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/sched.h>
20 #include <linux/crc32c.h>
24 #include "print-tree.h"
25 #include "transaction.h"
27 #define BLOCK_GROUP_DATA EXTENT_WRITEBACK
28 #define BLOCK_GROUP_METADATA EXTENT_UPTODATE
29 #define BLOCK_GROUP_DIRTY EXTENT_DIRTY
31 static int finish_current_insert(struct btrfs_trans_handle
*trans
, struct
32 btrfs_root
*extent_root
);
33 static int del_pending_extents(struct btrfs_trans_handle
*trans
, struct
34 btrfs_root
*extent_root
);
36 static int cache_block_group(struct btrfs_root
*root
,
37 struct btrfs_block_group_cache
*block_group
)
39 struct btrfs_path
*path
;
42 struct extent_buffer
*leaf
;
43 struct extent_map_tree
*free_space_cache
;
53 root
= root
->fs_info
->extent_root
;
54 free_space_cache
= &root
->fs_info
->free_space_cache
;
56 if (block_group
->cached
)
59 path
= btrfs_alloc_path();
64 first_free
= block_group
->key
.objectid
;
65 key
.objectid
= block_group
->key
.objectid
;
68 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
69 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
74 if (ret
&& path
->slots
[0] > 0)
78 leaf
= path
->nodes
[0];
79 slot
= path
->slots
[0];
80 if (slot
>= btrfs_header_nritems(leaf
)) {
81 ret
= btrfs_next_leaf(root
, path
);
91 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
92 if (key
.objectid
< block_group
->key
.objectid
) {
93 if (btrfs_key_type(&key
) != BTRFS_EXTENT_REF_KEY
&&
94 key
.objectid
+ key
.offset
> first_free
)
95 first_free
= key
.objectid
+ key
.offset
;
99 if (key
.objectid
>= block_group
->key
.objectid
+
100 block_group
->key
.offset
) {
104 if (btrfs_key_type(&key
) == BTRFS_EXTENT_ITEM_KEY
) {
109 if (key
.objectid
> last
) {
110 hole_size
= key
.objectid
- last
;
111 set_extent_dirty(free_space_cache
, last
,
112 last
+ hole_size
- 1,
115 last
= key
.objectid
+ key
.offset
;
123 if (block_group
->key
.objectid
+
124 block_group
->key
.offset
> last
) {
125 hole_size
= block_group
->key
.objectid
+
126 block_group
->key
.offset
- last
;
127 set_extent_dirty(free_space_cache
, last
,
128 last
+ hole_size
- 1, GFP_NOFS
);
130 block_group
->cached
= 1;
132 btrfs_free_path(path
);
136 struct btrfs_block_group_cache
*btrfs_lookup_block_group(struct
140 struct extent_map_tree
*block_group_cache
;
141 struct btrfs_block_group_cache
*block_group
= NULL
;
147 block_group_cache
= &info
->block_group_cache
;
148 ret
= find_first_extent_bit(block_group_cache
,
149 bytenr
, &start
, &end
,
150 BLOCK_GROUP_DATA
| BLOCK_GROUP_METADATA
);
154 ret
= get_state_private(block_group_cache
, start
, &ptr
);
158 block_group
= (struct btrfs_block_group_cache
*)(unsigned long)ptr
;
159 if (block_group
->key
.objectid
<= bytenr
&& bytenr
<
160 block_group
->key
.objectid
+ block_group
->key
.offset
)
164 static u64
find_search_start(struct btrfs_root
*root
,
165 struct btrfs_block_group_cache
**cache_ret
,
166 u64 search_start
, int num
,
167 int data
, int full_scan
)
170 struct btrfs_block_group_cache
*cache
= *cache_ret
;
181 ret
= cache_block_group(root
, cache
);
185 last
= max(search_start
, cache
->key
.objectid
);
188 ret
= find_first_extent_bit(&root
->fs_info
->free_space_cache
,
189 last
, &start
, &end
, EXTENT_DIRTY
);
196 start
= max(last
, start
);
198 if (last
- start
< num
) {
199 if (last
== cache
->key
.objectid
+ cache
->key
.offset
)
203 if (data
!= BTRFS_BLOCK_GROUP_MIXED
&&
204 start
+ num
> cache
->key
.objectid
+ cache
->key
.offset
)
209 cache
= btrfs_lookup_block_group(root
->fs_info
, search_start
);
211 printk("Unable to find block group for %Lu\n",
219 last
= cache
->key
.objectid
+ cache
->key
.offset
;
221 cache
= btrfs_lookup_block_group(root
->fs_info
, last
);
227 data
= BTRFS_BLOCK_GROUP_MIXED
;
232 if (cache_miss
&& !cache
->cached
) {
233 cache_block_group(root
, cache
);
235 cache
= btrfs_lookup_block_group(root
->fs_info
, last
);
237 cache
= btrfs_find_block_group(root
, cache
, last
, data
, 0);
245 static u64
div_factor(u64 num
, int factor
)
254 struct btrfs_block_group_cache
*btrfs_find_block_group(struct btrfs_root
*root
,
255 struct btrfs_block_group_cache
256 *hint
, u64 search_start
,
259 struct btrfs_block_group_cache
*cache
;
260 struct extent_map_tree
*block_group_cache
;
261 struct btrfs_block_group_cache
*found_group
= NULL
;
262 struct btrfs_fs_info
*info
= root
->fs_info
;
276 block_group_cache
= &info
->block_group_cache
;
281 if (data
== BTRFS_BLOCK_GROUP_MIXED
) {
282 bit
= BLOCK_GROUP_DATA
| BLOCK_GROUP_METADATA
;
285 bit
= BLOCK_GROUP_DATA
;
287 bit
= BLOCK_GROUP_METADATA
;
290 struct btrfs_block_group_cache
*shint
;
291 shint
= btrfs_lookup_block_group(info
, search_start
);
292 if (shint
&& (shint
->data
== data
||
293 shint
->data
== BTRFS_BLOCK_GROUP_MIXED
)) {
294 used
= btrfs_block_group_used(&shint
->item
);
295 if (used
+ shint
->pinned
<
296 div_factor(shint
->key
.offset
, factor
)) {
301 if (hint
&& (hint
->data
== data
||
302 hint
->data
== BTRFS_BLOCK_GROUP_MIXED
)) {
303 used
= btrfs_block_group_used(&hint
->item
);
304 if (used
+ hint
->pinned
<
305 div_factor(hint
->key
.offset
, factor
)) {
308 last
= hint
->key
.objectid
+ hint
->key
.offset
;
312 hint_last
= max(hint
->key
.objectid
, search_start
);
314 hint_last
= search_start
;
320 ret
= find_first_extent_bit(block_group_cache
, last
,
325 ret
= get_state_private(block_group_cache
, start
, &ptr
);
329 cache
= (struct btrfs_block_group_cache
*)(unsigned long)ptr
;
330 last
= cache
->key
.objectid
+ cache
->key
.offset
;
331 used
= btrfs_block_group_used(&cache
->item
);
334 free_check
= cache
->key
.offset
;
336 free_check
= div_factor(cache
->key
.offset
, factor
);
337 if (used
+ cache
->pinned
< free_check
) {
350 bit
= BLOCK_GROUP_DATA
| BLOCK_GROUP_METADATA
;
358 static u64
hash_extent_ref(u64 root_objectid
, u64 ref_generation
,
359 u64 owner
, u64 owner_offset
)
361 u32 high_crc
= ~(u32
)0;
362 u32 low_crc
= ~(u32
)0;
365 lenum
= cpu_to_le64(root_objectid
);
366 high_crc
= crc32c(high_crc
, &lenum
, sizeof(lenum
));
367 lenum
= cpu_to_le64(ref_generation
);
368 low_crc
= crc32c(low_crc
, &lenum
, sizeof(lenum
));
371 lenum
= cpu_to_le64(owner
);
372 low_crc
= crc32c(low_crc
, &lenum
, sizeof(lenum
));
373 lenum
= cpu_to_le64(owner_offset
);
374 low_crc
= crc32c(low_crc
, &lenum
, sizeof(lenum
));
376 return ((u64
)high_crc
<< 32) | (u64
)low_crc
;
379 static int match_extent_ref(struct extent_buffer
*leaf
,
380 struct btrfs_extent_ref
*disk_ref
,
381 struct btrfs_extent_ref
*cpu_ref
)
386 if (cpu_ref
->objectid
)
387 len
= sizeof(*cpu_ref
);
389 len
= 2 * sizeof(u64
);
390 ret
= memcmp_extent_buffer(leaf
, cpu_ref
, (unsigned long)disk_ref
,
395 static int lookup_extent_backref(struct btrfs_trans_handle
*trans
,
396 struct btrfs_root
*root
,
397 struct btrfs_path
*path
, u64 bytenr
,
398 u64 root_objectid
, u64 ref_generation
,
399 u64 owner
, u64 owner_offset
, int del
)
402 struct btrfs_key key
;
403 struct btrfs_key found_key
;
404 struct btrfs_extent_ref ref
;
405 struct extent_buffer
*leaf
;
406 struct btrfs_extent_ref
*disk_ref
;
410 btrfs_set_stack_ref_root(&ref
, root_objectid
);
411 btrfs_set_stack_ref_generation(&ref
, ref_generation
);
412 btrfs_set_stack_ref_objectid(&ref
, owner
);
413 btrfs_set_stack_ref_offset(&ref
, owner_offset
);
415 hash
= hash_extent_ref(root_objectid
, ref_generation
, owner
,
418 key
.objectid
= bytenr
;
419 key
.type
= BTRFS_EXTENT_REF_KEY
;
422 ret
= btrfs_search_slot(trans
, root
, &key
, path
,
426 leaf
= path
->nodes
[0];
428 u32 nritems
= btrfs_header_nritems(leaf
);
429 if (path
->slots
[0] >= nritems
) {
430 ret2
= btrfs_next_leaf(root
, path
);
433 leaf
= path
->nodes
[0];
435 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
436 if (found_key
.objectid
!= bytenr
||
437 found_key
.type
!= BTRFS_EXTENT_REF_KEY
)
439 key
.offset
= found_key
.offset
;
441 btrfs_release_path(root
, path
);
445 disk_ref
= btrfs_item_ptr(path
->nodes
[0],
447 struct btrfs_extent_ref
);
448 if (match_extent_ref(path
->nodes
[0], disk_ref
, &ref
)) {
452 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
453 key
.offset
= found_key
.offset
+ 1;
454 btrfs_release_path(root
, path
);
461 * Back reference rules. Back refs have three main goals:
463 * 1) differentiate between all holders of references to an extent so that
464 * when a reference is dropped we can make sure it was a valid reference
465 * before freeing the extent.
467 * 2) Provide enough information to quickly find the holders of an extent
468 * if we notice a given block is corrupted or bad.
470 * 3) Make it easy to migrate blocks for FS shrinking or storage pool
471 * maintenance. This is actually the same as #2, but with a slightly
472 * different use case.
474 * File extents can be referenced by:
476 * - multiple snapshots, subvolumes, or different generations in one subvol
477 * - different files inside a single subvolume (in theory, not implemented yet)
478 * - different offsets inside a file (bookend extents in file.c)
480 * The extent ref structure has fields for:
482 * - Objectid of the subvolume root
483 * - Generation number of the tree holding the reference
484 * - objectid of the file holding the reference
485 * - offset in the file corresponding to the key holding the reference
487 * When a file extent is allocated the fields are filled in:
488 * (root_key.objectid, trans->transid, inode objectid, offset in file)
490 * When a leaf is cow'd new references are added for every file extent found
491 * in the leaf. It looks the same as the create case, but trans->transid
492 * will be different when the block is cow'd.
494 * (root_key.objectid, trans->transid, inode objectid, offset in file)
496 * When a file extent is removed either during snapshot deletion or file
497 * truncation, the corresponding back reference is found
500 * (btrfs_header_owner(leaf), btrfs_header_generation(leaf),
501 * inode objectid, offset in file)
503 * Btree extents can be referenced by:
505 * - Different subvolumes
506 * - Different generations of the same subvolume
508 * Storing sufficient information for a full reverse mapping of a btree
509 * block would require storing the lowest key of the block in the backref,
510 * and it would require updating that lowest key either before write out or
511 * every time it changed. Instead, the objectid of the lowest key is stored
512 * along with the level of the tree block. This provides a hint
513 * about where in the btree the block can be found. Searches through the
514 * btree only need to look for a pointer to that block, so they stop one
515 * level higher than the level recorded in the backref.
517 * Some btrees do not do reference counting on their extents. These
518 * include the extent tree and the tree of tree roots. Backrefs for these
519 * trees always have a generation of zero.
521 * When a tree block is created, back references are inserted:
523 * (root->root_key.objectid, trans->transid or zero, level, lowest_key_objectid)
525 * When a tree block is cow'd in a reference counted root,
526 * new back references are added for all the blocks it points to.
527 * These are of the form (trans->transid will have increased since creation):
529 * (root->root_key.objectid, trans->transid, level, lowest_key_objectid)
531 * Because the lowest_key_objectid and the level are just hints
532 * they are not used when backrefs are deleted. When a backref is deleted:
534 * if backref was for a tree root:
535 * root_objectid = root->root_key.objectid
537 * root_objectid = btrfs_header_owner(parent)
539 * (root_objectid, btrfs_header_generation(parent) or zero, 0, 0)
541 * Back Reference Key hashing:
543 * Back references have four fields, each 64 bits long. Unfortunately,
544 * This is hashed into a single 64 bit number and placed into the key offset.
545 * The key objectid corresponds to the first byte in the extent, and the
546 * key type is set to BTRFS_EXTENT_REF_KEY
548 int btrfs_insert_extent_backref(struct btrfs_trans_handle
*trans
,
549 struct btrfs_root
*root
,
550 struct btrfs_path
*path
, u64 bytenr
,
551 u64 root_objectid
, u64 ref_generation
,
552 u64 owner
, u64 owner_offset
)
555 struct btrfs_key key
;
556 struct btrfs_extent_ref ref
;
557 struct btrfs_extent_ref
*disk_ref
;
560 btrfs_set_stack_ref_root(&ref
, root_objectid
);
561 btrfs_set_stack_ref_generation(&ref
, ref_generation
);
562 btrfs_set_stack_ref_objectid(&ref
, owner
);
563 btrfs_set_stack_ref_offset(&ref
, owner_offset
);
565 hash
= hash_extent_ref(root_objectid
, ref_generation
, owner
,
568 key
.objectid
= bytenr
;
569 key
.type
= BTRFS_EXTENT_REF_KEY
;
571 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
, sizeof(ref
));
572 while (ret
== -EEXIST
) {
573 disk_ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
574 struct btrfs_extent_ref
);
575 if (match_extent_ref(path
->nodes
[0], disk_ref
, &ref
))
578 btrfs_release_path(root
, path
);
579 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
584 disk_ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
585 struct btrfs_extent_ref
);
586 write_extent_buffer(path
->nodes
[0], &ref
, (unsigned long)disk_ref
,
588 btrfs_mark_buffer_dirty(path
->nodes
[0]);
590 btrfs_release_path(root
, path
);
594 int btrfs_inc_extent_ref(struct btrfs_trans_handle
*trans
,
595 struct btrfs_root
*root
,
596 u64 bytenr
, u64 num_bytes
,
597 u64 root_objectid
, u64 ref_generation
,
598 u64 owner
, u64 owner_offset
)
600 struct btrfs_path
*path
;
602 struct btrfs_key key
;
603 struct extent_buffer
*l
;
604 struct btrfs_extent_item
*item
;
607 WARN_ON(num_bytes
< root
->sectorsize
);
608 path
= btrfs_alloc_path();
612 key
.objectid
= bytenr
;
613 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
614 key
.offset
= num_bytes
;
615 ret
= btrfs_search_slot(trans
, root
->fs_info
->extent_root
, &key
, path
,
624 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
625 refs
= btrfs_extent_refs(l
, item
);
626 btrfs_set_extent_refs(l
, item
, refs
+ 1);
627 btrfs_mark_buffer_dirty(path
->nodes
[0]);
629 btrfs_release_path(root
->fs_info
->extent_root
, path
);
631 ret
= btrfs_insert_extent_backref(trans
, root
->fs_info
->extent_root
,
632 path
, bytenr
, root_objectid
,
633 ref_generation
, owner
, owner_offset
);
635 finish_current_insert(trans
, root
->fs_info
->extent_root
);
636 del_pending_extents(trans
, root
->fs_info
->extent_root
);
638 btrfs_free_path(path
);
642 int btrfs_extent_post_op(struct btrfs_trans_handle
*trans
,
643 struct btrfs_root
*root
)
645 finish_current_insert(trans
, root
->fs_info
->extent_root
);
646 del_pending_extents(trans
, root
->fs_info
->extent_root
);
650 static int lookup_extent_ref(struct btrfs_trans_handle
*trans
,
651 struct btrfs_root
*root
, u64 bytenr
,
652 u64 num_bytes
, u32
*refs
)
654 struct btrfs_path
*path
;
656 struct btrfs_key key
;
657 struct extent_buffer
*l
;
658 struct btrfs_extent_item
*item
;
660 WARN_ON(num_bytes
< root
->sectorsize
);
661 path
= btrfs_alloc_path();
662 key
.objectid
= bytenr
;
663 key
.offset
= num_bytes
;
664 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
665 ret
= btrfs_search_slot(trans
, root
->fs_info
->extent_root
, &key
, path
,
670 btrfs_print_leaf(root
, path
->nodes
[0]);
671 printk("failed to find block number %Lu\n", bytenr
);
675 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
676 *refs
= btrfs_extent_refs(l
, item
);
678 btrfs_free_path(path
);
682 u32
btrfs_count_snapshots_in_path(struct btrfs_root
*root
,
683 struct btrfs_path
*count_path
,
686 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
687 struct btrfs_path
*path
;
690 u64 root_objectid
= 0;
696 struct btrfs_key key
;
697 struct btrfs_key found_key
;
698 struct extent_buffer
*l
;
699 struct btrfs_extent_item
*item
;
700 struct btrfs_extent_ref
*ref_item
;
703 path
= btrfs_alloc_path();
706 bytenr
= first_extent
;
708 bytenr
= count_path
->nodes
[level
]->start
;
711 key
.objectid
= bytenr
;
714 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
715 ret
= btrfs_search_slot(NULL
, extent_root
, &key
, path
, 0, 0);
721 btrfs_item_key_to_cpu(l
, &found_key
, path
->slots
[0]);
723 if (found_key
.objectid
!= bytenr
||
724 found_key
.type
!= BTRFS_EXTENT_ITEM_KEY
) {
728 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
729 refs
= btrfs_extent_refs(l
, item
);
731 nritems
= btrfs_header_nritems(l
);
732 if (path
->slots
[0] >= nritems
) {
733 ret
= btrfs_next_leaf(extent_root
, path
);
738 btrfs_item_key_to_cpu(l
, &found_key
, path
->slots
[0]);
739 if (found_key
.objectid
!= bytenr
)
741 if (found_key
.type
!= BTRFS_EXTENT_REF_KEY
) {
747 ref_item
= btrfs_item_ptr(l
, path
->slots
[0],
748 struct btrfs_extent_ref
);
749 found_objectid
= btrfs_ref_root(l
, ref_item
);
751 if (found_objectid
!= root_objectid
)
757 if (root_objectid
== 0)
758 root_objectid
= found_objectid
;
762 if (cur_count
== 0) {
768 if (level
>= 0 && root
->node
== count_path
->nodes
[level
])
771 btrfs_release_path(root
, path
);
775 btrfs_free_path(path
);
780 int btrfs_inc_root_ref(struct btrfs_trans_handle
*trans
,
781 struct btrfs_root
*root
, u64 owner_objectid
)
787 struct btrfs_disk_key disk_key
;
789 level
= btrfs_header_level(root
->node
);
790 generation
= trans
->transid
;
791 nritems
= btrfs_header_nritems(root
->node
);
794 btrfs_item_key(root
->node
, &disk_key
, 0);
796 btrfs_node_key(root
->node
, &disk_key
, 0);
797 key_objectid
= btrfs_disk_key_objectid(&disk_key
);
801 return btrfs_inc_extent_ref(trans
, root
, root
->node
->start
,
802 root
->node
->len
, owner_objectid
,
803 generation
, level
, key_objectid
);
806 int btrfs_inc_ref(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
807 struct extent_buffer
*buf
)
811 struct btrfs_key key
;
812 struct btrfs_file_extent_item
*fi
;
821 level
= btrfs_header_level(buf
);
822 nritems
= btrfs_header_nritems(buf
);
823 for (i
= 0; i
< nritems
; i
++) {
826 btrfs_item_key_to_cpu(buf
, &key
, i
);
827 if (btrfs_key_type(&key
) != BTRFS_EXTENT_DATA_KEY
)
829 fi
= btrfs_item_ptr(buf
, i
,
830 struct btrfs_file_extent_item
);
831 if (btrfs_file_extent_type(buf
, fi
) ==
832 BTRFS_FILE_EXTENT_INLINE
)
834 disk_bytenr
= btrfs_file_extent_disk_bytenr(buf
, fi
);
835 if (disk_bytenr
== 0)
837 ret
= btrfs_inc_extent_ref(trans
, root
, disk_bytenr
,
838 btrfs_file_extent_disk_num_bytes(buf
, fi
),
839 root
->root_key
.objectid
, trans
->transid
,
840 key
.objectid
, key
.offset
);
846 bytenr
= btrfs_node_blockptr(buf
, i
);
847 btrfs_node_key_to_cpu(buf
, &key
, i
);
848 ret
= btrfs_inc_extent_ref(trans
, root
, bytenr
,
849 btrfs_level_size(root
, level
- 1),
850 root
->root_key
.objectid
,
852 level
- 1, key
.objectid
);
863 for (i
=0; i
< faili
; i
++) {
866 btrfs_item_key_to_cpu(buf
, &key
, i
);
867 if (btrfs_key_type(&key
) != BTRFS_EXTENT_DATA_KEY
)
869 fi
= btrfs_item_ptr(buf
, i
,
870 struct btrfs_file_extent_item
);
871 if (btrfs_file_extent_type(buf
, fi
) ==
872 BTRFS_FILE_EXTENT_INLINE
)
874 disk_bytenr
= btrfs_file_extent_disk_bytenr(buf
, fi
);
875 if (disk_bytenr
== 0)
877 err
= btrfs_free_extent(trans
, root
, disk_bytenr
,
878 btrfs_file_extent_disk_num_bytes(buf
,
882 bytenr
= btrfs_node_blockptr(buf
, i
);
883 err
= btrfs_free_extent(trans
, root
, bytenr
,
884 btrfs_level_size(root
, level
- 1), 0);
892 static int write_one_cache_group(struct btrfs_trans_handle
*trans
,
893 struct btrfs_root
*root
,
894 struct btrfs_path
*path
,
895 struct btrfs_block_group_cache
*cache
)
899 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
901 struct extent_buffer
*leaf
;
903 ret
= btrfs_search_slot(trans
, extent_root
, &cache
->key
, path
, 0, 1);
908 leaf
= path
->nodes
[0];
909 bi
= btrfs_item_ptr_offset(leaf
, path
->slots
[0]);
910 write_extent_buffer(leaf
, &cache
->item
, bi
, sizeof(cache
->item
));
911 btrfs_mark_buffer_dirty(leaf
);
912 btrfs_release_path(extent_root
, path
);
914 finish_current_insert(trans
, extent_root
);
915 pending_ret
= del_pending_extents(trans
, extent_root
);
924 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle
*trans
,
925 struct btrfs_root
*root
)
927 struct extent_map_tree
*block_group_cache
;
928 struct btrfs_block_group_cache
*cache
;
932 struct btrfs_path
*path
;
938 block_group_cache
= &root
->fs_info
->block_group_cache
;
939 path
= btrfs_alloc_path();
944 ret
= find_first_extent_bit(block_group_cache
, last
,
945 &start
, &end
, BLOCK_GROUP_DIRTY
);
950 ret
= get_state_private(block_group_cache
, start
, &ptr
);
954 cache
= (struct btrfs_block_group_cache
*)(unsigned long)ptr
;
955 err
= write_one_cache_group(trans
, root
,
958 * if we fail to write the cache group, we want
959 * to keep it marked dirty in hopes that a later
966 clear_extent_bits(block_group_cache
, start
, end
,
967 BLOCK_GROUP_DIRTY
, GFP_NOFS
);
969 btrfs_free_path(path
);
973 static int update_block_group(struct btrfs_trans_handle
*trans
,
974 struct btrfs_root
*root
,
975 u64 bytenr
, u64 num_bytes
, int alloc
,
976 int mark_free
, int data
)
978 struct btrfs_block_group_cache
*cache
;
979 struct btrfs_fs_info
*info
= root
->fs_info
;
980 u64 total
= num_bytes
;
987 cache
= btrfs_lookup_block_group(info
, bytenr
);
991 byte_in_group
= bytenr
- cache
->key
.objectid
;
992 WARN_ON(byte_in_group
> cache
->key
.offset
);
993 start
= cache
->key
.objectid
;
994 end
= start
+ cache
->key
.offset
- 1;
995 set_extent_bits(&info
->block_group_cache
, start
, end
,
996 BLOCK_GROUP_DIRTY
, GFP_NOFS
);
998 old_val
= btrfs_block_group_used(&cache
->item
);
999 num_bytes
= min(total
, cache
->key
.offset
- byte_in_group
);
1001 if (cache
->data
!= data
&&
1002 old_val
< (cache
->key
.offset
>> 1)) {
1007 bit_to_clear
= BLOCK_GROUP_METADATA
;
1008 bit_to_set
= BLOCK_GROUP_DATA
;
1009 cache
->item
.flags
&=
1010 ~BTRFS_BLOCK_GROUP_MIXED
;
1011 cache
->item
.flags
|=
1012 BTRFS_BLOCK_GROUP_DATA
;
1014 bit_to_clear
= BLOCK_GROUP_DATA
;
1015 bit_to_set
= BLOCK_GROUP_METADATA
;
1016 cache
->item
.flags
&=
1017 ~BTRFS_BLOCK_GROUP_MIXED
;
1018 cache
->item
.flags
&=
1019 ~BTRFS_BLOCK_GROUP_DATA
;
1021 clear_extent_bits(&info
->block_group_cache
,
1022 start
, end
, bit_to_clear
,
1024 set_extent_bits(&info
->block_group_cache
,
1025 start
, end
, bit_to_set
,
1027 } else if (cache
->data
!= data
&&
1028 cache
->data
!= BTRFS_BLOCK_GROUP_MIXED
) {
1029 cache
->data
= BTRFS_BLOCK_GROUP_MIXED
;
1030 set_extent_bits(&info
->block_group_cache
,
1033 BLOCK_GROUP_METADATA
,
1036 old_val
+= num_bytes
;
1038 old_val
-= num_bytes
;
1040 set_extent_dirty(&info
->free_space_cache
,
1041 bytenr
, bytenr
+ num_bytes
- 1,
1045 btrfs_set_block_group_used(&cache
->item
, old_val
);
1047 bytenr
+= num_bytes
;
1051 static int update_pinned_extents(struct btrfs_root
*root
,
1052 u64 bytenr
, u64 num
, int pin
)
1055 struct btrfs_block_group_cache
*cache
;
1056 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1059 set_extent_dirty(&fs_info
->pinned_extents
,
1060 bytenr
, bytenr
+ num
- 1, GFP_NOFS
);
1062 clear_extent_dirty(&fs_info
->pinned_extents
,
1063 bytenr
, bytenr
+ num
- 1, GFP_NOFS
);
1066 cache
= btrfs_lookup_block_group(fs_info
, bytenr
);
1068 len
= min(num
, cache
->key
.offset
-
1069 (bytenr
- cache
->key
.objectid
));
1071 cache
->pinned
+= len
;
1072 fs_info
->total_pinned
+= len
;
1074 cache
->pinned
-= len
;
1075 fs_info
->total_pinned
-= len
;
1083 int btrfs_copy_pinned(struct btrfs_root
*root
, struct extent_map_tree
*copy
)
1088 struct extent_map_tree
*pinned_extents
= &root
->fs_info
->pinned_extents
;
1092 ret
= find_first_extent_bit(pinned_extents
, last
,
1093 &start
, &end
, EXTENT_DIRTY
);
1096 set_extent_dirty(copy
, start
, end
, GFP_NOFS
);
1102 int btrfs_finish_extent_commit(struct btrfs_trans_handle
*trans
,
1103 struct btrfs_root
*root
,
1104 struct extent_map_tree
*unpin
)
1109 struct extent_map_tree
*free_space_cache
;
1110 free_space_cache
= &root
->fs_info
->free_space_cache
;
1113 ret
= find_first_extent_bit(unpin
, 0, &start
, &end
,
1117 update_pinned_extents(root
, start
, end
+ 1 - start
, 0);
1118 clear_extent_dirty(unpin
, start
, end
, GFP_NOFS
);
1119 set_extent_dirty(free_space_cache
, start
, end
, GFP_NOFS
);
1124 static int finish_current_insert(struct btrfs_trans_handle
*trans
, struct
1125 btrfs_root
*extent_root
)
1129 struct btrfs_fs_info
*info
= extent_root
->fs_info
;
1130 struct extent_buffer
*eb
;
1131 struct btrfs_path
*path
;
1132 struct btrfs_key ins
;
1133 struct btrfs_disk_key first
;
1134 struct btrfs_extent_item extent_item
;
1139 btrfs_set_stack_extent_refs(&extent_item
, 1);
1140 btrfs_set_key_type(&ins
, BTRFS_EXTENT_ITEM_KEY
);
1141 path
= btrfs_alloc_path();
1144 ret
= find_first_extent_bit(&info
->extent_ins
, 0, &start
,
1145 &end
, EXTENT_LOCKED
);
1149 ins
.objectid
= start
;
1150 ins
.offset
= end
+ 1 - start
;
1151 err
= btrfs_insert_item(trans
, extent_root
, &ins
,
1152 &extent_item
, sizeof(extent_item
));
1153 clear_extent_bits(&info
->extent_ins
, start
, end
, EXTENT_LOCKED
,
1155 eb
= read_tree_block(extent_root
, ins
.objectid
, ins
.offset
);
1156 level
= btrfs_header_level(eb
);
1158 btrfs_item_key(eb
, &first
, 0);
1160 btrfs_node_key(eb
, &first
, 0);
1162 err
= btrfs_insert_extent_backref(trans
, extent_root
, path
,
1163 start
, extent_root
->root_key
.objectid
,
1165 btrfs_disk_key_objectid(&first
));
1167 free_extent_buffer(eb
);
1169 btrfs_free_path(path
);
1173 static int pin_down_bytes(struct btrfs_root
*root
, u64 bytenr
, u32 num_bytes
,
1177 struct extent_buffer
*buf
;
1180 buf
= btrfs_find_tree_block(root
, bytenr
, num_bytes
);
1182 if (btrfs_buffer_uptodate(buf
)) {
1184 root
->fs_info
->running_transaction
->transid
;
1185 if (btrfs_header_generation(buf
) == transid
) {
1186 free_extent_buffer(buf
);
1190 free_extent_buffer(buf
);
1192 update_pinned_extents(root
, bytenr
, num_bytes
, 1);
1194 set_extent_bits(&root
->fs_info
->pending_del
,
1195 bytenr
, bytenr
+ num_bytes
- 1,
1196 EXTENT_LOCKED
, GFP_NOFS
);
1203 * remove an extent from the root, returns 0 on success
1205 static int __free_extent(struct btrfs_trans_handle
*trans
, struct btrfs_root
1206 *root
, u64 bytenr
, u64 num_bytes
,
1207 u64 root_objectid
, u64 ref_generation
,
1208 u64 owner_objectid
, u64 owner_offset
, int pin
,
1211 struct btrfs_path
*path
;
1212 struct btrfs_key key
;
1213 struct btrfs_fs_info
*info
= root
->fs_info
;
1214 struct btrfs_root
*extent_root
= info
->extent_root
;
1215 struct extent_buffer
*leaf
;
1217 struct btrfs_extent_item
*ei
;
1220 key
.objectid
= bytenr
;
1221 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
1222 key
.offset
= num_bytes
;
1224 path
= btrfs_alloc_path();
1228 ret
= lookup_extent_backref(trans
, extent_root
, path
,
1229 bytenr
, root_objectid
,
1231 owner_objectid
, owner_offset
, 1);
1233 ret
= btrfs_del_item(trans
, extent_root
, path
);
1235 btrfs_print_leaf(extent_root
, path
->nodes
[0]);
1237 printk("Unable to find ref byte nr %Lu root %Lu "
1238 " gen %Lu owner %Lu offset %Lu\n", bytenr
,
1239 root_objectid
, ref_generation
, owner_objectid
,
1242 btrfs_release_path(extent_root
, path
);
1243 ret
= btrfs_search_slot(trans
, extent_root
, &key
, path
, -1, 1);
1248 leaf
= path
->nodes
[0];
1249 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
1250 struct btrfs_extent_item
);
1251 refs
= btrfs_extent_refs(leaf
, ei
);
1254 btrfs_set_extent_refs(leaf
, ei
, refs
);
1255 btrfs_mark_buffer_dirty(leaf
);
1262 ret
= pin_down_bytes(root
, bytenr
, num_bytes
, 0);
1268 /* block accounting for super block */
1269 super_used
= btrfs_super_bytes_used(&info
->super_copy
);
1270 btrfs_set_super_bytes_used(&info
->super_copy
,
1271 super_used
- num_bytes
);
1273 /* block accounting for root item */
1274 root_used
= btrfs_root_used(&root
->root_item
);
1275 btrfs_set_root_used(&root
->root_item
,
1276 root_used
- num_bytes
);
1278 ret
= btrfs_del_item(trans
, extent_root
, path
);
1282 ret
= update_block_group(trans
, root
, bytenr
, num_bytes
, 0,
1286 btrfs_free_path(path
);
1287 finish_current_insert(trans
, extent_root
);
1292 * find all the blocks marked as pending in the radix tree and remove
1293 * them from the extent map
1295 static int del_pending_extents(struct btrfs_trans_handle
*trans
, struct
1296 btrfs_root
*extent_root
)
1302 struct extent_map_tree
*pending_del
;
1303 struct extent_map_tree
*pinned_extents
;
1305 pending_del
= &extent_root
->fs_info
->pending_del
;
1306 pinned_extents
= &extent_root
->fs_info
->pinned_extents
;
1309 ret
= find_first_extent_bit(pending_del
, 0, &start
, &end
,
1313 update_pinned_extents(extent_root
, start
, end
+ 1 - start
, 1);
1314 clear_extent_bits(pending_del
, start
, end
, EXTENT_LOCKED
,
1316 ret
= __free_extent(trans
, extent_root
,
1317 start
, end
+ 1 - start
,
1318 extent_root
->root_key
.objectid
,
1327 * remove an extent from the root, returns 0 on success
1329 int btrfs_free_extent(struct btrfs_trans_handle
*trans
, struct btrfs_root
1330 *root
, u64 bytenr
, u64 num_bytes
,
1331 u64 root_objectid
, u64 ref_generation
,
1332 u64 owner_objectid
, u64 owner_offset
, int pin
)
1334 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
1338 WARN_ON(num_bytes
< root
->sectorsize
);
1339 if (!root
->ref_cows
)
1342 if (root
== extent_root
) {
1343 pin_down_bytes(root
, bytenr
, num_bytes
, 1);
1346 ret
= __free_extent(trans
, root
, bytenr
, num_bytes
, root_objectid
,
1347 ref_generation
, owner_objectid
, owner_offset
,
1349 pending_ret
= del_pending_extents(trans
, root
->fs_info
->extent_root
);
1350 return ret
? ret
: pending_ret
;
1353 static u64
stripe_align(struct btrfs_root
*root
, u64 val
)
1355 u64 mask
= ((u64
)root
->stripesize
- 1);
1356 u64 ret
= (val
+ mask
) & ~mask
;
1361 * walks the btree of allocated extents and find a hole of a given size.
1362 * The key ins is changed to record the hole:
1363 * ins->objectid == block start
1364 * ins->flags = BTRFS_EXTENT_ITEM_KEY
1365 * ins->offset == number of blocks
1366 * Any available blocks before search_start are skipped.
1368 static int find_free_extent(struct btrfs_trans_handle
*trans
, struct btrfs_root
1369 *orig_root
, u64 num_bytes
, u64 empty_size
,
1370 u64 search_start
, u64 search_end
, u64 hint_byte
,
1371 struct btrfs_key
*ins
, u64 exclude_start
,
1372 u64 exclude_nr
, int data
)
1374 struct btrfs_path
*path
;
1375 struct btrfs_key key
;
1381 u64 orig_search_start
= search_start
;
1383 struct extent_buffer
*l
;
1384 struct btrfs_root
* root
= orig_root
->fs_info
->extent_root
;
1385 struct btrfs_fs_info
*info
= root
->fs_info
;
1386 u64 total_needed
= num_bytes
;
1388 struct btrfs_block_group_cache
*block_group
;
1393 WARN_ON(num_bytes
< root
->sectorsize
);
1394 btrfs_set_key_type(ins
, BTRFS_EXTENT_ITEM_KEY
);
1396 level
= btrfs_header_level(root
->node
);
1398 if (num_bytes
>= 32 * 1024 * 1024 && hint_byte
) {
1399 data
= BTRFS_BLOCK_GROUP_MIXED
;
1402 if (search_end
== (u64
)-1)
1403 search_end
= btrfs_super_total_bytes(&info
->super_copy
);
1405 block_group
= btrfs_lookup_block_group(info
, hint_byte
);
1407 hint_byte
= search_start
;
1408 block_group
= btrfs_find_block_group(root
, block_group
,
1409 hint_byte
, data
, 1);
1411 block_group
= btrfs_find_block_group(root
,
1413 search_start
, data
, 1);
1416 total_needed
+= empty_size
;
1417 path
= btrfs_alloc_path();
1420 block_group
= btrfs_lookup_block_group(info
, search_start
);
1422 block_group
= btrfs_lookup_block_group(info
,
1425 search_start
= find_search_start(root
, &block_group
, search_start
,
1426 total_needed
, data
, full_scan
);
1427 search_start
= stripe_align(root
, search_start
);
1428 cached_start
= search_start
;
1429 btrfs_init_path(path
);
1430 ins
->objectid
= search_start
;
1435 ret
= btrfs_search_slot(trans
, root
, ins
, path
, 0, 0);
1439 if (path
->slots
[0] > 0) {
1444 btrfs_item_key_to_cpu(l
, &key
, path
->slots
[0]);
1447 * walk backwards to find the first extent item key
1449 while(btrfs_key_type(&key
) != BTRFS_EXTENT_ITEM_KEY
) {
1450 if (path
->slots
[0] == 0) {
1451 ret
= btrfs_prev_leaf(root
, path
);
1453 ret
= btrfs_search_slot(trans
, root
, ins
,
1457 if (path
->slots
[0] > 0)
1465 btrfs_item_key_to_cpu(l
, &key
, path
->slots
[0]);
1469 slot
= path
->slots
[0];
1470 if (slot
>= btrfs_header_nritems(l
)) {
1471 ret
= btrfs_next_leaf(root
, path
);
1477 search_start
= max(search_start
,
1478 block_group
->key
.objectid
);
1480 aligned
= stripe_align(root
, search_start
);
1481 ins
->objectid
= aligned
;
1482 if (aligned
>= search_end
) {
1486 ins
->offset
= search_end
- aligned
;
1490 ins
->objectid
= stripe_align(root
,
1491 last_byte
> search_start
?
1492 last_byte
: search_start
);
1493 if (search_end
<= ins
->objectid
) {
1497 ins
->offset
= search_end
- ins
->objectid
;
1498 BUG_ON(ins
->objectid
>= search_end
);
1501 btrfs_item_key_to_cpu(l
, &key
, slot
);
1503 if (key
.objectid
>= search_start
&& key
.objectid
> last_byte
&&
1505 if (last_byte
< search_start
)
1506 last_byte
= search_start
;
1507 aligned
= stripe_align(root
, last_byte
);
1508 hole_size
= key
.objectid
- aligned
;
1509 if (key
.objectid
> aligned
&& hole_size
>= num_bytes
) {
1510 ins
->objectid
= aligned
;
1511 ins
->offset
= hole_size
;
1515 if (btrfs_key_type(&key
) != BTRFS_EXTENT_ITEM_KEY
) {
1516 if (!start_found
&& btrfs_key_type(&key
) ==
1517 BTRFS_BLOCK_GROUP_ITEM_KEY
) {
1518 last_byte
= key
.objectid
;
1526 last_byte
= key
.objectid
+ key
.offset
;
1528 if (!full_scan
&& data
!= BTRFS_BLOCK_GROUP_MIXED
&&
1529 last_byte
>= block_group
->key
.objectid
+
1530 block_group
->key
.offset
) {
1531 btrfs_release_path(root
, path
);
1532 search_start
= block_group
->key
.objectid
+
1533 block_group
->key
.offset
;
1541 /* we have to make sure we didn't find an extent that has already
1542 * been allocated by the map tree or the original allocation
1544 btrfs_release_path(root
, path
);
1545 BUG_ON(ins
->objectid
< search_start
);
1547 if (ins
->objectid
+ num_bytes
>= search_end
)
1549 if (!full_scan
&& data
!= BTRFS_BLOCK_GROUP_MIXED
&&
1550 ins
->objectid
+ num_bytes
> block_group
->
1551 key
.objectid
+ block_group
->key
.offset
) {
1552 search_start
= block_group
->key
.objectid
+
1553 block_group
->key
.offset
;
1556 if (test_range_bit(&info
->extent_ins
, ins
->objectid
,
1557 ins
->objectid
+ num_bytes
-1, EXTENT_LOCKED
, 0)) {
1558 search_start
= ins
->objectid
+ num_bytes
;
1561 if (test_range_bit(&info
->pinned_extents
, ins
->objectid
,
1562 ins
->objectid
+ num_bytes
-1, EXTENT_DIRTY
, 0)) {
1563 search_start
= ins
->objectid
+ num_bytes
;
1566 if (exclude_nr
> 0 && (ins
->objectid
+ num_bytes
> exclude_start
&&
1567 ins
->objectid
< exclude_start
+ exclude_nr
)) {
1568 search_start
= exclude_start
+ exclude_nr
;
1572 block_group
= btrfs_lookup_block_group(info
, ins
->objectid
);
1574 trans
->block_group
= block_group
;
1576 ins
->offset
= num_bytes
;
1577 btrfs_free_path(path
);
1581 if (search_start
+ num_bytes
>= search_end
) {
1583 search_start
= orig_search_start
;
1590 total_needed
-= empty_size
;
1592 data
= BTRFS_BLOCK_GROUP_MIXED
;
1596 block_group
= btrfs_lookup_block_group(info
, search_start
);
1598 block_group
= btrfs_find_block_group(root
, block_group
,
1599 search_start
, data
, 0);
1603 btrfs_release_path(root
, path
);
1604 btrfs_free_path(path
);
1608 * finds a free extent and does all the dirty work required for allocation
1609 * returns the key for the extent through ins, and a tree buffer for
1610 * the first block of the extent through buf.
1612 * returns 0 if everything worked, non-zero otherwise.
1614 int btrfs_alloc_extent(struct btrfs_trans_handle
*trans
,
1615 struct btrfs_root
*root
,
1616 u64 num_bytes
, u64 root_objectid
, u64 ref_generation
,
1617 u64 owner
, u64 owner_offset
,
1618 u64 empty_size
, u64 hint_byte
,
1619 u64 search_end
, struct btrfs_key
*ins
, int data
)
1623 u64 super_used
, root_used
;
1624 u64 search_start
= 0;
1625 struct btrfs_fs_info
*info
= root
->fs_info
;
1626 struct btrfs_root
*extent_root
= info
->extent_root
;
1627 struct btrfs_extent_item extent_item
;
1628 struct btrfs_path
*path
;
1630 btrfs_set_stack_extent_refs(&extent_item
, 1);
1632 WARN_ON(num_bytes
< root
->sectorsize
);
1633 ret
= find_free_extent(trans
, root
, num_bytes
, empty_size
,
1634 search_start
, search_end
, hint_byte
, ins
,
1635 trans
->alloc_exclude_start
,
1636 trans
->alloc_exclude_nr
, data
);
1641 /* block accounting for super block */
1642 super_used
= btrfs_super_bytes_used(&info
->super_copy
);
1643 btrfs_set_super_bytes_used(&info
->super_copy
, super_used
+ num_bytes
);
1645 /* block accounting for root item */
1646 root_used
= btrfs_root_used(&root
->root_item
);
1647 btrfs_set_root_used(&root
->root_item
, root_used
+ num_bytes
);
1649 clear_extent_dirty(&root
->fs_info
->free_space_cache
,
1650 ins
->objectid
, ins
->objectid
+ ins
->offset
- 1,
1653 if (root
== extent_root
) {
1654 set_extent_bits(&root
->fs_info
->extent_ins
, ins
->objectid
,
1655 ins
->objectid
+ ins
->offset
- 1,
1656 EXTENT_LOCKED
, GFP_NOFS
);
1661 WARN_ON(trans
->alloc_exclude_nr
);
1662 trans
->alloc_exclude_start
= ins
->objectid
;
1663 trans
->alloc_exclude_nr
= ins
->offset
;
1664 ret
= btrfs_insert_item(trans
, extent_root
, ins
, &extent_item
,
1665 sizeof(extent_item
));
1667 trans
->alloc_exclude_start
= 0;
1668 trans
->alloc_exclude_nr
= 0;
1671 path
= btrfs_alloc_path();
1673 ret
= btrfs_insert_extent_backref(trans
, extent_root
, path
,
1674 ins
->objectid
, root_objectid
,
1675 ref_generation
, owner
, owner_offset
);
1678 btrfs_free_path(path
);
1679 finish_current_insert(trans
, extent_root
);
1680 pending_ret
= del_pending_extents(trans
, extent_root
);
1690 ret
= update_block_group(trans
, root
, ins
->objectid
, ins
->offset
, 1, 0,
1697 * helper function to allocate a block for a given tree
1698 * returns the tree buffer or NULL.
1700 struct extent_buffer
*btrfs_alloc_free_block(struct btrfs_trans_handle
*trans
,
1701 struct btrfs_root
*root
,
1703 u64 root_objectid
, u64 hint
,
1709 ref_generation
= trans
->transid
;
1714 return __btrfs_alloc_free_block(trans
, root
, blocksize
, root_objectid
,
1715 ref_generation
, 0, 0, hint
, empty_size
);
1719 * helper function to allocate a block for a given tree
1720 * returns the tree buffer or NULL.
1722 struct extent_buffer
*__btrfs_alloc_free_block(struct btrfs_trans_handle
*trans
,
1723 struct btrfs_root
*root
,
1732 struct btrfs_key ins
;
1734 struct extent_buffer
*buf
;
1736 ret
= btrfs_alloc_extent(trans
, root
, blocksize
,
1737 root_objectid
, ref_generation
,
1738 level
, first_objectid
, empty_size
, hint
,
1742 return ERR_PTR(ret
);
1744 buf
= btrfs_find_create_tree_block(root
, ins
.objectid
, blocksize
);
1746 btrfs_free_extent(trans
, root
, ins
.objectid
, blocksize
,
1747 root
->root_key
.objectid
, ref_generation
,
1749 return ERR_PTR(-ENOMEM
);
1751 btrfs_set_buffer_uptodate(buf
);
1752 set_extent_dirty(&trans
->transaction
->dirty_pages
, buf
->start
,
1753 buf
->start
+ buf
->len
- 1, GFP_NOFS
);
1754 set_extent_bits(&BTRFS_I(root
->fs_info
->btree_inode
)->extent_tree
,
1755 buf
->start
, buf
->start
+ buf
->len
- 1,
1756 EXTENT_CSUM
, GFP_NOFS
);
1757 buf
->flags
|= EXTENT_CSUM
;
1758 btrfs_set_buffer_defrag(buf
);
1759 trans
->blocks_used
++;
1763 static int drop_leaf_ref(struct btrfs_trans_handle
*trans
,
1764 struct btrfs_root
*root
, struct extent_buffer
*leaf
)
1767 u64 leaf_generation
;
1768 struct btrfs_key key
;
1769 struct btrfs_file_extent_item
*fi
;
1774 BUG_ON(!btrfs_is_leaf(leaf
));
1775 nritems
= btrfs_header_nritems(leaf
);
1776 leaf_owner
= btrfs_header_owner(leaf
);
1777 leaf_generation
= btrfs_header_generation(leaf
);
1779 for (i
= 0; i
< nritems
; i
++) {
1782 btrfs_item_key_to_cpu(leaf
, &key
, i
);
1783 if (btrfs_key_type(&key
) != BTRFS_EXTENT_DATA_KEY
)
1785 fi
= btrfs_item_ptr(leaf
, i
, struct btrfs_file_extent_item
);
1786 if (btrfs_file_extent_type(leaf
, fi
) ==
1787 BTRFS_FILE_EXTENT_INLINE
)
1790 * FIXME make sure to insert a trans record that
1791 * repeats the snapshot del on crash
1793 disk_bytenr
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
1794 if (disk_bytenr
== 0)
1796 ret
= btrfs_free_extent(trans
, root
, disk_bytenr
,
1797 btrfs_file_extent_disk_num_bytes(leaf
, fi
),
1798 leaf_owner
, leaf_generation
,
1799 key
.objectid
, key
.offset
, 0);
1805 static void reada_walk_down(struct btrfs_root
*root
,
1806 struct extent_buffer
*node
)
1816 nritems
= btrfs_header_nritems(node
);
1817 level
= btrfs_header_level(node
);
1818 for (i
= 0; i
< nritems
; i
++) {
1819 bytenr
= btrfs_node_blockptr(node
, i
);
1820 blocksize
= btrfs_level_size(root
, level
- 1);
1821 ret
= lookup_extent_ref(NULL
, root
, bytenr
, blocksize
, &refs
);
1825 mutex_unlock(&root
->fs_info
->fs_mutex
);
1826 ret
= readahead_tree_block(root
, bytenr
, blocksize
);
1828 mutex_lock(&root
->fs_info
->fs_mutex
);
1835 * helper function for drop_snapshot, this walks down the tree dropping ref
1836 * counts as it goes.
1838 static int walk_down_tree(struct btrfs_trans_handle
*trans
, struct btrfs_root
1839 *root
, struct btrfs_path
*path
, int *level
)
1844 struct extent_buffer
*next
;
1845 struct extent_buffer
*cur
;
1846 struct extent_buffer
*parent
;
1851 WARN_ON(*level
< 0);
1852 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
1853 ret
= lookup_extent_ref(trans
, root
,
1854 path
->nodes
[*level
]->start
,
1855 path
->nodes
[*level
]->len
, &refs
);
1861 * walk down to the last node level and free all the leaves
1863 while(*level
>= 0) {
1864 WARN_ON(*level
< 0);
1865 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
1866 cur
= path
->nodes
[*level
];
1868 if (*level
> 0 && path
->slots
[*level
] == 0)
1869 reada_walk_down(root
, cur
);
1871 if (btrfs_header_level(cur
) != *level
)
1874 if (path
->slots
[*level
] >=
1875 btrfs_header_nritems(cur
))
1878 ret
= drop_leaf_ref(trans
, root
, cur
);
1882 bytenr
= btrfs_node_blockptr(cur
, path
->slots
[*level
]);
1883 blocksize
= btrfs_level_size(root
, *level
- 1);
1884 ret
= lookup_extent_ref(trans
, root
, bytenr
, blocksize
, &refs
);
1887 parent
= path
->nodes
[*level
];
1888 root_owner
= btrfs_header_owner(parent
);
1889 root_gen
= btrfs_header_generation(parent
);
1890 path
->slots
[*level
]++;
1891 ret
= btrfs_free_extent(trans
, root
, bytenr
,
1892 blocksize
, root_owner
,
1897 next
= btrfs_find_tree_block(root
, bytenr
, blocksize
);
1898 if (!next
|| !btrfs_buffer_uptodate(next
)) {
1899 free_extent_buffer(next
);
1900 mutex_unlock(&root
->fs_info
->fs_mutex
);
1901 next
= read_tree_block(root
, bytenr
, blocksize
);
1902 mutex_lock(&root
->fs_info
->fs_mutex
);
1904 /* we dropped the lock, check one more time */
1905 ret
= lookup_extent_ref(trans
, root
, bytenr
,
1909 parent
= path
->nodes
[*level
];
1910 root_owner
= btrfs_header_owner(parent
);
1911 root_gen
= btrfs_header_generation(parent
);
1913 path
->slots
[*level
]++;
1914 free_extent_buffer(next
);
1915 ret
= btrfs_free_extent(trans
, root
, bytenr
,
1923 WARN_ON(*level
<= 0);
1924 if (path
->nodes
[*level
-1])
1925 free_extent_buffer(path
->nodes
[*level
-1]);
1926 path
->nodes
[*level
-1] = next
;
1927 *level
= btrfs_header_level(next
);
1928 path
->slots
[*level
] = 0;
1931 WARN_ON(*level
< 0);
1932 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
1934 if (path
->nodes
[*level
] == root
->node
) {
1935 root_owner
= root
->root_key
.objectid
;
1936 parent
= path
->nodes
[*level
];
1938 parent
= path
->nodes
[*level
+ 1];
1939 root_owner
= btrfs_header_owner(parent
);
1942 root_gen
= btrfs_header_generation(parent
);
1943 ret
= btrfs_free_extent(trans
, root
, path
->nodes
[*level
]->start
,
1944 path
->nodes
[*level
]->len
,
1945 root_owner
, root_gen
, 0, 0, 1);
1946 free_extent_buffer(path
->nodes
[*level
]);
1947 path
->nodes
[*level
] = NULL
;
1954 * helper for dropping snapshots. This walks back up the tree in the path
1955 * to find the first node higher up where we haven't yet gone through
1958 static int walk_up_tree(struct btrfs_trans_handle
*trans
, struct btrfs_root
1959 *root
, struct btrfs_path
*path
, int *level
)
1963 struct btrfs_root_item
*root_item
= &root
->root_item
;
1968 for(i
= *level
; i
< BTRFS_MAX_LEVEL
- 1 && path
->nodes
[i
]; i
++) {
1969 slot
= path
->slots
[i
];
1970 if (slot
< btrfs_header_nritems(path
->nodes
[i
]) - 1) {
1971 struct extent_buffer
*node
;
1972 struct btrfs_disk_key disk_key
;
1973 node
= path
->nodes
[i
];
1976 WARN_ON(*level
== 0);
1977 btrfs_node_key(node
, &disk_key
, path
->slots
[i
]);
1978 memcpy(&root_item
->drop_progress
,
1979 &disk_key
, sizeof(disk_key
));
1980 root_item
->drop_level
= i
;
1983 if (path
->nodes
[*level
] == root
->node
) {
1984 root_owner
= root
->root_key
.objectid
;
1986 btrfs_header_generation(path
->nodes
[*level
]);
1988 struct extent_buffer
*node
;
1989 node
= path
->nodes
[*level
+ 1];
1990 root_owner
= btrfs_header_owner(node
);
1991 root_gen
= btrfs_header_generation(node
);
1993 ret
= btrfs_free_extent(trans
, root
,
1994 path
->nodes
[*level
]->start
,
1995 path
->nodes
[*level
]->len
,
1996 root_owner
, root_gen
, 0, 0, 1);
1998 free_extent_buffer(path
->nodes
[*level
]);
1999 path
->nodes
[*level
] = NULL
;
2007 * drop the reference count on the tree rooted at 'snap'. This traverses
2008 * the tree freeing any blocks that have a ref count of zero after being
2011 int btrfs_drop_snapshot(struct btrfs_trans_handle
*trans
, struct btrfs_root
2017 struct btrfs_path
*path
;
2020 struct btrfs_root_item
*root_item
= &root
->root_item
;
2022 path
= btrfs_alloc_path();
2025 level
= btrfs_header_level(root
->node
);
2027 if (btrfs_disk_key_objectid(&root_item
->drop_progress
) == 0) {
2028 path
->nodes
[level
] = root
->node
;
2029 extent_buffer_get(root
->node
);
2030 path
->slots
[level
] = 0;
2032 struct btrfs_key key
;
2033 struct btrfs_disk_key found_key
;
2034 struct extent_buffer
*node
;
2036 btrfs_disk_key_to_cpu(&key
, &root_item
->drop_progress
);
2037 level
= root_item
->drop_level
;
2038 path
->lowest_level
= level
;
2039 wret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2044 node
= path
->nodes
[level
];
2045 btrfs_node_key(node
, &found_key
, path
->slots
[level
]);
2046 WARN_ON(memcmp(&found_key
, &root_item
->drop_progress
,
2047 sizeof(found_key
)));
2050 wret
= walk_down_tree(trans
, root
, path
, &level
);
2056 wret
= walk_up_tree(trans
, root
, path
, &level
);
2064 for (i
= 0; i
<= orig_level
; i
++) {
2065 if (path
->nodes
[i
]) {
2066 free_extent_buffer(path
->nodes
[i
]);
2067 path
->nodes
[i
] = NULL
;
2071 btrfs_free_path(path
);
2075 int btrfs_free_block_groups(struct btrfs_fs_info
*info
)
2082 ret
= find_first_extent_bit(&info
->block_group_cache
, 0,
2083 &start
, &end
, (unsigned int)-1);
2086 ret
= get_state_private(&info
->block_group_cache
, start
, &ptr
);
2088 kfree((void *)(unsigned long)ptr
);
2089 clear_extent_bits(&info
->block_group_cache
, start
,
2090 end
, (unsigned int)-1, GFP_NOFS
);
2093 ret
= find_first_extent_bit(&info
->free_space_cache
, 0,
2094 &start
, &end
, EXTENT_DIRTY
);
2097 clear_extent_dirty(&info
->free_space_cache
, start
,
2103 int btrfs_read_block_groups(struct btrfs_root
*root
)
2105 struct btrfs_path
*path
;
2109 struct btrfs_block_group_cache
*cache
;
2110 struct btrfs_fs_info
*info
= root
->fs_info
;
2111 struct extent_map_tree
*block_group_cache
;
2112 struct btrfs_key key
;
2113 struct btrfs_key found_key
;
2114 struct extent_buffer
*leaf
;
2116 block_group_cache
= &info
->block_group_cache
;
2118 root
= info
->extent_root
;
2120 key
.offset
= BTRFS_BLOCK_GROUP_SIZE
;
2121 btrfs_set_key_type(&key
, BTRFS_BLOCK_GROUP_ITEM_KEY
);
2123 path
= btrfs_alloc_path();
2128 ret
= btrfs_search_slot(NULL
, info
->extent_root
,
2134 leaf
= path
->nodes
[0];
2135 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2136 cache
= kmalloc(sizeof(*cache
), GFP_NOFS
);
2142 read_extent_buffer(leaf
, &cache
->item
,
2143 btrfs_item_ptr_offset(leaf
, path
->slots
[0]),
2144 sizeof(cache
->item
));
2145 memcpy(&cache
->key
, &found_key
, sizeof(found_key
));
2148 key
.objectid
= found_key
.objectid
+ found_key
.offset
;
2149 btrfs_release_path(root
, path
);
2151 if (cache
->item
.flags
& BTRFS_BLOCK_GROUP_MIXED
) {
2152 bit
= BLOCK_GROUP_DATA
| BLOCK_GROUP_METADATA
;
2153 cache
->data
= BTRFS_BLOCK_GROUP_MIXED
;
2154 } else if (cache
->item
.flags
& BTRFS_BLOCK_GROUP_DATA
) {
2155 bit
= BLOCK_GROUP_DATA
;
2156 cache
->data
= BTRFS_BLOCK_GROUP_DATA
;
2158 bit
= BLOCK_GROUP_METADATA
;
2162 /* use EXTENT_LOCKED to prevent merging */
2163 set_extent_bits(block_group_cache
, found_key
.objectid
,
2164 found_key
.objectid
+ found_key
.offset
- 1,
2165 bit
| EXTENT_LOCKED
, GFP_NOFS
);
2166 set_state_private(block_group_cache
, found_key
.objectid
,
2167 (unsigned long)cache
);
2170 btrfs_super_total_bytes(&info
->super_copy
))
2174 btrfs_free_path(path
);