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
;
532 struct list_head list
;
533 struct btrfs_key ins
;
536 ret
= btrfs_dec_test_ordered_pending(inode
, start
, end
- start
+ 1);
540 trans
= btrfs_join_transaction(root
, 1);
542 ordered_extent
= btrfs_lookup_ordered_extent(inode
, start
);
543 BUG_ON(!ordered_extent
);
544 if (test_bit(BTRFS_ORDERED_NOCOW
, &ordered_extent
->flags
))
547 lock_extent(io_tree
, ordered_extent
->file_offset
,
548 ordered_extent
->file_offset
+ ordered_extent
->len
- 1,
551 INIT_LIST_HEAD(&list
);
553 ins
.objectid
= ordered_extent
->start
;
554 ins
.offset
= ordered_extent
->len
;
555 ins
.type
= BTRFS_EXTENT_ITEM_KEY
;
557 ret
= btrfs_alloc_reserved_extent(trans
, root
, root
->root_key
.objectid
,
558 trans
->transid
, inode
->i_ino
,
559 ordered_extent
->file_offset
, &ins
);
561 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
563 ret
= btrfs_drop_extents(trans
, root
, inode
,
564 ordered_extent
->file_offset
,
565 ordered_extent
->file_offset
+
567 ordered_extent
->file_offset
, &alloc_hint
);
569 ret
= btrfs_insert_file_extent(trans
, root
, inode
->i_ino
,
570 ordered_extent
->file_offset
,
571 ordered_extent
->start
,
573 ordered_extent
->len
, 0);
576 btrfs_drop_extent_cache(inode
, ordered_extent
->file_offset
,
577 ordered_extent
->file_offset
+
578 ordered_extent
->len
- 1);
579 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
581 inode
->i_blocks
+= ordered_extent
->len
>> 9;
582 unlock_extent(io_tree
, ordered_extent
->file_offset
,
583 ordered_extent
->file_offset
+ ordered_extent
->len
- 1,
586 add_pending_csums(trans
, inode
, ordered_extent
->file_offset
,
587 &ordered_extent
->list
);
589 btrfs_ordered_update_i_size(inode
, ordered_extent
);
590 btrfs_update_inode(trans
, root
, inode
);
591 btrfs_remove_ordered_extent(inode
, ordered_extent
);
594 btrfs_put_ordered_extent(ordered_extent
);
595 /* once for the tree */
596 btrfs_put_ordered_extent(ordered_extent
);
598 btrfs_end_transaction(trans
, root
);
602 int btrfs_writepage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
603 struct extent_state
*state
, int uptodate
)
605 return btrfs_finish_ordered_io(page
->mapping
->host
, start
, end
);
608 struct io_failure_record
{
616 int btrfs_io_failed_hook(struct bio
*failed_bio
,
617 struct page
*page
, u64 start
, u64 end
,
618 struct extent_state
*state
)
620 struct io_failure_record
*failrec
= NULL
;
622 struct extent_map
*em
;
623 struct inode
*inode
= page
->mapping
->host
;
624 struct extent_io_tree
*failure_tree
= &BTRFS_I(inode
)->io_failure_tree
;
625 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
632 ret
= get_state_private(failure_tree
, start
, &private);
634 failrec
= kmalloc(sizeof(*failrec
), GFP_NOFS
);
637 failrec
->start
= start
;
638 failrec
->len
= end
- start
+ 1;
639 failrec
->last_mirror
= 0;
641 spin_lock(&em_tree
->lock
);
642 em
= lookup_extent_mapping(em_tree
, start
, failrec
->len
);
643 if (em
->start
> start
|| em
->start
+ em
->len
< start
) {
647 spin_unlock(&em_tree
->lock
);
649 if (!em
|| IS_ERR(em
)) {
653 logical
= start
- em
->start
;
654 logical
= em
->block_start
+ logical
;
655 failrec
->logical
= logical
;
657 set_extent_bits(failure_tree
, start
, end
, EXTENT_LOCKED
|
658 EXTENT_DIRTY
, GFP_NOFS
);
659 set_state_private(failure_tree
, start
,
660 (u64
)(unsigned long)failrec
);
662 failrec
= (struct io_failure_record
*)(unsigned long)private;
664 num_copies
= btrfs_num_copies(
665 &BTRFS_I(inode
)->root
->fs_info
->mapping_tree
,
666 failrec
->logical
, failrec
->len
);
667 failrec
->last_mirror
++;
669 spin_lock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
670 state
= find_first_extent_bit_state(&BTRFS_I(inode
)->io_tree
,
673 if (state
&& state
->start
!= failrec
->start
)
675 spin_unlock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
677 if (!state
|| failrec
->last_mirror
> num_copies
) {
678 set_state_private(failure_tree
, failrec
->start
, 0);
679 clear_extent_bits(failure_tree
, failrec
->start
,
680 failrec
->start
+ failrec
->len
- 1,
681 EXTENT_LOCKED
| EXTENT_DIRTY
, GFP_NOFS
);
685 bio
= bio_alloc(GFP_NOFS
, 1);
686 bio
->bi_private
= state
;
687 bio
->bi_end_io
= failed_bio
->bi_end_io
;
688 bio
->bi_sector
= failrec
->logical
>> 9;
689 bio
->bi_bdev
= failed_bio
->bi_bdev
;
691 bio_add_page(bio
, page
, failrec
->len
, start
- page_offset(page
));
692 if (failed_bio
->bi_rw
& (1 << BIO_RW
))
697 BTRFS_I(inode
)->io_tree
.ops
->submit_bio_hook(inode
, rw
, bio
,
698 failrec
->last_mirror
);
702 int btrfs_clean_io_failures(struct inode
*inode
, u64 start
)
706 struct io_failure_record
*failure
;
710 if (count_range_bits(&BTRFS_I(inode
)->io_failure_tree
, &private,
711 (u64
)-1, 1, EXTENT_DIRTY
)) {
712 ret
= get_state_private(&BTRFS_I(inode
)->io_failure_tree
,
713 start
, &private_failure
);
715 failure
= (struct io_failure_record
*)(unsigned long)
717 set_state_private(&BTRFS_I(inode
)->io_failure_tree
,
719 clear_extent_bits(&BTRFS_I(inode
)->io_failure_tree
,
721 failure
->start
+ failure
->len
- 1,
722 EXTENT_DIRTY
| EXTENT_LOCKED
,
730 int btrfs_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
731 struct extent_state
*state
)
733 size_t offset
= start
- ((u64
)page
->index
<< PAGE_CACHE_SHIFT
);
734 struct inode
*inode
= page
->mapping
->host
;
735 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
737 u64
private = ~(u32
)0;
739 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
743 if (btrfs_test_opt(root
, NODATASUM
) ||
744 btrfs_test_flag(inode
, NODATASUM
))
746 if (state
&& state
->start
== start
) {
747 private = state
->private;
750 ret
= get_state_private(io_tree
, start
, &private);
752 local_irq_save(flags
);
753 kaddr
= kmap_atomic(page
, KM_IRQ0
);
757 csum
= btrfs_csum_data(root
, kaddr
+ offset
, csum
, end
- start
+ 1);
758 btrfs_csum_final(csum
, (char *)&csum
);
759 if (csum
!= private) {
762 kunmap_atomic(kaddr
, KM_IRQ0
);
763 local_irq_restore(flags
);
765 /* if the io failure tree for this inode is non-empty,
766 * check to see if we've recovered from a failed IO
768 btrfs_clean_io_failures(inode
, start
);
772 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
773 page
->mapping
->host
->i_ino
, (unsigned long long)start
, csum
,
775 memset(kaddr
+ offset
, 1, end
- start
+ 1);
776 flush_dcache_page(page
);
777 kunmap_atomic(kaddr
, KM_IRQ0
);
778 local_irq_restore(flags
);
785 * This creates an orphan entry for the given inode in case something goes
786 * wrong in the middle of an unlink/truncate.
788 int btrfs_orphan_add(struct btrfs_trans_handle
*trans
, struct inode
*inode
)
790 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
793 spin_lock(&root
->list_lock
);
795 /* already on the orphan list, we're good */
796 if (!list_empty(&BTRFS_I(inode
)->i_orphan
)) {
797 spin_unlock(&root
->list_lock
);
801 list_add(&BTRFS_I(inode
)->i_orphan
, &root
->orphan_list
);
803 spin_unlock(&root
->list_lock
);
806 * insert an orphan item to track this unlinked/truncated file
808 ret
= btrfs_insert_orphan_item(trans
, root
, inode
->i_ino
);
814 * We have done the truncate/delete so we can go ahead and remove the orphan
815 * item for this particular inode.
817 int btrfs_orphan_del(struct btrfs_trans_handle
*trans
, struct inode
*inode
)
819 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
822 spin_lock(&root
->list_lock
);
824 if (list_empty(&BTRFS_I(inode
)->i_orphan
)) {
825 spin_unlock(&root
->list_lock
);
829 list_del_init(&BTRFS_I(inode
)->i_orphan
);
831 spin_unlock(&root
->list_lock
);
835 spin_unlock(&root
->list_lock
);
837 ret
= btrfs_del_orphan_item(trans
, root
, inode
->i_ino
);
843 * this cleans up any orphans that may be left on the list from the last use
846 void btrfs_orphan_cleanup(struct btrfs_root
*root
)
848 struct btrfs_path
*path
;
849 struct extent_buffer
*leaf
;
850 struct btrfs_item
*item
;
851 struct btrfs_key key
, found_key
;
852 struct btrfs_trans_handle
*trans
;
854 int ret
= 0, nr_unlink
= 0, nr_truncate
= 0;
856 /* don't do orphan cleanup if the fs is readonly. */
857 if (root
->inode
->i_sb
->s_flags
& MS_RDONLY
)
860 path
= btrfs_alloc_path();
865 key
.objectid
= BTRFS_ORPHAN_OBJECTID
;
866 btrfs_set_key_type(&key
, BTRFS_ORPHAN_ITEM_KEY
);
867 key
.offset
= (u64
)-1;
869 trans
= btrfs_start_transaction(root
, 1);
870 btrfs_set_trans_block_group(trans
, root
->inode
);
873 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
875 printk(KERN_ERR
"Error searching slot for orphan: %d"
881 * if ret == 0 means we found what we were searching for, which
882 * is weird, but possible, so only screw with path if we didnt
883 * find the key and see if we have stuff that matches
886 if (path
->slots
[0] == 0)
891 /* pull out the item */
892 leaf
= path
->nodes
[0];
893 item
= btrfs_item_nr(leaf
, path
->slots
[0]);
894 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
896 /* make sure the item matches what we want */
897 if (found_key
.objectid
!= BTRFS_ORPHAN_OBJECTID
)
899 if (btrfs_key_type(&found_key
) != BTRFS_ORPHAN_ITEM_KEY
)
902 /* release the path since we're done with it */
903 btrfs_release_path(root
, path
);
906 * this is where we are basically btrfs_lookup, without the
907 * crossing root thing. we store the inode number in the
908 * offset of the orphan item.
910 inode
= btrfs_iget_locked(root
->inode
->i_sb
,
911 found_key
.offset
, root
);
915 if (inode
->i_state
& I_NEW
) {
916 BTRFS_I(inode
)->root
= root
;
918 /* have to set the location manually */
919 BTRFS_I(inode
)->location
.objectid
= inode
->i_ino
;
920 BTRFS_I(inode
)->location
.type
= BTRFS_INODE_ITEM_KEY
;
921 BTRFS_I(inode
)->location
.offset
= 0;
923 btrfs_read_locked_inode(inode
);
924 unlock_new_inode(inode
);
928 * add this inode to the orphan list so btrfs_orphan_del does
929 * the proper thing when we hit it
931 spin_lock(&root
->list_lock
);
932 list_add(&BTRFS_I(inode
)->i_orphan
, &root
->orphan_list
);
933 spin_unlock(&root
->list_lock
);
936 * if this is a bad inode, means we actually succeeded in
937 * removing the inode, but not the orphan record, which means
938 * we need to manually delete the orphan since iput will just
941 if (is_bad_inode(inode
)) {
942 btrfs_orphan_del(trans
, inode
);
947 /* if we have links, this was a truncate, lets do that */
948 if (inode
->i_nlink
) {
950 btrfs_truncate(inode
);
955 /* this will do delete_inode and everything for us */
960 printk(KERN_INFO
"btrfs: unlinked %d orphans\n", nr_unlink
);
962 printk(KERN_INFO
"btrfs: truncated %d orphans\n", nr_truncate
);
964 btrfs_free_path(path
);
965 btrfs_end_transaction(trans
, root
);
968 void btrfs_read_locked_inode(struct inode
*inode
)
970 struct btrfs_path
*path
;
971 struct extent_buffer
*leaf
;
972 struct btrfs_inode_item
*inode_item
;
973 struct btrfs_timespec
*tspec
;
974 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
975 struct btrfs_key location
;
976 u64 alloc_group_block
;
980 path
= btrfs_alloc_path();
982 memcpy(&location
, &BTRFS_I(inode
)->location
, sizeof(location
));
984 ret
= btrfs_lookup_inode(NULL
, root
, path
, &location
, 0);
988 leaf
= path
->nodes
[0];
989 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
990 struct btrfs_inode_item
);
992 inode
->i_mode
= btrfs_inode_mode(leaf
, inode_item
);
993 inode
->i_nlink
= btrfs_inode_nlink(leaf
, inode_item
);
994 inode
->i_uid
= btrfs_inode_uid(leaf
, inode_item
);
995 inode
->i_gid
= btrfs_inode_gid(leaf
, inode_item
);
996 btrfs_i_size_write(inode
, btrfs_inode_size(leaf
, inode_item
));
998 tspec
= btrfs_inode_atime(inode_item
);
999 inode
->i_atime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1000 inode
->i_atime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1002 tspec
= btrfs_inode_mtime(inode_item
);
1003 inode
->i_mtime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1004 inode
->i_mtime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1006 tspec
= btrfs_inode_ctime(inode_item
);
1007 inode
->i_ctime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1008 inode
->i_ctime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1010 inode
->i_blocks
= btrfs_inode_nblocks(leaf
, inode_item
);
1011 BTRFS_I(inode
)->generation
= btrfs_inode_generation(leaf
, inode_item
);
1012 inode
->i_generation
= BTRFS_I(inode
)->generation
;
1014 rdev
= btrfs_inode_rdev(leaf
, inode_item
);
1016 BTRFS_I(inode
)->index_cnt
= (u64
)-1;
1018 alloc_group_block
= btrfs_inode_block_group(leaf
, inode_item
);
1019 BTRFS_I(inode
)->block_group
= btrfs_lookup_block_group(root
->fs_info
,
1021 BTRFS_I(inode
)->flags
= btrfs_inode_flags(leaf
, inode_item
);
1022 if (!BTRFS_I(inode
)->block_group
) {
1023 BTRFS_I(inode
)->block_group
= btrfs_find_block_group(root
,
1025 BTRFS_BLOCK_GROUP_METADATA
, 0);
1027 btrfs_free_path(path
);
1030 switch (inode
->i_mode
& S_IFMT
) {
1032 inode
->i_mapping
->a_ops
= &btrfs_aops
;
1033 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1034 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
1035 inode
->i_fop
= &btrfs_file_operations
;
1036 inode
->i_op
= &btrfs_file_inode_operations
;
1039 inode
->i_fop
= &btrfs_dir_file_operations
;
1040 if (root
== root
->fs_info
->tree_root
)
1041 inode
->i_op
= &btrfs_dir_ro_inode_operations
;
1043 inode
->i_op
= &btrfs_dir_inode_operations
;
1046 inode
->i_op
= &btrfs_symlink_inode_operations
;
1047 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
1048 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1051 init_special_inode(inode
, inode
->i_mode
, rdev
);
1057 btrfs_free_path(path
);
1058 make_bad_inode(inode
);
1061 static void fill_inode_item(struct btrfs_trans_handle
*trans
,
1062 struct extent_buffer
*leaf
,
1063 struct btrfs_inode_item
*item
,
1064 struct inode
*inode
)
1066 btrfs_set_inode_uid(leaf
, item
, inode
->i_uid
);
1067 btrfs_set_inode_gid(leaf
, item
, inode
->i_gid
);
1068 btrfs_set_inode_size(leaf
, item
, BTRFS_I(inode
)->disk_i_size
);
1069 btrfs_set_inode_mode(leaf
, item
, inode
->i_mode
);
1070 btrfs_set_inode_nlink(leaf
, item
, inode
->i_nlink
);
1072 btrfs_set_timespec_sec(leaf
, btrfs_inode_atime(item
),
1073 inode
->i_atime
.tv_sec
);
1074 btrfs_set_timespec_nsec(leaf
, btrfs_inode_atime(item
),
1075 inode
->i_atime
.tv_nsec
);
1077 btrfs_set_timespec_sec(leaf
, btrfs_inode_mtime(item
),
1078 inode
->i_mtime
.tv_sec
);
1079 btrfs_set_timespec_nsec(leaf
, btrfs_inode_mtime(item
),
1080 inode
->i_mtime
.tv_nsec
);
1082 btrfs_set_timespec_sec(leaf
, btrfs_inode_ctime(item
),
1083 inode
->i_ctime
.tv_sec
);
1084 btrfs_set_timespec_nsec(leaf
, btrfs_inode_ctime(item
),
1085 inode
->i_ctime
.tv_nsec
);
1087 btrfs_set_inode_nblocks(leaf
, item
, inode
->i_blocks
);
1088 btrfs_set_inode_generation(leaf
, item
, BTRFS_I(inode
)->generation
);
1089 btrfs_set_inode_transid(leaf
, item
, trans
->transid
);
1090 btrfs_set_inode_rdev(leaf
, item
, inode
->i_rdev
);
1091 btrfs_set_inode_flags(leaf
, item
, BTRFS_I(inode
)->flags
);
1092 btrfs_set_inode_block_group(leaf
, item
,
1093 BTRFS_I(inode
)->block_group
->key
.objectid
);
1096 int noinline
btrfs_update_inode(struct btrfs_trans_handle
*trans
,
1097 struct btrfs_root
*root
,
1098 struct inode
*inode
)
1100 struct btrfs_inode_item
*inode_item
;
1101 struct btrfs_path
*path
;
1102 struct extent_buffer
*leaf
;
1105 path
= btrfs_alloc_path();
1107 ret
= btrfs_lookup_inode(trans
, root
, path
,
1108 &BTRFS_I(inode
)->location
, 1);
1115 leaf
= path
->nodes
[0];
1116 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
1117 struct btrfs_inode_item
);
1119 fill_inode_item(trans
, leaf
, inode_item
, inode
);
1120 btrfs_mark_buffer_dirty(leaf
);
1121 btrfs_set_inode_last_trans(trans
, inode
);
1124 btrfs_free_path(path
);
1129 int btrfs_unlink_inode(struct btrfs_trans_handle
*trans
,
1130 struct btrfs_root
*root
,
1131 struct inode
*dir
, struct inode
*inode
,
1132 const char *name
, int name_len
)
1134 struct btrfs_path
*path
;
1136 struct extent_buffer
*leaf
;
1137 struct btrfs_dir_item
*di
;
1138 struct btrfs_key key
;
1141 path
= btrfs_alloc_path();
1147 di
= btrfs_lookup_dir_item(trans
, root
, path
, dir
->i_ino
,
1148 name
, name_len
, -1);
1157 leaf
= path
->nodes
[0];
1158 btrfs_dir_item_key_to_cpu(leaf
, di
, &key
);
1159 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
1162 btrfs_release_path(root
, path
);
1164 ret
= btrfs_del_inode_ref(trans
, root
, name
, name_len
,
1166 dir
->i_ino
, &index
);
1168 printk("failed to delete reference to %.*s, "
1169 "inode %lu parent %lu\n", name_len
, name
,
1170 inode
->i_ino
, dir
->i_ino
);
1174 di
= btrfs_lookup_dir_index_item(trans
, root
, path
, dir
->i_ino
,
1175 index
, name
, name_len
, -1);
1184 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
1185 btrfs_release_path(root
, path
);
1187 ret
= btrfs_del_inode_ref_in_log(trans
, root
, name
, name_len
,
1189 BUG_ON(ret
!= 0 && ret
!= -ENOENT
);
1191 BTRFS_I(dir
)->log_dirty_trans
= trans
->transid
;
1193 ret
= btrfs_del_dir_entries_in_log(trans
, root
, name
, name_len
,
1197 btrfs_free_path(path
);
1201 btrfs_i_size_write(dir
, dir
->i_size
- name_len
* 2);
1202 inode
->i_ctime
= dir
->i_mtime
= dir
->i_ctime
= CURRENT_TIME
;
1203 btrfs_update_inode(trans
, root
, dir
);
1204 btrfs_drop_nlink(inode
);
1205 ret
= btrfs_update_inode(trans
, root
, inode
);
1206 dir
->i_sb
->s_dirt
= 1;
1211 static int btrfs_unlink(struct inode
*dir
, struct dentry
*dentry
)
1213 struct btrfs_root
*root
;
1214 struct btrfs_trans_handle
*trans
;
1215 struct inode
*inode
= dentry
->d_inode
;
1217 unsigned long nr
= 0;
1219 root
= BTRFS_I(dir
)->root
;
1221 ret
= btrfs_check_free_space(root
, 1, 1);
1225 trans
= btrfs_start_transaction(root
, 1);
1227 btrfs_set_trans_block_group(trans
, dir
);
1228 ret
= btrfs_unlink_inode(trans
, root
, dir
, dentry
->d_inode
,
1229 dentry
->d_name
.name
, dentry
->d_name
.len
);
1231 if (inode
->i_nlink
== 0)
1232 ret
= btrfs_orphan_add(trans
, inode
);
1234 nr
= trans
->blocks_used
;
1236 btrfs_end_transaction_throttle(trans
, root
);
1238 btrfs_btree_balance_dirty(root
, nr
);
1242 static int btrfs_rmdir(struct inode
*dir
, struct dentry
*dentry
)
1244 struct inode
*inode
= dentry
->d_inode
;
1247 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1248 struct btrfs_trans_handle
*trans
;
1249 unsigned long nr
= 0;
1251 if (inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
1255 ret
= btrfs_check_free_space(root
, 1, 1);
1259 trans
= btrfs_start_transaction(root
, 1);
1260 btrfs_set_trans_block_group(trans
, dir
);
1262 err
= btrfs_orphan_add(trans
, inode
);
1266 /* now the directory is empty */
1267 err
= btrfs_unlink_inode(trans
, root
, dir
, dentry
->d_inode
,
1268 dentry
->d_name
.name
, dentry
->d_name
.len
);
1270 btrfs_i_size_write(inode
, 0);
1274 nr
= trans
->blocks_used
;
1275 ret
= btrfs_end_transaction_throttle(trans
, root
);
1277 btrfs_btree_balance_dirty(root
, nr
);
1285 * this can truncate away extent items, csum items and directory items.
1286 * It starts at a high offset and removes keys until it can't find
1287 * any higher than i_size.
1289 * csum items that cross the new i_size are truncated to the new size
1292 * min_type is the minimum key type to truncate down to. If set to 0, this
1293 * will kill all the items on this inode, including the INODE_ITEM_KEY.
1295 noinline
int btrfs_truncate_inode_items(struct btrfs_trans_handle
*trans
,
1296 struct btrfs_root
*root
,
1297 struct inode
*inode
,
1298 u64 new_size
, u32 min_type
)
1301 struct btrfs_path
*path
;
1302 struct btrfs_key key
;
1303 struct btrfs_key found_key
;
1305 struct extent_buffer
*leaf
;
1306 struct btrfs_file_extent_item
*fi
;
1307 u64 extent_start
= 0;
1308 u64 extent_num_bytes
= 0;
1314 int pending_del_nr
= 0;
1315 int pending_del_slot
= 0;
1316 int extent_type
= -1;
1317 u64 mask
= root
->sectorsize
- 1;
1320 btrfs_drop_extent_cache(inode
,
1321 new_size
& (~mask
), (u64
)-1);
1322 path
= btrfs_alloc_path();
1326 /* FIXME, add redo link to tree so we don't leak on crash */
1327 key
.objectid
= inode
->i_ino
;
1328 key
.offset
= (u64
)-1;
1331 btrfs_init_path(path
);
1333 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
1338 /* there are no items in the tree for us to truncate, we're
1341 if (path
->slots
[0] == 0) {
1350 leaf
= path
->nodes
[0];
1351 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
1352 found_type
= btrfs_key_type(&found_key
);
1354 if (found_key
.objectid
!= inode
->i_ino
)
1357 if (found_type
< min_type
)
1360 item_end
= found_key
.offset
;
1361 if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
1362 fi
= btrfs_item_ptr(leaf
, path
->slots
[0],
1363 struct btrfs_file_extent_item
);
1364 extent_type
= btrfs_file_extent_type(leaf
, fi
);
1365 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1367 btrfs_file_extent_num_bytes(leaf
, fi
);
1368 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1369 struct btrfs_item
*item
= btrfs_item_nr(leaf
,
1371 item_end
+= btrfs_file_extent_inline_len(leaf
,
1376 if (found_type
== BTRFS_CSUM_ITEM_KEY
) {
1377 ret
= btrfs_csum_truncate(trans
, root
, path
,
1381 if (item_end
< new_size
) {
1382 if (found_type
== BTRFS_DIR_ITEM_KEY
) {
1383 found_type
= BTRFS_INODE_ITEM_KEY
;
1384 } else if (found_type
== BTRFS_EXTENT_ITEM_KEY
) {
1385 found_type
= BTRFS_CSUM_ITEM_KEY
;
1386 } else if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
1387 found_type
= BTRFS_XATTR_ITEM_KEY
;
1388 } else if (found_type
== BTRFS_XATTR_ITEM_KEY
) {
1389 found_type
= BTRFS_INODE_REF_KEY
;
1390 } else if (found_type
) {
1395 btrfs_set_key_type(&key
, found_type
);
1398 if (found_key
.offset
>= new_size
)
1404 /* FIXME, shrink the extent if the ref count is only 1 */
1405 if (found_type
!= BTRFS_EXTENT_DATA_KEY
)
1408 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1410 extent_start
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
1412 u64 orig_num_bytes
=
1413 btrfs_file_extent_num_bytes(leaf
, fi
);
1414 extent_num_bytes
= new_size
-
1415 found_key
.offset
+ root
->sectorsize
- 1;
1416 extent_num_bytes
= extent_num_bytes
&
1417 ~((u64
)root
->sectorsize
- 1);
1418 btrfs_set_file_extent_num_bytes(leaf
, fi
,
1420 num_dec
= (orig_num_bytes
-
1422 if (root
->ref_cows
&& extent_start
!= 0)
1423 dec_i_blocks(inode
, num_dec
);
1424 btrfs_mark_buffer_dirty(leaf
);
1427 btrfs_file_extent_disk_num_bytes(leaf
,
1429 /* FIXME blocksize != 4096 */
1430 num_dec
= btrfs_file_extent_num_bytes(leaf
, fi
);
1431 if (extent_start
!= 0) {
1434 dec_i_blocks(inode
, num_dec
);
1436 if (root
->ref_cows
) {
1438 btrfs_header_generation(leaf
);
1440 root_owner
= btrfs_header_owner(leaf
);
1442 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1444 u32 size
= new_size
- found_key
.offset
;
1446 if (root
->ref_cows
) {
1447 dec_i_blocks(inode
, item_end
+ 1 -
1448 found_key
.offset
- size
);
1451 btrfs_file_extent_calc_inline_size(size
);
1452 ret
= btrfs_truncate_item(trans
, root
, path
,
1455 } else if (root
->ref_cows
) {
1456 dec_i_blocks(inode
, item_end
+ 1 -
1462 if (!pending_del_nr
) {
1463 /* no pending yet, add ourselves */
1464 pending_del_slot
= path
->slots
[0];
1466 } else if (pending_del_nr
&&
1467 path
->slots
[0] + 1 == pending_del_slot
) {
1468 /* hop on the pending chunk */
1470 pending_del_slot
= path
->slots
[0];
1472 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path
->slots
[0], pending_del_nr
, pending_del_slot
);
1478 ret
= btrfs_free_extent(trans
, root
, extent_start
,
1481 root_gen
, inode
->i_ino
,
1482 found_key
.offset
, 0);
1486 if (path
->slots
[0] == 0) {
1489 btrfs_release_path(root
, path
);
1494 if (pending_del_nr
&&
1495 path
->slots
[0] + 1 != pending_del_slot
) {
1496 struct btrfs_key debug
;
1498 btrfs_item_key_to_cpu(path
->nodes
[0], &debug
,
1500 ret
= btrfs_del_items(trans
, root
, path
,
1505 btrfs_release_path(root
, path
);
1511 if (pending_del_nr
) {
1512 ret
= btrfs_del_items(trans
, root
, path
, pending_del_slot
,
1515 btrfs_free_path(path
);
1516 inode
->i_sb
->s_dirt
= 1;
1521 * taken from block_truncate_page, but does cow as it zeros out
1522 * any bytes left in the last page in the file.
1524 static int btrfs_truncate_page(struct address_space
*mapping
, loff_t from
)
1526 struct inode
*inode
= mapping
->host
;
1527 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1528 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1529 struct btrfs_ordered_extent
*ordered
;
1531 u32 blocksize
= root
->sectorsize
;
1532 pgoff_t index
= from
>> PAGE_CACHE_SHIFT
;
1533 unsigned offset
= from
& (PAGE_CACHE_SIZE
-1);
1539 if ((offset
& (blocksize
- 1)) == 0)
1544 page
= grab_cache_page(mapping
, index
);
1548 page_start
= page_offset(page
);
1549 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1551 if (!PageUptodate(page
)) {
1552 ret
= btrfs_readpage(NULL
, page
);
1554 if (page
->mapping
!= mapping
) {
1556 page_cache_release(page
);
1559 if (!PageUptodate(page
)) {
1564 wait_on_page_writeback(page
);
1566 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1567 set_page_extent_mapped(page
);
1569 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
1571 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1573 page_cache_release(page
);
1574 btrfs_start_ordered_extent(inode
, ordered
, 1);
1575 btrfs_put_ordered_extent(ordered
);
1579 btrfs_set_extent_delalloc(inode
, page_start
, page_end
);
1581 if (offset
!= PAGE_CACHE_SIZE
) {
1583 memset(kaddr
+ offset
, 0, PAGE_CACHE_SIZE
- offset
);
1584 flush_dcache_page(page
);
1587 ClearPageChecked(page
);
1588 set_page_dirty(page
);
1589 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1593 page_cache_release(page
);
1598 static int btrfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
1600 struct inode
*inode
= dentry
->d_inode
;
1603 err
= inode_change_ok(inode
, attr
);
1607 if (S_ISREG(inode
->i_mode
) &&
1608 attr
->ia_valid
& ATTR_SIZE
&& attr
->ia_size
> inode
->i_size
) {
1609 struct btrfs_trans_handle
*trans
;
1610 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1611 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1613 u64 mask
= root
->sectorsize
- 1;
1614 u64 hole_start
= (inode
->i_size
+ mask
) & ~mask
;
1615 u64 block_end
= (attr
->ia_size
+ mask
) & ~mask
;
1619 if (attr
->ia_size
<= hole_start
)
1622 err
= btrfs_check_free_space(root
, 1, 0);
1626 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
1628 hole_size
= block_end
- hole_start
;
1630 struct btrfs_ordered_extent
*ordered
;
1631 btrfs_wait_ordered_range(inode
, hole_start
, hole_size
);
1633 lock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1634 ordered
= btrfs_lookup_ordered_extent(inode
, hole_start
);
1636 unlock_extent(io_tree
, hole_start
,
1637 block_end
- 1, GFP_NOFS
);
1638 btrfs_put_ordered_extent(ordered
);
1644 trans
= btrfs_start_transaction(root
, 1);
1645 btrfs_set_trans_block_group(trans
, inode
);
1646 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
1647 err
= btrfs_drop_extents(trans
, root
, inode
,
1648 hole_start
, block_end
, hole_start
,
1651 if (alloc_hint
!= EXTENT_MAP_INLINE
) {
1652 err
= btrfs_insert_file_extent(trans
, root
,
1656 btrfs_drop_extent_cache(inode
, hole_start
,
1658 btrfs_check_file(root
, inode
);
1660 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
1661 btrfs_end_transaction(trans
, root
);
1662 unlock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1667 err
= inode_setattr(inode
, attr
);
1669 if (!err
&& ((attr
->ia_valid
& ATTR_MODE
)))
1670 err
= btrfs_acl_chmod(inode
);
1675 void btrfs_delete_inode(struct inode
*inode
)
1677 struct btrfs_trans_handle
*trans
;
1678 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1682 truncate_inode_pages(&inode
->i_data
, 0);
1683 if (is_bad_inode(inode
)) {
1684 btrfs_orphan_del(NULL
, inode
);
1687 btrfs_wait_ordered_range(inode
, 0, (u64
)-1);
1689 btrfs_i_size_write(inode
, 0);
1690 trans
= btrfs_start_transaction(root
, 1);
1692 btrfs_set_trans_block_group(trans
, inode
);
1693 ret
= btrfs_truncate_inode_items(trans
, root
, inode
, inode
->i_size
, 0);
1695 btrfs_orphan_del(NULL
, inode
);
1696 goto no_delete_lock
;
1699 btrfs_orphan_del(trans
, inode
);
1701 nr
= trans
->blocks_used
;
1704 btrfs_end_transaction(trans
, root
);
1705 btrfs_btree_balance_dirty(root
, nr
);
1709 nr
= trans
->blocks_used
;
1710 btrfs_end_transaction(trans
, root
);
1711 btrfs_btree_balance_dirty(root
, nr
);
1717 * this returns the key found in the dir entry in the location pointer.
1718 * If no dir entries were found, location->objectid is 0.
1720 static int btrfs_inode_by_name(struct inode
*dir
, struct dentry
*dentry
,
1721 struct btrfs_key
*location
)
1723 const char *name
= dentry
->d_name
.name
;
1724 int namelen
= dentry
->d_name
.len
;
1725 struct btrfs_dir_item
*di
;
1726 struct btrfs_path
*path
;
1727 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1730 path
= btrfs_alloc_path();
1733 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir
->i_ino
, name
,
1737 if (!di
|| IS_ERR(di
)) {
1740 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, location
);
1742 btrfs_free_path(path
);
1745 location
->objectid
= 0;
1750 * when we hit a tree root in a directory, the btrfs part of the inode
1751 * needs to be changed to reflect the root directory of the tree root. This
1752 * is kind of like crossing a mount point.
1754 static int fixup_tree_root_location(struct btrfs_root
*root
,
1755 struct btrfs_key
*location
,
1756 struct btrfs_root
**sub_root
,
1757 struct dentry
*dentry
)
1759 struct btrfs_root_item
*ri
;
1761 if (btrfs_key_type(location
) != BTRFS_ROOT_ITEM_KEY
)
1763 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1766 *sub_root
= btrfs_read_fs_root(root
->fs_info
, location
,
1767 dentry
->d_name
.name
,
1768 dentry
->d_name
.len
);
1769 if (IS_ERR(*sub_root
))
1770 return PTR_ERR(*sub_root
);
1772 ri
= &(*sub_root
)->root_item
;
1773 location
->objectid
= btrfs_root_dirid(ri
);
1774 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1775 location
->offset
= 0;
1780 static noinline
void init_btrfs_i(struct inode
*inode
)
1782 struct btrfs_inode
*bi
= BTRFS_I(inode
);
1785 bi
->i_default_acl
= NULL
;
1789 bi
->logged_trans
= 0;
1790 bi
->delalloc_bytes
= 0;
1791 bi
->disk_i_size
= 0;
1793 bi
->index_cnt
= (u64
)-1;
1794 bi
->log_dirty_trans
= 0;
1795 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1796 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1797 inode
->i_mapping
, GFP_NOFS
);
1798 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1799 inode
->i_mapping
, GFP_NOFS
);
1800 INIT_LIST_HEAD(&BTRFS_I(inode
)->delalloc_inodes
);
1801 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
1802 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
1803 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
1804 mutex_init(&BTRFS_I(inode
)->log_mutex
);
1807 static int btrfs_init_locked_inode(struct inode
*inode
, void *p
)
1809 struct btrfs_iget_args
*args
= p
;
1810 inode
->i_ino
= args
->ino
;
1811 init_btrfs_i(inode
);
1812 BTRFS_I(inode
)->root
= args
->root
;
1816 static int btrfs_find_actor(struct inode
*inode
, void *opaque
)
1818 struct btrfs_iget_args
*args
= opaque
;
1819 return (args
->ino
== inode
->i_ino
&&
1820 args
->root
== BTRFS_I(inode
)->root
);
1823 struct inode
*btrfs_iget_locked(struct super_block
*s
, u64 objectid
,
1824 struct btrfs_root
*root
)
1826 struct inode
*inode
;
1827 struct btrfs_iget_args args
;
1828 args
.ino
= objectid
;
1831 inode
= iget5_locked(s
, objectid
, btrfs_find_actor
,
1832 btrfs_init_locked_inode
,
1837 /* Get an inode object given its location and corresponding root.
1838 * Returns in *is_new if the inode was read from disk
1840 struct inode
*btrfs_iget(struct super_block
*s
, struct btrfs_key
*location
,
1841 struct btrfs_root
*root
, int *is_new
)
1843 struct inode
*inode
;
1845 inode
= btrfs_iget_locked(s
, location
->objectid
, root
);
1847 return ERR_PTR(-EACCES
);
1849 if (inode
->i_state
& I_NEW
) {
1850 BTRFS_I(inode
)->root
= root
;
1851 memcpy(&BTRFS_I(inode
)->location
, location
, sizeof(*location
));
1852 btrfs_read_locked_inode(inode
);
1853 unlock_new_inode(inode
);
1864 static struct dentry
*btrfs_lookup(struct inode
*dir
, struct dentry
*dentry
,
1865 struct nameidata
*nd
)
1867 struct inode
* inode
;
1868 struct btrfs_inode
*bi
= BTRFS_I(dir
);
1869 struct btrfs_root
*root
= bi
->root
;
1870 struct btrfs_root
*sub_root
= root
;
1871 struct btrfs_key location
;
1872 int ret
, new, do_orphan
= 0;
1874 if (dentry
->d_name
.len
> BTRFS_NAME_LEN
)
1875 return ERR_PTR(-ENAMETOOLONG
);
1877 ret
= btrfs_inode_by_name(dir
, dentry
, &location
);
1880 return ERR_PTR(ret
);
1883 if (location
.objectid
) {
1884 ret
= fixup_tree_root_location(root
, &location
, &sub_root
,
1887 return ERR_PTR(ret
);
1889 return ERR_PTR(-ENOENT
);
1890 inode
= btrfs_iget(dir
->i_sb
, &location
, sub_root
, &new);
1892 return ERR_CAST(inode
);
1894 /* the inode and parent dir are two different roots */
1895 if (new && root
!= sub_root
) {
1897 sub_root
->inode
= inode
;
1902 if (unlikely(do_orphan
))
1903 btrfs_orphan_cleanup(sub_root
);
1905 return d_splice_alias(inode
, dentry
);
1908 static unsigned char btrfs_filetype_table
[] = {
1909 DT_UNKNOWN
, DT_REG
, DT_DIR
, DT_CHR
, DT_BLK
, DT_FIFO
, DT_SOCK
, DT_LNK
1912 static int btrfs_real_readdir(struct file
*filp
, void *dirent
,
1915 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1916 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1917 struct btrfs_item
*item
;
1918 struct btrfs_dir_item
*di
;
1919 struct btrfs_key key
;
1920 struct btrfs_key found_key
;
1921 struct btrfs_path
*path
;
1924 struct extent_buffer
*leaf
;
1927 unsigned char d_type
;
1932 int key_type
= BTRFS_DIR_INDEX_KEY
;
1937 /* FIXME, use a real flag for deciding about the key type */
1938 if (root
->fs_info
->tree_root
== root
)
1939 key_type
= BTRFS_DIR_ITEM_KEY
;
1941 /* special case for "." */
1942 if (filp
->f_pos
== 0) {
1943 over
= filldir(dirent
, ".", 1,
1950 /* special case for .., just use the back ref */
1951 if (filp
->f_pos
== 1) {
1952 u64 pino
= parent_ino(filp
->f_path
.dentry
);
1953 over
= filldir(dirent
, "..", 2,
1960 path
= btrfs_alloc_path();
1963 btrfs_set_key_type(&key
, key_type
);
1964 key
.offset
= filp
->f_pos
;
1965 key
.objectid
= inode
->i_ino
;
1967 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1973 leaf
= path
->nodes
[0];
1974 nritems
= btrfs_header_nritems(leaf
);
1975 slot
= path
->slots
[0];
1976 if (advance
|| slot
>= nritems
) {
1977 if (slot
>= nritems
- 1) {
1978 ret
= btrfs_next_leaf(root
, path
);
1981 leaf
= path
->nodes
[0];
1982 nritems
= btrfs_header_nritems(leaf
);
1983 slot
= path
->slots
[0];
1990 item
= btrfs_item_nr(leaf
, slot
);
1991 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1993 if (found_key
.objectid
!= key
.objectid
)
1995 if (btrfs_key_type(&found_key
) != key_type
)
1997 if (found_key
.offset
< filp
->f_pos
)
2000 filp
->f_pos
= found_key
.offset
;
2002 di
= btrfs_item_ptr(leaf
, slot
, struct btrfs_dir_item
);
2004 di_total
= btrfs_item_size(leaf
, item
);
2006 while (di_cur
< di_total
) {
2007 struct btrfs_key location
;
2009 name_len
= btrfs_dir_name_len(leaf
, di
);
2010 if (name_len
<= sizeof(tmp_name
)) {
2011 name_ptr
= tmp_name
;
2013 name_ptr
= kmalloc(name_len
, GFP_NOFS
);
2019 read_extent_buffer(leaf
, name_ptr
,
2020 (unsigned long)(di
+ 1), name_len
);
2022 d_type
= btrfs_filetype_table
[btrfs_dir_type(leaf
, di
)];
2023 btrfs_dir_item_key_to_cpu(leaf
, di
, &location
);
2024 over
= filldir(dirent
, name_ptr
, name_len
,
2025 found_key
.offset
, location
.objectid
,
2028 if (name_ptr
!= tmp_name
)
2034 di_len
= btrfs_dir_name_len(leaf
, di
) +
2035 btrfs_dir_data_len(leaf
, di
) + sizeof(*di
);
2037 di
= (struct btrfs_dir_item
*)((char *)di
+ di_len
);
2041 /* Reached end of directory/root. Bump pos past the last item. */
2042 if (key_type
== BTRFS_DIR_INDEX_KEY
)
2043 filp
->f_pos
= INT_LIMIT(typeof(filp
->f_pos
));
2049 btrfs_free_path(path
);
2053 /* Kernels earlier than 2.6.28 still have the NFS deadlock where nfsd
2054 will call the file system's ->lookup() method from within its
2055 filldir callback, which in turn was called from the file system's
2056 ->readdir() method. And will deadlock for many file systems. */
2057 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28)
2059 struct nfshack_dirent
{
2063 unsigned int d_type
;
2067 struct nfshack_readdir
{
2075 static int btrfs_nfshack_filldir(void *__buf
, const char *name
, int namlen
,
2076 loff_t offset
, u64 ino
, unsigned int d_type
)
2078 struct nfshack_readdir
*buf
= __buf
;
2079 struct nfshack_dirent
*de
= (void *)(buf
->dirent
+ buf
->used
);
2080 unsigned int reclen
;
2082 reclen
= ALIGN(sizeof(struct nfshack_dirent
) + namlen
, sizeof(u64
));
2083 if (buf
->used
+ reclen
> PAGE_SIZE
) {
2088 de
->namlen
= namlen
;
2089 de
->offset
= offset
;
2091 de
->d_type
= d_type
;
2092 memcpy(de
->name
, name
, namlen
);
2093 buf
->used
+= reclen
;
2098 static int btrfs_nfshack_readdir(struct file
*file
, void *dirent
,
2101 struct nfshack_readdir buf
;
2102 struct nfshack_dirent
*de
;
2107 buf
.dirent
= (void *)__get_free_page(GFP_KERNEL
);
2111 offset
= file
->f_pos
;
2114 unsigned int reclen
;
2118 err
= btrfs_real_readdir(file
, &buf
, btrfs_nfshack_filldir
);
2127 de
= (struct nfshack_dirent
*)buf
.dirent
;
2129 offset
= de
->offset
;
2131 if (filldir(dirent
, de
->name
, de
->namlen
, de
->offset
,
2132 de
->ino
, de
->d_type
))
2134 offset
= file
->f_pos
;
2136 reclen
= ALIGN(sizeof(*de
) + de
->namlen
,
2139 de
= (struct nfshack_dirent
*)((char *)de
+ reclen
);
2144 free_page((unsigned long)buf
.dirent
);
2145 file
->f_pos
= offset
;
2151 int btrfs_write_inode(struct inode
*inode
, int wait
)
2153 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2154 struct btrfs_trans_handle
*trans
;
2157 if (root
->fs_info
->closing
> 1)
2161 trans
= btrfs_join_transaction(root
, 1);
2162 btrfs_set_trans_block_group(trans
, inode
);
2163 ret
= btrfs_commit_transaction(trans
, root
);
2169 * This is somewhat expensive, updating the tree every time the
2170 * inode changes. But, it is most likely to find the inode in cache.
2171 * FIXME, needs more benchmarking...there are no reasons other than performance
2172 * to keep or drop this code.
2174 void btrfs_dirty_inode(struct inode
*inode
)
2176 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2177 struct btrfs_trans_handle
*trans
;
2179 trans
= btrfs_join_transaction(root
, 1);
2180 btrfs_set_trans_block_group(trans
, inode
);
2181 btrfs_update_inode(trans
, root
, inode
);
2182 btrfs_end_transaction(trans
, root
);
2185 static int btrfs_set_inode_index_count(struct inode
*inode
)
2187 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2188 struct btrfs_key key
, found_key
;
2189 struct btrfs_path
*path
;
2190 struct extent_buffer
*leaf
;
2193 key
.objectid
= inode
->i_ino
;
2194 btrfs_set_key_type(&key
, BTRFS_DIR_INDEX_KEY
);
2195 key
.offset
= (u64
)-1;
2197 path
= btrfs_alloc_path();
2201 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2204 /* FIXME: we should be able to handle this */
2210 * MAGIC NUMBER EXPLANATION:
2211 * since we search a directory based on f_pos we have to start at 2
2212 * since '.' and '..' have f_pos of 0 and 1 respectively, so everybody
2213 * else has to start at 2
2215 if (path
->slots
[0] == 0) {
2216 BTRFS_I(inode
)->index_cnt
= 2;
2222 leaf
= path
->nodes
[0];
2223 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2225 if (found_key
.objectid
!= inode
->i_ino
||
2226 btrfs_key_type(&found_key
) != BTRFS_DIR_INDEX_KEY
) {
2227 BTRFS_I(inode
)->index_cnt
= 2;
2231 BTRFS_I(inode
)->index_cnt
= found_key
.offset
+ 1;
2233 btrfs_free_path(path
);
2237 static int btrfs_set_inode_index(struct inode
*dir
, struct inode
*inode
,
2242 if (BTRFS_I(dir
)->index_cnt
== (u64
)-1) {
2243 ret
= btrfs_set_inode_index_count(dir
);
2249 *index
= BTRFS_I(dir
)->index_cnt
;
2250 BTRFS_I(dir
)->index_cnt
++;
2255 static struct inode
*btrfs_new_inode(struct btrfs_trans_handle
*trans
,
2256 struct btrfs_root
*root
,
2258 const char *name
, int name_len
,
2261 struct btrfs_block_group_cache
*group
,
2262 int mode
, u64
*index
)
2264 struct inode
*inode
;
2265 struct btrfs_inode_item
*inode_item
;
2266 struct btrfs_block_group_cache
*new_inode_group
;
2267 struct btrfs_key
*location
;
2268 struct btrfs_path
*path
;
2269 struct btrfs_inode_ref
*ref
;
2270 struct btrfs_key key
[2];
2276 path
= btrfs_alloc_path();
2279 inode
= new_inode(root
->fs_info
->sb
);
2281 return ERR_PTR(-ENOMEM
);
2284 ret
= btrfs_set_inode_index(dir
, inode
, index
);
2286 return ERR_PTR(ret
);
2289 * index_cnt is ignored for everything but a dir,
2290 * btrfs_get_inode_index_count has an explanation for the magic
2293 init_btrfs_i(inode
);
2294 BTRFS_I(inode
)->index_cnt
= 2;
2295 BTRFS_I(inode
)->root
= root
;
2296 BTRFS_I(inode
)->generation
= trans
->transid
;
2302 new_inode_group
= btrfs_find_block_group(root
, group
, 0,
2303 BTRFS_BLOCK_GROUP_METADATA
, owner
);
2304 if (!new_inode_group
) {
2305 printk("find_block group failed\n");
2306 new_inode_group
= group
;
2308 BTRFS_I(inode
)->block_group
= new_inode_group
;
2310 key
[0].objectid
= objectid
;
2311 btrfs_set_key_type(&key
[0], BTRFS_INODE_ITEM_KEY
);
2314 key
[1].objectid
= objectid
;
2315 btrfs_set_key_type(&key
[1], BTRFS_INODE_REF_KEY
);
2316 key
[1].offset
= ref_objectid
;
2318 sizes
[0] = sizeof(struct btrfs_inode_item
);
2319 sizes
[1] = name_len
+ sizeof(*ref
);
2321 ret
= btrfs_insert_empty_items(trans
, root
, path
, key
, sizes
, 2);
2325 if (objectid
> root
->highest_inode
)
2326 root
->highest_inode
= objectid
;
2328 inode
->i_uid
= current
->fsuid
;
2329 inode
->i_gid
= current
->fsgid
;
2330 inode
->i_mode
= mode
;
2331 inode
->i_ino
= objectid
;
2332 inode
->i_blocks
= 0;
2333 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2334 inode_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
2335 struct btrfs_inode_item
);
2336 fill_inode_item(trans
, path
->nodes
[0], inode_item
, inode
);
2338 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0] + 1,
2339 struct btrfs_inode_ref
);
2340 btrfs_set_inode_ref_name_len(path
->nodes
[0], ref
, name_len
);
2341 btrfs_set_inode_ref_index(path
->nodes
[0], ref
, *index
);
2342 ptr
= (unsigned long)(ref
+ 1);
2343 write_extent_buffer(path
->nodes
[0], name
, ptr
, name_len
);
2345 btrfs_mark_buffer_dirty(path
->nodes
[0]);
2346 btrfs_free_path(path
);
2348 location
= &BTRFS_I(inode
)->location
;
2349 location
->objectid
= objectid
;
2350 location
->offset
= 0;
2351 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
2353 insert_inode_hash(inode
);
2357 BTRFS_I(dir
)->index_cnt
--;
2358 btrfs_free_path(path
);
2359 return ERR_PTR(ret
);
2362 static inline u8
btrfs_inode_type(struct inode
*inode
)
2364 return btrfs_type_by_mode
[(inode
->i_mode
& S_IFMT
) >> S_SHIFT
];
2367 int btrfs_add_link(struct btrfs_trans_handle
*trans
,
2368 struct inode
*parent_inode
, struct inode
*inode
,
2369 const char *name
, int name_len
, int add_backref
, u64 index
)
2372 struct btrfs_key key
;
2373 struct btrfs_root
*root
= BTRFS_I(parent_inode
)->root
;
2375 key
.objectid
= inode
->i_ino
;
2376 btrfs_set_key_type(&key
, BTRFS_INODE_ITEM_KEY
);
2379 ret
= btrfs_insert_dir_item(trans
, root
, name
, name_len
,
2380 parent_inode
->i_ino
,
2381 &key
, btrfs_inode_type(inode
),
2385 ret
= btrfs_insert_inode_ref(trans
, root
,
2388 parent_inode
->i_ino
,
2391 btrfs_i_size_write(parent_inode
, parent_inode
->i_size
+
2393 parent_inode
->i_mtime
= parent_inode
->i_ctime
= CURRENT_TIME
;
2394 ret
= btrfs_update_inode(trans
, root
, parent_inode
);
2399 static int btrfs_add_nondir(struct btrfs_trans_handle
*trans
,
2400 struct dentry
*dentry
, struct inode
*inode
,
2401 int backref
, u64 index
)
2403 int err
= btrfs_add_link(trans
, dentry
->d_parent
->d_inode
,
2404 inode
, dentry
->d_name
.name
,
2405 dentry
->d_name
.len
, backref
, index
);
2407 d_instantiate(dentry
, inode
);
2415 static int btrfs_mknod(struct inode
*dir
, struct dentry
*dentry
,
2416 int mode
, dev_t rdev
)
2418 struct btrfs_trans_handle
*trans
;
2419 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2420 struct inode
*inode
= NULL
;
2424 unsigned long nr
= 0;
2427 if (!new_valid_dev(rdev
))
2430 err
= btrfs_check_free_space(root
, 1, 0);
2434 trans
= btrfs_start_transaction(root
, 1);
2435 btrfs_set_trans_block_group(trans
, dir
);
2437 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2443 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2445 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2446 BTRFS_I(dir
)->block_group
, mode
, &index
);
2447 err
= PTR_ERR(inode
);
2451 err
= btrfs_init_acl(inode
, dir
);
2457 btrfs_set_trans_block_group(trans
, inode
);
2458 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
2462 inode
->i_op
= &btrfs_special_inode_operations
;
2463 init_special_inode(inode
, inode
->i_mode
, rdev
);
2464 btrfs_update_inode(trans
, root
, inode
);
2466 dir
->i_sb
->s_dirt
= 1;
2467 btrfs_update_inode_block_group(trans
, inode
);
2468 btrfs_update_inode_block_group(trans
, dir
);
2470 nr
= trans
->blocks_used
;
2471 btrfs_end_transaction_throttle(trans
, root
);
2474 inode_dec_link_count(inode
);
2477 btrfs_btree_balance_dirty(root
, nr
);
2481 static int btrfs_create(struct inode
*dir
, struct dentry
*dentry
,
2482 int mode
, struct nameidata
*nd
)
2484 struct btrfs_trans_handle
*trans
;
2485 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2486 struct inode
*inode
= NULL
;
2489 unsigned long nr
= 0;
2493 err
= btrfs_check_free_space(root
, 1, 0);
2496 trans
= btrfs_start_transaction(root
, 1);
2497 btrfs_set_trans_block_group(trans
, dir
);
2499 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2505 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2507 dentry
->d_parent
->d_inode
->i_ino
,
2508 objectid
, BTRFS_I(dir
)->block_group
, mode
,
2510 err
= PTR_ERR(inode
);
2514 err
= btrfs_init_acl(inode
, dir
);
2520 btrfs_set_trans_block_group(trans
, inode
);
2521 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
2525 inode
->i_mapping
->a_ops
= &btrfs_aops
;
2526 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
2527 inode
->i_fop
= &btrfs_file_operations
;
2528 inode
->i_op
= &btrfs_file_inode_operations
;
2529 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
2531 dir
->i_sb
->s_dirt
= 1;
2532 btrfs_update_inode_block_group(trans
, inode
);
2533 btrfs_update_inode_block_group(trans
, dir
);
2535 nr
= trans
->blocks_used
;
2536 btrfs_end_transaction_throttle(trans
, root
);
2539 inode_dec_link_count(inode
);
2542 btrfs_btree_balance_dirty(root
, nr
);
2546 static int btrfs_link(struct dentry
*old_dentry
, struct inode
*dir
,
2547 struct dentry
*dentry
)
2549 struct btrfs_trans_handle
*trans
;
2550 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2551 struct inode
*inode
= old_dentry
->d_inode
;
2553 unsigned long nr
= 0;
2557 if (inode
->i_nlink
== 0)
2560 btrfs_inc_nlink(inode
);
2561 err
= btrfs_check_free_space(root
, 1, 0);
2564 err
= btrfs_set_inode_index(dir
, inode
, &index
);
2568 trans
= btrfs_start_transaction(root
, 1);
2570 btrfs_set_trans_block_group(trans
, dir
);
2571 atomic_inc(&inode
->i_count
);
2573 err
= btrfs_add_nondir(trans
, dentry
, inode
, 1, index
);
2578 dir
->i_sb
->s_dirt
= 1;
2579 btrfs_update_inode_block_group(trans
, dir
);
2580 err
= btrfs_update_inode(trans
, root
, inode
);
2585 nr
= trans
->blocks_used
;
2586 btrfs_end_transaction_throttle(trans
, root
);
2589 inode_dec_link_count(inode
);
2592 btrfs_btree_balance_dirty(root
, nr
);
2596 static int btrfs_mkdir(struct inode
*dir
, struct dentry
*dentry
, int mode
)
2598 struct inode
*inode
= NULL
;
2599 struct btrfs_trans_handle
*trans
;
2600 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2602 int drop_on_err
= 0;
2605 unsigned long nr
= 1;
2607 err
= btrfs_check_free_space(root
, 1, 0);
2611 trans
= btrfs_start_transaction(root
, 1);
2612 btrfs_set_trans_block_group(trans
, dir
);
2614 if (IS_ERR(trans
)) {
2615 err
= PTR_ERR(trans
);
2619 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2625 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2627 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2628 BTRFS_I(dir
)->block_group
, S_IFDIR
| mode
,
2630 if (IS_ERR(inode
)) {
2631 err
= PTR_ERR(inode
);
2637 err
= btrfs_init_acl(inode
, dir
);
2641 inode
->i_op
= &btrfs_dir_inode_operations
;
2642 inode
->i_fop
= &btrfs_dir_file_operations
;
2643 btrfs_set_trans_block_group(trans
, inode
);
2645 btrfs_i_size_write(inode
, 0);
2646 err
= btrfs_update_inode(trans
, root
, inode
);
2650 err
= btrfs_add_link(trans
, dentry
->d_parent
->d_inode
,
2651 inode
, dentry
->d_name
.name
,
2652 dentry
->d_name
.len
, 0, index
);
2656 d_instantiate(dentry
, inode
);
2658 dir
->i_sb
->s_dirt
= 1;
2659 btrfs_update_inode_block_group(trans
, inode
);
2660 btrfs_update_inode_block_group(trans
, dir
);
2663 nr
= trans
->blocks_used
;
2664 btrfs_end_transaction_throttle(trans
, root
);
2669 btrfs_btree_balance_dirty(root
, nr
);
2673 static int merge_extent_mapping(struct extent_map_tree
*em_tree
,
2674 struct extent_map
*existing
,
2675 struct extent_map
*em
,
2676 u64 map_start
, u64 map_len
)
2680 BUG_ON(map_start
< em
->start
|| map_start
>= extent_map_end(em
));
2681 start_diff
= map_start
- em
->start
;
2682 em
->start
= map_start
;
2684 if (em
->block_start
< EXTENT_MAP_LAST_BYTE
)
2685 em
->block_start
+= start_diff
;
2686 return add_extent_mapping(em_tree
, em
);
2689 struct extent_map
*btrfs_get_extent(struct inode
*inode
, struct page
*page
,
2690 size_t pg_offset
, u64 start
, u64 len
,
2696 u64 extent_start
= 0;
2698 u64 objectid
= inode
->i_ino
;
2700 struct btrfs_path
*path
= NULL
;
2701 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2702 struct btrfs_file_extent_item
*item
;
2703 struct extent_buffer
*leaf
;
2704 struct btrfs_key found_key
;
2705 struct extent_map
*em
= NULL
;
2706 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
2707 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2708 struct btrfs_trans_handle
*trans
= NULL
;
2711 spin_lock(&em_tree
->lock
);
2712 em
= lookup_extent_mapping(em_tree
, start
, len
);
2714 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2715 spin_unlock(&em_tree
->lock
);
2718 if (em
->start
> start
|| em
->start
+ em
->len
<= start
)
2719 free_extent_map(em
);
2720 else if (em
->block_start
== EXTENT_MAP_INLINE
&& page
)
2721 free_extent_map(em
);
2725 em
= alloc_extent_map(GFP_NOFS
);
2730 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2731 em
->start
= EXTENT_MAP_HOLE
;
2735 path
= btrfs_alloc_path();
2739 ret
= btrfs_lookup_file_extent(trans
, root
, path
,
2740 objectid
, start
, trans
!= NULL
);
2747 if (path
->slots
[0] == 0)
2752 leaf
= path
->nodes
[0];
2753 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
2754 struct btrfs_file_extent_item
);
2755 /* are we inside the extent that was found? */
2756 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2757 found_type
= btrfs_key_type(&found_key
);
2758 if (found_key
.objectid
!= objectid
||
2759 found_type
!= BTRFS_EXTENT_DATA_KEY
) {
2763 found_type
= btrfs_file_extent_type(leaf
, item
);
2764 extent_start
= found_key
.offset
;
2765 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
2766 extent_end
= extent_start
+
2767 btrfs_file_extent_num_bytes(leaf
, item
);
2769 if (start
< extent_start
|| start
>= extent_end
) {
2771 if (start
< extent_start
) {
2772 if (start
+ len
<= extent_start
)
2774 em
->len
= extent_end
- extent_start
;
2780 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
2782 em
->start
= extent_start
;
2783 em
->len
= extent_end
- extent_start
;
2784 em
->block_start
= EXTENT_MAP_HOLE
;
2787 bytenr
+= btrfs_file_extent_offset(leaf
, item
);
2788 em
->block_start
= bytenr
;
2789 em
->start
= extent_start
;
2790 em
->len
= extent_end
- extent_start
;
2792 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
2797 size_t extent_offset
;
2800 size
= btrfs_file_extent_inline_len(leaf
, btrfs_item_nr(leaf
,
2802 extent_end
= (extent_start
+ size
+ root
->sectorsize
- 1) &
2803 ~((u64
)root
->sectorsize
- 1);
2804 if (start
< extent_start
|| start
>= extent_end
) {
2806 if (start
< extent_start
) {
2807 if (start
+ len
<= extent_start
)
2809 em
->len
= extent_end
- extent_start
;
2815 em
->block_start
= EXTENT_MAP_INLINE
;
2818 em
->start
= extent_start
;
2823 page_start
= page_offset(page
) + pg_offset
;
2824 extent_offset
= page_start
- extent_start
;
2825 copy_size
= min_t(u64
, PAGE_CACHE_SIZE
- pg_offset
,
2826 size
- extent_offset
);
2827 em
->start
= extent_start
+ extent_offset
;
2828 em
->len
= (copy_size
+ root
->sectorsize
- 1) &
2829 ~((u64
)root
->sectorsize
- 1);
2831 ptr
= btrfs_file_extent_inline_start(item
) + extent_offset
;
2832 if (create
== 0 && !PageUptodate(page
)) {
2833 read_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2835 flush_dcache_page(page
);
2836 } else if (create
&& PageUptodate(page
)) {
2839 free_extent_map(em
);
2841 btrfs_release_path(root
, path
);
2842 trans
= btrfs_join_transaction(root
, 1);
2845 write_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2847 btrfs_mark_buffer_dirty(leaf
);
2850 set_extent_uptodate(io_tree
, em
->start
,
2851 extent_map_end(em
) - 1, GFP_NOFS
);
2854 printk("unkknown found_type %d\n", found_type
);
2861 em
->block_start
= EXTENT_MAP_HOLE
;
2863 btrfs_release_path(root
, path
);
2864 if (em
->start
> start
|| extent_map_end(em
) <= start
) {
2865 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em
->start
, em
->len
, start
, len
);
2871 spin_lock(&em_tree
->lock
);
2872 ret
= add_extent_mapping(em_tree
, em
);
2873 /* it is possible that someone inserted the extent into the tree
2874 * while we had the lock dropped. It is also possible that
2875 * an overlapping map exists in the tree
2877 if (ret
== -EEXIST
) {
2878 struct extent_map
*existing
;
2882 existing
= lookup_extent_mapping(em_tree
, start
, len
);
2883 if (existing
&& (existing
->start
> start
||
2884 existing
->start
+ existing
->len
<= start
)) {
2885 free_extent_map(existing
);
2889 existing
= lookup_extent_mapping(em_tree
, em
->start
,
2892 err
= merge_extent_mapping(em_tree
, existing
,
2895 free_extent_map(existing
);
2897 free_extent_map(em
);
2902 printk("failing to insert %Lu %Lu\n",
2904 free_extent_map(em
);
2908 free_extent_map(em
);
2913 spin_unlock(&em_tree
->lock
);
2916 btrfs_free_path(path
);
2918 ret
= btrfs_end_transaction(trans
, root
);
2924 free_extent_map(em
);
2926 return ERR_PTR(err
);
2931 #if 0 /* waiting for O_DIRECT reads */
2932 static int btrfs_get_block(struct inode
*inode
, sector_t iblock
,
2933 struct buffer_head
*bh_result
, int create
)
2935 struct extent_map
*em
;
2936 u64 start
= (u64
)iblock
<< inode
->i_blkbits
;
2937 struct btrfs_multi_bio
*multi
= NULL
;
2938 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2944 em
= btrfs_get_extent(inode
, NULL
, 0, start
, bh_result
->b_size
, 0);
2946 if (!em
|| IS_ERR(em
))
2949 if (em
->start
> start
|| em
->start
+ em
->len
<= start
) {
2953 if (em
->block_start
== EXTENT_MAP_INLINE
) {
2958 len
= em
->start
+ em
->len
- start
;
2959 len
= min_t(u64
, len
, INT_LIMIT(typeof(bh_result
->b_size
)));
2961 if (em
->block_start
== EXTENT_MAP_HOLE
||
2962 em
->block_start
== EXTENT_MAP_DELALLOC
) {
2963 bh_result
->b_size
= len
;
2967 logical
= start
- em
->start
;
2968 logical
= em
->block_start
+ logical
;
2971 ret
= btrfs_map_block(&root
->fs_info
->mapping_tree
, READ
,
2972 logical
, &map_length
, &multi
, 0);
2974 bh_result
->b_blocknr
= multi
->stripes
[0].physical
>> inode
->i_blkbits
;
2975 bh_result
->b_size
= min(map_length
, len
);
2977 bh_result
->b_bdev
= multi
->stripes
[0].dev
->bdev
;
2978 set_buffer_mapped(bh_result
);
2981 free_extent_map(em
);
2986 static ssize_t
btrfs_direct_IO(int rw
, struct kiocb
*iocb
,
2987 const struct iovec
*iov
, loff_t offset
,
2988 unsigned long nr_segs
)
2992 struct file
*file
= iocb
->ki_filp
;
2993 struct inode
*inode
= file
->f_mapping
->host
;
2998 return blockdev_direct_IO(rw
, iocb
, inode
, inode
->i_sb
->s_bdev
, iov
,
2999 offset
, nr_segs
, btrfs_get_block
, NULL
);
3003 static sector_t
btrfs_bmap(struct address_space
*mapping
, sector_t iblock
)
3005 return extent_bmap(mapping
, iblock
, btrfs_get_extent
);
3008 int btrfs_readpage(struct file
*file
, struct page
*page
)
3010 struct extent_io_tree
*tree
;
3011 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3012 return extent_read_full_page(tree
, page
, btrfs_get_extent
);
3015 static int btrfs_writepage(struct page
*page
, struct writeback_control
*wbc
)
3017 struct extent_io_tree
*tree
;
3020 if (current
->flags
& PF_MEMALLOC
) {
3021 redirty_page_for_writepage(wbc
, page
);
3025 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3026 return extent_write_full_page(tree
, page
, btrfs_get_extent
, wbc
);
3029 int btrfs_writepages(struct address_space
*mapping
,
3030 struct writeback_control
*wbc
)
3032 struct extent_io_tree
*tree
;
3033 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
3034 return extent_writepages(tree
, mapping
, btrfs_get_extent
, wbc
);
3038 btrfs_readpages(struct file
*file
, struct address_space
*mapping
,
3039 struct list_head
*pages
, unsigned nr_pages
)
3041 struct extent_io_tree
*tree
;
3042 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
3043 return extent_readpages(tree
, mapping
, pages
, nr_pages
,
3046 static int __btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
3048 struct extent_io_tree
*tree
;
3049 struct extent_map_tree
*map
;
3052 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3053 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
3054 ret
= try_release_extent_mapping(map
, tree
, page
, gfp_flags
);
3056 ClearPagePrivate(page
);
3057 set_page_private(page
, 0);
3058 page_cache_release(page
);
3063 static int btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
3065 if (PageWriteback(page
) || PageDirty(page
))
3067 return __btrfs_releasepage(page
, gfp_flags
);
3070 static void btrfs_invalidatepage(struct page
*page
, unsigned long offset
)
3072 struct extent_io_tree
*tree
;
3073 struct btrfs_ordered_extent
*ordered
;
3074 u64 page_start
= page_offset(page
);
3075 u64 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
3077 wait_on_page_writeback(page
);
3078 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3080 btrfs_releasepage(page
, GFP_NOFS
);
3084 lock_extent(tree
, page_start
, page_end
, GFP_NOFS
);
3085 ordered
= btrfs_lookup_ordered_extent(page
->mapping
->host
,
3089 * IO on this page will never be started, so we need
3090 * to account for any ordered extents now
3092 clear_extent_bit(tree
, page_start
, page_end
,
3093 EXTENT_DIRTY
| EXTENT_DELALLOC
|
3094 EXTENT_LOCKED
, 1, 0, GFP_NOFS
);
3095 btrfs_finish_ordered_io(page
->mapping
->host
,
3096 page_start
, page_end
);
3097 btrfs_put_ordered_extent(ordered
);
3098 lock_extent(tree
, page_start
, page_end
, GFP_NOFS
);
3100 clear_extent_bit(tree
, page_start
, page_end
,
3101 EXTENT_LOCKED
| EXTENT_DIRTY
| EXTENT_DELALLOC
|
3104 __btrfs_releasepage(page
, GFP_NOFS
);
3106 ClearPageChecked(page
);
3107 if (PagePrivate(page
)) {
3108 ClearPagePrivate(page
);
3109 set_page_private(page
, 0);
3110 page_cache_release(page
);
3115 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
3116 * called from a page fault handler when a page is first dirtied. Hence we must
3117 * be careful to check for EOF conditions here. We set the page up correctly
3118 * for a written page which means we get ENOSPC checking when writing into
3119 * holes and correct delalloc and unwritten extent mapping on filesystems that
3120 * support these features.
3122 * We are not allowed to take the i_mutex here so we have to play games to
3123 * protect against truncate races as the page could now be beyond EOF. Because
3124 * vmtruncate() writes the inode size before removing pages, once we have the
3125 * page lock we can determine safely if the page is beyond EOF. If it is not
3126 * beyond EOF, then the page is guaranteed safe against truncation until we
3129 int btrfs_page_mkwrite(struct vm_area_struct
*vma
, struct page
*page
)
3131 struct inode
*inode
= fdentry(vma
->vm_file
)->d_inode
;
3132 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3133 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
3134 struct btrfs_ordered_extent
*ordered
;
3136 unsigned long zero_start
;
3142 ret
= btrfs_check_free_space(root
, PAGE_CACHE_SIZE
, 0);
3149 size
= i_size_read(inode
);
3150 page_start
= page_offset(page
);
3151 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
3153 if ((page
->mapping
!= inode
->i_mapping
) ||
3154 (page_start
>= size
)) {
3155 /* page got truncated out from underneath us */
3158 wait_on_page_writeback(page
);
3160 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3161 set_page_extent_mapped(page
);
3164 * we can't set the delalloc bits if there are pending ordered
3165 * extents. Drop our locks and wait for them to finish
3167 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
3169 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3171 btrfs_start_ordered_extent(inode
, ordered
, 1);
3172 btrfs_put_ordered_extent(ordered
);
3176 btrfs_set_extent_delalloc(inode
, page_start
, page_end
);
3179 /* page is wholly or partially inside EOF */
3180 if (page_start
+ PAGE_CACHE_SIZE
> size
)
3181 zero_start
= size
& ~PAGE_CACHE_MASK
;
3183 zero_start
= PAGE_CACHE_SIZE
;
3185 if (zero_start
!= PAGE_CACHE_SIZE
) {
3187 memset(kaddr
+ zero_start
, 0, PAGE_CACHE_SIZE
- zero_start
);
3188 flush_dcache_page(page
);
3191 ClearPageChecked(page
);
3192 set_page_dirty(page
);
3193 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3201 static void btrfs_truncate(struct inode
*inode
)
3203 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3205 struct btrfs_trans_handle
*trans
;
3207 u64 mask
= root
->sectorsize
- 1;
3209 if (!S_ISREG(inode
->i_mode
))
3211 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
3214 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
3215 btrfs_wait_ordered_range(inode
, inode
->i_size
& (~mask
), (u64
)-1);
3217 trans
= btrfs_start_transaction(root
, 1);
3218 btrfs_set_trans_block_group(trans
, inode
);
3219 btrfs_i_size_write(inode
, inode
->i_size
);
3221 ret
= btrfs_orphan_add(trans
, inode
);
3224 /* FIXME, add redo link to tree so we don't leak on crash */
3225 ret
= btrfs_truncate_inode_items(trans
, root
, inode
, inode
->i_size
,
3226 BTRFS_EXTENT_DATA_KEY
);
3227 btrfs_update_inode(trans
, root
, inode
);
3229 ret
= btrfs_orphan_del(trans
, inode
);
3233 nr
= trans
->blocks_used
;
3234 ret
= btrfs_end_transaction_throttle(trans
, root
);
3236 btrfs_btree_balance_dirty(root
, nr
);
3240 * Invalidate a single dcache entry at the root of the filesystem.
3241 * Needed after creation of snapshot or subvolume.
3243 void btrfs_invalidate_dcache_root(struct btrfs_root
*root
, char *name
,
3246 struct dentry
*alias
, *entry
;
3249 alias
= d_find_alias(root
->fs_info
->sb
->s_root
->d_inode
);
3253 /* change me if btrfs ever gets a d_hash operation */
3254 qstr
.hash
= full_name_hash(qstr
.name
, qstr
.len
);
3255 entry
= d_lookup(alias
, &qstr
);
3258 d_invalidate(entry
);
3264 int btrfs_create_subvol_root(struct btrfs_root
*new_root
,
3265 struct btrfs_trans_handle
*trans
, u64 new_dirid
,
3266 struct btrfs_block_group_cache
*block_group
)
3268 struct inode
*inode
;
3271 inode
= btrfs_new_inode(trans
, new_root
, NULL
, "..", 2, new_dirid
,
3272 new_dirid
, block_group
, S_IFDIR
| 0700, &index
);
3274 return PTR_ERR(inode
);
3275 inode
->i_op
= &btrfs_dir_inode_operations
;
3276 inode
->i_fop
= &btrfs_dir_file_operations
;
3277 new_root
->inode
= inode
;
3280 btrfs_i_size_write(inode
, 0);
3282 return btrfs_update_inode(trans
, new_root
, inode
);
3285 unsigned long btrfs_force_ra(struct address_space
*mapping
,
3286 struct file_ra_state
*ra
, struct file
*file
,
3287 pgoff_t offset
, pgoff_t last_index
)
3289 pgoff_t req_size
= last_index
- offset
+ 1;
3291 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3292 offset
= page_cache_readahead(mapping
, ra
, file
, offset
, req_size
);
3295 page_cache_sync_readahead(mapping
, ra
, file
, offset
, req_size
);
3296 return offset
+ req_size
;
3300 struct inode
*btrfs_alloc_inode(struct super_block
*sb
)
3302 struct btrfs_inode
*ei
;
3304 ei
= kmem_cache_alloc(btrfs_inode_cachep
, GFP_NOFS
);
3308 ei
->logged_trans
= 0;
3309 btrfs_ordered_inode_tree_init(&ei
->ordered_tree
);
3310 ei
->i_acl
= BTRFS_ACL_NOT_CACHED
;
3311 ei
->i_default_acl
= BTRFS_ACL_NOT_CACHED
;
3312 INIT_LIST_HEAD(&ei
->i_orphan
);
3313 return &ei
->vfs_inode
;
3316 void btrfs_destroy_inode(struct inode
*inode
)
3318 struct btrfs_ordered_extent
*ordered
;
3319 WARN_ON(!list_empty(&inode
->i_dentry
));
3320 WARN_ON(inode
->i_data
.nrpages
);
3322 if (BTRFS_I(inode
)->i_acl
&&
3323 BTRFS_I(inode
)->i_acl
!= BTRFS_ACL_NOT_CACHED
)
3324 posix_acl_release(BTRFS_I(inode
)->i_acl
);
3325 if (BTRFS_I(inode
)->i_default_acl
&&
3326 BTRFS_I(inode
)->i_default_acl
!= BTRFS_ACL_NOT_CACHED
)
3327 posix_acl_release(BTRFS_I(inode
)->i_default_acl
);
3329 spin_lock(&BTRFS_I(inode
)->root
->list_lock
);
3330 if (!list_empty(&BTRFS_I(inode
)->i_orphan
)) {
3331 printk(KERN_ERR
"BTRFS: inode %lu: inode still on the orphan"
3332 " list\n", inode
->i_ino
);
3335 spin_unlock(&BTRFS_I(inode
)->root
->list_lock
);
3338 ordered
= btrfs_lookup_first_ordered_extent(inode
, (u64
)-1);
3342 printk("found ordered extent %Lu %Lu\n",
3343 ordered
->file_offset
, ordered
->len
);
3344 btrfs_remove_ordered_extent(inode
, ordered
);
3345 btrfs_put_ordered_extent(ordered
);
3346 btrfs_put_ordered_extent(ordered
);
3349 btrfs_drop_extent_cache(inode
, 0, (u64
)-1);
3350 kmem_cache_free(btrfs_inode_cachep
, BTRFS_I(inode
));
3353 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3354 static void init_once(void *foo
)
3355 #elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3356 static void init_once(struct kmem_cache
* cachep
, void *foo
)
3358 static void init_once(void * foo
, struct kmem_cache
* cachep
,
3359 unsigned long flags
)
3362 struct btrfs_inode
*ei
= (struct btrfs_inode
*) foo
;
3364 inode_init_once(&ei
->vfs_inode
);
3367 void btrfs_destroy_cachep(void)
3369 if (btrfs_inode_cachep
)
3370 kmem_cache_destroy(btrfs_inode_cachep
);
3371 if (btrfs_trans_handle_cachep
)
3372 kmem_cache_destroy(btrfs_trans_handle_cachep
);
3373 if (btrfs_transaction_cachep
)
3374 kmem_cache_destroy(btrfs_transaction_cachep
);
3375 if (btrfs_bit_radix_cachep
)
3376 kmem_cache_destroy(btrfs_bit_radix_cachep
);
3377 if (btrfs_path_cachep
)
3378 kmem_cache_destroy(btrfs_path_cachep
);
3381 struct kmem_cache
*btrfs_cache_create(const char *name
, size_t size
,
3382 unsigned long extra_flags
,
3383 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3384 void (*ctor
)(void *)
3385 #elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3386 void (*ctor
)(struct kmem_cache
*, void *)
3388 void (*ctor
)(void *, struct kmem_cache
*,
3393 return kmem_cache_create(name
, size
, 0, (SLAB_RECLAIM_ACCOUNT
|
3394 SLAB_MEM_SPREAD
| extra_flags
), ctor
3395 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3401 int btrfs_init_cachep(void)
3403 btrfs_inode_cachep
= btrfs_cache_create("btrfs_inode_cache",
3404 sizeof(struct btrfs_inode
),
3406 if (!btrfs_inode_cachep
)
3408 btrfs_trans_handle_cachep
=
3409 btrfs_cache_create("btrfs_trans_handle_cache",
3410 sizeof(struct btrfs_trans_handle
),
3412 if (!btrfs_trans_handle_cachep
)
3414 btrfs_transaction_cachep
= btrfs_cache_create("btrfs_transaction_cache",
3415 sizeof(struct btrfs_transaction
),
3417 if (!btrfs_transaction_cachep
)
3419 btrfs_path_cachep
= btrfs_cache_create("btrfs_path_cache",
3420 sizeof(struct btrfs_path
),
3422 if (!btrfs_path_cachep
)
3424 btrfs_bit_radix_cachep
= btrfs_cache_create("btrfs_radix", 256,
3425 SLAB_DESTROY_BY_RCU
, NULL
);
3426 if (!btrfs_bit_radix_cachep
)
3430 btrfs_destroy_cachep();
3434 static int btrfs_getattr(struct vfsmount
*mnt
,
3435 struct dentry
*dentry
, struct kstat
*stat
)
3437 struct inode
*inode
= dentry
->d_inode
;
3438 generic_fillattr(inode
, stat
);
3439 stat
->blksize
= PAGE_CACHE_SIZE
;
3440 stat
->blocks
= inode
->i_blocks
+ (BTRFS_I(inode
)->delalloc_bytes
>> 9);
3444 static int btrfs_rename(struct inode
* old_dir
, struct dentry
*old_dentry
,
3445 struct inode
* new_dir
,struct dentry
*new_dentry
)
3447 struct btrfs_trans_handle
*trans
;
3448 struct btrfs_root
*root
= BTRFS_I(old_dir
)->root
;
3449 struct inode
*new_inode
= new_dentry
->d_inode
;
3450 struct inode
*old_inode
= old_dentry
->d_inode
;
3451 struct timespec ctime
= CURRENT_TIME
;
3455 if (S_ISDIR(old_inode
->i_mode
) && new_inode
&&
3456 new_inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
3460 ret
= btrfs_check_free_space(root
, 1, 0);
3464 trans
= btrfs_start_transaction(root
, 1);
3466 btrfs_set_trans_block_group(trans
, new_dir
);
3468 btrfs_inc_nlink(old_dentry
->d_inode
);
3469 old_dir
->i_ctime
= old_dir
->i_mtime
= ctime
;
3470 new_dir
->i_ctime
= new_dir
->i_mtime
= ctime
;
3471 old_inode
->i_ctime
= ctime
;
3473 ret
= btrfs_unlink_inode(trans
, root
, old_dir
, old_dentry
->d_inode
,
3474 old_dentry
->d_name
.name
,
3475 old_dentry
->d_name
.len
);
3480 new_inode
->i_ctime
= CURRENT_TIME
;
3481 ret
= btrfs_unlink_inode(trans
, root
, new_dir
,
3482 new_dentry
->d_inode
,
3483 new_dentry
->d_name
.name
,
3484 new_dentry
->d_name
.len
);
3487 if (new_inode
->i_nlink
== 0) {
3488 ret
= btrfs_orphan_add(trans
, new_dentry
->d_inode
);
3494 ret
= btrfs_set_inode_index(new_dir
, old_inode
, &index
);
3498 ret
= btrfs_add_link(trans
, new_dentry
->d_parent
->d_inode
,
3499 old_inode
, new_dentry
->d_name
.name
,
3500 new_dentry
->d_name
.len
, 1, index
);
3505 btrfs_end_transaction_throttle(trans
, root
);
3510 int btrfs_start_delalloc_inodes(struct btrfs_root
*root
)
3512 struct list_head
*head
= &root
->fs_info
->delalloc_inodes
;
3513 struct btrfs_inode
*binode
;
3514 unsigned long flags
;
3516 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
3517 while(!list_empty(head
)) {
3518 binode
= list_entry(head
->next
, struct btrfs_inode
,
3520 atomic_inc(&binode
->vfs_inode
.i_count
);
3521 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
3522 filemap_write_and_wait(binode
->vfs_inode
.i_mapping
);
3523 iput(&binode
->vfs_inode
);
3524 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
3526 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
3530 static int btrfs_symlink(struct inode
*dir
, struct dentry
*dentry
,
3531 const char *symname
)
3533 struct btrfs_trans_handle
*trans
;
3534 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
3535 struct btrfs_path
*path
;
3536 struct btrfs_key key
;
3537 struct inode
*inode
= NULL
;
3545 struct btrfs_file_extent_item
*ei
;
3546 struct extent_buffer
*leaf
;
3547 unsigned long nr
= 0;
3549 name_len
= strlen(symname
) + 1;
3550 if (name_len
> BTRFS_MAX_INLINE_DATA_SIZE(root
))
3551 return -ENAMETOOLONG
;
3553 err
= btrfs_check_free_space(root
, 1, 0);
3557 trans
= btrfs_start_transaction(root
, 1);
3558 btrfs_set_trans_block_group(trans
, dir
);
3560 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
3566 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
3568 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
3569 BTRFS_I(dir
)->block_group
, S_IFLNK
|S_IRWXUGO
,
3571 err
= PTR_ERR(inode
);
3575 err
= btrfs_init_acl(inode
, dir
);
3581 btrfs_set_trans_block_group(trans
, inode
);
3582 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
3586 inode
->i_mapping
->a_ops
= &btrfs_aops
;
3587 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3588 inode
->i_fop
= &btrfs_file_operations
;
3589 inode
->i_op
= &btrfs_file_inode_operations
;
3590 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
3592 dir
->i_sb
->s_dirt
= 1;
3593 btrfs_update_inode_block_group(trans
, inode
);
3594 btrfs_update_inode_block_group(trans
, dir
);
3598 path
= btrfs_alloc_path();
3600 key
.objectid
= inode
->i_ino
;
3602 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
3603 datasize
= btrfs_file_extent_calc_inline_size(name_len
);
3604 err
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
3610 leaf
= path
->nodes
[0];
3611 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
3612 struct btrfs_file_extent_item
);
3613 btrfs_set_file_extent_generation(leaf
, ei
, trans
->transid
);
3614 btrfs_set_file_extent_type(leaf
, ei
,
3615 BTRFS_FILE_EXTENT_INLINE
);
3616 ptr
= btrfs_file_extent_inline_start(ei
);
3617 write_extent_buffer(leaf
, symname
, ptr
, name_len
);
3618 btrfs_mark_buffer_dirty(leaf
);
3619 btrfs_free_path(path
);
3621 inode
->i_op
= &btrfs_symlink_inode_operations
;
3622 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
3623 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3624 btrfs_i_size_write(inode
, name_len
- 1);
3625 err
= btrfs_update_inode(trans
, root
, inode
);
3630 nr
= trans
->blocks_used
;
3631 btrfs_end_transaction_throttle(trans
, root
);
3634 inode_dec_link_count(inode
);
3637 btrfs_btree_balance_dirty(root
, nr
);
3641 static int btrfs_set_page_dirty(struct page
*page
)
3643 return __set_page_dirty_nobuffers(page
);
3646 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3647 static int btrfs_permission(struct inode
*inode
, int mask
)
3649 static int btrfs_permission(struct inode
*inode
, int mask
,
3650 struct nameidata
*nd
)
3653 if (btrfs_test_flag(inode
, READONLY
) && (mask
& MAY_WRITE
))
3655 return generic_permission(inode
, mask
, btrfs_check_acl
);
3658 static struct inode_operations btrfs_dir_inode_operations
= {
3659 .lookup
= btrfs_lookup
,
3660 .create
= btrfs_create
,
3661 .unlink
= btrfs_unlink
,
3663 .mkdir
= btrfs_mkdir
,
3664 .rmdir
= btrfs_rmdir
,
3665 .rename
= btrfs_rename
,
3666 .symlink
= btrfs_symlink
,
3667 .setattr
= btrfs_setattr
,
3668 .mknod
= btrfs_mknod
,
3669 .setxattr
= btrfs_setxattr
,
3670 .getxattr
= btrfs_getxattr
,
3671 .listxattr
= btrfs_listxattr
,
3672 .removexattr
= btrfs_removexattr
,
3673 .permission
= btrfs_permission
,
3675 static struct inode_operations btrfs_dir_ro_inode_operations
= {
3676 .lookup
= btrfs_lookup
,
3677 .permission
= btrfs_permission
,
3679 static struct file_operations btrfs_dir_file_operations
= {
3680 .llseek
= generic_file_llseek
,
3681 .read
= generic_read_dir
,
3682 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28)
3683 .readdir
= btrfs_nfshack_readdir
,
3684 #else /* NFSd readdir/lookup deadlock is fixed */
3685 .readdir
= btrfs_real_readdir
,
3687 .unlocked_ioctl
= btrfs_ioctl
,
3688 #ifdef CONFIG_COMPAT
3689 .compat_ioctl
= btrfs_ioctl
,
3691 .release
= btrfs_release_file
,
3692 .fsync
= btrfs_sync_file
,
3695 static struct extent_io_ops btrfs_extent_io_ops
= {
3696 .fill_delalloc
= run_delalloc_range
,
3697 .submit_bio_hook
= btrfs_submit_bio_hook
,
3698 .merge_bio_hook
= btrfs_merge_bio_hook
,
3699 .readpage_end_io_hook
= btrfs_readpage_end_io_hook
,
3700 .writepage_end_io_hook
= btrfs_writepage_end_io_hook
,
3701 .writepage_start_hook
= btrfs_writepage_start_hook
,
3702 .readpage_io_failed_hook
= btrfs_io_failed_hook
,
3703 .set_bit_hook
= btrfs_set_bit_hook
,
3704 .clear_bit_hook
= btrfs_clear_bit_hook
,
3707 static struct address_space_operations btrfs_aops
= {
3708 .readpage
= btrfs_readpage
,
3709 .writepage
= btrfs_writepage
,
3710 .writepages
= btrfs_writepages
,
3711 .readpages
= btrfs_readpages
,
3712 .sync_page
= block_sync_page
,
3714 .direct_IO
= btrfs_direct_IO
,
3715 .invalidatepage
= btrfs_invalidatepage
,
3716 .releasepage
= btrfs_releasepage
,
3717 .set_page_dirty
= btrfs_set_page_dirty
,
3720 static struct address_space_operations btrfs_symlink_aops
= {
3721 .readpage
= btrfs_readpage
,
3722 .writepage
= btrfs_writepage
,
3723 .invalidatepage
= btrfs_invalidatepage
,
3724 .releasepage
= btrfs_releasepage
,
3727 static struct inode_operations btrfs_file_inode_operations
= {
3728 .truncate
= btrfs_truncate
,
3729 .getattr
= btrfs_getattr
,
3730 .setattr
= btrfs_setattr
,
3731 .setxattr
= btrfs_setxattr
,
3732 .getxattr
= btrfs_getxattr
,
3733 .listxattr
= btrfs_listxattr
,
3734 .removexattr
= btrfs_removexattr
,
3735 .permission
= btrfs_permission
,
3737 static struct inode_operations btrfs_special_inode_operations
= {
3738 .getattr
= btrfs_getattr
,
3739 .setattr
= btrfs_setattr
,
3740 .permission
= btrfs_permission
,
3741 .setxattr
= btrfs_setxattr
,
3742 .getxattr
= btrfs_getxattr
,
3743 .listxattr
= btrfs_listxattr
,
3744 .removexattr
= btrfs_removexattr
,
3746 static struct inode_operations btrfs_symlink_inode_operations
= {
3747 .readlink
= generic_readlink
,
3748 .follow_link
= page_follow_link_light
,
3749 .put_link
= page_put_link
,
3750 .permission
= btrfs_permission
,