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/bio.h>
20 #include <linux/buffer_head.h>
22 #include <linux/pagemap.h>
23 #include <linux/highmem.h>
24 #include <linux/time.h>
25 #include <linux/init.h>
26 #include <linux/string.h>
27 #include <linux/smp_lock.h>
28 #include <linux/backing-dev.h>
29 #include <linux/mpage.h>
30 #include <linux/swap.h>
31 #include <linux/writeback.h>
32 #include <linux/statfs.h>
33 #include <linux/compat.h>
34 #include <linux/bit_spinlock.h>
35 #include <linux/version.h>
36 #include <linux/xattr.h>
39 #include "transaction.h"
40 #include "btrfs_inode.h"
42 #include "print-tree.h"
45 struct btrfs_iget_args
{
47 struct btrfs_root
*root
;
50 static struct inode_operations btrfs_dir_inode_operations
;
51 static struct inode_operations btrfs_symlink_inode_operations
;
52 static struct inode_operations btrfs_dir_ro_inode_operations
;
53 static struct inode_operations btrfs_special_inode_operations
;
54 static struct inode_operations btrfs_file_inode_operations
;
55 static struct address_space_operations btrfs_aops
;
56 static struct address_space_operations btrfs_symlink_aops
;
57 static struct file_operations btrfs_dir_file_operations
;
58 static struct extent_io_ops btrfs_extent_io_ops
;
60 static struct kmem_cache
*btrfs_inode_cachep
;
61 struct kmem_cache
*btrfs_trans_handle_cachep
;
62 struct kmem_cache
*btrfs_transaction_cachep
;
63 struct kmem_cache
*btrfs_bit_radix_cachep
;
64 struct kmem_cache
*btrfs_path_cachep
;
67 static unsigned char btrfs_type_by_mode
[S_IFMT
>> S_SHIFT
] = {
68 [S_IFREG
>> S_SHIFT
] = BTRFS_FT_REG_FILE
,
69 [S_IFDIR
>> S_SHIFT
] = BTRFS_FT_DIR
,
70 [S_IFCHR
>> S_SHIFT
] = BTRFS_FT_CHRDEV
,
71 [S_IFBLK
>> S_SHIFT
] = BTRFS_FT_BLKDEV
,
72 [S_IFIFO
>> S_SHIFT
] = BTRFS_FT_FIFO
,
73 [S_IFSOCK
>> S_SHIFT
] = BTRFS_FT_SOCK
,
74 [S_IFLNK
>> S_SHIFT
] = BTRFS_FT_SYMLINK
,
77 int btrfs_check_free_space(struct btrfs_root
*root
, u64 num_required
,
80 u64 total
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
81 u64 used
= btrfs_super_bytes_used(&root
->fs_info
->super_copy
);
92 spin_lock(&root
->fs_info
->delalloc_lock
);
93 if (used
+ root
->fs_info
->delalloc_bytes
+ num_required
> thresh
)
95 spin_unlock(&root
->fs_info
->delalloc_lock
);
99 static int cow_file_range(struct inode
*inode
, u64 start
, u64 end
)
101 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
102 struct btrfs_trans_handle
*trans
;
106 u64 blocksize
= root
->sectorsize
;
107 u64 orig_start
= start
;
109 struct btrfs_key ins
;
112 trans
= btrfs_start_transaction(root
, 1);
114 btrfs_set_trans_block_group(trans
, inode
);
116 num_bytes
= (end
- start
+ blocksize
) & ~(blocksize
- 1);
117 num_bytes
= max(blocksize
, num_bytes
);
118 ret
= btrfs_drop_extents(trans
, root
, inode
,
119 start
, start
+ num_bytes
, start
, &alloc_hint
);
120 orig_num_bytes
= num_bytes
;
122 if (alloc_hint
== EXTENT_MAP_INLINE
)
125 BUG_ON(num_bytes
> btrfs_super_total_bytes(&root
->fs_info
->super_copy
));
127 while(num_bytes
> 0) {
128 cur_alloc_size
= min(num_bytes
, root
->fs_info
->max_extent
);
129 ret
= btrfs_alloc_extent(trans
, root
, cur_alloc_size
,
131 root
->root_key
.objectid
,
133 inode
->i_ino
, start
, 0,
134 alloc_hint
, (u64
)-1, &ins
, 1);
139 cur_alloc_size
= ins
.offset
;
140 ret
= btrfs_insert_file_extent(trans
, root
, inode
->i_ino
,
141 start
, ins
.objectid
, ins
.offset
,
143 inode
->i_blocks
+= ins
.offset
>> 9;
144 btrfs_check_file(root
, inode
);
145 if (num_bytes
< cur_alloc_size
) {
146 printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes
,
150 num_bytes
-= cur_alloc_size
;
151 alloc_hint
= ins
.objectid
+ ins
.offset
;
152 start
+= cur_alloc_size
;
154 btrfs_drop_extent_cache(inode
, orig_start
,
155 orig_start
+ orig_num_bytes
- 1);
156 btrfs_add_ordered_inode(inode
);
157 btrfs_update_inode(trans
, root
, inode
);
159 btrfs_end_transaction(trans
, root
);
163 static int run_delalloc_nocow(struct inode
*inode
, u64 start
, u64 end
)
171 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
172 struct extent_buffer
*leaf
;
174 struct btrfs_path
*path
;
175 struct btrfs_file_extent_item
*item
;
178 struct btrfs_key found_key
;
180 total_fs_bytes
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
181 path
= btrfs_alloc_path();
184 ret
= btrfs_lookup_file_extent(NULL
, root
, path
,
185 inode
->i_ino
, start
, 0);
187 btrfs_free_path(path
);
193 if (path
->slots
[0] == 0)
198 leaf
= path
->nodes
[0];
199 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
200 struct btrfs_file_extent_item
);
202 /* are we inside the extent that was found? */
203 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
204 found_type
= btrfs_key_type(&found_key
);
205 if (found_key
.objectid
!= inode
->i_ino
||
206 found_type
!= BTRFS_EXTENT_DATA_KEY
) {
210 found_type
= btrfs_file_extent_type(leaf
, item
);
211 extent_start
= found_key
.offset
;
212 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
213 u64 extent_num_bytes
;
215 extent_num_bytes
= btrfs_file_extent_num_bytes(leaf
, item
);
216 extent_end
= extent_start
+ extent_num_bytes
;
219 if (loops
&& start
!= extent_start
)
222 if (start
< extent_start
|| start
>= extent_end
)
225 cow_end
= min(end
, extent_end
- 1);
226 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
231 * we may be called by the resizer, make sure we're inside
232 * the limits of the FS
234 if (bytenr
+ extent_num_bytes
> total_fs_bytes
)
237 if (btrfs_count_snapshots_in_path(root
, path
, bytenr
) != 1) {
247 btrfs_free_path(path
);
250 btrfs_release_path(root
, path
);
255 cow_file_range(inode
, start
, cow_end
);
260 static int run_delalloc_range(struct inode
*inode
, u64 start
, u64 end
)
262 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
264 mutex_lock(&root
->fs_info
->fs_mutex
);
265 if (btrfs_test_opt(root
, NODATACOW
) ||
266 btrfs_test_flag(inode
, NODATACOW
))
267 ret
= run_delalloc_nocow(inode
, start
, end
);
269 ret
= cow_file_range(inode
, start
, end
);
271 mutex_unlock(&root
->fs_info
->fs_mutex
);
275 int btrfs_set_bit_hook(struct inode
*inode
, u64 start
, u64 end
,
276 unsigned long old
, unsigned long bits
)
278 if (!(old
& EXTENT_DELALLOC
) && (bits
& EXTENT_DELALLOC
)) {
279 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
280 spin_lock(&root
->fs_info
->delalloc_lock
);
281 BTRFS_I(inode
)->delalloc_bytes
+= end
- start
+ 1;
282 root
->fs_info
->delalloc_bytes
+= end
- start
+ 1;
283 spin_unlock(&root
->fs_info
->delalloc_lock
);
288 int btrfs_clear_bit_hook(struct inode
*inode
, u64 start
, u64 end
,
289 unsigned long old
, unsigned long bits
)
291 if ((old
& EXTENT_DELALLOC
) && (bits
& EXTENT_DELALLOC
)) {
292 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
293 spin_lock(&root
->fs_info
->delalloc_lock
);
294 if (end
- start
+ 1 > root
->fs_info
->delalloc_bytes
) {
295 printk("warning: delalloc account %Lu %Lu\n",
296 end
- start
+ 1, root
->fs_info
->delalloc_bytes
);
297 root
->fs_info
->delalloc_bytes
= 0;
298 BTRFS_I(inode
)->delalloc_bytes
= 0;
300 root
->fs_info
->delalloc_bytes
-= end
- start
+ 1;
301 BTRFS_I(inode
)->delalloc_bytes
-= end
- start
+ 1;
303 spin_unlock(&root
->fs_info
->delalloc_lock
);
308 int btrfs_merge_bio_hook(struct page
*page
, unsigned long offset
,
309 size_t size
, struct bio
*bio
)
311 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
312 struct btrfs_mapping_tree
*map_tree
;
313 u64 logical
= bio
->bi_sector
<< 9;
318 length
= bio
->bi_size
;
319 map_tree
= &root
->fs_info
->mapping_tree
;
321 ret
= btrfs_map_block(map_tree
, READ
, logical
,
322 &map_length
, NULL
, 0);
324 if (map_length
< length
+ size
) {
330 int __btrfs_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
333 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
334 struct btrfs_trans_handle
*trans
;
338 ret
= btrfs_csum_one_bio(root
, bio
, &sums
);
341 mutex_lock(&root
->fs_info
->fs_mutex
);
342 trans
= btrfs_start_transaction(root
, 1);
344 btrfs_set_trans_block_group(trans
, inode
);
345 btrfs_csum_file_blocks(trans
, root
, inode
, bio
, sums
);
347 ret
= btrfs_end_transaction(trans
, root
);
349 mutex_unlock(&root
->fs_info
->fs_mutex
);
353 return btrfs_map_bio(root
, rw
, bio
, mirror_num
);
356 int btrfs_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
359 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
362 if (!(rw
& (1 << BIO_RW
))) {
363 ret
= btrfs_bio_wq_end_io(root
->fs_info
, bio
, 0);
368 if (btrfs_test_opt(root
, NODATASUM
) ||
369 btrfs_test_flag(inode
, NODATASUM
)) {
373 return btrfs_wq_submit_bio(BTRFS_I(inode
)->root
->fs_info
,
374 inode
, rw
, bio
, mirror_num
,
375 __btrfs_submit_bio_hook
);
377 return btrfs_map_bio(root
, rw
, bio
, mirror_num
);
380 int btrfs_readpage_io_hook(struct page
*page
, u64 start
, u64 end
)
383 struct inode
*inode
= page
->mapping
->host
;
384 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
385 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
386 struct btrfs_csum_item
*item
;
387 struct btrfs_path
*path
= NULL
;
390 if (btrfs_test_opt(root
, NODATASUM
) ||
391 btrfs_test_flag(inode
, NODATASUM
))
394 mutex_lock(&root
->fs_info
->fs_mutex
);
395 path
= btrfs_alloc_path();
396 item
= btrfs_lookup_csum(NULL
, root
, path
, inode
->i_ino
, start
, 0);
399 /* a csum that isn't present is a preallocated region. */
400 if (ret
== -ENOENT
|| ret
== -EFBIG
)
403 printk("no csum found for inode %lu start %Lu\n", inode
->i_ino
, start
);
406 read_extent_buffer(path
->nodes
[0], &csum
, (unsigned long)item
,
408 set_state_private(io_tree
, start
, csum
);
411 btrfs_free_path(path
);
412 mutex_unlock(&root
->fs_info
->fs_mutex
);
416 struct io_failure_record
{
424 int btrfs_readpage_io_failed_hook(struct bio
*failed_bio
,
425 struct page
*page
, u64 start
, u64 end
,
426 struct extent_state
*state
)
428 struct io_failure_record
*failrec
= NULL
;
430 struct extent_map
*em
;
431 struct inode
*inode
= page
->mapping
->host
;
432 struct extent_io_tree
*failure_tree
= &BTRFS_I(inode
)->io_failure_tree
;
433 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
439 ret
= get_state_private(failure_tree
, start
, &private);
441 failrec
= kmalloc(sizeof(*failrec
), GFP_NOFS
);
444 failrec
->start
= start
;
445 failrec
->len
= end
- start
+ 1;
446 failrec
->last_mirror
= 0;
448 spin_lock(&em_tree
->lock
);
449 em
= lookup_extent_mapping(em_tree
, start
, failrec
->len
);
450 if (em
->start
> start
|| em
->start
+ em
->len
< start
) {
454 spin_unlock(&em_tree
->lock
);
456 if (!em
|| IS_ERR(em
)) {
460 logical
= start
- em
->start
;
461 logical
= em
->block_start
+ logical
;
462 failrec
->logical
= logical
;
464 set_extent_bits(failure_tree
, start
, end
, EXTENT_LOCKED
|
465 EXTENT_DIRTY
, GFP_NOFS
);
466 set_state_private(failure_tree
, start
,
467 (u64
)(unsigned long)failrec
);
469 failrec
= (struct io_failure_record
*)(unsigned long)private;
471 num_copies
= btrfs_num_copies(
472 &BTRFS_I(inode
)->root
->fs_info
->mapping_tree
,
473 failrec
->logical
, failrec
->len
);
474 failrec
->last_mirror
++;
476 spin_lock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
477 state
= find_first_extent_bit_state(&BTRFS_I(inode
)->io_tree
,
480 if (state
&& state
->start
!= failrec
->start
)
482 spin_unlock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
484 if (!state
|| failrec
->last_mirror
> num_copies
) {
485 set_state_private(failure_tree
, failrec
->start
, 0);
486 clear_extent_bits(failure_tree
, failrec
->start
,
487 failrec
->start
+ failrec
->len
- 1,
488 EXTENT_LOCKED
| EXTENT_DIRTY
, GFP_NOFS
);
492 bio
= bio_alloc(GFP_NOFS
, 1);
493 bio
->bi_private
= state
;
494 bio
->bi_end_io
= failed_bio
->bi_end_io
;
495 bio
->bi_sector
= failrec
->logical
>> 9;
496 bio
->bi_bdev
= failed_bio
->bi_bdev
;
497 bio_add_page(bio
, page
, failrec
->len
, start
- page_offset(page
));
498 btrfs_submit_bio_hook(inode
, READ
, bio
, failrec
->last_mirror
);
502 int btrfs_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
503 struct extent_state
*state
)
505 size_t offset
= start
- ((u64
)page
->index
<< PAGE_CACHE_SHIFT
);
506 struct inode
*inode
= page
->mapping
->host
;
507 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
509 u64
private = ~(u32
)0;
511 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
515 if (btrfs_test_opt(root
, NODATASUM
) ||
516 btrfs_test_flag(inode
, NODATASUM
))
518 if (state
&& state
->start
== start
) {
519 private = state
->private;
522 ret
= get_state_private(io_tree
, start
, &private);
524 local_irq_save(flags
);
525 kaddr
= kmap_atomic(page
, KM_IRQ0
);
529 csum
= btrfs_csum_data(root
, kaddr
+ offset
, csum
, end
- start
+ 1);
530 btrfs_csum_final(csum
, (char *)&csum
);
531 if (csum
!= private) {
534 kunmap_atomic(kaddr
, KM_IRQ0
);
535 local_irq_restore(flags
);
537 /* if the io failure tree for this inode is non-empty,
538 * check to see if we've recovered from a failed IO
541 if (count_range_bits(&BTRFS_I(inode
)->io_failure_tree
, &private,
542 (u64
)-1, 1, EXTENT_DIRTY
)) {
544 struct io_failure_record
*failure
;
545 ret
= get_state_private(&BTRFS_I(inode
)->io_failure_tree
,
546 start
, &private_failure
);
548 failure
= (struct io_failure_record
*)(unsigned long)
550 set_state_private(&BTRFS_I(inode
)->io_failure_tree
,
552 clear_extent_bits(&BTRFS_I(inode
)->io_failure_tree
,
554 failure
->start
+ failure
->len
- 1,
555 EXTENT_DIRTY
| EXTENT_LOCKED
,
563 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
564 page
->mapping
->host
->i_ino
, (unsigned long long)start
, csum
,
566 memset(kaddr
+ offset
, 1, end
- start
+ 1);
567 flush_dcache_page(page
);
568 kunmap_atomic(kaddr
, KM_IRQ0
);
569 local_irq_restore(flags
);
575 void btrfs_read_locked_inode(struct inode
*inode
)
577 struct btrfs_path
*path
;
578 struct extent_buffer
*leaf
;
579 struct btrfs_inode_item
*inode_item
;
580 struct btrfs_timespec
*tspec
;
581 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
582 struct btrfs_key location
;
583 u64 alloc_group_block
;
587 path
= btrfs_alloc_path();
589 mutex_lock(&root
->fs_info
->fs_mutex
);
590 memcpy(&location
, &BTRFS_I(inode
)->location
, sizeof(location
));
592 ret
= btrfs_lookup_inode(NULL
, root
, path
, &location
, 0);
596 leaf
= path
->nodes
[0];
597 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
598 struct btrfs_inode_item
);
600 inode
->i_mode
= btrfs_inode_mode(leaf
, inode_item
);
601 inode
->i_nlink
= btrfs_inode_nlink(leaf
, inode_item
);
602 inode
->i_uid
= btrfs_inode_uid(leaf
, inode_item
);
603 inode
->i_gid
= btrfs_inode_gid(leaf
, inode_item
);
604 inode
->i_size
= btrfs_inode_size(leaf
, inode_item
);
606 tspec
= btrfs_inode_atime(inode_item
);
607 inode
->i_atime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
608 inode
->i_atime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
610 tspec
= btrfs_inode_mtime(inode_item
);
611 inode
->i_mtime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
612 inode
->i_mtime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
614 tspec
= btrfs_inode_ctime(inode_item
);
615 inode
->i_ctime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
616 inode
->i_ctime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
618 inode
->i_blocks
= btrfs_inode_nblocks(leaf
, inode_item
);
619 inode
->i_generation
= btrfs_inode_generation(leaf
, inode_item
);
621 rdev
= btrfs_inode_rdev(leaf
, inode_item
);
623 alloc_group_block
= btrfs_inode_block_group(leaf
, inode_item
);
624 BTRFS_I(inode
)->block_group
= btrfs_lookup_block_group(root
->fs_info
,
626 BTRFS_I(inode
)->flags
= btrfs_inode_flags(leaf
, inode_item
);
627 if (!BTRFS_I(inode
)->block_group
) {
628 BTRFS_I(inode
)->block_group
= btrfs_find_block_group(root
,
630 BTRFS_BLOCK_GROUP_METADATA
, 0);
632 btrfs_free_path(path
);
635 mutex_unlock(&root
->fs_info
->fs_mutex
);
637 switch (inode
->i_mode
& S_IFMT
) {
639 inode
->i_mapping
->a_ops
= &btrfs_aops
;
640 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
641 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
642 inode
->i_fop
= &btrfs_file_operations
;
643 inode
->i_op
= &btrfs_file_inode_operations
;
646 inode
->i_fop
= &btrfs_dir_file_operations
;
647 if (root
== root
->fs_info
->tree_root
)
648 inode
->i_op
= &btrfs_dir_ro_inode_operations
;
650 inode
->i_op
= &btrfs_dir_inode_operations
;
653 inode
->i_op
= &btrfs_symlink_inode_operations
;
654 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
655 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
658 init_special_inode(inode
, inode
->i_mode
, rdev
);
664 btrfs_release_path(root
, path
);
665 btrfs_free_path(path
);
666 mutex_unlock(&root
->fs_info
->fs_mutex
);
667 make_bad_inode(inode
);
670 static void fill_inode_item(struct extent_buffer
*leaf
,
671 struct btrfs_inode_item
*item
,
674 btrfs_set_inode_uid(leaf
, item
, inode
->i_uid
);
675 btrfs_set_inode_gid(leaf
, item
, inode
->i_gid
);
676 btrfs_set_inode_size(leaf
, item
, inode
->i_size
);
677 btrfs_set_inode_mode(leaf
, item
, inode
->i_mode
);
678 btrfs_set_inode_nlink(leaf
, item
, inode
->i_nlink
);
680 btrfs_set_timespec_sec(leaf
, btrfs_inode_atime(item
),
681 inode
->i_atime
.tv_sec
);
682 btrfs_set_timespec_nsec(leaf
, btrfs_inode_atime(item
),
683 inode
->i_atime
.tv_nsec
);
685 btrfs_set_timespec_sec(leaf
, btrfs_inode_mtime(item
),
686 inode
->i_mtime
.tv_sec
);
687 btrfs_set_timespec_nsec(leaf
, btrfs_inode_mtime(item
),
688 inode
->i_mtime
.tv_nsec
);
690 btrfs_set_timespec_sec(leaf
, btrfs_inode_ctime(item
),
691 inode
->i_ctime
.tv_sec
);
692 btrfs_set_timespec_nsec(leaf
, btrfs_inode_ctime(item
),
693 inode
->i_ctime
.tv_nsec
);
695 btrfs_set_inode_nblocks(leaf
, item
, inode
->i_blocks
);
696 btrfs_set_inode_generation(leaf
, item
, inode
->i_generation
);
697 btrfs_set_inode_rdev(leaf
, item
, inode
->i_rdev
);
698 btrfs_set_inode_flags(leaf
, item
, BTRFS_I(inode
)->flags
);
699 btrfs_set_inode_block_group(leaf
, item
,
700 BTRFS_I(inode
)->block_group
->key
.objectid
);
703 int btrfs_update_inode(struct btrfs_trans_handle
*trans
,
704 struct btrfs_root
*root
,
707 struct btrfs_inode_item
*inode_item
;
708 struct btrfs_path
*path
;
709 struct extent_buffer
*leaf
;
712 path
= btrfs_alloc_path();
714 ret
= btrfs_lookup_inode(trans
, root
, path
,
715 &BTRFS_I(inode
)->location
, 1);
722 leaf
= path
->nodes
[0];
723 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
724 struct btrfs_inode_item
);
726 fill_inode_item(leaf
, inode_item
, inode
);
727 btrfs_mark_buffer_dirty(leaf
);
728 btrfs_set_inode_last_trans(trans
, inode
);
731 btrfs_release_path(root
, path
);
732 btrfs_free_path(path
);
737 static int btrfs_unlink_trans(struct btrfs_trans_handle
*trans
,
738 struct btrfs_root
*root
,
740 struct dentry
*dentry
)
742 struct btrfs_path
*path
;
743 const char *name
= dentry
->d_name
.name
;
744 int name_len
= dentry
->d_name
.len
;
746 struct extent_buffer
*leaf
;
747 struct btrfs_dir_item
*di
;
748 struct btrfs_key key
;
750 path
= btrfs_alloc_path();
756 di
= btrfs_lookup_dir_item(trans
, root
, path
, dir
->i_ino
,
766 leaf
= path
->nodes
[0];
767 btrfs_dir_item_key_to_cpu(leaf
, di
, &key
);
768 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
771 btrfs_release_path(root
, path
);
773 di
= btrfs_lookup_dir_index_item(trans
, root
, path
, dir
->i_ino
,
774 key
.objectid
, name
, name_len
, -1);
783 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
785 dentry
->d_inode
->i_ctime
= dir
->i_ctime
;
786 ret
= btrfs_del_inode_ref(trans
, root
, name
, name_len
,
787 dentry
->d_inode
->i_ino
,
788 dentry
->d_parent
->d_inode
->i_ino
);
790 printk("failed to delete reference to %.*s, "
791 "inode %lu parent %lu\n", name_len
, name
,
792 dentry
->d_inode
->i_ino
,
793 dentry
->d_parent
->d_inode
->i_ino
);
796 btrfs_free_path(path
);
798 dir
->i_size
-= name_len
* 2;
799 dir
->i_mtime
= dir
->i_ctime
= CURRENT_TIME
;
800 btrfs_update_inode(trans
, root
, dir
);
801 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
802 dentry
->d_inode
->i_nlink
--;
804 drop_nlink(dentry
->d_inode
);
806 ret
= btrfs_update_inode(trans
, root
, dentry
->d_inode
);
807 dir
->i_sb
->s_dirt
= 1;
812 static int btrfs_unlink(struct inode
*dir
, struct dentry
*dentry
)
814 struct btrfs_root
*root
;
815 struct btrfs_trans_handle
*trans
;
816 struct inode
*inode
= dentry
->d_inode
;
818 unsigned long nr
= 0;
820 root
= BTRFS_I(dir
)->root
;
821 mutex_lock(&root
->fs_info
->fs_mutex
);
823 ret
= btrfs_check_free_space(root
, 1, 1);
827 trans
= btrfs_start_transaction(root
, 1);
829 btrfs_set_trans_block_group(trans
, dir
);
830 ret
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
831 nr
= trans
->blocks_used
;
833 if (inode
->i_nlink
== 0) {
835 /* if the inode isn't linked anywhere,
836 * we don't need to worry about
839 found
= btrfs_del_ordered_inode(inode
);
841 atomic_dec(&inode
->i_count
);
845 btrfs_end_transaction(trans
, root
);
847 mutex_unlock(&root
->fs_info
->fs_mutex
);
848 btrfs_btree_balance_dirty(root
, nr
);
849 btrfs_throttle(root
);
853 static int btrfs_rmdir(struct inode
*dir
, struct dentry
*dentry
)
855 struct inode
*inode
= dentry
->d_inode
;
858 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
859 struct btrfs_trans_handle
*trans
;
860 unsigned long nr
= 0;
862 if (inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
)
865 mutex_lock(&root
->fs_info
->fs_mutex
);
866 ret
= btrfs_check_free_space(root
, 1, 1);
870 trans
= btrfs_start_transaction(root
, 1);
871 btrfs_set_trans_block_group(trans
, dir
);
873 /* now the directory is empty */
874 err
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
879 nr
= trans
->blocks_used
;
880 ret
= btrfs_end_transaction(trans
, root
);
882 mutex_unlock(&root
->fs_info
->fs_mutex
);
883 btrfs_btree_balance_dirty(root
, nr
);
884 btrfs_throttle(root
);
892 * this can truncate away extent items, csum items and directory items.
893 * It starts at a high offset and removes keys until it can't find
894 * any higher than i_size.
896 * csum items that cross the new i_size are truncated to the new size
899 static int btrfs_truncate_in_trans(struct btrfs_trans_handle
*trans
,
900 struct btrfs_root
*root
,
905 struct btrfs_path
*path
;
906 struct btrfs_key key
;
907 struct btrfs_key found_key
;
909 struct extent_buffer
*leaf
;
910 struct btrfs_file_extent_item
*fi
;
911 u64 extent_start
= 0;
912 u64 extent_num_bytes
= 0;
918 int pending_del_nr
= 0;
919 int pending_del_slot
= 0;
920 int extent_type
= -1;
921 u64 mask
= root
->sectorsize
- 1;
923 btrfs_drop_extent_cache(inode
, inode
->i_size
& (~mask
), (u64
)-1);
924 path
= btrfs_alloc_path();
928 /* FIXME, add redo link to tree so we don't leak on crash */
929 key
.objectid
= inode
->i_ino
;
930 key
.offset
= (u64
)-1;
933 btrfs_init_path(path
);
935 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
940 BUG_ON(path
->slots
[0] == 0);
946 leaf
= path
->nodes
[0];
947 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
948 found_type
= btrfs_key_type(&found_key
);
950 if (found_key
.objectid
!= inode
->i_ino
)
953 if (found_type
< min_type
)
956 item_end
= found_key
.offset
;
957 if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
958 fi
= btrfs_item_ptr(leaf
, path
->slots
[0],
959 struct btrfs_file_extent_item
);
960 extent_type
= btrfs_file_extent_type(leaf
, fi
);
961 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
963 btrfs_file_extent_num_bytes(leaf
, fi
);
964 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
965 struct btrfs_item
*item
= btrfs_item_nr(leaf
,
967 item_end
+= btrfs_file_extent_inline_len(leaf
,
972 if (found_type
== BTRFS_CSUM_ITEM_KEY
) {
973 ret
= btrfs_csum_truncate(trans
, root
, path
,
977 if (item_end
< inode
->i_size
) {
978 if (found_type
== BTRFS_DIR_ITEM_KEY
) {
979 found_type
= BTRFS_INODE_ITEM_KEY
;
980 } else if (found_type
== BTRFS_EXTENT_ITEM_KEY
) {
981 found_type
= BTRFS_CSUM_ITEM_KEY
;
982 } else if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
983 found_type
= BTRFS_XATTR_ITEM_KEY
;
984 } else if (found_type
== BTRFS_XATTR_ITEM_KEY
) {
985 found_type
= BTRFS_INODE_REF_KEY
;
986 } else if (found_type
) {
991 btrfs_set_key_type(&key
, found_type
);
994 if (found_key
.offset
>= inode
->i_size
)
1000 /* FIXME, shrink the extent if the ref count is only 1 */
1001 if (found_type
!= BTRFS_EXTENT_DATA_KEY
)
1004 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1006 extent_start
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
1008 u64 orig_num_bytes
=
1009 btrfs_file_extent_num_bytes(leaf
, fi
);
1010 extent_num_bytes
= inode
->i_size
-
1011 found_key
.offset
+ root
->sectorsize
- 1;
1012 extent_num_bytes
= extent_num_bytes
&
1013 ~((u64
)root
->sectorsize
- 1);
1014 btrfs_set_file_extent_num_bytes(leaf
, fi
,
1016 num_dec
= (orig_num_bytes
-
1018 if (extent_start
!= 0)
1019 dec_i_blocks(inode
, num_dec
);
1020 btrfs_mark_buffer_dirty(leaf
);
1023 btrfs_file_extent_disk_num_bytes(leaf
,
1025 /* FIXME blocksize != 4096 */
1026 num_dec
= btrfs_file_extent_num_bytes(leaf
, fi
);
1027 if (extent_start
!= 0) {
1029 dec_i_blocks(inode
, num_dec
);
1031 root_gen
= btrfs_header_generation(leaf
);
1032 root_owner
= btrfs_header_owner(leaf
);
1034 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1036 u32 newsize
= inode
->i_size
- found_key
.offset
;
1037 dec_i_blocks(inode
, item_end
+ 1 -
1038 found_key
.offset
- newsize
);
1040 btrfs_file_extent_calc_inline_size(newsize
);
1041 ret
= btrfs_truncate_item(trans
, root
, path
,
1045 dec_i_blocks(inode
, item_end
+ 1 -
1051 if (!pending_del_nr
) {
1052 /* no pending yet, add ourselves */
1053 pending_del_slot
= path
->slots
[0];
1055 } else if (pending_del_nr
&&
1056 path
->slots
[0] + 1 == pending_del_slot
) {
1057 /* hop on the pending chunk */
1059 pending_del_slot
= path
->slots
[0];
1061 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path
->slots
[0], pending_del_nr
, pending_del_slot
);
1067 ret
= btrfs_free_extent(trans
, root
, extent_start
,
1070 root_gen
, inode
->i_ino
,
1071 found_key
.offset
, 0);
1075 if (path
->slots
[0] == 0) {
1078 btrfs_release_path(root
, path
);
1083 if (pending_del_nr
&&
1084 path
->slots
[0] + 1 != pending_del_slot
) {
1085 struct btrfs_key debug
;
1087 btrfs_item_key_to_cpu(path
->nodes
[0], &debug
,
1089 ret
= btrfs_del_items(trans
, root
, path
,
1094 btrfs_release_path(root
, path
);
1100 if (pending_del_nr
) {
1101 ret
= btrfs_del_items(trans
, root
, path
, pending_del_slot
,
1104 btrfs_release_path(root
, path
);
1105 btrfs_free_path(path
);
1106 inode
->i_sb
->s_dirt
= 1;
1110 static int btrfs_cow_one_page(struct inode
*inode
, struct page
*page
,
1114 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1115 u64 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
1116 u64 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1119 WARN_ON(!PageLocked(page
));
1120 set_page_extent_mapped(page
);
1122 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1123 set_extent_delalloc(&BTRFS_I(inode
)->io_tree
, page_start
,
1124 page_end
, GFP_NOFS
);
1126 if (zero_start
!= PAGE_CACHE_SIZE
) {
1128 memset(kaddr
+ zero_start
, 0, PAGE_CACHE_SIZE
- zero_start
);
1129 flush_dcache_page(page
);
1132 set_page_dirty(page
);
1133 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1139 * taken from block_truncate_page, but does cow as it zeros out
1140 * any bytes left in the last page in the file.
1142 static int btrfs_truncate_page(struct address_space
*mapping
, loff_t from
)
1144 struct inode
*inode
= mapping
->host
;
1145 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1146 u32 blocksize
= root
->sectorsize
;
1147 pgoff_t index
= from
>> PAGE_CACHE_SHIFT
;
1148 unsigned offset
= from
& (PAGE_CACHE_SIZE
-1);
1153 if ((offset
& (blocksize
- 1)) == 0)
1157 page
= grab_cache_page(mapping
, index
);
1160 if (!PageUptodate(page
)) {
1161 ret
= btrfs_readpage(NULL
, page
);
1163 if (!PageUptodate(page
)) {
1168 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
1170 ret
= btrfs_cow_one_page(inode
, page
, offset
);
1173 page_cache_release(page
);
1178 static int btrfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
1180 struct inode
*inode
= dentry
->d_inode
;
1183 err
= inode_change_ok(inode
, attr
);
1187 if (S_ISREG(inode
->i_mode
) &&
1188 attr
->ia_valid
& ATTR_SIZE
&& attr
->ia_size
> inode
->i_size
) {
1189 struct btrfs_trans_handle
*trans
;
1190 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1191 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1193 u64 mask
= root
->sectorsize
- 1;
1194 u64 hole_start
= (inode
->i_size
+ mask
) & ~mask
;
1195 u64 block_end
= (attr
->ia_size
+ mask
) & ~mask
;
1199 if (attr
->ia_size
<= hole_start
)
1202 mutex_lock(&root
->fs_info
->fs_mutex
);
1203 err
= btrfs_check_free_space(root
, 1, 0);
1204 mutex_unlock(&root
->fs_info
->fs_mutex
);
1208 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
1210 lock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1211 hole_size
= block_end
- hole_start
;
1213 mutex_lock(&root
->fs_info
->fs_mutex
);
1214 trans
= btrfs_start_transaction(root
, 1);
1215 btrfs_set_trans_block_group(trans
, inode
);
1216 err
= btrfs_drop_extents(trans
, root
, inode
,
1217 hole_start
, block_end
, hole_start
,
1220 if (alloc_hint
!= EXTENT_MAP_INLINE
) {
1221 err
= btrfs_insert_file_extent(trans
, root
,
1225 btrfs_drop_extent_cache(inode
, hole_start
,
1227 btrfs_check_file(root
, inode
);
1229 btrfs_end_transaction(trans
, root
);
1230 mutex_unlock(&root
->fs_info
->fs_mutex
);
1231 unlock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1236 err
= inode_setattr(inode
, attr
);
1241 void btrfs_put_inode(struct inode
*inode
)
1245 if (!BTRFS_I(inode
)->ordered_trans
) {
1249 if (mapping_tagged(inode
->i_mapping
, PAGECACHE_TAG_DIRTY
) ||
1250 mapping_tagged(inode
->i_mapping
, PAGECACHE_TAG_WRITEBACK
))
1253 ret
= btrfs_del_ordered_inode(inode
);
1255 atomic_dec(&inode
->i_count
);
1259 void btrfs_delete_inode(struct inode
*inode
)
1261 struct btrfs_trans_handle
*trans
;
1262 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1266 truncate_inode_pages(&inode
->i_data
, 0);
1267 if (is_bad_inode(inode
)) {
1272 mutex_lock(&root
->fs_info
->fs_mutex
);
1273 trans
= btrfs_start_transaction(root
, 1);
1275 btrfs_set_trans_block_group(trans
, inode
);
1276 ret
= btrfs_truncate_in_trans(trans
, root
, inode
, 0);
1278 goto no_delete_lock
;
1280 nr
= trans
->blocks_used
;
1283 btrfs_end_transaction(trans
, root
);
1284 mutex_unlock(&root
->fs_info
->fs_mutex
);
1285 btrfs_btree_balance_dirty(root
, nr
);
1286 btrfs_throttle(root
);
1290 nr
= trans
->blocks_used
;
1291 btrfs_end_transaction(trans
, root
);
1292 mutex_unlock(&root
->fs_info
->fs_mutex
);
1293 btrfs_btree_balance_dirty(root
, nr
);
1294 btrfs_throttle(root
);
1300 * this returns the key found in the dir entry in the location pointer.
1301 * If no dir entries were found, location->objectid is 0.
1303 static int btrfs_inode_by_name(struct inode
*dir
, struct dentry
*dentry
,
1304 struct btrfs_key
*location
)
1306 const char *name
= dentry
->d_name
.name
;
1307 int namelen
= dentry
->d_name
.len
;
1308 struct btrfs_dir_item
*di
;
1309 struct btrfs_path
*path
;
1310 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1313 if (namelen
== 1 && strcmp(name
, ".") == 0) {
1314 location
->objectid
= dir
->i_ino
;
1315 location
->type
= BTRFS_INODE_ITEM_KEY
;
1316 location
->offset
= 0;
1319 path
= btrfs_alloc_path();
1322 if (namelen
== 2 && strcmp(name
, "..") == 0) {
1323 struct btrfs_key key
;
1324 struct extent_buffer
*leaf
;
1328 key
.objectid
= dir
->i_ino
;
1329 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1331 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1335 leaf
= path
->nodes
[0];
1336 slot
= path
->slots
[0];
1337 nritems
= btrfs_header_nritems(leaf
);
1338 if (slot
>= nritems
)
1341 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
1342 if (key
.objectid
!= dir
->i_ino
||
1343 key
.type
!= BTRFS_INODE_REF_KEY
) {
1346 location
->objectid
= key
.offset
;
1347 location
->type
= BTRFS_INODE_ITEM_KEY
;
1348 location
->offset
= 0;
1352 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir
->i_ino
, name
,
1356 if (!di
|| IS_ERR(di
)) {
1359 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, location
);
1361 btrfs_free_path(path
);
1364 location
->objectid
= 0;
1369 * when we hit a tree root in a directory, the btrfs part of the inode
1370 * needs to be changed to reflect the root directory of the tree root. This
1371 * is kind of like crossing a mount point.
1373 static int fixup_tree_root_location(struct btrfs_root
*root
,
1374 struct btrfs_key
*location
,
1375 struct btrfs_root
**sub_root
,
1376 struct dentry
*dentry
)
1378 struct btrfs_path
*path
;
1379 struct btrfs_root_item
*ri
;
1381 if (btrfs_key_type(location
) != BTRFS_ROOT_ITEM_KEY
)
1383 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1386 path
= btrfs_alloc_path();
1388 mutex_lock(&root
->fs_info
->fs_mutex
);
1390 *sub_root
= btrfs_read_fs_root(root
->fs_info
, location
,
1391 dentry
->d_name
.name
,
1392 dentry
->d_name
.len
);
1393 if (IS_ERR(*sub_root
))
1394 return PTR_ERR(*sub_root
);
1396 ri
= &(*sub_root
)->root_item
;
1397 location
->objectid
= btrfs_root_dirid(ri
);
1398 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1399 location
->offset
= 0;
1401 btrfs_free_path(path
);
1402 mutex_unlock(&root
->fs_info
->fs_mutex
);
1406 static int btrfs_init_locked_inode(struct inode
*inode
, void *p
)
1408 struct btrfs_iget_args
*args
= p
;
1409 inode
->i_ino
= args
->ino
;
1410 BTRFS_I(inode
)->root
= args
->root
;
1411 BTRFS_I(inode
)->delalloc_bytes
= 0;
1412 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1413 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1414 inode
->i_mapping
, GFP_NOFS
);
1415 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1416 inode
->i_mapping
, GFP_NOFS
);
1420 static int btrfs_find_actor(struct inode
*inode
, void *opaque
)
1422 struct btrfs_iget_args
*args
= opaque
;
1423 return (args
->ino
== inode
->i_ino
&&
1424 args
->root
== BTRFS_I(inode
)->root
);
1427 struct inode
*btrfs_ilookup(struct super_block
*s
, u64 objectid
,
1430 struct btrfs_iget_args args
;
1431 args
.ino
= objectid
;
1432 args
.root
= btrfs_lookup_fs_root(btrfs_sb(s
)->fs_info
, root_objectid
);
1437 return ilookup5(s
, objectid
, btrfs_find_actor
, (void *)&args
);
1440 struct inode
*btrfs_iget_locked(struct super_block
*s
, u64 objectid
,
1441 struct btrfs_root
*root
)
1443 struct inode
*inode
;
1444 struct btrfs_iget_args args
;
1445 args
.ino
= objectid
;
1448 inode
= iget5_locked(s
, objectid
, btrfs_find_actor
,
1449 btrfs_init_locked_inode
,
1454 static struct dentry
*btrfs_lookup(struct inode
*dir
, struct dentry
*dentry
,
1455 struct nameidata
*nd
)
1457 struct inode
* inode
;
1458 struct btrfs_inode
*bi
= BTRFS_I(dir
);
1459 struct btrfs_root
*root
= bi
->root
;
1460 struct btrfs_root
*sub_root
= root
;
1461 struct btrfs_key location
;
1464 if (dentry
->d_name
.len
> BTRFS_NAME_LEN
)
1465 return ERR_PTR(-ENAMETOOLONG
);
1467 mutex_lock(&root
->fs_info
->fs_mutex
);
1468 ret
= btrfs_inode_by_name(dir
, dentry
, &location
);
1469 mutex_unlock(&root
->fs_info
->fs_mutex
);
1472 return ERR_PTR(ret
);
1475 if (location
.objectid
) {
1476 ret
= fixup_tree_root_location(root
, &location
, &sub_root
,
1479 return ERR_PTR(ret
);
1481 return ERR_PTR(-ENOENT
);
1482 inode
= btrfs_iget_locked(dir
->i_sb
, location
.objectid
,
1485 return ERR_PTR(-EACCES
);
1486 if (inode
->i_state
& I_NEW
) {
1487 /* the inode and parent dir are two different roots */
1488 if (sub_root
!= root
) {
1490 sub_root
->inode
= inode
;
1492 BTRFS_I(inode
)->root
= sub_root
;
1493 memcpy(&BTRFS_I(inode
)->location
, &location
,
1495 btrfs_read_locked_inode(inode
);
1496 unlock_new_inode(inode
);
1499 return d_splice_alias(inode
, dentry
);
1502 static unsigned char btrfs_filetype_table
[] = {
1503 DT_UNKNOWN
, DT_REG
, DT_DIR
, DT_CHR
, DT_BLK
, DT_FIFO
, DT_SOCK
, DT_LNK
1506 static int btrfs_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
1508 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1509 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1510 struct btrfs_item
*item
;
1511 struct btrfs_dir_item
*di
;
1512 struct btrfs_key key
;
1513 struct btrfs_key found_key
;
1514 struct btrfs_path
*path
;
1517 struct extent_buffer
*leaf
;
1520 unsigned char d_type
;
1525 int key_type
= BTRFS_DIR_INDEX_KEY
;
1530 /* FIXME, use a real flag for deciding about the key type */
1531 if (root
->fs_info
->tree_root
== root
)
1532 key_type
= BTRFS_DIR_ITEM_KEY
;
1534 /* special case for "." */
1535 if (filp
->f_pos
== 0) {
1536 over
= filldir(dirent
, ".", 1,
1544 mutex_lock(&root
->fs_info
->fs_mutex
);
1545 key
.objectid
= inode
->i_ino
;
1546 path
= btrfs_alloc_path();
1549 /* special case for .., just use the back ref */
1550 if (filp
->f_pos
== 1) {
1551 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1553 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1555 leaf
= path
->nodes
[0];
1556 slot
= path
->slots
[0];
1557 nritems
= btrfs_header_nritems(leaf
);
1558 if (slot
>= nritems
) {
1559 btrfs_release_path(root
, path
);
1560 goto read_dir_items
;
1562 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1563 btrfs_release_path(root
, path
);
1564 if (found_key
.objectid
!= key
.objectid
||
1565 found_key
.type
!= BTRFS_INODE_REF_KEY
)
1566 goto read_dir_items
;
1567 over
= filldir(dirent
, "..", 2,
1568 2, found_key
.offset
, DT_DIR
);
1575 btrfs_set_key_type(&key
, key_type
);
1576 key
.offset
= filp
->f_pos
;
1578 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1583 leaf
= path
->nodes
[0];
1584 nritems
= btrfs_header_nritems(leaf
);
1585 slot
= path
->slots
[0];
1586 if (advance
|| slot
>= nritems
) {
1587 if (slot
>= nritems
-1) {
1588 ret
= btrfs_next_leaf(root
, path
);
1591 leaf
= path
->nodes
[0];
1592 nritems
= btrfs_header_nritems(leaf
);
1593 slot
= path
->slots
[0];
1600 item
= btrfs_item_nr(leaf
, slot
);
1601 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1603 if (found_key
.objectid
!= key
.objectid
)
1605 if (btrfs_key_type(&found_key
) != key_type
)
1607 if (found_key
.offset
< filp
->f_pos
)
1610 filp
->f_pos
= found_key
.offset
;
1612 di
= btrfs_item_ptr(leaf
, slot
, struct btrfs_dir_item
);
1614 di_total
= btrfs_item_size(leaf
, item
);
1615 while(di_cur
< di_total
) {
1616 struct btrfs_key location
;
1618 name_len
= btrfs_dir_name_len(leaf
, di
);
1619 if (name_len
< 32) {
1620 name_ptr
= tmp_name
;
1622 name_ptr
= kmalloc(name_len
, GFP_NOFS
);
1625 read_extent_buffer(leaf
, name_ptr
,
1626 (unsigned long)(di
+ 1), name_len
);
1628 d_type
= btrfs_filetype_table
[btrfs_dir_type(leaf
, di
)];
1629 btrfs_dir_item_key_to_cpu(leaf
, di
, &location
);
1630 over
= filldir(dirent
, name_ptr
, name_len
,
1635 if (name_ptr
!= tmp_name
)
1640 di_len
= btrfs_dir_name_len(leaf
, di
) +
1641 btrfs_dir_data_len(leaf
, di
) +sizeof(*di
);
1643 di
= (struct btrfs_dir_item
*)((char *)di
+ di_len
);
1646 if (key_type
== BTRFS_DIR_INDEX_KEY
)
1647 filp
->f_pos
= INT_LIMIT(typeof(filp
->f_pos
));
1653 btrfs_release_path(root
, path
);
1654 btrfs_free_path(path
);
1655 mutex_unlock(&root
->fs_info
->fs_mutex
);
1659 int btrfs_write_inode(struct inode
*inode
, int wait
)
1661 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1662 struct btrfs_trans_handle
*trans
;
1666 mutex_lock(&root
->fs_info
->fs_mutex
);
1667 trans
= btrfs_start_transaction(root
, 1);
1668 btrfs_set_trans_block_group(trans
, inode
);
1669 ret
= btrfs_commit_transaction(trans
, root
);
1670 mutex_unlock(&root
->fs_info
->fs_mutex
);
1676 * This is somewhat expensive, updating the tree every time the
1677 * inode changes. But, it is most likely to find the inode in cache.
1678 * FIXME, needs more benchmarking...there are no reasons other than performance
1679 * to keep or drop this code.
1681 void btrfs_dirty_inode(struct inode
*inode
)
1683 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1684 struct btrfs_trans_handle
*trans
;
1686 mutex_lock(&root
->fs_info
->fs_mutex
);
1687 trans
= btrfs_start_transaction(root
, 1);
1688 btrfs_set_trans_block_group(trans
, inode
);
1689 btrfs_update_inode(trans
, root
, inode
);
1690 btrfs_end_transaction(trans
, root
);
1691 mutex_unlock(&root
->fs_info
->fs_mutex
);
1694 static struct inode
*btrfs_new_inode(struct btrfs_trans_handle
*trans
,
1695 struct btrfs_root
*root
,
1696 const char *name
, int name_len
,
1699 struct btrfs_block_group_cache
*group
,
1702 struct inode
*inode
;
1703 struct btrfs_inode_item
*inode_item
;
1704 struct btrfs_block_group_cache
*new_inode_group
;
1705 struct btrfs_key
*location
;
1706 struct btrfs_path
*path
;
1707 struct btrfs_inode_ref
*ref
;
1708 struct btrfs_key key
[2];
1714 path
= btrfs_alloc_path();
1717 inode
= new_inode(root
->fs_info
->sb
);
1719 return ERR_PTR(-ENOMEM
);
1721 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1722 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1723 inode
->i_mapping
, GFP_NOFS
);
1724 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1725 inode
->i_mapping
, GFP_NOFS
);
1726 BTRFS_I(inode
)->delalloc_bytes
= 0;
1727 BTRFS_I(inode
)->root
= root
;
1733 new_inode_group
= btrfs_find_block_group(root
, group
, 0,
1734 BTRFS_BLOCK_GROUP_METADATA
, owner
);
1735 if (!new_inode_group
) {
1736 printk("find_block group failed\n");
1737 new_inode_group
= group
;
1739 BTRFS_I(inode
)->block_group
= new_inode_group
;
1740 BTRFS_I(inode
)->flags
= 0;
1742 key
[0].objectid
= objectid
;
1743 btrfs_set_key_type(&key
[0], BTRFS_INODE_ITEM_KEY
);
1746 key
[1].objectid
= objectid
;
1747 btrfs_set_key_type(&key
[1], BTRFS_INODE_REF_KEY
);
1748 key
[1].offset
= ref_objectid
;
1750 sizes
[0] = sizeof(struct btrfs_inode_item
);
1751 sizes
[1] = name_len
+ sizeof(*ref
);
1753 ret
= btrfs_insert_empty_items(trans
, root
, path
, key
, sizes
, 2);
1757 if (objectid
> root
->highest_inode
)
1758 root
->highest_inode
= objectid
;
1760 inode
->i_uid
= current
->fsuid
;
1761 inode
->i_gid
= current
->fsgid
;
1762 inode
->i_mode
= mode
;
1763 inode
->i_ino
= objectid
;
1764 inode
->i_blocks
= 0;
1765 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1766 inode_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
1767 struct btrfs_inode_item
);
1768 fill_inode_item(path
->nodes
[0], inode_item
, inode
);
1770 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0] + 1,
1771 struct btrfs_inode_ref
);
1772 btrfs_set_inode_ref_name_len(path
->nodes
[0], ref
, name_len
);
1773 ptr
= (unsigned long)(ref
+ 1);
1774 write_extent_buffer(path
->nodes
[0], name
, ptr
, name_len
);
1776 btrfs_mark_buffer_dirty(path
->nodes
[0]);
1777 btrfs_free_path(path
);
1779 location
= &BTRFS_I(inode
)->location
;
1780 location
->objectid
= objectid
;
1781 location
->offset
= 0;
1782 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1784 insert_inode_hash(inode
);
1787 btrfs_free_path(path
);
1788 return ERR_PTR(ret
);
1791 static inline u8
btrfs_inode_type(struct inode
*inode
)
1793 return btrfs_type_by_mode
[(inode
->i_mode
& S_IFMT
) >> S_SHIFT
];
1796 static int btrfs_add_link(struct btrfs_trans_handle
*trans
,
1797 struct dentry
*dentry
, struct inode
*inode
,
1801 struct btrfs_key key
;
1802 struct btrfs_root
*root
= BTRFS_I(dentry
->d_parent
->d_inode
)->root
;
1803 struct inode
*parent_inode
;
1805 key
.objectid
= inode
->i_ino
;
1806 btrfs_set_key_type(&key
, BTRFS_INODE_ITEM_KEY
);
1809 ret
= btrfs_insert_dir_item(trans
, root
,
1810 dentry
->d_name
.name
, dentry
->d_name
.len
,
1811 dentry
->d_parent
->d_inode
->i_ino
,
1812 &key
, btrfs_inode_type(inode
));
1815 ret
= btrfs_insert_inode_ref(trans
, root
,
1816 dentry
->d_name
.name
,
1819 dentry
->d_parent
->d_inode
->i_ino
);
1821 parent_inode
= dentry
->d_parent
->d_inode
;
1822 parent_inode
->i_size
+= dentry
->d_name
.len
* 2;
1823 parent_inode
->i_mtime
= parent_inode
->i_ctime
= CURRENT_TIME
;
1824 ret
= btrfs_update_inode(trans
, root
,
1825 dentry
->d_parent
->d_inode
);
1830 static int btrfs_add_nondir(struct btrfs_trans_handle
*trans
,
1831 struct dentry
*dentry
, struct inode
*inode
,
1834 int err
= btrfs_add_link(trans
, dentry
, inode
, backref
);
1836 d_instantiate(dentry
, inode
);
1844 static int btrfs_mknod(struct inode
*dir
, struct dentry
*dentry
,
1845 int mode
, dev_t rdev
)
1847 struct btrfs_trans_handle
*trans
;
1848 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1849 struct inode
*inode
= NULL
;
1853 unsigned long nr
= 0;
1855 if (!new_valid_dev(rdev
))
1858 mutex_lock(&root
->fs_info
->fs_mutex
);
1859 err
= btrfs_check_free_space(root
, 1, 0);
1863 trans
= btrfs_start_transaction(root
, 1);
1864 btrfs_set_trans_block_group(trans
, dir
);
1866 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
1872 inode
= btrfs_new_inode(trans
, root
, dentry
->d_name
.name
,
1874 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
1875 BTRFS_I(dir
)->block_group
, mode
);
1876 err
= PTR_ERR(inode
);
1880 btrfs_set_trans_block_group(trans
, inode
);
1881 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
1885 inode
->i_op
= &btrfs_special_inode_operations
;
1886 init_special_inode(inode
, inode
->i_mode
, rdev
);
1887 btrfs_update_inode(trans
, root
, inode
);
1889 dir
->i_sb
->s_dirt
= 1;
1890 btrfs_update_inode_block_group(trans
, inode
);
1891 btrfs_update_inode_block_group(trans
, dir
);
1893 nr
= trans
->blocks_used
;
1894 btrfs_end_transaction(trans
, root
);
1896 mutex_unlock(&root
->fs_info
->fs_mutex
);
1899 inode_dec_link_count(inode
);
1902 btrfs_btree_balance_dirty(root
, nr
);
1903 btrfs_throttle(root
);
1907 static int btrfs_create(struct inode
*dir
, struct dentry
*dentry
,
1908 int mode
, struct nameidata
*nd
)
1910 struct btrfs_trans_handle
*trans
;
1911 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1912 struct inode
*inode
= NULL
;
1915 unsigned long nr
= 0;
1918 mutex_lock(&root
->fs_info
->fs_mutex
);
1919 err
= btrfs_check_free_space(root
, 1, 0);
1922 trans
= btrfs_start_transaction(root
, 1);
1923 btrfs_set_trans_block_group(trans
, dir
);
1925 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
1931 inode
= btrfs_new_inode(trans
, root
, dentry
->d_name
.name
,
1933 dentry
->d_parent
->d_inode
->i_ino
,
1934 objectid
, BTRFS_I(dir
)->block_group
, mode
);
1935 err
= PTR_ERR(inode
);
1939 btrfs_set_trans_block_group(trans
, inode
);
1940 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
1944 inode
->i_mapping
->a_ops
= &btrfs_aops
;
1945 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1946 inode
->i_fop
= &btrfs_file_operations
;
1947 inode
->i_op
= &btrfs_file_inode_operations
;
1948 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1949 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1950 inode
->i_mapping
, GFP_NOFS
);
1951 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1952 inode
->i_mapping
, GFP_NOFS
);
1953 BTRFS_I(inode
)->delalloc_bytes
= 0;
1954 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
1956 dir
->i_sb
->s_dirt
= 1;
1957 btrfs_update_inode_block_group(trans
, inode
);
1958 btrfs_update_inode_block_group(trans
, dir
);
1960 nr
= trans
->blocks_used
;
1961 btrfs_end_transaction(trans
, root
);
1963 mutex_unlock(&root
->fs_info
->fs_mutex
);
1966 inode_dec_link_count(inode
);
1969 btrfs_btree_balance_dirty(root
, nr
);
1970 btrfs_throttle(root
);
1974 static int btrfs_link(struct dentry
*old_dentry
, struct inode
*dir
,
1975 struct dentry
*dentry
)
1977 struct btrfs_trans_handle
*trans
;
1978 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1979 struct inode
*inode
= old_dentry
->d_inode
;
1980 unsigned long nr
= 0;
1984 if (inode
->i_nlink
== 0)
1987 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1992 mutex_lock(&root
->fs_info
->fs_mutex
);
1993 err
= btrfs_check_free_space(root
, 1, 0);
1996 trans
= btrfs_start_transaction(root
, 1);
1998 btrfs_set_trans_block_group(trans
, dir
);
1999 atomic_inc(&inode
->i_count
);
2000 err
= btrfs_add_nondir(trans
, dentry
, inode
, 1);
2005 dir
->i_sb
->s_dirt
= 1;
2006 btrfs_update_inode_block_group(trans
, dir
);
2007 err
= btrfs_update_inode(trans
, root
, inode
);
2012 nr
= trans
->blocks_used
;
2013 btrfs_end_transaction(trans
, root
);
2015 mutex_unlock(&root
->fs_info
->fs_mutex
);
2018 inode_dec_link_count(inode
);
2021 btrfs_btree_balance_dirty(root
, nr
);
2022 btrfs_throttle(root
);
2026 static int btrfs_mkdir(struct inode
*dir
, struct dentry
*dentry
, int mode
)
2028 struct inode
*inode
;
2029 struct btrfs_trans_handle
*trans
;
2030 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2032 int drop_on_err
= 0;
2034 unsigned long nr
= 1;
2036 mutex_lock(&root
->fs_info
->fs_mutex
);
2037 err
= btrfs_check_free_space(root
, 1, 0);
2041 trans
= btrfs_start_transaction(root
, 1);
2042 btrfs_set_trans_block_group(trans
, dir
);
2044 if (IS_ERR(trans
)) {
2045 err
= PTR_ERR(trans
);
2049 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2055 inode
= btrfs_new_inode(trans
, root
, dentry
->d_name
.name
,
2057 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2058 BTRFS_I(dir
)->block_group
, S_IFDIR
| mode
);
2059 if (IS_ERR(inode
)) {
2060 err
= PTR_ERR(inode
);
2065 inode
->i_op
= &btrfs_dir_inode_operations
;
2066 inode
->i_fop
= &btrfs_dir_file_operations
;
2067 btrfs_set_trans_block_group(trans
, inode
);
2070 err
= btrfs_update_inode(trans
, root
, inode
);
2074 err
= btrfs_add_link(trans
, dentry
, inode
, 0);
2078 d_instantiate(dentry
, inode
);
2080 dir
->i_sb
->s_dirt
= 1;
2081 btrfs_update_inode_block_group(trans
, inode
);
2082 btrfs_update_inode_block_group(trans
, dir
);
2085 nr
= trans
->blocks_used
;
2086 btrfs_end_transaction(trans
, root
);
2089 mutex_unlock(&root
->fs_info
->fs_mutex
);
2092 btrfs_btree_balance_dirty(root
, nr
);
2093 btrfs_throttle(root
);
2097 static int merge_extent_mapping(struct extent_map_tree
*em_tree
,
2098 struct extent_map
*existing
,
2099 struct extent_map
*em
)
2104 int real_blocks
= existing
->block_start
< EXTENT_MAP_LAST_BYTE
;
2106 if (real_blocks
&& em
->block_start
>= EXTENT_MAP_LAST_BYTE
)
2109 if (!real_blocks
&& em
->block_start
!= existing
->block_start
)
2112 new_end
= max(existing
->start
+ existing
->len
, em
->start
+ em
->len
);
2114 if (existing
->start
>= em
->start
) {
2115 if (em
->start
+ em
->len
< existing
->start
)
2118 start_diff
= existing
->start
- em
->start
;
2119 if (real_blocks
&& em
->block_start
+ start_diff
!=
2120 existing
->block_start
)
2123 em
->len
= new_end
- em
->start
;
2125 remove_extent_mapping(em_tree
, existing
);
2126 /* free for the tree */
2127 free_extent_map(existing
);
2128 ret
= add_extent_mapping(em_tree
, em
);
2130 } else if (em
->start
> existing
->start
) {
2132 if (existing
->start
+ existing
->len
< em
->start
)
2135 start_diff
= em
->start
- existing
->start
;
2136 if (real_blocks
&& existing
->block_start
+ start_diff
!=
2140 remove_extent_mapping(em_tree
, existing
);
2141 em
->block_start
= existing
->block_start
;
2142 em
->start
= existing
->start
;
2143 em
->len
= new_end
- existing
->start
;
2144 free_extent_map(existing
);
2146 ret
= add_extent_mapping(em_tree
, em
);
2153 printk("invalid extent map merge [%Lu %Lu %Lu] [%Lu %Lu %Lu]\n",
2154 existing
->start
, existing
->len
, existing
->block_start
,
2155 em
->start
, em
->len
, em
->block_start
);
2159 struct extent_map
*btrfs_get_extent(struct inode
*inode
, struct page
*page
,
2160 size_t pg_offset
, u64 start
, u64 len
,
2166 u64 extent_start
= 0;
2168 u64 objectid
= inode
->i_ino
;
2170 struct btrfs_path
*path
;
2171 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2172 struct btrfs_file_extent_item
*item
;
2173 struct extent_buffer
*leaf
;
2174 struct btrfs_key found_key
;
2175 struct extent_map
*em
= NULL
;
2176 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
2177 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2178 struct btrfs_trans_handle
*trans
= NULL
;
2180 path
= btrfs_alloc_path();
2182 mutex_lock(&root
->fs_info
->fs_mutex
);
2185 spin_lock(&em_tree
->lock
);
2186 em
= lookup_extent_mapping(em_tree
, start
, len
);
2187 spin_unlock(&em_tree
->lock
);
2190 if (em
->start
> start
) {
2191 printk("get_extent lookup [%Lu %Lu] em [%Lu %Lu]\n",
2192 start
, len
, em
->start
, em
->len
);
2195 if (em
->block_start
== EXTENT_MAP_INLINE
&& page
)
2196 free_extent_map(em
);
2200 em
= alloc_extent_map(GFP_NOFS
);
2206 em
->start
= EXTENT_MAP_HOLE
;
2208 em
->bdev
= inode
->i_sb
->s_bdev
;
2209 ret
= btrfs_lookup_file_extent(trans
, root
, path
,
2210 objectid
, start
, trans
!= NULL
);
2217 if (path
->slots
[0] == 0)
2222 leaf
= path
->nodes
[0];
2223 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
2224 struct btrfs_file_extent_item
);
2225 /* are we inside the extent that was found? */
2226 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2227 found_type
= btrfs_key_type(&found_key
);
2228 if (found_key
.objectid
!= objectid
||
2229 found_type
!= BTRFS_EXTENT_DATA_KEY
) {
2233 found_type
= btrfs_file_extent_type(leaf
, item
);
2234 extent_start
= found_key
.offset
;
2235 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
2236 extent_end
= extent_start
+
2237 btrfs_file_extent_num_bytes(leaf
, item
);
2239 if (start
< extent_start
|| start
>= extent_end
) {
2241 if (start
< extent_start
) {
2242 if (start
+ len
<= extent_start
)
2244 em
->len
= extent_end
- extent_start
;
2250 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
2252 em
->start
= extent_start
;
2253 em
->len
= extent_end
- extent_start
;
2254 em
->block_start
= EXTENT_MAP_HOLE
;
2257 bytenr
+= btrfs_file_extent_offset(leaf
, item
);
2258 em
->block_start
= bytenr
;
2259 em
->start
= extent_start
;
2260 em
->len
= extent_end
- extent_start
;
2262 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
2267 size_t extent_offset
;
2270 size
= btrfs_file_extent_inline_len(leaf
, btrfs_item_nr(leaf
,
2272 extent_end
= (extent_start
+ size
+ root
->sectorsize
- 1) &
2273 ~((u64
)root
->sectorsize
- 1);
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 em
->block_start
= EXTENT_MAP_INLINE
;
2288 em
->start
= extent_start
;
2293 page_start
= page_offset(page
) + pg_offset
;
2294 extent_offset
= page_start
- extent_start
;
2295 copy_size
= min_t(u64
, PAGE_CACHE_SIZE
- pg_offset
,
2296 size
- extent_offset
);
2297 em
->start
= extent_start
+ extent_offset
;
2298 em
->len
= (copy_size
+ root
->sectorsize
- 1) &
2299 ~((u64
)root
->sectorsize
- 1);
2301 ptr
= btrfs_file_extent_inline_start(item
) + extent_offset
;
2302 if (create
== 0 && !PageUptodate(page
)) {
2303 read_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2305 flush_dcache_page(page
);
2306 } else if (create
&& PageUptodate(page
)) {
2309 free_extent_map(em
);
2311 btrfs_release_path(root
, path
);
2312 trans
= btrfs_start_transaction(root
, 1);
2315 write_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2317 btrfs_mark_buffer_dirty(leaf
);
2320 set_extent_uptodate(io_tree
, em
->start
,
2321 extent_map_end(em
) - 1, GFP_NOFS
);
2324 printk("unkknown found_type %d\n", found_type
);
2331 em
->block_start
= EXTENT_MAP_HOLE
;
2333 btrfs_release_path(root
, path
);
2334 if (em
->start
> start
|| extent_map_end(em
) <= start
) {
2335 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em
->start
, em
->len
, start
, len
);
2341 spin_lock(&em_tree
->lock
);
2342 ret
= add_extent_mapping(em_tree
, em
);
2344 /* it is possible that someone inserted the extent into the tree
2345 * while we had the lock dropped. It is also possible that
2346 * an overlapping map exists in the tree
2348 if (ret
== -EEXIST
) {
2349 struct extent_map
*existing
;
2350 existing
= lookup_extent_mapping(em_tree
, start
, len
);
2352 existing
= lookup_extent_mapping(em_tree
, em
->start
,
2355 err
= merge_extent_mapping(em_tree
, existing
,
2357 free_extent_map(existing
);
2359 free_extent_map(em
);
2364 printk("failing to insert %Lu %Lu\n",
2366 free_extent_map(em
);
2370 free_extent_map(em
);
2374 spin_unlock(&em_tree
->lock
);
2376 btrfs_free_path(path
);
2378 ret
= btrfs_end_transaction(trans
, root
);
2382 mutex_unlock(&root
->fs_info
->fs_mutex
);
2384 free_extent_map(em
);
2386 return ERR_PTR(err
);
2391 static int btrfs_get_block(struct inode
*inode
, sector_t iblock
,
2392 struct buffer_head
*bh_result
, int create
)
2394 struct extent_map
*em
;
2395 u64 start
= (u64
)iblock
<< inode
->i_blkbits
;
2396 struct btrfs_multi_bio
*multi
= NULL
;
2397 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2403 em
= btrfs_get_extent(inode
, NULL
, 0, start
, bh_result
->b_size
, 0);
2405 if (!em
|| IS_ERR(em
))
2408 if (em
->start
> start
|| em
->start
+ em
->len
<= start
)
2411 if (em
->block_start
== EXTENT_MAP_INLINE
) {
2416 if (em
->block_start
== EXTENT_MAP_HOLE
||
2417 em
->block_start
== EXTENT_MAP_DELALLOC
) {
2421 len
= em
->start
+ em
->len
- start
;
2422 len
= min_t(u64
, len
, INT_LIMIT(typeof(bh_result
->b_size
)));
2424 logical
= start
- em
->start
;
2425 logical
= em
->block_start
+ logical
;
2428 ret
= btrfs_map_block(&root
->fs_info
->mapping_tree
, READ
,
2429 logical
, &map_length
, &multi
, 0);
2431 bh_result
->b_blocknr
= multi
->stripes
[0].physical
>> inode
->i_blkbits
;
2432 bh_result
->b_size
= min(map_length
, len
);
2433 bh_result
->b_bdev
= multi
->stripes
[0].dev
->bdev
;
2434 set_buffer_mapped(bh_result
);
2437 free_extent_map(em
);
2441 static ssize_t
btrfs_direct_IO(int rw
, struct kiocb
*iocb
,
2442 const struct iovec
*iov
, loff_t offset
,
2443 unsigned long nr_segs
)
2445 struct file
*file
= iocb
->ki_filp
;
2446 struct inode
*inode
= file
->f_mapping
->host
;
2451 return blockdev_direct_IO(rw
, iocb
, inode
, inode
->i_sb
->s_bdev
, iov
,
2452 offset
, nr_segs
, btrfs_get_block
, NULL
);
2455 static sector_t
btrfs_bmap(struct address_space
*mapping
, sector_t iblock
)
2457 return extent_bmap(mapping
, iblock
, btrfs_get_extent
);
2460 int btrfs_readpage(struct file
*file
, struct page
*page
)
2462 struct extent_io_tree
*tree
;
2463 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2464 return extent_read_full_page(tree
, page
, btrfs_get_extent
);
2467 static int btrfs_writepage(struct page
*page
, struct writeback_control
*wbc
)
2469 struct extent_io_tree
*tree
;
2472 if (current
->flags
& PF_MEMALLOC
) {
2473 redirty_page_for_writepage(wbc
, page
);
2477 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2478 return extent_write_full_page(tree
, page
, btrfs_get_extent
, wbc
);
2481 static int btrfs_writepages(struct address_space
*mapping
,
2482 struct writeback_control
*wbc
)
2484 struct extent_io_tree
*tree
;
2485 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
2486 return extent_writepages(tree
, mapping
, btrfs_get_extent
, wbc
);
2490 btrfs_readpages(struct file
*file
, struct address_space
*mapping
,
2491 struct list_head
*pages
, unsigned nr_pages
)
2493 struct extent_io_tree
*tree
;
2494 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
2495 return extent_readpages(tree
, mapping
, pages
, nr_pages
,
2499 static int btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
2501 struct extent_io_tree
*tree
;
2502 struct extent_map_tree
*map
;
2505 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2506 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
2507 ret
= try_release_extent_mapping(map
, tree
, page
, gfp_flags
);
2509 invalidate_extent_lru(tree
, page_offset(page
), PAGE_CACHE_SIZE
);
2510 ClearPagePrivate(page
);
2511 set_page_private(page
, 0);
2512 page_cache_release(page
);
2517 static void btrfs_invalidatepage(struct page
*page
, unsigned long offset
)
2519 struct extent_io_tree
*tree
;
2521 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2522 extent_invalidatepage(tree
, page
, offset
);
2523 btrfs_releasepage(page
, GFP_NOFS
);
2524 if (PagePrivate(page
)) {
2525 invalidate_extent_lru(tree
, page_offset(page
), PAGE_CACHE_SIZE
);
2526 ClearPagePrivate(page
);
2527 set_page_private(page
, 0);
2528 page_cache_release(page
);
2533 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
2534 * called from a page fault handler when a page is first dirtied. Hence we must
2535 * be careful to check for EOF conditions here. We set the page up correctly
2536 * for a written page which means we get ENOSPC checking when writing into
2537 * holes and correct delalloc and unwritten extent mapping on filesystems that
2538 * support these features.
2540 * We are not allowed to take the i_mutex here so we have to play games to
2541 * protect against truncate races as the page could now be beyond EOF. Because
2542 * vmtruncate() writes the inode size before removing pages, once we have the
2543 * page lock we can determine safely if the page is beyond EOF. If it is not
2544 * beyond EOF, then the page is guaranteed safe against truncation until we
2547 int btrfs_page_mkwrite(struct vm_area_struct
*vma
, struct page
*page
)
2549 struct inode
*inode
= fdentry(vma
->vm_file
)->d_inode
;
2550 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2556 mutex_lock(&root
->fs_info
->fs_mutex
);
2557 ret
= btrfs_check_free_space(root
, PAGE_CACHE_SIZE
, 0);
2558 mutex_unlock(&root
->fs_info
->fs_mutex
);
2565 wait_on_page_writeback(page
);
2566 size
= i_size_read(inode
);
2567 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
2569 if ((page
->mapping
!= inode
->i_mapping
) ||
2570 (page_start
> size
)) {
2571 /* page got truncated out from underneath us */
2575 /* page is wholly or partially inside EOF */
2576 if (page_start
+ PAGE_CACHE_SIZE
> size
)
2577 end
= size
& ~PAGE_CACHE_MASK
;
2579 end
= PAGE_CACHE_SIZE
;
2581 ret
= btrfs_cow_one_page(inode
, page
, end
);
2589 static void btrfs_truncate(struct inode
*inode
)
2591 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2593 struct btrfs_trans_handle
*trans
;
2596 if (!S_ISREG(inode
->i_mode
))
2598 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
2601 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
2603 mutex_lock(&root
->fs_info
->fs_mutex
);
2604 trans
= btrfs_start_transaction(root
, 1);
2605 btrfs_set_trans_block_group(trans
, inode
);
2607 /* FIXME, add redo link to tree so we don't leak on crash */
2608 ret
= btrfs_truncate_in_trans(trans
, root
, inode
,
2609 BTRFS_EXTENT_DATA_KEY
);
2610 btrfs_update_inode(trans
, root
, inode
);
2611 nr
= trans
->blocks_used
;
2613 ret
= btrfs_end_transaction(trans
, root
);
2615 mutex_unlock(&root
->fs_info
->fs_mutex
);
2616 btrfs_btree_balance_dirty(root
, nr
);
2617 btrfs_throttle(root
);
2620 static int noinline
create_subvol(struct btrfs_root
*root
, char *name
,
2623 struct btrfs_trans_handle
*trans
;
2624 struct btrfs_key key
;
2625 struct btrfs_root_item root_item
;
2626 struct btrfs_inode_item
*inode_item
;
2627 struct extent_buffer
*leaf
;
2628 struct btrfs_root
*new_root
= root
;
2629 struct inode
*inode
;
2634 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
2635 unsigned long nr
= 1;
2637 mutex_lock(&root
->fs_info
->fs_mutex
);
2638 ret
= btrfs_check_free_space(root
, 1, 0);
2642 trans
= btrfs_start_transaction(root
, 1);
2645 ret
= btrfs_find_free_objectid(trans
, root
->fs_info
->tree_root
,
2650 leaf
= __btrfs_alloc_free_block(trans
, root
, root
->leafsize
,
2651 objectid
, trans
->transid
, 0, 0,
2654 return PTR_ERR(leaf
);
2656 btrfs_set_header_nritems(leaf
, 0);
2657 btrfs_set_header_level(leaf
, 0);
2658 btrfs_set_header_bytenr(leaf
, leaf
->start
);
2659 btrfs_set_header_generation(leaf
, trans
->transid
);
2660 btrfs_set_header_owner(leaf
, objectid
);
2662 write_extent_buffer(leaf
, root
->fs_info
->fsid
,
2663 (unsigned long)btrfs_header_fsid(leaf
),
2665 btrfs_mark_buffer_dirty(leaf
);
2667 inode_item
= &root_item
.inode
;
2668 memset(inode_item
, 0, sizeof(*inode_item
));
2669 inode_item
->generation
= cpu_to_le64(1);
2670 inode_item
->size
= cpu_to_le64(3);
2671 inode_item
->nlink
= cpu_to_le32(1);
2672 inode_item
->nblocks
= cpu_to_le64(1);
2673 inode_item
->mode
= cpu_to_le32(S_IFDIR
| 0755);
2675 btrfs_set_root_bytenr(&root_item
, leaf
->start
);
2676 btrfs_set_root_level(&root_item
, 0);
2677 btrfs_set_root_refs(&root_item
, 1);
2678 btrfs_set_root_used(&root_item
, 0);
2680 memset(&root_item
.drop_progress
, 0, sizeof(root_item
.drop_progress
));
2681 root_item
.drop_level
= 0;
2683 free_extent_buffer(leaf
);
2686 btrfs_set_root_dirid(&root_item
, new_dirid
);
2688 key
.objectid
= objectid
;
2690 btrfs_set_key_type(&key
, BTRFS_ROOT_ITEM_KEY
);
2691 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
2697 * insert the directory item
2699 key
.offset
= (u64
)-1;
2700 dir
= root
->fs_info
->sb
->s_root
->d_inode
;
2701 ret
= btrfs_insert_dir_item(trans
, root
->fs_info
->tree_root
,
2702 name
, namelen
, dir
->i_ino
, &key
,
2707 ret
= btrfs_insert_inode_ref(trans
, root
->fs_info
->tree_root
,
2708 name
, namelen
, objectid
,
2709 root
->fs_info
->sb
->s_root
->d_inode
->i_ino
);
2713 ret
= btrfs_commit_transaction(trans
, root
);
2717 new_root
= btrfs_read_fs_root(root
->fs_info
, &key
, name
, namelen
);
2720 trans
= btrfs_start_transaction(new_root
, 1);
2723 inode
= btrfs_new_inode(trans
, new_root
, "..", 2, new_dirid
,
2725 BTRFS_I(dir
)->block_group
, S_IFDIR
| 0700);
2728 inode
->i_op
= &btrfs_dir_inode_operations
;
2729 inode
->i_fop
= &btrfs_dir_file_operations
;
2730 new_root
->inode
= inode
;
2732 ret
= btrfs_insert_inode_ref(trans
, new_root
, "..", 2, new_dirid
,
2736 ret
= btrfs_update_inode(trans
, new_root
, inode
);
2740 nr
= trans
->blocks_used
;
2741 err
= btrfs_commit_transaction(trans
, new_root
);
2745 mutex_unlock(&root
->fs_info
->fs_mutex
);
2746 btrfs_btree_balance_dirty(root
, nr
);
2747 btrfs_throttle(root
);
2751 static int create_snapshot(struct btrfs_root
*root
, char *name
, int namelen
)
2753 struct btrfs_pending_snapshot
*pending_snapshot
;
2754 struct btrfs_trans_handle
*trans
;
2757 unsigned long nr
= 0;
2759 if (!root
->ref_cows
)
2762 mutex_lock(&root
->fs_info
->fs_mutex
);
2763 ret
= btrfs_check_free_space(root
, 1, 0);
2767 pending_snapshot
= kmalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
2768 if (!pending_snapshot
) {
2772 pending_snapshot
->name
= kmalloc(namelen
+ 1, GFP_NOFS
);
2773 if (!pending_snapshot
->name
) {
2775 kfree(pending_snapshot
);
2778 memcpy(pending_snapshot
->name
, name
, namelen
);
2779 pending_snapshot
->name
[namelen
] = '\0';
2780 trans
= btrfs_start_transaction(root
, 1);
2782 pending_snapshot
->root
= root
;
2783 list_add(&pending_snapshot
->list
,
2784 &trans
->transaction
->pending_snapshots
);
2785 ret
= btrfs_update_inode(trans
, root
, root
->inode
);
2786 err
= btrfs_commit_transaction(trans
, root
);
2789 mutex_unlock(&root
->fs_info
->fs_mutex
);
2790 btrfs_btree_balance_dirty(root
, nr
);
2791 btrfs_throttle(root
);
2795 unsigned long btrfs_force_ra(struct address_space
*mapping
,
2796 struct file_ra_state
*ra
, struct file
*file
,
2797 pgoff_t offset
, pgoff_t last_index
)
2801 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2802 req_size
= last_index
- offset
+ 1;
2803 offset
= page_cache_readahead(mapping
, ra
, file
, offset
, req_size
);
2806 req_size
= min(last_index
- offset
+ 1, (pgoff_t
)128);
2807 page_cache_sync_readahead(mapping
, ra
, file
, offset
, req_size
);
2808 return offset
+ req_size
;
2812 int btrfs_defrag_file(struct file
*file
) {
2813 struct inode
*inode
= fdentry(file
)->d_inode
;
2814 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2815 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2817 unsigned long last_index
;
2818 unsigned long ra_index
= 0;
2824 mutex_lock(&root
->fs_info
->fs_mutex
);
2825 ret
= btrfs_check_free_space(root
, inode
->i_size
, 0);
2826 mutex_unlock(&root
->fs_info
->fs_mutex
);
2830 mutex_lock(&inode
->i_mutex
);
2831 last_index
= inode
->i_size
>> PAGE_CACHE_SHIFT
;
2832 for (i
= 0; i
<= last_index
; i
++) {
2833 if (i
== ra_index
) {
2834 ra_index
= btrfs_force_ra(inode
->i_mapping
,
2836 file
, ra_index
, last_index
);
2838 page
= grab_cache_page(inode
->i_mapping
, i
);
2841 if (!PageUptodate(page
)) {
2842 btrfs_readpage(NULL
, page
);
2844 if (!PageUptodate(page
)) {
2846 page_cache_release(page
);
2850 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
2851 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
2853 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
2854 set_extent_delalloc(io_tree
, page_start
,
2855 page_end
, GFP_NOFS
);
2857 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
2858 set_page_dirty(page
);
2860 page_cache_release(page
);
2861 balance_dirty_pages_ratelimited_nr(inode
->i_mapping
, 1);
2865 mutex_unlock(&inode
->i_mutex
);
2869 static int btrfs_ioctl_resize(struct btrfs_root
*root
, void __user
*arg
)
2873 struct btrfs_ioctl_vol_args
*vol_args
;
2874 struct btrfs_trans_handle
*trans
;
2880 vol_args
= kmalloc(sizeof(*vol_args
), GFP_NOFS
);
2885 if (copy_from_user(vol_args
, arg
, sizeof(*vol_args
))) {
2889 namelen
= strlen(vol_args
->name
);
2890 if (namelen
> BTRFS_VOL_NAME_MAX
) {
2895 sizestr
= vol_args
->name
;
2896 if (!strcmp(sizestr
, "max"))
2897 new_size
= root
->fs_info
->sb
->s_bdev
->bd_inode
->i_size
;
2899 if (sizestr
[0] == '-') {
2902 } else if (sizestr
[0] == '+') {
2906 new_size
= btrfs_parse_size(sizestr
);
2907 if (new_size
== 0) {
2913 mutex_lock(&root
->fs_info
->fs_mutex
);
2914 old_size
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
2917 if (new_size
> old_size
) {
2921 new_size
= old_size
- new_size
;
2922 } else if (mod
> 0) {
2923 new_size
= old_size
+ new_size
;
2926 if (new_size
< 256 * 1024 * 1024) {
2930 if (new_size
> root
->fs_info
->sb
->s_bdev
->bd_inode
->i_size
) {
2935 do_div(new_size
, root
->sectorsize
);
2936 new_size
*= root
->sectorsize
;
2938 printk("new size is %Lu\n", new_size
);
2939 if (new_size
> old_size
) {
2940 trans
= btrfs_start_transaction(root
, 1);
2941 ret
= btrfs_grow_extent_tree(trans
, root
, new_size
);
2942 btrfs_commit_transaction(trans
, root
);
2944 ret
= btrfs_shrink_extent_tree(root
, new_size
);
2948 mutex_unlock(&root
->fs_info
->fs_mutex
);
2954 static int noinline
btrfs_ioctl_snap_create(struct btrfs_root
*root
,
2957 struct btrfs_ioctl_vol_args
*vol_args
;
2958 struct btrfs_dir_item
*di
;
2959 struct btrfs_path
*path
;
2964 vol_args
= kmalloc(sizeof(*vol_args
), GFP_NOFS
);
2969 if (copy_from_user(vol_args
, arg
, sizeof(*vol_args
))) {
2974 namelen
= strlen(vol_args
->name
);
2975 if (namelen
> BTRFS_VOL_NAME_MAX
) {
2979 if (strchr(vol_args
->name
, '/')) {
2984 path
= btrfs_alloc_path();
2990 root_dirid
= root
->fs_info
->sb
->s_root
->d_inode
->i_ino
,
2991 mutex_lock(&root
->fs_info
->fs_mutex
);
2992 di
= btrfs_lookup_dir_item(NULL
, root
->fs_info
->tree_root
,
2994 vol_args
->name
, namelen
, 0);
2995 mutex_unlock(&root
->fs_info
->fs_mutex
);
2996 btrfs_free_path(path
);
2998 if (di
&& !IS_ERR(di
)) {
3008 if (root
== root
->fs_info
->tree_root
)
3009 ret
= create_subvol(root
, vol_args
->name
, namelen
);
3011 ret
= create_snapshot(root
, vol_args
->name
, namelen
);
3017 static int btrfs_ioctl_defrag(struct file
*file
)
3019 struct inode
*inode
= fdentry(file
)->d_inode
;
3020 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3022 switch (inode
->i_mode
& S_IFMT
) {
3024 mutex_lock(&root
->fs_info
->fs_mutex
);
3025 btrfs_defrag_root(root
, 0);
3026 btrfs_defrag_root(root
->fs_info
->extent_root
, 0);
3027 mutex_unlock(&root
->fs_info
->fs_mutex
);
3030 btrfs_defrag_file(file
);
3037 long btrfs_ioctl(struct file
*file
, unsigned int
3038 cmd
, unsigned long arg
)
3040 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3043 case BTRFS_IOC_SNAP_CREATE
:
3044 return btrfs_ioctl_snap_create(root
, (void __user
*)arg
);
3045 case BTRFS_IOC_DEFRAG
:
3046 return btrfs_ioctl_defrag(file
);
3047 case BTRFS_IOC_RESIZE
:
3048 return btrfs_ioctl_resize(root
, (void __user
*)arg
);
3055 * Called inside transaction, so use GFP_NOFS
3057 struct inode
*btrfs_alloc_inode(struct super_block
*sb
)
3059 struct btrfs_inode
*ei
;
3061 ei
= kmem_cache_alloc(btrfs_inode_cachep
, GFP_NOFS
);
3065 ei
->ordered_trans
= 0;
3066 return &ei
->vfs_inode
;
3069 void btrfs_destroy_inode(struct inode
*inode
)
3071 WARN_ON(!list_empty(&inode
->i_dentry
));
3072 WARN_ON(inode
->i_data
.nrpages
);
3074 btrfs_drop_extent_cache(inode
, 0, (u64
)-1);
3075 kmem_cache_free(btrfs_inode_cachep
, BTRFS_I(inode
));
3078 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3079 static void init_once(struct kmem_cache
* cachep
, void *foo
)
3081 static void init_once(void * foo
, struct kmem_cache
* cachep
,
3082 unsigned long flags
)
3085 struct btrfs_inode
*ei
= (struct btrfs_inode
*) foo
;
3087 inode_init_once(&ei
->vfs_inode
);
3090 void btrfs_destroy_cachep(void)
3092 if (btrfs_inode_cachep
)
3093 kmem_cache_destroy(btrfs_inode_cachep
);
3094 if (btrfs_trans_handle_cachep
)
3095 kmem_cache_destroy(btrfs_trans_handle_cachep
);
3096 if (btrfs_transaction_cachep
)
3097 kmem_cache_destroy(btrfs_transaction_cachep
);
3098 if (btrfs_bit_radix_cachep
)
3099 kmem_cache_destroy(btrfs_bit_radix_cachep
);
3100 if (btrfs_path_cachep
)
3101 kmem_cache_destroy(btrfs_path_cachep
);
3104 struct kmem_cache
*btrfs_cache_create(const char *name
, size_t size
,
3105 unsigned long extra_flags
,
3106 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3107 void (*ctor
)(struct kmem_cache
*, void *)
3109 void (*ctor
)(void *, struct kmem_cache
*,
3114 return kmem_cache_create(name
, size
, 0, (SLAB_RECLAIM_ACCOUNT
|
3115 SLAB_MEM_SPREAD
| extra_flags
), ctor
3116 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3122 int btrfs_init_cachep(void)
3124 btrfs_inode_cachep
= btrfs_cache_create("btrfs_inode_cache",
3125 sizeof(struct btrfs_inode
),
3127 if (!btrfs_inode_cachep
)
3129 btrfs_trans_handle_cachep
=
3130 btrfs_cache_create("btrfs_trans_handle_cache",
3131 sizeof(struct btrfs_trans_handle
),
3133 if (!btrfs_trans_handle_cachep
)
3135 btrfs_transaction_cachep
= btrfs_cache_create("btrfs_transaction_cache",
3136 sizeof(struct btrfs_transaction
),
3138 if (!btrfs_transaction_cachep
)
3140 btrfs_path_cachep
= btrfs_cache_create("btrfs_path_cache",
3141 sizeof(struct btrfs_path
),
3143 if (!btrfs_path_cachep
)
3145 btrfs_bit_radix_cachep
= btrfs_cache_create("btrfs_radix", 256,
3146 SLAB_DESTROY_BY_RCU
, NULL
);
3147 if (!btrfs_bit_radix_cachep
)
3151 btrfs_destroy_cachep();
3155 static int btrfs_getattr(struct vfsmount
*mnt
,
3156 struct dentry
*dentry
, struct kstat
*stat
)
3158 struct inode
*inode
= dentry
->d_inode
;
3159 generic_fillattr(inode
, stat
);
3160 stat
->blksize
= PAGE_CACHE_SIZE
;
3161 stat
->blocks
= inode
->i_blocks
+ (BTRFS_I(inode
)->delalloc_bytes
>> 9);
3165 static int btrfs_rename(struct inode
* old_dir
, struct dentry
*old_dentry
,
3166 struct inode
* new_dir
,struct dentry
*new_dentry
)
3168 struct btrfs_trans_handle
*trans
;
3169 struct btrfs_root
*root
= BTRFS_I(old_dir
)->root
;
3170 struct inode
*new_inode
= new_dentry
->d_inode
;
3171 struct inode
*old_inode
= old_dentry
->d_inode
;
3172 struct timespec ctime
= CURRENT_TIME
;
3173 struct btrfs_path
*path
;
3176 if (S_ISDIR(old_inode
->i_mode
) && new_inode
&&
3177 new_inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
3181 mutex_lock(&root
->fs_info
->fs_mutex
);
3182 ret
= btrfs_check_free_space(root
, 1, 0);
3186 trans
= btrfs_start_transaction(root
, 1);
3188 btrfs_set_trans_block_group(trans
, new_dir
);
3189 path
= btrfs_alloc_path();
3195 old_dentry
->d_inode
->i_nlink
++;
3196 old_dir
->i_ctime
= old_dir
->i_mtime
= ctime
;
3197 new_dir
->i_ctime
= new_dir
->i_mtime
= ctime
;
3198 old_inode
->i_ctime
= ctime
;
3200 ret
= btrfs_unlink_trans(trans
, root
, old_dir
, old_dentry
);
3205 new_inode
->i_ctime
= CURRENT_TIME
;
3206 ret
= btrfs_unlink_trans(trans
, root
, new_dir
, new_dentry
);
3210 ret
= btrfs_add_link(trans
, new_dentry
, old_inode
, 1);
3215 btrfs_free_path(path
);
3216 btrfs_end_transaction(trans
, root
);
3218 mutex_unlock(&root
->fs_info
->fs_mutex
);
3222 static int btrfs_symlink(struct inode
*dir
, struct dentry
*dentry
,
3223 const char *symname
)
3225 struct btrfs_trans_handle
*trans
;
3226 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
3227 struct btrfs_path
*path
;
3228 struct btrfs_key key
;
3229 struct inode
*inode
= NULL
;
3236 struct btrfs_file_extent_item
*ei
;
3237 struct extent_buffer
*leaf
;
3238 unsigned long nr
= 0;
3240 name_len
= strlen(symname
) + 1;
3241 if (name_len
> BTRFS_MAX_INLINE_DATA_SIZE(root
))
3242 return -ENAMETOOLONG
;
3244 mutex_lock(&root
->fs_info
->fs_mutex
);
3245 err
= btrfs_check_free_space(root
, 1, 0);
3249 trans
= btrfs_start_transaction(root
, 1);
3250 btrfs_set_trans_block_group(trans
, dir
);
3252 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
3258 inode
= btrfs_new_inode(trans
, root
, dentry
->d_name
.name
,
3260 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
3261 BTRFS_I(dir
)->block_group
, S_IFLNK
|S_IRWXUGO
);
3262 err
= PTR_ERR(inode
);
3266 btrfs_set_trans_block_group(trans
, inode
);
3267 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
3271 inode
->i_mapping
->a_ops
= &btrfs_aops
;
3272 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3273 inode
->i_fop
= &btrfs_file_operations
;
3274 inode
->i_op
= &btrfs_file_inode_operations
;
3275 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
3276 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
3277 inode
->i_mapping
, GFP_NOFS
);
3278 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
3279 inode
->i_mapping
, GFP_NOFS
);
3280 BTRFS_I(inode
)->delalloc_bytes
= 0;
3281 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
3283 dir
->i_sb
->s_dirt
= 1;
3284 btrfs_update_inode_block_group(trans
, inode
);
3285 btrfs_update_inode_block_group(trans
, dir
);
3289 path
= btrfs_alloc_path();
3291 key
.objectid
= inode
->i_ino
;
3293 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
3294 datasize
= btrfs_file_extent_calc_inline_size(name_len
);
3295 err
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
3301 leaf
= path
->nodes
[0];
3302 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
3303 struct btrfs_file_extent_item
);
3304 btrfs_set_file_extent_generation(leaf
, ei
, trans
->transid
);
3305 btrfs_set_file_extent_type(leaf
, ei
,
3306 BTRFS_FILE_EXTENT_INLINE
);
3307 ptr
= btrfs_file_extent_inline_start(ei
);
3308 write_extent_buffer(leaf
, symname
, ptr
, name_len
);
3309 btrfs_mark_buffer_dirty(leaf
);
3310 btrfs_free_path(path
);
3312 inode
->i_op
= &btrfs_symlink_inode_operations
;
3313 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
3314 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3315 inode
->i_size
= name_len
- 1;
3316 err
= btrfs_update_inode(trans
, root
, inode
);
3321 nr
= trans
->blocks_used
;
3322 btrfs_end_transaction(trans
, root
);
3324 mutex_unlock(&root
->fs_info
->fs_mutex
);
3326 inode_dec_link_count(inode
);
3329 btrfs_btree_balance_dirty(root
, nr
);
3330 btrfs_throttle(root
);
3334 static int btrfs_permission(struct inode
*inode
, int mask
,
3335 struct nameidata
*nd
)
3337 if (btrfs_test_flag(inode
, READONLY
) && (mask
& MAY_WRITE
))
3339 return generic_permission(inode
, mask
, NULL
);
3342 static struct inode_operations btrfs_dir_inode_operations
= {
3343 .lookup
= btrfs_lookup
,
3344 .create
= btrfs_create
,
3345 .unlink
= btrfs_unlink
,
3347 .mkdir
= btrfs_mkdir
,
3348 .rmdir
= btrfs_rmdir
,
3349 .rename
= btrfs_rename
,
3350 .symlink
= btrfs_symlink
,
3351 .setattr
= btrfs_setattr
,
3352 .mknod
= btrfs_mknod
,
3353 .setxattr
= generic_setxattr
,
3354 .getxattr
= generic_getxattr
,
3355 .listxattr
= btrfs_listxattr
,
3356 .removexattr
= generic_removexattr
,
3357 .permission
= btrfs_permission
,
3359 static struct inode_operations btrfs_dir_ro_inode_operations
= {
3360 .lookup
= btrfs_lookup
,
3361 .permission
= btrfs_permission
,
3363 static struct file_operations btrfs_dir_file_operations
= {
3364 .llseek
= generic_file_llseek
,
3365 .read
= generic_read_dir
,
3366 .readdir
= btrfs_readdir
,
3367 .unlocked_ioctl
= btrfs_ioctl
,
3368 #ifdef CONFIG_COMPAT
3369 .compat_ioctl
= btrfs_ioctl
,
3373 static struct extent_io_ops btrfs_extent_io_ops
= {
3374 .fill_delalloc
= run_delalloc_range
,
3375 .submit_bio_hook
= btrfs_submit_bio_hook
,
3376 .merge_bio_hook
= btrfs_merge_bio_hook
,
3377 .readpage_io_hook
= btrfs_readpage_io_hook
,
3378 .readpage_end_io_hook
= btrfs_readpage_end_io_hook
,
3379 .readpage_io_failed_hook
= btrfs_readpage_io_failed_hook
,
3380 .set_bit_hook
= btrfs_set_bit_hook
,
3381 .clear_bit_hook
= btrfs_clear_bit_hook
,
3384 static struct address_space_operations btrfs_aops
= {
3385 .readpage
= btrfs_readpage
,
3386 .writepage
= btrfs_writepage
,
3387 .writepages
= btrfs_writepages
,
3388 .readpages
= btrfs_readpages
,
3389 .sync_page
= block_sync_page
,
3391 .direct_IO
= btrfs_direct_IO
,
3392 .invalidatepage
= btrfs_invalidatepage
,
3393 .releasepage
= btrfs_releasepage
,
3394 .set_page_dirty
= __set_page_dirty_nobuffers
,
3397 static struct address_space_operations btrfs_symlink_aops
= {
3398 .readpage
= btrfs_readpage
,
3399 .writepage
= btrfs_writepage
,
3400 .invalidatepage
= btrfs_invalidatepage
,
3401 .releasepage
= btrfs_releasepage
,
3404 static struct inode_operations btrfs_file_inode_operations
= {
3405 .truncate
= btrfs_truncate
,
3406 .getattr
= btrfs_getattr
,
3407 .setattr
= btrfs_setattr
,
3408 .setxattr
= generic_setxattr
,
3409 .getxattr
= generic_getxattr
,
3410 .listxattr
= btrfs_listxattr
,
3411 .removexattr
= generic_removexattr
,
3412 .permission
= btrfs_permission
,
3414 static struct inode_operations btrfs_special_inode_operations
= {
3415 .getattr
= btrfs_getattr
,
3416 .setattr
= btrfs_setattr
,
3417 .permission
= btrfs_permission
,
3419 static struct inode_operations btrfs_symlink_inode_operations
= {
3420 .readlink
= generic_readlink
,
3421 .follow_link
= page_follow_link_light
,
3422 .put_link
= page_put_link
,
3423 .permission
= btrfs_permission
,