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
));
325 static int find_and_setup_root(int blocksize
,
326 struct btrfs_root
*tree_root
,
327 struct btrfs_fs_info
*fs_info
,
329 struct btrfs_root
*root
)
333 __setup_root(blocksize
, root
, fs_info
, objectid
);
334 ret
= btrfs_find_last_root(tree_root
, objectid
,
335 &root
->root_item
, &root
->root_key
);
338 root
->node
= read_tree_block(root
,
339 btrfs_root_blocknr(&root
->root_item
));
344 struct btrfs_root
*btrfs_read_fs_root(struct btrfs_fs_info
*fs_info
,
345 struct btrfs_key
*location
)
347 struct btrfs_root
*root
;
348 struct btrfs_root
*tree_root
= fs_info
->tree_root
;
349 struct btrfs_path
*path
;
350 struct btrfs_leaf
*l
;
354 printk("read_fs_root looking for %Lu %Lu %u\n", location
->objectid
, location
->offset
, location
->flags
);
355 root
= radix_tree_lookup(&fs_info
->fs_roots_radix
,
356 (unsigned long)location
->objectid
);
358 printk("found %p in cache\n", root
);
361 root
= kmalloc(sizeof(*root
), GFP_NOFS
);
364 return ERR_PTR(-ENOMEM
);
366 if (location
->offset
== (u64
)-1) {
367 ret
= find_and_setup_root(fs_info
->sb
->s_blocksize
,
368 fs_info
->tree_root
, fs_info
,
369 location
->objectid
, root
);
378 __setup_root(fs_info
->sb
->s_blocksize
, root
, fs_info
,
381 path
= btrfs_alloc_path();
383 ret
= btrfs_search_slot(NULL
, tree_root
, location
, path
, 0, 0);
385 printk("internal search_slot gives us %d\n", ret
);
390 l
= btrfs_buffer_leaf(path
->nodes
[0]);
391 memcpy(&root
->root_item
,
392 btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_root_item
),
393 sizeof(root
->root_item
));
394 memcpy(&root
->root_key
, location
, sizeof(*location
));
397 btrfs_release_path(root
, path
);
398 btrfs_free_path(path
);
403 root
->node
= read_tree_block(root
,
404 btrfs_root_blocknr(&root
->root_item
));
407 printk("inserting %p\n", root
);
409 ret
= radix_tree_insert(&fs_info
->fs_roots_radix
,
410 (unsigned long)root
->root_key
.objectid
,
413 printk("radix_tree_insert gives us %d\n", ret
);
418 ret
= btrfs_find_highest_inode(root
, &highest_inode
);
420 root
->highest_inode
= highest_inode
;
421 root
->last_inode_alloc
= highest_inode
;
422 printk("highest inode is %Lu\n", highest_inode
);
424 printk("all worked\n");
428 static int btrfs_open_disk(struct btrfs_root
*root
, u64 device_id
,
429 u64 block_start
, u64 num_blocks
,
430 char *filename
, int name_len
)
433 struct block_device
*bdev
;
436 null_filename
= kmalloc(name_len
+ 1, GFP_NOFS
);
439 memcpy(null_filename
, filename
, name_len
);
440 null_filename
[name_len
] = '\0';
442 bdev
= open_bdev_excl(null_filename
, O_RDWR
, root
->fs_info
->sb
);
447 set_blocksize(bdev
, root
->fs_info
->sb
->s_blocksize
);
448 ret
= btrfs_insert_dev_radix(root
, bdev
, device_id
,
449 block_start
, num_blocks
);
453 kfree(null_filename
);
457 static int read_device_info(struct btrfs_root
*root
)
459 struct btrfs_path
*path
;
461 struct btrfs_key key
;
462 struct btrfs_leaf
*leaf
;
463 struct btrfs_device_item
*dev_item
;
467 root
= root
->fs_info
->dev_root
;
469 path
= btrfs_alloc_path();
475 btrfs_set_key_type(&key
, BTRFS_DEV_ITEM_KEY
);
477 mutex_lock(&root
->fs_info
->fs_mutex
);
478 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
479 leaf
= btrfs_buffer_leaf(path
->nodes
[0]);
480 nritems
= btrfs_header_nritems(&leaf
->header
);
482 slot
= path
->slots
[0];
483 if (slot
>= nritems
) {
484 ret
= btrfs_next_leaf(root
, path
);
487 leaf
= btrfs_buffer_leaf(path
->nodes
[0]);
488 nritems
= btrfs_header_nritems(&leaf
->header
);
489 slot
= path
->slots
[0];
491 btrfs_disk_key_to_cpu(&key
, &leaf
->items
[slot
].key
);
492 if (btrfs_key_type(&key
) != BTRFS_DEV_ITEM_KEY
) {
496 dev_item
= btrfs_item_ptr(leaf
, slot
, struct btrfs_device_item
);
497 printk("found key %Lu %Lu\n", key
.objectid
, key
.offset
);
498 if (btrfs_device_id(dev_item
) !=
499 btrfs_super_device_id(root
->fs_info
->disk_super
)) {
500 ret
= btrfs_open_disk(root
, btrfs_device_id(dev_item
),
501 key
.objectid
, key
.offset
,
502 (char *)(dev_item
+ 1),
503 btrfs_device_pathlen(dev_item
));
508 btrfs_free_path(path
);
509 mutex_unlock(&root
->fs_info
->fs_mutex
);
513 struct btrfs_root
*open_ctree(struct super_block
*sb
)
515 struct btrfs_root
*extent_root
= kmalloc(sizeof(struct btrfs_root
),
517 struct btrfs_root
*dev_root
= kmalloc(sizeof(struct btrfs_root
),
519 struct btrfs_root
*tree_root
= kmalloc(sizeof(struct btrfs_root
),
521 struct btrfs_fs_info
*fs_info
= kmalloc(sizeof(*fs_info
),
524 struct btrfs_super_block
*disk_super
;
525 struct dev_lookup
*dev_lookup
;
527 init_bit_radix(&fs_info
->pinned_radix
);
528 init_bit_radix(&fs_info
->pending_del_radix
);
529 INIT_RADIX_TREE(&fs_info
->fs_roots_radix
, GFP_NOFS
);
530 INIT_RADIX_TREE(&fs_info
->dev_radix
, GFP_NOFS
);
531 INIT_LIST_HEAD(&fs_info
->trans_list
);
532 sb_set_blocksize(sb
, 4096);
533 fs_info
->running_transaction
= NULL
;
534 fs_info
->tree_root
= tree_root
;
535 fs_info
->extent_root
= extent_root
;
536 fs_info
->dev_root
= dev_root
;
538 fs_info
->btree_inode
= new_inode(sb
);
539 fs_info
->btree_inode
->i_ino
= 1;
540 fs_info
->btree_inode
->i_nlink
= 1;
541 fs_info
->btree_inode
->i_size
= sb
->s_bdev
->bd_inode
->i_size
;
542 fs_info
->btree_inode
->i_mapping
->a_ops
= &btree_aops
;
543 BTRFS_I(fs_info
->btree_inode
)->root
= tree_root
;
544 memset(&BTRFS_I(fs_info
->btree_inode
)->location
, 0,
545 sizeof(struct btrfs_key
));
546 insert_inode_hash(fs_info
->btree_inode
);
547 mapping_set_gfp_mask(fs_info
->btree_inode
->i_mapping
, GFP_NOFS
);
548 fs_info
->hash_tfm
= crypto_alloc_hash("sha256", 0, CRYPTO_ALG_ASYNC
);
549 spin_lock_init(&fs_info
->hash_lock
);
550 if (!fs_info
->hash_tfm
|| IS_ERR(fs_info
->hash_tfm
)) {
551 printk("failed to allocate sha256 hash\n");
554 mutex_init(&fs_info
->trans_mutex
);
555 mutex_init(&fs_info
->fs_mutex
);
556 memset(&fs_info
->current_insert
, 0, sizeof(fs_info
->current_insert
));
557 memset(&fs_info
->last_insert
, 0, sizeof(fs_info
->last_insert
));
559 __setup_root(sb
->s_blocksize
, dev_root
,
560 fs_info
, BTRFS_DEV_TREE_OBJECTID
);
562 __setup_root(sb
->s_blocksize
, tree_root
,
563 fs_info
, BTRFS_ROOT_TREE_OBJECTID
);
565 dev_lookup
= kmalloc(sizeof(*dev_lookup
), GFP_NOFS
);
566 dev_lookup
->block_start
= 0;
567 dev_lookup
->num_blocks
= (u32
)-2;
568 dev_lookup
->bdev
= sb
->s_bdev
;
569 dev_lookup
->device_id
= 0;
570 ret
= radix_tree_insert(&fs_info
->dev_radix
, (u32
)-2, dev_lookup
);
572 fs_info
->sb_buffer
= read_tree_block(tree_root
,
573 BTRFS_SUPER_INFO_OFFSET
/
576 if (!fs_info
->sb_buffer
)
578 disk_super
= (struct btrfs_super_block
*)fs_info
->sb_buffer
->b_data
;
579 if (!btrfs_super_root(disk_super
))
582 i_size_write(fs_info
->btree_inode
,
583 btrfs_super_total_blocks(disk_super
) <<
584 fs_info
->btree_inode
->i_blkbits
);
586 radix_tree_delete(&fs_info
->dev_radix
, (u32
)-2);
587 dev_lookup
->block_start
= btrfs_super_device_block_start(disk_super
);
588 dev_lookup
->num_blocks
= btrfs_super_device_num_blocks(disk_super
);
589 dev_lookup
->device_id
= btrfs_super_device_id(disk_super
);
591 ret
= radix_tree_insert(&fs_info
->dev_radix
,
592 dev_lookup
->block_start
+
593 dev_lookup
->num_blocks
- 1, dev_lookup
);
596 fs_info
->disk_super
= disk_super
;
598 dev_root
->node
= read_tree_block(tree_root
,
599 btrfs_super_device_root(disk_super
));
601 ret
= read_device_info(dev_root
);
604 tree_root
->node
= read_tree_block(tree_root
,
605 btrfs_super_root(disk_super
));
606 BUG_ON(!tree_root
->node
);
608 mutex_lock(&fs_info
->fs_mutex
);
609 ret
= find_and_setup_root(sb
->s_blocksize
, tree_root
, fs_info
,
610 BTRFS_EXTENT_TREE_OBJECTID
, extent_root
);
613 fs_info
->generation
= btrfs_super_generation(disk_super
) + 1;
614 memset(&fs_info
->kobj
, 0, sizeof(fs_info
->kobj
));
615 kobj_set_kset_s(fs_info
, btrfs_subsys
);
616 kobject_set_name(&fs_info
->kobj
, "%s", sb
->s_id
);
617 kobject_register(&fs_info
->kobj
);
618 mutex_unlock(&fs_info
->fs_mutex
);
622 int write_ctree_super(struct btrfs_trans_handle
*trans
, struct btrfs_root
625 struct buffer_head
*bh
= root
->fs_info
->sb_buffer
;
627 btrfs_set_super_root(root
->fs_info
->disk_super
,
628 bh_blocknr(root
->fs_info
->tree_root
->node
));
630 WARN_ON(atomic_read(&bh
->b_count
) < 1);
631 clear_buffer_dirty(bh
);
632 csum_tree_block(root
, bh
, 0);
633 bh
->b_end_io
= end_buffer_write_sync
;
635 submit_bh(WRITE
, bh
);
637 if (!buffer_uptodate(bh
)) {
644 static int free_fs_root(struct btrfs_fs_info
*fs_info
, struct btrfs_root
*root
)
646 radix_tree_delete(&fs_info
->fs_roots_radix
,
647 (unsigned long)root
->root_key
.objectid
);
652 if (root
->commit_root
)
653 brelse(root
->commit_root
);
658 int del_fs_roots(struct btrfs_fs_info
*fs_info
)
661 struct btrfs_root
*gang
[8];
665 ret
= radix_tree_gang_lookup(&fs_info
->fs_roots_radix
,
670 for (i
= 0; i
< ret
; i
++)
671 free_fs_root(fs_info
, gang
[i
]);
676 static int free_dev_radix(struct btrfs_fs_info
*fs_info
)
678 struct dev_lookup
*lookup
[8];
679 struct block_device
*super_bdev
= fs_info
->sb
->s_bdev
;
683 ret
= radix_tree_gang_lookup(&fs_info
->dev_radix
,
688 for (i
= 0; i
< ret
; i
++) {
689 if (lookup
[i
]->bdev
!= super_bdev
)
690 close_bdev_excl(lookup
[i
]->bdev
);
691 radix_tree_delete(&fs_info
->dev_radix
,
692 lookup
[i
]->block_start
+
693 lookup
[i
]->num_blocks
- 1);
700 int close_ctree(struct btrfs_root
*root
)
703 struct btrfs_trans_handle
*trans
;
704 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
706 mutex_lock(&fs_info
->fs_mutex
);
707 trans
= btrfs_start_transaction(root
, 1);
708 btrfs_commit_transaction(trans
, root
);
709 /* run commit again to drop the original snapshot */
710 trans
= btrfs_start_transaction(root
, 1);
711 btrfs_commit_transaction(trans
, root
);
712 ret
= btrfs_write_and_wait_transaction(NULL
, root
);
714 write_ctree_super(NULL
, root
);
715 mutex_unlock(&fs_info
->fs_mutex
);
717 if (fs_info
->extent_root
->node
)
718 btrfs_block_release(fs_info
->extent_root
,
719 fs_info
->extent_root
->node
);
720 if (fs_info
->dev_root
->node
)
721 btrfs_block_release(fs_info
->dev_root
,
722 fs_info
->dev_root
->node
);
723 if (fs_info
->tree_root
->node
)
724 btrfs_block_release(fs_info
->tree_root
,
725 fs_info
->tree_root
->node
);
726 btrfs_block_release(root
, fs_info
->sb_buffer
);
727 crypto_free_hash(fs_info
->hash_tfm
);
728 truncate_inode_pages(fs_info
->btree_inode
->i_mapping
, 0);
729 iput(fs_info
->btree_inode
);
731 free_dev_radix(fs_info
);
732 del_fs_roots(fs_info
);
733 kfree(fs_info
->extent_root
);
734 kfree(fs_info
->tree_root
);
735 kobject_unregister(&fs_info
->kobj
);
739 void btrfs_block_release(struct btrfs_root
*root
, struct buffer_head
*buf
)