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 printk("no csum found for inode %lu start %Lu\n", inode
->i_ino
,
648 read_extent_buffer(path
->nodes
[0], &csum
, (unsigned long)item
,
651 set_state_private(io_tree
, start
, csum
);
654 btrfs_free_path(path
);
658 struct io_failure_record
{
666 int btrfs_io_failed_hook(struct bio
*failed_bio
,
667 struct page
*page
, u64 start
, u64 end
,
668 struct extent_state
*state
)
670 struct io_failure_record
*failrec
= NULL
;
672 struct extent_map
*em
;
673 struct inode
*inode
= page
->mapping
->host
;
674 struct extent_io_tree
*failure_tree
= &BTRFS_I(inode
)->io_failure_tree
;
675 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
682 ret
= get_state_private(failure_tree
, start
, &private);
684 failrec
= kmalloc(sizeof(*failrec
), GFP_NOFS
);
687 failrec
->start
= start
;
688 failrec
->len
= end
- start
+ 1;
689 failrec
->last_mirror
= 0;
691 spin_lock(&em_tree
->lock
);
692 em
= lookup_extent_mapping(em_tree
, start
, failrec
->len
);
693 if (em
->start
> start
|| em
->start
+ em
->len
< start
) {
697 spin_unlock(&em_tree
->lock
);
699 if (!em
|| IS_ERR(em
)) {
703 logical
= start
- em
->start
;
704 logical
= em
->block_start
+ logical
;
705 failrec
->logical
= logical
;
707 set_extent_bits(failure_tree
, start
, end
, EXTENT_LOCKED
|
708 EXTENT_DIRTY
, GFP_NOFS
);
709 set_state_private(failure_tree
, start
,
710 (u64
)(unsigned long)failrec
);
712 failrec
= (struct io_failure_record
*)(unsigned long)private;
714 num_copies
= btrfs_num_copies(
715 &BTRFS_I(inode
)->root
->fs_info
->mapping_tree
,
716 failrec
->logical
, failrec
->len
);
717 failrec
->last_mirror
++;
719 spin_lock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
720 state
= find_first_extent_bit_state(&BTRFS_I(inode
)->io_tree
,
723 if (state
&& state
->start
!= failrec
->start
)
725 spin_unlock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
727 if (!state
|| failrec
->last_mirror
> num_copies
) {
728 set_state_private(failure_tree
, failrec
->start
, 0);
729 clear_extent_bits(failure_tree
, failrec
->start
,
730 failrec
->start
+ failrec
->len
- 1,
731 EXTENT_LOCKED
| EXTENT_DIRTY
, GFP_NOFS
);
735 bio
= bio_alloc(GFP_NOFS
, 1);
736 bio
->bi_private
= state
;
737 bio
->bi_end_io
= failed_bio
->bi_end_io
;
738 bio
->bi_sector
= failrec
->logical
>> 9;
739 bio
->bi_bdev
= failed_bio
->bi_bdev
;
741 bio_add_page(bio
, page
, failrec
->len
, start
- page_offset(page
));
742 if (failed_bio
->bi_rw
& (1 << BIO_RW
))
747 BTRFS_I(inode
)->io_tree
.ops
->submit_bio_hook(inode
, rw
, bio
,
748 failrec
->last_mirror
);
752 int btrfs_clean_io_failures(struct inode
*inode
, u64 start
)
756 struct io_failure_record
*failure
;
760 if (count_range_bits(&BTRFS_I(inode
)->io_failure_tree
, &private,
761 (u64
)-1, 1, EXTENT_DIRTY
)) {
762 ret
= get_state_private(&BTRFS_I(inode
)->io_failure_tree
,
763 start
, &private_failure
);
765 failure
= (struct io_failure_record
*)(unsigned long)
767 set_state_private(&BTRFS_I(inode
)->io_failure_tree
,
769 clear_extent_bits(&BTRFS_I(inode
)->io_failure_tree
,
771 failure
->start
+ failure
->len
- 1,
772 EXTENT_DIRTY
| EXTENT_LOCKED
,
780 int btrfs_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
781 struct extent_state
*state
)
783 size_t offset
= start
- ((u64
)page
->index
<< PAGE_CACHE_SHIFT
);
784 struct inode
*inode
= page
->mapping
->host
;
785 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
787 u64
private = ~(u32
)0;
789 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
793 if (btrfs_test_opt(root
, NODATASUM
) ||
794 btrfs_test_flag(inode
, NODATASUM
))
796 if (state
&& state
->start
== start
) {
797 private = state
->private;
800 ret
= get_state_private(io_tree
, start
, &private);
802 local_irq_save(flags
);
803 kaddr
= kmap_atomic(page
, KM_IRQ0
);
807 csum
= btrfs_csum_data(root
, kaddr
+ offset
, csum
, end
- start
+ 1);
808 btrfs_csum_final(csum
, (char *)&csum
);
809 if (csum
!= private) {
812 kunmap_atomic(kaddr
, KM_IRQ0
);
813 local_irq_restore(flags
);
815 /* if the io failure tree for this inode is non-empty,
816 * check to see if we've recovered from a failed IO
818 btrfs_clean_io_failures(inode
, start
);
822 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
823 page
->mapping
->host
->i_ino
, (unsigned long long)start
, csum
,
825 memset(kaddr
+ offset
, 1, end
- start
+ 1);
826 flush_dcache_page(page
);
827 kunmap_atomic(kaddr
, KM_IRQ0
);
828 local_irq_restore(flags
);
835 * This creates an orphan entry for the given inode in case something goes
836 * wrong in the middle of an unlink/truncate.
838 int btrfs_orphan_add(struct btrfs_trans_handle
*trans
, struct inode
*inode
)
840 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
843 spin_lock(&root
->list_lock
);
845 /* already on the orphan list, we're good */
846 if (!list_empty(&BTRFS_I(inode
)->i_orphan
)) {
847 spin_unlock(&root
->list_lock
);
851 list_add(&BTRFS_I(inode
)->i_orphan
, &root
->orphan_list
);
853 spin_unlock(&root
->list_lock
);
856 * insert an orphan item to track this unlinked/truncated file
858 ret
= btrfs_insert_orphan_item(trans
, root
, inode
->i_ino
);
864 * We have done the truncate/delete so we can go ahead and remove the orphan
865 * item for this particular inode.
867 int btrfs_orphan_del(struct btrfs_trans_handle
*trans
, struct inode
*inode
)
869 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
872 spin_lock(&root
->list_lock
);
874 if (list_empty(&BTRFS_I(inode
)->i_orphan
)) {
875 spin_unlock(&root
->list_lock
);
879 list_del_init(&BTRFS_I(inode
)->i_orphan
);
881 spin_unlock(&root
->list_lock
);
885 spin_unlock(&root
->list_lock
);
887 ret
= btrfs_del_orphan_item(trans
, root
, inode
->i_ino
);
893 * this cleans up any orphans that may be left on the list from the last use
896 void btrfs_orphan_cleanup(struct btrfs_root
*root
)
898 struct btrfs_path
*path
;
899 struct extent_buffer
*leaf
;
900 struct btrfs_item
*item
;
901 struct btrfs_key key
, found_key
;
902 struct btrfs_trans_handle
*trans
;
904 int ret
= 0, nr_unlink
= 0, nr_truncate
= 0;
906 /* don't do orphan cleanup if the fs is readonly. */
907 if (root
->inode
->i_sb
->s_flags
& MS_RDONLY
)
910 path
= btrfs_alloc_path();
915 key
.objectid
= BTRFS_ORPHAN_OBJECTID
;
916 btrfs_set_key_type(&key
, BTRFS_ORPHAN_ITEM_KEY
);
917 key
.offset
= (u64
)-1;
919 trans
= btrfs_start_transaction(root
, 1);
920 btrfs_set_trans_block_group(trans
, root
->inode
);
923 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
925 printk(KERN_ERR
"Error searching slot for orphan: %d"
931 * if ret == 0 means we found what we were searching for, which
932 * is weird, but possible, so only screw with path if we didnt
933 * find the key and see if we have stuff that matches
936 if (path
->slots
[0] == 0)
941 /* pull out the item */
942 leaf
= path
->nodes
[0];
943 item
= btrfs_item_nr(leaf
, path
->slots
[0]);
944 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
946 /* make sure the item matches what we want */
947 if (found_key
.objectid
!= BTRFS_ORPHAN_OBJECTID
)
949 if (btrfs_key_type(&found_key
) != BTRFS_ORPHAN_ITEM_KEY
)
952 /* release the path since we're done with it */
953 btrfs_release_path(root
, path
);
956 * this is where we are basically btrfs_lookup, without the
957 * crossing root thing. we store the inode number in the
958 * offset of the orphan item.
960 inode
= btrfs_iget_locked(root
->inode
->i_sb
,
961 found_key
.offset
, root
);
965 if (inode
->i_state
& I_NEW
) {
966 BTRFS_I(inode
)->root
= root
;
968 /* have to set the location manually */
969 BTRFS_I(inode
)->location
.objectid
= inode
->i_ino
;
970 BTRFS_I(inode
)->location
.type
= BTRFS_INODE_ITEM_KEY
;
971 BTRFS_I(inode
)->location
.offset
= 0;
973 btrfs_read_locked_inode(inode
);
974 unlock_new_inode(inode
);
978 * add this inode to the orphan list so btrfs_orphan_del does
979 * the proper thing when we hit it
981 spin_lock(&root
->list_lock
);
982 list_add(&BTRFS_I(inode
)->i_orphan
, &root
->orphan_list
);
983 spin_unlock(&root
->list_lock
);
986 * if this is a bad inode, means we actually succeeded in
987 * removing the inode, but not the orphan record, which means
988 * we need to manually delete the orphan since iput will just
991 if (is_bad_inode(inode
)) {
992 btrfs_orphan_del(trans
, inode
);
997 /* if we have links, this was a truncate, lets do that */
998 if (inode
->i_nlink
) {
1000 btrfs_truncate(inode
);
1005 /* this will do delete_inode and everything for us */
1010 printk(KERN_INFO
"btrfs: unlinked %d orphans\n", nr_unlink
);
1012 printk(KERN_INFO
"btrfs: truncated %d orphans\n", nr_truncate
);
1014 btrfs_free_path(path
);
1015 btrfs_end_transaction(trans
, root
);
1018 void btrfs_read_locked_inode(struct inode
*inode
)
1020 struct btrfs_path
*path
;
1021 struct extent_buffer
*leaf
;
1022 struct btrfs_inode_item
*inode_item
;
1023 struct btrfs_timespec
*tspec
;
1024 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1025 struct btrfs_key location
;
1026 u64 alloc_group_block
;
1030 path
= btrfs_alloc_path();
1032 memcpy(&location
, &BTRFS_I(inode
)->location
, sizeof(location
));
1034 ret
= btrfs_lookup_inode(NULL
, root
, path
, &location
, 0);
1038 leaf
= path
->nodes
[0];
1039 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
1040 struct btrfs_inode_item
);
1042 inode
->i_mode
= btrfs_inode_mode(leaf
, inode_item
);
1043 inode
->i_nlink
= btrfs_inode_nlink(leaf
, inode_item
);
1044 inode
->i_uid
= btrfs_inode_uid(leaf
, inode_item
);
1045 inode
->i_gid
= btrfs_inode_gid(leaf
, inode_item
);
1046 btrfs_i_size_write(inode
, btrfs_inode_size(leaf
, inode_item
));
1048 tspec
= btrfs_inode_atime(inode_item
);
1049 inode
->i_atime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1050 inode
->i_atime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1052 tspec
= btrfs_inode_mtime(inode_item
);
1053 inode
->i_mtime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1054 inode
->i_mtime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1056 tspec
= btrfs_inode_ctime(inode_item
);
1057 inode
->i_ctime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1058 inode
->i_ctime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1060 inode
->i_blocks
= btrfs_inode_nblocks(leaf
, inode_item
);
1061 inode
->i_generation
= btrfs_inode_generation(leaf
, inode_item
);
1063 rdev
= btrfs_inode_rdev(leaf
, inode_item
);
1065 BTRFS_I(inode
)->index_cnt
= (u64
)-1;
1067 alloc_group_block
= btrfs_inode_block_group(leaf
, inode_item
);
1068 BTRFS_I(inode
)->block_group
= btrfs_lookup_block_group(root
->fs_info
,
1070 BTRFS_I(inode
)->flags
= btrfs_inode_flags(leaf
, inode_item
);
1071 if (!BTRFS_I(inode
)->block_group
) {
1072 BTRFS_I(inode
)->block_group
= btrfs_find_block_group(root
,
1074 BTRFS_BLOCK_GROUP_METADATA
, 0);
1076 btrfs_free_path(path
);
1079 switch (inode
->i_mode
& S_IFMT
) {
1081 inode
->i_mapping
->a_ops
= &btrfs_aops
;
1082 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1083 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
1084 inode
->i_fop
= &btrfs_file_operations
;
1085 inode
->i_op
= &btrfs_file_inode_operations
;
1088 inode
->i_fop
= &btrfs_dir_file_operations
;
1089 if (root
== root
->fs_info
->tree_root
)
1090 inode
->i_op
= &btrfs_dir_ro_inode_operations
;
1092 inode
->i_op
= &btrfs_dir_inode_operations
;
1095 inode
->i_op
= &btrfs_symlink_inode_operations
;
1096 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
1097 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1100 init_special_inode(inode
, inode
->i_mode
, rdev
);
1106 btrfs_free_path(path
);
1107 make_bad_inode(inode
);
1110 static void fill_inode_item(struct extent_buffer
*leaf
,
1111 struct btrfs_inode_item
*item
,
1112 struct inode
*inode
)
1114 btrfs_set_inode_uid(leaf
, item
, inode
->i_uid
);
1115 btrfs_set_inode_gid(leaf
, item
, inode
->i_gid
);
1116 btrfs_set_inode_size(leaf
, item
, BTRFS_I(inode
)->disk_i_size
);
1117 btrfs_set_inode_mode(leaf
, item
, inode
->i_mode
);
1118 btrfs_set_inode_nlink(leaf
, item
, inode
->i_nlink
);
1120 btrfs_set_timespec_sec(leaf
, btrfs_inode_atime(item
),
1121 inode
->i_atime
.tv_sec
);
1122 btrfs_set_timespec_nsec(leaf
, btrfs_inode_atime(item
),
1123 inode
->i_atime
.tv_nsec
);
1125 btrfs_set_timespec_sec(leaf
, btrfs_inode_mtime(item
),
1126 inode
->i_mtime
.tv_sec
);
1127 btrfs_set_timespec_nsec(leaf
, btrfs_inode_mtime(item
),
1128 inode
->i_mtime
.tv_nsec
);
1130 btrfs_set_timespec_sec(leaf
, btrfs_inode_ctime(item
),
1131 inode
->i_ctime
.tv_sec
);
1132 btrfs_set_timespec_nsec(leaf
, btrfs_inode_ctime(item
),
1133 inode
->i_ctime
.tv_nsec
);
1135 btrfs_set_inode_nblocks(leaf
, item
, inode
->i_blocks
);
1136 btrfs_set_inode_generation(leaf
, item
, inode
->i_generation
);
1137 btrfs_set_inode_rdev(leaf
, item
, inode
->i_rdev
);
1138 btrfs_set_inode_flags(leaf
, item
, BTRFS_I(inode
)->flags
);
1139 btrfs_set_inode_block_group(leaf
, item
,
1140 BTRFS_I(inode
)->block_group
->key
.objectid
);
1143 int noinline
btrfs_update_inode(struct btrfs_trans_handle
*trans
,
1144 struct btrfs_root
*root
,
1145 struct inode
*inode
)
1147 struct btrfs_inode_item
*inode_item
;
1148 struct btrfs_path
*path
;
1149 struct extent_buffer
*leaf
;
1152 path
= btrfs_alloc_path();
1154 ret
= btrfs_lookup_inode(trans
, root
, path
,
1155 &BTRFS_I(inode
)->location
, 1);
1162 leaf
= path
->nodes
[0];
1163 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
1164 struct btrfs_inode_item
);
1166 fill_inode_item(leaf
, inode_item
, inode
);
1167 btrfs_mark_buffer_dirty(leaf
);
1168 btrfs_set_inode_last_trans(trans
, inode
);
1171 btrfs_free_path(path
);
1176 static int btrfs_unlink_trans(struct btrfs_trans_handle
*trans
,
1177 struct btrfs_root
*root
,
1179 struct dentry
*dentry
)
1181 struct btrfs_path
*path
;
1182 const char *name
= dentry
->d_name
.name
;
1183 int name_len
= dentry
->d_name
.len
;
1185 struct extent_buffer
*leaf
;
1186 struct btrfs_dir_item
*di
;
1187 struct btrfs_key key
;
1190 path
= btrfs_alloc_path();
1196 di
= btrfs_lookup_dir_item(trans
, root
, path
, dir
->i_ino
,
1197 name
, name_len
, -1);
1206 leaf
= path
->nodes
[0];
1207 btrfs_dir_item_key_to_cpu(leaf
, di
, &key
);
1208 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
1211 btrfs_release_path(root
, path
);
1213 ret
= btrfs_del_inode_ref(trans
, root
, name
, name_len
,
1214 dentry
->d_inode
->i_ino
,
1215 dentry
->d_parent
->d_inode
->i_ino
, &index
);
1217 printk("failed to delete reference to %.*s, "
1218 "inode %lu parent %lu\n", name_len
, name
,
1219 dentry
->d_inode
->i_ino
,
1220 dentry
->d_parent
->d_inode
->i_ino
);
1224 di
= btrfs_lookup_dir_index_item(trans
, root
, path
, dir
->i_ino
,
1225 index
, name
, name_len
, -1);
1234 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
1235 btrfs_release_path(root
, path
);
1237 dentry
->d_inode
->i_ctime
= dir
->i_ctime
;
1239 btrfs_free_path(path
);
1241 btrfs_i_size_write(dir
, dir
->i_size
- name_len
* 2);
1242 dir
->i_mtime
= dir
->i_ctime
= CURRENT_TIME
;
1243 btrfs_update_inode(trans
, root
, dir
);
1244 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1245 dentry
->d_inode
->i_nlink
--;
1247 drop_nlink(dentry
->d_inode
);
1249 ret
= btrfs_update_inode(trans
, root
, dentry
->d_inode
);
1250 dir
->i_sb
->s_dirt
= 1;
1255 static int btrfs_unlink(struct inode
*dir
, struct dentry
*dentry
)
1257 struct btrfs_root
*root
;
1258 struct btrfs_trans_handle
*trans
;
1259 struct inode
*inode
= dentry
->d_inode
;
1261 unsigned long nr
= 0;
1263 root
= BTRFS_I(dir
)->root
;
1265 ret
= btrfs_check_free_space(root
, 1, 1);
1269 trans
= btrfs_start_transaction(root
, 1);
1271 btrfs_set_trans_block_group(trans
, dir
);
1272 ret
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
1274 if (inode
->i_nlink
== 0)
1275 ret
= btrfs_orphan_add(trans
, inode
);
1277 nr
= trans
->blocks_used
;
1279 btrfs_end_transaction_throttle(trans
, root
);
1281 btrfs_btree_balance_dirty(root
, nr
);
1285 static int btrfs_rmdir(struct inode
*dir
, struct dentry
*dentry
)
1287 struct inode
*inode
= dentry
->d_inode
;
1290 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1291 struct btrfs_trans_handle
*trans
;
1292 unsigned long nr
= 0;
1294 if (inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
1298 ret
= btrfs_check_free_space(root
, 1, 1);
1302 trans
= btrfs_start_transaction(root
, 1);
1303 btrfs_set_trans_block_group(trans
, dir
);
1305 err
= btrfs_orphan_add(trans
, inode
);
1309 /* now the directory is empty */
1310 err
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
1312 btrfs_i_size_write(inode
, 0);
1316 nr
= trans
->blocks_used
;
1317 ret
= btrfs_end_transaction_throttle(trans
, root
);
1319 btrfs_btree_balance_dirty(root
, nr
);
1327 * this can truncate away extent items, csum items and directory items.
1328 * It starts at a high offset and removes keys until it can't find
1329 * any higher than i_size.
1331 * csum items that cross the new i_size are truncated to the new size
1334 * min_type is the minimum key type to truncate down to. If set to 0, this
1335 * will kill all the items on this inode, including the INODE_ITEM_KEY.
1337 static int btrfs_truncate_in_trans(struct btrfs_trans_handle
*trans
,
1338 struct btrfs_root
*root
,
1339 struct inode
*inode
,
1343 struct btrfs_path
*path
;
1344 struct btrfs_key key
;
1345 struct btrfs_key found_key
;
1347 struct extent_buffer
*leaf
;
1348 struct btrfs_file_extent_item
*fi
;
1349 u64 extent_start
= 0;
1350 u64 extent_num_bytes
= 0;
1356 int pending_del_nr
= 0;
1357 int pending_del_slot
= 0;
1358 int extent_type
= -1;
1359 u64 mask
= root
->sectorsize
- 1;
1361 btrfs_drop_extent_cache(inode
, inode
->i_size
& (~mask
), (u64
)-1);
1362 path
= btrfs_alloc_path();
1366 /* FIXME, add redo link to tree so we don't leak on crash */
1367 key
.objectid
= inode
->i_ino
;
1368 key
.offset
= (u64
)-1;
1371 btrfs_init_path(path
);
1373 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
1378 BUG_ON(path
->slots
[0] == 0);
1384 leaf
= path
->nodes
[0];
1385 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
1386 found_type
= btrfs_key_type(&found_key
);
1388 if (found_key
.objectid
!= inode
->i_ino
)
1391 if (found_type
< min_type
)
1394 item_end
= found_key
.offset
;
1395 if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
1396 fi
= btrfs_item_ptr(leaf
, path
->slots
[0],
1397 struct btrfs_file_extent_item
);
1398 extent_type
= btrfs_file_extent_type(leaf
, fi
);
1399 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1401 btrfs_file_extent_num_bytes(leaf
, fi
);
1402 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1403 struct btrfs_item
*item
= btrfs_item_nr(leaf
,
1405 item_end
+= btrfs_file_extent_inline_len(leaf
,
1410 if (found_type
== BTRFS_CSUM_ITEM_KEY
) {
1411 ret
= btrfs_csum_truncate(trans
, root
, path
,
1415 if (item_end
< inode
->i_size
) {
1416 if (found_type
== BTRFS_DIR_ITEM_KEY
) {
1417 found_type
= BTRFS_INODE_ITEM_KEY
;
1418 } else if (found_type
== BTRFS_EXTENT_ITEM_KEY
) {
1419 found_type
= BTRFS_CSUM_ITEM_KEY
;
1420 } else if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
1421 found_type
= BTRFS_XATTR_ITEM_KEY
;
1422 } else if (found_type
== BTRFS_XATTR_ITEM_KEY
) {
1423 found_type
= BTRFS_INODE_REF_KEY
;
1424 } else if (found_type
) {
1429 btrfs_set_key_type(&key
, found_type
);
1432 if (found_key
.offset
>= inode
->i_size
)
1438 /* FIXME, shrink the extent if the ref count is only 1 */
1439 if (found_type
!= BTRFS_EXTENT_DATA_KEY
)
1442 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1444 extent_start
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
1446 u64 orig_num_bytes
=
1447 btrfs_file_extent_num_bytes(leaf
, fi
);
1448 extent_num_bytes
= inode
->i_size
-
1449 found_key
.offset
+ root
->sectorsize
- 1;
1450 extent_num_bytes
= extent_num_bytes
&
1451 ~((u64
)root
->sectorsize
- 1);
1452 btrfs_set_file_extent_num_bytes(leaf
, fi
,
1454 num_dec
= (orig_num_bytes
-
1456 if (extent_start
!= 0)
1457 dec_i_blocks(inode
, num_dec
);
1458 btrfs_mark_buffer_dirty(leaf
);
1461 btrfs_file_extent_disk_num_bytes(leaf
,
1463 /* FIXME blocksize != 4096 */
1464 num_dec
= btrfs_file_extent_num_bytes(leaf
, fi
);
1465 if (extent_start
!= 0) {
1467 dec_i_blocks(inode
, num_dec
);
1469 root_gen
= btrfs_header_generation(leaf
);
1470 root_owner
= btrfs_header_owner(leaf
);
1472 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1474 u32 newsize
= inode
->i_size
- found_key
.offset
;
1475 dec_i_blocks(inode
, item_end
+ 1 -
1476 found_key
.offset
- newsize
);
1478 btrfs_file_extent_calc_inline_size(newsize
);
1479 ret
= btrfs_truncate_item(trans
, root
, path
,
1483 dec_i_blocks(inode
, item_end
+ 1 -
1489 if (!pending_del_nr
) {
1490 /* no pending yet, add ourselves */
1491 pending_del_slot
= path
->slots
[0];
1493 } else if (pending_del_nr
&&
1494 path
->slots
[0] + 1 == pending_del_slot
) {
1495 /* hop on the pending chunk */
1497 pending_del_slot
= path
->slots
[0];
1499 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path
->slots
[0], pending_del_nr
, pending_del_slot
);
1505 ret
= btrfs_free_extent(trans
, root
, extent_start
,
1508 root_gen
, inode
->i_ino
,
1509 found_key
.offset
, 0);
1513 if (path
->slots
[0] == 0) {
1516 btrfs_release_path(root
, path
);
1521 if (pending_del_nr
&&
1522 path
->slots
[0] + 1 != pending_del_slot
) {
1523 struct btrfs_key debug
;
1525 btrfs_item_key_to_cpu(path
->nodes
[0], &debug
,
1527 ret
= btrfs_del_items(trans
, root
, path
,
1532 btrfs_release_path(root
, path
);
1538 if (pending_del_nr
) {
1539 ret
= btrfs_del_items(trans
, root
, path
, pending_del_slot
,
1542 btrfs_free_path(path
);
1543 inode
->i_sb
->s_dirt
= 1;
1548 * taken from block_truncate_page, but does cow as it zeros out
1549 * any bytes left in the last page in the file.
1551 static int btrfs_truncate_page(struct address_space
*mapping
, loff_t from
)
1553 struct inode
*inode
= mapping
->host
;
1554 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1555 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1556 struct btrfs_ordered_extent
*ordered
;
1558 u32 blocksize
= root
->sectorsize
;
1559 pgoff_t index
= from
>> PAGE_CACHE_SHIFT
;
1560 unsigned offset
= from
& (PAGE_CACHE_SIZE
-1);
1566 if ((offset
& (blocksize
- 1)) == 0)
1571 page
= grab_cache_page(mapping
, index
);
1575 page_start
= page_offset(page
);
1576 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1578 if (!PageUptodate(page
)) {
1579 ret
= btrfs_readpage(NULL
, page
);
1581 if (page
->mapping
!= mapping
) {
1583 page_cache_release(page
);
1586 if (!PageUptodate(page
)) {
1591 wait_on_page_writeback(page
);
1593 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1594 set_page_extent_mapped(page
);
1596 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
1598 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1600 page_cache_release(page
);
1601 btrfs_start_ordered_extent(inode
, ordered
, 1);
1602 btrfs_put_ordered_extent(ordered
);
1606 btrfs_set_extent_delalloc(inode
, page_start
, page_end
);
1608 if (offset
!= PAGE_CACHE_SIZE
) {
1610 memset(kaddr
+ offset
, 0, PAGE_CACHE_SIZE
- offset
);
1611 flush_dcache_page(page
);
1614 ClearPageChecked(page
);
1615 set_page_dirty(page
);
1616 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1620 page_cache_release(page
);
1625 static int btrfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
1627 struct inode
*inode
= dentry
->d_inode
;
1630 err
= inode_change_ok(inode
, attr
);
1634 if (S_ISREG(inode
->i_mode
) &&
1635 attr
->ia_valid
& ATTR_SIZE
&& attr
->ia_size
> inode
->i_size
) {
1636 struct btrfs_trans_handle
*trans
;
1637 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1638 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1640 u64 mask
= root
->sectorsize
- 1;
1641 u64 hole_start
= (inode
->i_size
+ mask
) & ~mask
;
1642 u64 block_end
= (attr
->ia_size
+ mask
) & ~mask
;
1646 if (attr
->ia_size
<= hole_start
)
1649 err
= btrfs_check_free_space(root
, 1, 0);
1653 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
1655 hole_size
= block_end
- hole_start
;
1656 btrfs_wait_ordered_range(inode
, hole_start
, hole_size
);
1657 lock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1659 trans
= btrfs_start_transaction(root
, 1);
1660 btrfs_set_trans_block_group(trans
, inode
);
1661 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
1662 err
= btrfs_drop_extents(trans
, root
, inode
,
1663 hole_start
, block_end
, hole_start
,
1666 if (alloc_hint
!= EXTENT_MAP_INLINE
) {
1667 err
= btrfs_insert_file_extent(trans
, root
,
1671 btrfs_drop_extent_cache(inode
, hole_start
,
1673 btrfs_check_file(root
, inode
);
1675 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
1676 btrfs_end_transaction(trans
, root
);
1677 unlock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1682 err
= inode_setattr(inode
, attr
);
1684 if (!err
&& ((attr
->ia_valid
& ATTR_MODE
)))
1685 err
= btrfs_acl_chmod(inode
);
1690 void btrfs_delete_inode(struct inode
*inode
)
1692 struct btrfs_trans_handle
*trans
;
1693 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1697 truncate_inode_pages(&inode
->i_data
, 0);
1698 if (is_bad_inode(inode
)) {
1699 btrfs_orphan_del(NULL
, inode
);
1702 btrfs_wait_ordered_range(inode
, 0, (u64
)-1);
1704 btrfs_i_size_write(inode
, 0);
1705 trans
= btrfs_start_transaction(root
, 1);
1707 btrfs_set_trans_block_group(trans
, inode
);
1708 ret
= btrfs_truncate_in_trans(trans
, root
, inode
, 0);
1710 btrfs_orphan_del(NULL
, inode
);
1711 goto no_delete_lock
;
1714 btrfs_orphan_del(trans
, inode
);
1716 nr
= trans
->blocks_used
;
1719 btrfs_end_transaction(trans
, root
);
1720 btrfs_btree_balance_dirty(root
, nr
);
1724 nr
= trans
->blocks_used
;
1725 btrfs_end_transaction(trans
, root
);
1726 btrfs_btree_balance_dirty(root
, nr
);
1732 * this returns the key found in the dir entry in the location pointer.
1733 * If no dir entries were found, location->objectid is 0.
1735 static int btrfs_inode_by_name(struct inode
*dir
, struct dentry
*dentry
,
1736 struct btrfs_key
*location
)
1738 const char *name
= dentry
->d_name
.name
;
1739 int namelen
= dentry
->d_name
.len
;
1740 struct btrfs_dir_item
*di
;
1741 struct btrfs_path
*path
;
1742 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1745 if (namelen
== 1 && strcmp(name
, ".") == 0) {
1746 location
->objectid
= dir
->i_ino
;
1747 location
->type
= BTRFS_INODE_ITEM_KEY
;
1748 location
->offset
= 0;
1751 path
= btrfs_alloc_path();
1754 if (namelen
== 2 && strcmp(name
, "..") == 0) {
1755 struct btrfs_key key
;
1756 struct extent_buffer
*leaf
;
1759 key
.objectid
= dir
->i_ino
;
1760 key
.offset
= (u64
)-1;
1761 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1762 if (ret
< 0 || path
->slots
[0] == 0)
1764 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1767 leaf
= path
->nodes
[0];
1768 slot
= path
->slots
[0] - 1;
1770 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
1771 if (key
.objectid
!= dir
->i_ino
||
1772 key
.type
!= BTRFS_INODE_REF_KEY
) {
1775 location
->objectid
= key
.offset
;
1776 location
->type
= BTRFS_INODE_ITEM_KEY
;
1777 location
->offset
= 0;
1781 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir
->i_ino
, name
,
1785 if (!di
|| IS_ERR(di
)) {
1788 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, location
);
1790 btrfs_free_path(path
);
1793 location
->objectid
= 0;
1798 * when we hit a tree root in a directory, the btrfs part of the inode
1799 * needs to be changed to reflect the root directory of the tree root. This
1800 * is kind of like crossing a mount point.
1802 static int fixup_tree_root_location(struct btrfs_root
*root
,
1803 struct btrfs_key
*location
,
1804 struct btrfs_root
**sub_root
,
1805 struct dentry
*dentry
)
1807 struct btrfs_root_item
*ri
;
1809 if (btrfs_key_type(location
) != BTRFS_ROOT_ITEM_KEY
)
1811 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1814 *sub_root
= btrfs_read_fs_root(root
->fs_info
, location
,
1815 dentry
->d_name
.name
,
1816 dentry
->d_name
.len
);
1817 if (IS_ERR(*sub_root
))
1818 return PTR_ERR(*sub_root
);
1820 ri
= &(*sub_root
)->root_item
;
1821 location
->objectid
= btrfs_root_dirid(ri
);
1822 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1823 location
->offset
= 0;
1828 static int btrfs_init_locked_inode(struct inode
*inode
, void *p
)
1830 struct btrfs_iget_args
*args
= p
;
1831 inode
->i_ino
= args
->ino
;
1832 BTRFS_I(inode
)->root
= args
->root
;
1833 BTRFS_I(inode
)->delalloc_bytes
= 0;
1834 inode
->i_mapping
->writeback_index
= 0;
1835 BTRFS_I(inode
)->disk_i_size
= 0;
1836 BTRFS_I(inode
)->index_cnt
= (u64
)-1;
1837 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1838 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1839 inode
->i_mapping
, GFP_NOFS
);
1840 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1841 inode
->i_mapping
, GFP_NOFS
);
1842 INIT_LIST_HEAD(&BTRFS_I(inode
)->delalloc_inodes
);
1843 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
1844 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
1845 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
1849 static int btrfs_find_actor(struct inode
*inode
, void *opaque
)
1851 struct btrfs_iget_args
*args
= opaque
;
1852 return (args
->ino
== inode
->i_ino
&&
1853 args
->root
== BTRFS_I(inode
)->root
);
1856 struct inode
*btrfs_ilookup(struct super_block
*s
, u64 objectid
,
1859 struct btrfs_iget_args args
;
1860 args
.ino
= objectid
;
1861 args
.root
= btrfs_lookup_fs_root(btrfs_sb(s
)->fs_info
, root_objectid
);
1866 return ilookup5(s
, objectid
, btrfs_find_actor
, (void *)&args
);
1869 struct inode
*btrfs_iget_locked(struct super_block
*s
, u64 objectid
,
1870 struct btrfs_root
*root
)
1872 struct inode
*inode
;
1873 struct btrfs_iget_args args
;
1874 args
.ino
= objectid
;
1877 inode
= iget5_locked(s
, objectid
, btrfs_find_actor
,
1878 btrfs_init_locked_inode
,
1883 static struct dentry
*btrfs_lookup(struct inode
*dir
, struct dentry
*dentry
,
1884 struct nameidata
*nd
)
1886 struct inode
* inode
;
1887 struct btrfs_inode
*bi
= BTRFS_I(dir
);
1888 struct btrfs_root
*root
= bi
->root
;
1889 struct btrfs_root
*sub_root
= root
;
1890 struct btrfs_key location
;
1891 int ret
, do_orphan
= 0;
1893 if (dentry
->d_name
.len
> BTRFS_NAME_LEN
)
1894 return ERR_PTR(-ENAMETOOLONG
);
1896 ret
= btrfs_inode_by_name(dir
, dentry
, &location
);
1899 return ERR_PTR(ret
);
1902 if (location
.objectid
) {
1903 ret
= fixup_tree_root_location(root
, &location
, &sub_root
,
1906 return ERR_PTR(ret
);
1908 return ERR_PTR(-ENOENT
);
1910 inode
= btrfs_iget_locked(dir
->i_sb
, location
.objectid
,
1913 return ERR_PTR(-EACCES
);
1914 if (inode
->i_state
& I_NEW
) {
1915 /* the inode and parent dir are two different roots */
1916 if (sub_root
!= root
) {
1918 sub_root
->inode
= inode
;
1921 BTRFS_I(inode
)->root
= sub_root
;
1922 memcpy(&BTRFS_I(inode
)->location
, &location
,
1924 btrfs_read_locked_inode(inode
);
1925 unlock_new_inode(inode
);
1929 if (unlikely(do_orphan
))
1930 btrfs_orphan_cleanup(sub_root
);
1932 return d_splice_alias(inode
, dentry
);
1935 static unsigned char btrfs_filetype_table
[] = {
1936 DT_UNKNOWN
, DT_REG
, DT_DIR
, DT_CHR
, DT_BLK
, DT_FIFO
, DT_SOCK
, DT_LNK
1939 static int btrfs_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
1941 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1942 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1943 struct btrfs_item
*item
;
1944 struct btrfs_dir_item
*di
;
1945 struct btrfs_key key
;
1946 struct btrfs_key found_key
;
1947 struct btrfs_path
*path
;
1950 struct extent_buffer
*leaf
;
1953 unsigned char d_type
;
1958 int key_type
= BTRFS_DIR_INDEX_KEY
;
1963 /* FIXME, use a real flag for deciding about the key type */
1964 if (root
->fs_info
->tree_root
== root
)
1965 key_type
= BTRFS_DIR_ITEM_KEY
;
1967 /* special case for "." */
1968 if (filp
->f_pos
== 0) {
1969 over
= filldir(dirent
, ".", 1,
1977 key
.objectid
= inode
->i_ino
;
1978 path
= btrfs_alloc_path();
1981 /* special case for .., just use the back ref */
1982 if (filp
->f_pos
== 1) {
1983 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1984 key
.offset
= (u64
)-1;
1985 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1986 if (ret
< 0 || path
->slots
[0] == 0) {
1987 btrfs_release_path(root
, path
);
1988 goto read_dir_items
;
1991 leaf
= path
->nodes
[0];
1992 slot
= path
->slots
[0] - 1;
1993 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1994 btrfs_release_path(root
, path
);
1995 if (found_key
.objectid
!= key
.objectid
||
1996 found_key
.type
!= BTRFS_INODE_REF_KEY
)
1997 goto read_dir_items
;
1998 over
= filldir(dirent
, "..", 2,
1999 2, found_key
.offset
, DT_DIR
);
2006 btrfs_set_key_type(&key
, key_type
);
2007 key
.offset
= filp
->f_pos
;
2009 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2014 leaf
= path
->nodes
[0];
2015 nritems
= btrfs_header_nritems(leaf
);
2016 slot
= path
->slots
[0];
2017 if (advance
|| slot
>= nritems
) {
2018 if (slot
>= nritems
-1) {
2019 ret
= btrfs_next_leaf(root
, path
);
2022 leaf
= path
->nodes
[0];
2023 nritems
= btrfs_header_nritems(leaf
);
2024 slot
= path
->slots
[0];
2031 item
= btrfs_item_nr(leaf
, slot
);
2032 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
2034 if (found_key
.objectid
!= key
.objectid
)
2036 if (btrfs_key_type(&found_key
) != key_type
)
2038 if (found_key
.offset
< filp
->f_pos
)
2041 filp
->f_pos
= found_key
.offset
;
2043 di
= btrfs_item_ptr(leaf
, slot
, struct btrfs_dir_item
);
2045 di_total
= btrfs_item_size(leaf
, item
);
2046 while(di_cur
< di_total
) {
2047 struct btrfs_key location
;
2049 name_len
= btrfs_dir_name_len(leaf
, di
);
2050 if (name_len
< 32) {
2051 name_ptr
= tmp_name
;
2053 name_ptr
= kmalloc(name_len
, GFP_NOFS
);
2056 read_extent_buffer(leaf
, name_ptr
,
2057 (unsigned long)(di
+ 1), name_len
);
2059 d_type
= btrfs_filetype_table
[btrfs_dir_type(leaf
, di
)];
2060 btrfs_dir_item_key_to_cpu(leaf
, di
, &location
);
2061 over
= filldir(dirent
, name_ptr
, name_len
,
2066 if (name_ptr
!= tmp_name
)
2071 di_len
= btrfs_dir_name_len(leaf
, di
) +
2072 btrfs_dir_data_len(leaf
, di
) +sizeof(*di
);
2074 di
= (struct btrfs_dir_item
*)((char *)di
+ di_len
);
2077 if (key_type
== BTRFS_DIR_INDEX_KEY
)
2078 filp
->f_pos
= INT_LIMIT(typeof(filp
->f_pos
));
2084 btrfs_free_path(path
);
2088 int btrfs_write_inode(struct inode
*inode
, int wait
)
2090 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2091 struct btrfs_trans_handle
*trans
;
2094 if (root
->fs_info
->closing
> 1)
2098 trans
= btrfs_join_transaction(root
, 1);
2099 btrfs_set_trans_block_group(trans
, inode
);
2100 ret
= btrfs_commit_transaction(trans
, root
);
2106 * This is somewhat expensive, updating the tree every time the
2107 * inode changes. But, it is most likely to find the inode in cache.
2108 * FIXME, needs more benchmarking...there are no reasons other than performance
2109 * to keep or drop this code.
2111 void btrfs_dirty_inode(struct inode
*inode
)
2113 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2114 struct btrfs_trans_handle
*trans
;
2116 trans
= btrfs_join_transaction(root
, 1);
2117 btrfs_set_trans_block_group(trans
, inode
);
2118 btrfs_update_inode(trans
, root
, inode
);
2119 btrfs_end_transaction(trans
, root
);
2122 static int btrfs_set_inode_index_count(struct inode
*inode
)
2124 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2125 struct btrfs_key key
, found_key
;
2126 struct btrfs_path
*path
;
2127 struct extent_buffer
*leaf
;
2130 key
.objectid
= inode
->i_ino
;
2131 btrfs_set_key_type(&key
, BTRFS_DIR_INDEX_KEY
);
2132 key
.offset
= (u64
)-1;
2134 path
= btrfs_alloc_path();
2138 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2141 /* FIXME: we should be able to handle this */
2147 * MAGIC NUMBER EXPLANATION:
2148 * since we search a directory based on f_pos we have to start at 2
2149 * since '.' and '..' have f_pos of 0 and 1 respectively, so everybody
2150 * else has to start at 2
2152 if (path
->slots
[0] == 0) {
2153 BTRFS_I(inode
)->index_cnt
= 2;
2159 leaf
= path
->nodes
[0];
2160 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2162 if (found_key
.objectid
!= inode
->i_ino
||
2163 btrfs_key_type(&found_key
) != BTRFS_DIR_INDEX_KEY
) {
2164 BTRFS_I(inode
)->index_cnt
= 2;
2168 BTRFS_I(inode
)->index_cnt
= found_key
.offset
+ 1;
2170 btrfs_free_path(path
);
2174 static int btrfs_set_inode_index(struct inode
*dir
, struct inode
*inode
,
2179 if (BTRFS_I(dir
)->index_cnt
== (u64
)-1) {
2180 ret
= btrfs_set_inode_index_count(dir
);
2185 *index
= BTRFS_I(dir
)->index_cnt
;
2186 BTRFS_I(dir
)->index_cnt
++;
2191 static struct inode
*btrfs_new_inode(struct btrfs_trans_handle
*trans
,
2192 struct btrfs_root
*root
,
2194 const char *name
, int name_len
,
2197 struct btrfs_block_group_cache
*group
,
2198 int mode
, u64
*index
)
2200 struct inode
*inode
;
2201 struct btrfs_inode_item
*inode_item
;
2202 struct btrfs_block_group_cache
*new_inode_group
;
2203 struct btrfs_key
*location
;
2204 struct btrfs_path
*path
;
2205 struct btrfs_inode_ref
*ref
;
2206 struct btrfs_key key
[2];
2212 path
= btrfs_alloc_path();
2215 inode
= new_inode(root
->fs_info
->sb
);
2217 return ERR_PTR(-ENOMEM
);
2220 ret
= btrfs_set_inode_index(dir
, inode
, index
);
2222 return ERR_PTR(ret
);
2225 * index_cnt is ignored for everything but a dir,
2226 * btrfs_get_inode_index_count has an explanation for the magic
2229 BTRFS_I(inode
)->index_cnt
= 2;
2231 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
2232 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
2233 inode
->i_mapping
, GFP_NOFS
);
2234 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
2235 inode
->i_mapping
, GFP_NOFS
);
2236 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
2237 INIT_LIST_HEAD(&BTRFS_I(inode
)->delalloc_inodes
);
2238 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
2239 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
2240 BTRFS_I(inode
)->delalloc_bytes
= 0;
2241 inode
->i_mapping
->writeback_index
= 0;
2242 BTRFS_I(inode
)->disk_i_size
= 0;
2243 BTRFS_I(inode
)->root
= root
;
2249 new_inode_group
= btrfs_find_block_group(root
, group
, 0,
2250 BTRFS_BLOCK_GROUP_METADATA
, owner
);
2251 if (!new_inode_group
) {
2252 printk("find_block group failed\n");
2253 new_inode_group
= group
;
2255 BTRFS_I(inode
)->block_group
= new_inode_group
;
2256 BTRFS_I(inode
)->flags
= 0;
2258 key
[0].objectid
= objectid
;
2259 btrfs_set_key_type(&key
[0], BTRFS_INODE_ITEM_KEY
);
2262 key
[1].objectid
= objectid
;
2263 btrfs_set_key_type(&key
[1], BTRFS_INODE_REF_KEY
);
2264 key
[1].offset
= ref_objectid
;
2266 sizes
[0] = sizeof(struct btrfs_inode_item
);
2267 sizes
[1] = name_len
+ sizeof(*ref
);
2269 ret
= btrfs_insert_empty_items(trans
, root
, path
, key
, sizes
, 2);
2273 if (objectid
> root
->highest_inode
)
2274 root
->highest_inode
= objectid
;
2276 inode
->i_uid
= current
->fsuid
;
2277 inode
->i_gid
= current
->fsgid
;
2278 inode
->i_mode
= mode
;
2279 inode
->i_ino
= objectid
;
2280 inode
->i_blocks
= 0;
2281 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2282 inode_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
2283 struct btrfs_inode_item
);
2284 fill_inode_item(path
->nodes
[0], inode_item
, inode
);
2286 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0] + 1,
2287 struct btrfs_inode_ref
);
2288 btrfs_set_inode_ref_name_len(path
->nodes
[0], ref
, name_len
);
2289 btrfs_set_inode_ref_index(path
->nodes
[0], ref
, *index
);
2290 ptr
= (unsigned long)(ref
+ 1);
2291 write_extent_buffer(path
->nodes
[0], name
, ptr
, name_len
);
2293 btrfs_mark_buffer_dirty(path
->nodes
[0]);
2294 btrfs_free_path(path
);
2296 location
= &BTRFS_I(inode
)->location
;
2297 location
->objectid
= objectid
;
2298 location
->offset
= 0;
2299 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
2301 insert_inode_hash(inode
);
2305 BTRFS_I(dir
)->index_cnt
--;
2306 btrfs_free_path(path
);
2307 return ERR_PTR(ret
);
2310 static inline u8
btrfs_inode_type(struct inode
*inode
)
2312 return btrfs_type_by_mode
[(inode
->i_mode
& S_IFMT
) >> S_SHIFT
];
2315 static int btrfs_add_link(struct btrfs_trans_handle
*trans
,
2316 struct dentry
*dentry
, struct inode
*inode
,
2317 int add_backref
, u64 index
)
2320 struct btrfs_key key
;
2321 struct btrfs_root
*root
= BTRFS_I(dentry
->d_parent
->d_inode
)->root
;
2322 struct inode
*parent_inode
= dentry
->d_parent
->d_inode
;
2324 key
.objectid
= inode
->i_ino
;
2325 btrfs_set_key_type(&key
, BTRFS_INODE_ITEM_KEY
);
2328 ret
= btrfs_insert_dir_item(trans
, root
,
2329 dentry
->d_name
.name
, dentry
->d_name
.len
,
2330 dentry
->d_parent
->d_inode
->i_ino
,
2331 &key
, btrfs_inode_type(inode
),
2335 ret
= btrfs_insert_inode_ref(trans
, root
,
2336 dentry
->d_name
.name
,
2339 parent_inode
->i_ino
,
2342 btrfs_i_size_write(parent_inode
, parent_inode
->i_size
+
2343 dentry
->d_name
.len
* 2);
2344 parent_inode
->i_mtime
= parent_inode
->i_ctime
= CURRENT_TIME
;
2345 ret
= btrfs_update_inode(trans
, root
,
2346 dentry
->d_parent
->d_inode
);
2351 static int btrfs_add_nondir(struct btrfs_trans_handle
*trans
,
2352 struct dentry
*dentry
, struct inode
*inode
,
2353 int backref
, u64 index
)
2355 int err
= btrfs_add_link(trans
, dentry
, inode
, backref
, index
);
2357 d_instantiate(dentry
, inode
);
2365 static int btrfs_mknod(struct inode
*dir
, struct dentry
*dentry
,
2366 int mode
, dev_t rdev
)
2368 struct btrfs_trans_handle
*trans
;
2369 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2370 struct inode
*inode
= NULL
;
2374 unsigned long nr
= 0;
2377 if (!new_valid_dev(rdev
))
2380 err
= btrfs_check_free_space(root
, 1, 0);
2384 trans
= btrfs_start_transaction(root
, 1);
2385 btrfs_set_trans_block_group(trans
, dir
);
2387 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2393 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2395 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2396 BTRFS_I(dir
)->block_group
, mode
, &index
);
2397 err
= PTR_ERR(inode
);
2401 err
= btrfs_init_acl(inode
, dir
);
2407 btrfs_set_trans_block_group(trans
, inode
);
2408 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
2412 inode
->i_op
= &btrfs_special_inode_operations
;
2413 init_special_inode(inode
, inode
->i_mode
, rdev
);
2414 btrfs_update_inode(trans
, root
, inode
);
2416 dir
->i_sb
->s_dirt
= 1;
2417 btrfs_update_inode_block_group(trans
, inode
);
2418 btrfs_update_inode_block_group(trans
, dir
);
2420 nr
= trans
->blocks_used
;
2421 btrfs_end_transaction_throttle(trans
, root
);
2424 inode_dec_link_count(inode
);
2427 btrfs_btree_balance_dirty(root
, nr
);
2431 static int btrfs_create(struct inode
*dir
, struct dentry
*dentry
,
2432 int mode
, struct nameidata
*nd
)
2434 struct btrfs_trans_handle
*trans
;
2435 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2436 struct inode
*inode
= NULL
;
2439 unsigned long nr
= 0;
2443 err
= btrfs_check_free_space(root
, 1, 0);
2446 trans
= btrfs_start_transaction(root
, 1);
2447 btrfs_set_trans_block_group(trans
, dir
);
2449 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2455 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2457 dentry
->d_parent
->d_inode
->i_ino
,
2458 objectid
, BTRFS_I(dir
)->block_group
, mode
,
2460 err
= PTR_ERR(inode
);
2464 err
= btrfs_init_acl(inode
, dir
);
2470 btrfs_set_trans_block_group(trans
, inode
);
2471 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
2475 inode
->i_mapping
->a_ops
= &btrfs_aops
;
2476 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
2477 inode
->i_fop
= &btrfs_file_operations
;
2478 inode
->i_op
= &btrfs_file_inode_operations
;
2479 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
2480 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
2481 inode
->i_mapping
, GFP_NOFS
);
2482 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
2483 inode
->i_mapping
, GFP_NOFS
);
2484 INIT_LIST_HEAD(&BTRFS_I(inode
)->delalloc_inodes
);
2485 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
2486 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
2487 BTRFS_I(inode
)->delalloc_bytes
= 0;
2488 BTRFS_I(inode
)->disk_i_size
= 0;
2489 inode
->i_mapping
->writeback_index
= 0;
2490 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
2491 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
2493 dir
->i_sb
->s_dirt
= 1;
2494 btrfs_update_inode_block_group(trans
, inode
);
2495 btrfs_update_inode_block_group(trans
, dir
);
2497 nr
= trans
->blocks_used
;
2498 btrfs_end_transaction_throttle(trans
, root
);
2501 inode_dec_link_count(inode
);
2504 btrfs_btree_balance_dirty(root
, nr
);
2508 static int btrfs_link(struct dentry
*old_dentry
, struct inode
*dir
,
2509 struct dentry
*dentry
)
2511 struct btrfs_trans_handle
*trans
;
2512 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2513 struct inode
*inode
= old_dentry
->d_inode
;
2515 unsigned long nr
= 0;
2519 if (inode
->i_nlink
== 0)
2522 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
2527 err
= btrfs_check_free_space(root
, 1, 0);
2530 err
= btrfs_set_inode_index(dir
, inode
, &index
);
2534 trans
= btrfs_start_transaction(root
, 1);
2536 btrfs_set_trans_block_group(trans
, dir
);
2537 atomic_inc(&inode
->i_count
);
2539 err
= btrfs_add_nondir(trans
, dentry
, inode
, 1, index
);
2544 dir
->i_sb
->s_dirt
= 1;
2545 btrfs_update_inode_block_group(trans
, dir
);
2546 err
= btrfs_update_inode(trans
, root
, inode
);
2551 nr
= trans
->blocks_used
;
2552 btrfs_end_transaction_throttle(trans
, root
);
2555 inode_dec_link_count(inode
);
2558 btrfs_btree_balance_dirty(root
, nr
);
2562 static int btrfs_mkdir(struct inode
*dir
, struct dentry
*dentry
, int mode
)
2564 struct inode
*inode
= NULL
;
2565 struct btrfs_trans_handle
*trans
;
2566 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2568 int drop_on_err
= 0;
2571 unsigned long nr
= 1;
2573 err
= btrfs_check_free_space(root
, 1, 0);
2577 trans
= btrfs_start_transaction(root
, 1);
2578 btrfs_set_trans_block_group(trans
, dir
);
2580 if (IS_ERR(trans
)) {
2581 err
= PTR_ERR(trans
);
2585 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2591 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2593 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2594 BTRFS_I(dir
)->block_group
, S_IFDIR
| mode
,
2596 if (IS_ERR(inode
)) {
2597 err
= PTR_ERR(inode
);
2603 err
= btrfs_init_acl(inode
, dir
);
2607 inode
->i_op
= &btrfs_dir_inode_operations
;
2608 inode
->i_fop
= &btrfs_dir_file_operations
;
2609 btrfs_set_trans_block_group(trans
, inode
);
2611 btrfs_i_size_write(inode
, 0);
2612 err
= btrfs_update_inode(trans
, root
, inode
);
2616 err
= btrfs_add_link(trans
, dentry
, inode
, 0, index
);
2620 d_instantiate(dentry
, inode
);
2622 dir
->i_sb
->s_dirt
= 1;
2623 btrfs_update_inode_block_group(trans
, inode
);
2624 btrfs_update_inode_block_group(trans
, dir
);
2627 nr
= trans
->blocks_used
;
2628 btrfs_end_transaction_throttle(trans
, root
);
2633 btrfs_btree_balance_dirty(root
, nr
);
2637 static int merge_extent_mapping(struct extent_map_tree
*em_tree
,
2638 struct extent_map
*existing
,
2639 struct extent_map
*em
,
2640 u64 map_start
, u64 map_len
)
2644 BUG_ON(map_start
< em
->start
|| map_start
>= extent_map_end(em
));
2645 start_diff
= map_start
- em
->start
;
2646 em
->start
= map_start
;
2648 if (em
->block_start
< EXTENT_MAP_LAST_BYTE
)
2649 em
->block_start
+= start_diff
;
2650 return add_extent_mapping(em_tree
, em
);
2653 struct extent_map
*btrfs_get_extent(struct inode
*inode
, struct page
*page
,
2654 size_t pg_offset
, u64 start
, u64 len
,
2660 u64 extent_start
= 0;
2662 u64 objectid
= inode
->i_ino
;
2664 struct btrfs_path
*path
= NULL
;
2665 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2666 struct btrfs_file_extent_item
*item
;
2667 struct extent_buffer
*leaf
;
2668 struct btrfs_key found_key
;
2669 struct extent_map
*em
= NULL
;
2670 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
2671 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2672 struct btrfs_trans_handle
*trans
= NULL
;
2675 spin_lock(&em_tree
->lock
);
2676 em
= lookup_extent_mapping(em_tree
, start
, len
);
2678 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2679 spin_unlock(&em_tree
->lock
);
2682 if (em
->start
> start
|| em
->start
+ em
->len
<= start
)
2683 free_extent_map(em
);
2684 else if (em
->block_start
== EXTENT_MAP_INLINE
&& page
)
2685 free_extent_map(em
);
2689 em
= alloc_extent_map(GFP_NOFS
);
2694 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2695 em
->start
= EXTENT_MAP_HOLE
;
2699 path
= btrfs_alloc_path();
2703 ret
= btrfs_lookup_file_extent(trans
, root
, path
,
2704 objectid
, start
, trans
!= NULL
);
2711 if (path
->slots
[0] == 0)
2716 leaf
= path
->nodes
[0];
2717 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
2718 struct btrfs_file_extent_item
);
2719 /* are we inside the extent that was found? */
2720 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2721 found_type
= btrfs_key_type(&found_key
);
2722 if (found_key
.objectid
!= objectid
||
2723 found_type
!= BTRFS_EXTENT_DATA_KEY
) {
2727 found_type
= btrfs_file_extent_type(leaf
, item
);
2728 extent_start
= found_key
.offset
;
2729 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
2730 extent_end
= extent_start
+
2731 btrfs_file_extent_num_bytes(leaf
, item
);
2733 if (start
< extent_start
|| start
>= extent_end
) {
2735 if (start
< extent_start
) {
2736 if (start
+ len
<= extent_start
)
2738 em
->len
= extent_end
- extent_start
;
2744 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
2746 em
->start
= extent_start
;
2747 em
->len
= extent_end
- extent_start
;
2748 em
->block_start
= EXTENT_MAP_HOLE
;
2751 bytenr
+= btrfs_file_extent_offset(leaf
, item
);
2752 em
->block_start
= bytenr
;
2753 em
->start
= extent_start
;
2754 em
->len
= extent_end
- extent_start
;
2756 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
2761 size_t extent_offset
;
2764 size
= btrfs_file_extent_inline_len(leaf
, btrfs_item_nr(leaf
,
2766 extent_end
= (extent_start
+ size
+ root
->sectorsize
- 1) &
2767 ~((u64
)root
->sectorsize
- 1);
2768 if (start
< extent_start
|| start
>= extent_end
) {
2770 if (start
< extent_start
) {
2771 if (start
+ len
<= extent_start
)
2773 em
->len
= extent_end
- extent_start
;
2779 em
->block_start
= EXTENT_MAP_INLINE
;
2782 em
->start
= extent_start
;
2787 page_start
= page_offset(page
) + pg_offset
;
2788 extent_offset
= page_start
- extent_start
;
2789 copy_size
= min_t(u64
, PAGE_CACHE_SIZE
- pg_offset
,
2790 size
- extent_offset
);
2791 em
->start
= extent_start
+ extent_offset
;
2792 em
->len
= (copy_size
+ root
->sectorsize
- 1) &
2793 ~((u64
)root
->sectorsize
- 1);
2795 ptr
= btrfs_file_extent_inline_start(item
) + extent_offset
;
2796 if (create
== 0 && !PageUptodate(page
)) {
2797 read_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2799 flush_dcache_page(page
);
2800 } else if (create
&& PageUptodate(page
)) {
2803 free_extent_map(em
);
2805 btrfs_release_path(root
, path
);
2806 trans
= btrfs_join_transaction(root
, 1);
2809 write_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2811 btrfs_mark_buffer_dirty(leaf
);
2814 set_extent_uptodate(io_tree
, em
->start
,
2815 extent_map_end(em
) - 1, GFP_NOFS
);
2818 printk("unkknown found_type %d\n", found_type
);
2825 em
->block_start
= EXTENT_MAP_HOLE
;
2827 btrfs_release_path(root
, path
);
2828 if (em
->start
> start
|| extent_map_end(em
) <= start
) {
2829 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em
->start
, em
->len
, start
, len
);
2835 spin_lock(&em_tree
->lock
);
2836 ret
= add_extent_mapping(em_tree
, em
);
2837 /* it is possible that someone inserted the extent into the tree
2838 * while we had the lock dropped. It is also possible that
2839 * an overlapping map exists in the tree
2841 if (ret
== -EEXIST
) {
2842 struct extent_map
*existing
;
2846 existing
= lookup_extent_mapping(em_tree
, start
, len
);
2847 if (existing
&& (existing
->start
> start
||
2848 existing
->start
+ existing
->len
<= start
)) {
2849 free_extent_map(existing
);
2853 existing
= lookup_extent_mapping(em_tree
, em
->start
,
2856 err
= merge_extent_mapping(em_tree
, existing
,
2859 free_extent_map(existing
);
2861 free_extent_map(em
);
2866 printk("failing to insert %Lu %Lu\n",
2868 free_extent_map(em
);
2872 free_extent_map(em
);
2877 spin_unlock(&em_tree
->lock
);
2880 btrfs_free_path(path
);
2882 ret
= btrfs_end_transaction(trans
, root
);
2888 free_extent_map(em
);
2890 return ERR_PTR(err
);
2895 #if 0 /* waiting for O_DIRECT reads */
2896 static int btrfs_get_block(struct inode
*inode
, sector_t iblock
,
2897 struct buffer_head
*bh_result
, int create
)
2899 struct extent_map
*em
;
2900 u64 start
= (u64
)iblock
<< inode
->i_blkbits
;
2901 struct btrfs_multi_bio
*multi
= NULL
;
2902 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2908 em
= btrfs_get_extent(inode
, NULL
, 0, start
, bh_result
->b_size
, 0);
2910 if (!em
|| IS_ERR(em
))
2913 if (em
->start
> start
|| em
->start
+ em
->len
<= start
) {
2917 if (em
->block_start
== EXTENT_MAP_INLINE
) {
2922 len
= em
->start
+ em
->len
- start
;
2923 len
= min_t(u64
, len
, INT_LIMIT(typeof(bh_result
->b_size
)));
2925 if (em
->block_start
== EXTENT_MAP_HOLE
||
2926 em
->block_start
== EXTENT_MAP_DELALLOC
) {
2927 bh_result
->b_size
= len
;
2931 logical
= start
- em
->start
;
2932 logical
= em
->block_start
+ logical
;
2935 ret
= btrfs_map_block(&root
->fs_info
->mapping_tree
, READ
,
2936 logical
, &map_length
, &multi
, 0);
2938 bh_result
->b_blocknr
= multi
->stripes
[0].physical
>> inode
->i_blkbits
;
2939 bh_result
->b_size
= min(map_length
, len
);
2941 bh_result
->b_bdev
= multi
->stripes
[0].dev
->bdev
;
2942 set_buffer_mapped(bh_result
);
2945 free_extent_map(em
);
2950 static ssize_t
btrfs_direct_IO(int rw
, struct kiocb
*iocb
,
2951 const struct iovec
*iov
, loff_t offset
,
2952 unsigned long nr_segs
)
2956 struct file
*file
= iocb
->ki_filp
;
2957 struct inode
*inode
= file
->f_mapping
->host
;
2962 return blockdev_direct_IO(rw
, iocb
, inode
, inode
->i_sb
->s_bdev
, iov
,
2963 offset
, nr_segs
, btrfs_get_block
, NULL
);
2967 static sector_t
btrfs_bmap(struct address_space
*mapping
, sector_t iblock
)
2969 return extent_bmap(mapping
, iblock
, btrfs_get_extent
);
2972 int btrfs_readpage(struct file
*file
, struct page
*page
)
2974 struct extent_io_tree
*tree
;
2975 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2976 return extent_read_full_page(tree
, page
, btrfs_get_extent
);
2979 static int btrfs_writepage(struct page
*page
, struct writeback_control
*wbc
)
2981 struct extent_io_tree
*tree
;
2984 if (current
->flags
& PF_MEMALLOC
) {
2985 redirty_page_for_writepage(wbc
, page
);
2989 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2990 return extent_write_full_page(tree
, page
, btrfs_get_extent
, wbc
);
2993 int btrfs_writepages(struct address_space
*mapping
,
2994 struct writeback_control
*wbc
)
2996 struct extent_io_tree
*tree
;
2997 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
2998 return extent_writepages(tree
, mapping
, btrfs_get_extent
, wbc
);
3002 btrfs_readpages(struct file
*file
, struct address_space
*mapping
,
3003 struct list_head
*pages
, unsigned nr_pages
)
3005 struct extent_io_tree
*tree
;
3006 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
3007 return extent_readpages(tree
, mapping
, pages
, nr_pages
,
3010 static int __btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
3012 struct extent_io_tree
*tree
;
3013 struct extent_map_tree
*map
;
3016 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3017 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
3018 ret
= try_release_extent_mapping(map
, tree
, page
, gfp_flags
);
3020 ClearPagePrivate(page
);
3021 set_page_private(page
, 0);
3022 page_cache_release(page
);
3027 static int btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
3029 return __btrfs_releasepage(page
, gfp_flags
);
3032 static void btrfs_invalidatepage(struct page
*page
, unsigned long offset
)
3034 struct extent_io_tree
*tree
;
3035 struct btrfs_ordered_extent
*ordered
;
3036 u64 page_start
= page_offset(page
);
3037 u64 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
3039 wait_on_page_writeback(page
);
3040 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3042 btrfs_releasepage(page
, GFP_NOFS
);
3046 lock_extent(tree
, page_start
, page_end
, GFP_NOFS
);
3047 ordered
= btrfs_lookup_ordered_extent(page
->mapping
->host
,
3051 * IO on this page will never be started, so we need
3052 * to account for any ordered extents now
3054 clear_extent_bit(tree
, page_start
, page_end
,
3055 EXTENT_DIRTY
| EXTENT_DELALLOC
|
3056 EXTENT_LOCKED
, 1, 0, GFP_NOFS
);
3057 btrfs_finish_ordered_io(page
->mapping
->host
,
3058 page_start
, page_end
);
3059 btrfs_put_ordered_extent(ordered
);
3060 lock_extent(tree
, page_start
, page_end
, GFP_NOFS
);
3062 clear_extent_bit(tree
, page_start
, page_end
,
3063 EXTENT_LOCKED
| EXTENT_DIRTY
| EXTENT_DELALLOC
|
3066 __btrfs_releasepage(page
, GFP_NOFS
);
3068 ClearPageChecked(page
);
3069 if (PagePrivate(page
)) {
3070 ClearPagePrivate(page
);
3071 set_page_private(page
, 0);
3072 page_cache_release(page
);
3077 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
3078 * called from a page fault handler when a page is first dirtied. Hence we must
3079 * be careful to check for EOF conditions here. We set the page up correctly
3080 * for a written page which means we get ENOSPC checking when writing into
3081 * holes and correct delalloc and unwritten extent mapping on filesystems that
3082 * support these features.
3084 * We are not allowed to take the i_mutex here so we have to play games to
3085 * protect against truncate races as the page could now be beyond EOF. Because
3086 * vmtruncate() writes the inode size before removing pages, once we have the
3087 * page lock we can determine safely if the page is beyond EOF. If it is not
3088 * beyond EOF, then the page is guaranteed safe against truncation until we
3091 int btrfs_page_mkwrite(struct vm_area_struct
*vma
, struct page
*page
)
3093 struct inode
*inode
= fdentry(vma
->vm_file
)->d_inode
;
3094 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3095 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
3096 struct btrfs_ordered_extent
*ordered
;
3098 unsigned long zero_start
;
3104 ret
= btrfs_check_free_space(root
, PAGE_CACHE_SIZE
, 0);
3111 size
= i_size_read(inode
);
3112 page_start
= page_offset(page
);
3113 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
3115 if ((page
->mapping
!= inode
->i_mapping
) ||
3116 (page_start
>= size
)) {
3117 /* page got truncated out from underneath us */
3120 wait_on_page_writeback(page
);
3122 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3123 set_page_extent_mapped(page
);
3126 * we can't set the delalloc bits if there are pending ordered
3127 * extents. Drop our locks and wait for them to finish
3129 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
3131 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3133 btrfs_start_ordered_extent(inode
, ordered
, 1);
3134 btrfs_put_ordered_extent(ordered
);
3138 btrfs_set_extent_delalloc(inode
, page_start
, page_end
);
3141 /* page is wholly or partially inside EOF */
3142 if (page_start
+ PAGE_CACHE_SIZE
> size
)
3143 zero_start
= size
& ~PAGE_CACHE_MASK
;
3145 zero_start
= PAGE_CACHE_SIZE
;
3147 if (zero_start
!= PAGE_CACHE_SIZE
) {
3149 memset(kaddr
+ zero_start
, 0, PAGE_CACHE_SIZE
- zero_start
);
3150 flush_dcache_page(page
);
3153 ClearPageChecked(page
);
3154 set_page_dirty(page
);
3155 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3163 static void btrfs_truncate(struct inode
*inode
)
3165 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3167 struct btrfs_trans_handle
*trans
;
3169 u64 mask
= root
->sectorsize
- 1;
3171 if (!S_ISREG(inode
->i_mode
))
3173 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
3176 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
3177 btrfs_wait_ordered_range(inode
, inode
->i_size
& (~mask
), (u64
)-1);
3179 trans
= btrfs_start_transaction(root
, 1);
3180 btrfs_set_trans_block_group(trans
, inode
);
3181 btrfs_i_size_write(inode
, inode
->i_size
);
3183 ret
= btrfs_orphan_add(trans
, inode
);
3186 /* FIXME, add redo link to tree so we don't leak on crash */
3187 ret
= btrfs_truncate_in_trans(trans
, root
, inode
,
3188 BTRFS_EXTENT_DATA_KEY
);
3189 btrfs_update_inode(trans
, root
, inode
);
3191 ret
= btrfs_orphan_del(trans
, inode
);
3195 nr
= trans
->blocks_used
;
3196 ret
= btrfs_end_transaction_throttle(trans
, root
);
3198 btrfs_btree_balance_dirty(root
, nr
);
3202 * Invalidate a single dcache entry at the root of the filesystem.
3203 * Needed after creation of snapshot or subvolume.
3205 void btrfs_invalidate_dcache_root(struct btrfs_root
*root
, char *name
,
3208 struct dentry
*alias
, *entry
;
3211 alias
= d_find_alias(root
->fs_info
->sb
->s_root
->d_inode
);
3215 /* change me if btrfs ever gets a d_hash operation */
3216 qstr
.hash
= full_name_hash(qstr
.name
, qstr
.len
);
3217 entry
= d_lookup(alias
, &qstr
);
3220 d_invalidate(entry
);
3226 int btrfs_create_subvol_root(struct btrfs_root
*new_root
,
3227 struct btrfs_trans_handle
*trans
, u64 new_dirid
,
3228 struct btrfs_block_group_cache
*block_group
)
3230 struct inode
*inode
;
3233 inode
= btrfs_new_inode(trans
, new_root
, NULL
, "..", 2, new_dirid
,
3234 new_dirid
, block_group
, S_IFDIR
| 0700, &index
);
3236 return PTR_ERR(inode
);
3237 inode
->i_op
= &btrfs_dir_inode_operations
;
3238 inode
->i_fop
= &btrfs_dir_file_operations
;
3239 new_root
->inode
= inode
;
3242 btrfs_i_size_write(inode
, 0);
3244 return btrfs_update_inode(trans
, new_root
, inode
);
3247 unsigned long btrfs_force_ra(struct address_space
*mapping
,
3248 struct file_ra_state
*ra
, struct file
*file
,
3249 pgoff_t offset
, pgoff_t last_index
)
3251 pgoff_t req_size
= last_index
- offset
+ 1;
3253 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3254 offset
= page_cache_readahead(mapping
, ra
, file
, offset
, req_size
);
3257 page_cache_sync_readahead(mapping
, ra
, file
, offset
, req_size
);
3258 return offset
+ req_size
;
3262 struct inode
*btrfs_alloc_inode(struct super_block
*sb
)
3264 struct btrfs_inode
*ei
;
3266 ei
= kmem_cache_alloc(btrfs_inode_cachep
, GFP_NOFS
);
3270 btrfs_ordered_inode_tree_init(&ei
->ordered_tree
);
3271 ei
->i_acl
= BTRFS_ACL_NOT_CACHED
;
3272 ei
->i_default_acl
= BTRFS_ACL_NOT_CACHED
;
3273 INIT_LIST_HEAD(&ei
->i_orphan
);
3274 return &ei
->vfs_inode
;
3277 void btrfs_destroy_inode(struct inode
*inode
)
3279 struct btrfs_ordered_extent
*ordered
;
3280 WARN_ON(!list_empty(&inode
->i_dentry
));
3281 WARN_ON(inode
->i_data
.nrpages
);
3283 if (BTRFS_I(inode
)->i_acl
&&
3284 BTRFS_I(inode
)->i_acl
!= BTRFS_ACL_NOT_CACHED
)
3285 posix_acl_release(BTRFS_I(inode
)->i_acl
);
3286 if (BTRFS_I(inode
)->i_default_acl
&&
3287 BTRFS_I(inode
)->i_default_acl
!= BTRFS_ACL_NOT_CACHED
)
3288 posix_acl_release(BTRFS_I(inode
)->i_default_acl
);
3290 spin_lock(&BTRFS_I(inode
)->root
->list_lock
);
3291 if (!list_empty(&BTRFS_I(inode
)->i_orphan
)) {
3292 printk(KERN_ERR
"BTRFS: inode %lu: inode still on the orphan"
3293 " list\n", inode
->i_ino
);
3296 spin_unlock(&BTRFS_I(inode
)->root
->list_lock
);
3299 ordered
= btrfs_lookup_first_ordered_extent(inode
, (u64
)-1);
3303 printk("found ordered extent %Lu %Lu\n",
3304 ordered
->file_offset
, ordered
->len
);
3305 btrfs_remove_ordered_extent(inode
, ordered
);
3306 btrfs_put_ordered_extent(ordered
);
3307 btrfs_put_ordered_extent(ordered
);
3310 btrfs_drop_extent_cache(inode
, 0, (u64
)-1);
3311 kmem_cache_free(btrfs_inode_cachep
, BTRFS_I(inode
));
3314 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3315 static void init_once(void *foo
)
3316 #elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3317 static void init_once(struct kmem_cache
* cachep
, void *foo
)
3319 static void init_once(void * foo
, struct kmem_cache
* cachep
,
3320 unsigned long flags
)
3323 struct btrfs_inode
*ei
= (struct btrfs_inode
*) foo
;
3325 inode_init_once(&ei
->vfs_inode
);
3328 void btrfs_destroy_cachep(void)
3330 if (btrfs_inode_cachep
)
3331 kmem_cache_destroy(btrfs_inode_cachep
);
3332 if (btrfs_trans_handle_cachep
)
3333 kmem_cache_destroy(btrfs_trans_handle_cachep
);
3334 if (btrfs_transaction_cachep
)
3335 kmem_cache_destroy(btrfs_transaction_cachep
);
3336 if (btrfs_bit_radix_cachep
)
3337 kmem_cache_destroy(btrfs_bit_radix_cachep
);
3338 if (btrfs_path_cachep
)
3339 kmem_cache_destroy(btrfs_path_cachep
);
3342 struct kmem_cache
*btrfs_cache_create(const char *name
, size_t size
,
3343 unsigned long extra_flags
,
3344 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3345 void (*ctor
)(void *)
3346 #elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3347 void (*ctor
)(struct kmem_cache
*, void *)
3349 void (*ctor
)(void *, struct kmem_cache
*,
3354 return kmem_cache_create(name
, size
, 0, (SLAB_RECLAIM_ACCOUNT
|
3355 SLAB_MEM_SPREAD
| extra_flags
), ctor
3356 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3362 int btrfs_init_cachep(void)
3364 btrfs_inode_cachep
= btrfs_cache_create("btrfs_inode_cache",
3365 sizeof(struct btrfs_inode
),
3367 if (!btrfs_inode_cachep
)
3369 btrfs_trans_handle_cachep
=
3370 btrfs_cache_create("btrfs_trans_handle_cache",
3371 sizeof(struct btrfs_trans_handle
),
3373 if (!btrfs_trans_handle_cachep
)
3375 btrfs_transaction_cachep
= btrfs_cache_create("btrfs_transaction_cache",
3376 sizeof(struct btrfs_transaction
),
3378 if (!btrfs_transaction_cachep
)
3380 btrfs_path_cachep
= btrfs_cache_create("btrfs_path_cache",
3381 sizeof(struct btrfs_path
),
3383 if (!btrfs_path_cachep
)
3385 btrfs_bit_radix_cachep
= btrfs_cache_create("btrfs_radix", 256,
3386 SLAB_DESTROY_BY_RCU
, NULL
);
3387 if (!btrfs_bit_radix_cachep
)
3391 btrfs_destroy_cachep();
3395 static int btrfs_getattr(struct vfsmount
*mnt
,
3396 struct dentry
*dentry
, struct kstat
*stat
)
3398 struct inode
*inode
= dentry
->d_inode
;
3399 generic_fillattr(inode
, stat
);
3400 stat
->blksize
= PAGE_CACHE_SIZE
;
3401 stat
->blocks
= inode
->i_blocks
+ (BTRFS_I(inode
)->delalloc_bytes
>> 9);
3405 static int btrfs_rename(struct inode
* old_dir
, struct dentry
*old_dentry
,
3406 struct inode
* new_dir
,struct dentry
*new_dentry
)
3408 struct btrfs_trans_handle
*trans
;
3409 struct btrfs_root
*root
= BTRFS_I(old_dir
)->root
;
3410 struct inode
*new_inode
= new_dentry
->d_inode
;
3411 struct inode
*old_inode
= old_dentry
->d_inode
;
3412 struct timespec ctime
= CURRENT_TIME
;
3416 if (S_ISDIR(old_inode
->i_mode
) && new_inode
&&
3417 new_inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
3421 ret
= btrfs_check_free_space(root
, 1, 0);
3425 trans
= btrfs_start_transaction(root
, 1);
3427 btrfs_set_trans_block_group(trans
, new_dir
);
3429 old_dentry
->d_inode
->i_nlink
++;
3430 old_dir
->i_ctime
= old_dir
->i_mtime
= ctime
;
3431 new_dir
->i_ctime
= new_dir
->i_mtime
= ctime
;
3432 old_inode
->i_ctime
= ctime
;
3434 ret
= btrfs_unlink_trans(trans
, root
, old_dir
, old_dentry
);
3439 new_inode
->i_ctime
= CURRENT_TIME
;
3440 ret
= btrfs_unlink_trans(trans
, root
, new_dir
, new_dentry
);
3443 if (new_inode
->i_nlink
== 0) {
3444 ret
= btrfs_orphan_add(trans
, new_inode
);
3449 ret
= btrfs_set_inode_index(new_dir
, old_inode
, &index
);
3453 ret
= btrfs_add_link(trans
, new_dentry
, old_inode
, 1, index
);
3458 btrfs_end_transaction_throttle(trans
, root
);
3463 int btrfs_start_delalloc_inodes(struct btrfs_root
*root
)
3465 struct list_head
*head
= &root
->fs_info
->delalloc_inodes
;
3466 struct btrfs_inode
*binode
;
3467 unsigned long flags
;
3469 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
3470 while(!list_empty(head
)) {
3471 binode
= list_entry(head
->next
, struct btrfs_inode
,
3473 atomic_inc(&binode
->vfs_inode
.i_count
);
3474 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
3475 filemap_write_and_wait(binode
->vfs_inode
.i_mapping
);
3476 iput(&binode
->vfs_inode
);
3477 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
3479 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
3483 static int btrfs_symlink(struct inode
*dir
, struct dentry
*dentry
,
3484 const char *symname
)
3486 struct btrfs_trans_handle
*trans
;
3487 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
3488 struct btrfs_path
*path
;
3489 struct btrfs_key key
;
3490 struct inode
*inode
= NULL
;
3498 struct btrfs_file_extent_item
*ei
;
3499 struct extent_buffer
*leaf
;
3500 unsigned long nr
= 0;
3502 name_len
= strlen(symname
) + 1;
3503 if (name_len
> BTRFS_MAX_INLINE_DATA_SIZE(root
))
3504 return -ENAMETOOLONG
;
3506 err
= btrfs_check_free_space(root
, 1, 0);
3510 trans
= btrfs_start_transaction(root
, 1);
3511 btrfs_set_trans_block_group(trans
, dir
);
3513 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
3519 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
3521 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
3522 BTRFS_I(dir
)->block_group
, S_IFLNK
|S_IRWXUGO
,
3524 err
= PTR_ERR(inode
);
3528 err
= btrfs_init_acl(inode
, dir
);
3534 btrfs_set_trans_block_group(trans
, inode
);
3535 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
3539 inode
->i_mapping
->a_ops
= &btrfs_aops
;
3540 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3541 inode
->i_fop
= &btrfs_file_operations
;
3542 inode
->i_op
= &btrfs_file_inode_operations
;
3543 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
3544 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
3545 inode
->i_mapping
, GFP_NOFS
);
3546 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
3547 inode
->i_mapping
, GFP_NOFS
);
3548 INIT_LIST_HEAD(&BTRFS_I(inode
)->delalloc_inodes
);
3549 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
3550 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
3551 BTRFS_I(inode
)->delalloc_bytes
= 0;
3552 BTRFS_I(inode
)->disk_i_size
= 0;
3553 inode
->i_mapping
->writeback_index
= 0;
3554 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
3555 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
3557 dir
->i_sb
->s_dirt
= 1;
3558 btrfs_update_inode_block_group(trans
, inode
);
3559 btrfs_update_inode_block_group(trans
, dir
);
3563 path
= btrfs_alloc_path();
3565 key
.objectid
= inode
->i_ino
;
3567 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
3568 datasize
= btrfs_file_extent_calc_inline_size(name_len
);
3569 err
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
3575 leaf
= path
->nodes
[0];
3576 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
3577 struct btrfs_file_extent_item
);
3578 btrfs_set_file_extent_generation(leaf
, ei
, trans
->transid
);
3579 btrfs_set_file_extent_type(leaf
, ei
,
3580 BTRFS_FILE_EXTENT_INLINE
);
3581 ptr
= btrfs_file_extent_inline_start(ei
);
3582 write_extent_buffer(leaf
, symname
, ptr
, name_len
);
3583 btrfs_mark_buffer_dirty(leaf
);
3584 btrfs_free_path(path
);
3586 inode
->i_op
= &btrfs_symlink_inode_operations
;
3587 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
3588 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3589 btrfs_i_size_write(inode
, name_len
- 1);
3590 err
= btrfs_update_inode(trans
, root
, inode
);
3595 nr
= trans
->blocks_used
;
3596 btrfs_end_transaction_throttle(trans
, root
);
3599 inode_dec_link_count(inode
);
3602 btrfs_btree_balance_dirty(root
, nr
);
3606 static int btrfs_set_page_dirty(struct page
*page
)
3608 return __set_page_dirty_nobuffers(page
);
3611 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3612 static int btrfs_permission(struct inode
*inode
, int mask
)
3614 static int btrfs_permission(struct inode
*inode
, int mask
,
3615 struct nameidata
*nd
)
3618 if (btrfs_test_flag(inode
, READONLY
) && (mask
& MAY_WRITE
))
3620 return generic_permission(inode
, mask
, btrfs_check_acl
);
3623 static struct inode_operations btrfs_dir_inode_operations
= {
3624 .lookup
= btrfs_lookup
,
3625 .create
= btrfs_create
,
3626 .unlink
= btrfs_unlink
,
3628 .mkdir
= btrfs_mkdir
,
3629 .rmdir
= btrfs_rmdir
,
3630 .rename
= btrfs_rename
,
3631 .symlink
= btrfs_symlink
,
3632 .setattr
= btrfs_setattr
,
3633 .mknod
= btrfs_mknod
,
3634 .setxattr
= generic_setxattr
,
3635 .getxattr
= generic_getxattr
,
3636 .listxattr
= btrfs_listxattr
,
3637 .removexattr
= generic_removexattr
,
3638 .permission
= btrfs_permission
,
3640 static struct inode_operations btrfs_dir_ro_inode_operations
= {
3641 .lookup
= btrfs_lookup
,
3642 .permission
= btrfs_permission
,
3644 static struct file_operations btrfs_dir_file_operations
= {
3645 .llseek
= generic_file_llseek
,
3646 .read
= generic_read_dir
,
3647 .readdir
= btrfs_readdir
,
3648 .unlocked_ioctl
= btrfs_ioctl
,
3649 #ifdef CONFIG_COMPAT
3650 .compat_ioctl
= btrfs_ioctl
,
3652 .release
= btrfs_release_file
,
3655 static struct extent_io_ops btrfs_extent_io_ops
= {
3656 .fill_delalloc
= run_delalloc_range
,
3657 .submit_bio_hook
= btrfs_submit_bio_hook
,
3658 .merge_bio_hook
= btrfs_merge_bio_hook
,
3659 .readpage_io_hook
= btrfs_readpage_io_hook
,
3660 .readpage_end_io_hook
= btrfs_readpage_end_io_hook
,
3661 .writepage_end_io_hook
= btrfs_writepage_end_io_hook
,
3662 .writepage_start_hook
= btrfs_writepage_start_hook
,
3663 .readpage_io_failed_hook
= btrfs_io_failed_hook
,
3664 .set_bit_hook
= btrfs_set_bit_hook
,
3665 .clear_bit_hook
= btrfs_clear_bit_hook
,
3668 static struct address_space_operations btrfs_aops
= {
3669 .readpage
= btrfs_readpage
,
3670 .writepage
= btrfs_writepage
,
3671 .writepages
= btrfs_writepages
,
3672 .readpages
= btrfs_readpages
,
3673 .sync_page
= block_sync_page
,
3675 .direct_IO
= btrfs_direct_IO
,
3676 .invalidatepage
= btrfs_invalidatepage
,
3677 .releasepage
= btrfs_releasepage
,
3678 .set_page_dirty
= btrfs_set_page_dirty
,
3681 static struct address_space_operations btrfs_symlink_aops
= {
3682 .readpage
= btrfs_readpage
,
3683 .writepage
= btrfs_writepage
,
3684 .invalidatepage
= btrfs_invalidatepage
,
3685 .releasepage
= btrfs_releasepage
,
3688 static struct inode_operations btrfs_file_inode_operations
= {
3689 .truncate
= btrfs_truncate
,
3690 .getattr
= btrfs_getattr
,
3691 .setattr
= btrfs_setattr
,
3692 .setxattr
= generic_setxattr
,
3693 .getxattr
= generic_getxattr
,
3694 .listxattr
= btrfs_listxattr
,
3695 .removexattr
= generic_removexattr
,
3696 .permission
= btrfs_permission
,
3698 static struct inode_operations btrfs_special_inode_operations
= {
3699 .getattr
= btrfs_getattr
,
3700 .setattr
= btrfs_setattr
,
3701 .permission
= btrfs_permission
,
3702 .setxattr
= generic_setxattr
,
3703 .getxattr
= generic_getxattr
,
3704 .listxattr
= btrfs_listxattr
,
3705 .removexattr
= generic_removexattr
,
3707 static struct inode_operations btrfs_symlink_inode_operations
= {
3708 .readlink
= generic_readlink
,
3709 .follow_link
= page_follow_link_light
,
3710 .put_link
= page_put_link
,
3711 .permission
= btrfs_permission
,