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.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/pagemap.h>
25 #include <linux/highmem.h>
26 #include <linux/time.h>
27 #include <linux/init.h>
28 #include <linux/string.h>
29 #include <linux/smp_lock.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mpage.h>
32 #include <linux/swap.h>
33 #include <linux/writeback.h>
34 #include <linux/statfs.h>
35 #include <linux/compat.h>
36 #include <linux/bit_spinlock.h>
37 #include <linux/version.h>
38 #include <linux/xattr.h>
41 #include "transaction.h"
42 #include "btrfs_inode.h"
44 #include "print-tree.h"
47 struct btrfs_iget_args
{
49 struct btrfs_root
*root
;
52 static struct inode_operations btrfs_dir_inode_operations
;
53 static struct inode_operations btrfs_symlink_inode_operations
;
54 static struct inode_operations btrfs_dir_ro_inode_operations
;
55 static struct inode_operations btrfs_special_inode_operations
;
56 static struct inode_operations btrfs_file_inode_operations
;
57 static struct address_space_operations btrfs_aops
;
58 static struct address_space_operations btrfs_symlink_aops
;
59 static struct file_operations btrfs_dir_file_operations
;
60 static struct extent_io_ops btrfs_extent_io_ops
;
62 static struct kmem_cache
*btrfs_inode_cachep
;
63 struct kmem_cache
*btrfs_trans_handle_cachep
;
64 struct kmem_cache
*btrfs_transaction_cachep
;
65 struct kmem_cache
*btrfs_bit_radix_cachep
;
66 struct kmem_cache
*btrfs_path_cachep
;
69 static unsigned char btrfs_type_by_mode
[S_IFMT
>> S_SHIFT
] = {
70 [S_IFREG
>> S_SHIFT
] = BTRFS_FT_REG_FILE
,
71 [S_IFDIR
>> S_SHIFT
] = BTRFS_FT_DIR
,
72 [S_IFCHR
>> S_SHIFT
] = BTRFS_FT_CHRDEV
,
73 [S_IFBLK
>> S_SHIFT
] = BTRFS_FT_BLKDEV
,
74 [S_IFIFO
>> S_SHIFT
] = BTRFS_FT_FIFO
,
75 [S_IFSOCK
>> S_SHIFT
] = BTRFS_FT_SOCK
,
76 [S_IFLNK
>> S_SHIFT
] = BTRFS_FT_SYMLINK
,
79 int btrfs_check_free_space(struct btrfs_root
*root
, u64 num_required
,
88 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
89 total
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
90 used
= btrfs_super_bytes_used(&root
->fs_info
->super_copy
);
98 if (used
+ root
->fs_info
->delalloc_bytes
+ num_required
> thresh
)
100 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
104 static int cow_file_range(struct inode
*inode
, u64 start
, u64 end
)
106 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
107 struct btrfs_trans_handle
*trans
;
111 u64 blocksize
= root
->sectorsize
;
112 u64 orig_start
= start
;
114 struct btrfs_key ins
;
117 trans
= btrfs_start_transaction(root
, 1);
119 btrfs_set_trans_block_group(trans
, inode
);
121 num_bytes
= (end
- start
+ blocksize
) & ~(blocksize
- 1);
122 num_bytes
= max(blocksize
, num_bytes
);
123 ret
= btrfs_drop_extents(trans
, root
, inode
,
124 start
, start
+ num_bytes
, start
, &alloc_hint
);
125 orig_num_bytes
= num_bytes
;
127 if (alloc_hint
== EXTENT_MAP_INLINE
)
130 BUG_ON(num_bytes
> btrfs_super_total_bytes(&root
->fs_info
->super_copy
));
132 while(num_bytes
> 0) {
133 cur_alloc_size
= min(num_bytes
, root
->fs_info
->max_extent
);
134 ret
= btrfs_alloc_extent(trans
, root
, cur_alloc_size
,
136 root
->root_key
.objectid
,
138 inode
->i_ino
, start
, 0,
139 alloc_hint
, (u64
)-1, &ins
, 1);
144 cur_alloc_size
= ins
.offset
;
145 ret
= btrfs_insert_file_extent(trans
, root
, inode
->i_ino
,
146 start
, ins
.objectid
, ins
.offset
,
148 inode
->i_blocks
+= ins
.offset
>> 9;
149 btrfs_check_file(root
, inode
);
150 if (num_bytes
< cur_alloc_size
) {
151 printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes
,
155 num_bytes
-= cur_alloc_size
;
156 alloc_hint
= ins
.objectid
+ ins
.offset
;
157 start
+= cur_alloc_size
;
159 btrfs_drop_extent_cache(inode
, orig_start
,
160 orig_start
+ orig_num_bytes
- 1);
161 btrfs_add_ordered_inode(inode
);
162 btrfs_update_inode(trans
, root
, inode
);
164 btrfs_end_transaction(trans
, root
);
168 static int run_delalloc_nocow(struct inode
*inode
, u64 start
, u64 end
)
176 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
177 struct btrfs_block_group_cache
*block_group
;
178 struct extent_buffer
*leaf
;
180 struct btrfs_path
*path
;
181 struct btrfs_file_extent_item
*item
;
184 struct btrfs_key found_key
;
186 total_fs_bytes
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
187 path
= btrfs_alloc_path();
190 ret
= btrfs_lookup_file_extent(NULL
, root
, path
,
191 inode
->i_ino
, start
, 0);
193 btrfs_free_path(path
);
199 if (path
->slots
[0] == 0)
204 leaf
= path
->nodes
[0];
205 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
206 struct btrfs_file_extent_item
);
208 /* are we inside the extent that was found? */
209 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
210 found_type
= btrfs_key_type(&found_key
);
211 if (found_key
.objectid
!= inode
->i_ino
||
212 found_type
!= BTRFS_EXTENT_DATA_KEY
)
215 found_type
= btrfs_file_extent_type(leaf
, item
);
216 extent_start
= found_key
.offset
;
217 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
218 u64 extent_num_bytes
;
220 extent_num_bytes
= btrfs_file_extent_num_bytes(leaf
, item
);
221 extent_end
= extent_start
+ extent_num_bytes
;
224 if (loops
&& start
!= extent_start
)
227 if (start
< extent_start
|| start
>= extent_end
)
230 cow_end
= min(end
, extent_end
- 1);
231 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
235 if (btrfs_count_snapshots_in_path(root
, path
, inode
->i_ino
,
241 * we may be called by the resizer, make sure we're inside
242 * the limits of the FS
244 block_group
= btrfs_lookup_block_group(root
->fs_info
,
246 if (!block_group
|| block_group
->ro
)
255 btrfs_free_path(path
);
258 btrfs_release_path(root
, path
);
263 cow_file_range(inode
, start
, end
);
268 static int run_delalloc_range(struct inode
*inode
, u64 start
, u64 end
)
270 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
273 if (btrfs_test_opt(root
, NODATACOW
) ||
274 btrfs_test_flag(inode
, NODATACOW
))
275 ret
= run_delalloc_nocow(inode
, start
, end
);
277 ret
= cow_file_range(inode
, start
, end
);
282 int btrfs_set_bit_hook(struct inode
*inode
, u64 start
, u64 end
,
283 unsigned long old
, unsigned long bits
)
286 if (!(old
& EXTENT_DELALLOC
) && (bits
& EXTENT_DELALLOC
)) {
287 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
288 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
289 BTRFS_I(inode
)->delalloc_bytes
+= end
- start
+ 1;
290 root
->fs_info
->delalloc_bytes
+= end
- start
+ 1;
291 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
296 int btrfs_clear_bit_hook(struct inode
*inode
, u64 start
, u64 end
,
297 unsigned long old
, unsigned long bits
)
299 if ((old
& EXTENT_DELALLOC
) && (bits
& EXTENT_DELALLOC
)) {
300 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
303 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
304 if (end
- start
+ 1 > root
->fs_info
->delalloc_bytes
) {
305 printk("warning: delalloc account %Lu %Lu\n",
306 end
- start
+ 1, root
->fs_info
->delalloc_bytes
);
307 root
->fs_info
->delalloc_bytes
= 0;
308 BTRFS_I(inode
)->delalloc_bytes
= 0;
310 root
->fs_info
->delalloc_bytes
-= end
- start
+ 1;
311 BTRFS_I(inode
)->delalloc_bytes
-= end
- start
+ 1;
313 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
318 int btrfs_merge_bio_hook(struct page
*page
, unsigned long offset
,
319 size_t size
, struct bio
*bio
)
321 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
322 struct btrfs_mapping_tree
*map_tree
;
323 u64 logical
= bio
->bi_sector
<< 9;
328 length
= bio
->bi_size
;
329 map_tree
= &root
->fs_info
->mapping_tree
;
331 ret
= btrfs_map_block(map_tree
, READ
, logical
,
332 &map_length
, NULL
, 0);
334 if (map_length
< length
+ size
) {
340 int __btrfs_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
343 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
344 struct btrfs_trans_handle
*trans
;
348 ret
= btrfs_csum_one_bio(root
, bio
, &sums
);
351 trans
= btrfs_start_transaction(root
, 1);
353 btrfs_set_trans_block_group(trans
, inode
);
354 btrfs_csum_file_blocks(trans
, root
, inode
, bio
, sums
);
356 ret
= btrfs_end_transaction(trans
, root
);
361 return btrfs_map_bio(root
, rw
, bio
, mirror_num
, 1);
364 int btrfs_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
367 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
370 if (!(rw
& (1 << BIO_RW
))) {
371 ret
= btrfs_bio_wq_end_io(root
->fs_info
, bio
, 0);
376 if (btrfs_test_opt(root
, NODATASUM
) ||
377 btrfs_test_flag(inode
, NODATASUM
)) {
381 return btrfs_wq_submit_bio(BTRFS_I(inode
)->root
->fs_info
,
382 inode
, rw
, bio
, mirror_num
,
383 __btrfs_submit_bio_hook
);
385 return btrfs_map_bio(root
, rw
, bio
, mirror_num
, 0);
388 int btrfs_readpage_io_hook(struct page
*page
, u64 start
, u64 end
)
391 struct inode
*inode
= page
->mapping
->host
;
392 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
393 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
394 struct btrfs_csum_item
*item
;
395 struct btrfs_path
*path
= NULL
;
398 if (btrfs_test_opt(root
, NODATASUM
) ||
399 btrfs_test_flag(inode
, NODATASUM
))
402 path
= btrfs_alloc_path();
403 item
= btrfs_lookup_csum(NULL
, root
, path
, inode
->i_ino
, start
, 0);
406 /* a csum that isn't present is a preallocated region. */
407 if (ret
== -ENOENT
|| ret
== -EFBIG
)
410 printk("no csum found for inode %lu start %Lu\n", inode
->i_ino
, start
);
413 read_extent_buffer(path
->nodes
[0], &csum
, (unsigned long)item
,
415 set_state_private(io_tree
, start
, csum
);
418 btrfs_free_path(path
);
422 struct io_failure_record
{
430 int btrfs_io_failed_hook(struct bio
*failed_bio
,
431 struct page
*page
, u64 start
, u64 end
,
432 struct extent_state
*state
)
434 struct io_failure_record
*failrec
= NULL
;
436 struct extent_map
*em
;
437 struct inode
*inode
= page
->mapping
->host
;
438 struct extent_io_tree
*failure_tree
= &BTRFS_I(inode
)->io_failure_tree
;
439 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
446 ret
= get_state_private(failure_tree
, start
, &private);
448 failrec
= kmalloc(sizeof(*failrec
), GFP_NOFS
);
451 failrec
->start
= start
;
452 failrec
->len
= end
- start
+ 1;
453 failrec
->last_mirror
= 0;
455 spin_lock(&em_tree
->lock
);
456 em
= lookup_extent_mapping(em_tree
, start
, failrec
->len
);
457 if (em
->start
> start
|| em
->start
+ em
->len
< start
) {
461 spin_unlock(&em_tree
->lock
);
463 if (!em
|| IS_ERR(em
)) {
467 logical
= start
- em
->start
;
468 logical
= em
->block_start
+ logical
;
469 failrec
->logical
= logical
;
471 set_extent_bits(failure_tree
, start
, end
, EXTENT_LOCKED
|
472 EXTENT_DIRTY
, GFP_NOFS
);
473 set_state_private(failure_tree
, start
,
474 (u64
)(unsigned long)failrec
);
476 failrec
= (struct io_failure_record
*)(unsigned long)private;
478 num_copies
= btrfs_num_copies(
479 &BTRFS_I(inode
)->root
->fs_info
->mapping_tree
,
480 failrec
->logical
, failrec
->len
);
481 failrec
->last_mirror
++;
483 spin_lock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
484 state
= find_first_extent_bit_state(&BTRFS_I(inode
)->io_tree
,
487 if (state
&& state
->start
!= failrec
->start
)
489 spin_unlock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
491 if (!state
|| failrec
->last_mirror
> num_copies
) {
492 set_state_private(failure_tree
, failrec
->start
, 0);
493 clear_extent_bits(failure_tree
, failrec
->start
,
494 failrec
->start
+ failrec
->len
- 1,
495 EXTENT_LOCKED
| EXTENT_DIRTY
, GFP_NOFS
);
499 bio
= bio_alloc(GFP_NOFS
, 1);
500 bio
->bi_private
= state
;
501 bio
->bi_end_io
= failed_bio
->bi_end_io
;
502 bio
->bi_sector
= failrec
->logical
>> 9;
503 bio
->bi_bdev
= failed_bio
->bi_bdev
;
505 bio_add_page(bio
, page
, failrec
->len
, start
- page_offset(page
));
506 if (failed_bio
->bi_rw
& (1 << BIO_RW
))
511 BTRFS_I(inode
)->io_tree
.ops
->submit_bio_hook(inode
, rw
, bio
,
512 failrec
->last_mirror
);
516 int btrfs_clean_io_failures(struct inode
*inode
, u64 start
)
520 struct io_failure_record
*failure
;
524 if (count_range_bits(&BTRFS_I(inode
)->io_failure_tree
, &private,
525 (u64
)-1, 1, EXTENT_DIRTY
)) {
526 ret
= get_state_private(&BTRFS_I(inode
)->io_failure_tree
,
527 start
, &private_failure
);
529 failure
= (struct io_failure_record
*)(unsigned long)
531 set_state_private(&BTRFS_I(inode
)->io_failure_tree
,
533 clear_extent_bits(&BTRFS_I(inode
)->io_failure_tree
,
535 failure
->start
+ failure
->len
- 1,
536 EXTENT_DIRTY
| EXTENT_LOCKED
,
544 int btrfs_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
545 struct extent_state
*state
)
547 size_t offset
= start
- ((u64
)page
->index
<< PAGE_CACHE_SHIFT
);
548 struct inode
*inode
= page
->mapping
->host
;
549 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
551 u64
private = ~(u32
)0;
553 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
557 if (btrfs_test_opt(root
, NODATASUM
) ||
558 btrfs_test_flag(inode
, NODATASUM
))
560 if (state
&& state
->start
== start
) {
561 private = state
->private;
564 ret
= get_state_private(io_tree
, start
, &private);
566 local_irq_save(flags
);
567 kaddr
= kmap_atomic(page
, KM_IRQ0
);
571 csum
= btrfs_csum_data(root
, kaddr
+ offset
, csum
, end
- start
+ 1);
572 btrfs_csum_final(csum
, (char *)&csum
);
573 if (csum
!= private) {
576 kunmap_atomic(kaddr
, KM_IRQ0
);
577 local_irq_restore(flags
);
579 /* if the io failure tree for this inode is non-empty,
580 * check to see if we've recovered from a failed IO
582 btrfs_clean_io_failures(inode
, start
);
586 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
587 page
->mapping
->host
->i_ino
, (unsigned long long)start
, csum
,
589 memset(kaddr
+ offset
, 1, end
- start
+ 1);
590 flush_dcache_page(page
);
591 kunmap_atomic(kaddr
, KM_IRQ0
);
592 local_irq_restore(flags
);
598 void btrfs_read_locked_inode(struct inode
*inode
)
600 struct btrfs_path
*path
;
601 struct extent_buffer
*leaf
;
602 struct btrfs_inode_item
*inode_item
;
603 struct btrfs_timespec
*tspec
;
604 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
605 struct btrfs_key location
;
606 u64 alloc_group_block
;
610 path
= btrfs_alloc_path();
612 memcpy(&location
, &BTRFS_I(inode
)->location
, sizeof(location
));
614 ret
= btrfs_lookup_inode(NULL
, root
, path
, &location
, 0);
618 leaf
= path
->nodes
[0];
619 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
620 struct btrfs_inode_item
);
622 inode
->i_mode
= btrfs_inode_mode(leaf
, inode_item
);
623 inode
->i_nlink
= btrfs_inode_nlink(leaf
, inode_item
);
624 inode
->i_uid
= btrfs_inode_uid(leaf
, inode_item
);
625 inode
->i_gid
= btrfs_inode_gid(leaf
, inode_item
);
626 inode
->i_size
= btrfs_inode_size(leaf
, inode_item
);
628 tspec
= btrfs_inode_atime(inode_item
);
629 inode
->i_atime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
630 inode
->i_atime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
632 tspec
= btrfs_inode_mtime(inode_item
);
633 inode
->i_mtime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
634 inode
->i_mtime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
636 tspec
= btrfs_inode_ctime(inode_item
);
637 inode
->i_ctime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
638 inode
->i_ctime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
640 inode
->i_blocks
= btrfs_inode_nblocks(leaf
, inode_item
);
641 inode
->i_generation
= btrfs_inode_generation(leaf
, inode_item
);
643 rdev
= btrfs_inode_rdev(leaf
, inode_item
);
645 alloc_group_block
= btrfs_inode_block_group(leaf
, inode_item
);
646 BTRFS_I(inode
)->block_group
= btrfs_lookup_block_group(root
->fs_info
,
648 BTRFS_I(inode
)->flags
= btrfs_inode_flags(leaf
, inode_item
);
649 if (!BTRFS_I(inode
)->block_group
) {
650 BTRFS_I(inode
)->block_group
= btrfs_find_block_group(root
,
652 BTRFS_BLOCK_GROUP_METADATA
, 0);
654 btrfs_free_path(path
);
657 switch (inode
->i_mode
& S_IFMT
) {
659 inode
->i_mapping
->a_ops
= &btrfs_aops
;
660 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
661 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
662 inode
->i_fop
= &btrfs_file_operations
;
663 inode
->i_op
= &btrfs_file_inode_operations
;
666 inode
->i_fop
= &btrfs_dir_file_operations
;
667 if (root
== root
->fs_info
->tree_root
)
668 inode
->i_op
= &btrfs_dir_ro_inode_operations
;
670 inode
->i_op
= &btrfs_dir_inode_operations
;
673 inode
->i_op
= &btrfs_symlink_inode_operations
;
674 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
675 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
678 init_special_inode(inode
, inode
->i_mode
, rdev
);
684 btrfs_free_path(path
);
685 make_bad_inode(inode
);
688 static void fill_inode_item(struct extent_buffer
*leaf
,
689 struct btrfs_inode_item
*item
,
692 btrfs_set_inode_uid(leaf
, item
, inode
->i_uid
);
693 btrfs_set_inode_gid(leaf
, item
, inode
->i_gid
);
694 btrfs_set_inode_size(leaf
, item
, inode
->i_size
);
695 btrfs_set_inode_mode(leaf
, item
, inode
->i_mode
);
696 btrfs_set_inode_nlink(leaf
, item
, inode
->i_nlink
);
698 btrfs_set_timespec_sec(leaf
, btrfs_inode_atime(item
),
699 inode
->i_atime
.tv_sec
);
700 btrfs_set_timespec_nsec(leaf
, btrfs_inode_atime(item
),
701 inode
->i_atime
.tv_nsec
);
703 btrfs_set_timespec_sec(leaf
, btrfs_inode_mtime(item
),
704 inode
->i_mtime
.tv_sec
);
705 btrfs_set_timespec_nsec(leaf
, btrfs_inode_mtime(item
),
706 inode
->i_mtime
.tv_nsec
);
708 btrfs_set_timespec_sec(leaf
, btrfs_inode_ctime(item
),
709 inode
->i_ctime
.tv_sec
);
710 btrfs_set_timespec_nsec(leaf
, btrfs_inode_ctime(item
),
711 inode
->i_ctime
.tv_nsec
);
713 btrfs_set_inode_nblocks(leaf
, item
, inode
->i_blocks
);
714 btrfs_set_inode_generation(leaf
, item
, inode
->i_generation
);
715 btrfs_set_inode_rdev(leaf
, item
, inode
->i_rdev
);
716 btrfs_set_inode_flags(leaf
, item
, BTRFS_I(inode
)->flags
);
717 btrfs_set_inode_block_group(leaf
, item
,
718 BTRFS_I(inode
)->block_group
->key
.objectid
);
721 int btrfs_update_inode(struct btrfs_trans_handle
*trans
,
722 struct btrfs_root
*root
,
725 struct btrfs_inode_item
*inode_item
;
726 struct btrfs_path
*path
;
727 struct extent_buffer
*leaf
;
730 path
= btrfs_alloc_path();
732 ret
= btrfs_lookup_inode(trans
, root
, path
,
733 &BTRFS_I(inode
)->location
, 1);
740 leaf
= path
->nodes
[0];
741 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
742 struct btrfs_inode_item
);
744 fill_inode_item(leaf
, inode_item
, inode
);
745 btrfs_mark_buffer_dirty(leaf
);
746 btrfs_set_inode_last_trans(trans
, inode
);
749 btrfs_free_path(path
);
754 static int btrfs_unlink_trans(struct btrfs_trans_handle
*trans
,
755 struct btrfs_root
*root
,
757 struct dentry
*dentry
)
759 struct btrfs_path
*path
;
760 const char *name
= dentry
->d_name
.name
;
761 int name_len
= dentry
->d_name
.len
;
763 struct extent_buffer
*leaf
;
764 struct btrfs_dir_item
*di
;
765 struct btrfs_key key
;
767 path
= btrfs_alloc_path();
773 di
= btrfs_lookup_dir_item(trans
, root
, path
, dir
->i_ino
,
783 leaf
= path
->nodes
[0];
784 btrfs_dir_item_key_to_cpu(leaf
, di
, &key
);
785 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
788 btrfs_release_path(root
, path
);
790 di
= btrfs_lookup_dir_index_item(trans
, root
, path
, dir
->i_ino
,
791 key
.objectid
, name
, name_len
, -1);
800 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
801 btrfs_release_path(root
, path
);
803 dentry
->d_inode
->i_ctime
= dir
->i_ctime
;
804 ret
= btrfs_del_inode_ref(trans
, root
, name
, name_len
,
805 dentry
->d_inode
->i_ino
,
806 dentry
->d_parent
->d_inode
->i_ino
);
808 printk("failed to delete reference to %.*s, "
809 "inode %lu parent %lu\n", name_len
, name
,
810 dentry
->d_inode
->i_ino
,
811 dentry
->d_parent
->d_inode
->i_ino
);
814 btrfs_free_path(path
);
816 dir
->i_size
-= name_len
* 2;
817 dir
->i_mtime
= dir
->i_ctime
= CURRENT_TIME
;
818 btrfs_update_inode(trans
, root
, dir
);
819 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
820 dentry
->d_inode
->i_nlink
--;
822 drop_nlink(dentry
->d_inode
);
824 ret
= btrfs_update_inode(trans
, root
, dentry
->d_inode
);
825 dir
->i_sb
->s_dirt
= 1;
830 static int btrfs_unlink(struct inode
*dir
, struct dentry
*dentry
)
832 struct btrfs_root
*root
;
833 struct btrfs_trans_handle
*trans
;
834 struct inode
*inode
= dentry
->d_inode
;
836 unsigned long nr
= 0;
838 root
= BTRFS_I(dir
)->root
;
840 ret
= btrfs_check_free_space(root
, 1, 1);
844 trans
= btrfs_start_transaction(root
, 1);
846 btrfs_set_trans_block_group(trans
, dir
);
847 ret
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
848 nr
= trans
->blocks_used
;
850 if (inode
->i_nlink
== 0) {
851 /* if the inode isn't linked anywhere,
852 * we don't need to worry about
855 btrfs_del_ordered_inode(inode
, 1);
858 btrfs_end_transaction_throttle(trans
, root
);
860 btrfs_btree_balance_dirty(root
, nr
);
864 static int btrfs_rmdir(struct inode
*dir
, struct dentry
*dentry
)
866 struct inode
*inode
= dentry
->d_inode
;
869 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
870 struct btrfs_trans_handle
*trans
;
871 unsigned long nr
= 0;
873 if (inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
877 ret
= btrfs_check_free_space(root
, 1, 1);
881 trans
= btrfs_start_transaction(root
, 1);
882 btrfs_set_trans_block_group(trans
, dir
);
884 /* now the directory is empty */
885 err
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
890 nr
= trans
->blocks_used
;
891 ret
= btrfs_end_transaction_throttle(trans
, root
);
893 btrfs_btree_balance_dirty(root
, nr
);
901 * this can truncate away extent items, csum items and directory items.
902 * It starts at a high offset and removes keys until it can't find
903 * any higher than i_size.
905 * csum items that cross the new i_size are truncated to the new size
908 static int btrfs_truncate_in_trans(struct btrfs_trans_handle
*trans
,
909 struct btrfs_root
*root
,
914 struct btrfs_path
*path
;
915 struct btrfs_key key
;
916 struct btrfs_key found_key
;
918 struct extent_buffer
*leaf
;
919 struct btrfs_file_extent_item
*fi
;
920 u64 extent_start
= 0;
921 u64 extent_num_bytes
= 0;
927 int pending_del_nr
= 0;
928 int pending_del_slot
= 0;
929 int extent_type
= -1;
930 u64 mask
= root
->sectorsize
- 1;
932 btrfs_drop_extent_cache(inode
, inode
->i_size
& (~mask
), (u64
)-1);
933 path
= btrfs_alloc_path();
937 /* FIXME, add redo link to tree so we don't leak on crash */
938 key
.objectid
= inode
->i_ino
;
939 key
.offset
= (u64
)-1;
942 btrfs_init_path(path
);
944 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
949 BUG_ON(path
->slots
[0] == 0);
955 leaf
= path
->nodes
[0];
956 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
957 found_type
= btrfs_key_type(&found_key
);
959 if (found_key
.objectid
!= inode
->i_ino
)
962 if (found_type
< min_type
)
965 item_end
= found_key
.offset
;
966 if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
967 fi
= btrfs_item_ptr(leaf
, path
->slots
[0],
968 struct btrfs_file_extent_item
);
969 extent_type
= btrfs_file_extent_type(leaf
, fi
);
970 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
972 btrfs_file_extent_num_bytes(leaf
, fi
);
973 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
974 struct btrfs_item
*item
= btrfs_item_nr(leaf
,
976 item_end
+= btrfs_file_extent_inline_len(leaf
,
981 if (found_type
== BTRFS_CSUM_ITEM_KEY
) {
982 ret
= btrfs_csum_truncate(trans
, root
, path
,
986 if (item_end
< inode
->i_size
) {
987 if (found_type
== BTRFS_DIR_ITEM_KEY
) {
988 found_type
= BTRFS_INODE_ITEM_KEY
;
989 } else if (found_type
== BTRFS_EXTENT_ITEM_KEY
) {
990 found_type
= BTRFS_CSUM_ITEM_KEY
;
991 } else if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
992 found_type
= BTRFS_XATTR_ITEM_KEY
;
993 } else if (found_type
== BTRFS_XATTR_ITEM_KEY
) {
994 found_type
= BTRFS_INODE_REF_KEY
;
995 } else if (found_type
) {
1000 btrfs_set_key_type(&key
, found_type
);
1003 if (found_key
.offset
>= inode
->i_size
)
1009 /* FIXME, shrink the extent if the ref count is only 1 */
1010 if (found_type
!= BTRFS_EXTENT_DATA_KEY
)
1013 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1015 extent_start
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
1017 u64 orig_num_bytes
=
1018 btrfs_file_extent_num_bytes(leaf
, fi
);
1019 extent_num_bytes
= inode
->i_size
-
1020 found_key
.offset
+ root
->sectorsize
- 1;
1021 extent_num_bytes
= extent_num_bytes
&
1022 ~((u64
)root
->sectorsize
- 1);
1023 btrfs_set_file_extent_num_bytes(leaf
, fi
,
1025 num_dec
= (orig_num_bytes
-
1027 if (extent_start
!= 0)
1028 dec_i_blocks(inode
, num_dec
);
1029 btrfs_mark_buffer_dirty(leaf
);
1032 btrfs_file_extent_disk_num_bytes(leaf
,
1034 /* FIXME blocksize != 4096 */
1035 num_dec
= btrfs_file_extent_num_bytes(leaf
, fi
);
1036 if (extent_start
!= 0) {
1038 dec_i_blocks(inode
, num_dec
);
1040 root_gen
= btrfs_header_generation(leaf
);
1041 root_owner
= btrfs_header_owner(leaf
);
1043 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1045 u32 newsize
= inode
->i_size
- found_key
.offset
;
1046 dec_i_blocks(inode
, item_end
+ 1 -
1047 found_key
.offset
- newsize
);
1049 btrfs_file_extent_calc_inline_size(newsize
);
1050 ret
= btrfs_truncate_item(trans
, root
, path
,
1054 dec_i_blocks(inode
, item_end
+ 1 -
1060 if (!pending_del_nr
) {
1061 /* no pending yet, add ourselves */
1062 pending_del_slot
= path
->slots
[0];
1064 } else if (pending_del_nr
&&
1065 path
->slots
[0] + 1 == pending_del_slot
) {
1066 /* hop on the pending chunk */
1068 pending_del_slot
= path
->slots
[0];
1070 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path
->slots
[0], pending_del_nr
, pending_del_slot
);
1076 ret
= btrfs_free_extent(trans
, root
, extent_start
,
1079 root_gen
, inode
->i_ino
,
1080 found_key
.offset
, 0);
1084 if (path
->slots
[0] == 0) {
1087 btrfs_release_path(root
, path
);
1092 if (pending_del_nr
&&
1093 path
->slots
[0] + 1 != pending_del_slot
) {
1094 struct btrfs_key debug
;
1096 btrfs_item_key_to_cpu(path
->nodes
[0], &debug
,
1098 ret
= btrfs_del_items(trans
, root
, path
,
1103 btrfs_release_path(root
, path
);
1109 if (pending_del_nr
) {
1110 ret
= btrfs_del_items(trans
, root
, path
, pending_del_slot
,
1113 btrfs_free_path(path
);
1114 inode
->i_sb
->s_dirt
= 1;
1118 static int btrfs_cow_one_page(struct inode
*inode
, struct page
*page
,
1122 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1123 u64 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
1124 u64 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1127 WARN_ON(!PageLocked(page
));
1128 set_page_extent_mapped(page
);
1130 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1131 set_extent_delalloc(&BTRFS_I(inode
)->io_tree
, page_start
,
1132 page_end
, GFP_NOFS
);
1134 if (zero_start
!= PAGE_CACHE_SIZE
) {
1136 memset(kaddr
+ zero_start
, 0, PAGE_CACHE_SIZE
- zero_start
);
1137 flush_dcache_page(page
);
1140 set_page_dirty(page
);
1141 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1147 * taken from block_truncate_page, but does cow as it zeros out
1148 * any bytes left in the last page in the file.
1150 static int btrfs_truncate_page(struct address_space
*mapping
, loff_t from
)
1152 struct inode
*inode
= mapping
->host
;
1153 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1154 u32 blocksize
= root
->sectorsize
;
1155 pgoff_t index
= from
>> PAGE_CACHE_SHIFT
;
1156 unsigned offset
= from
& (PAGE_CACHE_SIZE
-1);
1161 if ((offset
& (blocksize
- 1)) == 0)
1166 page
= grab_cache_page(mapping
, index
);
1169 if (!PageUptodate(page
)) {
1170 ret
= btrfs_readpage(NULL
, page
);
1172 if (page
->mapping
!= mapping
) {
1174 page_cache_release(page
);
1177 if (!PageUptodate(page
)) {
1183 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
1184 wait_on_page_writeback(page
);
1185 ret
= btrfs_cow_one_page(inode
, page
, offset
);
1188 page_cache_release(page
);
1193 static int btrfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
1195 struct inode
*inode
= dentry
->d_inode
;
1198 err
= inode_change_ok(inode
, attr
);
1202 if (S_ISREG(inode
->i_mode
) &&
1203 attr
->ia_valid
& ATTR_SIZE
&& attr
->ia_size
> inode
->i_size
) {
1204 struct btrfs_trans_handle
*trans
;
1205 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1206 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1208 u64 mask
= root
->sectorsize
- 1;
1209 u64 hole_start
= (inode
->i_size
+ mask
) & ~mask
;
1210 u64 block_end
= (attr
->ia_size
+ mask
) & ~mask
;
1214 if (attr
->ia_size
<= hole_start
)
1217 err
= btrfs_check_free_space(root
, 1, 0);
1221 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
1223 lock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1224 hole_size
= block_end
- hole_start
;
1226 trans
= btrfs_start_transaction(root
, 1);
1227 btrfs_set_trans_block_group(trans
, inode
);
1228 err
= btrfs_drop_extents(trans
, root
, inode
,
1229 hole_start
, block_end
, hole_start
,
1232 if (alloc_hint
!= EXTENT_MAP_INLINE
) {
1233 err
= btrfs_insert_file_extent(trans
, root
,
1237 btrfs_drop_extent_cache(inode
, hole_start
,
1239 btrfs_check_file(root
, inode
);
1241 btrfs_end_transaction(trans
, root
);
1242 unlock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1247 err
= inode_setattr(inode
, attr
);
1252 void btrfs_delete_inode(struct inode
*inode
)
1254 struct btrfs_trans_handle
*trans
;
1255 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1259 truncate_inode_pages(&inode
->i_data
, 0);
1260 if (is_bad_inode(inode
)) {
1265 trans
= btrfs_start_transaction(root
, 1);
1267 btrfs_set_trans_block_group(trans
, inode
);
1268 ret
= btrfs_truncate_in_trans(trans
, root
, inode
, 0);
1270 goto no_delete_lock
;
1272 nr
= trans
->blocks_used
;
1275 btrfs_end_transaction(trans
, root
);
1276 btrfs_btree_balance_dirty(root
, nr
);
1280 nr
= trans
->blocks_used
;
1281 btrfs_end_transaction(trans
, root
);
1282 btrfs_btree_balance_dirty(root
, nr
);
1288 * this returns the key found in the dir entry in the location pointer.
1289 * If no dir entries were found, location->objectid is 0.
1291 static int btrfs_inode_by_name(struct inode
*dir
, struct dentry
*dentry
,
1292 struct btrfs_key
*location
)
1294 const char *name
= dentry
->d_name
.name
;
1295 int namelen
= dentry
->d_name
.len
;
1296 struct btrfs_dir_item
*di
;
1297 struct btrfs_path
*path
;
1298 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1301 if (namelen
== 1 && strcmp(name
, ".") == 0) {
1302 location
->objectid
= dir
->i_ino
;
1303 location
->type
= BTRFS_INODE_ITEM_KEY
;
1304 location
->offset
= 0;
1307 path
= btrfs_alloc_path();
1310 if (namelen
== 2 && strcmp(name
, "..") == 0) {
1311 struct btrfs_key key
;
1312 struct extent_buffer
*leaf
;
1316 key
.objectid
= dir
->i_ino
;
1317 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1319 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1323 leaf
= path
->nodes
[0];
1324 slot
= path
->slots
[0];
1325 nritems
= btrfs_header_nritems(leaf
);
1326 if (slot
>= nritems
)
1329 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
1330 if (key
.objectid
!= dir
->i_ino
||
1331 key
.type
!= BTRFS_INODE_REF_KEY
) {
1334 location
->objectid
= key
.offset
;
1335 location
->type
= BTRFS_INODE_ITEM_KEY
;
1336 location
->offset
= 0;
1340 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir
->i_ino
, name
,
1344 if (!di
|| IS_ERR(di
)) {
1347 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, location
);
1349 btrfs_free_path(path
);
1352 location
->objectid
= 0;
1357 * when we hit a tree root in a directory, the btrfs part of the inode
1358 * needs to be changed to reflect the root directory of the tree root. This
1359 * is kind of like crossing a mount point.
1361 static int fixup_tree_root_location(struct btrfs_root
*root
,
1362 struct btrfs_key
*location
,
1363 struct btrfs_root
**sub_root
,
1364 struct dentry
*dentry
)
1366 struct btrfs_path
*path
;
1367 struct btrfs_root_item
*ri
;
1369 if (btrfs_key_type(location
) != BTRFS_ROOT_ITEM_KEY
)
1371 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1374 path
= btrfs_alloc_path();
1377 *sub_root
= btrfs_read_fs_root(root
->fs_info
, location
,
1378 dentry
->d_name
.name
,
1379 dentry
->d_name
.len
);
1380 if (IS_ERR(*sub_root
))
1381 return PTR_ERR(*sub_root
);
1383 ri
= &(*sub_root
)->root_item
;
1384 location
->objectid
= btrfs_root_dirid(ri
);
1385 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1386 location
->offset
= 0;
1388 btrfs_free_path(path
);
1392 static int btrfs_init_locked_inode(struct inode
*inode
, void *p
)
1394 struct btrfs_iget_args
*args
= p
;
1395 inode
->i_ino
= args
->ino
;
1396 BTRFS_I(inode
)->root
= args
->root
;
1397 BTRFS_I(inode
)->delalloc_bytes
= 0;
1398 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1399 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1400 inode
->i_mapping
, GFP_NOFS
);
1401 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1402 inode
->i_mapping
, GFP_NOFS
);
1403 atomic_set(&BTRFS_I(inode
)->ordered_writeback
, 0);
1407 static int btrfs_find_actor(struct inode
*inode
, void *opaque
)
1409 struct btrfs_iget_args
*args
= opaque
;
1410 return (args
->ino
== inode
->i_ino
&&
1411 args
->root
== BTRFS_I(inode
)->root
);
1414 struct inode
*btrfs_ilookup(struct super_block
*s
, u64 objectid
,
1417 struct btrfs_iget_args args
;
1418 args
.ino
= objectid
;
1419 args
.root
= btrfs_lookup_fs_root(btrfs_sb(s
)->fs_info
, root_objectid
);
1424 return ilookup5(s
, objectid
, btrfs_find_actor
, (void *)&args
);
1427 struct inode
*btrfs_iget_locked(struct super_block
*s
, u64 objectid
,
1428 struct btrfs_root
*root
)
1430 struct inode
*inode
;
1431 struct btrfs_iget_args args
;
1432 args
.ino
= objectid
;
1435 inode
= iget5_locked(s
, objectid
, btrfs_find_actor
,
1436 btrfs_init_locked_inode
,
1441 static struct dentry
*btrfs_lookup(struct inode
*dir
, struct dentry
*dentry
,
1442 struct nameidata
*nd
)
1444 struct inode
* inode
;
1445 struct btrfs_inode
*bi
= BTRFS_I(dir
);
1446 struct btrfs_root
*root
= bi
->root
;
1447 struct btrfs_root
*sub_root
= root
;
1448 struct btrfs_key location
;
1451 if (dentry
->d_name
.len
> BTRFS_NAME_LEN
)
1452 return ERR_PTR(-ENAMETOOLONG
);
1454 ret
= btrfs_inode_by_name(dir
, dentry
, &location
);
1457 return ERR_PTR(ret
);
1460 if (location
.objectid
) {
1461 ret
= fixup_tree_root_location(root
, &location
, &sub_root
,
1464 return ERR_PTR(ret
);
1466 return ERR_PTR(-ENOENT
);
1467 inode
= btrfs_iget_locked(dir
->i_sb
, location
.objectid
,
1470 return ERR_PTR(-EACCES
);
1471 if (inode
->i_state
& I_NEW
) {
1472 /* the inode and parent dir are two different roots */
1473 if (sub_root
!= root
) {
1475 sub_root
->inode
= inode
;
1477 BTRFS_I(inode
)->root
= sub_root
;
1478 memcpy(&BTRFS_I(inode
)->location
, &location
,
1480 btrfs_read_locked_inode(inode
);
1481 unlock_new_inode(inode
);
1484 return d_splice_alias(inode
, dentry
);
1487 static unsigned char btrfs_filetype_table
[] = {
1488 DT_UNKNOWN
, DT_REG
, DT_DIR
, DT_CHR
, DT_BLK
, DT_FIFO
, DT_SOCK
, DT_LNK
1491 static int btrfs_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
1493 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1494 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1495 struct btrfs_item
*item
;
1496 struct btrfs_dir_item
*di
;
1497 struct btrfs_key key
;
1498 struct btrfs_key found_key
;
1499 struct btrfs_path
*path
;
1502 struct extent_buffer
*leaf
;
1505 unsigned char d_type
;
1510 int key_type
= BTRFS_DIR_INDEX_KEY
;
1515 /* FIXME, use a real flag for deciding about the key type */
1516 if (root
->fs_info
->tree_root
== root
)
1517 key_type
= BTRFS_DIR_ITEM_KEY
;
1519 /* special case for "." */
1520 if (filp
->f_pos
== 0) {
1521 over
= filldir(dirent
, ".", 1,
1529 key
.objectid
= inode
->i_ino
;
1530 path
= btrfs_alloc_path();
1533 /* special case for .., just use the back ref */
1534 if (filp
->f_pos
== 1) {
1535 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1537 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1539 leaf
= path
->nodes
[0];
1540 slot
= path
->slots
[0];
1541 nritems
= btrfs_header_nritems(leaf
);
1542 if (slot
>= nritems
) {
1543 btrfs_release_path(root
, path
);
1544 goto read_dir_items
;
1546 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1547 btrfs_release_path(root
, path
);
1548 if (found_key
.objectid
!= key
.objectid
||
1549 found_key
.type
!= BTRFS_INODE_REF_KEY
)
1550 goto read_dir_items
;
1551 over
= filldir(dirent
, "..", 2,
1552 2, found_key
.offset
, DT_DIR
);
1559 btrfs_set_key_type(&key
, key_type
);
1560 key
.offset
= filp
->f_pos
;
1562 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1567 leaf
= path
->nodes
[0];
1568 nritems
= btrfs_header_nritems(leaf
);
1569 slot
= path
->slots
[0];
1570 if (advance
|| slot
>= nritems
) {
1571 if (slot
>= nritems
-1) {
1572 ret
= btrfs_next_leaf(root
, path
);
1575 leaf
= path
->nodes
[0];
1576 nritems
= btrfs_header_nritems(leaf
);
1577 slot
= path
->slots
[0];
1584 item
= btrfs_item_nr(leaf
, slot
);
1585 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1587 if (found_key
.objectid
!= key
.objectid
)
1589 if (btrfs_key_type(&found_key
) != key_type
)
1591 if (found_key
.offset
< filp
->f_pos
)
1594 filp
->f_pos
= found_key
.offset
;
1596 di
= btrfs_item_ptr(leaf
, slot
, struct btrfs_dir_item
);
1598 di_total
= btrfs_item_size(leaf
, item
);
1599 while(di_cur
< di_total
) {
1600 struct btrfs_key location
;
1602 name_len
= btrfs_dir_name_len(leaf
, di
);
1603 if (name_len
< 32) {
1604 name_ptr
= tmp_name
;
1606 name_ptr
= kmalloc(name_len
, GFP_NOFS
);
1609 read_extent_buffer(leaf
, name_ptr
,
1610 (unsigned long)(di
+ 1), name_len
);
1612 d_type
= btrfs_filetype_table
[btrfs_dir_type(leaf
, di
)];
1613 btrfs_dir_item_key_to_cpu(leaf
, di
, &location
);
1614 over
= filldir(dirent
, name_ptr
, name_len
,
1619 if (name_ptr
!= tmp_name
)
1624 di_len
= btrfs_dir_name_len(leaf
, di
) +
1625 btrfs_dir_data_len(leaf
, di
) +sizeof(*di
);
1627 di
= (struct btrfs_dir_item
*)((char *)di
+ di_len
);
1630 if (key_type
== BTRFS_DIR_INDEX_KEY
)
1631 filp
->f_pos
= INT_LIMIT(typeof(filp
->f_pos
));
1637 btrfs_free_path(path
);
1641 int btrfs_write_inode(struct inode
*inode
, int wait
)
1643 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1644 struct btrfs_trans_handle
*trans
;
1648 trans
= btrfs_start_transaction(root
, 1);
1649 btrfs_set_trans_block_group(trans
, inode
);
1650 ret
= btrfs_commit_transaction(trans
, root
);
1656 * This is somewhat expensive, updating the tree every time the
1657 * inode changes. But, it is most likely to find the inode in cache.
1658 * FIXME, needs more benchmarking...there are no reasons other than performance
1659 * to keep or drop this code.
1661 void btrfs_dirty_inode(struct inode
*inode
)
1663 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1664 struct btrfs_trans_handle
*trans
;
1666 trans
= btrfs_start_transaction(root
, 1);
1667 btrfs_set_trans_block_group(trans
, inode
);
1668 btrfs_update_inode(trans
, root
, inode
);
1669 btrfs_end_transaction(trans
, root
);
1672 static struct inode
*btrfs_new_inode(struct btrfs_trans_handle
*trans
,
1673 struct btrfs_root
*root
,
1674 const char *name
, int name_len
,
1677 struct btrfs_block_group_cache
*group
,
1680 struct inode
*inode
;
1681 struct btrfs_inode_item
*inode_item
;
1682 struct btrfs_block_group_cache
*new_inode_group
;
1683 struct btrfs_key
*location
;
1684 struct btrfs_path
*path
;
1685 struct btrfs_inode_ref
*ref
;
1686 struct btrfs_key key
[2];
1692 path
= btrfs_alloc_path();
1695 inode
= new_inode(root
->fs_info
->sb
);
1697 return ERR_PTR(-ENOMEM
);
1699 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1700 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1701 inode
->i_mapping
, GFP_NOFS
);
1702 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1703 inode
->i_mapping
, GFP_NOFS
);
1704 atomic_set(&BTRFS_I(inode
)->ordered_writeback
, 0);
1705 BTRFS_I(inode
)->delalloc_bytes
= 0;
1706 BTRFS_I(inode
)->root
= root
;
1712 new_inode_group
= btrfs_find_block_group(root
, group
, 0,
1713 BTRFS_BLOCK_GROUP_METADATA
, owner
);
1714 if (!new_inode_group
) {
1715 printk("find_block group failed\n");
1716 new_inode_group
= group
;
1718 BTRFS_I(inode
)->block_group
= new_inode_group
;
1719 BTRFS_I(inode
)->flags
= 0;
1721 key
[0].objectid
= objectid
;
1722 btrfs_set_key_type(&key
[0], BTRFS_INODE_ITEM_KEY
);
1725 key
[1].objectid
= objectid
;
1726 btrfs_set_key_type(&key
[1], BTRFS_INODE_REF_KEY
);
1727 key
[1].offset
= ref_objectid
;
1729 sizes
[0] = sizeof(struct btrfs_inode_item
);
1730 sizes
[1] = name_len
+ sizeof(*ref
);
1732 ret
= btrfs_insert_empty_items(trans
, root
, path
, key
, sizes
, 2);
1736 if (objectid
> root
->highest_inode
)
1737 root
->highest_inode
= objectid
;
1739 inode
->i_uid
= current
->fsuid
;
1740 inode
->i_gid
= current
->fsgid
;
1741 inode
->i_mode
= mode
;
1742 inode
->i_ino
= objectid
;
1743 inode
->i_blocks
= 0;
1744 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1745 inode_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
1746 struct btrfs_inode_item
);
1747 fill_inode_item(path
->nodes
[0], inode_item
, inode
);
1749 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0] + 1,
1750 struct btrfs_inode_ref
);
1751 btrfs_set_inode_ref_name_len(path
->nodes
[0], ref
, name_len
);
1752 ptr
= (unsigned long)(ref
+ 1);
1753 write_extent_buffer(path
->nodes
[0], name
, ptr
, name_len
);
1755 btrfs_mark_buffer_dirty(path
->nodes
[0]);
1756 btrfs_free_path(path
);
1758 location
= &BTRFS_I(inode
)->location
;
1759 location
->objectid
= objectid
;
1760 location
->offset
= 0;
1761 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1763 insert_inode_hash(inode
);
1766 btrfs_free_path(path
);
1767 return ERR_PTR(ret
);
1770 static inline u8
btrfs_inode_type(struct inode
*inode
)
1772 return btrfs_type_by_mode
[(inode
->i_mode
& S_IFMT
) >> S_SHIFT
];
1775 static int btrfs_add_link(struct btrfs_trans_handle
*trans
,
1776 struct dentry
*dentry
, struct inode
*inode
,
1780 struct btrfs_key key
;
1781 struct btrfs_root
*root
= BTRFS_I(dentry
->d_parent
->d_inode
)->root
;
1782 struct inode
*parent_inode
;
1784 key
.objectid
= inode
->i_ino
;
1785 btrfs_set_key_type(&key
, BTRFS_INODE_ITEM_KEY
);
1788 ret
= btrfs_insert_dir_item(trans
, root
,
1789 dentry
->d_name
.name
, dentry
->d_name
.len
,
1790 dentry
->d_parent
->d_inode
->i_ino
,
1791 &key
, btrfs_inode_type(inode
));
1794 ret
= btrfs_insert_inode_ref(trans
, root
,
1795 dentry
->d_name
.name
,
1798 dentry
->d_parent
->d_inode
->i_ino
);
1800 parent_inode
= dentry
->d_parent
->d_inode
;
1801 parent_inode
->i_size
+= dentry
->d_name
.len
* 2;
1802 parent_inode
->i_mtime
= parent_inode
->i_ctime
= CURRENT_TIME
;
1803 ret
= btrfs_update_inode(trans
, root
,
1804 dentry
->d_parent
->d_inode
);
1809 static int btrfs_add_nondir(struct btrfs_trans_handle
*trans
,
1810 struct dentry
*dentry
, struct inode
*inode
,
1813 int err
= btrfs_add_link(trans
, dentry
, inode
, backref
);
1815 d_instantiate(dentry
, inode
);
1823 static int btrfs_mknod(struct inode
*dir
, struct dentry
*dentry
,
1824 int mode
, dev_t rdev
)
1826 struct btrfs_trans_handle
*trans
;
1827 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1828 struct inode
*inode
= NULL
;
1832 unsigned long nr
= 0;
1834 if (!new_valid_dev(rdev
))
1837 err
= btrfs_check_free_space(root
, 1, 0);
1841 trans
= btrfs_start_transaction(root
, 1);
1842 btrfs_set_trans_block_group(trans
, dir
);
1844 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
1850 inode
= btrfs_new_inode(trans
, root
, dentry
->d_name
.name
,
1852 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
1853 BTRFS_I(dir
)->block_group
, mode
);
1854 err
= PTR_ERR(inode
);
1858 btrfs_set_trans_block_group(trans
, inode
);
1859 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
1863 inode
->i_op
= &btrfs_special_inode_operations
;
1864 init_special_inode(inode
, inode
->i_mode
, rdev
);
1865 btrfs_update_inode(trans
, root
, inode
);
1867 dir
->i_sb
->s_dirt
= 1;
1868 btrfs_update_inode_block_group(trans
, inode
);
1869 btrfs_update_inode_block_group(trans
, dir
);
1871 nr
= trans
->blocks_used
;
1872 btrfs_end_transaction_throttle(trans
, root
);
1875 inode_dec_link_count(inode
);
1878 btrfs_btree_balance_dirty(root
, nr
);
1882 static int btrfs_create(struct inode
*dir
, struct dentry
*dentry
,
1883 int mode
, struct nameidata
*nd
)
1885 struct btrfs_trans_handle
*trans
;
1886 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1887 struct inode
*inode
= NULL
;
1890 unsigned long nr
= 0;
1893 err
= btrfs_check_free_space(root
, 1, 0);
1896 trans
= btrfs_start_transaction(root
, 1);
1897 btrfs_set_trans_block_group(trans
, dir
);
1899 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
1905 inode
= btrfs_new_inode(trans
, root
, dentry
->d_name
.name
,
1907 dentry
->d_parent
->d_inode
->i_ino
,
1908 objectid
, BTRFS_I(dir
)->block_group
, mode
);
1909 err
= PTR_ERR(inode
);
1913 btrfs_set_trans_block_group(trans
, inode
);
1914 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
1918 inode
->i_mapping
->a_ops
= &btrfs_aops
;
1919 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1920 inode
->i_fop
= &btrfs_file_operations
;
1921 inode
->i_op
= &btrfs_file_inode_operations
;
1922 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1923 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1924 inode
->i_mapping
, GFP_NOFS
);
1925 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1926 inode
->i_mapping
, GFP_NOFS
);
1927 BTRFS_I(inode
)->delalloc_bytes
= 0;
1928 atomic_set(&BTRFS_I(inode
)->ordered_writeback
, 0);
1929 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
1931 dir
->i_sb
->s_dirt
= 1;
1932 btrfs_update_inode_block_group(trans
, inode
);
1933 btrfs_update_inode_block_group(trans
, dir
);
1935 nr
= trans
->blocks_used
;
1936 btrfs_end_transaction_throttle(trans
, root
);
1939 inode_dec_link_count(inode
);
1942 btrfs_btree_balance_dirty(root
, nr
);
1946 static int btrfs_link(struct dentry
*old_dentry
, struct inode
*dir
,
1947 struct dentry
*dentry
)
1949 struct btrfs_trans_handle
*trans
;
1950 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1951 struct inode
*inode
= old_dentry
->d_inode
;
1952 unsigned long nr
= 0;
1956 if (inode
->i_nlink
== 0)
1959 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1964 err
= btrfs_check_free_space(root
, 1, 0);
1967 trans
= btrfs_start_transaction(root
, 1);
1969 btrfs_set_trans_block_group(trans
, dir
);
1970 atomic_inc(&inode
->i_count
);
1971 err
= btrfs_add_nondir(trans
, dentry
, inode
, 1);
1976 dir
->i_sb
->s_dirt
= 1;
1977 btrfs_update_inode_block_group(trans
, dir
);
1978 err
= btrfs_update_inode(trans
, root
, inode
);
1983 nr
= trans
->blocks_used
;
1984 btrfs_end_transaction_throttle(trans
, root
);
1987 inode_dec_link_count(inode
);
1990 btrfs_btree_balance_dirty(root
, nr
);
1994 static int btrfs_mkdir(struct inode
*dir
, struct dentry
*dentry
, int mode
)
1996 struct inode
*inode
= NULL
;
1997 struct btrfs_trans_handle
*trans
;
1998 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2000 int drop_on_err
= 0;
2002 unsigned long nr
= 1;
2004 err
= btrfs_check_free_space(root
, 1, 0);
2008 trans
= btrfs_start_transaction(root
, 1);
2009 btrfs_set_trans_block_group(trans
, dir
);
2011 if (IS_ERR(trans
)) {
2012 err
= PTR_ERR(trans
);
2016 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2022 inode
= btrfs_new_inode(trans
, root
, dentry
->d_name
.name
,
2024 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2025 BTRFS_I(dir
)->block_group
, S_IFDIR
| mode
);
2026 if (IS_ERR(inode
)) {
2027 err
= PTR_ERR(inode
);
2032 inode
->i_op
= &btrfs_dir_inode_operations
;
2033 inode
->i_fop
= &btrfs_dir_file_operations
;
2034 btrfs_set_trans_block_group(trans
, inode
);
2037 err
= btrfs_update_inode(trans
, root
, inode
);
2041 err
= btrfs_add_link(trans
, dentry
, inode
, 0);
2045 d_instantiate(dentry
, inode
);
2047 dir
->i_sb
->s_dirt
= 1;
2048 btrfs_update_inode_block_group(trans
, inode
);
2049 btrfs_update_inode_block_group(trans
, dir
);
2052 nr
= trans
->blocks_used
;
2053 btrfs_end_transaction_throttle(trans
, root
);
2058 btrfs_btree_balance_dirty(root
, nr
);
2062 static int merge_extent_mapping(struct extent_map_tree
*em_tree
,
2063 struct extent_map
*existing
,
2064 struct extent_map
*em
)
2069 int real_blocks
= existing
->block_start
< EXTENT_MAP_LAST_BYTE
;
2071 if (real_blocks
&& em
->block_start
>= EXTENT_MAP_LAST_BYTE
)
2074 if (!real_blocks
&& em
->block_start
!= existing
->block_start
)
2077 new_end
= max(existing
->start
+ existing
->len
, em
->start
+ em
->len
);
2079 if (existing
->start
>= em
->start
) {
2080 if (em
->start
+ em
->len
< existing
->start
)
2083 start_diff
= existing
->start
- em
->start
;
2084 if (real_blocks
&& em
->block_start
+ start_diff
!=
2085 existing
->block_start
)
2088 em
->len
= new_end
- em
->start
;
2090 remove_extent_mapping(em_tree
, existing
);
2091 /* free for the tree */
2092 free_extent_map(existing
);
2093 ret
= add_extent_mapping(em_tree
, em
);
2095 } else if (em
->start
> existing
->start
) {
2097 if (existing
->start
+ existing
->len
< em
->start
)
2100 start_diff
= em
->start
- existing
->start
;
2101 if (real_blocks
&& existing
->block_start
+ start_diff
!=
2105 remove_extent_mapping(em_tree
, existing
);
2106 em
->block_start
= existing
->block_start
;
2107 em
->start
= existing
->start
;
2108 em
->len
= new_end
- existing
->start
;
2109 free_extent_map(existing
);
2111 ret
= add_extent_mapping(em_tree
, em
);
2118 printk("invalid extent map merge [%Lu %Lu %Lu] [%Lu %Lu %Lu]\n",
2119 existing
->start
, existing
->len
, existing
->block_start
,
2120 em
->start
, em
->len
, em
->block_start
);
2124 struct extent_map
*btrfs_get_extent(struct inode
*inode
, struct page
*page
,
2125 size_t pg_offset
, u64 start
, u64 len
,
2131 u64 extent_start
= 0;
2133 u64 objectid
= inode
->i_ino
;
2135 struct btrfs_path
*path
;
2136 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2137 struct btrfs_file_extent_item
*item
;
2138 struct extent_buffer
*leaf
;
2139 struct btrfs_key found_key
;
2140 struct extent_map
*em
= NULL
;
2141 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
2142 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2143 struct btrfs_trans_handle
*trans
= NULL
;
2145 path
= btrfs_alloc_path();
2149 spin_lock(&em_tree
->lock
);
2150 em
= lookup_extent_mapping(em_tree
, start
, len
);
2152 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2153 spin_unlock(&em_tree
->lock
);
2156 if (em
->start
> start
|| em
->start
+ em
->len
<= start
)
2157 free_extent_map(em
);
2158 else if (em
->block_start
== EXTENT_MAP_INLINE
&& page
)
2159 free_extent_map(em
);
2163 em
= alloc_extent_map(GFP_NOFS
);
2169 em
->start
= EXTENT_MAP_HOLE
;
2171 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2172 ret
= btrfs_lookup_file_extent(trans
, root
, path
,
2173 objectid
, start
, trans
!= NULL
);
2180 if (path
->slots
[0] == 0)
2185 leaf
= path
->nodes
[0];
2186 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
2187 struct btrfs_file_extent_item
);
2188 /* are we inside the extent that was found? */
2189 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2190 found_type
= btrfs_key_type(&found_key
);
2191 if (found_key
.objectid
!= objectid
||
2192 found_type
!= BTRFS_EXTENT_DATA_KEY
) {
2196 found_type
= btrfs_file_extent_type(leaf
, item
);
2197 extent_start
= found_key
.offset
;
2198 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
2199 extent_end
= extent_start
+
2200 btrfs_file_extent_num_bytes(leaf
, item
);
2202 if (start
< extent_start
|| start
>= extent_end
) {
2204 if (start
< extent_start
) {
2205 if (start
+ len
<= extent_start
)
2207 em
->len
= extent_end
- extent_start
;
2213 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
2215 em
->start
= extent_start
;
2216 em
->len
= extent_end
- extent_start
;
2217 em
->block_start
= EXTENT_MAP_HOLE
;
2220 bytenr
+= btrfs_file_extent_offset(leaf
, item
);
2221 em
->block_start
= bytenr
;
2222 em
->start
= extent_start
;
2223 em
->len
= extent_end
- extent_start
;
2225 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
2230 size_t extent_offset
;
2233 size
= btrfs_file_extent_inline_len(leaf
, btrfs_item_nr(leaf
,
2235 extent_end
= (extent_start
+ size
+ root
->sectorsize
- 1) &
2236 ~((u64
)root
->sectorsize
- 1);
2237 if (start
< extent_start
|| start
>= extent_end
) {
2239 if (start
< extent_start
) {
2240 if (start
+ len
<= extent_start
)
2242 em
->len
= extent_end
- extent_start
;
2248 em
->block_start
= EXTENT_MAP_INLINE
;
2251 em
->start
= extent_start
;
2256 page_start
= page_offset(page
) + pg_offset
;
2257 extent_offset
= page_start
- extent_start
;
2258 copy_size
= min_t(u64
, PAGE_CACHE_SIZE
- pg_offset
,
2259 size
- extent_offset
);
2260 em
->start
= extent_start
+ extent_offset
;
2261 em
->len
= (copy_size
+ root
->sectorsize
- 1) &
2262 ~((u64
)root
->sectorsize
- 1);
2264 ptr
= btrfs_file_extent_inline_start(item
) + extent_offset
;
2265 if (create
== 0 && !PageUptodate(page
)) {
2266 read_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2268 flush_dcache_page(page
);
2269 } else if (create
&& PageUptodate(page
)) {
2272 free_extent_map(em
);
2274 btrfs_release_path(root
, path
);
2275 trans
= btrfs_start_transaction(root
, 1);
2278 write_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2280 btrfs_mark_buffer_dirty(leaf
);
2283 set_extent_uptodate(io_tree
, em
->start
,
2284 extent_map_end(em
) - 1, GFP_NOFS
);
2287 printk("unkknown found_type %d\n", found_type
);
2294 em
->block_start
= EXTENT_MAP_HOLE
;
2296 btrfs_release_path(root
, path
);
2297 if (em
->start
> start
|| extent_map_end(em
) <= start
) {
2298 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em
->start
, em
->len
, start
, len
);
2304 spin_lock(&em_tree
->lock
);
2305 ret
= add_extent_mapping(em_tree
, em
);
2306 /* it is possible that someone inserted the extent into the tree
2307 * while we had the lock dropped. It is also possible that
2308 * an overlapping map exists in the tree
2310 if (ret
== -EEXIST
) {
2311 struct extent_map
*existing
;
2312 existing
= lookup_extent_mapping(em_tree
, start
, len
);
2313 if (existing
&& (existing
->start
> start
||
2314 existing
->start
+ existing
->len
<= start
)) {
2315 free_extent_map(existing
);
2319 existing
= lookup_extent_mapping(em_tree
, em
->start
,
2322 err
= merge_extent_mapping(em_tree
, existing
,
2324 free_extent_map(existing
);
2326 free_extent_map(em
);
2331 printk("failing to insert %Lu %Lu\n",
2333 free_extent_map(em
);
2337 free_extent_map(em
);
2341 spin_unlock(&em_tree
->lock
);
2343 btrfs_free_path(path
);
2345 ret
= btrfs_end_transaction(trans
, root
);
2350 free_extent_map(em
);
2352 return ERR_PTR(err
);
2357 #if 0 /* waiting for O_DIRECT reads */
2358 static int btrfs_get_block(struct inode
*inode
, sector_t iblock
,
2359 struct buffer_head
*bh_result
, int create
)
2361 struct extent_map
*em
;
2362 u64 start
= (u64
)iblock
<< inode
->i_blkbits
;
2363 struct btrfs_multi_bio
*multi
= NULL
;
2364 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2370 em
= btrfs_get_extent(inode
, NULL
, 0, start
, bh_result
->b_size
, 0);
2372 if (!em
|| IS_ERR(em
))
2375 if (em
->start
> start
|| em
->start
+ em
->len
<= start
) {
2379 if (em
->block_start
== EXTENT_MAP_INLINE
) {
2384 len
= em
->start
+ em
->len
- start
;
2385 len
= min_t(u64
, len
, INT_LIMIT(typeof(bh_result
->b_size
)));
2387 if (em
->block_start
== EXTENT_MAP_HOLE
||
2388 em
->block_start
== EXTENT_MAP_DELALLOC
) {
2389 bh_result
->b_size
= len
;
2393 logical
= start
- em
->start
;
2394 logical
= em
->block_start
+ logical
;
2397 ret
= btrfs_map_block(&root
->fs_info
->mapping_tree
, READ
,
2398 logical
, &map_length
, &multi
, 0);
2400 bh_result
->b_blocknr
= multi
->stripes
[0].physical
>> inode
->i_blkbits
;
2401 bh_result
->b_size
= min(map_length
, len
);
2403 bh_result
->b_bdev
= multi
->stripes
[0].dev
->bdev
;
2404 set_buffer_mapped(bh_result
);
2407 free_extent_map(em
);
2412 static ssize_t
btrfs_direct_IO(int rw
, struct kiocb
*iocb
,
2413 const struct iovec
*iov
, loff_t offset
,
2414 unsigned long nr_segs
)
2418 struct file
*file
= iocb
->ki_filp
;
2419 struct inode
*inode
= file
->f_mapping
->host
;
2424 return blockdev_direct_IO(rw
, iocb
, inode
, inode
->i_sb
->s_bdev
, iov
,
2425 offset
, nr_segs
, btrfs_get_block
, NULL
);
2429 static sector_t
btrfs_bmap(struct address_space
*mapping
, sector_t iblock
)
2431 return extent_bmap(mapping
, iblock
, btrfs_get_extent
);
2434 int btrfs_readpage(struct file
*file
, struct page
*page
)
2436 struct extent_io_tree
*tree
;
2437 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2438 return extent_read_full_page(tree
, page
, btrfs_get_extent
);
2441 static int btrfs_writepage(struct page
*page
, struct writeback_control
*wbc
)
2443 struct extent_io_tree
*tree
;
2446 if (current
->flags
& PF_MEMALLOC
) {
2447 redirty_page_for_writepage(wbc
, page
);
2451 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2452 return extent_write_full_page(tree
, page
, btrfs_get_extent
, wbc
);
2455 static int btrfs_writepages(struct address_space
*mapping
,
2456 struct writeback_control
*wbc
)
2458 struct extent_io_tree
*tree
;
2459 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
2460 return extent_writepages(tree
, mapping
, btrfs_get_extent
, wbc
);
2464 btrfs_readpages(struct file
*file
, struct address_space
*mapping
,
2465 struct list_head
*pages
, unsigned nr_pages
)
2467 struct extent_io_tree
*tree
;
2468 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
2469 return extent_readpages(tree
, mapping
, pages
, nr_pages
,
2473 static int btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
2475 struct extent_io_tree
*tree
;
2476 struct extent_map_tree
*map
;
2479 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2480 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
2481 ret
= try_release_extent_mapping(map
, tree
, page
, gfp_flags
);
2483 invalidate_extent_lru(tree
, page_offset(page
), PAGE_CACHE_SIZE
);
2484 ClearPagePrivate(page
);
2485 set_page_private(page
, 0);
2486 page_cache_release(page
);
2491 static void btrfs_invalidatepage(struct page
*page
, unsigned long offset
)
2493 struct extent_io_tree
*tree
;
2495 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2496 extent_invalidatepage(tree
, page
, offset
);
2497 btrfs_releasepage(page
, GFP_NOFS
);
2498 if (PagePrivate(page
)) {
2499 invalidate_extent_lru(tree
, page_offset(page
), PAGE_CACHE_SIZE
);
2500 ClearPagePrivate(page
);
2501 set_page_private(page
, 0);
2502 page_cache_release(page
);
2507 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
2508 * called from a page fault handler when a page is first dirtied. Hence we must
2509 * be careful to check for EOF conditions here. We set the page up correctly
2510 * for a written page which means we get ENOSPC checking when writing into
2511 * holes and correct delalloc and unwritten extent mapping on filesystems that
2512 * support these features.
2514 * We are not allowed to take the i_mutex here so we have to play games to
2515 * protect against truncate races as the page could now be beyond EOF. Because
2516 * vmtruncate() writes the inode size before removing pages, once we have the
2517 * page lock we can determine safely if the page is beyond EOF. If it is not
2518 * beyond EOF, then the page is guaranteed safe against truncation until we
2521 int btrfs_page_mkwrite(struct vm_area_struct
*vma
, struct page
*page
)
2523 struct inode
*inode
= fdentry(vma
->vm_file
)->d_inode
;
2524 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2530 ret
= btrfs_check_free_space(root
, PAGE_CACHE_SIZE
, 0);
2537 wait_on_page_writeback(page
);
2538 size
= i_size_read(inode
);
2539 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
2541 if ((page
->mapping
!= inode
->i_mapping
) ||
2542 (page_start
> size
)) {
2543 /* page got truncated out from underneath us */
2547 /* page is wholly or partially inside EOF */
2548 if (page_start
+ PAGE_CACHE_SIZE
> size
)
2549 end
= size
& ~PAGE_CACHE_MASK
;
2551 end
= PAGE_CACHE_SIZE
;
2553 ret
= btrfs_cow_one_page(inode
, page
, end
);
2561 static void btrfs_truncate(struct inode
*inode
)
2563 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2565 struct btrfs_trans_handle
*trans
;
2568 if (!S_ISREG(inode
->i_mode
))
2570 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
2573 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
2575 trans
= btrfs_start_transaction(root
, 1);
2576 btrfs_set_trans_block_group(trans
, inode
);
2578 /* FIXME, add redo link to tree so we don't leak on crash */
2579 ret
= btrfs_truncate_in_trans(trans
, root
, inode
,
2580 BTRFS_EXTENT_DATA_KEY
);
2581 btrfs_update_inode(trans
, root
, inode
);
2582 nr
= trans
->blocks_used
;
2584 ret
= btrfs_end_transaction_throttle(trans
, root
);
2586 btrfs_btree_balance_dirty(root
, nr
);
2590 * Invalidate a single dcache entry at the root of the filesystem.
2591 * Needed after creation of snapshot or subvolume.
2593 void btrfs_invalidate_dcache_root(struct btrfs_root
*root
, char *name
,
2596 struct dentry
*alias
, *entry
;
2599 alias
= d_find_alias(root
->fs_info
->sb
->s_root
->d_inode
);
2603 /* change me if btrfs ever gets a d_hash operation */
2604 qstr
.hash
= full_name_hash(qstr
.name
, qstr
.len
);
2605 entry
= d_lookup(alias
, &qstr
);
2608 d_invalidate(entry
);
2614 int btrfs_create_subvol_root(struct btrfs_root
*new_root
,
2615 struct btrfs_trans_handle
*trans
, u64 new_dirid
,
2616 struct btrfs_block_group_cache
*block_group
)
2618 struct inode
*inode
;
2621 inode
= btrfs_new_inode(trans
, new_root
, "..", 2, new_dirid
,
2622 new_dirid
, block_group
, S_IFDIR
| 0700);
2624 return PTR_ERR(inode
);
2625 inode
->i_op
= &btrfs_dir_inode_operations
;
2626 inode
->i_fop
= &btrfs_dir_file_operations
;
2627 new_root
->inode
= inode
;
2629 ret
= btrfs_insert_inode_ref(trans
, new_root
, "..", 2, new_dirid
,
2634 return btrfs_update_inode(trans
, new_root
, inode
);
2637 unsigned long btrfs_force_ra(struct address_space
*mapping
,
2638 struct file_ra_state
*ra
, struct file
*file
,
2639 pgoff_t offset
, pgoff_t last_index
)
2641 pgoff_t req_size
= last_index
- offset
+ 1;
2643 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2644 offset
= page_cache_readahead(mapping
, ra
, file
, offset
, req_size
);
2647 page_cache_sync_readahead(mapping
, ra
, file
, offset
, req_size
);
2648 return offset
+ req_size
;
2652 struct inode
*btrfs_alloc_inode(struct super_block
*sb
)
2654 struct btrfs_inode
*ei
;
2656 ei
= kmem_cache_alloc(btrfs_inode_cachep
, GFP_NOFS
);
2660 ei
->ordered_trans
= 0;
2661 return &ei
->vfs_inode
;
2664 void btrfs_destroy_inode(struct inode
*inode
)
2666 WARN_ON(!list_empty(&inode
->i_dentry
));
2667 WARN_ON(inode
->i_data
.nrpages
);
2669 btrfs_drop_extent_cache(inode
, 0, (u64
)-1);
2670 kmem_cache_free(btrfs_inode_cachep
, BTRFS_I(inode
));
2673 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2674 static void init_once(struct kmem_cache
* cachep
, void *foo
)
2676 static void init_once(void * foo
, struct kmem_cache
* cachep
,
2677 unsigned long flags
)
2680 struct btrfs_inode
*ei
= (struct btrfs_inode
*) foo
;
2682 inode_init_once(&ei
->vfs_inode
);
2685 void btrfs_destroy_cachep(void)
2687 if (btrfs_inode_cachep
)
2688 kmem_cache_destroy(btrfs_inode_cachep
);
2689 if (btrfs_trans_handle_cachep
)
2690 kmem_cache_destroy(btrfs_trans_handle_cachep
);
2691 if (btrfs_transaction_cachep
)
2692 kmem_cache_destroy(btrfs_transaction_cachep
);
2693 if (btrfs_bit_radix_cachep
)
2694 kmem_cache_destroy(btrfs_bit_radix_cachep
);
2695 if (btrfs_path_cachep
)
2696 kmem_cache_destroy(btrfs_path_cachep
);
2699 struct kmem_cache
*btrfs_cache_create(const char *name
, size_t size
,
2700 unsigned long extra_flags
,
2701 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2702 void (*ctor
)(struct kmem_cache
*, void *)
2704 void (*ctor
)(void *, struct kmem_cache
*,
2709 return kmem_cache_create(name
, size
, 0, (SLAB_RECLAIM_ACCOUNT
|
2710 SLAB_MEM_SPREAD
| extra_flags
), ctor
2711 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2717 int btrfs_init_cachep(void)
2719 btrfs_inode_cachep
= btrfs_cache_create("btrfs_inode_cache",
2720 sizeof(struct btrfs_inode
),
2722 if (!btrfs_inode_cachep
)
2724 btrfs_trans_handle_cachep
=
2725 btrfs_cache_create("btrfs_trans_handle_cache",
2726 sizeof(struct btrfs_trans_handle
),
2728 if (!btrfs_trans_handle_cachep
)
2730 btrfs_transaction_cachep
= btrfs_cache_create("btrfs_transaction_cache",
2731 sizeof(struct btrfs_transaction
),
2733 if (!btrfs_transaction_cachep
)
2735 btrfs_path_cachep
= btrfs_cache_create("btrfs_path_cache",
2736 sizeof(struct btrfs_path
),
2738 if (!btrfs_path_cachep
)
2740 btrfs_bit_radix_cachep
= btrfs_cache_create("btrfs_radix", 256,
2741 SLAB_DESTROY_BY_RCU
, NULL
);
2742 if (!btrfs_bit_radix_cachep
)
2746 btrfs_destroy_cachep();
2750 static int btrfs_getattr(struct vfsmount
*mnt
,
2751 struct dentry
*dentry
, struct kstat
*stat
)
2753 struct inode
*inode
= dentry
->d_inode
;
2754 generic_fillattr(inode
, stat
);
2755 stat
->blksize
= PAGE_CACHE_SIZE
;
2756 stat
->blocks
= inode
->i_blocks
+ (BTRFS_I(inode
)->delalloc_bytes
>> 9);
2760 static int btrfs_rename(struct inode
* old_dir
, struct dentry
*old_dentry
,
2761 struct inode
* new_dir
,struct dentry
*new_dentry
)
2763 struct btrfs_trans_handle
*trans
;
2764 struct btrfs_root
*root
= BTRFS_I(old_dir
)->root
;
2765 struct inode
*new_inode
= new_dentry
->d_inode
;
2766 struct inode
*old_inode
= old_dentry
->d_inode
;
2767 struct timespec ctime
= CURRENT_TIME
;
2770 if (S_ISDIR(old_inode
->i_mode
) && new_inode
&&
2771 new_inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
2775 ret
= btrfs_check_free_space(root
, 1, 0);
2779 trans
= btrfs_start_transaction(root
, 1);
2781 btrfs_set_trans_block_group(trans
, new_dir
);
2783 old_dentry
->d_inode
->i_nlink
++;
2784 old_dir
->i_ctime
= old_dir
->i_mtime
= ctime
;
2785 new_dir
->i_ctime
= new_dir
->i_mtime
= ctime
;
2786 old_inode
->i_ctime
= ctime
;
2788 ret
= btrfs_unlink_trans(trans
, root
, old_dir
, old_dentry
);
2793 new_inode
->i_ctime
= CURRENT_TIME
;
2794 ret
= btrfs_unlink_trans(trans
, root
, new_dir
, new_dentry
);
2798 ret
= btrfs_add_link(trans
, new_dentry
, old_inode
, 1);
2803 btrfs_end_transaction(trans
, root
);
2808 static int btrfs_symlink(struct inode
*dir
, struct dentry
*dentry
,
2809 const char *symname
)
2811 struct btrfs_trans_handle
*trans
;
2812 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2813 struct btrfs_path
*path
;
2814 struct btrfs_key key
;
2815 struct inode
*inode
= NULL
;
2822 struct btrfs_file_extent_item
*ei
;
2823 struct extent_buffer
*leaf
;
2824 unsigned long nr
= 0;
2826 name_len
= strlen(symname
) + 1;
2827 if (name_len
> BTRFS_MAX_INLINE_DATA_SIZE(root
))
2828 return -ENAMETOOLONG
;
2830 err
= btrfs_check_free_space(root
, 1, 0);
2834 trans
= btrfs_start_transaction(root
, 1);
2835 btrfs_set_trans_block_group(trans
, dir
);
2837 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2843 inode
= btrfs_new_inode(trans
, root
, dentry
->d_name
.name
,
2845 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2846 BTRFS_I(dir
)->block_group
, S_IFLNK
|S_IRWXUGO
);
2847 err
= PTR_ERR(inode
);
2851 btrfs_set_trans_block_group(trans
, inode
);
2852 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
2856 inode
->i_mapping
->a_ops
= &btrfs_aops
;
2857 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
2858 inode
->i_fop
= &btrfs_file_operations
;
2859 inode
->i_op
= &btrfs_file_inode_operations
;
2860 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
2861 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
2862 inode
->i_mapping
, GFP_NOFS
);
2863 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
2864 inode
->i_mapping
, GFP_NOFS
);
2865 BTRFS_I(inode
)->delalloc_bytes
= 0;
2866 atomic_set(&BTRFS_I(inode
)->ordered_writeback
, 0);
2867 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
2869 dir
->i_sb
->s_dirt
= 1;
2870 btrfs_update_inode_block_group(trans
, inode
);
2871 btrfs_update_inode_block_group(trans
, dir
);
2875 path
= btrfs_alloc_path();
2877 key
.objectid
= inode
->i_ino
;
2879 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
2880 datasize
= btrfs_file_extent_calc_inline_size(name_len
);
2881 err
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
2887 leaf
= path
->nodes
[0];
2888 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
2889 struct btrfs_file_extent_item
);
2890 btrfs_set_file_extent_generation(leaf
, ei
, trans
->transid
);
2891 btrfs_set_file_extent_type(leaf
, ei
,
2892 BTRFS_FILE_EXTENT_INLINE
);
2893 ptr
= btrfs_file_extent_inline_start(ei
);
2894 write_extent_buffer(leaf
, symname
, ptr
, name_len
);
2895 btrfs_mark_buffer_dirty(leaf
);
2896 btrfs_free_path(path
);
2898 inode
->i_op
= &btrfs_symlink_inode_operations
;
2899 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
2900 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
2901 inode
->i_size
= name_len
- 1;
2902 err
= btrfs_update_inode(trans
, root
, inode
);
2907 nr
= trans
->blocks_used
;
2908 btrfs_end_transaction_throttle(trans
, root
);
2911 inode_dec_link_count(inode
);
2914 btrfs_btree_balance_dirty(root
, nr
);
2918 static int btrfs_permission(struct inode
*inode
, int mask
,
2919 struct nameidata
*nd
)
2921 if (btrfs_test_flag(inode
, READONLY
) && (mask
& MAY_WRITE
))
2923 return generic_permission(inode
, mask
, NULL
);
2926 static struct inode_operations btrfs_dir_inode_operations
= {
2927 .lookup
= btrfs_lookup
,
2928 .create
= btrfs_create
,
2929 .unlink
= btrfs_unlink
,
2931 .mkdir
= btrfs_mkdir
,
2932 .rmdir
= btrfs_rmdir
,
2933 .rename
= btrfs_rename
,
2934 .symlink
= btrfs_symlink
,
2935 .setattr
= btrfs_setattr
,
2936 .mknod
= btrfs_mknod
,
2937 .setxattr
= generic_setxattr
,
2938 .getxattr
= generic_getxattr
,
2939 .listxattr
= btrfs_listxattr
,
2940 .removexattr
= generic_removexattr
,
2941 .permission
= btrfs_permission
,
2943 static struct inode_operations btrfs_dir_ro_inode_operations
= {
2944 .lookup
= btrfs_lookup
,
2945 .permission
= btrfs_permission
,
2947 static struct file_operations btrfs_dir_file_operations
= {
2948 .llseek
= generic_file_llseek
,
2949 .read
= generic_read_dir
,
2950 .readdir
= btrfs_readdir
,
2951 .unlocked_ioctl
= btrfs_ioctl
,
2952 #ifdef CONFIG_COMPAT
2953 .compat_ioctl
= btrfs_ioctl
,
2955 .release
= btrfs_release_file
,
2958 static struct extent_io_ops btrfs_extent_io_ops
= {
2959 .fill_delalloc
= run_delalloc_range
,
2960 .submit_bio_hook
= btrfs_submit_bio_hook
,
2961 .merge_bio_hook
= btrfs_merge_bio_hook
,
2962 .readpage_io_hook
= btrfs_readpage_io_hook
,
2963 .readpage_end_io_hook
= btrfs_readpage_end_io_hook
,
2964 .readpage_io_failed_hook
= btrfs_io_failed_hook
,
2965 .set_bit_hook
= btrfs_set_bit_hook
,
2966 .clear_bit_hook
= btrfs_clear_bit_hook
,
2969 static struct address_space_operations btrfs_aops
= {
2970 .readpage
= btrfs_readpage
,
2971 .writepage
= btrfs_writepage
,
2972 .writepages
= btrfs_writepages
,
2973 .readpages
= btrfs_readpages
,
2974 .sync_page
= block_sync_page
,
2976 .direct_IO
= btrfs_direct_IO
,
2977 .invalidatepage
= btrfs_invalidatepage
,
2978 .releasepage
= btrfs_releasepage
,
2979 .set_page_dirty
= __set_page_dirty_nobuffers
,
2982 static struct address_space_operations btrfs_symlink_aops
= {
2983 .readpage
= btrfs_readpage
,
2984 .writepage
= btrfs_writepage
,
2985 .invalidatepage
= btrfs_invalidatepage
,
2986 .releasepage
= btrfs_releasepage
,
2989 static struct inode_operations btrfs_file_inode_operations
= {
2990 .truncate
= btrfs_truncate
,
2991 .getattr
= btrfs_getattr
,
2992 .setattr
= btrfs_setattr
,
2993 .setxattr
= generic_setxattr
,
2994 .getxattr
= generic_getxattr
,
2995 .listxattr
= btrfs_listxattr
,
2996 .removexattr
= generic_removexattr
,
2997 .permission
= btrfs_permission
,
2999 static struct inode_operations btrfs_special_inode_operations
= {
3000 .getattr
= btrfs_getattr
,
3001 .setattr
= btrfs_setattr
,
3002 .permission
= btrfs_permission
,
3004 static struct inode_operations btrfs_symlink_inode_operations
= {
3005 .readlink
= generic_readlink
,
3006 .follow_link
= page_follow_link_light
,
3007 .put_link
= page_put_link
,
3008 .permission
= btrfs_permission
,