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
);
83 BTRFS_I(inode
)->root
->fs_info
->fs_devices
->latest_bdev
;
84 spin_unlock(&em_tree
->lock
);
87 spin_unlock(&em_tree
->lock
);
89 em
= alloc_extent_map(GFP_NOFS
);
91 em
= ERR_PTR(-ENOMEM
);
97 em
->bdev
= BTRFS_I(inode
)->root
->fs_info
->fs_devices
->latest_bdev
;
99 spin_lock(&em_tree
->lock
);
100 ret
= add_extent_mapping(em_tree
, em
);
101 if (ret
== -EEXIST
) {
102 u64 failed_start
= em
->start
;
103 u64 failed_len
= em
->len
;
105 printk("failed to insert %Lu %Lu -> %Lu into tree\n",
106 em
->start
, em
->len
, em
->block_start
);
108 em
= lookup_extent_mapping(em_tree
, start
, len
);
110 printk("after failing, found %Lu %Lu %Lu\n",
111 em
->start
, em
->len
, em
->block_start
);
114 em
= lookup_extent_mapping(em_tree
, failed_start
,
117 printk("double failure lookup gives us "
118 "%Lu %Lu -> %Lu\n", em
->start
,
119 em
->len
, em
->block_start
);
128 spin_unlock(&em_tree
->lock
);
136 u32
btrfs_csum_data(struct btrfs_root
*root
, char *data
, u32 seed
, size_t len
)
138 return btrfs_crc32c(seed
, data
, len
);
141 void btrfs_csum_final(u32 crc
, char *result
)
143 *(__le32
*)result
= ~cpu_to_le32(crc
);
146 static int csum_tree_block(struct btrfs_root
*root
, struct extent_buffer
*buf
,
149 char result
[BTRFS_CRC32_SIZE
];
151 unsigned long cur_len
;
152 unsigned long offset
= BTRFS_CSUM_SIZE
;
153 char *map_token
= NULL
;
155 unsigned long map_start
;
156 unsigned long map_len
;
160 len
= buf
->len
- offset
;
162 err
= map_private_extent_buffer(buf
, offset
, 32,
164 &map_start
, &map_len
, KM_USER0
);
166 printk("failed to map extent buffer! %lu\n",
170 cur_len
= min(len
, map_len
- (offset
- map_start
));
171 crc
= btrfs_csum_data(root
, kaddr
+ offset
- map_start
,
175 unmap_extent_buffer(buf
, map_token
, KM_USER0
);
177 btrfs_csum_final(crc
, result
);
180 int from_this_trans
= 0;
182 if (root
->fs_info
->running_transaction
&&
183 btrfs_header_generation(buf
) ==
184 root
->fs_info
->running_transaction
->transid
)
187 /* FIXME, this is not good */
188 if (memcmp_extent_buffer(buf
, result
, 0, BTRFS_CRC32_SIZE
)) {
191 memcpy(&found
, result
, BTRFS_CRC32_SIZE
);
193 read_extent_buffer(buf
, &val
, 0, BTRFS_CRC32_SIZE
);
194 printk("btrfs: %s checksum verify failed on %llu "
195 "wanted %X found %X from_this_trans %d "
197 root
->fs_info
->sb
->s_id
,
198 buf
->start
, val
, found
, from_this_trans
,
199 btrfs_header_level(buf
));
203 write_extent_buffer(buf
, result
, 0, BTRFS_CRC32_SIZE
);
208 static int verify_parent_transid(struct extent_io_tree
*io_tree
,
209 struct extent_buffer
*eb
, u64 parent_transid
)
213 if (!parent_transid
|| btrfs_header_generation(eb
) == parent_transid
)
216 lock_extent(io_tree
, eb
->start
, eb
->start
+ eb
->len
- 1, GFP_NOFS
);
217 if (extent_buffer_uptodate(io_tree
, eb
) &&
218 btrfs_header_generation(eb
) == parent_transid
) {
222 printk("parent transid verify failed on %llu wanted %llu found %llu\n",
223 (unsigned long long)eb
->start
,
224 (unsigned long long)parent_transid
,
225 (unsigned long long)btrfs_header_generation(eb
));
228 clear_extent_buffer_uptodate(io_tree
, eb
);
229 unlock_extent(io_tree
, eb
->start
, eb
->start
+ eb
->len
- 1,
235 static int btree_read_extent_buffer_pages(struct btrfs_root
*root
,
236 struct extent_buffer
*eb
,
237 u64 start
, u64 parent_transid
)
239 struct extent_io_tree
*io_tree
;
244 io_tree
= &BTRFS_I(root
->fs_info
->btree_inode
)->io_tree
;
246 ret
= read_extent_buffer_pages(io_tree
, eb
, start
, 1,
247 btree_get_extent
, mirror_num
);
249 !verify_parent_transid(io_tree
, eb
, parent_transid
))
252 num_copies
= btrfs_num_copies(&root
->fs_info
->mapping_tree
,
258 if (mirror_num
> num_copies
)
264 int csum_dirty_buffer(struct btrfs_root
*root
, struct page
*page
)
266 struct extent_io_tree
*tree
;
267 u64 start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
271 struct extent_buffer
*eb
;
274 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
276 if (page
->private == EXTENT_PAGE_PRIVATE
)
280 len
= page
->private >> 2;
284 eb
= alloc_extent_buffer(tree
, start
, len
, page
, GFP_NOFS
);
285 ret
= btree_read_extent_buffer_pages(root
, eb
, start
+ PAGE_CACHE_SIZE
,
286 btrfs_header_generation(eb
));
288 btrfs_clear_buffer_defrag(eb
);
289 found_start
= btrfs_header_bytenr(eb
);
290 if (found_start
!= start
) {
291 printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
292 start
, found_start
, len
);
296 if (eb
->first_page
!= page
) {
297 printk("bad first page %lu %lu\n", eb
->first_page
->index
,
302 if (!PageUptodate(page
)) {
303 printk("csum not up to date page %lu\n", page
->index
);
307 found_level
= btrfs_header_level(eb
);
308 spin_lock(&root
->fs_info
->hash_lock
);
309 btrfs_set_header_flag(eb
, BTRFS_HEADER_FLAG_WRITTEN
);
310 spin_unlock(&root
->fs_info
->hash_lock
);
311 csum_tree_block(root
, eb
, 0);
313 free_extent_buffer(eb
);
318 static int btree_writepage_io_hook(struct page
*page
, u64 start
, u64 end
)
320 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
322 csum_dirty_buffer(root
, page
);
326 int btree_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
327 struct extent_state
*state
)
329 struct extent_io_tree
*tree
;
333 struct extent_buffer
*eb
;
334 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
337 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
338 if (page
->private == EXTENT_PAGE_PRIVATE
)
342 len
= page
->private >> 2;
346 eb
= alloc_extent_buffer(tree
, start
, len
, page
, GFP_NOFS
);
348 btrfs_clear_buffer_defrag(eb
);
349 found_start
= btrfs_header_bytenr(eb
);
350 if (found_start
!= start
) {
354 if (eb
->first_page
!= page
) {
355 printk("bad first page %lu %lu\n", eb
->first_page
->index
,
361 if (memcmp_extent_buffer(eb
, root
->fs_info
->fsid
,
362 (unsigned long)btrfs_header_fsid(eb
),
364 printk("bad fsid on block %Lu\n", eb
->start
);
368 found_level
= btrfs_header_level(eb
);
370 ret
= csum_tree_block(root
, eb
, 1);
374 end
= min_t(u64
, eb
->len
, PAGE_CACHE_SIZE
);
375 end
= eb
->start
+ end
- 1;
376 release_extent_buffer_tail_pages(eb
);
378 free_extent_buffer(eb
);
383 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
384 static void end_workqueue_bio(struct bio
*bio
, int err
)
386 static int end_workqueue_bio(struct bio
*bio
,
387 unsigned int bytes_done
, int err
)
390 struct end_io_wq
*end_io_wq
= bio
->bi_private
;
391 struct btrfs_fs_info
*fs_info
;
394 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
399 fs_info
= end_io_wq
->info
;
400 spin_lock_irqsave(&fs_info
->end_io_work_lock
, flags
);
401 end_io_wq
->error
= err
;
402 list_add_tail(&end_io_wq
->list
, &fs_info
->end_io_work_list
);
403 spin_unlock_irqrestore(&fs_info
->end_io_work_lock
, flags
);
404 queue_work(end_io_workqueue
, &fs_info
->end_io_work
);
406 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
411 int btrfs_bio_wq_end_io(struct btrfs_fs_info
*info
, struct bio
*bio
,
414 struct end_io_wq
*end_io_wq
;
415 end_io_wq
= kmalloc(sizeof(*end_io_wq
), GFP_NOFS
);
419 end_io_wq
->private = bio
->bi_private
;
420 end_io_wq
->end_io
= bio
->bi_end_io
;
421 end_io_wq
->info
= info
;
422 end_io_wq
->error
= 0;
423 end_io_wq
->bio
= bio
;
424 end_io_wq
->metadata
= metadata
;
426 bio
->bi_private
= end_io_wq
;
427 bio
->bi_end_io
= end_workqueue_bio
;
431 int btrfs_wq_submit_bio(struct btrfs_fs_info
*fs_info
, struct inode
*inode
,
432 int rw
, struct bio
*bio
, int mirror_num
,
433 extent_submit_bio_hook_t
*submit_bio_hook
)
435 struct async_submit_bio
*async
;
437 async
= kmalloc(sizeof(*async
), GFP_NOFS
);
441 async
->inode
= inode
;
444 async
->mirror_num
= mirror_num
;
445 async
->submit_bio_hook
= submit_bio_hook
;
447 spin_lock(&fs_info
->async_submit_work_lock
);
448 list_add_tail(&async
->list
, &fs_info
->async_submit_work_list
);
449 atomic_inc(&fs_info
->nr_async_submits
);
450 spin_unlock(&fs_info
->async_submit_work_lock
);
452 queue_work(async_submit_workqueue
, &fs_info
->async_submit_work
);
456 static int __btree_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
459 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
463 offset
= bio
->bi_sector
<< 9;
465 if (rw
& (1 << BIO_RW
)) {
466 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
, mirror_num
);
469 ret
= btrfs_bio_wq_end_io(root
->fs_info
, bio
, 1);
472 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
, mirror_num
);
475 static int btree_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
478 if (!(rw
& (1 << BIO_RW
))) {
479 return __btree_submit_bio_hook(inode
, rw
, bio
, mirror_num
);
481 return btrfs_wq_submit_bio(BTRFS_I(inode
)->root
->fs_info
,
482 inode
, rw
, bio
, mirror_num
,
483 __btree_submit_bio_hook
);
486 static int btree_writepage(struct page
*page
, struct writeback_control
*wbc
)
488 struct extent_io_tree
*tree
;
489 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
490 return extent_write_full_page(tree
, page
, btree_get_extent
, wbc
);
493 static int btree_writepages(struct address_space
*mapping
,
494 struct writeback_control
*wbc
)
496 struct extent_io_tree
*tree
;
497 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
498 if (wbc
->sync_mode
== WB_SYNC_NONE
) {
501 unsigned long thresh
= 96 * 1024 * 1024;
503 if (wbc
->for_kupdate
)
506 if (current_is_pdflush()) {
507 thresh
= 96 * 1024 * 1024;
509 thresh
= 8 * 1024 * 1024;
511 num_dirty
= count_range_bits(tree
, &start
, (u64
)-1,
512 thresh
, EXTENT_DIRTY
);
513 if (num_dirty
< thresh
) {
517 return extent_writepages(tree
, mapping
, btree_get_extent
, wbc
);
520 int btree_readpage(struct file
*file
, struct page
*page
)
522 struct extent_io_tree
*tree
;
523 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
524 return extent_read_full_page(tree
, page
, btree_get_extent
);
527 static int btree_releasepage(struct page
*page
, gfp_t gfp_flags
)
529 struct extent_io_tree
*tree
;
530 struct extent_map_tree
*map
;
533 if (page_count(page
) > 3) {
534 /* once for page->private, once for the caller, once
535 * once for the page cache
539 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
540 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
541 ret
= try_release_extent_state(map
, tree
, page
, gfp_flags
);
543 invalidate_extent_lru(tree
, page_offset(page
), PAGE_CACHE_SIZE
);
544 ClearPagePrivate(page
);
545 set_page_private(page
, 0);
546 page_cache_release(page
);
551 static void btree_invalidatepage(struct page
*page
, unsigned long offset
)
553 struct extent_io_tree
*tree
;
554 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
555 extent_invalidatepage(tree
, page
, offset
);
556 btree_releasepage(page
, GFP_NOFS
);
557 if (PagePrivate(page
)) {
558 invalidate_extent_lru(tree
, page_offset(page
), PAGE_CACHE_SIZE
);
559 ClearPagePrivate(page
);
560 set_page_private(page
, 0);
561 page_cache_release(page
);
566 static int btree_writepage(struct page
*page
, struct writeback_control
*wbc
)
568 struct buffer_head
*bh
;
569 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
570 struct buffer_head
*head
;
571 if (!page_has_buffers(page
)) {
572 create_empty_buffers(page
, root
->fs_info
->sb
->s_blocksize
,
573 (1 << BH_Dirty
)|(1 << BH_Uptodate
));
575 head
= page_buffers(page
);
578 if (buffer_dirty(bh
))
579 csum_tree_block(root
, bh
, 0);
580 bh
= bh
->b_this_page
;
581 } while (bh
!= head
);
582 return block_write_full_page(page
, btree_get_block
, wbc
);
586 static struct address_space_operations btree_aops
= {
587 .readpage
= btree_readpage
,
588 .writepage
= btree_writepage
,
589 .writepages
= btree_writepages
,
590 .releasepage
= btree_releasepage
,
591 .invalidatepage
= btree_invalidatepage
,
592 .sync_page
= block_sync_page
,
595 int readahead_tree_block(struct btrfs_root
*root
, u64 bytenr
, u32 blocksize
,
598 struct extent_buffer
*buf
= NULL
;
599 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
602 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
605 read_extent_buffer_pages(&BTRFS_I(btree_inode
)->io_tree
,
606 buf
, 0, 0, btree_get_extent
, 0);
607 free_extent_buffer(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
,
634 u32 blocksize
, u64 parent_transid
)
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, parent_transid
);
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
;
821 if (location
->objectid
== BTRFS_CHUNK_TREE_OBJECTID
)
822 return fs_info
->chunk_root
;
823 if (location
->objectid
== BTRFS_DEV_TREE_OBJECTID
)
824 return fs_info
->dev_root
;
826 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
827 (unsigned long)location
->objectid
);
831 root
= btrfs_read_fs_root_no_radix(fs_info
, location
);
834 ret
= radix_tree_insert(&fs_info
->fs_roots_radix
,
835 (unsigned long)root
->root_key
.objectid
,
838 free_extent_buffer(root
->node
);
842 ret
= btrfs_find_dead_roots(fs_info
->tree_root
,
843 root
->root_key
.objectid
, root
);
849 struct btrfs_root
*btrfs_read_fs_root(struct btrfs_fs_info
*fs_info
,
850 struct btrfs_key
*location
,
851 const char *name
, int namelen
)
853 struct btrfs_root
*root
;
856 root
= btrfs_read_fs_root_no_name(fs_info
, location
);
863 ret
= btrfs_set_root_name(root
, name
, namelen
);
865 free_extent_buffer(root
->node
);
870 ret
= btrfs_sysfs_add_root(root
);
872 free_extent_buffer(root
->node
);
881 static int add_hasher(struct btrfs_fs_info
*info
, char *type
) {
882 struct btrfs_hasher
*hasher
;
884 hasher
= kmalloc(sizeof(*hasher
), GFP_NOFS
);
887 hasher
->hash_tfm
= crypto_alloc_hash(type
, 0, CRYPTO_ALG_ASYNC
);
888 if (!hasher
->hash_tfm
) {
892 spin_lock(&info
->hash_lock
);
893 list_add(&hasher
->list
, &info
->hashers
);
894 spin_unlock(&info
->hash_lock
);
899 static int btrfs_congested_fn(void *congested_data
, int bdi_bits
)
901 struct btrfs_fs_info
*info
= (struct btrfs_fs_info
*)congested_data
;
903 int limit
= 256 * info
->fs_devices
->open_devices
;
904 struct list_head
*cur
;
905 struct btrfs_device
*device
;
906 struct backing_dev_info
*bdi
;
908 if ((bdi_bits
& (1 << BDI_write_congested
)) &&
909 atomic_read(&info
->nr_async_submits
) > limit
) {
913 list_for_each(cur
, &info
->fs_devices
->devices
) {
914 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
917 bdi
= blk_get_backing_dev_info(device
->bdev
);
918 if (bdi
&& bdi_congested(bdi
, bdi_bits
)) {
927 * this unplugs every device on the box, and it is only used when page
930 static void __unplug_io_fn(struct backing_dev_info
*bdi
, struct page
*page
)
932 struct list_head
*cur
;
933 struct btrfs_device
*device
;
934 struct btrfs_fs_info
*info
;
936 info
= (struct btrfs_fs_info
*)bdi
->unplug_io_data
;
937 list_for_each(cur
, &info
->fs_devices
->devices
) {
938 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
939 bdi
= blk_get_backing_dev_info(device
->bdev
);
940 if (bdi
->unplug_io_fn
) {
941 bdi
->unplug_io_fn(bdi
, page
);
946 void btrfs_unplug_io_fn(struct backing_dev_info
*bdi
, struct page
*page
)
949 struct extent_map_tree
*em_tree
;
950 struct extent_map
*em
;
951 struct address_space
*mapping
;
954 /* the generic O_DIRECT read code does this */
956 __unplug_io_fn(bdi
, page
);
961 * page->mapping may change at any time. Get a consistent copy
962 * and use that for everything below
965 mapping
= page
->mapping
;
969 inode
= mapping
->host
;
970 offset
= page_offset(page
);
972 em_tree
= &BTRFS_I(inode
)->extent_tree
;
973 spin_lock(&em_tree
->lock
);
974 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_CACHE_SIZE
);
975 spin_unlock(&em_tree
->lock
);
979 offset
= offset
- em
->start
;
980 btrfs_unplug_page(&BTRFS_I(inode
)->root
->fs_info
->mapping_tree
,
981 em
->block_start
+ offset
, page
);
985 static int setup_bdi(struct btrfs_fs_info
*info
, struct backing_dev_info
*bdi
)
987 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
990 bdi
->ra_pages
= default_backing_dev_info
.ra_pages
;
992 bdi
->capabilities
= default_backing_dev_info
.capabilities
;
993 bdi
->unplug_io_fn
= btrfs_unplug_io_fn
;
994 bdi
->unplug_io_data
= info
;
995 bdi
->congested_fn
= btrfs_congested_fn
;
996 bdi
->congested_data
= info
;
1000 static int bio_ready_for_csum(struct bio
*bio
)
1006 struct extent_io_tree
*io_tree
= NULL
;
1007 struct btrfs_fs_info
*info
= NULL
;
1008 struct bio_vec
*bvec
;
1012 bio_for_each_segment(bvec
, bio
, i
) {
1013 page
= bvec
->bv_page
;
1014 if (page
->private == EXTENT_PAGE_PRIVATE
) {
1015 length
+= bvec
->bv_len
;
1018 if (!page
->private) {
1019 length
+= bvec
->bv_len
;
1022 length
= bvec
->bv_len
;
1023 buf_len
= page
->private >> 2;
1024 start
= page_offset(page
) + bvec
->bv_offset
;
1025 io_tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
1026 info
= BTRFS_I(page
->mapping
->host
)->root
->fs_info
;
1028 /* are we fully contained in this bio? */
1029 if (buf_len
<= length
)
1032 ret
= extent_range_uptodate(io_tree
, start
+ length
,
1033 start
+ buf_len
- 1);
1039 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1040 static void btrfs_end_io_csum(void *p
)
1042 static void btrfs_end_io_csum(struct work_struct
*work
)
1045 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1046 struct btrfs_fs_info
*fs_info
= p
;
1048 struct btrfs_fs_info
*fs_info
= container_of(work
,
1049 struct btrfs_fs_info
,
1052 unsigned long flags
;
1053 struct end_io_wq
*end_io_wq
;
1055 struct list_head
*next
;
1060 spin_lock_irqsave(&fs_info
->end_io_work_lock
, flags
);
1061 if (list_empty(&fs_info
->end_io_work_list
)) {
1062 spin_unlock_irqrestore(&fs_info
->end_io_work_lock
,
1066 next
= fs_info
->end_io_work_list
.next
;
1068 spin_unlock_irqrestore(&fs_info
->end_io_work_lock
, flags
);
1070 end_io_wq
= list_entry(next
, struct end_io_wq
, list
);
1072 bio
= end_io_wq
->bio
;
1073 if (end_io_wq
->metadata
&& !bio_ready_for_csum(bio
)) {
1074 spin_lock_irqsave(&fs_info
->end_io_work_lock
, flags
);
1075 was_empty
= list_empty(&fs_info
->end_io_work_list
);
1076 list_add_tail(&end_io_wq
->list
,
1077 &fs_info
->end_io_work_list
);
1078 spin_unlock_irqrestore(&fs_info
->end_io_work_lock
,
1084 error
= end_io_wq
->error
;
1085 bio
->bi_private
= end_io_wq
->private;
1086 bio
->bi_end_io
= end_io_wq
->end_io
;
1088 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
1089 bio_endio(bio
, bio
->bi_size
, error
);
1091 bio_endio(bio
, error
);
1096 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1097 static void btrfs_async_submit_work(void *p
)
1099 static void btrfs_async_submit_work(struct work_struct
*work
)
1102 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1103 struct btrfs_fs_info
*fs_info
= p
;
1105 struct btrfs_fs_info
*fs_info
= container_of(work
,
1106 struct btrfs_fs_info
,
1109 struct async_submit_bio
*async
;
1110 struct list_head
*next
;
1113 spin_lock(&fs_info
->async_submit_work_lock
);
1114 if (list_empty(&fs_info
->async_submit_work_list
)) {
1115 spin_unlock(&fs_info
->async_submit_work_lock
);
1118 next
= fs_info
->async_submit_work_list
.next
;
1120 atomic_dec(&fs_info
->nr_async_submits
);
1121 spin_unlock(&fs_info
->async_submit_work_lock
);
1123 async
= list_entry(next
, struct async_submit_bio
, list
);
1124 async
->submit_bio_hook(async
->inode
, async
->rw
, async
->bio
,
1130 struct btrfs_root
*open_ctree(struct super_block
*sb
,
1131 struct btrfs_fs_devices
*fs_devices
,
1139 struct buffer_head
*bh
;
1140 struct btrfs_root
*extent_root
= kmalloc(sizeof(struct btrfs_root
),
1142 struct btrfs_root
*tree_root
= kmalloc(sizeof(struct btrfs_root
),
1144 struct btrfs_fs_info
*fs_info
= kzalloc(sizeof(*fs_info
),
1146 struct btrfs_root
*chunk_root
= kmalloc(sizeof(struct btrfs_root
),
1148 struct btrfs_root
*dev_root
= kmalloc(sizeof(struct btrfs_root
),
1152 struct btrfs_super_block
*disk_super
;
1154 if (!extent_root
|| !tree_root
|| !fs_info
) {
1158 end_io_workqueue
= create_workqueue("btrfs-end-io");
1159 BUG_ON(!end_io_workqueue
);
1160 async_submit_workqueue
= create_workqueue("btrfs-async-submit");
1162 INIT_RADIX_TREE(&fs_info
->fs_roots_radix
, GFP_NOFS
);
1163 INIT_LIST_HEAD(&fs_info
->trans_list
);
1164 INIT_LIST_HEAD(&fs_info
->dead_roots
);
1165 INIT_LIST_HEAD(&fs_info
->hashers
);
1166 INIT_LIST_HEAD(&fs_info
->end_io_work_list
);
1167 INIT_LIST_HEAD(&fs_info
->async_submit_work_list
);
1168 spin_lock_init(&fs_info
->hash_lock
);
1169 spin_lock_init(&fs_info
->end_io_work_lock
);
1170 spin_lock_init(&fs_info
->async_submit_work_lock
);
1171 spin_lock_init(&fs_info
->delalloc_lock
);
1172 spin_lock_init(&fs_info
->new_trans_lock
);
1174 init_completion(&fs_info
->kobj_unregister
);
1175 fs_info
->tree_root
= tree_root
;
1176 fs_info
->extent_root
= extent_root
;
1177 fs_info
->chunk_root
= chunk_root
;
1178 fs_info
->dev_root
= dev_root
;
1179 fs_info
->fs_devices
= fs_devices
;
1180 INIT_LIST_HEAD(&fs_info
->dirty_cowonly_roots
);
1181 INIT_LIST_HEAD(&fs_info
->space_info
);
1182 btrfs_mapping_init(&fs_info
->mapping_tree
);
1183 atomic_set(&fs_info
->nr_async_submits
, 0);
1185 fs_info
->max_extent
= (u64
)-1;
1186 fs_info
->max_inline
= 8192 * 1024;
1187 setup_bdi(fs_info
, &fs_info
->bdi
);
1188 fs_info
->btree_inode
= new_inode(sb
);
1189 fs_info
->btree_inode
->i_ino
= 1;
1190 fs_info
->btree_inode
->i_nlink
= 1;
1192 sb
->s_blocksize
= 4096;
1193 sb
->s_blocksize_bits
= blksize_bits(4096);
1196 * we set the i_size on the btree inode to the max possible int.
1197 * the real end of the address space is determined by all of
1198 * the devices in the system
1200 fs_info
->btree_inode
->i_size
= OFFSET_MAX
;
1201 fs_info
->btree_inode
->i_mapping
->a_ops
= &btree_aops
;
1202 fs_info
->btree_inode
->i_mapping
->backing_dev_info
= &fs_info
->bdi
;
1204 extent_io_tree_init(&BTRFS_I(fs_info
->btree_inode
)->io_tree
,
1205 fs_info
->btree_inode
->i_mapping
,
1207 extent_map_tree_init(&BTRFS_I(fs_info
->btree_inode
)->extent_tree
,
1210 BTRFS_I(fs_info
->btree_inode
)->io_tree
.ops
= &btree_extent_io_ops
;
1212 extent_io_tree_init(&fs_info
->free_space_cache
,
1213 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1214 extent_io_tree_init(&fs_info
->block_group_cache
,
1215 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1216 extent_io_tree_init(&fs_info
->pinned_extents
,
1217 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1218 extent_io_tree_init(&fs_info
->pending_del
,
1219 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1220 extent_io_tree_init(&fs_info
->extent_ins
,
1221 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1222 fs_info
->do_barriers
= 1;
1224 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1225 INIT_WORK(&fs_info
->end_io_work
, btrfs_end_io_csum
, fs_info
);
1226 INIT_WORK(&fs_info
->async_submit_work
, btrfs_async_submit_work
,
1228 INIT_WORK(&fs_info
->trans_work
, btrfs_transaction_cleaner
, fs_info
);
1230 INIT_WORK(&fs_info
->end_io_work
, btrfs_end_io_csum
);
1231 INIT_WORK(&fs_info
->async_submit_work
, btrfs_async_submit_work
);
1232 INIT_DELAYED_WORK(&fs_info
->trans_work
, btrfs_transaction_cleaner
);
1234 BTRFS_I(fs_info
->btree_inode
)->root
= tree_root
;
1235 memset(&BTRFS_I(fs_info
->btree_inode
)->location
, 0,
1236 sizeof(struct btrfs_key
));
1237 insert_inode_hash(fs_info
->btree_inode
);
1238 mapping_set_gfp_mask(fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1240 mutex_init(&fs_info
->trans_mutex
);
1241 mutex_init(&fs_info
->fs_mutex
);
1244 ret
= add_hasher(fs_info
, "crc32c");
1246 printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
1251 __setup_root(4096, 4096, 4096, 4096, tree_root
,
1252 fs_info
, BTRFS_ROOT_TREE_OBJECTID
);
1255 bh
= __bread(fs_devices
->latest_bdev
,
1256 BTRFS_SUPER_INFO_OFFSET
/ 4096, 4096);
1260 memcpy(&fs_info
->super_copy
, bh
->b_data
, sizeof(fs_info
->super_copy
));
1263 memcpy(fs_info
->fsid
, fs_info
->super_copy
.fsid
, BTRFS_FSID_SIZE
);
1265 disk_super
= &fs_info
->super_copy
;
1266 if (!btrfs_super_root(disk_super
))
1267 goto fail_sb_buffer
;
1269 err
= btrfs_parse_options(tree_root
, options
);
1271 goto fail_sb_buffer
;
1274 if (btrfs_super_num_devices(disk_super
) > fs_devices
->open_devices
) {
1275 printk("Btrfs: wanted %llu devices, but found %llu\n",
1276 (unsigned long long)btrfs_super_num_devices(disk_super
),
1277 (unsigned long long)fs_devices
->open_devices
);
1278 if (btrfs_test_opt(tree_root
, DEGRADED
))
1279 printk("continuing in degraded mode\n");
1281 goto fail_sb_buffer
;
1285 fs_info
->bdi
.ra_pages
*= btrfs_super_num_devices(disk_super
);
1287 nodesize
= btrfs_super_nodesize(disk_super
);
1288 leafsize
= btrfs_super_leafsize(disk_super
);
1289 sectorsize
= btrfs_super_sectorsize(disk_super
);
1290 stripesize
= btrfs_super_stripesize(disk_super
);
1291 tree_root
->nodesize
= nodesize
;
1292 tree_root
->leafsize
= leafsize
;
1293 tree_root
->sectorsize
= sectorsize
;
1294 tree_root
->stripesize
= stripesize
;
1296 sb
->s_blocksize
= sectorsize
;
1297 sb
->s_blocksize_bits
= blksize_bits(sectorsize
);
1299 if (strncmp((char *)(&disk_super
->magic
), BTRFS_MAGIC
,
1300 sizeof(disk_super
->magic
))) {
1301 printk("btrfs: valid FS not found on %s\n", sb
->s_id
);
1302 goto fail_sb_buffer
;
1305 mutex_lock(&fs_info
->fs_mutex
);
1307 ret
= btrfs_read_sys_array(tree_root
);
1309 printk("btrfs: failed to read the system array on %s\n",
1311 goto fail_sys_array
;
1314 blocksize
= btrfs_level_size(tree_root
,
1315 btrfs_super_chunk_root_level(disk_super
));
1317 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
1318 chunk_root
, fs_info
, BTRFS_CHUNK_TREE_OBJECTID
);
1320 chunk_root
->node
= read_tree_block(chunk_root
,
1321 btrfs_super_chunk_root(disk_super
),
1323 BUG_ON(!chunk_root
->node
);
1325 read_extent_buffer(chunk_root
->node
, fs_info
->chunk_tree_uuid
,
1326 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root
->node
),
1329 ret
= btrfs_read_chunk_tree(chunk_root
);
1332 btrfs_close_extra_devices(fs_devices
);
1334 blocksize
= btrfs_level_size(tree_root
,
1335 btrfs_super_root_level(disk_super
));
1338 tree_root
->node
= read_tree_block(tree_root
,
1339 btrfs_super_root(disk_super
),
1341 if (!tree_root
->node
)
1342 goto fail_sb_buffer
;
1345 ret
= find_and_setup_root(tree_root
, fs_info
,
1346 BTRFS_EXTENT_TREE_OBJECTID
, extent_root
);
1348 goto fail_tree_root
;
1349 extent_root
->track_dirty
= 1;
1351 ret
= find_and_setup_root(tree_root
, fs_info
,
1352 BTRFS_DEV_TREE_OBJECTID
, dev_root
);
1353 dev_root
->track_dirty
= 1;
1356 goto fail_extent_root
;
1358 btrfs_read_block_groups(extent_root
);
1360 fs_info
->generation
= btrfs_super_generation(disk_super
) + 1;
1361 fs_info
->data_alloc_profile
= (u64
)-1;
1362 fs_info
->metadata_alloc_profile
= (u64
)-1;
1363 fs_info
->system_alloc_profile
= fs_info
->metadata_alloc_profile
;
1365 mutex_unlock(&fs_info
->fs_mutex
);
1369 free_extent_buffer(extent_root
->node
);
1371 free_extent_buffer(tree_root
->node
);
1373 mutex_unlock(&fs_info
->fs_mutex
);
1375 extent_io_tree_empty_lru(&BTRFS_I(fs_info
->btree_inode
)->io_tree
);
1377 iput(fs_info
->btree_inode
);
1379 btrfs_close_devices(fs_info
->fs_devices
);
1380 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
1384 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
1385 bdi_destroy(&fs_info
->bdi
);
1388 return ERR_PTR(err
);
1391 static void btrfs_end_buffer_write_sync(struct buffer_head
*bh
, int uptodate
)
1393 char b
[BDEVNAME_SIZE
];
1396 set_buffer_uptodate(bh
);
1398 if (!buffer_eopnotsupp(bh
) && printk_ratelimit()) {
1399 printk(KERN_WARNING
"lost page write due to "
1400 "I/O error on %s\n",
1401 bdevname(bh
->b_bdev
, b
));
1403 /* note, we dont' set_buffer_write_io_error because we have
1404 * our own ways of dealing with the IO errors
1406 clear_buffer_uptodate(bh
);
1412 int write_all_supers(struct btrfs_root
*root
)
1414 struct list_head
*cur
;
1415 struct list_head
*head
= &root
->fs_info
->fs_devices
->devices
;
1416 struct btrfs_device
*dev
;
1417 struct btrfs_super_block
*sb
;
1418 struct btrfs_dev_item
*dev_item
;
1419 struct buffer_head
*bh
;
1423 int total_errors
= 0;
1427 max_errors
= btrfs_super_num_devices(&root
->fs_info
->super_copy
) - 1;
1428 do_barriers
= !btrfs_test_opt(root
, NOBARRIER
);
1430 sb
= &root
->fs_info
->super_for_commit
;
1431 dev_item
= &sb
->dev_item
;
1432 list_for_each(cur
, head
) {
1433 dev
= list_entry(cur
, struct btrfs_device
, dev_list
);
1438 if (!dev
->in_fs_metadata
)
1441 btrfs_set_stack_device_type(dev_item
, dev
->type
);
1442 btrfs_set_stack_device_id(dev_item
, dev
->devid
);
1443 btrfs_set_stack_device_total_bytes(dev_item
, dev
->total_bytes
);
1444 btrfs_set_stack_device_bytes_used(dev_item
, dev
->bytes_used
);
1445 btrfs_set_stack_device_io_align(dev_item
, dev
->io_align
);
1446 btrfs_set_stack_device_io_width(dev_item
, dev
->io_width
);
1447 btrfs_set_stack_device_sector_size(dev_item
, dev
->sector_size
);
1448 memcpy(dev_item
->uuid
, dev
->uuid
, BTRFS_UUID_SIZE
);
1449 flags
= btrfs_super_flags(sb
);
1450 btrfs_set_super_flags(sb
, flags
| BTRFS_HEADER_FLAG_WRITTEN
);
1454 crc
= btrfs_csum_data(root
, (char *)sb
+ BTRFS_CSUM_SIZE
, crc
,
1455 BTRFS_SUPER_INFO_SIZE
- BTRFS_CSUM_SIZE
);
1456 btrfs_csum_final(crc
, sb
->csum
);
1458 bh
= __getblk(dev
->bdev
, BTRFS_SUPER_INFO_OFFSET
/ 4096,
1459 BTRFS_SUPER_INFO_SIZE
);
1461 memcpy(bh
->b_data
, sb
, BTRFS_SUPER_INFO_SIZE
);
1462 dev
->pending_io
= bh
;
1465 set_buffer_uptodate(bh
);
1467 bh
->b_end_io
= btrfs_end_buffer_write_sync
;
1469 if (do_barriers
&& dev
->barriers
) {
1470 ret
= submit_bh(WRITE_BARRIER
, bh
);
1471 if (ret
== -EOPNOTSUPP
) {
1472 printk("btrfs: disabling barriers on dev %s\n",
1474 set_buffer_uptodate(bh
);
1478 ret
= submit_bh(WRITE
, bh
);
1481 ret
= submit_bh(WRITE
, bh
);
1486 if (total_errors
> max_errors
) {
1487 printk("btrfs: %d errors while writing supers\n", total_errors
);
1492 list_for_each(cur
, head
) {
1493 dev
= list_entry(cur
, struct btrfs_device
, dev_list
);
1496 if (!dev
->in_fs_metadata
)
1499 BUG_ON(!dev
->pending_io
);
1500 bh
= dev
->pending_io
;
1502 if (!buffer_uptodate(dev
->pending_io
)) {
1503 if (do_barriers
&& dev
->barriers
) {
1504 printk("btrfs: disabling barriers on dev %s\n",
1506 set_buffer_uptodate(bh
);
1510 ret
= submit_bh(WRITE
, bh
);
1513 if (!buffer_uptodate(bh
))
1520 dev
->pending_io
= NULL
;
1523 if (total_errors
> max_errors
) {
1524 printk("btrfs: %d errors while writing supers\n", total_errors
);
1530 int write_ctree_super(struct btrfs_trans_handle
*trans
, struct btrfs_root
1535 ret
= write_all_supers(root
);
1539 int btrfs_free_fs_root(struct btrfs_fs_info
*fs_info
, struct btrfs_root
*root
)
1541 radix_tree_delete(&fs_info
->fs_roots_radix
,
1542 (unsigned long)root
->root_key
.objectid
);
1544 btrfs_sysfs_del_root(root
);
1548 free_extent_buffer(root
->node
);
1549 if (root
->commit_root
)
1550 free_extent_buffer(root
->commit_root
);
1557 static int del_fs_roots(struct btrfs_fs_info
*fs_info
)
1560 struct btrfs_root
*gang
[8];
1564 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
1569 for (i
= 0; i
< ret
; i
++)
1570 btrfs_free_fs_root(fs_info
, gang
[i
]);
1575 int close_ctree(struct btrfs_root
*root
)
1578 struct btrfs_trans_handle
*trans
;
1579 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1581 fs_info
->closing
= 1;
1582 btrfs_transaction_flush_work(root
);
1583 mutex_lock(&fs_info
->fs_mutex
);
1584 btrfs_defrag_dirty_roots(root
->fs_info
);
1585 trans
= btrfs_start_transaction(root
, 1);
1586 ret
= btrfs_commit_transaction(trans
, root
);
1587 /* run commit again to drop the original snapshot */
1588 trans
= btrfs_start_transaction(root
, 1);
1589 btrfs_commit_transaction(trans
, root
);
1590 ret
= btrfs_write_and_wait_transaction(NULL
, root
);
1593 write_ctree_super(NULL
, root
);
1594 mutex_unlock(&fs_info
->fs_mutex
);
1596 btrfs_transaction_flush_work(root
);
1598 if (fs_info
->delalloc_bytes
) {
1599 printk("btrfs: at unmount delalloc count %Lu\n",
1600 fs_info
->delalloc_bytes
);
1602 if (fs_info
->extent_root
->node
)
1603 free_extent_buffer(fs_info
->extent_root
->node
);
1605 if (fs_info
->tree_root
->node
)
1606 free_extent_buffer(fs_info
->tree_root
->node
);
1608 if (root
->fs_info
->chunk_root
->node
);
1609 free_extent_buffer(root
->fs_info
->chunk_root
->node
);
1611 if (root
->fs_info
->dev_root
->node
);
1612 free_extent_buffer(root
->fs_info
->dev_root
->node
);
1614 btrfs_free_block_groups(root
->fs_info
);
1615 del_fs_roots(fs_info
);
1617 filemap_write_and_wait(fs_info
->btree_inode
->i_mapping
);
1619 extent_io_tree_empty_lru(&fs_info
->free_space_cache
);
1620 extent_io_tree_empty_lru(&fs_info
->block_group_cache
);
1621 extent_io_tree_empty_lru(&fs_info
->pinned_extents
);
1622 extent_io_tree_empty_lru(&fs_info
->pending_del
);
1623 extent_io_tree_empty_lru(&fs_info
->extent_ins
);
1624 extent_io_tree_empty_lru(&BTRFS_I(fs_info
->btree_inode
)->io_tree
);
1626 flush_workqueue(async_submit_workqueue
);
1627 flush_workqueue(end_io_workqueue
);
1629 truncate_inode_pages(fs_info
->btree_inode
->i_mapping
, 0);
1631 flush_workqueue(async_submit_workqueue
);
1632 destroy_workqueue(async_submit_workqueue
);
1634 flush_workqueue(end_io_workqueue
);
1635 destroy_workqueue(end_io_workqueue
);
1637 iput(fs_info
->btree_inode
);
1639 while(!list_empty(&fs_info
->hashers
)) {
1640 struct btrfs_hasher
*hasher
;
1641 hasher
= list_entry(fs_info
->hashers
.next
, struct btrfs_hasher
,
1643 list_del(&hasher
->hashers
);
1644 crypto_free_hash(&fs_info
->hash_tfm
);
1648 btrfs_close_devices(fs_info
->fs_devices
);
1649 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
1651 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
1652 bdi_destroy(&fs_info
->bdi
);
1655 kfree(fs_info
->extent_root
);
1656 kfree(fs_info
->tree_root
);
1657 kfree(fs_info
->chunk_root
);
1658 kfree(fs_info
->dev_root
);
1662 int btrfs_buffer_uptodate(struct extent_buffer
*buf
, u64 parent_transid
)
1665 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
1667 ret
= extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
, buf
);
1671 ret
= verify_parent_transid(&BTRFS_I(btree_inode
)->io_tree
, buf
,
1676 int btrfs_set_buffer_uptodate(struct extent_buffer
*buf
)
1678 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
1679 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
,
1683 void btrfs_mark_buffer_dirty(struct extent_buffer
*buf
)
1685 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1686 u64 transid
= btrfs_header_generation(buf
);
1687 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1689 if (transid
!= root
->fs_info
->generation
) {
1690 printk(KERN_CRIT
"transid mismatch buffer %llu, found %Lu running %Lu\n",
1691 (unsigned long long)buf
->start
,
1692 transid
, root
->fs_info
->generation
);
1695 set_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
, buf
);
1698 void btrfs_throttle(struct btrfs_root
*root
)
1700 struct backing_dev_info
*bdi
;
1702 bdi
= &root
->fs_info
->bdi
;
1703 if (root
->fs_info
->throttles
&& bdi_write_congested(bdi
)) {
1704 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18)
1705 congestion_wait(WRITE
, HZ
/20);
1707 blk_congestion_wait(WRITE
, HZ
/20);
1712 void btrfs_btree_balance_dirty(struct btrfs_root
*root
, unsigned long nr
)
1715 * looks as though older kernels can get into trouble with
1716 * this code, they end up stuck in balance_dirty_pages forever
1718 struct extent_io_tree
*tree
;
1721 unsigned long thresh
= 16 * 1024 * 1024;
1722 tree
= &BTRFS_I(root
->fs_info
->btree_inode
)->io_tree
;
1724 if (current_is_pdflush())
1727 num_dirty
= count_range_bits(tree
, &start
, (u64
)-1,
1728 thresh
, EXTENT_DIRTY
);
1729 if (num_dirty
> thresh
) {
1730 balance_dirty_pages_ratelimited_nr(
1731 root
->fs_info
->btree_inode
->i_mapping
, 1);
1736 void btrfs_set_buffer_defrag(struct extent_buffer
*buf
)
1738 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1739 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1740 set_extent_bits(&BTRFS_I(btree_inode
)->io_tree
, buf
->start
,
1741 buf
->start
+ buf
->len
- 1, EXTENT_DEFRAG
, GFP_NOFS
);
1744 void btrfs_set_buffer_defrag_done(struct extent_buffer
*buf
)
1746 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1747 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1748 set_extent_bits(&BTRFS_I(btree_inode
)->io_tree
, buf
->start
,
1749 buf
->start
+ buf
->len
- 1, EXTENT_DEFRAG_DONE
,
1753 int btrfs_buffer_defrag(struct extent_buffer
*buf
)
1755 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1756 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1757 return test_range_bit(&BTRFS_I(btree_inode
)->io_tree
,
1758 buf
->start
, buf
->start
+ buf
->len
- 1, EXTENT_DEFRAG
, 0);
1761 int btrfs_buffer_defrag_done(struct extent_buffer
*buf
)
1763 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1764 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1765 return test_range_bit(&BTRFS_I(btree_inode
)->io_tree
,
1766 buf
->start
, buf
->start
+ buf
->len
- 1,
1767 EXTENT_DEFRAG_DONE
, 0);
1770 int btrfs_clear_buffer_defrag_done(struct extent_buffer
*buf
)
1772 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1773 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1774 return clear_extent_bits(&BTRFS_I(btree_inode
)->io_tree
,
1775 buf
->start
, buf
->start
+ buf
->len
- 1,
1776 EXTENT_DEFRAG_DONE
, GFP_NOFS
);
1779 int btrfs_clear_buffer_defrag(struct extent_buffer
*buf
)
1781 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1782 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1783 return clear_extent_bits(&BTRFS_I(btree_inode
)->io_tree
,
1784 buf
->start
, buf
->start
+ buf
->len
- 1,
1785 EXTENT_DEFRAG
, GFP_NOFS
);
1788 int btrfs_read_buffer(struct extent_buffer
*buf
, u64 parent_transid
)
1790 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1792 ret
= btree_read_extent_buffer_pages(root
, buf
, 0, parent_transid
);
1794 buf
->flags
|= EXTENT_UPTODATE
;
1799 static struct extent_io_ops btree_extent_io_ops
= {
1800 .writepage_io_hook
= btree_writepage_io_hook
,
1801 .readpage_end_io_hook
= btree_readpage_end_io_hook
,
1802 .submit_bio_hook
= btree_submit_bio_hook
,
1803 /* note we're sharing with inode.c for the merge bio hook */
1804 .merge_bio_hook
= btrfs_merge_bio_hook
,