2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/pagemap.h>
25 #include <linux/highmem.h>
26 #include <linux/time.h>
27 #include <linux/init.h>
28 #include <linux/string.h>
29 #include <linux/smp_lock.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mpage.h>
32 #include <linux/swap.h>
33 #include <linux/writeback.h>
34 #include <linux/statfs.h>
35 #include <linux/compat.h>
36 #include <linux/bit_spinlock.h>
37 #include <linux/version.h>
38 #include <linux/xattr.h>
39 #include <linux/posix_acl.h>
42 #include "transaction.h"
43 #include "btrfs_inode.h"
45 #include "print-tree.h"
47 #include "ordered-data.h"
49 struct btrfs_iget_args
{
51 struct btrfs_root
*root
;
54 static struct inode_operations btrfs_dir_inode_operations
;
55 static struct inode_operations btrfs_symlink_inode_operations
;
56 static struct inode_operations btrfs_dir_ro_inode_operations
;
57 static struct inode_operations btrfs_special_inode_operations
;
58 static struct inode_operations btrfs_file_inode_operations
;
59 static struct address_space_operations btrfs_aops
;
60 static struct address_space_operations btrfs_symlink_aops
;
61 static struct file_operations btrfs_dir_file_operations
;
62 static struct extent_io_ops btrfs_extent_io_ops
;
64 static struct kmem_cache
*btrfs_inode_cachep
;
65 struct kmem_cache
*btrfs_trans_handle_cachep
;
66 struct kmem_cache
*btrfs_transaction_cachep
;
67 struct kmem_cache
*btrfs_bit_radix_cachep
;
68 struct kmem_cache
*btrfs_path_cachep
;
71 static unsigned char btrfs_type_by_mode
[S_IFMT
>> S_SHIFT
] = {
72 [S_IFREG
>> S_SHIFT
] = BTRFS_FT_REG_FILE
,
73 [S_IFDIR
>> S_SHIFT
] = BTRFS_FT_DIR
,
74 [S_IFCHR
>> S_SHIFT
] = BTRFS_FT_CHRDEV
,
75 [S_IFBLK
>> S_SHIFT
] = BTRFS_FT_BLKDEV
,
76 [S_IFIFO
>> S_SHIFT
] = BTRFS_FT_FIFO
,
77 [S_IFSOCK
>> S_SHIFT
] = BTRFS_FT_SOCK
,
78 [S_IFLNK
>> S_SHIFT
] = BTRFS_FT_SYMLINK
,
81 static void btrfs_truncate(struct inode
*inode
);
83 int btrfs_check_free_space(struct btrfs_root
*root
, u64 num_required
,
92 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
93 total
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
94 used
= btrfs_super_bytes_used(&root
->fs_info
->super_copy
);
102 if (used
+ root
->fs_info
->delalloc_bytes
+ num_required
> thresh
)
104 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
108 static int cow_file_range(struct inode
*inode
, u64 start
, u64 end
)
110 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
111 struct btrfs_trans_handle
*trans
;
115 u64 blocksize
= root
->sectorsize
;
117 struct btrfs_key ins
;
118 struct extent_map
*em
;
119 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
122 trans
= btrfs_join_transaction(root
, 1);
124 btrfs_set_trans_block_group(trans
, inode
);
126 num_bytes
= (end
- start
+ blocksize
) & ~(blocksize
- 1);
127 num_bytes
= max(blocksize
, num_bytes
);
128 orig_num_bytes
= num_bytes
;
130 if (alloc_hint
== EXTENT_MAP_INLINE
)
133 BUG_ON(num_bytes
> btrfs_super_total_bytes(&root
->fs_info
->super_copy
));
134 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
135 btrfs_drop_extent_cache(inode
, start
, start
+ num_bytes
- 1);
136 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
138 while(num_bytes
> 0) {
139 cur_alloc_size
= min(num_bytes
, root
->fs_info
->max_extent
);
140 ret
= btrfs_reserve_extent(trans
, root
, cur_alloc_size
,
141 root
->sectorsize
, 0, 0,
147 em
= alloc_extent_map(GFP_NOFS
);
149 em
->len
= ins
.offset
;
150 em
->block_start
= ins
.objectid
;
151 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
152 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
153 set_bit(EXTENT_FLAG_PINNED
, &em
->flags
);
155 spin_lock(&em_tree
->lock
);
156 ret
= add_extent_mapping(em_tree
, em
);
157 spin_unlock(&em_tree
->lock
);
158 if (ret
!= -EEXIST
) {
162 btrfs_drop_extent_cache(inode
, start
,
163 start
+ ins
.offset
- 1);
165 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
167 cur_alloc_size
= ins
.offset
;
168 ret
= btrfs_add_ordered_extent(inode
, start
, ins
.objectid
,
171 if (num_bytes
< cur_alloc_size
) {
172 printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes
,
176 num_bytes
-= cur_alloc_size
;
177 alloc_hint
= ins
.objectid
+ ins
.offset
;
178 start
+= cur_alloc_size
;
181 btrfs_end_transaction(trans
, root
);
185 static int run_delalloc_nocow(struct inode
*inode
, u64 start
, u64 end
)
192 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
193 struct btrfs_block_group_cache
*block_group
;
194 struct btrfs_trans_handle
*trans
;
195 struct extent_buffer
*leaf
;
197 struct btrfs_path
*path
;
198 struct btrfs_file_extent_item
*item
;
201 struct btrfs_key found_key
;
203 total_fs_bytes
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
204 path
= btrfs_alloc_path();
206 trans
= btrfs_join_transaction(root
, 1);
209 ret
= btrfs_lookup_file_extent(NULL
, root
, path
,
210 inode
->i_ino
, start
, 0);
217 if (path
->slots
[0] == 0)
222 leaf
= path
->nodes
[0];
223 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
224 struct btrfs_file_extent_item
);
226 /* are we inside the extent that was found? */
227 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
228 found_type
= btrfs_key_type(&found_key
);
229 if (found_key
.objectid
!= inode
->i_ino
||
230 found_type
!= BTRFS_EXTENT_DATA_KEY
)
233 found_type
= btrfs_file_extent_type(leaf
, item
);
234 extent_start
= found_key
.offset
;
235 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
236 u64 extent_num_bytes
;
238 extent_num_bytes
= btrfs_file_extent_num_bytes(leaf
, item
);
239 extent_end
= extent_start
+ extent_num_bytes
;
242 if (loops
&& start
!= extent_start
)
245 if (start
< extent_start
|| start
>= extent_end
)
248 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
252 if (btrfs_cross_ref_exists(trans
, root
, &found_key
, bytenr
))
255 * we may be called by the resizer, make sure we're inside
256 * the limits of the FS
258 block_group
= btrfs_lookup_block_group(root
->fs_info
,
260 if (!block_group
|| block_group
->ro
)
263 bytenr
+= btrfs_file_extent_offset(leaf
, item
);
264 extent_num_bytes
= min(end
+ 1, extent_end
) - start
;
265 ret
= btrfs_add_ordered_extent(inode
, start
, bytenr
,
266 extent_num_bytes
, 1);
272 btrfs_release_path(root
, path
);
280 btrfs_end_transaction(trans
, root
);
281 btrfs_free_path(path
);
282 return cow_file_range(inode
, start
, end
);
286 btrfs_end_transaction(trans
, root
);
287 btrfs_free_path(path
);
291 static int run_delalloc_range(struct inode
*inode
, u64 start
, u64 end
)
293 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
296 if (btrfs_test_opt(root
, NODATACOW
) ||
297 btrfs_test_flag(inode
, NODATACOW
))
298 ret
= run_delalloc_nocow(inode
, start
, end
);
300 ret
= cow_file_range(inode
, start
, end
);
305 int btrfs_set_bit_hook(struct inode
*inode
, u64 start
, u64 end
,
306 unsigned long old
, unsigned long bits
)
309 if (!(old
& EXTENT_DELALLOC
) && (bits
& EXTENT_DELALLOC
)) {
310 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
311 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
312 BTRFS_I(inode
)->delalloc_bytes
+= end
- start
+ 1;
313 root
->fs_info
->delalloc_bytes
+= end
- start
+ 1;
314 if (list_empty(&BTRFS_I(inode
)->delalloc_inodes
)) {
315 list_add_tail(&BTRFS_I(inode
)->delalloc_inodes
,
316 &root
->fs_info
->delalloc_inodes
);
318 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
323 int btrfs_clear_bit_hook(struct inode
*inode
, u64 start
, u64 end
,
324 unsigned long old
, unsigned long bits
)
326 if ((old
& EXTENT_DELALLOC
) && (bits
& EXTENT_DELALLOC
)) {
327 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
330 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
331 if (end
- start
+ 1 > root
->fs_info
->delalloc_bytes
) {
332 printk("warning: delalloc account %Lu %Lu\n",
333 end
- start
+ 1, root
->fs_info
->delalloc_bytes
);
334 root
->fs_info
->delalloc_bytes
= 0;
335 BTRFS_I(inode
)->delalloc_bytes
= 0;
337 root
->fs_info
->delalloc_bytes
-= end
- start
+ 1;
338 BTRFS_I(inode
)->delalloc_bytes
-= end
- start
+ 1;
340 if (BTRFS_I(inode
)->delalloc_bytes
== 0 &&
341 !list_empty(&BTRFS_I(inode
)->delalloc_inodes
)) {
342 list_del_init(&BTRFS_I(inode
)->delalloc_inodes
);
344 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
349 int btrfs_merge_bio_hook(struct page
*page
, unsigned long offset
,
350 size_t size
, struct bio
*bio
)
352 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
353 struct btrfs_mapping_tree
*map_tree
;
354 u64 logical
= bio
->bi_sector
<< 9;
359 length
= bio
->bi_size
;
360 map_tree
= &root
->fs_info
->mapping_tree
;
362 ret
= btrfs_map_block(map_tree
, READ
, logical
,
363 &map_length
, NULL
, 0);
365 if (map_length
< length
+ size
) {
371 int __btrfs_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
374 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
377 ret
= btrfs_csum_one_bio(root
, inode
, bio
);
380 return btrfs_map_bio(root
, rw
, bio
, mirror_num
, 1);
383 int btrfs_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
386 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
389 ret
= btrfs_bio_wq_end_io(root
->fs_info
, bio
, 0);
392 if (!(rw
& (1 << BIO_RW
))) {
396 if (btrfs_test_opt(root
, NODATASUM
) ||
397 btrfs_test_flag(inode
, NODATASUM
)) {
401 return btrfs_wq_submit_bio(BTRFS_I(inode
)->root
->fs_info
,
402 inode
, rw
, bio
, mirror_num
,
403 __btrfs_submit_bio_hook
);
405 return btrfs_map_bio(root
, rw
, bio
, mirror_num
, 0);
408 static noinline
int add_pending_csums(struct btrfs_trans_handle
*trans
,
409 struct inode
*inode
, u64 file_offset
,
410 struct list_head
*list
)
412 struct list_head
*cur
;
413 struct btrfs_ordered_sum
*sum
;
415 btrfs_set_trans_block_group(trans
, inode
);
416 list_for_each(cur
, list
) {
417 sum
= list_entry(cur
, struct btrfs_ordered_sum
, list
);
418 btrfs_csum_file_blocks(trans
, BTRFS_I(inode
)->root
,
424 int btrfs_set_extent_delalloc(struct inode
*inode
, u64 start
, u64 end
)
426 return set_extent_delalloc(&BTRFS_I(inode
)->io_tree
, start
, end
,
430 struct btrfs_writepage_fixup
{
432 struct btrfs_work work
;
435 /* see btrfs_writepage_start_hook for details on why this is required */
436 void btrfs_writepage_fixup_worker(struct btrfs_work
*work
)
438 struct btrfs_writepage_fixup
*fixup
;
439 struct btrfs_ordered_extent
*ordered
;
445 fixup
= container_of(work
, struct btrfs_writepage_fixup
, work
);
449 if (!page
->mapping
|| !PageDirty(page
) || !PageChecked(page
)) {
450 ClearPageChecked(page
);
454 inode
= page
->mapping
->host
;
455 page_start
= page_offset(page
);
456 page_end
= page_offset(page
) + PAGE_CACHE_SIZE
- 1;
458 lock_extent(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
, GFP_NOFS
);
460 /* already ordered? We're done */
461 if (test_range_bit(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
,
462 EXTENT_ORDERED
, 0)) {
466 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
468 unlock_extent(&BTRFS_I(inode
)->io_tree
, page_start
,
471 btrfs_start_ordered_extent(inode
, ordered
, 1);
475 btrfs_set_extent_delalloc(inode
, page_start
, page_end
);
476 ClearPageChecked(page
);
478 unlock_extent(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
, GFP_NOFS
);
481 page_cache_release(page
);
485 * There are a few paths in the higher layers of the kernel that directly
486 * set the page dirty bit without asking the filesystem if it is a
487 * good idea. This causes problems because we want to make sure COW
488 * properly happens and the data=ordered rules are followed.
490 * In our case any range that doesn't have the EXTENT_ORDERED bit set
491 * hasn't been properly setup for IO. We kick off an async process
492 * to fix it up. The async helper will wait for ordered extents, set
493 * the delalloc bit and make it safe to write the page.
495 int btrfs_writepage_start_hook(struct page
*page
, u64 start
, u64 end
)
497 struct inode
*inode
= page
->mapping
->host
;
498 struct btrfs_writepage_fixup
*fixup
;
499 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
502 ret
= test_range_bit(&BTRFS_I(inode
)->io_tree
, start
, end
,
507 if (PageChecked(page
))
510 fixup
= kzalloc(sizeof(*fixup
), GFP_NOFS
);
514 SetPageChecked(page
);
515 page_cache_get(page
);
516 fixup
->work
.func
= btrfs_writepage_fixup_worker
;
518 btrfs_queue_worker(&root
->fs_info
->fixup_workers
, &fixup
->work
);
522 static int btrfs_finish_ordered_io(struct inode
*inode
, u64 start
, u64 end
)
524 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
525 struct btrfs_trans_handle
*trans
;
526 struct btrfs_ordered_extent
*ordered_extent
;
527 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
529 struct list_head list
;
530 struct btrfs_key ins
;
533 ret
= btrfs_dec_test_ordered_pending(inode
, start
, end
- start
+ 1);
537 trans
= btrfs_join_transaction(root
, 1);
539 ordered_extent
= btrfs_lookup_ordered_extent(inode
, start
);
540 BUG_ON(!ordered_extent
);
541 if (test_bit(BTRFS_ORDERED_NOCOW
, &ordered_extent
->flags
))
544 lock_extent(io_tree
, ordered_extent
->file_offset
,
545 ordered_extent
->file_offset
+ ordered_extent
->len
- 1,
548 INIT_LIST_HEAD(&list
);
550 ins
.objectid
= ordered_extent
->start
;
551 ins
.offset
= ordered_extent
->len
;
552 ins
.type
= BTRFS_EXTENT_ITEM_KEY
;
554 ret
= btrfs_alloc_reserved_extent(trans
, root
, root
->root_key
.objectid
,
555 trans
->transid
, inode
->i_ino
,
556 ordered_extent
->file_offset
, &ins
);
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
);
567 ret
= btrfs_insert_file_extent(trans
, root
, inode
->i_ino
,
568 ordered_extent
->file_offset
,
569 ordered_extent
->start
,
571 ordered_extent
->len
, 0);
574 btrfs_drop_extent_cache(inode
, ordered_extent
->file_offset
,
575 ordered_extent
->file_offset
+
576 ordered_extent
->len
- 1);
577 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
579 inode
->i_blocks
+= ordered_extent
->len
>> 9;
580 unlock_extent(io_tree
, ordered_extent
->file_offset
,
581 ordered_extent
->file_offset
+ ordered_extent
->len
- 1,
584 add_pending_csums(trans
, inode
, ordered_extent
->file_offset
,
585 &ordered_extent
->list
);
587 btrfs_ordered_update_i_size(inode
, ordered_extent
);
588 btrfs_remove_ordered_extent(inode
, ordered_extent
);
591 btrfs_put_ordered_extent(ordered_extent
);
592 /* once for the tree */
593 btrfs_put_ordered_extent(ordered_extent
);
595 btrfs_update_inode(trans
, root
, inode
);
596 btrfs_end_transaction(trans
, root
);
600 int btrfs_writepage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
601 struct extent_state
*state
, int uptodate
)
603 return btrfs_finish_ordered_io(page
->mapping
->host
, start
, end
);
606 int btrfs_readpage_io_hook(struct page
*page
, u64 start
, u64 end
)
609 struct inode
*inode
= page
->mapping
->host
;
610 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
611 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
612 struct btrfs_csum_item
*item
;
613 struct btrfs_path
*path
= NULL
;
616 if (btrfs_test_opt(root
, NODATASUM
) ||
617 btrfs_test_flag(inode
, NODATASUM
))
621 * It is possible there is an ordered extent that has
622 * not yet finished for this range in the file. If so,
623 * that extent will have a csum cached, and it will insert
624 * the sum after all the blocks in the extent are fully
625 * on disk. So, look for an ordered extent and use the
626 * sum if found. We have to do this before looking in the
627 * btree because csum items are pre-inserted based on
628 * the file size. btrfs_lookup_csum might find an item
629 * that still hasn't been fully filled.
631 ret
= btrfs_find_ordered_sum(inode
, start
, &csum
);
636 path
= btrfs_alloc_path();
637 item
= btrfs_lookup_csum(NULL
, root
, path
, inode
->i_ino
, start
, 0);
640 /* a csum that isn't present is a preallocated region. */
641 if (ret
== -ENOENT
|| ret
== -EFBIG
)
644 if (printk_ratelimit())
645 printk("no csum found for inode %lu start %Lu\n", inode
->i_ino
,
649 read_extent_buffer(path
->nodes
[0], &csum
, (unsigned long)item
,
652 set_state_private(io_tree
, start
, csum
);
655 btrfs_free_path(path
);
659 struct io_failure_record
{
667 int btrfs_io_failed_hook(struct bio
*failed_bio
,
668 struct page
*page
, u64 start
, u64 end
,
669 struct extent_state
*state
)
671 struct io_failure_record
*failrec
= NULL
;
673 struct extent_map
*em
;
674 struct inode
*inode
= page
->mapping
->host
;
675 struct extent_io_tree
*failure_tree
= &BTRFS_I(inode
)->io_failure_tree
;
676 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
683 ret
= get_state_private(failure_tree
, start
, &private);
685 failrec
= kmalloc(sizeof(*failrec
), GFP_NOFS
);
688 failrec
->start
= start
;
689 failrec
->len
= end
- start
+ 1;
690 failrec
->last_mirror
= 0;
692 spin_lock(&em_tree
->lock
);
693 em
= lookup_extent_mapping(em_tree
, start
, failrec
->len
);
694 if (em
->start
> start
|| em
->start
+ em
->len
< start
) {
698 spin_unlock(&em_tree
->lock
);
700 if (!em
|| IS_ERR(em
)) {
704 logical
= start
- em
->start
;
705 logical
= em
->block_start
+ logical
;
706 failrec
->logical
= logical
;
708 set_extent_bits(failure_tree
, start
, end
, EXTENT_LOCKED
|
709 EXTENT_DIRTY
, GFP_NOFS
);
710 set_state_private(failure_tree
, start
,
711 (u64
)(unsigned long)failrec
);
713 failrec
= (struct io_failure_record
*)(unsigned long)private;
715 num_copies
= btrfs_num_copies(
716 &BTRFS_I(inode
)->root
->fs_info
->mapping_tree
,
717 failrec
->logical
, failrec
->len
);
718 failrec
->last_mirror
++;
720 spin_lock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
721 state
= find_first_extent_bit_state(&BTRFS_I(inode
)->io_tree
,
724 if (state
&& state
->start
!= failrec
->start
)
726 spin_unlock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
728 if (!state
|| failrec
->last_mirror
> num_copies
) {
729 set_state_private(failure_tree
, failrec
->start
, 0);
730 clear_extent_bits(failure_tree
, failrec
->start
,
731 failrec
->start
+ failrec
->len
- 1,
732 EXTENT_LOCKED
| EXTENT_DIRTY
, GFP_NOFS
);
736 bio
= bio_alloc(GFP_NOFS
, 1);
737 bio
->bi_private
= state
;
738 bio
->bi_end_io
= failed_bio
->bi_end_io
;
739 bio
->bi_sector
= failrec
->logical
>> 9;
740 bio
->bi_bdev
= failed_bio
->bi_bdev
;
742 bio_add_page(bio
, page
, failrec
->len
, start
- page_offset(page
));
743 if (failed_bio
->bi_rw
& (1 << BIO_RW
))
748 BTRFS_I(inode
)->io_tree
.ops
->submit_bio_hook(inode
, rw
, bio
,
749 failrec
->last_mirror
);
753 int btrfs_clean_io_failures(struct inode
*inode
, u64 start
)
757 struct io_failure_record
*failure
;
761 if (count_range_bits(&BTRFS_I(inode
)->io_failure_tree
, &private,
762 (u64
)-1, 1, EXTENT_DIRTY
)) {
763 ret
= get_state_private(&BTRFS_I(inode
)->io_failure_tree
,
764 start
, &private_failure
);
766 failure
= (struct io_failure_record
*)(unsigned long)
768 set_state_private(&BTRFS_I(inode
)->io_failure_tree
,
770 clear_extent_bits(&BTRFS_I(inode
)->io_failure_tree
,
772 failure
->start
+ failure
->len
- 1,
773 EXTENT_DIRTY
| EXTENT_LOCKED
,
781 int btrfs_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
782 struct extent_state
*state
)
784 size_t offset
= start
- ((u64
)page
->index
<< PAGE_CACHE_SHIFT
);
785 struct inode
*inode
= page
->mapping
->host
;
786 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
788 u64
private = ~(u32
)0;
790 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
794 if (btrfs_test_opt(root
, NODATASUM
) ||
795 btrfs_test_flag(inode
, NODATASUM
))
797 if (state
&& state
->start
== start
) {
798 private = state
->private;
801 ret
= get_state_private(io_tree
, start
, &private);
803 local_irq_save(flags
);
804 kaddr
= kmap_atomic(page
, KM_IRQ0
);
808 csum
= btrfs_csum_data(root
, kaddr
+ offset
, csum
, end
- start
+ 1);
809 btrfs_csum_final(csum
, (char *)&csum
);
810 if (csum
!= private) {
813 kunmap_atomic(kaddr
, KM_IRQ0
);
814 local_irq_restore(flags
);
816 /* if the io failure tree for this inode is non-empty,
817 * check to see if we've recovered from a failed IO
819 btrfs_clean_io_failures(inode
, start
);
823 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
824 page
->mapping
->host
->i_ino
, (unsigned long long)start
, csum
,
826 memset(kaddr
+ offset
, 1, end
- start
+ 1);
827 flush_dcache_page(page
);
828 kunmap_atomic(kaddr
, KM_IRQ0
);
829 local_irq_restore(flags
);
836 * This creates an orphan entry for the given inode in case something goes
837 * wrong in the middle of an unlink/truncate.
839 int btrfs_orphan_add(struct btrfs_trans_handle
*trans
, struct inode
*inode
)
841 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
844 spin_lock(&root
->list_lock
);
846 /* already on the orphan list, we're good */
847 if (!list_empty(&BTRFS_I(inode
)->i_orphan
)) {
848 spin_unlock(&root
->list_lock
);
852 list_add(&BTRFS_I(inode
)->i_orphan
, &root
->orphan_list
);
854 spin_unlock(&root
->list_lock
);
857 * insert an orphan item to track this unlinked/truncated file
859 ret
= btrfs_insert_orphan_item(trans
, root
, inode
->i_ino
);
865 * We have done the truncate/delete so we can go ahead and remove the orphan
866 * item for this particular inode.
868 int btrfs_orphan_del(struct btrfs_trans_handle
*trans
, struct inode
*inode
)
870 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
873 spin_lock(&root
->list_lock
);
875 if (list_empty(&BTRFS_I(inode
)->i_orphan
)) {
876 spin_unlock(&root
->list_lock
);
880 list_del_init(&BTRFS_I(inode
)->i_orphan
);
882 spin_unlock(&root
->list_lock
);
886 spin_unlock(&root
->list_lock
);
888 ret
= btrfs_del_orphan_item(trans
, root
, inode
->i_ino
);
894 * this cleans up any orphans that may be left on the list from the last use
897 void btrfs_orphan_cleanup(struct btrfs_root
*root
)
899 struct btrfs_path
*path
;
900 struct extent_buffer
*leaf
;
901 struct btrfs_item
*item
;
902 struct btrfs_key key
, found_key
;
903 struct btrfs_trans_handle
*trans
;
905 int ret
= 0, nr_unlink
= 0, nr_truncate
= 0;
907 /* don't do orphan cleanup if the fs is readonly. */
908 if (root
->inode
->i_sb
->s_flags
& MS_RDONLY
)
911 path
= btrfs_alloc_path();
916 key
.objectid
= BTRFS_ORPHAN_OBJECTID
;
917 btrfs_set_key_type(&key
, BTRFS_ORPHAN_ITEM_KEY
);
918 key
.offset
= (u64
)-1;
920 trans
= btrfs_start_transaction(root
, 1);
921 btrfs_set_trans_block_group(trans
, root
->inode
);
924 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
926 printk(KERN_ERR
"Error searching slot for orphan: %d"
932 * if ret == 0 means we found what we were searching for, which
933 * is weird, but possible, so only screw with path if we didnt
934 * find the key and see if we have stuff that matches
937 if (path
->slots
[0] == 0)
942 /* pull out the item */
943 leaf
= path
->nodes
[0];
944 item
= btrfs_item_nr(leaf
, path
->slots
[0]);
945 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
947 /* make sure the item matches what we want */
948 if (found_key
.objectid
!= BTRFS_ORPHAN_OBJECTID
)
950 if (btrfs_key_type(&found_key
) != BTRFS_ORPHAN_ITEM_KEY
)
953 /* release the path since we're done with it */
954 btrfs_release_path(root
, path
);
957 * this is where we are basically btrfs_lookup, without the
958 * crossing root thing. we store the inode number in the
959 * offset of the orphan item.
961 inode
= btrfs_iget_locked(root
->inode
->i_sb
,
962 found_key
.offset
, root
);
966 if (inode
->i_state
& I_NEW
) {
967 BTRFS_I(inode
)->root
= root
;
969 /* have to set the location manually */
970 BTRFS_I(inode
)->location
.objectid
= inode
->i_ino
;
971 BTRFS_I(inode
)->location
.type
= BTRFS_INODE_ITEM_KEY
;
972 BTRFS_I(inode
)->location
.offset
= 0;
974 btrfs_read_locked_inode(inode
);
975 unlock_new_inode(inode
);
979 * add this inode to the orphan list so btrfs_orphan_del does
980 * the proper thing when we hit it
982 spin_lock(&root
->list_lock
);
983 list_add(&BTRFS_I(inode
)->i_orphan
, &root
->orphan_list
);
984 spin_unlock(&root
->list_lock
);
987 * if this is a bad inode, means we actually succeeded in
988 * removing the inode, but not the orphan record, which means
989 * we need to manually delete the orphan since iput will just
992 if (is_bad_inode(inode
)) {
993 btrfs_orphan_del(trans
, inode
);
998 /* if we have links, this was a truncate, lets do that */
999 if (inode
->i_nlink
) {
1001 btrfs_truncate(inode
);
1006 /* this will do delete_inode and everything for us */
1011 printk(KERN_INFO
"btrfs: unlinked %d orphans\n", nr_unlink
);
1013 printk(KERN_INFO
"btrfs: truncated %d orphans\n", nr_truncate
);
1015 btrfs_free_path(path
);
1016 btrfs_end_transaction(trans
, root
);
1019 void btrfs_read_locked_inode(struct inode
*inode
)
1021 struct btrfs_path
*path
;
1022 struct extent_buffer
*leaf
;
1023 struct btrfs_inode_item
*inode_item
;
1024 struct btrfs_timespec
*tspec
;
1025 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1026 struct btrfs_key location
;
1027 u64 alloc_group_block
;
1031 path
= btrfs_alloc_path();
1033 memcpy(&location
, &BTRFS_I(inode
)->location
, sizeof(location
));
1035 ret
= btrfs_lookup_inode(NULL
, root
, path
, &location
, 0);
1039 leaf
= path
->nodes
[0];
1040 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
1041 struct btrfs_inode_item
);
1043 inode
->i_mode
= btrfs_inode_mode(leaf
, inode_item
);
1044 inode
->i_nlink
= btrfs_inode_nlink(leaf
, inode_item
);
1045 inode
->i_uid
= btrfs_inode_uid(leaf
, inode_item
);
1046 inode
->i_gid
= btrfs_inode_gid(leaf
, inode_item
);
1047 btrfs_i_size_write(inode
, btrfs_inode_size(leaf
, inode_item
));
1049 tspec
= btrfs_inode_atime(inode_item
);
1050 inode
->i_atime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1051 inode
->i_atime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1053 tspec
= btrfs_inode_mtime(inode_item
);
1054 inode
->i_mtime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1055 inode
->i_mtime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1057 tspec
= btrfs_inode_ctime(inode_item
);
1058 inode
->i_ctime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1059 inode
->i_ctime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1061 inode
->i_blocks
= btrfs_inode_nblocks(leaf
, inode_item
);
1062 inode
->i_generation
= btrfs_inode_generation(leaf
, inode_item
);
1064 rdev
= btrfs_inode_rdev(leaf
, inode_item
);
1066 BTRFS_I(inode
)->index_cnt
= (u64
)-1;
1068 alloc_group_block
= btrfs_inode_block_group(leaf
, inode_item
);
1069 BTRFS_I(inode
)->block_group
= btrfs_lookup_block_group(root
->fs_info
,
1071 BTRFS_I(inode
)->flags
= btrfs_inode_flags(leaf
, inode_item
);
1072 if (!BTRFS_I(inode
)->block_group
) {
1073 BTRFS_I(inode
)->block_group
= btrfs_find_block_group(root
,
1075 BTRFS_BLOCK_GROUP_METADATA
, 0);
1077 btrfs_free_path(path
);
1080 switch (inode
->i_mode
& S_IFMT
) {
1082 inode
->i_mapping
->a_ops
= &btrfs_aops
;
1083 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1084 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
1085 inode
->i_fop
= &btrfs_file_operations
;
1086 inode
->i_op
= &btrfs_file_inode_operations
;
1089 inode
->i_fop
= &btrfs_dir_file_operations
;
1090 if (root
== root
->fs_info
->tree_root
)
1091 inode
->i_op
= &btrfs_dir_ro_inode_operations
;
1093 inode
->i_op
= &btrfs_dir_inode_operations
;
1096 inode
->i_op
= &btrfs_symlink_inode_operations
;
1097 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
1098 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1101 init_special_inode(inode
, inode
->i_mode
, rdev
);
1107 btrfs_free_path(path
);
1108 make_bad_inode(inode
);
1111 static void fill_inode_item(struct extent_buffer
*leaf
,
1112 struct btrfs_inode_item
*item
,
1113 struct inode
*inode
)
1115 btrfs_set_inode_uid(leaf
, item
, inode
->i_uid
);
1116 btrfs_set_inode_gid(leaf
, item
, inode
->i_gid
);
1117 btrfs_set_inode_size(leaf
, item
, BTRFS_I(inode
)->disk_i_size
);
1118 btrfs_set_inode_mode(leaf
, item
, inode
->i_mode
);
1119 btrfs_set_inode_nlink(leaf
, item
, inode
->i_nlink
);
1121 btrfs_set_timespec_sec(leaf
, btrfs_inode_atime(item
),
1122 inode
->i_atime
.tv_sec
);
1123 btrfs_set_timespec_nsec(leaf
, btrfs_inode_atime(item
),
1124 inode
->i_atime
.tv_nsec
);
1126 btrfs_set_timespec_sec(leaf
, btrfs_inode_mtime(item
),
1127 inode
->i_mtime
.tv_sec
);
1128 btrfs_set_timespec_nsec(leaf
, btrfs_inode_mtime(item
),
1129 inode
->i_mtime
.tv_nsec
);
1131 btrfs_set_timespec_sec(leaf
, btrfs_inode_ctime(item
),
1132 inode
->i_ctime
.tv_sec
);
1133 btrfs_set_timespec_nsec(leaf
, btrfs_inode_ctime(item
),
1134 inode
->i_ctime
.tv_nsec
);
1136 btrfs_set_inode_nblocks(leaf
, item
, inode
->i_blocks
);
1137 btrfs_set_inode_generation(leaf
, item
, inode
->i_generation
);
1138 btrfs_set_inode_rdev(leaf
, item
, inode
->i_rdev
);
1139 btrfs_set_inode_flags(leaf
, item
, BTRFS_I(inode
)->flags
);
1140 btrfs_set_inode_block_group(leaf
, item
,
1141 BTRFS_I(inode
)->block_group
->key
.objectid
);
1144 int noinline
btrfs_update_inode(struct btrfs_trans_handle
*trans
,
1145 struct btrfs_root
*root
,
1146 struct inode
*inode
)
1148 struct btrfs_inode_item
*inode_item
;
1149 struct btrfs_path
*path
;
1150 struct extent_buffer
*leaf
;
1153 path
= btrfs_alloc_path();
1155 ret
= btrfs_lookup_inode(trans
, root
, path
,
1156 &BTRFS_I(inode
)->location
, 1);
1163 leaf
= path
->nodes
[0];
1164 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
1165 struct btrfs_inode_item
);
1167 fill_inode_item(leaf
, inode_item
, inode
);
1168 btrfs_mark_buffer_dirty(leaf
);
1169 btrfs_set_inode_last_trans(trans
, inode
);
1172 btrfs_free_path(path
);
1177 static int btrfs_unlink_trans(struct btrfs_trans_handle
*trans
,
1178 struct btrfs_root
*root
,
1180 struct dentry
*dentry
)
1182 struct btrfs_path
*path
;
1183 const char *name
= dentry
->d_name
.name
;
1184 int name_len
= dentry
->d_name
.len
;
1186 struct extent_buffer
*leaf
;
1187 struct btrfs_dir_item
*di
;
1188 struct btrfs_key key
;
1191 path
= btrfs_alloc_path();
1197 di
= btrfs_lookup_dir_item(trans
, root
, path
, dir
->i_ino
,
1198 name
, name_len
, -1);
1207 leaf
= path
->nodes
[0];
1208 btrfs_dir_item_key_to_cpu(leaf
, di
, &key
);
1209 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
1212 btrfs_release_path(root
, path
);
1214 ret
= btrfs_del_inode_ref(trans
, root
, name
, name_len
,
1215 dentry
->d_inode
->i_ino
,
1216 dentry
->d_parent
->d_inode
->i_ino
, &index
);
1218 printk("failed to delete reference to %.*s, "
1219 "inode %lu parent %lu\n", name_len
, name
,
1220 dentry
->d_inode
->i_ino
,
1221 dentry
->d_parent
->d_inode
->i_ino
);
1225 di
= btrfs_lookup_dir_index_item(trans
, root
, path
, dir
->i_ino
,
1226 index
, name
, name_len
, -1);
1235 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
1236 btrfs_release_path(root
, path
);
1238 dentry
->d_inode
->i_ctime
= dir
->i_ctime
;
1240 btrfs_free_path(path
);
1242 btrfs_i_size_write(dir
, dir
->i_size
- name_len
* 2);
1243 dir
->i_mtime
= dir
->i_ctime
= CURRENT_TIME
;
1244 btrfs_update_inode(trans
, root
, dir
);
1245 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1246 dentry
->d_inode
->i_nlink
--;
1248 drop_nlink(dentry
->d_inode
);
1250 ret
= btrfs_update_inode(trans
, root
, dentry
->d_inode
);
1251 dir
->i_sb
->s_dirt
= 1;
1256 static int btrfs_unlink(struct inode
*dir
, struct dentry
*dentry
)
1258 struct btrfs_root
*root
;
1259 struct btrfs_trans_handle
*trans
;
1260 struct inode
*inode
= dentry
->d_inode
;
1262 unsigned long nr
= 0;
1264 root
= BTRFS_I(dir
)->root
;
1266 ret
= btrfs_check_free_space(root
, 1, 1);
1270 trans
= btrfs_start_transaction(root
, 1);
1272 btrfs_set_trans_block_group(trans
, dir
);
1273 ret
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
1275 if (inode
->i_nlink
== 0)
1276 ret
= btrfs_orphan_add(trans
, inode
);
1278 nr
= trans
->blocks_used
;
1280 btrfs_end_transaction_throttle(trans
, root
);
1282 btrfs_btree_balance_dirty(root
, nr
);
1286 static int btrfs_rmdir(struct inode
*dir
, struct dentry
*dentry
)
1288 struct inode
*inode
= dentry
->d_inode
;
1291 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1292 struct btrfs_trans_handle
*trans
;
1293 unsigned long nr
= 0;
1295 if (inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
1299 ret
= btrfs_check_free_space(root
, 1, 1);
1303 trans
= btrfs_start_transaction(root
, 1);
1304 btrfs_set_trans_block_group(trans
, dir
);
1306 err
= btrfs_orphan_add(trans
, inode
);
1310 /* now the directory is empty */
1311 err
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
1313 btrfs_i_size_write(inode
, 0);
1317 nr
= trans
->blocks_used
;
1318 ret
= btrfs_end_transaction_throttle(trans
, root
);
1320 btrfs_btree_balance_dirty(root
, nr
);
1328 * this can truncate away extent items, csum items and directory items.
1329 * It starts at a high offset and removes keys until it can't find
1330 * any higher than i_size.
1332 * csum items that cross the new i_size are truncated to the new size
1335 * min_type is the minimum key type to truncate down to. If set to 0, this
1336 * will kill all the items on this inode, including the INODE_ITEM_KEY.
1338 static int btrfs_truncate_in_trans(struct btrfs_trans_handle
*trans
,
1339 struct btrfs_root
*root
,
1340 struct inode
*inode
,
1344 struct btrfs_path
*path
;
1345 struct btrfs_key key
;
1346 struct btrfs_key found_key
;
1348 struct extent_buffer
*leaf
;
1349 struct btrfs_file_extent_item
*fi
;
1350 u64 extent_start
= 0;
1351 u64 extent_num_bytes
= 0;
1357 int pending_del_nr
= 0;
1358 int pending_del_slot
= 0;
1359 int extent_type
= -1;
1360 u64 mask
= root
->sectorsize
- 1;
1362 btrfs_drop_extent_cache(inode
, inode
->i_size
& (~mask
), (u64
)-1);
1363 path
= btrfs_alloc_path();
1367 /* FIXME, add redo link to tree so we don't leak on crash */
1368 key
.objectid
= inode
->i_ino
;
1369 key
.offset
= (u64
)-1;
1372 btrfs_init_path(path
);
1374 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
1379 BUG_ON(path
->slots
[0] == 0);
1385 leaf
= path
->nodes
[0];
1386 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
1387 found_type
= btrfs_key_type(&found_key
);
1389 if (found_key
.objectid
!= inode
->i_ino
)
1392 if (found_type
< min_type
)
1395 item_end
= found_key
.offset
;
1396 if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
1397 fi
= btrfs_item_ptr(leaf
, path
->slots
[0],
1398 struct btrfs_file_extent_item
);
1399 extent_type
= btrfs_file_extent_type(leaf
, fi
);
1400 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1402 btrfs_file_extent_num_bytes(leaf
, fi
);
1403 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1404 struct btrfs_item
*item
= btrfs_item_nr(leaf
,
1406 item_end
+= btrfs_file_extent_inline_len(leaf
,
1411 if (found_type
== BTRFS_CSUM_ITEM_KEY
) {
1412 ret
= btrfs_csum_truncate(trans
, root
, path
,
1416 if (item_end
< inode
->i_size
) {
1417 if (found_type
== BTRFS_DIR_ITEM_KEY
) {
1418 found_type
= BTRFS_INODE_ITEM_KEY
;
1419 } else if (found_type
== BTRFS_EXTENT_ITEM_KEY
) {
1420 found_type
= BTRFS_CSUM_ITEM_KEY
;
1421 } else if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
1422 found_type
= BTRFS_XATTR_ITEM_KEY
;
1423 } else if (found_type
== BTRFS_XATTR_ITEM_KEY
) {
1424 found_type
= BTRFS_INODE_REF_KEY
;
1425 } else if (found_type
) {
1430 btrfs_set_key_type(&key
, found_type
);
1433 if (found_key
.offset
>= inode
->i_size
)
1439 /* FIXME, shrink the extent if the ref count is only 1 */
1440 if (found_type
!= BTRFS_EXTENT_DATA_KEY
)
1443 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1445 extent_start
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
1447 u64 orig_num_bytes
=
1448 btrfs_file_extent_num_bytes(leaf
, fi
);
1449 extent_num_bytes
= inode
->i_size
-
1450 found_key
.offset
+ root
->sectorsize
- 1;
1451 extent_num_bytes
= extent_num_bytes
&
1452 ~((u64
)root
->sectorsize
- 1);
1453 btrfs_set_file_extent_num_bytes(leaf
, fi
,
1455 num_dec
= (orig_num_bytes
-
1457 if (extent_start
!= 0)
1458 dec_i_blocks(inode
, num_dec
);
1459 btrfs_mark_buffer_dirty(leaf
);
1462 btrfs_file_extent_disk_num_bytes(leaf
,
1464 /* FIXME blocksize != 4096 */
1465 num_dec
= btrfs_file_extent_num_bytes(leaf
, fi
);
1466 if (extent_start
!= 0) {
1468 dec_i_blocks(inode
, num_dec
);
1470 root_gen
= btrfs_header_generation(leaf
);
1471 root_owner
= btrfs_header_owner(leaf
);
1473 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1475 u32 newsize
= inode
->i_size
- found_key
.offset
;
1476 dec_i_blocks(inode
, item_end
+ 1 -
1477 found_key
.offset
- newsize
);
1479 btrfs_file_extent_calc_inline_size(newsize
);
1480 ret
= btrfs_truncate_item(trans
, root
, path
,
1484 dec_i_blocks(inode
, item_end
+ 1 -
1490 if (!pending_del_nr
) {
1491 /* no pending yet, add ourselves */
1492 pending_del_slot
= path
->slots
[0];
1494 } else if (pending_del_nr
&&
1495 path
->slots
[0] + 1 == pending_del_slot
) {
1496 /* hop on the pending chunk */
1498 pending_del_slot
= path
->slots
[0];
1500 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path
->slots
[0], pending_del_nr
, pending_del_slot
);
1506 ret
= btrfs_free_extent(trans
, root
, extent_start
,
1509 root_gen
, inode
->i_ino
,
1510 found_key
.offset
, 0);
1514 if (path
->slots
[0] == 0) {
1517 btrfs_release_path(root
, path
);
1522 if (pending_del_nr
&&
1523 path
->slots
[0] + 1 != pending_del_slot
) {
1524 struct btrfs_key debug
;
1526 btrfs_item_key_to_cpu(path
->nodes
[0], &debug
,
1528 ret
= btrfs_del_items(trans
, root
, path
,
1533 btrfs_release_path(root
, path
);
1539 if (pending_del_nr
) {
1540 ret
= btrfs_del_items(trans
, root
, path
, pending_del_slot
,
1543 btrfs_free_path(path
);
1544 inode
->i_sb
->s_dirt
= 1;
1549 * taken from block_truncate_page, but does cow as it zeros out
1550 * any bytes left in the last page in the file.
1552 static int btrfs_truncate_page(struct address_space
*mapping
, loff_t from
)
1554 struct inode
*inode
= mapping
->host
;
1555 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1556 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1557 struct btrfs_ordered_extent
*ordered
;
1559 u32 blocksize
= root
->sectorsize
;
1560 pgoff_t index
= from
>> PAGE_CACHE_SHIFT
;
1561 unsigned offset
= from
& (PAGE_CACHE_SIZE
-1);
1567 if ((offset
& (blocksize
- 1)) == 0)
1572 page
= grab_cache_page(mapping
, index
);
1576 page_start
= page_offset(page
);
1577 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1579 if (!PageUptodate(page
)) {
1580 ret
= btrfs_readpage(NULL
, page
);
1582 if (page
->mapping
!= mapping
) {
1584 page_cache_release(page
);
1587 if (!PageUptodate(page
)) {
1592 wait_on_page_writeback(page
);
1594 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1595 set_page_extent_mapped(page
);
1597 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
1599 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1601 page_cache_release(page
);
1602 btrfs_start_ordered_extent(inode
, ordered
, 1);
1603 btrfs_put_ordered_extent(ordered
);
1607 btrfs_set_extent_delalloc(inode
, page_start
, page_end
);
1609 if (offset
!= PAGE_CACHE_SIZE
) {
1611 memset(kaddr
+ offset
, 0, PAGE_CACHE_SIZE
- offset
);
1612 flush_dcache_page(page
);
1615 ClearPageChecked(page
);
1616 set_page_dirty(page
);
1617 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1621 page_cache_release(page
);
1626 static int btrfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
1628 struct inode
*inode
= dentry
->d_inode
;
1631 err
= inode_change_ok(inode
, attr
);
1635 if (S_ISREG(inode
->i_mode
) &&
1636 attr
->ia_valid
& ATTR_SIZE
&& attr
->ia_size
> inode
->i_size
) {
1637 struct btrfs_trans_handle
*trans
;
1638 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1639 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1641 u64 mask
= root
->sectorsize
- 1;
1642 u64 hole_start
= (inode
->i_size
+ mask
) & ~mask
;
1643 u64 block_end
= (attr
->ia_size
+ mask
) & ~mask
;
1647 if (attr
->ia_size
<= hole_start
)
1650 err
= btrfs_check_free_space(root
, 1, 0);
1654 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
1656 hole_size
= block_end
- hole_start
;
1658 struct btrfs_ordered_extent
*ordered
;
1659 btrfs_wait_ordered_range(inode
, hole_start
, hole_size
);
1661 lock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1662 ordered
= btrfs_lookup_ordered_extent(inode
, hole_start
);
1664 unlock_extent(io_tree
, hole_start
,
1665 block_end
- 1, GFP_NOFS
);
1666 btrfs_put_ordered_extent(ordered
);
1672 trans
= btrfs_start_transaction(root
, 1);
1673 btrfs_set_trans_block_group(trans
, inode
);
1674 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
1675 err
= btrfs_drop_extents(trans
, root
, inode
,
1676 hole_start
, block_end
, hole_start
,
1679 if (alloc_hint
!= EXTENT_MAP_INLINE
) {
1680 err
= btrfs_insert_file_extent(trans
, root
,
1684 btrfs_drop_extent_cache(inode
, hole_start
,
1686 btrfs_check_file(root
, inode
);
1688 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
1689 btrfs_end_transaction(trans
, root
);
1690 unlock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1695 err
= inode_setattr(inode
, attr
);
1697 if (!err
&& ((attr
->ia_valid
& ATTR_MODE
)))
1698 err
= btrfs_acl_chmod(inode
);
1703 void btrfs_delete_inode(struct inode
*inode
)
1705 struct btrfs_trans_handle
*trans
;
1706 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1710 truncate_inode_pages(&inode
->i_data
, 0);
1711 if (is_bad_inode(inode
)) {
1712 btrfs_orphan_del(NULL
, inode
);
1715 btrfs_wait_ordered_range(inode
, 0, (u64
)-1);
1717 btrfs_i_size_write(inode
, 0);
1718 trans
= btrfs_start_transaction(root
, 1);
1720 btrfs_set_trans_block_group(trans
, inode
);
1721 ret
= btrfs_truncate_in_trans(trans
, root
, inode
, 0);
1723 btrfs_orphan_del(NULL
, inode
);
1724 goto no_delete_lock
;
1727 btrfs_orphan_del(trans
, inode
);
1729 nr
= trans
->blocks_used
;
1732 btrfs_end_transaction(trans
, root
);
1733 btrfs_btree_balance_dirty(root
, nr
);
1737 nr
= trans
->blocks_used
;
1738 btrfs_end_transaction(trans
, root
);
1739 btrfs_btree_balance_dirty(root
, nr
);
1745 * this returns the key found in the dir entry in the location pointer.
1746 * If no dir entries were found, location->objectid is 0.
1748 static int btrfs_inode_by_name(struct inode
*dir
, struct dentry
*dentry
,
1749 struct btrfs_key
*location
)
1751 const char *name
= dentry
->d_name
.name
;
1752 int namelen
= dentry
->d_name
.len
;
1753 struct btrfs_dir_item
*di
;
1754 struct btrfs_path
*path
;
1755 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1758 if (namelen
== 1 && strcmp(name
, ".") == 0) {
1759 location
->objectid
= dir
->i_ino
;
1760 location
->type
= BTRFS_INODE_ITEM_KEY
;
1761 location
->offset
= 0;
1764 path
= btrfs_alloc_path();
1767 if (namelen
== 2 && strcmp(name
, "..") == 0) {
1768 struct btrfs_key key
;
1769 struct extent_buffer
*leaf
;
1772 key
.objectid
= dir
->i_ino
;
1773 key
.offset
= (u64
)-1;
1774 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1775 if (ret
< 0 || path
->slots
[0] == 0)
1777 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1780 leaf
= path
->nodes
[0];
1781 slot
= path
->slots
[0] - 1;
1783 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
1784 if (key
.objectid
!= dir
->i_ino
||
1785 key
.type
!= BTRFS_INODE_REF_KEY
) {
1788 location
->objectid
= key
.offset
;
1789 location
->type
= BTRFS_INODE_ITEM_KEY
;
1790 location
->offset
= 0;
1794 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir
->i_ino
, name
,
1798 if (!di
|| IS_ERR(di
)) {
1801 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, location
);
1803 btrfs_free_path(path
);
1806 location
->objectid
= 0;
1811 * when we hit a tree root in a directory, the btrfs part of the inode
1812 * needs to be changed to reflect the root directory of the tree root. This
1813 * is kind of like crossing a mount point.
1815 static int fixup_tree_root_location(struct btrfs_root
*root
,
1816 struct btrfs_key
*location
,
1817 struct btrfs_root
**sub_root
,
1818 struct dentry
*dentry
)
1820 struct btrfs_root_item
*ri
;
1822 if (btrfs_key_type(location
) != BTRFS_ROOT_ITEM_KEY
)
1824 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1827 *sub_root
= btrfs_read_fs_root(root
->fs_info
, location
,
1828 dentry
->d_name
.name
,
1829 dentry
->d_name
.len
);
1830 if (IS_ERR(*sub_root
))
1831 return PTR_ERR(*sub_root
);
1833 ri
= &(*sub_root
)->root_item
;
1834 location
->objectid
= btrfs_root_dirid(ri
);
1835 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1836 location
->offset
= 0;
1841 static int btrfs_init_locked_inode(struct inode
*inode
, void *p
)
1843 struct btrfs_iget_args
*args
= p
;
1844 inode
->i_ino
= args
->ino
;
1845 BTRFS_I(inode
)->root
= args
->root
;
1846 BTRFS_I(inode
)->delalloc_bytes
= 0;
1847 inode
->i_mapping
->writeback_index
= 0;
1848 BTRFS_I(inode
)->disk_i_size
= 0;
1849 BTRFS_I(inode
)->index_cnt
= (u64
)-1;
1850 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1851 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1852 inode
->i_mapping
, GFP_NOFS
);
1853 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1854 inode
->i_mapping
, GFP_NOFS
);
1855 INIT_LIST_HEAD(&BTRFS_I(inode
)->delalloc_inodes
);
1856 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
1857 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
1858 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
1862 static int btrfs_find_actor(struct inode
*inode
, void *opaque
)
1864 struct btrfs_iget_args
*args
= opaque
;
1865 return (args
->ino
== inode
->i_ino
&&
1866 args
->root
== BTRFS_I(inode
)->root
);
1869 struct inode
*btrfs_ilookup(struct super_block
*s
, u64 objectid
,
1872 struct btrfs_iget_args args
;
1873 args
.ino
= objectid
;
1874 args
.root
= btrfs_lookup_fs_root(btrfs_sb(s
)->fs_info
, root_objectid
);
1879 return ilookup5(s
, objectid
, btrfs_find_actor
, (void *)&args
);
1882 struct inode
*btrfs_iget_locked(struct super_block
*s
, u64 objectid
,
1883 struct btrfs_root
*root
)
1885 struct inode
*inode
;
1886 struct btrfs_iget_args args
;
1887 args
.ino
= objectid
;
1890 inode
= iget5_locked(s
, objectid
, btrfs_find_actor
,
1891 btrfs_init_locked_inode
,
1896 static struct dentry
*btrfs_lookup(struct inode
*dir
, struct dentry
*dentry
,
1897 struct nameidata
*nd
)
1899 struct inode
* inode
;
1900 struct btrfs_inode
*bi
= BTRFS_I(dir
);
1901 struct btrfs_root
*root
= bi
->root
;
1902 struct btrfs_root
*sub_root
= root
;
1903 struct btrfs_key location
;
1904 int ret
, do_orphan
= 0;
1906 if (dentry
->d_name
.len
> BTRFS_NAME_LEN
)
1907 return ERR_PTR(-ENAMETOOLONG
);
1909 ret
= btrfs_inode_by_name(dir
, dentry
, &location
);
1912 return ERR_PTR(ret
);
1915 if (location
.objectid
) {
1916 ret
= fixup_tree_root_location(root
, &location
, &sub_root
,
1919 return ERR_PTR(ret
);
1921 return ERR_PTR(-ENOENT
);
1923 inode
= btrfs_iget_locked(dir
->i_sb
, location
.objectid
,
1926 return ERR_PTR(-EACCES
);
1927 if (inode
->i_state
& I_NEW
) {
1928 /* the inode and parent dir are two different roots */
1929 if (sub_root
!= root
) {
1931 sub_root
->inode
= inode
;
1934 BTRFS_I(inode
)->root
= sub_root
;
1935 memcpy(&BTRFS_I(inode
)->location
, &location
,
1937 btrfs_read_locked_inode(inode
);
1938 unlock_new_inode(inode
);
1942 if (unlikely(do_orphan
))
1943 btrfs_orphan_cleanup(sub_root
);
1945 return d_splice_alias(inode
, dentry
);
1948 static unsigned char btrfs_filetype_table
[] = {
1949 DT_UNKNOWN
, DT_REG
, DT_DIR
, DT_CHR
, DT_BLK
, DT_FIFO
, DT_SOCK
, DT_LNK
1952 static int btrfs_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
1954 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1955 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1956 struct btrfs_item
*item
;
1957 struct btrfs_dir_item
*di
;
1958 struct btrfs_key key
;
1959 struct btrfs_key found_key
;
1960 struct btrfs_path
*path
;
1963 struct extent_buffer
*leaf
;
1966 unsigned char d_type
;
1971 int key_type
= BTRFS_DIR_INDEX_KEY
;
1976 /* FIXME, use a real flag for deciding about the key type */
1977 if (root
->fs_info
->tree_root
== root
)
1978 key_type
= BTRFS_DIR_ITEM_KEY
;
1980 /* special case for "." */
1981 if (filp
->f_pos
== 0) {
1982 over
= filldir(dirent
, ".", 1,
1990 key
.objectid
= inode
->i_ino
;
1991 path
= btrfs_alloc_path();
1994 /* special case for .., just use the back ref */
1995 if (filp
->f_pos
== 1) {
1996 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1997 key
.offset
= (u64
)-1;
1998 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1999 if (ret
< 0 || path
->slots
[0] == 0) {
2000 btrfs_release_path(root
, path
);
2001 goto read_dir_items
;
2004 leaf
= path
->nodes
[0];
2005 slot
= path
->slots
[0] - 1;
2006 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
2007 btrfs_release_path(root
, path
);
2008 if (found_key
.objectid
!= key
.objectid
||
2009 found_key
.type
!= BTRFS_INODE_REF_KEY
)
2010 goto read_dir_items
;
2011 over
= filldir(dirent
, "..", 2,
2012 2, found_key
.offset
, DT_DIR
);
2019 btrfs_set_key_type(&key
, key_type
);
2020 key
.offset
= filp
->f_pos
;
2022 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2027 leaf
= path
->nodes
[0];
2028 nritems
= btrfs_header_nritems(leaf
);
2029 slot
= path
->slots
[0];
2030 if (advance
|| slot
>= nritems
) {
2031 if (slot
>= nritems
-1) {
2032 ret
= btrfs_next_leaf(root
, path
);
2035 leaf
= path
->nodes
[0];
2036 nritems
= btrfs_header_nritems(leaf
);
2037 slot
= path
->slots
[0];
2044 item
= btrfs_item_nr(leaf
, slot
);
2045 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
2047 if (found_key
.objectid
!= key
.objectid
)
2049 if (btrfs_key_type(&found_key
) != key_type
)
2051 if (found_key
.offset
< filp
->f_pos
)
2054 filp
->f_pos
= found_key
.offset
;
2056 di
= btrfs_item_ptr(leaf
, slot
, struct btrfs_dir_item
);
2058 di_total
= btrfs_item_size(leaf
, item
);
2059 while(di_cur
< di_total
) {
2060 struct btrfs_key location
;
2062 name_len
= btrfs_dir_name_len(leaf
, di
);
2063 if (name_len
< 32) {
2064 name_ptr
= tmp_name
;
2066 name_ptr
= kmalloc(name_len
, GFP_NOFS
);
2069 read_extent_buffer(leaf
, name_ptr
,
2070 (unsigned long)(di
+ 1), name_len
);
2072 d_type
= btrfs_filetype_table
[btrfs_dir_type(leaf
, di
)];
2073 btrfs_dir_item_key_to_cpu(leaf
, di
, &location
);
2074 over
= filldir(dirent
, name_ptr
, name_len
,
2079 if (name_ptr
!= tmp_name
)
2084 di_len
= btrfs_dir_name_len(leaf
, di
) +
2085 btrfs_dir_data_len(leaf
, di
) +sizeof(*di
);
2087 di
= (struct btrfs_dir_item
*)((char *)di
+ di_len
);
2090 if (key_type
== BTRFS_DIR_INDEX_KEY
)
2091 filp
->f_pos
= INT_LIMIT(typeof(filp
->f_pos
));
2097 btrfs_free_path(path
);
2101 int btrfs_write_inode(struct inode
*inode
, int wait
)
2103 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2104 struct btrfs_trans_handle
*trans
;
2107 if (root
->fs_info
->closing
> 1)
2111 trans
= btrfs_join_transaction(root
, 1);
2112 btrfs_set_trans_block_group(trans
, inode
);
2113 ret
= btrfs_commit_transaction(trans
, root
);
2119 * This is somewhat expensive, updating the tree every time the
2120 * inode changes. But, it is most likely to find the inode in cache.
2121 * FIXME, needs more benchmarking...there are no reasons other than performance
2122 * to keep or drop this code.
2124 void btrfs_dirty_inode(struct inode
*inode
)
2126 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2127 struct btrfs_trans_handle
*trans
;
2129 trans
= btrfs_join_transaction(root
, 1);
2130 btrfs_set_trans_block_group(trans
, inode
);
2131 btrfs_update_inode(trans
, root
, inode
);
2132 btrfs_end_transaction(trans
, root
);
2135 static int btrfs_set_inode_index_count(struct inode
*inode
)
2137 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2138 struct btrfs_key key
, found_key
;
2139 struct btrfs_path
*path
;
2140 struct extent_buffer
*leaf
;
2143 key
.objectid
= inode
->i_ino
;
2144 btrfs_set_key_type(&key
, BTRFS_DIR_INDEX_KEY
);
2145 key
.offset
= (u64
)-1;
2147 path
= btrfs_alloc_path();
2151 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2154 /* FIXME: we should be able to handle this */
2160 * MAGIC NUMBER EXPLANATION:
2161 * since we search a directory based on f_pos we have to start at 2
2162 * since '.' and '..' have f_pos of 0 and 1 respectively, so everybody
2163 * else has to start at 2
2165 if (path
->slots
[0] == 0) {
2166 BTRFS_I(inode
)->index_cnt
= 2;
2172 leaf
= path
->nodes
[0];
2173 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2175 if (found_key
.objectid
!= inode
->i_ino
||
2176 btrfs_key_type(&found_key
) != BTRFS_DIR_INDEX_KEY
) {
2177 BTRFS_I(inode
)->index_cnt
= 2;
2181 BTRFS_I(inode
)->index_cnt
= found_key
.offset
+ 1;
2183 btrfs_free_path(path
);
2187 static int btrfs_set_inode_index(struct inode
*dir
, struct inode
*inode
,
2192 if (BTRFS_I(dir
)->index_cnt
== (u64
)-1) {
2193 ret
= btrfs_set_inode_index_count(dir
);
2198 *index
= BTRFS_I(dir
)->index_cnt
;
2199 BTRFS_I(dir
)->index_cnt
++;
2204 static struct inode
*btrfs_new_inode(struct btrfs_trans_handle
*trans
,
2205 struct btrfs_root
*root
,
2207 const char *name
, int name_len
,
2210 struct btrfs_block_group_cache
*group
,
2211 int mode
, u64
*index
)
2213 struct inode
*inode
;
2214 struct btrfs_inode_item
*inode_item
;
2215 struct btrfs_block_group_cache
*new_inode_group
;
2216 struct btrfs_key
*location
;
2217 struct btrfs_path
*path
;
2218 struct btrfs_inode_ref
*ref
;
2219 struct btrfs_key key
[2];
2225 path
= btrfs_alloc_path();
2228 inode
= new_inode(root
->fs_info
->sb
);
2230 return ERR_PTR(-ENOMEM
);
2233 ret
= btrfs_set_inode_index(dir
, inode
, index
);
2235 return ERR_PTR(ret
);
2238 * index_cnt is ignored for everything but a dir,
2239 * btrfs_get_inode_index_count has an explanation for the magic
2242 BTRFS_I(inode
)->index_cnt
= 2;
2244 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
2245 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
2246 inode
->i_mapping
, GFP_NOFS
);
2247 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
2248 inode
->i_mapping
, GFP_NOFS
);
2249 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
2250 INIT_LIST_HEAD(&BTRFS_I(inode
)->delalloc_inodes
);
2251 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
2252 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
2253 BTRFS_I(inode
)->delalloc_bytes
= 0;
2254 inode
->i_mapping
->writeback_index
= 0;
2255 BTRFS_I(inode
)->disk_i_size
= 0;
2256 BTRFS_I(inode
)->root
= root
;
2262 new_inode_group
= btrfs_find_block_group(root
, group
, 0,
2263 BTRFS_BLOCK_GROUP_METADATA
, owner
);
2264 if (!new_inode_group
) {
2265 printk("find_block group failed\n");
2266 new_inode_group
= group
;
2268 BTRFS_I(inode
)->block_group
= new_inode_group
;
2269 BTRFS_I(inode
)->flags
= 0;
2271 key
[0].objectid
= objectid
;
2272 btrfs_set_key_type(&key
[0], BTRFS_INODE_ITEM_KEY
);
2275 key
[1].objectid
= objectid
;
2276 btrfs_set_key_type(&key
[1], BTRFS_INODE_REF_KEY
);
2277 key
[1].offset
= ref_objectid
;
2279 sizes
[0] = sizeof(struct btrfs_inode_item
);
2280 sizes
[1] = name_len
+ sizeof(*ref
);
2282 ret
= btrfs_insert_empty_items(trans
, root
, path
, key
, sizes
, 2);
2286 if (objectid
> root
->highest_inode
)
2287 root
->highest_inode
= objectid
;
2289 inode
->i_uid
= current
->fsuid
;
2290 inode
->i_gid
= current
->fsgid
;
2291 inode
->i_mode
= mode
;
2292 inode
->i_ino
= objectid
;
2293 inode
->i_blocks
= 0;
2294 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2295 inode_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
2296 struct btrfs_inode_item
);
2297 fill_inode_item(path
->nodes
[0], inode_item
, inode
);
2299 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0] + 1,
2300 struct btrfs_inode_ref
);
2301 btrfs_set_inode_ref_name_len(path
->nodes
[0], ref
, name_len
);
2302 btrfs_set_inode_ref_index(path
->nodes
[0], ref
, *index
);
2303 ptr
= (unsigned long)(ref
+ 1);
2304 write_extent_buffer(path
->nodes
[0], name
, ptr
, name_len
);
2306 btrfs_mark_buffer_dirty(path
->nodes
[0]);
2307 btrfs_free_path(path
);
2309 location
= &BTRFS_I(inode
)->location
;
2310 location
->objectid
= objectid
;
2311 location
->offset
= 0;
2312 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
2314 insert_inode_hash(inode
);
2318 BTRFS_I(dir
)->index_cnt
--;
2319 btrfs_free_path(path
);
2320 return ERR_PTR(ret
);
2323 static inline u8
btrfs_inode_type(struct inode
*inode
)
2325 return btrfs_type_by_mode
[(inode
->i_mode
& S_IFMT
) >> S_SHIFT
];
2328 static int btrfs_add_link(struct btrfs_trans_handle
*trans
,
2329 struct dentry
*dentry
, struct inode
*inode
,
2330 int add_backref
, u64 index
)
2333 struct btrfs_key key
;
2334 struct btrfs_root
*root
= BTRFS_I(dentry
->d_parent
->d_inode
)->root
;
2335 struct inode
*parent_inode
= dentry
->d_parent
->d_inode
;
2337 key
.objectid
= inode
->i_ino
;
2338 btrfs_set_key_type(&key
, BTRFS_INODE_ITEM_KEY
);
2341 ret
= btrfs_insert_dir_item(trans
, root
,
2342 dentry
->d_name
.name
, dentry
->d_name
.len
,
2343 dentry
->d_parent
->d_inode
->i_ino
,
2344 &key
, btrfs_inode_type(inode
),
2348 ret
= btrfs_insert_inode_ref(trans
, root
,
2349 dentry
->d_name
.name
,
2352 parent_inode
->i_ino
,
2355 btrfs_i_size_write(parent_inode
, parent_inode
->i_size
+
2356 dentry
->d_name
.len
* 2);
2357 parent_inode
->i_mtime
= parent_inode
->i_ctime
= CURRENT_TIME
;
2358 ret
= btrfs_update_inode(trans
, root
,
2359 dentry
->d_parent
->d_inode
);
2364 static int btrfs_add_nondir(struct btrfs_trans_handle
*trans
,
2365 struct dentry
*dentry
, struct inode
*inode
,
2366 int backref
, u64 index
)
2368 int err
= btrfs_add_link(trans
, dentry
, inode
, backref
, index
);
2370 d_instantiate(dentry
, inode
);
2378 static int btrfs_mknod(struct inode
*dir
, struct dentry
*dentry
,
2379 int mode
, dev_t rdev
)
2381 struct btrfs_trans_handle
*trans
;
2382 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2383 struct inode
*inode
= NULL
;
2387 unsigned long nr
= 0;
2390 if (!new_valid_dev(rdev
))
2393 err
= btrfs_check_free_space(root
, 1, 0);
2397 trans
= btrfs_start_transaction(root
, 1);
2398 btrfs_set_trans_block_group(trans
, dir
);
2400 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2406 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2408 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2409 BTRFS_I(dir
)->block_group
, mode
, &index
);
2410 err
= PTR_ERR(inode
);
2414 err
= btrfs_init_acl(inode
, dir
);
2420 btrfs_set_trans_block_group(trans
, inode
);
2421 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
2425 inode
->i_op
= &btrfs_special_inode_operations
;
2426 init_special_inode(inode
, inode
->i_mode
, rdev
);
2427 btrfs_update_inode(trans
, root
, inode
);
2429 dir
->i_sb
->s_dirt
= 1;
2430 btrfs_update_inode_block_group(trans
, inode
);
2431 btrfs_update_inode_block_group(trans
, dir
);
2433 nr
= trans
->blocks_used
;
2434 btrfs_end_transaction_throttle(trans
, root
);
2437 inode_dec_link_count(inode
);
2440 btrfs_btree_balance_dirty(root
, nr
);
2444 static int btrfs_create(struct inode
*dir
, struct dentry
*dentry
,
2445 int mode
, struct nameidata
*nd
)
2447 struct btrfs_trans_handle
*trans
;
2448 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2449 struct inode
*inode
= NULL
;
2452 unsigned long nr
= 0;
2456 err
= btrfs_check_free_space(root
, 1, 0);
2459 trans
= btrfs_start_transaction(root
, 1);
2460 btrfs_set_trans_block_group(trans
, dir
);
2462 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2468 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2470 dentry
->d_parent
->d_inode
->i_ino
,
2471 objectid
, BTRFS_I(dir
)->block_group
, mode
,
2473 err
= PTR_ERR(inode
);
2477 err
= btrfs_init_acl(inode
, dir
);
2483 btrfs_set_trans_block_group(trans
, inode
);
2484 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
2488 inode
->i_mapping
->a_ops
= &btrfs_aops
;
2489 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
2490 inode
->i_fop
= &btrfs_file_operations
;
2491 inode
->i_op
= &btrfs_file_inode_operations
;
2492 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
2493 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
2494 inode
->i_mapping
, GFP_NOFS
);
2495 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
2496 inode
->i_mapping
, GFP_NOFS
);
2497 INIT_LIST_HEAD(&BTRFS_I(inode
)->delalloc_inodes
);
2498 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
2499 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
2500 BTRFS_I(inode
)->delalloc_bytes
= 0;
2501 BTRFS_I(inode
)->disk_i_size
= 0;
2502 inode
->i_mapping
->writeback_index
= 0;
2503 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
2504 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
2506 dir
->i_sb
->s_dirt
= 1;
2507 btrfs_update_inode_block_group(trans
, inode
);
2508 btrfs_update_inode_block_group(trans
, dir
);
2510 nr
= trans
->blocks_used
;
2511 btrfs_end_transaction_throttle(trans
, root
);
2514 inode_dec_link_count(inode
);
2517 btrfs_btree_balance_dirty(root
, nr
);
2521 static int btrfs_link(struct dentry
*old_dentry
, struct inode
*dir
,
2522 struct dentry
*dentry
)
2524 struct btrfs_trans_handle
*trans
;
2525 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2526 struct inode
*inode
= old_dentry
->d_inode
;
2528 unsigned long nr
= 0;
2532 if (inode
->i_nlink
== 0)
2535 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
2540 err
= btrfs_check_free_space(root
, 1, 0);
2543 err
= btrfs_set_inode_index(dir
, inode
, &index
);
2547 trans
= btrfs_start_transaction(root
, 1);
2549 btrfs_set_trans_block_group(trans
, dir
);
2550 atomic_inc(&inode
->i_count
);
2552 err
= btrfs_add_nondir(trans
, dentry
, inode
, 1, index
);
2557 dir
->i_sb
->s_dirt
= 1;
2558 btrfs_update_inode_block_group(trans
, dir
);
2559 err
= btrfs_update_inode(trans
, root
, inode
);
2564 nr
= trans
->blocks_used
;
2565 btrfs_end_transaction_throttle(trans
, root
);
2568 inode_dec_link_count(inode
);
2571 btrfs_btree_balance_dirty(root
, nr
);
2575 static int btrfs_mkdir(struct inode
*dir
, struct dentry
*dentry
, int mode
)
2577 struct inode
*inode
= NULL
;
2578 struct btrfs_trans_handle
*trans
;
2579 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2581 int drop_on_err
= 0;
2584 unsigned long nr
= 1;
2586 err
= btrfs_check_free_space(root
, 1, 0);
2590 trans
= btrfs_start_transaction(root
, 1);
2591 btrfs_set_trans_block_group(trans
, dir
);
2593 if (IS_ERR(trans
)) {
2594 err
= PTR_ERR(trans
);
2598 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2604 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2606 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2607 BTRFS_I(dir
)->block_group
, S_IFDIR
| mode
,
2609 if (IS_ERR(inode
)) {
2610 err
= PTR_ERR(inode
);
2616 err
= btrfs_init_acl(inode
, dir
);
2620 inode
->i_op
= &btrfs_dir_inode_operations
;
2621 inode
->i_fop
= &btrfs_dir_file_operations
;
2622 btrfs_set_trans_block_group(trans
, inode
);
2624 btrfs_i_size_write(inode
, 0);
2625 err
= btrfs_update_inode(trans
, root
, inode
);
2629 err
= btrfs_add_link(trans
, dentry
, inode
, 0, index
);
2633 d_instantiate(dentry
, inode
);
2635 dir
->i_sb
->s_dirt
= 1;
2636 btrfs_update_inode_block_group(trans
, inode
);
2637 btrfs_update_inode_block_group(trans
, dir
);
2640 nr
= trans
->blocks_used
;
2641 btrfs_end_transaction_throttle(trans
, root
);
2646 btrfs_btree_balance_dirty(root
, nr
);
2650 static int merge_extent_mapping(struct extent_map_tree
*em_tree
,
2651 struct extent_map
*existing
,
2652 struct extent_map
*em
,
2653 u64 map_start
, u64 map_len
)
2657 BUG_ON(map_start
< em
->start
|| map_start
>= extent_map_end(em
));
2658 start_diff
= map_start
- em
->start
;
2659 em
->start
= map_start
;
2661 if (em
->block_start
< EXTENT_MAP_LAST_BYTE
)
2662 em
->block_start
+= start_diff
;
2663 return add_extent_mapping(em_tree
, em
);
2666 struct extent_map
*btrfs_get_extent(struct inode
*inode
, struct page
*page
,
2667 size_t pg_offset
, u64 start
, u64 len
,
2673 u64 extent_start
= 0;
2675 u64 objectid
= inode
->i_ino
;
2677 struct btrfs_path
*path
= NULL
;
2678 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2679 struct btrfs_file_extent_item
*item
;
2680 struct extent_buffer
*leaf
;
2681 struct btrfs_key found_key
;
2682 struct extent_map
*em
= NULL
;
2683 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
2684 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2685 struct btrfs_trans_handle
*trans
= NULL
;
2688 spin_lock(&em_tree
->lock
);
2689 em
= lookup_extent_mapping(em_tree
, start
, len
);
2691 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2692 spin_unlock(&em_tree
->lock
);
2695 if (em
->start
> start
|| em
->start
+ em
->len
<= start
)
2696 free_extent_map(em
);
2697 else if (em
->block_start
== EXTENT_MAP_INLINE
&& page
)
2698 free_extent_map(em
);
2702 em
= alloc_extent_map(GFP_NOFS
);
2707 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2708 em
->start
= EXTENT_MAP_HOLE
;
2712 path
= btrfs_alloc_path();
2716 ret
= btrfs_lookup_file_extent(trans
, root
, path
,
2717 objectid
, start
, trans
!= NULL
);
2724 if (path
->slots
[0] == 0)
2729 leaf
= path
->nodes
[0];
2730 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
2731 struct btrfs_file_extent_item
);
2732 /* are we inside the extent that was found? */
2733 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2734 found_type
= btrfs_key_type(&found_key
);
2735 if (found_key
.objectid
!= objectid
||
2736 found_type
!= BTRFS_EXTENT_DATA_KEY
) {
2740 found_type
= btrfs_file_extent_type(leaf
, item
);
2741 extent_start
= found_key
.offset
;
2742 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
2743 extent_end
= extent_start
+
2744 btrfs_file_extent_num_bytes(leaf
, item
);
2746 if (start
< extent_start
|| start
>= extent_end
) {
2748 if (start
< extent_start
) {
2749 if (start
+ len
<= extent_start
)
2751 em
->len
= extent_end
- extent_start
;
2757 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
2759 em
->start
= extent_start
;
2760 em
->len
= extent_end
- extent_start
;
2761 em
->block_start
= EXTENT_MAP_HOLE
;
2764 bytenr
+= btrfs_file_extent_offset(leaf
, item
);
2765 em
->block_start
= bytenr
;
2766 em
->start
= extent_start
;
2767 em
->len
= extent_end
- extent_start
;
2769 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
2774 size_t extent_offset
;
2777 size
= btrfs_file_extent_inline_len(leaf
, btrfs_item_nr(leaf
,
2779 extent_end
= (extent_start
+ size
+ root
->sectorsize
- 1) &
2780 ~((u64
)root
->sectorsize
- 1);
2781 if (start
< extent_start
|| start
>= extent_end
) {
2783 if (start
< extent_start
) {
2784 if (start
+ len
<= extent_start
)
2786 em
->len
= extent_end
- extent_start
;
2792 em
->block_start
= EXTENT_MAP_INLINE
;
2795 em
->start
= extent_start
;
2800 page_start
= page_offset(page
) + pg_offset
;
2801 extent_offset
= page_start
- extent_start
;
2802 copy_size
= min_t(u64
, PAGE_CACHE_SIZE
- pg_offset
,
2803 size
- extent_offset
);
2804 em
->start
= extent_start
+ extent_offset
;
2805 em
->len
= (copy_size
+ root
->sectorsize
- 1) &
2806 ~((u64
)root
->sectorsize
- 1);
2808 ptr
= btrfs_file_extent_inline_start(item
) + extent_offset
;
2809 if (create
== 0 && !PageUptodate(page
)) {
2810 read_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2812 flush_dcache_page(page
);
2813 } else if (create
&& PageUptodate(page
)) {
2816 free_extent_map(em
);
2818 btrfs_release_path(root
, path
);
2819 trans
= btrfs_join_transaction(root
, 1);
2822 write_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2824 btrfs_mark_buffer_dirty(leaf
);
2827 set_extent_uptodate(io_tree
, em
->start
,
2828 extent_map_end(em
) - 1, GFP_NOFS
);
2831 printk("unkknown found_type %d\n", found_type
);
2838 em
->block_start
= EXTENT_MAP_HOLE
;
2840 btrfs_release_path(root
, path
);
2841 if (em
->start
> start
|| extent_map_end(em
) <= start
) {
2842 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em
->start
, em
->len
, start
, len
);
2848 spin_lock(&em_tree
->lock
);
2849 ret
= add_extent_mapping(em_tree
, em
);
2850 /* it is possible that someone inserted the extent into the tree
2851 * while we had the lock dropped. It is also possible that
2852 * an overlapping map exists in the tree
2854 if (ret
== -EEXIST
) {
2855 struct extent_map
*existing
;
2859 existing
= lookup_extent_mapping(em_tree
, start
, len
);
2860 if (existing
&& (existing
->start
> start
||
2861 existing
->start
+ existing
->len
<= start
)) {
2862 free_extent_map(existing
);
2866 existing
= lookup_extent_mapping(em_tree
, em
->start
,
2869 err
= merge_extent_mapping(em_tree
, existing
,
2872 free_extent_map(existing
);
2874 free_extent_map(em
);
2879 printk("failing to insert %Lu %Lu\n",
2881 free_extent_map(em
);
2885 free_extent_map(em
);
2890 spin_unlock(&em_tree
->lock
);
2893 btrfs_free_path(path
);
2895 ret
= btrfs_end_transaction(trans
, root
);
2901 free_extent_map(em
);
2903 return ERR_PTR(err
);
2908 #if 0 /* waiting for O_DIRECT reads */
2909 static int btrfs_get_block(struct inode
*inode
, sector_t iblock
,
2910 struct buffer_head
*bh_result
, int create
)
2912 struct extent_map
*em
;
2913 u64 start
= (u64
)iblock
<< inode
->i_blkbits
;
2914 struct btrfs_multi_bio
*multi
= NULL
;
2915 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2921 em
= btrfs_get_extent(inode
, NULL
, 0, start
, bh_result
->b_size
, 0);
2923 if (!em
|| IS_ERR(em
))
2926 if (em
->start
> start
|| em
->start
+ em
->len
<= start
) {
2930 if (em
->block_start
== EXTENT_MAP_INLINE
) {
2935 len
= em
->start
+ em
->len
- start
;
2936 len
= min_t(u64
, len
, INT_LIMIT(typeof(bh_result
->b_size
)));
2938 if (em
->block_start
== EXTENT_MAP_HOLE
||
2939 em
->block_start
== EXTENT_MAP_DELALLOC
) {
2940 bh_result
->b_size
= len
;
2944 logical
= start
- em
->start
;
2945 logical
= em
->block_start
+ logical
;
2948 ret
= btrfs_map_block(&root
->fs_info
->mapping_tree
, READ
,
2949 logical
, &map_length
, &multi
, 0);
2951 bh_result
->b_blocknr
= multi
->stripes
[0].physical
>> inode
->i_blkbits
;
2952 bh_result
->b_size
= min(map_length
, len
);
2954 bh_result
->b_bdev
= multi
->stripes
[0].dev
->bdev
;
2955 set_buffer_mapped(bh_result
);
2958 free_extent_map(em
);
2963 static ssize_t
btrfs_direct_IO(int rw
, struct kiocb
*iocb
,
2964 const struct iovec
*iov
, loff_t offset
,
2965 unsigned long nr_segs
)
2969 struct file
*file
= iocb
->ki_filp
;
2970 struct inode
*inode
= file
->f_mapping
->host
;
2975 return blockdev_direct_IO(rw
, iocb
, inode
, inode
->i_sb
->s_bdev
, iov
,
2976 offset
, nr_segs
, btrfs_get_block
, NULL
);
2980 static sector_t
btrfs_bmap(struct address_space
*mapping
, sector_t iblock
)
2982 return extent_bmap(mapping
, iblock
, btrfs_get_extent
);
2985 int btrfs_readpage(struct file
*file
, struct page
*page
)
2987 struct extent_io_tree
*tree
;
2988 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2989 return extent_read_full_page(tree
, page
, btrfs_get_extent
);
2992 static int btrfs_writepage(struct page
*page
, struct writeback_control
*wbc
)
2994 struct extent_io_tree
*tree
;
2997 if (current
->flags
& PF_MEMALLOC
) {
2998 redirty_page_for_writepage(wbc
, page
);
3002 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3003 return extent_write_full_page(tree
, page
, btrfs_get_extent
, wbc
);
3006 int btrfs_writepages(struct address_space
*mapping
,
3007 struct writeback_control
*wbc
)
3009 struct extent_io_tree
*tree
;
3010 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
3011 return extent_writepages(tree
, mapping
, btrfs_get_extent
, wbc
);
3015 btrfs_readpages(struct file
*file
, struct address_space
*mapping
,
3016 struct list_head
*pages
, unsigned nr_pages
)
3018 struct extent_io_tree
*tree
;
3019 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
3020 return extent_readpages(tree
, mapping
, pages
, nr_pages
,
3023 static int __btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
3025 struct extent_io_tree
*tree
;
3026 struct extent_map_tree
*map
;
3029 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3030 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
3031 ret
= try_release_extent_mapping(map
, tree
, page
, gfp_flags
);
3033 ClearPagePrivate(page
);
3034 set_page_private(page
, 0);
3035 page_cache_release(page
);
3040 static int btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
3042 return __btrfs_releasepage(page
, gfp_flags
);
3045 static void btrfs_invalidatepage(struct page
*page
, unsigned long offset
)
3047 struct extent_io_tree
*tree
;
3048 struct btrfs_ordered_extent
*ordered
;
3049 u64 page_start
= page_offset(page
);
3050 u64 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
3052 wait_on_page_writeback(page
);
3053 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3055 btrfs_releasepage(page
, GFP_NOFS
);
3059 lock_extent(tree
, page_start
, page_end
, GFP_NOFS
);
3060 ordered
= btrfs_lookup_ordered_extent(page
->mapping
->host
,
3064 * IO on this page will never be started, so we need
3065 * to account for any ordered extents now
3067 clear_extent_bit(tree
, page_start
, page_end
,
3068 EXTENT_DIRTY
| EXTENT_DELALLOC
|
3069 EXTENT_LOCKED
, 1, 0, GFP_NOFS
);
3070 btrfs_finish_ordered_io(page
->mapping
->host
,
3071 page_start
, page_end
);
3072 btrfs_put_ordered_extent(ordered
);
3073 lock_extent(tree
, page_start
, page_end
, GFP_NOFS
);
3075 clear_extent_bit(tree
, page_start
, page_end
,
3076 EXTENT_LOCKED
| EXTENT_DIRTY
| EXTENT_DELALLOC
|
3079 __btrfs_releasepage(page
, GFP_NOFS
);
3081 ClearPageChecked(page
);
3082 if (PagePrivate(page
)) {
3083 ClearPagePrivate(page
);
3084 set_page_private(page
, 0);
3085 page_cache_release(page
);
3090 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
3091 * called from a page fault handler when a page is first dirtied. Hence we must
3092 * be careful to check for EOF conditions here. We set the page up correctly
3093 * for a written page which means we get ENOSPC checking when writing into
3094 * holes and correct delalloc and unwritten extent mapping on filesystems that
3095 * support these features.
3097 * We are not allowed to take the i_mutex here so we have to play games to
3098 * protect against truncate races as the page could now be beyond EOF. Because
3099 * vmtruncate() writes the inode size before removing pages, once we have the
3100 * page lock we can determine safely if the page is beyond EOF. If it is not
3101 * beyond EOF, then the page is guaranteed safe against truncation until we
3104 int btrfs_page_mkwrite(struct vm_area_struct
*vma
, struct page
*page
)
3106 struct inode
*inode
= fdentry(vma
->vm_file
)->d_inode
;
3107 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3108 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
3109 struct btrfs_ordered_extent
*ordered
;
3111 unsigned long zero_start
;
3117 ret
= btrfs_check_free_space(root
, PAGE_CACHE_SIZE
, 0);
3124 size
= i_size_read(inode
);
3125 page_start
= page_offset(page
);
3126 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
3128 if ((page
->mapping
!= inode
->i_mapping
) ||
3129 (page_start
>= size
)) {
3130 /* page got truncated out from underneath us */
3133 wait_on_page_writeback(page
);
3135 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3136 set_page_extent_mapped(page
);
3139 * we can't set the delalloc bits if there are pending ordered
3140 * extents. Drop our locks and wait for them to finish
3142 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
3144 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3146 btrfs_start_ordered_extent(inode
, ordered
, 1);
3147 btrfs_put_ordered_extent(ordered
);
3151 btrfs_set_extent_delalloc(inode
, page_start
, page_end
);
3154 /* page is wholly or partially inside EOF */
3155 if (page_start
+ PAGE_CACHE_SIZE
> size
)
3156 zero_start
= size
& ~PAGE_CACHE_MASK
;
3158 zero_start
= PAGE_CACHE_SIZE
;
3160 if (zero_start
!= PAGE_CACHE_SIZE
) {
3162 memset(kaddr
+ zero_start
, 0, PAGE_CACHE_SIZE
- zero_start
);
3163 flush_dcache_page(page
);
3166 ClearPageChecked(page
);
3167 set_page_dirty(page
);
3168 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3176 static void btrfs_truncate(struct inode
*inode
)
3178 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3180 struct btrfs_trans_handle
*trans
;
3182 u64 mask
= root
->sectorsize
- 1;
3184 if (!S_ISREG(inode
->i_mode
))
3186 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
3189 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
3190 btrfs_wait_ordered_range(inode
, inode
->i_size
& (~mask
), (u64
)-1);
3192 trans
= btrfs_start_transaction(root
, 1);
3193 btrfs_set_trans_block_group(trans
, inode
);
3194 btrfs_i_size_write(inode
, inode
->i_size
);
3196 ret
= btrfs_orphan_add(trans
, inode
);
3199 /* FIXME, add redo link to tree so we don't leak on crash */
3200 ret
= btrfs_truncate_in_trans(trans
, root
, inode
,
3201 BTRFS_EXTENT_DATA_KEY
);
3202 btrfs_update_inode(trans
, root
, inode
);
3204 ret
= btrfs_orphan_del(trans
, inode
);
3208 nr
= trans
->blocks_used
;
3209 ret
= btrfs_end_transaction_throttle(trans
, root
);
3211 btrfs_btree_balance_dirty(root
, nr
);
3215 * Invalidate a single dcache entry at the root of the filesystem.
3216 * Needed after creation of snapshot or subvolume.
3218 void btrfs_invalidate_dcache_root(struct btrfs_root
*root
, char *name
,
3221 struct dentry
*alias
, *entry
;
3224 alias
= d_find_alias(root
->fs_info
->sb
->s_root
->d_inode
);
3228 /* change me if btrfs ever gets a d_hash operation */
3229 qstr
.hash
= full_name_hash(qstr
.name
, qstr
.len
);
3230 entry
= d_lookup(alias
, &qstr
);
3233 d_invalidate(entry
);
3239 int btrfs_create_subvol_root(struct btrfs_root
*new_root
,
3240 struct btrfs_trans_handle
*trans
, u64 new_dirid
,
3241 struct btrfs_block_group_cache
*block_group
)
3243 struct inode
*inode
;
3246 inode
= btrfs_new_inode(trans
, new_root
, NULL
, "..", 2, new_dirid
,
3247 new_dirid
, block_group
, S_IFDIR
| 0700, &index
);
3249 return PTR_ERR(inode
);
3250 inode
->i_op
= &btrfs_dir_inode_operations
;
3251 inode
->i_fop
= &btrfs_dir_file_operations
;
3252 new_root
->inode
= inode
;
3255 btrfs_i_size_write(inode
, 0);
3257 return btrfs_update_inode(trans
, new_root
, inode
);
3260 unsigned long btrfs_force_ra(struct address_space
*mapping
,
3261 struct file_ra_state
*ra
, struct file
*file
,
3262 pgoff_t offset
, pgoff_t last_index
)
3264 pgoff_t req_size
= last_index
- offset
+ 1;
3266 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3267 offset
= page_cache_readahead(mapping
, ra
, file
, offset
, req_size
);
3270 page_cache_sync_readahead(mapping
, ra
, file
, offset
, req_size
);
3271 return offset
+ req_size
;
3275 struct inode
*btrfs_alloc_inode(struct super_block
*sb
)
3277 struct btrfs_inode
*ei
;
3279 ei
= kmem_cache_alloc(btrfs_inode_cachep
, GFP_NOFS
);
3283 btrfs_ordered_inode_tree_init(&ei
->ordered_tree
);
3284 ei
->i_acl
= BTRFS_ACL_NOT_CACHED
;
3285 ei
->i_default_acl
= BTRFS_ACL_NOT_CACHED
;
3286 INIT_LIST_HEAD(&ei
->i_orphan
);
3287 return &ei
->vfs_inode
;
3290 void btrfs_destroy_inode(struct inode
*inode
)
3292 struct btrfs_ordered_extent
*ordered
;
3293 WARN_ON(!list_empty(&inode
->i_dentry
));
3294 WARN_ON(inode
->i_data
.nrpages
);
3296 if (BTRFS_I(inode
)->i_acl
&&
3297 BTRFS_I(inode
)->i_acl
!= BTRFS_ACL_NOT_CACHED
)
3298 posix_acl_release(BTRFS_I(inode
)->i_acl
);
3299 if (BTRFS_I(inode
)->i_default_acl
&&
3300 BTRFS_I(inode
)->i_default_acl
!= BTRFS_ACL_NOT_CACHED
)
3301 posix_acl_release(BTRFS_I(inode
)->i_default_acl
);
3303 spin_lock(&BTRFS_I(inode
)->root
->list_lock
);
3304 if (!list_empty(&BTRFS_I(inode
)->i_orphan
)) {
3305 printk(KERN_ERR
"BTRFS: inode %lu: inode still on the orphan"
3306 " list\n", inode
->i_ino
);
3309 spin_unlock(&BTRFS_I(inode
)->root
->list_lock
);
3312 ordered
= btrfs_lookup_first_ordered_extent(inode
, (u64
)-1);
3316 printk("found ordered extent %Lu %Lu\n",
3317 ordered
->file_offset
, ordered
->len
);
3318 btrfs_remove_ordered_extent(inode
, ordered
);
3319 btrfs_put_ordered_extent(ordered
);
3320 btrfs_put_ordered_extent(ordered
);
3323 btrfs_drop_extent_cache(inode
, 0, (u64
)-1);
3324 kmem_cache_free(btrfs_inode_cachep
, BTRFS_I(inode
));
3327 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3328 static void init_once(void *foo
)
3329 #elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3330 static void init_once(struct kmem_cache
* cachep
, void *foo
)
3332 static void init_once(void * foo
, struct kmem_cache
* cachep
,
3333 unsigned long flags
)
3336 struct btrfs_inode
*ei
= (struct btrfs_inode
*) foo
;
3338 inode_init_once(&ei
->vfs_inode
);
3341 void btrfs_destroy_cachep(void)
3343 if (btrfs_inode_cachep
)
3344 kmem_cache_destroy(btrfs_inode_cachep
);
3345 if (btrfs_trans_handle_cachep
)
3346 kmem_cache_destroy(btrfs_trans_handle_cachep
);
3347 if (btrfs_transaction_cachep
)
3348 kmem_cache_destroy(btrfs_transaction_cachep
);
3349 if (btrfs_bit_radix_cachep
)
3350 kmem_cache_destroy(btrfs_bit_radix_cachep
);
3351 if (btrfs_path_cachep
)
3352 kmem_cache_destroy(btrfs_path_cachep
);
3355 struct kmem_cache
*btrfs_cache_create(const char *name
, size_t size
,
3356 unsigned long extra_flags
,
3357 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3358 void (*ctor
)(void *)
3359 #elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3360 void (*ctor
)(struct kmem_cache
*, void *)
3362 void (*ctor
)(void *, struct kmem_cache
*,
3367 return kmem_cache_create(name
, size
, 0, (SLAB_RECLAIM_ACCOUNT
|
3368 SLAB_MEM_SPREAD
| extra_flags
), ctor
3369 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3375 int btrfs_init_cachep(void)
3377 btrfs_inode_cachep
= btrfs_cache_create("btrfs_inode_cache",
3378 sizeof(struct btrfs_inode
),
3380 if (!btrfs_inode_cachep
)
3382 btrfs_trans_handle_cachep
=
3383 btrfs_cache_create("btrfs_trans_handle_cache",
3384 sizeof(struct btrfs_trans_handle
),
3386 if (!btrfs_trans_handle_cachep
)
3388 btrfs_transaction_cachep
= btrfs_cache_create("btrfs_transaction_cache",
3389 sizeof(struct btrfs_transaction
),
3391 if (!btrfs_transaction_cachep
)
3393 btrfs_path_cachep
= btrfs_cache_create("btrfs_path_cache",
3394 sizeof(struct btrfs_path
),
3396 if (!btrfs_path_cachep
)
3398 btrfs_bit_radix_cachep
= btrfs_cache_create("btrfs_radix", 256,
3399 SLAB_DESTROY_BY_RCU
, NULL
);
3400 if (!btrfs_bit_radix_cachep
)
3404 btrfs_destroy_cachep();
3408 static int btrfs_getattr(struct vfsmount
*mnt
,
3409 struct dentry
*dentry
, struct kstat
*stat
)
3411 struct inode
*inode
= dentry
->d_inode
;
3412 generic_fillattr(inode
, stat
);
3413 stat
->blksize
= PAGE_CACHE_SIZE
;
3414 stat
->blocks
= inode
->i_blocks
+ (BTRFS_I(inode
)->delalloc_bytes
>> 9);
3418 static int btrfs_rename(struct inode
* old_dir
, struct dentry
*old_dentry
,
3419 struct inode
* new_dir
,struct dentry
*new_dentry
)
3421 struct btrfs_trans_handle
*trans
;
3422 struct btrfs_root
*root
= BTRFS_I(old_dir
)->root
;
3423 struct inode
*new_inode
= new_dentry
->d_inode
;
3424 struct inode
*old_inode
= old_dentry
->d_inode
;
3425 struct timespec ctime
= CURRENT_TIME
;
3429 if (S_ISDIR(old_inode
->i_mode
) && new_inode
&&
3430 new_inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
3434 ret
= btrfs_check_free_space(root
, 1, 0);
3438 trans
= btrfs_start_transaction(root
, 1);
3440 btrfs_set_trans_block_group(trans
, new_dir
);
3442 old_dentry
->d_inode
->i_nlink
++;
3443 old_dir
->i_ctime
= old_dir
->i_mtime
= ctime
;
3444 new_dir
->i_ctime
= new_dir
->i_mtime
= ctime
;
3445 old_inode
->i_ctime
= ctime
;
3447 ret
= btrfs_unlink_trans(trans
, root
, old_dir
, old_dentry
);
3452 new_inode
->i_ctime
= CURRENT_TIME
;
3453 ret
= btrfs_unlink_trans(trans
, root
, new_dir
, new_dentry
);
3456 if (new_inode
->i_nlink
== 0) {
3457 ret
= btrfs_orphan_add(trans
, new_inode
);
3462 ret
= btrfs_set_inode_index(new_dir
, old_inode
, &index
);
3466 ret
= btrfs_add_link(trans
, new_dentry
, old_inode
, 1, index
);
3471 btrfs_end_transaction_throttle(trans
, root
);
3476 int btrfs_start_delalloc_inodes(struct btrfs_root
*root
)
3478 struct list_head
*head
= &root
->fs_info
->delalloc_inodes
;
3479 struct btrfs_inode
*binode
;
3480 unsigned long flags
;
3482 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
3483 while(!list_empty(head
)) {
3484 binode
= list_entry(head
->next
, struct btrfs_inode
,
3486 atomic_inc(&binode
->vfs_inode
.i_count
);
3487 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
3488 filemap_write_and_wait(binode
->vfs_inode
.i_mapping
);
3489 iput(&binode
->vfs_inode
);
3490 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
3492 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
3496 static int btrfs_symlink(struct inode
*dir
, struct dentry
*dentry
,
3497 const char *symname
)
3499 struct btrfs_trans_handle
*trans
;
3500 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
3501 struct btrfs_path
*path
;
3502 struct btrfs_key key
;
3503 struct inode
*inode
= NULL
;
3511 struct btrfs_file_extent_item
*ei
;
3512 struct extent_buffer
*leaf
;
3513 unsigned long nr
= 0;
3515 name_len
= strlen(symname
) + 1;
3516 if (name_len
> BTRFS_MAX_INLINE_DATA_SIZE(root
))
3517 return -ENAMETOOLONG
;
3519 err
= btrfs_check_free_space(root
, 1, 0);
3523 trans
= btrfs_start_transaction(root
, 1);
3524 btrfs_set_trans_block_group(trans
, dir
);
3526 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
3532 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
3534 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
3535 BTRFS_I(dir
)->block_group
, S_IFLNK
|S_IRWXUGO
,
3537 err
= PTR_ERR(inode
);
3541 err
= btrfs_init_acl(inode
, dir
);
3547 btrfs_set_trans_block_group(trans
, inode
);
3548 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
3552 inode
->i_mapping
->a_ops
= &btrfs_aops
;
3553 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3554 inode
->i_fop
= &btrfs_file_operations
;
3555 inode
->i_op
= &btrfs_file_inode_operations
;
3556 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
3557 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
3558 inode
->i_mapping
, GFP_NOFS
);
3559 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
3560 inode
->i_mapping
, GFP_NOFS
);
3561 INIT_LIST_HEAD(&BTRFS_I(inode
)->delalloc_inodes
);
3562 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
3563 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
3564 BTRFS_I(inode
)->delalloc_bytes
= 0;
3565 BTRFS_I(inode
)->disk_i_size
= 0;
3566 inode
->i_mapping
->writeback_index
= 0;
3567 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
3568 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
3570 dir
->i_sb
->s_dirt
= 1;
3571 btrfs_update_inode_block_group(trans
, inode
);
3572 btrfs_update_inode_block_group(trans
, dir
);
3576 path
= btrfs_alloc_path();
3578 key
.objectid
= inode
->i_ino
;
3580 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
3581 datasize
= btrfs_file_extent_calc_inline_size(name_len
);
3582 err
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
3588 leaf
= path
->nodes
[0];
3589 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
3590 struct btrfs_file_extent_item
);
3591 btrfs_set_file_extent_generation(leaf
, ei
, trans
->transid
);
3592 btrfs_set_file_extent_type(leaf
, ei
,
3593 BTRFS_FILE_EXTENT_INLINE
);
3594 ptr
= btrfs_file_extent_inline_start(ei
);
3595 write_extent_buffer(leaf
, symname
, ptr
, name_len
);
3596 btrfs_mark_buffer_dirty(leaf
);
3597 btrfs_free_path(path
);
3599 inode
->i_op
= &btrfs_symlink_inode_operations
;
3600 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
3601 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3602 btrfs_i_size_write(inode
, name_len
- 1);
3603 err
= btrfs_update_inode(trans
, root
, inode
);
3608 nr
= trans
->blocks_used
;
3609 btrfs_end_transaction_throttle(trans
, root
);
3612 inode_dec_link_count(inode
);
3615 btrfs_btree_balance_dirty(root
, nr
);
3619 static int btrfs_set_page_dirty(struct page
*page
)
3621 return __set_page_dirty_nobuffers(page
);
3624 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3625 static int btrfs_permission(struct inode
*inode
, int mask
)
3627 static int btrfs_permission(struct inode
*inode
, int mask
,
3628 struct nameidata
*nd
)
3631 if (btrfs_test_flag(inode
, READONLY
) && (mask
& MAY_WRITE
))
3633 return generic_permission(inode
, mask
, btrfs_check_acl
);
3636 static struct inode_operations btrfs_dir_inode_operations
= {
3637 .lookup
= btrfs_lookup
,
3638 .create
= btrfs_create
,
3639 .unlink
= btrfs_unlink
,
3641 .mkdir
= btrfs_mkdir
,
3642 .rmdir
= btrfs_rmdir
,
3643 .rename
= btrfs_rename
,
3644 .symlink
= btrfs_symlink
,
3645 .setattr
= btrfs_setattr
,
3646 .mknod
= btrfs_mknod
,
3647 .setxattr
= generic_setxattr
,
3648 .getxattr
= generic_getxattr
,
3649 .listxattr
= btrfs_listxattr
,
3650 .removexattr
= generic_removexattr
,
3651 .permission
= btrfs_permission
,
3653 static struct inode_operations btrfs_dir_ro_inode_operations
= {
3654 .lookup
= btrfs_lookup
,
3655 .permission
= btrfs_permission
,
3657 static struct file_operations btrfs_dir_file_operations
= {
3658 .llseek
= generic_file_llseek
,
3659 .read
= generic_read_dir
,
3660 .readdir
= btrfs_readdir
,
3661 .unlocked_ioctl
= btrfs_ioctl
,
3662 #ifdef CONFIG_COMPAT
3663 .compat_ioctl
= btrfs_ioctl
,
3665 .release
= btrfs_release_file
,
3668 static struct extent_io_ops btrfs_extent_io_ops
= {
3669 .fill_delalloc
= run_delalloc_range
,
3670 .submit_bio_hook
= btrfs_submit_bio_hook
,
3671 .merge_bio_hook
= btrfs_merge_bio_hook
,
3672 .readpage_io_hook
= btrfs_readpage_io_hook
,
3673 .readpage_end_io_hook
= btrfs_readpage_end_io_hook
,
3674 .writepage_end_io_hook
= btrfs_writepage_end_io_hook
,
3675 .writepage_start_hook
= btrfs_writepage_start_hook
,
3676 .readpage_io_failed_hook
= btrfs_io_failed_hook
,
3677 .set_bit_hook
= btrfs_set_bit_hook
,
3678 .clear_bit_hook
= btrfs_clear_bit_hook
,
3681 static struct address_space_operations btrfs_aops
= {
3682 .readpage
= btrfs_readpage
,
3683 .writepage
= btrfs_writepage
,
3684 .writepages
= btrfs_writepages
,
3685 .readpages
= btrfs_readpages
,
3686 .sync_page
= block_sync_page
,
3688 .direct_IO
= btrfs_direct_IO
,
3689 .invalidatepage
= btrfs_invalidatepage
,
3690 .releasepage
= btrfs_releasepage
,
3691 .set_page_dirty
= btrfs_set_page_dirty
,
3694 static struct address_space_operations btrfs_symlink_aops
= {
3695 .readpage
= btrfs_readpage
,
3696 .writepage
= btrfs_writepage
,
3697 .invalidatepage
= btrfs_invalidatepage
,
3698 .releasepage
= btrfs_releasepage
,
3701 static struct inode_operations btrfs_file_inode_operations
= {
3702 .truncate
= btrfs_truncate
,
3703 .getattr
= btrfs_getattr
,
3704 .setattr
= btrfs_setattr
,
3705 .setxattr
= generic_setxattr
,
3706 .getxattr
= generic_getxattr
,
3707 .listxattr
= btrfs_listxattr
,
3708 .removexattr
= generic_removexattr
,
3709 .permission
= btrfs_permission
,
3711 static struct inode_operations btrfs_special_inode_operations
= {
3712 .getattr
= btrfs_getattr
,
3713 .setattr
= btrfs_setattr
,
3714 .permission
= btrfs_permission
,
3715 .setxattr
= generic_setxattr
,
3716 .getxattr
= generic_getxattr
,
3717 .listxattr
= btrfs_listxattr
,
3718 .removexattr
= generic_removexattr
,
3720 static struct inode_operations btrfs_symlink_inode_operations
= {
3721 .readlink
= generic_readlink
,
3722 .follow_link
= page_follow_link_light
,
3723 .put_link
= page_put_link
,
3724 .permission
= btrfs_permission
,