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
,
82 u64 total
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
83 u64 used
= btrfs_super_bytes_used(&root
->fs_info
->super_copy
);
95 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
96 if (used
+ root
->fs_info
->delalloc_bytes
+ num_required
> thresh
)
98 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
102 static int cow_file_range(struct inode
*inode
, u64 start
, u64 end
)
104 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
105 struct btrfs_trans_handle
*trans
;
109 u64 blocksize
= root
->sectorsize
;
110 u64 orig_start
= start
;
112 struct btrfs_key ins
;
115 trans
= btrfs_start_transaction(root
, 1);
117 btrfs_set_trans_block_group(trans
, inode
);
119 num_bytes
= (end
- start
+ blocksize
) & ~(blocksize
- 1);
120 num_bytes
= max(blocksize
, num_bytes
);
121 ret
= btrfs_drop_extents(trans
, root
, inode
,
122 start
, start
+ num_bytes
, start
, &alloc_hint
);
123 orig_num_bytes
= num_bytes
;
125 if (alloc_hint
== EXTENT_MAP_INLINE
)
128 BUG_ON(num_bytes
> btrfs_super_total_bytes(&root
->fs_info
->super_copy
));
130 while(num_bytes
> 0) {
131 cur_alloc_size
= min(num_bytes
, root
->fs_info
->max_extent
);
132 ret
= btrfs_alloc_extent(trans
, root
, cur_alloc_size
,
134 root
->root_key
.objectid
,
136 inode
->i_ino
, start
, 0,
137 alloc_hint
, (u64
)-1, &ins
, 1);
142 cur_alloc_size
= ins
.offset
;
143 ret
= btrfs_insert_file_extent(trans
, root
, inode
->i_ino
,
144 start
, ins
.objectid
, ins
.offset
,
146 inode
->i_blocks
+= ins
.offset
>> 9;
147 btrfs_check_file(root
, inode
);
148 if (num_bytes
< cur_alloc_size
) {
149 printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes
,
153 num_bytes
-= cur_alloc_size
;
154 alloc_hint
= ins
.objectid
+ ins
.offset
;
155 start
+= cur_alloc_size
;
157 btrfs_drop_extent_cache(inode
, orig_start
,
158 orig_start
+ orig_num_bytes
- 1);
159 btrfs_add_ordered_inode(inode
);
160 btrfs_update_inode(trans
, root
, inode
);
162 btrfs_end_transaction(trans
, root
);
166 static int run_delalloc_nocow(struct inode
*inode
, u64 start
, u64 end
)
174 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
175 struct btrfs_block_group_cache
*block_group
;
176 struct extent_buffer
*leaf
;
178 struct btrfs_path
*path
;
179 struct btrfs_file_extent_item
*item
;
182 struct btrfs_key found_key
;
184 total_fs_bytes
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
185 path
= btrfs_alloc_path();
188 ret
= btrfs_lookup_file_extent(NULL
, root
, path
,
189 inode
->i_ino
, start
, 0);
191 btrfs_free_path(path
);
197 if (path
->slots
[0] == 0)
202 leaf
= path
->nodes
[0];
203 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
204 struct btrfs_file_extent_item
);
206 /* are we inside the extent that was found? */
207 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
208 found_type
= btrfs_key_type(&found_key
);
209 if (found_key
.objectid
!= inode
->i_ino
||
210 found_type
!= BTRFS_EXTENT_DATA_KEY
)
213 found_type
= btrfs_file_extent_type(leaf
, item
);
214 extent_start
= found_key
.offset
;
215 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
216 u64 extent_num_bytes
;
218 extent_num_bytes
= btrfs_file_extent_num_bytes(leaf
, item
);
219 extent_end
= extent_start
+ extent_num_bytes
;
222 if (loops
&& start
!= extent_start
)
225 if (start
< extent_start
|| start
>= extent_end
)
228 cow_end
= min(end
, extent_end
- 1);
229 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
233 if (btrfs_count_snapshots_in_path(root
, path
, inode
->i_ino
,
239 * we may be called by the resizer, make sure we're inside
240 * the limits of the FS
242 block_group
= btrfs_lookup_block_group(root
->fs_info
,
244 if (!block_group
|| block_group
->ro
)
253 btrfs_free_path(path
);
256 btrfs_release_path(root
, path
);
261 cow_file_range(inode
, start
, end
);
266 static int run_delalloc_range(struct inode
*inode
, u64 start
, u64 end
)
268 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
270 mutex_lock(&root
->fs_info
->fs_mutex
);
271 if (btrfs_test_opt(root
, NODATACOW
) ||
272 btrfs_test_flag(inode
, NODATACOW
))
273 ret
= run_delalloc_nocow(inode
, start
, end
);
275 ret
= cow_file_range(inode
, start
, end
);
277 mutex_unlock(&root
->fs_info
->fs_mutex
);
281 int btrfs_set_bit_hook(struct inode
*inode
, u64 start
, u64 end
,
282 unsigned long old
, unsigned long bits
)
285 if (!(old
& EXTENT_DELALLOC
) && (bits
& EXTENT_DELALLOC
)) {
286 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
287 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
288 BTRFS_I(inode
)->delalloc_bytes
+= end
- start
+ 1;
289 root
->fs_info
->delalloc_bytes
+= end
- start
+ 1;
290 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
295 int btrfs_clear_bit_hook(struct inode
*inode
, u64 start
, u64 end
,
296 unsigned long old
, unsigned long bits
)
298 if ((old
& EXTENT_DELALLOC
) && (bits
& EXTENT_DELALLOC
)) {
299 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
302 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
303 if (end
- start
+ 1 > root
->fs_info
->delalloc_bytes
) {
304 printk("warning: delalloc account %Lu %Lu\n",
305 end
- start
+ 1, root
->fs_info
->delalloc_bytes
);
306 root
->fs_info
->delalloc_bytes
= 0;
307 BTRFS_I(inode
)->delalloc_bytes
= 0;
309 root
->fs_info
->delalloc_bytes
-= end
- start
+ 1;
310 BTRFS_I(inode
)->delalloc_bytes
-= end
- start
+ 1;
312 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
317 int btrfs_merge_bio_hook(struct page
*page
, unsigned long offset
,
318 size_t size
, struct bio
*bio
)
320 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
321 struct btrfs_mapping_tree
*map_tree
;
322 u64 logical
= bio
->bi_sector
<< 9;
327 length
= bio
->bi_size
;
328 map_tree
= &root
->fs_info
->mapping_tree
;
330 ret
= btrfs_map_block(map_tree
, READ
, logical
,
331 &map_length
, NULL
, 0);
333 if (map_length
< length
+ size
) {
339 int __btrfs_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
342 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
343 struct btrfs_trans_handle
*trans
;
347 ret
= btrfs_csum_one_bio(root
, bio
, &sums
);
350 mutex_lock(&root
->fs_info
->fs_mutex
);
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
);
358 mutex_unlock(&root
->fs_info
->fs_mutex
);
362 return btrfs_map_bio(root
, rw
, bio
, mirror_num
);
365 int btrfs_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
368 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
371 if (!(rw
& (1 << BIO_RW
))) {
372 ret
= btrfs_bio_wq_end_io(root
->fs_info
, bio
, 0);
377 if (btrfs_test_opt(root
, NODATASUM
) ||
378 btrfs_test_flag(inode
, NODATASUM
)) {
382 return btrfs_wq_submit_bio(BTRFS_I(inode
)->root
->fs_info
,
383 inode
, rw
, bio
, mirror_num
,
384 __btrfs_submit_bio_hook
);
386 return btrfs_map_bio(root
, rw
, bio
, mirror_num
);
389 int btrfs_readpage_io_hook(struct page
*page
, u64 start
, u64 end
)
392 struct inode
*inode
= page
->mapping
->host
;
393 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
394 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
395 struct btrfs_csum_item
*item
;
396 struct btrfs_path
*path
= NULL
;
399 if (btrfs_test_opt(root
, NODATASUM
) ||
400 btrfs_test_flag(inode
, NODATASUM
))
403 mutex_lock(&root
->fs_info
->fs_mutex
);
404 path
= btrfs_alloc_path();
405 item
= btrfs_lookup_csum(NULL
, root
, path
, inode
->i_ino
, start
, 0);
408 /* a csum that isn't present is a preallocated region. */
409 if (ret
== -ENOENT
|| ret
== -EFBIG
)
412 printk("no csum found for inode %lu start %Lu\n", inode
->i_ino
, start
);
415 read_extent_buffer(path
->nodes
[0], &csum
, (unsigned long)item
,
417 set_state_private(io_tree
, start
, csum
);
420 btrfs_free_path(path
);
421 mutex_unlock(&root
->fs_info
->fs_mutex
);
425 struct io_failure_record
{
433 int btrfs_io_failed_hook(struct bio
*failed_bio
,
434 struct page
*page
, u64 start
, u64 end
,
435 struct extent_state
*state
)
437 struct io_failure_record
*failrec
= NULL
;
439 struct extent_map
*em
;
440 struct inode
*inode
= page
->mapping
->host
;
441 struct extent_io_tree
*failure_tree
= &BTRFS_I(inode
)->io_failure_tree
;
442 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
449 ret
= get_state_private(failure_tree
, start
, &private);
451 failrec
= kmalloc(sizeof(*failrec
), GFP_NOFS
);
454 failrec
->start
= start
;
455 failrec
->len
= end
- start
+ 1;
456 failrec
->last_mirror
= 0;
458 spin_lock(&em_tree
->lock
);
459 em
= lookup_extent_mapping(em_tree
, start
, failrec
->len
);
460 if (em
->start
> start
|| em
->start
+ em
->len
< start
) {
464 spin_unlock(&em_tree
->lock
);
466 if (!em
|| IS_ERR(em
)) {
470 logical
= start
- em
->start
;
471 logical
= em
->block_start
+ logical
;
472 failrec
->logical
= logical
;
474 set_extent_bits(failure_tree
, start
, end
, EXTENT_LOCKED
|
475 EXTENT_DIRTY
, GFP_NOFS
);
476 set_state_private(failure_tree
, start
,
477 (u64
)(unsigned long)failrec
);
479 failrec
= (struct io_failure_record
*)(unsigned long)private;
481 num_copies
= btrfs_num_copies(
482 &BTRFS_I(inode
)->root
->fs_info
->mapping_tree
,
483 failrec
->logical
, failrec
->len
);
484 failrec
->last_mirror
++;
486 spin_lock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
487 state
= find_first_extent_bit_state(&BTRFS_I(inode
)->io_tree
,
490 if (state
&& state
->start
!= failrec
->start
)
492 spin_unlock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
494 if (!state
|| failrec
->last_mirror
> num_copies
) {
495 set_state_private(failure_tree
, failrec
->start
, 0);
496 clear_extent_bits(failure_tree
, failrec
->start
,
497 failrec
->start
+ failrec
->len
- 1,
498 EXTENT_LOCKED
| EXTENT_DIRTY
, GFP_NOFS
);
502 bio
= bio_alloc(GFP_NOFS
, 1);
503 bio
->bi_private
= state
;
504 bio
->bi_end_io
= failed_bio
->bi_end_io
;
505 bio
->bi_sector
= failrec
->logical
>> 9;
506 bio
->bi_bdev
= failed_bio
->bi_bdev
;
508 bio_add_page(bio
, page
, failrec
->len
, start
- page_offset(page
));
509 if (failed_bio
->bi_rw
& (1 << BIO_RW
))
514 BTRFS_I(inode
)->io_tree
.ops
->submit_bio_hook(inode
, rw
, bio
,
515 failrec
->last_mirror
);
519 int btrfs_clean_io_failures(struct inode
*inode
, u64 start
)
523 struct io_failure_record
*failure
;
527 if (count_range_bits(&BTRFS_I(inode
)->io_failure_tree
, &private,
528 (u64
)-1, 1, EXTENT_DIRTY
)) {
529 ret
= get_state_private(&BTRFS_I(inode
)->io_failure_tree
,
530 start
, &private_failure
);
532 failure
= (struct io_failure_record
*)(unsigned long)
534 set_state_private(&BTRFS_I(inode
)->io_failure_tree
,
536 clear_extent_bits(&BTRFS_I(inode
)->io_failure_tree
,
538 failure
->start
+ failure
->len
- 1,
539 EXTENT_DIRTY
| EXTENT_LOCKED
,
547 int btrfs_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
548 struct extent_state
*state
)
550 size_t offset
= start
- ((u64
)page
->index
<< PAGE_CACHE_SHIFT
);
551 struct inode
*inode
= page
->mapping
->host
;
552 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
554 u64
private = ~(u32
)0;
556 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
560 if (btrfs_test_opt(root
, NODATASUM
) ||
561 btrfs_test_flag(inode
, NODATASUM
))
563 if (state
&& state
->start
== start
) {
564 private = state
->private;
567 ret
= get_state_private(io_tree
, start
, &private);
569 local_irq_save(flags
);
570 kaddr
= kmap_atomic(page
, KM_IRQ0
);
574 csum
= btrfs_csum_data(root
, kaddr
+ offset
, csum
, end
- start
+ 1);
575 btrfs_csum_final(csum
, (char *)&csum
);
576 if (csum
!= private) {
579 kunmap_atomic(kaddr
, KM_IRQ0
);
580 local_irq_restore(flags
);
582 /* if the io failure tree for this inode is non-empty,
583 * check to see if we've recovered from a failed IO
585 btrfs_clean_io_failures(inode
, start
);
589 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
590 page
->mapping
->host
->i_ino
, (unsigned long long)start
, csum
,
592 memset(kaddr
+ offset
, 1, end
- start
+ 1);
593 flush_dcache_page(page
);
594 kunmap_atomic(kaddr
, KM_IRQ0
);
595 local_irq_restore(flags
);
601 void btrfs_read_locked_inode(struct inode
*inode
)
603 struct btrfs_path
*path
;
604 struct extent_buffer
*leaf
;
605 struct btrfs_inode_item
*inode_item
;
606 struct btrfs_timespec
*tspec
;
607 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
608 struct btrfs_key location
;
609 u64 alloc_group_block
;
613 path
= btrfs_alloc_path();
615 mutex_lock(&root
->fs_info
->fs_mutex
);
616 memcpy(&location
, &BTRFS_I(inode
)->location
, sizeof(location
));
618 ret
= btrfs_lookup_inode(NULL
, root
, path
, &location
, 0);
622 leaf
= path
->nodes
[0];
623 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
624 struct btrfs_inode_item
);
626 inode
->i_mode
= btrfs_inode_mode(leaf
, inode_item
);
627 inode
->i_nlink
= btrfs_inode_nlink(leaf
, inode_item
);
628 inode
->i_uid
= btrfs_inode_uid(leaf
, inode_item
);
629 inode
->i_gid
= btrfs_inode_gid(leaf
, inode_item
);
630 inode
->i_size
= btrfs_inode_size(leaf
, inode_item
);
632 tspec
= btrfs_inode_atime(inode_item
);
633 inode
->i_atime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
634 inode
->i_atime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
636 tspec
= btrfs_inode_mtime(inode_item
);
637 inode
->i_mtime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
638 inode
->i_mtime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
640 tspec
= btrfs_inode_ctime(inode_item
);
641 inode
->i_ctime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
642 inode
->i_ctime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
644 inode
->i_blocks
= btrfs_inode_nblocks(leaf
, inode_item
);
645 inode
->i_generation
= btrfs_inode_generation(leaf
, inode_item
);
647 rdev
= btrfs_inode_rdev(leaf
, inode_item
);
649 alloc_group_block
= btrfs_inode_block_group(leaf
, inode_item
);
650 BTRFS_I(inode
)->block_group
= btrfs_lookup_block_group(root
->fs_info
,
652 BTRFS_I(inode
)->flags
= btrfs_inode_flags(leaf
, inode_item
);
653 if (!BTRFS_I(inode
)->block_group
) {
654 BTRFS_I(inode
)->block_group
= btrfs_find_block_group(root
,
656 BTRFS_BLOCK_GROUP_METADATA
, 0);
658 btrfs_free_path(path
);
661 mutex_unlock(&root
->fs_info
->fs_mutex
);
663 switch (inode
->i_mode
& S_IFMT
) {
665 inode
->i_mapping
->a_ops
= &btrfs_aops
;
666 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
667 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
668 inode
->i_fop
= &btrfs_file_operations
;
669 inode
->i_op
= &btrfs_file_inode_operations
;
672 inode
->i_fop
= &btrfs_dir_file_operations
;
673 if (root
== root
->fs_info
->tree_root
)
674 inode
->i_op
= &btrfs_dir_ro_inode_operations
;
676 inode
->i_op
= &btrfs_dir_inode_operations
;
679 inode
->i_op
= &btrfs_symlink_inode_operations
;
680 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
681 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
684 init_special_inode(inode
, inode
->i_mode
, rdev
);
690 btrfs_release_path(root
, path
);
691 btrfs_free_path(path
);
692 mutex_unlock(&root
->fs_info
->fs_mutex
);
693 make_bad_inode(inode
);
696 static void fill_inode_item(struct extent_buffer
*leaf
,
697 struct btrfs_inode_item
*item
,
700 btrfs_set_inode_uid(leaf
, item
, inode
->i_uid
);
701 btrfs_set_inode_gid(leaf
, item
, inode
->i_gid
);
702 btrfs_set_inode_size(leaf
, item
, inode
->i_size
);
703 btrfs_set_inode_mode(leaf
, item
, inode
->i_mode
);
704 btrfs_set_inode_nlink(leaf
, item
, inode
->i_nlink
);
706 btrfs_set_timespec_sec(leaf
, btrfs_inode_atime(item
),
707 inode
->i_atime
.tv_sec
);
708 btrfs_set_timespec_nsec(leaf
, btrfs_inode_atime(item
),
709 inode
->i_atime
.tv_nsec
);
711 btrfs_set_timespec_sec(leaf
, btrfs_inode_mtime(item
),
712 inode
->i_mtime
.tv_sec
);
713 btrfs_set_timespec_nsec(leaf
, btrfs_inode_mtime(item
),
714 inode
->i_mtime
.tv_nsec
);
716 btrfs_set_timespec_sec(leaf
, btrfs_inode_ctime(item
),
717 inode
->i_ctime
.tv_sec
);
718 btrfs_set_timespec_nsec(leaf
, btrfs_inode_ctime(item
),
719 inode
->i_ctime
.tv_nsec
);
721 btrfs_set_inode_nblocks(leaf
, item
, inode
->i_blocks
);
722 btrfs_set_inode_generation(leaf
, item
, inode
->i_generation
);
723 btrfs_set_inode_rdev(leaf
, item
, inode
->i_rdev
);
724 btrfs_set_inode_flags(leaf
, item
, BTRFS_I(inode
)->flags
);
725 btrfs_set_inode_block_group(leaf
, item
,
726 BTRFS_I(inode
)->block_group
->key
.objectid
);
729 int btrfs_update_inode(struct btrfs_trans_handle
*trans
,
730 struct btrfs_root
*root
,
733 struct btrfs_inode_item
*inode_item
;
734 struct btrfs_path
*path
;
735 struct extent_buffer
*leaf
;
738 path
= btrfs_alloc_path();
740 ret
= btrfs_lookup_inode(trans
, root
, path
,
741 &BTRFS_I(inode
)->location
, 1);
748 leaf
= path
->nodes
[0];
749 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
750 struct btrfs_inode_item
);
752 fill_inode_item(leaf
, inode_item
, inode
);
753 btrfs_mark_buffer_dirty(leaf
);
754 btrfs_set_inode_last_trans(trans
, inode
);
757 btrfs_release_path(root
, path
);
758 btrfs_free_path(path
);
763 static int btrfs_unlink_trans(struct btrfs_trans_handle
*trans
,
764 struct btrfs_root
*root
,
766 struct dentry
*dentry
)
768 struct btrfs_path
*path
;
769 const char *name
= dentry
->d_name
.name
;
770 int name_len
= dentry
->d_name
.len
;
772 struct extent_buffer
*leaf
;
773 struct btrfs_dir_item
*di
;
774 struct btrfs_key key
;
776 path
= btrfs_alloc_path();
782 di
= btrfs_lookup_dir_item(trans
, root
, path
, dir
->i_ino
,
792 leaf
= path
->nodes
[0];
793 btrfs_dir_item_key_to_cpu(leaf
, di
, &key
);
794 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
797 btrfs_release_path(root
, path
);
799 di
= btrfs_lookup_dir_index_item(trans
, root
, path
, dir
->i_ino
,
800 key
.objectid
, name
, name_len
, -1);
809 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
811 dentry
->d_inode
->i_ctime
= dir
->i_ctime
;
812 ret
= btrfs_del_inode_ref(trans
, root
, name
, name_len
,
813 dentry
->d_inode
->i_ino
,
814 dentry
->d_parent
->d_inode
->i_ino
);
816 printk("failed to delete reference to %.*s, "
817 "inode %lu parent %lu\n", name_len
, name
,
818 dentry
->d_inode
->i_ino
,
819 dentry
->d_parent
->d_inode
->i_ino
);
822 btrfs_free_path(path
);
824 dir
->i_size
-= name_len
* 2;
825 dir
->i_mtime
= dir
->i_ctime
= CURRENT_TIME
;
826 btrfs_update_inode(trans
, root
, dir
);
827 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
828 dentry
->d_inode
->i_nlink
--;
830 drop_nlink(dentry
->d_inode
);
832 ret
= btrfs_update_inode(trans
, root
, dentry
->d_inode
);
833 dir
->i_sb
->s_dirt
= 1;
838 static int btrfs_unlink(struct inode
*dir
, struct dentry
*dentry
)
840 struct btrfs_root
*root
;
841 struct btrfs_trans_handle
*trans
;
842 struct inode
*inode
= dentry
->d_inode
;
844 unsigned long nr
= 0;
846 root
= BTRFS_I(dir
)->root
;
847 mutex_lock(&root
->fs_info
->fs_mutex
);
849 ret
= btrfs_check_free_space(root
, 1, 1);
853 trans
= btrfs_start_transaction(root
, 1);
855 btrfs_set_trans_block_group(trans
, dir
);
856 ret
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
857 nr
= trans
->blocks_used
;
859 if (inode
->i_nlink
== 0) {
861 /* if the inode isn't linked anywhere,
862 * we don't need to worry about
865 found
= btrfs_del_ordered_inode(inode
);
867 atomic_dec(&inode
->i_count
);
871 btrfs_end_transaction(trans
, root
);
873 mutex_unlock(&root
->fs_info
->fs_mutex
);
874 btrfs_btree_balance_dirty(root
, nr
);
875 btrfs_throttle(root
);
879 static int btrfs_rmdir(struct inode
*dir
, struct dentry
*dentry
)
881 struct inode
*inode
= dentry
->d_inode
;
884 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
885 struct btrfs_trans_handle
*trans
;
886 unsigned long nr
= 0;
888 if (inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
)
891 mutex_lock(&root
->fs_info
->fs_mutex
);
892 ret
= btrfs_check_free_space(root
, 1, 1);
896 trans
= btrfs_start_transaction(root
, 1);
897 btrfs_set_trans_block_group(trans
, dir
);
899 /* now the directory is empty */
900 err
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
905 nr
= trans
->blocks_used
;
906 ret
= btrfs_end_transaction(trans
, root
);
908 mutex_unlock(&root
->fs_info
->fs_mutex
);
909 btrfs_btree_balance_dirty(root
, nr
);
910 btrfs_throttle(root
);
918 * this can truncate away extent items, csum items and directory items.
919 * It starts at a high offset and removes keys until it can't find
920 * any higher than i_size.
922 * csum items that cross the new i_size are truncated to the new size
925 static int btrfs_truncate_in_trans(struct btrfs_trans_handle
*trans
,
926 struct btrfs_root
*root
,
931 struct btrfs_path
*path
;
932 struct btrfs_key key
;
933 struct btrfs_key found_key
;
935 struct extent_buffer
*leaf
;
936 struct btrfs_file_extent_item
*fi
;
937 u64 extent_start
= 0;
938 u64 extent_num_bytes
= 0;
944 int pending_del_nr
= 0;
945 int pending_del_slot
= 0;
946 int extent_type
= -1;
947 u64 mask
= root
->sectorsize
- 1;
949 btrfs_drop_extent_cache(inode
, inode
->i_size
& (~mask
), (u64
)-1);
950 path
= btrfs_alloc_path();
954 /* FIXME, add redo link to tree so we don't leak on crash */
955 key
.objectid
= inode
->i_ino
;
956 key
.offset
= (u64
)-1;
959 btrfs_init_path(path
);
961 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
966 BUG_ON(path
->slots
[0] == 0);
972 leaf
= path
->nodes
[0];
973 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
974 found_type
= btrfs_key_type(&found_key
);
976 if (found_key
.objectid
!= inode
->i_ino
)
979 if (found_type
< min_type
)
982 item_end
= found_key
.offset
;
983 if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
984 fi
= btrfs_item_ptr(leaf
, path
->slots
[0],
985 struct btrfs_file_extent_item
);
986 extent_type
= btrfs_file_extent_type(leaf
, fi
);
987 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
989 btrfs_file_extent_num_bytes(leaf
, fi
);
990 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
991 struct btrfs_item
*item
= btrfs_item_nr(leaf
,
993 item_end
+= btrfs_file_extent_inline_len(leaf
,
998 if (found_type
== BTRFS_CSUM_ITEM_KEY
) {
999 ret
= btrfs_csum_truncate(trans
, root
, path
,
1003 if (item_end
< inode
->i_size
) {
1004 if (found_type
== BTRFS_DIR_ITEM_KEY
) {
1005 found_type
= BTRFS_INODE_ITEM_KEY
;
1006 } else if (found_type
== BTRFS_EXTENT_ITEM_KEY
) {
1007 found_type
= BTRFS_CSUM_ITEM_KEY
;
1008 } else if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
1009 found_type
= BTRFS_XATTR_ITEM_KEY
;
1010 } else if (found_type
== BTRFS_XATTR_ITEM_KEY
) {
1011 found_type
= BTRFS_INODE_REF_KEY
;
1012 } else if (found_type
) {
1017 btrfs_set_key_type(&key
, found_type
);
1020 if (found_key
.offset
>= inode
->i_size
)
1026 /* FIXME, shrink the extent if the ref count is only 1 */
1027 if (found_type
!= BTRFS_EXTENT_DATA_KEY
)
1030 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1032 extent_start
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
1034 u64 orig_num_bytes
=
1035 btrfs_file_extent_num_bytes(leaf
, fi
);
1036 extent_num_bytes
= inode
->i_size
-
1037 found_key
.offset
+ root
->sectorsize
- 1;
1038 extent_num_bytes
= extent_num_bytes
&
1039 ~((u64
)root
->sectorsize
- 1);
1040 btrfs_set_file_extent_num_bytes(leaf
, fi
,
1042 num_dec
= (orig_num_bytes
-
1044 if (extent_start
!= 0)
1045 dec_i_blocks(inode
, num_dec
);
1046 btrfs_mark_buffer_dirty(leaf
);
1049 btrfs_file_extent_disk_num_bytes(leaf
,
1051 /* FIXME blocksize != 4096 */
1052 num_dec
= btrfs_file_extent_num_bytes(leaf
, fi
);
1053 if (extent_start
!= 0) {
1055 dec_i_blocks(inode
, num_dec
);
1057 root_gen
= btrfs_header_generation(leaf
);
1058 root_owner
= btrfs_header_owner(leaf
);
1060 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1062 u32 newsize
= inode
->i_size
- found_key
.offset
;
1063 dec_i_blocks(inode
, item_end
+ 1 -
1064 found_key
.offset
- newsize
);
1066 btrfs_file_extent_calc_inline_size(newsize
);
1067 ret
= btrfs_truncate_item(trans
, root
, path
,
1071 dec_i_blocks(inode
, item_end
+ 1 -
1077 if (!pending_del_nr
) {
1078 /* no pending yet, add ourselves */
1079 pending_del_slot
= path
->slots
[0];
1081 } else if (pending_del_nr
&&
1082 path
->slots
[0] + 1 == pending_del_slot
) {
1083 /* hop on the pending chunk */
1085 pending_del_slot
= path
->slots
[0];
1087 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path
->slots
[0], pending_del_nr
, pending_del_slot
);
1093 ret
= btrfs_free_extent(trans
, root
, extent_start
,
1096 root_gen
, inode
->i_ino
,
1097 found_key
.offset
, 0);
1101 if (path
->slots
[0] == 0) {
1104 btrfs_release_path(root
, path
);
1109 if (pending_del_nr
&&
1110 path
->slots
[0] + 1 != pending_del_slot
) {
1111 struct btrfs_key debug
;
1113 btrfs_item_key_to_cpu(path
->nodes
[0], &debug
,
1115 ret
= btrfs_del_items(trans
, root
, path
,
1120 btrfs_release_path(root
, path
);
1126 if (pending_del_nr
) {
1127 ret
= btrfs_del_items(trans
, root
, path
, pending_del_slot
,
1130 btrfs_release_path(root
, path
);
1131 btrfs_free_path(path
);
1132 inode
->i_sb
->s_dirt
= 1;
1136 static int btrfs_cow_one_page(struct inode
*inode
, struct page
*page
,
1140 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1141 u64 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
1142 u64 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1145 WARN_ON(!PageLocked(page
));
1146 set_page_extent_mapped(page
);
1148 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1149 set_extent_delalloc(&BTRFS_I(inode
)->io_tree
, page_start
,
1150 page_end
, GFP_NOFS
);
1152 if (zero_start
!= PAGE_CACHE_SIZE
) {
1154 memset(kaddr
+ zero_start
, 0, PAGE_CACHE_SIZE
- zero_start
);
1155 flush_dcache_page(page
);
1158 set_page_dirty(page
);
1159 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1165 * taken from block_truncate_page, but does cow as it zeros out
1166 * any bytes left in the last page in the file.
1168 static int btrfs_truncate_page(struct address_space
*mapping
, loff_t from
)
1170 struct inode
*inode
= mapping
->host
;
1171 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1172 u32 blocksize
= root
->sectorsize
;
1173 pgoff_t index
= from
>> PAGE_CACHE_SHIFT
;
1174 unsigned offset
= from
& (PAGE_CACHE_SIZE
-1);
1179 if ((offset
& (blocksize
- 1)) == 0)
1184 page
= grab_cache_page(mapping
, index
);
1187 if (!PageUptodate(page
)) {
1188 ret
= btrfs_readpage(NULL
, page
);
1190 if (page
->mapping
!= mapping
) {
1192 page_cache_release(page
);
1195 if (!PageUptodate(page
)) {
1201 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
1202 wait_on_page_writeback(page
);
1203 ret
= btrfs_cow_one_page(inode
, page
, offset
);
1206 page_cache_release(page
);
1211 static int btrfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
1213 struct inode
*inode
= dentry
->d_inode
;
1216 err
= inode_change_ok(inode
, attr
);
1220 if (S_ISREG(inode
->i_mode
) &&
1221 attr
->ia_valid
& ATTR_SIZE
&& attr
->ia_size
> inode
->i_size
) {
1222 struct btrfs_trans_handle
*trans
;
1223 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1224 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1226 u64 mask
= root
->sectorsize
- 1;
1227 u64 hole_start
= (inode
->i_size
+ mask
) & ~mask
;
1228 u64 block_end
= (attr
->ia_size
+ mask
) & ~mask
;
1232 if (attr
->ia_size
<= hole_start
)
1235 mutex_lock(&root
->fs_info
->fs_mutex
);
1236 err
= btrfs_check_free_space(root
, 1, 0);
1237 mutex_unlock(&root
->fs_info
->fs_mutex
);
1241 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
1243 lock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1244 hole_size
= block_end
- hole_start
;
1246 mutex_lock(&root
->fs_info
->fs_mutex
);
1247 trans
= btrfs_start_transaction(root
, 1);
1248 btrfs_set_trans_block_group(trans
, inode
);
1249 err
= btrfs_drop_extents(trans
, root
, inode
,
1250 hole_start
, block_end
, hole_start
,
1253 if (alloc_hint
!= EXTENT_MAP_INLINE
) {
1254 err
= btrfs_insert_file_extent(trans
, root
,
1258 btrfs_drop_extent_cache(inode
, hole_start
,
1260 btrfs_check_file(root
, inode
);
1262 btrfs_end_transaction(trans
, root
);
1263 mutex_unlock(&root
->fs_info
->fs_mutex
);
1264 unlock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1269 err
= inode_setattr(inode
, attr
);
1274 void btrfs_put_inode(struct inode
*inode
)
1278 if (!BTRFS_I(inode
)->ordered_trans
) {
1282 if (mapping_tagged(inode
->i_mapping
, PAGECACHE_TAG_DIRTY
) ||
1283 mapping_tagged(inode
->i_mapping
, PAGECACHE_TAG_WRITEBACK
))
1286 ret
= btrfs_del_ordered_inode(inode
);
1288 atomic_dec(&inode
->i_count
);
1292 void btrfs_delete_inode(struct inode
*inode
)
1294 struct btrfs_trans_handle
*trans
;
1295 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1299 truncate_inode_pages(&inode
->i_data
, 0);
1300 if (is_bad_inode(inode
)) {
1305 mutex_lock(&root
->fs_info
->fs_mutex
);
1306 trans
= btrfs_start_transaction(root
, 1);
1308 btrfs_set_trans_block_group(trans
, inode
);
1309 ret
= btrfs_truncate_in_trans(trans
, root
, inode
, 0);
1311 goto no_delete_lock
;
1313 nr
= trans
->blocks_used
;
1316 btrfs_end_transaction(trans
, root
);
1317 mutex_unlock(&root
->fs_info
->fs_mutex
);
1318 btrfs_btree_balance_dirty(root
, nr
);
1319 btrfs_throttle(root
);
1323 nr
= trans
->blocks_used
;
1324 btrfs_end_transaction(trans
, root
);
1325 mutex_unlock(&root
->fs_info
->fs_mutex
);
1326 btrfs_btree_balance_dirty(root
, nr
);
1327 btrfs_throttle(root
);
1333 * this returns the key found in the dir entry in the location pointer.
1334 * If no dir entries were found, location->objectid is 0.
1336 static int btrfs_inode_by_name(struct inode
*dir
, struct dentry
*dentry
,
1337 struct btrfs_key
*location
)
1339 const char *name
= dentry
->d_name
.name
;
1340 int namelen
= dentry
->d_name
.len
;
1341 struct btrfs_dir_item
*di
;
1342 struct btrfs_path
*path
;
1343 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1346 if (namelen
== 1 && strcmp(name
, ".") == 0) {
1347 location
->objectid
= dir
->i_ino
;
1348 location
->type
= BTRFS_INODE_ITEM_KEY
;
1349 location
->offset
= 0;
1352 path
= btrfs_alloc_path();
1355 if (namelen
== 2 && strcmp(name
, "..") == 0) {
1356 struct btrfs_key key
;
1357 struct extent_buffer
*leaf
;
1361 key
.objectid
= dir
->i_ino
;
1362 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1364 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1368 leaf
= path
->nodes
[0];
1369 slot
= path
->slots
[0];
1370 nritems
= btrfs_header_nritems(leaf
);
1371 if (slot
>= nritems
)
1374 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
1375 if (key
.objectid
!= dir
->i_ino
||
1376 key
.type
!= BTRFS_INODE_REF_KEY
) {
1379 location
->objectid
= key
.offset
;
1380 location
->type
= BTRFS_INODE_ITEM_KEY
;
1381 location
->offset
= 0;
1385 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir
->i_ino
, name
,
1389 if (!di
|| IS_ERR(di
)) {
1392 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, location
);
1394 btrfs_free_path(path
);
1397 location
->objectid
= 0;
1402 * when we hit a tree root in a directory, the btrfs part of the inode
1403 * needs to be changed to reflect the root directory of the tree root. This
1404 * is kind of like crossing a mount point.
1406 static int fixup_tree_root_location(struct btrfs_root
*root
,
1407 struct btrfs_key
*location
,
1408 struct btrfs_root
**sub_root
,
1409 struct dentry
*dentry
)
1411 struct btrfs_path
*path
;
1412 struct btrfs_root_item
*ri
;
1414 if (btrfs_key_type(location
) != BTRFS_ROOT_ITEM_KEY
)
1416 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1419 path
= btrfs_alloc_path();
1421 mutex_lock(&root
->fs_info
->fs_mutex
);
1423 *sub_root
= btrfs_read_fs_root(root
->fs_info
, location
,
1424 dentry
->d_name
.name
,
1425 dentry
->d_name
.len
);
1426 if (IS_ERR(*sub_root
))
1427 return PTR_ERR(*sub_root
);
1429 ri
= &(*sub_root
)->root_item
;
1430 location
->objectid
= btrfs_root_dirid(ri
);
1431 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1432 location
->offset
= 0;
1434 btrfs_free_path(path
);
1435 mutex_unlock(&root
->fs_info
->fs_mutex
);
1439 static int btrfs_init_locked_inode(struct inode
*inode
, void *p
)
1441 struct btrfs_iget_args
*args
= p
;
1442 inode
->i_ino
= args
->ino
;
1443 BTRFS_I(inode
)->root
= args
->root
;
1444 BTRFS_I(inode
)->delalloc_bytes
= 0;
1445 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1446 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1447 inode
->i_mapping
, GFP_NOFS
);
1448 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1449 inode
->i_mapping
, GFP_NOFS
);
1450 atomic_set(&BTRFS_I(inode
)->ordered_writeback
, 0);
1454 static int btrfs_find_actor(struct inode
*inode
, void *opaque
)
1456 struct btrfs_iget_args
*args
= opaque
;
1457 return (args
->ino
== inode
->i_ino
&&
1458 args
->root
== BTRFS_I(inode
)->root
);
1461 struct inode
*btrfs_ilookup(struct super_block
*s
, u64 objectid
,
1464 struct btrfs_iget_args args
;
1465 args
.ino
= objectid
;
1466 args
.root
= btrfs_lookup_fs_root(btrfs_sb(s
)->fs_info
, root_objectid
);
1471 return ilookup5(s
, objectid
, btrfs_find_actor
, (void *)&args
);
1474 struct inode
*btrfs_iget_locked(struct super_block
*s
, u64 objectid
,
1475 struct btrfs_root
*root
)
1477 struct inode
*inode
;
1478 struct btrfs_iget_args args
;
1479 args
.ino
= objectid
;
1482 inode
= iget5_locked(s
, objectid
, btrfs_find_actor
,
1483 btrfs_init_locked_inode
,
1488 static struct dentry
*btrfs_lookup(struct inode
*dir
, struct dentry
*dentry
,
1489 struct nameidata
*nd
)
1491 struct inode
* inode
;
1492 struct btrfs_inode
*bi
= BTRFS_I(dir
);
1493 struct btrfs_root
*root
= bi
->root
;
1494 struct btrfs_root
*sub_root
= root
;
1495 struct btrfs_key location
;
1498 if (dentry
->d_name
.len
> BTRFS_NAME_LEN
)
1499 return ERR_PTR(-ENAMETOOLONG
);
1501 mutex_lock(&root
->fs_info
->fs_mutex
);
1502 ret
= btrfs_inode_by_name(dir
, dentry
, &location
);
1503 mutex_unlock(&root
->fs_info
->fs_mutex
);
1506 return ERR_PTR(ret
);
1509 if (location
.objectid
) {
1510 ret
= fixup_tree_root_location(root
, &location
, &sub_root
,
1513 return ERR_PTR(ret
);
1515 return ERR_PTR(-ENOENT
);
1516 inode
= btrfs_iget_locked(dir
->i_sb
, location
.objectid
,
1519 return ERR_PTR(-EACCES
);
1520 if (inode
->i_state
& I_NEW
) {
1521 /* the inode and parent dir are two different roots */
1522 if (sub_root
!= root
) {
1524 sub_root
->inode
= inode
;
1526 BTRFS_I(inode
)->root
= sub_root
;
1527 memcpy(&BTRFS_I(inode
)->location
, &location
,
1529 btrfs_read_locked_inode(inode
);
1530 unlock_new_inode(inode
);
1533 return d_splice_alias(inode
, dentry
);
1536 static unsigned char btrfs_filetype_table
[] = {
1537 DT_UNKNOWN
, DT_REG
, DT_DIR
, DT_CHR
, DT_BLK
, DT_FIFO
, DT_SOCK
, DT_LNK
1540 static int btrfs_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
1542 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1543 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1544 struct btrfs_item
*item
;
1545 struct btrfs_dir_item
*di
;
1546 struct btrfs_key key
;
1547 struct btrfs_key found_key
;
1548 struct btrfs_path
*path
;
1551 struct extent_buffer
*leaf
;
1554 unsigned char d_type
;
1559 int key_type
= BTRFS_DIR_INDEX_KEY
;
1564 /* FIXME, use a real flag for deciding about the key type */
1565 if (root
->fs_info
->tree_root
== root
)
1566 key_type
= BTRFS_DIR_ITEM_KEY
;
1568 /* special case for "." */
1569 if (filp
->f_pos
== 0) {
1570 over
= filldir(dirent
, ".", 1,
1578 mutex_lock(&root
->fs_info
->fs_mutex
);
1579 key
.objectid
= inode
->i_ino
;
1580 path
= btrfs_alloc_path();
1583 /* special case for .., just use the back ref */
1584 if (filp
->f_pos
== 1) {
1585 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1587 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1589 leaf
= path
->nodes
[0];
1590 slot
= path
->slots
[0];
1591 nritems
= btrfs_header_nritems(leaf
);
1592 if (slot
>= nritems
) {
1593 btrfs_release_path(root
, path
);
1594 goto read_dir_items
;
1596 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1597 btrfs_release_path(root
, path
);
1598 if (found_key
.objectid
!= key
.objectid
||
1599 found_key
.type
!= BTRFS_INODE_REF_KEY
)
1600 goto read_dir_items
;
1601 over
= filldir(dirent
, "..", 2,
1602 2, found_key
.offset
, DT_DIR
);
1609 btrfs_set_key_type(&key
, key_type
);
1610 key
.offset
= filp
->f_pos
;
1612 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1617 leaf
= path
->nodes
[0];
1618 nritems
= btrfs_header_nritems(leaf
);
1619 slot
= path
->slots
[0];
1620 if (advance
|| slot
>= nritems
) {
1621 if (slot
>= nritems
-1) {
1622 ret
= btrfs_next_leaf(root
, path
);
1625 leaf
= path
->nodes
[0];
1626 nritems
= btrfs_header_nritems(leaf
);
1627 slot
= path
->slots
[0];
1634 item
= btrfs_item_nr(leaf
, slot
);
1635 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1637 if (found_key
.objectid
!= key
.objectid
)
1639 if (btrfs_key_type(&found_key
) != key_type
)
1641 if (found_key
.offset
< filp
->f_pos
)
1644 filp
->f_pos
= found_key
.offset
;
1646 di
= btrfs_item_ptr(leaf
, slot
, struct btrfs_dir_item
);
1648 di_total
= btrfs_item_size(leaf
, item
);
1649 while(di_cur
< di_total
) {
1650 struct btrfs_key location
;
1652 name_len
= btrfs_dir_name_len(leaf
, di
);
1653 if (name_len
< 32) {
1654 name_ptr
= tmp_name
;
1656 name_ptr
= kmalloc(name_len
, GFP_NOFS
);
1659 read_extent_buffer(leaf
, name_ptr
,
1660 (unsigned long)(di
+ 1), name_len
);
1662 d_type
= btrfs_filetype_table
[btrfs_dir_type(leaf
, di
)];
1663 btrfs_dir_item_key_to_cpu(leaf
, di
, &location
);
1664 over
= filldir(dirent
, name_ptr
, name_len
,
1669 if (name_ptr
!= tmp_name
)
1674 di_len
= btrfs_dir_name_len(leaf
, di
) +
1675 btrfs_dir_data_len(leaf
, di
) +sizeof(*di
);
1677 di
= (struct btrfs_dir_item
*)((char *)di
+ di_len
);
1680 if (key_type
== BTRFS_DIR_INDEX_KEY
)
1681 filp
->f_pos
= INT_LIMIT(typeof(filp
->f_pos
));
1687 btrfs_release_path(root
, path
);
1688 btrfs_free_path(path
);
1689 mutex_unlock(&root
->fs_info
->fs_mutex
);
1693 int btrfs_write_inode(struct inode
*inode
, int wait
)
1695 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1696 struct btrfs_trans_handle
*trans
;
1700 mutex_lock(&root
->fs_info
->fs_mutex
);
1701 trans
= btrfs_start_transaction(root
, 1);
1702 btrfs_set_trans_block_group(trans
, inode
);
1703 ret
= btrfs_commit_transaction(trans
, root
);
1704 mutex_unlock(&root
->fs_info
->fs_mutex
);
1710 * This is somewhat expensive, updating the tree every time the
1711 * inode changes. But, it is most likely to find the inode in cache.
1712 * FIXME, needs more benchmarking...there are no reasons other than performance
1713 * to keep or drop this code.
1715 void btrfs_dirty_inode(struct inode
*inode
)
1717 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1718 struct btrfs_trans_handle
*trans
;
1720 mutex_lock(&root
->fs_info
->fs_mutex
);
1721 trans
= btrfs_start_transaction(root
, 1);
1722 btrfs_set_trans_block_group(trans
, inode
);
1723 btrfs_update_inode(trans
, root
, inode
);
1724 btrfs_end_transaction(trans
, root
);
1725 mutex_unlock(&root
->fs_info
->fs_mutex
);
1728 static struct inode
*btrfs_new_inode(struct btrfs_trans_handle
*trans
,
1729 struct btrfs_root
*root
,
1730 const char *name
, int name_len
,
1733 struct btrfs_block_group_cache
*group
,
1736 struct inode
*inode
;
1737 struct btrfs_inode_item
*inode_item
;
1738 struct btrfs_block_group_cache
*new_inode_group
;
1739 struct btrfs_key
*location
;
1740 struct btrfs_path
*path
;
1741 struct btrfs_inode_ref
*ref
;
1742 struct btrfs_key key
[2];
1748 path
= btrfs_alloc_path();
1751 inode
= new_inode(root
->fs_info
->sb
);
1753 return ERR_PTR(-ENOMEM
);
1755 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1756 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1757 inode
->i_mapping
, GFP_NOFS
);
1758 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1759 inode
->i_mapping
, GFP_NOFS
);
1760 atomic_set(&BTRFS_I(inode
)->ordered_writeback
, 0);
1761 BTRFS_I(inode
)->delalloc_bytes
= 0;
1762 BTRFS_I(inode
)->root
= root
;
1768 new_inode_group
= btrfs_find_block_group(root
, group
, 0,
1769 BTRFS_BLOCK_GROUP_METADATA
, owner
);
1770 if (!new_inode_group
) {
1771 printk("find_block group failed\n");
1772 new_inode_group
= group
;
1774 BTRFS_I(inode
)->block_group
= new_inode_group
;
1775 BTRFS_I(inode
)->flags
= 0;
1777 key
[0].objectid
= objectid
;
1778 btrfs_set_key_type(&key
[0], BTRFS_INODE_ITEM_KEY
);
1781 key
[1].objectid
= objectid
;
1782 btrfs_set_key_type(&key
[1], BTRFS_INODE_REF_KEY
);
1783 key
[1].offset
= ref_objectid
;
1785 sizes
[0] = sizeof(struct btrfs_inode_item
);
1786 sizes
[1] = name_len
+ sizeof(*ref
);
1788 ret
= btrfs_insert_empty_items(trans
, root
, path
, key
, sizes
, 2);
1792 if (objectid
> root
->highest_inode
)
1793 root
->highest_inode
= objectid
;
1795 inode
->i_uid
= current
->fsuid
;
1796 inode
->i_gid
= current
->fsgid
;
1797 inode
->i_mode
= mode
;
1798 inode
->i_ino
= objectid
;
1799 inode
->i_blocks
= 0;
1800 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1801 inode_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
1802 struct btrfs_inode_item
);
1803 fill_inode_item(path
->nodes
[0], inode_item
, inode
);
1805 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0] + 1,
1806 struct btrfs_inode_ref
);
1807 btrfs_set_inode_ref_name_len(path
->nodes
[0], ref
, name_len
);
1808 ptr
= (unsigned long)(ref
+ 1);
1809 write_extent_buffer(path
->nodes
[0], name
, ptr
, name_len
);
1811 btrfs_mark_buffer_dirty(path
->nodes
[0]);
1812 btrfs_free_path(path
);
1814 location
= &BTRFS_I(inode
)->location
;
1815 location
->objectid
= objectid
;
1816 location
->offset
= 0;
1817 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1819 insert_inode_hash(inode
);
1822 btrfs_free_path(path
);
1823 return ERR_PTR(ret
);
1826 static inline u8
btrfs_inode_type(struct inode
*inode
)
1828 return btrfs_type_by_mode
[(inode
->i_mode
& S_IFMT
) >> S_SHIFT
];
1831 static int btrfs_add_link(struct btrfs_trans_handle
*trans
,
1832 struct dentry
*dentry
, struct inode
*inode
,
1836 struct btrfs_key key
;
1837 struct btrfs_root
*root
= BTRFS_I(dentry
->d_parent
->d_inode
)->root
;
1838 struct inode
*parent_inode
;
1840 key
.objectid
= inode
->i_ino
;
1841 btrfs_set_key_type(&key
, BTRFS_INODE_ITEM_KEY
);
1844 ret
= btrfs_insert_dir_item(trans
, root
,
1845 dentry
->d_name
.name
, dentry
->d_name
.len
,
1846 dentry
->d_parent
->d_inode
->i_ino
,
1847 &key
, btrfs_inode_type(inode
));
1850 ret
= btrfs_insert_inode_ref(trans
, root
,
1851 dentry
->d_name
.name
,
1854 dentry
->d_parent
->d_inode
->i_ino
);
1856 parent_inode
= dentry
->d_parent
->d_inode
;
1857 parent_inode
->i_size
+= dentry
->d_name
.len
* 2;
1858 parent_inode
->i_mtime
= parent_inode
->i_ctime
= CURRENT_TIME
;
1859 ret
= btrfs_update_inode(trans
, root
,
1860 dentry
->d_parent
->d_inode
);
1865 static int btrfs_add_nondir(struct btrfs_trans_handle
*trans
,
1866 struct dentry
*dentry
, struct inode
*inode
,
1869 int err
= btrfs_add_link(trans
, dentry
, inode
, backref
);
1871 d_instantiate(dentry
, inode
);
1879 static int btrfs_mknod(struct inode
*dir
, struct dentry
*dentry
,
1880 int mode
, dev_t rdev
)
1882 struct btrfs_trans_handle
*trans
;
1883 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1884 struct inode
*inode
= NULL
;
1888 unsigned long nr
= 0;
1890 if (!new_valid_dev(rdev
))
1893 mutex_lock(&root
->fs_info
->fs_mutex
);
1894 err
= btrfs_check_free_space(root
, 1, 0);
1898 trans
= btrfs_start_transaction(root
, 1);
1899 btrfs_set_trans_block_group(trans
, dir
);
1901 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
1907 inode
= btrfs_new_inode(trans
, root
, dentry
->d_name
.name
,
1909 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
1910 BTRFS_I(dir
)->block_group
, mode
);
1911 err
= PTR_ERR(inode
);
1915 btrfs_set_trans_block_group(trans
, inode
);
1916 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
1920 inode
->i_op
= &btrfs_special_inode_operations
;
1921 init_special_inode(inode
, inode
->i_mode
, rdev
);
1922 btrfs_update_inode(trans
, root
, inode
);
1924 dir
->i_sb
->s_dirt
= 1;
1925 btrfs_update_inode_block_group(trans
, inode
);
1926 btrfs_update_inode_block_group(trans
, dir
);
1928 nr
= trans
->blocks_used
;
1929 btrfs_end_transaction(trans
, root
);
1931 mutex_unlock(&root
->fs_info
->fs_mutex
);
1934 inode_dec_link_count(inode
);
1937 btrfs_btree_balance_dirty(root
, nr
);
1938 btrfs_throttle(root
);
1942 static int btrfs_create(struct inode
*dir
, struct dentry
*dentry
,
1943 int mode
, struct nameidata
*nd
)
1945 struct btrfs_trans_handle
*trans
;
1946 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1947 struct inode
*inode
= NULL
;
1950 unsigned long nr
= 0;
1953 mutex_lock(&root
->fs_info
->fs_mutex
);
1954 err
= btrfs_check_free_space(root
, 1, 0);
1957 trans
= btrfs_start_transaction(root
, 1);
1958 btrfs_set_trans_block_group(trans
, dir
);
1960 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
1966 inode
= btrfs_new_inode(trans
, root
, dentry
->d_name
.name
,
1968 dentry
->d_parent
->d_inode
->i_ino
,
1969 objectid
, BTRFS_I(dir
)->block_group
, mode
);
1970 err
= PTR_ERR(inode
);
1974 btrfs_set_trans_block_group(trans
, inode
);
1975 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
1979 inode
->i_mapping
->a_ops
= &btrfs_aops
;
1980 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1981 inode
->i_fop
= &btrfs_file_operations
;
1982 inode
->i_op
= &btrfs_file_inode_operations
;
1983 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1984 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1985 inode
->i_mapping
, GFP_NOFS
);
1986 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1987 inode
->i_mapping
, GFP_NOFS
);
1988 BTRFS_I(inode
)->delalloc_bytes
= 0;
1989 atomic_set(&BTRFS_I(inode
)->ordered_writeback
, 0);
1990 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
1992 dir
->i_sb
->s_dirt
= 1;
1993 btrfs_update_inode_block_group(trans
, inode
);
1994 btrfs_update_inode_block_group(trans
, dir
);
1996 nr
= trans
->blocks_used
;
1997 btrfs_end_transaction(trans
, root
);
1999 mutex_unlock(&root
->fs_info
->fs_mutex
);
2002 inode_dec_link_count(inode
);
2005 btrfs_btree_balance_dirty(root
, nr
);
2006 btrfs_throttle(root
);
2010 static int btrfs_link(struct dentry
*old_dentry
, struct inode
*dir
,
2011 struct dentry
*dentry
)
2013 struct btrfs_trans_handle
*trans
;
2014 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2015 struct inode
*inode
= old_dentry
->d_inode
;
2016 unsigned long nr
= 0;
2020 if (inode
->i_nlink
== 0)
2023 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
2028 mutex_lock(&root
->fs_info
->fs_mutex
);
2029 err
= btrfs_check_free_space(root
, 1, 0);
2032 trans
= btrfs_start_transaction(root
, 1);
2034 btrfs_set_trans_block_group(trans
, dir
);
2035 atomic_inc(&inode
->i_count
);
2036 err
= btrfs_add_nondir(trans
, dentry
, inode
, 1);
2041 dir
->i_sb
->s_dirt
= 1;
2042 btrfs_update_inode_block_group(trans
, dir
);
2043 err
= btrfs_update_inode(trans
, root
, inode
);
2048 nr
= trans
->blocks_used
;
2049 btrfs_end_transaction(trans
, root
);
2051 mutex_unlock(&root
->fs_info
->fs_mutex
);
2054 inode_dec_link_count(inode
);
2057 btrfs_btree_balance_dirty(root
, nr
);
2058 btrfs_throttle(root
);
2062 static int btrfs_mkdir(struct inode
*dir
, struct dentry
*dentry
, int mode
)
2064 struct inode
*inode
= NULL
;
2065 struct btrfs_trans_handle
*trans
;
2066 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2068 int drop_on_err
= 0;
2070 unsigned long nr
= 1;
2072 mutex_lock(&root
->fs_info
->fs_mutex
);
2073 err
= btrfs_check_free_space(root
, 1, 0);
2077 trans
= btrfs_start_transaction(root
, 1);
2078 btrfs_set_trans_block_group(trans
, dir
);
2080 if (IS_ERR(trans
)) {
2081 err
= PTR_ERR(trans
);
2085 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2091 inode
= btrfs_new_inode(trans
, root
, dentry
->d_name
.name
,
2093 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2094 BTRFS_I(dir
)->block_group
, S_IFDIR
| mode
);
2095 if (IS_ERR(inode
)) {
2096 err
= PTR_ERR(inode
);
2101 inode
->i_op
= &btrfs_dir_inode_operations
;
2102 inode
->i_fop
= &btrfs_dir_file_operations
;
2103 btrfs_set_trans_block_group(trans
, inode
);
2106 err
= btrfs_update_inode(trans
, root
, inode
);
2110 err
= btrfs_add_link(trans
, dentry
, inode
, 0);
2114 d_instantiate(dentry
, inode
);
2116 dir
->i_sb
->s_dirt
= 1;
2117 btrfs_update_inode_block_group(trans
, inode
);
2118 btrfs_update_inode_block_group(trans
, dir
);
2121 nr
= trans
->blocks_used
;
2122 btrfs_end_transaction(trans
, root
);
2125 mutex_unlock(&root
->fs_info
->fs_mutex
);
2128 btrfs_btree_balance_dirty(root
, nr
);
2129 btrfs_throttle(root
);
2133 static int merge_extent_mapping(struct extent_map_tree
*em_tree
,
2134 struct extent_map
*existing
,
2135 struct extent_map
*em
)
2140 int real_blocks
= existing
->block_start
< EXTENT_MAP_LAST_BYTE
;
2142 if (real_blocks
&& em
->block_start
>= EXTENT_MAP_LAST_BYTE
)
2145 if (!real_blocks
&& em
->block_start
!= existing
->block_start
)
2148 new_end
= max(existing
->start
+ existing
->len
, em
->start
+ em
->len
);
2150 if (existing
->start
>= em
->start
) {
2151 if (em
->start
+ em
->len
< existing
->start
)
2154 start_diff
= existing
->start
- em
->start
;
2155 if (real_blocks
&& em
->block_start
+ start_diff
!=
2156 existing
->block_start
)
2159 em
->len
= new_end
- em
->start
;
2161 remove_extent_mapping(em_tree
, existing
);
2162 /* free for the tree */
2163 free_extent_map(existing
);
2164 ret
= add_extent_mapping(em_tree
, em
);
2166 } else if (em
->start
> existing
->start
) {
2168 if (existing
->start
+ existing
->len
< em
->start
)
2171 start_diff
= em
->start
- existing
->start
;
2172 if (real_blocks
&& existing
->block_start
+ start_diff
!=
2176 remove_extent_mapping(em_tree
, existing
);
2177 em
->block_start
= existing
->block_start
;
2178 em
->start
= existing
->start
;
2179 em
->len
= new_end
- existing
->start
;
2180 free_extent_map(existing
);
2182 ret
= add_extent_mapping(em_tree
, em
);
2189 printk("invalid extent map merge [%Lu %Lu %Lu] [%Lu %Lu %Lu]\n",
2190 existing
->start
, existing
->len
, existing
->block_start
,
2191 em
->start
, em
->len
, em
->block_start
);
2195 struct extent_map
*btrfs_get_extent(struct inode
*inode
, struct page
*page
,
2196 size_t pg_offset
, u64 start
, u64 len
,
2202 u64 extent_start
= 0;
2204 u64 objectid
= inode
->i_ino
;
2206 struct btrfs_path
*path
;
2207 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2208 struct btrfs_file_extent_item
*item
;
2209 struct extent_buffer
*leaf
;
2210 struct btrfs_key found_key
;
2211 struct extent_map
*em
= NULL
;
2212 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
2213 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2214 struct btrfs_trans_handle
*trans
= NULL
;
2216 path
= btrfs_alloc_path();
2218 mutex_lock(&root
->fs_info
->fs_mutex
);
2221 spin_lock(&em_tree
->lock
);
2222 em
= lookup_extent_mapping(em_tree
, start
, len
);
2224 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2225 spin_unlock(&em_tree
->lock
);
2228 if (em
->start
> start
|| em
->start
+ em
->len
<= start
)
2229 free_extent_map(em
);
2230 else if (em
->block_start
== EXTENT_MAP_INLINE
&& page
)
2231 free_extent_map(em
);
2235 em
= alloc_extent_map(GFP_NOFS
);
2241 em
->start
= EXTENT_MAP_HOLE
;
2243 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2244 ret
= btrfs_lookup_file_extent(trans
, root
, path
,
2245 objectid
, start
, trans
!= NULL
);
2252 if (path
->slots
[0] == 0)
2257 leaf
= path
->nodes
[0];
2258 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
2259 struct btrfs_file_extent_item
);
2260 /* are we inside the extent that was found? */
2261 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2262 found_type
= btrfs_key_type(&found_key
);
2263 if (found_key
.objectid
!= objectid
||
2264 found_type
!= BTRFS_EXTENT_DATA_KEY
) {
2268 found_type
= btrfs_file_extent_type(leaf
, item
);
2269 extent_start
= found_key
.offset
;
2270 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
2271 extent_end
= extent_start
+
2272 btrfs_file_extent_num_bytes(leaf
, item
);
2274 if (start
< extent_start
|| start
>= extent_end
) {
2276 if (start
< extent_start
) {
2277 if (start
+ len
<= extent_start
)
2279 em
->len
= extent_end
- extent_start
;
2285 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
2287 em
->start
= extent_start
;
2288 em
->len
= extent_end
- extent_start
;
2289 em
->block_start
= EXTENT_MAP_HOLE
;
2292 bytenr
+= btrfs_file_extent_offset(leaf
, item
);
2293 em
->block_start
= bytenr
;
2294 em
->start
= extent_start
;
2295 em
->len
= extent_end
- extent_start
;
2297 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
2302 size_t extent_offset
;
2305 size
= btrfs_file_extent_inline_len(leaf
, btrfs_item_nr(leaf
,
2307 extent_end
= (extent_start
+ size
+ root
->sectorsize
- 1) &
2308 ~((u64
)root
->sectorsize
- 1);
2309 if (start
< extent_start
|| start
>= extent_end
) {
2311 if (start
< extent_start
) {
2312 if (start
+ len
<= extent_start
)
2314 em
->len
= extent_end
- extent_start
;
2320 em
->block_start
= EXTENT_MAP_INLINE
;
2323 em
->start
= extent_start
;
2328 page_start
= page_offset(page
) + pg_offset
;
2329 extent_offset
= page_start
- extent_start
;
2330 copy_size
= min_t(u64
, PAGE_CACHE_SIZE
- pg_offset
,
2331 size
- extent_offset
);
2332 em
->start
= extent_start
+ extent_offset
;
2333 em
->len
= (copy_size
+ root
->sectorsize
- 1) &
2334 ~((u64
)root
->sectorsize
- 1);
2336 ptr
= btrfs_file_extent_inline_start(item
) + extent_offset
;
2337 if (create
== 0 && !PageUptodate(page
)) {
2338 read_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2340 flush_dcache_page(page
);
2341 } else if (create
&& PageUptodate(page
)) {
2344 free_extent_map(em
);
2346 btrfs_release_path(root
, path
);
2347 trans
= btrfs_start_transaction(root
, 1);
2350 write_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2352 btrfs_mark_buffer_dirty(leaf
);
2355 set_extent_uptodate(io_tree
, em
->start
,
2356 extent_map_end(em
) - 1, GFP_NOFS
);
2359 printk("unkknown found_type %d\n", found_type
);
2366 em
->block_start
= EXTENT_MAP_HOLE
;
2368 btrfs_release_path(root
, path
);
2369 if (em
->start
> start
|| extent_map_end(em
) <= start
) {
2370 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em
->start
, em
->len
, start
, len
);
2376 spin_lock(&em_tree
->lock
);
2377 ret
= add_extent_mapping(em_tree
, em
);
2378 /* it is possible that someone inserted the extent into the tree
2379 * while we had the lock dropped. It is also possible that
2380 * an overlapping map exists in the tree
2382 if (ret
== -EEXIST
) {
2383 struct extent_map
*existing
;
2384 existing
= lookup_extent_mapping(em_tree
, start
, len
);
2385 if (existing
&& (existing
->start
> start
||
2386 existing
->start
+ existing
->len
<= start
)) {
2387 free_extent_map(existing
);
2391 existing
= lookup_extent_mapping(em_tree
, em
->start
,
2394 err
= merge_extent_mapping(em_tree
, existing
,
2396 free_extent_map(existing
);
2398 free_extent_map(em
);
2403 printk("failing to insert %Lu %Lu\n",
2405 free_extent_map(em
);
2409 free_extent_map(em
);
2413 spin_unlock(&em_tree
->lock
);
2415 btrfs_free_path(path
);
2417 ret
= btrfs_end_transaction(trans
, root
);
2421 mutex_unlock(&root
->fs_info
->fs_mutex
);
2423 free_extent_map(em
);
2425 return ERR_PTR(err
);
2430 #if 0 /* waiting for O_DIRECT reads */
2431 static int btrfs_get_block(struct inode
*inode
, sector_t iblock
,
2432 struct buffer_head
*bh_result
, int create
)
2434 struct extent_map
*em
;
2435 u64 start
= (u64
)iblock
<< inode
->i_blkbits
;
2436 struct btrfs_multi_bio
*multi
= NULL
;
2437 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2443 em
= btrfs_get_extent(inode
, NULL
, 0, start
, bh_result
->b_size
, 0);
2445 if (!em
|| IS_ERR(em
))
2448 if (em
->start
> start
|| em
->start
+ em
->len
<= start
) {
2452 if (em
->block_start
== EXTENT_MAP_INLINE
) {
2457 len
= em
->start
+ em
->len
- start
;
2458 len
= min_t(u64
, len
, INT_LIMIT(typeof(bh_result
->b_size
)));
2460 if (em
->block_start
== EXTENT_MAP_HOLE
||
2461 em
->block_start
== EXTENT_MAP_DELALLOC
) {
2462 bh_result
->b_size
= len
;
2466 logical
= start
- em
->start
;
2467 logical
= em
->block_start
+ logical
;
2470 ret
= btrfs_map_block(&root
->fs_info
->mapping_tree
, READ
,
2471 logical
, &map_length
, &multi
, 0);
2473 bh_result
->b_blocknr
= multi
->stripes
[0].physical
>> inode
->i_blkbits
;
2474 bh_result
->b_size
= min(map_length
, len
);
2476 bh_result
->b_bdev
= multi
->stripes
[0].dev
->bdev
;
2477 set_buffer_mapped(bh_result
);
2480 free_extent_map(em
);
2485 static ssize_t
btrfs_direct_IO(int rw
, struct kiocb
*iocb
,
2486 const struct iovec
*iov
, loff_t offset
,
2487 unsigned long nr_segs
)
2491 struct file
*file
= iocb
->ki_filp
;
2492 struct inode
*inode
= file
->f_mapping
->host
;
2497 return blockdev_direct_IO(rw
, iocb
, inode
, inode
->i_sb
->s_bdev
, iov
,
2498 offset
, nr_segs
, btrfs_get_block
, NULL
);
2502 static sector_t
btrfs_bmap(struct address_space
*mapping
, sector_t iblock
)
2504 return extent_bmap(mapping
, iblock
, btrfs_get_extent
);
2507 int btrfs_readpage(struct file
*file
, struct page
*page
)
2509 struct extent_io_tree
*tree
;
2510 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2511 return extent_read_full_page(tree
, page
, btrfs_get_extent
);
2514 static int btrfs_writepage(struct page
*page
, struct writeback_control
*wbc
)
2516 struct extent_io_tree
*tree
;
2519 if (current
->flags
& PF_MEMALLOC
) {
2520 redirty_page_for_writepage(wbc
, page
);
2524 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2525 return extent_write_full_page(tree
, page
, btrfs_get_extent
, wbc
);
2528 static int btrfs_writepages(struct address_space
*mapping
,
2529 struct writeback_control
*wbc
)
2531 struct extent_io_tree
*tree
;
2532 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
2533 return extent_writepages(tree
, mapping
, btrfs_get_extent
, wbc
);
2537 btrfs_readpages(struct file
*file
, struct address_space
*mapping
,
2538 struct list_head
*pages
, unsigned nr_pages
)
2540 struct extent_io_tree
*tree
;
2541 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
2542 return extent_readpages(tree
, mapping
, pages
, nr_pages
,
2546 static int btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
2548 struct extent_io_tree
*tree
;
2549 struct extent_map_tree
*map
;
2552 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2553 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
2554 ret
= try_release_extent_mapping(map
, tree
, page
, gfp_flags
);
2556 invalidate_extent_lru(tree
, page_offset(page
), PAGE_CACHE_SIZE
);
2557 ClearPagePrivate(page
);
2558 set_page_private(page
, 0);
2559 page_cache_release(page
);
2564 static void btrfs_invalidatepage(struct page
*page
, unsigned long offset
)
2566 struct extent_io_tree
*tree
;
2568 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2569 extent_invalidatepage(tree
, page
, offset
);
2570 btrfs_releasepage(page
, GFP_NOFS
);
2571 if (PagePrivate(page
)) {
2572 invalidate_extent_lru(tree
, page_offset(page
), PAGE_CACHE_SIZE
);
2573 ClearPagePrivate(page
);
2574 set_page_private(page
, 0);
2575 page_cache_release(page
);
2580 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
2581 * called from a page fault handler when a page is first dirtied. Hence we must
2582 * be careful to check for EOF conditions here. We set the page up correctly
2583 * for a written page which means we get ENOSPC checking when writing into
2584 * holes and correct delalloc and unwritten extent mapping on filesystems that
2585 * support these features.
2587 * We are not allowed to take the i_mutex here so we have to play games to
2588 * protect against truncate races as the page could now be beyond EOF. Because
2589 * vmtruncate() writes the inode size before removing pages, once we have the
2590 * page lock we can determine safely if the page is beyond EOF. If it is not
2591 * beyond EOF, then the page is guaranteed safe against truncation until we
2594 int btrfs_page_mkwrite(struct vm_area_struct
*vma
, struct page
*page
)
2596 struct inode
*inode
= fdentry(vma
->vm_file
)->d_inode
;
2597 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2603 mutex_lock(&root
->fs_info
->fs_mutex
);
2604 ret
= btrfs_check_free_space(root
, PAGE_CACHE_SIZE
, 0);
2605 mutex_unlock(&root
->fs_info
->fs_mutex
);
2612 wait_on_page_writeback(page
);
2613 size
= i_size_read(inode
);
2614 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
2616 if ((page
->mapping
!= inode
->i_mapping
) ||
2617 (page_start
> size
)) {
2618 /* page got truncated out from underneath us */
2622 /* page is wholly or partially inside EOF */
2623 if (page_start
+ PAGE_CACHE_SIZE
> size
)
2624 end
= size
& ~PAGE_CACHE_MASK
;
2626 end
= PAGE_CACHE_SIZE
;
2628 ret
= btrfs_cow_one_page(inode
, page
, end
);
2636 static void btrfs_truncate(struct inode
*inode
)
2638 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2640 struct btrfs_trans_handle
*trans
;
2643 if (!S_ISREG(inode
->i_mode
))
2645 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
2648 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
2650 mutex_lock(&root
->fs_info
->fs_mutex
);
2651 trans
= btrfs_start_transaction(root
, 1);
2652 btrfs_set_trans_block_group(trans
, inode
);
2654 /* FIXME, add redo link to tree so we don't leak on crash */
2655 ret
= btrfs_truncate_in_trans(trans
, root
, inode
,
2656 BTRFS_EXTENT_DATA_KEY
);
2657 btrfs_update_inode(trans
, root
, inode
);
2658 nr
= trans
->blocks_used
;
2660 ret
= btrfs_end_transaction(trans
, root
);
2662 mutex_unlock(&root
->fs_info
->fs_mutex
);
2663 btrfs_btree_balance_dirty(root
, nr
);
2664 btrfs_throttle(root
);
2667 static int noinline
create_subvol(struct btrfs_root
*root
, char *name
,
2670 struct btrfs_trans_handle
*trans
;
2671 struct btrfs_key key
;
2672 struct btrfs_root_item root_item
;
2673 struct btrfs_inode_item
*inode_item
;
2674 struct extent_buffer
*leaf
;
2675 struct btrfs_root
*new_root
= root
;
2676 struct inode
*inode
;
2681 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
2682 unsigned long nr
= 1;
2684 mutex_lock(&root
->fs_info
->fs_mutex
);
2685 ret
= btrfs_check_free_space(root
, 1, 0);
2689 trans
= btrfs_start_transaction(root
, 1);
2692 ret
= btrfs_find_free_objectid(trans
, root
->fs_info
->tree_root
,
2697 leaf
= __btrfs_alloc_free_block(trans
, root
, root
->leafsize
,
2698 objectid
, trans
->transid
, 0, 0,
2701 return PTR_ERR(leaf
);
2703 btrfs_set_header_nritems(leaf
, 0);
2704 btrfs_set_header_level(leaf
, 0);
2705 btrfs_set_header_bytenr(leaf
, leaf
->start
);
2706 btrfs_set_header_generation(leaf
, trans
->transid
);
2707 btrfs_set_header_owner(leaf
, objectid
);
2709 write_extent_buffer(leaf
, root
->fs_info
->fsid
,
2710 (unsigned long)btrfs_header_fsid(leaf
),
2712 btrfs_mark_buffer_dirty(leaf
);
2714 inode_item
= &root_item
.inode
;
2715 memset(inode_item
, 0, sizeof(*inode_item
));
2716 inode_item
->generation
= cpu_to_le64(1);
2717 inode_item
->size
= cpu_to_le64(3);
2718 inode_item
->nlink
= cpu_to_le32(1);
2719 inode_item
->nblocks
= cpu_to_le64(1);
2720 inode_item
->mode
= cpu_to_le32(S_IFDIR
| 0755);
2722 btrfs_set_root_bytenr(&root_item
, leaf
->start
);
2723 btrfs_set_root_level(&root_item
, 0);
2724 btrfs_set_root_refs(&root_item
, 1);
2725 btrfs_set_root_used(&root_item
, 0);
2727 memset(&root_item
.drop_progress
, 0, sizeof(root_item
.drop_progress
));
2728 root_item
.drop_level
= 0;
2730 free_extent_buffer(leaf
);
2733 btrfs_set_root_dirid(&root_item
, new_dirid
);
2735 key
.objectid
= objectid
;
2737 btrfs_set_key_type(&key
, BTRFS_ROOT_ITEM_KEY
);
2738 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
2744 * insert the directory item
2746 key
.offset
= (u64
)-1;
2747 dir
= root
->fs_info
->sb
->s_root
->d_inode
;
2748 ret
= btrfs_insert_dir_item(trans
, root
->fs_info
->tree_root
,
2749 name
, namelen
, dir
->i_ino
, &key
,
2754 ret
= btrfs_insert_inode_ref(trans
, root
->fs_info
->tree_root
,
2755 name
, namelen
, objectid
,
2756 root
->fs_info
->sb
->s_root
->d_inode
->i_ino
);
2760 ret
= btrfs_commit_transaction(trans
, root
);
2764 new_root
= btrfs_read_fs_root(root
->fs_info
, &key
, name
, namelen
);
2767 trans
= btrfs_start_transaction(new_root
, 1);
2770 inode
= btrfs_new_inode(trans
, new_root
, "..", 2, new_dirid
,
2772 BTRFS_I(dir
)->block_group
, S_IFDIR
| 0700);
2775 inode
->i_op
= &btrfs_dir_inode_operations
;
2776 inode
->i_fop
= &btrfs_dir_file_operations
;
2777 new_root
->inode
= inode
;
2779 ret
= btrfs_insert_inode_ref(trans
, new_root
, "..", 2, new_dirid
,
2783 ret
= btrfs_update_inode(trans
, new_root
, inode
);
2787 nr
= trans
->blocks_used
;
2788 err
= btrfs_commit_transaction(trans
, new_root
);
2792 mutex_unlock(&root
->fs_info
->fs_mutex
);
2793 btrfs_btree_balance_dirty(root
, nr
);
2794 btrfs_throttle(root
);
2798 static int create_snapshot(struct btrfs_root
*root
, char *name
, int namelen
)
2800 struct btrfs_pending_snapshot
*pending_snapshot
;
2801 struct btrfs_trans_handle
*trans
;
2804 unsigned long nr
= 0;
2806 if (!root
->ref_cows
)
2809 mutex_lock(&root
->fs_info
->fs_mutex
);
2810 ret
= btrfs_check_free_space(root
, 1, 0);
2814 pending_snapshot
= kmalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
2815 if (!pending_snapshot
) {
2819 pending_snapshot
->name
= kmalloc(namelen
+ 1, GFP_NOFS
);
2820 if (!pending_snapshot
->name
) {
2822 kfree(pending_snapshot
);
2825 memcpy(pending_snapshot
->name
, name
, namelen
);
2826 pending_snapshot
->name
[namelen
] = '\0';
2827 trans
= btrfs_start_transaction(root
, 1);
2829 pending_snapshot
->root
= root
;
2830 list_add(&pending_snapshot
->list
,
2831 &trans
->transaction
->pending_snapshots
);
2832 ret
= btrfs_update_inode(trans
, root
, root
->inode
);
2833 err
= btrfs_commit_transaction(trans
, root
);
2836 mutex_unlock(&root
->fs_info
->fs_mutex
);
2837 btrfs_btree_balance_dirty(root
, nr
);
2838 btrfs_throttle(root
);
2842 unsigned long btrfs_force_ra(struct address_space
*mapping
,
2843 struct file_ra_state
*ra
, struct file
*file
,
2844 pgoff_t offset
, pgoff_t last_index
)
2846 pgoff_t req_size
= last_index
- offset
+ 1;
2848 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2849 offset
= page_cache_readahead(mapping
, ra
, file
, offset
, req_size
);
2852 page_cache_sync_readahead(mapping
, ra
, file
, offset
, req_size
);
2853 return offset
+ req_size
;
2857 int btrfs_defrag_file(struct file
*file
) {
2858 struct inode
*inode
= fdentry(file
)->d_inode
;
2859 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2860 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2862 unsigned long last_index
;
2863 unsigned long ra_pages
= root
->fs_info
->bdi
.ra_pages
;
2864 unsigned long total_read
= 0;
2870 mutex_lock(&root
->fs_info
->fs_mutex
);
2871 ret
= btrfs_check_free_space(root
, inode
->i_size
, 0);
2872 mutex_unlock(&root
->fs_info
->fs_mutex
);
2876 mutex_lock(&inode
->i_mutex
);
2877 last_index
= inode
->i_size
>> PAGE_CACHE_SHIFT
;
2878 for (i
= 0; i
<= last_index
; i
++) {
2879 if (total_read
% ra_pages
== 0) {
2880 btrfs_force_ra(inode
->i_mapping
, &file
->f_ra
, file
, i
,
2881 min(last_index
, i
+ ra_pages
- 1));
2884 page
= grab_cache_page(inode
->i_mapping
, i
);
2887 if (!PageUptodate(page
)) {
2888 btrfs_readpage(NULL
, page
);
2890 if (!PageUptodate(page
)) {
2892 page_cache_release(page
);
2897 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
2898 ClearPageDirty(page
);
2900 cancel_dirty_page(page
, PAGE_CACHE_SIZE
);
2902 wait_on_page_writeback(page
);
2903 set_page_extent_mapped(page
);
2905 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
2906 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
2908 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
2909 set_extent_delalloc(io_tree
, page_start
,
2910 page_end
, GFP_NOFS
);
2912 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
2913 set_page_dirty(page
);
2915 page_cache_release(page
);
2916 balance_dirty_pages_ratelimited_nr(inode
->i_mapping
, 1);
2920 mutex_unlock(&inode
->i_mutex
);
2924 static int btrfs_ioctl_resize(struct btrfs_root
*root
, void __user
*arg
)
2929 struct btrfs_ioctl_vol_args
*vol_args
;
2930 struct btrfs_trans_handle
*trans
;
2931 struct btrfs_device
*device
= NULL
;
2933 char *devstr
= NULL
;
2938 vol_args
= kmalloc(sizeof(*vol_args
), GFP_NOFS
);
2943 if (copy_from_user(vol_args
, arg
, sizeof(*vol_args
))) {
2947 namelen
= strlen(vol_args
->name
);
2948 if (namelen
> BTRFS_VOL_NAME_MAX
) {
2953 mutex_lock(&root
->fs_info
->fs_mutex
);
2954 sizestr
= vol_args
->name
;
2955 devstr
= strchr(sizestr
, ':');
2958 sizestr
= devstr
+ 1;
2960 devstr
= vol_args
->name
;
2961 devid
= simple_strtoull(devstr
, &end
, 10);
2962 printk("resizing devid %Lu\n", devid
);
2964 device
= btrfs_find_device(root
, devid
, NULL
);
2966 printk("resizer unable to find device %Lu\n", devid
);
2970 if (!strcmp(sizestr
, "max"))
2971 new_size
= device
->bdev
->bd_inode
->i_size
;
2973 if (sizestr
[0] == '-') {
2976 } else if (sizestr
[0] == '+') {
2980 new_size
= btrfs_parse_size(sizestr
);
2981 if (new_size
== 0) {
2987 old_size
= device
->total_bytes
;
2990 if (new_size
> old_size
) {
2994 new_size
= old_size
- new_size
;
2995 } else if (mod
> 0) {
2996 new_size
= old_size
+ new_size
;
2999 if (new_size
< 256 * 1024 * 1024) {
3003 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
3008 do_div(new_size
, root
->sectorsize
);
3009 new_size
*= root
->sectorsize
;
3011 printk("new size for %s is %llu\n", device
->name
, (unsigned long long)new_size
);
3013 if (new_size
> old_size
) {
3014 trans
= btrfs_start_transaction(root
, 1);
3015 ret
= btrfs_grow_device(trans
, device
, new_size
);
3016 btrfs_commit_transaction(trans
, root
);
3018 ret
= btrfs_shrink_device(device
, new_size
);
3022 mutex_unlock(&root
->fs_info
->fs_mutex
);
3028 static int noinline
btrfs_ioctl_snap_create(struct btrfs_root
*root
,
3031 struct btrfs_ioctl_vol_args
*vol_args
;
3032 struct btrfs_dir_item
*di
;
3033 struct btrfs_path
*path
;
3038 vol_args
= kmalloc(sizeof(*vol_args
), GFP_NOFS
);
3043 if (copy_from_user(vol_args
, arg
, sizeof(*vol_args
))) {
3048 namelen
= strlen(vol_args
->name
);
3049 if (namelen
> BTRFS_VOL_NAME_MAX
) {
3053 if (strchr(vol_args
->name
, '/')) {
3058 path
= btrfs_alloc_path();
3064 root_dirid
= root
->fs_info
->sb
->s_root
->d_inode
->i_ino
,
3065 mutex_lock(&root
->fs_info
->fs_mutex
);
3066 di
= btrfs_lookup_dir_item(NULL
, root
->fs_info
->tree_root
,
3068 vol_args
->name
, namelen
, 0);
3069 mutex_unlock(&root
->fs_info
->fs_mutex
);
3070 btrfs_free_path(path
);
3072 if (di
&& !IS_ERR(di
)) {
3082 if (root
== root
->fs_info
->tree_root
)
3083 ret
= create_subvol(root
, vol_args
->name
, namelen
);
3085 ret
= create_snapshot(root
, vol_args
->name
, namelen
);
3091 static int btrfs_ioctl_defrag(struct file
*file
)
3093 struct inode
*inode
= fdentry(file
)->d_inode
;
3094 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3096 switch (inode
->i_mode
& S_IFMT
) {
3098 mutex_lock(&root
->fs_info
->fs_mutex
);
3099 btrfs_defrag_root(root
, 0);
3100 btrfs_defrag_root(root
->fs_info
->extent_root
, 0);
3101 mutex_unlock(&root
->fs_info
->fs_mutex
);
3104 btrfs_defrag_file(file
);
3111 long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
3113 struct btrfs_ioctl_vol_args
*vol_args
;
3116 vol_args
= kmalloc(sizeof(*vol_args
), GFP_NOFS
);
3121 if (copy_from_user(vol_args
, arg
, sizeof(*vol_args
))) {
3125 ret
= btrfs_init_new_device(root
, vol_args
->name
);
3132 long btrfs_ioctl_rm_dev(struct btrfs_root
*root
, void __user
*arg
)
3134 struct btrfs_ioctl_vol_args
*vol_args
;
3137 vol_args
= kmalloc(sizeof(*vol_args
), GFP_NOFS
);
3142 if (copy_from_user(vol_args
, arg
, sizeof(*vol_args
))) {
3146 ret
= btrfs_rm_device(root
, vol_args
->name
);
3153 int dup_item_to_inode(struct btrfs_trans_handle
*trans
,
3154 struct btrfs_root
*root
,
3155 struct btrfs_path
*path
,
3156 struct extent_buffer
*leaf
,
3158 struct btrfs_key
*key
,
3162 int len
= btrfs_item_size_nr(leaf
, slot
);
3163 struct btrfs_key ckey
= *key
;
3166 dup
= kmalloc(len
, GFP_NOFS
);
3170 read_extent_buffer(leaf
, dup
, btrfs_item_ptr_offset(leaf
, slot
), len
);
3171 btrfs_release_path(root
, path
);
3173 ckey
.objectid
= destino
;
3174 ret
= btrfs_insert_item(trans
, root
, &ckey
, dup
, len
);
3179 long btrfs_ioctl_clone(struct file
*file
, unsigned long src_fd
)
3181 struct inode
*inode
= fdentry(file
)->d_inode
;
3182 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3183 struct file
*src_file
;
3185 struct btrfs_trans_handle
*trans
;
3188 struct btrfs_path
*path
;
3189 struct btrfs_key key
;
3190 struct extent_buffer
*leaf
;
3194 src_file
= fget(src_fd
);
3197 src
= src_file
->f_dentry
->d_inode
;
3200 if (src
->i_sb
!= inode
->i_sb
)
3204 mutex_lock(&inode
->i_mutex
);
3205 mutex_lock(&src
->i_mutex
);
3207 mutex_lock(&src
->i_mutex
);
3208 mutex_lock(&inode
->i_mutex
);
3215 /* do any pending delalloc/csum calc on src, one way or
3216 another, and lock file content */
3218 filemap_write_and_wait(src
->i_mapping
);
3219 lock_extent(&BTRFS_I(src
)->io_tree
, 0, (u64
)-1, GFP_NOFS
);
3220 if (BTRFS_I(src
)->delalloc_bytes
== 0)
3222 unlock_extent(&BTRFS_I(src
)->io_tree
, 0, (u64
)-1, GFP_NOFS
);
3225 mutex_lock(&root
->fs_info
->fs_mutex
);
3226 trans
= btrfs_start_transaction(root
, 0);
3227 path
= btrfs_alloc_path();
3233 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3234 key
.objectid
= src
->i_ino
;
3240 * note the key will change type as we walk through the
3243 ret
= btrfs_search_slot(trans
, root
, &key
, path
, 0, 0);
3247 if (path
->slots
[0] >= btrfs_header_nritems(path
->nodes
[0])) {
3248 ret
= btrfs_next_leaf(root
, path
);
3254 leaf
= path
->nodes
[0];
3255 slot
= path
->slots
[0];
3256 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
3257 nritems
= btrfs_header_nritems(leaf
);
3259 if (btrfs_key_type(&key
) > BTRFS_CSUM_ITEM_KEY
||
3260 key
.objectid
!= src
->i_ino
)
3263 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
) {
3264 struct btrfs_file_extent_item
*extent
;
3267 extent
= btrfs_item_ptr(leaf
, slot
,
3268 struct btrfs_file_extent_item
);
3269 found_type
= btrfs_file_extent_type(leaf
, extent
);
3270 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
3271 u64 len
= btrfs_file_extent_num_bytes(leaf
,
3273 u64 ds
= btrfs_file_extent_disk_bytenr(leaf
,
3275 u64 dl
= btrfs_file_extent_disk_num_bytes(leaf
,
3277 u64 off
= btrfs_file_extent_offset(leaf
,
3279 btrfs_insert_file_extent(trans
, root
,
3282 /* ds == 0 means there's a hole */
3284 btrfs_inc_extent_ref(trans
, root
,
3286 root
->root_key
.objectid
,
3290 pos
= key
.offset
+ len
;
3291 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
3292 ret
= dup_item_to_inode(trans
, root
, path
,
3297 pos
= key
.offset
+ btrfs_item_size_nr(leaf
,
3300 } else if (btrfs_key_type(&key
) == BTRFS_CSUM_ITEM_KEY
) {
3301 ret
= dup_item_to_inode(trans
, root
, path
, leaf
,
3302 slot
, &key
, inode
->i_ino
);
3308 btrfs_release_path(root
, path
);
3313 btrfs_free_path(path
);
3315 inode
->i_blocks
= src
->i_blocks
;
3316 i_size_write(inode
, src
->i_size
);
3317 btrfs_update_inode(trans
, root
, inode
);
3319 unlock_extent(&BTRFS_I(src
)->io_tree
, 0, (u64
)-1, GFP_NOFS
);
3321 btrfs_end_transaction(trans
, root
);
3322 mutex_unlock(&root
->fs_info
->fs_mutex
);
3325 mutex_unlock(&src
->i_mutex
);
3326 mutex_unlock(&inode
->i_mutex
);
3332 long btrfs_ioctl(struct file
*file
, unsigned int
3333 cmd
, unsigned long arg
)
3335 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3338 case BTRFS_IOC_SNAP_CREATE
:
3339 return btrfs_ioctl_snap_create(root
, (void __user
*)arg
);
3340 case BTRFS_IOC_DEFRAG
:
3341 return btrfs_ioctl_defrag(file
);
3342 case BTRFS_IOC_RESIZE
:
3343 return btrfs_ioctl_resize(root
, (void __user
*)arg
);
3344 case BTRFS_IOC_ADD_DEV
:
3345 return btrfs_ioctl_add_dev(root
, (void __user
*)arg
);
3346 case BTRFS_IOC_RM_DEV
:
3347 return btrfs_ioctl_rm_dev(root
, (void __user
*)arg
);
3348 case BTRFS_IOC_BALANCE
:
3349 return btrfs_balance(root
->fs_info
->dev_root
);
3350 case BTRFS_IOC_CLONE
:
3351 return btrfs_ioctl_clone(file
, arg
);
3358 * Called inside transaction, so use GFP_NOFS
3360 struct inode
*btrfs_alloc_inode(struct super_block
*sb
)
3362 struct btrfs_inode
*ei
;
3364 ei
= kmem_cache_alloc(btrfs_inode_cachep
, GFP_NOFS
);
3368 ei
->ordered_trans
= 0;
3369 return &ei
->vfs_inode
;
3372 void btrfs_destroy_inode(struct inode
*inode
)
3374 WARN_ON(!list_empty(&inode
->i_dentry
));
3375 WARN_ON(inode
->i_data
.nrpages
);
3377 btrfs_drop_extent_cache(inode
, 0, (u64
)-1);
3378 kmem_cache_free(btrfs_inode_cachep
, BTRFS_I(inode
));
3381 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3382 static void init_once(struct kmem_cache
* cachep
, void *foo
)
3384 static void init_once(void * foo
, struct kmem_cache
* cachep
,
3385 unsigned long flags
)
3388 struct btrfs_inode
*ei
= (struct btrfs_inode
*) foo
;
3390 inode_init_once(&ei
->vfs_inode
);
3393 void btrfs_destroy_cachep(void)
3395 if (btrfs_inode_cachep
)
3396 kmem_cache_destroy(btrfs_inode_cachep
);
3397 if (btrfs_trans_handle_cachep
)
3398 kmem_cache_destroy(btrfs_trans_handle_cachep
);
3399 if (btrfs_transaction_cachep
)
3400 kmem_cache_destroy(btrfs_transaction_cachep
);
3401 if (btrfs_bit_radix_cachep
)
3402 kmem_cache_destroy(btrfs_bit_radix_cachep
);
3403 if (btrfs_path_cachep
)
3404 kmem_cache_destroy(btrfs_path_cachep
);
3407 struct kmem_cache
*btrfs_cache_create(const char *name
, size_t size
,
3408 unsigned long extra_flags
,
3409 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3410 void (*ctor
)(struct kmem_cache
*, void *)
3412 void (*ctor
)(void *, struct kmem_cache
*,
3417 return kmem_cache_create(name
, size
, 0, (SLAB_RECLAIM_ACCOUNT
|
3418 SLAB_MEM_SPREAD
| extra_flags
), ctor
3419 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3425 int btrfs_init_cachep(void)
3427 btrfs_inode_cachep
= btrfs_cache_create("btrfs_inode_cache",
3428 sizeof(struct btrfs_inode
),
3430 if (!btrfs_inode_cachep
)
3432 btrfs_trans_handle_cachep
=
3433 btrfs_cache_create("btrfs_trans_handle_cache",
3434 sizeof(struct btrfs_trans_handle
),
3436 if (!btrfs_trans_handle_cachep
)
3438 btrfs_transaction_cachep
= btrfs_cache_create("btrfs_transaction_cache",
3439 sizeof(struct btrfs_transaction
),
3441 if (!btrfs_transaction_cachep
)
3443 btrfs_path_cachep
= btrfs_cache_create("btrfs_path_cache",
3444 sizeof(struct btrfs_path
),
3446 if (!btrfs_path_cachep
)
3448 btrfs_bit_radix_cachep
= btrfs_cache_create("btrfs_radix", 256,
3449 SLAB_DESTROY_BY_RCU
, NULL
);
3450 if (!btrfs_bit_radix_cachep
)
3454 btrfs_destroy_cachep();
3458 static int btrfs_getattr(struct vfsmount
*mnt
,
3459 struct dentry
*dentry
, struct kstat
*stat
)
3461 struct inode
*inode
= dentry
->d_inode
;
3462 generic_fillattr(inode
, stat
);
3463 stat
->blksize
= PAGE_CACHE_SIZE
;
3464 stat
->blocks
= inode
->i_blocks
+ (BTRFS_I(inode
)->delalloc_bytes
>> 9);
3468 static int btrfs_rename(struct inode
* old_dir
, struct dentry
*old_dentry
,
3469 struct inode
* new_dir
,struct dentry
*new_dentry
)
3471 struct btrfs_trans_handle
*trans
;
3472 struct btrfs_root
*root
= BTRFS_I(old_dir
)->root
;
3473 struct inode
*new_inode
= new_dentry
->d_inode
;
3474 struct inode
*old_inode
= old_dentry
->d_inode
;
3475 struct timespec ctime
= CURRENT_TIME
;
3476 struct btrfs_path
*path
;
3479 if (S_ISDIR(old_inode
->i_mode
) && new_inode
&&
3480 new_inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
3484 mutex_lock(&root
->fs_info
->fs_mutex
);
3485 ret
= btrfs_check_free_space(root
, 1, 0);
3489 trans
= btrfs_start_transaction(root
, 1);
3491 btrfs_set_trans_block_group(trans
, new_dir
);
3492 path
= btrfs_alloc_path();
3498 old_dentry
->d_inode
->i_nlink
++;
3499 old_dir
->i_ctime
= old_dir
->i_mtime
= ctime
;
3500 new_dir
->i_ctime
= new_dir
->i_mtime
= ctime
;
3501 old_inode
->i_ctime
= ctime
;
3503 ret
= btrfs_unlink_trans(trans
, root
, old_dir
, old_dentry
);
3508 new_inode
->i_ctime
= CURRENT_TIME
;
3509 ret
= btrfs_unlink_trans(trans
, root
, new_dir
, new_dentry
);
3513 ret
= btrfs_add_link(trans
, new_dentry
, old_inode
, 1);
3518 btrfs_free_path(path
);
3519 btrfs_end_transaction(trans
, root
);
3521 mutex_unlock(&root
->fs_info
->fs_mutex
);
3525 static int btrfs_symlink(struct inode
*dir
, struct dentry
*dentry
,
3526 const char *symname
)
3528 struct btrfs_trans_handle
*trans
;
3529 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
3530 struct btrfs_path
*path
;
3531 struct btrfs_key key
;
3532 struct inode
*inode
= NULL
;
3539 struct btrfs_file_extent_item
*ei
;
3540 struct extent_buffer
*leaf
;
3541 unsigned long nr
= 0;
3543 name_len
= strlen(symname
) + 1;
3544 if (name_len
> BTRFS_MAX_INLINE_DATA_SIZE(root
))
3545 return -ENAMETOOLONG
;
3547 mutex_lock(&root
->fs_info
->fs_mutex
);
3548 err
= btrfs_check_free_space(root
, 1, 0);
3552 trans
= btrfs_start_transaction(root
, 1);
3553 btrfs_set_trans_block_group(trans
, dir
);
3555 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
3561 inode
= btrfs_new_inode(trans
, root
, dentry
->d_name
.name
,
3563 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
3564 BTRFS_I(dir
)->block_group
, S_IFLNK
|S_IRWXUGO
);
3565 err
= PTR_ERR(inode
);
3569 btrfs_set_trans_block_group(trans
, inode
);
3570 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
3574 inode
->i_mapping
->a_ops
= &btrfs_aops
;
3575 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3576 inode
->i_fop
= &btrfs_file_operations
;
3577 inode
->i_op
= &btrfs_file_inode_operations
;
3578 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
3579 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
3580 inode
->i_mapping
, GFP_NOFS
);
3581 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
3582 inode
->i_mapping
, GFP_NOFS
);
3583 BTRFS_I(inode
)->delalloc_bytes
= 0;
3584 atomic_set(&BTRFS_I(inode
)->ordered_writeback
, 0);
3585 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
3587 dir
->i_sb
->s_dirt
= 1;
3588 btrfs_update_inode_block_group(trans
, inode
);
3589 btrfs_update_inode_block_group(trans
, dir
);
3593 path
= btrfs_alloc_path();
3595 key
.objectid
= inode
->i_ino
;
3597 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
3598 datasize
= btrfs_file_extent_calc_inline_size(name_len
);
3599 err
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
3605 leaf
= path
->nodes
[0];
3606 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
3607 struct btrfs_file_extent_item
);
3608 btrfs_set_file_extent_generation(leaf
, ei
, trans
->transid
);
3609 btrfs_set_file_extent_type(leaf
, ei
,
3610 BTRFS_FILE_EXTENT_INLINE
);
3611 ptr
= btrfs_file_extent_inline_start(ei
);
3612 write_extent_buffer(leaf
, symname
, ptr
, name_len
);
3613 btrfs_mark_buffer_dirty(leaf
);
3614 btrfs_free_path(path
);
3616 inode
->i_op
= &btrfs_symlink_inode_operations
;
3617 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
3618 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3619 inode
->i_size
= name_len
- 1;
3620 err
= btrfs_update_inode(trans
, root
, inode
);
3625 nr
= trans
->blocks_used
;
3626 btrfs_end_transaction(trans
, root
);
3628 mutex_unlock(&root
->fs_info
->fs_mutex
);
3630 inode_dec_link_count(inode
);
3633 btrfs_btree_balance_dirty(root
, nr
);
3634 btrfs_throttle(root
);
3638 static int btrfs_permission(struct inode
*inode
, int mask
,
3639 struct nameidata
*nd
)
3641 if (btrfs_test_flag(inode
, READONLY
) && (mask
& MAY_WRITE
))
3643 return generic_permission(inode
, mask
, NULL
);
3646 static struct inode_operations btrfs_dir_inode_operations
= {
3647 .lookup
= btrfs_lookup
,
3648 .create
= btrfs_create
,
3649 .unlink
= btrfs_unlink
,
3651 .mkdir
= btrfs_mkdir
,
3652 .rmdir
= btrfs_rmdir
,
3653 .rename
= btrfs_rename
,
3654 .symlink
= btrfs_symlink
,
3655 .setattr
= btrfs_setattr
,
3656 .mknod
= btrfs_mknod
,
3657 .setxattr
= generic_setxattr
,
3658 .getxattr
= generic_getxattr
,
3659 .listxattr
= btrfs_listxattr
,
3660 .removexattr
= generic_removexattr
,
3661 .permission
= btrfs_permission
,
3663 static struct inode_operations btrfs_dir_ro_inode_operations
= {
3664 .lookup
= btrfs_lookup
,
3665 .permission
= btrfs_permission
,
3667 static struct file_operations btrfs_dir_file_operations
= {
3668 .llseek
= generic_file_llseek
,
3669 .read
= generic_read_dir
,
3670 .readdir
= btrfs_readdir
,
3671 .unlocked_ioctl
= btrfs_ioctl
,
3672 #ifdef CONFIG_COMPAT
3673 .compat_ioctl
= btrfs_ioctl
,
3677 static struct extent_io_ops btrfs_extent_io_ops
= {
3678 .fill_delalloc
= run_delalloc_range
,
3679 .submit_bio_hook
= btrfs_submit_bio_hook
,
3680 .merge_bio_hook
= btrfs_merge_bio_hook
,
3681 .readpage_io_hook
= btrfs_readpage_io_hook
,
3682 .readpage_end_io_hook
= btrfs_readpage_end_io_hook
,
3683 .readpage_io_failed_hook
= btrfs_io_failed_hook
,
3684 .set_bit_hook
= btrfs_set_bit_hook
,
3685 .clear_bit_hook
= btrfs_clear_bit_hook
,
3688 static struct address_space_operations btrfs_aops
= {
3689 .readpage
= btrfs_readpage
,
3690 .writepage
= btrfs_writepage
,
3691 .writepages
= btrfs_writepages
,
3692 .readpages
= btrfs_readpages
,
3693 .sync_page
= block_sync_page
,
3695 .direct_IO
= btrfs_direct_IO
,
3696 .invalidatepage
= btrfs_invalidatepage
,
3697 .releasepage
= btrfs_releasepage
,
3698 .set_page_dirty
= __set_page_dirty_nobuffers
,
3701 static struct address_space_operations btrfs_symlink_aops
= {
3702 .readpage
= btrfs_readpage
,
3703 .writepage
= btrfs_writepage
,
3704 .invalidatepage
= btrfs_invalidatepage
,
3705 .releasepage
= btrfs_releasepage
,
3708 static struct inode_operations btrfs_file_inode_operations
= {
3709 .truncate
= btrfs_truncate
,
3710 .getattr
= btrfs_getattr
,
3711 .setattr
= btrfs_setattr
,
3712 .setxattr
= generic_setxattr
,
3713 .getxattr
= generic_getxattr
,
3714 .listxattr
= btrfs_listxattr
,
3715 .removexattr
= generic_removexattr
,
3716 .permission
= btrfs_permission
,
3718 static struct inode_operations btrfs_special_inode_operations
= {
3719 .getattr
= btrfs_getattr
,
3720 .setattr
= btrfs_setattr
,
3721 .permission
= btrfs_permission
,
3723 static struct inode_operations btrfs_symlink_inode_operations
= {
3724 .readlink
= generic_readlink
,
3725 .follow_link
= page_follow_link_light
,
3726 .put_link
= page_put_link
,
3727 .permission
= btrfs_permission
,