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/buffer_head.h>
21 #include <linux/pagemap.h>
22 #include <linux/highmem.h>
23 #include <linux/time.h>
24 #include <linux/init.h>
25 #include <linux/string.h>
26 #include <linux/smp_lock.h>
27 #include <linux/backing-dev.h>
28 #include <linux/mpage.h>
29 #include <linux/swap.h>
30 #include <linux/writeback.h>
31 #include <linux/statfs.h>
32 #include <linux/compat.h>
33 #include <linux/bit_spinlock.h>
34 #include <linux/version.h>
35 #include <linux/xattr.h>
38 #include "transaction.h"
39 #include "btrfs_inode.h"
41 #include "print-tree.h"
43 struct btrfs_iget_args
{
45 struct btrfs_root
*root
;
48 static struct inode_operations btrfs_dir_inode_operations
;
49 static struct inode_operations btrfs_symlink_inode_operations
;
50 static struct inode_operations btrfs_dir_ro_inode_operations
;
51 static struct inode_operations btrfs_special_inode_operations
;
52 static struct inode_operations btrfs_file_inode_operations
;
53 static struct address_space_operations btrfs_aops
;
54 static struct address_space_operations btrfs_symlink_aops
;
55 static struct file_operations btrfs_dir_file_operations
;
56 static struct extent_map_ops btrfs_extent_map_ops
;
58 static struct kmem_cache
*btrfs_inode_cachep
;
59 struct kmem_cache
*btrfs_trans_handle_cachep
;
60 struct kmem_cache
*btrfs_transaction_cachep
;
61 struct kmem_cache
*btrfs_bit_radix_cachep
;
62 struct kmem_cache
*btrfs_path_cachep
;
65 static unsigned char btrfs_type_by_mode
[S_IFMT
>> S_SHIFT
] = {
66 [S_IFREG
>> S_SHIFT
] = BTRFS_FT_REG_FILE
,
67 [S_IFDIR
>> S_SHIFT
] = BTRFS_FT_DIR
,
68 [S_IFCHR
>> S_SHIFT
] = BTRFS_FT_CHRDEV
,
69 [S_IFBLK
>> S_SHIFT
] = BTRFS_FT_BLKDEV
,
70 [S_IFIFO
>> S_SHIFT
] = BTRFS_FT_FIFO
,
71 [S_IFSOCK
>> S_SHIFT
] = BTRFS_FT_SOCK
,
72 [S_IFLNK
>> S_SHIFT
] = BTRFS_FT_SYMLINK
,
75 int btrfs_check_free_space(struct btrfs_root
*root
, u64 num_required
,
78 u64 total
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
79 u64 used
= btrfs_super_bytes_used(&root
->fs_info
->super_copy
);
84 thresh
= (total
* 90) / 100;
86 thresh
= (total
* 85) / 100;
88 spin_lock(&root
->fs_info
->delalloc_lock
);
89 if (used
+ root
->fs_info
->delalloc_bytes
+ num_required
> thresh
)
91 spin_unlock(&root
->fs_info
->delalloc_lock
);
95 static int cow_file_range(struct inode
*inode
, u64 start
, u64 end
)
97 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
98 struct btrfs_trans_handle
*trans
;
102 u64 blocksize
= root
->sectorsize
;
103 struct btrfs_key ins
;
106 trans
= btrfs_start_transaction(root
, 1);
108 btrfs_set_trans_block_group(trans
, inode
);
110 num_bytes
= (end
- start
+ blocksize
) & ~(blocksize
- 1);
111 num_bytes
= max(blocksize
, num_bytes
);
112 ret
= btrfs_drop_extents(trans
, root
, inode
,
113 start
, start
+ num_bytes
, start
, &alloc_hint
);
115 if (alloc_hint
== EXTENT_MAP_INLINE
)
118 while(num_bytes
> 0) {
119 cur_alloc_size
= min(num_bytes
, root
->fs_info
->max_extent
);
120 ret
= btrfs_alloc_extent(trans
, root
, cur_alloc_size
,
121 root
->root_key
.objectid
,
123 inode
->i_ino
, start
, 0,
124 alloc_hint
, (u64
)-1, &ins
, 1);
129 ret
= btrfs_insert_file_extent(trans
, root
, inode
->i_ino
,
130 start
, ins
.objectid
, ins
.offset
,
132 num_bytes
-= cur_alloc_size
;
133 alloc_hint
= ins
.objectid
+ ins
.offset
;
134 start
+= cur_alloc_size
;
137 btrfs_end_transaction(trans
, root
);
141 static int run_delalloc_nocow(struct inode
*inode
, u64 start
, u64 end
)
148 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
149 struct extent_buffer
*leaf
;
151 struct btrfs_path
*path
;
152 struct btrfs_file_extent_item
*item
;
155 struct btrfs_key found_key
;
157 path
= btrfs_alloc_path();
160 ret
= btrfs_lookup_file_extent(NULL
, root
, path
,
161 inode
->i_ino
, start
, 0);
163 btrfs_free_path(path
);
169 if (path
->slots
[0] == 0)
174 leaf
= path
->nodes
[0];
175 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
176 struct btrfs_file_extent_item
);
178 /* are we inside the extent that was found? */
179 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
180 found_type
= btrfs_key_type(&found_key
);
181 if (found_key
.objectid
!= inode
->i_ino
||
182 found_type
!= BTRFS_EXTENT_DATA_KEY
) {
186 found_type
= btrfs_file_extent_type(leaf
, item
);
187 extent_start
= found_key
.offset
;
188 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
189 extent_end
= extent_start
+
190 btrfs_file_extent_num_bytes(leaf
, item
);
193 if (loops
&& start
!= extent_start
)
196 if (start
< extent_start
|| start
>= extent_end
)
199 cow_end
= min(end
, extent_end
- 1);
200 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
204 if (btrfs_count_snapshots_in_path(root
, path
, bytenr
) != 1) {
209 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
214 btrfs_free_path(path
);
217 btrfs_release_path(root
, path
);
222 cow_file_range(inode
, start
, cow_end
);
227 static int run_delalloc_range(struct inode
*inode
, u64 start
, u64 end
)
229 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
233 mutex_lock(&root
->fs_info
->fs_mutex
);
234 if (btrfs_test_opt(root
, NODATACOW
))
235 ret
= run_delalloc_nocow(inode
, start
, end
);
237 ret
= cow_file_range(inode
, start
, end
);
239 spin_lock(&root
->fs_info
->delalloc_lock
);
240 num_bytes
= end
+ 1 - start
;
241 if (root
->fs_info
->delalloc_bytes
< num_bytes
) {
242 printk("delalloc accounting error total %llu sub %llu\n",
243 root
->fs_info
->delalloc_bytes
, num_bytes
);
245 root
->fs_info
->delalloc_bytes
-= num_bytes
;
247 spin_unlock(&root
->fs_info
->delalloc_lock
);
249 mutex_unlock(&root
->fs_info
->fs_mutex
);
253 int btrfs_writepage_io_hook(struct page
*page
, u64 start
, u64 end
)
255 struct inode
*inode
= page
->mapping
->host
;
256 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
257 struct btrfs_trans_handle
*trans
;
260 u64 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
261 size_t offset
= start
- page_start
;
263 if (btrfs_test_opt(root
, NODATASUM
))
266 mutex_lock(&root
->fs_info
->fs_mutex
);
267 trans
= btrfs_start_transaction(root
, 1);
268 btrfs_set_trans_block_group(trans
, inode
);
270 btrfs_csum_file_block(trans
, root
, inode
, inode
->i_ino
,
271 start
, kaddr
+ offset
, end
- start
+ 1);
273 ret
= btrfs_end_transaction(trans
, root
);
275 mutex_unlock(&root
->fs_info
->fs_mutex
);
279 int btrfs_readpage_io_hook(struct page
*page
, u64 start
, u64 end
)
282 struct inode
*inode
= page
->mapping
->host
;
283 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
284 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
285 struct btrfs_csum_item
*item
;
286 struct btrfs_path
*path
= NULL
;
289 if (btrfs_test_opt(root
, NODATASUM
))
292 mutex_lock(&root
->fs_info
->fs_mutex
);
293 path
= btrfs_alloc_path();
294 item
= btrfs_lookup_csum(NULL
, root
, path
, inode
->i_ino
, start
, 0);
297 /* a csum that isn't present is a preallocated region. */
298 if (ret
== -ENOENT
|| ret
== -EFBIG
)
303 read_extent_buffer(path
->nodes
[0], &csum
, (unsigned long)item
,
305 set_state_private(em_tree
, start
, csum
);
308 btrfs_free_path(path
);
309 mutex_unlock(&root
->fs_info
->fs_mutex
);
313 int btrfs_readpage_end_io_hook(struct page
*page
, u64 start
, u64 end
)
315 size_t offset
= start
- ((u64
)page
->index
<< PAGE_CACHE_SHIFT
);
316 struct inode
*inode
= page
->mapping
->host
;
317 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
321 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
325 if (btrfs_test_opt(root
, NODATASUM
))
328 ret
= get_state_private(em_tree
, start
, &private);
329 local_irq_save(flags
);
330 kaddr
= kmap_atomic(page
, KM_IRQ0
);
334 csum
= btrfs_csum_data(root
, kaddr
+ offset
, csum
, end
- start
+ 1);
335 btrfs_csum_final(csum
, (char *)&csum
);
336 if (csum
!= private) {
339 kunmap_atomic(kaddr
, KM_IRQ0
);
340 local_irq_restore(flags
);
344 printk("btrfs csum failed ino %lu off %llu\n",
345 page
->mapping
->host
->i_ino
, (unsigned long long)start
);
346 memset(kaddr
+ offset
, 1, end
- start
+ 1);
347 flush_dcache_page(page
);
348 kunmap_atomic(kaddr
, KM_IRQ0
);
349 local_irq_restore(flags
);
353 void btrfs_read_locked_inode(struct inode
*inode
)
355 struct btrfs_path
*path
;
356 struct extent_buffer
*leaf
;
357 struct btrfs_inode_item
*inode_item
;
358 struct btrfs_inode_timespec
*tspec
;
359 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
360 struct btrfs_key location
;
361 u64 alloc_group_block
;
365 path
= btrfs_alloc_path();
367 mutex_lock(&root
->fs_info
->fs_mutex
);
369 memcpy(&location
, &BTRFS_I(inode
)->location
, sizeof(location
));
370 ret
= btrfs_lookup_inode(NULL
, root
, path
, &location
, 0);
374 leaf
= path
->nodes
[0];
375 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
376 struct btrfs_inode_item
);
378 inode
->i_mode
= btrfs_inode_mode(leaf
, inode_item
);
379 inode
->i_nlink
= btrfs_inode_nlink(leaf
, inode_item
);
380 inode
->i_uid
= btrfs_inode_uid(leaf
, inode_item
);
381 inode
->i_gid
= btrfs_inode_gid(leaf
, inode_item
);
382 inode
->i_size
= btrfs_inode_size(leaf
, inode_item
);
384 tspec
= btrfs_inode_atime(inode_item
);
385 inode
->i_atime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
386 inode
->i_atime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
388 tspec
= btrfs_inode_mtime(inode_item
);
389 inode
->i_mtime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
390 inode
->i_mtime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
392 tspec
= btrfs_inode_ctime(inode_item
);
393 inode
->i_ctime
.tv_sec
= btrfs_timespec_sec(leaf
, tspec
);
394 inode
->i_ctime
.tv_nsec
= btrfs_timespec_nsec(leaf
, tspec
);
396 inode
->i_blocks
= btrfs_inode_nblocks(leaf
, inode_item
);
397 inode
->i_generation
= btrfs_inode_generation(leaf
, inode_item
);
399 rdev
= btrfs_inode_rdev(leaf
, inode_item
);
401 alloc_group_block
= btrfs_inode_block_group(leaf
, inode_item
);
402 BTRFS_I(inode
)->block_group
= btrfs_lookup_block_group(root
->fs_info
,
405 btrfs_free_path(path
);
408 mutex_unlock(&root
->fs_info
->fs_mutex
);
410 switch (inode
->i_mode
& S_IFMT
) {
412 inode
->i_mapping
->a_ops
= &btrfs_aops
;
413 BTRFS_I(inode
)->extent_tree
.ops
= &btrfs_extent_map_ops
;
414 inode
->i_fop
= &btrfs_file_operations
;
415 inode
->i_op
= &btrfs_file_inode_operations
;
418 inode
->i_fop
= &btrfs_dir_file_operations
;
419 if (root
== root
->fs_info
->tree_root
)
420 inode
->i_op
= &btrfs_dir_ro_inode_operations
;
422 inode
->i_op
= &btrfs_dir_inode_operations
;
425 inode
->i_op
= &btrfs_symlink_inode_operations
;
426 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
429 init_special_inode(inode
, inode
->i_mode
, rdev
);
435 btrfs_release_path(root
, path
);
436 btrfs_free_path(path
);
437 mutex_unlock(&root
->fs_info
->fs_mutex
);
438 make_bad_inode(inode
);
441 static void fill_inode_item(struct extent_buffer
*leaf
,
442 struct btrfs_inode_item
*item
,
445 btrfs_set_inode_uid(leaf
, item
, inode
->i_uid
);
446 btrfs_set_inode_gid(leaf
, item
, inode
->i_gid
);
447 btrfs_set_inode_size(leaf
, item
, inode
->i_size
);
448 btrfs_set_inode_mode(leaf
, item
, inode
->i_mode
);
449 btrfs_set_inode_nlink(leaf
, item
, inode
->i_nlink
);
451 btrfs_set_timespec_sec(leaf
, btrfs_inode_atime(item
),
452 inode
->i_atime
.tv_sec
);
453 btrfs_set_timespec_nsec(leaf
, btrfs_inode_atime(item
),
454 inode
->i_atime
.tv_nsec
);
456 btrfs_set_timespec_sec(leaf
, btrfs_inode_mtime(item
),
457 inode
->i_mtime
.tv_sec
);
458 btrfs_set_timespec_nsec(leaf
, btrfs_inode_mtime(item
),
459 inode
->i_mtime
.tv_nsec
);
461 btrfs_set_timespec_sec(leaf
, btrfs_inode_ctime(item
),
462 inode
->i_ctime
.tv_sec
);
463 btrfs_set_timespec_nsec(leaf
, btrfs_inode_ctime(item
),
464 inode
->i_ctime
.tv_nsec
);
466 btrfs_set_inode_nblocks(leaf
, item
, inode
->i_blocks
);
467 btrfs_set_inode_generation(leaf
, item
, inode
->i_generation
);
468 btrfs_set_inode_rdev(leaf
, item
, inode
->i_rdev
);
469 btrfs_set_inode_block_group(leaf
, item
,
470 BTRFS_I(inode
)->block_group
->key
.objectid
);
473 int btrfs_update_inode(struct btrfs_trans_handle
*trans
,
474 struct btrfs_root
*root
,
477 struct btrfs_inode_item
*inode_item
;
478 struct btrfs_path
*path
;
479 struct extent_buffer
*leaf
;
482 path
= btrfs_alloc_path();
484 ret
= btrfs_lookup_inode(trans
, root
, path
,
485 &BTRFS_I(inode
)->location
, 1);
492 leaf
= path
->nodes
[0];
493 inode_item
= btrfs_item_ptr(leaf
, path
->slots
[0],
494 struct btrfs_inode_item
);
496 fill_inode_item(leaf
, inode_item
, inode
);
497 btrfs_mark_buffer_dirty(leaf
);
498 btrfs_set_inode_last_trans(trans
, inode
);
501 btrfs_release_path(root
, path
);
502 btrfs_free_path(path
);
507 static int btrfs_unlink_trans(struct btrfs_trans_handle
*trans
,
508 struct btrfs_root
*root
,
510 struct dentry
*dentry
)
512 struct btrfs_path
*path
;
513 const char *name
= dentry
->d_name
.name
;
514 int name_len
= dentry
->d_name
.len
;
516 struct extent_buffer
*leaf
;
517 struct btrfs_dir_item
*di
;
518 struct btrfs_key key
;
520 path
= btrfs_alloc_path();
526 di
= btrfs_lookup_dir_item(trans
, root
, path
, dir
->i_ino
,
536 leaf
= path
->nodes
[0];
537 btrfs_dir_item_key_to_cpu(leaf
, di
, &key
);
538 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
541 btrfs_release_path(root
, path
);
543 di
= btrfs_lookup_dir_index_item(trans
, root
, path
, dir
->i_ino
,
544 key
.objectid
, name
, name_len
, -1);
553 ret
= btrfs_delete_one_dir_name(trans
, root
, path
, di
);
555 dentry
->d_inode
->i_ctime
= dir
->i_ctime
;
556 ret
= btrfs_del_inode_ref(trans
, root
, name
, name_len
,
557 dentry
->d_inode
->i_ino
,
558 dentry
->d_parent
->d_inode
->i_ino
);
560 printk("failed to delete reference to %.*s, "
561 "inode %lu parent %lu\n", name_len
, name
,
562 dentry
->d_inode
->i_ino
,
563 dentry
->d_parent
->d_inode
->i_ino
);
566 btrfs_free_path(path
);
568 dir
->i_size
-= name_len
* 2;
569 dir
->i_mtime
= dir
->i_ctime
= CURRENT_TIME
;
570 btrfs_update_inode(trans
, root
, dir
);
571 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
572 dentry
->d_inode
->i_nlink
--;
574 drop_nlink(dentry
->d_inode
);
576 ret
= btrfs_update_inode(trans
, root
, dentry
->d_inode
);
577 dir
->i_sb
->s_dirt
= 1;
582 static int btrfs_unlink(struct inode
*dir
, struct dentry
*dentry
)
584 struct btrfs_root
*root
;
585 struct btrfs_trans_handle
*trans
;
587 unsigned long nr
= 0;
589 root
= BTRFS_I(dir
)->root
;
590 mutex_lock(&root
->fs_info
->fs_mutex
);
592 ret
= btrfs_check_free_space(root
, 1, 1);
596 trans
= btrfs_start_transaction(root
, 1);
598 btrfs_set_trans_block_group(trans
, dir
);
599 ret
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
600 nr
= trans
->blocks_used
;
602 btrfs_end_transaction(trans
, root
);
604 mutex_unlock(&root
->fs_info
->fs_mutex
);
605 btrfs_btree_balance_dirty(root
, nr
);
609 static int btrfs_rmdir(struct inode
*dir
, struct dentry
*dentry
)
611 struct inode
*inode
= dentry
->d_inode
;
614 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
615 struct btrfs_trans_handle
*trans
;
616 unsigned long nr
= 0;
618 if (inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
)
621 mutex_lock(&root
->fs_info
->fs_mutex
);
622 ret
= btrfs_check_free_space(root
, 1, 1);
626 trans
= btrfs_start_transaction(root
, 1);
627 btrfs_set_trans_block_group(trans
, dir
);
629 /* now the directory is empty */
630 err
= btrfs_unlink_trans(trans
, root
, dir
, dentry
);
635 nr
= trans
->blocks_used
;
636 ret
= btrfs_end_transaction(trans
, root
);
638 mutex_unlock(&root
->fs_info
->fs_mutex
);
639 btrfs_btree_balance_dirty(root
, nr
);
646 static int btrfs_free_inode(struct btrfs_trans_handle
*trans
,
647 struct btrfs_root
*root
,
650 struct btrfs_path
*path
;
655 path
= btrfs_alloc_path();
657 ret
= btrfs_lookup_inode(trans
, root
, path
,
658 &BTRFS_I(inode
)->location
, -1);
662 ret
= btrfs_del_item(trans
, root
, path
);
663 btrfs_free_path(path
);
668 * this can truncate away extent items, csum items and directory items.
669 * It starts at a high offset and removes keys until it can't find
670 * any higher than i_size.
672 * csum items that cross the new i_size are truncated to the new size
675 static int btrfs_truncate_in_trans(struct btrfs_trans_handle
*trans
,
676 struct btrfs_root
*root
,
680 struct btrfs_path
*path
;
681 struct btrfs_key key
;
682 struct btrfs_key found_key
;
684 struct extent_buffer
*leaf
;
685 struct btrfs_file_extent_item
*fi
;
686 u64 extent_start
= 0;
687 u64 extent_num_bytes
= 0;
693 int extent_type
= -1;
695 btrfs_drop_extent_cache(inode
, inode
->i_size
, (u64
)-1);
696 path
= btrfs_alloc_path();
700 /* FIXME, add redo link to tree so we don't leak on crash */
701 key
.objectid
= inode
->i_ino
;
702 key
.offset
= (u64
)-1;
706 btrfs_init_path(path
);
708 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
713 BUG_ON(path
->slots
[0] == 0);
716 leaf
= path
->nodes
[0];
717 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
718 found_type
= btrfs_key_type(&found_key
);
720 if (found_key
.objectid
!= inode
->i_ino
)
723 if (found_type
!= BTRFS_CSUM_ITEM_KEY
&&
724 found_type
!= BTRFS_DIR_ITEM_KEY
&&
725 found_type
!= BTRFS_DIR_INDEX_KEY
&&
726 found_type
!= BTRFS_EXTENT_DATA_KEY
)
729 item_end
= found_key
.offset
;
730 if (found_type
== BTRFS_EXTENT_DATA_KEY
) {
731 fi
= btrfs_item_ptr(leaf
, path
->slots
[0],
732 struct btrfs_file_extent_item
);
733 extent_type
= btrfs_file_extent_type(leaf
, fi
);
734 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
736 btrfs_file_extent_num_bytes(leaf
, fi
);
737 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
) {
738 struct btrfs_item
*item
= btrfs_item_nr(leaf
,
740 item_end
+= btrfs_file_extent_inline_len(leaf
,
745 if (found_type
== BTRFS_CSUM_ITEM_KEY
) {
746 ret
= btrfs_csum_truncate(trans
, root
, path
,
750 if (item_end
< inode
->i_size
) {
751 if (found_type
== BTRFS_DIR_ITEM_KEY
) {
752 found_type
= BTRFS_INODE_ITEM_KEY
;
753 } else if (found_type
== BTRFS_EXTENT_ITEM_KEY
) {
754 found_type
= BTRFS_CSUM_ITEM_KEY
;
755 } else if (found_type
) {
760 btrfs_set_key_type(&key
, found_type
);
761 btrfs_release_path(root
, path
);
764 if (found_key
.offset
>= inode
->i_size
)
770 /* FIXME, shrink the extent if the ref count is only 1 */
771 if (found_type
!= BTRFS_EXTENT_DATA_KEY
)
774 if (extent_type
!= BTRFS_FILE_EXTENT_INLINE
) {
776 extent_start
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
779 btrfs_file_extent_num_bytes(leaf
, fi
);
780 extent_num_bytes
= inode
->i_size
-
781 found_key
.offset
+ root
->sectorsize
- 1;
782 btrfs_set_file_extent_num_bytes(leaf
, fi
,
784 num_dec
= (orig_num_bytes
-
785 extent_num_bytes
) >> 9;
786 if (extent_start
!= 0) {
787 inode
->i_blocks
-= num_dec
;
789 btrfs_mark_buffer_dirty(leaf
);
792 btrfs_file_extent_disk_num_bytes(leaf
,
794 /* FIXME blocksize != 4096 */
795 num_dec
= btrfs_file_extent_num_bytes(leaf
,
797 if (extent_start
!= 0) {
799 inode
->i_blocks
-= num_dec
;
801 root_gen
= btrfs_header_generation(leaf
);
802 root_owner
= btrfs_header_owner(leaf
);
804 } else if (extent_type
== BTRFS_FILE_EXTENT_INLINE
&&
806 u32 newsize
= inode
->i_size
- found_key
.offset
;
807 newsize
= btrfs_file_extent_calc_inline_size(newsize
);
808 ret
= btrfs_truncate_item(trans
, root
, path
,
814 ret
= btrfs_del_item(trans
, root
, path
);
820 btrfs_release_path(root
, path
);
822 ret
= btrfs_free_extent(trans
, root
, extent_start
,
825 root_gen
, inode
->i_ino
,
826 found_key
.offset
, 0);
832 btrfs_release_path(root
, path
);
833 btrfs_free_path(path
);
834 inode
->i_sb
->s_dirt
= 1;
838 static int btrfs_cow_one_page(struct inode
*inode
, struct page
*page
,
842 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
843 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
844 u64 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
845 u64 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
846 u64 existing_delalloc
;
850 WARN_ON(!PageLocked(page
));
851 set_page_extent_mapped(page
);
853 lock_extent(em_tree
, page_start
, page_end
, GFP_NOFS
);
854 delalloc_start
= page_start
;
855 existing_delalloc
= count_range_bits(&BTRFS_I(inode
)->extent_tree
,
856 &delalloc_start
, page_end
,
857 PAGE_CACHE_SIZE
, EXTENT_DELALLOC
);
858 set_extent_delalloc(&BTRFS_I(inode
)->extent_tree
, page_start
,
861 spin_lock(&root
->fs_info
->delalloc_lock
);
862 root
->fs_info
->delalloc_bytes
+= PAGE_CACHE_SIZE
- existing_delalloc
;
863 spin_unlock(&root
->fs_info
->delalloc_lock
);
865 if (zero_start
!= PAGE_CACHE_SIZE
) {
867 memset(kaddr
+ zero_start
, 0, PAGE_CACHE_SIZE
- zero_start
);
868 flush_dcache_page(page
);
871 set_page_dirty(page
);
872 unlock_extent(em_tree
, page_start
, page_end
, GFP_NOFS
);
878 * taken from block_truncate_page, but does cow as it zeros out
879 * any bytes left in the last page in the file.
881 static int btrfs_truncate_page(struct address_space
*mapping
, loff_t from
)
883 struct inode
*inode
= mapping
->host
;
884 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
885 u32 blocksize
= root
->sectorsize
;
886 pgoff_t index
= from
>> PAGE_CACHE_SHIFT
;
887 unsigned offset
= from
& (PAGE_CACHE_SIZE
-1);
892 if ((offset
& (blocksize
- 1)) == 0)
895 down_read(&root
->snap_sem
);
897 page
= grab_cache_page(mapping
, index
);
900 if (!PageUptodate(page
)) {
901 ret
= btrfs_readpage(NULL
, page
);
903 if (!PageUptodate(page
)) {
908 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
910 ret
= btrfs_cow_one_page(inode
, page
, offset
);
913 page_cache_release(page
);
914 up_read(&BTRFS_I(inode
)->root
->snap_sem
);
919 static int btrfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
921 struct inode
*inode
= dentry
->d_inode
;
924 err
= inode_change_ok(inode
, attr
);
928 if (S_ISREG(inode
->i_mode
) &&
929 attr
->ia_valid
& ATTR_SIZE
&& attr
->ia_size
> inode
->i_size
) {
930 struct btrfs_trans_handle
*trans
;
931 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
932 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
934 u64 mask
= root
->sectorsize
- 1;
935 u64 pos
= (inode
->i_size
+ mask
) & ~mask
;
936 u64 block_end
= attr
->ia_size
| mask
;
940 if (attr
->ia_size
<= pos
)
943 mutex_lock(&root
->fs_info
->fs_mutex
);
944 err
= btrfs_check_free_space(root
, 1, 0);
945 mutex_unlock(&root
->fs_info
->fs_mutex
);
949 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
951 lock_extent(em_tree
, pos
, block_end
, GFP_NOFS
);
952 hole_size
= (attr
->ia_size
- pos
+ mask
) & ~mask
;
954 mutex_lock(&root
->fs_info
->fs_mutex
);
955 trans
= btrfs_start_transaction(root
, 1);
956 btrfs_set_trans_block_group(trans
, inode
);
957 err
= btrfs_drop_extents(trans
, root
, inode
,
958 pos
, pos
+ hole_size
, pos
,
961 if (alloc_hint
!= EXTENT_MAP_INLINE
) {
962 err
= btrfs_insert_file_extent(trans
, root
,
964 pos
, 0, 0, hole_size
);
966 btrfs_end_transaction(trans
, root
);
967 mutex_unlock(&root
->fs_info
->fs_mutex
);
968 unlock_extent(em_tree
, pos
, block_end
, GFP_NOFS
);
973 err
= inode_setattr(inode
, attr
);
977 void btrfs_delete_inode(struct inode
*inode
)
979 struct btrfs_trans_handle
*trans
;
980 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
984 truncate_inode_pages(&inode
->i_data
, 0);
985 if (is_bad_inode(inode
)) {
990 mutex_lock(&root
->fs_info
->fs_mutex
);
991 trans
= btrfs_start_transaction(root
, 1);
993 btrfs_set_trans_block_group(trans
, inode
);
994 ret
= btrfs_truncate_in_trans(trans
, root
, inode
);
997 ret
= btrfs_delete_xattrs(trans
, root
, inode
);
1000 ret
= btrfs_free_inode(trans
, root
, inode
);
1002 goto no_delete_lock
;
1003 nr
= trans
->blocks_used
;
1005 btrfs_end_transaction(trans
, root
);
1006 mutex_unlock(&root
->fs_info
->fs_mutex
);
1007 btrfs_btree_balance_dirty(root
, nr
);
1011 nr
= trans
->blocks_used
;
1012 btrfs_end_transaction(trans
, root
);
1013 mutex_unlock(&root
->fs_info
->fs_mutex
);
1014 btrfs_btree_balance_dirty(root
, nr
);
1020 * this returns the key found in the dir entry in the location pointer.
1021 * If no dir entries were found, location->objectid is 0.
1023 static int btrfs_inode_by_name(struct inode
*dir
, struct dentry
*dentry
,
1024 struct btrfs_key
*location
)
1026 const char *name
= dentry
->d_name
.name
;
1027 int namelen
= dentry
->d_name
.len
;
1028 struct btrfs_dir_item
*di
;
1029 struct btrfs_path
*path
;
1030 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1033 if (namelen
== 1 && strcmp(name
, ".") == 0) {
1034 location
->objectid
= dir
->i_ino
;
1035 location
->type
= BTRFS_INODE_ITEM_KEY
;
1036 location
->offset
= 0;
1039 path
= btrfs_alloc_path();
1042 if (namelen
== 2 && strcmp(name
, "..") == 0) {
1043 struct btrfs_key key
;
1044 struct extent_buffer
*leaf
;
1048 key
.objectid
= dir
->i_ino
;
1049 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1051 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1055 leaf
= path
->nodes
[0];
1056 slot
= path
->slots
[0];
1057 nritems
= btrfs_header_nritems(leaf
);
1058 if (slot
>= nritems
)
1061 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
1062 if (key
.objectid
!= dir
->i_ino
||
1063 key
.type
!= BTRFS_INODE_REF_KEY
) {
1066 location
->objectid
= key
.offset
;
1067 location
->type
= BTRFS_INODE_ITEM_KEY
;
1068 location
->offset
= 0;
1072 di
= btrfs_lookup_dir_item(NULL
, root
, path
, dir
->i_ino
, name
,
1076 if (!di
|| IS_ERR(di
)) {
1079 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, location
);
1081 btrfs_free_path(path
);
1084 location
->objectid
= 0;
1089 * when we hit a tree root in a directory, the btrfs part of the inode
1090 * needs to be changed to reflect the root directory of the tree root. This
1091 * is kind of like crossing a mount point.
1093 static int fixup_tree_root_location(struct btrfs_root
*root
,
1094 struct btrfs_key
*location
,
1095 struct btrfs_root
**sub_root
,
1096 struct dentry
*dentry
)
1098 struct btrfs_path
*path
;
1099 struct btrfs_root_item
*ri
;
1101 if (btrfs_key_type(location
) != BTRFS_ROOT_ITEM_KEY
)
1103 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
1106 path
= btrfs_alloc_path();
1108 mutex_lock(&root
->fs_info
->fs_mutex
);
1110 *sub_root
= btrfs_read_fs_root(root
->fs_info
, location
,
1111 dentry
->d_name
.name
,
1112 dentry
->d_name
.len
);
1113 if (IS_ERR(*sub_root
))
1114 return PTR_ERR(*sub_root
);
1116 ri
= &(*sub_root
)->root_item
;
1117 location
->objectid
= btrfs_root_dirid(ri
);
1118 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1119 location
->offset
= 0;
1121 btrfs_free_path(path
);
1122 mutex_unlock(&root
->fs_info
->fs_mutex
);
1126 static int btrfs_init_locked_inode(struct inode
*inode
, void *p
)
1128 struct btrfs_iget_args
*args
= p
;
1129 inode
->i_ino
= args
->ino
;
1130 BTRFS_I(inode
)->root
= args
->root
;
1131 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
,
1132 inode
->i_mapping
, GFP_NOFS
);
1136 static int btrfs_find_actor(struct inode
*inode
, void *opaque
)
1138 struct btrfs_iget_args
*args
= opaque
;
1139 return (args
->ino
== inode
->i_ino
&&
1140 args
->root
== BTRFS_I(inode
)->root
);
1143 struct inode
*btrfs_iget_locked(struct super_block
*s
, u64 objectid
,
1144 struct btrfs_root
*root
)
1146 struct inode
*inode
;
1147 struct btrfs_iget_args args
;
1148 args
.ino
= objectid
;
1151 inode
= iget5_locked(s
, objectid
, btrfs_find_actor
,
1152 btrfs_init_locked_inode
,
1157 static struct dentry
*btrfs_lookup(struct inode
*dir
, struct dentry
*dentry
,
1158 struct nameidata
*nd
)
1160 struct inode
* inode
;
1161 struct btrfs_inode
*bi
= BTRFS_I(dir
);
1162 struct btrfs_root
*root
= bi
->root
;
1163 struct btrfs_root
*sub_root
= root
;
1164 struct btrfs_key location
;
1167 if (dentry
->d_name
.len
> BTRFS_NAME_LEN
)
1168 return ERR_PTR(-ENAMETOOLONG
);
1170 mutex_lock(&root
->fs_info
->fs_mutex
);
1171 ret
= btrfs_inode_by_name(dir
, dentry
, &location
);
1172 mutex_unlock(&root
->fs_info
->fs_mutex
);
1175 return ERR_PTR(ret
);
1178 if (location
.objectid
) {
1179 ret
= fixup_tree_root_location(root
, &location
, &sub_root
,
1182 return ERR_PTR(ret
);
1184 return ERR_PTR(-ENOENT
);
1185 inode
= btrfs_iget_locked(dir
->i_sb
, location
.objectid
,
1188 return ERR_PTR(-EACCES
);
1189 if (inode
->i_state
& I_NEW
) {
1190 /* the inode and parent dir are two different roots */
1191 if (sub_root
!= root
) {
1193 sub_root
->inode
= inode
;
1195 BTRFS_I(inode
)->root
= sub_root
;
1196 memcpy(&BTRFS_I(inode
)->location
, &location
,
1198 btrfs_read_locked_inode(inode
);
1199 unlock_new_inode(inode
);
1202 return d_splice_alias(inode
, dentry
);
1205 static unsigned char btrfs_filetype_table
[] = {
1206 DT_UNKNOWN
, DT_REG
, DT_DIR
, DT_CHR
, DT_BLK
, DT_FIFO
, DT_SOCK
, DT_LNK
1209 static int btrfs_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
1211 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1212 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1213 struct btrfs_item
*item
;
1214 struct btrfs_dir_item
*di
;
1215 struct btrfs_key key
;
1216 struct btrfs_key found_key
;
1217 struct btrfs_path
*path
;
1220 struct extent_buffer
*leaf
;
1223 unsigned char d_type
;
1228 int key_type
= BTRFS_DIR_INDEX_KEY
;
1233 /* FIXME, use a real flag for deciding about the key type */
1234 if (root
->fs_info
->tree_root
== root
)
1235 key_type
= BTRFS_DIR_ITEM_KEY
;
1237 /* special case for "." */
1238 if (filp
->f_pos
== 0) {
1239 over
= filldir(dirent
, ".", 1,
1247 mutex_lock(&root
->fs_info
->fs_mutex
);
1248 key
.objectid
= inode
->i_ino
;
1249 path
= btrfs_alloc_path();
1252 /* special case for .., just use the back ref */
1253 if (filp
->f_pos
== 1) {
1254 btrfs_set_key_type(&key
, BTRFS_INODE_REF_KEY
);
1256 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1258 leaf
= path
->nodes
[0];
1259 slot
= path
->slots
[0];
1260 nritems
= btrfs_header_nritems(leaf
);
1261 if (slot
>= nritems
) {
1262 btrfs_release_path(root
, path
);
1263 goto read_dir_items
;
1265 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1266 btrfs_release_path(root
, path
);
1267 if (found_key
.objectid
!= key
.objectid
||
1268 found_key
.type
!= BTRFS_INODE_REF_KEY
)
1269 goto read_dir_items
;
1270 over
= filldir(dirent
, "..", 2,
1271 2, found_key
.offset
, DT_DIR
);
1278 btrfs_set_key_type(&key
, key_type
);
1279 key
.offset
= filp
->f_pos
;
1281 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1286 leaf
= path
->nodes
[0];
1287 nritems
= btrfs_header_nritems(leaf
);
1288 slot
= path
->slots
[0];
1289 if (advance
|| slot
>= nritems
) {
1290 if (slot
>= nritems
-1) {
1291 ret
= btrfs_next_leaf(root
, path
);
1294 leaf
= path
->nodes
[0];
1295 nritems
= btrfs_header_nritems(leaf
);
1296 slot
= path
->slots
[0];
1303 item
= btrfs_item_nr(leaf
, slot
);
1304 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1306 if (found_key
.objectid
!= key
.objectid
)
1308 if (btrfs_key_type(&found_key
) != key_type
)
1310 if (found_key
.offset
< filp
->f_pos
)
1313 filp
->f_pos
= found_key
.offset
;
1315 di
= btrfs_item_ptr(leaf
, slot
, struct btrfs_dir_item
);
1317 di_total
= btrfs_item_size(leaf
, item
);
1318 while(di_cur
< di_total
) {
1319 struct btrfs_key location
;
1321 name_len
= btrfs_dir_name_len(leaf
, di
);
1322 if (name_len
< 32) {
1323 name_ptr
= tmp_name
;
1325 name_ptr
= kmalloc(name_len
, GFP_NOFS
);
1328 read_extent_buffer(leaf
, name_ptr
,
1329 (unsigned long)(di
+ 1), name_len
);
1331 d_type
= btrfs_filetype_table
[btrfs_dir_type(leaf
, di
)];
1332 btrfs_dir_item_key_to_cpu(leaf
, di
, &location
);
1334 over
= filldir(dirent
, name_ptr
, name_len
,
1339 if (name_ptr
!= tmp_name
)
1344 di_len
= btrfs_dir_name_len(leaf
, di
) +
1345 btrfs_dir_data_len(leaf
, di
) +sizeof(*di
);
1347 di
= (struct btrfs_dir_item
*)((char *)di
+ di_len
);
1354 btrfs_release_path(root
, path
);
1355 btrfs_free_path(path
);
1356 mutex_unlock(&root
->fs_info
->fs_mutex
);
1360 int btrfs_write_inode(struct inode
*inode
, int wait
)
1362 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1363 struct btrfs_trans_handle
*trans
;
1367 mutex_lock(&root
->fs_info
->fs_mutex
);
1368 trans
= btrfs_start_transaction(root
, 1);
1369 btrfs_set_trans_block_group(trans
, inode
);
1370 ret
= btrfs_commit_transaction(trans
, root
);
1371 mutex_unlock(&root
->fs_info
->fs_mutex
);
1377 * This is somewhat expensive, updating the tree every time the
1378 * inode changes. But, it is most likely to find the inode in cache.
1379 * FIXME, needs more benchmarking...there are no reasons other than performance
1380 * to keep or drop this code.
1382 void btrfs_dirty_inode(struct inode
*inode
)
1384 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1385 struct btrfs_trans_handle
*trans
;
1387 mutex_lock(&root
->fs_info
->fs_mutex
);
1388 trans
= btrfs_start_transaction(root
, 1);
1389 btrfs_set_trans_block_group(trans
, inode
);
1390 btrfs_update_inode(trans
, root
, inode
);
1391 btrfs_end_transaction(trans
, root
);
1392 mutex_unlock(&root
->fs_info
->fs_mutex
);
1395 static struct inode
*btrfs_new_inode(struct btrfs_trans_handle
*trans
,
1396 struct btrfs_root
*root
,
1398 struct btrfs_block_group_cache
*group
,
1401 struct inode
*inode
;
1402 struct btrfs_inode_item
*inode_item
;
1403 struct btrfs_key
*location
;
1404 struct btrfs_path
*path
;
1408 path
= btrfs_alloc_path();
1411 inode
= new_inode(root
->fs_info
->sb
);
1413 return ERR_PTR(-ENOMEM
);
1415 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
,
1416 inode
->i_mapping
, GFP_NOFS
);
1417 BTRFS_I(inode
)->root
= root
;
1423 group
= btrfs_find_block_group(root
, group
, 0, 0, owner
);
1424 BTRFS_I(inode
)->block_group
= group
;
1426 ret
= btrfs_insert_empty_inode(trans
, root
, path
, objectid
);
1430 inode
->i_uid
= current
->fsuid
;
1431 inode
->i_gid
= current
->fsgid
;
1432 inode
->i_mode
= mode
;
1433 inode
->i_ino
= objectid
;
1434 inode
->i_blocks
= 0;
1435 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1436 inode_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
1437 struct btrfs_inode_item
);
1438 fill_inode_item(path
->nodes
[0], inode_item
, inode
);
1439 btrfs_mark_buffer_dirty(path
->nodes
[0]);
1440 btrfs_free_path(path
);
1442 location
= &BTRFS_I(inode
)->location
;
1443 location
->objectid
= objectid
;
1444 location
->offset
= 0;
1445 btrfs_set_key_type(location
, BTRFS_INODE_ITEM_KEY
);
1447 insert_inode_hash(inode
);
1450 btrfs_free_path(path
);
1451 return ERR_PTR(ret
);
1454 static inline u8
btrfs_inode_type(struct inode
*inode
)
1456 return btrfs_type_by_mode
[(inode
->i_mode
& S_IFMT
) >> S_SHIFT
];
1459 static int btrfs_add_link(struct btrfs_trans_handle
*trans
,
1460 struct dentry
*dentry
, struct inode
*inode
)
1463 struct btrfs_key key
;
1464 struct btrfs_root
*root
= BTRFS_I(dentry
->d_parent
->d_inode
)->root
;
1465 struct inode
*parent_inode
;
1467 key
.objectid
= inode
->i_ino
;
1468 btrfs_set_key_type(&key
, BTRFS_INODE_ITEM_KEY
);
1471 ret
= btrfs_insert_dir_item(trans
, root
,
1472 dentry
->d_name
.name
, dentry
->d_name
.len
,
1473 dentry
->d_parent
->d_inode
->i_ino
,
1474 &key
, btrfs_inode_type(inode
));
1476 ret
= btrfs_insert_inode_ref(trans
, root
,
1477 dentry
->d_name
.name
,
1480 dentry
->d_parent
->d_inode
->i_ino
);
1481 parent_inode
= dentry
->d_parent
->d_inode
;
1482 parent_inode
->i_size
+= dentry
->d_name
.len
* 2;
1483 parent_inode
->i_mtime
= parent_inode
->i_ctime
= CURRENT_TIME
;
1484 ret
= btrfs_update_inode(trans
, root
,
1485 dentry
->d_parent
->d_inode
);
1490 static int btrfs_add_nondir(struct btrfs_trans_handle
*trans
,
1491 struct dentry
*dentry
, struct inode
*inode
)
1493 int err
= btrfs_add_link(trans
, dentry
, inode
);
1495 d_instantiate(dentry
, inode
);
1503 static int btrfs_mknod(struct inode
*dir
, struct dentry
*dentry
,
1504 int mode
, dev_t rdev
)
1506 struct btrfs_trans_handle
*trans
;
1507 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1508 struct inode
*inode
= NULL
;
1512 unsigned long nr
= 0;
1514 if (!new_valid_dev(rdev
))
1517 mutex_lock(&root
->fs_info
->fs_mutex
);
1518 err
= btrfs_check_free_space(root
, 1, 0);
1522 trans
= btrfs_start_transaction(root
, 1);
1523 btrfs_set_trans_block_group(trans
, dir
);
1525 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
1531 inode
= btrfs_new_inode(trans
, root
, objectid
,
1532 BTRFS_I(dir
)->block_group
, mode
);
1533 err
= PTR_ERR(inode
);
1537 btrfs_set_trans_block_group(trans
, inode
);
1538 err
= btrfs_add_nondir(trans
, dentry
, inode
);
1542 inode
->i_op
= &btrfs_special_inode_operations
;
1543 init_special_inode(inode
, inode
->i_mode
, rdev
);
1544 btrfs_update_inode(trans
, root
, inode
);
1546 dir
->i_sb
->s_dirt
= 1;
1547 btrfs_update_inode_block_group(trans
, inode
);
1548 btrfs_update_inode_block_group(trans
, dir
);
1550 nr
= trans
->blocks_used
;
1551 btrfs_end_transaction(trans
, root
);
1553 mutex_unlock(&root
->fs_info
->fs_mutex
);
1556 inode_dec_link_count(inode
);
1559 btrfs_btree_balance_dirty(root
, nr
);
1563 static int btrfs_create(struct inode
*dir
, struct dentry
*dentry
,
1564 int mode
, struct nameidata
*nd
)
1566 struct btrfs_trans_handle
*trans
;
1567 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1568 struct inode
*inode
= NULL
;
1571 unsigned long nr
= 0;
1574 mutex_lock(&root
->fs_info
->fs_mutex
);
1575 err
= btrfs_check_free_space(root
, 1, 0);
1578 trans
= btrfs_start_transaction(root
, 1);
1579 btrfs_set_trans_block_group(trans
, dir
);
1581 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
1587 inode
= btrfs_new_inode(trans
, root
, objectid
,
1588 BTRFS_I(dir
)->block_group
, mode
);
1589 err
= PTR_ERR(inode
);
1593 btrfs_set_trans_block_group(trans
, inode
);
1594 err
= btrfs_add_nondir(trans
, dentry
, inode
);
1598 inode
->i_mapping
->a_ops
= &btrfs_aops
;
1599 inode
->i_fop
= &btrfs_file_operations
;
1600 inode
->i_op
= &btrfs_file_inode_operations
;
1601 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
,
1602 inode
->i_mapping
, GFP_NOFS
);
1603 BTRFS_I(inode
)->extent_tree
.ops
= &btrfs_extent_map_ops
;
1605 dir
->i_sb
->s_dirt
= 1;
1606 btrfs_update_inode_block_group(trans
, inode
);
1607 btrfs_update_inode_block_group(trans
, dir
);
1609 nr
= trans
->blocks_used
;
1610 btrfs_end_transaction(trans
, root
);
1612 mutex_unlock(&root
->fs_info
->fs_mutex
);
1615 inode_dec_link_count(inode
);
1618 btrfs_btree_balance_dirty(root
, nr
);
1622 static int btrfs_link(struct dentry
*old_dentry
, struct inode
*dir
,
1623 struct dentry
*dentry
)
1625 struct btrfs_trans_handle
*trans
;
1626 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1627 struct inode
*inode
= old_dentry
->d_inode
;
1628 unsigned long nr
= 0;
1632 if (inode
->i_nlink
== 0)
1635 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1640 mutex_lock(&root
->fs_info
->fs_mutex
);
1641 err
= btrfs_check_free_space(root
, 1, 0);
1644 trans
= btrfs_start_transaction(root
, 1);
1646 btrfs_set_trans_block_group(trans
, dir
);
1647 atomic_inc(&inode
->i_count
);
1648 err
= btrfs_add_nondir(trans
, dentry
, inode
);
1653 dir
->i_sb
->s_dirt
= 1;
1654 btrfs_update_inode_block_group(trans
, dir
);
1655 err
= btrfs_update_inode(trans
, root
, inode
);
1660 nr
= trans
->blocks_used
;
1661 btrfs_end_transaction(trans
, root
);
1663 mutex_unlock(&root
->fs_info
->fs_mutex
);
1666 inode_dec_link_count(inode
);
1669 btrfs_btree_balance_dirty(root
, nr
);
1673 static int btrfs_mkdir(struct inode
*dir
, struct dentry
*dentry
, int mode
)
1675 struct inode
*inode
;
1676 struct btrfs_trans_handle
*trans
;
1677 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1679 int drop_on_err
= 0;
1681 unsigned long nr
= 1;
1683 mutex_lock(&root
->fs_info
->fs_mutex
);
1684 err
= btrfs_check_free_space(root
, 1, 0);
1688 trans
= btrfs_start_transaction(root
, 1);
1689 btrfs_set_trans_block_group(trans
, dir
);
1691 if (IS_ERR(trans
)) {
1692 err
= PTR_ERR(trans
);
1696 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
1702 inode
= btrfs_new_inode(trans
, root
, objectid
,
1703 BTRFS_I(dir
)->block_group
, S_IFDIR
| mode
);
1704 if (IS_ERR(inode
)) {
1705 err
= PTR_ERR(inode
);
1710 inode
->i_op
= &btrfs_dir_inode_operations
;
1711 inode
->i_fop
= &btrfs_dir_file_operations
;
1712 btrfs_set_trans_block_group(trans
, inode
);
1715 err
= btrfs_update_inode(trans
, root
, inode
);
1719 err
= btrfs_add_link(trans
, dentry
, inode
);
1723 d_instantiate(dentry
, inode
);
1725 dir
->i_sb
->s_dirt
= 1;
1726 btrfs_update_inode_block_group(trans
, inode
);
1727 btrfs_update_inode_block_group(trans
, dir
);
1730 nr
= trans
->blocks_used
;
1731 btrfs_end_transaction(trans
, root
);
1734 mutex_unlock(&root
->fs_info
->fs_mutex
);
1737 btrfs_btree_balance_dirty(root
, nr
);
1741 struct extent_map
*btrfs_get_extent(struct inode
*inode
, struct page
*page
,
1742 size_t page_offset
, u64 start
, u64 end
,
1748 u64 extent_start
= 0;
1750 u64 objectid
= inode
->i_ino
;
1752 int failed_insert
= 0;
1753 struct btrfs_path
*path
;
1754 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1755 struct btrfs_file_extent_item
*item
;
1756 struct extent_buffer
*leaf
;
1757 struct btrfs_key found_key
;
1758 struct extent_map
*em
= NULL
;
1759 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
1760 struct btrfs_trans_handle
*trans
= NULL
;
1762 path
= btrfs_alloc_path();
1764 mutex_lock(&root
->fs_info
->fs_mutex
);
1767 em
= lookup_extent_mapping(em_tree
, start
, end
);
1772 em
= alloc_extent_map(GFP_NOFS
);
1777 em
->start
= EXTENT_MAP_HOLE
;
1778 em
->end
= EXTENT_MAP_HOLE
;
1780 em
->bdev
= inode
->i_sb
->s_bdev
;
1781 ret
= btrfs_lookup_file_extent(trans
, root
, path
,
1782 objectid
, start
, trans
!= NULL
);
1789 if (path
->slots
[0] == 0)
1794 leaf
= path
->nodes
[0];
1795 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
1796 struct btrfs_file_extent_item
);
1797 /* are we inside the extent that was found? */
1798 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
1799 found_type
= btrfs_key_type(&found_key
);
1800 if (found_key
.objectid
!= objectid
||
1801 found_type
!= BTRFS_EXTENT_DATA_KEY
) {
1805 found_type
= btrfs_file_extent_type(leaf
, item
);
1806 extent_start
= found_key
.offset
;
1807 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
1808 extent_end
= extent_start
+
1809 btrfs_file_extent_num_bytes(leaf
, item
);
1811 if (start
< extent_start
|| start
>= extent_end
) {
1813 if (start
< extent_start
) {
1814 if (end
< extent_start
)
1816 em
->end
= extent_end
- 1;
1822 bytenr
= btrfs_file_extent_disk_bytenr(leaf
, item
);
1824 em
->start
= extent_start
;
1825 em
->end
= extent_end
- 1;
1826 em
->block_start
= EXTENT_MAP_HOLE
;
1827 em
->block_end
= EXTENT_MAP_HOLE
;
1830 bytenr
+= btrfs_file_extent_offset(leaf
, item
);
1831 em
->block_start
= bytenr
;
1832 em
->block_end
= em
->block_start
+
1833 btrfs_file_extent_num_bytes(leaf
, item
) - 1;
1834 em
->start
= extent_start
;
1835 em
->end
= extent_end
- 1;
1837 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
1841 size_t extent_offset
;
1844 size
= btrfs_file_extent_inline_len(leaf
, btrfs_item_nr(leaf
,
1846 extent_end
= (extent_start
+ size
- 1) |
1847 ((u64
)root
->sectorsize
- 1);
1848 if (start
< extent_start
|| start
>= extent_end
) {
1850 if (start
< extent_start
) {
1851 if (end
< extent_start
)
1853 em
->end
= extent_end
;
1859 em
->block_start
= EXTENT_MAP_INLINE
;
1860 em
->block_end
= EXTENT_MAP_INLINE
;
1863 em
->start
= extent_start
;
1864 em
->end
= extent_start
+ size
- 1;
1868 extent_offset
= ((u64
)page
->index
<< PAGE_CACHE_SHIFT
) -
1869 extent_start
+ page_offset
;
1870 copy_size
= min_t(u64
, PAGE_CACHE_SIZE
- page_offset
,
1871 size
- extent_offset
);
1872 em
->start
= extent_start
+ extent_offset
;
1873 em
->end
= (em
->start
+ copy_size
-1) |
1874 ((u64
)root
->sectorsize
-1);
1876 ptr
= btrfs_file_extent_inline_start(item
) + extent_offset
;
1877 if (create
== 0 && !PageUptodate(page
)) {
1878 read_extent_buffer(leaf
, map
+ page_offset
, ptr
,
1880 flush_dcache_page(page
);
1881 } else if (create
&& PageUptodate(page
)) {
1884 free_extent_map(em
);
1886 btrfs_release_path(root
, path
);
1887 trans
= btrfs_start_transaction(root
, 1);
1890 write_extent_buffer(leaf
, map
+ page_offset
, ptr
,
1892 btrfs_mark_buffer_dirty(leaf
);
1895 set_extent_uptodate(em_tree
, em
->start
, em
->end
, GFP_NOFS
);
1898 printk("unkknown found_type %d\n", found_type
);
1905 em
->block_start
= EXTENT_MAP_HOLE
;
1906 em
->block_end
= EXTENT_MAP_HOLE
;
1908 btrfs_release_path(root
, path
);
1909 if (em
->start
> start
|| em
->end
< start
) {
1910 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em
->start
, em
->end
, start
, end
);
1914 ret
= add_extent_mapping(em_tree
, em
);
1915 if (ret
== -EEXIST
) {
1916 free_extent_map(em
);
1918 if (0 && failed_insert
== 1) {
1919 btrfs_drop_extent_cache(inode
, start
, end
);
1922 if (failed_insert
> 5) {
1923 printk("failing to insert %Lu %Lu\n", start
, end
);
1931 btrfs_free_path(path
);
1933 ret
= btrfs_end_transaction(trans
, root
);
1937 mutex_unlock(&root
->fs_info
->fs_mutex
);
1939 free_extent_map(em
);
1941 return ERR_PTR(err
);
1946 static sector_t
btrfs_bmap(struct address_space
*mapping
, sector_t iblock
)
1948 return extent_bmap(mapping
, iblock
, btrfs_get_extent
);
1951 static int btrfs_prepare_write(struct file
*file
, struct page
*page
,
1952 unsigned from
, unsigned to
)
1954 struct btrfs_root
*root
= BTRFS_I(page
->mapping
->host
)->root
;
1957 mutex_lock(&root
->fs_info
->fs_mutex
);
1958 err
= btrfs_check_free_space(root
, PAGE_CACHE_SIZE
, 0);
1959 mutex_unlock(&root
->fs_info
->fs_mutex
);
1963 return extent_prepare_write(&BTRFS_I(page
->mapping
->host
)->extent_tree
,
1964 page
->mapping
->host
, page
, from
, to
,
1968 int btrfs_readpage(struct file
*file
, struct page
*page
)
1970 struct extent_map_tree
*tree
;
1971 tree
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
1972 return extent_read_full_page(tree
, page
, btrfs_get_extent
);
1975 static int btrfs_writepage(struct page
*page
, struct writeback_control
*wbc
)
1977 struct extent_map_tree
*tree
;
1980 if (current
->flags
& PF_MEMALLOC
) {
1981 redirty_page_for_writepage(wbc
, page
);
1985 tree
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
1986 return extent_write_full_page(tree
, page
, btrfs_get_extent
, wbc
);
1989 static int btrfs_writepages(struct address_space
*mapping
,
1990 struct writeback_control
*wbc
)
1992 struct extent_map_tree
*tree
;
1993 tree
= &BTRFS_I(mapping
->host
)->extent_tree
;
1994 return extent_writepages(tree
, mapping
, btrfs_get_extent
, wbc
);
1998 btrfs_readpages(struct file
*file
, struct address_space
*mapping
,
1999 struct list_head
*pages
, unsigned nr_pages
)
2001 struct extent_map_tree
*tree
;
2002 tree
= &BTRFS_I(mapping
->host
)->extent_tree
;
2003 return extent_readpages(tree
, mapping
, pages
, nr_pages
,
2007 static int btrfs_releasepage(struct page
*page
, gfp_t unused_gfp_flags
)
2009 struct extent_map_tree
*tree
;
2012 tree
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
2013 ret
= try_release_extent_mapping(tree
, page
);
2015 ClearPagePrivate(page
);
2016 set_page_private(page
, 0);
2017 page_cache_release(page
);
2022 static void btrfs_invalidatepage(struct page
*page
, unsigned long offset
)
2024 struct extent_map_tree
*tree
;
2026 tree
= &BTRFS_I(page
->mapping
->host
)->extent_tree
;
2027 extent_invalidatepage(tree
, page
, offset
);
2028 btrfs_releasepage(page
, GFP_NOFS
);
2032 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
2033 * called from a page fault handler when a page is first dirtied. Hence we must
2034 * be careful to check for EOF conditions here. We set the page up correctly
2035 * for a written page which means we get ENOSPC checking when writing into
2036 * holes and correct delalloc and unwritten extent mapping on filesystems that
2037 * support these features.
2039 * We are not allowed to take the i_mutex here so we have to play games to
2040 * protect against truncate races as the page could now be beyond EOF. Because
2041 * vmtruncate() writes the inode size before removing pages, once we have the
2042 * page lock we can determine safely if the page is beyond EOF. If it is not
2043 * beyond EOF, then the page is guaranteed safe against truncation until we
2046 int btrfs_page_mkwrite(struct vm_area_struct
*vma
, struct page
*page
)
2048 struct inode
*inode
= fdentry(vma
->vm_file
)->d_inode
;
2049 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2055 mutex_lock(&root
->fs_info
->fs_mutex
);
2056 ret
= btrfs_check_free_space(root
, PAGE_CACHE_SIZE
, 0);
2057 mutex_lock(&root
->fs_info
->fs_mutex
);
2063 down_read(&BTRFS_I(inode
)->root
->snap_sem
);
2065 wait_on_page_writeback(page
);
2066 size
= i_size_read(inode
);
2067 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
2069 if ((page
->mapping
!= inode
->i_mapping
) ||
2070 (page_start
> size
)) {
2071 /* page got truncated out from underneath us */
2075 /* page is wholly or partially inside EOF */
2076 if (page_start
+ PAGE_CACHE_SIZE
> size
)
2077 end
= size
& ~PAGE_CACHE_MASK
;
2079 end
= PAGE_CACHE_SIZE
;
2081 ret
= btrfs_cow_one_page(inode
, page
, end
);
2084 up_read(&BTRFS_I(inode
)->root
->snap_sem
);
2090 static void btrfs_truncate(struct inode
*inode
)
2092 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2094 struct btrfs_trans_handle
*trans
;
2097 if (!S_ISREG(inode
->i_mode
))
2099 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
2102 btrfs_truncate_page(inode
->i_mapping
, inode
->i_size
);
2104 mutex_lock(&root
->fs_info
->fs_mutex
);
2105 trans
= btrfs_start_transaction(root
, 1);
2106 btrfs_set_trans_block_group(trans
, inode
);
2108 /* FIXME, add redo link to tree so we don't leak on crash */
2109 ret
= btrfs_truncate_in_trans(trans
, root
, inode
);
2110 btrfs_update_inode(trans
, root
, inode
);
2111 nr
= trans
->blocks_used
;
2113 ret
= btrfs_end_transaction(trans
, root
);
2115 mutex_unlock(&root
->fs_info
->fs_mutex
);
2116 btrfs_btree_balance_dirty(root
, nr
);
2119 int btrfs_commit_write(struct file
*file
, struct page
*page
,
2120 unsigned from
, unsigned to
)
2122 loff_t pos
= ((loff_t
)page
->index
<< PAGE_CACHE_SHIFT
) + to
;
2123 struct inode
*inode
= page
->mapping
->host
;
2125 btrfs_cow_one_page(inode
, page
, PAGE_CACHE_SIZE
);
2127 if (pos
> inode
->i_size
) {
2128 i_size_write(inode
, pos
);
2129 mark_inode_dirty(inode
);
2134 static int create_subvol(struct btrfs_root
*root
, char *name
, int namelen
)
2136 struct btrfs_trans_handle
*trans
;
2137 struct btrfs_key key
;
2138 struct btrfs_root_item root_item
;
2139 struct btrfs_inode_item
*inode_item
;
2140 struct extent_buffer
*leaf
;
2141 struct btrfs_root
*new_root
;
2142 struct inode
*inode
;
2147 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
2148 unsigned long nr
= 1;
2150 mutex_lock(&root
->fs_info
->fs_mutex
);
2151 ret
= btrfs_check_free_space(root
, 1, 0);
2155 trans
= btrfs_start_transaction(root
, 1);
2158 ret
= btrfs_find_free_objectid(trans
, root
->fs_info
->tree_root
,
2163 leaf
= __btrfs_alloc_free_block(trans
, root
, root
->leafsize
,
2164 objectid
, trans
->transid
, 0, 0,
2167 return PTR_ERR(leaf
);
2169 btrfs_set_header_nritems(leaf
, 0);
2170 btrfs_set_header_level(leaf
, 0);
2171 btrfs_set_header_bytenr(leaf
, leaf
->start
);
2172 btrfs_set_header_generation(leaf
, trans
->transid
);
2173 btrfs_set_header_owner(leaf
, objectid
);
2175 write_extent_buffer(leaf
, root
->fs_info
->fsid
,
2176 (unsigned long)btrfs_header_fsid(leaf
),
2178 btrfs_mark_buffer_dirty(leaf
);
2180 inode_item
= &root_item
.inode
;
2181 memset(inode_item
, 0, sizeof(*inode_item
));
2182 inode_item
->generation
= cpu_to_le64(1);
2183 inode_item
->size
= cpu_to_le64(3);
2184 inode_item
->nlink
= cpu_to_le32(1);
2185 inode_item
->nblocks
= cpu_to_le64(1);
2186 inode_item
->mode
= cpu_to_le32(S_IFDIR
| 0755);
2188 btrfs_set_root_bytenr(&root_item
, leaf
->start
);
2189 btrfs_set_root_level(&root_item
, 0);
2190 btrfs_set_root_refs(&root_item
, 1);
2191 btrfs_set_root_used(&root_item
, 0);
2193 memset(&root_item
.drop_progress
, 0, sizeof(root_item
.drop_progress
));
2194 root_item
.drop_level
= 0;
2196 free_extent_buffer(leaf
);
2199 btrfs_set_root_dirid(&root_item
, new_dirid
);
2201 key
.objectid
= objectid
;
2203 btrfs_set_key_type(&key
, BTRFS_ROOT_ITEM_KEY
);
2204 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
2210 * insert the directory item
2212 key
.offset
= (u64
)-1;
2213 dir
= root
->fs_info
->sb
->s_root
->d_inode
;
2214 ret
= btrfs_insert_dir_item(trans
, root
->fs_info
->tree_root
,
2215 name
, namelen
, dir
->i_ino
, &key
,
2220 ret
= btrfs_insert_inode_ref(trans
, root
->fs_info
->tree_root
,
2221 name
, namelen
, objectid
,
2222 root
->fs_info
->sb
->s_root
->d_inode
->i_ino
);
2226 ret
= btrfs_commit_transaction(trans
, root
);
2230 new_root
= btrfs_read_fs_root(root
->fs_info
, &key
, name
, namelen
);
2233 trans
= btrfs_start_transaction(new_root
, 1);
2236 inode
= btrfs_new_inode(trans
, new_root
, new_dirid
,
2237 BTRFS_I(dir
)->block_group
, S_IFDIR
| 0700);
2240 inode
->i_op
= &btrfs_dir_inode_operations
;
2241 inode
->i_fop
= &btrfs_dir_file_operations
;
2242 new_root
->inode
= inode
;
2244 ret
= btrfs_insert_inode_ref(trans
, new_root
, "..", 2, new_dirid
,
2248 ret
= btrfs_update_inode(trans
, new_root
, inode
);
2252 nr
= trans
->blocks_used
;
2253 err
= btrfs_commit_transaction(trans
, root
);
2257 mutex_unlock(&root
->fs_info
->fs_mutex
);
2258 btrfs_btree_balance_dirty(root
, nr
);
2262 static int create_snapshot(struct btrfs_root
*root
, char *name
, int namelen
)
2264 struct btrfs_trans_handle
*trans
;
2265 struct btrfs_key key
;
2266 struct btrfs_root_item new_root_item
;
2267 struct extent_buffer
*tmp
;
2271 unsigned long nr
= 0;
2273 if (!root
->ref_cows
)
2276 down_write(&root
->snap_sem
);
2277 freeze_bdev(root
->fs_info
->sb
->s_bdev
);
2278 thaw_bdev(root
->fs_info
->sb
->s_bdev
, root
->fs_info
->sb
);
2280 mutex_lock(&root
->fs_info
->fs_mutex
);
2281 ret
= btrfs_check_free_space(root
, 1, 0);
2285 trans
= btrfs_start_transaction(root
, 1);
2288 ret
= btrfs_update_inode(trans
, root
, root
->inode
);
2292 ret
= btrfs_find_free_objectid(trans
, root
->fs_info
->tree_root
,
2297 memcpy(&new_root_item
, &root
->root_item
,
2298 sizeof(new_root_item
));
2300 key
.objectid
= objectid
;
2302 btrfs_set_key_type(&key
, BTRFS_ROOT_ITEM_KEY
);
2304 extent_buffer_get(root
->node
);
2305 btrfs_cow_block(trans
, root
, root
->node
, NULL
, 0, &tmp
);
2306 free_extent_buffer(tmp
);
2308 btrfs_copy_root(trans
, root
, root
->node
, &tmp
, objectid
);
2310 btrfs_set_root_bytenr(&new_root_item
, tmp
->start
);
2311 btrfs_set_root_level(&new_root_item
, btrfs_header_level(tmp
));
2312 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
2314 free_extent_buffer(tmp
);
2319 * insert the directory item
2321 key
.offset
= (u64
)-1;
2322 ret
= btrfs_insert_dir_item(trans
, root
->fs_info
->tree_root
,
2324 root
->fs_info
->sb
->s_root
->d_inode
->i_ino
,
2325 &key
, BTRFS_FT_DIR
);
2330 ret
= btrfs_insert_inode_ref(trans
, root
->fs_info
->tree_root
,
2331 name
, namelen
, objectid
,
2332 root
->fs_info
->sb
->s_root
->d_inode
->i_ino
);
2337 nr
= trans
->blocks_used
;
2338 err
= btrfs_commit_transaction(trans
, root
);
2343 mutex_unlock(&root
->fs_info
->fs_mutex
);
2344 up_write(&root
->snap_sem
);
2345 btrfs_btree_balance_dirty(root
, nr
);
2349 unsigned long btrfs_force_ra(struct address_space
*mapping
,
2350 struct file_ra_state
*ra
, struct file
*file
,
2351 pgoff_t offset
, pgoff_t last_index
)
2355 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2356 req_size
= last_index
- offset
+ 1;
2357 offset
= page_cache_readahead(mapping
, ra
, file
, offset
, req_size
);
2360 req_size
= min(last_index
- offset
+ 1, (pgoff_t
)128);
2361 page_cache_sync_readahead(mapping
, ra
, file
, offset
, req_size
);
2362 return offset
+ req_size
;
2366 int btrfs_defrag_file(struct file
*file
) {
2367 struct inode
*inode
= fdentry(file
)->d_inode
;
2368 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2369 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
2371 unsigned long last_index
;
2372 unsigned long ra_index
= 0;
2376 u64 existing_delalloc
;
2380 mutex_lock(&root
->fs_info
->fs_mutex
);
2381 ret
= btrfs_check_free_space(root
, inode
->i_size
, 0);
2382 mutex_unlock(&root
->fs_info
->fs_mutex
);
2386 mutex_lock(&inode
->i_mutex
);
2387 last_index
= inode
->i_size
>> PAGE_CACHE_SHIFT
;
2388 for (i
= 0; i
<= last_index
; i
++) {
2389 if (i
== ra_index
) {
2390 ra_index
= btrfs_force_ra(inode
->i_mapping
,
2392 file
, ra_index
, last_index
);
2394 page
= grab_cache_page(inode
->i_mapping
, i
);
2397 if (!PageUptodate(page
)) {
2398 btrfs_readpage(NULL
, page
);
2400 if (!PageUptodate(page
)) {
2402 page_cache_release(page
);
2406 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
2407 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
2409 lock_extent(em_tree
, page_start
, page_end
, GFP_NOFS
);
2410 delalloc_start
= page_start
;
2412 count_range_bits(&BTRFS_I(inode
)->extent_tree
,
2413 &delalloc_start
, page_end
,
2414 PAGE_CACHE_SIZE
, EXTENT_DELALLOC
);
2415 set_extent_delalloc(em_tree
, page_start
,
2416 page_end
, GFP_NOFS
);
2418 spin_lock(&root
->fs_info
->delalloc_lock
);
2419 root
->fs_info
->delalloc_bytes
+= PAGE_CACHE_SIZE
-
2421 spin_unlock(&root
->fs_info
->delalloc_lock
);
2423 unlock_extent(em_tree
, page_start
, page_end
, GFP_NOFS
);
2424 set_page_dirty(page
);
2426 page_cache_release(page
);
2427 balance_dirty_pages_ratelimited_nr(inode
->i_mapping
, 1);
2431 mutex_unlock(&inode
->i_mutex
);
2435 static int btrfs_ioctl_resize(struct btrfs_root
*root
, void __user
*arg
)
2439 struct btrfs_ioctl_vol_args
*vol_args
;
2440 struct btrfs_trans_handle
*trans
;
2446 vol_args
= kmalloc(sizeof(*vol_args
), GFP_NOFS
);
2451 if (copy_from_user(vol_args
, arg
, sizeof(*vol_args
))) {
2455 namelen
= strlen(vol_args
->name
);
2456 if (namelen
> BTRFS_VOL_NAME_MAX
) {
2461 sizestr
= vol_args
->name
;
2462 if (!strcmp(sizestr
, "max"))
2463 new_size
= root
->fs_info
->sb
->s_bdev
->bd_inode
->i_size
;
2465 if (sizestr
[0] == '-') {
2468 } else if (sizestr
[0] == '+') {
2472 new_size
= btrfs_parse_size(sizestr
);
2473 if (new_size
== 0) {
2479 mutex_lock(&root
->fs_info
->fs_mutex
);
2480 old_size
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
2483 if (new_size
> old_size
) {
2487 new_size
= old_size
- new_size
;
2488 } else if (mod
> 0) {
2489 new_size
= old_size
+ new_size
;
2492 if (new_size
< 256 * 1024 * 1024) {
2496 if (new_size
> root
->fs_info
->sb
->s_bdev
->bd_inode
->i_size
) {
2500 new_size
= (new_size
/ root
->sectorsize
) * root
->sectorsize
;
2502 printk("new size is %Lu\n", new_size
);
2503 if (new_size
> old_size
) {
2504 trans
= btrfs_start_transaction(root
, 1);
2505 ret
= btrfs_grow_extent_tree(trans
, root
, new_size
);
2506 btrfs_commit_transaction(trans
, root
);
2508 ret
= btrfs_shrink_extent_tree(root
, new_size
);
2512 mutex_unlock(&root
->fs_info
->fs_mutex
);
2518 static int btrfs_ioctl_snap_create(struct btrfs_root
*root
, void __user
*arg
)
2520 struct btrfs_ioctl_vol_args
*vol_args
;
2521 struct btrfs_dir_item
*di
;
2522 struct btrfs_path
*path
;
2527 vol_args
= kmalloc(sizeof(*vol_args
), GFP_NOFS
);
2532 if (copy_from_user(vol_args
, arg
, sizeof(*vol_args
))) {
2537 namelen
= strlen(vol_args
->name
);
2538 if (namelen
> BTRFS_VOL_NAME_MAX
) {
2542 if (strchr(vol_args
->name
, '/')) {
2547 path
= btrfs_alloc_path();
2553 root_dirid
= root
->fs_info
->sb
->s_root
->d_inode
->i_ino
,
2554 mutex_lock(&root
->fs_info
->fs_mutex
);
2555 di
= btrfs_lookup_dir_item(NULL
, root
->fs_info
->tree_root
,
2557 vol_args
->name
, namelen
, 0);
2558 mutex_unlock(&root
->fs_info
->fs_mutex
);
2559 btrfs_free_path(path
);
2561 if (di
&& !IS_ERR(di
)) {
2571 if (root
== root
->fs_info
->tree_root
)
2572 ret
= create_subvol(root
, vol_args
->name
, namelen
);
2574 ret
= create_snapshot(root
, vol_args
->name
, namelen
);
2580 static int btrfs_ioctl_defrag(struct file
*file
)
2582 struct inode
*inode
= fdentry(file
)->d_inode
;
2583 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2585 switch (inode
->i_mode
& S_IFMT
) {
2587 mutex_lock(&root
->fs_info
->fs_mutex
);
2588 btrfs_defrag_root(root
, 0);
2589 btrfs_defrag_root(root
->fs_info
->extent_root
, 0);
2590 mutex_unlock(&root
->fs_info
->fs_mutex
);
2593 btrfs_defrag_file(file
);
2600 long btrfs_ioctl(struct file
*file
, unsigned int
2601 cmd
, unsigned long arg
)
2603 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
2606 case BTRFS_IOC_SNAP_CREATE
:
2607 return btrfs_ioctl_snap_create(root
, (void __user
*)arg
);
2608 case BTRFS_IOC_DEFRAG
:
2609 return btrfs_ioctl_defrag(file
);
2610 case BTRFS_IOC_RESIZE
:
2611 return btrfs_ioctl_resize(root
, (void __user
*)arg
);
2618 * Called inside transaction, so use GFP_NOFS
2620 struct inode
*btrfs_alloc_inode(struct super_block
*sb
)
2622 struct btrfs_inode
*ei
;
2624 ei
= kmem_cache_alloc(btrfs_inode_cachep
, GFP_NOFS
);
2628 return &ei
->vfs_inode
;
2631 void btrfs_destroy_inode(struct inode
*inode
)
2633 WARN_ON(!list_empty(&inode
->i_dentry
));
2634 WARN_ON(inode
->i_data
.nrpages
);
2636 kmem_cache_free(btrfs_inode_cachep
, BTRFS_I(inode
));
2639 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2640 static void init_once(struct kmem_cache
* cachep
, void *foo
)
2642 static void init_once(void * foo
, struct kmem_cache
* cachep
,
2643 unsigned long flags
)
2646 struct btrfs_inode
*ei
= (struct btrfs_inode
*) foo
;
2648 inode_init_once(&ei
->vfs_inode
);
2651 void btrfs_destroy_cachep(void)
2653 if (btrfs_inode_cachep
)
2654 kmem_cache_destroy(btrfs_inode_cachep
);
2655 if (btrfs_trans_handle_cachep
)
2656 kmem_cache_destroy(btrfs_trans_handle_cachep
);
2657 if (btrfs_transaction_cachep
)
2658 kmem_cache_destroy(btrfs_transaction_cachep
);
2659 if (btrfs_bit_radix_cachep
)
2660 kmem_cache_destroy(btrfs_bit_radix_cachep
);
2661 if (btrfs_path_cachep
)
2662 kmem_cache_destroy(btrfs_path_cachep
);
2665 struct kmem_cache
*btrfs_cache_create(const char *name
, size_t size
,
2666 unsigned long extra_flags
,
2667 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2668 void (*ctor
)(struct kmem_cache
*, void *)
2670 void (*ctor
)(void *, struct kmem_cache
*,
2675 return kmem_cache_create(name
, size
, 0, (SLAB_RECLAIM_ACCOUNT
|
2676 SLAB_MEM_SPREAD
| extra_flags
), ctor
2677 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2683 int btrfs_init_cachep(void)
2685 btrfs_inode_cachep
= btrfs_cache_create("btrfs_inode_cache",
2686 sizeof(struct btrfs_inode
),
2688 if (!btrfs_inode_cachep
)
2690 btrfs_trans_handle_cachep
=
2691 btrfs_cache_create("btrfs_trans_handle_cache",
2692 sizeof(struct btrfs_trans_handle
),
2694 if (!btrfs_trans_handle_cachep
)
2696 btrfs_transaction_cachep
= btrfs_cache_create("btrfs_transaction_cache",
2697 sizeof(struct btrfs_transaction
),
2699 if (!btrfs_transaction_cachep
)
2701 btrfs_path_cachep
= btrfs_cache_create("btrfs_path_cache",
2702 sizeof(struct btrfs_path
),
2704 if (!btrfs_path_cachep
)
2706 btrfs_bit_radix_cachep
= btrfs_cache_create("btrfs_radix", 256,
2707 SLAB_DESTROY_BY_RCU
, NULL
);
2708 if (!btrfs_bit_radix_cachep
)
2712 btrfs_destroy_cachep();
2716 static int btrfs_getattr(struct vfsmount
*mnt
,
2717 struct dentry
*dentry
, struct kstat
*stat
)
2719 struct inode
*inode
= dentry
->d_inode
;
2720 generic_fillattr(inode
, stat
);
2721 stat
->blksize
= 256 * 1024;
2725 static int btrfs_rename(struct inode
* old_dir
, struct dentry
*old_dentry
,
2726 struct inode
* new_dir
,struct dentry
*new_dentry
)
2728 struct btrfs_trans_handle
*trans
;
2729 struct btrfs_root
*root
= BTRFS_I(old_dir
)->root
;
2730 struct inode
*new_inode
= new_dentry
->d_inode
;
2731 struct inode
*old_inode
= old_dentry
->d_inode
;
2732 struct timespec ctime
= CURRENT_TIME
;
2733 struct btrfs_path
*path
;
2736 if (S_ISDIR(old_inode
->i_mode
) && new_inode
&&
2737 new_inode
->i_size
> BTRFS_EMPTY_DIR_SIZE
) {
2741 mutex_lock(&root
->fs_info
->fs_mutex
);
2742 ret
= btrfs_check_free_space(root
, 1, 0);
2746 trans
= btrfs_start_transaction(root
, 1);
2748 btrfs_set_trans_block_group(trans
, new_dir
);
2749 path
= btrfs_alloc_path();
2755 old_dentry
->d_inode
->i_nlink
++;
2756 old_dir
->i_ctime
= old_dir
->i_mtime
= ctime
;
2757 new_dir
->i_ctime
= new_dir
->i_mtime
= ctime
;
2758 old_inode
->i_ctime
= ctime
;
2760 ret
= btrfs_unlink_trans(trans
, root
, old_dir
, old_dentry
);
2765 new_inode
->i_ctime
= CURRENT_TIME
;
2766 ret
= btrfs_unlink_trans(trans
, root
, new_dir
, new_dentry
);
2770 ret
= btrfs_add_link(trans
, new_dentry
, old_inode
);
2775 btrfs_free_path(path
);
2776 btrfs_end_transaction(trans
, root
);
2778 mutex_unlock(&root
->fs_info
->fs_mutex
);
2782 static int btrfs_symlink(struct inode
*dir
, struct dentry
*dentry
,
2783 const char *symname
)
2785 struct btrfs_trans_handle
*trans
;
2786 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2787 struct btrfs_path
*path
;
2788 struct btrfs_key key
;
2789 struct inode
*inode
= NULL
;
2796 struct btrfs_file_extent_item
*ei
;
2797 struct extent_buffer
*leaf
;
2798 unsigned long nr
= 0;
2800 name_len
= strlen(symname
) + 1;
2801 if (name_len
> BTRFS_MAX_INLINE_DATA_SIZE(root
))
2802 return -ENAMETOOLONG
;
2804 mutex_lock(&root
->fs_info
->fs_mutex
);
2805 err
= btrfs_check_free_space(root
, 1, 0);
2809 trans
= btrfs_start_transaction(root
, 1);
2810 btrfs_set_trans_block_group(trans
, dir
);
2812 err
= btrfs_find_free_objectid(trans
, root
, dir
->i_ino
, &objectid
);
2818 inode
= btrfs_new_inode(trans
, root
, objectid
,
2819 BTRFS_I(dir
)->block_group
, S_IFLNK
|S_IRWXUGO
);
2820 err
= PTR_ERR(inode
);
2824 btrfs_set_trans_block_group(trans
, inode
);
2825 err
= btrfs_add_nondir(trans
, dentry
, inode
);
2829 inode
->i_mapping
->a_ops
= &btrfs_aops
;
2830 inode
->i_fop
= &btrfs_file_operations
;
2831 inode
->i_op
= &btrfs_file_inode_operations
;
2832 extent_map_tree_init(&BTRFS_I(inode
)->extent_tree
,
2833 inode
->i_mapping
, GFP_NOFS
);
2834 BTRFS_I(inode
)->extent_tree
.ops
= &btrfs_extent_map_ops
;
2836 dir
->i_sb
->s_dirt
= 1;
2837 btrfs_update_inode_block_group(trans
, inode
);
2838 btrfs_update_inode_block_group(trans
, dir
);
2842 path
= btrfs_alloc_path();
2844 key
.objectid
= inode
->i_ino
;
2846 btrfs_set_key_type(&key
, BTRFS_EXTENT_DATA_KEY
);
2847 datasize
= btrfs_file_extent_calc_inline_size(name_len
);
2848 err
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
2854 leaf
= path
->nodes
[0];
2855 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
2856 struct btrfs_file_extent_item
);
2857 btrfs_set_file_extent_generation(leaf
, ei
, trans
->transid
);
2858 btrfs_set_file_extent_type(leaf
, ei
,
2859 BTRFS_FILE_EXTENT_INLINE
);
2860 ptr
= btrfs_file_extent_inline_start(ei
);
2861 write_extent_buffer(leaf
, symname
, ptr
, name_len
);
2862 btrfs_mark_buffer_dirty(leaf
);
2863 btrfs_free_path(path
);
2865 inode
->i_op
= &btrfs_symlink_inode_operations
;
2866 inode
->i_mapping
->a_ops
= &btrfs_symlink_aops
;
2867 inode
->i_size
= name_len
- 1;
2868 err
= btrfs_update_inode(trans
, root
, inode
);
2873 nr
= trans
->blocks_used
;
2874 btrfs_end_transaction(trans
, root
);
2876 mutex_unlock(&root
->fs_info
->fs_mutex
);
2878 inode_dec_link_count(inode
);
2881 btrfs_btree_balance_dirty(root
, nr
);
2885 static struct inode_operations btrfs_dir_inode_operations
= {
2886 .lookup
= btrfs_lookup
,
2887 .create
= btrfs_create
,
2888 .unlink
= btrfs_unlink
,
2890 .mkdir
= btrfs_mkdir
,
2891 .rmdir
= btrfs_rmdir
,
2892 .rename
= btrfs_rename
,
2893 .symlink
= btrfs_symlink
,
2894 .setattr
= btrfs_setattr
,
2895 .mknod
= btrfs_mknod
,
2896 .setxattr
= generic_setxattr
,
2897 .getxattr
= generic_getxattr
,
2898 .listxattr
= btrfs_listxattr
,
2899 .removexattr
= generic_removexattr
,
2902 static struct inode_operations btrfs_dir_ro_inode_operations
= {
2903 .lookup
= btrfs_lookup
,
2906 static struct file_operations btrfs_dir_file_operations
= {
2907 .llseek
= generic_file_llseek
,
2908 .read
= generic_read_dir
,
2909 .readdir
= btrfs_readdir
,
2910 .unlocked_ioctl
= btrfs_ioctl
,
2911 #ifdef CONFIG_COMPAT
2912 .compat_ioctl
= btrfs_ioctl
,
2916 static struct extent_map_ops btrfs_extent_map_ops
= {
2917 .fill_delalloc
= run_delalloc_range
,
2918 .writepage_io_hook
= btrfs_writepage_io_hook
,
2919 .readpage_io_hook
= btrfs_readpage_io_hook
,
2920 .readpage_end_io_hook
= btrfs_readpage_end_io_hook
,
2923 static struct address_space_operations btrfs_aops
= {
2924 .readpage
= btrfs_readpage
,
2925 .writepage
= btrfs_writepage
,
2926 .writepages
= btrfs_writepages
,
2927 .readpages
= btrfs_readpages
,
2928 .sync_page
= block_sync_page
,
2929 .prepare_write
= btrfs_prepare_write
,
2930 .commit_write
= btrfs_commit_write
,
2932 .invalidatepage
= btrfs_invalidatepage
,
2933 .releasepage
= btrfs_releasepage
,
2934 .set_page_dirty
= __set_page_dirty_nobuffers
,
2937 static struct address_space_operations btrfs_symlink_aops
= {
2938 .readpage
= btrfs_readpage
,
2939 .writepage
= btrfs_writepage
,
2940 .invalidatepage
= btrfs_invalidatepage
,
2941 .releasepage
= btrfs_releasepage
,
2944 static struct inode_operations btrfs_file_inode_operations
= {
2945 .truncate
= btrfs_truncate
,
2946 .getattr
= btrfs_getattr
,
2947 .setattr
= btrfs_setattr
,
2948 .setxattr
= generic_setxattr
,
2949 .getxattr
= generic_getxattr
,
2950 .listxattr
= btrfs_listxattr
,
2951 .removexattr
= generic_removexattr
,
2954 static struct inode_operations btrfs_special_inode_operations
= {
2955 .getattr
= btrfs_getattr
,
2956 .setattr
= btrfs_setattr
,
2959 static struct inode_operations btrfs_symlink_inode_operations
= {
2960 .readlink
= generic_readlink
,
2961 .follow_link
= page_follow_link_light
,
2962 .put_link
= page_put_link
,