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
, 1);
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
, 0);
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) {
860 /* if the inode isn't linked anywhere,
861 * we don't need to worry about
864 btrfs_del_ordered_inode(inode
);
867 btrfs_end_transaction(trans
, root
);
869 mutex_unlock(&root
->fs_info
->fs_mutex
);
870 btrfs_btree_balance_dirty(root
, nr
);
871 btrfs_throttle(root
);
875 static int btrfs_rmdir(struct inode
*dir
, struct dentry
*dentry
)
877 struct inode
*inode
= dentry
->d_inode
;
880 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
881 struct btrfs_trans_handle
*trans
;
882 unsigned long nr
= 0;
884 if (inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
)
887 mutex_lock(&root
->fs_info
->fs_mutex
);
888 ret
= btrfs_check_free_space(root
, 1, 1);
892 trans
= btrfs_start_transaction(root
, 1);
893 btrfs_set_trans_block_group(trans
, dir
);
895 /* now the directory is empty */
896 err
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
901 nr
= trans
->blocks_used
;
902 ret
= btrfs_end_transaction(trans
, root
);
904 mutex_unlock(&root
->fs_info
->fs_mutex
);
905 btrfs_btree_balance_dirty(root
, nr
);
906 btrfs_throttle(root
);
914 * this can truncate away extent items, csum items and directory items.
915 * It starts at a high offset and removes keys until it can't find
916 * any higher than i_size.
918 * csum items that cross the new i_size are truncated to the new size
921 static int btrfs_truncate_in_trans(struct btrfs_trans_handle
*trans
,
922 struct btrfs_root
*root
,
927 struct btrfs_path
*path
;
928 struct btrfs_key key
;
929 struct btrfs_key found_key
;
931 struct extent_buffer
*leaf
;
932 struct btrfs_file_extent_item
*fi
;
933 u64 extent_start
= 0;
934 u64 extent_num_bytes
= 0;
940 int pending_del_nr
= 0;
941 int pending_del_slot
= 0;
942 int extent_type
= -1;
943 u64 mask
= root
->sectorsize
- 1;
945 btrfs_drop_extent_cache(inode
, inode
->i_size
& (~mask
), (u64
)-1);
946 path
= btrfs_alloc_path();
950 /* FIXME, add redo link to tree so we don't leak on crash */
951 key
.objectid
= inode
->i_ino
;
952 key
.offset
= (u64
)-1;
955 btrfs_init_path(path
);
957 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
962 BUG_ON(path
->slots
[0] == 0);
968 leaf
= path
->nodes
[0];
969 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
970 found_type
= btrfs_key_type(&found_key
);
972 if (found_key
.objectid
!= inode
->i_ino
)
975 if (found_type
< min_type
)
978 item_end
= found_key
.offset
;
979 if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
980 fi
= btrfs_item_ptr(leaf
, path
->slots
[0],
981 struct btrfs_file_extent_item
);
982 extent_type
= btrfs_file_extent_type(leaf
, fi
);
983 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
985 btrfs_file_extent_num_bytes(leaf
, fi
);
986 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
987 struct btrfs_item
*item
= btrfs_item_nr(leaf
,
989 item_end
+= btrfs_file_extent_inline_len(leaf
,
994 if (found_type
== BTRFS_CSUM_ITEM_KEY
) {
995 ret
= btrfs_csum_truncate(trans
, root
, path
,
999 if (item_end
< inode
->i_size
) {
1000 if (found_type
== BTRFS_DIR_ITEM_KEY
) {
1001 found_type
= BTRFS_INODE_ITEM_KEY
;
1002 } else if (found_type
== BTRFS_EXTENT_ITEM_KEY
) {
1003 found_type
= BTRFS_CSUM_ITEM_KEY
;
1004 } else if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
1005 found_type
= BTRFS_XATTR_ITEM_KEY
;
1006 } else if (found_type
== BTRFS_XATTR_ITEM_KEY
) {
1007 found_type
= BTRFS_INODE_REF_KEY
;
1008 } else if (found_type
) {
1013 btrfs_set_key_type(&key
, found_type
);
1016 if (found_key
.offset
>= inode
->i_size
)
1022 /* FIXME, shrink the extent if the ref count is only 1 */
1023 if (found_type
!= BTRFS_EXTENT_DATA_KEY
)
1026 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1028 extent_start
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
1030 u64 orig_num_bytes
=
1031 btrfs_file_extent_num_bytes(leaf
, fi
);
1032 extent_num_bytes
= inode
->i_size
-
1033 found_key
.offset
+ root
->sectorsize
- 1;
1034 extent_num_bytes
= extent_num_bytes
&
1035 ~((u64
)root
->sectorsize
- 1);
1036 btrfs_set_file_extent_num_bytes(leaf
, fi
,
1038 num_dec
= (orig_num_bytes
-
1040 if (extent_start
!= 0)
1041 dec_i_blocks(inode
, num_dec
);
1042 btrfs_mark_buffer_dirty(leaf
);
1045 btrfs_file_extent_disk_num_bytes(leaf
,
1047 /* FIXME blocksize != 4096 */
1048 num_dec
= btrfs_file_extent_num_bytes(leaf
, fi
);
1049 if (extent_start
!= 0) {
1051 dec_i_blocks(inode
, num_dec
);
1053 root_gen
= btrfs_header_generation(leaf
);
1054 root_owner
= btrfs_header_owner(leaf
);
1056 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1058 u32 newsize
= inode
->i_size
- found_key
.offset
;
1059 dec_i_blocks(inode
, item_end
+ 1 -
1060 found_key
.offset
- newsize
);
1062 btrfs_file_extent_calc_inline_size(newsize
);
1063 ret
= btrfs_truncate_item(trans
, root
, path
,
1067 dec_i_blocks(inode
, item_end
+ 1 -
1073 if (!pending_del_nr
) {
1074 /* no pending yet, add ourselves */
1075 pending_del_slot
= path
->slots
[0];
1077 } else if (pending_del_nr
&&
1078 path
->slots
[0] + 1 == pending_del_slot
) {
1079 /* hop on the pending chunk */
1081 pending_del_slot
= path
->slots
[0];
1083 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path
->slots
[0], pending_del_nr
, pending_del_slot
);
1089 ret
= btrfs_free_extent(trans
, root
, extent_start
,
1092 root_gen
, inode
->i_ino
,
1093 found_key
.offset
, 0);
1097 if (path
->slots
[0] == 0) {
1100 btrfs_release_path(root
, path
);
1105 if (pending_del_nr
&&
1106 path
->slots
[0] + 1 != pending_del_slot
) {
1107 struct btrfs_key debug
;
1109 btrfs_item_key_to_cpu(path
->nodes
[0], &debug
,
1111 ret
= btrfs_del_items(trans
, root
, path
,
1116 btrfs_release_path(root
, path
);
1122 if (pending_del_nr
) {
1123 ret
= btrfs_del_items(trans
, root
, path
, pending_del_slot
,
1126 btrfs_release_path(root
, path
);
1127 btrfs_free_path(path
);
1128 inode
->i_sb
->s_dirt
= 1;
1132 static int btrfs_cow_one_page(struct inode
*inode
, struct page
*page
,
1136 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1137 u64 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
1138 u64 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1141 WARN_ON(!PageLocked(page
));
1142 set_page_extent_mapped(page
);
1144 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1145 set_extent_delalloc(&BTRFS_I(inode
)->io_tree
, page_start
,
1146 page_end
, GFP_NOFS
);
1148 if (zero_start
!= PAGE_CACHE_SIZE
) {
1150 memset(kaddr
+ zero_start
, 0, PAGE_CACHE_SIZE
- zero_start
);
1151 flush_dcache_page(page
);
1154 set_page_dirty(page
);
1155 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1161 * taken from block_truncate_page, but does cow as it zeros out
1162 * any bytes left in the last page in the file.
1164 static int btrfs_truncate_page(struct address_space
*mapping
, loff_t from
)
1166 struct inode
*inode
= mapping
->host
;
1167 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1168 u32 blocksize
= root
->sectorsize
;
1169 pgoff_t index
= from
>> PAGE_CACHE_SHIFT
;
1170 unsigned offset
= from
& (PAGE_CACHE_SIZE
-1);
1175 if ((offset
& (blocksize
- 1)) == 0)
1180 page
= grab_cache_page(mapping
, index
);
1183 if (!PageUptodate(page
)) {
1184 ret
= btrfs_readpage(NULL
, page
);
1186 if (page
->mapping
!= mapping
) {
1188 page_cache_release(page
);
1191 if (!PageUptodate(page
)) {
1197 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
1198 wait_on_page_writeback(page
);
1199 ret
= btrfs_cow_one_page(inode
, page
, offset
);
1202 page_cache_release(page
);
1207 static int btrfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
1209 struct inode
*inode
= dentry
->d_inode
;
1212 err
= inode_change_ok(inode
, attr
);
1216 if (S_ISREG(inode
->i_mode
) &&
1217 attr
->ia_valid
& ATTR_SIZE
&& attr
->ia_size
> inode
->i_size
) {
1218 struct btrfs_trans_handle
*trans
;
1219 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1220 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1222 u64 mask
= root
->sectorsize
- 1;
1223 u64 hole_start
= (inode
->i_size
+ mask
) & ~mask
;
1224 u64 block_end
= (attr
->ia_size
+ mask
) & ~mask
;
1228 if (attr
->ia_size
<= hole_start
)
1231 mutex_lock(&root
->fs_info
->fs_mutex
);
1232 err
= btrfs_check_free_space(root
, 1, 0);
1233 mutex_unlock(&root
->fs_info
->fs_mutex
);
1237 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
1239 lock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1240 hole_size
= block_end
- hole_start
;
1242 mutex_lock(&root
->fs_info
->fs_mutex
);
1243 trans
= btrfs_start_transaction(root
, 1);
1244 btrfs_set_trans_block_group(trans
, inode
);
1245 err
= btrfs_drop_extents(trans
, root
, inode
,
1246 hole_start
, block_end
, hole_start
,
1249 if (alloc_hint
!= EXTENT_MAP_INLINE
) {
1250 err
= btrfs_insert_file_extent(trans
, root
,
1254 btrfs_drop_extent_cache(inode
, hole_start
,
1256 btrfs_check_file(root
, inode
);
1258 btrfs_end_transaction(trans
, root
);
1259 mutex_unlock(&root
->fs_info
->fs_mutex
);
1260 unlock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1265 err
= inode_setattr(inode
, attr
);
1270 void btrfs_delete_inode(struct inode
*inode
)
1272 struct btrfs_trans_handle
*trans
;
1273 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1277 truncate_inode_pages(&inode
->i_data
, 0);
1278 if (is_bad_inode(inode
)) {
1283 mutex_lock(&root
->fs_info
->fs_mutex
);
1284 trans
= btrfs_start_transaction(root
, 1);
1286 btrfs_set_trans_block_group(trans
, inode
);
1287 ret
= btrfs_truncate_in_trans(trans
, root
, inode
, 0);
1289 goto no_delete_lock
;
1291 nr
= trans
->blocks_used
;
1294 btrfs_end_transaction(trans
, root
);
1295 mutex_unlock(&root
->fs_info
->fs_mutex
);
1296 btrfs_btree_balance_dirty(root
, nr
);
1297 btrfs_throttle(root
);
1301 nr
= trans
->blocks_used
;
1302 btrfs_end_transaction(trans
, root
);
1303 mutex_unlock(&root
->fs_info
->fs_mutex
);
1304 btrfs_btree_balance_dirty(root
, nr
);
1305 btrfs_throttle(root
);
1311 * this returns the key found in the dir entry in the location pointer.
1312 * If no dir entries were found, location->objectid is 0.
1314 static int btrfs_inode_by_name(struct inode
*dir
, struct dentry
*dentry
,
1315 struct btrfs_key
*location
)
1317 const char *name
= dentry
->d_name
.name
;
1318 int namelen
= dentry
->d_name
.len
;
1319 struct btrfs_dir_item
*di
;
1320 struct btrfs_path
*path
;
1321 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1324 if (namelen
== 1 && strcmp(name
, ".") == 0) {
1325 location
->objectid
= dir
->i_ino
;
1326 location
->type
= BTRFS_INODE_ITEM_KEY
;
1327 location
->offset
= 0;
1330 path
= btrfs_alloc_path();
1333 if (namelen
== 2 && strcmp(name
, "..") == 0) {
1334 struct btrfs_key key
;
1335 struct extent_buffer
*leaf
;
1339 key
.objectid
= dir
->i_ino
;
1340 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1342 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1346 leaf
= path
->nodes
[0];
1347 slot
= path
->slots
[0];
1348 nritems
= btrfs_header_nritems(leaf
);
1349 if (slot
>= nritems
)
1352 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
1353 if (key
.objectid
!= dir
->i_ino
||
1354 key
.type
!= BTRFS_INODE_REF_KEY
) {
1357 location
->objectid
= key
.offset
;
1358 location
->type
= BTRFS_INODE_ITEM_KEY
;
1359 location
->offset
= 0;
1363 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir
->i_ino
, name
,
1367 if (!di
|| IS_ERR(di
)) {
1370 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, location
);
1372 btrfs_free_path(path
);
1375 location
->objectid
= 0;
1380 * when we hit a tree root in a directory, the btrfs part of the inode
1381 * needs to be changed to reflect the root directory of the tree root. This
1382 * is kind of like crossing a mount point.
1384 static int fixup_tree_root_location(struct btrfs_root
*root
,
1385 struct btrfs_key
*location
,
1386 struct btrfs_root
**sub_root
,
1387 struct dentry
*dentry
)
1389 struct btrfs_path
*path
;
1390 struct btrfs_root_item
*ri
;
1392 if (btrfs_key_type(location
) != BTRFS_ROOT_ITEM_KEY
)
1394 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1397 path
= btrfs_alloc_path();
1399 mutex_lock(&root
->fs_info
->fs_mutex
);
1401 *sub_root
= btrfs_read_fs_root(root
->fs_info
, location
,
1402 dentry
->d_name
.name
,
1403 dentry
->d_name
.len
);
1404 if (IS_ERR(*sub_root
))
1405 return PTR_ERR(*sub_root
);
1407 ri
= &(*sub_root
)->root_item
;
1408 location
->objectid
= btrfs_root_dirid(ri
);
1409 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1410 location
->offset
= 0;
1412 btrfs_free_path(path
);
1413 mutex_unlock(&root
->fs_info
->fs_mutex
);
1417 static int btrfs_init_locked_inode(struct inode
*inode
, void *p
)
1419 struct btrfs_iget_args
*args
= p
;
1420 inode
->i_ino
= args
->ino
;
1421 BTRFS_I(inode
)->root
= args
->root
;
1422 BTRFS_I(inode
)->delalloc_bytes
= 0;
1423 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1424 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1425 inode
->i_mapping
, GFP_NOFS
);
1426 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1427 inode
->i_mapping
, GFP_NOFS
);
1428 atomic_set(&BTRFS_I(inode
)->ordered_writeback
, 0);
1432 static int btrfs_find_actor(struct inode
*inode
, void *opaque
)
1434 struct btrfs_iget_args
*args
= opaque
;
1435 return (args
->ino
== inode
->i_ino
&&
1436 args
->root
== BTRFS_I(inode
)->root
);
1439 struct inode
*btrfs_ilookup(struct super_block
*s
, u64 objectid
,
1442 struct btrfs_iget_args args
;
1443 args
.ino
= objectid
;
1444 args
.root
= btrfs_lookup_fs_root(btrfs_sb(s
)->fs_info
, root_objectid
);
1449 return ilookup5(s
, objectid
, btrfs_find_actor
, (void *)&args
);
1452 struct inode
*btrfs_iget_locked(struct super_block
*s
, u64 objectid
,
1453 struct btrfs_root
*root
)
1455 struct inode
*inode
;
1456 struct btrfs_iget_args args
;
1457 args
.ino
= objectid
;
1460 inode
= iget5_locked(s
, objectid
, btrfs_find_actor
,
1461 btrfs_init_locked_inode
,
1466 static struct dentry
*btrfs_lookup(struct inode
*dir
, struct dentry
*dentry
,
1467 struct nameidata
*nd
)
1469 struct inode
* inode
;
1470 struct btrfs_inode
*bi
= BTRFS_I(dir
);
1471 struct btrfs_root
*root
= bi
->root
;
1472 struct btrfs_root
*sub_root
= root
;
1473 struct btrfs_key location
;
1476 if (dentry
->d_name
.len
> BTRFS_NAME_LEN
)
1477 return ERR_PTR(-ENAMETOOLONG
);
1479 mutex_lock(&root
->fs_info
->fs_mutex
);
1480 ret
= btrfs_inode_by_name(dir
, dentry
, &location
);
1481 mutex_unlock(&root
->fs_info
->fs_mutex
);
1484 return ERR_PTR(ret
);
1487 if (location
.objectid
) {
1488 ret
= fixup_tree_root_location(root
, &location
, &sub_root
,
1491 return ERR_PTR(ret
);
1493 return ERR_PTR(-ENOENT
);
1494 inode
= btrfs_iget_locked(dir
->i_sb
, location
.objectid
,
1497 return ERR_PTR(-EACCES
);
1498 if (inode
->i_state
& I_NEW
) {
1499 /* the inode and parent dir are two different roots */
1500 if (sub_root
!= root
) {
1502 sub_root
->inode
= inode
;
1504 BTRFS_I(inode
)->root
= sub_root
;
1505 memcpy(&BTRFS_I(inode
)->location
, &location
,
1507 btrfs_read_locked_inode(inode
);
1508 unlock_new_inode(inode
);
1511 return d_splice_alias(inode
, dentry
);
1514 static unsigned char btrfs_filetype_table
[] = {
1515 DT_UNKNOWN
, DT_REG
, DT_DIR
, DT_CHR
, DT_BLK
, DT_FIFO
, DT_SOCK
, DT_LNK
1518 static int btrfs_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
1520 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1521 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1522 struct btrfs_item
*item
;
1523 struct btrfs_dir_item
*di
;
1524 struct btrfs_key key
;
1525 struct btrfs_key found_key
;
1526 struct btrfs_path
*path
;
1529 struct extent_buffer
*leaf
;
1532 unsigned char d_type
;
1537 int key_type
= BTRFS_DIR_INDEX_KEY
;
1542 /* FIXME, use a real flag for deciding about the key type */
1543 if (root
->fs_info
->tree_root
== root
)
1544 key_type
= BTRFS_DIR_ITEM_KEY
;
1546 /* special case for "." */
1547 if (filp
->f_pos
== 0) {
1548 over
= filldir(dirent
, ".", 1,
1556 mutex_lock(&root
->fs_info
->fs_mutex
);
1557 key
.objectid
= inode
->i_ino
;
1558 path
= btrfs_alloc_path();
1561 /* special case for .., just use the back ref */
1562 if (filp
->f_pos
== 1) {
1563 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1565 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1567 leaf
= path
->nodes
[0];
1568 slot
= path
->slots
[0];
1569 nritems
= btrfs_header_nritems(leaf
);
1570 if (slot
>= nritems
) {
1571 btrfs_release_path(root
, path
);
1572 goto read_dir_items
;
1574 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1575 btrfs_release_path(root
, path
);
1576 if (found_key
.objectid
!= key
.objectid
||
1577 found_key
.type
!= BTRFS_INODE_REF_KEY
)
1578 goto read_dir_items
;
1579 over
= filldir(dirent
, "..", 2,
1580 2, found_key
.offset
, DT_DIR
);
1587 btrfs_set_key_type(&key
, key_type
);
1588 key
.offset
= filp
->f_pos
;
1590 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1595 leaf
= path
->nodes
[0];
1596 nritems
= btrfs_header_nritems(leaf
);
1597 slot
= path
->slots
[0];
1598 if (advance
|| slot
>= nritems
) {
1599 if (slot
>= nritems
-1) {
1600 ret
= btrfs_next_leaf(root
, path
);
1603 leaf
= path
->nodes
[0];
1604 nritems
= btrfs_header_nritems(leaf
);
1605 slot
= path
->slots
[0];
1612 item
= btrfs_item_nr(leaf
, slot
);
1613 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1615 if (found_key
.objectid
!= key
.objectid
)
1617 if (btrfs_key_type(&found_key
) != key_type
)
1619 if (found_key
.offset
< filp
->f_pos
)
1622 filp
->f_pos
= found_key
.offset
;
1624 di
= btrfs_item_ptr(leaf
, slot
, struct btrfs_dir_item
);
1626 di_total
= btrfs_item_size(leaf
, item
);
1627 while(di_cur
< di_total
) {
1628 struct btrfs_key location
;
1630 name_len
= btrfs_dir_name_len(leaf
, di
);
1631 if (name_len
< 32) {
1632 name_ptr
= tmp_name
;
1634 name_ptr
= kmalloc(name_len
, GFP_NOFS
);
1637 read_extent_buffer(leaf
, name_ptr
,
1638 (unsigned long)(di
+ 1), name_len
);
1640 d_type
= btrfs_filetype_table
[btrfs_dir_type(leaf
, di
)];
1641 btrfs_dir_item_key_to_cpu(leaf
, di
, &location
);
1642 over
= filldir(dirent
, name_ptr
, name_len
,
1647 if (name_ptr
!= tmp_name
)
1652 di_len
= btrfs_dir_name_len(leaf
, di
) +
1653 btrfs_dir_data_len(leaf
, di
) +sizeof(*di
);
1655 di
= (struct btrfs_dir_item
*)((char *)di
+ di_len
);
1658 if (key_type
== BTRFS_DIR_INDEX_KEY
)
1659 filp
->f_pos
= INT_LIMIT(typeof(filp
->f_pos
));
1665 btrfs_release_path(root
, path
);
1666 btrfs_free_path(path
);
1667 mutex_unlock(&root
->fs_info
->fs_mutex
);
1671 int btrfs_write_inode(struct inode
*inode
, int wait
)
1673 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1674 struct btrfs_trans_handle
*trans
;
1678 mutex_lock(&root
->fs_info
->fs_mutex
);
1679 trans
= btrfs_start_transaction(root
, 1);
1680 btrfs_set_trans_block_group(trans
, inode
);
1681 ret
= btrfs_commit_transaction(trans
, root
);
1682 mutex_unlock(&root
->fs_info
->fs_mutex
);
1688 * This is somewhat expensive, updating the tree every time the
1689 * inode changes. But, it is most likely to find the inode in cache.
1690 * FIXME, needs more benchmarking...there are no reasons other than performance
1691 * to keep or drop this code.
1693 void btrfs_dirty_inode(struct inode
*inode
)
1695 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1696 struct btrfs_trans_handle
*trans
;
1698 mutex_lock(&root
->fs_info
->fs_mutex
);
1699 trans
= btrfs_start_transaction(root
, 1);
1700 btrfs_set_trans_block_group(trans
, inode
);
1701 btrfs_update_inode(trans
, root
, inode
);
1702 btrfs_end_transaction(trans
, root
);
1703 mutex_unlock(&root
->fs_info
->fs_mutex
);
1706 static struct inode
*btrfs_new_inode(struct btrfs_trans_handle
*trans
,
1707 struct btrfs_root
*root
,
1708 const char *name
, int name_len
,
1711 struct btrfs_block_group_cache
*group
,
1714 struct inode
*inode
;
1715 struct btrfs_inode_item
*inode_item
;
1716 struct btrfs_block_group_cache
*new_inode_group
;
1717 struct btrfs_key
*location
;
1718 struct btrfs_path
*path
;
1719 struct btrfs_inode_ref
*ref
;
1720 struct btrfs_key key
[2];
1726 path
= btrfs_alloc_path();
1729 inode
= new_inode(root
->fs_info
->sb
);
1731 return ERR_PTR(-ENOMEM
);
1733 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1734 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1735 inode
->i_mapping
, GFP_NOFS
);
1736 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1737 inode
->i_mapping
, GFP_NOFS
);
1738 atomic_set(&BTRFS_I(inode
)->ordered_writeback
, 0);
1739 BTRFS_I(inode
)->delalloc_bytes
= 0;
1740 BTRFS_I(inode
)->root
= root
;
1746 new_inode_group
= btrfs_find_block_group(root
, group
, 0,
1747 BTRFS_BLOCK_GROUP_METADATA
, owner
);
1748 if (!new_inode_group
) {
1749 printk("find_block group failed\n");
1750 new_inode_group
= group
;
1752 BTRFS_I(inode
)->block_group
= new_inode_group
;
1753 BTRFS_I(inode
)->flags
= 0;
1755 key
[0].objectid
= objectid
;
1756 btrfs_set_key_type(&key
[0], BTRFS_INODE_ITEM_KEY
);
1759 key
[1].objectid
= objectid
;
1760 btrfs_set_key_type(&key
[1], BTRFS_INODE_REF_KEY
);
1761 key
[1].offset
= ref_objectid
;
1763 sizes
[0] = sizeof(struct btrfs_inode_item
);
1764 sizes
[1] = name_len
+ sizeof(*ref
);
1766 ret
= btrfs_insert_empty_items(trans
, root
, path
, key
, sizes
, 2);
1770 if (objectid
> root
->highest_inode
)
1771 root
->highest_inode
= objectid
;
1773 inode
->i_uid
= current
->fsuid
;
1774 inode
->i_gid
= current
->fsgid
;
1775 inode
->i_mode
= mode
;
1776 inode
->i_ino
= objectid
;
1777 inode
->i_blocks
= 0;
1778 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1779 inode_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
1780 struct btrfs_inode_item
);
1781 fill_inode_item(path
->nodes
[0], inode_item
, inode
);
1783 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0] + 1,
1784 struct btrfs_inode_ref
);
1785 btrfs_set_inode_ref_name_len(path
->nodes
[0], ref
, name_len
);
1786 ptr
= (unsigned long)(ref
+ 1);
1787 write_extent_buffer(path
->nodes
[0], name
, ptr
, name_len
);
1789 btrfs_mark_buffer_dirty(path
->nodes
[0]);
1790 btrfs_free_path(path
);
1792 location
= &BTRFS_I(inode
)->location
;
1793 location
->objectid
= objectid
;
1794 location
->offset
= 0;
1795 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1797 insert_inode_hash(inode
);
1800 btrfs_free_path(path
);
1801 return ERR_PTR(ret
);
1804 static inline u8
btrfs_inode_type(struct inode
*inode
)
1806 return btrfs_type_by_mode
[(inode
->i_mode
& S_IFMT
) >> S_SHIFT
];
1809 static int btrfs_add_link(struct btrfs_trans_handle
*trans
,
1810 struct dentry
*dentry
, struct inode
*inode
,
1814 struct btrfs_key key
;
1815 struct btrfs_root
*root
= BTRFS_I(dentry
->d_parent
->d_inode
)->root
;
1816 struct inode
*parent_inode
;
1818 key
.objectid
= inode
->i_ino
;
1819 btrfs_set_key_type(&key
, BTRFS_INODE_ITEM_KEY
);
1822 ret
= btrfs_insert_dir_item(trans
, root
,
1823 dentry
->d_name
.name
, dentry
->d_name
.len
,
1824 dentry
->d_parent
->d_inode
->i_ino
,
1825 &key
, btrfs_inode_type(inode
));
1828 ret
= btrfs_insert_inode_ref(trans
, root
,
1829 dentry
->d_name
.name
,
1832 dentry
->d_parent
->d_inode
->i_ino
);
1834 parent_inode
= dentry
->d_parent
->d_inode
;
1835 parent_inode
->i_size
+= dentry
->d_name
.len
* 2;
1836 parent_inode
->i_mtime
= parent_inode
->i_ctime
= CURRENT_TIME
;
1837 ret
= btrfs_update_inode(trans
, root
,
1838 dentry
->d_parent
->d_inode
);
1843 static int btrfs_add_nondir(struct btrfs_trans_handle
*trans
,
1844 struct dentry
*dentry
, struct inode
*inode
,
1847 int err
= btrfs_add_link(trans
, dentry
, inode
, backref
);
1849 d_instantiate(dentry
, inode
);
1857 static int btrfs_mknod(struct inode
*dir
, struct dentry
*dentry
,
1858 int mode
, dev_t rdev
)
1860 struct btrfs_trans_handle
*trans
;
1861 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1862 struct inode
*inode
= NULL
;
1866 unsigned long nr
= 0;
1868 if (!new_valid_dev(rdev
))
1871 mutex_lock(&root
->fs_info
->fs_mutex
);
1872 err
= btrfs_check_free_space(root
, 1, 0);
1876 trans
= btrfs_start_transaction(root
, 1);
1877 btrfs_set_trans_block_group(trans
, dir
);
1879 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
1885 inode
= btrfs_new_inode(trans
, root
, dentry
->d_name
.name
,
1887 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
1888 BTRFS_I(dir
)->block_group
, mode
);
1889 err
= PTR_ERR(inode
);
1893 btrfs_set_trans_block_group(trans
, inode
);
1894 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
1898 inode
->i_op
= &btrfs_special_inode_operations
;
1899 init_special_inode(inode
, inode
->i_mode
, rdev
);
1900 btrfs_update_inode(trans
, root
, inode
);
1902 dir
->i_sb
->s_dirt
= 1;
1903 btrfs_update_inode_block_group(trans
, inode
);
1904 btrfs_update_inode_block_group(trans
, dir
);
1906 nr
= trans
->blocks_used
;
1907 btrfs_end_transaction(trans
, root
);
1909 mutex_unlock(&root
->fs_info
->fs_mutex
);
1912 inode_dec_link_count(inode
);
1915 btrfs_btree_balance_dirty(root
, nr
);
1916 btrfs_throttle(root
);
1920 static int btrfs_create(struct inode
*dir
, struct dentry
*dentry
,
1921 int mode
, struct nameidata
*nd
)
1923 struct btrfs_trans_handle
*trans
;
1924 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1925 struct inode
*inode
= NULL
;
1928 unsigned long nr
= 0;
1931 mutex_lock(&root
->fs_info
->fs_mutex
);
1932 err
= btrfs_check_free_space(root
, 1, 0);
1935 trans
= btrfs_start_transaction(root
, 1);
1936 btrfs_set_trans_block_group(trans
, dir
);
1938 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
1944 inode
= btrfs_new_inode(trans
, root
, dentry
->d_name
.name
,
1946 dentry
->d_parent
->d_inode
->i_ino
,
1947 objectid
, BTRFS_I(dir
)->block_group
, mode
);
1948 err
= PTR_ERR(inode
);
1952 btrfs_set_trans_block_group(trans
, inode
);
1953 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
1957 inode
->i_mapping
->a_ops
= &btrfs_aops
;
1958 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1959 inode
->i_fop
= &btrfs_file_operations
;
1960 inode
->i_op
= &btrfs_file_inode_operations
;
1961 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1962 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1963 inode
->i_mapping
, GFP_NOFS
);
1964 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1965 inode
->i_mapping
, GFP_NOFS
);
1966 BTRFS_I(inode
)->delalloc_bytes
= 0;
1967 atomic_set(&BTRFS_I(inode
)->ordered_writeback
, 0);
1968 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
1970 dir
->i_sb
->s_dirt
= 1;
1971 btrfs_update_inode_block_group(trans
, inode
);
1972 btrfs_update_inode_block_group(trans
, dir
);
1974 nr
= trans
->blocks_used
;
1975 btrfs_end_transaction(trans
, root
);
1977 mutex_unlock(&root
->fs_info
->fs_mutex
);
1980 inode_dec_link_count(inode
);
1983 btrfs_btree_balance_dirty(root
, nr
);
1984 btrfs_throttle(root
);
1988 static int btrfs_link(struct dentry
*old_dentry
, struct inode
*dir
,
1989 struct dentry
*dentry
)
1991 struct btrfs_trans_handle
*trans
;
1992 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1993 struct inode
*inode
= old_dentry
->d_inode
;
1994 unsigned long nr
= 0;
1998 if (inode
->i_nlink
== 0)
2001 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
2006 mutex_lock(&root
->fs_info
->fs_mutex
);
2007 err
= btrfs_check_free_space(root
, 1, 0);
2010 trans
= btrfs_start_transaction(root
, 1);
2012 btrfs_set_trans_block_group(trans
, dir
);
2013 atomic_inc(&inode
->i_count
);
2014 err
= btrfs_add_nondir(trans
, dentry
, inode
, 1);
2019 dir
->i_sb
->s_dirt
= 1;
2020 btrfs_update_inode_block_group(trans
, dir
);
2021 err
= btrfs_update_inode(trans
, root
, inode
);
2026 nr
= trans
->blocks_used
;
2027 btrfs_end_transaction(trans
, root
);
2029 mutex_unlock(&root
->fs_info
->fs_mutex
);
2032 inode_dec_link_count(inode
);
2035 btrfs_btree_balance_dirty(root
, nr
);
2036 btrfs_throttle(root
);
2040 static int btrfs_mkdir(struct inode
*dir
, struct dentry
*dentry
, int mode
)
2042 struct inode
*inode
= NULL
;
2043 struct btrfs_trans_handle
*trans
;
2044 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2046 int drop_on_err
= 0;
2048 unsigned long nr
= 1;
2050 mutex_lock(&root
->fs_info
->fs_mutex
);
2051 err
= btrfs_check_free_space(root
, 1, 0);
2055 trans
= btrfs_start_transaction(root
, 1);
2056 btrfs_set_trans_block_group(trans
, dir
);
2058 if (IS_ERR(trans
)) {
2059 err
= PTR_ERR(trans
);
2063 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2069 inode
= btrfs_new_inode(trans
, root
, dentry
->d_name
.name
,
2071 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2072 BTRFS_I(dir
)->block_group
, S_IFDIR
| mode
);
2073 if (IS_ERR(inode
)) {
2074 err
= PTR_ERR(inode
);
2079 inode
->i_op
= &btrfs_dir_inode_operations
;
2080 inode
->i_fop
= &btrfs_dir_file_operations
;
2081 btrfs_set_trans_block_group(trans
, inode
);
2084 err
= btrfs_update_inode(trans
, root
, inode
);
2088 err
= btrfs_add_link(trans
, dentry
, inode
, 0);
2092 d_instantiate(dentry
, inode
);
2094 dir
->i_sb
->s_dirt
= 1;
2095 btrfs_update_inode_block_group(trans
, inode
);
2096 btrfs_update_inode_block_group(trans
, dir
);
2099 nr
= trans
->blocks_used
;
2100 btrfs_end_transaction(trans
, root
);
2103 mutex_unlock(&root
->fs_info
->fs_mutex
);
2106 btrfs_btree_balance_dirty(root
, nr
);
2107 btrfs_throttle(root
);
2111 static int merge_extent_mapping(struct extent_map_tree
*em_tree
,
2112 struct extent_map
*existing
,
2113 struct extent_map
*em
)
2118 int real_blocks
= existing
->block_start
< EXTENT_MAP_LAST_BYTE
;
2120 if (real_blocks
&& em
->block_start
>= EXTENT_MAP_LAST_BYTE
)
2123 if (!real_blocks
&& em
->block_start
!= existing
->block_start
)
2126 new_end
= max(existing
->start
+ existing
->len
, em
->start
+ em
->len
);
2128 if (existing
->start
>= em
->start
) {
2129 if (em
->start
+ em
->len
< existing
->start
)
2132 start_diff
= existing
->start
- em
->start
;
2133 if (real_blocks
&& em
->block_start
+ start_diff
!=
2134 existing
->block_start
)
2137 em
->len
= new_end
- em
->start
;
2139 remove_extent_mapping(em_tree
, existing
);
2140 /* free for the tree */
2141 free_extent_map(existing
);
2142 ret
= add_extent_mapping(em_tree
, em
);
2144 } else if (em
->start
> existing
->start
) {
2146 if (existing
->start
+ existing
->len
< em
->start
)
2149 start_diff
= em
->start
- existing
->start
;
2150 if (real_blocks
&& existing
->block_start
+ start_diff
!=
2154 remove_extent_mapping(em_tree
, existing
);
2155 em
->block_start
= existing
->block_start
;
2156 em
->start
= existing
->start
;
2157 em
->len
= new_end
- existing
->start
;
2158 free_extent_map(existing
);
2160 ret
= add_extent_mapping(em_tree
, em
);
2167 printk("invalid extent map merge [%Lu %Lu %Lu] [%Lu %Lu %Lu]\n",
2168 existing
->start
, existing
->len
, existing
->block_start
,
2169 em
->start
, em
->len
, em
->block_start
);
2173 struct extent_map
*btrfs_get_extent(struct inode
*inode
, struct page
*page
,
2174 size_t pg_offset
, u64 start
, u64 len
,
2180 u64 extent_start
= 0;
2182 u64 objectid
= inode
->i_ino
;
2184 struct btrfs_path
*path
;
2185 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2186 struct btrfs_file_extent_item
*item
;
2187 struct extent_buffer
*leaf
;
2188 struct btrfs_key found_key
;
2189 struct extent_map
*em
= NULL
;
2190 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
2191 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2192 struct btrfs_trans_handle
*trans
= NULL
;
2194 path
= btrfs_alloc_path();
2196 mutex_lock(&root
->fs_info
->fs_mutex
);
2199 spin_lock(&em_tree
->lock
);
2200 em
= lookup_extent_mapping(em_tree
, start
, len
);
2202 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2203 spin_unlock(&em_tree
->lock
);
2206 if (em
->start
> start
|| em
->start
+ em
->len
<= start
)
2207 free_extent_map(em
);
2208 else if (em
->block_start
== EXTENT_MAP_INLINE
&& page
)
2209 free_extent_map(em
);
2213 em
= alloc_extent_map(GFP_NOFS
);
2219 em
->start
= EXTENT_MAP_HOLE
;
2221 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2222 ret
= btrfs_lookup_file_extent(trans
, root
, path
,
2223 objectid
, start
, trans
!= NULL
);
2230 if (path
->slots
[0] == 0)
2235 leaf
= path
->nodes
[0];
2236 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
2237 struct btrfs_file_extent_item
);
2238 /* are we inside the extent that was found? */
2239 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2240 found_type
= btrfs_key_type(&found_key
);
2241 if (found_key
.objectid
!= objectid
||
2242 found_type
!= BTRFS_EXTENT_DATA_KEY
) {
2246 found_type
= btrfs_file_extent_type(leaf
, item
);
2247 extent_start
= found_key
.offset
;
2248 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
2249 extent_end
= extent_start
+
2250 btrfs_file_extent_num_bytes(leaf
, item
);
2252 if (start
< extent_start
|| start
>= extent_end
) {
2254 if (start
< extent_start
) {
2255 if (start
+ len
<= extent_start
)
2257 em
->len
= extent_end
- extent_start
;
2263 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
2265 em
->start
= extent_start
;
2266 em
->len
= extent_end
- extent_start
;
2267 em
->block_start
= EXTENT_MAP_HOLE
;
2270 bytenr
+= btrfs_file_extent_offset(leaf
, item
);
2271 em
->block_start
= bytenr
;
2272 em
->start
= extent_start
;
2273 em
->len
= extent_end
- extent_start
;
2275 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
2280 size_t extent_offset
;
2283 size
= btrfs_file_extent_inline_len(leaf
, btrfs_item_nr(leaf
,
2285 extent_end
= (extent_start
+ size
+ root
->sectorsize
- 1) &
2286 ~((u64
)root
->sectorsize
- 1);
2287 if (start
< extent_start
|| start
>= extent_end
) {
2289 if (start
< extent_start
) {
2290 if (start
+ len
<= extent_start
)
2292 em
->len
= extent_end
- extent_start
;
2298 em
->block_start
= EXTENT_MAP_INLINE
;
2301 em
->start
= extent_start
;
2306 page_start
= page_offset(page
) + pg_offset
;
2307 extent_offset
= page_start
- extent_start
;
2308 copy_size
= min_t(u64
, PAGE_CACHE_SIZE
- pg_offset
,
2309 size
- extent_offset
);
2310 em
->start
= extent_start
+ extent_offset
;
2311 em
->len
= (copy_size
+ root
->sectorsize
- 1) &
2312 ~((u64
)root
->sectorsize
- 1);
2314 ptr
= btrfs_file_extent_inline_start(item
) + extent_offset
;
2315 if (create
== 0 && !PageUptodate(page
)) {
2316 read_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2318 flush_dcache_page(page
);
2319 } else if (create
&& PageUptodate(page
)) {
2322 free_extent_map(em
);
2324 btrfs_release_path(root
, path
);
2325 trans
= btrfs_start_transaction(root
, 1);
2328 write_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2330 btrfs_mark_buffer_dirty(leaf
);
2333 set_extent_uptodate(io_tree
, em
->start
,
2334 extent_map_end(em
) - 1, GFP_NOFS
);
2337 printk("unkknown found_type %d\n", found_type
);
2344 em
->block_start
= EXTENT_MAP_HOLE
;
2346 btrfs_release_path(root
, path
);
2347 if (em
->start
> start
|| extent_map_end(em
) <= start
) {
2348 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em
->start
, em
->len
, start
, len
);
2354 spin_lock(&em_tree
->lock
);
2355 ret
= add_extent_mapping(em_tree
, em
);
2356 /* it is possible that someone inserted the extent into the tree
2357 * while we had the lock dropped. It is also possible that
2358 * an overlapping map exists in the tree
2360 if (ret
== -EEXIST
) {
2361 struct extent_map
*existing
;
2362 existing
= lookup_extent_mapping(em_tree
, start
, len
);
2363 if (existing
&& (existing
->start
> start
||
2364 existing
->start
+ existing
->len
<= start
)) {
2365 free_extent_map(existing
);
2369 existing
= lookup_extent_mapping(em_tree
, em
->start
,
2372 err
= merge_extent_mapping(em_tree
, existing
,
2374 free_extent_map(existing
);
2376 free_extent_map(em
);
2381 printk("failing to insert %Lu %Lu\n",
2383 free_extent_map(em
);
2387 free_extent_map(em
);
2391 spin_unlock(&em_tree
->lock
);
2393 btrfs_free_path(path
);
2395 ret
= btrfs_end_transaction(trans
, root
);
2399 mutex_unlock(&root
->fs_info
->fs_mutex
);
2401 free_extent_map(em
);
2403 return ERR_PTR(err
);
2408 #if 0 /* waiting for O_DIRECT reads */
2409 static int btrfs_get_block(struct inode
*inode
, sector_t iblock
,
2410 struct buffer_head
*bh_result
, int create
)
2412 struct extent_map
*em
;
2413 u64 start
= (u64
)iblock
<< inode
->i_blkbits
;
2414 struct btrfs_multi_bio
*multi
= NULL
;
2415 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2421 em
= btrfs_get_extent(inode
, NULL
, 0, start
, bh_result
->b_size
, 0);
2423 if (!em
|| IS_ERR(em
))
2426 if (em
->start
> start
|| em
->start
+ em
->len
<= start
) {
2430 if (em
->block_start
== EXTENT_MAP_INLINE
) {
2435 len
= em
->start
+ em
->len
- start
;
2436 len
= min_t(u64
, len
, INT_LIMIT(typeof(bh_result
->b_size
)));
2438 if (em
->block_start
== EXTENT_MAP_HOLE
||
2439 em
->block_start
== EXTENT_MAP_DELALLOC
) {
2440 bh_result
->b_size
= len
;
2444 logical
= start
- em
->start
;
2445 logical
= em
->block_start
+ logical
;
2448 ret
= btrfs_map_block(&root
->fs_info
->mapping_tree
, READ
,
2449 logical
, &map_length
, &multi
, 0);
2451 bh_result
->b_blocknr
= multi
->stripes
[0].physical
>> inode
->i_blkbits
;
2452 bh_result
->b_size
= min(map_length
, len
);
2454 bh_result
->b_bdev
= multi
->stripes
[0].dev
->bdev
;
2455 set_buffer_mapped(bh_result
);
2458 free_extent_map(em
);
2463 static ssize_t
btrfs_direct_IO(int rw
, struct kiocb
*iocb
,
2464 const struct iovec
*iov
, loff_t offset
,
2465 unsigned long nr_segs
)
2469 struct file
*file
= iocb
->ki_filp
;
2470 struct inode
*inode
= file
->f_mapping
->host
;
2475 return blockdev_direct_IO(rw
, iocb
, inode
, inode
->i_sb
->s_bdev
, iov
,
2476 offset
, nr_segs
, btrfs_get_block
, NULL
);
2480 static sector_t
btrfs_bmap(struct address_space
*mapping
, sector_t iblock
)
2482 return extent_bmap(mapping
, iblock
, btrfs_get_extent
);
2485 int btrfs_readpage(struct file
*file
, struct page
*page
)
2487 struct extent_io_tree
*tree
;
2488 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2489 return extent_read_full_page(tree
, page
, btrfs_get_extent
);
2492 static int btrfs_writepage(struct page
*page
, struct writeback_control
*wbc
)
2494 struct extent_io_tree
*tree
;
2497 if (current
->flags
& PF_MEMALLOC
) {
2498 redirty_page_for_writepage(wbc
, page
);
2502 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2503 return extent_write_full_page(tree
, page
, btrfs_get_extent
, wbc
);
2506 static int btrfs_writepages(struct address_space
*mapping
,
2507 struct writeback_control
*wbc
)
2509 struct extent_io_tree
*tree
;
2510 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
2511 return extent_writepages(tree
, mapping
, btrfs_get_extent
, wbc
);
2515 btrfs_readpages(struct file
*file
, struct address_space
*mapping
,
2516 struct list_head
*pages
, unsigned nr_pages
)
2518 struct extent_io_tree
*tree
;
2519 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
2520 return extent_readpages(tree
, mapping
, pages
, nr_pages
,
2524 static int btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
2526 struct extent_io_tree
*tree
;
2527 struct extent_map_tree
*map
;
2530 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2531 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
2532 ret
= try_release_extent_mapping(map
, tree
, page
, gfp_flags
);
2534 invalidate_extent_lru(tree
, page_offset(page
), PAGE_CACHE_SIZE
);
2535 ClearPagePrivate(page
);
2536 set_page_private(page
, 0);
2537 page_cache_release(page
);
2542 static void btrfs_invalidatepage(struct page
*page
, unsigned long offset
)
2544 struct extent_io_tree
*tree
;
2546 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2547 extent_invalidatepage(tree
, page
, offset
);
2548 btrfs_releasepage(page
, GFP_NOFS
);
2549 if (PagePrivate(page
)) {
2550 invalidate_extent_lru(tree
, page_offset(page
), PAGE_CACHE_SIZE
);
2551 ClearPagePrivate(page
);
2552 set_page_private(page
, 0);
2553 page_cache_release(page
);
2558 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
2559 * called from a page fault handler when a page is first dirtied. Hence we must
2560 * be careful to check for EOF conditions here. We set the page up correctly
2561 * for a written page which means we get ENOSPC checking when writing into
2562 * holes and correct delalloc and unwritten extent mapping on filesystems that
2563 * support these features.
2565 * We are not allowed to take the i_mutex here so we have to play games to
2566 * protect against truncate races as the page could now be beyond EOF. Because
2567 * vmtruncate() writes the inode size before removing pages, once we have the
2568 * page lock we can determine safely if the page is beyond EOF. If it is not
2569 * beyond EOF, then the page is guaranteed safe against truncation until we
2572 int btrfs_page_mkwrite(struct vm_area_struct
*vma
, struct page
*page
)
2574 struct inode
*inode
= fdentry(vma
->vm_file
)->d_inode
;
2575 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2581 mutex_lock(&root
->fs_info
->fs_mutex
);
2582 ret
= btrfs_check_free_space(root
, PAGE_CACHE_SIZE
, 0);
2583 mutex_unlock(&root
->fs_info
->fs_mutex
);
2590 wait_on_page_writeback(page
);
2591 size
= i_size_read(inode
);
2592 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
2594 if ((page
->mapping
!= inode
->i_mapping
) ||
2595 (page_start
> size
)) {
2596 /* page got truncated out from underneath us */
2600 /* page is wholly or partially inside EOF */
2601 if (page_start
+ PAGE_CACHE_SIZE
> size
)
2602 end
= size
& ~PAGE_CACHE_MASK
;
2604 end
= PAGE_CACHE_SIZE
;
2606 ret
= btrfs_cow_one_page(inode
, page
, end
);
2614 static void btrfs_truncate(struct inode
*inode
)
2616 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2618 struct btrfs_trans_handle
*trans
;
2621 if (!S_ISREG(inode
->i_mode
))
2623 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
2626 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
2628 mutex_lock(&root
->fs_info
->fs_mutex
);
2629 trans
= btrfs_start_transaction(root
, 1);
2630 btrfs_set_trans_block_group(trans
, inode
);
2632 /* FIXME, add redo link to tree so we don't leak on crash */
2633 ret
= btrfs_truncate_in_trans(trans
, root
, inode
,
2634 BTRFS_EXTENT_DATA_KEY
);
2635 btrfs_update_inode(trans
, root
, inode
);
2636 nr
= trans
->blocks_used
;
2638 ret
= btrfs_end_transaction(trans
, root
);
2640 mutex_unlock(&root
->fs_info
->fs_mutex
);
2641 btrfs_btree_balance_dirty(root
, nr
);
2642 btrfs_throttle(root
);
2646 * Invalidate a single dcache entry at the root of the filesystem.
2647 * Needed after creation of snapshot or subvolume.
2649 void btrfs_invalidate_dcache_root(struct btrfs_root
*root
, char *name
,
2652 struct dentry
*alias
, *entry
;
2655 alias
= d_find_alias(root
->fs_info
->sb
->s_root
->d_inode
);
2659 /* change me if btrfs ever gets a d_hash operation */
2660 qstr
.hash
= full_name_hash(qstr
.name
, qstr
.len
);
2661 entry
= d_lookup(alias
, &qstr
);
2664 d_invalidate(entry
);
2670 static int noinline
create_subvol(struct btrfs_root
*root
, char *name
,
2673 struct btrfs_trans_handle
*trans
;
2674 struct btrfs_key key
;
2675 struct btrfs_root_item root_item
;
2676 struct btrfs_inode_item
*inode_item
;
2677 struct extent_buffer
*leaf
;
2678 struct btrfs_root
*new_root
= root
;
2679 struct inode
*inode
;
2684 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
2685 unsigned long nr
= 1;
2687 mutex_lock(&root
->fs_info
->fs_mutex
);
2688 ret
= btrfs_check_free_space(root
, 1, 0);
2692 trans
= btrfs_start_transaction(root
, 1);
2695 ret
= btrfs_find_free_objectid(trans
, root
->fs_info
->tree_root
,
2700 leaf
= __btrfs_alloc_free_block(trans
, root
, root
->leafsize
,
2701 objectid
, trans
->transid
, 0, 0,
2704 return PTR_ERR(leaf
);
2706 btrfs_set_header_nritems(leaf
, 0);
2707 btrfs_set_header_level(leaf
, 0);
2708 btrfs_set_header_bytenr(leaf
, leaf
->start
);
2709 btrfs_set_header_generation(leaf
, trans
->transid
);
2710 btrfs_set_header_owner(leaf
, objectid
);
2712 write_extent_buffer(leaf
, root
->fs_info
->fsid
,
2713 (unsigned long)btrfs_header_fsid(leaf
),
2715 btrfs_mark_buffer_dirty(leaf
);
2717 inode_item
= &root_item
.inode
;
2718 memset(inode_item
, 0, sizeof(*inode_item
));
2719 inode_item
->generation
= cpu_to_le64(1);
2720 inode_item
->size
= cpu_to_le64(3);
2721 inode_item
->nlink
= cpu_to_le32(1);
2722 inode_item
->nblocks
= cpu_to_le64(1);
2723 inode_item
->mode
= cpu_to_le32(S_IFDIR
| 0755);
2725 btrfs_set_root_bytenr(&root_item
, leaf
->start
);
2726 btrfs_set_root_level(&root_item
, 0);
2727 btrfs_set_root_refs(&root_item
, 1);
2728 btrfs_set_root_used(&root_item
, 0);
2730 memset(&root_item
.drop_progress
, 0, sizeof(root_item
.drop_progress
));
2731 root_item
.drop_level
= 0;
2733 free_extent_buffer(leaf
);
2736 btrfs_set_root_dirid(&root_item
, new_dirid
);
2738 key
.objectid
= objectid
;
2740 btrfs_set_key_type(&key
, BTRFS_ROOT_ITEM_KEY
);
2741 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
2747 * insert the directory item
2749 key
.offset
= (u64
)-1;
2750 dir
= root
->fs_info
->sb
->s_root
->d_inode
;
2751 ret
= btrfs_insert_dir_item(trans
, root
->fs_info
->tree_root
,
2752 name
, namelen
, dir
->i_ino
, &key
,
2757 ret
= btrfs_insert_inode_ref(trans
, root
->fs_info
->tree_root
,
2758 name
, namelen
, objectid
,
2759 root
->fs_info
->sb
->s_root
->d_inode
->i_ino
);
2763 ret
= btrfs_commit_transaction(trans
, root
);
2767 new_root
= btrfs_read_fs_root(root
->fs_info
, &key
, name
, namelen
);
2770 trans
= btrfs_start_transaction(new_root
, 1);
2773 inode
= btrfs_new_inode(trans
, new_root
, "..", 2, new_dirid
,
2775 BTRFS_I(dir
)->block_group
, S_IFDIR
| 0700);
2778 inode
->i_op
= &btrfs_dir_inode_operations
;
2779 inode
->i_fop
= &btrfs_dir_file_operations
;
2780 new_root
->inode
= inode
;
2782 ret
= btrfs_insert_inode_ref(trans
, new_root
, "..", 2, new_dirid
,
2786 ret
= btrfs_update_inode(trans
, new_root
, inode
);
2790 /* Invalidate existing dcache entry for new subvolume. */
2791 btrfs_invalidate_dcache_root(root
, name
, namelen
);
2794 nr
= trans
->blocks_used
;
2795 err
= btrfs_commit_transaction(trans
, new_root
);
2799 mutex_unlock(&root
->fs_info
->fs_mutex
);
2800 btrfs_btree_balance_dirty(root
, nr
);
2801 btrfs_throttle(root
);
2805 static int create_snapshot(struct btrfs_root
*root
, char *name
, int namelen
)
2807 struct btrfs_pending_snapshot
*pending_snapshot
;
2808 struct btrfs_trans_handle
*trans
;
2811 unsigned long nr
= 0;
2813 if (!root
->ref_cows
)
2816 mutex_lock(&root
->fs_info
->fs_mutex
);
2817 ret
= btrfs_check_free_space(root
, 1, 0);
2821 pending_snapshot
= kmalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
2822 if (!pending_snapshot
) {
2826 pending_snapshot
->name
= kmalloc(namelen
+ 1, GFP_NOFS
);
2827 if (!pending_snapshot
->name
) {
2829 kfree(pending_snapshot
);
2832 memcpy(pending_snapshot
->name
, name
, namelen
);
2833 pending_snapshot
->name
[namelen
] = '\0';
2834 trans
= btrfs_start_transaction(root
, 1);
2836 pending_snapshot
->root
= root
;
2837 list_add(&pending_snapshot
->list
,
2838 &trans
->transaction
->pending_snapshots
);
2839 ret
= btrfs_update_inode(trans
, root
, root
->inode
);
2840 err
= btrfs_commit_transaction(trans
, root
);
2843 mutex_unlock(&root
->fs_info
->fs_mutex
);
2844 btrfs_btree_balance_dirty(root
, nr
);
2845 btrfs_throttle(root
);
2849 unsigned long btrfs_force_ra(struct address_space
*mapping
,
2850 struct file_ra_state
*ra
, struct file
*file
,
2851 pgoff_t offset
, pgoff_t last_index
)
2853 pgoff_t req_size
= last_index
- offset
+ 1;
2855 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2856 offset
= page_cache_readahead(mapping
, ra
, file
, offset
, req_size
);
2859 page_cache_sync_readahead(mapping
, ra
, file
, offset
, req_size
);
2860 return offset
+ req_size
;
2864 int btrfs_defrag_file(struct file
*file
) {
2865 struct inode
*inode
= fdentry(file
)->d_inode
;
2866 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2867 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2869 unsigned long last_index
;
2870 unsigned long ra_pages
= root
->fs_info
->bdi
.ra_pages
;
2871 unsigned long total_read
= 0;
2877 mutex_lock(&root
->fs_info
->fs_mutex
);
2878 ret
= btrfs_check_free_space(root
, inode
->i_size
, 0);
2879 mutex_unlock(&root
->fs_info
->fs_mutex
);
2883 mutex_lock(&inode
->i_mutex
);
2884 last_index
= inode
->i_size
>> PAGE_CACHE_SHIFT
;
2885 for (i
= 0; i
<= last_index
; i
++) {
2886 if (total_read
% ra_pages
== 0) {
2887 btrfs_force_ra(inode
->i_mapping
, &file
->f_ra
, file
, i
,
2888 min(last_index
, i
+ ra_pages
- 1));
2891 page
= grab_cache_page(inode
->i_mapping
, i
);
2894 if (!PageUptodate(page
)) {
2895 btrfs_readpage(NULL
, page
);
2897 if (!PageUptodate(page
)) {
2899 page_cache_release(page
);
2904 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
2905 ClearPageDirty(page
);
2907 cancel_dirty_page(page
, PAGE_CACHE_SIZE
);
2909 wait_on_page_writeback(page
);
2910 set_page_extent_mapped(page
);
2912 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
2913 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
2915 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
2916 set_extent_delalloc(io_tree
, page_start
,
2917 page_end
, GFP_NOFS
);
2919 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
2920 set_page_dirty(page
);
2922 page_cache_release(page
);
2923 balance_dirty_pages_ratelimited_nr(inode
->i_mapping
, 1);
2927 mutex_unlock(&inode
->i_mutex
);
2931 static int btrfs_ioctl_resize(struct btrfs_root
*root
, void __user
*arg
)
2936 struct btrfs_ioctl_vol_args
*vol_args
;
2937 struct btrfs_trans_handle
*trans
;
2938 struct btrfs_device
*device
= NULL
;
2940 char *devstr
= NULL
;
2945 vol_args
= kmalloc(sizeof(*vol_args
), GFP_NOFS
);
2950 if (copy_from_user(vol_args
, arg
, sizeof(*vol_args
))) {
2954 namelen
= strlen(vol_args
->name
);
2955 if (namelen
> BTRFS_VOL_NAME_MAX
) {
2960 mutex_lock(&root
->fs_info
->fs_mutex
);
2961 sizestr
= vol_args
->name
;
2962 devstr
= strchr(sizestr
, ':');
2965 sizestr
= devstr
+ 1;
2967 devstr
= vol_args
->name
;
2968 devid
= simple_strtoull(devstr
, &end
, 10);
2969 printk("resizing devid %Lu\n", devid
);
2971 device
= btrfs_find_device(root
, devid
, NULL
);
2973 printk("resizer unable to find device %Lu\n", devid
);
2977 if (!strcmp(sizestr
, "max"))
2978 new_size
= device
->bdev
->bd_inode
->i_size
;
2980 if (sizestr
[0] == '-') {
2983 } else if (sizestr
[0] == '+') {
2987 new_size
= btrfs_parse_size(sizestr
);
2988 if (new_size
== 0) {
2994 old_size
= device
->total_bytes
;
2997 if (new_size
> old_size
) {
3001 new_size
= old_size
- new_size
;
3002 } else if (mod
> 0) {
3003 new_size
= old_size
+ new_size
;
3006 if (new_size
< 256 * 1024 * 1024) {
3010 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
3015 do_div(new_size
, root
->sectorsize
);
3016 new_size
*= root
->sectorsize
;
3018 printk("new size for %s is %llu\n", device
->name
, (unsigned long long)new_size
);
3020 if (new_size
> old_size
) {
3021 trans
= btrfs_start_transaction(root
, 1);
3022 ret
= btrfs_grow_device(trans
, device
, new_size
);
3023 btrfs_commit_transaction(trans
, root
);
3025 ret
= btrfs_shrink_device(device
, new_size
);
3029 mutex_unlock(&root
->fs_info
->fs_mutex
);
3035 static int noinline
btrfs_ioctl_snap_create(struct btrfs_root
*root
,
3038 struct btrfs_ioctl_vol_args
*vol_args
;
3039 struct btrfs_dir_item
*di
;
3040 struct btrfs_path
*path
;
3045 vol_args
= kmalloc(sizeof(*vol_args
), GFP_NOFS
);
3050 if (copy_from_user(vol_args
, arg
, sizeof(*vol_args
))) {
3055 namelen
= strlen(vol_args
->name
);
3056 if (namelen
> BTRFS_VOL_NAME_MAX
) {
3060 if (strchr(vol_args
->name
, '/')) {
3065 path
= btrfs_alloc_path();
3071 root_dirid
= root
->fs_info
->sb
->s_root
->d_inode
->i_ino
,
3072 mutex_lock(&root
->fs_info
->fs_mutex
);
3073 di
= btrfs_lookup_dir_item(NULL
, root
->fs_info
->tree_root
,
3075 vol_args
->name
, namelen
, 0);
3076 mutex_unlock(&root
->fs_info
->fs_mutex
);
3077 btrfs_free_path(path
);
3079 if (di
&& !IS_ERR(di
)) {
3089 if (root
== root
->fs_info
->tree_root
)
3090 ret
= create_subvol(root
, vol_args
->name
, namelen
);
3092 ret
= create_snapshot(root
, vol_args
->name
, namelen
);
3098 static int btrfs_ioctl_defrag(struct file
*file
)
3100 struct inode
*inode
= fdentry(file
)->d_inode
;
3101 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3103 switch (inode
->i_mode
& S_IFMT
) {
3105 mutex_lock(&root
->fs_info
->fs_mutex
);
3106 btrfs_defrag_root(root
, 0);
3107 btrfs_defrag_root(root
->fs_info
->extent_root
, 0);
3108 mutex_unlock(&root
->fs_info
->fs_mutex
);
3111 btrfs_defrag_file(file
);
3118 long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
3120 struct btrfs_ioctl_vol_args
*vol_args
;
3123 vol_args
= kmalloc(sizeof(*vol_args
), GFP_NOFS
);
3128 if (copy_from_user(vol_args
, arg
, sizeof(*vol_args
))) {
3132 ret
= btrfs_init_new_device(root
, vol_args
->name
);
3139 long btrfs_ioctl_rm_dev(struct btrfs_root
*root
, void __user
*arg
)
3141 struct btrfs_ioctl_vol_args
*vol_args
;
3144 vol_args
= kmalloc(sizeof(*vol_args
), GFP_NOFS
);
3149 if (copy_from_user(vol_args
, arg
, sizeof(*vol_args
))) {
3153 ret
= btrfs_rm_device(root
, vol_args
->name
);
3160 int dup_item_to_inode(struct btrfs_trans_handle
*trans
,
3161 struct btrfs_root
*root
,
3162 struct btrfs_path
*path
,
3163 struct extent_buffer
*leaf
,
3165 struct btrfs_key
*key
,
3169 int len
= btrfs_item_size_nr(leaf
, slot
);
3170 struct btrfs_key ckey
= *key
;
3173 dup
= kmalloc(len
, GFP_NOFS
);
3177 read_extent_buffer(leaf
, dup
, btrfs_item_ptr_offset(leaf
, slot
), len
);
3178 btrfs_release_path(root
, path
);
3180 ckey
.objectid
= destino
;
3181 ret
= btrfs_insert_item(trans
, root
, &ckey
, dup
, len
);
3186 long btrfs_ioctl_clone(struct file
*file
, unsigned long src_fd
)
3188 struct inode
*inode
= fdentry(file
)->d_inode
;
3189 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3190 struct file
*src_file
;
3192 struct btrfs_trans_handle
*trans
;
3195 struct btrfs_path
*path
;
3196 struct btrfs_key key
;
3197 struct extent_buffer
*leaf
;
3201 src_file
= fget(src_fd
);
3204 src
= src_file
->f_dentry
->d_inode
;
3207 if (src
->i_sb
!= inode
->i_sb
)
3211 mutex_lock(&inode
->i_mutex
);
3212 mutex_lock(&src
->i_mutex
);
3214 mutex_lock(&src
->i_mutex
);
3215 mutex_lock(&inode
->i_mutex
);
3222 /* do any pending delalloc/csum calc on src, one way or
3223 another, and lock file content */
3225 filemap_write_and_wait(src
->i_mapping
);
3226 lock_extent(&BTRFS_I(src
)->io_tree
, 0, (u64
)-1, GFP_NOFS
);
3227 if (BTRFS_I(src
)->delalloc_bytes
== 0)
3229 unlock_extent(&BTRFS_I(src
)->io_tree
, 0, (u64
)-1, GFP_NOFS
);
3232 mutex_lock(&root
->fs_info
->fs_mutex
);
3233 trans
= btrfs_start_transaction(root
, 0);
3234 path
= btrfs_alloc_path();
3240 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3241 key
.objectid
= src
->i_ino
;
3247 * note the key will change type as we walk through the
3250 ret
= btrfs_search_slot(trans
, root
, &key
, path
, 0, 0);
3254 if (path
->slots
[0] >= btrfs_header_nritems(path
->nodes
[0])) {
3255 ret
= btrfs_next_leaf(root
, path
);
3261 leaf
= path
->nodes
[0];
3262 slot
= path
->slots
[0];
3263 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
3264 nritems
= btrfs_header_nritems(leaf
);
3266 if (btrfs_key_type(&key
) > BTRFS_CSUM_ITEM_KEY
||
3267 key
.objectid
!= src
->i_ino
)
3270 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
) {
3271 struct btrfs_file_extent_item
*extent
;
3274 extent
= btrfs_item_ptr(leaf
, slot
,
3275 struct btrfs_file_extent_item
);
3276 found_type
= btrfs_file_extent_type(leaf
, extent
);
3277 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
3278 u64 len
= btrfs_file_extent_num_bytes(leaf
,
3280 u64 ds
= btrfs_file_extent_disk_bytenr(leaf
,
3282 u64 dl
= btrfs_file_extent_disk_num_bytes(leaf
,
3284 u64 off
= btrfs_file_extent_offset(leaf
,
3286 btrfs_insert_file_extent(trans
, root
,
3289 /* ds == 0 means there's a hole */
3291 btrfs_inc_extent_ref(trans
, root
,
3293 root
->root_key
.objectid
,
3297 pos
= key
.offset
+ len
;
3298 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
3299 ret
= dup_item_to_inode(trans
, root
, path
,
3304 pos
= key
.offset
+ btrfs_item_size_nr(leaf
,
3307 } else if (btrfs_key_type(&key
) == BTRFS_CSUM_ITEM_KEY
) {
3308 ret
= dup_item_to_inode(trans
, root
, path
, leaf
,
3309 slot
, &key
, inode
->i_ino
);
3315 btrfs_release_path(root
, path
);
3320 btrfs_free_path(path
);
3322 inode
->i_blocks
= src
->i_blocks
;
3323 i_size_write(inode
, src
->i_size
);
3324 btrfs_update_inode(trans
, root
, inode
);
3326 unlock_extent(&BTRFS_I(src
)->io_tree
, 0, (u64
)-1, GFP_NOFS
);
3328 btrfs_end_transaction(trans
, root
);
3329 mutex_unlock(&root
->fs_info
->fs_mutex
);
3332 mutex_unlock(&src
->i_mutex
);
3333 mutex_unlock(&inode
->i_mutex
);
3340 * there are many ways the trans_start and trans_end ioctls can lead
3341 * to deadlocks. They should only be used by applications that
3342 * basically own the machine, and have a very in depth understanding
3343 * of all the possible deadlocks and enospc problems.
3345 long btrfs_ioctl_trans_start(struct file
*file
)
3347 struct inode
*inode
= fdentry(file
)->d_inode
;
3348 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3349 struct btrfs_trans_handle
*trans
;
3352 mutex_lock(&root
->fs_info
->fs_mutex
);
3353 if (file
->private_data
) {
3357 trans
= btrfs_start_transaction(root
, 0);
3359 file
->private_data
= trans
;
3362 /*printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);*/
3364 mutex_unlock(&root
->fs_info
->fs_mutex
);
3369 * there are many ways the trans_start and trans_end ioctls can lead
3370 * to deadlocks. They should only be used by applications that
3371 * basically own the machine, and have a very in depth understanding
3372 * of all the possible deadlocks and enospc problems.
3374 long btrfs_ioctl_trans_end(struct file
*file
)
3376 struct inode
*inode
= fdentry(file
)->d_inode
;
3377 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3378 struct btrfs_trans_handle
*trans
;
3381 mutex_lock(&root
->fs_info
->fs_mutex
);
3382 trans
= file
->private_data
;
3387 btrfs_end_transaction(trans
, root
);
3388 file
->private_data
= 0;
3390 mutex_unlock(&root
->fs_info
->fs_mutex
);
3394 long btrfs_ioctl(struct file
*file
, unsigned int
3395 cmd
, unsigned long arg
)
3397 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3400 case BTRFS_IOC_SNAP_CREATE
:
3401 return btrfs_ioctl_snap_create(root
, (void __user
*)arg
);
3402 case BTRFS_IOC_DEFRAG
:
3403 return btrfs_ioctl_defrag(file
);
3404 case BTRFS_IOC_RESIZE
:
3405 return btrfs_ioctl_resize(root
, (void __user
*)arg
);
3406 case BTRFS_IOC_ADD_DEV
:
3407 return btrfs_ioctl_add_dev(root
, (void __user
*)arg
);
3408 case BTRFS_IOC_RM_DEV
:
3409 return btrfs_ioctl_rm_dev(root
, (void __user
*)arg
);
3410 case BTRFS_IOC_BALANCE
:
3411 return btrfs_balance(root
->fs_info
->dev_root
);
3412 case BTRFS_IOC_CLONE
:
3413 return btrfs_ioctl_clone(file
, arg
);
3414 case BTRFS_IOC_TRANS_START
:
3415 return btrfs_ioctl_trans_start(file
);
3416 case BTRFS_IOC_TRANS_END
:
3417 return btrfs_ioctl_trans_end(file
);
3418 case BTRFS_IOC_SYNC
:
3419 btrfs_sync_fs(file
->f_dentry
->d_sb
, 1);
3427 * Called inside transaction, so use GFP_NOFS
3429 struct inode
*btrfs_alloc_inode(struct super_block
*sb
)
3431 struct btrfs_inode
*ei
;
3433 ei
= kmem_cache_alloc(btrfs_inode_cachep
, GFP_NOFS
);
3437 ei
->ordered_trans
= 0;
3438 return &ei
->vfs_inode
;
3441 void btrfs_destroy_inode(struct inode
*inode
)
3443 WARN_ON(!list_empty(&inode
->i_dentry
));
3444 WARN_ON(inode
->i_data
.nrpages
);
3446 btrfs_drop_extent_cache(inode
, 0, (u64
)-1);
3447 kmem_cache_free(btrfs_inode_cachep
, BTRFS_I(inode
));
3450 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3451 static void init_once(struct kmem_cache
* cachep
, void *foo
)
3453 static void init_once(void * foo
, struct kmem_cache
* cachep
,
3454 unsigned long flags
)
3457 struct btrfs_inode
*ei
= (struct btrfs_inode
*) foo
;
3459 inode_init_once(&ei
->vfs_inode
);
3462 void btrfs_destroy_cachep(void)
3464 if (btrfs_inode_cachep
)
3465 kmem_cache_destroy(btrfs_inode_cachep
);
3466 if (btrfs_trans_handle_cachep
)
3467 kmem_cache_destroy(btrfs_trans_handle_cachep
);
3468 if (btrfs_transaction_cachep
)
3469 kmem_cache_destroy(btrfs_transaction_cachep
);
3470 if (btrfs_bit_radix_cachep
)
3471 kmem_cache_destroy(btrfs_bit_radix_cachep
);
3472 if (btrfs_path_cachep
)
3473 kmem_cache_destroy(btrfs_path_cachep
);
3476 struct kmem_cache
*btrfs_cache_create(const char *name
, size_t size
,
3477 unsigned long extra_flags
,
3478 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3479 void (*ctor
)(struct kmem_cache
*, void *)
3481 void (*ctor
)(void *, struct kmem_cache
*,
3486 return kmem_cache_create(name
, size
, 0, (SLAB_RECLAIM_ACCOUNT
|
3487 SLAB_MEM_SPREAD
| extra_flags
), ctor
3488 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3494 int btrfs_init_cachep(void)
3496 btrfs_inode_cachep
= btrfs_cache_create("btrfs_inode_cache",
3497 sizeof(struct btrfs_inode
),
3499 if (!btrfs_inode_cachep
)
3501 btrfs_trans_handle_cachep
=
3502 btrfs_cache_create("btrfs_trans_handle_cache",
3503 sizeof(struct btrfs_trans_handle
),
3505 if (!btrfs_trans_handle_cachep
)
3507 btrfs_transaction_cachep
= btrfs_cache_create("btrfs_transaction_cache",
3508 sizeof(struct btrfs_transaction
),
3510 if (!btrfs_transaction_cachep
)
3512 btrfs_path_cachep
= btrfs_cache_create("btrfs_path_cache",
3513 sizeof(struct btrfs_path
),
3515 if (!btrfs_path_cachep
)
3517 btrfs_bit_radix_cachep
= btrfs_cache_create("btrfs_radix", 256,
3518 SLAB_DESTROY_BY_RCU
, NULL
);
3519 if (!btrfs_bit_radix_cachep
)
3523 btrfs_destroy_cachep();
3527 static int btrfs_getattr(struct vfsmount
*mnt
,
3528 struct dentry
*dentry
, struct kstat
*stat
)
3530 struct inode
*inode
= dentry
->d_inode
;
3531 generic_fillattr(inode
, stat
);
3532 stat
->blksize
= PAGE_CACHE_SIZE
;
3533 stat
->blocks
= inode
->i_blocks
+ (BTRFS_I(inode
)->delalloc_bytes
>> 9);
3537 static int btrfs_rename(struct inode
* old_dir
, struct dentry
*old_dentry
,
3538 struct inode
* new_dir
,struct dentry
*new_dentry
)
3540 struct btrfs_trans_handle
*trans
;
3541 struct btrfs_root
*root
= BTRFS_I(old_dir
)->root
;
3542 struct inode
*new_inode
= new_dentry
->d_inode
;
3543 struct inode
*old_inode
= old_dentry
->d_inode
;
3544 struct timespec ctime
= CURRENT_TIME
;
3545 struct btrfs_path
*path
;
3548 if (S_ISDIR(old_inode
->i_mode
) && new_inode
&&
3549 new_inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
3553 mutex_lock(&root
->fs_info
->fs_mutex
);
3554 ret
= btrfs_check_free_space(root
, 1, 0);
3558 trans
= btrfs_start_transaction(root
, 1);
3560 btrfs_set_trans_block_group(trans
, new_dir
);
3561 path
= btrfs_alloc_path();
3567 old_dentry
->d_inode
->i_nlink
++;
3568 old_dir
->i_ctime
= old_dir
->i_mtime
= ctime
;
3569 new_dir
->i_ctime
= new_dir
->i_mtime
= ctime
;
3570 old_inode
->i_ctime
= ctime
;
3572 ret
= btrfs_unlink_trans(trans
, root
, old_dir
, old_dentry
);
3577 new_inode
->i_ctime
= CURRENT_TIME
;
3578 ret
= btrfs_unlink_trans(trans
, root
, new_dir
, new_dentry
);
3582 ret
= btrfs_add_link(trans
, new_dentry
, old_inode
, 1);
3587 btrfs_free_path(path
);
3588 btrfs_end_transaction(trans
, root
);
3590 mutex_unlock(&root
->fs_info
->fs_mutex
);
3594 static int btrfs_symlink(struct inode
*dir
, struct dentry
*dentry
,
3595 const char *symname
)
3597 struct btrfs_trans_handle
*trans
;
3598 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
3599 struct btrfs_path
*path
;
3600 struct btrfs_key key
;
3601 struct inode
*inode
= NULL
;
3608 struct btrfs_file_extent_item
*ei
;
3609 struct extent_buffer
*leaf
;
3610 unsigned long nr
= 0;
3612 name_len
= strlen(symname
) + 1;
3613 if (name_len
> BTRFS_MAX_INLINE_DATA_SIZE(root
))
3614 return -ENAMETOOLONG
;
3616 mutex_lock(&root
->fs_info
->fs_mutex
);
3617 err
= btrfs_check_free_space(root
, 1, 0);
3621 trans
= btrfs_start_transaction(root
, 1);
3622 btrfs_set_trans_block_group(trans
, dir
);
3624 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
3630 inode
= btrfs_new_inode(trans
, root
, dentry
->d_name
.name
,
3632 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
3633 BTRFS_I(dir
)->block_group
, S_IFLNK
|S_IRWXUGO
);
3634 err
= PTR_ERR(inode
);
3638 btrfs_set_trans_block_group(trans
, inode
);
3639 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
3643 inode
->i_mapping
->a_ops
= &btrfs_aops
;
3644 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3645 inode
->i_fop
= &btrfs_file_operations
;
3646 inode
->i_op
= &btrfs_file_inode_operations
;
3647 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
3648 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
3649 inode
->i_mapping
, GFP_NOFS
);
3650 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
3651 inode
->i_mapping
, GFP_NOFS
);
3652 BTRFS_I(inode
)->delalloc_bytes
= 0;
3653 atomic_set(&BTRFS_I(inode
)->ordered_writeback
, 0);
3654 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
3656 dir
->i_sb
->s_dirt
= 1;
3657 btrfs_update_inode_block_group(trans
, inode
);
3658 btrfs_update_inode_block_group(trans
, dir
);
3662 path
= btrfs_alloc_path();
3664 key
.objectid
= inode
->i_ino
;
3666 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
3667 datasize
= btrfs_file_extent_calc_inline_size(name_len
);
3668 err
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
3674 leaf
= path
->nodes
[0];
3675 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
3676 struct btrfs_file_extent_item
);
3677 btrfs_set_file_extent_generation(leaf
, ei
, trans
->transid
);
3678 btrfs_set_file_extent_type(leaf
, ei
,
3679 BTRFS_FILE_EXTENT_INLINE
);
3680 ptr
= btrfs_file_extent_inline_start(ei
);
3681 write_extent_buffer(leaf
, symname
, ptr
, name_len
);
3682 btrfs_mark_buffer_dirty(leaf
);
3683 btrfs_free_path(path
);
3685 inode
->i_op
= &btrfs_symlink_inode_operations
;
3686 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
3687 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3688 inode
->i_size
= name_len
- 1;
3689 err
= btrfs_update_inode(trans
, root
, inode
);
3694 nr
= trans
->blocks_used
;
3695 btrfs_end_transaction(trans
, root
);
3697 mutex_unlock(&root
->fs_info
->fs_mutex
);
3699 inode_dec_link_count(inode
);
3702 btrfs_btree_balance_dirty(root
, nr
);
3703 btrfs_throttle(root
);
3707 static int btrfs_permission(struct inode
*inode
, int mask
,
3708 struct nameidata
*nd
)
3710 if (btrfs_test_flag(inode
, READONLY
) && (mask
& MAY_WRITE
))
3712 return generic_permission(inode
, mask
, NULL
);
3715 static struct inode_operations btrfs_dir_inode_operations
= {
3716 .lookup
= btrfs_lookup
,
3717 .create
= btrfs_create
,
3718 .unlink
= btrfs_unlink
,
3720 .mkdir
= btrfs_mkdir
,
3721 .rmdir
= btrfs_rmdir
,
3722 .rename
= btrfs_rename
,
3723 .symlink
= btrfs_symlink
,
3724 .setattr
= btrfs_setattr
,
3725 .mknod
= btrfs_mknod
,
3726 .setxattr
= generic_setxattr
,
3727 .getxattr
= generic_getxattr
,
3728 .listxattr
= btrfs_listxattr
,
3729 .removexattr
= generic_removexattr
,
3730 .permission
= btrfs_permission
,
3732 static struct inode_operations btrfs_dir_ro_inode_operations
= {
3733 .lookup
= btrfs_lookup
,
3734 .permission
= btrfs_permission
,
3736 static struct file_operations btrfs_dir_file_operations
= {
3737 .llseek
= generic_file_llseek
,
3738 .read
= generic_read_dir
,
3739 .readdir
= btrfs_readdir
,
3740 .unlocked_ioctl
= btrfs_ioctl
,
3741 #ifdef CONFIG_COMPAT
3742 .compat_ioctl
= btrfs_ioctl
,
3744 .release
= btrfs_release_file
,
3747 static struct extent_io_ops btrfs_extent_io_ops
= {
3748 .fill_delalloc
= run_delalloc_range
,
3749 .submit_bio_hook
= btrfs_submit_bio_hook
,
3750 .merge_bio_hook
= btrfs_merge_bio_hook
,
3751 .readpage_io_hook
= btrfs_readpage_io_hook
,
3752 .readpage_end_io_hook
= btrfs_readpage_end_io_hook
,
3753 .readpage_io_failed_hook
= btrfs_io_failed_hook
,
3754 .set_bit_hook
= btrfs_set_bit_hook
,
3755 .clear_bit_hook
= btrfs_clear_bit_hook
,
3758 static struct address_space_operations btrfs_aops
= {
3759 .readpage
= btrfs_readpage
,
3760 .writepage
= btrfs_writepage
,
3761 .writepages
= btrfs_writepages
,
3762 .readpages
= btrfs_readpages
,
3763 .sync_page
= block_sync_page
,
3765 .direct_IO
= btrfs_direct_IO
,
3766 .invalidatepage
= btrfs_invalidatepage
,
3767 .releasepage
= btrfs_releasepage
,
3768 .set_page_dirty
= __set_page_dirty_nobuffers
,
3771 static struct address_space_operations btrfs_symlink_aops
= {
3772 .readpage
= btrfs_readpage
,
3773 .writepage
= btrfs_writepage
,
3774 .invalidatepage
= btrfs_invalidatepage
,
3775 .releasepage
= btrfs_releasepage
,
3778 static struct inode_operations btrfs_file_inode_operations
= {
3779 .truncate
= btrfs_truncate
,
3780 .getattr
= btrfs_getattr
,
3781 .setattr
= btrfs_setattr
,
3782 .setxattr
= generic_setxattr
,
3783 .getxattr
= generic_getxattr
,
3784 .listxattr
= btrfs_listxattr
,
3785 .removexattr
= generic_removexattr
,
3786 .permission
= btrfs_permission
,
3788 static struct inode_operations btrfs_special_inode_operations
= {
3789 .getattr
= btrfs_getattr
,
3790 .setattr
= btrfs_setattr
,
3791 .permission
= btrfs_permission
,
3793 static struct inode_operations btrfs_symlink_inode_operations
= {
3794 .readlink
= generic_readlink
,
3795 .follow_link
= page_follow_link_light
,
3796 .put_link
= page_put_link
,
3797 .permission
= btrfs_permission
,