1 #include <linux/module.h>
3 #include <linux/blkdev.h>
4 #include <linux/crypto.h>
5 #include <linux/scatterlist.h>
6 #include <linux/swap.h>
7 #include <linux/radix-tree.h>
10 #include "transaction.h"
11 #include "btrfs_inode.h"
17 struct block_device
*bdev
;
20 int btrfs_insert_dev_radix(struct btrfs_root
*root
,
21 struct block_device
*bdev
,
26 struct dev_lookup
*lookup
;
29 lookup
= kmalloc(sizeof(*lookup
), GFP_NOFS
);
32 lookup
->block_start
= block_start
;
33 lookup
->num_blocks
= num_blocks
;
35 lookup
->device_id
= device_id
;
37 ret
= radix_tree_insert(&root
->fs_info
->dev_radix
, block_start
+
38 num_blocks
- 1, lookup
);
42 u64
bh_blocknr(struct buffer_head
*bh
)
44 int blkbits
= bh
->b_page
->mapping
->host
->i_blkbits
;
45 u64 blocknr
= bh
->b_page
->index
<< (PAGE_CACHE_SHIFT
- blkbits
);
48 if (PageHighMem(bh
->b_page
))
49 offset
= (unsigned long)bh
->b_data
;
51 offset
= bh
->b_data
- (char *)page_address(bh
->b_page
);
52 blocknr
+= offset
>> (PAGE_CACHE_SHIFT
- blkbits
);
56 static int check_tree_block(struct btrfs_root
*root
, struct buffer_head
*buf
)
58 struct btrfs_node
*node
= btrfs_buffer_node(buf
);
59 if (bh_blocknr(buf
) != btrfs_header_blocknr(&node
->header
)) {
60 printk(KERN_CRIT
"bh_blocknr(buf) is %Lu, header is %Lu\n",
61 bh_blocknr(buf
), btrfs_header_blocknr(&node
->header
));
67 struct buffer_head
*btrfs_find_tree_block(struct btrfs_root
*root
, u64 blocknr
)
69 struct address_space
*mapping
= root
->fs_info
->btree_inode
->i_mapping
;
70 int blockbits
= root
->fs_info
->sb
->s_blocksize_bits
;
71 unsigned long index
= blocknr
>> (PAGE_CACHE_SHIFT
- blockbits
);
73 struct buffer_head
*bh
;
74 struct buffer_head
*head
;
75 struct buffer_head
*ret
= NULL
;
78 page
= find_lock_page(mapping
, index
);
82 if (!page_has_buffers(page
))
85 head
= page_buffers(page
);
88 if (buffer_mapped(bh
) && bh_blocknr(bh
) == blocknr
) {
100 page_cache_release(page
);
104 int btrfs_map_bh_to_logical(struct btrfs_root
*root
, struct buffer_head
*bh
,
107 struct dev_lookup
*lookup
[2];
114 set_buffer_mapped(bh
);
117 root
= root
->fs_info
->dev_root
;
118 ret
= radix_tree_gang_lookup(&root
->fs_info
->dev_radix
,
120 (unsigned long)logical
,
122 if (ret
== 0 || lookup
[0]->block_start
> logical
||
123 lookup
[0]->block_start
+ lookup
[0]->num_blocks
<= logical
) {
127 bh
->b_bdev
= lookup
[0]->bdev
;
128 bh
->b_blocknr
= logical
- lookup
[0]->block_start
;
129 set_buffer_mapped(bh
);
135 struct buffer_head
*btrfs_find_create_tree_block(struct btrfs_root
*root
,
138 struct address_space
*mapping
= root
->fs_info
->btree_inode
->i_mapping
;
139 int blockbits
= root
->fs_info
->sb
->s_blocksize_bits
;
140 unsigned long index
= blocknr
>> (PAGE_CACHE_SHIFT
- blockbits
);
142 struct buffer_head
*bh
;
143 struct buffer_head
*head
;
144 struct buffer_head
*ret
= NULL
;
146 u64 first_block
= index
<< (PAGE_CACHE_SHIFT
- blockbits
);
148 page
= grab_cache_page(mapping
, index
);
152 if (!page_has_buffers(page
))
153 create_empty_buffers(page
, root
->fs_info
->sb
->s_blocksize
, 0);
154 head
= page_buffers(page
);
157 if (!buffer_mapped(bh
)) {
158 err
= btrfs_map_bh_to_logical(root
, bh
, first_block
);
161 if (bh_blocknr(bh
) == blocknr
) {
166 bh
= bh
->b_this_page
;
168 } while (bh
!= head
);
173 page_cache_release(page
);
177 static int btree_get_block(struct inode
*inode
, sector_t iblock
,
178 struct buffer_head
*bh
, int create
)
181 struct btrfs_root
*root
= BTRFS_I(bh
->b_page
->mapping
->host
)->root
;
182 err
= btrfs_map_bh_to_logical(root
, bh
, iblock
);
186 int btrfs_csum_data(struct btrfs_root
* root
, char *data
, size_t len
,
189 struct scatterlist sg
;
190 struct crypto_hash
*tfm
= root
->fs_info
->hash_tfm
;
191 struct hash_desc desc
;
196 sg_init_one(&sg
, data
, len
);
197 spin_lock(&root
->fs_info
->hash_lock
);
198 ret
= crypto_hash_digest(&desc
, &sg
, 1, result
);
199 spin_unlock(&root
->fs_info
->hash_lock
);
201 printk("sha256 digest failed\n");
205 static int csum_tree_block(struct btrfs_root
*root
, struct buffer_head
*bh
,
208 char result
[BTRFS_CSUM_SIZE
];
210 struct btrfs_node
*node
;
212 ret
= btrfs_csum_data(root
, bh
->b_data
+ BTRFS_CSUM_SIZE
,
213 bh
->b_size
- BTRFS_CSUM_SIZE
, result
);
217 if (memcmp(bh
->b_data
, result
, BTRFS_CSUM_SIZE
)) {
218 printk("checksum verify failed on %Lu\n",
223 node
= btrfs_buffer_node(bh
);
224 memcpy(node
->header
.csum
, result
, BTRFS_CSUM_SIZE
);
229 static int btree_writepage(struct page
*page
, struct writeback_control
*wbc
)
231 struct buffer_head
*bh
;
232 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
233 struct buffer_head
*head
;
234 if (!page_has_buffers(page
)) {
235 create_empty_buffers(page
, root
->fs_info
->sb
->s_blocksize
,
236 (1 << BH_Dirty
)|(1 << BH_Uptodate
));
238 head
= page_buffers(page
);
241 if (buffer_dirty(bh
))
242 csum_tree_block(root
, bh
, 0);
243 bh
= bh
->b_this_page
;
244 } while (bh
!= head
);
245 return block_write_full_page(page
, btree_get_block
, wbc
);
248 static int btree_readpage(struct file
* file
, struct page
* page
)
250 return block_read_full_page(page
, btree_get_block
);
253 static struct address_space_operations btree_aops
= {
254 .readpage
= btree_readpage
,
255 .writepage
= btree_writepage
,
256 .sync_page
= block_sync_page
,
259 struct buffer_head
*read_tree_block(struct btrfs_root
*root
, u64 blocknr
)
261 struct buffer_head
*bh
= NULL
;
263 bh
= btrfs_find_create_tree_block(root
, blocknr
);
266 if (buffer_uptodate(bh
))
269 if (!buffer_uptodate(bh
)) {
271 bh
->b_end_io
= end_buffer_read_sync
;
274 if (!buffer_uptodate(bh
))
276 csum_tree_block(root
, bh
, 1);
281 if (check_tree_block(root
, bh
))
289 int dirty_tree_block(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
290 struct buffer_head
*buf
)
292 WARN_ON(atomic_read(&buf
->b_count
) == 0);
293 mark_buffer_dirty(buf
);
297 int clean_tree_block(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
298 struct buffer_head
*buf
)
300 WARN_ON(atomic_read(&buf
->b_count
) == 0);
301 clear_buffer_dirty(buf
);
305 static int __setup_root(int blocksize
,
306 struct btrfs_root
*root
,
307 struct btrfs_fs_info
*fs_info
,
312 root
->commit_root
= NULL
;
313 root
->blocksize
= blocksize
;
315 root
->fs_info
= fs_info
;
316 root
->objectid
= objectid
;
317 root
->last_trans
= 0;
318 root
->highest_inode
= 0;
319 root
->last_inode_alloc
= 0;
320 memset(&root
->root_key
, 0, sizeof(root
->root_key
));
321 memset(&root
->root_item
, 0, sizeof(root
->root_item
));
322 root
->root_key
.objectid
= objectid
;
326 static int find_and_setup_root(int blocksize
,
327 struct btrfs_root
*tree_root
,
328 struct btrfs_fs_info
*fs_info
,
330 struct btrfs_root
*root
)
334 __setup_root(blocksize
, root
, fs_info
, objectid
);
335 ret
= btrfs_find_last_root(tree_root
, objectid
,
336 &root
->root_item
, &root
->root_key
);
339 root
->node
= read_tree_block(root
,
340 btrfs_root_blocknr(&root
->root_item
));
345 struct btrfs_root
*btrfs_read_fs_root(struct btrfs_fs_info
*fs_info
,
346 struct btrfs_key
*location
)
348 struct btrfs_root
*root
;
349 struct btrfs_root
*tree_root
= fs_info
->tree_root
;
350 struct btrfs_path
*path
;
351 struct btrfs_leaf
*l
;
355 printk("read_fs_root looking for %Lu %Lu %u\n", location
->objectid
, location
->offset
, location
->flags
);
356 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
357 (unsigned long)location
->objectid
);
359 printk("found %p in cache\n", root
);
362 root
= kmalloc(sizeof(*root
), GFP_NOFS
);
365 return ERR_PTR(-ENOMEM
);
367 if (location
->offset
== (u64
)-1) {
368 ret
= find_and_setup_root(fs_info
->sb
->s_blocksize
,
369 fs_info
->tree_root
, fs_info
,
370 location
->objectid
, root
);
379 __setup_root(fs_info
->sb
->s_blocksize
, root
, fs_info
,
382 path
= btrfs_alloc_path();
384 ret
= btrfs_search_slot(NULL
, tree_root
, location
, path
, 0, 0);
386 printk("internal search_slot gives us %d\n", ret
);
391 l
= btrfs_buffer_leaf(path
->nodes
[0]);
392 memcpy(&root
->root_item
,
393 btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_root_item
),
394 sizeof(root
->root_item
));
395 memcpy(&root
->root_key
, location
, sizeof(*location
));
398 btrfs_release_path(root
, path
);
399 btrfs_free_path(path
);
404 root
->node
= read_tree_block(root
,
405 btrfs_root_blocknr(&root
->root_item
));
408 printk("inserting %p\n", root
);
410 ret
= radix_tree_insert(&fs_info
->fs_roots_radix
,
411 (unsigned long)root
->root_key
.objectid
,
414 printk("radix_tree_insert gives us %d\n", ret
);
419 ret
= btrfs_find_highest_inode(root
, &highest_inode
);
421 root
->highest_inode
= highest_inode
;
422 root
->last_inode_alloc
= highest_inode
;
423 printk("highest inode is %Lu\n", highest_inode
);
425 printk("all worked\n");
429 static int btrfs_open_disk(struct btrfs_root
*root
, u64 device_id
,
430 u64 block_start
, u64 num_blocks
,
431 char *filename
, int name_len
)
434 struct block_device
*bdev
;
437 null_filename
= kmalloc(name_len
+ 1, GFP_NOFS
);
440 memcpy(null_filename
, filename
, name_len
);
441 null_filename
[name_len
] = '\0';
443 bdev
= open_bdev_excl(null_filename
, O_RDWR
, root
->fs_info
->sb
);
448 set_blocksize(bdev
, root
->fs_info
->sb
->s_blocksize
);
449 ret
= btrfs_insert_dev_radix(root
, bdev
, device_id
,
450 block_start
, num_blocks
);
454 kfree(null_filename
);
458 static int read_device_info(struct btrfs_root
*root
)
460 struct btrfs_path
*path
;
462 struct btrfs_key key
;
463 struct btrfs_leaf
*leaf
;
464 struct btrfs_device_item
*dev_item
;
468 root
= root
->fs_info
->dev_root
;
470 path
= btrfs_alloc_path();
476 btrfs_set_key_type(&key
, BTRFS_DEV_ITEM_KEY
);
478 mutex_lock(&root
->fs_info
->fs_mutex
);
479 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
480 leaf
= btrfs_buffer_leaf(path
->nodes
[0]);
481 nritems
= btrfs_header_nritems(&leaf
->header
);
483 slot
= path
->slots
[0];
484 if (slot
>= nritems
) {
485 ret
= btrfs_next_leaf(root
, path
);
488 leaf
= btrfs_buffer_leaf(path
->nodes
[0]);
489 nritems
= btrfs_header_nritems(&leaf
->header
);
490 slot
= path
->slots
[0];
492 btrfs_disk_key_to_cpu(&key
, &leaf
->items
[slot
].key
);
493 if (btrfs_key_type(&key
) != BTRFS_DEV_ITEM_KEY
) {
497 dev_item
= btrfs_item_ptr(leaf
, slot
, struct btrfs_device_item
);
498 printk("found key %Lu %Lu\n", key
.objectid
, key
.offset
);
499 if (btrfs_device_id(dev_item
) !=
500 btrfs_super_device_id(root
->fs_info
->disk_super
)) {
501 ret
= btrfs_open_disk(root
, btrfs_device_id(dev_item
),
502 key
.objectid
, key
.offset
,
503 (char *)(dev_item
+ 1),
504 btrfs_device_pathlen(dev_item
));
509 btrfs_free_path(path
);
510 mutex_unlock(&root
->fs_info
->fs_mutex
);
514 struct btrfs_root
*open_ctree(struct super_block
*sb
)
516 struct btrfs_root
*extent_root
= kmalloc(sizeof(struct btrfs_root
),
518 struct btrfs_root
*dev_root
= kmalloc(sizeof(struct btrfs_root
),
520 struct btrfs_root
*tree_root
= kmalloc(sizeof(struct btrfs_root
),
522 struct btrfs_fs_info
*fs_info
= kmalloc(sizeof(*fs_info
),
525 struct btrfs_super_block
*disk_super
;
526 struct dev_lookup
*dev_lookup
;
528 init_bit_radix(&fs_info
->pinned_radix
);
529 init_bit_radix(&fs_info
->pending_del_radix
);
530 INIT_RADIX_TREE(&fs_info
->fs_roots_radix
, GFP_NOFS
);
531 INIT_RADIX_TREE(&fs_info
->dev_radix
, GFP_NOFS
);
532 INIT_LIST_HEAD(&fs_info
->trans_list
);
533 sb_set_blocksize(sb
, 4096);
534 fs_info
->running_transaction
= NULL
;
535 fs_info
->tree_root
= tree_root
;
536 fs_info
->extent_root
= extent_root
;
537 fs_info
->dev_root
= dev_root
;
539 fs_info
->btree_inode
= new_inode(sb
);
540 fs_info
->btree_inode
->i_ino
= 1;
541 fs_info
->btree_inode
->i_nlink
= 1;
542 fs_info
->btree_inode
->i_size
= sb
->s_bdev
->bd_inode
->i_size
;
543 fs_info
->btree_inode
->i_mapping
->a_ops
= &btree_aops
;
544 fs_info
->do_barriers
= 1;
545 BTRFS_I(fs_info
->btree_inode
)->root
= tree_root
;
546 memset(&BTRFS_I(fs_info
->btree_inode
)->location
, 0,
547 sizeof(struct btrfs_key
));
548 insert_inode_hash(fs_info
->btree_inode
);
549 mapping_set_gfp_mask(fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
550 fs_info
->hash_tfm
= crypto_alloc_hash("sha256", 0, CRYPTO_ALG_ASYNC
);
551 spin_lock_init(&fs_info
->hash_lock
);
552 if (!fs_info
->hash_tfm
|| IS_ERR(fs_info
->hash_tfm
)) {
553 printk("failed to allocate sha256 hash\n");
556 mutex_init(&fs_info
->trans_mutex
);
557 mutex_init(&fs_info
->fs_mutex
);
558 memset(&fs_info
->current_insert
, 0, sizeof(fs_info
->current_insert
));
559 memset(&fs_info
->last_insert
, 0, sizeof(fs_info
->last_insert
));
561 __setup_root(sb
->s_blocksize
, dev_root
,
562 fs_info
, BTRFS_DEV_TREE_OBJECTID
);
564 __setup_root(sb
->s_blocksize
, tree_root
,
565 fs_info
, BTRFS_ROOT_TREE_OBJECTID
);
567 dev_lookup
= kmalloc(sizeof(*dev_lookup
), GFP_NOFS
);
568 dev_lookup
->block_start
= 0;
569 dev_lookup
->num_blocks
= (u32
)-2;
570 dev_lookup
->bdev
= sb
->s_bdev
;
571 dev_lookup
->device_id
= 0;
572 ret
= radix_tree_insert(&fs_info
->dev_radix
, (u32
)-2, dev_lookup
);
574 fs_info
->sb_buffer
= read_tree_block(tree_root
,
575 BTRFS_SUPER_INFO_OFFSET
/
578 if (!fs_info
->sb_buffer
)
580 disk_super
= (struct btrfs_super_block
*)fs_info
->sb_buffer
->b_data
;
581 if (!btrfs_super_root(disk_super
))
584 i_size_write(fs_info
->btree_inode
,
585 btrfs_super_total_blocks(disk_super
) <<
586 fs_info
->btree_inode
->i_blkbits
);
588 radix_tree_delete(&fs_info
->dev_radix
, (u32
)-2);
589 dev_lookup
->block_start
= btrfs_super_device_block_start(disk_super
);
590 dev_lookup
->num_blocks
= btrfs_super_device_num_blocks(disk_super
);
591 dev_lookup
->device_id
= btrfs_super_device_id(disk_super
);
593 ret
= radix_tree_insert(&fs_info
->dev_radix
,
594 dev_lookup
->block_start
+
595 dev_lookup
->num_blocks
- 1, dev_lookup
);
598 fs_info
->disk_super
= disk_super
;
600 dev_root
->node
= read_tree_block(tree_root
,
601 btrfs_super_device_root(disk_super
));
603 ret
= read_device_info(dev_root
);
606 tree_root
->node
= read_tree_block(tree_root
,
607 btrfs_super_root(disk_super
));
608 BUG_ON(!tree_root
->node
);
610 mutex_lock(&fs_info
->fs_mutex
);
611 ret
= find_and_setup_root(sb
->s_blocksize
, tree_root
, fs_info
,
612 BTRFS_EXTENT_TREE_OBJECTID
, extent_root
);
615 fs_info
->generation
= btrfs_super_generation(disk_super
) + 1;
616 memset(&fs_info
->kobj
, 0, sizeof(fs_info
->kobj
));
617 kobj_set_kset_s(fs_info
, btrfs_subsys
);
618 kobject_set_name(&fs_info
->kobj
, "%s", sb
->s_id
);
619 kobject_register(&fs_info
->kobj
);
620 mutex_unlock(&fs_info
->fs_mutex
);
624 int write_ctree_super(struct btrfs_trans_handle
*trans
, struct btrfs_root
628 struct buffer_head
*bh
= root
->fs_info
->sb_buffer
;
630 btrfs_set_super_root(root
->fs_info
->disk_super
,
631 bh_blocknr(root
->fs_info
->tree_root
->node
));
633 WARN_ON(atomic_read(&bh
->b_count
) < 1);
634 clear_buffer_dirty(bh
);
635 csum_tree_block(root
, bh
, 0);
636 bh
->b_end_io
= end_buffer_write_sync
;
638 if (root
->fs_info
->do_barriers
)
639 ret
= submit_bh(WRITE_BARRIER
, bh
);
641 ret
= submit_bh(WRITE
, bh
);
642 if (ret
== -EOPNOTSUPP
) {
643 set_buffer_uptodate(bh
);
644 root
->fs_info
->do_barriers
= 0;
645 ret
= submit_bh(WRITE
, bh
);
648 if (!buffer_uptodate(bh
)) {
655 static int free_fs_root(struct btrfs_fs_info
*fs_info
, struct btrfs_root
*root
)
657 radix_tree_delete(&fs_info
->fs_roots_radix
,
658 (unsigned long)root
->root_key
.objectid
);
663 if (root
->commit_root
)
664 brelse(root
->commit_root
);
669 int del_fs_roots(struct btrfs_fs_info
*fs_info
)
672 struct btrfs_root
*gang
[8];
676 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
681 for (i
= 0; i
< ret
; i
++)
682 free_fs_root(fs_info
, gang
[i
]);
687 static int free_dev_radix(struct btrfs_fs_info
*fs_info
)
689 struct dev_lookup
*lookup
[8];
690 struct block_device
*super_bdev
= fs_info
->sb
->s_bdev
;
694 ret
= radix_tree_gang_lookup(&fs_info
->dev_radix
,
699 for (i
= 0; i
< ret
; i
++) {
700 if (lookup
[i
]->bdev
!= super_bdev
)
701 close_bdev_excl(lookup
[i
]->bdev
);
702 radix_tree_delete(&fs_info
->dev_radix
,
703 lookup
[i
]->block_start
+
704 lookup
[i
]->num_blocks
- 1);
711 int close_ctree(struct btrfs_root
*root
)
714 struct btrfs_trans_handle
*trans
;
715 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
717 mutex_lock(&fs_info
->fs_mutex
);
718 trans
= btrfs_start_transaction(root
, 1);
719 btrfs_commit_transaction(trans
, root
);
720 /* run commit again to drop the original snapshot */
721 trans
= btrfs_start_transaction(root
, 1);
722 btrfs_commit_transaction(trans
, root
);
723 ret
= btrfs_write_and_wait_transaction(NULL
, root
);
725 write_ctree_super(NULL
, root
);
726 mutex_unlock(&fs_info
->fs_mutex
);
728 if (fs_info
->extent_root
->node
)
729 btrfs_block_release(fs_info
->extent_root
,
730 fs_info
->extent_root
->node
);
731 if (fs_info
->dev_root
->node
)
732 btrfs_block_release(fs_info
->dev_root
,
733 fs_info
->dev_root
->node
);
734 if (fs_info
->tree_root
->node
)
735 btrfs_block_release(fs_info
->tree_root
,
736 fs_info
->tree_root
->node
);
737 btrfs_block_release(root
, fs_info
->sb_buffer
);
738 crypto_free_hash(fs_info
->hash_tfm
);
739 truncate_inode_pages(fs_info
->btree_inode
->i_mapping
, 0);
740 iput(fs_info
->btree_inode
);
742 free_dev_radix(fs_info
);
743 del_fs_roots(fs_info
);
744 kfree(fs_info
->extent_root
);
745 kfree(fs_info
->tree_root
);
746 kobject_unregister(&fs_info
->kobj
);
750 void btrfs_block_release(struct btrfs_root
*root
, struct buffer_head
*buf
)