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"
52 struct btrfs_iget_args
{
54 struct btrfs_root
*root
;
57 static struct inode_operations btrfs_dir_inode_operations
;
58 static struct inode_operations btrfs_symlink_inode_operations
;
59 static struct inode_operations btrfs_dir_ro_inode_operations
;
60 static struct inode_operations btrfs_special_inode_operations
;
61 static struct inode_operations btrfs_file_inode_operations
;
62 static struct address_space_operations btrfs_aops
;
63 static struct address_space_operations btrfs_symlink_aops
;
64 static struct file_operations btrfs_dir_file_operations
;
65 static struct extent_io_ops btrfs_extent_io_ops
;
67 static struct kmem_cache
*btrfs_inode_cachep
;
68 struct kmem_cache
*btrfs_trans_handle_cachep
;
69 struct kmem_cache
*btrfs_transaction_cachep
;
70 struct kmem_cache
*btrfs_bit_radix_cachep
;
71 struct kmem_cache
*btrfs_path_cachep
;
74 static unsigned char btrfs_type_by_mode
[S_IFMT
>> S_SHIFT
] = {
75 [S_IFREG
>> S_SHIFT
] = BTRFS_FT_REG_FILE
,
76 [S_IFDIR
>> S_SHIFT
] = BTRFS_FT_DIR
,
77 [S_IFCHR
>> S_SHIFT
] = BTRFS_FT_CHRDEV
,
78 [S_IFBLK
>> S_SHIFT
] = BTRFS_FT_BLKDEV
,
79 [S_IFIFO
>> S_SHIFT
] = BTRFS_FT_FIFO
,
80 [S_IFSOCK
>> S_SHIFT
] = BTRFS_FT_SOCK
,
81 [S_IFLNK
>> S_SHIFT
] = BTRFS_FT_SYMLINK
,
84 static void btrfs_truncate(struct inode
*inode
);
86 int btrfs_check_free_space(struct btrfs_root
*root
, u64 num_required
,
95 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
96 total
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
97 used
= btrfs_super_bytes_used(&root
->fs_info
->super_copy
);
105 if (used
+ root
->fs_info
->delalloc_bytes
+ num_required
> thresh
)
107 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
111 static int cow_file_range(struct inode
*inode
, u64 start
, u64 end
)
113 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
114 struct btrfs_trans_handle
*trans
;
118 u64 blocksize
= root
->sectorsize
;
120 struct btrfs_key ins
;
121 struct extent_map
*em
;
122 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
125 trans
= btrfs_join_transaction(root
, 1);
127 btrfs_set_trans_block_group(trans
, inode
);
129 num_bytes
= (end
- start
+ blocksize
) & ~(blocksize
- 1);
130 num_bytes
= max(blocksize
, num_bytes
);
131 orig_num_bytes
= num_bytes
;
133 if (alloc_hint
== EXTENT_MAP_INLINE
)
136 BUG_ON(num_bytes
> btrfs_super_total_bytes(&root
->fs_info
->super_copy
));
137 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
138 btrfs_drop_extent_cache(inode
, start
, start
+ num_bytes
- 1, 0);
139 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
141 while(num_bytes
> 0) {
142 cur_alloc_size
= min(num_bytes
, root
->fs_info
->max_extent
);
143 ret
= btrfs_reserve_extent(trans
, root
, cur_alloc_size
,
144 root
->sectorsize
, 0, alloc_hint
,
150 em
= alloc_extent_map(GFP_NOFS
);
152 em
->len
= ins
.offset
;
153 em
->block_start
= ins
.objectid
;
154 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
155 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
156 set_bit(EXTENT_FLAG_PINNED
, &em
->flags
);
158 spin_lock(&em_tree
->lock
);
159 ret
= add_extent_mapping(em_tree
, em
);
160 spin_unlock(&em_tree
->lock
);
161 if (ret
!= -EEXIST
) {
165 btrfs_drop_extent_cache(inode
, start
,
166 start
+ ins
.offset
- 1, 0);
168 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
170 cur_alloc_size
= ins
.offset
;
171 ret
= btrfs_add_ordered_extent(inode
, start
, ins
.objectid
,
174 if (num_bytes
< cur_alloc_size
) {
175 printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes
,
179 num_bytes
-= cur_alloc_size
;
180 alloc_hint
= ins
.objectid
+ ins
.offset
;
181 start
+= cur_alloc_size
;
184 btrfs_end_transaction(trans
, root
);
188 static int run_delalloc_nocow(struct inode
*inode
, u64 start
, u64 end
)
195 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
196 struct btrfs_block_group_cache
*block_group
;
197 struct btrfs_trans_handle
*trans
;
198 struct extent_buffer
*leaf
;
200 struct btrfs_path
*path
;
201 struct btrfs_file_extent_item
*item
;
204 struct btrfs_key found_key
;
206 total_fs_bytes
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
207 path
= btrfs_alloc_path();
209 trans
= btrfs_join_transaction(root
, 1);
212 ret
= btrfs_lookup_file_extent(NULL
, root
, path
,
213 inode
->i_ino
, start
, 0);
220 if (path
->slots
[0] == 0)
225 leaf
= path
->nodes
[0];
226 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
227 struct btrfs_file_extent_item
);
229 /* are we inside the extent that was found? */
230 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
231 found_type
= btrfs_key_type(&found_key
);
232 if (found_key
.objectid
!= inode
->i_ino
||
233 found_type
!= BTRFS_EXTENT_DATA_KEY
)
236 found_type
= btrfs_file_extent_type(leaf
, item
);
237 extent_start
= found_key
.offset
;
238 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
239 u64 extent_num_bytes
;
241 extent_num_bytes
= btrfs_file_extent_num_bytes(leaf
, item
);
242 extent_end
= extent_start
+ extent_num_bytes
;
245 if (loops
&& start
!= extent_start
)
248 if (start
< extent_start
|| start
>= extent_end
)
251 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
255 if (btrfs_cross_ref_exists(trans
, root
, &found_key
, bytenr
))
258 * we may be called by the resizer, make sure we're inside
259 * the limits of the FS
261 block_group
= btrfs_lookup_block_group(root
->fs_info
,
263 if (!block_group
|| block_group
->ro
)
266 bytenr
+= btrfs_file_extent_offset(leaf
, item
);
267 extent_num_bytes
= min(end
+ 1, extent_end
) - start
;
268 ret
= btrfs_add_ordered_extent(inode
, start
, bytenr
,
269 extent_num_bytes
, 1);
275 btrfs_release_path(root
, path
);
283 btrfs_end_transaction(trans
, root
);
284 btrfs_free_path(path
);
285 return cow_file_range(inode
, start
, end
);
289 btrfs_end_transaction(trans
, root
);
290 btrfs_free_path(path
);
294 static int run_delalloc_range(struct inode
*inode
, u64 start
, u64 end
)
296 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
299 if (btrfs_test_opt(root
, NODATACOW
) ||
300 btrfs_test_flag(inode
, NODATACOW
))
301 ret
= run_delalloc_nocow(inode
, start
, end
);
303 ret
= cow_file_range(inode
, start
, end
);
308 int btrfs_set_bit_hook(struct inode
*inode
, u64 start
, u64 end
,
309 unsigned long old
, unsigned long bits
)
312 if (!(old
& EXTENT_DELALLOC
) && (bits
& EXTENT_DELALLOC
)) {
313 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
314 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
315 BTRFS_I(inode
)->delalloc_bytes
+= end
- start
+ 1;
316 root
->fs_info
->delalloc_bytes
+= end
- start
+ 1;
317 if (list_empty(&BTRFS_I(inode
)->delalloc_inodes
)) {
318 list_add_tail(&BTRFS_I(inode
)->delalloc_inodes
,
319 &root
->fs_info
->delalloc_inodes
);
321 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
326 int btrfs_clear_bit_hook(struct inode
*inode
, u64 start
, u64 end
,
327 unsigned long old
, unsigned long bits
)
329 if ((old
& EXTENT_DELALLOC
) && (bits
& EXTENT_DELALLOC
)) {
330 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
333 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
334 if (end
- start
+ 1 > root
->fs_info
->delalloc_bytes
) {
335 printk("warning: delalloc account %Lu %Lu\n",
336 end
- start
+ 1, root
->fs_info
->delalloc_bytes
);
337 root
->fs_info
->delalloc_bytes
= 0;
338 BTRFS_I(inode
)->delalloc_bytes
= 0;
340 root
->fs_info
->delalloc_bytes
-= end
- start
+ 1;
341 BTRFS_I(inode
)->delalloc_bytes
-= end
- start
+ 1;
343 if (BTRFS_I(inode
)->delalloc_bytes
== 0 &&
344 !list_empty(&BTRFS_I(inode
)->delalloc_inodes
)) {
345 list_del_init(&BTRFS_I(inode
)->delalloc_inodes
);
347 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
352 int btrfs_merge_bio_hook(struct page
*page
, unsigned long offset
,
353 size_t size
, struct bio
*bio
)
355 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
356 struct btrfs_mapping_tree
*map_tree
;
357 u64 logical
= bio
->bi_sector
<< 9;
362 length
= bio
->bi_size
;
363 map_tree
= &root
->fs_info
->mapping_tree
;
365 ret
= btrfs_map_block(map_tree
, READ
, logical
,
366 &map_length
, NULL
, 0);
368 if (map_length
< length
+ size
) {
374 int __btrfs_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
377 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
380 ret
= btrfs_csum_one_bio(root
, inode
, bio
);
383 return btrfs_map_bio(root
, rw
, bio
, mirror_num
, 1);
386 int btrfs_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
389 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
392 ret
= btrfs_bio_wq_end_io(root
->fs_info
, bio
, 0);
395 if (btrfs_test_opt(root
, NODATASUM
) ||
396 btrfs_test_flag(inode
, NODATASUM
)) {
400 if (!(rw
& (1 << BIO_RW
))) {
401 btrfs_lookup_bio_sums(root
, inode
, bio
);
404 return btrfs_wq_submit_bio(BTRFS_I(inode
)->root
->fs_info
,
405 inode
, rw
, bio
, mirror_num
,
406 __btrfs_submit_bio_hook
);
408 return btrfs_map_bio(root
, rw
, bio
, mirror_num
, 0);
411 static noinline
int add_pending_csums(struct btrfs_trans_handle
*trans
,
412 struct inode
*inode
, u64 file_offset
,
413 struct list_head
*list
)
415 struct list_head
*cur
;
416 struct btrfs_ordered_sum
*sum
;
418 btrfs_set_trans_block_group(trans
, inode
);
419 list_for_each(cur
, list
) {
420 sum
= list_entry(cur
, struct btrfs_ordered_sum
, list
);
421 btrfs_csum_file_blocks(trans
, BTRFS_I(inode
)->root
,
427 int btrfs_set_extent_delalloc(struct inode
*inode
, u64 start
, u64 end
)
429 return set_extent_delalloc(&BTRFS_I(inode
)->io_tree
, start
, end
,
433 struct btrfs_writepage_fixup
{
435 struct btrfs_work work
;
438 /* see btrfs_writepage_start_hook for details on why this is required */
439 void btrfs_writepage_fixup_worker(struct btrfs_work
*work
)
441 struct btrfs_writepage_fixup
*fixup
;
442 struct btrfs_ordered_extent
*ordered
;
448 fixup
= container_of(work
, struct btrfs_writepage_fixup
, work
);
452 if (!page
->mapping
|| !PageDirty(page
) || !PageChecked(page
)) {
453 ClearPageChecked(page
);
457 inode
= page
->mapping
->host
;
458 page_start
= page_offset(page
);
459 page_end
= page_offset(page
) + PAGE_CACHE_SIZE
- 1;
461 lock_extent(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
, GFP_NOFS
);
463 /* already ordered? We're done */
464 if (test_range_bit(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
,
465 EXTENT_ORDERED
, 0)) {
469 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
471 unlock_extent(&BTRFS_I(inode
)->io_tree
, page_start
,
474 btrfs_start_ordered_extent(inode
, ordered
, 1);
478 btrfs_set_extent_delalloc(inode
, page_start
, page_end
);
479 ClearPageChecked(page
);
481 unlock_extent(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
, GFP_NOFS
);
484 page_cache_release(page
);
488 * There are a few paths in the higher layers of the kernel that directly
489 * set the page dirty bit without asking the filesystem if it is a
490 * good idea. This causes problems because we want to make sure COW
491 * properly happens and the data=ordered rules are followed.
493 * In our case any range that doesn't have the EXTENT_ORDERED bit set
494 * hasn't been properly setup for IO. We kick off an async process
495 * to fix it up. The async helper will wait for ordered extents, set
496 * the delalloc bit and make it safe to write the page.
498 int btrfs_writepage_start_hook(struct page
*page
, u64 start
, u64 end
)
500 struct inode
*inode
= page
->mapping
->host
;
501 struct btrfs_writepage_fixup
*fixup
;
502 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
505 ret
= test_range_bit(&BTRFS_I(inode
)->io_tree
, start
, end
,
510 if (PageChecked(page
))
513 fixup
= kzalloc(sizeof(*fixup
), GFP_NOFS
);
517 SetPageChecked(page
);
518 page_cache_get(page
);
519 fixup
->work
.func
= btrfs_writepage_fixup_worker
;
521 btrfs_queue_worker(&root
->fs_info
->fixup_workers
, &fixup
->work
);
525 static int btrfs_finish_ordered_io(struct inode
*inode
, u64 start
, u64 end
)
527 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
528 struct btrfs_trans_handle
*trans
;
529 struct btrfs_ordered_extent
*ordered_extent
;
530 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
531 struct btrfs_file_extent_item
*extent_item
;
532 struct btrfs_path
*path
= NULL
;
533 struct extent_buffer
*leaf
;
535 struct list_head list
;
536 struct btrfs_key ins
;
539 ret
= btrfs_dec_test_ordered_pending(inode
, start
, end
- start
+ 1);
543 trans
= btrfs_join_transaction(root
, 1);
545 ordered_extent
= btrfs_lookup_ordered_extent(inode
, start
);
546 BUG_ON(!ordered_extent
);
547 if (test_bit(BTRFS_ORDERED_NOCOW
, &ordered_extent
->flags
))
550 path
= btrfs_alloc_path();
553 lock_extent(io_tree
, ordered_extent
->file_offset
,
554 ordered_extent
->file_offset
+ ordered_extent
->len
- 1,
557 INIT_LIST_HEAD(&list
);
559 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
561 ret
= btrfs_drop_extents(trans
, root
, inode
,
562 ordered_extent
->file_offset
,
563 ordered_extent
->file_offset
+
565 ordered_extent
->file_offset
, &alloc_hint
);
568 ins
.objectid
= inode
->i_ino
;
569 ins
.offset
= ordered_extent
->file_offset
;
570 ins
.type
= BTRFS_EXTENT_DATA_KEY
;
571 ret
= btrfs_insert_empty_item(trans
, root
, path
, &ins
,
572 sizeof(*extent_item
));
574 leaf
= path
->nodes
[0];
575 extent_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
576 struct btrfs_file_extent_item
);
577 btrfs_set_file_extent_generation(leaf
, extent_item
, trans
->transid
);
578 btrfs_set_file_extent_type(leaf
, extent_item
, BTRFS_FILE_EXTENT_REG
);
579 btrfs_set_file_extent_disk_bytenr(leaf
, extent_item
,
580 ordered_extent
->start
);
581 btrfs_set_file_extent_disk_num_bytes(leaf
, extent_item
,
582 ordered_extent
->len
);
583 btrfs_set_file_extent_offset(leaf
, extent_item
, 0);
584 btrfs_set_file_extent_num_bytes(leaf
, extent_item
,
585 ordered_extent
->len
);
586 btrfs_mark_buffer_dirty(leaf
);
588 btrfs_drop_extent_cache(inode
, ordered_extent
->file_offset
,
589 ordered_extent
->file_offset
+
590 ordered_extent
->len
- 1, 0);
591 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
593 ins
.objectid
= ordered_extent
->start
;
594 ins
.offset
= ordered_extent
->len
;
595 ins
.type
= BTRFS_EXTENT_ITEM_KEY
;
596 ret
= btrfs_alloc_reserved_extent(trans
, root
, leaf
->start
,
597 root
->root_key
.objectid
,
598 trans
->transid
, inode
->i_ino
,
599 ordered_extent
->file_offset
, &ins
);
601 btrfs_release_path(root
, path
);
603 inode
->i_blocks
+= ordered_extent
->len
>> 9;
604 unlock_extent(io_tree
, ordered_extent
->file_offset
,
605 ordered_extent
->file_offset
+ ordered_extent
->len
- 1,
608 add_pending_csums(trans
, inode
, ordered_extent
->file_offset
,
609 &ordered_extent
->list
);
611 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
612 btrfs_ordered_update_i_size(inode
, ordered_extent
);
613 btrfs_update_inode(trans
, root
, inode
);
614 btrfs_remove_ordered_extent(inode
, ordered_extent
);
615 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
618 btrfs_put_ordered_extent(ordered_extent
);
619 /* once for the tree */
620 btrfs_put_ordered_extent(ordered_extent
);
622 btrfs_end_transaction(trans
, root
);
624 btrfs_free_path(path
);
628 int btrfs_writepage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
629 struct extent_state
*state
, int uptodate
)
631 return btrfs_finish_ordered_io(page
->mapping
->host
, start
, end
);
634 struct io_failure_record
{
642 int btrfs_io_failed_hook(struct bio
*failed_bio
,
643 struct page
*page
, u64 start
, u64 end
,
644 struct extent_state
*state
)
646 struct io_failure_record
*failrec
= NULL
;
648 struct extent_map
*em
;
649 struct inode
*inode
= page
->mapping
->host
;
650 struct extent_io_tree
*failure_tree
= &BTRFS_I(inode
)->io_failure_tree
;
651 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
658 ret
= get_state_private(failure_tree
, start
, &private);
660 failrec
= kmalloc(sizeof(*failrec
), GFP_NOFS
);
663 failrec
->start
= start
;
664 failrec
->len
= end
- start
+ 1;
665 failrec
->last_mirror
= 0;
667 spin_lock(&em_tree
->lock
);
668 em
= lookup_extent_mapping(em_tree
, start
, failrec
->len
);
669 if (em
->start
> start
|| em
->start
+ em
->len
< start
) {
673 spin_unlock(&em_tree
->lock
);
675 if (!em
|| IS_ERR(em
)) {
679 logical
= start
- em
->start
;
680 logical
= em
->block_start
+ logical
;
681 failrec
->logical
= logical
;
683 set_extent_bits(failure_tree
, start
, end
, EXTENT_LOCKED
|
684 EXTENT_DIRTY
, GFP_NOFS
);
685 set_state_private(failure_tree
, start
,
686 (u64
)(unsigned long)failrec
);
688 failrec
= (struct io_failure_record
*)(unsigned long)private;
690 num_copies
= btrfs_num_copies(
691 &BTRFS_I(inode
)->root
->fs_info
->mapping_tree
,
692 failrec
->logical
, failrec
->len
);
693 failrec
->last_mirror
++;
695 spin_lock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
696 state
= find_first_extent_bit_state(&BTRFS_I(inode
)->io_tree
,
699 if (state
&& state
->start
!= failrec
->start
)
701 spin_unlock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
703 if (!state
|| failrec
->last_mirror
> num_copies
) {
704 set_state_private(failure_tree
, failrec
->start
, 0);
705 clear_extent_bits(failure_tree
, failrec
->start
,
706 failrec
->start
+ failrec
->len
- 1,
707 EXTENT_LOCKED
| EXTENT_DIRTY
, GFP_NOFS
);
711 bio
= bio_alloc(GFP_NOFS
, 1);
712 bio
->bi_private
= state
;
713 bio
->bi_end_io
= failed_bio
->bi_end_io
;
714 bio
->bi_sector
= failrec
->logical
>> 9;
715 bio
->bi_bdev
= failed_bio
->bi_bdev
;
717 bio_add_page(bio
, page
, failrec
->len
, start
- page_offset(page
));
718 if (failed_bio
->bi_rw
& (1 << BIO_RW
))
723 BTRFS_I(inode
)->io_tree
.ops
->submit_bio_hook(inode
, rw
, bio
,
724 failrec
->last_mirror
);
728 int btrfs_clean_io_failures(struct inode
*inode
, u64 start
)
732 struct io_failure_record
*failure
;
736 if (count_range_bits(&BTRFS_I(inode
)->io_failure_tree
, &private,
737 (u64
)-1, 1, EXTENT_DIRTY
)) {
738 ret
= get_state_private(&BTRFS_I(inode
)->io_failure_tree
,
739 start
, &private_failure
);
741 failure
= (struct io_failure_record
*)(unsigned long)
743 set_state_private(&BTRFS_I(inode
)->io_failure_tree
,
745 clear_extent_bits(&BTRFS_I(inode
)->io_failure_tree
,
747 failure
->start
+ failure
->len
- 1,
748 EXTENT_DIRTY
| EXTENT_LOCKED
,
756 int btrfs_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
757 struct extent_state
*state
)
759 size_t offset
= start
- ((u64
)page
->index
<< PAGE_CACHE_SHIFT
);
760 struct inode
*inode
= page
->mapping
->host
;
761 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
763 u64
private = ~(u32
)0;
765 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
769 if (btrfs_test_opt(root
, NODATASUM
) ||
770 btrfs_test_flag(inode
, NODATASUM
))
772 if (state
&& state
->start
== start
) {
773 private = state
->private;
776 ret
= get_state_private(io_tree
, start
, &private);
778 local_irq_save(flags
);
779 kaddr
= kmap_atomic(page
, KM_IRQ0
);
783 csum
= btrfs_csum_data(root
, kaddr
+ offset
, csum
, end
- start
+ 1);
784 btrfs_csum_final(csum
, (char *)&csum
);
785 if (csum
!= private) {
788 kunmap_atomic(kaddr
, KM_IRQ0
);
789 local_irq_restore(flags
);
791 /* if the io failure tree for this inode is non-empty,
792 * check to see if we've recovered from a failed IO
794 btrfs_clean_io_failures(inode
, start
);
798 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
799 page
->mapping
->host
->i_ino
, (unsigned long long)start
, csum
,
801 memset(kaddr
+ offset
, 1, end
- start
+ 1);
802 flush_dcache_page(page
);
803 kunmap_atomic(kaddr
, KM_IRQ0
);
804 local_irq_restore(flags
);
811 * This creates an orphan entry for the given inode in case something goes
812 * wrong in the middle of an unlink/truncate.
814 int btrfs_orphan_add(struct btrfs_trans_handle
*trans
, struct inode
*inode
)
816 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
819 spin_lock(&root
->list_lock
);
821 /* already on the orphan list, we're good */
822 if (!list_empty(&BTRFS_I(inode
)->i_orphan
)) {
823 spin_unlock(&root
->list_lock
);
827 list_add(&BTRFS_I(inode
)->i_orphan
, &root
->orphan_list
);
829 spin_unlock(&root
->list_lock
);
832 * insert an orphan item to track this unlinked/truncated file
834 ret
= btrfs_insert_orphan_item(trans
, root
, inode
->i_ino
);
840 * We have done the truncate/delete so we can go ahead and remove the orphan
841 * item for this particular inode.
843 int btrfs_orphan_del(struct btrfs_trans_handle
*trans
, struct inode
*inode
)
845 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
848 spin_lock(&root
->list_lock
);
850 if (list_empty(&BTRFS_I(inode
)->i_orphan
)) {
851 spin_unlock(&root
->list_lock
);
855 list_del_init(&BTRFS_I(inode
)->i_orphan
);
857 spin_unlock(&root
->list_lock
);
861 spin_unlock(&root
->list_lock
);
863 ret
= btrfs_del_orphan_item(trans
, root
, inode
->i_ino
);
869 * this cleans up any orphans that may be left on the list from the last use
872 void btrfs_orphan_cleanup(struct btrfs_root
*root
)
874 struct btrfs_path
*path
;
875 struct extent_buffer
*leaf
;
876 struct btrfs_item
*item
;
877 struct btrfs_key key
, found_key
;
878 struct btrfs_trans_handle
*trans
;
880 int ret
= 0, nr_unlink
= 0, nr_truncate
= 0;
882 /* don't do orphan cleanup if the fs is readonly. */
883 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
)
886 path
= btrfs_alloc_path();
891 key
.objectid
= BTRFS_ORPHAN_OBJECTID
;
892 btrfs_set_key_type(&key
, BTRFS_ORPHAN_ITEM_KEY
);
893 key
.offset
= (u64
)-1;
897 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
899 printk(KERN_ERR
"Error searching slot for orphan: %d"
905 * if ret == 0 means we found what we were searching for, which
906 * is weird, but possible, so only screw with path if we didnt
907 * find the key and see if we have stuff that matches
910 if (path
->slots
[0] == 0)
915 /* pull out the item */
916 leaf
= path
->nodes
[0];
917 item
= btrfs_item_nr(leaf
, path
->slots
[0]);
918 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
920 /* make sure the item matches what we want */
921 if (found_key
.objectid
!= BTRFS_ORPHAN_OBJECTID
)
923 if (btrfs_key_type(&found_key
) != BTRFS_ORPHAN_ITEM_KEY
)
926 /* release the path since we're done with it */
927 btrfs_release_path(root
, path
);
930 * this is where we are basically btrfs_lookup, without the
931 * crossing root thing. we store the inode number in the
932 * offset of the orphan item.
934 inode
= btrfs_iget_locked(root
->fs_info
->sb
,
935 found_key
.offset
, root
);
939 if (inode
->i_state
& I_NEW
) {
940 BTRFS_I(inode
)->root
= root
;
942 /* have to set the location manually */
943 BTRFS_I(inode
)->location
.objectid
= inode
->i_ino
;
944 BTRFS_I(inode
)->location
.type
= BTRFS_INODE_ITEM_KEY
;
945 BTRFS_I(inode
)->location
.offset
= 0;
947 btrfs_read_locked_inode(inode
);
948 unlock_new_inode(inode
);
952 * add this inode to the orphan list so btrfs_orphan_del does
953 * the proper thing when we hit it
955 spin_lock(&root
->list_lock
);
956 list_add(&BTRFS_I(inode
)->i_orphan
, &root
->orphan_list
);
957 spin_unlock(&root
->list_lock
);
960 * if this is a bad inode, means we actually succeeded in
961 * removing the inode, but not the orphan record, which means
962 * we need to manually delete the orphan since iput will just
965 if (is_bad_inode(inode
)) {
966 trans
= btrfs_start_transaction(root
, 1);
967 btrfs_orphan_del(trans
, inode
);
968 btrfs_end_transaction(trans
, root
);
973 /* if we have links, this was a truncate, lets do that */
974 if (inode
->i_nlink
) {
976 btrfs_truncate(inode
);
981 /* this will do delete_inode and everything for us */
986 printk(KERN_INFO
"btrfs: unlinked %d orphans\n", nr_unlink
);
988 printk(KERN_INFO
"btrfs: truncated %d orphans\n", nr_truncate
);
990 btrfs_free_path(path
);
993 void btrfs_read_locked_inode(struct inode
*inode
)
995 struct btrfs_path
*path
;
996 struct extent_buffer
*leaf
;
997 struct btrfs_inode_item
*inode_item
;
998 struct btrfs_timespec
*tspec
;
999 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1000 struct btrfs_key location
;
1001 u64 alloc_group_block
;
1005 path
= btrfs_alloc_path();
1007 memcpy(&location
, &BTRFS_I(inode
)->location
, sizeof(location
));
1009 ret
= btrfs_lookup_inode(NULL
, root
, path
, &location
, 0);
1013 leaf
= path
->nodes
[0];
1014 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
1015 struct btrfs_inode_item
);
1017 inode
->i_mode
= btrfs_inode_mode(leaf
, inode_item
);
1018 inode
->i_nlink
= btrfs_inode_nlink(leaf
, inode_item
);
1019 inode
->i_uid
= btrfs_inode_uid(leaf
, inode_item
);
1020 inode
->i_gid
= btrfs_inode_gid(leaf
, inode_item
);
1021 btrfs_i_size_write(inode
, btrfs_inode_size(leaf
, inode_item
));
1023 tspec
= btrfs_inode_atime(inode_item
);
1024 inode
->i_atime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1025 inode
->i_atime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1027 tspec
= btrfs_inode_mtime(inode_item
);
1028 inode
->i_mtime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1029 inode
->i_mtime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1031 tspec
= btrfs_inode_ctime(inode_item
);
1032 inode
->i_ctime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1033 inode
->i_ctime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1035 inode
->i_blocks
= btrfs_inode_nblocks(leaf
, inode_item
);
1036 BTRFS_I(inode
)->generation
= btrfs_inode_generation(leaf
, inode_item
);
1037 inode
->i_generation
= BTRFS_I(inode
)->generation
;
1039 rdev
= btrfs_inode_rdev(leaf
, inode_item
);
1041 BTRFS_I(inode
)->index_cnt
= (u64
)-1;
1043 alloc_group_block
= btrfs_inode_block_group(leaf
, inode_item
);
1044 BTRFS_I(inode
)->block_group
= btrfs_lookup_block_group(root
->fs_info
,
1046 BTRFS_I(inode
)->flags
= btrfs_inode_flags(leaf
, inode_item
);
1047 if (!BTRFS_I(inode
)->block_group
) {
1048 BTRFS_I(inode
)->block_group
= btrfs_find_block_group(root
,
1050 BTRFS_BLOCK_GROUP_METADATA
, 0);
1052 btrfs_free_path(path
);
1055 switch (inode
->i_mode
& S_IFMT
) {
1057 inode
->i_mapping
->a_ops
= &btrfs_aops
;
1058 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1059 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
1060 inode
->i_fop
= &btrfs_file_operations
;
1061 inode
->i_op
= &btrfs_file_inode_operations
;
1064 inode
->i_fop
= &btrfs_dir_file_operations
;
1065 if (root
== root
->fs_info
->tree_root
)
1066 inode
->i_op
= &btrfs_dir_ro_inode_operations
;
1068 inode
->i_op
= &btrfs_dir_inode_operations
;
1071 inode
->i_op
= &btrfs_symlink_inode_operations
;
1072 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
1073 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1076 init_special_inode(inode
, inode
->i_mode
, rdev
);
1082 btrfs_free_path(path
);
1083 make_bad_inode(inode
);
1086 static void fill_inode_item(struct btrfs_trans_handle
*trans
,
1087 struct extent_buffer
*leaf
,
1088 struct btrfs_inode_item
*item
,
1089 struct inode
*inode
)
1091 btrfs_set_inode_uid(leaf
, item
, inode
->i_uid
);
1092 btrfs_set_inode_gid(leaf
, item
, inode
->i_gid
);
1093 btrfs_set_inode_size(leaf
, item
, BTRFS_I(inode
)->disk_i_size
);
1094 btrfs_set_inode_mode(leaf
, item
, inode
->i_mode
);
1095 btrfs_set_inode_nlink(leaf
, item
, inode
->i_nlink
);
1097 btrfs_set_timespec_sec(leaf
, btrfs_inode_atime(item
),
1098 inode
->i_atime
.tv_sec
);
1099 btrfs_set_timespec_nsec(leaf
, btrfs_inode_atime(item
),
1100 inode
->i_atime
.tv_nsec
);
1102 btrfs_set_timespec_sec(leaf
, btrfs_inode_mtime(item
),
1103 inode
->i_mtime
.tv_sec
);
1104 btrfs_set_timespec_nsec(leaf
, btrfs_inode_mtime(item
),
1105 inode
->i_mtime
.tv_nsec
);
1107 btrfs_set_timespec_sec(leaf
, btrfs_inode_ctime(item
),
1108 inode
->i_ctime
.tv_sec
);
1109 btrfs_set_timespec_nsec(leaf
, btrfs_inode_ctime(item
),
1110 inode
->i_ctime
.tv_nsec
);
1112 btrfs_set_inode_nblocks(leaf
, item
, inode
->i_blocks
);
1113 btrfs_set_inode_generation(leaf
, item
, BTRFS_I(inode
)->generation
);
1114 btrfs_set_inode_transid(leaf
, item
, trans
->transid
);
1115 btrfs_set_inode_rdev(leaf
, item
, inode
->i_rdev
);
1116 btrfs_set_inode_flags(leaf
, item
, BTRFS_I(inode
)->flags
);
1117 btrfs_set_inode_block_group(leaf
, item
,
1118 BTRFS_I(inode
)->block_group
->key
.objectid
);
1121 int noinline
btrfs_update_inode(struct btrfs_trans_handle
*trans
,
1122 struct btrfs_root
*root
,
1123 struct inode
*inode
)
1125 struct btrfs_inode_item
*inode_item
;
1126 struct btrfs_path
*path
;
1127 struct extent_buffer
*leaf
;
1130 path
= btrfs_alloc_path();
1132 ret
= btrfs_lookup_inode(trans
, root
, path
,
1133 &BTRFS_I(inode
)->location
, 1);
1140 leaf
= path
->nodes
[0];
1141 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
1142 struct btrfs_inode_item
);
1144 fill_inode_item(trans
, leaf
, inode_item
, inode
);
1145 btrfs_mark_buffer_dirty(leaf
);
1146 btrfs_set_inode_last_trans(trans
, inode
);
1149 btrfs_free_path(path
);
1154 int btrfs_unlink_inode(struct btrfs_trans_handle
*trans
,
1155 struct btrfs_root
*root
,
1156 struct inode
*dir
, struct inode
*inode
,
1157 const char *name
, int name_len
)
1159 struct btrfs_path
*path
;
1161 struct extent_buffer
*leaf
;
1162 struct btrfs_dir_item
*di
;
1163 struct btrfs_key key
;
1166 path
= btrfs_alloc_path();
1172 di
= btrfs_lookup_dir_item(trans
, root
, path
, dir
->i_ino
,
1173 name
, name_len
, -1);
1182 leaf
= path
->nodes
[0];
1183 btrfs_dir_item_key_to_cpu(leaf
, di
, &key
);
1184 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
1187 btrfs_release_path(root
, path
);
1189 ret
= btrfs_del_inode_ref(trans
, root
, name
, name_len
,
1191 dir
->i_ino
, &index
);
1193 printk("failed to delete reference to %.*s, "
1194 "inode %lu parent %lu\n", name_len
, name
,
1195 inode
->i_ino
, dir
->i_ino
);
1199 di
= btrfs_lookup_dir_index_item(trans
, root
, path
, dir
->i_ino
,
1200 index
, name
, name_len
, -1);
1209 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
1210 btrfs_release_path(root
, path
);
1212 ret
= btrfs_del_inode_ref_in_log(trans
, root
, name
, name_len
,
1214 BUG_ON(ret
!= 0 && ret
!= -ENOENT
);
1216 BTRFS_I(dir
)->log_dirty_trans
= trans
->transid
;
1218 ret
= btrfs_del_dir_entries_in_log(trans
, root
, name
, name_len
,
1222 btrfs_free_path(path
);
1226 btrfs_i_size_write(dir
, dir
->i_size
- name_len
* 2);
1227 inode
->i_ctime
= dir
->i_mtime
= dir
->i_ctime
= CURRENT_TIME
;
1228 btrfs_update_inode(trans
, root
, dir
);
1229 btrfs_drop_nlink(inode
);
1230 ret
= btrfs_update_inode(trans
, root
, inode
);
1231 dir
->i_sb
->s_dirt
= 1;
1236 static int btrfs_unlink(struct inode
*dir
, struct dentry
*dentry
)
1238 struct btrfs_root
*root
;
1239 struct btrfs_trans_handle
*trans
;
1240 struct inode
*inode
= dentry
->d_inode
;
1242 unsigned long nr
= 0;
1244 root
= BTRFS_I(dir
)->root
;
1246 ret
= btrfs_check_free_space(root
, 1, 1);
1250 trans
= btrfs_start_transaction(root
, 1);
1252 btrfs_set_trans_block_group(trans
, dir
);
1253 ret
= btrfs_unlink_inode(trans
, root
, dir
, dentry
->d_inode
,
1254 dentry
->d_name
.name
, dentry
->d_name
.len
);
1256 if (inode
->i_nlink
== 0)
1257 ret
= btrfs_orphan_add(trans
, inode
);
1259 nr
= trans
->blocks_used
;
1261 btrfs_end_transaction_throttle(trans
, root
);
1263 btrfs_btree_balance_dirty(root
, nr
);
1267 static int btrfs_rmdir(struct inode
*dir
, struct dentry
*dentry
)
1269 struct inode
*inode
= dentry
->d_inode
;
1272 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1273 struct btrfs_trans_handle
*trans
;
1274 unsigned long nr
= 0;
1276 if (inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
1280 ret
= btrfs_check_free_space(root
, 1, 1);
1284 trans
= btrfs_start_transaction(root
, 1);
1285 btrfs_set_trans_block_group(trans
, dir
);
1287 err
= btrfs_orphan_add(trans
, inode
);
1291 /* now the directory is empty */
1292 err
= btrfs_unlink_inode(trans
, root
, dir
, dentry
->d_inode
,
1293 dentry
->d_name
.name
, dentry
->d_name
.len
);
1295 btrfs_i_size_write(inode
, 0);
1299 nr
= trans
->blocks_used
;
1300 ret
= btrfs_end_transaction_throttle(trans
, root
);
1302 btrfs_btree_balance_dirty(root
, nr
);
1310 * this can truncate away extent items, csum items and directory items.
1311 * It starts at a high offset and removes keys until it can't find
1312 * any higher than i_size.
1314 * csum items that cross the new i_size are truncated to the new size
1317 * min_type is the minimum key type to truncate down to. If set to 0, this
1318 * will kill all the items on this inode, including the INODE_ITEM_KEY.
1320 noinline
int btrfs_truncate_inode_items(struct btrfs_trans_handle
*trans
,
1321 struct btrfs_root
*root
,
1322 struct inode
*inode
,
1323 u64 new_size
, u32 min_type
)
1326 struct btrfs_path
*path
;
1327 struct btrfs_key key
;
1328 struct btrfs_key found_key
;
1330 struct extent_buffer
*leaf
;
1331 struct btrfs_file_extent_item
*fi
;
1332 u64 extent_start
= 0;
1333 u64 extent_num_bytes
= 0;
1339 int pending_del_nr
= 0;
1340 int pending_del_slot
= 0;
1341 int extent_type
= -1;
1342 u64 mask
= root
->sectorsize
- 1;
1345 btrfs_drop_extent_cache(inode
, new_size
& (~mask
), (u64
)-1, 0);
1346 path
= btrfs_alloc_path();
1350 /* FIXME, add redo link to tree so we don't leak on crash */
1351 key
.objectid
= inode
->i_ino
;
1352 key
.offset
= (u64
)-1;
1355 btrfs_init_path(path
);
1357 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
1362 /* there are no items in the tree for us to truncate, we're
1365 if (path
->slots
[0] == 0) {
1374 leaf
= path
->nodes
[0];
1375 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
1376 found_type
= btrfs_key_type(&found_key
);
1378 if (found_key
.objectid
!= inode
->i_ino
)
1381 if (found_type
< min_type
)
1384 item_end
= found_key
.offset
;
1385 if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
1386 fi
= btrfs_item_ptr(leaf
, path
->slots
[0],
1387 struct btrfs_file_extent_item
);
1388 extent_type
= btrfs_file_extent_type(leaf
, fi
);
1389 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1391 btrfs_file_extent_num_bytes(leaf
, fi
);
1392 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1393 struct btrfs_item
*item
= btrfs_item_nr(leaf
,
1395 item_end
+= btrfs_file_extent_inline_len(leaf
,
1400 if (found_type
== BTRFS_CSUM_ITEM_KEY
) {
1401 ret
= btrfs_csum_truncate(trans
, root
, path
,
1405 if (item_end
< new_size
) {
1406 if (found_type
== BTRFS_DIR_ITEM_KEY
) {
1407 found_type
= BTRFS_INODE_ITEM_KEY
;
1408 } else if (found_type
== BTRFS_EXTENT_ITEM_KEY
) {
1409 found_type
= BTRFS_CSUM_ITEM_KEY
;
1410 } else if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
1411 found_type
= BTRFS_XATTR_ITEM_KEY
;
1412 } else if (found_type
== BTRFS_XATTR_ITEM_KEY
) {
1413 found_type
= BTRFS_INODE_REF_KEY
;
1414 } else if (found_type
) {
1419 btrfs_set_key_type(&key
, found_type
);
1422 if (found_key
.offset
>= new_size
)
1428 /* FIXME, shrink the extent if the ref count is only 1 */
1429 if (found_type
!= BTRFS_EXTENT_DATA_KEY
)
1432 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1434 extent_start
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
1436 u64 orig_num_bytes
=
1437 btrfs_file_extent_num_bytes(leaf
, fi
);
1438 extent_num_bytes
= new_size
-
1439 found_key
.offset
+ root
->sectorsize
- 1;
1440 extent_num_bytes
= extent_num_bytes
&
1441 ~((u64
)root
->sectorsize
- 1);
1442 btrfs_set_file_extent_num_bytes(leaf
, fi
,
1444 num_dec
= (orig_num_bytes
-
1446 if (root
->ref_cows
&& extent_start
!= 0)
1447 dec_i_blocks(inode
, num_dec
);
1448 btrfs_mark_buffer_dirty(leaf
);
1451 btrfs_file_extent_disk_num_bytes(leaf
,
1453 /* FIXME blocksize != 4096 */
1454 num_dec
= btrfs_file_extent_num_bytes(leaf
, fi
);
1455 if (extent_start
!= 0) {
1458 dec_i_blocks(inode
, num_dec
);
1460 root_gen
= btrfs_header_generation(leaf
);
1461 root_owner
= btrfs_header_owner(leaf
);
1463 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1465 u32 size
= new_size
- found_key
.offset
;
1467 if (root
->ref_cows
) {
1468 dec_i_blocks(inode
, item_end
+ 1 -
1469 found_key
.offset
- size
);
1472 btrfs_file_extent_calc_inline_size(size
);
1473 ret
= btrfs_truncate_item(trans
, root
, path
,
1476 } else if (root
->ref_cows
) {
1477 dec_i_blocks(inode
, item_end
+ 1 -
1483 if (!pending_del_nr
) {
1484 /* no pending yet, add ourselves */
1485 pending_del_slot
= path
->slots
[0];
1487 } else if (pending_del_nr
&&
1488 path
->slots
[0] + 1 == pending_del_slot
) {
1489 /* hop on the pending chunk */
1491 pending_del_slot
= path
->slots
[0];
1493 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path
->slots
[0], pending_del_nr
, pending_del_slot
);
1499 ret
= btrfs_free_extent(trans
, root
, extent_start
,
1501 leaf
->start
, root_owner
,
1502 root_gen
, inode
->i_ino
,
1503 found_key
.offset
, 0);
1507 if (path
->slots
[0] == 0) {
1510 btrfs_release_path(root
, path
);
1515 if (pending_del_nr
&&
1516 path
->slots
[0] + 1 != pending_del_slot
) {
1517 struct btrfs_key debug
;
1519 btrfs_item_key_to_cpu(path
->nodes
[0], &debug
,
1521 ret
= btrfs_del_items(trans
, root
, path
,
1526 btrfs_release_path(root
, path
);
1532 if (pending_del_nr
) {
1533 ret
= btrfs_del_items(trans
, root
, path
, pending_del_slot
,
1536 btrfs_free_path(path
);
1537 inode
->i_sb
->s_dirt
= 1;
1542 * taken from block_truncate_page, but does cow as it zeros out
1543 * any bytes left in the last page in the file.
1545 static int btrfs_truncate_page(struct address_space
*mapping
, loff_t from
)
1547 struct inode
*inode
= mapping
->host
;
1548 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1549 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1550 struct btrfs_ordered_extent
*ordered
;
1552 u32 blocksize
= root
->sectorsize
;
1553 pgoff_t index
= from
>> PAGE_CACHE_SHIFT
;
1554 unsigned offset
= from
& (PAGE_CACHE_SIZE
-1);
1560 if ((offset
& (blocksize
- 1)) == 0)
1565 page
= grab_cache_page(mapping
, index
);
1569 page_start
= page_offset(page
);
1570 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1572 if (!PageUptodate(page
)) {
1573 ret
= btrfs_readpage(NULL
, page
);
1575 if (page
->mapping
!= mapping
) {
1577 page_cache_release(page
);
1580 if (!PageUptodate(page
)) {
1585 wait_on_page_writeback(page
);
1587 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1588 set_page_extent_mapped(page
);
1590 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
1592 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1594 page_cache_release(page
);
1595 btrfs_start_ordered_extent(inode
, ordered
, 1);
1596 btrfs_put_ordered_extent(ordered
);
1600 btrfs_set_extent_delalloc(inode
, page_start
, page_end
);
1602 if (offset
!= PAGE_CACHE_SIZE
) {
1604 memset(kaddr
+ offset
, 0, PAGE_CACHE_SIZE
- offset
);
1605 flush_dcache_page(page
);
1608 ClearPageChecked(page
);
1609 set_page_dirty(page
);
1610 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1614 page_cache_release(page
);
1619 static int btrfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
1621 struct inode
*inode
= dentry
->d_inode
;
1624 err
= inode_change_ok(inode
, attr
);
1628 if (S_ISREG(inode
->i_mode
) &&
1629 attr
->ia_valid
& ATTR_SIZE
&& attr
->ia_size
> inode
->i_size
) {
1630 struct btrfs_trans_handle
*trans
;
1631 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1632 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1634 u64 mask
= root
->sectorsize
- 1;
1635 u64 hole_start
= (inode
->i_size
+ mask
) & ~mask
;
1636 u64 block_end
= (attr
->ia_size
+ mask
) & ~mask
;
1640 if (attr
->ia_size
<= hole_start
)
1643 err
= btrfs_check_free_space(root
, 1, 0);
1647 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
1649 hole_size
= block_end
- hole_start
;
1651 struct btrfs_ordered_extent
*ordered
;
1652 btrfs_wait_ordered_range(inode
, hole_start
, hole_size
);
1654 lock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1655 ordered
= btrfs_lookup_ordered_extent(inode
, hole_start
);
1657 unlock_extent(io_tree
, hole_start
,
1658 block_end
- 1, GFP_NOFS
);
1659 btrfs_put_ordered_extent(ordered
);
1665 trans
= btrfs_start_transaction(root
, 1);
1666 btrfs_set_trans_block_group(trans
, inode
);
1667 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
1668 err
= btrfs_drop_extents(trans
, root
, inode
,
1669 hole_start
, block_end
, hole_start
,
1672 if (alloc_hint
!= EXTENT_MAP_INLINE
) {
1673 err
= btrfs_insert_file_extent(trans
, root
,
1677 btrfs_drop_extent_cache(inode
, hole_start
,
1679 btrfs_check_file(root
, inode
);
1681 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
1682 btrfs_end_transaction(trans
, root
);
1683 unlock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1688 err
= inode_setattr(inode
, attr
);
1690 if (!err
&& ((attr
->ia_valid
& ATTR_MODE
)))
1691 err
= btrfs_acl_chmod(inode
);
1696 void btrfs_delete_inode(struct inode
*inode
)
1698 struct btrfs_trans_handle
*trans
;
1699 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1703 truncate_inode_pages(&inode
->i_data
, 0);
1704 if (is_bad_inode(inode
)) {
1705 btrfs_orphan_del(NULL
, inode
);
1708 btrfs_wait_ordered_range(inode
, 0, (u64
)-1);
1710 btrfs_i_size_write(inode
, 0);
1711 trans
= btrfs_start_transaction(root
, 1);
1713 btrfs_set_trans_block_group(trans
, inode
);
1714 ret
= btrfs_truncate_inode_items(trans
, root
, inode
, inode
->i_size
, 0);
1716 btrfs_orphan_del(NULL
, inode
);
1717 goto no_delete_lock
;
1720 btrfs_orphan_del(trans
, inode
);
1722 nr
= trans
->blocks_used
;
1725 btrfs_end_transaction(trans
, root
);
1726 btrfs_btree_balance_dirty(root
, nr
);
1730 nr
= trans
->blocks_used
;
1731 btrfs_end_transaction(trans
, root
);
1732 btrfs_btree_balance_dirty(root
, nr
);
1738 * this returns the key found in the dir entry in the location pointer.
1739 * If no dir entries were found, location->objectid is 0.
1741 static int btrfs_inode_by_name(struct inode
*dir
, struct dentry
*dentry
,
1742 struct btrfs_key
*location
)
1744 const char *name
= dentry
->d_name
.name
;
1745 int namelen
= dentry
->d_name
.len
;
1746 struct btrfs_dir_item
*di
;
1747 struct btrfs_path
*path
;
1748 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1751 path
= btrfs_alloc_path();
1754 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir
->i_ino
, name
,
1758 if (!di
|| IS_ERR(di
)) {
1761 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, location
);
1763 btrfs_free_path(path
);
1766 location
->objectid
= 0;
1771 * when we hit a tree root in a directory, the btrfs part of the inode
1772 * needs to be changed to reflect the root directory of the tree root. This
1773 * is kind of like crossing a mount point.
1775 static int fixup_tree_root_location(struct btrfs_root
*root
,
1776 struct btrfs_key
*location
,
1777 struct btrfs_root
**sub_root
,
1778 struct dentry
*dentry
)
1780 struct btrfs_root_item
*ri
;
1782 if (btrfs_key_type(location
) != BTRFS_ROOT_ITEM_KEY
)
1784 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1787 *sub_root
= btrfs_read_fs_root(root
->fs_info
, location
,
1788 dentry
->d_name
.name
,
1789 dentry
->d_name
.len
);
1790 if (IS_ERR(*sub_root
))
1791 return PTR_ERR(*sub_root
);
1793 ri
= &(*sub_root
)->root_item
;
1794 location
->objectid
= btrfs_root_dirid(ri
);
1795 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1796 location
->offset
= 0;
1801 static noinline
void init_btrfs_i(struct inode
*inode
)
1803 struct btrfs_inode
*bi
= BTRFS_I(inode
);
1806 bi
->i_default_acl
= NULL
;
1810 bi
->logged_trans
= 0;
1811 bi
->delalloc_bytes
= 0;
1812 bi
->disk_i_size
= 0;
1814 bi
->index_cnt
= (u64
)-1;
1815 bi
->log_dirty_trans
= 0;
1816 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1817 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1818 inode
->i_mapping
, GFP_NOFS
);
1819 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1820 inode
->i_mapping
, GFP_NOFS
);
1821 INIT_LIST_HEAD(&BTRFS_I(inode
)->delalloc_inodes
);
1822 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
1823 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
1824 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
1825 mutex_init(&BTRFS_I(inode
)->log_mutex
);
1828 static int btrfs_init_locked_inode(struct inode
*inode
, void *p
)
1830 struct btrfs_iget_args
*args
= p
;
1831 inode
->i_ino
= args
->ino
;
1832 init_btrfs_i(inode
);
1833 BTRFS_I(inode
)->root
= args
->root
;
1837 static int btrfs_find_actor(struct inode
*inode
, void *opaque
)
1839 struct btrfs_iget_args
*args
= opaque
;
1840 return (args
->ino
== inode
->i_ino
&&
1841 args
->root
== BTRFS_I(inode
)->root
);
1844 struct inode
*btrfs_ilookup(struct super_block
*s
, u64 objectid
,
1845 struct btrfs_root
*root
, int wait
)
1847 struct inode
*inode
;
1848 struct btrfs_iget_args args
;
1849 args
.ino
= objectid
;
1853 inode
= ilookup5(s
, objectid
, btrfs_find_actor
,
1856 inode
= ilookup5_nowait(s
, objectid
, btrfs_find_actor
,
1862 struct inode
*btrfs_iget_locked(struct super_block
*s
, u64 objectid
,
1863 struct btrfs_root
*root
)
1865 struct inode
*inode
;
1866 struct btrfs_iget_args args
;
1867 args
.ino
= objectid
;
1870 inode
= iget5_locked(s
, objectid
, btrfs_find_actor
,
1871 btrfs_init_locked_inode
,
1876 /* Get an inode object given its location and corresponding root.
1877 * Returns in *is_new if the inode was read from disk
1879 struct inode
*btrfs_iget(struct super_block
*s
, struct btrfs_key
*location
,
1880 struct btrfs_root
*root
, int *is_new
)
1882 struct inode
*inode
;
1884 inode
= btrfs_iget_locked(s
, location
->objectid
, root
);
1886 return ERR_PTR(-EACCES
);
1888 if (inode
->i_state
& I_NEW
) {
1889 BTRFS_I(inode
)->root
= root
;
1890 memcpy(&BTRFS_I(inode
)->location
, location
, sizeof(*location
));
1891 btrfs_read_locked_inode(inode
);
1892 unlock_new_inode(inode
);
1903 static struct dentry
*btrfs_lookup(struct inode
*dir
, struct dentry
*dentry
,
1904 struct nameidata
*nd
)
1906 struct inode
* inode
;
1907 struct btrfs_inode
*bi
= BTRFS_I(dir
);
1908 struct btrfs_root
*root
= bi
->root
;
1909 struct btrfs_root
*sub_root
= root
;
1910 struct btrfs_key location
;
1911 int ret
, new, do_orphan
= 0;
1913 if (dentry
->d_name
.len
> BTRFS_NAME_LEN
)
1914 return ERR_PTR(-ENAMETOOLONG
);
1916 ret
= btrfs_inode_by_name(dir
, dentry
, &location
);
1919 return ERR_PTR(ret
);
1922 if (location
.objectid
) {
1923 ret
= fixup_tree_root_location(root
, &location
, &sub_root
,
1926 return ERR_PTR(ret
);
1928 return ERR_PTR(-ENOENT
);
1929 inode
= btrfs_iget(dir
->i_sb
, &location
, sub_root
, &new);
1931 return ERR_CAST(inode
);
1933 /* the inode and parent dir are two different roots */
1934 if (new && root
!= sub_root
) {
1936 sub_root
->inode
= inode
;
1941 if (unlikely(do_orphan
))
1942 btrfs_orphan_cleanup(sub_root
);
1944 return d_splice_alias(inode
, dentry
);
1947 static unsigned char btrfs_filetype_table
[] = {
1948 DT_UNKNOWN
, DT_REG
, DT_DIR
, DT_CHR
, DT_BLK
, DT_FIFO
, DT_SOCK
, DT_LNK
1951 static int btrfs_real_readdir(struct file
*filp
, void *dirent
,
1954 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1955 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1956 struct btrfs_item
*item
;
1957 struct btrfs_dir_item
*di
;
1958 struct btrfs_key key
;
1959 struct btrfs_key found_key
;
1960 struct btrfs_path
*path
;
1963 struct extent_buffer
*leaf
;
1966 unsigned char d_type
;
1971 int key_type
= BTRFS_DIR_INDEX_KEY
;
1976 /* FIXME, use a real flag for deciding about the key type */
1977 if (root
->fs_info
->tree_root
== root
)
1978 key_type
= BTRFS_DIR_ITEM_KEY
;
1980 /* special case for "." */
1981 if (filp
->f_pos
== 0) {
1982 over
= filldir(dirent
, ".", 1,
1989 /* special case for .., just use the back ref */
1990 if (filp
->f_pos
== 1) {
1991 u64 pino
= parent_ino(filp
->f_path
.dentry
);
1992 over
= filldir(dirent
, "..", 2,
1999 path
= btrfs_alloc_path();
2002 btrfs_set_key_type(&key
, key_type
);
2003 key
.offset
= filp
->f_pos
;
2004 key
.objectid
= inode
->i_ino
;
2006 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2012 leaf
= path
->nodes
[0];
2013 nritems
= btrfs_header_nritems(leaf
);
2014 slot
= path
->slots
[0];
2015 if (advance
|| slot
>= nritems
) {
2016 if (slot
>= nritems
- 1) {
2017 ret
= btrfs_next_leaf(root
, path
);
2020 leaf
= path
->nodes
[0];
2021 nritems
= btrfs_header_nritems(leaf
);
2022 slot
= path
->slots
[0];
2029 item
= btrfs_item_nr(leaf
, slot
);
2030 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
2032 if (found_key
.objectid
!= key
.objectid
)
2034 if (btrfs_key_type(&found_key
) != key_type
)
2036 if (found_key
.offset
< filp
->f_pos
)
2039 filp
->f_pos
= found_key
.offset
;
2041 di
= btrfs_item_ptr(leaf
, slot
, struct btrfs_dir_item
);
2043 di_total
= btrfs_item_size(leaf
, item
);
2045 while (di_cur
< di_total
) {
2046 struct btrfs_key location
;
2048 name_len
= btrfs_dir_name_len(leaf
, di
);
2049 if (name_len
<= sizeof(tmp_name
)) {
2050 name_ptr
= tmp_name
;
2052 name_ptr
= kmalloc(name_len
, GFP_NOFS
);
2058 read_extent_buffer(leaf
, name_ptr
,
2059 (unsigned long)(di
+ 1), name_len
);
2061 d_type
= btrfs_filetype_table
[btrfs_dir_type(leaf
, di
)];
2062 btrfs_dir_item_key_to_cpu(leaf
, di
, &location
);
2063 over
= filldir(dirent
, name_ptr
, name_len
,
2064 found_key
.offset
, location
.objectid
,
2067 if (name_ptr
!= tmp_name
)
2073 di_len
= btrfs_dir_name_len(leaf
, di
) +
2074 btrfs_dir_data_len(leaf
, di
) + sizeof(*di
);
2076 di
= (struct btrfs_dir_item
*)((char *)di
+ di_len
);
2080 /* Reached end of directory/root. Bump pos past the last item. */
2081 if (key_type
== BTRFS_DIR_INDEX_KEY
)
2082 filp
->f_pos
= INT_LIMIT(typeof(filp
->f_pos
));
2088 btrfs_free_path(path
);
2092 int btrfs_write_inode(struct inode
*inode
, int wait
)
2094 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2095 struct btrfs_trans_handle
*trans
;
2098 if (root
->fs_info
->closing
> 1)
2102 trans
= btrfs_join_transaction(root
, 1);
2103 btrfs_set_trans_block_group(trans
, inode
);
2104 ret
= btrfs_commit_transaction(trans
, root
);
2110 * This is somewhat expensive, updating the tree every time the
2111 * inode changes. But, it is most likely to find the inode in cache.
2112 * FIXME, needs more benchmarking...there are no reasons other than performance
2113 * to keep or drop this code.
2115 void btrfs_dirty_inode(struct inode
*inode
)
2117 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2118 struct btrfs_trans_handle
*trans
;
2120 trans
= btrfs_join_transaction(root
, 1);
2121 btrfs_set_trans_block_group(trans
, inode
);
2122 btrfs_update_inode(trans
, root
, inode
);
2123 btrfs_end_transaction(trans
, root
);
2126 static int btrfs_set_inode_index_count(struct inode
*inode
)
2128 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2129 struct btrfs_key key
, found_key
;
2130 struct btrfs_path
*path
;
2131 struct extent_buffer
*leaf
;
2134 key
.objectid
= inode
->i_ino
;
2135 btrfs_set_key_type(&key
, BTRFS_DIR_INDEX_KEY
);
2136 key
.offset
= (u64
)-1;
2138 path
= btrfs_alloc_path();
2142 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2145 /* FIXME: we should be able to handle this */
2151 * MAGIC NUMBER EXPLANATION:
2152 * since we search a directory based on f_pos we have to start at 2
2153 * since '.' and '..' have f_pos of 0 and 1 respectively, so everybody
2154 * else has to start at 2
2156 if (path
->slots
[0] == 0) {
2157 BTRFS_I(inode
)->index_cnt
= 2;
2163 leaf
= path
->nodes
[0];
2164 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2166 if (found_key
.objectid
!= inode
->i_ino
||
2167 btrfs_key_type(&found_key
) != BTRFS_DIR_INDEX_KEY
) {
2168 BTRFS_I(inode
)->index_cnt
= 2;
2172 BTRFS_I(inode
)->index_cnt
= found_key
.offset
+ 1;
2174 btrfs_free_path(path
);
2178 static int btrfs_set_inode_index(struct inode
*dir
, struct inode
*inode
,
2183 if (BTRFS_I(dir
)->index_cnt
== (u64
)-1) {
2184 ret
= btrfs_set_inode_index_count(dir
);
2190 *index
= BTRFS_I(dir
)->index_cnt
;
2191 BTRFS_I(dir
)->index_cnt
++;
2196 static struct inode
*btrfs_new_inode(struct btrfs_trans_handle
*trans
,
2197 struct btrfs_root
*root
,
2199 const char *name
, int name_len
,
2202 struct btrfs_block_group_cache
*group
,
2203 int mode
, u64
*index
)
2205 struct inode
*inode
;
2206 struct btrfs_inode_item
*inode_item
;
2207 struct btrfs_block_group_cache
*new_inode_group
;
2208 struct btrfs_key
*location
;
2209 struct btrfs_path
*path
;
2210 struct btrfs_inode_ref
*ref
;
2211 struct btrfs_key key
[2];
2217 path
= btrfs_alloc_path();
2220 inode
= new_inode(root
->fs_info
->sb
);
2222 return ERR_PTR(-ENOMEM
);
2225 ret
= btrfs_set_inode_index(dir
, inode
, index
);
2227 return ERR_PTR(ret
);
2230 * index_cnt is ignored for everything but a dir,
2231 * btrfs_get_inode_index_count has an explanation for the magic
2234 init_btrfs_i(inode
);
2235 BTRFS_I(inode
)->index_cnt
= 2;
2236 BTRFS_I(inode
)->root
= root
;
2237 BTRFS_I(inode
)->generation
= trans
->transid
;
2243 new_inode_group
= btrfs_find_block_group(root
, group
, 0,
2244 BTRFS_BLOCK_GROUP_METADATA
, owner
);
2245 if (!new_inode_group
) {
2246 printk("find_block group failed\n");
2247 new_inode_group
= group
;
2249 BTRFS_I(inode
)->block_group
= new_inode_group
;
2251 key
[0].objectid
= objectid
;
2252 btrfs_set_key_type(&key
[0], BTRFS_INODE_ITEM_KEY
);
2255 key
[1].objectid
= objectid
;
2256 btrfs_set_key_type(&key
[1], BTRFS_INODE_REF_KEY
);
2257 key
[1].offset
= ref_objectid
;
2259 sizes
[0] = sizeof(struct btrfs_inode_item
);
2260 sizes
[1] = name_len
+ sizeof(*ref
);
2262 ret
= btrfs_insert_empty_items(trans
, root
, path
, key
, sizes
, 2);
2266 if (objectid
> root
->highest_inode
)
2267 root
->highest_inode
= objectid
;
2269 inode
->i_uid
= current
->fsuid
;
2270 inode
->i_gid
= current
->fsgid
;
2271 inode
->i_mode
= mode
;
2272 inode
->i_ino
= objectid
;
2273 inode
->i_blocks
= 0;
2274 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2275 inode_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
2276 struct btrfs_inode_item
);
2277 fill_inode_item(trans
, path
->nodes
[0], inode_item
, inode
);
2279 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0] + 1,
2280 struct btrfs_inode_ref
);
2281 btrfs_set_inode_ref_name_len(path
->nodes
[0], ref
, name_len
);
2282 btrfs_set_inode_ref_index(path
->nodes
[0], ref
, *index
);
2283 ptr
= (unsigned long)(ref
+ 1);
2284 write_extent_buffer(path
->nodes
[0], name
, ptr
, name_len
);
2286 btrfs_mark_buffer_dirty(path
->nodes
[0]);
2287 btrfs_free_path(path
);
2289 location
= &BTRFS_I(inode
)->location
;
2290 location
->objectid
= objectid
;
2291 location
->offset
= 0;
2292 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
2294 insert_inode_hash(inode
);
2298 BTRFS_I(dir
)->index_cnt
--;
2299 btrfs_free_path(path
);
2300 return ERR_PTR(ret
);
2303 static inline u8
btrfs_inode_type(struct inode
*inode
)
2305 return btrfs_type_by_mode
[(inode
->i_mode
& S_IFMT
) >> S_SHIFT
];
2308 int btrfs_add_link(struct btrfs_trans_handle
*trans
,
2309 struct inode
*parent_inode
, struct inode
*inode
,
2310 const char *name
, int name_len
, int add_backref
, u64 index
)
2313 struct btrfs_key key
;
2314 struct btrfs_root
*root
= BTRFS_I(parent_inode
)->root
;
2316 key
.objectid
= inode
->i_ino
;
2317 btrfs_set_key_type(&key
, BTRFS_INODE_ITEM_KEY
);
2320 ret
= btrfs_insert_dir_item(trans
, root
, name
, name_len
,
2321 parent_inode
->i_ino
,
2322 &key
, btrfs_inode_type(inode
),
2326 ret
= btrfs_insert_inode_ref(trans
, root
,
2329 parent_inode
->i_ino
,
2332 btrfs_i_size_write(parent_inode
, parent_inode
->i_size
+
2334 parent_inode
->i_mtime
= parent_inode
->i_ctime
= CURRENT_TIME
;
2335 ret
= btrfs_update_inode(trans
, root
, parent_inode
);
2340 static int btrfs_add_nondir(struct btrfs_trans_handle
*trans
,
2341 struct dentry
*dentry
, struct inode
*inode
,
2342 int backref
, u64 index
)
2344 int err
= btrfs_add_link(trans
, dentry
->d_parent
->d_inode
,
2345 inode
, dentry
->d_name
.name
,
2346 dentry
->d_name
.len
, backref
, index
);
2348 d_instantiate(dentry
, inode
);
2356 static int btrfs_mknod(struct inode
*dir
, struct dentry
*dentry
,
2357 int mode
, dev_t rdev
)
2359 struct btrfs_trans_handle
*trans
;
2360 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2361 struct inode
*inode
= NULL
;
2365 unsigned long nr
= 0;
2368 if (!new_valid_dev(rdev
))
2371 err
= btrfs_check_free_space(root
, 1, 0);
2375 trans
= btrfs_start_transaction(root
, 1);
2376 btrfs_set_trans_block_group(trans
, dir
);
2378 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2384 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2386 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2387 BTRFS_I(dir
)->block_group
, mode
, &index
);
2388 err
= PTR_ERR(inode
);
2392 err
= btrfs_init_acl(inode
, dir
);
2398 btrfs_set_trans_block_group(trans
, inode
);
2399 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
2403 inode
->i_op
= &btrfs_special_inode_operations
;
2404 init_special_inode(inode
, inode
->i_mode
, rdev
);
2405 btrfs_update_inode(trans
, root
, inode
);
2407 dir
->i_sb
->s_dirt
= 1;
2408 btrfs_update_inode_block_group(trans
, inode
);
2409 btrfs_update_inode_block_group(trans
, dir
);
2411 nr
= trans
->blocks_used
;
2412 btrfs_end_transaction_throttle(trans
, root
);
2415 inode_dec_link_count(inode
);
2418 btrfs_btree_balance_dirty(root
, nr
);
2422 static int btrfs_create(struct inode
*dir
, struct dentry
*dentry
,
2423 int mode
, struct nameidata
*nd
)
2425 struct btrfs_trans_handle
*trans
;
2426 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2427 struct inode
*inode
= NULL
;
2430 unsigned long nr
= 0;
2434 err
= btrfs_check_free_space(root
, 1, 0);
2437 trans
= btrfs_start_transaction(root
, 1);
2438 btrfs_set_trans_block_group(trans
, dir
);
2440 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2446 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2448 dentry
->d_parent
->d_inode
->i_ino
,
2449 objectid
, BTRFS_I(dir
)->block_group
, mode
,
2451 err
= PTR_ERR(inode
);
2455 err
= btrfs_init_acl(inode
, dir
);
2461 btrfs_set_trans_block_group(trans
, inode
);
2462 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
2466 inode
->i_mapping
->a_ops
= &btrfs_aops
;
2467 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
2468 inode
->i_fop
= &btrfs_file_operations
;
2469 inode
->i_op
= &btrfs_file_inode_operations
;
2470 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
2472 dir
->i_sb
->s_dirt
= 1;
2473 btrfs_update_inode_block_group(trans
, inode
);
2474 btrfs_update_inode_block_group(trans
, dir
);
2476 nr
= trans
->blocks_used
;
2477 btrfs_end_transaction_throttle(trans
, root
);
2480 inode_dec_link_count(inode
);
2483 btrfs_btree_balance_dirty(root
, nr
);
2487 static int btrfs_link(struct dentry
*old_dentry
, struct inode
*dir
,
2488 struct dentry
*dentry
)
2490 struct btrfs_trans_handle
*trans
;
2491 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2492 struct inode
*inode
= old_dentry
->d_inode
;
2494 unsigned long nr
= 0;
2498 if (inode
->i_nlink
== 0)
2501 btrfs_inc_nlink(inode
);
2502 err
= btrfs_check_free_space(root
, 1, 0);
2505 err
= btrfs_set_inode_index(dir
, inode
, &index
);
2509 trans
= btrfs_start_transaction(root
, 1);
2511 btrfs_set_trans_block_group(trans
, dir
);
2512 atomic_inc(&inode
->i_count
);
2514 err
= btrfs_add_nondir(trans
, dentry
, inode
, 1, index
);
2519 dir
->i_sb
->s_dirt
= 1;
2520 btrfs_update_inode_block_group(trans
, dir
);
2521 err
= btrfs_update_inode(trans
, root
, inode
);
2526 nr
= trans
->blocks_used
;
2527 btrfs_end_transaction_throttle(trans
, root
);
2530 inode_dec_link_count(inode
);
2533 btrfs_btree_balance_dirty(root
, nr
);
2537 static int btrfs_mkdir(struct inode
*dir
, struct dentry
*dentry
, int mode
)
2539 struct inode
*inode
= NULL
;
2540 struct btrfs_trans_handle
*trans
;
2541 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2543 int drop_on_err
= 0;
2546 unsigned long nr
= 1;
2548 err
= btrfs_check_free_space(root
, 1, 0);
2552 trans
= btrfs_start_transaction(root
, 1);
2553 btrfs_set_trans_block_group(trans
, dir
);
2555 if (IS_ERR(trans
)) {
2556 err
= PTR_ERR(trans
);
2560 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2566 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2568 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2569 BTRFS_I(dir
)->block_group
, S_IFDIR
| mode
,
2571 if (IS_ERR(inode
)) {
2572 err
= PTR_ERR(inode
);
2578 err
= btrfs_init_acl(inode
, dir
);
2582 inode
->i_op
= &btrfs_dir_inode_operations
;
2583 inode
->i_fop
= &btrfs_dir_file_operations
;
2584 btrfs_set_trans_block_group(trans
, inode
);
2586 btrfs_i_size_write(inode
, 0);
2587 err
= btrfs_update_inode(trans
, root
, inode
);
2591 err
= btrfs_add_link(trans
, dentry
->d_parent
->d_inode
,
2592 inode
, dentry
->d_name
.name
,
2593 dentry
->d_name
.len
, 0, index
);
2597 d_instantiate(dentry
, inode
);
2599 dir
->i_sb
->s_dirt
= 1;
2600 btrfs_update_inode_block_group(trans
, inode
);
2601 btrfs_update_inode_block_group(trans
, dir
);
2604 nr
= trans
->blocks_used
;
2605 btrfs_end_transaction_throttle(trans
, root
);
2610 btrfs_btree_balance_dirty(root
, nr
);
2614 static int merge_extent_mapping(struct extent_map_tree
*em_tree
,
2615 struct extent_map
*existing
,
2616 struct extent_map
*em
,
2617 u64 map_start
, u64 map_len
)
2621 BUG_ON(map_start
< em
->start
|| map_start
>= extent_map_end(em
));
2622 start_diff
= map_start
- em
->start
;
2623 em
->start
= map_start
;
2625 if (em
->block_start
< EXTENT_MAP_LAST_BYTE
)
2626 em
->block_start
+= start_diff
;
2627 return add_extent_mapping(em_tree
, em
);
2630 struct extent_map
*btrfs_get_extent(struct inode
*inode
, struct page
*page
,
2631 size_t pg_offset
, u64 start
, u64 len
,
2637 u64 extent_start
= 0;
2639 u64 objectid
= inode
->i_ino
;
2641 struct btrfs_path
*path
= NULL
;
2642 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2643 struct btrfs_file_extent_item
*item
;
2644 struct extent_buffer
*leaf
;
2645 struct btrfs_key found_key
;
2646 struct extent_map
*em
= NULL
;
2647 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
2648 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2649 struct btrfs_trans_handle
*trans
= NULL
;
2652 spin_lock(&em_tree
->lock
);
2653 em
= lookup_extent_mapping(em_tree
, start
, len
);
2655 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2656 spin_unlock(&em_tree
->lock
);
2659 if (em
->start
> start
|| em
->start
+ em
->len
<= start
)
2660 free_extent_map(em
);
2661 else if (em
->block_start
== EXTENT_MAP_INLINE
&& page
)
2662 free_extent_map(em
);
2666 em
= alloc_extent_map(GFP_NOFS
);
2671 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2672 em
->start
= EXTENT_MAP_HOLE
;
2676 path
= btrfs_alloc_path();
2680 ret
= btrfs_lookup_file_extent(trans
, root
, path
,
2681 objectid
, start
, trans
!= NULL
);
2688 if (path
->slots
[0] == 0)
2693 leaf
= path
->nodes
[0];
2694 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
2695 struct btrfs_file_extent_item
);
2696 /* are we inside the extent that was found? */
2697 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2698 found_type
= btrfs_key_type(&found_key
);
2699 if (found_key
.objectid
!= objectid
||
2700 found_type
!= BTRFS_EXTENT_DATA_KEY
) {
2704 found_type
= btrfs_file_extent_type(leaf
, item
);
2705 extent_start
= found_key
.offset
;
2706 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
2707 extent_end
= extent_start
+
2708 btrfs_file_extent_num_bytes(leaf
, item
);
2710 if (start
< extent_start
|| start
>= extent_end
) {
2712 if (start
< extent_start
) {
2713 if (start
+ len
<= extent_start
)
2715 em
->len
= extent_end
- extent_start
;
2721 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
2723 em
->start
= extent_start
;
2724 em
->len
= extent_end
- extent_start
;
2725 em
->block_start
= EXTENT_MAP_HOLE
;
2728 bytenr
+= btrfs_file_extent_offset(leaf
, item
);
2729 em
->block_start
= bytenr
;
2730 em
->start
= extent_start
;
2731 em
->len
= extent_end
- extent_start
;
2733 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
2738 size_t extent_offset
;
2741 size
= btrfs_file_extent_inline_len(leaf
, btrfs_item_nr(leaf
,
2743 extent_end
= (extent_start
+ size
+ root
->sectorsize
- 1) &
2744 ~((u64
)root
->sectorsize
- 1);
2745 if (start
< extent_start
|| start
>= extent_end
) {
2747 if (start
< extent_start
) {
2748 if (start
+ len
<= extent_start
)
2750 em
->len
= extent_end
- extent_start
;
2756 em
->block_start
= EXTENT_MAP_INLINE
;
2759 em
->start
= extent_start
;
2764 page_start
= page_offset(page
) + pg_offset
;
2765 extent_offset
= page_start
- extent_start
;
2766 copy_size
= min_t(u64
, PAGE_CACHE_SIZE
- pg_offset
,
2767 size
- extent_offset
);
2768 em
->start
= extent_start
+ extent_offset
;
2769 em
->len
= (copy_size
+ root
->sectorsize
- 1) &
2770 ~((u64
)root
->sectorsize
- 1);
2772 ptr
= btrfs_file_extent_inline_start(item
) + extent_offset
;
2773 if (create
== 0 && !PageUptodate(page
)) {
2774 read_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2776 flush_dcache_page(page
);
2777 } else if (create
&& PageUptodate(page
)) {
2780 free_extent_map(em
);
2782 btrfs_release_path(root
, path
);
2783 trans
= btrfs_join_transaction(root
, 1);
2786 write_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2788 btrfs_mark_buffer_dirty(leaf
);
2791 set_extent_uptodate(io_tree
, em
->start
,
2792 extent_map_end(em
) - 1, GFP_NOFS
);
2795 printk("unkknown found_type %d\n", found_type
);
2802 em
->block_start
= EXTENT_MAP_HOLE
;
2804 btrfs_release_path(root
, path
);
2805 if (em
->start
> start
|| extent_map_end(em
) <= start
) {
2806 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em
->start
, em
->len
, start
, len
);
2812 spin_lock(&em_tree
->lock
);
2813 ret
= add_extent_mapping(em_tree
, em
);
2814 /* it is possible that someone inserted the extent into the tree
2815 * while we had the lock dropped. It is also possible that
2816 * an overlapping map exists in the tree
2818 if (ret
== -EEXIST
) {
2819 struct extent_map
*existing
;
2823 existing
= lookup_extent_mapping(em_tree
, start
, len
);
2824 if (existing
&& (existing
->start
> start
||
2825 existing
->start
+ existing
->len
<= start
)) {
2826 free_extent_map(existing
);
2830 existing
= lookup_extent_mapping(em_tree
, em
->start
,
2833 err
= merge_extent_mapping(em_tree
, existing
,
2836 free_extent_map(existing
);
2838 free_extent_map(em
);
2843 printk("failing to insert %Lu %Lu\n",
2845 free_extent_map(em
);
2849 free_extent_map(em
);
2854 spin_unlock(&em_tree
->lock
);
2857 btrfs_free_path(path
);
2859 ret
= btrfs_end_transaction(trans
, root
);
2865 free_extent_map(em
);
2867 return ERR_PTR(err
);
2872 #if 0 /* waiting for O_DIRECT reads */
2873 static int btrfs_get_block(struct inode
*inode
, sector_t iblock
,
2874 struct buffer_head
*bh_result
, int create
)
2876 struct extent_map
*em
;
2877 u64 start
= (u64
)iblock
<< inode
->i_blkbits
;
2878 struct btrfs_multi_bio
*multi
= NULL
;
2879 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2885 em
= btrfs_get_extent(inode
, NULL
, 0, start
, bh_result
->b_size
, 0);
2887 if (!em
|| IS_ERR(em
))
2890 if (em
->start
> start
|| em
->start
+ em
->len
<= start
) {
2894 if (em
->block_start
== EXTENT_MAP_INLINE
) {
2899 len
= em
->start
+ em
->len
- start
;
2900 len
= min_t(u64
, len
, INT_LIMIT(typeof(bh_result
->b_size
)));
2902 if (em
->block_start
== EXTENT_MAP_HOLE
||
2903 em
->block_start
== EXTENT_MAP_DELALLOC
) {
2904 bh_result
->b_size
= len
;
2908 logical
= start
- em
->start
;
2909 logical
= em
->block_start
+ logical
;
2912 ret
= btrfs_map_block(&root
->fs_info
->mapping_tree
, READ
,
2913 logical
, &map_length
, &multi
, 0);
2915 bh_result
->b_blocknr
= multi
->stripes
[0].physical
>> inode
->i_blkbits
;
2916 bh_result
->b_size
= min(map_length
, len
);
2918 bh_result
->b_bdev
= multi
->stripes
[0].dev
->bdev
;
2919 set_buffer_mapped(bh_result
);
2922 free_extent_map(em
);
2927 static ssize_t
btrfs_direct_IO(int rw
, struct kiocb
*iocb
,
2928 const struct iovec
*iov
, loff_t offset
,
2929 unsigned long nr_segs
)
2933 struct file
*file
= iocb
->ki_filp
;
2934 struct inode
*inode
= file
->f_mapping
->host
;
2939 return blockdev_direct_IO(rw
, iocb
, inode
, inode
->i_sb
->s_bdev
, iov
,
2940 offset
, nr_segs
, btrfs_get_block
, NULL
);
2944 static sector_t
btrfs_bmap(struct address_space
*mapping
, sector_t iblock
)
2946 return extent_bmap(mapping
, iblock
, btrfs_get_extent
);
2949 int btrfs_readpage(struct file
*file
, struct page
*page
)
2951 struct extent_io_tree
*tree
;
2952 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2953 return extent_read_full_page(tree
, page
, btrfs_get_extent
);
2956 static int btrfs_writepage(struct page
*page
, struct writeback_control
*wbc
)
2958 struct extent_io_tree
*tree
;
2961 if (current
->flags
& PF_MEMALLOC
) {
2962 redirty_page_for_writepage(wbc
, page
);
2966 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2967 return extent_write_full_page(tree
, page
, btrfs_get_extent
, wbc
);
2970 int btrfs_writepages(struct address_space
*mapping
,
2971 struct writeback_control
*wbc
)
2973 struct extent_io_tree
*tree
;
2974 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
2975 return extent_writepages(tree
, mapping
, btrfs_get_extent
, wbc
);
2979 btrfs_readpages(struct file
*file
, struct address_space
*mapping
,
2980 struct list_head
*pages
, unsigned nr_pages
)
2982 struct extent_io_tree
*tree
;
2983 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
2984 return extent_readpages(tree
, mapping
, pages
, nr_pages
,
2987 static int __btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
2989 struct extent_io_tree
*tree
;
2990 struct extent_map_tree
*map
;
2993 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2994 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
2995 ret
= try_release_extent_mapping(map
, tree
, page
, gfp_flags
);
2997 ClearPagePrivate(page
);
2998 set_page_private(page
, 0);
2999 page_cache_release(page
);
3004 static int btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
3006 if (PageWriteback(page
) || PageDirty(page
))
3008 return __btrfs_releasepage(page
, gfp_flags
);
3011 static void btrfs_invalidatepage(struct page
*page
, unsigned long offset
)
3013 struct extent_io_tree
*tree
;
3014 struct btrfs_ordered_extent
*ordered
;
3015 u64 page_start
= page_offset(page
);
3016 u64 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
3018 wait_on_page_writeback(page
);
3019 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3021 btrfs_releasepage(page
, GFP_NOFS
);
3025 lock_extent(tree
, page_start
, page_end
, GFP_NOFS
);
3026 ordered
= btrfs_lookup_ordered_extent(page
->mapping
->host
,
3030 * IO on this page will never be started, so we need
3031 * to account for any ordered extents now
3033 clear_extent_bit(tree
, page_start
, page_end
,
3034 EXTENT_DIRTY
| EXTENT_DELALLOC
|
3035 EXTENT_LOCKED
, 1, 0, GFP_NOFS
);
3036 btrfs_finish_ordered_io(page
->mapping
->host
,
3037 page_start
, page_end
);
3038 btrfs_put_ordered_extent(ordered
);
3039 lock_extent(tree
, page_start
, page_end
, GFP_NOFS
);
3041 clear_extent_bit(tree
, page_start
, page_end
,
3042 EXTENT_LOCKED
| EXTENT_DIRTY
| EXTENT_DELALLOC
|
3045 __btrfs_releasepage(page
, GFP_NOFS
);
3047 ClearPageChecked(page
);
3048 if (PagePrivate(page
)) {
3049 ClearPagePrivate(page
);
3050 set_page_private(page
, 0);
3051 page_cache_release(page
);
3056 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
3057 * called from a page fault handler when a page is first dirtied. Hence we must
3058 * be careful to check for EOF conditions here. We set the page up correctly
3059 * for a written page which means we get ENOSPC checking when writing into
3060 * holes and correct delalloc and unwritten extent mapping on filesystems that
3061 * support these features.
3063 * We are not allowed to take the i_mutex here so we have to play games to
3064 * protect against truncate races as the page could now be beyond EOF. Because
3065 * vmtruncate() writes the inode size before removing pages, once we have the
3066 * page lock we can determine safely if the page is beyond EOF. If it is not
3067 * beyond EOF, then the page is guaranteed safe against truncation until we
3070 int btrfs_page_mkwrite(struct vm_area_struct
*vma
, struct page
*page
)
3072 struct inode
*inode
= fdentry(vma
->vm_file
)->d_inode
;
3073 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3074 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
3075 struct btrfs_ordered_extent
*ordered
;
3077 unsigned long zero_start
;
3083 ret
= btrfs_check_free_space(root
, PAGE_CACHE_SIZE
, 0);
3090 size
= i_size_read(inode
);
3091 page_start
= page_offset(page
);
3092 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
3094 if ((page
->mapping
!= inode
->i_mapping
) ||
3095 (page_start
>= size
)) {
3096 /* page got truncated out from underneath us */
3099 wait_on_page_writeback(page
);
3101 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3102 set_page_extent_mapped(page
);
3105 * we can't set the delalloc bits if there are pending ordered
3106 * extents. Drop our locks and wait for them to finish
3108 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
3110 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3112 btrfs_start_ordered_extent(inode
, ordered
, 1);
3113 btrfs_put_ordered_extent(ordered
);
3117 btrfs_set_extent_delalloc(inode
, page_start
, page_end
);
3120 /* page is wholly or partially inside EOF */
3121 if (page_start
+ PAGE_CACHE_SIZE
> size
)
3122 zero_start
= size
& ~PAGE_CACHE_MASK
;
3124 zero_start
= PAGE_CACHE_SIZE
;
3126 if (zero_start
!= PAGE_CACHE_SIZE
) {
3128 memset(kaddr
+ zero_start
, 0, PAGE_CACHE_SIZE
- zero_start
);
3129 flush_dcache_page(page
);
3132 ClearPageChecked(page
);
3133 set_page_dirty(page
);
3134 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3142 static void btrfs_truncate(struct inode
*inode
)
3144 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3146 struct btrfs_trans_handle
*trans
;
3148 u64 mask
= root
->sectorsize
- 1;
3150 if (!S_ISREG(inode
->i_mode
))
3152 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
3155 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
3156 btrfs_wait_ordered_range(inode
, inode
->i_size
& (~mask
), (u64
)-1);
3158 trans
= btrfs_start_transaction(root
, 1);
3159 btrfs_set_trans_block_group(trans
, inode
);
3160 btrfs_i_size_write(inode
, inode
->i_size
);
3162 ret
= btrfs_orphan_add(trans
, inode
);
3165 /* FIXME, add redo link to tree so we don't leak on crash */
3166 ret
= btrfs_truncate_inode_items(trans
, root
, inode
, inode
->i_size
,
3167 BTRFS_EXTENT_DATA_KEY
);
3168 btrfs_update_inode(trans
, root
, inode
);
3170 ret
= btrfs_orphan_del(trans
, inode
);
3174 nr
= trans
->blocks_used
;
3175 ret
= btrfs_end_transaction_throttle(trans
, root
);
3177 btrfs_btree_balance_dirty(root
, nr
);
3181 * Invalidate a single dcache entry at the root of the filesystem.
3182 * Needed after creation of snapshot or subvolume.
3184 void btrfs_invalidate_dcache_root(struct btrfs_root
*root
, char *name
,
3187 struct dentry
*alias
, *entry
;
3190 alias
= d_find_alias(root
->fs_info
->sb
->s_root
->d_inode
);
3194 /* change me if btrfs ever gets a d_hash operation */
3195 qstr
.hash
= full_name_hash(qstr
.name
, qstr
.len
);
3196 entry
= d_lookup(alias
, &qstr
);
3199 d_invalidate(entry
);
3205 int btrfs_create_subvol_root(struct btrfs_root
*new_root
,
3206 struct btrfs_trans_handle
*trans
, u64 new_dirid
,
3207 struct btrfs_block_group_cache
*block_group
)
3209 struct inode
*inode
;
3212 inode
= btrfs_new_inode(trans
, new_root
, NULL
, "..", 2, new_dirid
,
3213 new_dirid
, block_group
, S_IFDIR
| 0700, &index
);
3215 return PTR_ERR(inode
);
3216 inode
->i_op
= &btrfs_dir_inode_operations
;
3217 inode
->i_fop
= &btrfs_dir_file_operations
;
3218 new_root
->inode
= inode
;
3221 btrfs_i_size_write(inode
, 0);
3223 return btrfs_update_inode(trans
, new_root
, inode
);
3226 unsigned long btrfs_force_ra(struct address_space
*mapping
,
3227 struct file_ra_state
*ra
, struct file
*file
,
3228 pgoff_t offset
, pgoff_t last_index
)
3230 pgoff_t req_size
= last_index
- offset
+ 1;
3232 page_cache_sync_readahead(mapping
, ra
, file
, offset
, req_size
);
3233 return offset
+ req_size
;
3236 struct inode
*btrfs_alloc_inode(struct super_block
*sb
)
3238 struct btrfs_inode
*ei
;
3240 ei
= kmem_cache_alloc(btrfs_inode_cachep
, GFP_NOFS
);
3244 ei
->logged_trans
= 0;
3245 btrfs_ordered_inode_tree_init(&ei
->ordered_tree
);
3246 ei
->i_acl
= BTRFS_ACL_NOT_CACHED
;
3247 ei
->i_default_acl
= BTRFS_ACL_NOT_CACHED
;
3248 INIT_LIST_HEAD(&ei
->i_orphan
);
3249 return &ei
->vfs_inode
;
3252 void btrfs_destroy_inode(struct inode
*inode
)
3254 struct btrfs_ordered_extent
*ordered
;
3255 WARN_ON(!list_empty(&inode
->i_dentry
));
3256 WARN_ON(inode
->i_data
.nrpages
);
3258 if (BTRFS_I(inode
)->i_acl
&&
3259 BTRFS_I(inode
)->i_acl
!= BTRFS_ACL_NOT_CACHED
)
3260 posix_acl_release(BTRFS_I(inode
)->i_acl
);
3261 if (BTRFS_I(inode
)->i_default_acl
&&
3262 BTRFS_I(inode
)->i_default_acl
!= BTRFS_ACL_NOT_CACHED
)
3263 posix_acl_release(BTRFS_I(inode
)->i_default_acl
);
3265 spin_lock(&BTRFS_I(inode
)->root
->list_lock
);
3266 if (!list_empty(&BTRFS_I(inode
)->i_orphan
)) {
3267 printk(KERN_ERR
"BTRFS: inode %lu: inode still on the orphan"
3268 " list\n", inode
->i_ino
);
3271 spin_unlock(&BTRFS_I(inode
)->root
->list_lock
);
3274 ordered
= btrfs_lookup_first_ordered_extent(inode
, (u64
)-1);
3278 printk("found ordered extent %Lu %Lu\n",
3279 ordered
->file_offset
, ordered
->len
);
3280 btrfs_remove_ordered_extent(inode
, ordered
);
3281 btrfs_put_ordered_extent(ordered
);
3282 btrfs_put_ordered_extent(ordered
);
3285 btrfs_drop_extent_cache(inode
, 0, (u64
)-1, 0);
3286 kmem_cache_free(btrfs_inode_cachep
, BTRFS_I(inode
));
3289 static void init_once(void *foo
)
3291 struct btrfs_inode
*ei
= (struct btrfs_inode
*) foo
;
3293 inode_init_once(&ei
->vfs_inode
);
3296 void btrfs_destroy_cachep(void)
3298 if (btrfs_inode_cachep
)
3299 kmem_cache_destroy(btrfs_inode_cachep
);
3300 if (btrfs_trans_handle_cachep
)
3301 kmem_cache_destroy(btrfs_trans_handle_cachep
);
3302 if (btrfs_transaction_cachep
)
3303 kmem_cache_destroy(btrfs_transaction_cachep
);
3304 if (btrfs_bit_radix_cachep
)
3305 kmem_cache_destroy(btrfs_bit_radix_cachep
);
3306 if (btrfs_path_cachep
)
3307 kmem_cache_destroy(btrfs_path_cachep
);
3310 struct kmem_cache
*btrfs_cache_create(const char *name
, size_t size
,
3311 unsigned long extra_flags
,
3312 void (*ctor
)(void *))
3314 return kmem_cache_create(name
, size
, 0, (SLAB_RECLAIM_ACCOUNT
|
3315 SLAB_MEM_SPREAD
| extra_flags
), ctor
);
3318 int btrfs_init_cachep(void)
3320 btrfs_inode_cachep
= btrfs_cache_create("btrfs_inode_cache",
3321 sizeof(struct btrfs_inode
),
3323 if (!btrfs_inode_cachep
)
3325 btrfs_trans_handle_cachep
=
3326 btrfs_cache_create("btrfs_trans_handle_cache",
3327 sizeof(struct btrfs_trans_handle
),
3329 if (!btrfs_trans_handle_cachep
)
3331 btrfs_transaction_cachep
= btrfs_cache_create("btrfs_transaction_cache",
3332 sizeof(struct btrfs_transaction
),
3334 if (!btrfs_transaction_cachep
)
3336 btrfs_path_cachep
= btrfs_cache_create("btrfs_path_cache",
3337 sizeof(struct btrfs_path
),
3339 if (!btrfs_path_cachep
)
3341 btrfs_bit_radix_cachep
= btrfs_cache_create("btrfs_radix", 256,
3342 SLAB_DESTROY_BY_RCU
, NULL
);
3343 if (!btrfs_bit_radix_cachep
)
3347 btrfs_destroy_cachep();
3351 static int btrfs_getattr(struct vfsmount
*mnt
,
3352 struct dentry
*dentry
, struct kstat
*stat
)
3354 struct inode
*inode
= dentry
->d_inode
;
3355 generic_fillattr(inode
, stat
);
3356 stat
->blksize
= PAGE_CACHE_SIZE
;
3357 stat
->blocks
= inode
->i_blocks
+ (BTRFS_I(inode
)->delalloc_bytes
>> 9);
3361 static int btrfs_rename(struct inode
* old_dir
, struct dentry
*old_dentry
,
3362 struct inode
* new_dir
,struct dentry
*new_dentry
)
3364 struct btrfs_trans_handle
*trans
;
3365 struct btrfs_root
*root
= BTRFS_I(old_dir
)->root
;
3366 struct inode
*new_inode
= new_dentry
->d_inode
;
3367 struct inode
*old_inode
= old_dentry
->d_inode
;
3368 struct timespec ctime
= CURRENT_TIME
;
3372 if (S_ISDIR(old_inode
->i_mode
) && new_inode
&&
3373 new_inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
3377 ret
= btrfs_check_free_space(root
, 1, 0);
3381 trans
= btrfs_start_transaction(root
, 1);
3383 btrfs_set_trans_block_group(trans
, new_dir
);
3385 btrfs_inc_nlink(old_dentry
->d_inode
);
3386 old_dir
->i_ctime
= old_dir
->i_mtime
= ctime
;
3387 new_dir
->i_ctime
= new_dir
->i_mtime
= ctime
;
3388 old_inode
->i_ctime
= ctime
;
3390 ret
= btrfs_unlink_inode(trans
, root
, old_dir
, old_dentry
->d_inode
,
3391 old_dentry
->d_name
.name
,
3392 old_dentry
->d_name
.len
);
3397 new_inode
->i_ctime
= CURRENT_TIME
;
3398 ret
= btrfs_unlink_inode(trans
, root
, new_dir
,
3399 new_dentry
->d_inode
,
3400 new_dentry
->d_name
.name
,
3401 new_dentry
->d_name
.len
);
3404 if (new_inode
->i_nlink
== 0) {
3405 ret
= btrfs_orphan_add(trans
, new_dentry
->d_inode
);
3411 ret
= btrfs_set_inode_index(new_dir
, old_inode
, &index
);
3415 ret
= btrfs_add_link(trans
, new_dentry
->d_parent
->d_inode
,
3416 old_inode
, new_dentry
->d_name
.name
,
3417 new_dentry
->d_name
.len
, 1, index
);
3422 btrfs_end_transaction_throttle(trans
, root
);
3427 int btrfs_start_delalloc_inodes(struct btrfs_root
*root
)
3429 struct list_head
*head
= &root
->fs_info
->delalloc_inodes
;
3430 struct btrfs_inode
*binode
;
3431 struct inode
*inode
;
3432 unsigned long flags
;
3434 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
3435 while(!list_empty(head
)) {
3436 binode
= list_entry(head
->next
, struct btrfs_inode
,
3438 inode
= igrab(&binode
->vfs_inode
);
3440 list_del_init(&binode
->delalloc_inodes
);
3441 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
3443 filemap_write_and_wait(inode
->i_mapping
);
3447 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
3449 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
3453 static int btrfs_symlink(struct inode
*dir
, struct dentry
*dentry
,
3454 const char *symname
)
3456 struct btrfs_trans_handle
*trans
;
3457 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
3458 struct btrfs_path
*path
;
3459 struct btrfs_key key
;
3460 struct inode
*inode
= NULL
;
3468 struct btrfs_file_extent_item
*ei
;
3469 struct extent_buffer
*leaf
;
3470 unsigned long nr
= 0;
3472 name_len
= strlen(symname
) + 1;
3473 if (name_len
> BTRFS_MAX_INLINE_DATA_SIZE(root
))
3474 return -ENAMETOOLONG
;
3476 err
= btrfs_check_free_space(root
, 1, 0);
3480 trans
= btrfs_start_transaction(root
, 1);
3481 btrfs_set_trans_block_group(trans
, dir
);
3483 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
3489 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
3491 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
3492 BTRFS_I(dir
)->block_group
, S_IFLNK
|S_IRWXUGO
,
3494 err
= PTR_ERR(inode
);
3498 err
= btrfs_init_acl(inode
, dir
);
3504 btrfs_set_trans_block_group(trans
, inode
);
3505 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
3509 inode
->i_mapping
->a_ops
= &btrfs_aops
;
3510 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3511 inode
->i_fop
= &btrfs_file_operations
;
3512 inode
->i_op
= &btrfs_file_inode_operations
;
3513 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
3515 dir
->i_sb
->s_dirt
= 1;
3516 btrfs_update_inode_block_group(trans
, inode
);
3517 btrfs_update_inode_block_group(trans
, dir
);
3521 path
= btrfs_alloc_path();
3523 key
.objectid
= inode
->i_ino
;
3525 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
3526 datasize
= btrfs_file_extent_calc_inline_size(name_len
);
3527 err
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
3533 leaf
= path
->nodes
[0];
3534 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
3535 struct btrfs_file_extent_item
);
3536 btrfs_set_file_extent_generation(leaf
, ei
, trans
->transid
);
3537 btrfs_set_file_extent_type(leaf
, ei
,
3538 BTRFS_FILE_EXTENT_INLINE
);
3539 ptr
= btrfs_file_extent_inline_start(ei
);
3540 write_extent_buffer(leaf
, symname
, ptr
, name_len
);
3541 btrfs_mark_buffer_dirty(leaf
);
3542 btrfs_free_path(path
);
3544 inode
->i_op
= &btrfs_symlink_inode_operations
;
3545 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
3546 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3547 btrfs_i_size_write(inode
, name_len
- 1);
3548 err
= btrfs_update_inode(trans
, root
, inode
);
3553 nr
= trans
->blocks_used
;
3554 btrfs_end_transaction_throttle(trans
, root
);
3557 inode_dec_link_count(inode
);
3560 btrfs_btree_balance_dirty(root
, nr
);
3564 static int btrfs_set_page_dirty(struct page
*page
)
3566 return __set_page_dirty_nobuffers(page
);
3569 static int btrfs_permission(struct inode
*inode
, int mask
)
3571 if (btrfs_test_flag(inode
, READONLY
) && (mask
& MAY_WRITE
))
3573 return generic_permission(inode
, mask
, btrfs_check_acl
);
3576 static struct inode_operations btrfs_dir_inode_operations
= {
3577 .lookup
= btrfs_lookup
,
3578 .create
= btrfs_create
,
3579 .unlink
= btrfs_unlink
,
3581 .mkdir
= btrfs_mkdir
,
3582 .rmdir
= btrfs_rmdir
,
3583 .rename
= btrfs_rename
,
3584 .symlink
= btrfs_symlink
,
3585 .setattr
= btrfs_setattr
,
3586 .mknod
= btrfs_mknod
,
3587 .setxattr
= btrfs_setxattr
,
3588 .getxattr
= btrfs_getxattr
,
3589 .listxattr
= btrfs_listxattr
,
3590 .removexattr
= btrfs_removexattr
,
3591 .permission
= btrfs_permission
,
3593 static struct inode_operations btrfs_dir_ro_inode_operations
= {
3594 .lookup
= btrfs_lookup
,
3595 .permission
= btrfs_permission
,
3597 static struct file_operations btrfs_dir_file_operations
= {
3598 .llseek
= generic_file_llseek
,
3599 .read
= generic_read_dir
,
3600 .readdir
= btrfs_real_readdir
,
3601 .unlocked_ioctl
= btrfs_ioctl
,
3602 #ifdef CONFIG_COMPAT
3603 .compat_ioctl
= btrfs_ioctl
,
3605 .release
= btrfs_release_file
,
3606 .fsync
= btrfs_sync_file
,
3609 static struct extent_io_ops btrfs_extent_io_ops
= {
3610 .fill_delalloc
= run_delalloc_range
,
3611 .submit_bio_hook
= btrfs_submit_bio_hook
,
3612 .merge_bio_hook
= btrfs_merge_bio_hook
,
3613 .readpage_end_io_hook
= btrfs_readpage_end_io_hook
,
3614 .writepage_end_io_hook
= btrfs_writepage_end_io_hook
,
3615 .writepage_start_hook
= btrfs_writepage_start_hook
,
3616 .readpage_io_failed_hook
= btrfs_io_failed_hook
,
3617 .set_bit_hook
= btrfs_set_bit_hook
,
3618 .clear_bit_hook
= btrfs_clear_bit_hook
,
3621 static struct address_space_operations btrfs_aops
= {
3622 .readpage
= btrfs_readpage
,
3623 .writepage
= btrfs_writepage
,
3624 .writepages
= btrfs_writepages
,
3625 .readpages
= btrfs_readpages
,
3626 .sync_page
= block_sync_page
,
3628 .direct_IO
= btrfs_direct_IO
,
3629 .invalidatepage
= btrfs_invalidatepage
,
3630 .releasepage
= btrfs_releasepage
,
3631 .set_page_dirty
= btrfs_set_page_dirty
,
3634 static struct address_space_operations btrfs_symlink_aops
= {
3635 .readpage
= btrfs_readpage
,
3636 .writepage
= btrfs_writepage
,
3637 .invalidatepage
= btrfs_invalidatepage
,
3638 .releasepage
= btrfs_releasepage
,
3641 static struct inode_operations btrfs_file_inode_operations
= {
3642 .truncate
= btrfs_truncate
,
3643 .getattr
= btrfs_getattr
,
3644 .setattr
= btrfs_setattr
,
3645 .setxattr
= btrfs_setxattr
,
3646 .getxattr
= btrfs_getxattr
,
3647 .listxattr
= btrfs_listxattr
,
3648 .removexattr
= btrfs_removexattr
,
3649 .permission
= btrfs_permission
,
3651 static struct inode_operations btrfs_special_inode_operations
= {
3652 .getattr
= btrfs_getattr
,
3653 .setattr
= btrfs_setattr
,
3654 .permission
= btrfs_permission
,
3655 .setxattr
= btrfs_setxattr
,
3656 .getxattr
= btrfs_getxattr
,
3657 .listxattr
= btrfs_listxattr
,
3658 .removexattr
= btrfs_removexattr
,
3660 static struct inode_operations btrfs_symlink_inode_operations
= {
3661 .readlink
= generic_readlink
,
3662 .follow_link
= page_follow_link_light
,
3663 .put_link
= page_put_link
,
3664 .permission
= btrfs_permission
,