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/crc32c.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>
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
;
55 struct btrfs_fs_info
*info
;
58 struct list_head list
;
61 struct extent_map
*btree_get_extent(struct inode
*inode
, struct page
*page
,
62 size_t page_offset
, u64 start
, u64 len
,
65 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
66 struct extent_map
*em
;
70 spin_lock(&em_tree
->lock
);
71 em
= lookup_extent_mapping(em_tree
, start
, len
);
72 spin_unlock(&em_tree
->lock
);
76 em
= alloc_extent_map(GFP_NOFS
);
78 em
= ERR_PTR(-ENOMEM
);
82 em
->len
= i_size_read(inode
);
84 em
->bdev
= inode
->i_sb
->s_bdev
;
86 spin_lock(&em_tree
->lock
);
87 ret
= add_extent_mapping(em_tree
, em
);
88 spin_unlock(&em_tree
->lock
);
101 u32
btrfs_csum_data(struct btrfs_root
*root
, char *data
, u32 seed
, size_t len
)
103 return crc32c(seed
, data
, len
);
106 void btrfs_csum_final(u32 crc
, char *result
)
108 *(__le32
*)result
= ~cpu_to_le32(crc
);
111 static int csum_tree_block(struct btrfs_root
*root
, struct extent_buffer
*buf
,
114 char result
[BTRFS_CRC32_SIZE
];
116 unsigned long cur_len
;
117 unsigned long offset
= BTRFS_CSUM_SIZE
;
118 char *map_token
= NULL
;
120 unsigned long map_start
;
121 unsigned long map_len
;
125 len
= buf
->len
- offset
;
127 err
= map_private_extent_buffer(buf
, offset
, 32,
129 &map_start
, &map_len
, KM_USER0
);
131 printk("failed to map extent buffer! %lu\n",
135 cur_len
= min(len
, map_len
- (offset
- map_start
));
136 crc
= btrfs_csum_data(root
, kaddr
+ offset
- map_start
,
140 unmap_extent_buffer(buf
, map_token
, KM_USER0
);
142 btrfs_csum_final(crc
, result
);
145 int from_this_trans
= 0;
147 if (root
->fs_info
->running_transaction
&&
148 btrfs_header_generation(buf
) ==
149 root
->fs_info
->running_transaction
->transid
)
152 /* FIXME, this is not good */
153 if (memcmp_extent_buffer(buf
, result
, 0, BTRFS_CRC32_SIZE
)) {
156 memcpy(&found
, result
, BTRFS_CRC32_SIZE
);
158 read_extent_buffer(buf
, &val
, 0, BTRFS_CRC32_SIZE
);
160 printk("btrfs: %s checksum verify failed on %llu "
161 "wanted %X found %X from_this_trans %d "
163 root
->fs_info
->sb
->s_id
,
164 buf
->start
, val
, found
, from_this_trans
,
165 btrfs_header_level(buf
));
169 write_extent_buffer(buf
, result
, 0, BTRFS_CRC32_SIZE
);
175 int csum_dirty_buffer(struct btrfs_root
*root
, struct page
*page
)
177 struct extent_io_tree
*tree
;
178 u64 start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
182 struct extent_buffer
*eb
;
183 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
185 if (page
->private == EXTENT_PAGE_PRIVATE
)
189 len
= page
->private >> 2;
193 eb
= alloc_extent_buffer(tree
, start
, len
, page
, GFP_NOFS
);
194 read_extent_buffer_pages(tree
, eb
, start
+ PAGE_CACHE_SIZE
, 1,
196 btrfs_clear_buffer_defrag(eb
);
197 found_start
= btrfs_header_bytenr(eb
);
198 if (found_start
!= start
) {
199 printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
200 start
, found_start
, len
);
204 if (eb
->first_page
!= page
) {
205 printk("bad first page %lu %lu\n", eb
->first_page
->index
,
210 if (!PageUptodate(page
)) {
211 printk("csum not up to date page %lu\n", page
->index
);
215 found_level
= btrfs_header_level(eb
);
216 spin_lock(&root
->fs_info
->hash_lock
);
217 btrfs_set_header_flag(eb
, BTRFS_HEADER_FLAG_WRITTEN
);
218 spin_unlock(&root
->fs_info
->hash_lock
);
219 csum_tree_block(root
, eb
, 0);
221 free_extent_buffer(eb
);
226 static int btree_writepage_io_hook(struct page
*page
, u64 start
, u64 end
)
228 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
230 csum_dirty_buffer(root
, page
);
234 int btree_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
235 struct extent_state
*state
)
237 struct extent_io_tree
*tree
;
241 struct extent_buffer
*eb
;
242 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
245 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
246 if (page
->private == EXTENT_PAGE_PRIVATE
)
250 len
= page
->private >> 2;
254 eb
= alloc_extent_buffer(tree
, start
, len
, page
, GFP_NOFS
);
255 read_extent_buffer_pages(tree
, eb
, start
+ PAGE_CACHE_SIZE
, 1,
257 btrfs_clear_buffer_defrag(eb
);
258 found_start
= btrfs_header_bytenr(eb
);
259 if (found_start
!= start
) {
260 printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
261 start
, found_start
, len
);
265 if (eb
->first_page
!= page
) {
266 printk("bad first page %lu %lu\n", eb
->first_page
->index
,
271 found_level
= btrfs_header_level(eb
);
273 ret
= csum_tree_block(root
, eb
, 1);
275 end
= min_t(u64
, eb
->len
, PAGE_CACHE_SIZE
);
276 end
= eb
->start
+ end
- 1;
277 release_extent_buffer_tail_pages(eb
);
279 free_extent_buffer(eb
);
284 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
285 static void end_workqueue_bio(struct bio
*bio
, int err
)
287 static int end_workqueue_bio(struct bio
*bio
,
288 unsigned int bytes_done
, int err
)
291 struct end_io_wq
*end_io_wq
= bio
->bi_private
;
292 struct btrfs_fs_info
*fs_info
;
295 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
300 fs_info
= end_io_wq
->info
;
301 spin_lock_irqsave(&fs_info
->end_io_work_lock
, flags
);
302 end_io_wq
->error
= err
;
303 list_add_tail(&end_io_wq
->list
, &fs_info
->end_io_work_list
);
304 spin_unlock_irqrestore(&fs_info
->end_io_work_lock
, flags
);
305 queue_work(end_io_workqueue
, &fs_info
->end_io_work
);
307 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
312 int btrfs_bio_wq_end_io(struct btrfs_fs_info
*info
, struct bio
*bio
,
315 struct end_io_wq
*end_io_wq
;
316 end_io_wq
= kmalloc(sizeof(*end_io_wq
), GFP_NOFS
);
320 end_io_wq
->private = bio
->bi_private
;
321 end_io_wq
->end_io
= bio
->bi_end_io
;
322 end_io_wq
->info
= info
;
323 end_io_wq
->error
= 0;
324 end_io_wq
->bio
= bio
;
325 end_io_wq
->metadata
= metadata
;
327 bio
->bi_private
= end_io_wq
;
328 bio
->bi_end_io
= end_workqueue_bio
;
332 static int btree_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
)
334 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
338 offset
= bio
->bi_sector
<< 9;
340 if (rw
& (1 << BIO_RW
)) {
341 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
);
344 ret
= btrfs_bio_wq_end_io(root
->fs_info
, bio
, 1);
347 if (offset
== BTRFS_SUPER_INFO_OFFSET
) {
348 bio
->bi_bdev
= root
->fs_info
->sb
->s_bdev
;
352 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
);
355 static int btree_writepage(struct page
*page
, struct writeback_control
*wbc
)
357 struct extent_io_tree
*tree
;
358 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
359 return extent_write_full_page(tree
, page
, btree_get_extent
, wbc
);
362 static int btree_writepages(struct address_space
*mapping
,
363 struct writeback_control
*wbc
)
365 struct extent_io_tree
*tree
;
366 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
367 if (wbc
->sync_mode
== WB_SYNC_NONE
) {
370 unsigned long thresh
= 96 * 1024 * 1024;
372 if (wbc
->for_kupdate
)
375 if (current_is_pdflush()) {
376 thresh
= 96 * 1024 * 1024;
378 thresh
= 8 * 1024 * 1024;
380 num_dirty
= count_range_bits(tree
, &start
, (u64
)-1,
381 thresh
, EXTENT_DIRTY
);
382 if (num_dirty
< thresh
) {
386 return extent_writepages(tree
, mapping
, btree_get_extent
, wbc
);
389 int btree_readpage(struct file
*file
, struct page
*page
)
391 struct extent_io_tree
*tree
;
392 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
393 return extent_read_full_page(tree
, page
, btree_get_extent
);
396 static int btree_releasepage(struct page
*page
, gfp_t gfp_flags
)
398 struct extent_io_tree
*tree
;
399 struct extent_map_tree
*map
;
402 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
403 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
404 ret
= try_release_extent_mapping(map
, tree
, page
, gfp_flags
);
406 invalidate_extent_lru(tree
, page_offset(page
), PAGE_CACHE_SIZE
);
407 ClearPagePrivate(page
);
408 set_page_private(page
, 0);
409 page_cache_release(page
);
414 static void btree_invalidatepage(struct page
*page
, unsigned long offset
)
416 struct extent_io_tree
*tree
;
417 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
418 extent_invalidatepage(tree
, page
, offset
);
419 btree_releasepage(page
, GFP_NOFS
);
423 static int btree_writepage(struct page
*page
, struct writeback_control
*wbc
)
425 struct buffer_head
*bh
;
426 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
427 struct buffer_head
*head
;
428 if (!page_has_buffers(page
)) {
429 create_empty_buffers(page
, root
->fs_info
->sb
->s_blocksize
,
430 (1 << BH_Dirty
)|(1 << BH_Uptodate
));
432 head
= page_buffers(page
);
435 if (buffer_dirty(bh
))
436 csum_tree_block(root
, bh
, 0);
437 bh
= bh
->b_this_page
;
438 } while (bh
!= head
);
439 return block_write_full_page(page
, btree_get_block
, wbc
);
443 static struct address_space_operations btree_aops
= {
444 .readpage
= btree_readpage
,
445 .writepage
= btree_writepage
,
446 .writepages
= btree_writepages
,
447 .releasepage
= btree_releasepage
,
448 .invalidatepage
= btree_invalidatepage
,
449 .sync_page
= block_sync_page
,
452 int readahead_tree_block(struct btrfs_root
*root
, u64 bytenr
, u32 blocksize
)
454 struct extent_buffer
*buf
= NULL
;
455 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
458 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
461 read_extent_buffer_pages(&BTRFS_I(btree_inode
)->io_tree
,
462 buf
, 0, 0, btree_get_extent
);
463 free_extent_buffer(buf
);
467 static int close_all_devices(struct btrfs_fs_info
*fs_info
)
469 struct list_head
*list
;
470 struct list_head
*next
;
471 struct btrfs_device
*device
;
473 list
= &fs_info
->fs_devices
->devices
;
474 list_for_each(next
, list
) {
475 device
= list_entry(next
, struct btrfs_device
, dev_list
);
476 if (device
->bdev
&& device
->bdev
!= fs_info
->sb
->s_bdev
)
477 close_bdev_excl(device
->bdev
);
483 int btrfs_verify_block_csum(struct btrfs_root
*root
,
484 struct extent_buffer
*buf
)
486 return btrfs_buffer_uptodate(buf
);
489 struct extent_buffer
*btrfs_find_tree_block(struct btrfs_root
*root
,
490 u64 bytenr
, u32 blocksize
)
492 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
493 struct extent_buffer
*eb
;
494 eb
= find_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
495 bytenr
, blocksize
, GFP_NOFS
);
499 struct extent_buffer
*btrfs_find_create_tree_block(struct btrfs_root
*root
,
500 u64 bytenr
, u32 blocksize
)
502 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
503 struct extent_buffer
*eb
;
505 eb
= alloc_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
506 bytenr
, blocksize
, NULL
, GFP_NOFS
);
511 struct extent_buffer
*read_tree_block(struct btrfs_root
*root
, u64 bytenr
,
514 struct extent_buffer
*buf
= NULL
;
515 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
516 struct extent_io_tree
*io_tree
;
519 io_tree
= &BTRFS_I(btree_inode
)->io_tree
;
521 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
525 ret
= read_extent_buffer_pages(&BTRFS_I(btree_inode
)->io_tree
, buf
, 0,
526 1, btree_get_extent
);
529 buf
->flags
|= EXTENT_UPTODATE
;
535 int clean_tree_block(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
536 struct extent_buffer
*buf
)
538 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
539 if (btrfs_header_generation(buf
) ==
540 root
->fs_info
->running_transaction
->transid
)
541 clear_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
,
546 int wait_on_tree_block_writeback(struct btrfs_root
*root
,
547 struct extent_buffer
*buf
)
549 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
550 wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode
)->io_tree
,
555 static int __setup_root(u32 nodesize
, u32 leafsize
, u32 sectorsize
,
556 u32 stripesize
, struct btrfs_root
*root
,
557 struct btrfs_fs_info
*fs_info
,
562 root
->commit_root
= NULL
;
563 root
->sectorsize
= sectorsize
;
564 root
->nodesize
= nodesize
;
565 root
->leafsize
= leafsize
;
566 root
->stripesize
= stripesize
;
568 root
->track_dirty
= 0;
570 root
->fs_info
= fs_info
;
571 root
->objectid
= objectid
;
572 root
->last_trans
= 0;
573 root
->highest_inode
= 0;
574 root
->last_inode_alloc
= 0;
578 INIT_LIST_HEAD(&root
->dirty_list
);
579 memset(&root
->root_key
, 0, sizeof(root
->root_key
));
580 memset(&root
->root_item
, 0, sizeof(root
->root_item
));
581 memset(&root
->defrag_progress
, 0, sizeof(root
->defrag_progress
));
582 memset(&root
->root_kobj
, 0, sizeof(root
->root_kobj
));
583 init_completion(&root
->kobj_unregister
);
584 root
->defrag_running
= 0;
585 root
->defrag_level
= 0;
586 root
->root_key
.objectid
= objectid
;
590 static int find_and_setup_root(struct btrfs_root
*tree_root
,
591 struct btrfs_fs_info
*fs_info
,
593 struct btrfs_root
*root
)
598 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
599 tree_root
->sectorsize
, tree_root
->stripesize
,
600 root
, fs_info
, objectid
);
601 ret
= btrfs_find_last_root(tree_root
, objectid
,
602 &root
->root_item
, &root
->root_key
);
605 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
606 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
612 struct btrfs_root
*btrfs_read_fs_root_no_radix(struct btrfs_fs_info
*fs_info
,
613 struct btrfs_key
*location
)
615 struct btrfs_root
*root
;
616 struct btrfs_root
*tree_root
= fs_info
->tree_root
;
617 struct btrfs_path
*path
;
618 struct extent_buffer
*l
;
623 root
= kzalloc(sizeof(*root
), GFP_NOFS
);
625 return ERR_PTR(-ENOMEM
);
626 if (location
->offset
== (u64
)-1) {
627 ret
= find_and_setup_root(tree_root
, fs_info
,
628 location
->objectid
, root
);
636 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
637 tree_root
->sectorsize
, tree_root
->stripesize
,
638 root
, fs_info
, location
->objectid
);
640 path
= btrfs_alloc_path();
642 ret
= btrfs_search_slot(NULL
, tree_root
, location
, path
, 0, 0);
649 read_extent_buffer(l
, &root
->root_item
,
650 btrfs_item_ptr_offset(l
, path
->slots
[0]),
651 sizeof(root
->root_item
));
652 memcpy(&root
->root_key
, location
, sizeof(*location
));
655 btrfs_release_path(root
, path
);
656 btrfs_free_path(path
);
661 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
662 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
667 ret
= btrfs_find_highest_inode(root
, &highest_inode
);
669 root
->highest_inode
= highest_inode
;
670 root
->last_inode_alloc
= highest_inode
;
675 struct btrfs_root
*btrfs_lookup_fs_root(struct btrfs_fs_info
*fs_info
,
678 struct btrfs_root
*root
;
680 if (root_objectid
== BTRFS_ROOT_TREE_OBJECTID
)
681 return fs_info
->tree_root
;
682 if (root_objectid
== BTRFS_EXTENT_TREE_OBJECTID
)
683 return fs_info
->extent_root
;
685 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
686 (unsigned long)root_objectid
);
690 struct btrfs_root
*btrfs_read_fs_root_no_name(struct btrfs_fs_info
*fs_info
,
691 struct btrfs_key
*location
)
693 struct btrfs_root
*root
;
696 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
697 return fs_info
->tree_root
;
698 if (location
->objectid
== BTRFS_EXTENT_TREE_OBJECTID
)
699 return fs_info
->extent_root
;
701 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
702 (unsigned long)location
->objectid
);
706 root
= btrfs_read_fs_root_no_radix(fs_info
, location
);
709 ret
= radix_tree_insert(&fs_info
->fs_roots_radix
,
710 (unsigned long)root
->root_key
.objectid
,
713 free_extent_buffer(root
->node
);
717 ret
= btrfs_find_dead_roots(fs_info
->tree_root
,
718 root
->root_key
.objectid
, root
);
724 struct btrfs_root
*btrfs_read_fs_root(struct btrfs_fs_info
*fs_info
,
725 struct btrfs_key
*location
,
726 const char *name
, int namelen
)
728 struct btrfs_root
*root
;
731 root
= btrfs_read_fs_root_no_name(fs_info
, location
);
738 ret
= btrfs_set_root_name(root
, name
, namelen
);
740 free_extent_buffer(root
->node
);
745 ret
= btrfs_sysfs_add_root(root
);
747 free_extent_buffer(root
->node
);
756 static int add_hasher(struct btrfs_fs_info
*info
, char *type
) {
757 struct btrfs_hasher
*hasher
;
759 hasher
= kmalloc(sizeof(*hasher
), GFP_NOFS
);
762 hasher
->hash_tfm
= crypto_alloc_hash(type
, 0, CRYPTO_ALG_ASYNC
);
763 if (!hasher
->hash_tfm
) {
767 spin_lock(&info
->hash_lock
);
768 list_add(&hasher
->list
, &info
->hashers
);
769 spin_unlock(&info
->hash_lock
);
774 static int btrfs_congested_fn(void *congested_data
, int bdi_bits
)
776 struct btrfs_fs_info
*info
= (struct btrfs_fs_info
*)congested_data
;
778 struct list_head
*cur
;
779 struct btrfs_device
*device
;
780 struct backing_dev_info
*bdi
;
782 list_for_each(cur
, &info
->fs_devices
->devices
) {
783 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
784 bdi
= blk_get_backing_dev_info(device
->bdev
);
785 if (bdi
&& bdi_congested(bdi
, bdi_bits
)) {
793 void btrfs_unplug_io_fn(struct backing_dev_info
*bdi
, struct page
*page
)
795 struct list_head
*cur
;
796 struct btrfs_device
*device
;
797 struct btrfs_fs_info
*info
;
799 info
= (struct btrfs_fs_info
*)bdi
->unplug_io_data
;
800 list_for_each(cur
, &info
->fs_devices
->devices
) {
801 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
802 bdi
= blk_get_backing_dev_info(device
->bdev
);
803 if (bdi
->unplug_io_fn
) {
804 bdi
->unplug_io_fn(bdi
, page
);
809 static int setup_bdi(struct btrfs_fs_info
*info
, struct backing_dev_info
*bdi
)
812 bdi
->ra_pages
= default_backing_dev_info
.ra_pages
* 4;
814 bdi
->capabilities
= default_backing_dev_info
.capabilities
;
815 bdi
->unplug_io_fn
= btrfs_unplug_io_fn
;
816 bdi
->unplug_io_data
= info
;
817 bdi
->congested_fn
= btrfs_congested_fn
;
818 bdi
->congested_data
= info
;
822 static int bio_ready_for_csum(struct bio
*bio
)
828 struct extent_io_tree
*io_tree
= NULL
;
829 struct btrfs_fs_info
*info
= NULL
;
830 struct bio_vec
*bvec
;
834 bio_for_each_segment(bvec
, bio
, i
) {
835 page
= bvec
->bv_page
;
836 if (page
->private == EXTENT_PAGE_PRIVATE
) {
837 length
+= bvec
->bv_len
;
840 if (!page
->private) {
841 length
+= bvec
->bv_len
;
844 length
= bvec
->bv_len
;
845 buf_len
= page
->private >> 2;
846 start
= page_offset(page
) + bvec
->bv_offset
;
847 io_tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
848 info
= BTRFS_I(page
->mapping
->host
)->root
->fs_info
;
850 /* are we fully contained in this bio? */
851 if (buf_len
<= length
)
854 ret
= extent_range_uptodate(io_tree
, start
+ length
,
855 start
+ buf_len
- 1);
861 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
862 void btrfs_end_io_csum(void *p
)
864 void btrfs_end_io_csum(struct work_struct
*work
)
867 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
868 struct btrfs_fs_info
*fs_info
= p
;
870 struct btrfs_fs_info
*fs_info
= container_of(work
,
871 struct btrfs_fs_info
,
875 struct end_io_wq
*end_io_wq
;
877 struct list_head
*next
;
882 spin_lock_irqsave(&fs_info
->end_io_work_lock
, flags
);
883 if (list_empty(&fs_info
->end_io_work_list
)) {
884 spin_unlock_irqrestore(&fs_info
->end_io_work_lock
,
888 next
= fs_info
->end_io_work_list
.next
;
890 spin_unlock_irqrestore(&fs_info
->end_io_work_lock
, flags
);
892 end_io_wq
= list_entry(next
, struct end_io_wq
, list
);
894 bio
= end_io_wq
->bio
;
895 if (end_io_wq
->metadata
&& !bio_ready_for_csum(bio
)) {
896 spin_lock_irqsave(&fs_info
->end_io_work_lock
, flags
);
897 was_empty
= list_empty(&fs_info
->end_io_work_list
);
898 list_add_tail(&end_io_wq
->list
,
899 &fs_info
->end_io_work_list
);
900 spin_unlock_irqrestore(&fs_info
->end_io_work_lock
,
906 error
= end_io_wq
->error
;
907 bio
->bi_private
= end_io_wq
->private;
908 bio
->bi_end_io
= end_io_wq
->end_io
;
910 bio_endio(bio
, error
);
915 struct btrfs_root
*open_ctree(struct super_block
*sb
,
916 struct btrfs_fs_devices
*fs_devices
)
923 struct btrfs_root
*extent_root
= kmalloc(sizeof(struct btrfs_root
),
925 struct btrfs_root
*tree_root
= kmalloc(sizeof(struct btrfs_root
),
927 struct btrfs_fs_info
*fs_info
= kzalloc(sizeof(*fs_info
),
929 struct btrfs_root
*chunk_root
= kmalloc(sizeof(struct btrfs_root
),
931 struct btrfs_root
*dev_root
= kmalloc(sizeof(struct btrfs_root
),
935 struct btrfs_super_block
*disk_super
;
937 if (!extent_root
|| !tree_root
|| !fs_info
) {
941 end_io_workqueue
= create_workqueue("btrfs-end-io");
942 BUG_ON(!end_io_workqueue
);
944 INIT_RADIX_TREE(&fs_info
->fs_roots_radix
, GFP_NOFS
);
945 INIT_LIST_HEAD(&fs_info
->trans_list
);
946 INIT_LIST_HEAD(&fs_info
->dead_roots
);
947 INIT_LIST_HEAD(&fs_info
->hashers
);
948 INIT_LIST_HEAD(&fs_info
->end_io_work_list
);
949 spin_lock_init(&fs_info
->hash_lock
);
950 spin_lock_init(&fs_info
->end_io_work_lock
);
951 spin_lock_init(&fs_info
->delalloc_lock
);
952 spin_lock_init(&fs_info
->new_trans_lock
);
954 init_completion(&fs_info
->kobj_unregister
);
955 sb_set_blocksize(sb
, 4096);
956 fs_info
->tree_root
= tree_root
;
957 fs_info
->extent_root
= extent_root
;
958 fs_info
->chunk_root
= chunk_root
;
959 fs_info
->dev_root
= dev_root
;
960 fs_info
->fs_devices
= fs_devices
;
961 INIT_LIST_HEAD(&fs_info
->dirty_cowonly_roots
);
962 INIT_LIST_HEAD(&fs_info
->space_info
);
963 btrfs_mapping_init(&fs_info
->mapping_tree
);
965 fs_info
->max_extent
= (u64
)-1;
966 fs_info
->max_inline
= 8192 * 1024;
967 setup_bdi(fs_info
, &fs_info
->bdi
);
968 fs_info
->btree_inode
= new_inode(sb
);
969 fs_info
->btree_inode
->i_ino
= 1;
970 fs_info
->btree_inode
->i_nlink
= 1;
971 fs_info
->btree_inode
->i_size
= sb
->s_bdev
->bd_inode
->i_size
;
972 fs_info
->btree_inode
->i_mapping
->a_ops
= &btree_aops
;
973 fs_info
->btree_inode
->i_mapping
->backing_dev_info
= &fs_info
->bdi
;
975 extent_io_tree_init(&BTRFS_I(fs_info
->btree_inode
)->io_tree
,
976 fs_info
->btree_inode
->i_mapping
,
978 extent_map_tree_init(&BTRFS_I(fs_info
->btree_inode
)->extent_tree
,
981 BTRFS_I(fs_info
->btree_inode
)->io_tree
.ops
= &btree_extent_io_ops
;
983 extent_io_tree_init(&fs_info
->free_space_cache
,
984 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
985 extent_io_tree_init(&fs_info
->block_group_cache
,
986 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
987 extent_io_tree_init(&fs_info
->pinned_extents
,
988 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
989 extent_io_tree_init(&fs_info
->pending_del
,
990 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
991 extent_io_tree_init(&fs_info
->extent_ins
,
992 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
993 fs_info
->do_barriers
= 1;
995 INIT_WORK(&fs_info
->end_io_work
, btrfs_end_io_csum
);
996 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
997 INIT_WORK(&fs_info
->trans_work
, btrfs_transaction_cleaner
, fs_info
);
999 INIT_DELAYED_WORK(&fs_info
->trans_work
, btrfs_transaction_cleaner
);
1001 BTRFS_I(fs_info
->btree_inode
)->root
= tree_root
;
1002 memset(&BTRFS_I(fs_info
->btree_inode
)->location
, 0,
1003 sizeof(struct btrfs_key
));
1004 insert_inode_hash(fs_info
->btree_inode
);
1005 mapping_set_gfp_mask(fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
1007 mutex_init(&fs_info
->trans_mutex
);
1008 mutex_init(&fs_info
->fs_mutex
);
1011 ret
= add_hasher(fs_info
, "crc32c");
1013 printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
1018 __setup_root(4096, 4096, 4096, 4096, tree_root
,
1019 fs_info
, BTRFS_ROOT_TREE_OBJECTID
);
1021 fs_info
->sb_buffer
= read_tree_block(tree_root
,
1022 BTRFS_SUPER_INFO_OFFSET
,
1025 if (!fs_info
->sb_buffer
)
1028 read_extent_buffer(fs_info
->sb_buffer
, &fs_info
->super_copy
, 0,
1029 sizeof(fs_info
->super_copy
));
1031 read_extent_buffer(fs_info
->sb_buffer
, fs_info
->fsid
,
1032 (unsigned long)btrfs_super_fsid(fs_info
->sb_buffer
),
1035 disk_super
= &fs_info
->super_copy
;
1036 if (!btrfs_super_root(disk_super
))
1037 goto fail_sb_buffer
;
1039 if (btrfs_super_num_devices(disk_super
) != fs_devices
->num_devices
) {
1040 printk("Btrfs: wanted %llu devices, but found %llu\n",
1041 (unsigned long long)btrfs_super_num_devices(disk_super
),
1042 (unsigned long long)fs_devices
->num_devices
);
1043 goto fail_sb_buffer
;
1045 nodesize
= btrfs_super_nodesize(disk_super
);
1046 leafsize
= btrfs_super_leafsize(disk_super
);
1047 sectorsize
= btrfs_super_sectorsize(disk_super
);
1048 stripesize
= btrfs_super_stripesize(disk_super
);
1049 tree_root
->nodesize
= nodesize
;
1050 tree_root
->leafsize
= leafsize
;
1051 tree_root
->sectorsize
= sectorsize
;
1052 tree_root
->stripesize
= stripesize
;
1053 sb_set_blocksize(sb
, sectorsize
);
1055 i_size_write(fs_info
->btree_inode
,
1056 btrfs_super_total_bytes(disk_super
));
1058 if (strncmp((char *)(&disk_super
->magic
), BTRFS_MAGIC
,
1059 sizeof(disk_super
->magic
))) {
1060 printk("btrfs: valid FS not found on %s\n", sb
->s_id
);
1061 goto fail_sb_buffer
;
1064 mutex_lock(&fs_info
->fs_mutex
);
1066 ret
= btrfs_read_sys_array(tree_root
);
1069 blocksize
= btrfs_level_size(tree_root
,
1070 btrfs_super_chunk_root_level(disk_super
));
1072 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
1073 chunk_root
, fs_info
, BTRFS_CHUNK_TREE_OBJECTID
);
1075 chunk_root
->node
= read_tree_block(chunk_root
,
1076 btrfs_super_chunk_root(disk_super
),
1078 BUG_ON(!chunk_root
->node
);
1080 ret
= btrfs_read_chunk_tree(chunk_root
);
1083 blocksize
= btrfs_level_size(tree_root
,
1084 btrfs_super_root_level(disk_super
));
1087 tree_root
->node
= read_tree_block(tree_root
,
1088 btrfs_super_root(disk_super
),
1090 if (!tree_root
->node
)
1091 goto fail_sb_buffer
;
1094 ret
= find_and_setup_root(tree_root
, fs_info
,
1095 BTRFS_EXTENT_TREE_OBJECTID
, extent_root
);
1097 goto fail_tree_root
;
1098 extent_root
->track_dirty
= 1;
1100 ret
= find_and_setup_root(tree_root
, fs_info
,
1101 BTRFS_DEV_TREE_OBJECTID
, dev_root
);
1102 dev_root
->track_dirty
= 1;
1105 goto fail_extent_root
;
1107 btrfs_read_block_groups(extent_root
);
1109 fs_info
->generation
= btrfs_super_generation(disk_super
) + 1;
1110 fs_info
->data_alloc_profile
= (u64
)-1;
1111 fs_info
->metadata_alloc_profile
= (u64
)-1;
1112 fs_info
->system_alloc_profile
= fs_info
->metadata_alloc_profile
;
1114 mutex_unlock(&fs_info
->fs_mutex
);
1118 free_extent_buffer(extent_root
->node
);
1120 mutex_unlock(&fs_info
->fs_mutex
);
1121 free_extent_buffer(tree_root
->node
);
1123 free_extent_buffer(fs_info
->sb_buffer
);
1124 extent_io_tree_empty_lru(&BTRFS_I(fs_info
->btree_inode
)->io_tree
);
1126 iput(fs_info
->btree_inode
);
1128 close_all_devices(fs_info
);
1131 bdi_destroy(&fs_info
->bdi
);
1133 return ERR_PTR(err
);
1136 int write_ctree_super(struct btrfs_trans_handle
*trans
, struct btrfs_root
1140 struct extent_buffer
*super
= root
->fs_info
->sb_buffer
;
1141 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1142 struct super_block
*sb
= root
->fs_info
->sb
;
1144 if (!btrfs_test_opt(root
, NOBARRIER
))
1145 blkdev_issue_flush(sb
->s_bdev
, NULL
);
1146 set_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
, super
);
1147 ret
= sync_page_range_nolock(btree_inode
, btree_inode
->i_mapping
,
1148 super
->start
, super
->len
);
1149 if (!btrfs_test_opt(root
, NOBARRIER
))
1150 blkdev_issue_flush(sb
->s_bdev
, NULL
);
1154 int btrfs_free_fs_root(struct btrfs_fs_info
*fs_info
, struct btrfs_root
*root
)
1156 radix_tree_delete(&fs_info
->fs_roots_radix
,
1157 (unsigned long)root
->root_key
.objectid
);
1159 btrfs_sysfs_del_root(root
);
1163 free_extent_buffer(root
->node
);
1164 if (root
->commit_root
)
1165 free_extent_buffer(root
->commit_root
);
1172 static int del_fs_roots(struct btrfs_fs_info
*fs_info
)
1175 struct btrfs_root
*gang
[8];
1179 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
1184 for (i
= 0; i
< ret
; i
++)
1185 btrfs_free_fs_root(fs_info
, gang
[i
]);
1190 int close_ctree(struct btrfs_root
*root
)
1193 struct btrfs_trans_handle
*trans
;
1194 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1196 fs_info
->closing
= 1;
1197 btrfs_transaction_flush_work(root
);
1198 mutex_lock(&fs_info
->fs_mutex
);
1199 btrfs_defrag_dirty_roots(root
->fs_info
);
1200 trans
= btrfs_start_transaction(root
, 1);
1201 ret
= btrfs_commit_transaction(trans
, root
);
1202 /* run commit again to drop the original snapshot */
1203 trans
= btrfs_start_transaction(root
, 1);
1204 btrfs_commit_transaction(trans
, root
);
1205 ret
= btrfs_write_and_wait_transaction(NULL
, root
);
1207 write_ctree_super(NULL
, root
);
1208 mutex_unlock(&fs_info
->fs_mutex
);
1210 if (fs_info
->delalloc_bytes
) {
1211 printk("btrfs: at unmount delalloc count %Lu\n",
1212 fs_info
->delalloc_bytes
);
1214 if (fs_info
->extent_root
->node
)
1215 free_extent_buffer(fs_info
->extent_root
->node
);
1217 if (fs_info
->tree_root
->node
)
1218 free_extent_buffer(fs_info
->tree_root
->node
);
1220 if (root
->fs_info
->chunk_root
->node
);
1221 free_extent_buffer(root
->fs_info
->chunk_root
->node
);
1223 if (root
->fs_info
->dev_root
->node
);
1224 free_extent_buffer(root
->fs_info
->dev_root
->node
);
1226 free_extent_buffer(fs_info
->sb_buffer
);
1228 btrfs_free_block_groups(root
->fs_info
);
1229 del_fs_roots(fs_info
);
1231 filemap_write_and_wait(fs_info
->btree_inode
->i_mapping
);
1233 extent_io_tree_empty_lru(&fs_info
->free_space_cache
);
1234 extent_io_tree_empty_lru(&fs_info
->block_group_cache
);
1235 extent_io_tree_empty_lru(&fs_info
->pinned_extents
);
1236 extent_io_tree_empty_lru(&fs_info
->pending_del
);
1237 extent_io_tree_empty_lru(&fs_info
->extent_ins
);
1238 extent_io_tree_empty_lru(&BTRFS_I(fs_info
->btree_inode
)->io_tree
);
1240 truncate_inode_pages(fs_info
->btree_inode
->i_mapping
, 0);
1241 flush_workqueue(end_io_workqueue
);
1242 destroy_workqueue(end_io_workqueue
);
1244 iput(fs_info
->btree_inode
);
1246 while(!list_empty(&fs_info
->hashers
)) {
1247 struct btrfs_hasher
*hasher
;
1248 hasher
= list_entry(fs_info
->hashers
.next
, struct btrfs_hasher
,
1250 list_del(&hasher
->hashers
);
1251 crypto_free_hash(&fs_info
->hash_tfm
);
1255 close_all_devices(fs_info
);
1256 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
1257 bdi_destroy(&fs_info
->bdi
);
1259 kfree(fs_info
->extent_root
);
1260 kfree(fs_info
->tree_root
);
1261 kfree(fs_info
->chunk_root
);
1262 kfree(fs_info
->dev_root
);
1266 int btrfs_buffer_uptodate(struct extent_buffer
*buf
)
1268 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
1269 return extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
, buf
);
1272 int btrfs_set_buffer_uptodate(struct extent_buffer
*buf
)
1274 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
1275 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
,
1279 void btrfs_mark_buffer_dirty(struct extent_buffer
*buf
)
1281 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1282 u64 transid
= btrfs_header_generation(buf
);
1283 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1285 if (transid
!= root
->fs_info
->generation
) {
1286 printk(KERN_CRIT
"transid mismatch buffer %llu, found %Lu running %Lu\n",
1287 (unsigned long long)buf
->start
,
1288 transid
, root
->fs_info
->generation
);
1291 set_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
, buf
);
1294 void btrfs_throttle(struct btrfs_root
*root
)
1296 struct backing_dev_info
*bdi
;
1298 bdi
= root
->fs_info
->sb
->s_bdev
->bd_inode
->i_mapping
->backing_dev_info
;
1299 if (root
->fs_info
->throttles
&& bdi_write_congested(bdi
)) {
1300 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18)
1301 congestion_wait(WRITE
, HZ
/20);
1303 blk_congestion_wait(WRITE
, HZ
/20);
1308 void btrfs_btree_balance_dirty(struct btrfs_root
*root
, unsigned long nr
)
1310 balance_dirty_pages_ratelimited_nr(
1311 root
->fs_info
->btree_inode
->i_mapping
, 1);
1314 void btrfs_set_buffer_defrag(struct extent_buffer
*buf
)
1316 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1317 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1318 set_extent_bits(&BTRFS_I(btree_inode
)->io_tree
, buf
->start
,
1319 buf
->start
+ buf
->len
- 1, EXTENT_DEFRAG
, GFP_NOFS
);
1322 void btrfs_set_buffer_defrag_done(struct extent_buffer
*buf
)
1324 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1325 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1326 set_extent_bits(&BTRFS_I(btree_inode
)->io_tree
, buf
->start
,
1327 buf
->start
+ buf
->len
- 1, EXTENT_DEFRAG_DONE
,
1331 int btrfs_buffer_defrag(struct extent_buffer
*buf
)
1333 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1334 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1335 return test_range_bit(&BTRFS_I(btree_inode
)->io_tree
,
1336 buf
->start
, buf
->start
+ buf
->len
- 1, EXTENT_DEFRAG
, 0);
1339 int btrfs_buffer_defrag_done(struct extent_buffer
*buf
)
1341 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1342 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1343 return test_range_bit(&BTRFS_I(btree_inode
)->io_tree
,
1344 buf
->start
, buf
->start
+ buf
->len
- 1,
1345 EXTENT_DEFRAG_DONE
, 0);
1348 int btrfs_clear_buffer_defrag_done(struct extent_buffer
*buf
)
1350 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1351 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1352 return clear_extent_bits(&BTRFS_I(btree_inode
)->io_tree
,
1353 buf
->start
, buf
->start
+ buf
->len
- 1,
1354 EXTENT_DEFRAG_DONE
, GFP_NOFS
);
1357 int btrfs_clear_buffer_defrag(struct extent_buffer
*buf
)
1359 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1360 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1361 return clear_extent_bits(&BTRFS_I(btree_inode
)->io_tree
,
1362 buf
->start
, buf
->start
+ buf
->len
- 1,
1363 EXTENT_DEFRAG
, GFP_NOFS
);
1366 int btrfs_read_buffer(struct extent_buffer
*buf
)
1368 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1369 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1371 ret
= read_extent_buffer_pages(&BTRFS_I(btree_inode
)->io_tree
,
1372 buf
, 0, 1, btree_get_extent
);
1374 buf
->flags
|= EXTENT_UPTODATE
;
1379 static struct extent_io_ops btree_extent_io_ops
= {
1380 .writepage_io_hook
= btree_writepage_io_hook
,
1381 .readpage_end_io_hook
= btree_readpage_end_io_hook
,
1382 .submit_bio_hook
= btree_submit_bio_hook
,
1383 /* note we're sharing with inode.c for the merge bio hook */
1384 .merge_bio_hook
= btrfs_merge_bio_hook
,