_DELAYED_WORK(&fs_info
->trans_work
, btrfs_transaction_cleaner
);
1216 BTRFS_I(fs_info
->btree_inode
)->root
= tree_root
;
1217 memset(&BTRFS_I(fs_info
->btree_inode
)->location
, 0,
1218 sizeof(struct btrfs_key
));
1219 insert_inode_hash(fs_info
->btree_inode
);
1220 mapping_set_gfp_mask(fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1222 mutex_init(&fs_info
->trans_mutex
);
1223 mutex_init(&fs_info
->fs_mutex
);
1226 ret
= add_hasher(fs_info
, "crc32c");
1228 printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
1233 __setup_root(4096, 4096, 4096, 4096, tree_root
,
1234 fs_info
, BTRFS_ROOT_TREE_OBJECTID
);
1236 fs_info
->sb_buffer
= read_tree_block(tree_root
,
1237 BTRFS_SUPER_INFO_OFFSET
,
1240 if (!fs_info
->sb_buffer
)
1243 read_extent_buffer(fs_info
->sb_buffer
, &fs_info
->super_copy
, 0,
1244 sizeof(fs_info
->super_copy
));
1246 read_extent_buffer(fs_info
->sb_buffer
, fs_info
->fsid
,
1247 (unsigned long)btrfs_super_fsid(fs_info
->sb_buffer
),
1250 disk_super
= &fs_info
->super_copy
;
1251 if (!btrfs_super_root(disk_super
))
1252 goto fail_sb_buffer
;
1254 if (btrfs_super_num_devices(disk_super
) != fs_devices
->num_devices
) {
1255 printk("Btrfs: wanted %llu devices, but found %llu\n",
1256 (unsigned long long)btrfs_super_num_devices(disk_super
),
1257 (unsigned long long)fs_devices
->num_devices
);
1258 goto fail_sb_buffer
;
1260 fs_info
->bdi
.ra_pages
*= btrfs_super_num_devices(disk_super
);
1262 nodesize
= btrfs_super_nodesize(disk_super
);
1263 leafsize
= btrfs_super_leafsize(disk_super
);
1264 sectorsize
= btrfs_super_sectorsize(disk_super
);
1265 stripesize
= btrfs_super_stripesize(disk_super
);
1266 tree_root
->nodesize
= nodesize
;
1267 tree_root
->leafsize
= leafsize
;
1268 tree_root
->sectorsize
= sectorsize
;
1269 tree_root
->stripesize
= stripesize
;
1270 sb_set_blocksize(sb
, sectorsize
);
1272 if (strncmp((char *)(&disk_super
->magic
), BTRFS_MAGIC
,
1273 sizeof(disk_super
->magic
))) {
1274 printk("btrfs: valid FS not found on %s\n", sb
->s_id
);
1275 goto fail_sb_buffer
;
1278 mutex_lock(&fs_info
->fs_mutex
);
1280 ret
= btrfs_read_sys_array(tree_root
);
1282 printk("btrfs: failed to read the system array on %s\n",
1284 goto fail_sys_array
;
1287 blocksize
= btrfs_level_size(tree_root
,
1288 btrfs_super_chunk_root_level(disk_super
));
1290 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
1291 chunk_root
, fs_info
, BTRFS_CHUNK_TREE_OBJECTID
);
1293 chunk_root
->node
= read_tree_block(chunk_root
,
1294 btrfs_super_chunk_root(disk_super
),
1296 BUG_ON(!chunk_root
->node
);
1298 read_extent_buffer(chunk_root
->node
, fs_info
->chunk_tree_uuid
,
1299 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root
->node
),
1302 ret
= btrfs_read_chunk_tree(chunk_root
);
1305 blocksize
= btrfs_level_size(tree_root
,
1306 btrfs_super_root_level(disk_super
));
1309 tree_root
->node
= read_tree_block(tree_root
,
1310 btrfs_super_root(disk_super
),
1312 if (!tree_root
->node
)
1313 goto fail_sb_buffer
;
1316 ret
= find_and_setup_root(tree_root
, fs_info
,
1317 BTRFS_EXTENT_TREE_OBJECTID
, extent_root
);
1319 goto fail_tree_root
;
1320 extent_root
->track_dirty
= 1;
1322 ret
= find_and_setup_root(tree_root
, fs_info
,
1323 BTRFS_DEV_TREE_OBJECTID
, dev_root
);
1324 dev_root
->track_dirty
= 1;
1327 goto fail_extent_root
;
1329 btrfs_read_block_groups(extent_root
);
1331 fs_info
->generation
= btrfs_super_generation(disk_super
) + 1;
1332 fs_info
->data_alloc_profile
= (u64
)-1;
1333 fs_info
->metadata_alloc_profile
= (u64
)-1;
1334 fs_info
->system_alloc_profile
= fs_info
->metadata_alloc_profile
;
1336 mutex_unlock(&fs_info
->fs_mutex
);
1340 free_extent_buffer(extent_root
->node
);
1342 free_extent_buffer(tree_root
->node
);
1344 mutex_unlock(&fs_info
->fs_mutex
);
1346 free_extent_buffer(fs_info
->sb_buffer
);
1347 extent_io_tree_empty_lru(&BTRFS_I(fs_info
->btree_inode
)->io_tree
);
1349 iput(fs_info
->btree_inode
);
1351 close_all_devices(fs_info
);
1352 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
1356 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
1357 bdi_destroy(&fs_info
->bdi
);
1360 return ERR_PTR(err
);
1363 static void btrfs_end_buffer_write_sync(struct buffer_head
*bh
, int uptodate
)
1365 char b
[BDEVNAME_SIZE
];
1368 set_buffer_uptodate(bh
);
1370 if (!buffer_eopnotsupp(bh
) && printk_ratelimit()) {
1371 printk(KERN_WARNING
"lost page write due to "
1372 "I/O error on %s\n",
1373 bdevname(bh
->b_bdev
, b
));
1375 set_buffer_write_io_error(bh
);
1376 clear_buffer_uptodate(bh
);
1382 int write_all_supers(struct btrfs_root
*root
)
1384 struct list_head
*cur
;
1385 struct list_head
*head
= &root
->fs_info
->fs_devices
->devices
;
1386 struct btrfs_device
*dev
;
1387 struct extent_buffer
*sb
;
1388 struct btrfs_dev_item
*dev_item
;
1389 struct buffer_head
*bh
;
1393 do_barriers
= !btrfs_test_opt(root
, NOBARRIER
);
1395 sb
= root
->fs_info
->sb_buffer
;
1396 dev_item
= (struct btrfs_dev_item
*)offsetof(struct btrfs_super_block
,
1398 list_for_each(cur
, head
) {
1399 dev
= list_entry(cur
, struct btrfs_device
, dev_list
);
1400 btrfs_set_device_type(sb
, dev_item
, dev
->type
);
1401 btrfs_set_device_id(sb
, dev_item
, dev
->devid
);
1402 btrfs_set_device_total_bytes(sb
, dev_item
, dev
->total_bytes
);
1403 btrfs_set_device_bytes_used(sb
, dev_item
, dev
->bytes_used
);
1404 btrfs_set_device_io_align(sb
, dev_item
, dev
->io_align
);
1405 btrfs_set_device_io_width(sb
, dev_item
, dev
->io_width
);
1406 btrfs_set_device_sector_size(sb
, dev_item
, dev
->sector_size
);
1407 write_extent_buffer(sb
, dev
->uuid
,
1408 (unsigned long)btrfs_device_uuid(dev_item
),
1411 btrfs_set_header_flag(sb
, BTRFS_HEADER_FLAG_WRITTEN
);
1412 csum_tree_block(root
, sb
, 0);
1414 bh
= __getblk(dev
->bdev
, BTRFS_SUPER_INFO_OFFSET
/
1415 root
->fs_info
->sb
->s_blocksize
,
1416 BTRFS_SUPER_INFO_SIZE
);
1418 read_extent_buffer(sb
, bh
->b_data
, 0, BTRFS_SUPER_INFO_SIZE
);
1419 dev
->pending_io
= bh
;
1422 set_buffer_uptodate(bh
);
1424 bh
->b_end_io
= btrfs_end_buffer_write_sync
;
1426 if (do_barriers
&& dev
->barriers
) {
1427 ret
= submit_bh(WRITE_BARRIER
, bh
);
1428 if (ret
== -EOPNOTSUPP
) {
1429 printk("btrfs: disabling barriers on dev %s\n",
1431 set_buffer_uptodate(bh
);
1435 ret
= submit_bh(WRITE
, bh
);
1438 ret
= submit_bh(WRITE
, bh
);
1443 list_for_each(cur
, head
) {
1444 dev
= list_entry(cur
, struct btrfs_device
, dev_list
);
1445 BUG_ON(!dev
->pending_io
);
1446 bh
= dev
->pending_io
;
1448 if (!buffer_uptodate(dev
->pending_io
)) {
1449 if (do_barriers
&& dev
->barriers
) {
1450 printk("btrfs: disabling barriers on dev %s\n",
1452 set_buffer_uptodate(bh
);
1456 ret
= submit_bh(WRITE
, bh
);
1459 BUG_ON(!buffer_uptodate(bh
));
1465 dev
->pending_io
= NULL
;
1471 int write_ctree_super(struct btrfs_trans_handle
*trans
, struct btrfs_root
1476 ret
= write_all_supers(root
);
1478 if (!btrfs_test_opt(root
, NOBARRIER
))
1479 blkdev_issue_flush(sb
->s_bdev
, NULL
);
1480 set_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
, super
);
1481 ret
= sync_page_range_nolock(btree_inode
, btree_inode
->i_mapping
,
1482 super
->start
, super
->len
);
1483 if (!btrfs_test_opt(root
, NOBARRIER
))
1484 blkdev_issue_flush(sb
->s_bdev
, NULL
);
1489 int btrfs_free_fs_root(struct btrfs_fs_info
*fs_info
, struct btrfs_root
*root
)
1491 radix_tree_delete(&fs_info
->fs_roots_radix
,
1492 (unsigned long)root
->root_key
.objectid
);
1494 btrfs_sysfs_del_root(root
);
1498 free_extent_buffer(root
->node
);
1499 if (root
->commit_root
)
1500 free_extent_buffer(root
->commit_root
);
1507 static int del_fs_roots(struct btrfs_fs_info
*fs_info
)
1510 struct btrfs_root
*gang
[8];
1514 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
1519 for (i
= 0; i
< ret
; i
++)
1520 btrfs_free_fs_root(fs_info
, gang
[i
]);
1525 int close_ctree(struct btrfs_root
*root
)
1528 struct btrfs_trans_handle
*trans
;
1529 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1531 fs_info
->closing
= 1;
1532 btrfs_transaction_flush_work(root
);
1533 mutex_lock(&fs_info
->fs_mutex
);
1534 btrfs_defrag_dirty_roots(root
->fs_info
);
1535 trans
= btrfs_start_transaction(root
, 1);
1536 ret
= btrfs_commit_transaction(trans
, root
);
1537 /* run commit again to drop the original snapshot */
1538 trans
= btrfs_start_transaction(root
, 1);
1539 btrfs_commit_transaction(trans
, root
);
1540 ret
= btrfs_write_and_wait_transaction(NULL
, root
);
1542 write_ctree_super(NULL
, root
);
1543 mutex_unlock(&fs_info
->fs_mutex
);
1545 btrfs_transaction_flush_work(root
);
1547 if (fs_info
->delalloc_bytes
) {
1548 printk("btrfs: at unmount delalloc count %Lu\n",
1549 fs_info
->delalloc_bytes
);
1551 if (fs_info
->extent_root
->node
)
1552 free_extent_buffer(fs_info
->extent_root
->node
);
1554 if (fs_info
->tree_root
->node
)
1555 free_extent_buffer(fs_info
->tree_root
->node
);
1557 if (root
->fs_info
->chunk_root
->node
);
1558 free_extent_buffer(root
->fs_info
->chunk_root
->node
);
1560 if (root
->fs_info
->dev_root
->node
);
1561 free_extent_buffer(root
->fs_info
->dev_root
->node
);
1563 free_extent_buffer(fs_info
->sb_buffer
);
1565 btrfs_free_block_groups(root
->fs_info
);
1566 del_fs_roots(fs_info
);
1568 filemap_write_and_wait(fs_info
->btree_inode
->i_mapping
);
1570 extent_io_tree_empty_lru(&fs_info
->free_space_cache
);
1571 extent_io_tree_empty_lru(&fs_info
->block_group_cache
);
1572 extent_io_tree_empty_lru(&fs_info
->pinned_extents
);
1573 extent_io_tree_empty_lru(&fs_info
->pending_del
);
1574 extent_io_tree_empty_lru(&fs_info
->extent_ins
);
1575 extent_io_tree_empty_lru(&BTRFS_I(fs_info
->btree_inode
)->io_tree
);
1577 flush_workqueue(end_io_workqueue
);
1578 flush_workqueue(async_submit_workqueue
);
1580 truncate_inode_pages(fs_info
->btree_inode
->i_mapping
, 0);
1582 flush_workqueue(end_io_workqueue
);
1583 destroy_workqueue(end_io_workqueue
);
1585 flush_workqueue(async_submit_workqueue
);
1586 destroy_workqueue(async_submit_workqueue
);
1588 iput(fs_info
->btree_inode
);
1590 while(!list_empty(&fs_info
->hashers
)) {
1591 struct btrfs_hasher
*hasher
;
1592 hasher
= list_entry(fs_info
->hashers
.next
, struct btrfs_hasher
,
1594 list_del(&hasher
->hashers
);
1595 crypto_free_hash(&fs_info
->hash_tfm
);
1599 close_all_devices(fs_info
);
1600 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
1602 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
1603 bdi_destroy(&fs_info
->bdi
);
1606 kfree(fs_info
->extent_root
);
1607 kfree(fs_info
->tree_root
);
1608 kfree(fs_info
->chunk_root
);
1609 kfree(fs_info
->dev_root
);
1613 int btrfs_buffer_uptodate(struct extent_buffer
*buf
)
1615 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
1616 return extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
, buf
);
1619 int btrfs_set_buffer_uptodate(struct extent_buffer
*buf
)
1621 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
1622 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
,
1626 void btrfs_mark_buffer_dirty(struct extent_buffer
*buf
)
1628 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1629 u64 transid
= btrfs_header_generation(buf
);
1630 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1632 if (transid
!= root
->fs_info
->generation
) {
1633 printk(KERN_CRIT
"transid mismatch buffer %llu, found %Lu running %Lu\n",
1634 (unsigned long long)buf
->start
,
1635 transid
, root
->fs_info
->generation
);
1638 set_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
, buf
);
1641 void btrfs_throttle(struct btrfs_root
*root
)
1643 struct backing_dev_info
*bdi
;
1645 bdi
= root
->fs_info
->sb
->s_bdev
->bd_inode
->i_mapping
->backing_dev_info
;
1646 if (root
->fs_info
->throttles
&& bdi_write_congested(bdi
)) {
1647 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18)
1648 congestion_wait(WRITE
, HZ
/20);
1650 blk_congestion_wait(WRITE
, HZ
/20);
1655 void btrfs_btree_balance_dirty(struct btrfs_root
*root
, unsigned long nr
)
1657 balance_dirty_pages_ratelimited_nr(
1658 root
->fs_info
->btree_inode
->i_mapping
, 1);
1661 void btrfs_set_buffer_defrag(struct extent_buffer
*buf
)
1663 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1664 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1665 set_extent_bits(&BTRFS_I(btree_inode
)->io_tree
, buf
->start
,
1666 buf
->start
+ buf
->len
- 1, EXTENT_DEFRAG
, GFP_NOFS
);
1669 void btrfs_set_buffer_defrag_done(struct extent_buffer
*buf
)
1671 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1672 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1673 set_extent_bits(&BTRFS_I(btree_inode
)->io_tree
, buf
->start
,
1674 buf
->start
+ buf
->len
- 1, EXTENT_DEFRAG_DONE
,
1678 int btrfs_buffer_defrag(struct extent_buffer
*buf
)
1680 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1681 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1682 return test_range_bit(&BTRFS_I(btree_inode
)->io_tree
,
1683 buf
->start
, buf
->start
+ buf
->len
- 1, EXTENT_DEFRAG
, 0);
1686 int btrfs_buffer_defrag_done(struct extent_buffer
*buf
)
1688 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1689 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1690 return test_range_bit(&BTRFS_I(btree_inode
)->io_tree
,
1691 buf
->start
, buf
->start
+ buf
->len
- 1,
1692 EXTENT_DEFRAG_DONE
, 0);
1695 int btrfs_clear_buffer_defrag_done(struct extent_buffer
*buf
)
1697 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1698 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1699 return clear_extent_bits(&BTRFS_I(btree_inode
)->io_tree
,
1700 buf
->start
, buf
->start
+ buf
->len
- 1,
1701 EXTENT_DEFRAG_DONE
, GFP_NOFS
);
1704 int btrfs_clear_buffer_defrag(struct extent_buffer
*buf
)
1706 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1707 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1708 return clear_extent_bits(&BTRFS_I(btree_inode
)->io_tree
,
1709 buf
->start
, buf
->start
+ buf
->len
- 1,
1710 EXTENT_DEFRAG
, GFP_NOFS
);
1713 int btrfs_read_buffer(struct extent_buffer
*buf
)
1715 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1717 ret
= btree_read_extent_buffer_pages(root
, buf
, 0);
1719 buf
->flags
|= EXTENT_UPTODATE
;
1724 static struct extent_io_ops btree_extent_io_ops
= {
1725 .writepage_io_hook
= btree_writepage_io_hook
,
1726 .readpage_end_io_hook
= btree_readpage_end_io_hook
,
1727 .submit_bio_hook
= btree_submit_bio_hook
,
1728 /* note we're sharing with inode.c for the merge bio hook */
1729 .merge_bio_hook
= btrfs_merge_bio_hook
,