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/module.h>
20 #include <linux/buffer_head.h>
22 #include <linux/pagemap.h>
23 #include <linux/highmem.h>
24 #include <linux/time.h>
25 #include <linux/init.h>
26 #include <linux/string.h>
27 #include <linux/smp_lock.h>
28 #include <linux/backing-dev.h>
29 #include <linux/mpage.h>
30 #include <linux/swap.h>
31 #include <linux/writeback.h>
32 #include <linux/statfs.h>
33 #include <linux/compat.h>
34 #include <linux/bit_spinlock.h>
37 #include "transaction.h"
38 #include "btrfs_inode.h"
40 #include "print-tree.h"
42 struct btrfs_iget_args
{
44 struct btrfs_root
*root
;
47 static struct inode_operations btrfs_dir_inode_operations
;
48 static struct inode_operations btrfs_symlink_inode_operations
;
49 static struct inode_operations btrfs_dir_ro_inode_operations
;
50 static struct inode_operations btrfs_file_inode_operations
;
51 static struct address_space_operations btrfs_aops
;
52 static struct address_space_operations btrfs_symlink_aops
;
53 static struct file_operations btrfs_dir_file_operations
;
55 static struct kmem_cache
*btrfs_inode_cachep
;
56 struct kmem_cache
*btrfs_trans_handle_cachep
;
57 struct kmem_cache
*btrfs_transaction_cachep
;
58 struct kmem_cache
*btrfs_bit_radix_cachep
;
59 struct kmem_cache
*btrfs_path_cachep
;
62 static unsigned char btrfs_type_by_mode
[S_IFMT
>> S_SHIFT
] = {
63 [S_IFREG
>> S_SHIFT
] = BTRFS_FT_REG_FILE
,
64 [S_IFDIR
>> S_SHIFT
] = BTRFS_FT_DIR
,
65 [S_IFCHR
>> S_SHIFT
] = BTRFS_FT_CHRDEV
,
66 [S_IFBLK
>> S_SHIFT
] = BTRFS_FT_BLKDEV
,
67 [S_IFIFO
>> S_SHIFT
] = BTRFS_FT_FIFO
,
68 [S_IFSOCK
>> S_SHIFT
] = BTRFS_FT_SOCK
,
69 [S_IFLNK
>> S_SHIFT
] = BTRFS_FT_SYMLINK
,
72 void btrfs_read_locked_inode(struct inode
*inode
)
74 struct btrfs_path
*path
;
75 struct btrfs_inode_item
*inode_item
;
76 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
77 struct btrfs_key location
;
78 u64 alloc_group_block
;
81 path
= btrfs_alloc_path();
83 mutex_lock(&root
->fs_info
->fs_mutex
);
85 memcpy(&location
, &BTRFS_I(inode
)->location
, sizeof(location
));
86 ret
= btrfs_lookup_inode(NULL
, root
, path
, &location
, 0);
88 btrfs_free_path(path
);
91 inode_item
= btrfs_item_ptr(btrfs_buffer_leaf(path
->nodes
[0]),
93 struct btrfs_inode_item
);
95 inode
->i_mode
= btrfs_inode_mode(inode_item
);
96 inode
->i_nlink
= btrfs_inode_nlink(inode_item
);
97 inode
->i_uid
= btrfs_inode_uid(inode_item
);
98 inode
->i_gid
= btrfs_inode_gid(inode_item
);
99 inode
->i_size
= btrfs_inode_size(inode_item
);
100 inode
->i_atime
.tv_sec
= btrfs_timespec_sec(&inode_item
->atime
);
101 inode
->i_atime
.tv_nsec
= btrfs_timespec_nsec(&inode_item
->atime
);
102 inode
->i_mtime
.tv_sec
= btrfs_timespec_sec(&inode_item
->mtime
);
103 inode
->i_mtime
.tv_nsec
= btrfs_timespec_nsec(&inode_item
->mtime
);
104 inode
->i_ctime
.tv_sec
= btrfs_timespec_sec(&inode_item
->ctime
);
105 inode
->i_ctime
.tv_nsec
= btrfs_timespec_nsec(&inode_item
->ctime
);
106 inode
->i_blocks
= btrfs_inode_nblocks(inode_item
);
107 inode
->i_generation
= btrfs_inode_generation(inode_item
);
108 alloc_group_block
= btrfs_inode_block_group(inode_item
);
109 BTRFS_I(inode
)->block_group
= btrfs_lookup_block_group(root
->fs_info
,
112 btrfs_free_path(path
);
115 mutex_unlock(&root
->fs_info
->fs_mutex
);
117 switch (inode
->i_mode
& S_IFMT
) {
120 init_special_inode(inode
, inode
->i_mode
,
121 btrfs_inode_rdev(inode_item
));
125 inode
->i_mapping
->a_ops
= &btrfs_aops
;
126 inode
->i_fop
= &btrfs_file_operations
;
127 inode
->i_op
= &btrfs_file_inode_operations
;
130 inode
->i_fop
= &btrfs_dir_file_operations
;
131 if (root
== root
->fs_info
->tree_root
)
132 inode
->i_op
= &btrfs_dir_ro_inode_operations
;
134 inode
->i_op
= &btrfs_dir_inode_operations
;
137 inode
->i_op
= &btrfs_symlink_inode_operations
;
138 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
144 btrfs_release_path(root
, path
);
145 btrfs_free_path(path
);
146 mutex_unlock(&root
->fs_info
->fs_mutex
);
147 make_bad_inode(inode
);
150 static void fill_inode_item(struct btrfs_inode_item
*item
,
153 btrfs_set_inode_uid(item
, inode
->i_uid
);
154 btrfs_set_inode_gid(item
, inode
->i_gid
);
155 btrfs_set_inode_size(item
, inode
->i_size
);
156 btrfs_set_inode_mode(item
, inode
->i_mode
);
157 btrfs_set_inode_nlink(item
, inode
->i_nlink
);
158 btrfs_set_timespec_sec(&item
->atime
, inode
->i_atime
.tv_sec
);
159 btrfs_set_timespec_nsec(&item
->atime
, inode
->i_atime
.tv_nsec
);
160 btrfs_set_timespec_sec(&item
->mtime
, inode
->i_mtime
.tv_sec
);
161 btrfs_set_timespec_nsec(&item
->mtime
, inode
->i_mtime
.tv_nsec
);
162 btrfs_set_timespec_sec(&item
->ctime
, inode
->i_ctime
.tv_sec
);
163 btrfs_set_timespec_nsec(&item
->ctime
, inode
->i_ctime
.tv_nsec
);
164 btrfs_set_inode_nblocks(item
, inode
->i_blocks
);
165 btrfs_set_inode_generation(item
, inode
->i_generation
);
166 btrfs_set_inode_block_group(item
,
167 BTRFS_I(inode
)->block_group
->key
.objectid
);
170 static int btrfs_update_inode(struct btrfs_trans_handle
*trans
,
171 struct btrfs_root
*root
,
174 struct btrfs_inode_item
*inode_item
;
175 struct btrfs_path
*path
;
178 path
= btrfs_alloc_path();
180 ret
= btrfs_lookup_inode(trans
, root
, path
,
181 &BTRFS_I(inode
)->location
, 1);
188 inode_item
= btrfs_item_ptr(btrfs_buffer_leaf(path
->nodes
[0]),
190 struct btrfs_inode_item
);
192 fill_inode_item(inode_item
, inode
);
193 btrfs_mark_buffer_dirty(path
->nodes
[0]);
196 btrfs_release_path(root
, path
);
197 btrfs_free_path(path
);
202 static int btrfs_unlink_trans(struct btrfs_trans_handle
*trans
,
203 struct btrfs_root
*root
,
205 struct dentry
*dentry
)
207 struct btrfs_path
*path
;
208 const char *name
= dentry
->d_name
.name
;
209 int name_len
= dentry
->d_name
.len
;
212 struct btrfs_dir_item
*di
;
214 path
= btrfs_alloc_path();
220 di
= btrfs_lookup_dir_item(trans
, root
, path
, dir
->i_ino
,
230 objectid
= btrfs_disk_key_objectid(&di
->location
);
231 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
234 btrfs_release_path(root
, path
);
236 di
= btrfs_lookup_dir_index_item(trans
, root
, path
, dir
->i_ino
,
237 objectid
, name
, name_len
, -1);
246 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
248 dentry
->d_inode
->i_ctime
= dir
->i_ctime
;
250 btrfs_free_path(path
);
252 dir
->i_size
-= name_len
* 2;
253 btrfs_update_inode(trans
, root
, dir
);
254 drop_nlink(dentry
->d_inode
);
255 ret
= btrfs_update_inode(trans
, root
, dentry
->d_inode
);
256 dir
->i_sb
->s_dirt
= 1;
261 static int btrfs_unlink(struct inode
*dir
, struct dentry
*dentry
)
263 struct btrfs_root
*root
;
264 struct btrfs_trans_handle
*trans
;
267 root
= BTRFS_I(dir
)->root
;
268 mutex_lock(&root
->fs_info
->fs_mutex
);
269 trans
= btrfs_start_transaction(root
, 1);
270 btrfs_set_trans_block_group(trans
, dir
);
271 ret
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
272 btrfs_end_transaction(trans
, root
);
273 mutex_unlock(&root
->fs_info
->fs_mutex
);
274 btrfs_btree_balance_dirty(root
);
278 static int btrfs_rmdir(struct inode
*dir
, struct dentry
*dentry
)
280 struct inode
*inode
= dentry
->d_inode
;
283 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
284 struct btrfs_path
*path
;
285 struct btrfs_key key
;
286 struct btrfs_trans_handle
*trans
;
287 struct btrfs_key found_key
;
289 struct btrfs_leaf
*leaf
;
290 char *goodnames
= "..";
292 path
= btrfs_alloc_path();
294 mutex_lock(&root
->fs_info
->fs_mutex
);
295 trans
= btrfs_start_transaction(root
, 1);
296 btrfs_set_trans_block_group(trans
, dir
);
297 key
.objectid
= inode
->i_ino
;
298 key
.offset
= (u64
)-1;
301 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
307 if (path
->slots
[0] == 0) {
312 leaf
= btrfs_buffer_leaf(path
->nodes
[0]);
313 btrfs_disk_key_to_cpu(&found_key
,
314 &leaf
->items
[path
->slots
[0]].key
);
315 found_type
= btrfs_key_type(&found_key
);
316 if (found_key
.objectid
!= inode
->i_ino
) {
320 if ((found_type
!= BTRFS_DIR_ITEM_KEY
&&
321 found_type
!= BTRFS_DIR_INDEX_KEY
) ||
322 (!btrfs_match_dir_item_name(root
, path
, goodnames
, 2) &&
323 !btrfs_match_dir_item_name(root
, path
, goodnames
, 1))) {
327 ret
= btrfs_del_item(trans
, root
, path
);
330 if (found_type
== BTRFS_DIR_ITEM_KEY
&& found_key
.offset
== 1)
332 btrfs_release_path(root
, path
);
335 btrfs_release_path(root
, path
);
337 /* now the directory is empty */
338 err
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
343 btrfs_release_path(root
, path
);
344 btrfs_free_path(path
);
345 mutex_unlock(&root
->fs_info
->fs_mutex
);
346 ret
= btrfs_end_transaction(trans
, root
);
347 btrfs_btree_balance_dirty(root
);
353 static int btrfs_free_inode(struct btrfs_trans_handle
*trans
,
354 struct btrfs_root
*root
,
357 struct btrfs_path
*path
;
362 path
= btrfs_alloc_path();
364 ret
= btrfs_lookup_inode(trans
, root
, path
,
365 &BTRFS_I(inode
)->location
, -1);
369 ret
= btrfs_del_item(trans
, root
, path
);
370 btrfs_free_path(path
);
375 * truncates go from a high offset to a low offset. So, walk
376 * from hi to lo in the node and issue readas. Stop when you find
377 * keys from a different objectid
379 static void reada_truncate(struct btrfs_root
*root
, struct btrfs_path
*path
,
382 struct btrfs_node
*node
;
392 node
= btrfs_buffer_node(path
->nodes
[1]);
393 slot
= path
->slots
[1];
396 nritems
= btrfs_header_nritems(&node
->header
);
397 for (i
= slot
- 1; i
>= 0; i
--) {
398 item_objectid
= btrfs_disk_key_objectid(&node
->ptrs
[i
].key
);
399 if (item_objectid
!= objectid
)
401 blocknr
= btrfs_node_blockptr(node
, i
);
402 ret
= readahead_tree_block(root
, blocknr
);
409 * this can truncate away extent items, csum items and directory items.
410 * It starts at a high offset and removes keys until it can't find
411 * any higher than i_size.
413 * csum items that cross the new i_size are truncated to the new size
416 static int btrfs_truncate_in_trans(struct btrfs_trans_handle
*trans
,
417 struct btrfs_root
*root
,
421 struct btrfs_path
*path
;
422 struct btrfs_key key
;
423 struct btrfs_disk_key
*found_key
;
425 struct btrfs_leaf
*leaf
;
426 struct btrfs_file_extent_item
*fi
;
427 u64 extent_start
= 0;
428 u64 extent_num_blocks
= 0;
433 path
= btrfs_alloc_path();
435 /* FIXME, add redo link to tree so we don't leak on crash */
436 key
.objectid
= inode
->i_ino
;
437 key
.offset
= (u64
)-1;
440 btrfs_init_path(path
);
442 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
447 BUG_ON(path
->slots
[0] == 0);
450 reada_truncate(root
, path
, inode
->i_ino
);
451 leaf
= btrfs_buffer_leaf(path
->nodes
[0]);
452 found_key
= &leaf
->items
[path
->slots
[0]].key
;
453 found_type
= btrfs_disk_key_type(found_key
);
455 if (btrfs_disk_key_objectid(found_key
) != inode
->i_ino
)
457 if (found_type
!= BTRFS_CSUM_ITEM_KEY
&&
458 found_type
!= BTRFS_DIR_ITEM_KEY
&&
459 found_type
!= BTRFS_DIR_INDEX_KEY
&&
460 found_type
!= BTRFS_EXTENT_DATA_KEY
)
463 item_end
= btrfs_disk_key_offset(found_key
);
464 if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
465 fi
= btrfs_item_ptr(btrfs_buffer_leaf(path
->nodes
[0]),
467 struct btrfs_file_extent_item
);
468 if (btrfs_file_extent_type(fi
) !=
469 BTRFS_FILE_EXTENT_INLINE
) {
470 item_end
+= btrfs_file_extent_num_blocks(fi
) <<
474 if (found_type
== BTRFS_CSUM_ITEM_KEY
) {
475 ret
= btrfs_csum_truncate(trans
, root
, path
,
479 if (item_end
< inode
->i_size
) {
481 btrfs_set_key_type(&key
, found_type
- 1);
486 if (btrfs_disk_key_offset(found_key
) >= inode
->i_size
)
492 /* FIXME, shrink the extent if the ref count is only 1 */
493 if (found_type
== BTRFS_EXTENT_DATA_KEY
&&
494 btrfs_file_extent_type(fi
) !=
495 BTRFS_FILE_EXTENT_INLINE
) {
498 u64 orig_num_blocks
=
499 btrfs_file_extent_num_blocks(fi
);
500 extent_num_blocks
= inode
->i_size
-
501 btrfs_disk_key_offset(found_key
) +
503 extent_num_blocks
>>= inode
->i_blkbits
;
504 btrfs_set_file_extent_num_blocks(fi
,
506 inode
->i_blocks
-= (orig_num_blocks
-
507 extent_num_blocks
) << 3;
508 mark_buffer_dirty(path
->nodes
[0]);
511 btrfs_file_extent_disk_blocknr(fi
);
513 btrfs_file_extent_disk_num_blocks(fi
);
514 /* FIXME blocksize != 4096 */
515 num_dec
= btrfs_file_extent_num_blocks(fi
) << 3;
516 if (extent_start
!= 0) {
518 inode
->i_blocks
-= num_dec
;
523 ret
= btrfs_del_item(trans
, root
, path
);
529 btrfs_release_path(root
, path
);
531 ret
= btrfs_free_extent(trans
, root
, extent_start
,
532 extent_num_blocks
, 0);
538 btrfs_release_path(root
, path
);
539 btrfs_free_path(path
);
540 inode
->i_sb
->s_dirt
= 1;
545 * taken from block_truncate_page, but does cow as it zeros out
546 * any bytes left in the last page in the file.
548 static int btrfs_truncate_page(struct address_space
*mapping
, loff_t from
)
550 struct inode
*inode
= mapping
->host
;
551 unsigned blocksize
= 1 << inode
->i_blkbits
;
552 pgoff_t index
= from
>> PAGE_CACHE_SHIFT
;
553 unsigned offset
= from
& (PAGE_CACHE_SIZE
-1);
557 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
559 struct btrfs_key ins
;
560 struct btrfs_trans_handle
*trans
;
562 if ((offset
& (blocksize
- 1)) == 0)
566 page
= grab_cache_page(mapping
, index
);
570 if (!PageUptodate(page
)) {
571 ret
= btrfs_readpage(NULL
, page
);
573 if (!PageUptodate(page
)) {
578 mutex_lock(&root
->fs_info
->fs_mutex
);
579 trans
= btrfs_start_transaction(root
, 1);
580 btrfs_set_trans_block_group(trans
, inode
);
582 ret
= btrfs_drop_extents(trans
, root
, inode
,
583 page
->index
<< PAGE_CACHE_SHIFT
,
584 (page
->index
+ 1) << PAGE_CACHE_SHIFT
,
588 ret
= btrfs_alloc_extent(trans
, root
, inode
->i_ino
, 1,
589 alloc_hint
, (u64
)-1, &ins
, 1);
592 ret
= btrfs_insert_file_extent(trans
, root
, inode
->i_ino
,
593 page
->index
<< PAGE_CACHE_SHIFT
,
597 SetPageChecked(page
);
599 memset(kaddr
+ offset
, 0, PAGE_CACHE_SIZE
- offset
);
600 flush_dcache_page(page
);
601 ret
= btrfs_csum_file_block(trans
, root
, inode
->i_ino
,
602 page
->index
<< PAGE_CACHE_SHIFT
,
603 kaddr
, PAGE_CACHE_SIZE
);
605 btrfs_end_transaction(trans
, root
);
606 mutex_unlock(&root
->fs_info
->fs_mutex
);
608 set_page_dirty(page
);
610 page_cache_release(page
);
615 static int btrfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
617 struct inode
*inode
= dentry
->d_inode
;
620 err
= inode_change_ok(inode
, attr
);
624 if (S_ISREG(inode
->i_mode
) &&
625 attr
->ia_valid
& ATTR_SIZE
&& attr
->ia_size
> inode
->i_size
) {
626 struct btrfs_trans_handle
*trans
;
627 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
628 u64 mask
= root
->blocksize
- 1;
629 u64 pos
= (inode
->i_size
+ mask
) & ~mask
;
632 if (attr
->ia_size
<= pos
)
635 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
637 hole_size
= (attr
->ia_size
- pos
+ mask
) & ~mask
;
638 hole_size
>>= inode
->i_blkbits
;
640 mutex_lock(&root
->fs_info
->fs_mutex
);
641 trans
= btrfs_start_transaction(root
, 1);
642 btrfs_set_trans_block_group(trans
, inode
);
643 err
= btrfs_insert_file_extent(trans
, root
, inode
->i_ino
,
644 pos
, 0, 0, hole_size
);
645 btrfs_end_transaction(trans
, root
);
646 mutex_unlock(&root
->fs_info
->fs_mutex
);
651 err
= inode_setattr(inode
, attr
);
655 void btrfs_delete_inode(struct inode
*inode
)
657 struct btrfs_trans_handle
*trans
;
658 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
661 truncate_inode_pages(&inode
->i_data
, 0);
662 if (is_bad_inode(inode
)) {
666 mutex_lock(&root
->fs_info
->fs_mutex
);
667 trans
= btrfs_start_transaction(root
, 1);
668 btrfs_set_trans_block_group(trans
, inode
);
669 ret
= btrfs_truncate_in_trans(trans
, root
, inode
);
672 ret
= btrfs_free_inode(trans
, root
, inode
);
675 btrfs_end_transaction(trans
, root
);
676 mutex_unlock(&root
->fs_info
->fs_mutex
);
677 btrfs_btree_balance_dirty(root
);
681 btrfs_end_transaction(trans
, root
);
682 mutex_unlock(&root
->fs_info
->fs_mutex
);
683 btrfs_btree_balance_dirty(root
);
689 * this returns the key found in the dir entry in the location pointer.
690 * If no dir entries were found, location->objectid is 0.
692 static int btrfs_inode_by_name(struct inode
*dir
, struct dentry
*dentry
,
693 struct btrfs_key
*location
)
695 const char *name
= dentry
->d_name
.name
;
696 int namelen
= dentry
->d_name
.len
;
697 struct btrfs_dir_item
*di
;
698 struct btrfs_path
*path
;
699 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
702 path
= btrfs_alloc_path();
704 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir
->i_ino
, name
,
706 if (!di
|| IS_ERR(di
)) {
707 location
->objectid
= 0;
711 btrfs_disk_key_to_cpu(location
, &di
->location
);
713 btrfs_release_path(root
, path
);
714 btrfs_free_path(path
);
719 * when we hit a tree root in a directory, the btrfs part of the inode
720 * needs to be changed to reflect the root directory of the tree root. This
721 * is kind of like crossing a mount point.
723 static int fixup_tree_root_location(struct btrfs_root
*root
,
724 struct btrfs_key
*location
,
725 struct btrfs_root
**sub_root
)
727 struct btrfs_path
*path
;
728 struct btrfs_root_item
*ri
;
730 if (btrfs_key_type(location
) != BTRFS_ROOT_ITEM_KEY
)
732 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
735 path
= btrfs_alloc_path();
737 mutex_lock(&root
->fs_info
->fs_mutex
);
739 *sub_root
= btrfs_read_fs_root(root
->fs_info
, location
);
740 if (IS_ERR(*sub_root
))
741 return PTR_ERR(*sub_root
);
743 ri
= &(*sub_root
)->root_item
;
744 location
->objectid
= btrfs_root_dirid(ri
);
746 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
747 location
->offset
= 0;
749 btrfs_free_path(path
);
750 mutex_unlock(&root
->fs_info
->fs_mutex
);
754 static int btrfs_init_locked_inode(struct inode
*inode
, void *p
)
756 struct btrfs_iget_args
*args
= p
;
757 inode
->i_ino
= args
->ino
;
758 BTRFS_I(inode
)->root
= args
->root
;
762 static int btrfs_find_actor(struct inode
*inode
, void *opaque
)
764 struct btrfs_iget_args
*args
= opaque
;
765 return (args
->ino
== inode
->i_ino
&&
766 args
->root
== BTRFS_I(inode
)->root
);
769 struct inode
*btrfs_iget_locked(struct super_block
*s
, u64 objectid
,
770 struct btrfs_root
*root
)
773 struct btrfs_iget_args args
;
777 inode
= iget5_locked(s
, objectid
, btrfs_find_actor
,
778 btrfs_init_locked_inode
,
783 static struct dentry
*btrfs_lookup(struct inode
*dir
, struct dentry
*dentry
,
784 struct nameidata
*nd
)
786 struct inode
* inode
;
787 struct btrfs_inode
*bi
= BTRFS_I(dir
);
788 struct btrfs_root
*root
= bi
->root
;
789 struct btrfs_root
*sub_root
= root
;
790 struct btrfs_key location
;
793 if (dentry
->d_name
.len
> BTRFS_NAME_LEN
)
794 return ERR_PTR(-ENAMETOOLONG
);
795 mutex_lock(&root
->fs_info
->fs_mutex
);
796 ret
= btrfs_inode_by_name(dir
, dentry
, &location
);
797 mutex_unlock(&root
->fs_info
->fs_mutex
);
801 if (location
.objectid
) {
802 ret
= fixup_tree_root_location(root
, &location
, &sub_root
);
806 return ERR_PTR(-ENOENT
);
807 inode
= btrfs_iget_locked(dir
->i_sb
, location
.objectid
,
810 return ERR_PTR(-EACCES
);
811 if (inode
->i_state
& I_NEW
) {
812 /* the inode and parent dir are two different roots */
813 if (sub_root
!= root
) {
815 sub_root
->inode
= inode
;
817 BTRFS_I(inode
)->root
= sub_root
;
818 memcpy(&BTRFS_I(inode
)->location
, &location
,
820 btrfs_read_locked_inode(inode
);
821 unlock_new_inode(inode
);
824 return d_splice_alias(inode
, dentry
);
828 * readahead one full node of leaves as long as their keys include
829 * the objectid supplied
831 static void reada_leaves(struct btrfs_root
*root
, struct btrfs_path
*path
,
834 struct btrfs_node
*node
;
844 node
= btrfs_buffer_node(path
->nodes
[1]);
845 slot
= path
->slots
[1];
846 nritems
= btrfs_header_nritems(&node
->header
);
847 for (i
= slot
+ 1; i
< nritems
; i
++) {
848 item_objectid
= btrfs_disk_key_objectid(&node
->ptrs
[i
].key
);
849 if (item_objectid
!= objectid
)
851 blocknr
= btrfs_node_blockptr(node
, i
);
852 ret
= readahead_tree_block(root
, blocknr
);
857 static unsigned char btrfs_filetype_table
[] = {
858 DT_UNKNOWN
, DT_REG
, DT_DIR
, DT_CHR
, DT_BLK
, DT_FIFO
, DT_SOCK
, DT_LNK
861 static int btrfs_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
863 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
864 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
865 struct btrfs_item
*item
;
866 struct btrfs_dir_item
*di
;
867 struct btrfs_key key
;
868 struct btrfs_path
*path
;
871 struct btrfs_leaf
*leaf
;
874 unsigned char d_type
;
879 int key_type
= BTRFS_DIR_INDEX_KEY
;
881 /* FIXME, use a real flag for deciding about the key type */
882 if (root
->fs_info
->tree_root
== root
)
883 key_type
= BTRFS_DIR_ITEM_KEY
;
884 mutex_lock(&root
->fs_info
->fs_mutex
);
885 key
.objectid
= inode
->i_ino
;
887 btrfs_set_key_type(&key
, key_type
);
888 key
.offset
= filp
->f_pos
;
889 path
= btrfs_alloc_path();
890 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
894 reada_leaves(root
, path
, inode
->i_ino
);
896 leaf
= btrfs_buffer_leaf(path
->nodes
[0]);
897 nritems
= btrfs_header_nritems(&leaf
->header
);
898 slot
= path
->slots
[0];
899 if (advance
|| slot
>= nritems
) {
900 if (slot
>= nritems
-1) {
901 reada_leaves(root
, path
, inode
->i_ino
);
902 ret
= btrfs_next_leaf(root
, path
);
905 leaf
= btrfs_buffer_leaf(path
->nodes
[0]);
906 nritems
= btrfs_header_nritems(&leaf
->header
);
907 slot
= path
->slots
[0];
914 item
= leaf
->items
+ slot
;
915 if (btrfs_disk_key_objectid(&item
->key
) != key
.objectid
)
917 if (btrfs_disk_key_type(&item
->key
) != key_type
)
919 if (btrfs_disk_key_offset(&item
->key
) < filp
->f_pos
)
921 filp
->f_pos
= btrfs_disk_key_offset(&item
->key
);
923 di
= btrfs_item_ptr(leaf
, slot
, struct btrfs_dir_item
);
925 di_total
= btrfs_item_size(leaf
->items
+ slot
);
926 while(di_cur
< di_total
) {
927 d_type
= btrfs_filetype_table
[btrfs_dir_type(di
)];
928 over
= filldir(dirent
, (const char *)(di
+ 1),
929 btrfs_dir_name_len(di
),
930 btrfs_disk_key_offset(&item
->key
),
931 btrfs_disk_key_objectid(&di
->location
),
935 di_len
= btrfs_dir_name_len(di
) + sizeof(*di
);
937 di
= (struct btrfs_dir_item
*)((char *)di
+ di_len
);
944 btrfs_release_path(root
, path
);
945 btrfs_free_path(path
);
946 mutex_unlock(&root
->fs_info
->fs_mutex
);
950 int btrfs_write_inode(struct inode
*inode
, int wait
)
952 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
953 struct btrfs_trans_handle
*trans
;
957 mutex_lock(&root
->fs_info
->fs_mutex
);
958 trans
= btrfs_start_transaction(root
, 1);
959 btrfs_set_trans_block_group(trans
, inode
);
960 ret
= btrfs_commit_transaction(trans
, root
);
961 mutex_unlock(&root
->fs_info
->fs_mutex
);
967 * This is somewhat expensive, updating the tree every time the
968 * inode changes. But, it is most likely to find the inode in cache.
969 * FIXME, needs more benchmarking...there are no reasons other than performance
970 * to keep or drop this code.
972 void btrfs_dirty_inode(struct inode
*inode
)
974 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
975 struct btrfs_trans_handle
*trans
;
977 mutex_lock(&root
->fs_info
->fs_mutex
);
978 trans
= btrfs_start_transaction(root
, 1);
979 btrfs_set_trans_block_group(trans
, inode
);
980 btrfs_update_inode(trans
, root
, inode
);
981 btrfs_end_transaction(trans
, root
);
982 mutex_unlock(&root
->fs_info
->fs_mutex
);
985 static struct inode
*btrfs_new_inode(struct btrfs_trans_handle
*trans
,
986 struct btrfs_root
*root
,
988 struct btrfs_block_group_cache
*group
,
992 struct btrfs_inode_item inode_item
;
993 struct btrfs_key
*location
;
997 inode
= new_inode(root
->fs_info
->sb
);
999 return ERR_PTR(-ENOMEM
);
1001 BTRFS_I(inode
)->root
= root
;
1006 group
= btrfs_find_block_group(root
, group
, 0, 0, owner
);
1007 BTRFS_I(inode
)->block_group
= group
;
1009 inode
->i_uid
= current
->fsuid
;
1010 inode
->i_gid
= current
->fsgid
;
1011 inode
->i_mode
= mode
;
1012 inode
->i_ino
= objectid
;
1013 inode
->i_blocks
= 0;
1014 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1015 fill_inode_item(&inode_item
, inode
);
1016 location
= &BTRFS_I(inode
)->location
;
1017 location
->objectid
= objectid
;
1018 location
->flags
= 0;
1019 location
->offset
= 0;
1020 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1022 ret
= btrfs_insert_inode(trans
, root
, objectid
, &inode_item
);
1024 return ERR_PTR(ret
);
1025 insert_inode_hash(inode
);
1029 static inline u8
btrfs_inode_type(struct inode
*inode
)
1031 return btrfs_type_by_mode
[(inode
->i_mode
& S_IFMT
) >> S_SHIFT
];
1034 static int btrfs_add_link(struct btrfs_trans_handle
*trans
,
1035 struct dentry
*dentry
, struct inode
*inode
)
1038 struct btrfs_key key
;
1039 struct btrfs_root
*root
= BTRFS_I(dentry
->d_parent
->d_inode
)->root
;
1040 key
.objectid
= inode
->i_ino
;
1042 btrfs_set_key_type(&key
, BTRFS_INODE_ITEM_KEY
);
1045 ret
= btrfs_insert_dir_item(trans
, root
,
1046 dentry
->d_name
.name
, dentry
->d_name
.len
,
1047 dentry
->d_parent
->d_inode
->i_ino
,
1048 &key
, btrfs_inode_type(inode
));
1050 dentry
->d_parent
->d_inode
->i_size
+= dentry
->d_name
.len
* 2;
1051 ret
= btrfs_update_inode(trans
, root
,
1052 dentry
->d_parent
->d_inode
);
1057 static int btrfs_add_nondir(struct btrfs_trans_handle
*trans
,
1058 struct dentry
*dentry
, struct inode
*inode
)
1060 int err
= btrfs_add_link(trans
, dentry
, inode
);
1062 d_instantiate(dentry
, inode
);
1070 static int btrfs_create(struct inode
*dir
, struct dentry
*dentry
,
1071 int mode
, struct nameidata
*nd
)
1073 struct btrfs_trans_handle
*trans
;
1074 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1075 struct inode
*inode
;
1080 mutex_lock(&root
->fs_info
->fs_mutex
);
1081 trans
= btrfs_start_transaction(root
, 1);
1082 btrfs_set_trans_block_group(trans
, dir
);
1084 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
1090 inode
= btrfs_new_inode(trans
, root
, objectid
,
1091 BTRFS_I(dir
)->block_group
, mode
);
1092 err
= PTR_ERR(inode
);
1096 btrfs_set_trans_block_group(trans
, inode
);
1097 err
= btrfs_add_nondir(trans
, dentry
, inode
);
1101 inode
->i_mapping
->a_ops
= &btrfs_aops
;
1102 inode
->i_fop
= &btrfs_file_operations
;
1103 inode
->i_op
= &btrfs_file_inode_operations
;
1105 dir
->i_sb
->s_dirt
= 1;
1106 btrfs_update_inode_block_group(trans
, inode
);
1107 btrfs_update_inode_block_group(trans
, dir
);
1109 btrfs_end_transaction(trans
, root
);
1110 mutex_unlock(&root
->fs_info
->fs_mutex
);
1113 inode_dec_link_count(inode
);
1116 btrfs_btree_balance_dirty(root
);
1120 static int btrfs_link(struct dentry
*old_dentry
, struct inode
*dir
,
1121 struct dentry
*dentry
)
1123 struct btrfs_trans_handle
*trans
;
1124 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1125 struct inode
*inode
= old_dentry
->d_inode
;
1129 if (inode
->i_nlink
== 0)
1133 mutex_lock(&root
->fs_info
->fs_mutex
);
1134 trans
= btrfs_start_transaction(root
, 1);
1135 btrfs_set_trans_block_group(trans
, dir
);
1136 atomic_inc(&inode
->i_count
);
1137 err
= btrfs_add_nondir(trans
, dentry
, inode
);
1140 dir
->i_sb
->s_dirt
= 1;
1141 btrfs_update_inode_block_group(trans
, dir
);
1142 err
= btrfs_update_inode(trans
, root
, inode
);
1146 btrfs_end_transaction(trans
, root
);
1147 mutex_unlock(&root
->fs_info
->fs_mutex
);
1150 inode_dec_link_count(inode
);
1153 btrfs_btree_balance_dirty(root
);
1157 static int btrfs_make_empty_dir(struct btrfs_trans_handle
*trans
,
1158 struct btrfs_root
*root
,
1159 u64 objectid
, u64 dirid
)
1163 struct btrfs_key key
;
1168 key
.objectid
= objectid
;
1171 btrfs_set_key_type(&key
, BTRFS_INODE_ITEM_KEY
);
1173 ret
= btrfs_insert_dir_item(trans
, root
, buf
, 1, objectid
,
1174 &key
, BTRFS_FT_DIR
);
1177 key
.objectid
= dirid
;
1178 ret
= btrfs_insert_dir_item(trans
, root
, buf
, 2, objectid
,
1179 &key
, BTRFS_FT_DIR
);
1186 static int btrfs_mkdir(struct inode
*dir
, struct dentry
*dentry
, int mode
)
1188 struct inode
*inode
;
1189 struct btrfs_trans_handle
*trans
;
1190 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1192 int drop_on_err
= 0;
1195 mutex_lock(&root
->fs_info
->fs_mutex
);
1196 trans
= btrfs_start_transaction(root
, 1);
1197 btrfs_set_trans_block_group(trans
, dir
);
1198 if (IS_ERR(trans
)) {
1199 err
= PTR_ERR(trans
);
1203 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
1209 inode
= btrfs_new_inode(trans
, root
, objectid
,
1210 BTRFS_I(dir
)->block_group
, S_IFDIR
| mode
);
1211 if (IS_ERR(inode
)) {
1212 err
= PTR_ERR(inode
);
1216 inode
->i_op
= &btrfs_dir_inode_operations
;
1217 inode
->i_fop
= &btrfs_dir_file_operations
;
1218 btrfs_set_trans_block_group(trans
, inode
);
1220 err
= btrfs_make_empty_dir(trans
, root
, inode
->i_ino
, dir
->i_ino
);
1225 err
= btrfs_update_inode(trans
, root
, inode
);
1228 err
= btrfs_add_link(trans
, dentry
, inode
);
1231 d_instantiate(dentry
, inode
);
1233 dir
->i_sb
->s_dirt
= 1;
1234 btrfs_update_inode_block_group(trans
, inode
);
1235 btrfs_update_inode_block_group(trans
, dir
);
1238 btrfs_end_transaction(trans
, root
);
1240 mutex_unlock(&root
->fs_info
->fs_mutex
);
1243 btrfs_btree_balance_dirty(root
);
1248 * FIBMAP and others want to pass in a fake buffer head. They need to
1249 * use BTRFS_GET_BLOCK_NO_DIRECT to make sure we don't try to memcpy
1250 * any packed file data into the fake bh
1252 #define BTRFS_GET_BLOCK_NO_CREATE 0
1253 #define BTRFS_GET_BLOCK_CREATE 1
1254 #define BTRFS_GET_BLOCK_NO_DIRECT 2
1257 * FIXME create==1 doe not work.
1259 static int btrfs_get_block_lock(struct inode
*inode
, sector_t iblock
,
1260 struct buffer_head
*result
, int create
)
1265 u64 extent_start
= 0;
1267 u64 objectid
= inode
->i_ino
;
1270 struct btrfs_path
*path
;
1271 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1272 struct btrfs_file_extent_item
*item
;
1273 struct btrfs_leaf
*leaf
;
1274 struct btrfs_disk_key
*found_key
;
1275 struct btrfs_trans_handle
*trans
= NULL
;
1277 path
= btrfs_alloc_path();
1279 if (create
& BTRFS_GET_BLOCK_CREATE
) {
1281 * danger!, this only works if the page is properly up
1284 trans
= btrfs_start_transaction(root
, 1);
1289 ret
= btrfs_drop_extents(trans
, root
, inode
,
1290 iblock
<< inode
->i_blkbits
,
1291 (iblock
+ 1) << inode
->i_blkbits
,
1296 ret
= btrfs_lookup_file_extent(NULL
, root
, path
,
1298 iblock
<< inode
->i_blkbits
, 0);
1305 if (path
->slots
[0] == 0) {
1306 btrfs_release_path(root
, path
);
1312 item
= btrfs_item_ptr(btrfs_buffer_leaf(path
->nodes
[0]), path
->slots
[0],
1313 struct btrfs_file_extent_item
);
1314 leaf
= btrfs_buffer_leaf(path
->nodes
[0]);
1315 blocknr
= btrfs_file_extent_disk_blocknr(item
);
1316 blocknr
+= btrfs_file_extent_offset(item
);
1318 /* are we inside the extent that was found? */
1319 found_key
= &leaf
->items
[path
->slots
[0]].key
;
1320 found_type
= btrfs_disk_key_type(found_key
);
1321 if (btrfs_disk_key_objectid(found_key
) != objectid
||
1322 found_type
!= BTRFS_EXTENT_DATA_KEY
) {
1327 found_type
= btrfs_file_extent_type(item
);
1328 extent_start
= btrfs_disk_key_offset(&leaf
->items
[path
->slots
[0]].key
);
1329 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
1330 extent_start
= extent_start
>> inode
->i_blkbits
;
1331 extent_end
= extent_start
+ btrfs_file_extent_num_blocks(item
);
1333 if (btrfs_file_extent_disk_blocknr(item
) == 0)
1335 if (iblock
>= extent_start
&& iblock
< extent_end
) {
1336 btrfs_map_bh_to_logical(root
, result
, blocknr
+
1337 iblock
- extent_start
);
1340 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
1345 if (create
& BTRFS_GET_BLOCK_NO_DIRECT
) {
1349 size
= btrfs_file_extent_inline_len(leaf
->items
+
1351 extent_end
= (extent_start
+ size
) >> inode
->i_blkbits
;
1352 extent_start
>>= inode
->i_blkbits
;
1353 if (iblock
< extent_start
|| iblock
> extent_end
) {
1356 ptr
= btrfs_file_extent_inline_start(item
);
1357 map
= kmap(result
->b_page
);
1358 memcpy(map
, ptr
, size
);
1359 memset(map
+ size
, 0, PAGE_CACHE_SIZE
- size
);
1360 flush_dcache_page(result
->b_page
);
1361 kunmap(result
->b_page
);
1362 set_buffer_uptodate(result
);
1363 SetPageChecked(result
->b_page
);
1364 btrfs_map_bh_to_logical(root
, result
, 0);
1367 if (create
& BTRFS_GET_BLOCK_CREATE
) {
1368 struct btrfs_key ins
;
1369 ret
= btrfs_alloc_extent(trans
, root
, inode
->i_ino
,
1370 1, alloc_hint
, (u64
)-1,
1376 ret
= btrfs_insert_file_extent(trans
, root
, inode
->i_ino
,
1377 iblock
<< inode
->i_blkbits
,
1378 ins
.objectid
, ins
.offset
,
1384 btrfs_map_bh_to_logical(root
, result
, ins
.objectid
);
1388 ret
= btrfs_end_transaction(trans
, root
);
1392 btrfs_free_path(path
);
1396 int btrfs_get_block(struct inode
*inode
, sector_t iblock
,
1397 struct buffer_head
*result
, int create
)
1400 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1401 mutex_lock(&root
->fs_info
->fs_mutex
);
1402 err
= btrfs_get_block_lock(inode
, iblock
, result
, create
);
1403 mutex_unlock(&root
->fs_info
->fs_mutex
);
1407 static int btrfs_get_block_csum(struct inode
*inode
, sector_t iblock
,
1408 struct buffer_head
*result
, int create
)
1411 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1412 struct page
*page
= result
->b_page
;
1413 u64 offset
= (page
->index
<< PAGE_CACHE_SHIFT
) + bh_offset(result
);
1414 struct btrfs_csum_item
*item
;
1415 struct btrfs_path
*path
= NULL
;
1417 mutex_lock(&root
->fs_info
->fs_mutex
);
1418 ret
= btrfs_get_block_lock(inode
, iblock
, result
, create
);
1422 path
= btrfs_alloc_path();
1423 item
= btrfs_lookup_csum(NULL
, root
, path
, inode
->i_ino
, offset
, 0);
1425 ret
= PTR_ERR(item
);
1426 /* a csum that isn't present is a preallocated region. */
1427 if (ret
== -ENOENT
|| ret
== -EFBIG
)
1429 result
->b_private
= NULL
;
1432 memcpy((char *)&result
->b_private
, &item
->csum
, BTRFS_CRC32_SIZE
);
1435 btrfs_free_path(path
);
1436 mutex_unlock(&root
->fs_info
->fs_mutex
);
1440 static int btrfs_get_block_bmap(struct inode
*inode
, sector_t iblock
,
1441 struct buffer_head
*result
, int create
)
1443 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1444 mutex_lock(&root
->fs_info
->fs_mutex
);
1445 btrfs_get_block_lock(inode
, iblock
, result
, BTRFS_GET_BLOCK_NO_DIRECT
);
1446 mutex_unlock(&root
->fs_info
->fs_mutex
);
1450 static sector_t
btrfs_bmap(struct address_space
*as
, sector_t block
)
1452 return generic_block_bmap(as
, block
, btrfs_get_block_bmap
);
1455 static int btrfs_prepare_write(struct file
*file
, struct page
*page
,
1456 unsigned from
, unsigned to
)
1458 return block_prepare_write(page
, from
, to
, btrfs_get_block
);
1461 static void buffer_io_error(struct buffer_head
*bh
)
1463 char b
[BDEVNAME_SIZE
];
1465 printk(KERN_ERR
"Buffer I/O error on device %s, logical block %Lu\n",
1466 bdevname(bh
->b_bdev
, b
),
1467 (unsigned long long)bh
->b_blocknr
);
1471 * I/O completion handler for block_read_full_page() - pages
1472 * which come unlocked at the end of I/O.
1474 static void btrfs_end_buffer_async_read(struct buffer_head
*bh
, int uptodate
)
1476 unsigned long flags
;
1477 struct buffer_head
*first
;
1478 struct buffer_head
*tmp
;
1480 int page_uptodate
= 1;
1481 struct inode
*inode
;
1484 BUG_ON(!buffer_async_read(bh
));
1487 inode
= page
->mapping
->host
;
1490 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
1491 if (bh
->b_private
) {
1492 char csum
[BTRFS_CRC32_SIZE
];
1493 kaddr
= kmap_atomic(page
, KM_IRQ0
);
1494 ret
= btrfs_csum_data(root
, kaddr
+ bh_offset(bh
),
1497 if (memcmp(csum
, &bh
->b_private
, BTRFS_CRC32_SIZE
)) {
1499 offset
= (page
->index
<< PAGE_CACHE_SHIFT
) +
1501 printk("btrfs csum failed ino %lu off %llu\n",
1502 page
->mapping
->host
->i_ino
,
1503 (unsigned long long)offset
);
1504 memset(kaddr
+ bh_offset(bh
), 1, bh
->b_size
);
1505 flush_dcache_page(page
);
1507 kunmap_atomic(kaddr
, KM_IRQ0
);
1509 set_buffer_uptodate(bh
);
1511 clear_buffer_uptodate(bh
);
1512 if (printk_ratelimit())
1513 buffer_io_error(bh
);
1518 * Be _very_ careful from here on. Bad things can happen if
1519 * two buffer heads end IO at almost the same time and both
1520 * decide that the page is now completely done.
1522 first
= page_buffers(page
);
1523 local_irq_save(flags
);
1524 bit_spin_lock(BH_Uptodate_Lock
, &first
->b_state
);
1525 clear_buffer_async_read(bh
);
1529 if (!buffer_uptodate(tmp
))
1531 if (buffer_async_read(tmp
)) {
1532 BUG_ON(!buffer_locked(tmp
));
1535 tmp
= tmp
->b_this_page
;
1536 } while (tmp
!= bh
);
1537 bit_spin_unlock(BH_Uptodate_Lock
, &first
->b_state
);
1538 local_irq_restore(flags
);
1541 * If none of the buffers had errors and they are all
1542 * uptodate then we can set the page uptodate.
1544 if (page_uptodate
&& !PageError(page
))
1545 SetPageUptodate(page
);
1550 bit_spin_unlock(BH_Uptodate_Lock
, &first
->b_state
);
1551 local_irq_restore(flags
);
1556 * Generic "read page" function for block devices that have the normal
1557 * get_block functionality. This is most of the block device filesystems.
1558 * Reads the page asynchronously --- the unlock_buffer() and
1559 * set/clear_buffer_uptodate() functions propagate buffer state into the
1560 * page struct once IO has completed.
1562 int btrfs_readpage(struct file
*file
, struct page
*page
)
1564 struct inode
*inode
= page
->mapping
->host
;
1565 sector_t iblock
, lblock
;
1566 struct buffer_head
*bh
, *head
, *arr
[MAX_BUF_PER_PAGE
];
1567 unsigned int blocksize
;
1569 int fully_mapped
= 1;
1571 BUG_ON(!PageLocked(page
));
1572 blocksize
= 1 << inode
->i_blkbits
;
1573 if (!page_has_buffers(page
))
1574 create_empty_buffers(page
, blocksize
, 0);
1575 head
= page_buffers(page
);
1577 iblock
= (sector_t
)page
->index
<< (PAGE_CACHE_SHIFT
- inode
->i_blkbits
);
1578 lblock
= (i_size_read(inode
)+blocksize
-1) >> inode
->i_blkbits
;
1584 if (buffer_uptodate(bh
))
1587 if (!buffer_mapped(bh
)) {
1591 if (iblock
< lblock
) {
1592 WARN_ON(bh
->b_size
!= blocksize
);
1593 err
= btrfs_get_block_csum(inode
, iblock
,
1598 if (!buffer_mapped(bh
)) {
1599 void *kaddr
= kmap_atomic(page
, KM_USER0
);
1600 memset(kaddr
+ i
* blocksize
, 0, blocksize
);
1601 flush_dcache_page(page
);
1602 kunmap_atomic(kaddr
, KM_USER0
);
1604 set_buffer_uptodate(bh
);
1608 * get_block() might have updated the buffer
1611 if (buffer_uptodate(bh
))
1615 } while (i
++, iblock
++, (bh
= bh
->b_this_page
) != head
);
1618 SetPageMappedToDisk(page
);
1622 * All buffers are uptodate - we can set the page uptodate
1623 * as well. But not if get_block() returned an error.
1625 if (!PageError(page
))
1626 SetPageUptodate(page
);
1631 /* Stage two: lock the buffers */
1632 for (i
= 0; i
< nr
; i
++) {
1635 bh
->b_end_io
= btrfs_end_buffer_async_read
;
1636 set_buffer_async_read(bh
);
1640 * Stage 3: start the IO. Check for uptodateness
1641 * inside the buffer lock in case another process reading
1642 * the underlying blockdev brought it uptodate (the sct fix).
1644 for (i
= 0; i
< nr
; i
++) {
1646 if (buffer_uptodate(bh
))
1647 btrfs_end_buffer_async_read(bh
, 1);
1649 submit_bh(READ
, bh
);
1655 * Aside from a tiny bit of packed file data handling, this is the
1656 * same as the generic code.
1658 * While block_write_full_page is writing back the dirty buffers under
1659 * the page lock, whoever dirtied the buffers may decide to clean them
1660 * again at any time. We handle that by only looking at the buffer
1661 * state inside lock_buffer().
1663 * If block_write_full_page() is called for regular writeback
1664 * (wbc->sync_mode == WB_SYNC_NONE) then it will redirty a page which has a
1665 * locked buffer. This only can happen if someone has written the buffer
1666 * directly, with submit_bh(). At the address_space level PageWriteback
1667 * prevents this contention from occurring.
1669 static int __btrfs_write_full_page(struct inode
*inode
, struct page
*page
,
1670 struct writeback_control
*wbc
)
1674 sector_t last_block
;
1675 struct buffer_head
*bh
, *head
;
1676 const unsigned blocksize
= 1 << inode
->i_blkbits
;
1677 int nr_underway
= 0;
1678 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1680 BUG_ON(!PageLocked(page
));
1682 last_block
= (i_size_read(inode
) - 1) >> inode
->i_blkbits
;
1684 /* no csumming allowed when from PF_MEMALLOC */
1685 if (current
->flags
& PF_MEMALLOC
) {
1686 redirty_page_for_writepage(wbc
, page
);
1691 if (!page_has_buffers(page
)) {
1692 create_empty_buffers(page
, blocksize
,
1693 (1 << BH_Dirty
)|(1 << BH_Uptodate
));
1697 * Be very careful. We have no exclusion from __set_page_dirty_buffers
1698 * here, and the (potentially unmapped) buffers may become dirty at
1699 * any time. If a buffer becomes dirty here after we've inspected it
1700 * then we just miss that fact, and the page stays dirty.
1702 * Buffers outside i_size may be dirtied by __set_page_dirty_buffers;
1703 * handle that here by just cleaning them.
1706 block
= (sector_t
)page
->index
<< (PAGE_CACHE_SHIFT
- inode
->i_blkbits
);
1707 head
= page_buffers(page
);
1711 * Get all the dirty buffers mapped to disk addresses and
1712 * handle any aliases from the underlying blockdev's mapping.
1715 if (block
> last_block
) {
1717 * mapped buffers outside i_size will occur, because
1718 * this page can be outside i_size when there is a
1719 * truncate in progress.
1722 * The buffer was zeroed by block_write_full_page()
1724 clear_buffer_dirty(bh
);
1725 set_buffer_uptodate(bh
);
1726 } else if (!buffer_mapped(bh
) && buffer_dirty(bh
)) {
1727 WARN_ON(bh
->b_size
!= blocksize
);
1728 err
= btrfs_get_block(inode
, block
, bh
, 0);
1732 if (buffer_new(bh
)) {
1733 /* blockdev mappings never come here */
1734 clear_buffer_new(bh
);
1737 bh
= bh
->b_this_page
;
1739 } while (bh
!= head
);
1742 if (!buffer_mapped(bh
))
1745 * If it's a fully non-blocking write attempt and we cannot
1746 * lock the buffer then redirty the page. Note that this can
1747 * potentially cause a busy-wait loop from pdflush and kswapd
1748 * activity, but those code paths have their own higher-level
1751 if (wbc
->sync_mode
!= WB_SYNC_NONE
|| !wbc
->nonblocking
) {
1753 } else if (test_set_buffer_locked(bh
)) {
1754 redirty_page_for_writepage(wbc
, page
);
1757 if (test_clear_buffer_dirty(bh
) && bh
->b_blocknr
!= 0) {
1758 struct btrfs_trans_handle
*trans
;
1760 u64 off
= page
->index
<< PAGE_CACHE_SHIFT
;
1763 off
+= bh_offset(bh
);
1764 mutex_lock(&root
->fs_info
->fs_mutex
);
1765 trans
= btrfs_start_transaction(root
, 1);
1766 btrfs_set_trans_block_group(trans
, inode
);
1768 btrfs_csum_file_block(trans
, root
, inode
->i_ino
,
1769 off
, kaddr
+ bh_offset(bh
),
1772 ret
= btrfs_end_transaction(trans
, root
);
1774 mutex_unlock(&root
->fs_info
->fs_mutex
);
1775 mark_buffer_async_write(bh
);
1779 } while ((bh
= bh
->b_this_page
) != head
);
1782 * The page and its buffers are protected by PageWriteback(), so we can
1783 * drop the bh refcounts early.
1785 BUG_ON(PageWriteback(page
));
1786 set_page_writeback(page
);
1789 struct buffer_head
*next
= bh
->b_this_page
;
1790 if (buffer_async_write(bh
)) {
1791 submit_bh(WRITE
, bh
);
1795 } while (bh
!= head
);
1800 if (nr_underway
== 0) {
1802 * The page was marked dirty, but the buffers were
1803 * clean. Someone wrote them back by hand with
1804 * ll_rw_block/submit_bh. A rare case.
1808 if (!buffer_uptodate(bh
)) {
1812 bh
= bh
->b_this_page
;
1813 } while (bh
!= head
);
1815 SetPageUptodate(page
);
1816 end_page_writeback(page
);
1822 * ENOSPC, or some other error. We may already have added some
1823 * blocks to the file, so we need to write these out to avoid
1824 * exposing stale data.
1825 * The page is currently locked and not marked for writeback
1828 /* Recovery: lock and submit the mapped buffers */
1830 if (buffer_mapped(bh
) && buffer_dirty(bh
)) {
1832 mark_buffer_async_write(bh
);
1835 * The buffer may have been set dirty during
1836 * attachment to a dirty page.
1838 clear_buffer_dirty(bh
);
1840 } while ((bh
= bh
->b_this_page
) != head
);
1842 BUG_ON(PageWriteback(page
));
1843 set_page_writeback(page
);
1845 struct buffer_head
*next
= bh
->b_this_page
;
1846 if (buffer_async_write(bh
)) {
1847 clear_buffer_dirty(bh
);
1848 submit_bh(WRITE
, bh
);
1852 } while (bh
!= head
);
1857 static int btrfs_writepage(struct page
*page
, struct writeback_control
*wbc
)
1859 struct inode
* const inode
= page
->mapping
->host
;
1860 loff_t i_size
= i_size_read(inode
);
1861 const pgoff_t end_index
= i_size
>> PAGE_CACHE_SHIFT
;
1865 /* Is the page fully inside i_size? */
1866 if (page
->index
< end_index
)
1867 return __btrfs_write_full_page(inode
, page
, wbc
);
1869 /* Is the page fully outside i_size? (truncate in progress) */
1870 offset
= i_size
& (PAGE_CACHE_SIZE
-1);
1871 if (page
->index
>= end_index
+1 || !offset
) {
1873 * The page may have dirty, unmapped buffers. For example,
1874 * they may have been added in ext3_writepage(). Make them
1875 * freeable here, so the page does not leak.
1877 block_invalidatepage(page
, 0);
1879 return 0; /* don't care */
1883 * The page straddles i_size. It must be zeroed out on each and every
1884 * writepage invokation because it may be mmapped. "A file is mapped
1885 * in multiples of the page size. For a file that is not a multiple of
1886 * the page size, the remaining memory is zeroed when mapped, and
1887 * writes to that region are not written out to the file."
1889 kaddr
= kmap_atomic(page
, KM_USER0
);
1890 memset(kaddr
+ offset
, 0, PAGE_CACHE_SIZE
- offset
);
1891 flush_dcache_page(page
);
1892 kunmap_atomic(kaddr
, KM_USER0
);
1893 return __btrfs_write_full_page(inode
, page
, wbc
);
1897 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
1898 * called from a page fault handler when a page is first dirtied. Hence we must
1899 * be careful to check for EOF conditions here. We set the page up correctly
1900 * for a written page which means we get ENOSPC checking when writing into
1901 * holes and correct delalloc and unwritten extent mapping on filesystems that
1902 * support these features.
1904 * We are not allowed to take the i_mutex here so we have to play games to
1905 * protect against truncate races as the page could now be beyond EOF. Because
1906 * vmtruncate() writes the inode size before removing pages, once we have the
1907 * page lock we can determine safely if the page is beyond EOF. If it is not
1908 * beyond EOF, then the page is guaranteed safe against truncation until we
1911 int btrfs_page_mkwrite(struct vm_area_struct
*vma
, struct page
*page
)
1913 struct inode
*inode
= vma
->vm_file
->f_path
.dentry
->d_inode
;
1919 wait_on_page_writeback(page
);
1920 size
= i_size_read(inode
);
1921 if ((page
->mapping
!= inode
->i_mapping
) ||
1922 ((page
->index
<< PAGE_CACHE_SHIFT
) > size
)) {
1923 /* page got truncated out from underneath us */
1927 /* page is wholly or partially inside EOF */
1928 if (((page
->index
+ 1) << PAGE_CACHE_SHIFT
) > size
)
1929 end
= size
& ~PAGE_CACHE_MASK
;
1931 end
= PAGE_CACHE_SIZE
;
1933 ret
= btrfs_prepare_write(NULL
, page
, 0, end
);
1935 ret
= btrfs_commit_write(NULL
, page
, 0, end
);
1942 static void btrfs_truncate(struct inode
*inode
)
1944 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1946 struct btrfs_trans_handle
*trans
;
1948 if (!S_ISREG(inode
->i_mode
))
1950 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
1953 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
1955 mutex_lock(&root
->fs_info
->fs_mutex
);
1956 trans
= btrfs_start_transaction(root
, 1);
1957 btrfs_set_trans_block_group(trans
, inode
);
1959 /* FIXME, add redo link to tree so we don't leak on crash */
1960 ret
= btrfs_truncate_in_trans(trans
, root
, inode
);
1961 btrfs_update_inode(trans
, root
, inode
);
1962 ret
= btrfs_end_transaction(trans
, root
);
1964 mutex_unlock(&root
->fs_info
->fs_mutex
);
1965 btrfs_btree_balance_dirty(root
);
1968 int btrfs_commit_write(struct file
*file
, struct page
*page
,
1969 unsigned from
, unsigned to
)
1971 struct inode
*inode
= page
->mapping
->host
;
1972 struct buffer_head
*bh
;
1973 loff_t pos
= ((loff_t
)page
->index
<< PAGE_CACHE_SHIFT
) + to
;
1975 SetPageUptodate(page
);
1976 bh
= page_buffers(page
);
1977 set_buffer_uptodate(bh
);
1978 if (buffer_mapped(bh
) && bh
->b_blocknr
!= 0) {
1979 set_page_dirty(page
);
1981 if (pos
> inode
->i_size
) {
1982 i_size_write(inode
, pos
);
1983 mark_inode_dirty(inode
);
1988 static int create_subvol(struct btrfs_root
*root
, char *name
, int namelen
)
1990 struct btrfs_trans_handle
*trans
;
1991 struct btrfs_key key
;
1992 struct btrfs_root_item root_item
;
1993 struct btrfs_inode_item
*inode_item
;
1994 struct buffer_head
*subvol
;
1995 struct btrfs_leaf
*leaf
;
1996 struct btrfs_root
*new_root
;
1997 struct inode
*inode
;
2002 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
2004 mutex_lock(&root
->fs_info
->fs_mutex
);
2005 trans
= btrfs_start_transaction(root
, 1);
2008 subvol
= btrfs_alloc_free_block(trans
, root
, 0);
2010 return PTR_ERR(subvol
);
2011 leaf
= btrfs_buffer_leaf(subvol
);
2012 btrfs_set_header_nritems(&leaf
->header
, 0);
2013 btrfs_set_header_level(&leaf
->header
, 0);
2014 btrfs_set_header_blocknr(&leaf
->header
, bh_blocknr(subvol
));
2015 btrfs_set_header_generation(&leaf
->header
, trans
->transid
);
2016 btrfs_set_header_owner(&leaf
->header
, root
->root_key
.objectid
);
2017 memcpy(leaf
->header
.fsid
, root
->fs_info
->disk_super
->fsid
,
2018 sizeof(leaf
->header
.fsid
));
2019 mark_buffer_dirty(subvol
);
2021 inode_item
= &root_item
.inode
;
2022 memset(inode_item
, 0, sizeof(*inode_item
));
2023 btrfs_set_inode_generation(inode_item
, 1);
2024 btrfs_set_inode_size(inode_item
, 3);
2025 btrfs_set_inode_nlink(inode_item
, 1);
2026 btrfs_set_inode_nblocks(inode_item
, 1);
2027 btrfs_set_inode_mode(inode_item
, S_IFDIR
| 0755);
2029 btrfs_set_root_blocknr(&root_item
, bh_blocknr(subvol
));
2030 btrfs_set_root_refs(&root_item
, 1);
2034 ret
= btrfs_find_free_objectid(trans
, root
->fs_info
->tree_root
,
2039 btrfs_set_root_dirid(&root_item
, new_dirid
);
2041 key
.objectid
= objectid
;
2044 btrfs_set_key_type(&key
, BTRFS_ROOT_ITEM_KEY
);
2045 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
2051 * insert the directory item
2053 key
.offset
= (u64
)-1;
2054 dir
= root
->fs_info
->sb
->s_root
->d_inode
;
2055 ret
= btrfs_insert_dir_item(trans
, root
->fs_info
->tree_root
,
2056 name
, namelen
, dir
->i_ino
, &key
,
2061 ret
= btrfs_commit_transaction(trans
, root
);
2065 new_root
= btrfs_read_fs_root(root
->fs_info
, &key
);
2068 trans
= btrfs_start_transaction(new_root
, 1);
2071 inode
= btrfs_new_inode(trans
, new_root
, new_dirid
,
2072 BTRFS_I(dir
)->block_group
, S_IFDIR
| 0700);
2075 inode
->i_op
= &btrfs_dir_inode_operations
;
2076 inode
->i_fop
= &btrfs_dir_file_operations
;
2077 new_root
->inode
= inode
;
2079 ret
= btrfs_make_empty_dir(trans
, new_root
, new_dirid
, new_dirid
);
2085 ret
= btrfs_update_inode(trans
, new_root
, inode
);
2089 err
= btrfs_commit_transaction(trans
, root
);
2093 mutex_unlock(&root
->fs_info
->fs_mutex
);
2094 btrfs_btree_balance_dirty(root
);
2098 static int create_snapshot(struct btrfs_root
*root
, char *name
, int namelen
)
2100 struct btrfs_trans_handle
*trans
;
2101 struct btrfs_key key
;
2102 struct btrfs_root_item new_root_item
;
2107 if (!root
->ref_cows
)
2110 mutex_lock(&root
->fs_info
->fs_mutex
);
2111 trans
= btrfs_start_transaction(root
, 1);
2114 ret
= btrfs_update_inode(trans
, root
, root
->inode
);
2118 ret
= btrfs_find_free_objectid(trans
, root
->fs_info
->tree_root
,
2123 memcpy(&new_root_item
, &root
->root_item
,
2124 sizeof(new_root_item
));
2126 key
.objectid
= objectid
;
2129 btrfs_set_key_type(&key
, BTRFS_ROOT_ITEM_KEY
);
2130 btrfs_set_root_blocknr(&new_root_item
, bh_blocknr(root
->node
));
2132 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
2138 * insert the directory item
2140 key
.offset
= (u64
)-1;
2141 ret
= btrfs_insert_dir_item(trans
, root
->fs_info
->tree_root
,
2143 root
->fs_info
->sb
->s_root
->d_inode
->i_ino
,
2144 &key
, BTRFS_FT_DIR
);
2149 ret
= btrfs_inc_root_ref(trans
, root
);
2154 err
= btrfs_commit_transaction(trans
, root
);
2157 mutex_unlock(&root
->fs_info
->fs_mutex
);
2158 btrfs_btree_balance_dirty(root
);
2162 int btrfs_ioctl(struct inode
*inode
, struct file
*filp
, unsigned int
2163 cmd
, unsigned long arg
)
2165 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2166 struct btrfs_ioctl_vol_args vol_args
;
2168 struct btrfs_dir_item
*di
;
2170 struct btrfs_path
*path
;
2174 case BTRFS_IOC_SNAP_CREATE
:
2175 if (copy_from_user(&vol_args
,
2176 (struct btrfs_ioctl_vol_args __user
*)arg
,
2179 namelen
= strlen(vol_args
.name
);
2180 if (namelen
> BTRFS_VOL_NAME_MAX
)
2182 if (strchr(vol_args
.name
, '/'))
2184 path
= btrfs_alloc_path();
2187 root_dirid
= root
->fs_info
->sb
->s_root
->d_inode
->i_ino
,
2188 mutex_lock(&root
->fs_info
->fs_mutex
);
2189 di
= btrfs_lookup_dir_item(NULL
, root
->fs_info
->tree_root
,
2191 vol_args
.name
, namelen
, 0);
2192 mutex_unlock(&root
->fs_info
->fs_mutex
);
2193 btrfs_free_path(path
);
2194 if (di
&& !IS_ERR(di
))
2199 if (root
== root
->fs_info
->tree_root
)
2200 ret
= create_subvol(root
, vol_args
.name
, namelen
);
2202 ret
= create_snapshot(root
, vol_args
.name
, namelen
);
2210 #ifdef CONFIG_COMPAT
2211 long btrfs_compat_ioctl(struct file
*file
, unsigned int cmd
,
2214 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
2217 ret
= btrfs_ioctl(inode
, file
, cmd
, (unsigned long) compat_ptr(arg
));
2225 * Called inside transaction, so use GFP_NOFS
2227 struct inode
*btrfs_alloc_inode(struct super_block
*sb
)
2229 struct btrfs_inode
*ei
;
2231 ei
= kmem_cache_alloc(btrfs_inode_cachep
, GFP_NOFS
);
2234 return &ei
->vfs_inode
;
2237 void btrfs_destroy_inode(struct inode
*inode
)
2239 WARN_ON(!list_empty(&inode
->i_dentry
));
2240 WARN_ON(inode
->i_data
.nrpages
);
2242 kmem_cache_free(btrfs_inode_cachep
, BTRFS_I(inode
));
2245 static void init_once(void * foo
, struct kmem_cache
* cachep
,
2246 unsigned long flags
)
2248 struct btrfs_inode
*ei
= (struct btrfs_inode
*) foo
;
2250 inode_init_once(&ei
->vfs_inode
);
2253 void btrfs_destroy_cachep(void)
2255 if (btrfs_inode_cachep
)
2256 kmem_cache_destroy(btrfs_inode_cachep
);
2257 if (btrfs_trans_handle_cachep
)
2258 kmem_cache_destroy(btrfs_trans_handle_cachep
);
2259 if (btrfs_transaction_cachep
)
2260 kmem_cache_destroy(btrfs_transaction_cachep
);
2261 if (btrfs_bit_radix_cachep
)
2262 kmem_cache_destroy(btrfs_bit_radix_cachep
);
2263 if (btrfs_path_cachep
)
2264 kmem_cache_destroy(btrfs_path_cachep
);
2267 int btrfs_init_cachep(void)
2269 btrfs_inode_cachep
= kmem_cache_create("btrfs_inode_cache",
2270 sizeof(struct btrfs_inode
),
2271 0, (SLAB_RECLAIM_ACCOUNT
|
2274 if (!btrfs_inode_cachep
)
2276 btrfs_trans_handle_cachep
= kmem_cache_create("btrfs_trans_handle_cache",
2277 sizeof(struct btrfs_trans_handle
),
2278 0, (SLAB_RECLAIM_ACCOUNT
|
2281 if (!btrfs_trans_handle_cachep
)
2283 btrfs_transaction_cachep
= kmem_cache_create("btrfs_transaction_cache",
2284 sizeof(struct btrfs_transaction
),
2285 0, (SLAB_RECLAIM_ACCOUNT
|
2288 if (!btrfs_transaction_cachep
)
2290 btrfs_path_cachep
= kmem_cache_create("btrfs_path_cache",
2291 sizeof(struct btrfs_transaction
),
2292 0, (SLAB_RECLAIM_ACCOUNT
|
2295 if (!btrfs_path_cachep
)
2297 btrfs_bit_radix_cachep
= kmem_cache_create("btrfs_radix",
2299 0, (SLAB_RECLAIM_ACCOUNT
|
2301 SLAB_DESTROY_BY_RCU
),
2303 if (!btrfs_bit_radix_cachep
)
2307 btrfs_destroy_cachep();
2311 static int btrfs_getattr(struct vfsmount
*mnt
,
2312 struct dentry
*dentry
, struct kstat
*stat
)
2314 struct inode
*inode
= dentry
->d_inode
;
2315 generic_fillattr(inode
, stat
);
2316 stat
->blksize
= 256 * 1024;
2320 static int btrfs_rename(struct inode
* old_dir
, struct dentry
*old_dentry
,
2321 struct inode
* new_dir
,struct dentry
*new_dentry
)
2323 struct btrfs_trans_handle
*trans
;
2324 struct btrfs_root
*root
= BTRFS_I(old_dir
)->root
;
2325 struct inode
*new_inode
= new_dentry
->d_inode
;
2326 struct inode
*old_inode
= old_dentry
->d_inode
;
2327 struct timespec ctime
= CURRENT_TIME
;
2328 struct btrfs_path
*path
;
2329 struct btrfs_dir_item
*di
;
2332 if (S_ISDIR(old_inode
->i_mode
) && new_inode
&&
2333 new_inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
2336 mutex_lock(&root
->fs_info
->fs_mutex
);
2337 trans
= btrfs_start_transaction(root
, 1);
2338 btrfs_set_trans_block_group(trans
, new_dir
);
2339 path
= btrfs_alloc_path();
2345 old_dentry
->d_inode
->i_nlink
++;
2346 old_dir
->i_ctime
= old_dir
->i_mtime
= ctime
;
2347 new_dir
->i_ctime
= new_dir
->i_mtime
= ctime
;
2348 old_inode
->i_ctime
= ctime
;
2349 if (S_ISDIR(old_inode
->i_mode
) && old_dir
!= new_dir
) {
2350 struct btrfs_key
*location
= &BTRFS_I(new_dir
)->location
;
2352 di
= btrfs_lookup_dir_item(trans
, root
, path
, old_inode
->i_ino
,
2362 old_parent_oid
= btrfs_disk_key_objectid(&di
->location
);
2363 ret
= btrfs_del_item(trans
, root
, path
);
2367 btrfs_release_path(root
, path
);
2369 di
= btrfs_lookup_dir_index_item(trans
, root
, path
,
2381 ret
= btrfs_del_item(trans
, root
, path
);
2385 btrfs_release_path(root
, path
);
2387 ret
= btrfs_insert_dir_item(trans
, root
, "..", 2,
2388 old_inode
->i_ino
, location
,
2395 ret
= btrfs_unlink_trans(trans
, root
, old_dir
, old_dentry
);
2400 new_inode
->i_ctime
= CURRENT_TIME
;
2401 ret
= btrfs_unlink_trans(trans
, root
, new_dir
, new_dentry
);
2404 if (S_ISDIR(new_inode
->i_mode
))
2405 clear_nlink(new_inode
);
2407 drop_nlink(new_inode
);
2408 ret
= btrfs_update_inode(trans
, root
, new_inode
);
2412 ret
= btrfs_add_link(trans
, new_dentry
, old_inode
);
2417 btrfs_free_path(path
);
2418 btrfs_end_transaction(trans
, root
);
2419 mutex_unlock(&root
->fs_info
->fs_mutex
);
2423 static int btrfs_symlink(struct inode
*dir
, struct dentry
*dentry
,
2424 const char *symname
)
2426 struct btrfs_trans_handle
*trans
;
2427 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2428 struct btrfs_path
*path
;
2429 struct btrfs_key key
;
2430 struct inode
*inode
;
2437 struct btrfs_file_extent_item
*ei
;
2439 name_len
= strlen(symname
) + 1;
2440 if (name_len
> BTRFS_MAX_INLINE_DATA_SIZE(root
))
2441 return -ENAMETOOLONG
;
2442 mutex_lock(&root
->fs_info
->fs_mutex
);
2443 trans
= btrfs_start_transaction(root
, 1);
2444 btrfs_set_trans_block_group(trans
, dir
);
2446 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2452 inode
= btrfs_new_inode(trans
, root
, objectid
,
2453 BTRFS_I(dir
)->block_group
, S_IFLNK
|S_IRWXUGO
);
2454 err
= PTR_ERR(inode
);
2458 btrfs_set_trans_block_group(trans
, inode
);
2459 err
= btrfs_add_nondir(trans
, dentry
, inode
);
2463 inode
->i_mapping
->a_ops
= &btrfs_aops
;
2464 inode
->i_fop
= &btrfs_file_operations
;
2465 inode
->i_op
= &btrfs_file_inode_operations
;
2467 dir
->i_sb
->s_dirt
= 1;
2468 btrfs_update_inode_block_group(trans
, inode
);
2469 btrfs_update_inode_block_group(trans
, dir
);
2473 path
= btrfs_alloc_path();
2475 key
.objectid
= inode
->i_ino
;
2478 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
2479 datasize
= btrfs_file_extent_calc_inline_size(name_len
);
2480 err
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
2486 ei
= btrfs_item_ptr(btrfs_buffer_leaf(path
->nodes
[0]),
2487 path
->slots
[0], struct btrfs_file_extent_item
);
2488 btrfs_set_file_extent_generation(ei
, trans
->transid
);
2489 btrfs_set_file_extent_type(ei
,
2490 BTRFS_FILE_EXTENT_INLINE
);
2491 ptr
= btrfs_file_extent_inline_start(ei
);
2492 btrfs_memcpy(root
, path
->nodes
[0]->b_data
,
2493 ptr
, symname
, name_len
);
2494 mark_buffer_dirty(path
->nodes
[0]);
2495 btrfs_free_path(path
);
2496 inode
->i_op
= &btrfs_symlink_inode_operations
;
2497 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
2498 inode
->i_size
= name_len
- 1;
2499 err
= btrfs_update_inode(trans
, root
, inode
);
2504 btrfs_end_transaction(trans
, root
);
2505 mutex_unlock(&root
->fs_info
->fs_mutex
);
2507 inode_dec_link_count(inode
);
2510 btrfs_btree_balance_dirty(root
);
2514 static struct inode_operations btrfs_dir_inode_operations
= {
2515 .lookup
= btrfs_lookup
,
2516 .create
= btrfs_create
,
2517 .unlink
= btrfs_unlink
,
2519 .mkdir
= btrfs_mkdir
,
2520 .rmdir
= btrfs_rmdir
,
2521 .rename
= btrfs_rename
,
2522 .symlink
= btrfs_symlink
,
2523 .setattr
= btrfs_setattr
,
2526 static struct inode_operations btrfs_dir_ro_inode_operations
= {
2527 .lookup
= btrfs_lookup
,
2530 static struct file_operations btrfs_dir_file_operations
= {
2531 .llseek
= generic_file_llseek
,
2532 .read
= generic_read_dir
,
2533 .readdir
= btrfs_readdir
,
2534 .ioctl
= btrfs_ioctl
,
2535 #ifdef CONFIG_COMPAT
2536 .compat_ioctl
= btrfs_compat_ioctl
,
2540 static struct address_space_operations btrfs_aops
= {
2541 .readpage
= btrfs_readpage
,
2542 .writepage
= btrfs_writepage
,
2543 .sync_page
= block_sync_page
,
2544 .prepare_write
= btrfs_prepare_write
,
2545 .commit_write
= btrfs_commit_write
,
2549 static struct address_space_operations btrfs_symlink_aops
= {
2550 .readpage
= btrfs_readpage
,
2551 .writepage
= btrfs_writepage
,
2554 static struct inode_operations btrfs_file_inode_operations
= {
2555 .truncate
= btrfs_truncate
,
2556 .getattr
= btrfs_getattr
,
2557 .setattr
= btrfs_setattr
,
2560 static struct inode_operations btrfs_symlink_inode_operations
= {
2561 .readlink
= generic_readlink
,
2562 .follow_link
= page_follow_link_light
,
2563 .put_link
= page_put_link
,