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);
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);
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);
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 btrfs_ordered_update_i_size(inode
, ordered_extent
);
612 btrfs_update_inode(trans
, root
, inode
);
613 btrfs_remove_ordered_extent(inode
, ordered_extent
);
616 btrfs_put_ordered_extent(ordered_extent
);
617 /* once for the tree */
618 btrfs_put_ordered_extent(ordered_extent
);
620 btrfs_end_transaction(trans
, root
);
622 btrfs_free_path(path
);
626 int btrfs_writepage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
627 struct extent_state
*state
, int uptodate
)
629 return btrfs_finish_ordered_io(page
->mapping
->host
, start
, end
);
632 struct io_failure_record
{
640 int btrfs_io_failed_hook(struct bio
*failed_bio
,
641 struct page
*page
, u64 start
, u64 end
,
642 struct extent_state
*state
)
644 struct io_failure_record
*failrec
= NULL
;
646 struct extent_map
*em
;
647 struct inode
*inode
= page
->mapping
->host
;
648 struct extent_io_tree
*failure_tree
= &BTRFS_I(inode
)->io_failure_tree
;
649 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
656 ret
= get_state_private(failure_tree
, start
, &private);
658 failrec
= kmalloc(sizeof(*failrec
), GFP_NOFS
);
661 failrec
->start
= start
;
662 failrec
->len
= end
- start
+ 1;
663 failrec
->last_mirror
= 0;
665 spin_lock(&em_tree
->lock
);
666 em
= lookup_extent_mapping(em_tree
, start
, failrec
->len
);
667 if (em
->start
> start
|| em
->start
+ em
->len
< start
) {
671 spin_unlock(&em_tree
->lock
);
673 if (!em
|| IS_ERR(em
)) {
677 logical
= start
- em
->start
;
678 logical
= em
->block_start
+ logical
;
679 failrec
->logical
= logical
;
681 set_extent_bits(failure_tree
, start
, end
, EXTENT_LOCKED
|
682 EXTENT_DIRTY
, GFP_NOFS
);
683 set_state_private(failure_tree
, start
,
684 (u64
)(unsigned long)failrec
);
686 failrec
= (struct io_failure_record
*)(unsigned long)private;
688 num_copies
= btrfs_num_copies(
689 &BTRFS_I(inode
)->root
->fs_info
->mapping_tree
,
690 failrec
->logical
, failrec
->len
);
691 failrec
->last_mirror
++;
693 spin_lock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
694 state
= find_first_extent_bit_state(&BTRFS_I(inode
)->io_tree
,
697 if (state
&& state
->start
!= failrec
->start
)
699 spin_unlock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
701 if (!state
|| failrec
->last_mirror
> num_copies
) {
702 set_state_private(failure_tree
, failrec
->start
, 0);
703 clear_extent_bits(failure_tree
, failrec
->start
,
704 failrec
->start
+ failrec
->len
- 1,
705 EXTENT_LOCKED
| EXTENT_DIRTY
, GFP_NOFS
);
709 bio
= bio_alloc(GFP_NOFS
, 1);
710 bio
->bi_private
= state
;
711 bio
->bi_end_io
= failed_bio
->bi_end_io
;
712 bio
->bi_sector
= failrec
->logical
>> 9;
713 bio
->bi_bdev
= failed_bio
->bi_bdev
;
715 bio_add_page(bio
, page
, failrec
->len
, start
- page_offset(page
));
716 if (failed_bio
->bi_rw
& (1 << BIO_RW
))
721 BTRFS_I(inode
)->io_tree
.ops
->submit_bio_hook(inode
, rw
, bio
,
722 failrec
->last_mirror
);
726 int btrfs_clean_io_failures(struct inode
*inode
, u64 start
)
730 struct io_failure_record
*failure
;
734 if (count_range_bits(&BTRFS_I(inode
)->io_failure_tree
, &private,
735 (u64
)-1, 1, EXTENT_DIRTY
)) {
736 ret
= get_state_private(&BTRFS_I(inode
)->io_failure_tree
,
737 start
, &private_failure
);
739 failure
= (struct io_failure_record
*)(unsigned long)
741 set_state_private(&BTRFS_I(inode
)->io_failure_tree
,
743 clear_extent_bits(&BTRFS_I(inode
)->io_failure_tree
,
745 failure
->start
+ failure
->len
- 1,
746 EXTENT_DIRTY
| EXTENT_LOCKED
,
754 int btrfs_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
755 struct extent_state
*state
)
757 size_t offset
= start
- ((u64
)page
->index
<< PAGE_CACHE_SHIFT
);
758 struct inode
*inode
= page
->mapping
->host
;
759 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
761 u64
private = ~(u32
)0;
763 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
767 if (btrfs_test_opt(root
, NODATASUM
) ||
768 btrfs_test_flag(inode
, NODATASUM
))
770 if (state
&& state
->start
== start
) {
771 private = state
->private;
774 ret
= get_state_private(io_tree
, start
, &private);
776 local_irq_save(flags
);
777 kaddr
= kmap_atomic(page
, KM_IRQ0
);
781 csum
= btrfs_csum_data(root
, kaddr
+ offset
, csum
, end
- start
+ 1);
782 btrfs_csum_final(csum
, (char *)&csum
);
783 if (csum
!= private) {
786 kunmap_atomic(kaddr
, KM_IRQ0
);
787 local_irq_restore(flags
);
789 /* if the io failure tree for this inode is non-empty,
790 * check to see if we've recovered from a failed IO
792 btrfs_clean_io_failures(inode
, start
);
796 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
797 page
->mapping
->host
->i_ino
, (unsigned long long)start
, csum
,
799 memset(kaddr
+ offset
, 1, end
- start
+ 1);
800 flush_dcache_page(page
);
801 kunmap_atomic(kaddr
, KM_IRQ0
);
802 local_irq_restore(flags
);
809 * This creates an orphan entry for the given inode in case something goes
810 * wrong in the middle of an unlink/truncate.
812 int btrfs_orphan_add(struct btrfs_trans_handle
*trans
, struct inode
*inode
)
814 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
817 spin_lock(&root
->list_lock
);
819 /* already on the orphan list, we're good */
820 if (!list_empty(&BTRFS_I(inode
)->i_orphan
)) {
821 spin_unlock(&root
->list_lock
);
825 list_add(&BTRFS_I(inode
)->i_orphan
, &root
->orphan_list
);
827 spin_unlock(&root
->list_lock
);
830 * insert an orphan item to track this unlinked/truncated file
832 ret
= btrfs_insert_orphan_item(trans
, root
, inode
->i_ino
);
838 * We have done the truncate/delete so we can go ahead and remove the orphan
839 * item for this particular inode.
841 int btrfs_orphan_del(struct btrfs_trans_handle
*trans
, struct inode
*inode
)
843 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
846 spin_lock(&root
->list_lock
);
848 if (list_empty(&BTRFS_I(inode
)->i_orphan
)) {
849 spin_unlock(&root
->list_lock
);
853 list_del_init(&BTRFS_I(inode
)->i_orphan
);
855 spin_unlock(&root
->list_lock
);
859 spin_unlock(&root
->list_lock
);
861 ret
= btrfs_del_orphan_item(trans
, root
, inode
->i_ino
);
867 * this cleans up any orphans that may be left on the list from the last use
870 void btrfs_orphan_cleanup(struct btrfs_root
*root
)
872 struct btrfs_path
*path
;
873 struct extent_buffer
*leaf
;
874 struct btrfs_item
*item
;
875 struct btrfs_key key
, found_key
;
876 struct btrfs_trans_handle
*trans
;
878 int ret
= 0, nr_unlink
= 0, nr_truncate
= 0;
880 /* don't do orphan cleanup if the fs is readonly. */
881 if (root
->inode
->i_sb
->s_flags
& MS_RDONLY
)
884 path
= btrfs_alloc_path();
889 key
.objectid
= BTRFS_ORPHAN_OBJECTID
;
890 btrfs_set_key_type(&key
, BTRFS_ORPHAN_ITEM_KEY
);
891 key
.offset
= (u64
)-1;
893 trans
= btrfs_start_transaction(root
, 1);
894 btrfs_set_trans_block_group(trans
, root
->inode
);
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
->inode
->i_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 btrfs_orphan_del(trans
, inode
);
971 /* if we have links, this was a truncate, lets do that */
972 if (inode
->i_nlink
) {
974 btrfs_truncate(inode
);
979 /* this will do delete_inode and everything for us */
984 printk(KERN_INFO
"btrfs: unlinked %d orphans\n", nr_unlink
);
986 printk(KERN_INFO
"btrfs: truncated %d orphans\n", nr_truncate
);
988 btrfs_free_path(path
);
989 btrfs_end_transaction(trans
, root
);
992 void btrfs_read_locked_inode(struct inode
*inode
)
994 struct btrfs_path
*path
;
995 struct extent_buffer
*leaf
;
996 struct btrfs_inode_item
*inode_item
;
997 struct btrfs_timespec
*tspec
;
998 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
999 struct btrfs_key location
;
1000 u64 alloc_group_block
;
1004 path
= btrfs_alloc_path();
1006 memcpy(&location
, &BTRFS_I(inode
)->location
, sizeof(location
));
1008 ret
= btrfs_lookup_inode(NULL
, root
, path
, &location
, 0);
1012 leaf
= path
->nodes
[0];
1013 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
1014 struct btrfs_inode_item
);
1016 inode
->i_mode
= btrfs_inode_mode(leaf
, inode_item
);
1017 inode
->i_nlink
= btrfs_inode_nlink(leaf
, inode_item
);
1018 inode
->i_uid
= btrfs_inode_uid(leaf
, inode_item
);
1019 inode
->i_gid
= btrfs_inode_gid(leaf
, inode_item
);
1020 btrfs_i_size_write(inode
, btrfs_inode_size(leaf
, inode_item
));
1022 tspec
= btrfs_inode_atime(inode_item
);
1023 inode
->i_atime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1024 inode
->i_atime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1026 tspec
= btrfs_inode_mtime(inode_item
);
1027 inode
->i_mtime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1028 inode
->i_mtime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1030 tspec
= btrfs_inode_ctime(inode_item
);
1031 inode
->i_ctime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1032 inode
->i_ctime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1034 inode
->i_blocks
= btrfs_inode_nblocks(leaf
, inode_item
);
1035 BTRFS_I(inode
)->generation
= btrfs_inode_generation(leaf
, inode_item
);
1036 inode
->i_generation
= BTRFS_I(inode
)->generation
;
1038 rdev
= btrfs_inode_rdev(leaf
, inode_item
);
1040 BTRFS_I(inode
)->index_cnt
= (u64
)-1;
1042 alloc_group_block
= btrfs_inode_block_group(leaf
, inode_item
);
1043 BTRFS_I(inode
)->block_group
= btrfs_lookup_block_group(root
->fs_info
,
1045 BTRFS_I(inode
)->flags
= btrfs_inode_flags(leaf
, inode_item
);
1046 if (!BTRFS_I(inode
)->block_group
) {
1047 BTRFS_I(inode
)->block_group
= btrfs_find_block_group(root
,
1049 BTRFS_BLOCK_GROUP_METADATA
, 0);
1051 btrfs_free_path(path
);
1054 switch (inode
->i_mode
& S_IFMT
) {
1056 inode
->i_mapping
->a_ops
= &btrfs_aops
;
1057 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1058 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
1059 inode
->i_fop
= &btrfs_file_operations
;
1060 inode
->i_op
= &btrfs_file_inode_operations
;
1063 inode
->i_fop
= &btrfs_dir_file_operations
;
1064 if (root
== root
->fs_info
->tree_root
)
1065 inode
->i_op
= &btrfs_dir_ro_inode_operations
;
1067 inode
->i_op
= &btrfs_dir_inode_operations
;
1070 inode
->i_op
= &btrfs_symlink_inode_operations
;
1071 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
1072 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1075 init_special_inode(inode
, inode
->i_mode
, rdev
);
1081 btrfs_free_path(path
);
1082 make_bad_inode(inode
);
1085 static void fill_inode_item(struct btrfs_trans_handle
*trans
,
1086 struct extent_buffer
*leaf
,
1087 struct btrfs_inode_item
*item
,
1088 struct inode
*inode
)
1090 btrfs_set_inode_uid(leaf
, item
, inode
->i_uid
);
1091 btrfs_set_inode_gid(leaf
, item
, inode
->i_gid
);
1092 btrfs_set_inode_size(leaf
, item
, BTRFS_I(inode
)->disk_i_size
);
1093 btrfs_set_inode_mode(leaf
, item
, inode
->i_mode
);
1094 btrfs_set_inode_nlink(leaf
, item
, inode
->i_nlink
);
1096 btrfs_set_timespec_sec(leaf
, btrfs_inode_atime(item
),
1097 inode
->i_atime
.tv_sec
);
1098 btrfs_set_timespec_nsec(leaf
, btrfs_inode_atime(item
),
1099 inode
->i_atime
.tv_nsec
);
1101 btrfs_set_timespec_sec(leaf
, btrfs_inode_mtime(item
),
1102 inode
->i_mtime
.tv_sec
);
1103 btrfs_set_timespec_nsec(leaf
, btrfs_inode_mtime(item
),
1104 inode
->i_mtime
.tv_nsec
);
1106 btrfs_set_timespec_sec(leaf
, btrfs_inode_ctime(item
),
1107 inode
->i_ctime
.tv_sec
);
1108 btrfs_set_timespec_nsec(leaf
, btrfs_inode_ctime(item
),
1109 inode
->i_ctime
.tv_nsec
);
1111 btrfs_set_inode_nblocks(leaf
, item
, inode
->i_blocks
);
1112 btrfs_set_inode_generation(leaf
, item
, BTRFS_I(inode
)->generation
);
1113 btrfs_set_inode_transid(leaf
, item
, trans
->transid
);
1114 btrfs_set_inode_rdev(leaf
, item
, inode
->i_rdev
);
1115 btrfs_set_inode_flags(leaf
, item
, BTRFS_I(inode
)->flags
);
1116 btrfs_set_inode_block_group(leaf
, item
,
1117 BTRFS_I(inode
)->block_group
->key
.objectid
);
1120 int noinline
btrfs_update_inode(struct btrfs_trans_handle
*trans
,
1121 struct btrfs_root
*root
,
1122 struct inode
*inode
)
1124 struct btrfs_inode_item
*inode_item
;
1125 struct btrfs_path
*path
;
1126 struct extent_buffer
*leaf
;
1129 path
= btrfs_alloc_path();
1131 ret
= btrfs_lookup_inode(trans
, root
, path
,
1132 &BTRFS_I(inode
)->location
, 1);
1139 leaf
= path
->nodes
[0];
1140 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
1141 struct btrfs_inode_item
);
1143 fill_inode_item(trans
, leaf
, inode_item
, inode
);
1144 btrfs_mark_buffer_dirty(leaf
);
1145 btrfs_set_inode_last_trans(trans
, inode
);
1148 btrfs_free_path(path
);
1153 int btrfs_unlink_inode(struct btrfs_trans_handle
*trans
,
1154 struct btrfs_root
*root
,
1155 struct inode
*dir
, struct inode
*inode
,
1156 const char *name
, int name_len
)
1158 struct btrfs_path
*path
;
1160 struct extent_buffer
*leaf
;
1161 struct btrfs_dir_item
*di
;
1162 struct btrfs_key key
;
1165 path
= btrfs_alloc_path();
1171 di
= btrfs_lookup_dir_item(trans
, root
, path
, dir
->i_ino
,
1172 name
, name_len
, -1);
1181 leaf
= path
->nodes
[0];
1182 btrfs_dir_item_key_to_cpu(leaf
, di
, &key
);
1183 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
1186 btrfs_release_path(root
, path
);
1188 ret
= btrfs_del_inode_ref(trans
, root
, name
, name_len
,
1190 dir
->i_ino
, &index
);
1192 printk("failed to delete reference to %.*s, "
1193 "inode %lu parent %lu\n", name_len
, name
,
1194 inode
->i_ino
, dir
->i_ino
);
1198 di
= btrfs_lookup_dir_index_item(trans
, root
, path
, dir
->i_ino
,
1199 index
, name
, name_len
, -1);
1208 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
1209 btrfs_release_path(root
, path
);
1211 ret
= btrfs_del_inode_ref_in_log(trans
, root
, name
, name_len
,
1213 BUG_ON(ret
!= 0 && ret
!= -ENOENT
);
1215 BTRFS_I(dir
)->log_dirty_trans
= trans
->transid
;
1217 ret
= btrfs_del_dir_entries_in_log(trans
, root
, name
, name_len
,
1221 btrfs_free_path(path
);
1225 btrfs_i_size_write(dir
, dir
->i_size
- name_len
* 2);
1226 inode
->i_ctime
= dir
->i_mtime
= dir
->i_ctime
= CURRENT_TIME
;
1227 btrfs_update_inode(trans
, root
, dir
);
1228 btrfs_drop_nlink(inode
);
1229 ret
= btrfs_update_inode(trans
, root
, inode
);
1230 dir
->i_sb
->s_dirt
= 1;
1235 static int btrfs_unlink(struct inode
*dir
, struct dentry
*dentry
)
1237 struct btrfs_root
*root
;
1238 struct btrfs_trans_handle
*trans
;
1239 struct inode
*inode
= dentry
->d_inode
;
1241 unsigned long nr
= 0;
1243 root
= BTRFS_I(dir
)->root
;
1245 ret
= btrfs_check_free_space(root
, 1, 1);
1249 trans
= btrfs_start_transaction(root
, 1);
1251 btrfs_set_trans_block_group(trans
, dir
);
1252 ret
= btrfs_unlink_inode(trans
, root
, dir
, dentry
->d_inode
,
1253 dentry
->d_name
.name
, dentry
->d_name
.len
);
1255 if (inode
->i_nlink
== 0)
1256 ret
= btrfs_orphan_add(trans
, inode
);
1258 nr
= trans
->blocks_used
;
1260 btrfs_end_transaction_throttle(trans
, root
);
1262 btrfs_btree_balance_dirty(root
, nr
);
1266 static int btrfs_rmdir(struct inode
*dir
, struct dentry
*dentry
)
1268 struct inode
*inode
= dentry
->d_inode
;
1271 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1272 struct btrfs_trans_handle
*trans
;
1273 unsigned long nr
= 0;
1275 if (inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
1279 ret
= btrfs_check_free_space(root
, 1, 1);
1283 trans
= btrfs_start_transaction(root
, 1);
1284 btrfs_set_trans_block_group(trans
, dir
);
1286 err
= btrfs_orphan_add(trans
, inode
);
1290 /* now the directory is empty */
1291 err
= btrfs_unlink_inode(trans
, root
, dir
, dentry
->d_inode
,
1292 dentry
->d_name
.name
, dentry
->d_name
.len
);
1294 btrfs_i_size_write(inode
, 0);
1298 nr
= trans
->blocks_used
;
1299 ret
= btrfs_end_transaction_throttle(trans
, root
);
1301 btrfs_btree_balance_dirty(root
, nr
);
1309 * this can truncate away extent items, csum items and directory items.
1310 * It starts at a high offset and removes keys until it can't find
1311 * any higher than i_size.
1313 * csum items that cross the new i_size are truncated to the new size
1316 * min_type is the minimum key type to truncate down to. If set to 0, this
1317 * will kill all the items on this inode, including the INODE_ITEM_KEY.
1319 noinline
int btrfs_truncate_inode_items(struct btrfs_trans_handle
*trans
,
1320 struct btrfs_root
*root
,
1321 struct inode
*inode
,
1322 u64 new_size
, u32 min_type
)
1325 struct btrfs_path
*path
;
1326 struct btrfs_key key
;
1327 struct btrfs_key found_key
;
1329 struct extent_buffer
*leaf
;
1330 struct btrfs_file_extent_item
*fi
;
1331 u64 extent_start
= 0;
1332 u64 extent_num_bytes
= 0;
1338 int pending_del_nr
= 0;
1339 int pending_del_slot
= 0;
1340 int extent_type
= -1;
1341 u64 mask
= root
->sectorsize
- 1;
1344 btrfs_drop_extent_cache(inode
,
1345 new_size
& (~mask
), (u64
)-1);
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_iget_locked(struct super_block
*s
, u64 objectid
,
1845 struct btrfs_root
*root
)
1847 struct inode
*inode
;
1848 struct btrfs_iget_args args
;
1849 args
.ino
= objectid
;
1852 inode
= iget5_locked(s
, objectid
, btrfs_find_actor
,
1853 btrfs_init_locked_inode
,
1858 /* Get an inode object given its location and corresponding root.
1859 * Returns in *is_new if the inode was read from disk
1861 struct inode
*btrfs_iget(struct super_block
*s
, struct btrfs_key
*location
,
1862 struct btrfs_root
*root
, int *is_new
)
1864 struct inode
*inode
;
1866 inode
= btrfs_iget_locked(s
, location
->objectid
, root
);
1868 return ERR_PTR(-EACCES
);
1870 if (inode
->i_state
& I_NEW
) {
1871 BTRFS_I(inode
)->root
= root
;
1872 memcpy(&BTRFS_I(inode
)->location
, location
, sizeof(*location
));
1873 btrfs_read_locked_inode(inode
);
1874 unlock_new_inode(inode
);
1885 static struct dentry
*btrfs_lookup(struct inode
*dir
, struct dentry
*dentry
,
1886 struct nameidata
*nd
)
1888 struct inode
* inode
;
1889 struct btrfs_inode
*bi
= BTRFS_I(dir
);
1890 struct btrfs_root
*root
= bi
->root
;
1891 struct btrfs_root
*sub_root
= root
;
1892 struct btrfs_key location
;
1893 int ret
, new, do_orphan
= 0;
1895 if (dentry
->d_name
.len
> BTRFS_NAME_LEN
)
1896 return ERR_PTR(-ENAMETOOLONG
);
1898 ret
= btrfs_inode_by_name(dir
, dentry
, &location
);
1901 return ERR_PTR(ret
);
1904 if (location
.objectid
) {
1905 ret
= fixup_tree_root_location(root
, &location
, &sub_root
,
1908 return ERR_PTR(ret
);
1910 return ERR_PTR(-ENOENT
);
1911 inode
= btrfs_iget(dir
->i_sb
, &location
, sub_root
, &new);
1913 return ERR_CAST(inode
);
1915 /* the inode and parent dir are two different roots */
1916 if (new && root
!= sub_root
) {
1918 sub_root
->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_real_readdir(struct file
*filp
, void *dirent
,
1936 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1937 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1938 struct btrfs_item
*item
;
1939 struct btrfs_dir_item
*di
;
1940 struct btrfs_key key
;
1941 struct btrfs_key found_key
;
1942 struct btrfs_path
*path
;
1945 struct extent_buffer
*leaf
;
1948 unsigned char d_type
;
1953 int key_type
= BTRFS_DIR_INDEX_KEY
;
1958 /* FIXME, use a real flag for deciding about the key type */
1959 if (root
->fs_info
->tree_root
== root
)
1960 key_type
= BTRFS_DIR_ITEM_KEY
;
1962 /* special case for "." */
1963 if (filp
->f_pos
== 0) {
1964 over
= filldir(dirent
, ".", 1,
1971 /* special case for .., just use the back ref */
1972 if (filp
->f_pos
== 1) {
1973 u64 pino
= parent_ino(filp
->f_path
.dentry
);
1974 over
= filldir(dirent
, "..", 2,
1981 path
= btrfs_alloc_path();
1984 btrfs_set_key_type(&key
, key_type
);
1985 key
.offset
= filp
->f_pos
;
1986 key
.objectid
= inode
->i_ino
;
1988 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1994 leaf
= path
->nodes
[0];
1995 nritems
= btrfs_header_nritems(leaf
);
1996 slot
= path
->slots
[0];
1997 if (advance
|| slot
>= nritems
) {
1998 if (slot
>= nritems
- 1) {
1999 ret
= btrfs_next_leaf(root
, path
);
2002 leaf
= path
->nodes
[0];
2003 nritems
= btrfs_header_nritems(leaf
);
2004 slot
= path
->slots
[0];
2011 item
= btrfs_item_nr(leaf
, slot
);
2012 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
2014 if (found_key
.objectid
!= key
.objectid
)
2016 if (btrfs_key_type(&found_key
) != key_type
)
2018 if (found_key
.offset
< filp
->f_pos
)
2021 filp
->f_pos
= found_key
.offset
;
2023 di
= btrfs_item_ptr(leaf
, slot
, struct btrfs_dir_item
);
2025 di_total
= btrfs_item_size(leaf
, item
);
2027 while (di_cur
< di_total
) {
2028 struct btrfs_key location
;
2030 name_len
= btrfs_dir_name_len(leaf
, di
);
2031 if (name_len
<= sizeof(tmp_name
)) {
2032 name_ptr
= tmp_name
;
2034 name_ptr
= kmalloc(name_len
, GFP_NOFS
);
2040 read_extent_buffer(leaf
, name_ptr
,
2041 (unsigned long)(di
+ 1), name_len
);
2043 d_type
= btrfs_filetype_table
[btrfs_dir_type(leaf
, di
)];
2044 btrfs_dir_item_key_to_cpu(leaf
, di
, &location
);
2045 over
= filldir(dirent
, name_ptr
, name_len
,
2046 found_key
.offset
, location
.objectid
,
2049 if (name_ptr
!= tmp_name
)
2055 di_len
= btrfs_dir_name_len(leaf
, di
) +
2056 btrfs_dir_data_len(leaf
, di
) + sizeof(*di
);
2058 di
= (struct btrfs_dir_item
*)((char *)di
+ di_len
);
2062 /* Reached end of directory/root. Bump pos past the last item. */
2063 if (key_type
== BTRFS_DIR_INDEX_KEY
)
2064 filp
->f_pos
= INT_LIMIT(typeof(filp
->f_pos
));
2070 btrfs_free_path(path
);
2074 /* Kernels earlier than 2.6.28 still have the NFS deadlock where nfsd
2075 will call the file system's ->lookup() method from within its
2076 filldir callback, which in turn was called from the file system's
2077 ->readdir() method. And will deadlock for many file systems. */
2078 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28)
2080 struct nfshack_dirent
{
2084 unsigned int d_type
;
2088 struct nfshack_readdir
{
2096 static int btrfs_nfshack_filldir(void *__buf
, const char *name
, int namlen
,
2097 loff_t offset
, u64 ino
, unsigned int d_type
)
2099 struct nfshack_readdir
*buf
= __buf
;
2100 struct nfshack_dirent
*de
= (void *)(buf
->dirent
+ buf
->used
);
2101 unsigned int reclen
;
2103 reclen
= ALIGN(sizeof(struct nfshack_dirent
) + namlen
, sizeof(u64
));
2104 if (buf
->used
+ reclen
> PAGE_SIZE
) {
2109 de
->namlen
= namlen
;
2110 de
->offset
= offset
;
2112 de
->d_type
= d_type
;
2113 memcpy(de
->name
, name
, namlen
);
2114 buf
->used
+= reclen
;
2119 static int btrfs_nfshack_readdir(struct file
*file
, void *dirent
,
2122 struct nfshack_readdir buf
;
2123 struct nfshack_dirent
*de
;
2128 buf
.dirent
= (void *)__get_free_page(GFP_KERNEL
);
2132 offset
= file
->f_pos
;
2135 unsigned int reclen
;
2139 err
= btrfs_real_readdir(file
, &buf
, btrfs_nfshack_filldir
);
2148 de
= (struct nfshack_dirent
*)buf
.dirent
;
2150 offset
= de
->offset
;
2152 if (filldir(dirent
, de
->name
, de
->namlen
, de
->offset
,
2153 de
->ino
, de
->d_type
))
2155 offset
= file
->f_pos
;
2157 reclen
= ALIGN(sizeof(*de
) + de
->namlen
,
2160 de
= (struct nfshack_dirent
*)((char *)de
+ reclen
);
2165 free_page((unsigned long)buf
.dirent
);
2166 file
->f_pos
= offset
;
2172 int btrfs_write_inode(struct inode
*inode
, int wait
)
2174 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2175 struct btrfs_trans_handle
*trans
;
2178 if (root
->fs_info
->closing
> 1)
2182 trans
= btrfs_join_transaction(root
, 1);
2183 btrfs_set_trans_block_group(trans
, inode
);
2184 ret
= btrfs_commit_transaction(trans
, root
);
2190 * This is somewhat expensive, updating the tree every time the
2191 * inode changes. But, it is most likely to find the inode in cache.
2192 * FIXME, needs more benchmarking...there are no reasons other than performance
2193 * to keep or drop this code.
2195 void btrfs_dirty_inode(struct inode
*inode
)
2197 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2198 struct btrfs_trans_handle
*trans
;
2200 trans
= btrfs_join_transaction(root
, 1);
2201 btrfs_set_trans_block_group(trans
, inode
);
2202 btrfs_update_inode(trans
, root
, inode
);
2203 btrfs_end_transaction(trans
, root
);
2206 static int btrfs_set_inode_index_count(struct inode
*inode
)
2208 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2209 struct btrfs_key key
, found_key
;
2210 struct btrfs_path
*path
;
2211 struct extent_buffer
*leaf
;
2214 key
.objectid
= inode
->i_ino
;
2215 btrfs_set_key_type(&key
, BTRFS_DIR_INDEX_KEY
);
2216 key
.offset
= (u64
)-1;
2218 path
= btrfs_alloc_path();
2222 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2225 /* FIXME: we should be able to handle this */
2231 * MAGIC NUMBER EXPLANATION:
2232 * since we search a directory based on f_pos we have to start at 2
2233 * since '.' and '..' have f_pos of 0 and 1 respectively, so everybody
2234 * else has to start at 2
2236 if (path
->slots
[0] == 0) {
2237 BTRFS_I(inode
)->index_cnt
= 2;
2243 leaf
= path
->nodes
[0];
2244 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2246 if (found_key
.objectid
!= inode
->i_ino
||
2247 btrfs_key_type(&found_key
) != BTRFS_DIR_INDEX_KEY
) {
2248 BTRFS_I(inode
)->index_cnt
= 2;
2252 BTRFS_I(inode
)->index_cnt
= found_key
.offset
+ 1;
2254 btrfs_free_path(path
);
2258 static int btrfs_set_inode_index(struct inode
*dir
, struct inode
*inode
,
2263 if (BTRFS_I(dir
)->index_cnt
== (u64
)-1) {
2264 ret
= btrfs_set_inode_index_count(dir
);
2270 *index
= BTRFS_I(dir
)->index_cnt
;
2271 BTRFS_I(dir
)->index_cnt
++;
2276 static struct inode
*btrfs_new_inode(struct btrfs_trans_handle
*trans
,
2277 struct btrfs_root
*root
,
2279 const char *name
, int name_len
,
2282 struct btrfs_block_group_cache
*group
,
2283 int mode
, u64
*index
)
2285 struct inode
*inode
;
2286 struct btrfs_inode_item
*inode_item
;
2287 struct btrfs_block_group_cache
*new_inode_group
;
2288 struct btrfs_key
*location
;
2289 struct btrfs_path
*path
;
2290 struct btrfs_inode_ref
*ref
;
2291 struct btrfs_key key
[2];
2297 path
= btrfs_alloc_path();
2300 inode
= new_inode(root
->fs_info
->sb
);
2302 return ERR_PTR(-ENOMEM
);
2305 ret
= btrfs_set_inode_index(dir
, inode
, index
);
2307 return ERR_PTR(ret
);
2310 * index_cnt is ignored for everything but a dir,
2311 * btrfs_get_inode_index_count has an explanation for the magic
2314 init_btrfs_i(inode
);
2315 BTRFS_I(inode
)->index_cnt
= 2;
2316 BTRFS_I(inode
)->root
= root
;
2317 BTRFS_I(inode
)->generation
= trans
->transid
;
2323 new_inode_group
= btrfs_find_block_group(root
, group
, 0,
2324 BTRFS_BLOCK_GROUP_METADATA
, owner
);
2325 if (!new_inode_group
) {
2326 printk("find_block group failed\n");
2327 new_inode_group
= group
;
2329 BTRFS_I(inode
)->block_group
= new_inode_group
;
2331 key
[0].objectid
= objectid
;
2332 btrfs_set_key_type(&key
[0], BTRFS_INODE_ITEM_KEY
);
2335 key
[1].objectid
= objectid
;
2336 btrfs_set_key_type(&key
[1], BTRFS_INODE_REF_KEY
);
2337 key
[1].offset
= ref_objectid
;
2339 sizes
[0] = sizeof(struct btrfs_inode_item
);
2340 sizes
[1] = name_len
+ sizeof(*ref
);
2342 ret
= btrfs_insert_empty_items(trans
, root
, path
, key
, sizes
, 2);
2346 if (objectid
> root
->highest_inode
)
2347 root
->highest_inode
= objectid
;
2349 inode
->i_uid
= current
->fsuid
;
2350 inode
->i_gid
= current
->fsgid
;
2351 inode
->i_mode
= mode
;
2352 inode
->i_ino
= objectid
;
2353 inode
->i_blocks
= 0;
2354 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2355 inode_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
2356 struct btrfs_inode_item
);
2357 fill_inode_item(trans
, path
->nodes
[0], inode_item
, inode
);
2359 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0] + 1,
2360 struct btrfs_inode_ref
);
2361 btrfs_set_inode_ref_name_len(path
->nodes
[0], ref
, name_len
);
2362 btrfs_set_inode_ref_index(path
->nodes
[0], ref
, *index
);
2363 ptr
= (unsigned long)(ref
+ 1);
2364 write_extent_buffer(path
->nodes
[0], name
, ptr
, name_len
);
2366 btrfs_mark_buffer_dirty(path
->nodes
[0]);
2367 btrfs_free_path(path
);
2369 location
= &BTRFS_I(inode
)->location
;
2370 location
->objectid
= objectid
;
2371 location
->offset
= 0;
2372 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
2374 insert_inode_hash(inode
);
2378 BTRFS_I(dir
)->index_cnt
--;
2379 btrfs_free_path(path
);
2380 return ERR_PTR(ret
);
2383 static inline u8
btrfs_inode_type(struct inode
*inode
)
2385 return btrfs_type_by_mode
[(inode
->i_mode
& S_IFMT
) >> S_SHIFT
];
2388 int btrfs_add_link(struct btrfs_trans_handle
*trans
,
2389 struct inode
*parent_inode
, struct inode
*inode
,
2390 const char *name
, int name_len
, int add_backref
, u64 index
)
2393 struct btrfs_key key
;
2394 struct btrfs_root
*root
= BTRFS_I(parent_inode
)->root
;
2396 key
.objectid
= inode
->i_ino
;
2397 btrfs_set_key_type(&key
, BTRFS_INODE_ITEM_KEY
);
2400 ret
= btrfs_insert_dir_item(trans
, root
, name
, name_len
,
2401 parent_inode
->i_ino
,
2402 &key
, btrfs_inode_type(inode
),
2406 ret
= btrfs_insert_inode_ref(trans
, root
,
2409 parent_inode
->i_ino
,
2412 btrfs_i_size_write(parent_inode
, parent_inode
->i_size
+
2414 parent_inode
->i_mtime
= parent_inode
->i_ctime
= CURRENT_TIME
;
2415 ret
= btrfs_update_inode(trans
, root
, parent_inode
);
2420 static int btrfs_add_nondir(struct btrfs_trans_handle
*trans
,
2421 struct dentry
*dentry
, struct inode
*inode
,
2422 int backref
, u64 index
)
2424 int err
= btrfs_add_link(trans
, dentry
->d_parent
->d_inode
,
2425 inode
, dentry
->d_name
.name
,
2426 dentry
->d_name
.len
, backref
, index
);
2428 d_instantiate(dentry
, inode
);
2436 static int btrfs_mknod(struct inode
*dir
, struct dentry
*dentry
,
2437 int mode
, dev_t rdev
)
2439 struct btrfs_trans_handle
*trans
;
2440 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2441 struct inode
*inode
= NULL
;
2445 unsigned long nr
= 0;
2448 if (!new_valid_dev(rdev
))
2451 err
= btrfs_check_free_space(root
, 1, 0);
2455 trans
= btrfs_start_transaction(root
, 1);
2456 btrfs_set_trans_block_group(trans
, dir
);
2458 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2464 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2466 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2467 BTRFS_I(dir
)->block_group
, mode
, &index
);
2468 err
= PTR_ERR(inode
);
2472 err
= btrfs_init_acl(inode
, dir
);
2478 btrfs_set_trans_block_group(trans
, inode
);
2479 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
2483 inode
->i_op
= &btrfs_special_inode_operations
;
2484 init_special_inode(inode
, inode
->i_mode
, rdev
);
2485 btrfs_update_inode(trans
, root
, inode
);
2487 dir
->i_sb
->s_dirt
= 1;
2488 btrfs_update_inode_block_group(trans
, inode
);
2489 btrfs_update_inode_block_group(trans
, dir
);
2491 nr
= trans
->blocks_used
;
2492 btrfs_end_transaction_throttle(trans
, root
);
2495 inode_dec_link_count(inode
);
2498 btrfs_btree_balance_dirty(root
, nr
);
2502 static int btrfs_create(struct inode
*dir
, struct dentry
*dentry
,
2503 int mode
, struct nameidata
*nd
)
2505 struct btrfs_trans_handle
*trans
;
2506 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2507 struct inode
*inode
= NULL
;
2510 unsigned long nr
= 0;
2514 err
= btrfs_check_free_space(root
, 1, 0);
2517 trans
= btrfs_start_transaction(root
, 1);
2518 btrfs_set_trans_block_group(trans
, dir
);
2520 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2526 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2528 dentry
->d_parent
->d_inode
->i_ino
,
2529 objectid
, BTRFS_I(dir
)->block_group
, mode
,
2531 err
= PTR_ERR(inode
);
2535 err
= btrfs_init_acl(inode
, dir
);
2541 btrfs_set_trans_block_group(trans
, inode
);
2542 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
2546 inode
->i_mapping
->a_ops
= &btrfs_aops
;
2547 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
2548 inode
->i_fop
= &btrfs_file_operations
;
2549 inode
->i_op
= &btrfs_file_inode_operations
;
2550 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
2552 dir
->i_sb
->s_dirt
= 1;
2553 btrfs_update_inode_block_group(trans
, inode
);
2554 btrfs_update_inode_block_group(trans
, dir
);
2556 nr
= trans
->blocks_used
;
2557 btrfs_end_transaction_throttle(trans
, root
);
2560 inode_dec_link_count(inode
);
2563 btrfs_btree_balance_dirty(root
, nr
);
2567 static int btrfs_link(struct dentry
*old_dentry
, struct inode
*dir
,
2568 struct dentry
*dentry
)
2570 struct btrfs_trans_handle
*trans
;
2571 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2572 struct inode
*inode
= old_dentry
->d_inode
;
2574 unsigned long nr
= 0;
2578 if (inode
->i_nlink
== 0)
2581 btrfs_inc_nlink(inode
);
2582 err
= btrfs_check_free_space(root
, 1, 0);
2585 err
= btrfs_set_inode_index(dir
, inode
, &index
);
2589 trans
= btrfs_start_transaction(root
, 1);
2591 btrfs_set_trans_block_group(trans
, dir
);
2592 atomic_inc(&inode
->i_count
);
2594 err
= btrfs_add_nondir(trans
, dentry
, inode
, 1, index
);
2599 dir
->i_sb
->s_dirt
= 1;
2600 btrfs_update_inode_block_group(trans
, dir
);
2601 err
= btrfs_update_inode(trans
, root
, inode
);
2606 nr
= trans
->blocks_used
;
2607 btrfs_end_transaction_throttle(trans
, root
);
2610 inode_dec_link_count(inode
);
2613 btrfs_btree_balance_dirty(root
, nr
);
2617 static int btrfs_mkdir(struct inode
*dir
, struct dentry
*dentry
, int mode
)
2619 struct inode
*inode
= NULL
;
2620 struct btrfs_trans_handle
*trans
;
2621 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2623 int drop_on_err
= 0;
2626 unsigned long nr
= 1;
2628 err
= btrfs_check_free_space(root
, 1, 0);
2632 trans
= btrfs_start_transaction(root
, 1);
2633 btrfs_set_trans_block_group(trans
, dir
);
2635 if (IS_ERR(trans
)) {
2636 err
= PTR_ERR(trans
);
2640 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2646 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2648 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2649 BTRFS_I(dir
)->block_group
, S_IFDIR
| mode
,
2651 if (IS_ERR(inode
)) {
2652 err
= PTR_ERR(inode
);
2658 err
= btrfs_init_acl(inode
, dir
);
2662 inode
->i_op
= &btrfs_dir_inode_operations
;
2663 inode
->i_fop
= &btrfs_dir_file_operations
;
2664 btrfs_set_trans_block_group(trans
, inode
);
2666 btrfs_i_size_write(inode
, 0);
2667 err
= btrfs_update_inode(trans
, root
, inode
);
2671 err
= btrfs_add_link(trans
, dentry
->d_parent
->d_inode
,
2672 inode
, dentry
->d_name
.name
,
2673 dentry
->d_name
.len
, 0, index
);
2677 d_instantiate(dentry
, inode
);
2679 dir
->i_sb
->s_dirt
= 1;
2680 btrfs_update_inode_block_group(trans
, inode
);
2681 btrfs_update_inode_block_group(trans
, dir
);
2684 nr
= trans
->blocks_used
;
2685 btrfs_end_transaction_throttle(trans
, root
);
2690 btrfs_btree_balance_dirty(root
, nr
);
2694 static int merge_extent_mapping(struct extent_map_tree
*em_tree
,
2695 struct extent_map
*existing
,
2696 struct extent_map
*em
,
2697 u64 map_start
, u64 map_len
)
2701 BUG_ON(map_start
< em
->start
|| map_start
>= extent_map_end(em
));
2702 start_diff
= map_start
- em
->start
;
2703 em
->start
= map_start
;
2705 if (em
->block_start
< EXTENT_MAP_LAST_BYTE
)
2706 em
->block_start
+= start_diff
;
2707 return add_extent_mapping(em_tree
, em
);
2710 struct extent_map
*btrfs_get_extent(struct inode
*inode
, struct page
*page
,
2711 size_t pg_offset
, u64 start
, u64 len
,
2717 u64 extent_start
= 0;
2719 u64 objectid
= inode
->i_ino
;
2721 struct btrfs_path
*path
= NULL
;
2722 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2723 struct btrfs_file_extent_item
*item
;
2724 struct extent_buffer
*leaf
;
2725 struct btrfs_key found_key
;
2726 struct extent_map
*em
= NULL
;
2727 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
2728 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2729 struct btrfs_trans_handle
*trans
= NULL
;
2732 spin_lock(&em_tree
->lock
);
2733 em
= lookup_extent_mapping(em_tree
, start
, len
);
2735 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2736 spin_unlock(&em_tree
->lock
);
2739 if (em
->start
> start
|| em
->start
+ em
->len
<= start
)
2740 free_extent_map(em
);
2741 else if (em
->block_start
== EXTENT_MAP_INLINE
&& page
)
2742 free_extent_map(em
);
2746 em
= alloc_extent_map(GFP_NOFS
);
2751 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2752 em
->start
= EXTENT_MAP_HOLE
;
2756 path
= btrfs_alloc_path();
2760 ret
= btrfs_lookup_file_extent(trans
, root
, path
,
2761 objectid
, start
, trans
!= NULL
);
2768 if (path
->slots
[0] == 0)
2773 leaf
= path
->nodes
[0];
2774 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
2775 struct btrfs_file_extent_item
);
2776 /* are we inside the extent that was found? */
2777 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2778 found_type
= btrfs_key_type(&found_key
);
2779 if (found_key
.objectid
!= objectid
||
2780 found_type
!= BTRFS_EXTENT_DATA_KEY
) {
2784 found_type
= btrfs_file_extent_type(leaf
, item
);
2785 extent_start
= found_key
.offset
;
2786 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
2787 extent_end
= extent_start
+
2788 btrfs_file_extent_num_bytes(leaf
, item
);
2790 if (start
< extent_start
|| start
>= extent_end
) {
2792 if (start
< extent_start
) {
2793 if (start
+ len
<= extent_start
)
2795 em
->len
= extent_end
- extent_start
;
2801 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
2803 em
->start
= extent_start
;
2804 em
->len
= extent_end
- extent_start
;
2805 em
->block_start
= EXTENT_MAP_HOLE
;
2808 bytenr
+= btrfs_file_extent_offset(leaf
, item
);
2809 em
->block_start
= bytenr
;
2810 em
->start
= extent_start
;
2811 em
->len
= extent_end
- extent_start
;
2813 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
2818 size_t extent_offset
;
2821 size
= btrfs_file_extent_inline_len(leaf
, btrfs_item_nr(leaf
,
2823 extent_end
= (extent_start
+ size
+ root
->sectorsize
- 1) &
2824 ~((u64
)root
->sectorsize
- 1);
2825 if (start
< extent_start
|| start
>= extent_end
) {
2827 if (start
< extent_start
) {
2828 if (start
+ len
<= extent_start
)
2830 em
->len
= extent_end
- extent_start
;
2836 em
->block_start
= EXTENT_MAP_INLINE
;
2839 em
->start
= extent_start
;
2844 page_start
= page_offset(page
) + pg_offset
;
2845 extent_offset
= page_start
- extent_start
;
2846 copy_size
= min_t(u64
, PAGE_CACHE_SIZE
- pg_offset
,
2847 size
- extent_offset
);
2848 em
->start
= extent_start
+ extent_offset
;
2849 em
->len
= (copy_size
+ root
->sectorsize
- 1) &
2850 ~((u64
)root
->sectorsize
- 1);
2852 ptr
= btrfs_file_extent_inline_start(item
) + extent_offset
;
2853 if (create
== 0 && !PageUptodate(page
)) {
2854 read_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2856 flush_dcache_page(page
);
2857 } else if (create
&& PageUptodate(page
)) {
2860 free_extent_map(em
);
2862 btrfs_release_path(root
, path
);
2863 trans
= btrfs_join_transaction(root
, 1);
2866 write_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2868 btrfs_mark_buffer_dirty(leaf
);
2871 set_extent_uptodate(io_tree
, em
->start
,
2872 extent_map_end(em
) - 1, GFP_NOFS
);
2875 printk("unkknown found_type %d\n", found_type
);
2882 em
->block_start
= EXTENT_MAP_HOLE
;
2884 btrfs_release_path(root
, path
);
2885 if (em
->start
> start
|| extent_map_end(em
) <= start
) {
2886 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em
->start
, em
->len
, start
, len
);
2892 spin_lock(&em_tree
->lock
);
2893 ret
= add_extent_mapping(em_tree
, em
);
2894 /* it is possible that someone inserted the extent into the tree
2895 * while we had the lock dropped. It is also possible that
2896 * an overlapping map exists in the tree
2898 if (ret
== -EEXIST
) {
2899 struct extent_map
*existing
;
2903 existing
= lookup_extent_mapping(em_tree
, start
, len
);
2904 if (existing
&& (existing
->start
> start
||
2905 existing
->start
+ existing
->len
<= start
)) {
2906 free_extent_map(existing
);
2910 existing
= lookup_extent_mapping(em_tree
, em
->start
,
2913 err
= merge_extent_mapping(em_tree
, existing
,
2916 free_extent_map(existing
);
2918 free_extent_map(em
);
2923 printk("failing to insert %Lu %Lu\n",
2925 free_extent_map(em
);
2929 free_extent_map(em
);
2934 spin_unlock(&em_tree
->lock
);
2937 btrfs_free_path(path
);
2939 ret
= btrfs_end_transaction(trans
, root
);
2945 free_extent_map(em
);
2947 return ERR_PTR(err
);
2952 #if 0 /* waiting for O_DIRECT reads */
2953 static int btrfs_get_block(struct inode
*inode
, sector_t iblock
,
2954 struct buffer_head
*bh_result
, int create
)
2956 struct extent_map
*em
;
2957 u64 start
= (u64
)iblock
<< inode
->i_blkbits
;
2958 struct btrfs_multi_bio
*multi
= NULL
;
2959 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2965 em
= btrfs_get_extent(inode
, NULL
, 0, start
, bh_result
->b_size
, 0);
2967 if (!em
|| IS_ERR(em
))
2970 if (em
->start
> start
|| em
->start
+ em
->len
<= start
) {
2974 if (em
->block_start
== EXTENT_MAP_INLINE
) {
2979 len
= em
->start
+ em
->len
- start
;
2980 len
= min_t(u64
, len
, INT_LIMIT(typeof(bh_result
->b_size
)));
2982 if (em
->block_start
== EXTENT_MAP_HOLE
||
2983 em
->block_start
== EXTENT_MAP_DELALLOC
) {
2984 bh_result
->b_size
= len
;
2988 logical
= start
- em
->start
;
2989 logical
= em
->block_start
+ logical
;
2992 ret
= btrfs_map_block(&root
->fs_info
->mapping_tree
, READ
,
2993 logical
, &map_length
, &multi
, 0);
2995 bh_result
->b_blocknr
= multi
->stripes
[0].physical
>> inode
->i_blkbits
;
2996 bh_result
->b_size
= min(map_length
, len
);
2998 bh_result
->b_bdev
= multi
->stripes
[0].dev
->bdev
;
2999 set_buffer_mapped(bh_result
);
3002 free_extent_map(em
);
3007 static ssize_t
btrfs_direct_IO(int rw
, struct kiocb
*iocb
,
3008 const struct iovec
*iov
, loff_t offset
,
3009 unsigned long nr_segs
)
3013 struct file
*file
= iocb
->ki_filp
;
3014 struct inode
*inode
= file
->f_mapping
->host
;
3019 return blockdev_direct_IO(rw
, iocb
, inode
, inode
->i_sb
->s_bdev
, iov
,
3020 offset
, nr_segs
, btrfs_get_block
, NULL
);
3024 static sector_t
btrfs_bmap(struct address_space
*mapping
, sector_t iblock
)
3026 return extent_bmap(mapping
, iblock
, btrfs_get_extent
);
3029 int btrfs_readpage(struct file
*file
, struct page
*page
)
3031 struct extent_io_tree
*tree
;
3032 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3033 return extent_read_full_page(tree
, page
, btrfs_get_extent
);
3036 static int btrfs_writepage(struct page
*page
, struct writeback_control
*wbc
)
3038 struct extent_io_tree
*tree
;
3041 if (current
->flags
& PF_MEMALLOC
) {
3042 redirty_page_for_writepage(wbc
, page
);
3046 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3047 return extent_write_full_page(tree
, page
, btrfs_get_extent
, wbc
);
3050 int btrfs_writepages(struct address_space
*mapping
,
3051 struct writeback_control
*wbc
)
3053 struct extent_io_tree
*tree
;
3054 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
3055 return extent_writepages(tree
, mapping
, btrfs_get_extent
, wbc
);
3059 btrfs_readpages(struct file
*file
, struct address_space
*mapping
,
3060 struct list_head
*pages
, unsigned nr_pages
)
3062 struct extent_io_tree
*tree
;
3063 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
3064 return extent_readpages(tree
, mapping
, pages
, nr_pages
,
3067 static int __btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
3069 struct extent_io_tree
*tree
;
3070 struct extent_map_tree
*map
;
3073 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3074 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
3075 ret
= try_release_extent_mapping(map
, tree
, page
, gfp_flags
);
3077 ClearPagePrivate(page
);
3078 set_page_private(page
, 0);
3079 page_cache_release(page
);
3084 static int btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
3086 if (PageWriteback(page
) || PageDirty(page
))
3088 return __btrfs_releasepage(page
, gfp_flags
);
3091 static void btrfs_invalidatepage(struct page
*page
, unsigned long offset
)
3093 struct extent_io_tree
*tree
;
3094 struct btrfs_ordered_extent
*ordered
;
3095 u64 page_start
= page_offset(page
);
3096 u64 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
3098 wait_on_page_writeback(page
);
3099 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3101 btrfs_releasepage(page
, GFP_NOFS
);
3105 lock_extent(tree
, page_start
, page_end
, GFP_NOFS
);
3106 ordered
= btrfs_lookup_ordered_extent(page
->mapping
->host
,
3110 * IO on this page will never be started, so we need
3111 * to account for any ordered extents now
3113 clear_extent_bit(tree
, page_start
, page_end
,
3114 EXTENT_DIRTY
| EXTENT_DELALLOC
|
3115 EXTENT_LOCKED
, 1, 0, GFP_NOFS
);
3116 btrfs_finish_ordered_io(page
->mapping
->host
,
3117 page_start
, page_end
);
3118 btrfs_put_ordered_extent(ordered
);
3119 lock_extent(tree
, page_start
, page_end
, GFP_NOFS
);
3121 clear_extent_bit(tree
, page_start
, page_end
,
3122 EXTENT_LOCKED
| EXTENT_DIRTY
| EXTENT_DELALLOC
|
3125 __btrfs_releasepage(page
, GFP_NOFS
);
3127 ClearPageChecked(page
);
3128 if (PagePrivate(page
)) {
3129 ClearPagePrivate(page
);
3130 set_page_private(page
, 0);
3131 page_cache_release(page
);
3136 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
3137 * called from a page fault handler when a page is first dirtied. Hence we must
3138 * be careful to check for EOF conditions here. We set the page up correctly
3139 * for a written page which means we get ENOSPC checking when writing into
3140 * holes and correct delalloc and unwritten extent mapping on filesystems that
3141 * support these features.
3143 * We are not allowed to take the i_mutex here so we have to play games to
3144 * protect against truncate races as the page could now be beyond EOF. Because
3145 * vmtruncate() writes the inode size before removing pages, once we have the
3146 * page lock we can determine safely if the page is beyond EOF. If it is not
3147 * beyond EOF, then the page is guaranteed safe against truncation until we
3150 int btrfs_page_mkwrite(struct vm_area_struct
*vma
, struct page
*page
)
3152 struct inode
*inode
= fdentry(vma
->vm_file
)->d_inode
;
3153 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3154 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
3155 struct btrfs_ordered_extent
*ordered
;
3157 unsigned long zero_start
;
3163 ret
= btrfs_check_free_space(root
, PAGE_CACHE_SIZE
, 0);
3170 size
= i_size_read(inode
);
3171 page_start
= page_offset(page
);
3172 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
3174 if ((page
->mapping
!= inode
->i_mapping
) ||
3175 (page_start
>= size
)) {
3176 /* page got truncated out from underneath us */
3179 wait_on_page_writeback(page
);
3181 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3182 set_page_extent_mapped(page
);
3185 * we can't set the delalloc bits if there are pending ordered
3186 * extents. Drop our locks and wait for them to finish
3188 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
3190 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3192 btrfs_start_ordered_extent(inode
, ordered
, 1);
3193 btrfs_put_ordered_extent(ordered
);
3197 btrfs_set_extent_delalloc(inode
, page_start
, page_end
);
3200 /* page is wholly or partially inside EOF */
3201 if (page_start
+ PAGE_CACHE_SIZE
> size
)
3202 zero_start
= size
& ~PAGE_CACHE_MASK
;
3204 zero_start
= PAGE_CACHE_SIZE
;
3206 if (zero_start
!= PAGE_CACHE_SIZE
) {
3208 memset(kaddr
+ zero_start
, 0, PAGE_CACHE_SIZE
- zero_start
);
3209 flush_dcache_page(page
);
3212 ClearPageChecked(page
);
3213 set_page_dirty(page
);
3214 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3222 static void btrfs_truncate(struct inode
*inode
)
3224 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3226 struct btrfs_trans_handle
*trans
;
3228 u64 mask
= root
->sectorsize
- 1;
3230 if (!S_ISREG(inode
->i_mode
))
3232 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
3235 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
3236 btrfs_wait_ordered_range(inode
, inode
->i_size
& (~mask
), (u64
)-1);
3238 trans
= btrfs_start_transaction(root
, 1);
3239 btrfs_set_trans_block_group(trans
, inode
);
3240 btrfs_i_size_write(inode
, inode
->i_size
);
3242 ret
= btrfs_orphan_add(trans
, inode
);
3245 /* FIXME, add redo link to tree so we don't leak on crash */
3246 ret
= btrfs_truncate_inode_items(trans
, root
, inode
, inode
->i_size
,
3247 BTRFS_EXTENT_DATA_KEY
);
3248 btrfs_update_inode(trans
, root
, inode
);
3250 ret
= btrfs_orphan_del(trans
, inode
);
3254 nr
= trans
->blocks_used
;
3255 ret
= btrfs_end_transaction_throttle(trans
, root
);
3257 btrfs_btree_balance_dirty(root
, nr
);
3261 * Invalidate a single dcache entry at the root of the filesystem.
3262 * Needed after creation of snapshot or subvolume.
3264 void btrfs_invalidate_dcache_root(struct btrfs_root
*root
, char *name
,
3267 struct dentry
*alias
, *entry
;
3270 alias
= d_find_alias(root
->fs_info
->sb
->s_root
->d_inode
);
3274 /* change me if btrfs ever gets a d_hash operation */
3275 qstr
.hash
= full_name_hash(qstr
.name
, qstr
.len
);
3276 entry
= d_lookup(alias
, &qstr
);
3279 d_invalidate(entry
);
3285 int btrfs_create_subvol_root(struct btrfs_root
*new_root
,
3286 struct btrfs_trans_handle
*trans
, u64 new_dirid
,
3287 struct btrfs_block_group_cache
*block_group
)
3289 struct inode
*inode
;
3292 inode
= btrfs_new_inode(trans
, new_root
, NULL
, "..", 2, new_dirid
,
3293 new_dirid
, block_group
, S_IFDIR
| 0700, &index
);
3295 return PTR_ERR(inode
);
3296 inode
->i_op
= &btrfs_dir_inode_operations
;
3297 inode
->i_fop
= &btrfs_dir_file_operations
;
3298 new_root
->inode
= inode
;
3301 btrfs_i_size_write(inode
, 0);
3303 return btrfs_update_inode(trans
, new_root
, inode
);
3306 unsigned long btrfs_force_ra(struct address_space
*mapping
,
3307 struct file_ra_state
*ra
, struct file
*file
,
3308 pgoff_t offset
, pgoff_t last_index
)
3310 pgoff_t req_size
= last_index
- offset
+ 1;
3312 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3313 offset
= page_cache_readahead(mapping
, ra
, file
, offset
, req_size
);
3316 page_cache_sync_readahead(mapping
, ra
, file
, offset
, req_size
);
3317 return offset
+ req_size
;
3321 struct inode
*btrfs_alloc_inode(struct super_block
*sb
)
3323 struct btrfs_inode
*ei
;
3325 ei
= kmem_cache_alloc(btrfs_inode_cachep
, GFP_NOFS
);
3329 ei
->logged_trans
= 0;
3330 btrfs_ordered_inode_tree_init(&ei
->ordered_tree
);
3331 ei
->i_acl
= BTRFS_ACL_NOT_CACHED
;
3332 ei
->i_default_acl
= BTRFS_ACL_NOT_CACHED
;
3333 INIT_LIST_HEAD(&ei
->i_orphan
);
3334 return &ei
->vfs_inode
;
3337 void btrfs_destroy_inode(struct inode
*inode
)
3339 struct btrfs_ordered_extent
*ordered
;
3340 WARN_ON(!list_empty(&inode
->i_dentry
));
3341 WARN_ON(inode
->i_data
.nrpages
);
3343 if (BTRFS_I(inode
)->i_acl
&&
3344 BTRFS_I(inode
)->i_acl
!= BTRFS_ACL_NOT_CACHED
)
3345 posix_acl_release(BTRFS_I(inode
)->i_acl
);
3346 if (BTRFS_I(inode
)->i_default_acl
&&
3347 BTRFS_I(inode
)->i_default_acl
!= BTRFS_ACL_NOT_CACHED
)
3348 posix_acl_release(BTRFS_I(inode
)->i_default_acl
);
3350 spin_lock(&BTRFS_I(inode
)->root
->list_lock
);
3351 if (!list_empty(&BTRFS_I(inode
)->i_orphan
)) {
3352 printk(KERN_ERR
"BTRFS: inode %lu: inode still on the orphan"
3353 " list\n", inode
->i_ino
);
3356 spin_unlock(&BTRFS_I(inode
)->root
->list_lock
);
3359 ordered
= btrfs_lookup_first_ordered_extent(inode
, (u64
)-1);
3363 printk("found ordered extent %Lu %Lu\n",
3364 ordered
->file_offset
, ordered
->len
);
3365 btrfs_remove_ordered_extent(inode
, ordered
);
3366 btrfs_put_ordered_extent(ordered
);
3367 btrfs_put_ordered_extent(ordered
);
3370 btrfs_drop_extent_cache(inode
, 0, (u64
)-1);
3371 kmem_cache_free(btrfs_inode_cachep
, BTRFS_I(inode
));
3374 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3375 static void init_once(void *foo
)
3376 #elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3377 static void init_once(struct kmem_cache
* cachep
, void *foo
)
3379 static void init_once(void * foo
, struct kmem_cache
* cachep
,
3380 unsigned long flags
)
3383 struct btrfs_inode
*ei
= (struct btrfs_inode
*) foo
;
3385 inode_init_once(&ei
->vfs_inode
);
3388 void btrfs_destroy_cachep(void)
3390 if (btrfs_inode_cachep
)
3391 kmem_cache_destroy(btrfs_inode_cachep
);
3392 if (btrfs_trans_handle_cachep
)
3393 kmem_cache_destroy(btrfs_trans_handle_cachep
);
3394 if (btrfs_transaction_cachep
)
3395 kmem_cache_destroy(btrfs_transaction_cachep
);
3396 if (btrfs_bit_radix_cachep
)
3397 kmem_cache_destroy(btrfs_bit_radix_cachep
);
3398 if (btrfs_path_cachep
)
3399 kmem_cache_destroy(btrfs_path_cachep
);
3402 struct kmem_cache
*btrfs_cache_create(const char *name
, size_t size
,
3403 unsigned long extra_flags
,
3404 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3405 void (*ctor
)(void *)
3406 #elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3407 void (*ctor
)(struct kmem_cache
*, void *)
3409 void (*ctor
)(void *, struct kmem_cache
*,
3414 return kmem_cache_create(name
, size
, 0, (SLAB_RECLAIM_ACCOUNT
|
3415 SLAB_MEM_SPREAD
| extra_flags
), ctor
3416 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3422 int btrfs_init_cachep(void)
3424 btrfs_inode_cachep
= btrfs_cache_create("btrfs_inode_cache",
3425 sizeof(struct btrfs_inode
),
3427 if (!btrfs_inode_cachep
)
3429 btrfs_trans_handle_cachep
=
3430 btrfs_cache_create("btrfs_trans_handle_cache",
3431 sizeof(struct btrfs_trans_handle
),
3433 if (!btrfs_trans_handle_cachep
)
3435 btrfs_transaction_cachep
= btrfs_cache_create("btrfs_transaction_cache",
3436 sizeof(struct btrfs_transaction
),
3438 if (!btrfs_transaction_cachep
)
3440 btrfs_path_cachep
= btrfs_cache_create("btrfs_path_cache",
3441 sizeof(struct btrfs_path
),
3443 if (!btrfs_path_cachep
)
3445 btrfs_bit_radix_cachep
= btrfs_cache_create("btrfs_radix", 256,
3446 SLAB_DESTROY_BY_RCU
, NULL
);
3447 if (!btrfs_bit_radix_cachep
)
3451 btrfs_destroy_cachep();
3455 static int btrfs_getattr(struct vfsmount
*mnt
,
3456 struct dentry
*dentry
, struct kstat
*stat
)
3458 struct inode
*inode
= dentry
->d_inode
;
3459 generic_fillattr(inode
, stat
);
3460 stat
->blksize
= PAGE_CACHE_SIZE
;
3461 stat
->blocks
= inode
->i_blocks
+ (BTRFS_I(inode
)->delalloc_bytes
>> 9);
3465 static int btrfs_rename(struct inode
* old_dir
, struct dentry
*old_dentry
,
3466 struct inode
* new_dir
,struct dentry
*new_dentry
)
3468 struct btrfs_trans_handle
*trans
;
3469 struct btrfs_root
*root
= BTRFS_I(old_dir
)->root
;
3470 struct inode
*new_inode
= new_dentry
->d_inode
;
3471 struct inode
*old_inode
= old_dentry
->d_inode
;
3472 struct timespec ctime
= CURRENT_TIME
;
3476 if (S_ISDIR(old_inode
->i_mode
) && new_inode
&&
3477 new_inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
3481 ret
= btrfs_check_free_space(root
, 1, 0);
3485 trans
= btrfs_start_transaction(root
, 1);
3487 btrfs_set_trans_block_group(trans
, new_dir
);
3489 btrfs_inc_nlink(old_dentry
->d_inode
);
3490 old_dir
->i_ctime
= old_dir
->i_mtime
= ctime
;
3491 new_dir
->i_ctime
= new_dir
->i_mtime
= ctime
;
3492 old_inode
->i_ctime
= ctime
;
3494 ret
= btrfs_unlink_inode(trans
, root
, old_dir
, old_dentry
->d_inode
,
3495 old_dentry
->d_name
.name
,
3496 old_dentry
->d_name
.len
);
3501 new_inode
->i_ctime
= CURRENT_TIME
;
3502 ret
= btrfs_unlink_inode(trans
, root
, new_dir
,
3503 new_dentry
->d_inode
,
3504 new_dentry
->d_name
.name
,
3505 new_dentry
->d_name
.len
);
3508 if (new_inode
->i_nlink
== 0) {
3509 ret
= btrfs_orphan_add(trans
, new_dentry
->d_inode
);
3515 ret
= btrfs_set_inode_index(new_dir
, old_inode
, &index
);
3519 ret
= btrfs_add_link(trans
, new_dentry
->d_parent
->d_inode
,
3520 old_inode
, new_dentry
->d_name
.name
,
3521 new_dentry
->d_name
.len
, 1, index
);
3526 btrfs_end_transaction_throttle(trans
, root
);
3531 int btrfs_start_delalloc_inodes(struct btrfs_root
*root
)
3533 struct list_head
*head
= &root
->fs_info
->delalloc_inodes
;
3534 struct btrfs_inode
*binode
;
3535 unsigned long flags
;
3537 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
3538 while(!list_empty(head
)) {
3539 binode
= list_entry(head
->next
, struct btrfs_inode
,
3541 atomic_inc(&binode
->vfs_inode
.i_count
);
3542 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
3543 filemap_write_and_wait(binode
->vfs_inode
.i_mapping
);
3544 iput(&binode
->vfs_inode
);
3545 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
3547 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
3551 static int btrfs_symlink(struct inode
*dir
, struct dentry
*dentry
,
3552 const char *symname
)
3554 struct btrfs_trans_handle
*trans
;
3555 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
3556 struct btrfs_path
*path
;
3557 struct btrfs_key key
;
3558 struct inode
*inode
= NULL
;
3566 struct btrfs_file_extent_item
*ei
;
3567 struct extent_buffer
*leaf
;
3568 unsigned long nr
= 0;
3570 name_len
= strlen(symname
) + 1;
3571 if (name_len
> BTRFS_MAX_INLINE_DATA_SIZE(root
))
3572 return -ENAMETOOLONG
;
3574 err
= btrfs_check_free_space(root
, 1, 0);
3578 trans
= btrfs_start_transaction(root
, 1);
3579 btrfs_set_trans_block_group(trans
, dir
);
3581 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
3587 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
3589 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
3590 BTRFS_I(dir
)->block_group
, S_IFLNK
|S_IRWXUGO
,
3592 err
= PTR_ERR(inode
);
3596 err
= btrfs_init_acl(inode
, dir
);
3602 btrfs_set_trans_block_group(trans
, inode
);
3603 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
3607 inode
->i_mapping
->a_ops
= &btrfs_aops
;
3608 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3609 inode
->i_fop
= &btrfs_file_operations
;
3610 inode
->i_op
= &btrfs_file_inode_operations
;
3611 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
3613 dir
->i_sb
->s_dirt
= 1;
3614 btrfs_update_inode_block_group(trans
, inode
);
3615 btrfs_update_inode_block_group(trans
, dir
);
3619 path
= btrfs_alloc_path();
3621 key
.objectid
= inode
->i_ino
;
3623 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
3624 datasize
= btrfs_file_extent_calc_inline_size(name_len
);
3625 err
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
3631 leaf
= path
->nodes
[0];
3632 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
3633 struct btrfs_file_extent_item
);
3634 btrfs_set_file_extent_generation(leaf
, ei
, trans
->transid
);
3635 btrfs_set_file_extent_type(leaf
, ei
,
3636 BTRFS_FILE_EXTENT_INLINE
);
3637 ptr
= btrfs_file_extent_inline_start(ei
);
3638 write_extent_buffer(leaf
, symname
, ptr
, name_len
);
3639 btrfs_mark_buffer_dirty(leaf
);
3640 btrfs_free_path(path
);
3642 inode
->i_op
= &btrfs_symlink_inode_operations
;
3643 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
3644 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3645 btrfs_i_size_write(inode
, name_len
- 1);
3646 err
= btrfs_update_inode(trans
, root
, inode
);
3651 nr
= trans
->blocks_used
;
3652 btrfs_end_transaction_throttle(trans
, root
);
3655 inode_dec_link_count(inode
);
3658 btrfs_btree_balance_dirty(root
, nr
);
3662 static int btrfs_set_page_dirty(struct page
*page
)
3664 return __set_page_dirty_nobuffers(page
);
3667 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3668 static int btrfs_permission(struct inode
*inode
, int mask
)
3670 static int btrfs_permission(struct inode
*inode
, int mask
,
3671 struct nameidata
*nd
)
3674 if (btrfs_test_flag(inode
, READONLY
) && (mask
& MAY_WRITE
))
3676 return generic_permission(inode
, mask
, btrfs_check_acl
);
3679 static struct inode_operations btrfs_dir_inode_operations
= {
3680 .lookup
= btrfs_lookup
,
3681 .create
= btrfs_create
,
3682 .unlink
= btrfs_unlink
,
3684 .mkdir
= btrfs_mkdir
,
3685 .rmdir
= btrfs_rmdir
,
3686 .rename
= btrfs_rename
,
3687 .symlink
= btrfs_symlink
,
3688 .setattr
= btrfs_setattr
,
3689 .mknod
= btrfs_mknod
,
3690 .setxattr
= btrfs_setxattr
,
3691 .getxattr
= btrfs_getxattr
,
3692 .listxattr
= btrfs_listxattr
,
3693 .removexattr
= btrfs_removexattr
,
3694 .permission
= btrfs_permission
,
3696 static struct inode_operations btrfs_dir_ro_inode_operations
= {
3697 .lookup
= btrfs_lookup
,
3698 .permission
= btrfs_permission
,
3700 static struct file_operations btrfs_dir_file_operations
= {
3701 .llseek
= generic_file_llseek
,
3702 .read
= generic_read_dir
,
3703 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28)
3704 .readdir
= btrfs_nfshack_readdir
,
3705 #else /* NFSd readdir/lookup deadlock is fixed */
3706 .readdir
= btrfs_real_readdir
,
3708 .unlocked_ioctl
= btrfs_ioctl
,
3709 #ifdef CONFIG_COMPAT
3710 .compat_ioctl
= btrfs_ioctl
,
3712 .release
= btrfs_release_file
,
3713 .fsync
= btrfs_sync_file
,
3716 static struct extent_io_ops btrfs_extent_io_ops
= {
3717 .fill_delalloc
= run_delalloc_range
,
3718 .submit_bio_hook
= btrfs_submit_bio_hook
,
3719 .merge_bio_hook
= btrfs_merge_bio_hook
,
3720 .readpage_end_io_hook
= btrfs_readpage_end_io_hook
,
3721 .writepage_end_io_hook
= btrfs_writepage_end_io_hook
,
3722 .writepage_start_hook
= btrfs_writepage_start_hook
,
3723 .readpage_io_failed_hook
= btrfs_io_failed_hook
,
3724 .set_bit_hook
= btrfs_set_bit_hook
,
3725 .clear_bit_hook
= btrfs_clear_bit_hook
,
3728 static struct address_space_operations btrfs_aops
= {
3729 .readpage
= btrfs_readpage
,
3730 .writepage
= btrfs_writepage
,
3731 .writepages
= btrfs_writepages
,
3732 .readpages
= btrfs_readpages
,
3733 .sync_page
= block_sync_page
,
3735 .direct_IO
= btrfs_direct_IO
,
3736 .invalidatepage
= btrfs_invalidatepage
,
3737 .releasepage
= btrfs_releasepage
,
3738 .set_page_dirty
= btrfs_set_page_dirty
,
3741 static struct address_space_operations btrfs_symlink_aops
= {
3742 .readpage
= btrfs_readpage
,
3743 .writepage
= btrfs_writepage
,
3744 .invalidatepage
= btrfs_invalidatepage
,
3745 .releasepage
= btrfs_releasepage
,
3748 static struct inode_operations btrfs_file_inode_operations
= {
3749 .truncate
= btrfs_truncate
,
3750 .getattr
= btrfs_getattr
,
3751 .setattr
= btrfs_setattr
,
3752 .setxattr
= btrfs_setxattr
,
3753 .getxattr
= btrfs_getxattr
,
3754 .listxattr
= btrfs_listxattr
,
3755 .removexattr
= btrfs_removexattr
,
3756 .permission
= btrfs_permission
,
3758 static struct inode_operations btrfs_special_inode_operations
= {
3759 .getattr
= btrfs_getattr
,
3760 .setattr
= btrfs_setattr
,
3761 .permission
= btrfs_permission
,
3762 .setxattr
= btrfs_setxattr
,
3763 .getxattr
= btrfs_getxattr
,
3764 .listxattr
= btrfs_listxattr
,
3765 .removexattr
= btrfs_removexattr
,
3767 static struct inode_operations btrfs_symlink_inode_operations
= {
3768 .readlink
= generic_readlink
,
3769 .follow_link
= page_follow_link_light
,
3770 .put_link
= page_put_link
,
3771 .permission
= btrfs_permission
,