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>
39 #include <linux/posix_acl.h>
42 #include "transaction.h"
43 #include "btrfs_inode.h"
45 #include "print-tree.h"
47 #include "ordered-data.h"
49 struct btrfs_iget_args
{
51 struct btrfs_root
*root
;
54 static struct inode_operations btrfs_dir_inode_operations
;
55 static struct inode_operations btrfs_symlink_inode_operations
;
56 static struct inode_operations btrfs_dir_ro_inode_operations
;
57 static struct inode_operations btrfs_special_inode_operations
;
58 static struct inode_operations btrfs_file_inode_operations
;
59 static struct address_space_operations btrfs_aops
;
60 static struct address_space_operations btrfs_symlink_aops
;
61 static struct file_operations btrfs_dir_file_operations
;
62 static struct extent_io_ops btrfs_extent_io_ops
;
64 static struct kmem_cache
*btrfs_inode_cachep
;
65 struct kmem_cache
*btrfs_trans_handle_cachep
;
66 struct kmem_cache
*btrfs_transaction_cachep
;
67 struct kmem_cache
*btrfs_bit_radix_cachep
;
68 struct kmem_cache
*btrfs_path_cachep
;
71 static unsigned char btrfs_type_by_mode
[S_IFMT
>> S_SHIFT
] = {
72 [S_IFREG
>> S_SHIFT
] = BTRFS_FT_REG_FILE
,
73 [S_IFDIR
>> S_SHIFT
] = BTRFS_FT_DIR
,
74 [S_IFCHR
>> S_SHIFT
] = BTRFS_FT_CHRDEV
,
75 [S_IFBLK
>> S_SHIFT
] = BTRFS_FT_BLKDEV
,
76 [S_IFIFO
>> S_SHIFT
] = BTRFS_FT_FIFO
,
77 [S_IFSOCK
>> S_SHIFT
] = BTRFS_FT_SOCK
,
78 [S_IFLNK
>> S_SHIFT
] = BTRFS_FT_SYMLINK
,
81 static void btrfs_truncate(struct inode
*inode
);
83 int btrfs_check_free_space(struct btrfs_root
*root
, u64 num_required
,
92 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
93 total
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
94 used
= btrfs_super_bytes_used(&root
->fs_info
->super_copy
);
102 if (used
+ root
->fs_info
->delalloc_bytes
+ num_required
> thresh
)
104 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
108 static int cow_file_range(struct inode
*inode
, u64 start
, u64 end
)
110 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
111 struct btrfs_trans_handle
*trans
;
115 u64 blocksize
= root
->sectorsize
;
117 struct btrfs_key ins
;
118 struct extent_map
*em
;
119 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
122 trans
= btrfs_join_transaction(root
, 1);
124 btrfs_set_trans_block_group(trans
, inode
);
126 num_bytes
= (end
- start
+ blocksize
) & ~(blocksize
- 1);
127 num_bytes
= max(blocksize
, num_bytes
);
128 orig_num_bytes
= num_bytes
;
130 if (alloc_hint
== EXTENT_MAP_INLINE
)
133 BUG_ON(num_bytes
> btrfs_super_total_bytes(&root
->fs_info
->super_copy
));
134 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
135 btrfs_drop_extent_cache(inode
, start
, start
+ num_bytes
- 1);
136 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
138 while(num_bytes
> 0) {
139 cur_alloc_size
= min(num_bytes
, root
->fs_info
->max_extent
);
140 ret
= btrfs_reserve_extent(trans
, root
, cur_alloc_size
,
141 root
->sectorsize
, 0, 0,
147 em
= alloc_extent_map(GFP_NOFS
);
149 em
->len
= ins
.offset
;
150 em
->block_start
= ins
.objectid
;
151 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
152 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
153 set_bit(EXTENT_FLAG_PINNED
, &em
->flags
);
155 spin_lock(&em_tree
->lock
);
156 ret
= add_extent_mapping(em_tree
, em
);
157 spin_unlock(&em_tree
->lock
);
158 if (ret
!= -EEXIST
) {
162 btrfs_drop_extent_cache(inode
, start
,
163 start
+ ins
.offset
- 1);
165 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
167 cur_alloc_size
= ins
.offset
;
168 ret
= btrfs_add_ordered_extent(inode
, start
, ins
.objectid
,
171 if (num_bytes
< cur_alloc_size
) {
172 printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes
,
176 num_bytes
-= cur_alloc_size
;
177 alloc_hint
= ins
.objectid
+ ins
.offset
;
178 start
+= cur_alloc_size
;
181 btrfs_end_transaction(trans
, root
);
185 static int run_delalloc_nocow(struct inode
*inode
, u64 start
, u64 end
)
193 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
194 struct btrfs_block_group_cache
*block_group
;
195 struct extent_buffer
*leaf
;
197 struct btrfs_path
*path
;
198 struct btrfs_file_extent_item
*item
;
201 struct btrfs_key found_key
;
203 total_fs_bytes
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
204 path
= btrfs_alloc_path();
207 ret
= btrfs_lookup_file_extent(NULL
, root
, path
,
208 inode
->i_ino
, start
, 0);
210 btrfs_free_path(path
);
216 if (path
->slots
[0] == 0)
221 leaf
= path
->nodes
[0];
222 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
223 struct btrfs_file_extent_item
);
225 /* are we inside the extent that was found? */
226 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
227 found_type
= btrfs_key_type(&found_key
);
228 if (found_key
.objectid
!= inode
->i_ino
||
229 found_type
!= BTRFS_EXTENT_DATA_KEY
)
232 found_type
= btrfs_file_extent_type(leaf
, item
);
233 extent_start
= found_key
.offset
;
234 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
235 u64 extent_num_bytes
;
237 extent_num_bytes
= btrfs_file_extent_num_bytes(leaf
, item
);
238 extent_end
= extent_start
+ extent_num_bytes
;
241 if (loops
&& start
!= extent_start
)
244 if (start
< extent_start
|| start
>= extent_end
)
247 cow_end
= min(end
, extent_end
- 1);
248 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
252 if (btrfs_cross_ref_exists(root
, &found_key
, bytenr
))
255 * we may be called by the resizer, make sure we're inside
256 * the limits of the FS
258 block_group
= btrfs_lookup_block_group(root
->fs_info
,
260 if (!block_group
|| block_group
->ro
)
269 btrfs_free_path(path
);
272 btrfs_release_path(root
, path
);
277 btrfs_release_path(root
, path
);
278 cow_file_range(inode
, start
, end
);
283 static int run_delalloc_range(struct inode
*inode
, u64 start
, u64 end
)
285 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
288 if (btrfs_test_opt(root
, NODATACOW
) ||
289 btrfs_test_flag(inode
, NODATACOW
))
290 ret
= run_delalloc_nocow(inode
, start
, end
);
292 ret
= cow_file_range(inode
, start
, end
);
297 int btrfs_set_bit_hook(struct inode
*inode
, u64 start
, u64 end
,
298 unsigned long old
, unsigned long bits
)
301 if (!(old
& EXTENT_DELALLOC
) && (bits
& EXTENT_DELALLOC
)) {
302 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
303 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
304 BTRFS_I(inode
)->delalloc_bytes
+= end
- start
+ 1;
305 root
->fs_info
->delalloc_bytes
+= end
- start
+ 1;
306 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
311 int btrfs_clear_bit_hook(struct inode
*inode
, u64 start
, u64 end
,
312 unsigned long old
, unsigned long bits
)
314 if ((old
& EXTENT_DELALLOC
) && (bits
& EXTENT_DELALLOC
)) {
315 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
318 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
319 if (end
- start
+ 1 > root
->fs_info
->delalloc_bytes
) {
320 printk("warning: delalloc account %Lu %Lu\n",
321 end
- start
+ 1, root
->fs_info
->delalloc_bytes
);
322 root
->fs_info
->delalloc_bytes
= 0;
323 BTRFS_I(inode
)->delalloc_bytes
= 0;
325 root
->fs_info
->delalloc_bytes
-= end
- start
+ 1;
326 BTRFS_I(inode
)->delalloc_bytes
-= end
- start
+ 1;
328 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
333 int btrfs_merge_bio_hook(struct page
*page
, unsigned long offset
,
334 size_t size
, struct bio
*bio
)
336 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
337 struct btrfs_mapping_tree
*map_tree
;
338 u64 logical
= bio
->bi_sector
<< 9;
343 length
= bio
->bi_size
;
344 map_tree
= &root
->fs_info
->mapping_tree
;
346 ret
= btrfs_map_block(map_tree
, READ
, logical
,
347 &map_length
, NULL
, 0);
349 if (map_length
< length
+ size
) {
355 int __btrfs_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
358 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
361 ret
= btrfs_csum_one_bio(root
, inode
, bio
);
364 return btrfs_map_bio(root
, rw
, bio
, mirror_num
, 1);
367 int btrfs_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
370 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
373 ret
= btrfs_bio_wq_end_io(root
->fs_info
, bio
, 0);
376 if (!(rw
& (1 << BIO_RW
))) {
380 return btrfs_wq_submit_bio(BTRFS_I(inode
)->root
->fs_info
,
381 inode
, rw
, bio
, mirror_num
,
382 __btrfs_submit_bio_hook
);
384 return btrfs_map_bio(root
, rw
, bio
, mirror_num
, 0);
387 static noinline
int add_pending_csums(struct btrfs_trans_handle
*trans
,
388 struct inode
*inode
, u64 file_offset
,
389 struct list_head
*list
)
391 struct list_head
*cur
;
392 struct btrfs_ordered_sum
*sum
;
394 btrfs_set_trans_block_group(trans
, inode
);
395 list_for_each(cur
, list
) {
396 sum
= list_entry(cur
, struct btrfs_ordered_sum
, list
);
397 mutex_lock(&BTRFS_I(inode
)->csum_mutex
);
398 btrfs_csum_file_blocks(trans
, BTRFS_I(inode
)->root
,
400 mutex_unlock(&BTRFS_I(inode
)->csum_mutex
);
405 struct btrfs_writepage_fixup
{
407 struct btrfs_work work
;
410 /* see btrfs_writepage_start_hook for details on why this is required */
411 void btrfs_writepage_fixup_worker(struct btrfs_work
*work
)
413 struct btrfs_writepage_fixup
*fixup
;
414 struct btrfs_ordered_extent
*ordered
;
420 fixup
= container_of(work
, struct btrfs_writepage_fixup
, work
);
424 if (!page
->mapping
|| !PageDirty(page
) || !PageChecked(page
)) {
425 ClearPageChecked(page
);
429 inode
= page
->mapping
->host
;
430 page_start
= page_offset(page
);
431 page_end
= page_offset(page
) + PAGE_CACHE_SIZE
- 1;
433 lock_extent(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
, GFP_NOFS
);
435 /* already ordered? We're done */
436 if (test_range_bit(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
,
437 EXTENT_ORDERED
, 0)) {
441 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
443 unlock_extent(&BTRFS_I(inode
)->io_tree
, page_start
,
446 btrfs_start_ordered_extent(inode
, ordered
, 1);
450 set_extent_delalloc(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
,
452 ClearPageChecked(page
);
454 unlock_extent(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
, GFP_NOFS
);
457 page_cache_release(page
);
461 * There are a few paths in the higher layers of the kernel that directly
462 * set the page dirty bit without asking the filesystem if it is a
463 * good idea. This causes problems because we want to make sure COW
464 * properly happens and the data=ordered rules are followed.
466 * In our case any range that doesn't have the EXTENT_ORDERED bit set
467 * hasn't been properly setup for IO. We kick off an async process
468 * to fix it up. The async helper will wait for ordered extents, set
469 * the delalloc bit and make it safe to write the page.
471 int btrfs_writepage_start_hook(struct page
*page
, u64 start
, u64 end
)
473 struct inode
*inode
= page
->mapping
->host
;
474 struct btrfs_writepage_fixup
*fixup
;
475 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
478 ret
= test_range_bit(&BTRFS_I(inode
)->io_tree
, start
, end
,
483 if (PageChecked(page
))
486 fixup
= kzalloc(sizeof(*fixup
), GFP_NOFS
);
490 SetPageChecked(page
);
491 page_cache_get(page
);
492 fixup
->work
.func
= btrfs_writepage_fixup_worker
;
494 btrfs_queue_worker(&root
->fs_info
->fixup_workers
, &fixup
->work
);
498 static int btrfs_finish_ordered_io(struct inode
*inode
, u64 start
, u64 end
)
500 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
501 struct btrfs_trans_handle
*trans
;
502 struct btrfs_ordered_extent
*ordered_extent
;
503 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
504 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
505 struct extent_map
*em
;
506 struct extent_map
*em_orig
;
510 struct list_head list
;
511 struct btrfs_key ins
;
515 ret
= btrfs_dec_test_ordered_pending(inode
, start
, end
- start
+ 1);
519 trans
= btrfs_join_transaction(root
, 1);
521 ordered_extent
= btrfs_lookup_ordered_extent(inode
, start
);
522 BUG_ON(!ordered_extent
);
524 lock_extent(io_tree
, ordered_extent
->file_offset
,
525 ordered_extent
->file_offset
+ ordered_extent
->len
- 1,
528 INIT_LIST_HEAD(&list
);
530 ins
.objectid
= ordered_extent
->start
;
531 ins
.offset
= ordered_extent
->len
;
532 ins
.type
= BTRFS_EXTENT_ITEM_KEY
;
534 ret
= btrfs_alloc_reserved_extent(trans
, root
, root
->root_key
.objectid
,
535 trans
->transid
, inode
->i_ino
,
536 ordered_extent
->file_offset
, &ins
);
539 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
541 spin_lock(&em_tree
->lock
);
542 clear_start
= ordered_extent
->file_offset
;
543 clear_end
= ordered_extent
->file_offset
+ ordered_extent
->len
;
544 em
= lookup_extent_mapping(em_tree
, clear_start
,
545 ordered_extent
->len
);
547 while(em
&& clear_start
< extent_map_end(em
) && clear_end
> em
->start
) {
548 clear_bit(EXTENT_FLAG_PINNED
, &em
->flags
);
549 rb
= rb_next(&em
->rb_node
);
552 em
= rb_entry(rb
, struct extent_map
, rb_node
);
554 free_extent_map(em_orig
);
555 spin_unlock(&em_tree
->lock
);
557 ret
= btrfs_drop_extents(trans
, root
, inode
,
558 ordered_extent
->file_offset
,
559 ordered_extent
->file_offset
+
561 ordered_extent
->file_offset
, &alloc_hint
);
563 ret
= btrfs_insert_file_extent(trans
, root
, inode
->i_ino
,
564 ordered_extent
->file_offset
,
565 ordered_extent
->start
,
567 ordered_extent
->len
, 0);
570 btrfs_drop_extent_cache(inode
, ordered_extent
->file_offset
,
571 ordered_extent
->file_offset
+
572 ordered_extent
->len
- 1);
573 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
575 inode
->i_blocks
+= ordered_extent
->len
>> 9;
576 unlock_extent(io_tree
, ordered_extent
->file_offset
,
577 ordered_extent
->file_offset
+ ordered_extent
->len
- 1,
579 add_pending_csums(trans
, inode
, ordered_extent
->file_offset
,
580 &ordered_extent
->list
);
582 btrfs_ordered_update_i_size(inode
, ordered_extent
);
583 btrfs_remove_ordered_extent(inode
, ordered_extent
);
586 btrfs_put_ordered_extent(ordered_extent
);
587 /* once for the tree */
588 btrfs_put_ordered_extent(ordered_extent
);
590 btrfs_update_inode(trans
, root
, inode
);
591 btrfs_end_transaction(trans
, root
);
595 int btrfs_writepage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
596 struct extent_state
*state
, int uptodate
)
598 return btrfs_finish_ordered_io(page
->mapping
->host
, start
, end
);
601 int btrfs_readpage_io_hook(struct page
*page
, u64 start
, u64 end
)
604 struct inode
*inode
= page
->mapping
->host
;
605 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
606 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
607 struct btrfs_csum_item
*item
;
608 struct btrfs_path
*path
= NULL
;
611 if (btrfs_test_opt(root
, NODATASUM
) ||
612 btrfs_test_flag(inode
, NODATASUM
))
616 * It is possible there is an ordered extent that has
617 * not yet finished for this range in the file. If so,
618 * that extent will have a csum cached, and it will insert
619 * the sum after all the blocks in the extent are fully
620 * on disk. So, look for an ordered extent and use the
621 * sum if found. We have to do this before looking in the
622 * btree because csum items are pre-inserted based on
623 * the file size. btrfs_lookup_csum might find an item
624 * that still hasn't been fully filled.
626 ret
= btrfs_find_ordered_sum(inode
, start
, &csum
);
631 path
= btrfs_alloc_path();
632 item
= btrfs_lookup_csum(NULL
, root
, path
, inode
->i_ino
, start
, 0);
635 /* a csum that isn't present is a preallocated region. */
636 if (ret
== -ENOENT
|| ret
== -EFBIG
)
639 printk("no csum found for inode %lu start %Lu\n", inode
->i_ino
,
643 read_extent_buffer(path
->nodes
[0], &csum
, (unsigned long)item
,
646 set_state_private(io_tree
, start
, csum
);
649 btrfs_free_path(path
);
653 struct io_failure_record
{
661 int btrfs_io_failed_hook(struct bio
*failed_bio
,
662 struct page
*page
, u64 start
, u64 end
,
663 struct extent_state
*state
)
665 struct io_failure_record
*failrec
= NULL
;
667 struct extent_map
*em
;
668 struct inode
*inode
= page
->mapping
->host
;
669 struct extent_io_tree
*failure_tree
= &BTRFS_I(inode
)->io_failure_tree
;
670 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
677 ret
= get_state_private(failure_tree
, start
, &private);
679 failrec
= kmalloc(sizeof(*failrec
), GFP_NOFS
);
682 failrec
->start
= start
;
683 failrec
->len
= end
- start
+ 1;
684 failrec
->last_mirror
= 0;
686 spin_lock(&em_tree
->lock
);
687 em
= lookup_extent_mapping(em_tree
, start
, failrec
->len
);
688 if (em
->start
> start
|| em
->start
+ em
->len
< start
) {
692 spin_unlock(&em_tree
->lock
);
694 if (!em
|| IS_ERR(em
)) {
698 logical
= start
- em
->start
;
699 logical
= em
->block_start
+ logical
;
700 failrec
->logical
= logical
;
702 set_extent_bits(failure_tree
, start
, end
, EXTENT_LOCKED
|
703 EXTENT_DIRTY
, GFP_NOFS
);
704 set_state_private(failure_tree
, start
,
705 (u64
)(unsigned long)failrec
);
707 failrec
= (struct io_failure_record
*)(unsigned long)private;
709 num_copies
= btrfs_num_copies(
710 &BTRFS_I(inode
)->root
->fs_info
->mapping_tree
,
711 failrec
->logical
, failrec
->len
);
712 failrec
->last_mirror
++;
714 spin_lock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
715 state
= find_first_extent_bit_state(&BTRFS_I(inode
)->io_tree
,
718 if (state
&& state
->start
!= failrec
->start
)
720 spin_unlock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
722 if (!state
|| failrec
->last_mirror
> num_copies
) {
723 set_state_private(failure_tree
, failrec
->start
, 0);
724 clear_extent_bits(failure_tree
, failrec
->start
,
725 failrec
->start
+ failrec
->len
- 1,
726 EXTENT_LOCKED
| EXTENT_DIRTY
, GFP_NOFS
);
730 bio
= bio_alloc(GFP_NOFS
, 1);
731 bio
->bi_private
= state
;
732 bio
->bi_end_io
= failed_bio
->bi_end_io
;
733 bio
->bi_sector
= failrec
->logical
>> 9;
734 bio
->bi_bdev
= failed_bio
->bi_bdev
;
736 bio_add_page(bio
, page
, failrec
->len
, start
- page_offset(page
));
737 if (failed_bio
->bi_rw
& (1 << BIO_RW
))
742 BTRFS_I(inode
)->io_tree
.ops
->submit_bio_hook(inode
, rw
, bio
,
743 failrec
->last_mirror
);
747 int btrfs_clean_io_failures(struct inode
*inode
, u64 start
)
751 struct io_failure_record
*failure
;
755 if (count_range_bits(&BTRFS_I(inode
)->io_failure_tree
, &private,
756 (u64
)-1, 1, EXTENT_DIRTY
)) {
757 ret
= get_state_private(&BTRFS_I(inode
)->io_failure_tree
,
758 start
, &private_failure
);
760 failure
= (struct io_failure_record
*)(unsigned long)
762 set_state_private(&BTRFS_I(inode
)->io_failure_tree
,
764 clear_extent_bits(&BTRFS_I(inode
)->io_failure_tree
,
766 failure
->start
+ failure
->len
- 1,
767 EXTENT_DIRTY
| EXTENT_LOCKED
,
775 int btrfs_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
776 struct extent_state
*state
)
778 size_t offset
= start
- ((u64
)page
->index
<< PAGE_CACHE_SHIFT
);
779 struct inode
*inode
= page
->mapping
->host
;
780 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
782 u64
private = ~(u32
)0;
784 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
788 if (btrfs_test_opt(root
, NODATASUM
) ||
789 btrfs_test_flag(inode
, NODATASUM
))
791 if (state
&& state
->start
== start
) {
792 private = state
->private;
795 ret
= get_state_private(io_tree
, start
, &private);
797 local_irq_save(flags
);
798 kaddr
= kmap_atomic(page
, KM_IRQ0
);
802 csum
= btrfs_csum_data(root
, kaddr
+ offset
, csum
, end
- start
+ 1);
803 btrfs_csum_final(csum
, (char *)&csum
);
804 if (csum
!= private) {
807 kunmap_atomic(kaddr
, KM_IRQ0
);
808 local_irq_restore(flags
);
810 /* if the io failure tree for this inode is non-empty,
811 * check to see if we've recovered from a failed IO
813 btrfs_clean_io_failures(inode
, start
);
817 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
818 page
->mapping
->host
->i_ino
, (unsigned long long)start
, csum
,
820 memset(kaddr
+ offset
, 1, end
- start
+ 1);
821 flush_dcache_page(page
);
822 kunmap_atomic(kaddr
, KM_IRQ0
);
823 local_irq_restore(flags
);
830 * This creates an orphan entry for the given inode in case something goes
831 * wrong in the middle of an unlink/truncate.
833 int btrfs_orphan_add(struct btrfs_trans_handle
*trans
, struct inode
*inode
)
835 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
838 spin_lock(&root
->list_lock
);
840 /* already on the orphan list, we're good */
841 if (!list_empty(&BTRFS_I(inode
)->i_orphan
)) {
842 spin_unlock(&root
->list_lock
);
846 list_add(&BTRFS_I(inode
)->i_orphan
, &root
->orphan_list
);
848 spin_unlock(&root
->list_lock
);
851 * insert an orphan item to track this unlinked/truncated file
853 ret
= btrfs_insert_orphan_item(trans
, root
, inode
->i_ino
);
859 * We have done the truncate/delete so we can go ahead and remove the orphan
860 * item for this particular inode.
862 int btrfs_orphan_del(struct btrfs_trans_handle
*trans
, struct inode
*inode
)
864 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
867 spin_lock(&root
->list_lock
);
869 if (list_empty(&BTRFS_I(inode
)->i_orphan
)) {
870 spin_unlock(&root
->list_lock
);
874 list_del_init(&BTRFS_I(inode
)->i_orphan
);
876 spin_unlock(&root
->list_lock
);
880 spin_unlock(&root
->list_lock
);
882 ret
= btrfs_del_orphan_item(trans
, root
, inode
->i_ino
);
888 * this cleans up any orphans that may be left on the list from the last use
891 void btrfs_orphan_cleanup(struct btrfs_root
*root
)
893 struct btrfs_path
*path
;
894 struct extent_buffer
*leaf
;
895 struct btrfs_item
*item
;
896 struct btrfs_key key
, found_key
;
897 struct btrfs_trans_handle
*trans
;
899 int ret
= 0, nr_unlink
= 0, nr_truncate
= 0;
901 /* don't do orphan cleanup if the fs is readonly. */
902 if (root
->inode
->i_sb
->s_flags
& MS_RDONLY
)
905 path
= btrfs_alloc_path();
910 key
.objectid
= BTRFS_ORPHAN_OBJECTID
;
911 btrfs_set_key_type(&key
, BTRFS_ORPHAN_ITEM_KEY
);
912 key
.offset
= (u64
)-1;
914 trans
= btrfs_start_transaction(root
, 1);
915 btrfs_set_trans_block_group(trans
, root
->inode
);
918 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
920 printk(KERN_ERR
"Error searching slot for orphan: %d"
926 * if ret == 0 means we found what we were searching for, which
927 * is weird, but possible, so only screw with path if we didnt
928 * find the key and see if we have stuff that matches
931 if (path
->slots
[0] == 0)
936 /* pull out the item */
937 leaf
= path
->nodes
[0];
938 item
= btrfs_item_nr(leaf
, path
->slots
[0]);
939 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
941 /* make sure the item matches what we want */
942 if (found_key
.objectid
!= BTRFS_ORPHAN_OBJECTID
)
944 if (btrfs_key_type(&found_key
) != BTRFS_ORPHAN_ITEM_KEY
)
947 /* release the path since we're done with it */
948 btrfs_release_path(root
, path
);
951 * this is where we are basically btrfs_lookup, without the
952 * crossing root thing. we store the inode number in the
953 * offset of the orphan item.
955 inode
= btrfs_iget_locked(root
->inode
->i_sb
,
956 found_key
.offset
, root
);
960 if (inode
->i_state
& I_NEW
) {
961 BTRFS_I(inode
)->root
= root
;
963 /* have to set the location manually */
964 BTRFS_I(inode
)->location
.objectid
= inode
->i_ino
;
965 BTRFS_I(inode
)->location
.type
= BTRFS_INODE_ITEM_KEY
;
966 BTRFS_I(inode
)->location
.offset
= 0;
968 btrfs_read_locked_inode(inode
);
969 unlock_new_inode(inode
);
973 * add this inode to the orphan list so btrfs_orphan_del does
974 * the proper thing when we hit it
976 spin_lock(&root
->list_lock
);
977 list_add(&BTRFS_I(inode
)->i_orphan
, &root
->orphan_list
);
978 spin_unlock(&root
->list_lock
);
981 * if this is a bad inode, means we actually succeeded in
982 * removing the inode, but not the orphan record, which means
983 * we need to manually delete the orphan since iput will just
986 if (is_bad_inode(inode
)) {
987 btrfs_orphan_del(trans
, inode
);
992 /* if we have links, this was a truncate, lets do that */
993 if (inode
->i_nlink
) {
995 btrfs_truncate(inode
);
1000 /* this will do delete_inode and everything for us */
1005 printk(KERN_INFO
"btrfs: unlinked %d orphans\n", nr_unlink
);
1007 printk(KERN_INFO
"btrfs: truncated %d orphans\n", nr_truncate
);
1009 btrfs_free_path(path
);
1010 btrfs_end_transaction(trans
, root
);
1013 void btrfs_read_locked_inode(struct inode
*inode
)
1015 struct btrfs_path
*path
;
1016 struct extent_buffer
*leaf
;
1017 struct btrfs_inode_item
*inode_item
;
1018 struct btrfs_timespec
*tspec
;
1019 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1020 struct btrfs_key location
;
1021 u64 alloc_group_block
;
1025 path
= btrfs_alloc_path();
1027 memcpy(&location
, &BTRFS_I(inode
)->location
, sizeof(location
));
1029 ret
= btrfs_lookup_inode(NULL
, root
, path
, &location
, 0);
1033 leaf
= path
->nodes
[0];
1034 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
1035 struct btrfs_inode_item
);
1037 inode
->i_mode
= btrfs_inode_mode(leaf
, inode_item
);
1038 inode
->i_nlink
= btrfs_inode_nlink(leaf
, inode_item
);
1039 inode
->i_uid
= btrfs_inode_uid(leaf
, inode_item
);
1040 inode
->i_gid
= btrfs_inode_gid(leaf
, inode_item
);
1041 btrfs_i_size_write(inode
, btrfs_inode_size(leaf
, inode_item
));
1043 tspec
= btrfs_inode_atime(inode_item
);
1044 inode
->i_atime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1045 inode
->i_atime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1047 tspec
= btrfs_inode_mtime(inode_item
);
1048 inode
->i_mtime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1049 inode
->i_mtime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1051 tspec
= btrfs_inode_ctime(inode_item
);
1052 inode
->i_ctime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1053 inode
->i_ctime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1055 inode
->i_blocks
= btrfs_inode_nblocks(leaf
, inode_item
);
1056 inode
->i_generation
= btrfs_inode_generation(leaf
, inode_item
);
1058 rdev
= btrfs_inode_rdev(leaf
, inode_item
);
1060 BTRFS_I(inode
)->index_cnt
= (u64
)-1;
1062 alloc_group_block
= btrfs_inode_block_group(leaf
, inode_item
);
1063 BTRFS_I(inode
)->block_group
= btrfs_lookup_block_group(root
->fs_info
,
1065 BTRFS_I(inode
)->flags
= btrfs_inode_flags(leaf
, inode_item
);
1066 if (!BTRFS_I(inode
)->block_group
) {
1067 BTRFS_I(inode
)->block_group
= btrfs_find_block_group(root
,
1069 BTRFS_BLOCK_GROUP_METADATA
, 0);
1071 btrfs_free_path(path
);
1074 switch (inode
->i_mode
& S_IFMT
) {
1076 inode
->i_mapping
->a_ops
= &btrfs_aops
;
1077 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1078 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
1079 inode
->i_fop
= &btrfs_file_operations
;
1080 inode
->i_op
= &btrfs_file_inode_operations
;
1083 inode
->i_fop
= &btrfs_dir_file_operations
;
1084 if (root
== root
->fs_info
->tree_root
)
1085 inode
->i_op
= &btrfs_dir_ro_inode_operations
;
1087 inode
->i_op
= &btrfs_dir_inode_operations
;
1090 inode
->i_op
= &btrfs_symlink_inode_operations
;
1091 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
1092 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1095 init_special_inode(inode
, inode
->i_mode
, rdev
);
1101 btrfs_free_path(path
);
1102 make_bad_inode(inode
);
1105 static void fill_inode_item(struct extent_buffer
*leaf
,
1106 struct btrfs_inode_item
*item
,
1107 struct inode
*inode
)
1109 btrfs_set_inode_uid(leaf
, item
, inode
->i_uid
);
1110 btrfs_set_inode_gid(leaf
, item
, inode
->i_gid
);
1111 btrfs_set_inode_size(leaf
, item
, BTRFS_I(inode
)->disk_i_size
);
1112 btrfs_set_inode_mode(leaf
, item
, inode
->i_mode
);
1113 btrfs_set_inode_nlink(leaf
, item
, inode
->i_nlink
);
1115 btrfs_set_timespec_sec(leaf
, btrfs_inode_atime(item
),
1116 inode
->i_atime
.tv_sec
);
1117 btrfs_set_timespec_nsec(leaf
, btrfs_inode_atime(item
),
1118 inode
->i_atime
.tv_nsec
);
1120 btrfs_set_timespec_sec(leaf
, btrfs_inode_mtime(item
),
1121 inode
->i_mtime
.tv_sec
);
1122 btrfs_set_timespec_nsec(leaf
, btrfs_inode_mtime(item
),
1123 inode
->i_mtime
.tv_nsec
);
1125 btrfs_set_timespec_sec(leaf
, btrfs_inode_ctime(item
),
1126 inode
->i_ctime
.tv_sec
);
1127 btrfs_set_timespec_nsec(leaf
, btrfs_inode_ctime(item
),
1128 inode
->i_ctime
.tv_nsec
);
1130 btrfs_set_inode_nblocks(leaf
, item
, inode
->i_blocks
);
1131 btrfs_set_inode_generation(leaf
, item
, inode
->i_generation
);
1132 btrfs_set_inode_rdev(leaf
, item
, inode
->i_rdev
);
1133 btrfs_set_inode_flags(leaf
, item
, BTRFS_I(inode
)->flags
);
1134 btrfs_set_inode_block_group(leaf
, item
,
1135 BTRFS_I(inode
)->block_group
->key
.objectid
);
1138 int noinline
btrfs_update_inode(struct btrfs_trans_handle
*trans
,
1139 struct btrfs_root
*root
,
1140 struct inode
*inode
)
1142 struct btrfs_inode_item
*inode_item
;
1143 struct btrfs_path
*path
;
1144 struct extent_buffer
*leaf
;
1147 path
= btrfs_alloc_path();
1149 ret
= btrfs_lookup_inode(trans
, root
, path
,
1150 &BTRFS_I(inode
)->location
, 1);
1157 leaf
= path
->nodes
[0];
1158 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
1159 struct btrfs_inode_item
);
1161 fill_inode_item(leaf
, inode_item
, inode
);
1162 btrfs_mark_buffer_dirty(leaf
);
1163 btrfs_set_inode_last_trans(trans
, inode
);
1166 btrfs_free_path(path
);
1171 static int btrfs_unlink_trans(struct btrfs_trans_handle
*trans
,
1172 struct btrfs_root
*root
,
1174 struct dentry
*dentry
)
1176 struct btrfs_path
*path
;
1177 const char *name
= dentry
->d_name
.name
;
1178 int name_len
= dentry
->d_name
.len
;
1180 struct extent_buffer
*leaf
;
1181 struct btrfs_dir_item
*di
;
1182 struct btrfs_key key
;
1185 path
= btrfs_alloc_path();
1191 di
= btrfs_lookup_dir_item(trans
, root
, path
, dir
->i_ino
,
1192 name
, name_len
, -1);
1201 leaf
= path
->nodes
[0];
1202 btrfs_dir_item_key_to_cpu(leaf
, di
, &key
);
1203 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
1206 btrfs_release_path(root
, path
);
1208 ret
= btrfs_del_inode_ref(trans
, root
, name
, name_len
,
1209 dentry
->d_inode
->i_ino
,
1210 dentry
->d_parent
->d_inode
->i_ino
, &index
);
1212 printk("failed to delete reference to %.*s, "
1213 "inode %lu parent %lu\n", name_len
, name
,
1214 dentry
->d_inode
->i_ino
,
1215 dentry
->d_parent
->d_inode
->i_ino
);
1219 di
= btrfs_lookup_dir_index_item(trans
, root
, path
, dir
->i_ino
,
1220 index
, name
, name_len
, -1);
1229 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
1230 btrfs_release_path(root
, path
);
1232 dentry
->d_inode
->i_ctime
= dir
->i_ctime
;
1234 btrfs_free_path(path
);
1236 btrfs_i_size_write(dir
, dir
->i_size
- name_len
* 2);
1237 dir
->i_mtime
= dir
->i_ctime
= CURRENT_TIME
;
1238 btrfs_update_inode(trans
, root
, dir
);
1239 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1240 dentry
->d_inode
->i_nlink
--;
1242 drop_nlink(dentry
->d_inode
);
1244 ret
= btrfs_update_inode(trans
, root
, dentry
->d_inode
);
1245 dir
->i_sb
->s_dirt
= 1;
1250 static int btrfs_unlink(struct inode
*dir
, struct dentry
*dentry
)
1252 struct btrfs_root
*root
;
1253 struct btrfs_trans_handle
*trans
;
1254 struct inode
*inode
= dentry
->d_inode
;
1256 unsigned long nr
= 0;
1258 root
= BTRFS_I(dir
)->root
;
1260 ret
= btrfs_check_free_space(root
, 1, 1);
1264 trans
= btrfs_start_transaction(root
, 1);
1266 btrfs_set_trans_block_group(trans
, dir
);
1267 ret
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
1269 if (inode
->i_nlink
== 0)
1270 ret
= btrfs_orphan_add(trans
, inode
);
1272 nr
= trans
->blocks_used
;
1274 btrfs_end_transaction_throttle(trans
, root
);
1276 btrfs_btree_balance_dirty(root
, nr
);
1280 static int btrfs_rmdir(struct inode
*dir
, struct dentry
*dentry
)
1282 struct inode
*inode
= dentry
->d_inode
;
1285 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1286 struct btrfs_trans_handle
*trans
;
1287 unsigned long nr
= 0;
1289 if (inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
1293 ret
= btrfs_check_free_space(root
, 1, 1);
1297 trans
= btrfs_start_transaction(root
, 1);
1298 btrfs_set_trans_block_group(trans
, dir
);
1300 err
= btrfs_orphan_add(trans
, inode
);
1304 /* now the directory is empty */
1305 err
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
1307 btrfs_i_size_write(inode
, 0);
1311 nr
= trans
->blocks_used
;
1312 ret
= btrfs_end_transaction_throttle(trans
, root
);
1314 btrfs_btree_balance_dirty(root
, nr
);
1322 * this can truncate away extent items, csum items and directory items.
1323 * It starts at a high offset and removes keys until it can't find
1324 * any higher than i_size.
1326 * csum items that cross the new i_size are truncated to the new size
1329 * min_type is the minimum key type to truncate down to. If set to 0, this
1330 * will kill all the items on this inode, including the INODE_ITEM_KEY.
1332 static int btrfs_truncate_in_trans(struct btrfs_trans_handle
*trans
,
1333 struct btrfs_root
*root
,
1334 struct inode
*inode
,
1338 struct btrfs_path
*path
;
1339 struct btrfs_key key
;
1340 struct btrfs_key found_key
;
1342 struct extent_buffer
*leaf
;
1343 struct btrfs_file_extent_item
*fi
;
1344 u64 extent_start
= 0;
1345 u64 extent_num_bytes
= 0;
1351 int pending_del_nr
= 0;
1352 int pending_del_slot
= 0;
1353 int extent_type
= -1;
1354 u64 mask
= root
->sectorsize
- 1;
1356 btrfs_drop_extent_cache(inode
, inode
->i_size
& (~mask
), (u64
)-1);
1357 path
= btrfs_alloc_path();
1361 /* FIXME, add redo link to tree so we don't leak on crash */
1362 key
.objectid
= inode
->i_ino
;
1363 key
.offset
= (u64
)-1;
1366 btrfs_init_path(path
);
1368 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
1373 BUG_ON(path
->slots
[0] == 0);
1379 leaf
= path
->nodes
[0];
1380 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
1381 found_type
= btrfs_key_type(&found_key
);
1383 if (found_key
.objectid
!= inode
->i_ino
)
1386 if (found_type
< min_type
)
1389 item_end
= found_key
.offset
;
1390 if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
1391 fi
= btrfs_item_ptr(leaf
, path
->slots
[0],
1392 struct btrfs_file_extent_item
);
1393 extent_type
= btrfs_file_extent_type(leaf
, fi
);
1394 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1396 btrfs_file_extent_num_bytes(leaf
, fi
);
1397 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1398 struct btrfs_item
*item
= btrfs_item_nr(leaf
,
1400 item_end
+= btrfs_file_extent_inline_len(leaf
,
1405 if (found_type
== BTRFS_CSUM_ITEM_KEY
) {
1406 ret
= btrfs_csum_truncate(trans
, root
, path
,
1410 if (item_end
< inode
->i_size
) {
1411 if (found_type
== BTRFS_DIR_ITEM_KEY
) {
1412 found_type
= BTRFS_INODE_ITEM_KEY
;
1413 } else if (found_type
== BTRFS_EXTENT_ITEM_KEY
) {
1414 found_type
= BTRFS_CSUM_ITEM_KEY
;
1415 } else if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
1416 found_type
= BTRFS_XATTR_ITEM_KEY
;
1417 } else if (found_type
== BTRFS_XATTR_ITEM_KEY
) {
1418 found_type
= BTRFS_INODE_REF_KEY
;
1419 } else if (found_type
) {
1424 btrfs_set_key_type(&key
, found_type
);
1427 if (found_key
.offset
>= inode
->i_size
)
1433 /* FIXME, shrink the extent if the ref count is only 1 */
1434 if (found_type
!= BTRFS_EXTENT_DATA_KEY
)
1437 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1439 extent_start
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
1441 u64 orig_num_bytes
=
1442 btrfs_file_extent_num_bytes(leaf
, fi
);
1443 extent_num_bytes
= inode
->i_size
-
1444 found_key
.offset
+ root
->sectorsize
- 1;
1445 extent_num_bytes
= extent_num_bytes
&
1446 ~((u64
)root
->sectorsize
- 1);
1447 btrfs_set_file_extent_num_bytes(leaf
, fi
,
1449 num_dec
= (orig_num_bytes
-
1451 if (extent_start
!= 0)
1452 dec_i_blocks(inode
, num_dec
);
1453 btrfs_mark_buffer_dirty(leaf
);
1456 btrfs_file_extent_disk_num_bytes(leaf
,
1458 /* FIXME blocksize != 4096 */
1459 num_dec
= btrfs_file_extent_num_bytes(leaf
, fi
);
1460 if (extent_start
!= 0) {
1462 dec_i_blocks(inode
, num_dec
);
1464 root_gen
= btrfs_header_generation(leaf
);
1465 root_owner
= btrfs_header_owner(leaf
);
1467 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1469 u32 newsize
= inode
->i_size
- found_key
.offset
;
1470 dec_i_blocks(inode
, item_end
+ 1 -
1471 found_key
.offset
- newsize
);
1473 btrfs_file_extent_calc_inline_size(newsize
);
1474 ret
= btrfs_truncate_item(trans
, root
, path
,
1478 dec_i_blocks(inode
, item_end
+ 1 -
1484 if (!pending_del_nr
) {
1485 /* no pending yet, add ourselves */
1486 pending_del_slot
= path
->slots
[0];
1488 } else if (pending_del_nr
&&
1489 path
->slots
[0] + 1 == pending_del_slot
) {
1490 /* hop on the pending chunk */
1492 pending_del_slot
= path
->slots
[0];
1494 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path
->slots
[0], pending_del_nr
, pending_del_slot
);
1500 ret
= btrfs_free_extent(trans
, root
, extent_start
,
1503 root_gen
, inode
->i_ino
,
1504 found_key
.offset
, 0);
1508 if (path
->slots
[0] == 0) {
1511 btrfs_release_path(root
, path
);
1516 if (pending_del_nr
&&
1517 path
->slots
[0] + 1 != pending_del_slot
) {
1518 struct btrfs_key debug
;
1520 btrfs_item_key_to_cpu(path
->nodes
[0], &debug
,
1522 ret
= btrfs_del_items(trans
, root
, path
,
1527 btrfs_release_path(root
, path
);
1533 if (pending_del_nr
) {
1534 ret
= btrfs_del_items(trans
, root
, path
, pending_del_slot
,
1537 btrfs_free_path(path
);
1538 inode
->i_sb
->s_dirt
= 1;
1543 * taken from block_truncate_page, but does cow as it zeros out
1544 * any bytes left in the last page in the file.
1546 static int btrfs_truncate_page(struct address_space
*mapping
, loff_t from
)
1548 struct inode
*inode
= mapping
->host
;
1549 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1550 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1551 struct btrfs_ordered_extent
*ordered
;
1553 u32 blocksize
= root
->sectorsize
;
1554 pgoff_t index
= from
>> PAGE_CACHE_SHIFT
;
1555 unsigned offset
= from
& (PAGE_CACHE_SIZE
-1);
1561 if ((offset
& (blocksize
- 1)) == 0)
1566 page
= grab_cache_page(mapping
, index
);
1570 page_start
= page_offset(page
);
1571 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1573 if (!PageUptodate(page
)) {
1574 ret
= btrfs_readpage(NULL
, page
);
1576 if (page
->mapping
!= mapping
) {
1578 page_cache_release(page
);
1581 if (!PageUptodate(page
)) {
1586 wait_on_page_writeback(page
);
1588 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1589 set_page_extent_mapped(page
);
1591 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
1593 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1595 page_cache_release(page
);
1596 btrfs_start_ordered_extent(inode
, ordered
, 1);
1597 btrfs_put_ordered_extent(ordered
);
1601 set_extent_delalloc(&BTRFS_I(inode
)->io_tree
, page_start
,
1602 page_end
, GFP_NOFS
);
1604 if (offset
!= PAGE_CACHE_SIZE
) {
1606 memset(kaddr
+ offset
, 0, PAGE_CACHE_SIZE
- offset
);
1607 flush_dcache_page(page
);
1610 ClearPageChecked(page
);
1611 set_page_dirty(page
);
1612 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1616 page_cache_release(page
);
1621 static int btrfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
1623 struct inode
*inode
= dentry
->d_inode
;
1626 err
= inode_change_ok(inode
, attr
);
1630 if (S_ISREG(inode
->i_mode
) &&
1631 attr
->ia_valid
& ATTR_SIZE
&& attr
->ia_size
> inode
->i_size
) {
1632 struct btrfs_trans_handle
*trans
;
1633 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1634 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1636 u64 mask
= root
->sectorsize
- 1;
1637 u64 hole_start
= (inode
->i_size
+ mask
) & ~mask
;
1638 u64 block_end
= (attr
->ia_size
+ mask
) & ~mask
;
1642 if (attr
->ia_size
<= hole_start
)
1645 err
= btrfs_check_free_space(root
, 1, 0);
1649 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
1651 hole_size
= block_end
- hole_start
;
1652 btrfs_wait_ordered_range(inode
, hole_start
, hole_size
);
1653 lock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1655 trans
= btrfs_start_transaction(root
, 1);
1656 btrfs_set_trans_block_group(trans
, inode
);
1657 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
1658 err
= btrfs_drop_extents(trans
, root
, inode
,
1659 hole_start
, block_end
, hole_start
,
1662 if (alloc_hint
!= EXTENT_MAP_INLINE
) {
1663 err
= btrfs_insert_file_extent(trans
, root
,
1667 btrfs_drop_extent_cache(inode
, hole_start
,
1669 btrfs_check_file(root
, inode
);
1671 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
1672 btrfs_end_transaction(trans
, root
);
1673 unlock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1678 err
= inode_setattr(inode
, attr
);
1680 if (!err
&& ((attr
->ia_valid
& ATTR_MODE
)))
1681 err
= btrfs_acl_chmod(inode
);
1686 void btrfs_delete_inode(struct inode
*inode
)
1688 struct btrfs_trans_handle
*trans
;
1689 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1693 truncate_inode_pages(&inode
->i_data
, 0);
1694 if (is_bad_inode(inode
)) {
1695 btrfs_orphan_del(NULL
, inode
);
1698 btrfs_wait_ordered_range(inode
, 0, (u64
)-1);
1700 btrfs_i_size_write(inode
, 0);
1701 trans
= btrfs_start_transaction(root
, 1);
1703 btrfs_set_trans_block_group(trans
, inode
);
1704 ret
= btrfs_truncate_in_trans(trans
, root
, inode
, 0);
1706 btrfs_orphan_del(NULL
, inode
);
1707 goto no_delete_lock
;
1710 btrfs_orphan_del(trans
, inode
);
1712 nr
= trans
->blocks_used
;
1715 btrfs_end_transaction(trans
, root
);
1716 btrfs_btree_balance_dirty(root
, nr
);
1720 nr
= trans
->blocks_used
;
1721 btrfs_end_transaction(trans
, root
);
1722 btrfs_btree_balance_dirty(root
, nr
);
1728 * this returns the key found in the dir entry in the location pointer.
1729 * If no dir entries were found, location->objectid is 0.
1731 static int btrfs_inode_by_name(struct inode
*dir
, struct dentry
*dentry
,
1732 struct btrfs_key
*location
)
1734 const char *name
= dentry
->d_name
.name
;
1735 int namelen
= dentry
->d_name
.len
;
1736 struct btrfs_dir_item
*di
;
1737 struct btrfs_path
*path
;
1738 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1741 if (namelen
== 1 && strcmp(name
, ".") == 0) {
1742 location
->objectid
= dir
->i_ino
;
1743 location
->type
= BTRFS_INODE_ITEM_KEY
;
1744 location
->offset
= 0;
1747 path
= btrfs_alloc_path();
1750 if (namelen
== 2 && strcmp(name
, "..") == 0) {
1751 struct btrfs_key key
;
1752 struct extent_buffer
*leaf
;
1755 key
.objectid
= dir
->i_ino
;
1756 key
.offset
= (u64
)-1;
1757 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1758 if (ret
< 0 || path
->slots
[0] == 0)
1760 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1763 leaf
= path
->nodes
[0];
1764 slot
= path
->slots
[0] - 1;
1766 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
1767 if (key
.objectid
!= dir
->i_ino
||
1768 key
.type
!= BTRFS_INODE_REF_KEY
) {
1771 location
->objectid
= key
.offset
;
1772 location
->type
= BTRFS_INODE_ITEM_KEY
;
1773 location
->offset
= 0;
1777 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir
->i_ino
, name
,
1781 if (!di
|| IS_ERR(di
)) {
1784 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, location
);
1786 btrfs_free_path(path
);
1789 location
->objectid
= 0;
1794 * when we hit a tree root in a directory, the btrfs part of the inode
1795 * needs to be changed to reflect the root directory of the tree root. This
1796 * is kind of like crossing a mount point.
1798 static int fixup_tree_root_location(struct btrfs_root
*root
,
1799 struct btrfs_key
*location
,
1800 struct btrfs_root
**sub_root
,
1801 struct dentry
*dentry
)
1803 struct btrfs_root_item
*ri
;
1805 if (btrfs_key_type(location
) != BTRFS_ROOT_ITEM_KEY
)
1807 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1810 *sub_root
= btrfs_read_fs_root(root
->fs_info
, location
,
1811 dentry
->d_name
.name
,
1812 dentry
->d_name
.len
);
1813 if (IS_ERR(*sub_root
))
1814 return PTR_ERR(*sub_root
);
1816 ri
= &(*sub_root
)->root_item
;
1817 location
->objectid
= btrfs_root_dirid(ri
);
1818 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1819 location
->offset
= 0;
1824 static int btrfs_init_locked_inode(struct inode
*inode
, void *p
)
1826 struct btrfs_iget_args
*args
= p
;
1827 inode
->i_ino
= args
->ino
;
1828 BTRFS_I(inode
)->root
= args
->root
;
1829 BTRFS_I(inode
)->delalloc_bytes
= 0;
1830 BTRFS_I(inode
)->disk_i_size
= 0;
1831 BTRFS_I(inode
)->index_cnt
= (u64
)-1;
1832 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1833 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1834 inode
->i_mapping
, GFP_NOFS
);
1835 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1836 inode
->i_mapping
, GFP_NOFS
);
1837 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
1838 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
1839 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
1843 static int btrfs_find_actor(struct inode
*inode
, void *opaque
)
1845 struct btrfs_iget_args
*args
= opaque
;
1846 return (args
->ino
== inode
->i_ino
&&
1847 args
->root
== BTRFS_I(inode
)->root
);
1850 struct inode
*btrfs_ilookup(struct super_block
*s
, u64 objectid
,
1853 struct btrfs_iget_args args
;
1854 args
.ino
= objectid
;
1855 args
.root
= btrfs_lookup_fs_root(btrfs_sb(s
)->fs_info
, root_objectid
);
1860 return ilookup5(s
, objectid
, btrfs_find_actor
, (void *)&args
);
1863 struct inode
*btrfs_iget_locked(struct super_block
*s
, u64 objectid
,
1864 struct btrfs_root
*root
)
1866 struct inode
*inode
;
1867 struct btrfs_iget_args args
;
1868 args
.ino
= objectid
;
1871 inode
= iget5_locked(s
, objectid
, btrfs_find_actor
,
1872 btrfs_init_locked_inode
,
1877 static struct dentry
*btrfs_lookup(struct inode
*dir
, struct dentry
*dentry
,
1878 struct nameidata
*nd
)
1880 struct inode
* inode
;
1881 struct btrfs_inode
*bi
= BTRFS_I(dir
);
1882 struct btrfs_root
*root
= bi
->root
;
1883 struct btrfs_root
*sub_root
= root
;
1884 struct btrfs_key location
;
1885 int ret
, do_orphan
= 0;
1887 if (dentry
->d_name
.len
> BTRFS_NAME_LEN
)
1888 return ERR_PTR(-ENAMETOOLONG
);
1890 ret
= btrfs_inode_by_name(dir
, dentry
, &location
);
1893 return ERR_PTR(ret
);
1896 if (location
.objectid
) {
1897 ret
= fixup_tree_root_location(root
, &location
, &sub_root
,
1900 return ERR_PTR(ret
);
1902 return ERR_PTR(-ENOENT
);
1904 inode
= btrfs_iget_locked(dir
->i_sb
, location
.objectid
,
1907 return ERR_PTR(-EACCES
);
1908 if (inode
->i_state
& I_NEW
) {
1909 /* the inode and parent dir are two different roots */
1910 if (sub_root
!= root
) {
1912 sub_root
->inode
= inode
;
1915 BTRFS_I(inode
)->root
= sub_root
;
1916 memcpy(&BTRFS_I(inode
)->location
, &location
,
1918 btrfs_read_locked_inode(inode
);
1919 unlock_new_inode(inode
);
1923 if (unlikely(do_orphan
))
1924 btrfs_orphan_cleanup(sub_root
);
1926 return d_splice_alias(inode
, dentry
);
1929 static unsigned char btrfs_filetype_table
[] = {
1930 DT_UNKNOWN
, DT_REG
, DT_DIR
, DT_CHR
, DT_BLK
, DT_FIFO
, DT_SOCK
, DT_LNK
1933 static int btrfs_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
1935 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1936 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1937 struct btrfs_item
*item
;
1938 struct btrfs_dir_item
*di
;
1939 struct btrfs_key key
;
1940 struct btrfs_key found_key
;
1941 struct btrfs_path
*path
;
1944 struct extent_buffer
*leaf
;
1947 unsigned char d_type
;
1952 int key_type
= BTRFS_DIR_INDEX_KEY
;
1957 /* FIXME, use a real flag for deciding about the key type */
1958 if (root
->fs_info
->tree_root
== root
)
1959 key_type
= BTRFS_DIR_ITEM_KEY
;
1961 /* special case for "." */
1962 if (filp
->f_pos
== 0) {
1963 over
= filldir(dirent
, ".", 1,
1971 key
.objectid
= inode
->i_ino
;
1972 path
= btrfs_alloc_path();
1975 /* special case for .., just use the back ref */
1976 if (filp
->f_pos
== 1) {
1977 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1978 key
.offset
= (u64
)-1;
1979 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1980 if (ret
< 0 || path
->slots
[0] == 0) {
1981 btrfs_release_path(root
, path
);
1982 goto read_dir_items
;
1985 leaf
= path
->nodes
[0];
1986 slot
= path
->slots
[0] - 1;
1987 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1988 btrfs_release_path(root
, path
);
1989 if (found_key
.objectid
!= key
.objectid
||
1990 found_key
.type
!= BTRFS_INODE_REF_KEY
)
1991 goto read_dir_items
;
1992 over
= filldir(dirent
, "..", 2,
1993 2, found_key
.offset
, DT_DIR
);
2000 btrfs_set_key_type(&key
, key_type
);
2001 key
.offset
= filp
->f_pos
;
2003 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2008 leaf
= path
->nodes
[0];
2009 nritems
= btrfs_header_nritems(leaf
);
2010 slot
= path
->slots
[0];
2011 if (advance
|| slot
>= nritems
) {
2012 if (slot
>= nritems
-1) {
2013 ret
= btrfs_next_leaf(root
, path
);
2016 leaf
= path
->nodes
[0];
2017 nritems
= btrfs_header_nritems(leaf
);
2018 slot
= path
->slots
[0];
2025 item
= btrfs_item_nr(leaf
, slot
);
2026 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
2028 if (found_key
.objectid
!= key
.objectid
)
2030 if (btrfs_key_type(&found_key
) != key_type
)
2032 if (found_key
.offset
< filp
->f_pos
)
2035 filp
->f_pos
= found_key
.offset
;
2037 di
= btrfs_item_ptr(leaf
, slot
, struct btrfs_dir_item
);
2039 di_total
= btrfs_item_size(leaf
, item
);
2040 while(di_cur
< di_total
) {
2041 struct btrfs_key location
;
2043 name_len
= btrfs_dir_name_len(leaf
, di
);
2044 if (name_len
< 32) {
2045 name_ptr
= tmp_name
;
2047 name_ptr
= kmalloc(name_len
, GFP_NOFS
);
2050 read_extent_buffer(leaf
, name_ptr
,
2051 (unsigned long)(di
+ 1), name_len
);
2053 d_type
= btrfs_filetype_table
[btrfs_dir_type(leaf
, di
)];
2054 btrfs_dir_item_key_to_cpu(leaf
, di
, &location
);
2055 over
= filldir(dirent
, name_ptr
, name_len
,
2060 if (name_ptr
!= tmp_name
)
2065 di_len
= btrfs_dir_name_len(leaf
, di
) +
2066 btrfs_dir_data_len(leaf
, di
) +sizeof(*di
);
2068 di
= (struct btrfs_dir_item
*)((char *)di
+ di_len
);
2071 if (key_type
== BTRFS_DIR_INDEX_KEY
)
2072 filp
->f_pos
= INT_LIMIT(typeof(filp
->f_pos
));
2078 btrfs_free_path(path
);
2082 int btrfs_write_inode(struct inode
*inode
, int wait
)
2084 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2085 struct btrfs_trans_handle
*trans
;
2089 trans
= btrfs_join_transaction(root
, 1);
2090 btrfs_set_trans_block_group(trans
, inode
);
2091 ret
= btrfs_commit_transaction(trans
, root
);
2097 * This is somewhat expensive, updating the tree every time the
2098 * inode changes. But, it is most likely to find the inode in cache.
2099 * FIXME, needs more benchmarking...there are no reasons other than performance
2100 * to keep or drop this code.
2102 void btrfs_dirty_inode(struct inode
*inode
)
2104 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2105 struct btrfs_trans_handle
*trans
;
2107 trans
= btrfs_join_transaction(root
, 1);
2108 btrfs_set_trans_block_group(trans
, inode
);
2109 btrfs_update_inode(trans
, root
, inode
);
2110 btrfs_end_transaction(trans
, root
);
2113 static int btrfs_set_inode_index_count(struct inode
*inode
)
2115 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2116 struct btrfs_key key
, found_key
;
2117 struct btrfs_path
*path
;
2118 struct extent_buffer
*leaf
;
2121 key
.objectid
= inode
->i_ino
;
2122 btrfs_set_key_type(&key
, BTRFS_DIR_INDEX_KEY
);
2123 key
.offset
= (u64
)-1;
2125 path
= btrfs_alloc_path();
2129 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2132 /* FIXME: we should be able to handle this */
2138 * MAGIC NUMBER EXPLANATION:
2139 * since we search a directory based on f_pos we have to start at 2
2140 * since '.' and '..' have f_pos of 0 and 1 respectively, so everybody
2141 * else has to start at 2
2143 if (path
->slots
[0] == 0) {
2144 BTRFS_I(inode
)->index_cnt
= 2;
2150 leaf
= path
->nodes
[0];
2151 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2153 if (found_key
.objectid
!= inode
->i_ino
||
2154 btrfs_key_type(&found_key
) != BTRFS_DIR_INDEX_KEY
) {
2155 BTRFS_I(inode
)->index_cnt
= 2;
2159 BTRFS_I(inode
)->index_cnt
= found_key
.offset
+ 1;
2161 btrfs_free_path(path
);
2165 static int btrfs_set_inode_index(struct inode
*dir
, struct inode
*inode
)
2169 if (BTRFS_I(dir
)->index_cnt
== (u64
)-1) {
2170 ret
= btrfs_set_inode_index_count(dir
);
2175 BTRFS_I(inode
)->index
= BTRFS_I(dir
)->index_cnt
;
2176 BTRFS_I(dir
)->index_cnt
++;
2181 static struct inode
*btrfs_new_inode(struct btrfs_trans_handle
*trans
,
2182 struct btrfs_root
*root
,
2184 const char *name
, int name_len
,
2187 struct btrfs_block_group_cache
*group
,
2190 struct inode
*inode
;
2191 struct btrfs_inode_item
*inode_item
;
2192 struct btrfs_block_group_cache
*new_inode_group
;
2193 struct btrfs_key
*location
;
2194 struct btrfs_path
*path
;
2195 struct btrfs_inode_ref
*ref
;
2196 struct btrfs_key key
[2];
2202 path
= btrfs_alloc_path();
2205 inode
= new_inode(root
->fs_info
->sb
);
2207 return ERR_PTR(-ENOMEM
);
2210 ret
= btrfs_set_inode_index(dir
, inode
);
2212 return ERR_PTR(ret
);
2214 BTRFS_I(inode
)->index
= 0;
2217 * index_cnt is ignored for everything but a dir,
2218 * btrfs_get_inode_index_count has an explanation for the magic
2221 BTRFS_I(inode
)->index_cnt
= 2;
2223 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
2224 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
2225 inode
->i_mapping
, GFP_NOFS
);
2226 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
2227 inode
->i_mapping
, GFP_NOFS
);
2228 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
2229 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
2230 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
2231 BTRFS_I(inode
)->delalloc_bytes
= 0;
2232 BTRFS_I(inode
)->disk_i_size
= 0;
2233 BTRFS_I(inode
)->root
= root
;
2239 new_inode_group
= btrfs_find_block_group(root
, group
, 0,
2240 BTRFS_BLOCK_GROUP_METADATA
, owner
);
2241 if (!new_inode_group
) {
2242 printk("find_block group failed\n");
2243 new_inode_group
= group
;
2245 BTRFS_I(inode
)->block_group
= new_inode_group
;
2246 BTRFS_I(inode
)->flags
= 0;
2248 key
[0].objectid
= objectid
;
2249 btrfs_set_key_type(&key
[0], BTRFS_INODE_ITEM_KEY
);
2252 key
[1].objectid
= objectid
;
2253 btrfs_set_key_type(&key
[1], BTRFS_INODE_REF_KEY
);
2254 key
[1].offset
= ref_objectid
;
2256 sizes
[0] = sizeof(struct btrfs_inode_item
);
2257 sizes
[1] = name_len
+ sizeof(*ref
);
2259 ret
= btrfs_insert_empty_items(trans
, root
, path
, key
, sizes
, 2);
2263 if (objectid
> root
->highest_inode
)
2264 root
->highest_inode
= objectid
;
2266 inode
->i_uid
= current
->fsuid
;
2267 inode
->i_gid
= current
->fsgid
;
2268 inode
->i_mode
= mode
;
2269 inode
->i_ino
= objectid
;
2270 inode
->i_blocks
= 0;
2271 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2272 inode_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
2273 struct btrfs_inode_item
);
2274 fill_inode_item(path
->nodes
[0], inode_item
, inode
);
2276 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0] + 1,
2277 struct btrfs_inode_ref
);
2278 btrfs_set_inode_ref_name_len(path
->nodes
[0], ref
, name_len
);
2279 btrfs_set_inode_ref_index(path
->nodes
[0], ref
, BTRFS_I(inode
)->index
);
2280 ptr
= (unsigned long)(ref
+ 1);
2281 write_extent_buffer(path
->nodes
[0], name
, ptr
, name_len
);
2283 btrfs_mark_buffer_dirty(path
->nodes
[0]);
2284 btrfs_free_path(path
);
2286 location
= &BTRFS_I(inode
)->location
;
2287 location
->objectid
= objectid
;
2288 location
->offset
= 0;
2289 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
2291 insert_inode_hash(inode
);
2295 BTRFS_I(dir
)->index_cnt
--;
2296 btrfs_free_path(path
);
2297 return ERR_PTR(ret
);
2300 static inline u8
btrfs_inode_type(struct inode
*inode
)
2302 return btrfs_type_by_mode
[(inode
->i_mode
& S_IFMT
) >> S_SHIFT
];
2305 static int btrfs_add_link(struct btrfs_trans_handle
*trans
,
2306 struct dentry
*dentry
, struct inode
*inode
,
2310 struct btrfs_key key
;
2311 struct btrfs_root
*root
= BTRFS_I(dentry
->d_parent
->d_inode
)->root
;
2312 struct inode
*parent_inode
= dentry
->d_parent
->d_inode
;
2314 key
.objectid
= inode
->i_ino
;
2315 btrfs_set_key_type(&key
, BTRFS_INODE_ITEM_KEY
);
2318 ret
= btrfs_insert_dir_item(trans
, root
,
2319 dentry
->d_name
.name
, dentry
->d_name
.len
,
2320 dentry
->d_parent
->d_inode
->i_ino
,
2321 &key
, btrfs_inode_type(inode
),
2322 BTRFS_I(inode
)->index
);
2325 ret
= btrfs_insert_inode_ref(trans
, root
,
2326 dentry
->d_name
.name
,
2329 parent_inode
->i_ino
,
2330 BTRFS_I(inode
)->index
);
2332 btrfs_i_size_write(parent_inode
, parent_inode
->i_size
+
2333 dentry
->d_name
.len
* 2);
2334 parent_inode
->i_mtime
= parent_inode
->i_ctime
= CURRENT_TIME
;
2335 ret
= btrfs_update_inode(trans
, root
,
2336 dentry
->d_parent
->d_inode
);
2341 static int btrfs_add_nondir(struct btrfs_trans_handle
*trans
,
2342 struct dentry
*dentry
, struct inode
*inode
,
2345 int err
= btrfs_add_link(trans
, dentry
, inode
, backref
);
2347 d_instantiate(dentry
, inode
);
2355 static int btrfs_mknod(struct inode
*dir
, struct dentry
*dentry
,
2356 int mode
, dev_t rdev
)
2358 struct btrfs_trans_handle
*trans
;
2359 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2360 struct inode
*inode
= NULL
;
2364 unsigned long nr
= 0;
2366 if (!new_valid_dev(rdev
))
2369 err
= btrfs_check_free_space(root
, 1, 0);
2373 trans
= btrfs_start_transaction(root
, 1);
2374 btrfs_set_trans_block_group(trans
, dir
);
2376 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2382 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2384 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2385 BTRFS_I(dir
)->block_group
, mode
);
2386 err
= PTR_ERR(inode
);
2390 err
= btrfs_init_acl(inode
, dir
);
2396 btrfs_set_trans_block_group(trans
, inode
);
2397 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
2401 inode
->i_op
= &btrfs_special_inode_operations
;
2402 init_special_inode(inode
, inode
->i_mode
, rdev
);
2403 btrfs_update_inode(trans
, root
, inode
);
2405 dir
->i_sb
->s_dirt
= 1;
2406 btrfs_update_inode_block_group(trans
, inode
);
2407 btrfs_update_inode_block_group(trans
, dir
);
2409 nr
= trans
->blocks_used
;
2410 btrfs_end_transaction_throttle(trans
, root
);
2413 inode_dec_link_count(inode
);
2416 btrfs_btree_balance_dirty(root
, nr
);
2420 static int btrfs_create(struct inode
*dir
, struct dentry
*dentry
,
2421 int mode
, struct nameidata
*nd
)
2423 struct btrfs_trans_handle
*trans
;
2424 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2425 struct inode
*inode
= NULL
;
2428 unsigned long nr
= 0;
2431 err
= btrfs_check_free_space(root
, 1, 0);
2434 trans
= btrfs_start_transaction(root
, 1);
2435 btrfs_set_trans_block_group(trans
, dir
);
2437 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2443 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2445 dentry
->d_parent
->d_inode
->i_ino
,
2446 objectid
, BTRFS_I(dir
)->block_group
, mode
);
2447 err
= PTR_ERR(inode
);
2451 err
= btrfs_init_acl(inode
, dir
);
2457 btrfs_set_trans_block_group(trans
, inode
);
2458 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
2462 inode
->i_mapping
->a_ops
= &btrfs_aops
;
2463 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
2464 inode
->i_fop
= &btrfs_file_operations
;
2465 inode
->i_op
= &btrfs_file_inode_operations
;
2466 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
2467 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
2468 inode
->i_mapping
, GFP_NOFS
);
2469 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
2470 inode
->i_mapping
, GFP_NOFS
);
2471 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
2472 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
2473 BTRFS_I(inode
)->delalloc_bytes
= 0;
2474 BTRFS_I(inode
)->disk_i_size
= 0;
2475 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
2476 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
2478 dir
->i_sb
->s_dirt
= 1;
2479 btrfs_update_inode_block_group(trans
, inode
);
2480 btrfs_update_inode_block_group(trans
, dir
);
2482 nr
= trans
->blocks_used
;
2483 btrfs_end_transaction_throttle(trans
, root
);
2486 inode_dec_link_count(inode
);
2489 btrfs_btree_balance_dirty(root
, nr
);
2493 static int btrfs_link(struct dentry
*old_dentry
, struct inode
*dir
,
2494 struct dentry
*dentry
)
2496 struct btrfs_trans_handle
*trans
;
2497 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2498 struct inode
*inode
= old_dentry
->d_inode
;
2499 unsigned long nr
= 0;
2503 if (inode
->i_nlink
== 0)
2506 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
2511 err
= btrfs_check_free_space(root
, 1, 0);
2514 err
= btrfs_set_inode_index(dir
, inode
);
2518 trans
= btrfs_start_transaction(root
, 1);
2520 btrfs_set_trans_block_group(trans
, dir
);
2521 atomic_inc(&inode
->i_count
);
2523 err
= btrfs_add_nondir(trans
, dentry
, inode
, 1);
2528 dir
->i_sb
->s_dirt
= 1;
2529 btrfs_update_inode_block_group(trans
, dir
);
2530 err
= btrfs_update_inode(trans
, root
, inode
);
2535 nr
= trans
->blocks_used
;
2536 btrfs_end_transaction_throttle(trans
, root
);
2539 inode_dec_link_count(inode
);
2542 btrfs_btree_balance_dirty(root
, nr
);
2546 static int btrfs_mkdir(struct inode
*dir
, struct dentry
*dentry
, int mode
)
2548 struct inode
*inode
= NULL
;
2549 struct btrfs_trans_handle
*trans
;
2550 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2552 int drop_on_err
= 0;
2554 unsigned long nr
= 1;
2556 err
= btrfs_check_free_space(root
, 1, 0);
2560 trans
= btrfs_start_transaction(root
, 1);
2561 btrfs_set_trans_block_group(trans
, dir
);
2563 if (IS_ERR(trans
)) {
2564 err
= PTR_ERR(trans
);
2568 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2574 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2576 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2577 BTRFS_I(dir
)->block_group
, S_IFDIR
| mode
);
2578 if (IS_ERR(inode
)) {
2579 err
= PTR_ERR(inode
);
2585 err
= btrfs_init_acl(inode
, dir
);
2589 inode
->i_op
= &btrfs_dir_inode_operations
;
2590 inode
->i_fop
= &btrfs_dir_file_operations
;
2591 btrfs_set_trans_block_group(trans
, inode
);
2593 btrfs_i_size_write(inode
, 0);
2594 err
= btrfs_update_inode(trans
, root
, inode
);
2598 err
= btrfs_add_link(trans
, dentry
, inode
, 0);
2602 d_instantiate(dentry
, inode
);
2604 dir
->i_sb
->s_dirt
= 1;
2605 btrfs_update_inode_block_group(trans
, inode
);
2606 btrfs_update_inode_block_group(trans
, dir
);
2609 nr
= trans
->blocks_used
;
2610 btrfs_end_transaction_throttle(trans
, root
);
2615 btrfs_btree_balance_dirty(root
, nr
);
2619 static int merge_extent_mapping(struct extent_map_tree
*em_tree
,
2620 struct extent_map
*existing
,
2621 struct extent_map
*em
,
2622 u64 map_start
, u64 map_len
)
2626 BUG_ON(map_start
< em
->start
|| map_start
>= extent_map_end(em
));
2627 start_diff
= map_start
- em
->start
;
2628 em
->start
= map_start
;
2630 if (em
->block_start
< EXTENT_MAP_LAST_BYTE
)
2631 em
->block_start
+= start_diff
;
2632 return add_extent_mapping(em_tree
, em
);
2635 struct extent_map
*btrfs_get_extent(struct inode
*inode
, struct page
*page
,
2636 size_t pg_offset
, u64 start
, u64 len
,
2642 u64 extent_start
= 0;
2644 u64 objectid
= inode
->i_ino
;
2646 struct btrfs_path
*path
= NULL
;
2647 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2648 struct btrfs_file_extent_item
*item
;
2649 struct extent_buffer
*leaf
;
2650 struct btrfs_key found_key
;
2651 struct extent_map
*em
= NULL
;
2652 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
2653 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2654 struct btrfs_trans_handle
*trans
= NULL
;
2657 spin_lock(&em_tree
->lock
);
2658 em
= lookup_extent_mapping(em_tree
, start
, len
);
2660 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2661 spin_unlock(&em_tree
->lock
);
2664 if (em
->start
> start
|| em
->start
+ em
->len
<= start
)
2665 free_extent_map(em
);
2666 else if (em
->block_start
== EXTENT_MAP_INLINE
&& page
)
2667 free_extent_map(em
);
2671 em
= alloc_extent_map(GFP_NOFS
);
2676 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2677 em
->start
= EXTENT_MAP_HOLE
;
2681 path
= btrfs_alloc_path();
2685 ret
= btrfs_lookup_file_extent(trans
, root
, path
,
2686 objectid
, start
, trans
!= NULL
);
2693 if (path
->slots
[0] == 0)
2698 leaf
= path
->nodes
[0];
2699 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
2700 struct btrfs_file_extent_item
);
2701 /* are we inside the extent that was found? */
2702 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2703 found_type
= btrfs_key_type(&found_key
);
2704 if (found_key
.objectid
!= objectid
||
2705 found_type
!= BTRFS_EXTENT_DATA_KEY
) {
2709 found_type
= btrfs_file_extent_type(leaf
, item
);
2710 extent_start
= found_key
.offset
;
2711 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
2712 extent_end
= extent_start
+
2713 btrfs_file_extent_num_bytes(leaf
, item
);
2715 if (start
< extent_start
|| start
>= extent_end
) {
2717 if (start
< extent_start
) {
2718 if (start
+ len
<= extent_start
)
2720 em
->len
= extent_end
- extent_start
;
2726 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
2728 em
->start
= extent_start
;
2729 em
->len
= extent_end
- extent_start
;
2730 em
->block_start
= EXTENT_MAP_HOLE
;
2733 bytenr
+= btrfs_file_extent_offset(leaf
, item
);
2734 em
->block_start
= bytenr
;
2735 em
->start
= extent_start
;
2736 em
->len
= extent_end
- extent_start
;
2738 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
2743 size_t extent_offset
;
2746 size
= btrfs_file_extent_inline_len(leaf
, btrfs_item_nr(leaf
,
2748 extent_end
= (extent_start
+ size
+ root
->sectorsize
- 1) &
2749 ~((u64
)root
->sectorsize
- 1);
2750 if (start
< extent_start
|| start
>= extent_end
) {
2752 if (start
< extent_start
) {
2753 if (start
+ len
<= extent_start
)
2755 em
->len
= extent_end
- extent_start
;
2761 em
->block_start
= EXTENT_MAP_INLINE
;
2764 em
->start
= extent_start
;
2769 page_start
= page_offset(page
) + pg_offset
;
2770 extent_offset
= page_start
- extent_start
;
2771 copy_size
= min_t(u64
, PAGE_CACHE_SIZE
- pg_offset
,
2772 size
- extent_offset
);
2773 em
->start
= extent_start
+ extent_offset
;
2774 em
->len
= (copy_size
+ root
->sectorsize
- 1) &
2775 ~((u64
)root
->sectorsize
- 1);
2777 ptr
= btrfs_file_extent_inline_start(item
) + extent_offset
;
2778 if (create
== 0 && !PageUptodate(page
)) {
2779 read_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2781 flush_dcache_page(page
);
2782 } else if (create
&& PageUptodate(page
)) {
2785 free_extent_map(em
);
2787 btrfs_release_path(root
, path
);
2788 trans
= btrfs_join_transaction(root
, 1);
2791 write_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2793 btrfs_mark_buffer_dirty(leaf
);
2796 set_extent_uptodate(io_tree
, em
->start
,
2797 extent_map_end(em
) - 1, GFP_NOFS
);
2800 printk("unkknown found_type %d\n", found_type
);
2807 em
->block_start
= EXTENT_MAP_HOLE
;
2809 btrfs_release_path(root
, path
);
2810 if (em
->start
> start
|| extent_map_end(em
) <= start
) {
2811 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em
->start
, em
->len
, start
, len
);
2817 spin_lock(&em_tree
->lock
);
2818 ret
= add_extent_mapping(em_tree
, em
);
2819 /* it is possible that someone inserted the extent into the tree
2820 * while we had the lock dropped. It is also possible that
2821 * an overlapping map exists in the tree
2823 if (ret
== -EEXIST
) {
2824 struct extent_map
*existing
;
2828 existing
= lookup_extent_mapping(em_tree
, start
, len
);
2829 if (existing
&& (existing
->start
> start
||
2830 existing
->start
+ existing
->len
<= start
)) {
2831 free_extent_map(existing
);
2835 existing
= lookup_extent_mapping(em_tree
, em
->start
,
2838 err
= merge_extent_mapping(em_tree
, existing
,
2841 free_extent_map(existing
);
2843 free_extent_map(em
);
2848 printk("failing to insert %Lu %Lu\n",
2850 free_extent_map(em
);
2854 free_extent_map(em
);
2859 spin_unlock(&em_tree
->lock
);
2862 btrfs_free_path(path
);
2864 ret
= btrfs_end_transaction(trans
, root
);
2870 free_extent_map(em
);
2872 return ERR_PTR(err
);
2877 #if 0 /* waiting for O_DIRECT reads */
2878 static int btrfs_get_block(struct inode
*inode
, sector_t iblock
,
2879 struct buffer_head
*bh_result
, int create
)
2881 struct extent_map
*em
;
2882 u64 start
= (u64
)iblock
<< inode
->i_blkbits
;
2883 struct btrfs_multi_bio
*multi
= NULL
;
2884 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2890 em
= btrfs_get_extent(inode
, NULL
, 0, start
, bh_result
->b_size
, 0);
2892 if (!em
|| IS_ERR(em
))
2895 if (em
->start
> start
|| em
->start
+ em
->len
<= start
) {
2899 if (em
->block_start
== EXTENT_MAP_INLINE
) {
2904 len
= em
->start
+ em
->len
- start
;
2905 len
= min_t(u64
, len
, INT_LIMIT(typeof(bh_result
->b_size
)));
2907 if (em
->block_start
== EXTENT_MAP_HOLE
||
2908 em
->block_start
== EXTENT_MAP_DELALLOC
) {
2909 bh_result
->b_size
= len
;
2913 logical
= start
- em
->start
;
2914 logical
= em
->block_start
+ logical
;
2917 ret
= btrfs_map_block(&root
->fs_info
->mapping_tree
, READ
,
2918 logical
, &map_length
, &multi
, 0);
2920 bh_result
->b_blocknr
= multi
->stripes
[0].physical
>> inode
->i_blkbits
;
2921 bh_result
->b_size
= min(map_length
, len
);
2923 bh_result
->b_bdev
= multi
->stripes
[0].dev
->bdev
;
2924 set_buffer_mapped(bh_result
);
2927 free_extent_map(em
);
2932 static ssize_t
btrfs_direct_IO(int rw
, struct kiocb
*iocb
,
2933 const struct iovec
*iov
, loff_t offset
,
2934 unsigned long nr_segs
)
2938 struct file
*file
= iocb
->ki_filp
;
2939 struct inode
*inode
= file
->f_mapping
->host
;
2944 return blockdev_direct_IO(rw
, iocb
, inode
, inode
->i_sb
->s_bdev
, iov
,
2945 offset
, nr_segs
, btrfs_get_block
, NULL
);
2949 static sector_t
btrfs_bmap(struct address_space
*mapping
, sector_t iblock
)
2951 return extent_bmap(mapping
, iblock
, btrfs_get_extent
);
2954 int btrfs_readpage(struct file
*file
, struct page
*page
)
2956 struct extent_io_tree
*tree
;
2957 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2958 return extent_read_full_page(tree
, page
, btrfs_get_extent
);
2961 static int btrfs_writepage(struct page
*page
, struct writeback_control
*wbc
)
2963 struct extent_io_tree
*tree
;
2966 if (current
->flags
& PF_MEMALLOC
) {
2967 redirty_page_for_writepage(wbc
, page
);
2971 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2972 return extent_write_full_page(tree
, page
, btrfs_get_extent
, wbc
);
2975 int btrfs_writepages(struct address_space
*mapping
,
2976 struct writeback_control
*wbc
)
2978 struct extent_io_tree
*tree
;
2979 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
2980 return extent_writepages(tree
, mapping
, btrfs_get_extent
, wbc
);
2984 btrfs_readpages(struct file
*file
, struct address_space
*mapping
,
2985 struct list_head
*pages
, unsigned nr_pages
)
2987 struct extent_io_tree
*tree
;
2988 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
2989 return extent_readpages(tree
, mapping
, pages
, nr_pages
,
2992 static int __btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
2994 struct extent_io_tree
*tree
;
2995 struct extent_map_tree
*map
;
2998 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2999 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
3000 ret
= try_release_extent_mapping(map
, tree
, page
, gfp_flags
);
3002 ClearPagePrivate(page
);
3003 set_page_private(page
, 0);
3004 page_cache_release(page
);
3009 static int btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
3011 return __btrfs_releasepage(page
, gfp_flags
);
3014 static void btrfs_invalidatepage(struct page
*page
, unsigned long offset
)
3016 struct extent_io_tree
*tree
;
3017 struct btrfs_ordered_extent
*ordered
;
3018 u64 page_start
= page_offset(page
);
3019 u64 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
3021 wait_on_page_writeback(page
);
3022 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3024 btrfs_releasepage(page
, GFP_NOFS
);
3028 lock_extent(tree
, page_start
, page_end
, GFP_NOFS
);
3029 ordered
= btrfs_lookup_ordered_extent(page
->mapping
->host
,
3033 * IO on this page will never be started, so we need
3034 * to account for any ordered extents now
3036 clear_extent_bit(tree
, page_start
, page_end
,
3037 EXTENT_DIRTY
| EXTENT_DELALLOC
|
3038 EXTENT_LOCKED
, 1, 0, GFP_NOFS
);
3039 btrfs_finish_ordered_io(page
->mapping
->host
,
3040 page_start
, page_end
);
3041 btrfs_put_ordered_extent(ordered
);
3042 lock_extent(tree
, page_start
, page_end
, GFP_NOFS
);
3044 clear_extent_bit(tree
, page_start
, page_end
,
3045 EXTENT_LOCKED
| EXTENT_DIRTY
| EXTENT_DELALLOC
|
3048 __btrfs_releasepage(page
, GFP_NOFS
);
3050 ClearPageChecked(page
);
3051 if (PagePrivate(page
)) {
3052 ClearPagePrivate(page
);
3053 set_page_private(page
, 0);
3054 page_cache_release(page
);
3059 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
3060 * called from a page fault handler when a page is first dirtied. Hence we must
3061 * be careful to check for EOF conditions here. We set the page up correctly
3062 * for a written page which means we get ENOSPC checking when writing into
3063 * holes and correct delalloc and unwritten extent mapping on filesystems that
3064 * support these features.
3066 * We are not allowed to take the i_mutex here so we have to play games to
3067 * protect against truncate races as the page could now be beyond EOF. Because
3068 * vmtruncate() writes the inode size before removing pages, once we have the
3069 * page lock we can determine safely if the page is beyond EOF. If it is not
3070 * beyond EOF, then the page is guaranteed safe against truncation until we
3073 int btrfs_page_mkwrite(struct vm_area_struct
*vma
, struct page
*page
)
3075 struct inode
*inode
= fdentry(vma
->vm_file
)->d_inode
;
3076 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3077 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
3078 struct btrfs_ordered_extent
*ordered
;
3080 unsigned long zero_start
;
3086 ret
= btrfs_check_free_space(root
, PAGE_CACHE_SIZE
, 0);
3093 size
= i_size_read(inode
);
3094 page_start
= page_offset(page
);
3095 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
3097 if ((page
->mapping
!= inode
->i_mapping
) ||
3098 (page_start
>= size
)) {
3099 /* page got truncated out from underneath us */
3102 wait_on_page_writeback(page
);
3104 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3105 set_page_extent_mapped(page
);
3108 * we can't set the delalloc bits if there are pending ordered
3109 * extents. Drop our locks and wait for them to finish
3111 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
3113 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3115 btrfs_start_ordered_extent(inode
, ordered
, 1);
3116 btrfs_put_ordered_extent(ordered
);
3120 set_extent_delalloc(&BTRFS_I(inode
)->io_tree
, page_start
,
3121 page_end
, GFP_NOFS
);
3124 /* page is wholly or partially inside EOF */
3125 if (page_start
+ PAGE_CACHE_SIZE
> size
)
3126 zero_start
= size
& ~PAGE_CACHE_MASK
;
3128 zero_start
= PAGE_CACHE_SIZE
;
3130 if (zero_start
!= PAGE_CACHE_SIZE
) {
3132 memset(kaddr
+ zero_start
, 0, PAGE_CACHE_SIZE
- zero_start
);
3133 flush_dcache_page(page
);
3136 ClearPageChecked(page
);
3137 set_page_dirty(page
);
3138 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3146 static void btrfs_truncate(struct inode
*inode
)
3148 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3150 struct btrfs_trans_handle
*trans
;
3152 u64 mask
= root
->sectorsize
- 1;
3154 if (!S_ISREG(inode
->i_mode
))
3156 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
3159 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
3160 btrfs_wait_ordered_range(inode
, inode
->i_size
& (~mask
), (u64
)-1);
3162 trans
= btrfs_start_transaction(root
, 1);
3163 btrfs_set_trans_block_group(trans
, inode
);
3164 btrfs_i_size_write(inode
, inode
->i_size
);
3166 ret
= btrfs_orphan_add(trans
, inode
);
3169 /* FIXME, add redo link to tree so we don't leak on crash */
3170 ret
= btrfs_truncate_in_trans(trans
, root
, inode
,
3171 BTRFS_EXTENT_DATA_KEY
);
3172 btrfs_update_inode(trans
, root
, inode
);
3174 ret
= btrfs_orphan_del(trans
, inode
);
3178 nr
= trans
->blocks_used
;
3179 ret
= btrfs_end_transaction_throttle(trans
, root
);
3181 btrfs_btree_balance_dirty(root
, nr
);
3185 * Invalidate a single dcache entry at the root of the filesystem.
3186 * Needed after creation of snapshot or subvolume.
3188 void btrfs_invalidate_dcache_root(struct btrfs_root
*root
, char *name
,
3191 struct dentry
*alias
, *entry
;
3194 alias
= d_find_alias(root
->fs_info
->sb
->s_root
->d_inode
);
3198 /* change me if btrfs ever gets a d_hash operation */
3199 qstr
.hash
= full_name_hash(qstr
.name
, qstr
.len
);
3200 entry
= d_lookup(alias
, &qstr
);
3203 d_invalidate(entry
);
3209 int btrfs_create_subvol_root(struct btrfs_root
*new_root
,
3210 struct btrfs_trans_handle
*trans
, u64 new_dirid
,
3211 struct btrfs_block_group_cache
*block_group
)
3213 struct inode
*inode
;
3215 inode
= btrfs_new_inode(trans
, new_root
, NULL
, "..", 2, new_dirid
,
3216 new_dirid
, block_group
, S_IFDIR
| 0700);
3218 return PTR_ERR(inode
);
3219 inode
->i_op
= &btrfs_dir_inode_operations
;
3220 inode
->i_fop
= &btrfs_dir_file_operations
;
3221 new_root
->inode
= inode
;
3224 btrfs_i_size_write(inode
, 0);
3226 return btrfs_update_inode(trans
, new_root
, inode
);
3229 unsigned long btrfs_force_ra(struct address_space
*mapping
,
3230 struct file_ra_state
*ra
, struct file
*file
,
3231 pgoff_t offset
, pgoff_t last_index
)
3233 pgoff_t req_size
= last_index
- offset
+ 1;
3235 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3236 offset
= page_cache_readahead(mapping
, ra
, file
, offset
, req_size
);
3239 page_cache_sync_readahead(mapping
, ra
, file
, offset
, req_size
);
3240 return offset
+ req_size
;
3244 struct inode
*btrfs_alloc_inode(struct super_block
*sb
)
3246 struct btrfs_inode
*ei
;
3248 ei
= kmem_cache_alloc(btrfs_inode_cachep
, GFP_NOFS
);
3252 btrfs_ordered_inode_tree_init(&ei
->ordered_tree
);
3253 ei
->i_acl
= BTRFS_ACL_NOT_CACHED
;
3254 ei
->i_default_acl
= BTRFS_ACL_NOT_CACHED
;
3255 INIT_LIST_HEAD(&ei
->i_orphan
);
3256 return &ei
->vfs_inode
;
3259 void btrfs_destroy_inode(struct inode
*inode
)
3261 struct btrfs_ordered_extent
*ordered
;
3262 WARN_ON(!list_empty(&inode
->i_dentry
));
3263 WARN_ON(inode
->i_data
.nrpages
);
3265 if (BTRFS_I(inode
)->i_acl
&&
3266 BTRFS_I(inode
)->i_acl
!= BTRFS_ACL_NOT_CACHED
)
3267 posix_acl_release(BTRFS_I(inode
)->i_acl
);
3268 if (BTRFS_I(inode
)->i_default_acl
&&
3269 BTRFS_I(inode
)->i_default_acl
!= BTRFS_ACL_NOT_CACHED
)
3270 posix_acl_release(BTRFS_I(inode
)->i_default_acl
);
3272 spin_lock(&BTRFS_I(inode
)->root
->list_lock
);
3273 if (!list_empty(&BTRFS_I(inode
)->i_orphan
)) {
3274 printk(KERN_ERR
"BTRFS: inode %lu: inode still on the orphan"
3275 " list\n", inode
->i_ino
);
3278 spin_unlock(&BTRFS_I(inode
)->root
->list_lock
);
3281 ordered
= btrfs_lookup_first_ordered_extent(inode
, (u64
)-1);
3285 printk("found ordered extent %Lu %Lu\n",
3286 ordered
->file_offset
, ordered
->len
);
3287 btrfs_remove_ordered_extent(inode
, ordered
);
3288 btrfs_put_ordered_extent(ordered
);
3289 btrfs_put_ordered_extent(ordered
);
3292 btrfs_drop_extent_cache(inode
, 0, (u64
)-1);
3293 kmem_cache_free(btrfs_inode_cachep
, BTRFS_I(inode
));
3296 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3297 static void init_once(struct kmem_cache
* cachep
, void *foo
)
3299 static void init_once(void * foo
, struct kmem_cache
* cachep
,
3300 unsigned long flags
)
3303 struct btrfs_inode
*ei
= (struct btrfs_inode
*) foo
;
3305 inode_init_once(&ei
->vfs_inode
);
3308 void btrfs_destroy_cachep(void)
3310 if (btrfs_inode_cachep
)
3311 kmem_cache_destroy(btrfs_inode_cachep
);
3312 if (btrfs_trans_handle_cachep
)
3313 kmem_cache_destroy(btrfs_trans_handle_cachep
);
3314 if (btrfs_transaction_cachep
)
3315 kmem_cache_destroy(btrfs_transaction_cachep
);
3316 if (btrfs_bit_radix_cachep
)
3317 kmem_cache_destroy(btrfs_bit_radix_cachep
);
3318 if (btrfs_path_cachep
)
3319 kmem_cache_destroy(btrfs_path_cachep
);
3322 struct kmem_cache
*btrfs_cache_create(const char *name
, size_t size
,
3323 unsigned long extra_flags
,
3324 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3325 void (*ctor
)(struct kmem_cache
*, void *)
3327 void (*ctor
)(void *, struct kmem_cache
*,
3332 return kmem_cache_create(name
, size
, 0, (SLAB_RECLAIM_ACCOUNT
|
3333 SLAB_MEM_SPREAD
| extra_flags
), ctor
3334 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3340 int btrfs_init_cachep(void)
3342 btrfs_inode_cachep
= btrfs_cache_create("btrfs_inode_cache",
3343 sizeof(struct btrfs_inode
),
3345 if (!btrfs_inode_cachep
)
3347 btrfs_trans_handle_cachep
=
3348 btrfs_cache_create("btrfs_trans_handle_cache",
3349 sizeof(struct btrfs_trans_handle
),
3351 if (!btrfs_trans_handle_cachep
)
3353 btrfs_transaction_cachep
= btrfs_cache_create("btrfs_transaction_cache",
3354 sizeof(struct btrfs_transaction
),
3356 if (!btrfs_transaction_cachep
)
3358 btrfs_path_cachep
= btrfs_cache_create("btrfs_path_cache",
3359 sizeof(struct btrfs_path
),
3361 if (!btrfs_path_cachep
)
3363 btrfs_bit_radix_cachep
= btrfs_cache_create("btrfs_radix", 256,
3364 SLAB_DESTROY_BY_RCU
, NULL
);
3365 if (!btrfs_bit_radix_cachep
)
3369 btrfs_destroy_cachep();
3373 static int btrfs_getattr(struct vfsmount
*mnt
,
3374 struct dentry
*dentry
, struct kstat
*stat
)
3376 struct inode
*inode
= dentry
->d_inode
;
3377 generic_fillattr(inode
, stat
);
3378 stat
->blksize
= PAGE_CACHE_SIZE
;
3379 stat
->blocks
= inode
->i_blocks
+ (BTRFS_I(inode
)->delalloc_bytes
>> 9);
3383 static int btrfs_rename(struct inode
* old_dir
, struct dentry
*old_dentry
,
3384 struct inode
* new_dir
,struct dentry
*new_dentry
)
3386 struct btrfs_trans_handle
*trans
;
3387 struct btrfs_root
*root
= BTRFS_I(old_dir
)->root
;
3388 struct inode
*new_inode
= new_dentry
->d_inode
;
3389 struct inode
*old_inode
= old_dentry
->d_inode
;
3390 struct timespec ctime
= CURRENT_TIME
;
3393 if (S_ISDIR(old_inode
->i_mode
) && new_inode
&&
3394 new_inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
3398 ret
= btrfs_check_free_space(root
, 1, 0);
3402 trans
= btrfs_start_transaction(root
, 1);
3404 btrfs_set_trans_block_group(trans
, new_dir
);
3406 old_dentry
->d_inode
->i_nlink
++;
3407 old_dir
->i_ctime
= old_dir
->i_mtime
= ctime
;
3408 new_dir
->i_ctime
= new_dir
->i_mtime
= ctime
;
3409 old_inode
->i_ctime
= ctime
;
3411 ret
= btrfs_unlink_trans(trans
, root
, old_dir
, old_dentry
);
3416 new_inode
->i_ctime
= CURRENT_TIME
;
3417 ret
= btrfs_unlink_trans(trans
, root
, new_dir
, new_dentry
);
3420 if (new_inode
->i_nlink
== 0) {
3421 ret
= btrfs_orphan_add(trans
, new_inode
);
3426 ret
= btrfs_set_inode_index(new_dir
, old_inode
);
3430 ret
= btrfs_add_link(trans
, new_dentry
, old_inode
, 1);
3435 btrfs_end_transaction_throttle(trans
, root
);
3440 static int btrfs_symlink(struct inode
*dir
, struct dentry
*dentry
,
3441 const char *symname
)
3443 struct btrfs_trans_handle
*trans
;
3444 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
3445 struct btrfs_path
*path
;
3446 struct btrfs_key key
;
3447 struct inode
*inode
= NULL
;
3454 struct btrfs_file_extent_item
*ei
;
3455 struct extent_buffer
*leaf
;
3456 unsigned long nr
= 0;
3458 name_len
= strlen(symname
) + 1;
3459 if (name_len
> BTRFS_MAX_INLINE_DATA_SIZE(root
))
3460 return -ENAMETOOLONG
;
3462 err
= btrfs_check_free_space(root
, 1, 0);
3466 trans
= btrfs_start_transaction(root
, 1);
3467 btrfs_set_trans_block_group(trans
, dir
);
3469 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
3475 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
3477 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
3478 BTRFS_I(dir
)->block_group
, S_IFLNK
|S_IRWXUGO
);
3479 err
= PTR_ERR(inode
);
3483 err
= btrfs_init_acl(inode
, dir
);
3489 btrfs_set_trans_block_group(trans
, inode
);
3490 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
3494 inode
->i_mapping
->a_ops
= &btrfs_aops
;
3495 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3496 inode
->i_fop
= &btrfs_file_operations
;
3497 inode
->i_op
= &btrfs_file_inode_operations
;
3498 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
3499 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
3500 inode
->i_mapping
, GFP_NOFS
);
3501 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
3502 inode
->i_mapping
, GFP_NOFS
);
3503 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
3504 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
3505 BTRFS_I(inode
)->delalloc_bytes
= 0;
3506 BTRFS_I(inode
)->disk_i_size
= 0;
3507 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
3508 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
3510 dir
->i_sb
->s_dirt
= 1;
3511 btrfs_update_inode_block_group(trans
, inode
);
3512 btrfs_update_inode_block_group(trans
, dir
);
3516 path
= btrfs_alloc_path();
3518 key
.objectid
= inode
->i_ino
;
3520 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
3521 datasize
= btrfs_file_extent_calc_inline_size(name_len
);
3522 err
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
3528 leaf
= path
->nodes
[0];
3529 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
3530 struct btrfs_file_extent_item
);
3531 btrfs_set_file_extent_generation(leaf
, ei
, trans
->transid
);
3532 btrfs_set_file_extent_type(leaf
, ei
,
3533 BTRFS_FILE_EXTENT_INLINE
);
3534 ptr
= btrfs_file_extent_inline_start(ei
);
3535 write_extent_buffer(leaf
, symname
, ptr
, name_len
);
3536 btrfs_mark_buffer_dirty(leaf
);
3537 btrfs_free_path(path
);
3539 inode
->i_op
= &btrfs_symlink_inode_operations
;
3540 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
3541 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3542 btrfs_i_size_write(inode
, name_len
- 1);
3543 err
= btrfs_update_inode(trans
, root
, inode
);
3548 nr
= trans
->blocks_used
;
3549 btrfs_end_transaction_throttle(trans
, root
);
3552 inode_dec_link_count(inode
);
3555 btrfs_btree_balance_dirty(root
, nr
);
3559 static int btrfs_set_page_dirty(struct page
*page
)
3561 return __set_page_dirty_nobuffers(page
);
3564 static int btrfs_permission(struct inode
*inode
, int mask
,
3565 struct nameidata
*nd
)
3567 if (btrfs_test_flag(inode
, READONLY
) && (mask
& MAY_WRITE
))
3569 return generic_permission(inode
, mask
, btrfs_check_acl
);
3572 static struct inode_operations btrfs_dir_inode_operations
= {
3573 .lookup
= btrfs_lookup
,
3574 .create
= btrfs_create
,
3575 .unlink
= btrfs_unlink
,
3577 .mkdir
= btrfs_mkdir
,
3578 .rmdir
= btrfs_rmdir
,
3579 .rename
= btrfs_rename
,
3580 .symlink
= btrfs_symlink
,
3581 .setattr
= btrfs_setattr
,
3582 .mknod
= btrfs_mknod
,
3583 .setxattr
= generic_setxattr
,
3584 .getxattr
= generic_getxattr
,
3585 .listxattr
= btrfs_listxattr
,
3586 .removexattr
= generic_removexattr
,
3587 .permission
= btrfs_permission
,
3589 static struct inode_operations btrfs_dir_ro_inode_operations
= {
3590 .lookup
= btrfs_lookup
,
3591 .permission
= btrfs_permission
,
3593 static struct file_operations btrfs_dir_file_operations
= {
3594 .llseek
= generic_file_llseek
,
3595 .read
= generic_read_dir
,
3596 .readdir
= btrfs_readdir
,
3597 .unlocked_ioctl
= btrfs_ioctl
,
3598 #ifdef CONFIG_COMPAT
3599 .compat_ioctl
= btrfs_ioctl
,
3601 .release
= btrfs_release_file
,
3604 static struct extent_io_ops btrfs_extent_io_ops
= {
3605 .fill_delalloc
= run_delalloc_range
,
3606 .submit_bio_hook
= btrfs_submit_bio_hook
,
3607 .merge_bio_hook
= btrfs_merge_bio_hook
,
3608 .readpage_io_hook
= btrfs_readpage_io_hook
,
3609 .readpage_end_io_hook
= btrfs_readpage_end_io_hook
,
3610 .writepage_end_io_hook
= btrfs_writepage_end_io_hook
,
3611 .writepage_start_hook
= btrfs_writepage_start_hook
,
3612 .readpage_io_failed_hook
= btrfs_io_failed_hook
,
3613 .set_bit_hook
= btrfs_set_bit_hook
,
3614 .clear_bit_hook
= btrfs_clear_bit_hook
,
3617 static struct address_space_operations btrfs_aops
= {
3618 .readpage
= btrfs_readpage
,
3619 .writepage
= btrfs_writepage
,
3620 .writepages
= btrfs_writepages
,
3621 .readpages
= btrfs_readpages
,
3622 .sync_page
= block_sync_page
,
3624 .direct_IO
= btrfs_direct_IO
,
3625 .invalidatepage
= btrfs_invalidatepage
,
3626 .releasepage
= btrfs_releasepage
,
3627 .set_page_dirty
= btrfs_set_page_dirty
,
3630 static struct address_space_operations btrfs_symlink_aops
= {
3631 .readpage
= btrfs_readpage
,
3632 .writepage
= btrfs_writepage
,
3633 .invalidatepage
= btrfs_invalidatepage
,
3634 .releasepage
= btrfs_releasepage
,
3637 static struct inode_operations btrfs_file_inode_operations
= {
3638 .truncate
= btrfs_truncate
,
3639 .getattr
= btrfs_getattr
,
3640 .setattr
= btrfs_setattr
,
3641 .setxattr
= generic_setxattr
,
3642 .getxattr
= generic_getxattr
,
3643 .listxattr
= btrfs_listxattr
,
3644 .removexattr
= generic_removexattr
,
3645 .permission
= btrfs_permission
,
3647 static struct inode_operations btrfs_special_inode_operations
= {
3648 .getattr
= btrfs_getattr
,
3649 .setattr
= btrfs_setattr
,
3650 .permission
= btrfs_permission
,
3651 .setxattr
= generic_setxattr
,
3652 .getxattr
= generic_getxattr
,
3653 .listxattr
= btrfs_listxattr
,
3654 .removexattr
= generic_removexattr
,
3656 static struct inode_operations btrfs_symlink_inode_operations
= {
3657 .readlink
= generic_readlink
,
3658 .follow_link
= page_follow_link_light
,
3659 .put_link
= page_put_link
,
3660 .permission
= btrfs_permission
,