2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/pagemap.h>
25 #include <linux/highmem.h>
26 #include <linux/time.h>
27 #include <linux/init.h>
28 #include <linux/string.h>
29 #include <linux/smp_lock.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mpage.h>
32 #include <linux/swap.h>
33 #include <linux/writeback.h>
34 #include <linux/statfs.h>
35 #include <linux/compat.h>
36 #include <linux/bit_spinlock.h>
37 #include <linux/version.h>
38 #include <linux/xattr.h>
39 #include <linux/posix_acl.h>
42 #include "transaction.h"
43 #include "btrfs_inode.h"
45 #include "print-tree.h"
47 #include "ordered-data.h"
52 struct btrfs_iget_args
{
54 struct btrfs_root
*root
;
57 static struct inode_operations btrfs_dir_inode_operations
;
58 static struct inode_operations btrfs_symlink_inode_operations
;
59 static struct inode_operations btrfs_dir_ro_inode_operations
;
60 static struct inode_operations btrfs_special_inode_operations
;
61 static struct inode_operations btrfs_file_inode_operations
;
62 static struct address_space_operations btrfs_aops
;
63 static struct address_space_operations btrfs_symlink_aops
;
64 static struct file_operations btrfs_dir_file_operations
;
65 static struct extent_io_ops btrfs_extent_io_ops
;
67 static struct kmem_cache
*btrfs_inode_cachep
;
68 struct kmem_cache
*btrfs_trans_handle_cachep
;
69 struct kmem_cache
*btrfs_transaction_cachep
;
70 struct kmem_cache
*btrfs_bit_radix_cachep
;
71 struct kmem_cache
*btrfs_path_cachep
;
74 static unsigned char btrfs_type_by_mode
[S_IFMT
>> S_SHIFT
] = {
75 [S_IFREG
>> S_SHIFT
] = BTRFS_FT_REG_FILE
,
76 [S_IFDIR
>> S_SHIFT
] = BTRFS_FT_DIR
,
77 [S_IFCHR
>> S_SHIFT
] = BTRFS_FT_CHRDEV
,
78 [S_IFBLK
>> S_SHIFT
] = BTRFS_FT_BLKDEV
,
79 [S_IFIFO
>> S_SHIFT
] = BTRFS_FT_FIFO
,
80 [S_IFSOCK
>> S_SHIFT
] = BTRFS_FT_SOCK
,
81 [S_IFLNK
>> S_SHIFT
] = BTRFS_FT_SYMLINK
,
84 static void btrfs_truncate(struct inode
*inode
);
86 int btrfs_check_free_space(struct btrfs_root
*root
, u64 num_required
,
95 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
96 total
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
97 used
= btrfs_super_bytes_used(&root
->fs_info
->super_copy
);
105 if (used
+ root
->fs_info
->delalloc_bytes
+ num_required
> thresh
)
107 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
111 static int cow_file_range(struct inode
*inode
, u64 start
, u64 end
)
113 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
114 struct btrfs_trans_handle
*trans
;
118 u64 blocksize
= root
->sectorsize
;
120 struct btrfs_key ins
;
121 struct extent_map
*em
;
122 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
125 trans
= btrfs_join_transaction(root
, 1);
127 btrfs_set_trans_block_group(trans
, inode
);
129 num_bytes
= (end
- start
+ blocksize
) & ~(blocksize
- 1);
130 num_bytes
= max(blocksize
, num_bytes
);
131 orig_num_bytes
= num_bytes
;
133 if (alloc_hint
== EXTENT_MAP_INLINE
)
136 BUG_ON(num_bytes
> btrfs_super_total_bytes(&root
->fs_info
->super_copy
));
137 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
138 btrfs_drop_extent_cache(inode
, start
, start
+ num_bytes
- 1);
139 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
141 while(num_bytes
> 0) {
142 cur_alloc_size
= min(num_bytes
, root
->fs_info
->max_extent
);
143 ret
= btrfs_reserve_extent(trans
, root
, cur_alloc_size
,
144 root
->sectorsize
, 0, 0,
150 em
= alloc_extent_map(GFP_NOFS
);
152 em
->len
= ins
.offset
;
153 em
->block_start
= ins
.objectid
;
154 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
155 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
156 set_bit(EXTENT_FLAG_PINNED
, &em
->flags
);
158 spin_lock(&em_tree
->lock
);
159 ret
= add_extent_mapping(em_tree
, em
);
160 spin_unlock(&em_tree
->lock
);
161 if (ret
!= -EEXIST
) {
165 btrfs_drop_extent_cache(inode
, start
,
166 start
+ ins
.offset
- 1);
168 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
170 cur_alloc_size
= ins
.offset
;
171 ret
= btrfs_add_ordered_extent(inode
, start
, ins
.objectid
,
174 if (num_bytes
< cur_alloc_size
) {
175 printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes
,
179 num_bytes
-= cur_alloc_size
;
180 alloc_hint
= ins
.objectid
+ ins
.offset
;
181 start
+= cur_alloc_size
;
184 btrfs_end_transaction(trans
, root
);
188 static int run_delalloc_nocow(struct inode
*inode
, u64 start
, u64 end
)
195 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
196 struct btrfs_block_group_cache
*block_group
;
197 struct btrfs_trans_handle
*trans
;
198 struct extent_buffer
*leaf
;
200 struct btrfs_path
*path
;
201 struct btrfs_file_extent_item
*item
;
204 struct btrfs_key found_key
;
206 total_fs_bytes
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
207 path
= btrfs_alloc_path();
209 trans
= btrfs_join_transaction(root
, 1);
212 ret
= btrfs_lookup_file_extent(NULL
, root
, path
,
213 inode
->i_ino
, start
, 0);
220 if (path
->slots
[0] == 0)
225 leaf
= path
->nodes
[0];
226 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
227 struct btrfs_file_extent_item
);
229 /* are we inside the extent that was found? */
230 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
231 found_type
= btrfs_key_type(&found_key
);
232 if (found_key
.objectid
!= inode
->i_ino
||
233 found_type
!= BTRFS_EXTENT_DATA_KEY
)
236 found_type
= btrfs_file_extent_type(leaf
, item
);
237 extent_start
= found_key
.offset
;
238 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
239 u64 extent_num_bytes
;
241 extent_num_bytes
= btrfs_file_extent_num_bytes(leaf
, item
);
242 extent_end
= extent_start
+ extent_num_bytes
;
245 if (loops
&& start
!= extent_start
)
248 if (start
< extent_start
|| start
>= extent_end
)
251 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
255 if (btrfs_cross_ref_exists(trans
, root
, &found_key
, bytenr
))
258 * we may be called by the resizer, make sure we're inside
259 * the limits of the FS
261 block_group
= btrfs_lookup_block_group(root
->fs_info
,
263 if (!block_group
|| block_group
->ro
)
266 bytenr
+= btrfs_file_extent_offset(leaf
, item
);
267 extent_num_bytes
= min(end
+ 1, extent_end
) - start
;
268 ret
= btrfs_add_ordered_extent(inode
, start
, bytenr
,
269 extent_num_bytes
, 1);
275 btrfs_release_path(root
, path
);
283 btrfs_end_transaction(trans
, root
);
284 btrfs_free_path(path
);
285 return cow_file_range(inode
, start
, end
);
289 btrfs_end_transaction(trans
, root
);
290 btrfs_free_path(path
);
294 static int run_delalloc_range(struct inode
*inode
, u64 start
, u64 end
)
296 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
299 if (btrfs_test_opt(root
, NODATACOW
) ||
300 btrfs_test_flag(inode
, NODATACOW
))
301 ret
= run_delalloc_nocow(inode
, start
, end
);
303 ret
= cow_file_range(inode
, start
, end
);
308 int btrfs_set_bit_hook(struct inode
*inode
, u64 start
, u64 end
,
309 unsigned long old
, unsigned long bits
)
312 if (!(old
& EXTENT_DELALLOC
) && (bits
& EXTENT_DELALLOC
)) {
313 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
314 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
315 BTRFS_I(inode
)->delalloc_bytes
+= end
- start
+ 1;
316 root
->fs_info
->delalloc_bytes
+= end
- start
+ 1;
317 if (list_empty(&BTRFS_I(inode
)->delalloc_inodes
)) {
318 list_add_tail(&BTRFS_I(inode
)->delalloc_inodes
,
319 &root
->fs_info
->delalloc_inodes
);
321 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
326 int btrfs_clear_bit_hook(struct inode
*inode
, u64 start
, u64 end
,
327 unsigned long old
, unsigned long bits
)
329 if ((old
& EXTENT_DELALLOC
) && (bits
& EXTENT_DELALLOC
)) {
330 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
333 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
334 if (end
- start
+ 1 > root
->fs_info
->delalloc_bytes
) {
335 printk("warning: delalloc account %Lu %Lu\n",
336 end
- start
+ 1, root
->fs_info
->delalloc_bytes
);
337 root
->fs_info
->delalloc_bytes
= 0;
338 BTRFS_I(inode
)->delalloc_bytes
= 0;
340 root
->fs_info
->delalloc_bytes
-= end
- start
+ 1;
341 BTRFS_I(inode
)->delalloc_bytes
-= end
- start
+ 1;
343 if (BTRFS_I(inode
)->delalloc_bytes
== 0 &&
344 !list_empty(&BTRFS_I(inode
)->delalloc_inodes
)) {
345 list_del_init(&BTRFS_I(inode
)->delalloc_inodes
);
347 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
352 int btrfs_merge_bio_hook(struct page
*page
, unsigned long offset
,
353 size_t size
, struct bio
*bio
)
355 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
356 struct btrfs_mapping_tree
*map_tree
;
357 u64 logical
= bio
->bi_sector
<< 9;
362 length
= bio
->bi_size
;
363 map_tree
= &root
->fs_info
->mapping_tree
;
365 ret
= btrfs_map_block(map_tree
, READ
, logical
,
366 &map_length
, NULL
, 0);
368 if (map_length
< length
+ size
) {
374 int __btrfs_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
377 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
380 ret
= btrfs_csum_one_bio(root
, inode
, bio
);
383 return btrfs_map_bio(root
, rw
, bio
, mirror_num
, 1);
386 int btrfs_submit_bio_hook(struct inode
*inode
, int rw
, struct bio
*bio
,
389 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
392 ret
= btrfs_bio_wq_end_io(root
->fs_info
, bio
, 0);
395 if (btrfs_test_opt(root
, NODATASUM
) ||
396 btrfs_test_flag(inode
, NODATASUM
)) {
400 if (!(rw
& (1 << BIO_RW
))) {
401 btrfs_lookup_bio_sums(root
, inode
, bio
);
404 return btrfs_wq_submit_bio(BTRFS_I(inode
)->root
->fs_info
,
405 inode
, rw
, bio
, mirror_num
,
406 __btrfs_submit_bio_hook
);
408 return btrfs_map_bio(root
, rw
, bio
, mirror_num
, 0);
411 static noinline
int add_pending_csums(struct btrfs_trans_handle
*trans
,
412 struct inode
*inode
, u64 file_offset
,
413 struct list_head
*list
)
415 struct list_head
*cur
;
416 struct btrfs_ordered_sum
*sum
;
418 btrfs_set_trans_block_group(trans
, inode
);
419 list_for_each(cur
, list
) {
420 sum
= list_entry(cur
, struct btrfs_ordered_sum
, list
);
421 btrfs_csum_file_blocks(trans
, BTRFS_I(inode
)->root
,
427 int btrfs_set_extent_delalloc(struct inode
*inode
, u64 start
, u64 end
)
429 return set_extent_delalloc(&BTRFS_I(inode
)->io_tree
, start
, end
,
433 struct btrfs_writepage_fixup
{
435 struct btrfs_work work
;
438 /* see btrfs_writepage_start_hook for details on why this is required */
439 void btrfs_writepage_fixup_worker(struct btrfs_work
*work
)
441 struct btrfs_writepage_fixup
*fixup
;
442 struct btrfs_ordered_extent
*ordered
;
448 fixup
= container_of(work
, struct btrfs_writepage_fixup
, work
);
452 if (!page
->mapping
|| !PageDirty(page
) || !PageChecked(page
)) {
453 ClearPageChecked(page
);
457 inode
= page
->mapping
->host
;
458 page_start
= page_offset(page
);
459 page_end
= page_offset(page
) + PAGE_CACHE_SIZE
- 1;
461 lock_extent(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
, GFP_NOFS
);
463 /* already ordered? We're done */
464 if (test_range_bit(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
,
465 EXTENT_ORDERED
, 0)) {
469 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
471 unlock_extent(&BTRFS_I(inode
)->io_tree
, page_start
,
474 btrfs_start_ordered_extent(inode
, ordered
, 1);
478 btrfs_set_extent_delalloc(inode
, page_start
, page_end
);
479 ClearPageChecked(page
);
481 unlock_extent(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
, GFP_NOFS
);
484 page_cache_release(page
);
488 * There are a few paths in the higher layers of the kernel that directly
489 * set the page dirty bit without asking the filesystem if it is a
490 * good idea. This causes problems because we want to make sure COW
491 * properly happens and the data=ordered rules are followed.
493 * In our case any range that doesn't have the EXTENT_ORDERED bit set
494 * hasn't been properly setup for IO. We kick off an async process
495 * to fix it up. The async helper will wait for ordered extents, set
496 * the delalloc bit and make it safe to write the page.
498 int btrfs_writepage_start_hook(struct page
*page
, u64 start
, u64 end
)
500 struct inode
*inode
= page
->mapping
->host
;
501 struct btrfs_writepage_fixup
*fixup
;
502 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
505 ret
= test_range_bit(&BTRFS_I(inode
)->io_tree
, start
, end
,
510 if (PageChecked(page
))
513 fixup
= kzalloc(sizeof(*fixup
), GFP_NOFS
);
517 SetPageChecked(page
);
518 page_cache_get(page
);
519 fixup
->work
.func
= btrfs_writepage_fixup_worker
;
521 btrfs_queue_worker(&root
->fs_info
->fixup_workers
, &fixup
->work
);
525 static int btrfs_finish_ordered_io(struct inode
*inode
, u64 start
, u64 end
)
527 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
528 struct btrfs_trans_handle
*trans
;
529 struct btrfs_ordered_extent
*ordered_extent
;
530 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
532 struct list_head list
;
533 struct btrfs_key ins
;
536 ret
= btrfs_dec_test_ordered_pending(inode
, start
, end
- start
+ 1);
540 trans
= btrfs_join_transaction(root
, 1);
542 ordered_extent
= btrfs_lookup_ordered_extent(inode
, start
);
543 BUG_ON(!ordered_extent
);
544 if (test_bit(BTRFS_ORDERED_NOCOW
, &ordered_extent
->flags
))
547 lock_extent(io_tree
, ordered_extent
->file_offset
,
548 ordered_extent
->file_offset
+ ordered_extent
->len
- 1,
551 INIT_LIST_HEAD(&list
);
553 ins
.objectid
= ordered_extent
->start
;
554 ins
.offset
= ordered_extent
->len
;
555 ins
.type
= BTRFS_EXTENT_ITEM_KEY
;
557 ret
= btrfs_alloc_reserved_extent(trans
, root
, root
->root_key
.objectid
,
558 trans
->transid
, inode
->i_ino
,
559 ordered_extent
->file_offset
, &ins
);
562 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
564 ret
= btrfs_drop_extents(trans
, root
, inode
,
565 ordered_extent
->file_offset
,
566 ordered_extent
->file_offset
+
568 ordered_extent
->file_offset
, &alloc_hint
);
570 ret
= btrfs_insert_file_extent(trans
, root
, inode
->i_ino
,
571 ordered_extent
->file_offset
,
572 ordered_extent
->start
,
574 ordered_extent
->len
, 0);
577 btrfs_drop_extent_cache(inode
, ordered_extent
->file_offset
,
578 ordered_extent
->file_offset
+
579 ordered_extent
->len
- 1);
580 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
582 inode
->i_blocks
+= ordered_extent
->len
>> 9;
583 unlock_extent(io_tree
, ordered_extent
->file_offset
,
584 ordered_extent
->file_offset
+ ordered_extent
->len
- 1,
587 add_pending_csums(trans
, inode
, ordered_extent
->file_offset
,
588 &ordered_extent
->list
);
590 btrfs_ordered_update_i_size(inode
, ordered_extent
);
591 btrfs_update_inode(trans
, root
, inode
);
592 btrfs_remove_ordered_extent(inode
, ordered_extent
);
595 btrfs_put_ordered_extent(ordered_extent
);
596 /* once for the tree */
597 btrfs_put_ordered_extent(ordered_extent
);
599 btrfs_end_transaction(trans
, root
);
603 int btrfs_writepage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
604 struct extent_state
*state
, int uptodate
)
606 return btrfs_finish_ordered_io(page
->mapping
->host
, start
, end
);
609 struct io_failure_record
{
617 int btrfs_io_failed_hook(struct bio
*failed_bio
,
618 struct page
*page
, u64 start
, u64 end
,
619 struct extent_state
*state
)
621 struct io_failure_record
*failrec
= NULL
;
623 struct extent_map
*em
;
624 struct inode
*inode
= page
->mapping
->host
;
625 struct extent_io_tree
*failure_tree
= &BTRFS_I(inode
)->io_failure_tree
;
626 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
633 ret
= get_state_private(failure_tree
, start
, &private);
635 failrec
= kmalloc(sizeof(*failrec
), GFP_NOFS
);
638 failrec
->start
= start
;
639 failrec
->len
= end
- start
+ 1;
640 failrec
->last_mirror
= 0;
642 spin_lock(&em_tree
->lock
);
643 em
= lookup_extent_mapping(em_tree
, start
, failrec
->len
);
644 if (em
->start
> start
|| em
->start
+ em
->len
< start
) {
648 spin_unlock(&em_tree
->lock
);
650 if (!em
|| IS_ERR(em
)) {
654 logical
= start
- em
->start
;
655 logical
= em
->block_start
+ logical
;
656 failrec
->logical
= logical
;
658 set_extent_bits(failure_tree
, start
, end
, EXTENT_LOCKED
|
659 EXTENT_DIRTY
, GFP_NOFS
);
660 set_state_private(failure_tree
, start
,
661 (u64
)(unsigned long)failrec
);
663 failrec
= (struct io_failure_record
*)(unsigned long)private;
665 num_copies
= btrfs_num_copies(
666 &BTRFS_I(inode
)->root
->fs_info
->mapping_tree
,
667 failrec
->logical
, failrec
->len
);
668 failrec
->last_mirror
++;
670 spin_lock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
671 state
= find_first_extent_bit_state(&BTRFS_I(inode
)->io_tree
,
674 if (state
&& state
->start
!= failrec
->start
)
676 spin_unlock_irq(&BTRFS_I(inode
)->io_tree
.lock
);
678 if (!state
|| failrec
->last_mirror
> num_copies
) {
679 set_state_private(failure_tree
, failrec
->start
, 0);
680 clear_extent_bits(failure_tree
, failrec
->start
,
681 failrec
->start
+ failrec
->len
- 1,
682 EXTENT_LOCKED
| EXTENT_DIRTY
, GFP_NOFS
);
686 bio
= bio_alloc(GFP_NOFS
, 1);
687 bio
->bi_private
= state
;
688 bio
->bi_end_io
= failed_bio
->bi_end_io
;
689 bio
->bi_sector
= failrec
->logical
>> 9;
690 bio
->bi_bdev
= failed_bio
->bi_bdev
;
692 bio_add_page(bio
, page
, failrec
->len
, start
- page_offset(page
));
693 if (failed_bio
->bi_rw
& (1 << BIO_RW
))
698 BTRFS_I(inode
)->io_tree
.ops
->submit_bio_hook(inode
, rw
, bio
,
699 failrec
->last_mirror
);
703 int btrfs_clean_io_failures(struct inode
*inode
, u64 start
)
707 struct io_failure_record
*failure
;
711 if (count_range_bits(&BTRFS_I(inode
)->io_failure_tree
, &private,
712 (u64
)-1, 1, EXTENT_DIRTY
)) {
713 ret
= get_state_private(&BTRFS_I(inode
)->io_failure_tree
,
714 start
, &private_failure
);
716 failure
= (struct io_failure_record
*)(unsigned long)
718 set_state_private(&BTRFS_I(inode
)->io_failure_tree
,
720 clear_extent_bits(&BTRFS_I(inode
)->io_failure_tree
,
722 failure
->start
+ failure
->len
- 1,
723 EXTENT_DIRTY
| EXTENT_LOCKED
,
731 int btrfs_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
,
732 struct extent_state
*state
)
734 size_t offset
= start
- ((u64
)page
->index
<< PAGE_CACHE_SHIFT
);
735 struct inode
*inode
= page
->mapping
->host
;
736 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
738 u64
private = ~(u32
)0;
740 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
744 if (btrfs_test_opt(root
, NODATASUM
) ||
745 btrfs_test_flag(inode
, NODATASUM
))
747 if (state
&& state
->start
== start
) {
748 private = state
->private;
751 ret
= get_state_private(io_tree
, start
, &private);
753 local_irq_save(flags
);
754 kaddr
= kmap_atomic(page
, KM_IRQ0
);
758 csum
= btrfs_csum_data(root
, kaddr
+ offset
, csum
, end
- start
+ 1);
759 btrfs_csum_final(csum
, (char *)&csum
);
760 if (csum
!= private) {
763 kunmap_atomic(kaddr
, KM_IRQ0
);
764 local_irq_restore(flags
);
766 /* if the io failure tree for this inode is non-empty,
767 * check to see if we've recovered from a failed IO
769 btrfs_clean_io_failures(inode
, start
);
773 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
774 page
->mapping
->host
->i_ino
, (unsigned long long)start
, csum
,
776 memset(kaddr
+ offset
, 1, end
- start
+ 1);
777 flush_dcache_page(page
);
778 kunmap_atomic(kaddr
, KM_IRQ0
);
779 local_irq_restore(flags
);
786 * This creates an orphan entry for the given inode in case something goes
787 * wrong in the middle of an unlink/truncate.
789 int btrfs_orphan_add(struct btrfs_trans_handle
*trans
, struct inode
*inode
)
791 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
794 spin_lock(&root
->list_lock
);
796 /* already on the orphan list, we're good */
797 if (!list_empty(&BTRFS_I(inode
)->i_orphan
)) {
798 spin_unlock(&root
->list_lock
);
802 list_add(&BTRFS_I(inode
)->i_orphan
, &root
->orphan_list
);
804 spin_unlock(&root
->list_lock
);
807 * insert an orphan item to track this unlinked/truncated file
809 ret
= btrfs_insert_orphan_item(trans
, root
, inode
->i_ino
);
815 * We have done the truncate/delete so we can go ahead and remove the orphan
816 * item for this particular inode.
818 int btrfs_orphan_del(struct btrfs_trans_handle
*trans
, struct inode
*inode
)
820 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
823 spin_lock(&root
->list_lock
);
825 if (list_empty(&BTRFS_I(inode
)->i_orphan
)) {
826 spin_unlock(&root
->list_lock
);
830 list_del_init(&BTRFS_I(inode
)->i_orphan
);
832 spin_unlock(&root
->list_lock
);
836 spin_unlock(&root
->list_lock
);
838 ret
= btrfs_del_orphan_item(trans
, root
, inode
->i_ino
);
844 * this cleans up any orphans that may be left on the list from the last use
847 void btrfs_orphan_cleanup(struct btrfs_root
*root
)
849 struct btrfs_path
*path
;
850 struct extent_buffer
*leaf
;
851 struct btrfs_item
*item
;
852 struct btrfs_key key
, found_key
;
853 struct btrfs_trans_handle
*trans
;
855 int ret
= 0, nr_unlink
= 0, nr_truncate
= 0;
857 /* don't do orphan cleanup if the fs is readonly. */
858 if (root
->inode
->i_sb
->s_flags
& MS_RDONLY
)
861 path
= btrfs_alloc_path();
866 key
.objectid
= BTRFS_ORPHAN_OBJECTID
;
867 btrfs_set_key_type(&key
, BTRFS_ORPHAN_ITEM_KEY
);
868 key
.offset
= (u64
)-1;
870 trans
= btrfs_start_transaction(root
, 1);
871 btrfs_set_trans_block_group(trans
, root
->inode
);
874 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
876 printk(KERN_ERR
"Error searching slot for orphan: %d"
882 * if ret == 0 means we found what we were searching for, which
883 * is weird, but possible, so only screw with path if we didnt
884 * find the key and see if we have stuff that matches
887 if (path
->slots
[0] == 0)
892 /* pull out the item */
893 leaf
= path
->nodes
[0];
894 item
= btrfs_item_nr(leaf
, path
->slots
[0]);
895 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
897 /* make sure the item matches what we want */
898 if (found_key
.objectid
!= BTRFS_ORPHAN_OBJECTID
)
900 if (btrfs_key_type(&found_key
) != BTRFS_ORPHAN_ITEM_KEY
)
903 /* release the path since we're done with it */
904 btrfs_release_path(root
, path
);
907 * this is where we are basically btrfs_lookup, without the
908 * crossing root thing. we store the inode number in the
909 * offset of the orphan item.
911 inode
= btrfs_iget_locked(root
->inode
->i_sb
,
912 found_key
.offset
, root
);
916 if (inode
->i_state
& I_NEW
) {
917 BTRFS_I(inode
)->root
= root
;
919 /* have to set the location manually */
920 BTRFS_I(inode
)->location
.objectid
= inode
->i_ino
;
921 BTRFS_I(inode
)->location
.type
= BTRFS_INODE_ITEM_KEY
;
922 BTRFS_I(inode
)->location
.offset
= 0;
924 btrfs_read_locked_inode(inode
);
925 unlock_new_inode(inode
);
929 * add this inode to the orphan list so btrfs_orphan_del does
930 * the proper thing when we hit it
932 spin_lock(&root
->list_lock
);
933 list_add(&BTRFS_I(inode
)->i_orphan
, &root
->orphan_list
);
934 spin_unlock(&root
->list_lock
);
937 * if this is a bad inode, means we actually succeeded in
938 * removing the inode, but not the orphan record, which means
939 * we need to manually delete the orphan since iput will just
942 if (is_bad_inode(inode
)) {
943 btrfs_orphan_del(trans
, inode
);
948 /* if we have links, this was a truncate, lets do that */
949 if (inode
->i_nlink
) {
951 btrfs_truncate(inode
);
956 /* this will do delete_inode and everything for us */
961 printk(KERN_INFO
"btrfs: unlinked %d orphans\n", nr_unlink
);
963 printk(KERN_INFO
"btrfs: truncated %d orphans\n", nr_truncate
);
965 btrfs_free_path(path
);
966 btrfs_end_transaction(trans
, root
);
969 void btrfs_read_locked_inode(struct inode
*inode
)
971 struct btrfs_path
*path
;
972 struct extent_buffer
*leaf
;
973 struct btrfs_inode_item
*inode_item
;
974 struct btrfs_timespec
*tspec
;
975 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
976 struct btrfs_key location
;
977 u64 alloc_group_block
;
981 path
= btrfs_alloc_path();
983 memcpy(&location
, &BTRFS_I(inode
)->location
, sizeof(location
));
985 ret
= btrfs_lookup_inode(NULL
, root
, path
, &location
, 0);
989 leaf
= path
->nodes
[0];
990 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
991 struct btrfs_inode_item
);
993 inode
->i_mode
= btrfs_inode_mode(leaf
, inode_item
);
994 inode
->i_nlink
= btrfs_inode_nlink(leaf
, inode_item
);
995 inode
->i_uid
= btrfs_inode_uid(leaf
, inode_item
);
996 inode
->i_gid
= btrfs_inode_gid(leaf
, inode_item
);
997 btrfs_i_size_write(inode
, btrfs_inode_size(leaf
, inode_item
));
999 tspec
= btrfs_inode_atime(inode_item
);
1000 inode
->i_atime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1001 inode
->i_atime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1003 tspec
= btrfs_inode_mtime(inode_item
);
1004 inode
->i_mtime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1005 inode
->i_mtime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1007 tspec
= btrfs_inode_ctime(inode_item
);
1008 inode
->i_ctime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
1009 inode
->i_ctime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
1011 inode
->i_blocks
= btrfs_inode_nblocks(leaf
, inode_item
);
1012 BTRFS_I(inode
)->generation
= btrfs_inode_generation(leaf
, inode_item
);
1013 inode
->i_generation
= BTRFS_I(inode
)->generation
;
1015 rdev
= btrfs_inode_rdev(leaf
, inode_item
);
1017 BTRFS_I(inode
)->index_cnt
= (u64
)-1;
1019 alloc_group_block
= btrfs_inode_block_group(leaf
, inode_item
);
1020 BTRFS_I(inode
)->block_group
= btrfs_lookup_block_group(root
->fs_info
,
1022 BTRFS_I(inode
)->flags
= btrfs_inode_flags(leaf
, inode_item
);
1023 if (!BTRFS_I(inode
)->block_group
) {
1024 BTRFS_I(inode
)->block_group
= btrfs_find_block_group(root
,
1026 BTRFS_BLOCK_GROUP_METADATA
, 0);
1028 btrfs_free_path(path
);
1031 switch (inode
->i_mode
& S_IFMT
) {
1033 inode
->i_mapping
->a_ops
= &btrfs_aops
;
1034 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1035 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
1036 inode
->i_fop
= &btrfs_file_operations
;
1037 inode
->i_op
= &btrfs_file_inode_operations
;
1040 inode
->i_fop
= &btrfs_dir_file_operations
;
1041 if (root
== root
->fs_info
->tree_root
)
1042 inode
->i_op
= &btrfs_dir_ro_inode_operations
;
1044 inode
->i_op
= &btrfs_dir_inode_operations
;
1047 inode
->i_op
= &btrfs_symlink_inode_operations
;
1048 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
1049 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
1052 init_special_inode(inode
, inode
->i_mode
, rdev
);
1058 btrfs_free_path(path
);
1059 make_bad_inode(inode
);
1062 static void fill_inode_item(struct btrfs_trans_handle
*trans
,
1063 struct extent_buffer
*leaf
,
1064 struct btrfs_inode_item
*item
,
1065 struct inode
*inode
)
1067 btrfs_set_inode_uid(leaf
, item
, inode
->i_uid
);
1068 btrfs_set_inode_gid(leaf
, item
, inode
->i_gid
);
1069 btrfs_set_inode_size(leaf
, item
, BTRFS_I(inode
)->disk_i_size
);
1070 btrfs_set_inode_mode(leaf
, item
, inode
->i_mode
);
1071 btrfs_set_inode_nlink(leaf
, item
, inode
->i_nlink
);
1073 btrfs_set_timespec_sec(leaf
, btrfs_inode_atime(item
),
1074 inode
->i_atime
.tv_sec
);
1075 btrfs_set_timespec_nsec(leaf
, btrfs_inode_atime(item
),
1076 inode
->i_atime
.tv_nsec
);
1078 btrfs_set_timespec_sec(leaf
, btrfs_inode_mtime(item
),
1079 inode
->i_mtime
.tv_sec
);
1080 btrfs_set_timespec_nsec(leaf
, btrfs_inode_mtime(item
),
1081 inode
->i_mtime
.tv_nsec
);
1083 btrfs_set_timespec_sec(leaf
, btrfs_inode_ctime(item
),
1084 inode
->i_ctime
.tv_sec
);
1085 btrfs_set_timespec_nsec(leaf
, btrfs_inode_ctime(item
),
1086 inode
->i_ctime
.tv_nsec
);
1088 btrfs_set_inode_nblocks(leaf
, item
, inode
->i_blocks
);
1089 btrfs_set_inode_generation(leaf
, item
, BTRFS_I(inode
)->generation
);
1090 btrfs_set_inode_transid(leaf
, item
, trans
->transid
);
1091 btrfs_set_inode_rdev(leaf
, item
, inode
->i_rdev
);
1092 btrfs_set_inode_flags(leaf
, item
, BTRFS_I(inode
)->flags
);
1093 btrfs_set_inode_block_group(leaf
, item
,
1094 BTRFS_I(inode
)->block_group
->key
.objectid
);
1097 int noinline
btrfs_update_inode(struct btrfs_trans_handle
*trans
,
1098 struct btrfs_root
*root
,
1099 struct inode
*inode
)
1101 struct btrfs_inode_item
*inode_item
;
1102 struct btrfs_path
*path
;
1103 struct extent_buffer
*leaf
;
1106 path
= btrfs_alloc_path();
1108 ret
= btrfs_lookup_inode(trans
, root
, path
,
1109 &BTRFS_I(inode
)->location
, 1);
1116 leaf
= path
->nodes
[0];
1117 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
1118 struct btrfs_inode_item
);
1120 fill_inode_item(trans
, leaf
, inode_item
, inode
);
1121 btrfs_mark_buffer_dirty(leaf
);
1122 btrfs_set_inode_last_trans(trans
, inode
);
1125 btrfs_free_path(path
);
1130 int btrfs_unlink_inode(struct btrfs_trans_handle
*trans
,
1131 struct btrfs_root
*root
,
1132 struct inode
*dir
, struct inode
*inode
,
1133 const char *name
, int name_len
)
1135 struct btrfs_path
*path
;
1137 struct extent_buffer
*leaf
;
1138 struct btrfs_dir_item
*di
;
1139 struct btrfs_key key
;
1142 path
= btrfs_alloc_path();
1148 di
= btrfs_lookup_dir_item(trans
, root
, path
, dir
->i_ino
,
1149 name
, name_len
, -1);
1158 leaf
= path
->nodes
[0];
1159 btrfs_dir_item_key_to_cpu(leaf
, di
, &key
);
1160 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
1163 btrfs_release_path(root
, path
);
1165 ret
= btrfs_del_inode_ref(trans
, root
, name
, name_len
,
1167 dir
->i_ino
, &index
);
1169 printk("failed to delete reference to %.*s, "
1170 "inode %lu parent %lu\n", name_len
, name
,
1171 inode
->i_ino
, dir
->i_ino
);
1175 di
= btrfs_lookup_dir_index_item(trans
, root
, path
, dir
->i_ino
,
1176 index
, name
, name_len
, -1);
1185 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
1186 btrfs_release_path(root
, path
);
1188 ret
= btrfs_del_inode_ref_in_log(trans
, root
, name
, name_len
,
1192 ret
= btrfs_del_dir_entries_in_log(trans
, root
, name
, name_len
,
1196 btrfs_free_path(path
);
1200 btrfs_i_size_write(dir
, dir
->i_size
- name_len
* 2);
1201 inode
->i_ctime
= dir
->i_mtime
= dir
->i_ctime
= CURRENT_TIME
;
1202 btrfs_update_inode(trans
, root
, dir
);
1203 btrfs_drop_nlink(inode
);
1204 ret
= btrfs_update_inode(trans
, root
, inode
);
1205 dir
->i_sb
->s_dirt
= 1;
1210 static int btrfs_unlink(struct inode
*dir
, struct dentry
*dentry
)
1212 struct btrfs_root
*root
;
1213 struct btrfs_trans_handle
*trans
;
1214 struct inode
*inode
= dentry
->d_inode
;
1216 unsigned long nr
= 0;
1218 root
= BTRFS_I(dir
)->root
;
1220 ret
= btrfs_check_free_space(root
, 1, 1);
1224 trans
= btrfs_start_transaction(root
, 1);
1226 btrfs_set_trans_block_group(trans
, dir
);
1227 ret
= btrfs_unlink_inode(trans
, root
, dir
, dentry
->d_inode
,
1228 dentry
->d_name
.name
, dentry
->d_name
.len
);
1230 if (inode
->i_nlink
== 0)
1231 ret
= btrfs_orphan_add(trans
, inode
);
1233 nr
= trans
->blocks_used
;
1235 btrfs_end_transaction_throttle(trans
, root
);
1237 btrfs_btree_balance_dirty(root
, nr
);
1241 static int btrfs_rmdir(struct inode
*dir
, struct dentry
*dentry
)
1243 struct inode
*inode
= dentry
->d_inode
;
1246 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1247 struct btrfs_trans_handle
*trans
;
1248 unsigned long nr
= 0;
1250 if (inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
1254 ret
= btrfs_check_free_space(root
, 1, 1);
1258 trans
= btrfs_start_transaction(root
, 1);
1259 btrfs_set_trans_block_group(trans
, dir
);
1261 err
= btrfs_orphan_add(trans
, inode
);
1265 /* now the directory is empty */
1266 err
= btrfs_unlink_inode(trans
, root
, dir
, dentry
->d_inode
,
1267 dentry
->d_name
.name
, dentry
->d_name
.len
);
1269 btrfs_i_size_write(inode
, 0);
1273 nr
= trans
->blocks_used
;
1274 ret
= btrfs_end_transaction_throttle(trans
, root
);
1276 btrfs_btree_balance_dirty(root
, nr
);
1284 * this can truncate away extent items, csum items and directory items.
1285 * It starts at a high offset and removes keys until it can't find
1286 * any higher than i_size.
1288 * csum items that cross the new i_size are truncated to the new size
1291 * min_type is the minimum key type to truncate down to. If set to 0, this
1292 * will kill all the items on this inode, including the INODE_ITEM_KEY.
1294 noinline
int btrfs_truncate_inode_items(struct btrfs_trans_handle
*trans
,
1295 struct btrfs_root
*root
,
1296 struct inode
*inode
,
1297 u64 new_size
, u32 min_type
)
1300 struct btrfs_path
*path
;
1301 struct btrfs_key key
;
1302 struct btrfs_key found_key
;
1304 struct extent_buffer
*leaf
;
1305 struct btrfs_file_extent_item
*fi
;
1306 u64 extent_start
= 0;
1307 u64 extent_num_bytes
= 0;
1313 int pending_del_nr
= 0;
1314 int pending_del_slot
= 0;
1315 int extent_type
= -1;
1316 u64 mask
= root
->sectorsize
- 1;
1319 btrfs_drop_extent_cache(inode
,
1320 new_size
& (~mask
), (u64
)-1);
1321 path
= btrfs_alloc_path();
1325 /* FIXME, add redo link to tree so we don't leak on crash */
1326 key
.objectid
= inode
->i_ino
;
1327 key
.offset
= (u64
)-1;
1330 btrfs_init_path(path
);
1332 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
1337 /* there are no items in the tree for us to truncate, we're
1340 if (path
->slots
[0] == 0) {
1349 leaf
= path
->nodes
[0];
1350 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
1351 found_type
= btrfs_key_type(&found_key
);
1353 if (found_key
.objectid
!= inode
->i_ino
)
1356 if (found_type
< min_type
)
1359 item_end
= found_key
.offset
;
1360 if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
1361 fi
= btrfs_item_ptr(leaf
, path
->slots
[0],
1362 struct btrfs_file_extent_item
);
1363 extent_type
= btrfs_file_extent_type(leaf
, fi
);
1364 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1366 btrfs_file_extent_num_bytes(leaf
, fi
);
1367 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1368 struct btrfs_item
*item
= btrfs_item_nr(leaf
,
1370 item_end
+= btrfs_file_extent_inline_len(leaf
,
1375 if (found_type
== BTRFS_CSUM_ITEM_KEY
) {
1376 ret
= btrfs_csum_truncate(trans
, root
, path
,
1380 if (item_end
< new_size
) {
1381 if (found_type
== BTRFS_DIR_ITEM_KEY
) {
1382 found_type
= BTRFS_INODE_ITEM_KEY
;
1383 } else if (found_type
== BTRFS_EXTENT_ITEM_KEY
) {
1384 found_type
= BTRFS_CSUM_ITEM_KEY
;
1385 } else if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
1386 found_type
= BTRFS_XATTR_ITEM_KEY
;
1387 } else if (found_type
== BTRFS_XATTR_ITEM_KEY
) {
1388 found_type
= BTRFS_INODE_REF_KEY
;
1389 } else if (found_type
) {
1394 btrfs_set_key_type(&key
, found_type
);
1397 if (found_key
.offset
>= new_size
)
1403 /* FIXME, shrink the extent if the ref count is only 1 */
1404 if (found_type
!= BTRFS_EXTENT_DATA_KEY
)
1407 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
1409 extent_start
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
1411 u64 orig_num_bytes
=
1412 btrfs_file_extent_num_bytes(leaf
, fi
);
1413 extent_num_bytes
= new_size
-
1414 found_key
.offset
+ root
->sectorsize
- 1;
1415 extent_num_bytes
= extent_num_bytes
&
1416 ~((u64
)root
->sectorsize
- 1);
1417 btrfs_set_file_extent_num_bytes(leaf
, fi
,
1419 num_dec
= (orig_num_bytes
-
1421 if (root
->ref_cows
&& extent_start
!= 0)
1422 dec_i_blocks(inode
, num_dec
);
1423 btrfs_mark_buffer_dirty(leaf
);
1426 btrfs_file_extent_disk_num_bytes(leaf
,
1428 /* FIXME blocksize != 4096 */
1429 num_dec
= btrfs_file_extent_num_bytes(leaf
, fi
);
1430 if (extent_start
!= 0) {
1433 dec_i_blocks(inode
, num_dec
);
1435 if (root
->ref_cows
) {
1437 btrfs_header_generation(leaf
);
1439 root_owner
= btrfs_header_owner(leaf
);
1441 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
1443 u32 size
= new_size
- found_key
.offset
;
1445 if (root
->ref_cows
) {
1446 dec_i_blocks(inode
, item_end
+ 1 -
1447 found_key
.offset
- size
);
1450 btrfs_file_extent_calc_inline_size(size
);
1451 ret
= btrfs_truncate_item(trans
, root
, path
,
1454 } else if (root
->ref_cows
) {
1455 dec_i_blocks(inode
, item_end
+ 1 -
1461 if (!pending_del_nr
) {
1462 /* no pending yet, add ourselves */
1463 pending_del_slot
= path
->slots
[0];
1465 } else if (pending_del_nr
&&
1466 path
->slots
[0] + 1 == pending_del_slot
) {
1467 /* hop on the pending chunk */
1469 pending_del_slot
= path
->slots
[0];
1471 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path
->slots
[0], pending_del_nr
, pending_del_slot
);
1477 ret
= btrfs_free_extent(trans
, root
, extent_start
,
1480 root_gen
, inode
->i_ino
,
1481 found_key
.offset
, 0);
1485 if (path
->slots
[0] == 0) {
1488 btrfs_release_path(root
, path
);
1493 if (pending_del_nr
&&
1494 path
->slots
[0] + 1 != pending_del_slot
) {
1495 struct btrfs_key debug
;
1497 btrfs_item_key_to_cpu(path
->nodes
[0], &debug
,
1499 ret
= btrfs_del_items(trans
, root
, path
,
1504 btrfs_release_path(root
, path
);
1510 if (pending_del_nr
) {
1511 ret
= btrfs_del_items(trans
, root
, path
, pending_del_slot
,
1514 btrfs_free_path(path
);
1515 inode
->i_sb
->s_dirt
= 1;
1520 * taken from block_truncate_page, but does cow as it zeros out
1521 * any bytes left in the last page in the file.
1523 static int btrfs_truncate_page(struct address_space
*mapping
, loff_t from
)
1525 struct inode
*inode
= mapping
->host
;
1526 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1527 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1528 struct btrfs_ordered_extent
*ordered
;
1530 u32 blocksize
= root
->sectorsize
;
1531 pgoff_t index
= from
>> PAGE_CACHE_SHIFT
;
1532 unsigned offset
= from
& (PAGE_CACHE_SIZE
-1);
1538 if ((offset
& (blocksize
- 1)) == 0)
1543 page
= grab_cache_page(mapping
, index
);
1547 page_start
= page_offset(page
);
1548 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1550 if (!PageUptodate(page
)) {
1551 ret
= btrfs_readpage(NULL
, page
);
1553 if (page
->mapping
!= mapping
) {
1555 page_cache_release(page
);
1558 if (!PageUptodate(page
)) {
1563 wait_on_page_writeback(page
);
1565 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1566 set_page_extent_mapped(page
);
1568 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
1570 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1572 page_cache_release(page
);
1573 btrfs_start_ordered_extent(inode
, ordered
, 1);
1574 btrfs_put_ordered_extent(ordered
);
1578 btrfs_set_extent_delalloc(inode
, page_start
, page_end
);
1580 if (offset
!= PAGE_CACHE_SIZE
) {
1582 memset(kaddr
+ offset
, 0, PAGE_CACHE_SIZE
- offset
);
1583 flush_dcache_page(page
);
1586 ClearPageChecked(page
);
1587 set_page_dirty(page
);
1588 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
1592 page_cache_release(page
);
1597 static int btrfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
1599 struct inode
*inode
= dentry
->d_inode
;
1602 err
= inode_change_ok(inode
, attr
);
1606 if (S_ISREG(inode
->i_mode
) &&
1607 attr
->ia_valid
& ATTR_SIZE
&& attr
->ia_size
> inode
->i_size
) {
1608 struct btrfs_trans_handle
*trans
;
1609 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1610 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1612 u64 mask
= root
->sectorsize
- 1;
1613 u64 hole_start
= (inode
->i_size
+ mask
) & ~mask
;
1614 u64 block_end
= (attr
->ia_size
+ mask
) & ~mask
;
1618 if (attr
->ia_size
<= hole_start
)
1621 err
= btrfs_check_free_space(root
, 1, 0);
1625 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
1627 hole_size
= block_end
- hole_start
;
1629 struct btrfs_ordered_extent
*ordered
;
1630 btrfs_wait_ordered_range(inode
, hole_start
, hole_size
);
1632 lock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1633 ordered
= btrfs_lookup_ordered_extent(inode
, hole_start
);
1635 unlock_extent(io_tree
, hole_start
,
1636 block_end
- 1, GFP_NOFS
);
1637 btrfs_put_ordered_extent(ordered
);
1643 trans
= btrfs_start_transaction(root
, 1);
1644 btrfs_set_trans_block_group(trans
, inode
);
1645 mutex_lock(&BTRFS_I(inode
)->extent_mutex
);
1646 err
= btrfs_drop_extents(trans
, root
, inode
,
1647 hole_start
, block_end
, hole_start
,
1650 if (alloc_hint
!= EXTENT_MAP_INLINE
) {
1651 err
= btrfs_insert_file_extent(trans
, root
,
1655 btrfs_drop_extent_cache(inode
, hole_start
,
1657 btrfs_check_file(root
, inode
);
1659 mutex_unlock(&BTRFS_I(inode
)->extent_mutex
);
1660 btrfs_end_transaction(trans
, root
);
1661 unlock_extent(io_tree
, hole_start
, block_end
- 1, GFP_NOFS
);
1666 err
= inode_setattr(inode
, attr
);
1668 if (!err
&& ((attr
->ia_valid
& ATTR_MODE
)))
1669 err
= btrfs_acl_chmod(inode
);
1674 void btrfs_delete_inode(struct inode
*inode
)
1676 struct btrfs_trans_handle
*trans
;
1677 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1681 truncate_inode_pages(&inode
->i_data
, 0);
1682 if (is_bad_inode(inode
)) {
1683 btrfs_orphan_del(NULL
, inode
);
1686 btrfs_wait_ordered_range(inode
, 0, (u64
)-1);
1688 btrfs_i_size_write(inode
, 0);
1689 trans
= btrfs_start_transaction(root
, 1);
1691 btrfs_set_trans_block_group(trans
, inode
);
1692 ret
= btrfs_truncate_inode_items(trans
, root
, inode
, inode
->i_size
, 0);
1694 btrfs_orphan_del(NULL
, inode
);
1695 goto no_delete_lock
;
1698 btrfs_orphan_del(trans
, inode
);
1700 nr
= trans
->blocks_used
;
1703 btrfs_end_transaction(trans
, root
);
1704 btrfs_btree_balance_dirty(root
, nr
);
1708 nr
= trans
->blocks_used
;
1709 btrfs_end_transaction(trans
, root
);
1710 btrfs_btree_balance_dirty(root
, nr
);
1716 * this returns the key found in the dir entry in the location pointer.
1717 * If no dir entries were found, location->objectid is 0.
1719 static int btrfs_inode_by_name(struct inode
*dir
, struct dentry
*dentry
,
1720 struct btrfs_key
*location
)
1722 const char *name
= dentry
->d_name
.name
;
1723 int namelen
= dentry
->d_name
.len
;
1724 struct btrfs_dir_item
*di
;
1725 struct btrfs_path
*path
;
1726 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1729 path
= btrfs_alloc_path();
1732 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir
->i_ino
, name
,
1736 if (!di
|| IS_ERR(di
)) {
1739 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, location
);
1741 btrfs_free_path(path
);
1744 location
->objectid
= 0;
1749 * when we hit a tree root in a directory, the btrfs part of the inode
1750 * needs to be changed to reflect the root directory of the tree root. This
1751 * is kind of like crossing a mount point.
1753 static int fixup_tree_root_location(struct btrfs_root
*root
,
1754 struct btrfs_key
*location
,
1755 struct btrfs_root
**sub_root
,
1756 struct dentry
*dentry
)
1758 struct btrfs_root_item
*ri
;
1760 if (btrfs_key_type(location
) != BTRFS_ROOT_ITEM_KEY
)
1762 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1765 *sub_root
= btrfs_read_fs_root(root
->fs_info
, location
,
1766 dentry
->d_name
.name
,
1767 dentry
->d_name
.len
);
1768 if (IS_ERR(*sub_root
))
1769 return PTR_ERR(*sub_root
);
1771 ri
= &(*sub_root
)->root_item
;
1772 location
->objectid
= btrfs_root_dirid(ri
);
1773 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1774 location
->offset
= 0;
1779 static noinline
void init_btrfs_i(struct inode
*inode
)
1781 struct btrfs_inode
*bi
= BTRFS_I(inode
);
1784 bi
->i_default_acl
= NULL
;
1788 bi
->logged_trans
= 0;
1789 bi
->delalloc_bytes
= 0;
1790 bi
->disk_i_size
= 0;
1792 bi
->index_cnt
= (u64
)-1;
1793 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
, GFP_NOFS
);
1794 extent_io_tree_init(&BTRFS_I(inode
)->io_tree
,
1795 inode
->i_mapping
, GFP_NOFS
);
1796 extent_io_tree_init(&BTRFS_I(inode
)->io_failure_tree
,
1797 inode
->i_mapping
, GFP_NOFS
);
1798 INIT_LIST_HEAD(&BTRFS_I(inode
)->delalloc_inodes
);
1799 btrfs_ordered_inode_tree_init(&BTRFS_I(inode
)->ordered_tree
);
1800 mutex_init(&BTRFS_I(inode
)->csum_mutex
);
1801 mutex_init(&BTRFS_I(inode
)->extent_mutex
);
1802 mutex_init(&BTRFS_I(inode
)->log_mutex
);
1805 static int btrfs_init_locked_inode(struct inode
*inode
, void *p
)
1807 struct btrfs_iget_args
*args
= p
;
1808 inode
->i_ino
= args
->ino
;
1809 init_btrfs_i(inode
);
1810 BTRFS_I(inode
)->root
= args
->root
;
1814 static int btrfs_find_actor(struct inode
*inode
, void *opaque
)
1816 struct btrfs_iget_args
*args
= opaque
;
1817 return (args
->ino
== inode
->i_ino
&&
1818 args
->root
== BTRFS_I(inode
)->root
);
1821 struct inode
*btrfs_iget_locked(struct super_block
*s
, u64 objectid
,
1822 struct btrfs_root
*root
)
1824 struct inode
*inode
;
1825 struct btrfs_iget_args args
;
1826 args
.ino
= objectid
;
1829 inode
= iget5_locked(s
, objectid
, btrfs_find_actor
,
1830 btrfs_init_locked_inode
,
1835 /* Get an inode object given its location and corresponding root.
1836 * Returns in *is_new if the inode was read from disk
1838 struct inode
*btrfs_iget(struct super_block
*s
, struct btrfs_key
*location
,
1839 struct btrfs_root
*root
, int *is_new
)
1841 struct inode
*inode
;
1843 inode
= btrfs_iget_locked(s
, location
->objectid
, root
);
1845 return ERR_PTR(-EACCES
);
1847 if (inode
->i_state
& I_NEW
) {
1848 BTRFS_I(inode
)->root
= root
;
1849 memcpy(&BTRFS_I(inode
)->location
, location
, sizeof(*location
));
1850 btrfs_read_locked_inode(inode
);
1851 unlock_new_inode(inode
);
1862 static struct dentry
*btrfs_lookup(struct inode
*dir
, struct dentry
*dentry
,
1863 struct nameidata
*nd
)
1865 struct inode
* inode
;
1866 struct btrfs_inode
*bi
= BTRFS_I(dir
);
1867 struct btrfs_root
*root
= bi
->root
;
1868 struct btrfs_root
*sub_root
= root
;
1869 struct btrfs_key location
;
1870 int ret
, new, do_orphan
= 0;
1872 if (dentry
->d_name
.len
> BTRFS_NAME_LEN
)
1873 return ERR_PTR(-ENAMETOOLONG
);
1875 ret
= btrfs_inode_by_name(dir
, dentry
, &location
);
1878 return ERR_PTR(ret
);
1881 if (location
.objectid
) {
1882 ret
= fixup_tree_root_location(root
, &location
, &sub_root
,
1885 return ERR_PTR(ret
);
1887 return ERR_PTR(-ENOENT
);
1888 inode
= btrfs_iget(dir
->i_sb
, &location
, sub_root
, &new);
1890 return ERR_CAST(inode
);
1892 /* the inode and parent dir are two different roots */
1893 if (new && root
!= sub_root
) {
1895 sub_root
->inode
= inode
;
1900 if (unlikely(do_orphan
))
1901 btrfs_orphan_cleanup(sub_root
);
1903 return d_splice_alias(inode
, dentry
);
1906 static unsigned char btrfs_filetype_table
[] = {
1907 DT_UNKNOWN
, DT_REG
, DT_DIR
, DT_CHR
, DT_BLK
, DT_FIFO
, DT_SOCK
, DT_LNK
1910 static int btrfs_real_readdir(struct file
*filp
, void *dirent
,
1913 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1914 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1915 struct btrfs_item
*item
;
1916 struct btrfs_dir_item
*di
;
1917 struct btrfs_key key
;
1918 struct btrfs_key found_key
;
1919 struct btrfs_path
*path
;
1922 struct extent_buffer
*leaf
;
1925 unsigned char d_type
;
1930 int key_type
= BTRFS_DIR_INDEX_KEY
;
1935 /* FIXME, use a real flag for deciding about the key type */
1936 if (root
->fs_info
->tree_root
== root
)
1937 key_type
= BTRFS_DIR_ITEM_KEY
;
1939 /* special case for "." */
1940 if (filp
->f_pos
== 0) {
1941 over
= filldir(dirent
, ".", 1,
1948 /* special case for .., just use the back ref */
1949 if (filp
->f_pos
== 1) {
1950 u64 pino
= parent_ino(filp
->f_path
.dentry
);
1951 over
= filldir(dirent
, "..", 2,
1958 path
= btrfs_alloc_path();
1961 btrfs_set_key_type(&key
, key_type
);
1962 key
.offset
= filp
->f_pos
;
1963 key
.objectid
= inode
->i_ino
;
1965 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1971 leaf
= path
->nodes
[0];
1972 nritems
= btrfs_header_nritems(leaf
);
1973 slot
= path
->slots
[0];
1974 if (advance
|| slot
>= nritems
) {
1975 if (slot
>= nritems
- 1) {
1976 ret
= btrfs_next_leaf(root
, path
);
1979 leaf
= path
->nodes
[0];
1980 nritems
= btrfs_header_nritems(leaf
);
1981 slot
= path
->slots
[0];
1988 item
= btrfs_item_nr(leaf
, slot
);
1989 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1991 if (found_key
.objectid
!= key
.objectid
)
1993 if (btrfs_key_type(&found_key
) != key_type
)
1995 if (found_key
.offset
< filp
->f_pos
)
1998 filp
->f_pos
= found_key
.offset
;
2000 di
= btrfs_item_ptr(leaf
, slot
, struct btrfs_dir_item
);
2002 di_total
= btrfs_item_size(leaf
, item
);
2004 while (di_cur
< di_total
) {
2005 struct btrfs_key location
;
2007 name_len
= btrfs_dir_name_len(leaf
, di
);
2008 if (name_len
<= sizeof(tmp_name
)) {
2009 name_ptr
= tmp_name
;
2011 name_ptr
= kmalloc(name_len
, GFP_NOFS
);
2017 read_extent_buffer(leaf
, name_ptr
,
2018 (unsigned long)(di
+ 1), name_len
);
2020 d_type
= btrfs_filetype_table
[btrfs_dir_type(leaf
, di
)];
2021 btrfs_dir_item_key_to_cpu(leaf
, di
, &location
);
2022 over
= filldir(dirent
, name_ptr
, name_len
,
2023 found_key
.offset
, location
.objectid
,
2026 if (name_ptr
!= tmp_name
)
2032 di_len
= btrfs_dir_name_len(leaf
, di
) +
2033 btrfs_dir_data_len(leaf
, di
) + sizeof(*di
);
2035 di
= (struct btrfs_dir_item
*)((char *)di
+ di_len
);
2039 /* Reached end of directory/root. Bump pos past the last item. */
2040 if (key_type
== BTRFS_DIR_INDEX_KEY
)
2041 filp
->f_pos
= INT_LIMIT(typeof(filp
->f_pos
));
2047 btrfs_free_path(path
);
2051 /* Kernels earlier than 2.6.28 still have the NFS deadlock where nfsd
2052 will call the file system's ->lookup() method from within its
2053 filldir callback, which in turn was called from the file system's
2054 ->readdir() method. And will deadlock for many file systems. */
2055 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28)
2057 struct nfshack_dirent
{
2061 unsigned int d_type
;
2065 struct nfshack_readdir
{
2073 static int btrfs_nfshack_filldir(void *__buf
, const char *name
, int namlen
,
2074 loff_t offset
, u64 ino
, unsigned int d_type
)
2076 struct nfshack_readdir
*buf
= __buf
;
2077 struct nfshack_dirent
*de
= (void *)(buf
->dirent
+ buf
->used
);
2078 unsigned int reclen
;
2080 reclen
= ALIGN(sizeof(struct nfshack_dirent
) + namlen
, sizeof(u64
));
2081 if (buf
->used
+ reclen
> PAGE_SIZE
) {
2086 de
->namlen
= namlen
;
2087 de
->offset
= offset
;
2089 de
->d_type
= d_type
;
2090 memcpy(de
->name
, name
, namlen
);
2091 buf
->used
+= reclen
;
2096 static int btrfs_nfshack_readdir(struct file
*file
, void *dirent
,
2099 struct nfshack_readdir buf
;
2100 struct nfshack_dirent
*de
;
2105 buf
.dirent
= (void *)__get_free_page(GFP_KERNEL
);
2109 offset
= file
->f_pos
;
2112 unsigned int reclen
;
2116 err
= btrfs_real_readdir(file
, &buf
, btrfs_nfshack_filldir
);
2125 de
= (struct nfshack_dirent
*)buf
.dirent
;
2127 offset
= de
->offset
;
2129 if (filldir(dirent
, de
->name
, de
->namlen
, de
->offset
,
2130 de
->ino
, de
->d_type
))
2132 offset
= file
->f_pos
;
2134 reclen
= ALIGN(sizeof(*de
) + de
->namlen
,
2137 de
= (struct nfshack_dirent
*)((char *)de
+ reclen
);
2142 free_page((unsigned long)buf
.dirent
);
2143 file
->f_pos
= offset
;
2149 int btrfs_write_inode(struct inode
*inode
, int wait
)
2151 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2152 struct btrfs_trans_handle
*trans
;
2155 if (root
->fs_info
->closing
> 1)
2159 trans
= btrfs_join_transaction(root
, 1);
2160 btrfs_set_trans_block_group(trans
, inode
);
2161 ret
= btrfs_commit_transaction(trans
, root
);
2167 * This is somewhat expensive, updating the tree every time the
2168 * inode changes. But, it is most likely to find the inode in cache.
2169 * FIXME, needs more benchmarking...there are no reasons other than performance
2170 * to keep or drop this code.
2172 void btrfs_dirty_inode(struct inode
*inode
)
2174 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2175 struct btrfs_trans_handle
*trans
;
2177 trans
= btrfs_join_transaction(root
, 1);
2178 btrfs_set_trans_block_group(trans
, inode
);
2179 btrfs_update_inode(trans
, root
, inode
);
2180 btrfs_end_transaction(trans
, root
);
2183 static int btrfs_set_inode_index_count(struct inode
*inode
)
2185 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2186 struct btrfs_key key
, found_key
;
2187 struct btrfs_path
*path
;
2188 struct extent_buffer
*leaf
;
2191 key
.objectid
= inode
->i_ino
;
2192 btrfs_set_key_type(&key
, BTRFS_DIR_INDEX_KEY
);
2193 key
.offset
= (u64
)-1;
2195 path
= btrfs_alloc_path();
2199 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2202 /* FIXME: we should be able to handle this */
2208 * MAGIC NUMBER EXPLANATION:
2209 * since we search a directory based on f_pos we have to start at 2
2210 * since '.' and '..' have f_pos of 0 and 1 respectively, so everybody
2211 * else has to start at 2
2213 if (path
->slots
[0] == 0) {
2214 BTRFS_I(inode
)->index_cnt
= 2;
2220 leaf
= path
->nodes
[0];
2221 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2223 if (found_key
.objectid
!= inode
->i_ino
||
2224 btrfs_key_type(&found_key
) != BTRFS_DIR_INDEX_KEY
) {
2225 BTRFS_I(inode
)->index_cnt
= 2;
2229 BTRFS_I(inode
)->index_cnt
= found_key
.offset
+ 1;
2231 btrfs_free_path(path
);
2235 static int btrfs_set_inode_index(struct inode
*dir
, struct inode
*inode
,
2240 if (BTRFS_I(dir
)->index_cnt
== (u64
)-1) {
2241 ret
= btrfs_set_inode_index_count(dir
);
2247 *index
= BTRFS_I(dir
)->index_cnt
;
2248 BTRFS_I(dir
)->index_cnt
++;
2253 static struct inode
*btrfs_new_inode(struct btrfs_trans_handle
*trans
,
2254 struct btrfs_root
*root
,
2256 const char *name
, int name_len
,
2259 struct btrfs_block_group_cache
*group
,
2260 int mode
, u64
*index
)
2262 struct inode
*inode
;
2263 struct btrfs_inode_item
*inode_item
;
2264 struct btrfs_block_group_cache
*new_inode_group
;
2265 struct btrfs_key
*location
;
2266 struct btrfs_path
*path
;
2267 struct btrfs_inode_ref
*ref
;
2268 struct btrfs_key key
[2];
2274 path
= btrfs_alloc_path();
2277 inode
= new_inode(root
->fs_info
->sb
);
2279 return ERR_PTR(-ENOMEM
);
2282 ret
= btrfs_set_inode_index(dir
, inode
, index
);
2284 return ERR_PTR(ret
);
2287 * index_cnt is ignored for everything but a dir,
2288 * btrfs_get_inode_index_count has an explanation for the magic
2291 init_btrfs_i(inode
);
2292 BTRFS_I(inode
)->index_cnt
= 2;
2293 BTRFS_I(inode
)->root
= root
;
2294 BTRFS_I(inode
)->generation
= trans
->transid
;
2300 new_inode_group
= btrfs_find_block_group(root
, group
, 0,
2301 BTRFS_BLOCK_GROUP_METADATA
, owner
);
2302 if (!new_inode_group
) {
2303 printk("find_block group failed\n");
2304 new_inode_group
= group
;
2306 BTRFS_I(inode
)->block_group
= new_inode_group
;
2308 key
[0].objectid
= objectid
;
2309 btrfs_set_key_type(&key
[0], BTRFS_INODE_ITEM_KEY
);
2312 key
[1].objectid
= objectid
;
2313 btrfs_set_key_type(&key
[1], BTRFS_INODE_REF_KEY
);
2314 key
[1].offset
= ref_objectid
;
2316 sizes
[0] = sizeof(struct btrfs_inode_item
);
2317 sizes
[1] = name_len
+ sizeof(*ref
);
2319 ret
= btrfs_insert_empty_items(trans
, root
, path
, key
, sizes
, 2);
2323 if (objectid
> root
->highest_inode
)
2324 root
->highest_inode
= objectid
;
2326 inode
->i_uid
= current
->fsuid
;
2327 inode
->i_gid
= current
->fsgid
;
2328 inode
->i_mode
= mode
;
2329 inode
->i_ino
= objectid
;
2330 inode
->i_blocks
= 0;
2331 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2332 inode_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
2333 struct btrfs_inode_item
);
2334 fill_inode_item(trans
, path
->nodes
[0], inode_item
, inode
);
2336 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0] + 1,
2337 struct btrfs_inode_ref
);
2338 btrfs_set_inode_ref_name_len(path
->nodes
[0], ref
, name_len
);
2339 btrfs_set_inode_ref_index(path
->nodes
[0], ref
, *index
);
2340 ptr
= (unsigned long)(ref
+ 1);
2341 write_extent_buffer(path
->nodes
[0], name
, ptr
, name_len
);
2343 btrfs_mark_buffer_dirty(path
->nodes
[0]);
2344 btrfs_free_path(path
);
2346 location
= &BTRFS_I(inode
)->location
;
2347 location
->objectid
= objectid
;
2348 location
->offset
= 0;
2349 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
2351 insert_inode_hash(inode
);
2355 BTRFS_I(dir
)->index_cnt
--;
2356 btrfs_free_path(path
);
2357 return ERR_PTR(ret
);
2360 static inline u8
btrfs_inode_type(struct inode
*inode
)
2362 return btrfs_type_by_mode
[(inode
->i_mode
& S_IFMT
) >> S_SHIFT
];
2365 int btrfs_add_link(struct btrfs_trans_handle
*trans
,
2366 struct inode
*parent_inode
, struct inode
*inode
,
2367 const char *name
, int name_len
, int add_backref
, u64 index
)
2370 struct btrfs_key key
;
2371 struct btrfs_root
*root
= BTRFS_I(parent_inode
)->root
;
2373 key
.objectid
= inode
->i_ino
;
2374 btrfs_set_key_type(&key
, BTRFS_INODE_ITEM_KEY
);
2377 ret
= btrfs_insert_dir_item(trans
, root
, name
, name_len
,
2378 parent_inode
->i_ino
,
2379 &key
, btrfs_inode_type(inode
),
2383 ret
= btrfs_insert_inode_ref(trans
, root
,
2386 parent_inode
->i_ino
,
2389 btrfs_i_size_write(parent_inode
, parent_inode
->i_size
+
2391 parent_inode
->i_mtime
= parent_inode
->i_ctime
= CURRENT_TIME
;
2392 ret
= btrfs_update_inode(trans
, root
, parent_inode
);
2397 static int btrfs_add_nondir(struct btrfs_trans_handle
*trans
,
2398 struct dentry
*dentry
, struct inode
*inode
,
2399 int backref
, u64 index
)
2401 int err
= btrfs_add_link(trans
, dentry
->d_parent
->d_inode
,
2402 inode
, dentry
->d_name
.name
,
2403 dentry
->d_name
.len
, backref
, index
);
2405 d_instantiate(dentry
, inode
);
2413 static int btrfs_mknod(struct inode
*dir
, struct dentry
*dentry
,
2414 int mode
, dev_t rdev
)
2416 struct btrfs_trans_handle
*trans
;
2417 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2418 struct inode
*inode
= NULL
;
2422 unsigned long nr
= 0;
2425 if (!new_valid_dev(rdev
))
2428 err
= btrfs_check_free_space(root
, 1, 0);
2432 trans
= btrfs_start_transaction(root
, 1);
2433 btrfs_set_trans_block_group(trans
, dir
);
2435 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2441 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2443 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2444 BTRFS_I(dir
)->block_group
, mode
, &index
);
2445 err
= PTR_ERR(inode
);
2449 err
= btrfs_init_acl(inode
, dir
);
2455 btrfs_set_trans_block_group(trans
, inode
);
2456 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
2460 inode
->i_op
= &btrfs_special_inode_operations
;
2461 init_special_inode(inode
, inode
->i_mode
, rdev
);
2462 btrfs_update_inode(trans
, root
, inode
);
2464 dir
->i_sb
->s_dirt
= 1;
2465 btrfs_update_inode_block_group(trans
, inode
);
2466 btrfs_update_inode_block_group(trans
, dir
);
2468 nr
= trans
->blocks_used
;
2469 btrfs_end_transaction_throttle(trans
, root
);
2472 inode_dec_link_count(inode
);
2475 btrfs_btree_balance_dirty(root
, nr
);
2479 static int btrfs_create(struct inode
*dir
, struct dentry
*dentry
,
2480 int mode
, struct nameidata
*nd
)
2482 struct btrfs_trans_handle
*trans
;
2483 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2484 struct inode
*inode
= NULL
;
2487 unsigned long nr
= 0;
2491 err
= btrfs_check_free_space(root
, 1, 0);
2494 trans
= btrfs_start_transaction(root
, 1);
2495 btrfs_set_trans_block_group(trans
, dir
);
2497 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2503 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2505 dentry
->d_parent
->d_inode
->i_ino
,
2506 objectid
, BTRFS_I(dir
)->block_group
, mode
,
2508 err
= PTR_ERR(inode
);
2512 err
= btrfs_init_acl(inode
, dir
);
2518 btrfs_set_trans_block_group(trans
, inode
);
2519 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
2523 inode
->i_mapping
->a_ops
= &btrfs_aops
;
2524 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
2525 inode
->i_fop
= &btrfs_file_operations
;
2526 inode
->i_op
= &btrfs_file_inode_operations
;
2527 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
2529 dir
->i_sb
->s_dirt
= 1;
2530 btrfs_update_inode_block_group(trans
, inode
);
2531 btrfs_update_inode_block_group(trans
, dir
);
2533 nr
= trans
->blocks_used
;
2534 btrfs_end_transaction_throttle(trans
, root
);
2537 inode_dec_link_count(inode
);
2540 btrfs_btree_balance_dirty(root
, nr
);
2544 static int btrfs_link(struct dentry
*old_dentry
, struct inode
*dir
,
2545 struct dentry
*dentry
)
2547 struct btrfs_trans_handle
*trans
;
2548 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2549 struct inode
*inode
= old_dentry
->d_inode
;
2551 unsigned long nr
= 0;
2555 if (inode
->i_nlink
== 0)
2558 btrfs_inc_nlink(inode
);
2559 err
= btrfs_check_free_space(root
, 1, 0);
2562 err
= btrfs_set_inode_index(dir
, inode
, &index
);
2566 trans
= btrfs_start_transaction(root
, 1);
2568 btrfs_set_trans_block_group(trans
, dir
);
2569 atomic_inc(&inode
->i_count
);
2571 err
= btrfs_add_nondir(trans
, dentry
, inode
, 1, index
);
2576 dir
->i_sb
->s_dirt
= 1;
2577 btrfs_update_inode_block_group(trans
, dir
);
2578 err
= btrfs_update_inode(trans
, root
, inode
);
2583 nr
= trans
->blocks_used
;
2584 btrfs_end_transaction_throttle(trans
, root
);
2587 inode_dec_link_count(inode
);
2590 btrfs_btree_balance_dirty(root
, nr
);
2594 static int btrfs_mkdir(struct inode
*dir
, struct dentry
*dentry
, int mode
)
2596 struct inode
*inode
= NULL
;
2597 struct btrfs_trans_handle
*trans
;
2598 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2600 int drop_on_err
= 0;
2603 unsigned long nr
= 1;
2605 err
= btrfs_check_free_space(root
, 1, 0);
2609 trans
= btrfs_start_transaction(root
, 1);
2610 btrfs_set_trans_block_group(trans
, dir
);
2612 if (IS_ERR(trans
)) {
2613 err
= PTR_ERR(trans
);
2617 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2623 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
2625 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
2626 BTRFS_I(dir
)->block_group
, S_IFDIR
| mode
,
2628 if (IS_ERR(inode
)) {
2629 err
= PTR_ERR(inode
);
2635 err
= btrfs_init_acl(inode
, dir
);
2639 inode
->i_op
= &btrfs_dir_inode_operations
;
2640 inode
->i_fop
= &btrfs_dir_file_operations
;
2641 btrfs_set_trans_block_group(trans
, inode
);
2643 btrfs_i_size_write(inode
, 0);
2644 err
= btrfs_update_inode(trans
, root
, inode
);
2648 err
= btrfs_add_link(trans
, dentry
->d_parent
->d_inode
,
2649 inode
, dentry
->d_name
.name
,
2650 dentry
->d_name
.len
, 0, index
);
2654 d_instantiate(dentry
, inode
);
2656 dir
->i_sb
->s_dirt
= 1;
2657 btrfs_update_inode_block_group(trans
, inode
);
2658 btrfs_update_inode_block_group(trans
, dir
);
2661 nr
= trans
->blocks_used
;
2662 btrfs_end_transaction_throttle(trans
, root
);
2667 btrfs_btree_balance_dirty(root
, nr
);
2671 static int merge_extent_mapping(struct extent_map_tree
*em_tree
,
2672 struct extent_map
*existing
,
2673 struct extent_map
*em
,
2674 u64 map_start
, u64 map_len
)
2678 BUG_ON(map_start
< em
->start
|| map_start
>= extent_map_end(em
));
2679 start_diff
= map_start
- em
->start
;
2680 em
->start
= map_start
;
2682 if (em
->block_start
< EXTENT_MAP_LAST_BYTE
)
2683 em
->block_start
+= start_diff
;
2684 return add_extent_mapping(em_tree
, em
);
2687 struct extent_map
*btrfs_get_extent(struct inode
*inode
, struct page
*page
,
2688 size_t pg_offset
, u64 start
, u64 len
,
2694 u64 extent_start
= 0;
2696 u64 objectid
= inode
->i_ino
;
2698 struct btrfs_path
*path
= NULL
;
2699 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2700 struct btrfs_file_extent_item
*item
;
2701 struct extent_buffer
*leaf
;
2702 struct btrfs_key found_key
;
2703 struct extent_map
*em
= NULL
;
2704 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
2705 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2706 struct btrfs_trans_handle
*trans
= NULL
;
2709 spin_lock(&em_tree
->lock
);
2710 em
= lookup_extent_mapping(em_tree
, start
, len
);
2712 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2713 spin_unlock(&em_tree
->lock
);
2716 if (em
->start
> start
|| em
->start
+ em
->len
<= start
)
2717 free_extent_map(em
);
2718 else if (em
->block_start
== EXTENT_MAP_INLINE
&& page
)
2719 free_extent_map(em
);
2723 em
= alloc_extent_map(GFP_NOFS
);
2728 em
->bdev
= root
->fs_info
->fs_devices
->latest_bdev
;
2729 em
->start
= EXTENT_MAP_HOLE
;
2733 path
= btrfs_alloc_path();
2737 ret
= btrfs_lookup_file_extent(trans
, root
, path
,
2738 objectid
, start
, trans
!= NULL
);
2745 if (path
->slots
[0] == 0)
2750 leaf
= path
->nodes
[0];
2751 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
2752 struct btrfs_file_extent_item
);
2753 /* are we inside the extent that was found? */
2754 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2755 found_type
= btrfs_key_type(&found_key
);
2756 if (found_key
.objectid
!= objectid
||
2757 found_type
!= BTRFS_EXTENT_DATA_KEY
) {
2761 found_type
= btrfs_file_extent_type(leaf
, item
);
2762 extent_start
= found_key
.offset
;
2763 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
2764 extent_end
= extent_start
+
2765 btrfs_file_extent_num_bytes(leaf
, item
);
2767 if (start
< extent_start
|| start
>= extent_end
) {
2769 if (start
< extent_start
) {
2770 if (start
+ len
<= extent_start
)
2772 em
->len
= extent_end
- extent_start
;
2778 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
2780 em
->start
= extent_start
;
2781 em
->len
= extent_end
- extent_start
;
2782 em
->block_start
= EXTENT_MAP_HOLE
;
2785 bytenr
+= btrfs_file_extent_offset(leaf
, item
);
2786 em
->block_start
= bytenr
;
2787 em
->start
= extent_start
;
2788 em
->len
= extent_end
- extent_start
;
2790 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
2795 size_t extent_offset
;
2798 size
= btrfs_file_extent_inline_len(leaf
, btrfs_item_nr(leaf
,
2800 extent_end
= (extent_start
+ size
+ root
->sectorsize
- 1) &
2801 ~((u64
)root
->sectorsize
- 1);
2802 if (start
< extent_start
|| start
>= extent_end
) {
2804 if (start
< extent_start
) {
2805 if (start
+ len
<= extent_start
)
2807 em
->len
= extent_end
- extent_start
;
2813 em
->block_start
= EXTENT_MAP_INLINE
;
2816 em
->start
= extent_start
;
2821 page_start
= page_offset(page
) + pg_offset
;
2822 extent_offset
= page_start
- extent_start
;
2823 copy_size
= min_t(u64
, PAGE_CACHE_SIZE
- pg_offset
,
2824 size
- extent_offset
);
2825 em
->start
= extent_start
+ extent_offset
;
2826 em
->len
= (copy_size
+ root
->sectorsize
- 1) &
2827 ~((u64
)root
->sectorsize
- 1);
2829 ptr
= btrfs_file_extent_inline_start(item
) + extent_offset
;
2830 if (create
== 0 && !PageUptodate(page
)) {
2831 read_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2833 flush_dcache_page(page
);
2834 } else if (create
&& PageUptodate(page
)) {
2837 free_extent_map(em
);
2839 btrfs_release_path(root
, path
);
2840 trans
= btrfs_join_transaction(root
, 1);
2843 write_extent_buffer(leaf
, map
+ pg_offset
, ptr
,
2845 btrfs_mark_buffer_dirty(leaf
);
2848 set_extent_uptodate(io_tree
, em
->start
,
2849 extent_map_end(em
) - 1, GFP_NOFS
);
2852 printk("unkknown found_type %d\n", found_type
);
2859 em
->block_start
= EXTENT_MAP_HOLE
;
2861 btrfs_release_path(root
, path
);
2862 if (em
->start
> start
|| extent_map_end(em
) <= start
) {
2863 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em
->start
, em
->len
, start
, len
);
2869 spin_lock(&em_tree
->lock
);
2870 ret
= add_extent_mapping(em_tree
, em
);
2871 /* it is possible that someone inserted the extent into the tree
2872 * while we had the lock dropped. It is also possible that
2873 * an overlapping map exists in the tree
2875 if (ret
== -EEXIST
) {
2876 struct extent_map
*existing
;
2880 existing
= lookup_extent_mapping(em_tree
, start
, len
);
2881 if (existing
&& (existing
->start
> start
||
2882 existing
->start
+ existing
->len
<= start
)) {
2883 free_extent_map(existing
);
2887 existing
= lookup_extent_mapping(em_tree
, em
->start
,
2890 err
= merge_extent_mapping(em_tree
, existing
,
2893 free_extent_map(existing
);
2895 free_extent_map(em
);
2900 printk("failing to insert %Lu %Lu\n",
2902 free_extent_map(em
);
2906 free_extent_map(em
);
2911 spin_unlock(&em_tree
->lock
);
2914 btrfs_free_path(path
);
2916 ret
= btrfs_end_transaction(trans
, root
);
2922 free_extent_map(em
);
2924 return ERR_PTR(err
);
2929 #if 0 /* waiting for O_DIRECT reads */
2930 static int btrfs_get_block(struct inode
*inode
, sector_t iblock
,
2931 struct buffer_head
*bh_result
, int create
)
2933 struct extent_map
*em
;
2934 u64 start
= (u64
)iblock
<< inode
->i_blkbits
;
2935 struct btrfs_multi_bio
*multi
= NULL
;
2936 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2942 em
= btrfs_get_extent(inode
, NULL
, 0, start
, bh_result
->b_size
, 0);
2944 if (!em
|| IS_ERR(em
))
2947 if (em
->start
> start
|| em
->start
+ em
->len
<= start
) {
2951 if (em
->block_start
== EXTENT_MAP_INLINE
) {
2956 len
= em
->start
+ em
->len
- start
;
2957 len
= min_t(u64
, len
, INT_LIMIT(typeof(bh_result
->b_size
)));
2959 if (em
->block_start
== EXTENT_MAP_HOLE
||
2960 em
->block_start
== EXTENT_MAP_DELALLOC
) {
2961 bh_result
->b_size
= len
;
2965 logical
= start
- em
->start
;
2966 logical
= em
->block_start
+ logical
;
2969 ret
= btrfs_map_block(&root
->fs_info
->mapping_tree
, READ
,
2970 logical
, &map_length
, &multi
, 0);
2972 bh_result
->b_blocknr
= multi
->stripes
[0].physical
>> inode
->i_blkbits
;
2973 bh_result
->b_size
= min(map_length
, len
);
2975 bh_result
->b_bdev
= multi
->stripes
[0].dev
->bdev
;
2976 set_buffer_mapped(bh_result
);
2979 free_extent_map(em
);
2984 static ssize_t
btrfs_direct_IO(int rw
, struct kiocb
*iocb
,
2985 const struct iovec
*iov
, loff_t offset
,
2986 unsigned long nr_segs
)
2990 struct file
*file
= iocb
->ki_filp
;
2991 struct inode
*inode
= file
->f_mapping
->host
;
2996 return blockdev_direct_IO(rw
, iocb
, inode
, inode
->i_sb
->s_bdev
, iov
,
2997 offset
, nr_segs
, btrfs_get_block
, NULL
);
3001 static sector_t
btrfs_bmap(struct address_space
*mapping
, sector_t iblock
)
3003 return extent_bmap(mapping
, iblock
, btrfs_get_extent
);
3006 int btrfs_readpage(struct file
*file
, struct page
*page
)
3008 struct extent_io_tree
*tree
;
3009 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3010 return extent_read_full_page(tree
, page
, btrfs_get_extent
);
3013 static int btrfs_writepage(struct page
*page
, struct writeback_control
*wbc
)
3015 struct extent_io_tree
*tree
;
3018 if (current
->flags
& PF_MEMALLOC
) {
3019 redirty_page_for_writepage(wbc
, page
);
3023 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3024 return extent_write_full_page(tree
, page
, btrfs_get_extent
, wbc
);
3027 int btrfs_writepages(struct address_space
*mapping
,
3028 struct writeback_control
*wbc
)
3030 struct extent_io_tree
*tree
;
3031 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
3032 return extent_writepages(tree
, mapping
, btrfs_get_extent
, wbc
);
3036 btrfs_readpages(struct file
*file
, struct address_space
*mapping
,
3037 struct list_head
*pages
, unsigned nr_pages
)
3039 struct extent_io_tree
*tree
;
3040 tree
= &BTRFS_I(mapping
->host
)->io_tree
;
3041 return extent_readpages(tree
, mapping
, pages
, nr_pages
,
3044 static int __btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
3046 struct extent_io_tree
*tree
;
3047 struct extent_map_tree
*map
;
3050 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3051 map
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
3052 ret
= try_release_extent_mapping(map
, tree
, page
, gfp_flags
);
3054 ClearPagePrivate(page
);
3055 set_page_private(page
, 0);
3056 page_cache_release(page
);
3061 static int btrfs_releasepage(struct page
*page
, gfp_t gfp_flags
)
3063 if (PageWriteback(page
) || PageDirty(page
))
3065 return __btrfs_releasepage(page
, gfp_flags
);
3068 static void btrfs_invalidatepage(struct page
*page
, unsigned long offset
)
3070 struct extent_io_tree
*tree
;
3071 struct btrfs_ordered_extent
*ordered
;
3072 u64 page_start
= page_offset(page
);
3073 u64 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
3075 wait_on_page_writeback(page
);
3076 tree
= &BTRFS_I(page
->mapping
->host
)->io_tree
;
3078 btrfs_releasepage(page
, GFP_NOFS
);
3082 lock_extent(tree
, page_start
, page_end
, GFP_NOFS
);
3083 ordered
= btrfs_lookup_ordered_extent(page
->mapping
->host
,
3087 * IO on this page will never be started, so we need
3088 * to account for any ordered extents now
3090 clear_extent_bit(tree
, page_start
, page_end
,
3091 EXTENT_DIRTY
| EXTENT_DELALLOC
|
3092 EXTENT_LOCKED
, 1, 0, GFP_NOFS
);
3093 btrfs_finish_ordered_io(page
->mapping
->host
,
3094 page_start
, page_end
);
3095 btrfs_put_ordered_extent(ordered
);
3096 lock_extent(tree
, page_start
, page_end
, GFP_NOFS
);
3098 clear_extent_bit(tree
, page_start
, page_end
,
3099 EXTENT_LOCKED
| EXTENT_DIRTY
| EXTENT_DELALLOC
|
3102 __btrfs_releasepage(page
, GFP_NOFS
);
3104 ClearPageChecked(page
);
3105 if (PagePrivate(page
)) {
3106 ClearPagePrivate(page
);
3107 set_page_private(page
, 0);
3108 page_cache_release(page
);
3113 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
3114 * called from a page fault handler when a page is first dirtied. Hence we must
3115 * be careful to check for EOF conditions here. We set the page up correctly
3116 * for a written page which means we get ENOSPC checking when writing into
3117 * holes and correct delalloc and unwritten extent mapping on filesystems that
3118 * support these features.
3120 * We are not allowed to take the i_mutex here so we have to play games to
3121 * protect against truncate races as the page could now be beyond EOF. Because
3122 * vmtruncate() writes the inode size before removing pages, once we have the
3123 * page lock we can determine safely if the page is beyond EOF. If it is not
3124 * beyond EOF, then the page is guaranteed safe against truncation until we
3127 int btrfs_page_mkwrite(struct vm_area_struct
*vma
, struct page
*page
)
3129 struct inode
*inode
= fdentry(vma
->vm_file
)->d_inode
;
3130 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3131 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
3132 struct btrfs_ordered_extent
*ordered
;
3134 unsigned long zero_start
;
3140 ret
= btrfs_check_free_space(root
, PAGE_CACHE_SIZE
, 0);
3147 size
= i_size_read(inode
);
3148 page_start
= page_offset(page
);
3149 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
3151 if ((page
->mapping
!= inode
->i_mapping
) ||
3152 (page_start
>= size
)) {
3153 /* page got truncated out from underneath us */
3156 wait_on_page_writeback(page
);
3158 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3159 set_page_extent_mapped(page
);
3162 * we can't set the delalloc bits if there are pending ordered
3163 * extents. Drop our locks and wait for them to finish
3165 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
3167 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3169 btrfs_start_ordered_extent(inode
, ordered
, 1);
3170 btrfs_put_ordered_extent(ordered
);
3174 btrfs_set_extent_delalloc(inode
, page_start
, page_end
);
3177 /* page is wholly or partially inside EOF */
3178 if (page_start
+ PAGE_CACHE_SIZE
> size
)
3179 zero_start
= size
& ~PAGE_CACHE_MASK
;
3181 zero_start
= PAGE_CACHE_SIZE
;
3183 if (zero_start
!= PAGE_CACHE_SIZE
) {
3185 memset(kaddr
+ zero_start
, 0, PAGE_CACHE_SIZE
- zero_start
);
3186 flush_dcache_page(page
);
3189 ClearPageChecked(page
);
3190 set_page_dirty(page
);
3191 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
3199 static void btrfs_truncate(struct inode
*inode
)
3201 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3203 struct btrfs_trans_handle
*trans
;
3205 u64 mask
= root
->sectorsize
- 1;
3207 if (!S_ISREG(inode
->i_mode
))
3209 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
3212 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
3213 btrfs_wait_ordered_range(inode
, inode
->i_size
& (~mask
), (u64
)-1);
3215 trans
= btrfs_start_transaction(root
, 1);
3216 btrfs_set_trans_block_group(trans
, inode
);
3217 btrfs_i_size_write(inode
, inode
->i_size
);
3219 ret
= btrfs_orphan_add(trans
, inode
);
3222 /* FIXME, add redo link to tree so we don't leak on crash */
3223 ret
= btrfs_truncate_inode_items(trans
, root
, inode
, inode
->i_size
,
3224 BTRFS_EXTENT_DATA_KEY
);
3225 btrfs_update_inode(trans
, root
, inode
);
3227 ret
= btrfs_orphan_del(trans
, inode
);
3231 nr
= trans
->blocks_used
;
3232 ret
= btrfs_end_transaction_throttle(trans
, root
);
3234 btrfs_btree_balance_dirty(root
, nr
);
3238 * Invalidate a single dcache entry at the root of the filesystem.
3239 * Needed after creation of snapshot or subvolume.
3241 void btrfs_invalidate_dcache_root(struct btrfs_root
*root
, char *name
,
3244 struct dentry
*alias
, *entry
;
3247 alias
= d_find_alias(root
->fs_info
->sb
->s_root
->d_inode
);
3251 /* change me if btrfs ever gets a d_hash operation */
3252 qstr
.hash
= full_name_hash(qstr
.name
, qstr
.len
);
3253 entry
= d_lookup(alias
, &qstr
);
3256 d_invalidate(entry
);
3262 int btrfs_create_subvol_root(struct btrfs_root
*new_root
,
3263 struct btrfs_trans_handle
*trans
, u64 new_dirid
,
3264 struct btrfs_block_group_cache
*block_group
)
3266 struct inode
*inode
;
3269 inode
= btrfs_new_inode(trans
, new_root
, NULL
, "..", 2, new_dirid
,
3270 new_dirid
, block_group
, S_IFDIR
| 0700, &index
);
3272 return PTR_ERR(inode
);
3273 inode
->i_op
= &btrfs_dir_inode_operations
;
3274 inode
->i_fop
= &btrfs_dir_file_operations
;
3275 new_root
->inode
= inode
;
3278 btrfs_i_size_write(inode
, 0);
3280 return btrfs_update_inode(trans
, new_root
, inode
);
3283 unsigned long btrfs_force_ra(struct address_space
*mapping
,
3284 struct file_ra_state
*ra
, struct file
*file
,
3285 pgoff_t offset
, pgoff_t last_index
)
3287 pgoff_t req_size
= last_index
- offset
+ 1;
3289 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3290 offset
= page_cache_readahead(mapping
, ra
, file
, offset
, req_size
);
3293 page_cache_sync_readahead(mapping
, ra
, file
, offset
, req_size
);
3294 return offset
+ req_size
;
3298 struct inode
*btrfs_alloc_inode(struct super_block
*sb
)
3300 struct btrfs_inode
*ei
;
3302 ei
= kmem_cache_alloc(btrfs_inode_cachep
, GFP_NOFS
);
3306 ei
->logged_trans
= 0;
3307 btrfs_ordered_inode_tree_init(&ei
->ordered_tree
);
3308 ei
->i_acl
= BTRFS_ACL_NOT_CACHED
;
3309 ei
->i_default_acl
= BTRFS_ACL_NOT_CACHED
;
3310 INIT_LIST_HEAD(&ei
->i_orphan
);
3311 return &ei
->vfs_inode
;
3314 void btrfs_destroy_inode(struct inode
*inode
)
3316 struct btrfs_ordered_extent
*ordered
;
3317 WARN_ON(!list_empty(&inode
->i_dentry
));
3318 WARN_ON(inode
->i_data
.nrpages
);
3320 if (BTRFS_I(inode
)->i_acl
&&
3321 BTRFS_I(inode
)->i_acl
!= BTRFS_ACL_NOT_CACHED
)
3322 posix_acl_release(BTRFS_I(inode
)->i_acl
);
3323 if (BTRFS_I(inode
)->i_default_acl
&&
3324 BTRFS_I(inode
)->i_default_acl
!= BTRFS_ACL_NOT_CACHED
)
3325 posix_acl_release(BTRFS_I(inode
)->i_default_acl
);
3327 spin_lock(&BTRFS_I(inode
)->root
->list_lock
);
3328 if (!list_empty(&BTRFS_I(inode
)->i_orphan
)) {
3329 printk(KERN_ERR
"BTRFS: inode %lu: inode still on the orphan"
3330 " list\n", inode
->i_ino
);
3333 spin_unlock(&BTRFS_I(inode
)->root
->list_lock
);
3336 ordered
= btrfs_lookup_first_ordered_extent(inode
, (u64
)-1);
3340 printk("found ordered extent %Lu %Lu\n",
3341 ordered
->file_offset
, ordered
->len
);
3342 btrfs_remove_ordered_extent(inode
, ordered
);
3343 btrfs_put_ordered_extent(ordered
);
3344 btrfs_put_ordered_extent(ordered
);
3347 btrfs_drop_extent_cache(inode
, 0, (u64
)-1);
3348 kmem_cache_free(btrfs_inode_cachep
, BTRFS_I(inode
));
3351 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3352 static void init_once(void *foo
)
3353 #elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3354 static void init_once(struct kmem_cache
* cachep
, void *foo
)
3356 static void init_once(void * foo
, struct kmem_cache
* cachep
,
3357 unsigned long flags
)
3360 struct btrfs_inode
*ei
= (struct btrfs_inode
*) foo
;
3362 inode_init_once(&ei
->vfs_inode
);
3365 void btrfs_destroy_cachep(void)
3367 if (btrfs_inode_cachep
)
3368 kmem_cache_destroy(btrfs_inode_cachep
);
3369 if (btrfs_trans_handle_cachep
)
3370 kmem_cache_destroy(btrfs_trans_handle_cachep
);
3371 if (btrfs_transaction_cachep
)
3372 kmem_cache_destroy(btrfs_transaction_cachep
);
3373 if (btrfs_bit_radix_cachep
)
3374 kmem_cache_destroy(btrfs_bit_radix_cachep
);
3375 if (btrfs_path_cachep
)
3376 kmem_cache_destroy(btrfs_path_cachep
);
3379 struct kmem_cache
*btrfs_cache_create(const char *name
, size_t size
,
3380 unsigned long extra_flags
,
3381 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3382 void (*ctor
)(void *)
3383 #elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3384 void (*ctor
)(struct kmem_cache
*, void *)
3386 void (*ctor
)(void *, struct kmem_cache
*,
3391 return kmem_cache_create(name
, size
, 0, (SLAB_RECLAIM_ACCOUNT
|
3392 SLAB_MEM_SPREAD
| extra_flags
), ctor
3393 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3399 int btrfs_init_cachep(void)
3401 btrfs_inode_cachep
= btrfs_cache_create("btrfs_inode_cache",
3402 sizeof(struct btrfs_inode
),
3404 if (!btrfs_inode_cachep
)
3406 btrfs_trans_handle_cachep
=
3407 btrfs_cache_create("btrfs_trans_handle_cache",
3408 sizeof(struct btrfs_trans_handle
),
3410 if (!btrfs_trans_handle_cachep
)
3412 btrfs_transaction_cachep
= btrfs_cache_create("btrfs_transaction_cache",
3413 sizeof(struct btrfs_transaction
),
3415 if (!btrfs_transaction_cachep
)
3417 btrfs_path_cachep
= btrfs_cache_create("btrfs_path_cache",
3418 sizeof(struct btrfs_path
),
3420 if (!btrfs_path_cachep
)
3422 btrfs_bit_radix_cachep
= btrfs_cache_create("btrfs_radix", 256,
3423 SLAB_DESTROY_BY_RCU
, NULL
);
3424 if (!btrfs_bit_radix_cachep
)
3428 btrfs_destroy_cachep();
3432 static int btrfs_getattr(struct vfsmount
*mnt
,
3433 struct dentry
*dentry
, struct kstat
*stat
)
3435 struct inode
*inode
= dentry
->d_inode
;
3436 generic_fillattr(inode
, stat
);
3437 stat
->blksize
= PAGE_CACHE_SIZE
;
3438 stat
->blocks
= inode
->i_blocks
+ (BTRFS_I(inode
)->delalloc_bytes
>> 9);
3442 static int btrfs_rename(struct inode
* old_dir
, struct dentry
*old_dentry
,
3443 struct inode
* new_dir
,struct dentry
*new_dentry
)
3445 struct btrfs_trans_handle
*trans
;
3446 struct btrfs_root
*root
= BTRFS_I(old_dir
)->root
;
3447 struct inode
*new_inode
= new_dentry
->d_inode
;
3448 struct inode
*old_inode
= old_dentry
->d_inode
;
3449 struct timespec ctime
= CURRENT_TIME
;
3453 if (S_ISDIR(old_inode
->i_mode
) && new_inode
&&
3454 new_inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
3458 ret
= btrfs_check_free_space(root
, 1, 0);
3462 trans
= btrfs_start_transaction(root
, 1);
3464 btrfs_set_trans_block_group(trans
, new_dir
);
3466 btrfs_inc_nlink(old_dentry
->d_inode
);
3467 old_dir
->i_ctime
= old_dir
->i_mtime
= ctime
;
3468 new_dir
->i_ctime
= new_dir
->i_mtime
= ctime
;
3469 old_inode
->i_ctime
= ctime
;
3471 ret
= btrfs_unlink_inode(trans
, root
, old_dir
, old_dentry
->d_inode
,
3472 old_dentry
->d_name
.name
,
3473 old_dentry
->d_name
.len
);
3478 new_inode
->i_ctime
= CURRENT_TIME
;
3479 ret
= btrfs_unlink_inode(trans
, root
, new_dir
,
3480 new_dentry
->d_inode
,
3481 new_dentry
->d_name
.name
,
3482 new_dentry
->d_name
.len
);
3485 if (new_inode
->i_nlink
== 0) {
3486 ret
= btrfs_orphan_add(trans
, new_dentry
->d_inode
);
3492 ret
= btrfs_set_inode_index(new_dir
, old_inode
, &index
);
3496 ret
= btrfs_add_link(trans
, new_dentry
->d_parent
->d_inode
,
3497 old_inode
, new_dentry
->d_name
.name
,
3498 new_dentry
->d_name
.len
, 1, index
);
3503 btrfs_end_transaction_throttle(trans
, root
);
3508 int btrfs_start_delalloc_inodes(struct btrfs_root
*root
)
3510 struct list_head
*head
= &root
->fs_info
->delalloc_inodes
;
3511 struct btrfs_inode
*binode
;
3512 unsigned long flags
;
3514 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
3515 while(!list_empty(head
)) {
3516 binode
= list_entry(head
->next
, struct btrfs_inode
,
3518 atomic_inc(&binode
->vfs_inode
.i_count
);
3519 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
3520 filemap_write_and_wait(binode
->vfs_inode
.i_mapping
);
3521 iput(&binode
->vfs_inode
);
3522 spin_lock_irqsave(&root
->fs_info
->delalloc_lock
, flags
);
3524 spin_unlock_irqrestore(&root
->fs_info
->delalloc_lock
, flags
);
3528 static int btrfs_symlink(struct inode
*dir
, struct dentry
*dentry
,
3529 const char *symname
)
3531 struct btrfs_trans_handle
*trans
;
3532 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
3533 struct btrfs_path
*path
;
3534 struct btrfs_key key
;
3535 struct inode
*inode
= NULL
;
3543 struct btrfs_file_extent_item
*ei
;
3544 struct extent_buffer
*leaf
;
3545 unsigned long nr
= 0;
3547 name_len
= strlen(symname
) + 1;
3548 if (name_len
> BTRFS_MAX_INLINE_DATA_SIZE(root
))
3549 return -ENAMETOOLONG
;
3551 err
= btrfs_check_free_space(root
, 1, 0);
3555 trans
= btrfs_start_transaction(root
, 1);
3556 btrfs_set_trans_block_group(trans
, dir
);
3558 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
3564 inode
= btrfs_new_inode(trans
, root
, dir
, dentry
->d_name
.name
,
3566 dentry
->d_parent
->d_inode
->i_ino
, objectid
,
3567 BTRFS_I(dir
)->block_group
, S_IFLNK
|S_IRWXUGO
,
3569 err
= PTR_ERR(inode
);
3573 err
= btrfs_init_acl(inode
, dir
);
3579 btrfs_set_trans_block_group(trans
, inode
);
3580 err
= btrfs_add_nondir(trans
, dentry
, inode
, 0, index
);
3584 inode
->i_mapping
->a_ops
= &btrfs_aops
;
3585 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3586 inode
->i_fop
= &btrfs_file_operations
;
3587 inode
->i_op
= &btrfs_file_inode_operations
;
3588 BTRFS_I(inode
)->io_tree
.ops
= &btrfs_extent_io_ops
;
3590 dir
->i_sb
->s_dirt
= 1;
3591 btrfs_update_inode_block_group(trans
, inode
);
3592 btrfs_update_inode_block_group(trans
, dir
);
3596 path
= btrfs_alloc_path();
3598 key
.objectid
= inode
->i_ino
;
3600 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
3601 datasize
= btrfs_file_extent_calc_inline_size(name_len
);
3602 err
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
3608 leaf
= path
->nodes
[0];
3609 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
3610 struct btrfs_file_extent_item
);
3611 btrfs_set_file_extent_generation(leaf
, ei
, trans
->transid
);
3612 btrfs_set_file_extent_type(leaf
, ei
,
3613 BTRFS_FILE_EXTENT_INLINE
);
3614 ptr
= btrfs_file_extent_inline_start(ei
);
3615 write_extent_buffer(leaf
, symname
, ptr
, name_len
);
3616 btrfs_mark_buffer_dirty(leaf
);
3617 btrfs_free_path(path
);
3619 inode
->i_op
= &btrfs_symlink_inode_operations
;
3620 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
3621 inode
->i_mapping
->backing_dev_info
= &root
->fs_info
->bdi
;
3622 btrfs_i_size_write(inode
, name_len
- 1);
3623 err
= btrfs_update_inode(trans
, root
, inode
);
3628 nr
= trans
->blocks_used
;
3629 btrfs_end_transaction_throttle(trans
, root
);
3632 inode_dec_link_count(inode
);
3635 btrfs_btree_balance_dirty(root
, nr
);
3639 static int btrfs_set_page_dirty(struct page
*page
)
3641 return __set_page_dirty_nobuffers(page
);
3644 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3645 static int btrfs_permission(struct inode
*inode
, int mask
)
3647 static int btrfs_permission(struct inode
*inode
, int mask
,
3648 struct nameidata
*nd
)
3651 if (btrfs_test_flag(inode
, READONLY
) && (mask
& MAY_WRITE
))
3653 return generic_permission(inode
, mask
, btrfs_check_acl
);
3656 static struct inode_operations btrfs_dir_inode_operations
= {
3657 .lookup
= btrfs_lookup
,
3658 .create
= btrfs_create
,
3659 .unlink
= btrfs_unlink
,
3661 .mkdir
= btrfs_mkdir
,
3662 .rmdir
= btrfs_rmdir
,
3663 .rename
= btrfs_rename
,
3664 .symlink
= btrfs_symlink
,
3665 .setattr
= btrfs_setattr
,
3666 .mknod
= btrfs_mknod
,
3667 .setxattr
= btrfs_setxattr
,
3668 .getxattr
= btrfs_getxattr
,
3669 .listxattr
= btrfs_listxattr
,
3670 .removexattr
= btrfs_removexattr
,
3671 .permission
= btrfs_permission
,
3673 static struct inode_operations btrfs_dir_ro_inode_operations
= {
3674 .lookup
= btrfs_lookup
,
3675 .permission
= btrfs_permission
,
3677 static struct file_operations btrfs_dir_file_operations
= {
3678 .llseek
= generic_file_llseek
,
3679 .read
= generic_read_dir
,
3680 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28)
3681 .readdir
= btrfs_nfshack_readdir
,
3682 #else /* NFSd readdir/lookup deadlock is fixed */
3683 .readdir
= btrfs_real_readdir
,
3685 .unlocked_ioctl
= btrfs_ioctl
,
3686 #ifdef CONFIG_COMPAT
3687 .compat_ioctl
= btrfs_ioctl
,
3689 .release
= btrfs_release_file
,
3690 .fsync
= btrfs_sync_file
,
3693 static struct extent_io_ops btrfs_extent_io_ops
= {
3694 .fill_delalloc
= run_delalloc_range
,
3695 .submit_bio_hook
= btrfs_submit_bio_hook
,
3696 .merge_bio_hook
= btrfs_merge_bio_hook
,
3697 .readpage_end_io_hook
= btrfs_readpage_end_io_hook
,
3698 .writepage_end_io_hook
= btrfs_writepage_end_io_hook
,
3699 .writepage_start_hook
= btrfs_writepage_start_hook
,
3700 .readpage_io_failed_hook
= btrfs_io_failed_hook
,
3701 .set_bit_hook
= btrfs_set_bit_hook
,
3702 .clear_bit_hook
= btrfs_clear_bit_hook
,
3705 static struct address_space_operations btrfs_aops
= {
3706 .readpage
= btrfs_readpage
,
3707 .writepage
= btrfs_writepage
,
3708 .writepages
= btrfs_writepages
,
3709 .readpages
= btrfs_readpages
,
3710 .sync_page
= block_sync_page
,
3712 .direct_IO
= btrfs_direct_IO
,
3713 .invalidatepage
= btrfs_invalidatepage
,
3714 .releasepage
= btrfs_releasepage
,
3715 .set_page_dirty
= btrfs_set_page_dirty
,
3718 static struct address_space_operations btrfs_symlink_aops
= {
3719 .readpage
= btrfs_readpage
,
3720 .writepage
= btrfs_writepage
,
3721 .invalidatepage
= btrfs_invalidatepage
,
3722 .releasepage
= btrfs_releasepage
,
3725 static struct inode_operations btrfs_file_inode_operations
= {
3726 .truncate
= btrfs_truncate
,
3727 .getattr
= btrfs_getattr
,
3728 .setattr
= btrfs_setattr
,
3729 .setxattr
= btrfs_setxattr
,
3730 .getxattr
= btrfs_getxattr
,
3731 .listxattr
= btrfs_listxattr
,
3732 .removexattr
= btrfs_removexattr
,
3733 .permission
= btrfs_permission
,
3735 static struct inode_operations btrfs_special_inode_operations
= {
3736 .getattr
= btrfs_getattr
,
3737 .setattr
= btrfs_setattr
,
3738 .permission
= btrfs_permission
,
3739 .setxattr
= btrfs_setxattr
,
3740 .getxattr
= btrfs_getxattr
,
3741 .listxattr
= btrfs_listxattr
,
3742 .removexattr
= btrfs_removexattr
,
3744 static struct inode_operations btrfs_symlink_inode_operations
= {
3745 .readlink
= generic_readlink
,
3746 .follow_link
= page_follow_link_light
,
3747 .put_link
= page_put_link
,
3748 .permission
= btrfs_permission
,