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>
21 #include <linux/pagemap.h>
25 #include "print-tree.h"
26 #include "transaction.h"
29 #define BLOCK_GROUP_DATA EXTENT_WRITEBACK
30 #define BLOCK_GROUP_METADATA EXTENT_UPTODATE
31 #define BLOCK_GROUP_SYSTEM EXTENT_NEW
33 #define BLOCK_GROUP_DIRTY EXTENT_DIRTY
35 static int finish_current_insert(struct btrfs_trans_handle
*trans
, struct
36 btrfs_root
*extent_root
);
37 static int del_pending_extents(struct btrfs_trans_handle
*trans
, struct
38 btrfs_root
*extent_root
);
39 int btrfs_make_block_group(struct btrfs_trans_handle
*trans
,
40 struct btrfs_root
*root
, u64 bytes_used
,
41 u64 type
, u64 chunk_tree
, u64 chunk_objectid
,
45 static int cache_block_group(struct btrfs_root
*root
,
46 struct btrfs_block_group_cache
*block_group
)
48 struct btrfs_path
*path
;
51 struct extent_buffer
*leaf
;
52 struct extent_io_tree
*free_space_cache
;
62 root
= root
->fs_info
->extent_root
;
63 free_space_cache
= &root
->fs_info
->free_space_cache
;
65 if (block_group
->cached
)
68 path
= btrfs_alloc_path();
73 first_free
= block_group
->key
.objectid
;
74 key
.objectid
= block_group
->key
.objectid
;
76 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
77 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
80 ret
= btrfs_previous_item(root
, path
, 0, BTRFS_EXTENT_ITEM_KEY
);
84 leaf
= path
->nodes
[0];
85 btrfs_item_key_to_cpu(leaf
, &key
, path
->slots
[0]);
86 if (key
.objectid
+ key
.offset
> first_free
)
87 first_free
= key
.objectid
+ key
.offset
;
90 leaf
= path
->nodes
[0];
91 slot
= path
->slots
[0];
92 if (slot
>= btrfs_header_nritems(leaf
)) {
93 ret
= btrfs_next_leaf(root
, path
);
102 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
103 if (key
.objectid
< block_group
->key
.objectid
) {
106 if (key
.objectid
>= block_group
->key
.objectid
+
107 block_group
->key
.offset
) {
111 if (btrfs_key_type(&key
) == BTRFS_EXTENT_ITEM_KEY
) {
116 if (key
.objectid
> last
) {
117 hole_size
= key
.objectid
- last
;
118 set_extent_dirty(free_space_cache
, last
,
119 last
+ hole_size
- 1,
122 last
= key
.objectid
+ key
.offset
;
130 if (block_group
->key
.objectid
+
131 block_group
->key
.offset
> last
) {
132 hole_size
= block_group
->key
.objectid
+
133 block_group
->key
.offset
- last
;
134 set_extent_dirty(free_space_cache
, last
,
135 last
+ hole_size
- 1, GFP_NOFS
);
137 block_group
->cached
= 1;
139 btrfs_free_path(path
);
143 struct btrfs_block_group_cache
*btrfs_lookup_block_group(struct
147 struct extent_io_tree
*block_group_cache
;
148 struct btrfs_block_group_cache
*block_group
= NULL
;
154 block_group_cache
= &info
->block_group_cache
;
155 ret
= find_first_extent_bit(block_group_cache
,
156 bytenr
, &start
, &end
,
157 BLOCK_GROUP_DATA
| BLOCK_GROUP_METADATA
|
162 ret
= get_state_private(block_group_cache
, start
, &ptr
);
166 block_group
= (struct btrfs_block_group_cache
*)(unsigned long)ptr
;
167 if (block_group
->key
.objectid
<= bytenr
&& bytenr
<
168 block_group
->key
.objectid
+ block_group
->key
.offset
)
173 static int block_group_bits(struct btrfs_block_group_cache
*cache
, u64 bits
)
175 return (cache
->flags
& bits
) == bits
;
178 static int noinline
find_search_start(struct btrfs_root
*root
,
179 struct btrfs_block_group_cache
**cache_ret
,
180 u64
*start_ret
, int num
, int data
)
183 struct btrfs_block_group_cache
*cache
= *cache_ret
;
184 struct extent_io_tree
*free_space_cache
;
185 struct extent_state
*state
;
190 u64 search_start
= *start_ret
;
195 total_fs_bytes
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
196 free_space_cache
= &root
->fs_info
->free_space_cache
;
199 ret
= cache_block_group(root
, cache
);
203 last
= max(search_start
, cache
->key
.objectid
);
204 if (!block_group_bits(cache
, data
)) {
208 spin_lock_irq(&free_space_cache
->lock
);
209 state
= find_first_extent_bit_state(free_space_cache
, last
, EXTENT_DIRTY
);
214 spin_unlock_irq(&free_space_cache
->lock
);
218 start
= max(last
, state
->start
);
219 last
= state
->end
+ 1;
220 if (last
- start
< num
) {
221 if (last
== cache
->key
.objectid
+ cache
->key
.offset
)
224 state
= extent_state_next(state
);
225 } while(state
&& !(state
->state
& EXTENT_DIRTY
));
228 spin_unlock_irq(&free_space_cache
->lock
);
229 if (start
+ num
> cache
->key
.objectid
+ cache
->key
.offset
)
231 if (start
+ num
> total_fs_bytes
)
236 cache
= btrfs_lookup_block_group(root
->fs_info
, search_start
);
238 printk("Unable to find block group for %Lu\n", search_start
);
244 last
= cache
->key
.objectid
+ cache
->key
.offset
;
246 cache
= btrfs_lookup_block_group(root
->fs_info
, last
);
247 if (!cache
|| cache
->key
.objectid
>= total_fs_bytes
) {
256 if (cache_miss
&& !cache
->cached
) {
257 cache_block_group(root
, cache
);
259 cache
= btrfs_lookup_block_group(root
->fs_info
, last
);
261 cache
= btrfs_find_block_group(root
, cache
, last
, data
, 0);
269 static u64
div_factor(u64 num
, int factor
)
278 static int block_group_state_bits(u64 flags
)
281 if (flags
& BTRFS_BLOCK_GROUP_DATA
)
282 bits
|= BLOCK_GROUP_DATA
;
283 if (flags
& BTRFS_BLOCK_GROUP_METADATA
)
284 bits
|= BLOCK_GROUP_METADATA
;
285 if (flags
& BTRFS_BLOCK_GROUP_SYSTEM
)
286 bits
|= BLOCK_GROUP_SYSTEM
;
290 struct btrfs_block_group_cache
*btrfs_find_block_group(struct btrfs_root
*root
,
291 struct btrfs_block_group_cache
292 *hint
, u64 search_start
,
295 struct btrfs_block_group_cache
*cache
;
296 struct extent_io_tree
*block_group_cache
;
297 struct btrfs_block_group_cache
*found_group
= NULL
;
298 struct btrfs_fs_info
*info
= root
->fs_info
;
312 block_group_cache
= &info
->block_group_cache
;
313 total_fs_bytes
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
318 bit
= block_group_state_bits(data
);
320 if (search_start
&& search_start
< total_fs_bytes
) {
321 struct btrfs_block_group_cache
*shint
;
322 shint
= btrfs_lookup_block_group(info
, search_start
);
323 if (shint
&& block_group_bits(shint
, data
)) {
324 used
= btrfs_block_group_used(&shint
->item
);
325 if (used
+ shint
->pinned
<
326 div_factor(shint
->key
.offset
, factor
)) {
331 if (hint
&& block_group_bits(hint
, data
) &&
332 hint
->key
.objectid
< total_fs_bytes
) {
333 used
= btrfs_block_group_used(&hint
->item
);
334 if (used
+ hint
->pinned
<
335 div_factor(hint
->key
.offset
, factor
)) {
338 last
= hint
->key
.objectid
+ hint
->key
.offset
;
342 hint_last
= max(hint
->key
.objectid
, search_start
);
344 hint_last
= search_start
;
346 if (hint_last
>= total_fs_bytes
)
347 hint_last
= search_start
;
352 ret
= find_first_extent_bit(block_group_cache
, last
,
357 ret
= get_state_private(block_group_cache
, start
, &ptr
);
361 cache
= (struct btrfs_block_group_cache
*)(unsigned long)ptr
;
362 last
= cache
->key
.objectid
+ cache
->key
.offset
;
363 used
= btrfs_block_group_used(&cache
->item
);
365 if (cache
->key
.objectid
> total_fs_bytes
)
369 free_check
= cache
->key
.offset
;
371 free_check
= div_factor(cache
->key
.offset
, factor
);
373 if (used
+ cache
->pinned
< free_check
) {
388 static u64
hash_extent_ref(u64 root_objectid
, u64 ref_generation
,
389 u64 owner
, u64 owner_offset
)
391 u32 high_crc
= ~(u32
)0;
392 u32 low_crc
= ~(u32
)0;
395 lenum
= cpu_to_le64(root_objectid
);
396 high_crc
= crc32c(high_crc
, &lenum
, sizeof(lenum
));
397 lenum
= cpu_to_le64(ref_generation
);
398 low_crc
= crc32c(low_crc
, &lenum
, sizeof(lenum
));
399 if (owner
>= BTRFS_FIRST_FREE_OBJECTID
) {
400 lenum
= cpu_to_le64(owner
);
401 low_crc
= crc32c(low_crc
, &lenum
, sizeof(lenum
));
402 lenum
= cpu_to_le64(owner_offset
);
403 low_crc
= crc32c(low_crc
, &lenum
, sizeof(lenum
));
405 return ((u64
)high_crc
<< 32) | (u64
)low_crc
;
408 static int match_extent_ref(struct extent_buffer
*leaf
,
409 struct btrfs_extent_ref
*disk_ref
,
410 struct btrfs_extent_ref
*cpu_ref
)
415 if (cpu_ref
->objectid
)
416 len
= sizeof(*cpu_ref
);
418 len
= 2 * sizeof(u64
);
419 ret
= memcmp_extent_buffer(leaf
, cpu_ref
, (unsigned long)disk_ref
,
424 static int noinline
lookup_extent_backref(struct btrfs_trans_handle
*trans
,
425 struct btrfs_root
*root
,
426 struct btrfs_path
*path
, u64 bytenr
,
428 u64 ref_generation
, u64 owner
,
429 u64 owner_offset
, int del
)
432 struct btrfs_key key
;
433 struct btrfs_key found_key
;
434 struct btrfs_extent_ref ref
;
435 struct extent_buffer
*leaf
;
436 struct btrfs_extent_ref
*disk_ref
;
440 btrfs_set_stack_ref_root(&ref
, root_objectid
);
441 btrfs_set_stack_ref_generation(&ref
, ref_generation
);
442 btrfs_set_stack_ref_objectid(&ref
, owner
);
443 btrfs_set_stack_ref_offset(&ref
, owner_offset
);
445 hash
= hash_extent_ref(root_objectid
, ref_generation
, owner
,
448 key
.objectid
= bytenr
;
449 key
.type
= BTRFS_EXTENT_REF_KEY
;
452 ret
= btrfs_search_slot(trans
, root
, &key
, path
,
456 leaf
= path
->nodes
[0];
458 u32 nritems
= btrfs_header_nritems(leaf
);
459 if (path
->slots
[0] >= nritems
) {
460 ret2
= btrfs_next_leaf(root
, path
);
463 leaf
= path
->nodes
[0];
465 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
466 if (found_key
.objectid
!= bytenr
||
467 found_key
.type
!= BTRFS_EXTENT_REF_KEY
)
469 key
.offset
= found_key
.offset
;
471 btrfs_release_path(root
, path
);
475 disk_ref
= btrfs_item_ptr(path
->nodes
[0],
477 struct btrfs_extent_ref
);
478 if (match_extent_ref(path
->nodes
[0], disk_ref
, &ref
)) {
482 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
483 key
.offset
= found_key
.offset
+ 1;
484 btrfs_release_path(root
, path
);
491 * Back reference rules. Back refs have three main goals:
493 * 1) differentiate between all holders of references to an extent so that
494 * when a reference is dropped we can make sure it was a valid reference
495 * before freeing the extent.
497 * 2) Provide enough information to quickly find the holders of an extent
498 * if we notice a given block is corrupted or bad.
500 * 3) Make it easy to migrate blocks for FS shrinking or storage pool
501 * maintenance. This is actually the same as #2, but with a slightly
502 * different use case.
504 * File extents can be referenced by:
506 * - multiple snapshots, subvolumes, or different generations in one subvol
507 * - different files inside a single subvolume (in theory, not implemented yet)
508 * - different offsets inside a file (bookend extents in file.c)
510 * The extent ref structure has fields for:
512 * - Objectid of the subvolume root
513 * - Generation number of the tree holding the reference
514 * - objectid of the file holding the reference
515 * - offset in the file corresponding to the key holding the reference
517 * When a file extent is allocated the fields are filled in:
518 * (root_key.objectid, trans->transid, inode objectid, offset in file)
520 * When a leaf is cow'd new references are added for every file extent found
521 * in the leaf. It looks the same as the create case, but trans->transid
522 * will be different when the block is cow'd.
524 * (root_key.objectid, trans->transid, inode objectid, offset in file)
526 * When a file extent is removed either during snapshot deletion or file
527 * truncation, the corresponding back reference is found
530 * (btrfs_header_owner(leaf), btrfs_header_generation(leaf),
531 * inode objectid, offset in file)
533 * Btree extents can be referenced by:
535 * - Different subvolumes
536 * - Different generations of the same subvolume
538 * Storing sufficient information for a full reverse mapping of a btree
539 * block would require storing the lowest key of the block in the backref,
540 * and it would require updating that lowest key either before write out or
541 * every time it changed. Instead, the objectid of the lowest key is stored
542 * along with the level of the tree block. This provides a hint
543 * about where in the btree the block can be found. Searches through the
544 * btree only need to look for a pointer to that block, so they stop one
545 * level higher than the level recorded in the backref.
547 * Some btrees do not do reference counting on their extents. These
548 * include the extent tree and the tree of tree roots. Backrefs for these
549 * trees always have a generation of zero.
551 * When a tree block is created, back references are inserted:
553 * (root->root_key.objectid, trans->transid or zero, level, lowest_key_objectid)
555 * When a tree block is cow'd in a reference counted root,
556 * new back references are added for all the blocks it points to.
557 * These are of the form (trans->transid will have increased since creation):
559 * (root->root_key.objectid, trans->transid, level, lowest_key_objectid)
561 * Because the lowest_key_objectid and the level are just hints
562 * they are not used when backrefs are deleted. When a backref is deleted:
564 * if backref was for a tree root:
565 * root_objectid = root->root_key.objectid
567 * root_objectid = btrfs_header_owner(parent)
569 * (root_objectid, btrfs_header_generation(parent) or zero, 0, 0)
571 * Back Reference Key hashing:
573 * Back references have four fields, each 64 bits long. Unfortunately,
574 * This is hashed into a single 64 bit number and placed into the key offset.
575 * The key objectid corresponds to the first byte in the extent, and the
576 * key type is set to BTRFS_EXTENT_REF_KEY
578 int btrfs_insert_extent_backref(struct btrfs_trans_handle
*trans
,
579 struct btrfs_root
*root
,
580 struct btrfs_path
*path
, u64 bytenr
,
581 u64 root_objectid
, u64 ref_generation
,
582 u64 owner
, u64 owner_offset
)
585 struct btrfs_key key
;
586 struct btrfs_extent_ref ref
;
587 struct btrfs_extent_ref
*disk_ref
;
590 btrfs_set_stack_ref_root(&ref
, root_objectid
);
591 btrfs_set_stack_ref_generation(&ref
, ref_generation
);
592 btrfs_set_stack_ref_objectid(&ref
, owner
);
593 btrfs_set_stack_ref_offset(&ref
, owner_offset
);
595 hash
= hash_extent_ref(root_objectid
, ref_generation
, owner
,
598 key
.objectid
= bytenr
;
599 key
.type
= BTRFS_EXTENT_REF_KEY
;
601 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
, sizeof(ref
));
602 while (ret
== -EEXIST
) {
603 disk_ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
604 struct btrfs_extent_ref
);
605 if (match_extent_ref(path
->nodes
[0], disk_ref
, &ref
))
608 btrfs_release_path(root
, path
);
609 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
614 disk_ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
615 struct btrfs_extent_ref
);
616 write_extent_buffer(path
->nodes
[0], &ref
, (unsigned long)disk_ref
,
618 btrfs_mark_buffer_dirty(path
->nodes
[0]);
620 btrfs_release_path(root
, path
);
624 int btrfs_inc_extent_ref(struct btrfs_trans_handle
*trans
,
625 struct btrfs_root
*root
,
626 u64 bytenr
, u64 num_bytes
,
627 u64 root_objectid
, u64 ref_generation
,
628 u64 owner
, u64 owner_offset
)
630 struct btrfs_path
*path
;
632 struct btrfs_key key
;
633 struct extent_buffer
*l
;
634 struct btrfs_extent_item
*item
;
637 WARN_ON(num_bytes
< root
->sectorsize
);
638 path
= btrfs_alloc_path();
643 key
.objectid
= bytenr
;
644 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
645 key
.offset
= num_bytes
;
646 ret
= btrfs_search_slot(trans
, root
->fs_info
->extent_root
, &key
, path
,
655 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
656 refs
= btrfs_extent_refs(l
, item
);
657 btrfs_set_extent_refs(l
, item
, refs
+ 1);
658 btrfs_mark_buffer_dirty(path
->nodes
[0]);
660 btrfs_release_path(root
->fs_info
->extent_root
, path
);
663 ret
= btrfs_insert_extent_backref(trans
, root
->fs_info
->extent_root
,
664 path
, bytenr
, root_objectid
,
665 ref_generation
, owner
, owner_offset
);
667 finish_current_insert(trans
, root
->fs_info
->extent_root
);
668 del_pending_extents(trans
, root
->fs_info
->extent_root
);
670 btrfs_free_path(path
);
674 int btrfs_extent_post_op(struct btrfs_trans_handle
*trans
,
675 struct btrfs_root
*root
)
677 finish_current_insert(trans
, root
->fs_info
->extent_root
);
678 del_pending_extents(trans
, root
->fs_info
->extent_root
);
682 static int lookup_extent_ref(struct btrfs_trans_handle
*trans
,
683 struct btrfs_root
*root
, u64 bytenr
,
684 u64 num_bytes
, u32
*refs
)
686 struct btrfs_path
*path
;
688 struct btrfs_key key
;
689 struct extent_buffer
*l
;
690 struct btrfs_extent_item
*item
;
692 WARN_ON(num_bytes
< root
->sectorsize
);
693 path
= btrfs_alloc_path();
695 key
.objectid
= bytenr
;
696 key
.offset
= num_bytes
;
697 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
698 ret
= btrfs_search_slot(trans
, root
->fs_info
->extent_root
, &key
, path
,
703 btrfs_print_leaf(root
, path
->nodes
[0]);
704 printk("failed to find block number %Lu\n", bytenr
);
708 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
709 *refs
= btrfs_extent_refs(l
, item
);
711 btrfs_free_path(path
);
715 u32
btrfs_count_snapshots_in_path(struct btrfs_root
*root
,
716 struct btrfs_path
*count_path
,
719 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
720 struct btrfs_path
*path
;
723 u64 root_objectid
= root
->root_key
.objectid
;
728 struct btrfs_key key
;
729 struct btrfs_key found_key
;
730 struct extent_buffer
*l
;
731 struct btrfs_extent_item
*item
;
732 struct btrfs_extent_ref
*ref_item
;
735 path
= btrfs_alloc_path();
738 bytenr
= first_extent
;
740 bytenr
= count_path
->nodes
[level
]->start
;
743 key
.objectid
= bytenr
;
746 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
747 ret
= btrfs_search_slot(NULL
, extent_root
, &key
, path
, 0, 0);
753 btrfs_item_key_to_cpu(l
, &found_key
, path
->slots
[0]);
755 if (found_key
.objectid
!= bytenr
||
756 found_key
.type
!= BTRFS_EXTENT_ITEM_KEY
) {
760 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
763 nritems
= btrfs_header_nritems(l
);
764 if (path
->slots
[0] >= nritems
) {
765 ret
= btrfs_next_leaf(extent_root
, path
);
770 btrfs_item_key_to_cpu(l
, &found_key
, path
->slots
[0]);
771 if (found_key
.objectid
!= bytenr
)
774 if (found_key
.type
!= BTRFS_EXTENT_REF_KEY
) {
780 ref_item
= btrfs_item_ptr(l
, path
->slots
[0],
781 struct btrfs_extent_ref
);
782 found_objectid
= btrfs_ref_root(l
, ref_item
);
784 if (found_objectid
!= root_objectid
) {
791 if (cur_count
== 0) {
795 if (level
>= 0 && root
->node
== count_path
->nodes
[level
])
798 btrfs_release_path(root
, path
);
802 btrfs_free_path(path
);
805 int btrfs_inc_root_ref(struct btrfs_trans_handle
*trans
,
806 struct btrfs_root
*root
, u64 owner_objectid
)
812 struct btrfs_disk_key disk_key
;
814 level
= btrfs_header_level(root
->node
);
815 generation
= trans
->transid
;
816 nritems
= btrfs_header_nritems(root
->node
);
819 btrfs_item_key(root
->node
, &disk_key
, 0);
821 btrfs_node_key(root
->node
, &disk_key
, 0);
822 key_objectid
= btrfs_disk_key_objectid(&disk_key
);
826 return btrfs_inc_extent_ref(trans
, root
, root
->node
->start
,
827 root
->node
->len
, owner_objectid
,
828 generation
, level
, key_objectid
);
831 int btrfs_inc_ref(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
832 struct extent_buffer
*buf
)
836 struct btrfs_key key
;
837 struct btrfs_file_extent_item
*fi
;
846 level
= btrfs_header_level(buf
);
847 nritems
= btrfs_header_nritems(buf
);
848 for (i
= 0; i
< nritems
; i
++) {
851 btrfs_item_key_to_cpu(buf
, &key
, i
);
852 if (btrfs_key_type(&key
) != BTRFS_EXTENT_DATA_KEY
)
854 fi
= btrfs_item_ptr(buf
, i
,
855 struct btrfs_file_extent_item
);
856 if (btrfs_file_extent_type(buf
, fi
) ==
857 BTRFS_FILE_EXTENT_INLINE
)
859 disk_bytenr
= btrfs_file_extent_disk_bytenr(buf
, fi
);
860 if (disk_bytenr
== 0)
862 ret
= btrfs_inc_extent_ref(trans
, root
, disk_bytenr
,
863 btrfs_file_extent_disk_num_bytes(buf
, fi
),
864 root
->root_key
.objectid
, trans
->transid
,
865 key
.objectid
, key
.offset
);
871 bytenr
= btrfs_node_blockptr(buf
, i
);
872 btrfs_node_key_to_cpu(buf
, &key
, i
);
873 ret
= btrfs_inc_extent_ref(trans
, root
, bytenr
,
874 btrfs_level_size(root
, level
- 1),
875 root
->root_key
.objectid
,
877 level
- 1, key
.objectid
);
888 for (i
=0; i
< faili
; i
++) {
891 btrfs_item_key_to_cpu(buf
, &key
, i
);
892 if (btrfs_key_type(&key
) != BTRFS_EXTENT_DATA_KEY
)
894 fi
= btrfs_item_ptr(buf
, i
,
895 struct btrfs_file_extent_item
);
896 if (btrfs_file_extent_type(buf
, fi
) ==
897 BTRFS_FILE_EXTENT_INLINE
)
899 disk_bytenr
= btrfs_file_extent_disk_bytenr(buf
, fi
);
900 if (disk_bytenr
== 0)
902 err
= btrfs_free_extent(trans
, root
, disk_bytenr
,
903 btrfs_file_extent_disk_num_bytes(buf
,
907 bytenr
= btrfs_node_blockptr(buf
, i
);
908 err
= btrfs_free_extent(trans
, root
, bytenr
,
909 btrfs_level_size(root
, level
- 1), 0);
917 static int write_one_cache_group(struct btrfs_trans_handle
*trans
,
918 struct btrfs_root
*root
,
919 struct btrfs_path
*path
,
920 struct btrfs_block_group_cache
*cache
)
924 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
926 struct extent_buffer
*leaf
;
928 ret
= btrfs_search_slot(trans
, extent_root
, &cache
->key
, path
, 0, 1);
933 leaf
= path
->nodes
[0];
934 bi
= btrfs_item_ptr_offset(leaf
, path
->slots
[0]);
935 write_extent_buffer(leaf
, &cache
->item
, bi
, sizeof(cache
->item
));
936 btrfs_mark_buffer_dirty(leaf
);
937 btrfs_release_path(extent_root
, path
);
939 finish_current_insert(trans
, extent_root
);
940 pending_ret
= del_pending_extents(trans
, extent_root
);
949 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle
*trans
,
950 struct btrfs_root
*root
)
952 struct extent_io_tree
*block_group_cache
;
953 struct btrfs_block_group_cache
*cache
;
957 struct btrfs_path
*path
;
963 block_group_cache
= &root
->fs_info
->block_group_cache
;
964 path
= btrfs_alloc_path();
969 ret
= find_first_extent_bit(block_group_cache
, last
,
970 &start
, &end
, BLOCK_GROUP_DIRTY
);
975 ret
= get_state_private(block_group_cache
, start
, &ptr
);
979 cache
= (struct btrfs_block_group_cache
*)(unsigned long)ptr
;
980 err
= write_one_cache_group(trans
, root
,
983 * if we fail to write the cache group, we want
984 * to keep it marked dirty in hopes that a later
991 clear_extent_bits(block_group_cache
, start
, end
,
992 BLOCK_GROUP_DIRTY
, GFP_NOFS
);
994 btrfs_free_path(path
);
998 static struct btrfs_space_info
*__find_space_info(struct btrfs_fs_info
*info
,
1001 struct list_head
*head
= &info
->space_info
;
1002 struct list_head
*cur
;
1003 struct btrfs_space_info
*found
;
1004 list_for_each(cur
, head
) {
1005 found
= list_entry(cur
, struct btrfs_space_info
, list
);
1006 if (found
->flags
== flags
)
1013 static int update_space_info(struct btrfs_fs_info
*info
, u64 flags
,
1014 u64 total_bytes
, u64 bytes_used
,
1015 struct btrfs_space_info
**space_info
)
1017 struct btrfs_space_info
*found
;
1019 found
= __find_space_info(info
, flags
);
1021 found
->total_bytes
+= total_bytes
;
1022 found
->bytes_used
+= bytes_used
;
1023 WARN_ON(found
->total_bytes
< found
->bytes_used
);
1024 *space_info
= found
;
1027 found
= kmalloc(sizeof(*found
), GFP_NOFS
);
1031 list_add(&found
->list
, &info
->space_info
);
1032 found
->flags
= flags
;
1033 found
->total_bytes
= total_bytes
;
1034 found
->bytes_used
= bytes_used
;
1035 found
->bytes_pinned
= 0;
1037 *space_info
= found
;
1042 static int do_chunk_alloc(struct btrfs_trans_handle
*trans
,
1043 struct btrfs_root
*extent_root
, u64 alloc_bytes
,
1046 struct btrfs_space_info
*space_info
;
1052 space_info
= __find_space_info(extent_root
->fs_info
, flags
);
1054 ret
= update_space_info(extent_root
->fs_info
, flags
,
1058 BUG_ON(!space_info
);
1060 if (space_info
->full
)
1063 thresh
= div_factor(space_info
->total_bytes
, 7);
1064 if ((space_info
->bytes_used
+ space_info
->bytes_pinned
+ alloc_bytes
) <
1068 ret
= btrfs_alloc_chunk(trans
, extent_root
, &start
, &num_bytes
, flags
);
1069 if (ret
== -ENOSPC
) {
1070 printk("space info full %Lu\n", flags
);
1071 space_info
->full
= 1;
1077 ret
= btrfs_make_block_group(trans
, extent_root
, 0, flags
,
1078 extent_root
->fs_info
->chunk_root
->root_key
.objectid
,
1082 if (flags
& BTRFS_BLOCK_GROUP_RAID0
) {
1083 if (flags
& BTRFS_BLOCK_GROUP_DATA
) {
1084 extent_root
->fs_info
->extra_data_alloc_bits
=
1085 BTRFS_BLOCK_GROUP_RAID0
;
1087 if (flags
& BTRFS_BLOCK_GROUP_METADATA
) {
1088 extent_root
->fs_info
->extra_alloc_bits
=
1089 BTRFS_BLOCK_GROUP_RAID0
;
1095 static int update_block_group(struct btrfs_trans_handle
*trans
,
1096 struct btrfs_root
*root
,
1097 u64 bytenr
, u64 num_bytes
, int alloc
,
1100 struct btrfs_block_group_cache
*cache
;
1101 struct btrfs_fs_info
*info
= root
->fs_info
;
1102 u64 total
= num_bytes
;
1109 cache
= btrfs_lookup_block_group(info
, bytenr
);
1113 byte_in_group
= bytenr
- cache
->key
.objectid
;
1114 WARN_ON(byte_in_group
> cache
->key
.offset
);
1115 start
= cache
->key
.objectid
;
1116 end
= start
+ cache
->key
.offset
- 1;
1117 set_extent_bits(&info
->block_group_cache
, start
, end
,
1118 BLOCK_GROUP_DIRTY
, GFP_NOFS
);
1120 old_val
= btrfs_block_group_used(&cache
->item
);
1121 num_bytes
= min(total
, cache
->key
.offset
- byte_in_group
);
1123 old_val
+= num_bytes
;
1124 cache
->space_info
->bytes_used
+= num_bytes
;
1126 old_val
-= num_bytes
;
1127 cache
->space_info
->bytes_used
-= num_bytes
;
1129 set_extent_dirty(&info
->free_space_cache
,
1130 bytenr
, bytenr
+ num_bytes
- 1,
1134 btrfs_set_block_group_used(&cache
->item
, old_val
);
1136 bytenr
+= num_bytes
;
1141 static int update_pinned_extents(struct btrfs_root
*root
,
1142 u64 bytenr
, u64 num
, int pin
)
1145 struct btrfs_block_group_cache
*cache
;
1146 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1149 set_extent_dirty(&fs_info
->pinned_extents
,
1150 bytenr
, bytenr
+ num
- 1, GFP_NOFS
);
1152 clear_extent_dirty(&fs_info
->pinned_extents
,
1153 bytenr
, bytenr
+ num
- 1, GFP_NOFS
);
1156 cache
= btrfs_lookup_block_group(fs_info
, bytenr
);
1158 len
= min(num
, cache
->key
.offset
-
1159 (bytenr
- cache
->key
.objectid
));
1161 cache
->pinned
+= len
;
1162 cache
->space_info
->bytes_pinned
+= len
;
1163 fs_info
->total_pinned
+= len
;
1165 cache
->pinned
-= len
;
1166 cache
->space_info
->bytes_pinned
-= len
;
1167 fs_info
->total_pinned
-= len
;
1175 int btrfs_copy_pinned(struct btrfs_root
*root
, struct extent_io_tree
*copy
)
1180 struct extent_io_tree
*pinned_extents
= &root
->fs_info
->pinned_extents
;
1184 ret
= find_first_extent_bit(pinned_extents
, last
,
1185 &start
, &end
, EXTENT_DIRTY
);
1188 set_extent_dirty(copy
, start
, end
, GFP_NOFS
);
1194 int btrfs_finish_extent_commit(struct btrfs_trans_handle
*trans
,
1195 struct btrfs_root
*root
,
1196 struct extent_io_tree
*unpin
)
1201 struct extent_io_tree
*free_space_cache
;
1202 free_space_cache
= &root
->fs_info
->free_space_cache
;
1205 ret
= find_first_extent_bit(unpin
, 0, &start
, &end
,
1209 update_pinned_extents(root
, start
, end
+ 1 - start
, 0);
1210 clear_extent_dirty(unpin
, start
, end
, GFP_NOFS
);
1211 set_extent_dirty(free_space_cache
, start
, end
, GFP_NOFS
);
1216 static int finish_current_insert(struct btrfs_trans_handle
*trans
,
1217 struct btrfs_root
*extent_root
)
1221 struct btrfs_fs_info
*info
= extent_root
->fs_info
;
1222 struct extent_buffer
*eb
;
1223 struct btrfs_path
*path
;
1224 struct btrfs_key ins
;
1225 struct btrfs_disk_key first
;
1226 struct btrfs_extent_item extent_item
;
1231 btrfs_set_stack_extent_refs(&extent_item
, 1);
1232 btrfs_set_key_type(&ins
, BTRFS_EXTENT_ITEM_KEY
);
1233 path
= btrfs_alloc_path();
1236 ret
= find_first_extent_bit(&info
->extent_ins
, 0, &start
,
1237 &end
, EXTENT_LOCKED
);
1241 ins
.objectid
= start
;
1242 ins
.offset
= end
+ 1 - start
;
1243 err
= btrfs_insert_item(trans
, extent_root
, &ins
,
1244 &extent_item
, sizeof(extent_item
));
1245 clear_extent_bits(&info
->extent_ins
, start
, end
, EXTENT_LOCKED
,
1247 eb
= read_tree_block(extent_root
, ins
.objectid
, ins
.offset
);
1248 level
= btrfs_header_level(eb
);
1250 btrfs_item_key(eb
, &first
, 0);
1252 btrfs_node_key(eb
, &first
, 0);
1254 err
= btrfs_insert_extent_backref(trans
, extent_root
, path
,
1255 start
, extent_root
->root_key
.objectid
,
1257 btrfs_disk_key_objectid(&first
));
1259 free_extent_buffer(eb
);
1261 btrfs_free_path(path
);
1265 static int pin_down_bytes(struct btrfs_root
*root
, u64 bytenr
, u32 num_bytes
,
1269 struct extent_buffer
*buf
;
1272 buf
= btrfs_find_tree_block(root
, bytenr
, num_bytes
);
1274 if (btrfs_buffer_uptodate(buf
)) {
1276 root
->fs_info
->running_transaction
->transid
;
1277 u64 header_transid
=
1278 btrfs_header_generation(buf
);
1279 if (header_transid
== transid
) {
1280 clean_tree_block(NULL
, root
, buf
);
1281 free_extent_buffer(buf
);
1285 free_extent_buffer(buf
);
1287 update_pinned_extents(root
, bytenr
, num_bytes
, 1);
1289 set_extent_bits(&root
->fs_info
->pending_del
,
1290 bytenr
, bytenr
+ num_bytes
- 1,
1291 EXTENT_LOCKED
, GFP_NOFS
);
1298 * remove an extent from the root, returns 0 on success
1300 static int __free_extent(struct btrfs_trans_handle
*trans
, struct btrfs_root
1301 *root
, u64 bytenr
, u64 num_bytes
,
1302 u64 root_objectid
, u64 ref_generation
,
1303 u64 owner_objectid
, u64 owner_offset
, int pin
,
1306 struct btrfs_path
*path
;
1307 struct btrfs_key key
;
1308 struct btrfs_fs_info
*info
= root
->fs_info
;
1309 struct btrfs_root
*extent_root
= info
->extent_root
;
1310 struct extent_buffer
*leaf
;
1312 int extent_slot
= 0;
1313 int found_extent
= 0;
1315 struct btrfs_extent_item
*ei
;
1318 key
.objectid
= bytenr
;
1319 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
1320 key
.offset
= num_bytes
;
1321 path
= btrfs_alloc_path();
1326 ret
= lookup_extent_backref(trans
, extent_root
, path
,
1327 bytenr
, root_objectid
,
1329 owner_objectid
, owner_offset
, 1);
1331 struct btrfs_key found_key
;
1332 extent_slot
= path
->slots
[0];
1333 while(extent_slot
> 0) {
1335 btrfs_item_key_to_cpu(path
->nodes
[0], &found_key
,
1337 if (found_key
.objectid
!= bytenr
)
1339 if (found_key
.type
== BTRFS_EXTENT_ITEM_KEY
&&
1340 found_key
.offset
== num_bytes
) {
1344 if (path
->slots
[0] - extent_slot
> 5)
1348 ret
= btrfs_del_item(trans
, extent_root
, path
);
1350 btrfs_print_leaf(extent_root
, path
->nodes
[0]);
1352 printk("Unable to find ref byte nr %Lu root %Lu "
1353 " gen %Lu owner %Lu offset %Lu\n", bytenr
,
1354 root_objectid
, ref_generation
, owner_objectid
,
1357 if (!found_extent
) {
1358 btrfs_release_path(extent_root
, path
);
1359 ret
= btrfs_search_slot(trans
, extent_root
, &key
, path
, -1, 1);
1363 extent_slot
= path
->slots
[0];
1366 leaf
= path
->nodes
[0];
1367 ei
= btrfs_item_ptr(leaf
, extent_slot
,
1368 struct btrfs_extent_item
);
1369 refs
= btrfs_extent_refs(leaf
, ei
);
1372 btrfs_set_extent_refs(leaf
, ei
, refs
);
1374 btrfs_mark_buffer_dirty(leaf
);
1376 if (refs
== 0 && found_extent
&& path
->slots
[0] == extent_slot
+ 1) {
1377 /* if the back ref and the extent are next to each other
1378 * they get deleted below in one shot
1380 path
->slots
[0] = extent_slot
;
1382 } else if (found_extent
) {
1383 /* otherwise delete the extent back ref */
1384 ret
= btrfs_del_item(trans
, extent_root
, path
);
1386 /* if refs are 0, we need to setup the path for deletion */
1388 btrfs_release_path(extent_root
, path
);
1389 ret
= btrfs_search_slot(trans
, extent_root
, &key
, path
,
1402 ret
= pin_down_bytes(root
, bytenr
, num_bytes
, 0);
1408 /* block accounting for super block */
1409 super_used
= btrfs_super_bytes_used(&info
->super_copy
);
1410 btrfs_set_super_bytes_used(&info
->super_copy
,
1411 super_used
- num_bytes
);
1413 /* block accounting for root item */
1414 root_used
= btrfs_root_used(&root
->root_item
);
1415 btrfs_set_root_used(&root
->root_item
,
1416 root_used
- num_bytes
);
1417 ret
= btrfs_del_items(trans
, extent_root
, path
, path
->slots
[0],
1422 ret
= update_block_group(trans
, root
, bytenr
, num_bytes
, 0,
1426 btrfs_free_path(path
);
1427 finish_current_insert(trans
, extent_root
);
1432 * find all the blocks marked as pending in the radix tree and remove
1433 * them from the extent map
1435 static int del_pending_extents(struct btrfs_trans_handle
*trans
, struct
1436 btrfs_root
*extent_root
)
1442 struct extent_io_tree
*pending_del
;
1443 struct extent_io_tree
*pinned_extents
;
1445 pending_del
= &extent_root
->fs_info
->pending_del
;
1446 pinned_extents
= &extent_root
->fs_info
->pinned_extents
;
1449 ret
= find_first_extent_bit(pending_del
, 0, &start
, &end
,
1453 update_pinned_extents(extent_root
, start
, end
+ 1 - start
, 1);
1454 clear_extent_bits(pending_del
, start
, end
, EXTENT_LOCKED
,
1456 ret
= __free_extent(trans
, extent_root
,
1457 start
, end
+ 1 - start
,
1458 extent_root
->root_key
.objectid
,
1467 * remove an extent from the root, returns 0 on success
1469 int btrfs_free_extent(struct btrfs_trans_handle
*trans
, struct btrfs_root
1470 *root
, u64 bytenr
, u64 num_bytes
,
1471 u64 root_objectid
, u64 ref_generation
,
1472 u64 owner_objectid
, u64 owner_offset
, int pin
)
1474 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
1478 WARN_ON(num_bytes
< root
->sectorsize
);
1479 if (!root
->ref_cows
)
1482 if (root
== extent_root
) {
1483 pin_down_bytes(root
, bytenr
, num_bytes
, 1);
1486 ret
= __free_extent(trans
, root
, bytenr
, num_bytes
, root_objectid
,
1487 ref_generation
, owner_objectid
, owner_offset
,
1489 pending_ret
= del_pending_extents(trans
, root
->fs_info
->extent_root
);
1490 return ret
? ret
: pending_ret
;
1493 static u64
stripe_align(struct btrfs_root
*root
, u64 val
)
1495 u64 mask
= ((u64
)root
->stripesize
- 1);
1496 u64 ret
= (val
+ mask
) & ~mask
;
1501 * walks the btree of allocated extents and find a hole of a given size.
1502 * The key ins is changed to record the hole:
1503 * ins->objectid == block start
1504 * ins->flags = BTRFS_EXTENT_ITEM_KEY
1505 * ins->offset == number of blocks
1506 * Any available blocks before search_start are skipped.
1508 static int noinline
find_free_extent(struct btrfs_trans_handle
*trans
,
1509 struct btrfs_root
*orig_root
,
1510 u64 num_bytes
, u64 empty_size
,
1511 u64 search_start
, u64 search_end
,
1512 u64 hint_byte
, struct btrfs_key
*ins
,
1513 u64 exclude_start
, u64 exclude_nr
,
1517 u64 orig_search_start
= search_start
;
1518 struct btrfs_root
* root
= orig_root
->fs_info
->extent_root
;
1519 struct btrfs_fs_info
*info
= root
->fs_info
;
1520 u64 total_needed
= num_bytes
;
1521 u64
*last_ptr
= NULL
;
1522 struct btrfs_block_group_cache
*block_group
;
1525 int empty_cluster
= 2 * 1024 * 1024;
1527 WARN_ON(num_bytes
< root
->sectorsize
);
1528 btrfs_set_key_type(ins
, BTRFS_EXTENT_ITEM_KEY
);
1530 if (data
& BTRFS_BLOCK_GROUP_METADATA
) {
1531 last_ptr
= &root
->fs_info
->last_alloc
;
1534 if ((data
& BTRFS_BLOCK_GROUP_DATA
) && btrfs_test_opt(root
, SSD
)) {
1535 last_ptr
= &root
->fs_info
->last_data_alloc
;
1540 hint_byte
= *last_ptr
;
1542 empty_size
+= empty_cluster
;
1546 if (search_end
== (u64
)-1)
1547 search_end
= btrfs_super_total_bytes(&info
->super_copy
);
1550 block_group
= btrfs_lookup_block_group(info
, hint_byte
);
1552 hint_byte
= search_start
;
1553 block_group
= btrfs_find_block_group(root
, block_group
,
1554 hint_byte
, data
, 1);
1555 if (last_ptr
&& *last_ptr
== 0 && block_group
)
1556 hint_byte
= block_group
->key
.objectid
;
1558 block_group
= btrfs_find_block_group(root
,
1560 search_start
, data
, 1);
1562 search_start
= max(search_start
, hint_byte
);
1564 total_needed
+= empty_size
;
1568 block_group
= btrfs_lookup_block_group(info
, search_start
);
1570 block_group
= btrfs_lookup_block_group(info
,
1573 ret
= find_search_start(root
, &block_group
, &search_start
,
1574 total_needed
, data
);
1575 if (ret
== -ENOSPC
&& last_ptr
&& *last_ptr
) {
1577 block_group
= btrfs_lookup_block_group(info
,
1579 search_start
= orig_search_start
;
1580 ret
= find_search_start(root
, &block_group
, &search_start
,
1581 total_needed
, data
);
1588 if (last_ptr
&& *last_ptr
&& search_start
!= *last_ptr
) {
1591 empty_size
+= empty_cluster
;
1592 total_needed
+= empty_size
;
1594 block_group
= btrfs_lookup_block_group(info
,
1596 search_start
= orig_search_start
;
1597 ret
= find_search_start(root
, &block_group
,
1598 &search_start
, total_needed
, data
);
1605 search_start
= stripe_align(root
, search_start
);
1606 ins
->objectid
= search_start
;
1607 ins
->offset
= num_bytes
;
1609 if (ins
->objectid
+ num_bytes
>= search_end
)
1612 if (ins
->objectid
+ num_bytes
>
1613 block_group
->key
.objectid
+ block_group
->key
.offset
) {
1614 search_start
= block_group
->key
.objectid
+
1615 block_group
->key
.offset
;
1619 if (test_range_bit(&info
->extent_ins
, ins
->objectid
,
1620 ins
->objectid
+ num_bytes
-1, EXTENT_LOCKED
, 0)) {
1621 search_start
= ins
->objectid
+ num_bytes
;
1625 if (test_range_bit(&info
->pinned_extents
, ins
->objectid
,
1626 ins
->objectid
+ num_bytes
-1, EXTENT_DIRTY
, 0)) {
1627 search_start
= ins
->objectid
+ num_bytes
;
1631 if (exclude_nr
> 0 && (ins
->objectid
+ num_bytes
> exclude_start
&&
1632 ins
->objectid
< exclude_start
+ exclude_nr
)) {
1633 search_start
= exclude_start
+ exclude_nr
;
1637 if (!(data
& BTRFS_BLOCK_GROUP_DATA
)) {
1638 block_group
= btrfs_lookup_block_group(info
, ins
->objectid
);
1640 trans
->block_group
= block_group
;
1642 ins
->offset
= num_bytes
;
1644 *last_ptr
= ins
->objectid
+ ins
->offset
;
1646 btrfs_super_total_bytes(&root
->fs_info
->super_copy
)) {
1653 if (search_start
+ num_bytes
>= search_end
) {
1655 search_start
= orig_search_start
;
1662 total_needed
-= empty_size
;
1667 block_group
= btrfs_lookup_block_group(info
, search_start
);
1669 block_group
= btrfs_find_block_group(root
, block_group
,
1670 search_start
, data
, 0);
1677 * finds a free extent and does all the dirty work required for allocation
1678 * returns the key for the extent through ins, and a tree buffer for
1679 * the first block of the extent through buf.
1681 * returns 0 if everything worked, non-zero otherwise.
1683 int btrfs_alloc_extent(struct btrfs_trans_handle
*trans
,
1684 struct btrfs_root
*root
,
1685 u64 num_bytes
, u64 root_objectid
, u64 ref_generation
,
1686 u64 owner
, u64 owner_offset
,
1687 u64 empty_size
, u64 hint_byte
,
1688 u64 search_end
, struct btrfs_key
*ins
, int data
)
1694 u64 search_start
= 0;
1697 struct btrfs_fs_info
*info
= root
->fs_info
;
1698 struct btrfs_root
*extent_root
= info
->extent_root
;
1699 struct btrfs_extent_item
*extent_item
;
1700 struct btrfs_extent_ref
*ref
;
1701 struct btrfs_path
*path
;
1702 struct btrfs_key keys
[2];
1703 int extra_chunk_alloc_bits
= 0;
1706 data
= BTRFS_BLOCK_GROUP_DATA
| info
->extra_data_alloc_bits
;
1707 } else if (root
== root
->fs_info
->chunk_root
) {
1708 data
= BTRFS_BLOCK_GROUP_SYSTEM
;
1710 data
= BTRFS_BLOCK_GROUP_METADATA
| info
->extra_alloc_bits
;
1712 if (btrfs_super_num_devices(&info
->super_copy
) > 1 &&
1713 !(data
& BTRFS_BLOCK_GROUP_SYSTEM
))
1714 extra_chunk_alloc_bits
= BTRFS_BLOCK_GROUP_RAID0
;
1716 if (root
->ref_cows
) {
1717 if (!(data
& BTRFS_BLOCK_GROUP_METADATA
)) {
1718 ret
= do_chunk_alloc(trans
, root
->fs_info
->extent_root
,
1720 BTRFS_BLOCK_GROUP_METADATA
|
1721 info
->extra_alloc_bits
|
1722 extra_chunk_alloc_bits
);
1725 ret
= do_chunk_alloc(trans
, root
->fs_info
->extent_root
,
1726 num_bytes
+ 2 * 1024 * 1024, data
|
1727 extra_chunk_alloc_bits
);
1731 new_hint
= max(hint_byte
, root
->fs_info
->alloc_start
);
1732 if (new_hint
< btrfs_super_total_bytes(&info
->super_copy
))
1733 hint_byte
= new_hint
;
1735 WARN_ON(num_bytes
< root
->sectorsize
);
1736 ret
= find_free_extent(trans
, root
, num_bytes
, empty_size
,
1737 search_start
, search_end
, hint_byte
, ins
,
1738 trans
->alloc_exclude_start
,
1739 trans
->alloc_exclude_nr
, data
);
1744 /* block accounting for super block */
1745 super_used
= btrfs_super_bytes_used(&info
->super_copy
);
1746 btrfs_set_super_bytes_used(&info
->super_copy
, super_used
+ num_bytes
);
1748 /* block accounting for root item */
1749 root_used
= btrfs_root_used(&root
->root_item
);
1750 btrfs_set_root_used(&root
->root_item
, root_used
+ num_bytes
);
1752 clear_extent_dirty(&root
->fs_info
->free_space_cache
,
1753 ins
->objectid
, ins
->objectid
+ ins
->offset
- 1,
1756 if (root
== extent_root
) {
1757 set_extent_bits(&root
->fs_info
->extent_ins
, ins
->objectid
,
1758 ins
->objectid
+ ins
->offset
- 1,
1759 EXTENT_LOCKED
, GFP_NOFS
);
1763 WARN_ON(trans
->alloc_exclude_nr
);
1764 trans
->alloc_exclude_start
= ins
->objectid
;
1765 trans
->alloc_exclude_nr
= ins
->offset
;
1767 memcpy(&keys
[0], ins
, sizeof(*ins
));
1768 keys
[1].offset
= hash_extent_ref(root_objectid
, ref_generation
,
1769 owner
, owner_offset
);
1770 keys
[1].objectid
= ins
->objectid
;
1771 keys
[1].type
= BTRFS_EXTENT_REF_KEY
;
1772 sizes
[0] = sizeof(*extent_item
);
1773 sizes
[1] = sizeof(*ref
);
1775 path
= btrfs_alloc_path();
1778 ret
= btrfs_insert_empty_items(trans
, extent_root
, path
, keys
,
1782 extent_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
1783 struct btrfs_extent_item
);
1784 btrfs_set_extent_refs(path
->nodes
[0], extent_item
, 1);
1785 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0] + 1,
1786 struct btrfs_extent_ref
);
1788 btrfs_set_ref_root(path
->nodes
[0], ref
, root_objectid
);
1789 btrfs_set_ref_generation(path
->nodes
[0], ref
, ref_generation
);
1790 btrfs_set_ref_objectid(path
->nodes
[0], ref
, owner
);
1791 btrfs_set_ref_offset(path
->nodes
[0], ref
, owner_offset
);
1793 btrfs_mark_buffer_dirty(path
->nodes
[0]);
1795 trans
->alloc_exclude_start
= 0;
1796 trans
->alloc_exclude_nr
= 0;
1797 btrfs_free_path(path
);
1798 finish_current_insert(trans
, extent_root
);
1799 pending_ret
= del_pending_extents(trans
, extent_root
);
1809 ret
= update_block_group(trans
, root
, ins
->objectid
, ins
->offset
, 1, 0);
1811 printk("update block group failed for %Lu %Lu\n",
1812 ins
->objectid
, ins
->offset
);
1819 * helper function to allocate a block for a given tree
1820 * returns the tree buffer or NULL.
1822 struct extent_buffer
*btrfs_alloc_free_block(struct btrfs_trans_handle
*trans
,
1823 struct btrfs_root
*root
,
1825 u64 root_objectid
, u64 hint
,
1831 ref_generation
= trans
->transid
;
1836 return __btrfs_alloc_free_block(trans
, root
, blocksize
, root_objectid
,
1837 ref_generation
, 0, 0, hint
, empty_size
);
1841 * helper function to allocate a block for a given tree
1842 * returns the tree buffer or NULL.
1844 struct extent_buffer
*__btrfs_alloc_free_block(struct btrfs_trans_handle
*trans
,
1845 struct btrfs_root
*root
,
1854 struct btrfs_key ins
;
1856 struct extent_buffer
*buf
;
1858 ret
= btrfs_alloc_extent(trans
, root
, blocksize
,
1859 root_objectid
, ref_generation
,
1860 level
, first_objectid
, empty_size
, hint
,
1864 return ERR_PTR(ret
);
1866 buf
= btrfs_find_create_tree_block(root
, ins
.objectid
, blocksize
);
1868 btrfs_free_extent(trans
, root
, ins
.objectid
, blocksize
,
1869 root
->root_key
.objectid
, ref_generation
,
1871 return ERR_PTR(-ENOMEM
);
1873 btrfs_set_header_generation(buf
, trans
->transid
);
1874 clean_tree_block(trans
, root
, buf
);
1875 wait_on_tree_block_writeback(root
, buf
);
1876 btrfs_set_buffer_uptodate(buf
);
1878 if (PageDirty(buf
->first_page
)) {
1879 printk("page %lu dirty\n", buf
->first_page
->index
);
1883 set_extent_dirty(&trans
->transaction
->dirty_pages
, buf
->start
,
1884 buf
->start
+ buf
->len
- 1, GFP_NOFS
);
1885 set_extent_bits(&BTRFS_I(root
->fs_info
->btree_inode
)->io_tree
,
1886 buf
->start
, buf
->start
+ buf
->len
- 1,
1887 EXTENT_CSUM
, GFP_NOFS
);
1888 buf
->flags
|= EXTENT_CSUM
;
1889 if (!btrfs_test_opt(root
, SSD
))
1890 btrfs_set_buffer_defrag(buf
);
1891 trans
->blocks_used
++;
1895 static int noinline
drop_leaf_ref(struct btrfs_trans_handle
*trans
,
1896 struct btrfs_root
*root
,
1897 struct extent_buffer
*leaf
)
1900 u64 leaf_generation
;
1901 struct btrfs_key key
;
1902 struct btrfs_file_extent_item
*fi
;
1907 BUG_ON(!btrfs_is_leaf(leaf
));
1908 nritems
= btrfs_header_nritems(leaf
);
1909 leaf_owner
= btrfs_header_owner(leaf
);
1910 leaf_generation
= btrfs_header_generation(leaf
);
1912 for (i
= 0; i
< nritems
; i
++) {
1915 btrfs_item_key_to_cpu(leaf
, &key
, i
);
1916 if (btrfs_key_type(&key
) != BTRFS_EXTENT_DATA_KEY
)
1918 fi
= btrfs_item_ptr(leaf
, i
, struct btrfs_file_extent_item
);
1919 if (btrfs_file_extent_type(leaf
, fi
) ==
1920 BTRFS_FILE_EXTENT_INLINE
)
1923 * FIXME make sure to insert a trans record that
1924 * repeats the snapshot del on crash
1926 disk_bytenr
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
1927 if (disk_bytenr
== 0)
1929 ret
= btrfs_free_extent(trans
, root
, disk_bytenr
,
1930 btrfs_file_extent_disk_num_bytes(leaf
, fi
),
1931 leaf_owner
, leaf_generation
,
1932 key
.objectid
, key
.offset
, 0);
1938 static void noinline
reada_walk_down(struct btrfs_root
*root
,
1939 struct extent_buffer
*node
,
1952 nritems
= btrfs_header_nritems(node
);
1953 level
= btrfs_header_level(node
);
1957 for (i
= slot
; i
< nritems
&& skipped
< 32; i
++) {
1958 bytenr
= btrfs_node_blockptr(node
, i
);
1959 if (last
&& ((bytenr
> last
&& bytenr
- last
> 32 * 1024) ||
1960 (last
> bytenr
&& last
- bytenr
> 32 * 1024))) {
1964 blocksize
= btrfs_level_size(root
, level
- 1);
1966 ret
= lookup_extent_ref(NULL
, root
, bytenr
,
1974 mutex_unlock(&root
->fs_info
->fs_mutex
);
1975 ret
= readahead_tree_block(root
, bytenr
, blocksize
);
1976 last
= bytenr
+ blocksize
;
1978 mutex_lock(&root
->fs_info
->fs_mutex
);
1985 * helper function for drop_snapshot, this walks down the tree dropping ref
1986 * counts as it goes.
1988 static int noinline
walk_down_tree(struct btrfs_trans_handle
*trans
,
1989 struct btrfs_root
*root
,
1990 struct btrfs_path
*path
, int *level
)
1995 struct extent_buffer
*next
;
1996 struct extent_buffer
*cur
;
1997 struct extent_buffer
*parent
;
2002 WARN_ON(*level
< 0);
2003 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
2004 ret
= lookup_extent_ref(trans
, root
,
2005 path
->nodes
[*level
]->start
,
2006 path
->nodes
[*level
]->len
, &refs
);
2012 * walk down to the last node level and free all the leaves
2014 while(*level
>= 0) {
2015 WARN_ON(*level
< 0);
2016 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
2017 cur
= path
->nodes
[*level
];
2019 if (btrfs_header_level(cur
) != *level
)
2022 if (path
->slots
[*level
] >=
2023 btrfs_header_nritems(cur
))
2026 ret
= drop_leaf_ref(trans
, root
, cur
);
2030 bytenr
= btrfs_node_blockptr(cur
, path
->slots
[*level
]);
2031 blocksize
= btrfs_level_size(root
, *level
- 1);
2032 ret
= lookup_extent_ref(trans
, root
, bytenr
, blocksize
, &refs
);
2035 parent
= path
->nodes
[*level
];
2036 root_owner
= btrfs_header_owner(parent
);
2037 root_gen
= btrfs_header_generation(parent
);
2038 path
->slots
[*level
]++;
2039 ret
= btrfs_free_extent(trans
, root
, bytenr
,
2040 blocksize
, root_owner
,
2045 next
= btrfs_find_tree_block(root
, bytenr
, blocksize
);
2046 if (!next
|| !btrfs_buffer_uptodate(next
)) {
2047 free_extent_buffer(next
);
2048 reada_walk_down(root
, cur
, path
->slots
[*level
]);
2049 next
= read_tree_block(root
, bytenr
, blocksize
);
2051 /* we used to drop the lock above, keep the
2052 * code to double check so that we won't forget
2053 * when we drop the lock again in the future
2055 ret
= lookup_extent_ref(trans
, root
, bytenr
,
2059 parent
= path
->nodes
[*level
];
2060 root_owner
= btrfs_header_owner(parent
);
2061 root_gen
= btrfs_header_generation(parent
);
2063 path
->slots
[*level
]++;
2064 free_extent_buffer(next
);
2065 ret
= btrfs_free_extent(trans
, root
, bytenr
,
2073 btrfs_verify_block_csum(root
, next
);
2075 WARN_ON(*level
<= 0);
2076 if (path
->nodes
[*level
-1])
2077 free_extent_buffer(path
->nodes
[*level
-1]);
2078 path
->nodes
[*level
-1] = next
;
2079 *level
= btrfs_header_level(next
);
2080 path
->slots
[*level
] = 0;
2083 WARN_ON(*level
< 0);
2084 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
2086 if (path
->nodes
[*level
] == root
->node
) {
2087 root_owner
= root
->root_key
.objectid
;
2088 parent
= path
->nodes
[*level
];
2090 parent
= path
->nodes
[*level
+ 1];
2091 root_owner
= btrfs_header_owner(parent
);
2094 root_gen
= btrfs_header_generation(parent
);
2095 ret
= btrfs_free_extent(trans
, root
, path
->nodes
[*level
]->start
,
2096 path
->nodes
[*level
]->len
,
2097 root_owner
, root_gen
, 0, 0, 1);
2098 free_extent_buffer(path
->nodes
[*level
]);
2099 path
->nodes
[*level
] = NULL
;
2106 * helper for dropping snapshots. This walks back up the tree in the path
2107 * to find the first node higher up where we haven't yet gone through
2110 static int noinline
walk_up_tree(struct btrfs_trans_handle
*trans
,
2111 struct btrfs_root
*root
,
2112 struct btrfs_path
*path
, int *level
)
2116 struct btrfs_root_item
*root_item
= &root
->root_item
;
2121 for(i
= *level
; i
< BTRFS_MAX_LEVEL
- 1 && path
->nodes
[i
]; i
++) {
2122 slot
= path
->slots
[i
];
2123 if (slot
< btrfs_header_nritems(path
->nodes
[i
]) - 1) {
2124 struct extent_buffer
*node
;
2125 struct btrfs_disk_key disk_key
;
2126 node
= path
->nodes
[i
];
2129 WARN_ON(*level
== 0);
2130 btrfs_node_key(node
, &disk_key
, path
->slots
[i
]);
2131 memcpy(&root_item
->drop_progress
,
2132 &disk_key
, sizeof(disk_key
));
2133 root_item
->drop_level
= i
;
2136 if (path
->nodes
[*level
] == root
->node
) {
2137 root_owner
= root
->root_key
.objectid
;
2139 btrfs_header_generation(path
->nodes
[*level
]);
2141 struct extent_buffer
*node
;
2142 node
= path
->nodes
[*level
+ 1];
2143 root_owner
= btrfs_header_owner(node
);
2144 root_gen
= btrfs_header_generation(node
);
2146 ret
= btrfs_free_extent(trans
, root
,
2147 path
->nodes
[*level
]->start
,
2148 path
->nodes
[*level
]->len
,
2149 root_owner
, root_gen
, 0, 0, 1);
2151 free_extent_buffer(path
->nodes
[*level
]);
2152 path
->nodes
[*level
] = NULL
;
2160 * drop the reference count on the tree rooted at 'snap'. This traverses
2161 * the tree freeing any blocks that have a ref count of zero after being
2164 int btrfs_drop_snapshot(struct btrfs_trans_handle
*trans
, struct btrfs_root
2170 struct btrfs_path
*path
;
2173 struct btrfs_root_item
*root_item
= &root
->root_item
;
2175 path
= btrfs_alloc_path();
2178 level
= btrfs_header_level(root
->node
);
2180 if (btrfs_disk_key_objectid(&root_item
->drop_progress
) == 0) {
2181 path
->nodes
[level
] = root
->node
;
2182 extent_buffer_get(root
->node
);
2183 path
->slots
[level
] = 0;
2185 struct btrfs_key key
;
2186 struct btrfs_disk_key found_key
;
2187 struct extent_buffer
*node
;
2189 btrfs_disk_key_to_cpu(&key
, &root_item
->drop_progress
);
2190 level
= root_item
->drop_level
;
2191 path
->lowest_level
= level
;
2192 wret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2197 node
= path
->nodes
[level
];
2198 btrfs_node_key(node
, &found_key
, path
->slots
[level
]);
2199 WARN_ON(memcmp(&found_key
, &root_item
->drop_progress
,
2200 sizeof(found_key
)));
2203 wret
= walk_down_tree(trans
, root
, path
, &level
);
2209 wret
= walk_up_tree(trans
, root
, path
, &level
);
2217 for (i
= 0; i
<= orig_level
; i
++) {
2218 if (path
->nodes
[i
]) {
2219 free_extent_buffer(path
->nodes
[i
]);
2220 path
->nodes
[i
] = NULL
;
2224 btrfs_free_path(path
);
2228 int btrfs_free_block_groups(struct btrfs_fs_info
*info
)
2235 ret
= find_first_extent_bit(&info
->block_group_cache
, 0,
2236 &start
, &end
, (unsigned int)-1);
2239 ret
= get_state_private(&info
->block_group_cache
, start
, &ptr
);
2241 kfree((void *)(unsigned long)ptr
);
2242 clear_extent_bits(&info
->block_group_cache
, start
,
2243 end
, (unsigned int)-1, GFP_NOFS
);
2246 ret
= find_first_extent_bit(&info
->free_space_cache
, 0,
2247 &start
, &end
, EXTENT_DIRTY
);
2250 clear_extent_dirty(&info
->free_space_cache
, start
,
2256 static int noinline
relocate_inode_pages(struct inode
*inode
, u64 start
,
2262 u64 existing_delalloc
;
2263 unsigned long last_index
;
2266 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2267 struct file_ra_state
*ra
;
2269 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
2271 mutex_lock(&inode
->i_mutex
);
2272 i
= start
>> PAGE_CACHE_SHIFT
;
2273 last_index
= (start
+ len
- 1) >> PAGE_CACHE_SHIFT
;
2275 file_ra_state_init(ra
, inode
->i_mapping
);
2276 btrfs_force_ra(inode
->i_mapping
, ra
, NULL
, i
, last_index
);
2279 for (; i
<= last_index
; i
++) {
2280 page
= grab_cache_page(inode
->i_mapping
, i
);
2283 if (!PageUptodate(page
)) {
2284 btrfs_readpage(NULL
, page
);
2286 if (!PageUptodate(page
)) {
2288 page_cache_release(page
);
2292 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
2293 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
2295 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
2297 delalloc_start
= page_start
;
2298 existing_delalloc
= count_range_bits(io_tree
,
2299 &delalloc_start
, page_end
,
2300 PAGE_CACHE_SIZE
, EXTENT_DELALLOC
);
2302 set_extent_delalloc(io_tree
, page_start
,
2303 page_end
, GFP_NOFS
);
2305 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
2306 set_page_dirty(page
);
2308 page_cache_release(page
);
2312 mutex_unlock(&inode
->i_mutex
);
2317 * note, this releases the path
2319 static int noinline
relocate_one_reference(struct btrfs_root
*extent_root
,
2320 struct btrfs_path
*path
,
2321 struct btrfs_key
*extent_key
)
2323 struct inode
*inode
;
2324 struct btrfs_root
*found_root
;
2325 struct btrfs_key
*root_location
;
2326 struct btrfs_extent_ref
*ref
;
2333 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
2334 struct btrfs_extent_ref
);
2335 ref_root
= btrfs_ref_root(path
->nodes
[0], ref
);
2336 ref_gen
= btrfs_ref_generation(path
->nodes
[0], ref
);
2337 ref_objectid
= btrfs_ref_objectid(path
->nodes
[0], ref
);
2338 ref_offset
= btrfs_ref_offset(path
->nodes
[0], ref
);
2339 btrfs_release_path(extent_root
, path
);
2341 root_location
= kmalloc(sizeof(*root_location
), GFP_NOFS
);
2342 root_location
->objectid
= ref_root
;
2344 root_location
->offset
= 0;
2346 root_location
->offset
= (u64
)-1;
2347 root_location
->type
= BTRFS_ROOT_ITEM_KEY
;
2349 found_root
= btrfs_read_fs_root_no_name(extent_root
->fs_info
,
2351 BUG_ON(!found_root
);
2352 kfree(root_location
);
2354 if (ref_objectid
>= BTRFS_FIRST_FREE_OBJECTID
) {
2355 mutex_unlock(&extent_root
->fs_info
->fs_mutex
);
2356 inode
= btrfs_iget_locked(extent_root
->fs_info
->sb
,
2357 ref_objectid
, found_root
);
2358 if (inode
->i_state
& I_NEW
) {
2359 /* the inode and parent dir are two different roots */
2360 BTRFS_I(inode
)->root
= found_root
;
2361 BTRFS_I(inode
)->location
.objectid
= ref_objectid
;
2362 BTRFS_I(inode
)->location
.type
= BTRFS_INODE_ITEM_KEY
;
2363 BTRFS_I(inode
)->location
.offset
= 0;
2364 btrfs_read_locked_inode(inode
);
2365 unlock_new_inode(inode
);
2368 /* this can happen if the reference is not against
2369 * the latest version of the tree root
2371 if (is_bad_inode(inode
)) {
2372 mutex_lock(&extent_root
->fs_info
->fs_mutex
);
2375 relocate_inode_pages(inode
, ref_offset
, extent_key
->offset
);
2376 /* FIXME, data=ordered will help get rid of this */
2377 filemap_fdatawrite(inode
->i_mapping
);
2379 mutex_lock(&extent_root
->fs_info
->fs_mutex
);
2381 struct btrfs_trans_handle
*trans
;
2382 struct btrfs_key found_key
;
2383 struct extent_buffer
*eb
;
2387 trans
= btrfs_start_transaction(found_root
, 1);
2388 eb
= read_tree_block(found_root
, extent_key
->objectid
,
2389 extent_key
->offset
);
2390 level
= btrfs_header_level(eb
);
2393 btrfs_item_key_to_cpu(eb
, &found_key
, 0);
2395 btrfs_node_key_to_cpu(eb
, &found_key
, 0);
2397 free_extent_buffer(eb
);
2399 path
->lowest_level
= level
;
2401 ret
= btrfs_search_slot(trans
, found_root
, &found_key
, path
,
2403 path
->lowest_level
= 0;
2404 for (i
= level
; i
< BTRFS_MAX_LEVEL
; i
++) {
2405 if (!path
->nodes
[i
])
2407 free_extent_buffer(path
->nodes
[i
]);
2408 path
->nodes
[i
] = NULL
;
2410 btrfs_release_path(found_root
, path
);
2411 btrfs_end_transaction(trans
, found_root
);
2418 static int noinline
relocate_one_extent(struct btrfs_root
*extent_root
,
2419 struct btrfs_path
*path
,
2420 struct btrfs_key
*extent_key
)
2422 struct btrfs_key key
;
2423 struct btrfs_key found_key
;
2424 struct extent_buffer
*leaf
;
2429 key
.objectid
= extent_key
->objectid
;
2430 key
.type
= BTRFS_EXTENT_REF_KEY
;
2434 ret
= btrfs_search_slot(NULL
, extent_root
, &key
, path
, 0, 0);
2440 leaf
= path
->nodes
[0];
2441 nritems
= btrfs_header_nritems(leaf
);
2442 if (path
->slots
[0] == nritems
)
2445 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2446 if (found_key
.objectid
!= extent_key
->objectid
)
2449 if (found_key
.type
!= BTRFS_EXTENT_REF_KEY
)
2452 key
.offset
= found_key
.offset
+ 1;
2453 item_size
= btrfs_item_size_nr(leaf
, path
->slots
[0]);
2455 ret
= relocate_one_reference(extent_root
, path
, extent_key
);
2461 btrfs_release_path(extent_root
, path
);
2465 int btrfs_shrink_extent_tree(struct btrfs_root
*root
, u64 new_size
)
2467 struct btrfs_trans_handle
*trans
;
2468 struct btrfs_root
*tree_root
= root
->fs_info
->tree_root
;
2469 struct btrfs_path
*path
;
2472 struct btrfs_fs_info
*info
= root
->fs_info
;
2473 struct extent_io_tree
*block_group_cache
;
2474 struct btrfs_key key
;
2475 struct btrfs_key found_key
;
2476 struct extent_buffer
*leaf
;
2481 btrfs_set_super_total_bytes(&info
->super_copy
, new_size
);
2482 clear_extent_dirty(&info
->free_space_cache
, new_size
, (u64
)-1,
2484 block_group_cache
= &info
->block_group_cache
;
2485 path
= btrfs_alloc_path();
2486 root
= root
->fs_info
->extent_root
;
2491 key
.objectid
= new_size
;
2494 cur_byte
= key
.objectid
;
2496 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2500 ret
= btrfs_previous_item(root
, path
, 0, BTRFS_EXTENT_ITEM_KEY
);
2504 leaf
= path
->nodes
[0];
2505 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2506 if (found_key
.objectid
+ found_key
.offset
> new_size
) {
2507 cur_byte
= found_key
.objectid
;
2508 key
.objectid
= cur_byte
;
2511 btrfs_release_path(root
, path
);
2514 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2518 leaf
= path
->nodes
[0];
2519 nritems
= btrfs_header_nritems(leaf
);
2521 if (path
->slots
[0] >= nritems
) {
2522 ret
= btrfs_next_leaf(root
, path
);
2529 leaf
= path
->nodes
[0];
2530 nritems
= btrfs_header_nritems(leaf
);
2533 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2535 if (progress
&& need_resched()) {
2536 memcpy(&key
, &found_key
, sizeof(key
));
2537 mutex_unlock(&root
->fs_info
->fs_mutex
);
2539 mutex_lock(&root
->fs_info
->fs_mutex
);
2540 btrfs_release_path(root
, path
);
2541 btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2547 if (btrfs_key_type(&found_key
) != BTRFS_EXTENT_ITEM_KEY
||
2548 found_key
.objectid
+ found_key
.offset
<= cur_byte
) {
2554 cur_byte
= found_key
.objectid
+ found_key
.offset
;
2555 key
.objectid
= cur_byte
;
2556 btrfs_release_path(root
, path
);
2557 ret
= relocate_one_extent(root
, path
, &found_key
);
2560 btrfs_release_path(root
, path
);
2562 if (total_found
> 0) {
2563 trans
= btrfs_start_transaction(tree_root
, 1);
2564 btrfs_commit_transaction(trans
, tree_root
);
2566 mutex_unlock(&root
->fs_info
->fs_mutex
);
2567 btrfs_clean_old_snapshots(tree_root
);
2568 mutex_lock(&root
->fs_info
->fs_mutex
);
2570 trans
= btrfs_start_transaction(tree_root
, 1);
2571 btrfs_commit_transaction(trans
, tree_root
);
2575 trans
= btrfs_start_transaction(root
, 1);
2576 key
.objectid
= new_size
;
2582 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
2586 leaf
= path
->nodes
[0];
2587 nritems
= btrfs_header_nritems(leaf
);
2589 if (path
->slots
[0] >= nritems
) {
2590 ret
= btrfs_next_leaf(root
, path
);
2597 leaf
= path
->nodes
[0];
2598 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2601 * btrfs_next_leaf doesn't cow buffers, we have to
2602 * do the search again
2604 memcpy(&key
, &found_key
, sizeof(key
));
2605 btrfs_release_path(root
, path
);
2609 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2610 if (btrfs_key_type(&found_key
) != BTRFS_BLOCK_GROUP_ITEM_KEY
) {
2611 printk("shrinker found key %Lu %u %Lu\n",
2612 found_key
.objectid
, found_key
.type
,
2617 ret
= get_state_private(&info
->block_group_cache
,
2618 found_key
.objectid
, &ptr
);
2620 kfree((void *)(unsigned long)ptr
);
2622 clear_extent_bits(&info
->block_group_cache
, found_key
.objectid
,
2623 found_key
.objectid
+ found_key
.offset
- 1,
2624 (unsigned int)-1, GFP_NOFS
);
2626 key
.objectid
= found_key
.objectid
+ 1;
2627 btrfs_del_item(trans
, root
, path
);
2628 btrfs_release_path(root
, path
);
2630 if (need_resched()) {
2631 mutex_unlock(&root
->fs_info
->fs_mutex
);
2633 mutex_lock(&root
->fs_info
->fs_mutex
);
2636 clear_extent_dirty(&info
->free_space_cache
, new_size
, (u64
)-1,
2638 btrfs_commit_transaction(trans
, root
);
2640 btrfs_free_path(path
);
2644 int btrfs_grow_extent_tree(struct btrfs_trans_handle
*trans
,
2645 struct btrfs_root
*root
, u64 new_size
)
2647 btrfs_set_super_total_bytes(&root
->fs_info
->super_copy
, new_size
);
2651 int find_first_block_group(struct btrfs_root
*root
, struct btrfs_path
*path
,
2652 struct btrfs_key
*key
)
2655 struct btrfs_key found_key
;
2656 struct extent_buffer
*leaf
;
2659 ret
= btrfs_search_slot(NULL
, root
, key
, path
, 0, 0);
2663 slot
= path
->slots
[0];
2664 leaf
= path
->nodes
[0];
2665 if (slot
>= btrfs_header_nritems(leaf
)) {
2666 ret
= btrfs_next_leaf(root
, path
);
2673 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
2675 if (found_key
.objectid
>= key
->objectid
&&
2676 found_key
.type
== BTRFS_BLOCK_GROUP_ITEM_KEY
)
2685 int btrfs_read_block_groups(struct btrfs_root
*root
)
2687 struct btrfs_path
*path
;
2690 struct btrfs_block_group_cache
*cache
;
2691 struct btrfs_fs_info
*info
= root
->fs_info
;
2692 struct btrfs_space_info
*space_info
;
2693 struct extent_io_tree
*block_group_cache
;
2694 struct btrfs_key key
;
2695 struct btrfs_key found_key
;
2696 struct extent_buffer
*leaf
;
2698 block_group_cache
= &info
->block_group_cache
;
2699 root
= info
->extent_root
;
2702 btrfs_set_key_type(&key
, BTRFS_BLOCK_GROUP_ITEM_KEY
);
2703 path
= btrfs_alloc_path();
2708 ret
= find_first_block_group(root
, path
, &key
);
2716 leaf
= path
->nodes
[0];
2717 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2718 cache
= kmalloc(sizeof(*cache
), GFP_NOFS
);
2724 read_extent_buffer(leaf
, &cache
->item
,
2725 btrfs_item_ptr_offset(leaf
, path
->slots
[0]),
2726 sizeof(cache
->item
));
2727 memcpy(&cache
->key
, &found_key
, sizeof(found_key
));
2731 key
.objectid
= found_key
.objectid
+ found_key
.offset
;
2732 btrfs_release_path(root
, path
);
2733 cache
->flags
= btrfs_block_group_flags(&cache
->item
);
2735 if (cache
->flags
& BTRFS_BLOCK_GROUP_DATA
) {
2736 bit
= BLOCK_GROUP_DATA
;
2737 } else if (cache
->flags
& BTRFS_BLOCK_GROUP_SYSTEM
) {
2738 bit
= BLOCK_GROUP_SYSTEM
;
2739 } else if (cache
->flags
& BTRFS_BLOCK_GROUP_METADATA
) {
2740 bit
= BLOCK_GROUP_METADATA
;
2742 if (cache
->flags
& BTRFS_BLOCK_GROUP_RAID0
) {
2743 if (cache
->flags
& BTRFS_BLOCK_GROUP_DATA
) {
2744 info
->extra_data_alloc_bits
=
2745 BTRFS_BLOCK_GROUP_RAID0
;
2747 if (cache
->flags
& BTRFS_BLOCK_GROUP_METADATA
) {
2748 info
->extra_alloc_bits
=
2749 BTRFS_BLOCK_GROUP_RAID0
;
2753 ret
= update_space_info(info
, cache
->flags
, found_key
.offset
,
2754 btrfs_block_group_used(&cache
->item
),
2757 cache
->space_info
= space_info
;
2759 /* use EXTENT_LOCKED to prevent merging */
2760 set_extent_bits(block_group_cache
, found_key
.objectid
,
2761 found_key
.objectid
+ found_key
.offset
- 1,
2762 bit
| EXTENT_LOCKED
, GFP_NOFS
);
2763 set_state_private(block_group_cache
, found_key
.objectid
,
2764 (unsigned long)cache
);
2767 btrfs_super_total_bytes(&info
->super_copy
))
2772 btrfs_free_path(path
);
2776 int btrfs_make_block_group(struct btrfs_trans_handle
*trans
,
2777 struct btrfs_root
*root
, u64 bytes_used
,
2778 u64 type
, u64 chunk_tree
, u64 chunk_objectid
,
2783 struct btrfs_root
*extent_root
;
2784 struct btrfs_block_group_cache
*cache
;
2785 struct extent_io_tree
*block_group_cache
;
2787 extent_root
= root
->fs_info
->extent_root
;
2788 block_group_cache
= &root
->fs_info
->block_group_cache
;
2790 cache
= kmalloc(sizeof(*cache
), GFP_NOFS
);
2792 cache
->key
.objectid
= chunk_objectid
;
2793 cache
->key
.offset
= size
;
2796 btrfs_set_key_type(&cache
->key
, BTRFS_BLOCK_GROUP_ITEM_KEY
);
2797 memset(&cache
->item
, 0, sizeof(cache
->item
));
2798 btrfs_set_block_group_used(&cache
->item
, bytes_used
);
2799 btrfs_set_block_group_chunk_tree(&cache
->item
, chunk_tree
);
2800 btrfs_set_block_group_chunk_objectid(&cache
->item
, chunk_objectid
);
2801 cache
->flags
= type
;
2802 btrfs_set_block_group_flags(&cache
->item
, type
);
2804 ret
= update_space_info(root
->fs_info
, cache
->flags
, size
, bytes_used
,
2805 &cache
->space_info
);
2808 if (type
& BTRFS_BLOCK_GROUP_DATA
) {
2809 bit
= BLOCK_GROUP_DATA
;
2810 } else if (type
& BTRFS_BLOCK_GROUP_SYSTEM
) {
2811 bit
= BLOCK_GROUP_SYSTEM
;
2812 } else if (type
& BTRFS_BLOCK_GROUP_METADATA
) {
2813 bit
= BLOCK_GROUP_METADATA
;
2815 set_extent_bits(block_group_cache
, chunk_objectid
,
2816 chunk_objectid
+ size
- 1,
2817 bit
| EXTENT_LOCKED
, GFP_NOFS
);
2818 set_state_private(block_group_cache
, chunk_objectid
,
2819 (unsigned long)cache
);
2821 ret
= btrfs_insert_item(trans
, extent_root
, &cache
->key
, &cache
->item
,
2822 sizeof(cache
->item
));
2825 finish_current_insert(trans
, extent_root
);
2826 ret
= del_pending_extents(trans
, extent_root
);