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
29 #include "transaction.h"
30 #include "btrfs_inode.h"
32 #include "print-tree.h"
35 static int check_tree_block(struct btrfs_root
*root
, struct extent_buffer
*buf
)
37 if (extent_buffer_blocknr(buf
) != btrfs_header_blocknr(buf
)) {
38 printk(KERN_CRIT
"buf blocknr(buf) is %llu, header is %llu\n",
39 (unsigned long long)extent_buffer_blocknr(buf
),
40 (unsigned long long)btrfs_header_blocknr(buf
));
47 static struct extent_io_ops btree_extent_io_ops
;
49 struct extent_map
*btree_get_extent(struct inode
*inode
, struct page
*page
,
50 size_t page_offset
, u64 start
, u64 len
,
53 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
54 struct extent_map
*em
;
58 spin_lock(&em_tree
->lock
);
59 em
= lookup_extent_mapping(em_tree
, start
, len
);
60 spin_unlock(&em_tree
->lock
);
64 em
= alloc_extent_map(GFP_NOFS
);
66 em
= ERR_PTR(-ENOMEM
);
70 em
->len
= i_size_read(inode
);
72 em
->bdev
= inode
->i_sb
->s_bdev
;
74 spin_lock(&em_tree
->lock
);
75 ret
= add_extent_mapping(em_tree
, em
);
76 spin_unlock(&em_tree
->lock
);
89 u32
btrfs_csum_data(struct btrfs_root
*root
, char *data
, u32 seed
, size_t len
)
91 return crc32c(seed
, data
, len
);
94 void btrfs_csum_final(u32 crc
, char *result
)
96 *(__le32
*)result
= ~cpu_to_le32(crc
);
99 static int csum_tree_block(struct btrfs_root
*root
, struct extent_buffer
*buf
,
102 char result
[BTRFS_CRC32_SIZE
];
104 unsigned long cur_len
;
105 unsigned long offset
= BTRFS_CSUM_SIZE
;
106 char *map_token
= NULL
;
108 unsigned long map_start
;
109 unsigned long map_len
;
113 len
= buf
->len
- offset
;
115 err
= map_private_extent_buffer(buf
, offset
, 32,
117 &map_start
, &map_len
, KM_USER0
);
119 printk("failed to map extent buffer! %lu\n",
123 cur_len
= min(len
, map_len
- (offset
- map_start
));
124 crc
= btrfs_csum_data(root
, kaddr
+ offset
- map_start
,
128 unmap_extent_buffer(buf
, map_token
, KM_USER0
);
130 btrfs_csum_final(crc
, result
);
133 int from_this_trans
= 0;
135 if (root
->fs_info
->running_transaction
&&
136 btrfs_header_generation(buf
) ==
137 root
->fs_info
->running_transaction
->transid
)
140 /* FIXME, this is not good */
141 if (memcmp_extent_buffer(buf
, result
, 0, BTRFS_CRC32_SIZE
)) {
144 memcpy(&found
, result
, BTRFS_CRC32_SIZE
);
146 read_extent_buffer(buf
, &val
, 0, BTRFS_CRC32_SIZE
);
148 printk("btrfs: %s checksum verify failed on %llu "
149 "wanted %X found %X from_this_trans %d "
151 root
->fs_info
->sb
->s_id
,
152 buf
->start
, val
, found
, from_this_trans
,
153 btrfs_header_level(buf
));
157 write_extent_buffer(buf
, result
, 0, BTRFS_CRC32_SIZE
);
163 int csum_dirty_buffer(struct btrfs_root
*root
, struct page
*page
)
165 struct extent_io_tree
*tree
;
166 u64 start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
170 struct extent_buffer
*eb
;
171 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
173 if (page
->private == EXTENT_PAGE_PRIVATE
)
177 len
= page
->private >> 2;
181 eb
= alloc_extent_buffer(tree
, start
, len
, page
, GFP_NOFS
);
182 read_extent_buffer_pages(tree
, eb
, start
+ PAGE_CACHE_SIZE
, 1,
184 btrfs_clear_buffer_defrag(eb
);
185 found_start
= btrfs_header_bytenr(eb
);
186 if (found_start
!= start
) {
187 printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
188 start
, found_start
, len
);
192 if (eb
->first_page
!= page
) {
193 printk("bad first page %lu %lu\n", eb
->first_page
->index
,
198 if (!PageUptodate(page
)) {
199 printk("csum not up to date page %lu\n", page
->index
);
203 found_level
= btrfs_header_level(eb
);
204 spin_lock(&root
->fs_info
->hash_lock
);
205 btrfs_set_header_flag(eb
, BTRFS_HEADER_FLAG_WRITTEN
);
206 spin_unlock(&root
->fs_info
->hash_lock
);
207 csum_tree_block(root
, eb
, 0);
209 free_extent_buffer(eb
);
214 static int btree_writepage_io_hook(struct page
*page
, u64 start
, u64 end
)
216 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
218 csum_dirty_buffer(root
, page
);
222 static int btree_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
)
224 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
226 offset
= bio
->bi_sector
<< 9;
227 if (offset
== BTRFS_SUPER_INFO_OFFSET
) {
228 bio
->bi_bdev
= root
->fs_info
->sb
->s_bdev
;
232 return btrfs_map_bio(BTRFS_I(inode
)->root
, rw
, bio
);
235 static int btree_writepage(struct page
*page
, struct writeback_control
*wbc
)
237 struct extent_io_tree
*tree
;
238 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
239 return extent_write_full_page(tree
, page
, btree_get_extent
, wbc
);
242 static int btree_writepages(struct address_space
*mapping
,
243 struct writeback_control
*wbc
)
245 struct extent_io_tree
*tree
;
246 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
247 if (wbc
->sync_mode
== WB_SYNC_NONE
) {
250 unsigned long thresh
= 96 * 1024 * 1024;
252 if (wbc
->for_kupdate
)
255 if (current_is_pdflush()) {
256 thresh
= 96 * 1024 * 1024;
258 thresh
= 8 * 1024 * 1024;
260 num_dirty
= count_range_bits(tree
, &start
, (u64
)-1,
261 thresh
, EXTENT_DIRTY
);
262 if (num_dirty
< thresh
) {
266 return extent_writepages(tree
, mapping
, btree_get_extent
, wbc
);
269 int btree_readpage(struct file
*file
, struct page
*page
)
271 struct extent_io_tree
*tree
;
272 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
273 return extent_read_full_page(tree
, page
, btree_get_extent
);
276 static int btree_releasepage(struct page
*page
, gfp_t gfp_flags
)
278 struct extent_io_tree
*tree
;
279 struct extent_map_tree
*map
;
282 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
283 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
284 ret
= try_release_extent_mapping(map
, tree
, page
, gfp_flags
);
286 invalidate_extent_lru(tree
, page_offset(page
), PAGE_CACHE_SIZE
);
287 ClearPagePrivate(page
);
288 set_page_private(page
, 0);
289 page_cache_release(page
);
294 static void btree_invalidatepage(struct page
*page
, unsigned long offset
)
296 struct extent_io_tree
*tree
;
297 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
298 extent_invalidatepage(tree
, page
, offset
);
299 btree_releasepage(page
, GFP_NOFS
);
303 static int btree_writepage(struct page
*page
, struct writeback_control
*wbc
)
305 struct buffer_head
*bh
;
306 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
307 struct buffer_head
*head
;
308 if (!page_has_buffers(page
)) {
309 create_empty_buffers(page
, root
->fs_info
->sb
->s_blocksize
,
310 (1 << BH_Dirty
)|(1 << BH_Uptodate
));
312 head
= page_buffers(page
);
315 if (buffer_dirty(bh
))
316 csum_tree_block(root
, bh
, 0);
317 bh
= bh
->b_this_page
;
318 } while (bh
!= head
);
319 return block_write_full_page(page
, btree_get_block
, wbc
);
323 static struct address_space_operations btree_aops
= {
324 .readpage
= btree_readpage
,
325 .writepage
= btree_writepage
,
326 .writepages
= btree_writepages
,
327 .releasepage
= btree_releasepage
,
328 .invalidatepage
= btree_invalidatepage
,
329 .sync_page
= block_sync_page
,
332 int readahead_tree_block(struct btrfs_root
*root
, u64 bytenr
, u32 blocksize
)
334 struct extent_buffer
*buf
= NULL
;
335 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
338 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
341 read_extent_buffer_pages(&BTRFS_I(btree_inode
)->io_tree
,
342 buf
, 0, 0, btree_get_extent
);
343 free_extent_buffer(buf
);
347 static int close_all_devices(struct btrfs_fs_info
*fs_info
)
349 struct list_head
*list
;
350 struct list_head
*next
;
351 struct btrfs_device
*device
;
353 list
= &fs_info
->fs_devices
->devices
;
354 list_for_each(next
, list
) {
355 device
= list_entry(next
, struct btrfs_device
, dev_list
);
356 if (device
->bdev
&& device
->bdev
!= fs_info
->sb
->s_bdev
)
357 close_bdev_excl(device
->bdev
);
363 int btrfs_verify_block_csum(struct btrfs_root
*root
,
364 struct extent_buffer
*buf
)
366 struct extent_io_tree
*io_tree
;
370 io_tree
= &BTRFS_I(root
->fs_info
->btree_inode
)->io_tree
;
371 if (buf
->flags
& EXTENT_CSUM
)
374 end
= min_t(u64
, buf
->len
, PAGE_CACHE_SIZE
);
375 end
= buf
->start
+ end
- 1;
376 if (test_range_bit(io_tree
, buf
->start
, end
, EXTENT_CSUM
, 1)) {
377 buf
->flags
|= EXTENT_CSUM
;
380 lock_extent(io_tree
, buf
->start
, end
, GFP_NOFS
);
382 if (test_range_bit(io_tree
, buf
->start
, end
, EXTENT_CSUM
, 1)) {
383 buf
->flags
|= EXTENT_CSUM
;
387 WARN_ON(buf
->flags
& EXTENT_CSUM
);
389 ret
= csum_tree_block(root
, buf
, 1);
390 set_extent_bits(io_tree
, buf
->start
, end
, EXTENT_CSUM
, GFP_NOFS
);
391 buf
->flags
|= EXTENT_CSUM
;
394 unlock_extent(io_tree
, buf
->start
, end
, GFP_NOFS
);
398 struct extent_buffer
*btrfs_find_tree_block(struct btrfs_root
*root
,
399 u64 bytenr
, u32 blocksize
)
401 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
402 struct extent_buffer
*eb
;
403 eb
= find_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
404 bytenr
, blocksize
, GFP_NOFS
);
408 struct extent_buffer
*btrfs_find_create_tree_block(struct btrfs_root
*root
,
409 u64 bytenr
, u32 blocksize
)
411 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
412 struct extent_buffer
*eb
;
414 eb
= alloc_extent_buffer(&BTRFS_I(btree_inode
)->io_tree
,
415 bytenr
, blocksize
, NULL
, GFP_NOFS
);
420 struct extent_buffer
*read_tree_block(struct btrfs_root
*root
, u64 bytenr
,
423 struct extent_buffer
*buf
= NULL
;
424 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
425 struct extent_io_tree
*io_tree
;
428 io_tree
= &BTRFS_I(btree_inode
)->io_tree
;
430 buf
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
433 read_extent_buffer_pages(&BTRFS_I(btree_inode
)->io_tree
, buf
, 0, 1,
436 ret
= btrfs_verify_block_csum(root
, buf
);
440 int clean_tree_block(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
441 struct extent_buffer
*buf
)
443 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
444 if (btrfs_header_generation(buf
) ==
445 root
->fs_info
->running_transaction
->transid
)
446 clear_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
,
451 int wait_on_tree_block_writeback(struct btrfs_root
*root
,
452 struct extent_buffer
*buf
)
454 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
455 wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode
)->io_tree
,
460 static int __setup_root(u32 nodesize
, u32 leafsize
, u32 sectorsize
,
461 u32 stripesize
, struct btrfs_root
*root
,
462 struct btrfs_fs_info
*fs_info
,
467 root
->commit_root
= NULL
;
468 root
->sectorsize
= sectorsize
;
469 root
->nodesize
= nodesize
;
470 root
->leafsize
= leafsize
;
471 root
->stripesize
= stripesize
;
473 root
->track_dirty
= 0;
475 root
->fs_info
= fs_info
;
476 root
->objectid
= objectid
;
477 root
->last_trans
= 0;
478 root
->highest_inode
= 0;
479 root
->last_inode_alloc
= 0;
483 INIT_LIST_HEAD(&root
->dirty_list
);
484 memset(&root
->root_key
, 0, sizeof(root
->root_key
));
485 memset(&root
->root_item
, 0, sizeof(root
->root_item
));
486 memset(&root
->defrag_progress
, 0, sizeof(root
->defrag_progress
));
487 memset(&root
->root_kobj
, 0, sizeof(root
->root_kobj
));
488 init_completion(&root
->kobj_unregister
);
489 root
->defrag_running
= 0;
490 root
->defrag_level
= 0;
491 root
->root_key
.objectid
= objectid
;
495 static int find_and_setup_root(struct btrfs_root
*tree_root
,
496 struct btrfs_fs_info
*fs_info
,
498 struct btrfs_root
*root
)
503 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
504 tree_root
->sectorsize
, tree_root
->stripesize
,
505 root
, fs_info
, objectid
);
506 ret
= btrfs_find_last_root(tree_root
, objectid
,
507 &root
->root_item
, &root
->root_key
);
510 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
511 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
517 struct btrfs_root
*btrfs_read_fs_root_no_radix(struct btrfs_fs_info
*fs_info
,
518 struct btrfs_key
*location
)
520 struct btrfs_root
*root
;
521 struct btrfs_root
*tree_root
= fs_info
->tree_root
;
522 struct btrfs_path
*path
;
523 struct extent_buffer
*l
;
528 root
= kzalloc(sizeof(*root
), GFP_NOFS
);
530 return ERR_PTR(-ENOMEM
);
531 if (location
->offset
== (u64
)-1) {
532 ret
= find_and_setup_root(tree_root
, fs_info
,
533 location
->objectid
, root
);
541 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
542 tree_root
->sectorsize
, tree_root
->stripesize
,
543 root
, fs_info
, location
->objectid
);
545 path
= btrfs_alloc_path();
547 ret
= btrfs_search_slot(NULL
, tree_root
, location
, path
, 0, 0);
554 read_extent_buffer(l
, &root
->root_item
,
555 btrfs_item_ptr_offset(l
, path
->slots
[0]),
556 sizeof(root
->root_item
));
557 memcpy(&root
->root_key
, location
, sizeof(*location
));
560 btrfs_release_path(root
, path
);
561 btrfs_free_path(path
);
566 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
567 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
572 ret
= btrfs_find_highest_inode(root
, &highest_inode
);
574 root
->highest_inode
= highest_inode
;
575 root
->last_inode_alloc
= highest_inode
;
580 struct btrfs_root
*btrfs_lookup_fs_root(struct btrfs_fs_info
*fs_info
,
583 struct btrfs_root
*root
;
585 if (root_objectid
== BTRFS_ROOT_TREE_OBJECTID
)
586 return fs_info
->tree_root
;
587 if (root_objectid
== BTRFS_EXTENT_TREE_OBJECTID
)
588 return fs_info
->extent_root
;
590 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
591 (unsigned long)root_objectid
);
595 struct btrfs_root
*btrfs_read_fs_root_no_name(struct btrfs_fs_info
*fs_info
,
596 struct btrfs_key
*location
)
598 struct btrfs_root
*root
;
601 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
602 return fs_info
->tree_root
;
603 if (location
->objectid
== BTRFS_EXTENT_TREE_OBJECTID
)
604 return fs_info
->extent_root
;
606 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
607 (unsigned long)location
->objectid
);
611 root
= btrfs_read_fs_root_no_radix(fs_info
, location
);
614 ret
= radix_tree_insert(&fs_info
->fs_roots_radix
,
615 (unsigned long)root
->root_key
.objectid
,
618 free_extent_buffer(root
->node
);
622 ret
= btrfs_find_dead_roots(fs_info
->tree_root
,
623 root
->root_key
.objectid
, root
);
629 struct btrfs_root
*btrfs_read_fs_root(struct btrfs_fs_info
*fs_info
,
630 struct btrfs_key
*location
,
631 const char *name
, int namelen
)
633 struct btrfs_root
*root
;
636 root
= btrfs_read_fs_root_no_name(fs_info
, location
);
643 ret
= btrfs_set_root_name(root
, name
, namelen
);
645 free_extent_buffer(root
->node
);
650 ret
= btrfs_sysfs_add_root(root
);
652 free_extent_buffer(root
->node
);
661 static int add_hasher(struct btrfs_fs_info
*info
, char *type
) {
662 struct btrfs_hasher
*hasher
;
664 hasher
= kmalloc(sizeof(*hasher
), GFP_NOFS
);
667 hasher
->hash_tfm
= crypto_alloc_hash(type
, 0, CRYPTO_ALG_ASYNC
);
668 if (!hasher
->hash_tfm
) {
672 spin_lock(&info
->hash_lock
);
673 list_add(&hasher
->list
, &info
->hashers
);
674 spin_unlock(&info
->hash_lock
);
679 static int btrfs_congested_fn(void *congested_data
, int bdi_bits
)
681 struct btrfs_fs_info
*info
= (struct btrfs_fs_info
*)congested_data
;
683 struct list_head
*cur
;
684 struct btrfs_device
*device
;
685 struct backing_dev_info
*bdi
;
687 list_for_each(cur
, &info
->fs_devices
->devices
) {
688 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
689 bdi
= blk_get_backing_dev_info(device
->bdev
);
690 if (bdi
&& bdi_congested(bdi
, bdi_bits
)) {
698 void btrfs_unplug_io_fn(struct backing_dev_info
*bdi
, struct page
*page
)
700 struct list_head
*cur
;
701 struct btrfs_device
*device
;
702 struct btrfs_fs_info
*info
;
704 info
= (struct btrfs_fs_info
*)bdi
->unplug_io_data
;
705 list_for_each(cur
, &info
->fs_devices
->devices
) {
706 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
707 bdi
= blk_get_backing_dev_info(device
->bdev
);
708 if (bdi
->unplug_io_fn
) {
709 bdi
->unplug_io_fn(bdi
, page
);
714 static int setup_bdi(struct btrfs_fs_info
*info
, struct backing_dev_info
*bdi
)
717 bdi
->ra_pages
= default_backing_dev_info
.ra_pages
* 4;
719 bdi
->capabilities
= default_backing_dev_info
.capabilities
;
720 bdi
->unplug_io_fn
= btrfs_unplug_io_fn
;
721 bdi
->unplug_io_data
= info
;
722 bdi
->congested_fn
= btrfs_congested_fn
;
723 bdi
->congested_data
= info
;
727 struct btrfs_root
*open_ctree(struct super_block
*sb
,
728 struct btrfs_fs_devices
*fs_devices
)
735 struct btrfs_root
*extent_root
= kmalloc(sizeof(struct btrfs_root
),
737 struct btrfs_root
*tree_root
= kmalloc(sizeof(struct btrfs_root
),
739 struct btrfs_fs_info
*fs_info
= kzalloc(sizeof(*fs_info
),
741 struct btrfs_root
*chunk_root
= kmalloc(sizeof(struct btrfs_root
),
743 struct btrfs_root
*dev_root
= kmalloc(sizeof(struct btrfs_root
),
747 struct btrfs_super_block
*disk_super
;
749 if (!extent_root
|| !tree_root
|| !fs_info
) {
753 INIT_RADIX_TREE(&fs_info
->fs_roots_radix
, GFP_NOFS
);
754 INIT_LIST_HEAD(&fs_info
->trans_list
);
755 INIT_LIST_HEAD(&fs_info
->dead_roots
);
756 INIT_LIST_HEAD(&fs_info
->hashers
);
757 spin_lock_init(&fs_info
->hash_lock
);
758 spin_lock_init(&fs_info
->delalloc_lock
);
759 spin_lock_init(&fs_info
->new_trans_lock
);
761 init_completion(&fs_info
->kobj_unregister
);
762 sb_set_blocksize(sb
, 4096);
763 fs_info
->tree_root
= tree_root
;
764 fs_info
->extent_root
= extent_root
;
765 fs_info
->chunk_root
= chunk_root
;
766 fs_info
->dev_root
= dev_root
;
767 fs_info
->fs_devices
= fs_devices
;
768 INIT_LIST_HEAD(&fs_info
->dirty_cowonly_roots
);
769 INIT_LIST_HEAD(&fs_info
->space_info
);
770 btrfs_mapping_init(&fs_info
->mapping_tree
);
772 fs_info
->max_extent
= (u64
)-1;
773 fs_info
->max_inline
= 8192 * 1024;
774 setup_bdi(fs_info
, &fs_info
->bdi
);
775 fs_info
->btree_inode
= new_inode(sb
);
776 fs_info
->btree_inode
->i_ino
= 1;
777 fs_info
->btree_inode
->i_nlink
= 1;
778 fs_info
->btree_inode
->i_size
= sb
->s_bdev
->bd_inode
->i_size
;
779 fs_info
->btree_inode
->i_mapping
->a_ops
= &btree_aops
;
780 fs_info
->btree_inode
->i_mapping
->backing_dev_info
= &fs_info
->bdi
;
782 extent_io_tree_init(&BTRFS_I(fs_info
->btree_inode
)->io_tree
,
783 fs_info
->btree_inode
->i_mapping
,
785 extent_map_tree_init(&BTRFS_I(fs_info
->btree_inode
)->extent_tree
,
788 BTRFS_I(fs_info
->btree_inode
)->io_tree
.ops
= &btree_extent_io_ops
;
790 extent_io_tree_init(&fs_info
->free_space_cache
,
791 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
792 extent_io_tree_init(&fs_info
->block_group_cache
,
793 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
794 extent_io_tree_init(&fs_info
->pinned_extents
,
795 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
796 extent_io_tree_init(&fs_info
->pending_del
,
797 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
798 extent_io_tree_init(&fs_info
->extent_ins
,
799 fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
800 fs_info
->do_barriers
= 1;
802 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
803 INIT_WORK(&fs_info
->trans_work
, btrfs_transaction_cleaner
, fs_info
);
805 INIT_DELAYED_WORK(&fs_info
->trans_work
, btrfs_transaction_cleaner
);
807 BTRFS_I(fs_info
->btree_inode
)->root
= tree_root
;
808 memset(&BTRFS_I(fs_info
->btree_inode
)->location
, 0,
809 sizeof(struct btrfs_key
));
810 insert_inode_hash(fs_info
->btree_inode
);
811 mapping_set_gfp_mask(fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
813 mutex_init(&fs_info
->trans_mutex
);
814 mutex_init(&fs_info
->fs_mutex
);
817 ret
= add_hasher(fs_info
, "crc32c");
819 printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
824 __setup_root(4096, 4096, 4096, 4096, tree_root
,
825 fs_info
, BTRFS_ROOT_TREE_OBJECTID
);
827 fs_info
->sb_buffer
= read_tree_block(tree_root
,
828 BTRFS_SUPER_INFO_OFFSET
,
831 if (!fs_info
->sb_buffer
)
834 read_extent_buffer(fs_info
->sb_buffer
, &fs_info
->super_copy
, 0,
835 sizeof(fs_info
->super_copy
));
837 read_extent_buffer(fs_info
->sb_buffer
, fs_info
->fsid
,
838 (unsigned long)btrfs_super_fsid(fs_info
->sb_buffer
),
841 disk_super
= &fs_info
->super_copy
;
842 if (!btrfs_super_root(disk_super
))
845 if (btrfs_super_num_devices(disk_super
) != fs_devices
->num_devices
) {
846 printk("Btrfs: wanted %llu devices, but found %llu\n",
847 (unsigned long long)btrfs_super_num_devices(disk_super
),
848 (unsigned long long)fs_devices
->num_devices
);
851 nodesize
= btrfs_super_nodesize(disk_super
);
852 leafsize
= btrfs_super_leafsize(disk_super
);
853 sectorsize
= btrfs_super_sectorsize(disk_super
);
854 stripesize
= btrfs_super_stripesize(disk_super
);
855 tree_root
->nodesize
= nodesize
;
856 tree_root
->leafsize
= leafsize
;
857 tree_root
->sectorsize
= sectorsize
;
858 tree_root
->stripesize
= stripesize
;
859 sb_set_blocksize(sb
, sectorsize
);
861 i_size_write(fs_info
->btree_inode
,
862 btrfs_super_total_bytes(disk_super
));
864 if (strncmp((char *)(&disk_super
->magic
), BTRFS_MAGIC
,
865 sizeof(disk_super
->magic
))) {
866 printk("btrfs: valid FS not found on %s\n", sb
->s_id
);
870 mutex_lock(&fs_info
->fs_mutex
);
872 ret
= btrfs_read_sys_array(tree_root
);
875 blocksize
= btrfs_level_size(tree_root
,
876 btrfs_super_chunk_root_level(disk_super
));
878 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
879 chunk_root
, fs_info
, BTRFS_CHUNK_TREE_OBJECTID
);
881 chunk_root
->node
= read_tree_block(chunk_root
,
882 btrfs_super_chunk_root(disk_super
),
884 BUG_ON(!chunk_root
->node
);
886 ret
= btrfs_read_chunk_tree(chunk_root
);
889 blocksize
= btrfs_level_size(tree_root
,
890 btrfs_super_root_level(disk_super
));
893 tree_root
->node
= read_tree_block(tree_root
,
894 btrfs_super_root(disk_super
),
896 if (!tree_root
->node
)
900 ret
= find_and_setup_root(tree_root
, fs_info
,
901 BTRFS_EXTENT_TREE_OBJECTID
, extent_root
);
904 extent_root
->track_dirty
= 1;
906 ret
= find_and_setup_root(tree_root
, fs_info
,
907 BTRFS_DEV_TREE_OBJECTID
, dev_root
);
908 dev_root
->track_dirty
= 1;
911 goto fail_extent_root
;
913 btrfs_read_block_groups(extent_root
);
915 fs_info
->generation
= btrfs_super_generation(disk_super
) + 1;
916 fs_info
->data_alloc_profile
= (u64
)-1;
917 fs_info
->metadata_alloc_profile
= (u64
)-1;
918 fs_info
->system_alloc_profile
= fs_info
->metadata_alloc_profile
;
920 mutex_unlock(&fs_info
->fs_mutex
);
924 free_extent_buffer(extent_root
->node
);
926 mutex_unlock(&fs_info
->fs_mutex
);
927 free_extent_buffer(tree_root
->node
);
929 free_extent_buffer(fs_info
->sb_buffer
);
930 extent_io_tree_empty_lru(&BTRFS_I(fs_info
->btree_inode
)->io_tree
);
932 iput(fs_info
->btree_inode
);
934 close_all_devices(fs_info
);
937 bdi_destroy(&fs_info
->bdi
);
942 int write_ctree_super(struct btrfs_trans_handle
*trans
, struct btrfs_root
946 struct extent_buffer
*super
= root
->fs_info
->sb_buffer
;
947 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
948 struct super_block
*sb
= root
->fs_info
->sb
;
950 if (!btrfs_test_opt(root
, NOBARRIER
))
951 blkdev_issue_flush(sb
->s_bdev
, NULL
);
952 set_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
, super
);
953 ret
= sync_page_range_nolock(btree_inode
, btree_inode
->i_mapping
,
954 super
->start
, super
->len
);
955 if (!btrfs_test_opt(root
, NOBARRIER
))
956 blkdev_issue_flush(sb
->s_bdev
, NULL
);
960 int btrfs_free_fs_root(struct btrfs_fs_info
*fs_info
, struct btrfs_root
*root
)
962 radix_tree_delete(&fs_info
->fs_roots_radix
,
963 (unsigned long)root
->root_key
.objectid
);
965 btrfs_sysfs_del_root(root
);
969 free_extent_buffer(root
->node
);
970 if (root
->commit_root
)
971 free_extent_buffer(root
->commit_root
);
978 static int del_fs_roots(struct btrfs_fs_info
*fs_info
)
981 struct btrfs_root
*gang
[8];
985 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
990 for (i
= 0; i
< ret
; i
++)
991 btrfs_free_fs_root(fs_info
, gang
[i
]);
996 int close_ctree(struct btrfs_root
*root
)
999 struct btrfs_trans_handle
*trans
;
1000 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1002 fs_info
->closing
= 1;
1003 btrfs_transaction_flush_work(root
);
1004 mutex_lock(&fs_info
->fs_mutex
);
1005 btrfs_defrag_dirty_roots(root
->fs_info
);
1006 trans
= btrfs_start_transaction(root
, 1);
1007 ret
= btrfs_commit_transaction(trans
, root
);
1008 /* run commit again to drop the original snapshot */
1009 trans
= btrfs_start_transaction(root
, 1);
1010 btrfs_commit_transaction(trans
, root
);
1011 ret
= btrfs_write_and_wait_transaction(NULL
, root
);
1013 write_ctree_super(NULL
, root
);
1014 mutex_unlock(&fs_info
->fs_mutex
);
1016 if (fs_info
->delalloc_bytes
) {
1017 printk("btrfs: at unmount delalloc count %Lu\n",
1018 fs_info
->delalloc_bytes
);
1020 if (fs_info
->extent_root
->node
)
1021 free_extent_buffer(fs_info
->extent_root
->node
);
1023 if (fs_info
->tree_root
->node
)
1024 free_extent_buffer(fs_info
->tree_root
->node
);
1026 if (root
->fs_info
->chunk_root
->node
);
1027 free_extent_buffer(root
->fs_info
->chunk_root
->node
);
1029 if (root
->fs_info
->dev_root
->node
);
1030 free_extent_buffer(root
->fs_info
->dev_root
->node
);
1032 free_extent_buffer(fs_info
->sb_buffer
);
1034 btrfs_free_block_groups(root
->fs_info
);
1035 del_fs_roots(fs_info
);
1037 filemap_write_and_wait(fs_info
->btree_inode
->i_mapping
);
1039 extent_io_tree_empty_lru(&fs_info
->free_space_cache
);
1040 extent_io_tree_empty_lru(&fs_info
->block_group_cache
);
1041 extent_io_tree_empty_lru(&fs_info
->pinned_extents
);
1042 extent_io_tree_empty_lru(&fs_info
->pending_del
);
1043 extent_io_tree_empty_lru(&fs_info
->extent_ins
);
1044 extent_io_tree_empty_lru(&BTRFS_I(fs_info
->btree_inode
)->io_tree
);
1046 truncate_inode_pages(fs_info
->btree_inode
->i_mapping
, 0);
1048 iput(fs_info
->btree_inode
);
1050 while(!list_empty(&fs_info
->hashers
)) {
1051 struct btrfs_hasher
*hasher
;
1052 hasher
= list_entry(fs_info
->hashers
.next
, struct btrfs_hasher
,
1054 list_del(&hasher
->hashers
);
1055 crypto_free_hash(&fs_info
->hash_tfm
);
1059 close_all_devices(fs_info
);
1060 btrfs_mapping_tree_free(&fs_info
->mapping_tree
);
1061 bdi_destroy(&fs_info
->bdi
);
1063 kfree(fs_info
->extent_root
);
1064 kfree(fs_info
->tree_root
);
1065 kfree(fs_info
->chunk_root
);
1066 kfree(fs_info
->dev_root
);
1070 int btrfs_buffer_uptodate(struct extent_buffer
*buf
)
1072 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
1073 return extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
, buf
);
1076 int btrfs_set_buffer_uptodate(struct extent_buffer
*buf
)
1078 struct inode
*btree_inode
= buf
->first_page
->mapping
->host
;
1079 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode
)->io_tree
,
1083 void btrfs_mark_buffer_dirty(struct extent_buffer
*buf
)
1085 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1086 u64 transid
= btrfs_header_generation(buf
);
1087 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1089 if (transid
!= root
->fs_info
->generation
) {
1090 printk(KERN_CRIT
"transid mismatch buffer %llu, found %Lu running %Lu\n",
1091 (unsigned long long)buf
->start
,
1092 transid
, root
->fs_info
->generation
);
1095 set_extent_buffer_dirty(&BTRFS_I(btree_inode
)->io_tree
, buf
);
1098 void btrfs_throttle(struct btrfs_root
*root
)
1100 struct backing_dev_info
*bdi
;
1102 bdi
= root
->fs_info
->sb
->s_bdev
->bd_inode
->i_mapping
->backing_dev_info
;
1103 if (root
->fs_info
->throttles
&& bdi_write_congested(bdi
)) {
1104 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18)
1105 congestion_wait(WRITE
, HZ
/20);
1107 blk_congestion_wait(WRITE
, HZ
/20);
1112 void btrfs_btree_balance_dirty(struct btrfs_root
*root
, unsigned long nr
)
1114 balance_dirty_pages_ratelimited_nr(
1115 root
->fs_info
->btree_inode
->i_mapping
, 1);
1118 void btrfs_set_buffer_defrag(struct extent_buffer
*buf
)
1120 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1121 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1122 set_extent_bits(&BTRFS_I(btree_inode
)->io_tree
, buf
->start
,
1123 buf
->start
+ buf
->len
- 1, EXTENT_DEFRAG
, GFP_NOFS
);
1126 void btrfs_set_buffer_defrag_done(struct extent_buffer
*buf
)
1128 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1129 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1130 set_extent_bits(&BTRFS_I(btree_inode
)->io_tree
, buf
->start
,
1131 buf
->start
+ buf
->len
- 1, EXTENT_DEFRAG_DONE
,
1135 int btrfs_buffer_defrag(struct extent_buffer
*buf
)
1137 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1138 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1139 return test_range_bit(&BTRFS_I(btree_inode
)->io_tree
,
1140 buf
->start
, buf
->start
+ buf
->len
- 1, EXTENT_DEFRAG
, 0);
1143 int btrfs_buffer_defrag_done(struct extent_buffer
*buf
)
1145 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1146 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1147 return test_range_bit(&BTRFS_I(btree_inode
)->io_tree
,
1148 buf
->start
, buf
->start
+ buf
->len
- 1,
1149 EXTENT_DEFRAG_DONE
, 0);
1152 int btrfs_clear_buffer_defrag_done(struct extent_buffer
*buf
)
1154 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1155 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1156 return clear_extent_bits(&BTRFS_I(btree_inode
)->io_tree
,
1157 buf
->start
, buf
->start
+ buf
->len
- 1,
1158 EXTENT_DEFRAG_DONE
, GFP_NOFS
);
1161 int btrfs_clear_buffer_defrag(struct extent_buffer
*buf
)
1163 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1164 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1165 return clear_extent_bits(&BTRFS_I(btree_inode
)->io_tree
,
1166 buf
->start
, buf
->start
+ buf
->len
- 1,
1167 EXTENT_DEFRAG
, GFP_NOFS
);
1170 int btrfs_read_buffer(struct extent_buffer
*buf
)
1172 struct btrfs_root
*root
= BTRFS_I(buf
->first_page
->mapping
->host
)->root
;
1173 struct inode
*btree_inode
= root
->fs_info
->btree_inode
;
1174 return read_extent_buffer_pages(&BTRFS_I(btree_inode
)->io_tree
,
1175 buf
, 0, 1, btree_get_extent
);
1178 static struct extent_io_ops btree_extent_io_ops
= {
1179 .writepage_io_hook
= btree_writepage_io_hook
,
1180 .submit_bio_hook
= btree_submit_bio_hook
,
1181 /* note we're sharing with inode.c for the merge bio hook */
1182 .merge_bio_hook
= btrfs_merge_bio_hook
,