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
)
193 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
194 struct btrfs_block_group_cache
*block_group
;
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();
207 ret
= btrfs_lookup_file_extent(NULL
, root
, path
,
208 inode
->i_ino
, start
, 0);
210 btrfs_free_path(path
);
216 if (path
->slots
[0] == 0)
221 leaf
= path
->nodes
[0];
222 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
223 struct btrfs_file_extent_item
);
225 /* are we inside the extent that was found? */
226 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
227 found_type
= btrfs_key_type(&found_key
);
228 if (found_key
.objectid
!= inode
->i_ino
||
229 found_type
!= BTRFS_EXTENT_DATA_KEY
)
232 found_type
= btrfs_file_extent_type(leaf
, item
);
233 extent_start
= found_key
.offset
;
234 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
235 u64 extent_num_bytes
;
237 extent_num_bytes
= btrfs_file_extent_num_bytes(leaf
, item
);
238 extent_end
= extent_start
+ extent_num_bytes
;
241 if (loops
&& start
!= extent_start
)
244 if (start
< extent_start
|| start
>= extent_end
)
247 cow_end
= min(end
, extent_end
- 1);
248 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
252 if (btrfs_count_snapshots_in_path(root
, path
, inode
->i_ino
,
258 * we may be called by the resizer, make sure we're inside
259 * the limits of the FS
261 block_group
= btrfs_lookup_block_group(root
->fs_info
,
263 if (!block_group
|| block_group
->ro
)
272 btrfs_free_path(path
);
275 btrfs_release_path(root
, path
);
280 cow_file_range(inode
, start
, end
);
285 static int run_delalloc_range(struct inode
*inode
, u64 start
, u64 end
)
287 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
290 if (btrfs_test_opt(root
, NODATACOW
) ||
291 btrfs_test_flag(inode
, NODATACOW
))
292 ret
= run_delalloc_nocow(inode
, start
, end
);
294 ret
= cow_file_range(inode
, start
, end
);
299 int btrfs_set_bit_hook(struct inode
*inode
, u64 start
, u64 end
,
300 unsigned long old
, unsigned long bits
)
303 if (!(old
& EXTENT_DELALLOC
) && (bits
& EXTENT_DELALLOC
)) {
304 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
305 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
306 BTRFS_I(inode
)->delalloc_bytes
+= end
- start
+ 1;
307 root
->fs_info
->delalloc_bytes
+= end
- start
+ 1;
308 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
313 int btrfs_clear_bit_hook(struct inode
*inode
, u64 start
, u64 end
,
314 unsigned long old
, unsigned long bits
)
316 if ((old
& EXTENT_DELALLOC
) && (bits
& EXTENT_DELALLOC
)) {
317 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
320 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
321 if (end
- start
+ 1 > root
->fs_info
->delalloc_bytes
) {
322 printk("warning: delalloc account %Lu %Lu\n",
323 end
- start
+ 1, root
->fs_info
->delalloc_bytes
);
324 root
->fs_info
->delalloc_bytes
= 0;
325 BTRFS_I(inode
)->delalloc_bytes
= 0;
327 root
->fs_info
->delalloc_bytes
-= end
- start
+ 1;
328 BTRFS_I(inode
)->delalloc_bytes
-= end
- start
+ 1;
330 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
335 int btrfs_merge_bio_hook(struct page
*page
, unsigned long offset
,
336 size_t size
, struct bio
*bio
)
338 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
339 struct btrfs_mapping_tree
*map_tree
;
340 u64 logical
= bio
->bi_sector
<< 9;
345 length
= bio
->bi_size
;
346 map_tree
= &root
->fs_info
->mapping_tree
;
348 ret
= btrfs_map_block(map_tree
, READ
, logical
,
349 &map_length
, NULL
, 0);
351 if (map_length
< length
+ size
) {
357 int __btrfs_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
360 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
363 ret
= btrfs_csum_one_bio(root
, inode
, bio
);
366 return btrfs_map_bio(root
, rw
, bio
, mirror_num
, 1);
369 int btrfs_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
372 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
375 ret
= btrfs_bio_wq_end_io(root
->fs_info
, bio
, 0);
378 if (!(rw
& (1 << BIO_RW
))) {
382 return btrfs_wq_submit_bio(BTRFS_I(inode
)->root
->fs_info
,
383 inode
, rw
, bio
, mirror_num
,
384 __btrfs_submit_bio_hook
);
386 return btrfs_map_bio(root
, rw
, bio
, mirror_num
, 0);
389 static noinline
int add_pending_csums(struct btrfs_trans_handle
*trans
,
390 struct inode
*inode
, u64 file_offset
,
391 struct list_head
*list
)
393 struct list_head
*cur
;
394 struct btrfs_ordered_sum
*sum
;
396 btrfs_set_trans_block_group(trans
, inode
);
397 list_for_each(cur
, list
) {
398 sum
= list_entry(cur
, struct btrfs_ordered_sum
, list
);
399 mutex_lock(&BTRFS_I(inode
)->csum_mutex
);
400 btrfs_csum_file_blocks(trans
, BTRFS_I(inode
)->root
,
402 mutex_unlock(&BTRFS_I(inode
)->csum_mutex
);
407 struct btrfs_writepage_fixup
{
409 struct btrfs_work work
;
412 /* see btrfs_writepage_start_hook for details on why this is required */
413 void btrfs_writepage_fixup_worker(struct btrfs_work
*work
)
415 struct btrfs_writepage_fixup
*fixup
;
416 struct btrfs_ordered_extent
*ordered
;
422 fixup
= container_of(work
, struct btrfs_writepage_fixup
, work
);
426 if (!page
->mapping
|| !PageDirty(page
) || !PageChecked(page
)) {
427 ClearPageChecked(page
);
431 inode
= page
->mapping
->host
;
432 page_start
= page_offset(page
);
433 page_end
= page_offset(page
) + PAGE_CACHE_SIZE
- 1;
435 lock_extent(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
, GFP_NOFS
);
437 /* already ordered? We're done */
438 if (test_range_bit(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
,
439 EXTENT_ORDERED
, 0)) {
443 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
445 unlock_extent(&BTRFS_I(inode
)->io_tree
, page_start
,
448 btrfs_start_ordered_extent(inode
, ordered
, 1);
452 set_extent_delalloc(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
,
454 ClearPageChecked(page
);
456 unlock_extent(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
, GFP_NOFS
);
459 page_cache_release(page
);
463 * There are a few paths in the higher layers of the kernel that directly
464 * set the page dirty bit without asking the filesystem if it is a
465 * good idea. This causes problems because we want to make sure COW
466 * properly happens and the data=ordered rules are followed.
468 * In our case any range that doesn't have the EXTENT_ORDERED bit set
469 * hasn't been properly setup for IO. We kick off an async process
470 * to fix it up. The async helper will wait for ordered extents, set
471 * the delalloc bit and make it safe to write the page.
473 int btrfs_writepage_start_hook(struct page
*page
, u64 start
, u64 end
)
475 struct inode
*inode
= page
->mapping
->host
;
476 struct btrfs_writepage_fixup
*fixup
;
477 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
480 ret
= test_range_bit(&BTRFS_I(inode
)->io_tree
, start
, end
,
485 if (PageChecked(page
))
488 fixup
= kzalloc(sizeof(*fixup
), GFP_NOFS
);
492 SetPageChecked(page
);
493 page_cache_get(page
);
494 fixup
->work
.func
= btrfs_writepage_fixup_worker
;
496 btrfs_queue_worker(&root
->fs_info
->fixup_workers
, &fixup
->work
);
500 static int btrfs_finish_ordered_io(struct inode
*inode
, u64 start
, u64 end
)
502 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
503 struct btrfs_trans_handle
*trans
;
504 struct btrfs_ordered_extent
*ordered_extent
;
505 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
506 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
507 struct extent_map
*em
;
508 struct extent_map
*em_orig
;
512 struct list_head list
;
513 struct btrfs_key ins
;
517 ret
= btrfs_dec_test_ordered_pending(inode
, start
, end
- start
+ 1);
521 trans
= btrfs_join_transaction(root
, 1);
523 ordered_extent
= btrfs_lookup_ordered_extent(inode
, start
);
524 BUG_ON(!ordered_extent
);
526 lock_extent(io_tree
, ordered_extent
->file_offset
,
527 ordered_extent
->file_offset
+ ordered_extent
->len
- 1,
530 INIT_LIST_HEAD(&list
);
532 ins
.objectid
= ordered_extent
->start
;
533 ins
.offset
= ordered_extent
->len
;
534 ins
.type
= BTRFS_EXTENT_ITEM_KEY
;
536 ret
= btrfs_alloc_reserved_extent(trans
, root
, root
->root_key
.objectid
,
537 trans
->transid
, inode
->i_ino
,
538 ordered_extent
->file_offset
, &ins
);
541 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
543 spin_lock(&em_tree
->lock
);
544 clear_start
= ordered_extent
->file_offset
;
545 clear_end
= ordered_extent
->file_offset
+ ordered_extent
->len
;
546 em
= lookup_extent_mapping(em_tree
, clear_start
,
547 ordered_extent
->len
);
549 while(em
&& clear_start
< extent_map_end(em
) && clear_end
> em
->start
) {
550 clear_bit(EXTENT_FLAG_PINNED
, &em
->flags
);
551 rb
= rb_next(&em
->rb_node
);
554 em
= rb_entry(rb
, struct extent_map
, rb_node
);
556 free_extent_map(em_orig
);
557 spin_unlock(&em_tree
->lock
);
559 ret
= btrfs_drop_extents(trans
, root
, inode
,
560 ordered_extent
->file_offset
,
561 ordered_extent
->file_offset
+
563 ordered_extent
->file_offset
, &alloc_hint
);
565 ret
= btrfs_insert_file_extent(trans
, root
, inode
->i_ino
,
566 ordered_extent
->file_offset
,
567 ordered_extent
->start
,
569 ordered_extent
->len
, 0);
572 btrfs_drop_extent_cache(inode
, ordered_extent
->file_offset
,
573 ordered_extent
->file_offset
+
574 ordered_extent
->len
- 1);
575 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
577 inode
->i_blocks
+= ordered_extent
->len
>> 9;
578 unlock_extent(io_tree
, ordered_extent
->file_offset
,
579 ordered_extent
->file_offset
+ ordered_extent
->len
- 1,
581 add_pending_csums(trans
, inode
, ordered_extent
->file_offset
,
582 &ordered_extent
->list
);
584 btrfs_ordered_update_i_size(inode
, ordered_extent
);
585 btrfs_remove_ordered_extent(inode
, ordered_extent
);
588 btrfs_put_ordered_extent(ordered_extent
);
589 /* once for the tree */
590 btrfs_put_ordered_extent(ordered_extent
);
592 btrfs_update_inode(trans
, root
, inode
);
593 btrfs_end_transaction(trans
, root
);
597 int btrfs_writepage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
598 struct extent_state
*state
, int uptodate
)
600 return btrfs_finish_ordered_io(page
->mapping
->host
, start
, end
);
603 int btrfs_readpage_io_hook(struct page
*page
, u64 start
, u64 end
)
606 struct inode
*inode
= page
->mapping
->host
;
607 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
608 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
609 struct btrfs_csum_item
*item
;
610 struct btrfs_path
*path
= NULL
;
613 if (btrfs_test_opt(root
, NODATASUM
) ||
614 btrfs_test_flag(inode
, NODATASUM
))
618 * It is possible there is an ordered extent that has
619 * not yet finished for this range in the file. If so,
620 * that extent will have a csum cached, and it will insert
621 * the sum after all the blocks in the extent are fully
622 * on disk. So, look for an ordered extent and use the
623 * sum if found. We have to do this before looking in the
624 * btree because csum items are pre-inserted based on
625 * the file size. btrfs_lookup_csum might find an item
626 * that still hasn't been fully filled.
628 ret
= btrfs_find_ordered_sum(inode
, start
, &csum
);
633 path
= btrfs_alloc_path();
634 item
= btrfs_lookup_csum(NULL
, root
, path
, inode
->i_ino
, start
, 0);
637 /* a csum that isn't present is a preallocated region. */
638 if (ret
== -ENOENT
|| ret
== -EFBIG
)
641 printk("no csum found for inode %lu start %Lu\n", inode
->i_ino
,
645 read_extent_buffer(path
->nodes
[0], &csum
, (unsigned long)item
,
648 set_state_private(io_tree
, start
, csum
);
651 btrfs_free_path(path
);
655 struct io_failure_record
{
663 int btrfs_io_failed_hook(struct bio
*failed_bio
,
664 struct page
*page
, u64 start
, u64 end
,
665 struct extent_state
*state
)
667 struct io_failure_record
*failrec
= NULL
;
669 struct extent_map
*em
;
670 struct inode
*inode
= page
->mapping
->host
;
671 struct extent_io_tree
*failure_tree
= &BTRFS_I(inode
)->io_failure_tree
;
672 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
679 ret
= get_state_private(failure_tree
, start
, &private);
681 failrec
= kmalloc(sizeof(*failrec
), GFP_NOFS
);
684 failrec
->start
= start
;
685 failrec
->len
= end
- start
+ 1;
686 failrec
->last_mirror
= 0;
688 spin_lock(&em_tree
->lock
);
689 em
= lookup_extent_mapping(em_tree
, start
, failrec
->len
);
690 if (em
->start
> start
|| em
->start
+ em
->len
< start
) {
694 spin_unlock(&em_tree
->lock
);
696 if (!em
|| IS_ERR(em
)) {
700 logical
= start
- em
->start
;
701 logical
= em
->block_start
+ logical
;
702 failrec
->logical
= logical
;
704 set_extent_bits(failure_tree
, start
, end
, EXTENT_LOCKED
|
705 EXTENT_DIRTY
, GFP_NOFS
);
706 set_state_private(failure_tree
, start
,
707 (u64
)(unsigned long)failrec
);
709 failrec
= (struct io_failure_record
*)(unsigned long)private;
711 num_copies
= btrfs_num_copies(
712 &BTRFS_I(inode
)->root
->fs_info
->mapping_tree
,
713 failrec
->logical
, failrec
->len
);
714 failrec
->last_mirror
++;
716 spin_lock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
717 state
= find_first_extent_bit_state(&BTRFS_I(inode
)->io_tree
,
720 if (state
&& state
->start
!= failrec
->start
)
722 spin_unlock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
724 if (!state
|| failrec
->last_mirror
> num_copies
) {
725 set_state_private(failure_tree
, failrec
->start
, 0);
726 clear_extent_bits(failure_tree
, failrec
->start
,
727 failrec
->start
+ failrec
->len
- 1,
728 EXTENT_LOCKED
| EXTENT_DIRTY
, GFP_NOFS
);
732 bio
= bio_alloc(GFP_NOFS
, 1);
733 bio
->bi_private
= state
;
734 bio
->bi_end_io
= failed_bio
->bi_end_io
;
735 bio
->bi_sector
= failrec
->logical
>> 9;
736 bio
->bi_bdev
= failed_bio
->bi_bdev
;
738 bio_add_page(bio
, page
, failrec
->len
, start
- page_offset(page
));
739 if (failed_bio
->bi_rw
& (1 << BIO_RW
))
744 BTRFS_I(inode
)->io_tree
.ops
->submit_bio_hook(inode
, rw
, bio
,
745 failrec
->last_mirror
);
749 int btrfs_clean_io_failures(struct inode
*inode
, u64 start
)
753 struct io_failure_record
*failure
;
757 if (count_range_bits(&BTRFS_I(inode
)->io_failure_tree
, &private,
758 (u64
)-1, 1, EXTENT_DIRTY
)) {
759 ret
= get_state_private(&BTRFS_I(inode
)->io_failure_tree
,
760 start
, &private_failure
);
762 failure
= (struct io_failure_record
*)(unsigned long)
764 set_state_private(&BTRFS_I(inode
)->io_failure_tree
,
766 clear_extent_bits(&BTRFS_I(inode
)->io_failure_tree
,
768 failure
->start
+ failure
->len
- 1,
769 EXTENT_DIRTY
| EXTENT_LOCKED
,
777 int btrfs_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
778 struct extent_state
*state
)
780 size_t offset
= start
- ((u64
)page
->index
<< PAGE_CACHE_SHIFT
);
781 struct inode
*inode
= page
->mapping
->host
;
782 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
784 u64
private = ~(u32
)0;
786 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
790 if (btrfs_test_opt(root
, NODATASUM
) ||
791 btrfs_test_flag(inode
, NODATASUM
))
793 if (state
&& state
->start
== start
) {
794 private = state
->private;
797 ret
= get_state_private(io_tree
, start
, &private);
799 local_irq_save(flags
);
800 kaddr
= kmap_atomic(page
, KM_IRQ0
);
804 csum
= btrfs_csum_data(root
, kaddr
+ offset
, csum
, end
- start
+ 1);
805 btrfs_csum_final(csum
, (char *)&csum
);
806 if (csum
!= private) {
809 kunmap_atomic(kaddr
, KM_IRQ0
);
810 local_irq_restore(flags
);
812 /* if the io failure tree for this inode is non-empty,
813 * check to see if we've recovered from a failed IO
815 btrfs_clean_io_failures(inode
, start
);
819 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
820 page
->mapping
->host
->i_ino
, (unsigned long long)start
, csum
,
822 memset(kaddr
+ offset
, 1, end
- start
+ 1);
823 flush_dcache_page(page
);
824 kunmap_atomic(kaddr
, KM_IRQ0
);
825 local_irq_restore(flags
);
832 * This creates an orphan entry for the given inode in case something goes
833 * wrong in the middle of an unlink/truncate.
835 int btrfs_orphan_add(struct btrfs_trans_handle
*trans
, struct inode
*inode
)
837 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
840 spin_lock(&root
->orphan_lock
);
842 /* already on the orphan list, we're good */
843 if (!list_empty(&BTRFS_I(inode
)->i_orphan
)) {
844 spin_unlock(&root
->orphan_lock
);
848 list_add(&BTRFS_I(inode
)->i_orphan
, &root
->orphan_list
);
850 spin_unlock(&root
->orphan_lock
);
853 * insert an orphan item to track this unlinked/truncated file
855 ret
= btrfs_insert_orphan_item(trans
, root
, inode
->i_ino
);
861 * We have done the truncate/delete so we can go ahead and remove the orphan
862 * item for this particular inode.
864 int btrfs_orphan_del(struct btrfs_trans_handle
*trans
, struct inode
*inode
)
866 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
869 spin_lock(&root
->orphan_lock
);
871 if (list_empty(&BTRFS_I(inode
)->i_orphan
)) {
872 spin_unlock(&root
->orphan_lock
);
876 list_del_init(&BTRFS_I(inode
)->i_orphan
);
878 spin_unlock(&root
->orphan_lock
);
882 spin_unlock(&root
->orphan_lock
);
884 ret
= btrfs_del_orphan_item(trans
, root
, inode
->i_ino
);
890 * this cleans up any orphans that may be left on the list from the last use
893 void btrfs_orphan_cleanup(struct btrfs_root
*root
)
895 struct btrfs_path
*path
;
896 struct extent_buffer
*leaf
;
897 struct btrfs_item
*item
;
898 struct btrfs_key key
, found_key
;
899 struct btrfs_trans_handle
*trans
;
901 int ret
= 0, nr_unlink
= 0, nr_truncate
= 0;
903 /* don't do orphan cleanup if the fs is readonly. */
904 if (root
->inode
->i_sb
->s_flags
& MS_RDONLY
)
907 path
= btrfs_alloc_path();
912 key
.objectid
= BTRFS_ORPHAN_OBJECTID
;
913 btrfs_set_key_type(&key
, BTRFS_ORPHAN_ITEM_KEY
);
914 key
.offset
= (u64
)-1;
916 trans
= btrfs_start_transaction(root
, 1);
917 btrfs_set_trans_block_group(trans
, root
->inode
);
920 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
922 printk(KERN_ERR
"Error searching slot for orphan: %d"
928 * if ret == 0 means we found what we were searching for, which
929 * is weird, but possible, so only screw with path if we didnt
930 * find the key and see if we have stuff that matches
933 if (path
->slots
[0] == 0)
938 /* pull out the item */
939 leaf
= path
->nodes
[0];
940 item
= btrfs_item_nr(leaf
, path
->slots
[0]);
941 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
943 /* make sure the item matches what we want */
944 if (found_key
.objectid
!= BTRFS_ORPHAN_OBJECTID
)
946 if (btrfs_key_type(&found_key
) != BTRFS_ORPHAN_ITEM_KEY
)
949 /* release the path since we're done with it */
950 btrfs_release_path(root
, path
);
953 * this is where we are basically btrfs_lookup, without the
954 * crossing root thing. we store the inode number in the
955 * offset of the orphan item.
957 inode
= btrfs_iget_locked(root
->inode
->i_sb
,
958 found_key
.offset
, root
);
962 if (inode
->i_state
& I_NEW
) {
963 BTRFS_I(inode
)->root
= root
;
965 /* have to set the location manually */
966 BTRFS_I(inode
)->location
.objectid
= inode
->i_ino
;
967 BTRFS_I(inode
)->location
.type
= BTRFS_INODE_ITEM_KEY
;
968 BTRFS_I(inode
)->location
.offset
= 0;
970 btrfs_read_locked_inode(inode
);
971 unlock_new_inode(inode
);
975 * add this inode to the orphan list so btrfs_orphan_del does
976 * the proper thing when we hit it
978 spin_lock(&root
->orphan_lock
);
979 list_add(&BTRFS_I(inode
)->i_orphan
, &root
->orphan_list
);
980 spin_unlock(&root
->orphan_lock
);
983 * if this is a bad inode, means we actually succeeded in
984 * removing the inode, but not the orphan record, which means
985 * we need to manually delete the orphan since iput will just
988 if (is_bad_inode(inode
)) {
989 btrfs_orphan_del(trans
, inode
);
994 /* if we have links, this was a truncate, lets do that */
995 if (inode
->i_nlink
) {
997 btrfs_truncate(inode
);
1002 /* this will do delete_inode and everything for us */
1007 printk(KERN_INFO
"btrfs: unlinked %d orphans\n", nr_unlink
);
1009 printk(KERN_INFO
"btrfs: truncated %d orphans\n", nr_truncate
);
1011 btrfs_free_path(path
);
1012 btrfs_end_transaction(trans
, root
);
1015 void btrfs_read_locked_inode(struct inode
*inode
)
1017 struct btrfs_path
*path
;
1018 struct extent_buffer
*leaf
;
1019 struct btrfs_inode_item
*inode_item
;
1020 struct btrfs_timespec
*tspec
;
1021 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1022 struct btrfs_key location
;
1023 u64 alloc_group_block
;
1027 path
= btrfs_alloc_path();
1029 memcpy(&location
, &BTRFS_I(inode
)->location
, sizeof(location
));
1031 ret
= btrfs_lookup_inode(NULL
, root
, path
, &location
, 0);
1035 leaf
= path
->nodes
[0];
1036 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
1037 struct btrfs_inode_item
);
1039 inode
->i_mode
= btrfs_inode_mode(leaf
, inode_item
);
1040 inode
->i_nlink
= btrfs_inode_nlink(leaf
, inode_item
);
1041 inode
->i_uid
= btrfs_inode_uid(leaf
, inode_item
);
1042 inode
->i_gid
= btrfs_inode_gid(leaf
, inode_item
);
1043 btrfs_i_size_write(inode
, btrfs_inode_size(leaf
, inode_item
));
1045 tspec
= btrfs_inode_atime(inode_item
);
1046 inode
->i_atime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1047 inode
->i_atime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1049 tspec
= btrfs_inode_mtime(inode_item
);
1050 inode
->i_mtime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1051 inode
->i_mtime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1053 tspec
= btrfs_inode_ctime(inode_item
);
1054 inode
->i_ctime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1055 inode
->i_ctime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1057 inode
->i_blocks
= btrfs_inode_nblocks(leaf
, inode_item
);
1058 inode
->i_generation
= btrfs_inode_generation(leaf
, inode_item
);
1060 rdev
= btrfs_inode_rdev(leaf
, inode_item
);
1062 BTRFS_I(inode
)->index_cnt
= (u64
)-1;
1064 alloc_group_block
= btrfs_inode_block_group(leaf
, inode_item
);
1065 BTRFS_I(inode
)->block_group
= btrfs_lookup_block_group(root
->fs_info
,
1067 BTRFS_I(inode
)->flags
= btrfs_inode_flags(leaf
, inode_item
);
1068 if (!BTRFS_I(inode
)->block_group
) {
1069 BTRFS_I(inode
)->block_group
= btrfs_find_block_group(root
,
1071 BTRFS_BLOCK_GROUP_METADATA
, 0);
1073 btrfs_free_path(path
);
1076 switch (inode
->i_mode
& S_IFMT
) {
1078 inode
->i_mapping
->a_ops
= &btrfs_aops
;
1079 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1080 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
1081 inode
->i_fop
= &btrfs_file_operations
;
1082 inode
->i_op
= &btrfs_file_inode_operations
;
1085 inode
->i_fop
= &btrfs_dir_file_operations
;
1086 if (root
== root
->fs_info
->tree_root
)
1087 inode
->i_op
= &btrfs_dir_ro_inode_operations
;
1089 inode
->i_op
= &btrfs_dir_inode_operations
;
1092 inode
->i_op
= &btrfs_symlink_inode_operations
;
1093 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
1094 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1097 init_special_inode(inode
, inode
->i_mode
, rdev
);
1103 btrfs_free_path(path
);
1104 make_bad_inode(inode
);
1107 static void fill_inode_item(struct extent_buffer
*leaf
,
1108 struct btrfs_inode_item
*item
,
1109 struct inode
*inode
)
1111 btrfs_set_inode_uid(leaf
, item
, inode
->i_uid
);
1112 btrfs_set_inode_gid(leaf
, item
, inode
->i_gid
);
1113 btrfs_set_inode_size(leaf
, item
, BTRFS_I(inode
)->disk_i_size
);
1114 btrfs_set_inode_mode(leaf
, item
, inode
->i_mode
);
1115 btrfs_set_inode_nlink(leaf
, item
, inode
->i_nlink
);
1117 btrfs_set_timespec_sec(leaf
, btrfs_inode_atime(item
),
1118 inode
->i_atime
.tv_sec
);
1119 btrfs_set_timespec_nsec(leaf
, btrfs_inode_atime(item
),
1120 inode
->i_atime
.tv_nsec
);
1122 btrfs_set_timespec_sec(leaf
, btrfs_inode_mtime(item
),
1123 inode
->i_mtime
.tv_sec
);
1124 btrfs_set_timespec_nsec(leaf
, btrfs_inode_mtime(item
),
1125 inode
->i_mtime
.tv_nsec
);
1127 btrfs_set_timespec_sec(leaf
, btrfs_inode_ctime(item
),
1128 inode
->i_ctime
.tv_sec
);
1129 btrfs_set_timespec_nsec(leaf
, btrfs_inode_ctime(item
),
1130 inode
->i_ctime
.tv_nsec
);
1132 btrfs_set_inode_nblocks(leaf
, item
, inode
->i_blocks
);
1133 btrfs_set_inode_generation(leaf
, item
, inode
->i_generation
);
1134 btrfs_set_inode_rdev(leaf
, item
, inode
->i_rdev
);
1135 btrfs_set_inode_flags(leaf
, item
, BTRFS_I(inode
)->flags
);
1136 btrfs_set_inode_block_group(leaf
, item
,
1137 BTRFS_I(inode
)->block_group
->key
.objectid
);
1140 int noinline
btrfs_update_inode(struct btrfs_trans_handle
*trans
,
1141 struct btrfs_root
*root
,
1142 struct inode
*inode
)
1144 struct btrfs_inode_item
*inode_item
;
1145 struct btrfs_path
*path
;
1146 struct extent_buffer
*leaf
;
1149 path
= btrfs_alloc_path();
1151 ret
= btrfs_lookup_inode(trans
, root
, path
,
1152 &BTRFS_I(inode
)->location
, 1);
1159 leaf
= path
->nodes
[0];
1160 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
1161 struct btrfs_inode_item
);
1163 fill_inode_item(leaf
, inode_item
, inode
);
1164 btrfs_mark_buffer_dirty(leaf
);
1165 btrfs_set_inode_last_trans(trans
, inode
);
1168 btrfs_free_path(path
);
1173 static int btrfs_unlink_trans(struct btrfs_trans_handle
*trans
,
1174 struct btrfs_root
*root
,
1176 struct dentry
*dentry
)
1178 struct btrfs_path
*path
;
1179 const char *name
= dentry
->d_name
.name
;
1180 int name_len
= dentry
->d_name
.len
;
1182 struct extent_buffer
*leaf
;
1183 struct btrfs_dir_item
*di
;
1184 struct btrfs_key key
;
1187 path
= btrfs_alloc_path();
1193 di
= btrfs_lookup_dir_item(trans
, root
, path
, dir
->i_ino
,
1194 name
, name_len
, -1);
1203 leaf
= path
->nodes
[0];
1204 btrfs_dir_item_key_to_cpu(leaf
, di
, &key
);
1205 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
1208 btrfs_release_path(root
, path
);
1210 ret
= btrfs_del_inode_ref(trans
, root
, name
, name_len
,
1211 dentry
->d_inode
->i_ino
,
1212 dentry
->d_parent
->d_inode
->i_ino
, &index
);
1214 printk("failed to delete reference to %.*s, "
1215 "inode %lu parent %lu\n", name_len
, name
,
1216 dentry
->d_inode
->i_ino
,
1217 dentry
->d_parent
->d_inode
->i_ino
);
1221 di
= btrfs_lookup_dir_index_item(trans
, root
, path
, dir
->i_ino
,
1222 index
, name
, name_len
, -1);
1231 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
1232 btrfs_release_path(root
, path
);
1234 dentry
->d_inode
->i_ctime
= dir
->i_ctime
;
1236 btrfs_free_path(path
);
1238 btrfs_i_size_write(dir
, dir
->i_size
- name_len
* 2);
1239 dir
->i_mtime
= dir
->i_ctime
= CURRENT_TIME
;
1240 btrfs_update_inode(trans
, root
, dir
);
1241 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1242 dentry
->d_inode
->i_nlink
--;
1244 drop_nlink(dentry
->d_inode
);
1246 ret
= btrfs_update_inode(trans
, root
, dentry
->d_inode
);
1247 dir
->i_sb
->s_dirt
= 1;
1252 static int btrfs_unlink(struct inode
*dir
, struct dentry
*dentry
)
1254 struct btrfs_root
*root
;
1255 struct btrfs_trans_handle
*trans
;
1256 struct inode
*inode
= dentry
->d_inode
;
1258 unsigned long nr
= 0;
1260 root
= BTRFS_I(dir
)->root
;
1262 ret
= btrfs_check_free_space(root
, 1, 1);
1266 trans
= btrfs_start_transaction(root
, 1);
1268 btrfs_set_trans_block_group(trans
, dir
);
1269 ret
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
1271 if (inode
->i_nlink
== 0)
1272 ret
= btrfs_orphan_add(trans
, inode
);
1274 nr
= trans
->blocks_used
;
1276 btrfs_end_transaction_throttle(trans
, root
);
1278 btrfs_btree_balance_dirty(root
, nr
);
1282 static int btrfs_rmdir(struct inode
*dir
, struct dentry
*dentry
)
1284 struct inode
*inode
= dentry
->d_inode
;
1287 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1288 struct btrfs_trans_handle
*trans
;
1289 unsigned long nr
= 0;
1291 if (inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
1295 ret
= btrfs_check_free_space(root
, 1, 1);
1299 trans
= btrfs_start_transaction(root
, 1);
1300 btrfs_set_trans_block_group(trans
, dir
);
1302 err
= btrfs_orphan_add(trans
, inode
);
1306 /* now the directory is empty */
1307 err
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
1309 btrfs_i_size_write(inode
, 0);
1313 nr
= trans
->blocks_used
;
1314 ret
= btrfs_end_transaction_throttle(trans
, root
);
1316 btrfs_btree_balance_dirty(root
, nr
);
1324 * this can truncate away extent items, csum items and directory items.
1325 * It starts at a high offset and removes keys until it can't find
1326 * any higher than i_size.
1328 * csum items that cross the new i_size are truncated to the new size
1331 * min_type is the minimum key type to truncate down to. If set to 0, this
1332 * will kill all the items on this inode, including the INODE_ITEM_KEY.
1334 static int btrfs_truncate_in_trans(struct btrfs_trans_handle
*trans
,
1335 struct btrfs_root
*root
,
1336 struct inode
*inode
,
1340 struct btrfs_path
*path
;
1341 struct btrfs_key key
;
1342 struct btrfs_key found_key
;
1344 struct extent_buffer
*leaf
;
1345 struct btrfs_file_extent_item
*fi
;
1346 u64 extent_start
= 0;
1347 u64 extent_num_bytes
= 0;
1353 int pending_del_nr
= 0;
1354 int pending_del_slot
= 0;
1355 int extent_type
= -1;
1356 u64 mask
= root
->sectorsize
- 1;
1358 btrfs_drop_extent_cache(inode
, inode
->i_size
& (~mask
), (u64
)-1);
1359 path
= btrfs_alloc_path();
1363 /* FIXME, add redo link to tree so we don't leak on crash */
1364 key
.objectid
= inode
->i_ino
;
1365 key
.offset
= (u64
)-1;
1368 btrfs_init_path(path
);
1370 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
1375 BUG_ON(path
->slots
[0] == 0);
1381 leaf
= path
->nodes
[0];
1382 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
1383 found_type
= btrfs_key_type(&found_key
);
1385 if (found_key
.objectid
!= inode
->i_ino
)
1388 if (found_type
< min_type
)
1391 item_end
= found_key
.offset
;
1392 if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
1393 fi
= btrfs_item_ptr(leaf
, path
->slots
[0],
1394 struct btrfs_file_extent_item
);
1395 extent_type
= btrfs_file_extent_type(leaf
, fi
);
1396 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1398 btrfs_file_extent_num_bytes(leaf
, fi
);
1399 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1400 struct btrfs_item
*item
= btrfs_item_nr(leaf
,
1402 item_end
+= btrfs_file_extent_inline_len(leaf
,
1407 if (found_type
== BTRFS_CSUM_ITEM_KEY
) {
1408 ret
= btrfs_csum_truncate(trans
, root
, path
,
1412 if (item_end
< inode
->i_size
) {
1413 if (found_type
== BTRFS_DIR_ITEM_KEY
) {
1414 found_type
= BTRFS_INODE_ITEM_KEY
;
1415 } else if (found_type
== BTRFS_EXTENT_ITEM_KEY
) {
1416 found_type
= BTRFS_CSUM_ITEM_KEY
;
1417 } else if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
1418 found_type
= BTRFS_XATTR_ITEM_KEY
;
1419 } else if (found_type
== BTRFS_XATTR_ITEM_KEY
) {
1420 found_type
= BTRFS_INODE_REF_KEY
;
1421 } else if (found_type
) {
1426 btrfs_set_key_type(&key
, found_type
);
1429 if (found_key
.offset
>= inode
->i_size
)
1435 /* FIXME, shrink the extent if the ref count is only 1 */
1436 if (found_type
!= BTRFS_EXTENT_DATA_KEY
)
1439 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1441 extent_start
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
1443 u64 orig_num_bytes
=
1444 btrfs_file_extent_num_bytes(leaf
, fi
);
1445 extent_num_bytes
= inode
->i_size
-
1446 found_key
.offset
+ root
->sectorsize
- 1;
1447 extent_num_bytes
= extent_num_bytes
&
1448 ~((u64
)root
->sectorsize
- 1);
1449 btrfs_set_file_extent_num_bytes(leaf
, fi
,
1451 num_dec
= (orig_num_bytes
-
1453 if (extent_start
!= 0)
1454 dec_i_blocks(inode
, num_dec
);
1455 btrfs_mark_buffer_dirty(leaf
);
1458 btrfs_file_extent_disk_num_bytes(leaf
,
1460 /* FIXME blocksize != 4096 */
1461 num_dec
= btrfs_file_extent_num_bytes(leaf
, fi
);
1462 if (extent_start
!= 0) {
1464 dec_i_blocks(inode
, num_dec
);
1466 root_gen
= btrfs_header_generation(leaf
);
1467 root_owner
= btrfs_header_owner(leaf
);
1469 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1471 u32 newsize
= inode
->i_size
- found_key
.offset
;
1472 dec_i_blocks(inode
, item_end
+ 1 -
1473 found_key
.offset
- newsize
);
1475 btrfs_file_extent_calc_inline_size(newsize
);
1476 ret
= btrfs_truncate_item(trans
, root
, path
,
1480 dec_i_blocks(inode
, item_end
+ 1 -
1486 if (!pending_del_nr
) {
1487 /* no pending yet, add ourselves */
1488 pending_del_slot
= path
->slots
[0];
1490 } else if (pending_del_nr
&&
1491 path
->slots
[0] + 1 == pending_del_slot
) {
1492 /* hop on the pending chunk */
1494 pending_del_slot
= path
->slots
[0];
1496 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path
->slots
[0], pending_del_nr
, pending_del_slot
);
1502 ret
= btrfs_free_extent(trans
, root
, extent_start
,
1505 root_gen
, inode
->i_ino
,
1506 found_key
.offset
, 0);
1510 if (path
->slots
[0] == 0) {
1513 btrfs_release_path(root
, path
);
1518 if (pending_del_nr
&&
1519 path
->slots
[0] + 1 != pending_del_slot
) {
1520 struct btrfs_key debug
;
1522 btrfs_item_key_to_cpu(path
->nodes
[0], &debug
,
1524 ret
= btrfs_del_items(trans
, root
, path
,
1529 btrfs_release_path(root
, path
);
1535 if (pending_del_nr
) {
1536 ret
= btrfs_del_items(trans
, root
, path
, pending_del_slot
,
1539 btrfs_free_path(path
);
1540 inode
->i_sb
->s_dirt
= 1;
1545 * taken from block_truncate_page, but does cow as it zeros out
1546 * any bytes left in the last page in the file.
1548 static int btrfs_truncate_page(struct address_space
*mapping
, loff_t from
)
1550 struct inode
*inode
= mapping
->host
;
1551 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1552 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1553 struct btrfs_ordered_extent
*ordered
;
1555 u32 blocksize
= root
->sectorsize
;
1556 pgoff_t index
= from
>> PAGE_CACHE_SHIFT
;
1557 unsigned offset
= from
& (PAGE_CACHE_SIZE
-1);
1563 if ((offset
& (blocksize
- 1)) == 0)
1568 page
= grab_cache_page(mapping
, index
);
1572 page_start
= page_offset(page
);
1573 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1575 if (!PageUptodate(page
)) {
1576 ret
= btrfs_readpage(NULL
, page
);
1578 if (page
->mapping
!= mapping
) {
1580 page_cache_release(page
);
1583 if (!PageUptodate(page
)) {
1588 wait_on_page_writeback(page
);
1590 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1591 set_page_extent_mapped(page
);
1593 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
1595 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1597 page_cache_release(page
);
1598 btrfs_start_ordered_extent(inode
, ordered
, 1);
1599 btrfs_put_ordered_extent(ordered
);
1603 set_extent_delalloc(&BTRFS_I(inode
)->io_tree
, page_start
,
1604 page_end
, GFP_NOFS
);
1606 if (offset
!= PAGE_CACHE_SIZE
) {
1608 memset(kaddr
+ offset
, 0, PAGE_CACHE_SIZE
- offset
);
1609 flush_dcache_page(page
);
1612 ClearPageChecked(page
);
1613 set_page_dirty(page
);
1614 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1618 page_cache_release(page
);
1623 static int btrfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
1625 struct inode
*inode
= dentry
->d_inode
;
1628 err
= inode_change_ok(inode
, attr
);
1632 if (S_ISREG(inode
->i_mode
) &&
1633 attr
->ia_valid
& ATTR_SIZE
&& attr
->ia_size
> inode
->i_size
) {
1634 struct btrfs_trans_handle
*trans
;
1635 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1636 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1638 u64 mask
= root
->sectorsize
- 1;
1639 u64 hole_start
= (inode
->i_size
+ mask
) & ~mask
;
1640 u64 block_end
= (attr
->ia_size
+ mask
) & ~mask
;
1644 if (attr
->ia_size
<= hole_start
)
1647 err
= btrfs_check_free_space(root
, 1, 0);
1651 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
1653 hole_size
= block_end
- hole_start
;
1654 btrfs_wait_ordered_range(inode
, hole_start
, hole_size
);
1655 lock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1657 trans
= btrfs_start_transaction(root
, 1);
1658 btrfs_set_trans_block_group(trans
, inode
);
1659 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
1660 err
= btrfs_drop_extents(trans
, root
, inode
,
1661 hole_start
, block_end
, hole_start
,
1664 if (alloc_hint
!= EXTENT_MAP_INLINE
) {
1665 err
= btrfs_insert_file_extent(trans
, root
,
1669 btrfs_drop_extent_cache(inode
, hole_start
,
1671 btrfs_check_file(root
, inode
);
1673 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
1674 btrfs_end_transaction(trans
, root
);
1675 unlock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1680 err
= inode_setattr(inode
, attr
);
1682 if (!err
&& ((attr
->ia_valid
& ATTR_MODE
)))
1683 err
= btrfs_acl_chmod(inode
);
1688 void btrfs_delete_inode(struct inode
*inode
)
1690 struct btrfs_trans_handle
*trans
;
1691 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1695 truncate_inode_pages(&inode
->i_data
, 0);
1696 if (is_bad_inode(inode
)) {
1697 btrfs_orphan_del(NULL
, inode
);
1700 btrfs_wait_ordered_range(inode
, 0, (u64
)-1);
1702 btrfs_i_size_write(inode
, 0);
1703 trans
= btrfs_start_transaction(root
, 1);
1705 btrfs_set_trans_block_group(trans
, inode
);
1706 ret
= btrfs_truncate_in_trans(trans
, root
, inode
, 0);
1708 btrfs_orphan_del(NULL
, inode
);
1709 goto no_delete_lock
;
1712 btrfs_orphan_del(trans
, inode
);
1714 nr
= trans
->blocks_used
;
1717 btrfs_end_transaction(trans
, root
);
1718 btrfs_btree_balance_dirty(root
, nr
);
1722 nr
= trans
->blocks_used
;
1723 btrfs_end_transaction(trans
, root
);
1724 btrfs_btree_balance_dirty(root
, nr
);
1730 * this returns the key found in the dir entry in the location pointer.
1731 * If no dir entries were found, location->objectid is 0.
1733 static int btrfs_inode_by_name(struct inode
*dir
, struct dentry
*dentry
,
1734 struct btrfs_key
*location
)
1736 const char *name
= dentry
->d_name
.name
;
1737 int namelen
= dentry
->d_name
.len
;
1738 struct btrfs_dir_item
*di
;
1739 struct btrfs_path
*path
;
1740 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1743 if (namelen
== 1 && strcmp(name
, ".") == 0) {
1744 location
->objectid
= dir
->i_ino
;
1745 location
->type
= BTRFS_INODE_ITEM_KEY
;
1746 location
->offset
= 0;
1749 path
= btrfs_alloc_path();
1752 if (namelen
== 2 && strcmp(name
, "..") == 0) {
1753 struct btrfs_key key
;
1754 struct extent_buffer
*leaf
;
1757 key
.objectid
= dir
->i_ino
;
1758 key
.offset
= (u64
)-1;
1759 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1760 if (ret
< 0 || path
->slots
[0] == 0)
1762 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1765 leaf
= path
->nodes
[0];
1766 slot
= path
->slots
[0] - 1;
1768 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
1769 if (key
.objectid
!= dir
->i_ino
||
1770 key
.type
!= BTRFS_INODE_REF_KEY
) {
1773 location
->objectid
= key
.offset
;
1774 location
->type
= BTRFS_INODE_ITEM_KEY
;
1775 location
->offset
= 0;
1779 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir
->i_ino
, name
,
1783 if (!di
|| IS_ERR(di
)) {
1786 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, location
);
1788 btrfs_free_path(path
);
1791 location
->objectid
= 0;
1796 * when we hit a tree root in a directory, the btrfs part of the inode
1797 * needs to be changed to reflect the root directory of the tree root. This
1798 * is kind of like crossing a mount point.
1800 static int fixup_tree_root_location(struct btrfs_root
*root
,
1801 struct btrfs_key
*location
,
1802 struct btrfs_root
**sub_root
,
1803 struct dentry
*dentry
)
1805 struct btrfs_root_item
*ri
;
1807 if (btrfs_key_type(location
) != BTRFS_ROOT_ITEM_KEY
)
1809 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1812 *sub_root
= btrfs_read_fs_root(root
->fs_info
, location
,
1813 dentry
->d_name
.name
,
1814 dentry
->d_name
.len
);
1815 if (IS_ERR(*sub_root
))
1816 return PTR_ERR(*sub_root
);
1818 ri
= &(*sub_root
)->root_item
;
1819 location
->objectid
= btrfs_root_dirid(ri
);
1820 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1821 location
->offset
= 0;
1826 static int btrfs_init_locked_inode(struct inode
*inode
, void *p
)
1828 struct btrfs_iget_args
*args
= p
;
1829 inode
->i_ino
= args
->ino
;
1830 BTRFS_I(inode
)->root
= args
->root
;
1831 BTRFS_I(inode
)->delalloc_bytes
= 0;
1832 BTRFS_I(inode
)->disk_i_size
= 0;
1833 BTRFS_I(inode
)->index_cnt
= (u64
)-1;
1834 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1835 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1836 inode
->i_mapping
, GFP_NOFS
);
1837 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1838 inode
->i_mapping
, GFP_NOFS
);
1839 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
1840 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
1841 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
1845 static int btrfs_find_actor(struct inode
*inode
, void *opaque
)
1847 struct btrfs_iget_args
*args
= opaque
;
1848 return (args
->ino
== inode
->i_ino
&&
1849 args
->root
== BTRFS_I(inode
)->root
);
1852 struct inode
*btrfs_ilookup(struct super_block
*s
, u64 objectid
,
1855 struct btrfs_iget_args args
;
1856 args
.ino
= objectid
;
1857 args
.root
= btrfs_lookup_fs_root(btrfs_sb(s
)->fs_info
, root_objectid
);
1862 return ilookup5(s
, objectid
, btrfs_find_actor
, (void *)&args
);
1865 struct inode
*btrfs_iget_locked(struct super_block
*s
, u64 objectid
,
1866 struct btrfs_root
*root
)
1868 struct inode
*inode
;
1869 struct btrfs_iget_args args
;
1870 args
.ino
= objectid
;
1873 inode
= iget5_locked(s
, objectid
, btrfs_find_actor
,
1874 btrfs_init_locked_inode
,
1879 static struct dentry
*btrfs_lookup(struct inode
*dir
, struct dentry
*dentry
,
1880 struct nameidata
*nd
)
1882 struct inode
* inode
;
1883 struct btrfs_inode
*bi
= BTRFS_I(dir
);
1884 struct btrfs_root
*root
= bi
->root
;
1885 struct btrfs_root
*sub_root
= root
;
1886 struct btrfs_key location
;
1887 int ret
, do_orphan
= 0;
1889 if (dentry
->d_name
.len
> BTRFS_NAME_LEN
)
1890 return ERR_PTR(-ENAMETOOLONG
);
1892 ret
= btrfs_inode_by_name(dir
, dentry
, &location
);
1895 return ERR_PTR(ret
);
1898 if (location
.objectid
) {
1899 ret
= fixup_tree_root_location(root
, &location
, &sub_root
,
1902 return ERR_PTR(ret
);
1904 return ERR_PTR(-ENOENT
);
1906 inode
= btrfs_iget_locked(dir
->i_sb
, location
.objectid
,
1909 return ERR_PTR(-EACCES
);
1910 if (inode
->i_state
& I_NEW
) {
1911 /* the inode and parent dir are two different roots */
1912 if (sub_root
!= root
) {
1914 sub_root
->inode
= inode
;
1917 BTRFS_I(inode
)->root
= sub_root
;
1918 memcpy(&BTRFS_I(inode
)->location
, &location
,
1920 btrfs_read_locked_inode(inode
);
1921 unlock_new_inode(inode
);
1925 if (unlikely(do_orphan
))
1926 btrfs_orphan_cleanup(sub_root
);
1928 return d_splice_alias(inode
, dentry
);
1931 static unsigned char btrfs_filetype_table
[] = {
1932 DT_UNKNOWN
, DT_REG
, DT_DIR
, DT_CHR
, DT_BLK
, DT_FIFO
, DT_SOCK
, DT_LNK
1935 static int btrfs_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
1937 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1938 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1939 struct btrfs_item
*item
;
1940 struct btrfs_dir_item
*di
;
1941 struct btrfs_key key
;
1942 struct btrfs_key found_key
;
1943 struct btrfs_path
*path
;
1946 struct extent_buffer
*leaf
;
1949 unsigned char d_type
;
1954 int key_type
= BTRFS_DIR_INDEX_KEY
;
1959 /* FIXME, use a real flag for deciding about the key type */
1960 if (root
->fs_info
->tree_root
== root
)
1961 key_type
= BTRFS_DIR_ITEM_KEY
;
1963 /* special case for "." */
1964 if (filp
->f_pos
== 0) {
1965 over
= filldir(dirent
, ".", 1,
1973 key
.objectid
= inode
->i_ino
;
1974 path
= btrfs_alloc_path();
1977 /* special case for .., just use the back ref */
1978 if (filp
->f_pos
== 1) {
1979 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1980 key
.offset
= (u64
)-1;
1981 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1982 if (ret
< 0 || path
->slots
[0] == 0) {
1983 btrfs_release_path(root
, path
);
1984 goto read_dir_items
;
1987 leaf
= path
->nodes
[0];
1988 slot
= path
->slots
[0] - 1;
1989 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1990 btrfs_release_path(root
, path
);
1991 if (found_key
.objectid
!= key
.objectid
||
1992 found_key
.type
!= BTRFS_INODE_REF_KEY
)
1993 goto read_dir_items
;
1994 over
= filldir(dirent
, "..", 2,
1995 2, found_key
.offset
, DT_DIR
);
2002 btrfs_set_key_type(&key
, key_type
);
2003 key
.offset
= filp
->f_pos
;
2005 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2010 leaf
= path
->nodes
[0];
2011 nritems
= btrfs_header_nritems(leaf
);
2012 slot
= path
->slots
[0];
2013 if (advance
|| slot
>= nritems
) {
2014 if (slot
>= nritems
-1) {
2015 ret
= btrfs_next_leaf(root
, path
);
2018 leaf
= path
->nodes
[0];
2019 nritems
= btrfs_header_nritems(leaf
);
2020 slot
= path
->slots
[0];
2027 item
= btrfs_item_nr(leaf
, slot
);
2028 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
2030 if (found_key
.objectid
!= key
.objectid
)
2032 if (btrfs_key_type(&found_key
) != key_type
)
2034 if (found_key
.offset
< filp
->f_pos
)
2037 filp
->f_pos
= found_key
.offset
;
2039 di
= btrfs_item_ptr(leaf
, slot
, struct btrfs_dir_item
);
2041 di_total
= btrfs_item_size(leaf
, item
);
2042 while(di_cur
< di_total
) {
2043 struct btrfs_key location
;
2045 name_len
= btrfs_dir_name_len(leaf
, di
);
2046 if (name_len
< 32) {
2047 name_ptr
= tmp_name
;
2049 name_ptr
= kmalloc(name_len
, GFP_NOFS
);
2052 read_extent_buffer(leaf
, name_ptr
,
2053 (unsigned long)(di
+ 1), name_len
);
2055 d_type
= btrfs_filetype_table
[btrfs_dir_type(leaf
, di
)];
2056 btrfs_dir_item_key_to_cpu(leaf
, di
, &location
);
2057 over
= filldir(dirent
, name_ptr
, name_len
,
2062 if (name_ptr
!= tmp_name
)
2067 di_len
= btrfs_dir_name_len(leaf
, di
) +
2068 btrfs_dir_data_len(leaf
, di
) +sizeof(*di
);
2070 di
= (struct btrfs_dir_item
*)((char *)di
+ di_len
);
2073 if (key_type
== BTRFS_DIR_INDEX_KEY
)
2074 filp
->f_pos
= INT_LIMIT(typeof(filp
->f_pos
));
2080 btrfs_free_path(path
);
2084 int btrfs_write_inode(struct inode
*inode
, int wait
)
2086 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2087 struct btrfs_trans_handle
*trans
;
2091 trans
= btrfs_join_transaction(root
, 1);
2092 btrfs_set_trans_block_group(trans
, inode
);
2093 ret
= btrfs_commit_transaction(trans
, root
);
2099 * This is somewhat expensive, updating the tree every time the
2100 * inode changes. But, it is most likely to find the inode in cache.
2101 * FIXME, needs more benchmarking...there are no reasons other than performance
2102 * to keep or drop this code.
2104 void btrfs_dirty_inode(struct inode
*inode
)
2106 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2107 struct btrfs_trans_handle
*trans
;
2109 trans
= btrfs_join_transaction(root
, 1);
2110 btrfs_set_trans_block_group(trans
, inode
);
2111 btrfs_update_inode(trans
, root
, inode
);
2112 btrfs_end_transaction(trans
, root
);
2115 static int btrfs_set_inode_index_count(struct inode
*inode
)
2117 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2118 struct btrfs_key key
, found_key
;
2119 struct btrfs_path
*path
;
2120 struct extent_buffer
*leaf
;
2123 key
.objectid
= inode
->i_ino
;
2124 btrfs_set_key_type(&key
, BTRFS_DIR_INDEX_KEY
);
2125 key
.offset
= (u64
)-1;
2127 path
= btrfs_alloc_path();
2131 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2134 /* FIXME: we should be able to handle this */
2140 * MAGIC NUMBER EXPLANATION:
2141 * since we search a directory based on f_pos we have to start at 2
2142 * since '.' and '..' have f_pos of 0 and 1 respectively, so everybody
2143 * else has to start at 2
2145 if (path
->slots
[0] == 0) {
2146 BTRFS_I(inode
)->index_cnt
= 2;
2152 leaf
= path
->nodes
[0];
2153 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2155 if (found_key
.objectid
!= inode
->i_ino
||
2156 btrfs_key_type(&found_key
) != BTRFS_DIR_INDEX_KEY
) {
2157 BTRFS_I(inode
)->index_cnt
= 2;
2161 BTRFS_I(inode
)->index_cnt
= found_key
.offset
+ 1;
2163 btrfs_free_path(path
);
2167 static int btrfs_set_inode_index(struct inode
*dir
, struct inode
*inode
)
2171 if (BTRFS_I(dir
)->index_cnt
== (u64
)-1) {
2172 ret
= btrfs_set_inode_index_count(dir
);
2177 BTRFS_I(inode
)->index
= BTRFS_I(dir
)->index_cnt
;
2178 BTRFS_I(dir
)->index_cnt
++;
2183 static struct inode
*btrfs_new_inode(struct btrfs_trans_handle
*trans
,
2184 struct btrfs_root
*root
,
2186 const char *name
, int name_len
,
2189 struct btrfs_block_group_cache
*group
,
2192 struct inode
*inode
;
2193 struct btrfs_inode_item
*inode_item
;
2194 struct btrfs_block_group_cache
*new_inode_group
;
2195 struct btrfs_key
*location
;
2196 struct btrfs_path
*path
;
2197 struct btrfs_inode_ref
*ref
;
2198 struct btrfs_key key
[2];
2204 path
= btrfs_alloc_path();
2207 inode
= new_inode(root
->fs_info
->sb
);
2209 return ERR_PTR(-ENOMEM
);
2212 ret
= btrfs_set_inode_index(dir
, inode
);
2214 return ERR_PTR(ret
);
2216 BTRFS_I(inode
)->index
= 0;
2219 * index_cnt is ignored for everything but a dir,
2220 * btrfs_get_inode_index_count has an explanation for the magic
2223 BTRFS_I(inode
)->index_cnt
= 2;
2225 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
2226 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
2227 inode
->i_mapping
, GFP_NOFS
);
2228 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
2229 inode
->i_mapping
, GFP_NOFS
);
2230 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
2231 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
2232 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
2233 BTRFS_I(inode
)->delalloc_bytes
= 0;
2234 BTRFS_I(inode
)->disk_i_size
= 0;
2235 BTRFS_I(inode
)->root
= root
;
2241 new_inode_group
= btrfs_find_block_group(root
, group
, 0,
2242 BTRFS_BLOCK_GROUP_METADATA
, owner
);
2243 if (!new_inode_group
) {
2244 printk("find_block group failed\n");
2245 new_inode_group
= group
;
2247 BTRFS_I(inode
)->block_group
= new_inode_group
;
2248 BTRFS_I(inode
)->flags
= 0;
2250 key
[0].objectid
= objectid
;
2251 btrfs_set_key_type(&key
[0], BTRFS_INODE_ITEM_KEY
);
2254 key
[1].objectid
= objectid
;
2255 btrfs_set_key_type(&key
[1], BTRFS_INODE_REF_KEY
);
2256 key
[1].offset
= ref_objectid
;
2258 sizes
[0] = sizeof(struct btrfs_inode_item
);
2259 sizes
[1] = name_len
+ sizeof(*ref
);
2261 ret
= btrfs_insert_empty_items(trans
, root
, path
, key
, sizes
, 2);
2265 if (objectid
> root
->highest_inode
)
2266 root
->highest_inode
= objectid
;
2268 inode
->i_uid
= current
->fsuid
;
2269 inode
->i_gid
= current
->fsgid
;
2270 inode
->i_mode
= mode
;
2271 inode
->i_ino
= objectid
;
2272 inode
->i_blocks
= 0;
2273 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2274 inode_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
2275 struct btrfs_inode_item
);
2276 fill_inode_item(path
->nodes
[0], inode_item
, inode
);
2278 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0] + 1,
2279 struct btrfs_inode_ref
);
2280 btrfs_set_inode_ref_name_len(path
->nodes
[0], ref
, name_len
);
2281 btrfs_set_inode_ref_index(path
->nodes
[0], ref
, BTRFS_I(inode
)->index
);
2282 ptr
= (unsigned long)(ref
+ 1);
2283 write_extent_buffer(path
->nodes
[0], name
, ptr
, name_len
);
2285 btrfs_mark_buffer_dirty(path
->nodes
[0]);
2286 btrfs_free_path(path
);
2288 location
= &BTRFS_I(inode
)->location
;
2289 location
->objectid
= objectid
;
2290 location
->offset
= 0;
2291 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
2293 insert_inode_hash(inode
);
2297 BTRFS_I(dir
)->index_cnt
--;
2298 btrfs_free_path(path
);
2299 return ERR_PTR(ret
);
2302 static inline u8
btrfs_inode_type(struct inode
*inode
)
2304 return btrfs_type_by_mode
[(inode
->i_mode
& S_IFMT
) >> S_SHIFT
];
2307 static int btrfs_add_link(struct btrfs_trans_handle
*trans
,
2308 struct dentry
*dentry
, struct inode
*inode
,
2312 struct btrfs_key key
;
2313 struct btrfs_root
*root
= BTRFS_I(dentry
->d_parent
->d_inode
)->root
;
2314 struct inode
*parent_inode
= dentry
->d_parent
->d_inode
;
2316 key
.objectid
= inode
->i_ino
;
2317 btrfs_set_key_type(&key
, BTRFS_INODE_ITEM_KEY
);
2320 ret
= btrfs_insert_dir_item(trans
, root
,
2321 dentry
->d_name
.name
, dentry
->d_name
.len
,
2322 dentry
->d_parent
->d_inode
->i_ino
,
2323 &key
, btrfs_inode_type(inode
),
2324 BTRFS_I(inode
)->index
);
2327 ret
= btrfs_insert_inode_ref(trans
, root
,
2328 dentry
->d_name
.name
,
2331 parent_inode
->i_ino
,
2332 BTRFS_I(inode
)->index
);
2334 btrfs_i_size_write(parent_inode
, parent_inode
->i_size
+
2335 dentry
->d_name
.len
* 2);
2336 parent_inode
->i_mtime
= parent_inode
->i_ctime
= CURRENT_TIME
;
2337 ret
= btrfs_update_inode(trans
, root
,
2338 dentry
->d_parent
->d_inode
);
2343 static int btrfs_add_nondir(struct btrfs_trans_handle
*trans
,
2344 struct dentry
*dentry
, struct inode
*inode
,
2347 int err
= btrfs_add_link(trans
, dentry
, inode
, backref
);
2349 d_instantiate(dentry
, inode
);
2357 static int btrfs_mknod(struct inode
*dir
, struct dentry
*dentry
,
2358 int mode
, dev_t rdev
)
2360 struct btrfs_trans_handle
*trans
;
2361 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2362 struct inode
*inode
= NULL
;
2366 unsigned long nr
= 0;
2368 if (!new_valid_dev(rdev
))
2371 err
= btrfs_check_free_space(root
, 1, 0);
2375 trans
= btrfs_start_transaction(root
, 1);
2376 btrfs_set_trans_block_group(trans
, dir
);
2378 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2384 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2386 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2387 BTRFS_I(dir
)->block_group
, mode
);
2388 err
= PTR_ERR(inode
);
2392 err
= btrfs_init_acl(inode
, dir
);
2398 btrfs_set_trans_block_group(trans
, inode
);
2399 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
2403 inode
->i_op
= &btrfs_special_inode_operations
;
2404 init_special_inode(inode
, inode
->i_mode
, rdev
);
2405 btrfs_update_inode(trans
, root
, inode
);
2407 dir
->i_sb
->s_dirt
= 1;
2408 btrfs_update_inode_block_group(trans
, inode
);
2409 btrfs_update_inode_block_group(trans
, dir
);
2411 nr
= trans
->blocks_used
;
2412 btrfs_end_transaction_throttle(trans
, root
);
2415 inode_dec_link_count(inode
);
2418 btrfs_btree_balance_dirty(root
, nr
);
2422 static int btrfs_create(struct inode
*dir
, struct dentry
*dentry
,
2423 int mode
, struct nameidata
*nd
)
2425 struct btrfs_trans_handle
*trans
;
2426 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2427 struct inode
*inode
= NULL
;
2430 unsigned long nr
= 0;
2433 err
= btrfs_check_free_space(root
, 1, 0);
2436 trans
= btrfs_start_transaction(root
, 1);
2437 btrfs_set_trans_block_group(trans
, dir
);
2439 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2445 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2447 dentry
->d_parent
->d_inode
->i_ino
,
2448 objectid
, BTRFS_I(dir
)->block_group
, mode
);
2449 err
= PTR_ERR(inode
);
2453 err
= btrfs_init_acl(inode
, dir
);
2459 btrfs_set_trans_block_group(trans
, inode
);
2460 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
2464 inode
->i_mapping
->a_ops
= &btrfs_aops
;
2465 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
2466 inode
->i_fop
= &btrfs_file_operations
;
2467 inode
->i_op
= &btrfs_file_inode_operations
;
2468 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
2469 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
2470 inode
->i_mapping
, GFP_NOFS
);
2471 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
2472 inode
->i_mapping
, GFP_NOFS
);
2473 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
2474 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
2475 BTRFS_I(inode
)->delalloc_bytes
= 0;
2476 BTRFS_I(inode
)->disk_i_size
= 0;
2477 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
2478 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
2480 dir
->i_sb
->s_dirt
= 1;
2481 btrfs_update_inode_block_group(trans
, inode
);
2482 btrfs_update_inode_block_group(trans
, dir
);
2484 nr
= trans
->blocks_used
;
2485 btrfs_end_transaction_throttle(trans
, root
);
2488 inode_dec_link_count(inode
);
2491 btrfs_btree_balance_dirty(root
, nr
);
2495 static int btrfs_link(struct dentry
*old_dentry
, struct inode
*dir
,
2496 struct dentry
*dentry
)
2498 struct btrfs_trans_handle
*trans
;
2499 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2500 struct inode
*inode
= old_dentry
->d_inode
;
2501 unsigned long nr
= 0;
2505 if (inode
->i_nlink
== 0)
2508 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
2513 err
= btrfs_check_free_space(root
, 1, 0);
2516 err
= btrfs_set_inode_index(dir
, inode
);
2520 trans
= btrfs_start_transaction(root
, 1);
2522 btrfs_set_trans_block_group(trans
, dir
);
2523 atomic_inc(&inode
->i_count
);
2525 err
= btrfs_add_nondir(trans
, dentry
, inode
, 1);
2530 dir
->i_sb
->s_dirt
= 1;
2531 btrfs_update_inode_block_group(trans
, dir
);
2532 err
= btrfs_update_inode(trans
, root
, inode
);
2537 nr
= trans
->blocks_used
;
2538 btrfs_end_transaction_throttle(trans
, root
);
2541 inode_dec_link_count(inode
);
2544 btrfs_btree_balance_dirty(root
, nr
);
2548 static int btrfs_mkdir(struct inode
*dir
, struct dentry
*dentry
, int mode
)
2550 struct inode
*inode
= NULL
;
2551 struct btrfs_trans_handle
*trans
;
2552 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2554 int drop_on_err
= 0;
2556 unsigned long nr
= 1;
2558 err
= btrfs_check_free_space(root
, 1, 0);
2562 trans
= btrfs_start_transaction(root
, 1);
2563 btrfs_set_trans_block_group(trans
, dir
);
2565 if (IS_ERR(trans
)) {
2566 err
= PTR_ERR(trans
);
2570 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2576 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2578 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2579 BTRFS_I(dir
)->block_group
, S_IFDIR
| mode
);
2580 if (IS_ERR(inode
)) {
2581 err
= PTR_ERR(inode
);
2587 err
= btrfs_init_acl(inode
, dir
);
2591 inode
->i_op
= &btrfs_dir_inode_operations
;
2592 inode
->i_fop
= &btrfs_dir_file_operations
;
2593 btrfs_set_trans_block_group(trans
, inode
);
2595 btrfs_i_size_write(inode
, 0);
2596 err
= btrfs_update_inode(trans
, root
, inode
);
2600 err
= btrfs_add_link(trans
, dentry
, inode
, 0);
2604 d_instantiate(dentry
, inode
);
2606 dir
->i_sb
->s_dirt
= 1;
2607 btrfs_update_inode_block_group(trans
, inode
);
2608 btrfs_update_inode_block_group(trans
, dir
);
2611 nr
= trans
->blocks_used
;
2612 btrfs_end_transaction_throttle(trans
, root
);
2617 btrfs_btree_balance_dirty(root
, nr
);
2621 static int merge_extent_mapping(struct extent_map_tree
*em_tree
,
2622 struct extent_map
*existing
,
2623 struct extent_map
*em
,
2624 u64 map_start
, u64 map_len
)
2628 BUG_ON(map_start
< em
->start
|| map_start
>= extent_map_end(em
));
2629 start_diff
= map_start
- em
->start
;
2630 em
->start
= map_start
;
2632 if (em
->block_start
< EXTENT_MAP_LAST_BYTE
)
2633 em
->block_start
+= start_diff
;
2634 return add_extent_mapping(em_tree
, em
);
2637 struct extent_map
*btrfs_get_extent(struct inode
*inode
, struct page
*page
,
2638 size_t pg_offset
, u64 start
, u64 len
,
2644 u64 extent_start
= 0;
2646 u64 objectid
= inode
->i_ino
;
2648 struct btrfs_path
*path
= NULL
;
2649 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2650 struct btrfs_file_extent_item
*item
;
2651 struct extent_buffer
*leaf
;
2652 struct btrfs_key found_key
;
2653 struct extent_map
*em
= NULL
;
2654 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
2655 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2656 struct btrfs_trans_handle
*trans
= NULL
;
2659 spin_lock(&em_tree
->lock
);
2660 em
= lookup_extent_mapping(em_tree
, start
, len
);
2662 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2663 spin_unlock(&em_tree
->lock
);
2666 if (em
->start
> start
|| em
->start
+ em
->len
<= start
)
2667 free_extent_map(em
);
2668 else if (em
->block_start
== EXTENT_MAP_INLINE
&& page
)
2669 free_extent_map(em
);
2673 em
= alloc_extent_map(GFP_NOFS
);
2678 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2679 em
->start
= EXTENT_MAP_HOLE
;
2683 path
= btrfs_alloc_path();
2687 ret
= btrfs_lookup_file_extent(trans
, root
, path
,
2688 objectid
, start
, trans
!= NULL
);
2695 if (path
->slots
[0] == 0)
2700 leaf
= path
->nodes
[0];
2701 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
2702 struct btrfs_file_extent_item
);
2703 /* are we inside the extent that was found? */
2704 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2705 found_type
= btrfs_key_type(&found_key
);
2706 if (found_key
.objectid
!= objectid
||
2707 found_type
!= BTRFS_EXTENT_DATA_KEY
) {
2711 found_type
= btrfs_file_extent_type(leaf
, item
);
2712 extent_start
= found_key
.offset
;
2713 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
2714 extent_end
= extent_start
+
2715 btrfs_file_extent_num_bytes(leaf
, item
);
2717 if (start
< extent_start
|| start
>= extent_end
) {
2719 if (start
< extent_start
) {
2720 if (start
+ len
<= extent_start
)
2722 em
->len
= extent_end
- extent_start
;
2728 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
2730 em
->start
= extent_start
;
2731 em
->len
= extent_end
- extent_start
;
2732 em
->block_start
= EXTENT_MAP_HOLE
;
2735 bytenr
+= btrfs_file_extent_offset(leaf
, item
);
2736 em
->block_start
= bytenr
;
2737 em
->start
= extent_start
;
2738 em
->len
= extent_end
- extent_start
;
2740 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
2745 size_t extent_offset
;
2748 size
= btrfs_file_extent_inline_len(leaf
, btrfs_item_nr(leaf
,
2750 extent_end
= (extent_start
+ size
+ root
->sectorsize
- 1) &
2751 ~((u64
)root
->sectorsize
- 1);
2752 if (start
< extent_start
|| start
>= extent_end
) {
2754 if (start
< extent_start
) {
2755 if (start
+ len
<= extent_start
)
2757 em
->len
= extent_end
- extent_start
;
2763 em
->block_start
= EXTENT_MAP_INLINE
;
2766 em
->start
= extent_start
;
2771 page_start
= page_offset(page
) + pg_offset
;
2772 extent_offset
= page_start
- extent_start
;
2773 copy_size
= min_t(u64
, PAGE_CACHE_SIZE
- pg_offset
,
2774 size
- extent_offset
);
2775 em
->start
= extent_start
+ extent_offset
;
2776 em
->len
= (copy_size
+ root
->sectorsize
- 1) &
2777 ~((u64
)root
->sectorsize
- 1);
2779 ptr
= btrfs_file_extent_inline_start(item
) + extent_offset
;
2780 if (create
== 0 && !PageUptodate(page
)) {
2781 read_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2783 flush_dcache_page(page
);
2784 } else if (create
&& PageUptodate(page
)) {
2787 free_extent_map(em
);
2789 btrfs_release_path(root
, path
);
2790 trans
= btrfs_join_transaction(root
, 1);
2793 write_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2795 btrfs_mark_buffer_dirty(leaf
);
2798 set_extent_uptodate(io_tree
, em
->start
,
2799 extent_map_end(em
) - 1, GFP_NOFS
);
2802 printk("unkknown found_type %d\n", found_type
);
2809 em
->block_start
= EXTENT_MAP_HOLE
;
2811 btrfs_release_path(root
, path
);
2812 if (em
->start
> start
|| extent_map_end(em
) <= start
) {
2813 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em
->start
, em
->len
, start
, len
);
2819 spin_lock(&em_tree
->lock
);
2820 ret
= add_extent_mapping(em_tree
, em
);
2821 /* it is possible that someone inserted the extent into the tree
2822 * while we had the lock dropped. It is also possible that
2823 * an overlapping map exists in the tree
2825 if (ret
== -EEXIST
) {
2826 struct extent_map
*existing
;
2830 existing
= lookup_extent_mapping(em_tree
, start
, len
);
2831 if (existing
&& (existing
->start
> start
||
2832 existing
->start
+ existing
->len
<= start
)) {
2833 free_extent_map(existing
);
2837 existing
= lookup_extent_mapping(em_tree
, em
->start
,
2840 err
= merge_extent_mapping(em_tree
, existing
,
2843 free_extent_map(existing
);
2845 free_extent_map(em
);
2850 printk("failing to insert %Lu %Lu\n",
2852 free_extent_map(em
);
2856 free_extent_map(em
);
2861 spin_unlock(&em_tree
->lock
);
2864 btrfs_free_path(path
);
2866 ret
= btrfs_end_transaction(trans
, root
);
2872 free_extent_map(em
);
2874 return ERR_PTR(err
);
2879 #if 0 /* waiting for O_DIRECT reads */
2880 static int btrfs_get_block(struct inode
*inode
, sector_t iblock
,
2881 struct buffer_head
*bh_result
, int create
)
2883 struct extent_map
*em
;
2884 u64 start
= (u64
)iblock
<< inode
->i_blkbits
;
2885 struct btrfs_multi_bio
*multi
= NULL
;
2886 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2892 em
= btrfs_get_extent(inode
, NULL
, 0, start
, bh_result
->b_size
, 0);
2894 if (!em
|| IS_ERR(em
))
2897 if (em
->start
> start
|| em
->start
+ em
->len
<= start
) {
2901 if (em
->block_start
== EXTENT_MAP_INLINE
) {
2906 len
= em
->start
+ em
->len
- start
;
2907 len
= min_t(u64
, len
, INT_LIMIT(typeof(bh_result
->b_size
)));
2909 if (em
->block_start
== EXTENT_MAP_HOLE
||
2910 em
->block_start
== EXTENT_MAP_DELALLOC
) {
2911 bh_result
->b_size
= len
;
2915 logical
= start
- em
->start
;
2916 logical
= em
->block_start
+ logical
;
2919 ret
= btrfs_map_block(&root
->fs_info
->mapping_tree
, READ
,
2920 logical
, &map_length
, &multi
, 0);
2922 bh_result
->b_blocknr
= multi
->stripes
[0].physical
>> inode
->i_blkbits
;
2923 bh_result
->b_size
= min(map_length
, len
);
2925 bh_result
->b_bdev
= multi
->stripes
[0].dev
->bdev
;
2926 set_buffer_mapped(bh_result
);
2929 free_extent_map(em
);
2934 static ssize_t
btrfs_direct_IO(int rw
, struct kiocb
*iocb
,
2935 const struct iovec
*iov
, loff_t offset
,
2936 unsigned long nr_segs
)
2940 struct file
*file
= iocb
->ki_filp
;
2941 struct inode
*inode
= file
->f_mapping
->host
;
2946 return blockdev_direct_IO(rw
, iocb
, inode
, inode
->i_sb
->s_bdev
, iov
,
2947 offset
, nr_segs
, btrfs_get_block
, NULL
);
2951 static sector_t
btrfs_bmap(struct address_space
*mapping
, sector_t iblock
)
2953 return extent_bmap(mapping
, iblock
, btrfs_get_extent
);
2956 int btrfs_readpage(struct file
*file
, struct page
*page
)
2958 struct extent_io_tree
*tree
;
2959 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2960 return extent_read_full_page(tree
, page
, btrfs_get_extent
);
2963 static int btrfs_writepage(struct page
*page
, struct writeback_control
*wbc
)
2965 struct extent_io_tree
*tree
;
2968 if (current
->flags
& PF_MEMALLOC
) {
2969 redirty_page_for_writepage(wbc
, page
);
2973 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
2974 return extent_write_full_page(tree
, page
, btrfs_get_extent
, wbc
);
2977 int btrfs_writepages(struct address_space
*mapping
,
2978 struct writeback_control
*wbc
)
2980 struct extent_io_tree
*tree
;
2981 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
2982 return extent_writepages(tree
, mapping
, btrfs_get_extent
, wbc
);
2986 btrfs_readpages(struct file
*file
, struct address_space
*mapping
,
2987 struct list_head
*pages
, unsigned nr_pages
)
2989 struct extent_io_tree
*tree
;
2990 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
2991 return extent_readpages(tree
, mapping
, pages
, nr_pages
,
2994 static int __btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
2996 struct extent_io_tree
*tree
;
2997 struct extent_map_tree
*map
;
3000 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3001 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
3002 ret
= try_release_extent_mapping(map
, tree
, page
, gfp_flags
);
3004 ClearPagePrivate(page
);
3005 set_page_private(page
, 0);
3006 page_cache_release(page
);
3011 static int btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
3013 return __btrfs_releasepage(page
, gfp_flags
);
3016 static void btrfs_invalidatepage(struct page
*page
, unsigned long offset
)
3018 struct extent_io_tree
*tree
;
3019 struct btrfs_ordered_extent
*ordered
;
3020 u64 page_start
= page_offset(page
);
3021 u64 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
3023 wait_on_page_writeback(page
);
3024 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3026 btrfs_releasepage(page
, GFP_NOFS
);
3030 lock_extent(tree
, page_start
, page_end
, GFP_NOFS
);
3031 ordered
= btrfs_lookup_ordered_extent(page
->mapping
->host
,
3035 * IO on this page will never be started, so we need
3036 * to account for any ordered extents now
3038 clear_extent_bit(tree
, page_start
, page_end
,
3039 EXTENT_DIRTY
| EXTENT_DELALLOC
|
3040 EXTENT_LOCKED
, 1, 0, GFP_NOFS
);
3041 btrfs_finish_ordered_io(page
->mapping
->host
,
3042 page_start
, page_end
);
3043 btrfs_put_ordered_extent(ordered
);
3044 lock_extent(tree
, page_start
, page_end
, GFP_NOFS
);
3046 clear_extent_bit(tree
, page_start
, page_end
,
3047 EXTENT_LOCKED
| EXTENT_DIRTY
| EXTENT_DELALLOC
|
3050 __btrfs_releasepage(page
, GFP_NOFS
);
3052 ClearPageChecked(page
);
3053 if (PagePrivate(page
)) {
3054 ClearPagePrivate(page
);
3055 set_page_private(page
, 0);
3056 page_cache_release(page
);
3061 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
3062 * called from a page fault handler when a page is first dirtied. Hence we must
3063 * be careful to check for EOF conditions here. We set the page up correctly
3064 * for a written page which means we get ENOSPC checking when writing into
3065 * holes and correct delalloc and unwritten extent mapping on filesystems that
3066 * support these features.
3068 * We are not allowed to take the i_mutex here so we have to play games to
3069 * protect against truncate races as the page could now be beyond EOF. Because
3070 * vmtruncate() writes the inode size before removing pages, once we have the
3071 * page lock we can determine safely if the page is beyond EOF. If it is not
3072 * beyond EOF, then the page is guaranteed safe against truncation until we
3075 int btrfs_page_mkwrite(struct vm_area_struct
*vma
, struct page
*page
)
3077 struct inode
*inode
= fdentry(vma
->vm_file
)->d_inode
;
3078 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3079 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
3080 struct btrfs_ordered_extent
*ordered
;
3082 unsigned long zero_start
;
3088 ret
= btrfs_check_free_space(root
, PAGE_CACHE_SIZE
, 0);
3095 size
= i_size_read(inode
);
3096 page_start
= page_offset(page
);
3097 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
3099 if ((page
->mapping
!= inode
->i_mapping
) ||
3100 (page_start
>= size
)) {
3101 /* page got truncated out from underneath us */
3104 wait_on_page_writeback(page
);
3106 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3107 set_page_extent_mapped(page
);
3110 * we can't set the delalloc bits if there are pending ordered
3111 * extents. Drop our locks and wait for them to finish
3113 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
3115 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3117 btrfs_start_ordered_extent(inode
, ordered
, 1);
3118 btrfs_put_ordered_extent(ordered
);
3122 set_extent_delalloc(&BTRFS_I(inode
)->io_tree
, page_start
,
3123 page_end
, GFP_NOFS
);
3126 /* page is wholly or partially inside EOF */
3127 if (page_start
+ PAGE_CACHE_SIZE
> size
)
3128 zero_start
= size
& ~PAGE_CACHE_MASK
;
3130 zero_start
= PAGE_CACHE_SIZE
;
3132 if (zero_start
!= PAGE_CACHE_SIZE
) {
3134 memset(kaddr
+ zero_start
, 0, PAGE_CACHE_SIZE
- zero_start
);
3135 flush_dcache_page(page
);
3138 ClearPageChecked(page
);
3139 set_page_dirty(page
);
3140 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3148 static void btrfs_truncate(struct inode
*inode
)
3150 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3152 struct btrfs_trans_handle
*trans
;
3154 u64 mask
= root
->sectorsize
- 1;
3156 if (!S_ISREG(inode
->i_mode
))
3158 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
3161 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
3162 btrfs_wait_ordered_range(inode
, inode
->i_size
& (~mask
), (u64
)-1);
3164 trans
= btrfs_start_transaction(root
, 1);
3165 btrfs_set_trans_block_group(trans
, inode
);
3166 btrfs_i_size_write(inode
, inode
->i_size
);
3168 ret
= btrfs_orphan_add(trans
, inode
);
3171 /* FIXME, add redo link to tree so we don't leak on crash */
3172 ret
= btrfs_truncate_in_trans(trans
, root
, inode
,
3173 BTRFS_EXTENT_DATA_KEY
);
3174 btrfs_update_inode(trans
, root
, inode
);
3176 ret
= btrfs_orphan_del(trans
, inode
);
3180 nr
= trans
->blocks_used
;
3181 ret
= btrfs_end_transaction_throttle(trans
, root
);
3183 btrfs_btree_balance_dirty(root
, nr
);
3187 * Invalidate a single dcache entry at the root of the filesystem.
3188 * Needed after creation of snapshot or subvolume.
3190 void btrfs_invalidate_dcache_root(struct btrfs_root
*root
, char *name
,
3193 struct dentry
*alias
, *entry
;
3196 alias
= d_find_alias(root
->fs_info
->sb
->s_root
->d_inode
);
3200 /* change me if btrfs ever gets a d_hash operation */
3201 qstr
.hash
= full_name_hash(qstr
.name
, qstr
.len
);
3202 entry
= d_lookup(alias
, &qstr
);
3205 d_invalidate(entry
);
3211 int btrfs_create_subvol_root(struct btrfs_root
*new_root
,
3212 struct btrfs_trans_handle
*trans
, u64 new_dirid
,
3213 struct btrfs_block_group_cache
*block_group
)
3215 struct inode
*inode
;
3217 inode
= btrfs_new_inode(trans
, new_root
, NULL
, "..", 2, new_dirid
,
3218 new_dirid
, block_group
, S_IFDIR
| 0700);
3220 return PTR_ERR(inode
);
3221 inode
->i_op
= &btrfs_dir_inode_operations
;
3222 inode
->i_fop
= &btrfs_dir_file_operations
;
3223 new_root
->inode
= inode
;
3226 btrfs_i_size_write(inode
, 0);
3228 return btrfs_update_inode(trans
, new_root
, inode
);
3231 unsigned long btrfs_force_ra(struct address_space
*mapping
,
3232 struct file_ra_state
*ra
, struct file
*file
,
3233 pgoff_t offset
, pgoff_t last_index
)
3235 pgoff_t req_size
= last_index
- offset
+ 1;
3237 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3238 offset
= page_cache_readahead(mapping
, ra
, file
, offset
, req_size
);
3241 page_cache_sync_readahead(mapping
, ra
, file
, offset
, req_size
);
3242 return offset
+ req_size
;
3246 struct inode
*btrfs_alloc_inode(struct super_block
*sb
)
3248 struct btrfs_inode
*ei
;
3250 ei
= kmem_cache_alloc(btrfs_inode_cachep
, GFP_NOFS
);
3254 btrfs_ordered_inode_tree_init(&ei
->ordered_tree
);
3255 ei
->i_acl
= BTRFS_ACL_NOT_CACHED
;
3256 ei
->i_default_acl
= BTRFS_ACL_NOT_CACHED
;
3257 INIT_LIST_HEAD(&ei
->i_orphan
);
3258 return &ei
->vfs_inode
;
3261 void btrfs_destroy_inode(struct inode
*inode
)
3263 struct btrfs_ordered_extent
*ordered
;
3264 WARN_ON(!list_empty(&inode
->i_dentry
));
3265 WARN_ON(inode
->i_data
.nrpages
);
3267 if (BTRFS_I(inode
)->i_acl
&&
3268 BTRFS_I(inode
)->i_acl
!= BTRFS_ACL_NOT_CACHED
)
3269 posix_acl_release(BTRFS_I(inode
)->i_acl
);
3270 if (BTRFS_I(inode
)->i_default_acl
&&
3271 BTRFS_I(inode
)->i_default_acl
!= BTRFS_ACL_NOT_CACHED
)
3272 posix_acl_release(BTRFS_I(inode
)->i_default_acl
);
3274 spin_lock(&BTRFS_I(inode
)->root
->orphan_lock
);
3275 if (!list_empty(&BTRFS_I(inode
)->i_orphan
)) {
3276 printk(KERN_ERR
"BTRFS: inode %lu: inode still on the orphan"
3277 " list\n", inode
->i_ino
);
3280 spin_unlock(&BTRFS_I(inode
)->root
->orphan_lock
);
3283 ordered
= btrfs_lookup_first_ordered_extent(inode
, (u64
)-1);
3287 printk("found ordered extent %Lu %Lu\n",
3288 ordered
->file_offset
, ordered
->len
);
3289 btrfs_remove_ordered_extent(inode
, ordered
);
3290 btrfs_put_ordered_extent(ordered
);
3291 btrfs_put_ordered_extent(ordered
);
3294 btrfs_drop_extent_cache(inode
, 0, (u64
)-1);
3295 kmem_cache_free(btrfs_inode_cachep
, BTRFS_I(inode
));
3298 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3299 static void init_once(struct kmem_cache
* cachep
, void *foo
)
3301 static void init_once(void * foo
, struct kmem_cache
* cachep
,
3302 unsigned long flags
)
3305 struct btrfs_inode
*ei
= (struct btrfs_inode
*) foo
;
3307 inode_init_once(&ei
->vfs_inode
);
3310 void btrfs_destroy_cachep(void)
3312 if (btrfs_inode_cachep
)
3313 kmem_cache_destroy(btrfs_inode_cachep
);
3314 if (btrfs_trans_handle_cachep
)
3315 kmem_cache_destroy(btrfs_trans_handle_cachep
);
3316 if (btrfs_transaction_cachep
)
3317 kmem_cache_destroy(btrfs_transaction_cachep
);
3318 if (btrfs_bit_radix_cachep
)
3319 kmem_cache_destroy(btrfs_bit_radix_cachep
);
3320 if (btrfs_path_cachep
)
3321 kmem_cache_destroy(btrfs_path_cachep
);
3324 struct kmem_cache
*btrfs_cache_create(const char *name
, size_t size
,
3325 unsigned long extra_flags
,
3326 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3327 void (*ctor
)(struct kmem_cache
*, void *)
3329 void (*ctor
)(void *, struct kmem_cache
*,
3334 return kmem_cache_create(name
, size
, 0, (SLAB_RECLAIM_ACCOUNT
|
3335 SLAB_MEM_SPREAD
| extra_flags
), ctor
3336 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3342 int btrfs_init_cachep(void)
3344 btrfs_inode_cachep
= btrfs_cache_create("btrfs_inode_cache",
3345 sizeof(struct btrfs_inode
),
3347 if (!btrfs_inode_cachep
)
3349 btrfs_trans_handle_cachep
=
3350 btrfs_cache_create("btrfs_trans_handle_cache",
3351 sizeof(struct btrfs_trans_handle
),
3353 if (!btrfs_trans_handle_cachep
)
3355 btrfs_transaction_cachep
= btrfs_cache_create("btrfs_transaction_cache",
3356 sizeof(struct btrfs_transaction
),
3358 if (!btrfs_transaction_cachep
)
3360 btrfs_path_cachep
= btrfs_cache_create("btrfs_path_cache",
3361 sizeof(struct btrfs_path
),
3363 if (!btrfs_path_cachep
)
3365 btrfs_bit_radix_cachep
= btrfs_cache_create("btrfs_radix", 256,
3366 SLAB_DESTROY_BY_RCU
, NULL
);
3367 if (!btrfs_bit_radix_cachep
)
3371 btrfs_destroy_cachep();
3375 static int btrfs_getattr(struct vfsmount
*mnt
,
3376 struct dentry
*dentry
, struct kstat
*stat
)
3378 struct inode
*inode
= dentry
->d_inode
;
3379 generic_fillattr(inode
, stat
);
3380 stat
->blksize
= PAGE_CACHE_SIZE
;
3381 stat
->blocks
= inode
->i_blocks
+ (BTRFS_I(inode
)->delalloc_bytes
>> 9);
3385 static int btrfs_rename(struct inode
* old_dir
, struct dentry
*old_dentry
,
3386 struct inode
* new_dir
,struct dentry
*new_dentry
)
3388 struct btrfs_trans_handle
*trans
;
3389 struct btrfs_root
*root
= BTRFS_I(old_dir
)->root
;
3390 struct inode
*new_inode
= new_dentry
->d_inode
;
3391 struct inode
*old_inode
= old_dentry
->d_inode
;
3392 struct timespec ctime
= CURRENT_TIME
;
3395 if (S_ISDIR(old_inode
->i_mode
) && new_inode
&&
3396 new_inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
3400 ret
= btrfs_check_free_space(root
, 1, 0);
3404 trans
= btrfs_start_transaction(root
, 1);
3406 btrfs_set_trans_block_group(trans
, new_dir
);
3408 old_dentry
->d_inode
->i_nlink
++;
3409 old_dir
->i_ctime
= old_dir
->i_mtime
= ctime
;
3410 new_dir
->i_ctime
= new_dir
->i_mtime
= ctime
;
3411 old_inode
->i_ctime
= ctime
;
3413 ret
= btrfs_unlink_trans(trans
, root
, old_dir
, old_dentry
);
3418 new_inode
->i_ctime
= CURRENT_TIME
;
3419 ret
= btrfs_unlink_trans(trans
, root
, new_dir
, new_dentry
);
3422 if (new_inode
->i_nlink
== 0) {
3423 ret
= btrfs_orphan_add(trans
, new_inode
);
3428 ret
= btrfs_set_inode_index(new_dir
, old_inode
);
3432 ret
= btrfs_add_link(trans
, new_dentry
, old_inode
, 1);
3437 btrfs_end_transaction_throttle(trans
, root
);
3442 static int btrfs_symlink(struct inode
*dir
, struct dentry
*dentry
,
3443 const char *symname
)
3445 struct btrfs_trans_handle
*trans
;
3446 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
3447 struct btrfs_path
*path
;
3448 struct btrfs_key key
;
3449 struct inode
*inode
= NULL
;
3456 struct btrfs_file_extent_item
*ei
;
3457 struct extent_buffer
*leaf
;
3458 unsigned long nr
= 0;
3460 name_len
= strlen(symname
) + 1;
3461 if (name_len
> BTRFS_MAX_INLINE_DATA_SIZE(root
))
3462 return -ENAMETOOLONG
;
3464 err
= btrfs_check_free_space(root
, 1, 0);
3468 trans
= btrfs_start_transaction(root
, 1);
3469 btrfs_set_trans_block_group(trans
, dir
);
3471 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
3477 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
3479 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
3480 BTRFS_I(dir
)->block_group
, S_IFLNK
|S_IRWXUGO
);
3481 err
= PTR_ERR(inode
);
3485 err
= btrfs_init_acl(inode
, dir
);
3491 btrfs_set_trans_block_group(trans
, inode
);
3492 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0);
3496 inode
->i_mapping
->a_ops
= &btrfs_aops
;
3497 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3498 inode
->i_fop
= &btrfs_file_operations
;
3499 inode
->i_op
= &btrfs_file_inode_operations
;
3500 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
3501 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
3502 inode
->i_mapping
, GFP_NOFS
);
3503 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
3504 inode
->i_mapping
, GFP_NOFS
);
3505 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
3506 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
3507 BTRFS_I(inode
)->delalloc_bytes
= 0;
3508 BTRFS_I(inode
)->disk_i_size
= 0;
3509 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
3510 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
3512 dir
->i_sb
->s_dirt
= 1;
3513 btrfs_update_inode_block_group(trans
, inode
);
3514 btrfs_update_inode_block_group(trans
, dir
);
3518 path
= btrfs_alloc_path();
3520 key
.objectid
= inode
->i_ino
;
3522 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
3523 datasize
= btrfs_file_extent_calc_inline_size(name_len
);
3524 err
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
3530 leaf
= path
->nodes
[0];
3531 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
3532 struct btrfs_file_extent_item
);
3533 btrfs_set_file_extent_generation(leaf
, ei
, trans
->transid
);
3534 btrfs_set_file_extent_type(leaf
, ei
,
3535 BTRFS_FILE_EXTENT_INLINE
);
3536 ptr
= btrfs_file_extent_inline_start(ei
);
3537 write_extent_buffer(leaf
, symname
, ptr
, name_len
);
3538 btrfs_mark_buffer_dirty(leaf
);
3539 btrfs_free_path(path
);
3541 inode
->i_op
= &btrfs_symlink_inode_operations
;
3542 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
3543 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3544 btrfs_i_size_write(inode
, name_len
- 1);
3545 err
= btrfs_update_inode(trans
, root
, inode
);
3550 nr
= trans
->blocks_used
;
3551 btrfs_end_transaction_throttle(trans
, root
);
3554 inode_dec_link_count(inode
);
3557 btrfs_btree_balance_dirty(root
, nr
);
3561 static int btrfs_set_page_dirty(struct page
*page
)
3563 return __set_page_dirty_nobuffers(page
);
3566 static int btrfs_permission(struct inode
*inode
, int mask
,
3567 struct nameidata
*nd
)
3569 if (btrfs_test_flag(inode
, READONLY
) && (mask
& MAY_WRITE
))
3571 return generic_permission(inode
, mask
, btrfs_check_acl
);
3574 static struct inode_operations btrfs_dir_inode_operations
= {
3575 .lookup
= btrfs_lookup
,
3576 .create
= btrfs_create
,
3577 .unlink
= btrfs_unlink
,
3579 .mkdir
= btrfs_mkdir
,
3580 .rmdir
= btrfs_rmdir
,
3581 .rename
= btrfs_rename
,
3582 .symlink
= btrfs_symlink
,
3583 .setattr
= btrfs_setattr
,
3584 .mknod
= btrfs_mknod
,
3585 .setxattr
= generic_setxattr
,
3586 .getxattr
= generic_getxattr
,
3587 .listxattr
= btrfs_listxattr
,
3588 .removexattr
= generic_removexattr
,
3589 .permission
= btrfs_permission
,
3591 static struct inode_operations btrfs_dir_ro_inode_operations
= {
3592 .lookup
= btrfs_lookup
,
3593 .permission
= btrfs_permission
,
3595 static struct file_operations btrfs_dir_file_operations
= {
3596 .llseek
= generic_file_llseek
,
3597 .read
= generic_read_dir
,
3598 .readdir
= btrfs_readdir
,
3599 .unlocked_ioctl
= btrfs_ioctl
,
3600 #ifdef CONFIG_COMPAT
3601 .compat_ioctl
= btrfs_ioctl
,
3603 .release
= btrfs_release_file
,
3606 static struct extent_io_ops btrfs_extent_io_ops
= {
3607 .fill_delalloc
= run_delalloc_range
,
3608 .submit_bio_hook
= btrfs_submit_bio_hook
,
3609 .merge_bio_hook
= btrfs_merge_bio_hook
,
3610 .readpage_io_hook
= btrfs_readpage_io_hook
,
3611 .readpage_end_io_hook
= btrfs_readpage_end_io_hook
,
3612 .writepage_end_io_hook
= btrfs_writepage_end_io_hook
,
3613 .writepage_start_hook
= btrfs_writepage_start_hook
,
3614 .readpage_io_failed_hook
= btrfs_io_failed_hook
,
3615 .set_bit_hook
= btrfs_set_bit_hook
,
3616 .clear_bit_hook
= btrfs_clear_bit_hook
,
3619 static struct address_space_operations btrfs_aops
= {
3620 .readpage
= btrfs_readpage
,
3621 .writepage
= btrfs_writepage
,
3622 .writepages
= btrfs_writepages
,
3623 .readpages
= btrfs_readpages
,
3624 .sync_page
= block_sync_page
,
3626 .direct_IO
= btrfs_direct_IO
,
3627 .invalidatepage
= btrfs_invalidatepage
,
3628 .releasepage
= btrfs_releasepage
,
3629 .set_page_dirty
= btrfs_set_page_dirty
,
3632 static struct address_space_operations btrfs_symlink_aops
= {
3633 .readpage
= btrfs_readpage
,
3634 .writepage
= btrfs_writepage
,
3635 .invalidatepage
= btrfs_invalidatepage
,
3636 .releasepage
= btrfs_releasepage
,
3639 static struct inode_operations btrfs_file_inode_operations
= {
3640 .truncate
= btrfs_truncate
,
3641 .getattr
= btrfs_getattr
,
3642 .setattr
= btrfs_setattr
,
3643 .setxattr
= generic_setxattr
,
3644 .getxattr
= generic_getxattr
,
3645 .listxattr
= btrfs_listxattr
,
3646 .removexattr
= generic_removexattr
,
3647 .permission
= btrfs_permission
,
3649 static struct inode_operations btrfs_special_inode_operations
= {
3650 .getattr
= btrfs_getattr
,
3651 .setattr
= btrfs_setattr
,
3652 .permission
= btrfs_permission
,
3653 .setxattr
= generic_setxattr
,
3654 .getxattr
= generic_getxattr
,
3655 .listxattr
= btrfs_listxattr
,
3656 .removexattr
= generic_removexattr
,
3658 static struct inode_operations btrfs_symlink_inode_operations
= {
3659 .readlink
= generic_readlink
,
3660 .follow_link
= page_follow_link_light
,
3661 .put_link
= page_put_link
,
3662 .permission
= btrfs_permission
,