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.
18 #include <linux/sched.h>
19 #include <linux/pagemap.h>
24 #include "print-tree.h"
25 #include "transaction.h"
28 #define BLOCK_GROUP_DATA EXTENT_WRITEBACK
29 #define BLOCK_GROUP_METADATA EXTENT_UPTODATE
30 #define BLOCK_GROUP_SYSTEM EXTENT_NEW
32 #define BLOCK_GROUP_DIRTY EXTENT_DIRTY
34 static int finish_current_insert(struct btrfs_trans_handle
*trans
, struct
35 btrfs_root
*extent_root
);
36 static int del_pending_extents(struct btrfs_trans_handle
*trans
, struct
37 btrfs_root
*extent_root
);
38 int btrfs_make_block_group(struct btrfs_trans_handle
*trans
,
39 struct btrfs_root
*root
, u64 bytes_used
,
40 u64 type
, u64 chunk_tree
, u64 chunk_objectid
,
44 static int cache_block_group(struct btrfs_root
*root
,
45 struct btrfs_block_group_cache
*block_group
)
47 struct btrfs_path
*path
;
50 struct extent_buffer
*leaf
;
51 struct extent_io_tree
*free_space_cache
;
61 root
= root
->fs_info
->extent_root
;
62 free_space_cache
= &root
->fs_info
->free_space_cache
;
64 if (block_group
->cached
)
67 path
= btrfs_alloc_path();
72 first_free
= block_group
->key
.objectid
;
73 key
.objectid
= block_group
->key
.objectid
;
75 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
76 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
79 ret
= btrfs_previous_item(root
, path
, 0, BTRFS_EXTENT_ITEM_KEY
);
83 leaf
= path
->nodes
[0];
84 btrfs_item_key_to_cpu(leaf
, &key
, path
->slots
[0]);
85 if (key
.objectid
+ key
.offset
> first_free
)
86 first_free
= key
.objectid
+ key
.offset
;
89 leaf
= path
->nodes
[0];
90 slot
= path
->slots
[0];
91 if (slot
>= btrfs_header_nritems(leaf
)) {
92 ret
= btrfs_next_leaf(root
, path
);
101 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
102 if (key
.objectid
< block_group
->key
.objectid
) {
105 if (key
.objectid
>= block_group
->key
.objectid
+
106 block_group
->key
.offset
) {
110 if (btrfs_key_type(&key
) == BTRFS_EXTENT_ITEM_KEY
) {
115 if (key
.objectid
> last
) {
116 hole_size
= key
.objectid
- last
;
117 set_extent_dirty(free_space_cache
, last
,
118 last
+ hole_size
- 1,
121 last
= key
.objectid
+ key
.offset
;
129 if (block_group
->key
.objectid
+
130 block_group
->key
.offset
> last
) {
131 hole_size
= block_group
->key
.objectid
+
132 block_group
->key
.offset
- last
;
133 set_extent_dirty(free_space_cache
, last
,
134 last
+ hole_size
- 1, GFP_NOFS
);
136 block_group
->cached
= 1;
138 btrfs_free_path(path
);
142 struct btrfs_block_group_cache
*btrfs_lookup_block_group(struct
146 struct extent_io_tree
*block_group_cache
;
147 struct btrfs_block_group_cache
*block_group
= NULL
;
153 block_group_cache
= &info
->block_group_cache
;
154 ret
= find_first_extent_bit(block_group_cache
,
155 bytenr
, &start
, &end
,
156 BLOCK_GROUP_DATA
| BLOCK_GROUP_METADATA
|
161 ret
= get_state_private(block_group_cache
, start
, &ptr
);
165 block_group
= (struct btrfs_block_group_cache
*)(unsigned long)ptr
;
166 if (block_group
->key
.objectid
<= bytenr
&& bytenr
<
167 block_group
->key
.objectid
+ block_group
->key
.offset
)
172 static int block_group_bits(struct btrfs_block_group_cache
*cache
, u64 bits
)
174 return (cache
->flags
& bits
) == bits
;
177 static int noinline
find_search_start(struct btrfs_root
*root
,
178 struct btrfs_block_group_cache
**cache_ret
,
179 u64
*start_ret
, int num
, int data
)
182 struct btrfs_block_group_cache
*cache
= *cache_ret
;
183 struct extent_io_tree
*free_space_cache
;
184 struct extent_state
*state
;
189 u64 search_start
= *start_ret
;
194 total_fs_bytes
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
195 free_space_cache
= &root
->fs_info
->free_space_cache
;
198 ret
= cache_block_group(root
, cache
);
202 last
= max(search_start
, cache
->key
.objectid
);
203 if (!block_group_bits(cache
, data
)) {
207 spin_lock_irq(&free_space_cache
->lock
);
208 state
= find_first_extent_bit_state(free_space_cache
, last
, EXTENT_DIRTY
);
213 spin_unlock_irq(&free_space_cache
->lock
);
217 start
= max(last
, state
->start
);
218 last
= state
->end
+ 1;
219 if (last
- start
< num
) {
220 if (last
== cache
->key
.objectid
+ cache
->key
.offset
)
223 state
= extent_state_next(state
);
224 } while(state
&& !(state
->state
& EXTENT_DIRTY
));
227 spin_unlock_irq(&free_space_cache
->lock
);
228 if (start
+ num
> cache
->key
.objectid
+ cache
->key
.offset
)
230 if (start
+ num
> total_fs_bytes
)
232 if (!block_group_bits(cache
, data
)) {
233 printk("block group bits don't match %Lu %d\n", cache
->flags
, data
);
239 cache
= btrfs_lookup_block_group(root
->fs_info
, search_start
);
241 printk("Unable to find block group for %Lu\n", search_start
);
247 last
= cache
->key
.objectid
+ cache
->key
.offset
;
249 cache
= btrfs_lookup_block_group(root
->fs_info
, last
);
250 if (!cache
|| cache
->key
.objectid
>= total_fs_bytes
) {
259 if (cache_miss
&& !cache
->cached
) {
260 cache_block_group(root
, cache
);
262 cache
= btrfs_lookup_block_group(root
->fs_info
, last
);
264 cache
= btrfs_find_block_group(root
, cache
, last
, data
, 0);
272 static u64
div_factor(u64 num
, int factor
)
281 static int block_group_state_bits(u64 flags
)
284 if (flags
& BTRFS_BLOCK_GROUP_DATA
)
285 bits
|= BLOCK_GROUP_DATA
;
286 if (flags
& BTRFS_BLOCK_GROUP_METADATA
)
287 bits
|= BLOCK_GROUP_METADATA
;
288 if (flags
& BTRFS_BLOCK_GROUP_SYSTEM
)
289 bits
|= BLOCK_GROUP_SYSTEM
;
293 struct btrfs_block_group_cache
*btrfs_find_block_group(struct btrfs_root
*root
,
294 struct btrfs_block_group_cache
295 *hint
, u64 search_start
,
298 struct btrfs_block_group_cache
*cache
;
299 struct extent_io_tree
*block_group_cache
;
300 struct btrfs_block_group_cache
*found_group
= NULL
;
301 struct btrfs_fs_info
*info
= root
->fs_info
;
315 block_group_cache
= &info
->block_group_cache
;
316 total_fs_bytes
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
321 bit
= block_group_state_bits(data
);
323 if (search_start
&& search_start
< total_fs_bytes
) {
324 struct btrfs_block_group_cache
*shint
;
325 shint
= btrfs_lookup_block_group(info
, search_start
);
326 if (shint
&& block_group_bits(shint
, data
)) {
327 used
= btrfs_block_group_used(&shint
->item
);
328 if (used
+ shint
->pinned
<
329 div_factor(shint
->key
.offset
, factor
)) {
334 if (hint
&& block_group_bits(hint
, data
) &&
335 hint
->key
.objectid
< total_fs_bytes
) {
336 used
= btrfs_block_group_used(&hint
->item
);
337 if (used
+ hint
->pinned
<
338 div_factor(hint
->key
.offset
, factor
)) {
341 last
= hint
->key
.objectid
+ hint
->key
.offset
;
345 hint_last
= max(hint
->key
.objectid
, search_start
);
347 hint_last
= search_start
;
349 if (hint_last
>= total_fs_bytes
)
350 hint_last
= search_start
;
355 ret
= find_first_extent_bit(block_group_cache
, last
,
360 ret
= get_state_private(block_group_cache
, start
, &ptr
);
364 cache
= (struct btrfs_block_group_cache
*)(unsigned long)ptr
;
365 last
= cache
->key
.objectid
+ cache
->key
.offset
;
366 used
= btrfs_block_group_used(&cache
->item
);
368 if (cache
->key
.objectid
> total_fs_bytes
)
371 if (block_group_bits(cache
, data
)) {
373 free_check
= cache
->key
.offset
;
375 free_check
= div_factor(cache
->key
.offset
,
378 if (used
+ cache
->pinned
< free_check
) {
394 static u64
hash_extent_ref(u64 root_objectid
, u64 ref_generation
,
395 u64 owner
, u64 owner_offset
)
397 u32 high_crc
= ~(u32
)0;
398 u32 low_crc
= ~(u32
)0;
400 lenum
= cpu_to_le64(root_objectid
);
401 high_crc
= btrfs_crc32c(high_crc
, &lenum
, sizeof(lenum
));
402 lenum
= cpu_to_le64(ref_generation
);
403 low_crc
= btrfs_crc32c(low_crc
, &lenum
, sizeof(lenum
));
404 if (owner
>= BTRFS_FIRST_FREE_OBJECTID
) {
405 lenum
= cpu_to_le64(owner
);
406 low_crc
= btrfs_crc32c(low_crc
, &lenum
, sizeof(lenum
));
407 lenum
= cpu_to_le64(owner_offset
);
408 low_crc
= btrfs_crc32c(low_crc
, &lenum
, sizeof(lenum
));
410 return ((u64
)high_crc
<< 32) | (u64
)low_crc
;
413 static int match_extent_ref(struct extent_buffer
*leaf
,
414 struct btrfs_extent_ref
*disk_ref
,
415 struct btrfs_extent_ref
*cpu_ref
)
420 if (cpu_ref
->objectid
)
421 len
= sizeof(*cpu_ref
);
423 len
= 2 * sizeof(u64
);
424 ret
= memcmp_extent_buffer(leaf
, cpu_ref
, (unsigned long)disk_ref
,
429 static int noinline
lookup_extent_backref(struct btrfs_trans_handle
*trans
,
430 struct btrfs_root
*root
,
431 struct btrfs_path
*path
, u64 bytenr
,
433 u64 ref_generation
, u64 owner
,
434 u64 owner_offset
, int del
)
437 struct btrfs_key key
;
438 struct btrfs_key found_key
;
439 struct btrfs_extent_ref ref
;
440 struct extent_buffer
*leaf
;
441 struct btrfs_extent_ref
*disk_ref
;
445 btrfs_set_stack_ref_root(&ref
, root_objectid
);
446 btrfs_set_stack_ref_generation(&ref
, ref_generation
);
447 btrfs_set_stack_ref_objectid(&ref
, owner
);
448 btrfs_set_stack_ref_offset(&ref
, owner_offset
);
450 hash
= hash_extent_ref(root_objectid
, ref_generation
, owner
,
453 key
.objectid
= bytenr
;
454 key
.type
= BTRFS_EXTENT_REF_KEY
;
457 ret
= btrfs_search_slot(trans
, root
, &key
, path
,
461 leaf
= path
->nodes
[0];
463 u32 nritems
= btrfs_header_nritems(leaf
);
464 if (path
->slots
[0] >= nritems
) {
465 ret2
= btrfs_next_leaf(root
, path
);
468 leaf
= path
->nodes
[0];
470 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
471 if (found_key
.objectid
!= bytenr
||
472 found_key
.type
!= BTRFS_EXTENT_REF_KEY
)
474 key
.offset
= found_key
.offset
;
476 btrfs_release_path(root
, path
);
480 disk_ref
= btrfs_item_ptr(path
->nodes
[0],
482 struct btrfs_extent_ref
);
483 if (match_extent_ref(path
->nodes
[0], disk_ref
, &ref
)) {
487 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
488 key
.offset
= found_key
.offset
+ 1;
489 btrfs_release_path(root
, path
);
496 * Back reference rules. Back refs have three main goals:
498 * 1) differentiate between all holders of references to an extent so that
499 * when a reference is dropped we can make sure it was a valid reference
500 * before freeing the extent.
502 * 2) Provide enough information to quickly find the holders of an extent
503 * if we notice a given block is corrupted or bad.
505 * 3) Make it easy to migrate blocks for FS shrinking or storage pool
506 * maintenance. This is actually the same as #2, but with a slightly
507 * different use case.
509 * File extents can be referenced by:
511 * - multiple snapshots, subvolumes, or different generations in one subvol
512 * - different files inside a single subvolume (in theory, not implemented yet)
513 * - different offsets inside a file (bookend extents in file.c)
515 * The extent ref structure has fields for:
517 * - Objectid of the subvolume root
518 * - Generation number of the tree holding the reference
519 * - objectid of the file holding the reference
520 * - offset in the file corresponding to the key holding the reference
522 * When a file extent is allocated the fields are filled in:
523 * (root_key.objectid, trans->transid, inode objectid, offset in file)
525 * When a leaf is cow'd new references are added for every file extent found
526 * in the leaf. It looks the same as the create case, but trans->transid
527 * will be different when the block is cow'd.
529 * (root_key.objectid, trans->transid, inode objectid, offset in file)
531 * When a file extent is removed either during snapshot deletion or file
532 * truncation, the corresponding back reference is found
535 * (btrfs_header_owner(leaf), btrfs_header_generation(leaf),
536 * inode objectid, offset in file)
538 * Btree extents can be referenced by:
540 * - Different subvolumes
541 * - Different generations of the same subvolume
543 * Storing sufficient information for a full reverse mapping of a btree
544 * block would require storing the lowest key of the block in the backref,
545 * and it would require updating that lowest key either before write out or
546 * every time it changed. Instead, the objectid of the lowest key is stored
547 * along with the level of the tree block. This provides a hint
548 * about where in the btree the block can be found. Searches through the
549 * btree only need to look for a pointer to that block, so they stop one
550 * level higher than the level recorded in the backref.
552 * Some btrees do not do reference counting on their extents. These
553 * include the extent tree and the tree of tree roots. Backrefs for these
554 * trees always have a generation of zero.
556 * When a tree block is created, back references are inserted:
558 * (root->root_key.objectid, trans->transid or zero, level, lowest_key_objectid)
560 * When a tree block is cow'd in a reference counted root,
561 * new back references are added for all the blocks it points to.
562 * These are of the form (trans->transid will have increased since creation):
564 * (root->root_key.objectid, trans->transid, level, lowest_key_objectid)
566 * Because the lowest_key_objectid and the level are just hints
567 * they are not used when backrefs are deleted. When a backref is deleted:
569 * if backref was for a tree root:
570 * root_objectid = root->root_key.objectid
572 * root_objectid = btrfs_header_owner(parent)
574 * (root_objectid, btrfs_header_generation(parent) or zero, 0, 0)
576 * Back Reference Key hashing:
578 * Back references have four fields, each 64 bits long. Unfortunately,
579 * This is hashed into a single 64 bit number and placed into the key offset.
580 * The key objectid corresponds to the first byte in the extent, and the
581 * key type is set to BTRFS_EXTENT_REF_KEY
583 int btrfs_insert_extent_backref(struct btrfs_trans_handle
*trans
,
584 struct btrfs_root
*root
,
585 struct btrfs_path
*path
, u64 bytenr
,
586 u64 root_objectid
, u64 ref_generation
,
587 u64 owner
, u64 owner_offset
)
590 struct btrfs_key key
;
591 struct btrfs_extent_ref ref
;
592 struct btrfs_extent_ref
*disk_ref
;
595 btrfs_set_stack_ref_root(&ref
, root_objectid
);
596 btrfs_set_stack_ref_generation(&ref
, ref_generation
);
597 btrfs_set_stack_ref_objectid(&ref
, owner
);
598 btrfs_set_stack_ref_offset(&ref
, owner_offset
);
600 hash
= hash_extent_ref(root_objectid
, ref_generation
, owner
,
603 key
.objectid
= bytenr
;
604 key
.type
= BTRFS_EXTENT_REF_KEY
;
606 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
, sizeof(ref
));
607 while (ret
== -EEXIST
) {
608 disk_ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
609 struct btrfs_extent_ref
);
610 if (match_extent_ref(path
->nodes
[0], disk_ref
, &ref
))
613 btrfs_release_path(root
, path
);
614 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
619 disk_ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
620 struct btrfs_extent_ref
);
621 write_extent_buffer(path
->nodes
[0], &ref
, (unsigned long)disk_ref
,
623 btrfs_mark_buffer_dirty(path
->nodes
[0]);
625 btrfs_release_path(root
, path
);
629 int btrfs_inc_extent_ref(struct btrfs_trans_handle
*trans
,
630 struct btrfs_root
*root
,
631 u64 bytenr
, u64 num_bytes
,
632 u64 root_objectid
, u64 ref_generation
,
633 u64 owner
, u64 owner_offset
)
635 struct btrfs_path
*path
;
637 struct btrfs_key key
;
638 struct extent_buffer
*l
;
639 struct btrfs_extent_item
*item
;
642 WARN_ON(num_bytes
< root
->sectorsize
);
643 path
= btrfs_alloc_path();
648 key
.objectid
= bytenr
;
649 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
650 key
.offset
= num_bytes
;
651 ret
= btrfs_search_slot(trans
, root
->fs_info
->extent_root
, &key
, path
,
660 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
661 refs
= btrfs_extent_refs(l
, item
);
662 btrfs_set_extent_refs(l
, item
, refs
+ 1);
663 btrfs_mark_buffer_dirty(path
->nodes
[0]);
665 btrfs_release_path(root
->fs_info
->extent_root
, path
);
668 ret
= btrfs_insert_extent_backref(trans
, root
->fs_info
->extent_root
,
669 path
, bytenr
, root_objectid
,
670 ref_generation
, owner
, owner_offset
);
672 finish_current_insert(trans
, root
->fs_info
->extent_root
);
673 del_pending_extents(trans
, root
->fs_info
->extent_root
);
675 btrfs_free_path(path
);
679 int btrfs_extent_post_op(struct btrfs_trans_handle
*trans
,
680 struct btrfs_root
*root
)
682 finish_current_insert(trans
, root
->fs_info
->extent_root
);
683 del_pending_extents(trans
, root
->fs_info
->extent_root
);
687 static int lookup_extent_ref(struct btrfs_trans_handle
*trans
,
688 struct btrfs_root
*root
, u64 bytenr
,
689 u64 num_bytes
, u32
*refs
)
691 struct btrfs_path
*path
;
693 struct btrfs_key key
;
694 struct extent_buffer
*l
;
695 struct btrfs_extent_item
*item
;
697 WARN_ON(num_bytes
< root
->sectorsize
);
698 path
= btrfs_alloc_path();
700 key
.objectid
= bytenr
;
701 key
.offset
= num_bytes
;
702 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
703 ret
= btrfs_search_slot(trans
, root
->fs_info
->extent_root
, &key
, path
,
708 btrfs_print_leaf(root
, path
->nodes
[0]);
709 printk("failed to find block number %Lu\n", bytenr
);
713 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
714 *refs
= btrfs_extent_refs(l
, item
);
716 btrfs_free_path(path
);
720 u32
btrfs_count_snapshots_in_path(struct btrfs_root
*root
,
721 struct btrfs_path
*count_path
,
724 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
725 struct btrfs_path
*path
;
728 u64 root_objectid
= root
->root_key
.objectid
;
733 struct btrfs_key key
;
734 struct btrfs_key found_key
;
735 struct extent_buffer
*l
;
736 struct btrfs_extent_item
*item
;
737 struct btrfs_extent_ref
*ref_item
;
740 path
= btrfs_alloc_path();
743 bytenr
= first_extent
;
745 bytenr
= count_path
->nodes
[level
]->start
;
748 key
.objectid
= bytenr
;
751 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
752 ret
= btrfs_search_slot(NULL
, extent_root
, &key
, path
, 0, 0);
758 btrfs_item_key_to_cpu(l
, &found_key
, path
->slots
[0]);
760 if (found_key
.objectid
!= bytenr
||
761 found_key
.type
!= BTRFS_EXTENT_ITEM_KEY
) {
765 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
768 nritems
= btrfs_header_nritems(l
);
769 if (path
->slots
[0] >= nritems
) {
770 ret
= btrfs_next_leaf(extent_root
, path
);
775 btrfs_item_key_to_cpu(l
, &found_key
, path
->slots
[0]);
776 if (found_key
.objectid
!= bytenr
)
779 if (found_key
.type
!= BTRFS_EXTENT_REF_KEY
) {
785 ref_item
= btrfs_item_ptr(l
, path
->slots
[0],
786 struct btrfs_extent_ref
);
787 found_objectid
= btrfs_ref_root(l
, ref_item
);
789 if (found_objectid
!= root_objectid
) {
796 if (cur_count
== 0) {
800 if (level
>= 0 && root
->node
== count_path
->nodes
[level
])
803 btrfs_release_path(root
, path
);
807 btrfs_free_path(path
);
810 int btrfs_inc_root_ref(struct btrfs_trans_handle
*trans
,
811 struct btrfs_root
*root
, u64 owner_objectid
)
817 struct btrfs_disk_key disk_key
;
819 level
= btrfs_header_level(root
->node
);
820 generation
= trans
->transid
;
821 nritems
= btrfs_header_nritems(root
->node
);
824 btrfs_item_key(root
->node
, &disk_key
, 0);
826 btrfs_node_key(root
->node
, &disk_key
, 0);
827 key_objectid
= btrfs_disk_key_objectid(&disk_key
);
831 return btrfs_inc_extent_ref(trans
, root
, root
->node
->start
,
832 root
->node
->len
, owner_objectid
,
833 generation
, level
, key_objectid
);
836 int btrfs_inc_ref(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
837 struct extent_buffer
*buf
)
841 struct btrfs_key key
;
842 struct btrfs_file_extent_item
*fi
;
851 level
= btrfs_header_level(buf
);
852 nritems
= btrfs_header_nritems(buf
);
853 for (i
= 0; i
< nritems
; i
++) {
856 btrfs_item_key_to_cpu(buf
, &key
, i
);
857 if (btrfs_key_type(&key
) != BTRFS_EXTENT_DATA_KEY
)
859 fi
= btrfs_item_ptr(buf
, i
,
860 struct btrfs_file_extent_item
);
861 if (btrfs_file_extent_type(buf
, fi
) ==
862 BTRFS_FILE_EXTENT_INLINE
)
864 disk_bytenr
= btrfs_file_extent_disk_bytenr(buf
, fi
);
865 if (disk_bytenr
== 0)
867 ret
= btrfs_inc_extent_ref(trans
, root
, disk_bytenr
,
868 btrfs_file_extent_disk_num_bytes(buf
, fi
),
869 root
->root_key
.objectid
, trans
->transid
,
870 key
.objectid
, key
.offset
);
876 bytenr
= btrfs_node_blockptr(buf
, i
);
877 btrfs_node_key_to_cpu(buf
, &key
, i
);
878 ret
= btrfs_inc_extent_ref(trans
, root
, bytenr
,
879 btrfs_level_size(root
, level
- 1),
880 root
->root_key
.objectid
,
882 level
- 1, key
.objectid
);
893 for (i
=0; i
< faili
; i
++) {
896 btrfs_item_key_to_cpu(buf
, &key
, i
);
897 if (btrfs_key_type(&key
) != BTRFS_EXTENT_DATA_KEY
)
899 fi
= btrfs_item_ptr(buf
, i
,
900 struct btrfs_file_extent_item
);
901 if (btrfs_file_extent_type(buf
, fi
) ==
902 BTRFS_FILE_EXTENT_INLINE
)
904 disk_bytenr
= btrfs_file_extent_disk_bytenr(buf
, fi
);
905 if (disk_bytenr
== 0)
907 err
= btrfs_free_extent(trans
, root
, disk_bytenr
,
908 btrfs_file_extent_disk_num_bytes(buf
,
912 bytenr
= btrfs_node_blockptr(buf
, i
);
913 err
= btrfs_free_extent(trans
, root
, bytenr
,
914 btrfs_level_size(root
, level
- 1), 0);
922 static int write_one_cache_group(struct btrfs_trans_handle
*trans
,
923 struct btrfs_root
*root
,
924 struct btrfs_path
*path
,
925 struct btrfs_block_group_cache
*cache
)
929 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
931 struct extent_buffer
*leaf
;
933 ret
= btrfs_search_slot(trans
, extent_root
, &cache
->key
, path
, 0, 1);
938 leaf
= path
->nodes
[0];
939 bi
= btrfs_item_ptr_offset(leaf
, path
->slots
[0]);
940 write_extent_buffer(leaf
, &cache
->item
, bi
, sizeof(cache
->item
));
941 btrfs_mark_buffer_dirty(leaf
);
942 btrfs_release_path(extent_root
, path
);
944 finish_current_insert(trans
, extent_root
);
945 pending_ret
= del_pending_extents(trans
, extent_root
);
954 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle
*trans
,
955 struct btrfs_root
*root
)
957 struct extent_io_tree
*block_group_cache
;
958 struct btrfs_block_group_cache
*cache
;
962 struct btrfs_path
*path
;
968 block_group_cache
= &root
->fs_info
->block_group_cache
;
969 path
= btrfs_alloc_path();
974 ret
= find_first_extent_bit(block_group_cache
, last
,
975 &start
, &end
, BLOCK_GROUP_DIRTY
);
980 ret
= get_state_private(block_group_cache
, start
, &ptr
);
984 cache
= (struct btrfs_block_group_cache
*)(unsigned long)ptr
;
985 err
= write_one_cache_group(trans
, root
,
988 * if we fail to write the cache group, we want
989 * to keep it marked dirty in hopes that a later
996 clear_extent_bits(block_group_cache
, start
, end
,
997 BLOCK_GROUP_DIRTY
, GFP_NOFS
);
999 btrfs_free_path(path
);
1003 static struct btrfs_space_info
*__find_space_info(struct btrfs_fs_info
*info
,
1006 struct list_head
*head
= &info
->space_info
;
1007 struct list_head
*cur
;
1008 struct btrfs_space_info
*found
;
1009 list_for_each(cur
, head
) {
1010 found
= list_entry(cur
, struct btrfs_space_info
, list
);
1011 if (found
->flags
== flags
)
1018 static int update_space_info(struct btrfs_fs_info
*info
, u64 flags
,
1019 u64 total_bytes
, u64 bytes_used
,
1020 struct btrfs_space_info
**space_info
)
1022 struct btrfs_space_info
*found
;
1024 found
= __find_space_info(info
, flags
);
1026 found
->total_bytes
+= total_bytes
;
1027 found
->bytes_used
+= bytes_used
;
1028 WARN_ON(found
->total_bytes
< found
->bytes_used
);
1029 *space_info
= found
;
1032 found
= kmalloc(sizeof(*found
), GFP_NOFS
);
1036 list_add(&found
->list
, &info
->space_info
);
1037 found
->flags
= flags
;
1038 found
->total_bytes
= total_bytes
;
1039 found
->bytes_used
= bytes_used
;
1040 found
->bytes_pinned
= 0;
1042 *space_info
= found
;
1046 static void set_avail_alloc_bits(struct btrfs_fs_info
*fs_info
, u64 flags
)
1048 u64 extra_flags
= flags
& (BTRFS_BLOCK_GROUP_RAID0
|
1049 BTRFS_BLOCK_GROUP_RAID1
|
1050 BTRFS_BLOCK_GROUP_DUP
);
1052 if (flags
& BTRFS_BLOCK_GROUP_DATA
)
1053 fs_info
->avail_data_alloc_bits
|= extra_flags
;
1054 if (flags
& BTRFS_BLOCK_GROUP_METADATA
)
1055 fs_info
->avail_metadata_alloc_bits
|= extra_flags
;
1056 if (flags
& BTRFS_BLOCK_GROUP_SYSTEM
)
1057 fs_info
->avail_system_alloc_bits
|= extra_flags
;
1061 static int do_chunk_alloc(struct btrfs_trans_handle
*trans
,
1062 struct btrfs_root
*extent_root
, u64 alloc_bytes
,
1065 struct btrfs_space_info
*space_info
;
1071 space_info
= __find_space_info(extent_root
->fs_info
, flags
);
1073 ret
= update_space_info(extent_root
->fs_info
, flags
,
1077 BUG_ON(!space_info
);
1079 if (space_info
->full
)
1082 thresh
= div_factor(space_info
->total_bytes
, 6);
1083 if ((space_info
->bytes_used
+ space_info
->bytes_pinned
+ alloc_bytes
) <
1087 ret
= btrfs_alloc_chunk(trans
, extent_root
, &start
, &num_bytes
, flags
);
1088 if (ret
== -ENOSPC
) {
1089 printk("space info full %Lu\n", flags
);
1090 space_info
->full
= 1;
1096 ret
= btrfs_make_block_group(trans
, extent_root
, 0, flags
,
1097 extent_root
->fs_info
->chunk_root
->root_key
.objectid
,
1104 static int update_block_group(struct btrfs_trans_handle
*trans
,
1105 struct btrfs_root
*root
,
1106 u64 bytenr
, u64 num_bytes
, int alloc
,
1109 struct btrfs_block_group_cache
*cache
;
1110 struct btrfs_fs_info
*info
= root
->fs_info
;
1111 u64 total
= num_bytes
;
1118 cache
= btrfs_lookup_block_group(info
, bytenr
);
1122 byte_in_group
= bytenr
- cache
->key
.objectid
;
1123 WARN_ON(byte_in_group
> cache
->key
.offset
);
1124 start
= cache
->key
.objectid
;
1125 end
= start
+ cache
->key
.offset
- 1;
1126 set_extent_bits(&info
->block_group_cache
, start
, end
,
1127 BLOCK_GROUP_DIRTY
, GFP_NOFS
);
1129 old_val
= btrfs_block_group_used(&cache
->item
);
1130 num_bytes
= min(total
, cache
->key
.offset
- byte_in_group
);
1132 old_val
+= num_bytes
;
1133 cache
->space_info
->bytes_used
+= num_bytes
;
1135 old_val
-= num_bytes
;
1136 cache
->space_info
->bytes_used
-= num_bytes
;
1138 set_extent_dirty(&info
->free_space_cache
,
1139 bytenr
, bytenr
+ num_bytes
- 1,
1143 btrfs_set_block_group_used(&cache
->item
, old_val
);
1145 bytenr
+= num_bytes
;
1150 static int update_pinned_extents(struct btrfs_root
*root
,
1151 u64 bytenr
, u64 num
, int pin
)
1154 struct btrfs_block_group_cache
*cache
;
1155 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1158 set_extent_dirty(&fs_info
->pinned_extents
,
1159 bytenr
, bytenr
+ num
- 1, GFP_NOFS
);
1161 clear_extent_dirty(&fs_info
->pinned_extents
,
1162 bytenr
, bytenr
+ num
- 1, GFP_NOFS
);
1165 cache
= btrfs_lookup_block_group(fs_info
, bytenr
);
1167 len
= min(num
, cache
->key
.offset
-
1168 (bytenr
- cache
->key
.objectid
));
1170 cache
->pinned
+= len
;
1171 cache
->space_info
->bytes_pinned
+= len
;
1172 fs_info
->total_pinned
+= len
;
1174 cache
->pinned
-= len
;
1175 cache
->space_info
->bytes_pinned
-= len
;
1176 fs_info
->total_pinned
-= len
;
1184 int btrfs_copy_pinned(struct btrfs_root
*root
, struct extent_io_tree
*copy
)
1189 struct extent_io_tree
*pinned_extents
= &root
->fs_info
->pinned_extents
;
1193 ret
= find_first_extent_bit(pinned_extents
, last
,
1194 &start
, &end
, EXTENT_DIRTY
);
1197 set_extent_dirty(copy
, start
, end
, GFP_NOFS
);
1203 int btrfs_finish_extent_commit(struct btrfs_trans_handle
*trans
,
1204 struct btrfs_root
*root
,
1205 struct extent_io_tree
*unpin
)
1210 struct extent_io_tree
*free_space_cache
;
1211 free_space_cache
= &root
->fs_info
->free_space_cache
;
1214 ret
= find_first_extent_bit(unpin
, 0, &start
, &end
,
1218 update_pinned_extents(root
, start
, end
+ 1 - start
, 0);
1219 clear_extent_dirty(unpin
, start
, end
, GFP_NOFS
);
1220 set_extent_dirty(free_space_cache
, start
, end
, GFP_NOFS
);
1225 static int finish_current_insert(struct btrfs_trans_handle
*trans
,
1226 struct btrfs_root
*extent_root
)
1230 struct btrfs_fs_info
*info
= extent_root
->fs_info
;
1231 struct extent_buffer
*eb
;
1232 struct btrfs_path
*path
;
1233 struct btrfs_key ins
;
1234 struct btrfs_disk_key first
;
1235 struct btrfs_extent_item extent_item
;
1240 btrfs_set_stack_extent_refs(&extent_item
, 1);
1241 btrfs_set_key_type(&ins
, BTRFS_EXTENT_ITEM_KEY
);
1242 path
= btrfs_alloc_path();
1245 ret
= find_first_extent_bit(&info
->extent_ins
, 0, &start
,
1246 &end
, EXTENT_LOCKED
);
1250 ins
.objectid
= start
;
1251 ins
.offset
= end
+ 1 - start
;
1252 err
= btrfs_insert_item(trans
, extent_root
, &ins
,
1253 &extent_item
, sizeof(extent_item
));
1254 clear_extent_bits(&info
->extent_ins
, start
, end
, EXTENT_LOCKED
,
1256 eb
= read_tree_block(extent_root
, ins
.objectid
, ins
.offset
);
1257 level
= btrfs_header_level(eb
);
1259 btrfs_item_key(eb
, &first
, 0);
1261 btrfs_node_key(eb
, &first
, 0);
1263 err
= btrfs_insert_extent_backref(trans
, extent_root
, path
,
1264 start
, extent_root
->root_key
.objectid
,
1266 btrfs_disk_key_objectid(&first
));
1268 free_extent_buffer(eb
);
1270 btrfs_free_path(path
);
1274 static int pin_down_bytes(struct btrfs_root
*root
, u64 bytenr
, u32 num_bytes
,
1278 struct extent_buffer
*buf
;
1281 buf
= btrfs_find_tree_block(root
, bytenr
, num_bytes
);
1283 if (btrfs_buffer_uptodate(buf
)) {
1285 root
->fs_info
->running_transaction
->transid
;
1286 u64 header_transid
=
1287 btrfs_header_generation(buf
);
1288 if (header_transid
== transid
&&
1289 !btrfs_header_flag(buf
,
1290 BTRFS_HEADER_FLAG_WRITTEN
)) {
1291 clean_tree_block(NULL
, root
, buf
);
1292 free_extent_buffer(buf
);
1296 free_extent_buffer(buf
);
1298 update_pinned_extents(root
, bytenr
, num_bytes
, 1);
1300 set_extent_bits(&root
->fs_info
->pending_del
,
1301 bytenr
, bytenr
+ num_bytes
- 1,
1302 EXTENT_LOCKED
, GFP_NOFS
);
1309 * remove an extent from the root, returns 0 on success
1311 static int __free_extent(struct btrfs_trans_handle
*trans
, struct btrfs_root
1312 *root
, u64 bytenr
, u64 num_bytes
,
1313 u64 root_objectid
, u64 ref_generation
,
1314 u64 owner_objectid
, u64 owner_offset
, int pin
,
1317 struct btrfs_path
*path
;
1318 struct btrfs_key key
;
1319 struct btrfs_fs_info
*info
= root
->fs_info
;
1320 struct btrfs_root
*extent_root
= info
->extent_root
;
1321 struct extent_buffer
*leaf
;
1323 int extent_slot
= 0;
1324 int found_extent
= 0;
1326 struct btrfs_extent_item
*ei
;
1329 key
.objectid
= bytenr
;
1330 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
1331 key
.offset
= num_bytes
;
1332 path
= btrfs_alloc_path();
1337 ret
= lookup_extent_backref(trans
, extent_root
, path
,
1338 bytenr
, root_objectid
,
1340 owner_objectid
, owner_offset
, 1);
1342 struct btrfs_key found_key
;
1343 extent_slot
= path
->slots
[0];
1344 while(extent_slot
> 0) {
1346 btrfs_item_key_to_cpu(path
->nodes
[0], &found_key
,
1348 if (found_key
.objectid
!= bytenr
)
1350 if (found_key
.type
== BTRFS_EXTENT_ITEM_KEY
&&
1351 found_key
.offset
== num_bytes
) {
1355 if (path
->slots
[0] - extent_slot
> 5)
1359 ret
= btrfs_del_item(trans
, extent_root
, path
);
1361 btrfs_print_leaf(extent_root
, path
->nodes
[0]);
1363 printk("Unable to find ref byte nr %Lu root %Lu "
1364 " gen %Lu owner %Lu offset %Lu\n", bytenr
,
1365 root_objectid
, ref_generation
, owner_objectid
,
1368 if (!found_extent
) {
1369 btrfs_release_path(extent_root
, path
);
1370 ret
= btrfs_search_slot(trans
, extent_root
, &key
, path
, -1, 1);
1374 extent_slot
= path
->slots
[0];
1377 leaf
= path
->nodes
[0];
1378 ei
= btrfs_item_ptr(leaf
, extent_slot
,
1379 struct btrfs_extent_item
);
1380 refs
= btrfs_extent_refs(leaf
, ei
);
1383 btrfs_set_extent_refs(leaf
, ei
, refs
);
1385 btrfs_mark_buffer_dirty(leaf
);
1387 if (refs
== 0 && found_extent
&& path
->slots
[0] == extent_slot
+ 1) {
1388 /* if the back ref and the extent are next to each other
1389 * they get deleted below in one shot
1391 path
->slots
[0] = extent_slot
;
1393 } else if (found_extent
) {
1394 /* otherwise delete the extent back ref */
1395 ret
= btrfs_del_item(trans
, extent_root
, path
);
1397 /* if refs are 0, we need to setup the path for deletion */
1399 btrfs_release_path(extent_root
, path
);
1400 ret
= btrfs_search_slot(trans
, extent_root
, &key
, path
,
1413 ret
= pin_down_bytes(root
, bytenr
, num_bytes
, 0);
1419 /* block accounting for super block */
1420 super_used
= btrfs_super_bytes_used(&info
->super_copy
);
1421 btrfs_set_super_bytes_used(&info
->super_copy
,
1422 super_used
- num_bytes
);
1424 /* block accounting for root item */
1425 root_used
= btrfs_root_used(&root
->root_item
);
1426 btrfs_set_root_used(&root
->root_item
,
1427 root_used
- num_bytes
);
1428 ret
= btrfs_del_items(trans
, extent_root
, path
, path
->slots
[0],
1433 ret
= update_block_group(trans
, root
, bytenr
, num_bytes
, 0,
1437 btrfs_free_path(path
);
1438 finish_current_insert(trans
, extent_root
);
1443 * find all the blocks marked as pending in the radix tree and remove
1444 * them from the extent map
1446 static int del_pending_extents(struct btrfs_trans_handle
*trans
, struct
1447 btrfs_root
*extent_root
)
1453 struct extent_io_tree
*pending_del
;
1454 struct extent_io_tree
*pinned_extents
;
1456 pending_del
= &extent_root
->fs_info
->pending_del
;
1457 pinned_extents
= &extent_root
->fs_info
->pinned_extents
;
1460 ret
= find_first_extent_bit(pending_del
, 0, &start
, &end
,
1464 update_pinned_extents(extent_root
, start
, end
+ 1 - start
, 1);
1465 clear_extent_bits(pending_del
, start
, end
, EXTENT_LOCKED
,
1467 ret
= __free_extent(trans
, extent_root
,
1468 start
, end
+ 1 - start
,
1469 extent_root
->root_key
.objectid
,
1478 * remove an extent from the root, returns 0 on success
1480 int btrfs_free_extent(struct btrfs_trans_handle
*trans
, struct btrfs_root
1481 *root
, u64 bytenr
, u64 num_bytes
,
1482 u64 root_objectid
, u64 ref_generation
,
1483 u64 owner_objectid
, u64 owner_offset
, int pin
)
1485 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
1489 WARN_ON(num_bytes
< root
->sectorsize
);
1490 if (!root
->ref_cows
)
1493 if (root
== extent_root
) {
1494 pin_down_bytes(root
, bytenr
, num_bytes
, 1);
1497 ret
= __free_extent(trans
, root
, bytenr
, num_bytes
, root_objectid
,
1498 ref_generation
, owner_objectid
, owner_offset
,
1500 pending_ret
= del_pending_extents(trans
, root
->fs_info
->extent_root
);
1501 return ret
? ret
: pending_ret
;
1504 static u64
stripe_align(struct btrfs_root
*root
, u64 val
)
1506 u64 mask
= ((u64
)root
->stripesize
- 1);
1507 u64 ret
= (val
+ mask
) & ~mask
;
1512 * walks the btree of allocated extents and find a hole of a given size.
1513 * The key ins is changed to record the hole:
1514 * ins->objectid == block start
1515 * ins->flags = BTRFS_EXTENT_ITEM_KEY
1516 * ins->offset == number of blocks
1517 * Any available blocks before search_start are skipped.
1519 static int noinline
find_free_extent(struct btrfs_trans_handle
*trans
,
1520 struct btrfs_root
*orig_root
,
1521 u64 num_bytes
, u64 empty_size
,
1522 u64 search_start
, u64 search_end
,
1523 u64 hint_byte
, struct btrfs_key
*ins
,
1524 u64 exclude_start
, u64 exclude_nr
,
1528 u64 orig_search_start
= search_start
;
1529 struct btrfs_root
* root
= orig_root
->fs_info
->extent_root
;
1530 struct btrfs_fs_info
*info
= root
->fs_info
;
1531 u64 total_needed
= num_bytes
;
1532 u64
*last_ptr
= NULL
;
1533 struct btrfs_block_group_cache
*block_group
;
1536 int empty_cluster
= 2 * 1024 * 1024;
1538 WARN_ON(num_bytes
< root
->sectorsize
);
1539 btrfs_set_key_type(ins
, BTRFS_EXTENT_ITEM_KEY
);
1541 if (data
& BTRFS_BLOCK_GROUP_METADATA
) {
1542 last_ptr
= &root
->fs_info
->last_alloc
;
1543 empty_cluster
= 256 * 1024;
1546 if ((data
& BTRFS_BLOCK_GROUP_DATA
) && btrfs_test_opt(root
, SSD
)) {
1547 last_ptr
= &root
->fs_info
->last_data_alloc
;
1552 hint_byte
= *last_ptr
;
1554 empty_size
+= empty_cluster
;
1558 if (search_end
== (u64
)-1)
1559 search_end
= btrfs_super_total_bytes(&info
->super_copy
);
1562 block_group
= btrfs_lookup_block_group(info
, hint_byte
);
1564 hint_byte
= search_start
;
1565 block_group
= btrfs_find_block_group(root
, block_group
,
1566 hint_byte
, data
, 1);
1567 if (last_ptr
&& *last_ptr
== 0 && block_group
)
1568 hint_byte
= block_group
->key
.objectid
;
1570 block_group
= btrfs_find_block_group(root
,
1572 search_start
, data
, 1);
1574 search_start
= max(search_start
, hint_byte
);
1576 total_needed
+= empty_size
;
1580 block_group
= btrfs_lookup_block_group(info
, search_start
);
1582 block_group
= btrfs_lookup_block_group(info
,
1585 ret
= find_search_start(root
, &block_group
, &search_start
,
1586 total_needed
, data
);
1587 if (ret
== -ENOSPC
&& last_ptr
&& *last_ptr
) {
1589 block_group
= btrfs_lookup_block_group(info
,
1591 search_start
= orig_search_start
;
1592 ret
= find_search_start(root
, &block_group
, &search_start
,
1593 total_needed
, data
);
1600 if (last_ptr
&& *last_ptr
&& search_start
!= *last_ptr
) {
1603 empty_size
+= empty_cluster
;
1604 total_needed
+= empty_size
;
1606 block_group
= btrfs_lookup_block_group(info
,
1608 search_start
= orig_search_start
;
1609 ret
= find_search_start(root
, &block_group
,
1610 &search_start
, total_needed
, data
);
1617 search_start
= stripe_align(root
, search_start
);
1618 ins
->objectid
= search_start
;
1619 ins
->offset
= num_bytes
;
1621 if (ins
->objectid
+ num_bytes
>= search_end
)
1624 if (ins
->objectid
+ num_bytes
>
1625 block_group
->key
.objectid
+ block_group
->key
.offset
) {
1626 search_start
= block_group
->key
.objectid
+
1627 block_group
->key
.offset
;
1631 if (test_range_bit(&info
->extent_ins
, ins
->objectid
,
1632 ins
->objectid
+ num_bytes
-1, EXTENT_LOCKED
, 0)) {
1633 search_start
= ins
->objectid
+ num_bytes
;
1637 if (test_range_bit(&info
->pinned_extents
, ins
->objectid
,
1638 ins
->objectid
+ num_bytes
-1, EXTENT_DIRTY
, 0)) {
1639 search_start
= ins
->objectid
+ num_bytes
;
1643 if (exclude_nr
> 0 && (ins
->objectid
+ num_bytes
> exclude_start
&&
1644 ins
->objectid
< exclude_start
+ exclude_nr
)) {
1645 search_start
= exclude_start
+ exclude_nr
;
1649 if (!(data
& BTRFS_BLOCK_GROUP_DATA
)) {
1650 block_group
= btrfs_lookup_block_group(info
, ins
->objectid
);
1652 trans
->block_group
= block_group
;
1654 ins
->offset
= num_bytes
;
1656 *last_ptr
= ins
->objectid
+ ins
->offset
;
1658 btrfs_super_total_bytes(&root
->fs_info
->super_copy
)) {
1665 if (search_start
+ num_bytes
>= search_end
) {
1667 search_start
= orig_search_start
;
1674 total_needed
-= empty_size
;
1679 block_group
= btrfs_lookup_block_group(info
, search_start
);
1681 block_group
= btrfs_find_block_group(root
, block_group
,
1682 search_start
, data
, 0);
1689 * finds a free extent and does all the dirty work required for allocation
1690 * returns the key for the extent through ins, and a tree buffer for
1691 * the first block of the extent through buf.
1693 * returns 0 if everything worked, non-zero otherwise.
1695 int btrfs_alloc_extent(struct btrfs_trans_handle
*trans
,
1696 struct btrfs_root
*root
,
1697 u64 num_bytes
, u64 root_objectid
, u64 ref_generation
,
1698 u64 owner
, u64 owner_offset
,
1699 u64 empty_size
, u64 hint_byte
,
1700 u64 search_end
, struct btrfs_key
*ins
, int data
)
1706 u64 search_start
= 0;
1710 struct btrfs_fs_info
*info
= root
->fs_info
;
1711 struct btrfs_root
*extent_root
= info
->extent_root
;
1712 struct btrfs_extent_item
*extent_item
;
1713 struct btrfs_extent_ref
*ref
;
1714 struct btrfs_path
*path
;
1715 struct btrfs_key keys
[2];
1718 alloc_profile
= info
->avail_data_alloc_bits
&
1719 info
->data_alloc_profile
;
1720 data
= BTRFS_BLOCK_GROUP_DATA
| alloc_profile
;
1721 } else if (root
== root
->fs_info
->chunk_root
) {
1722 alloc_profile
= info
->avail_system_alloc_bits
&
1723 info
->system_alloc_profile
;
1724 data
= BTRFS_BLOCK_GROUP_SYSTEM
| alloc_profile
;
1726 alloc_profile
= info
->avail_metadata_alloc_bits
&
1727 info
->metadata_alloc_profile
;
1728 data
= BTRFS_BLOCK_GROUP_METADATA
| alloc_profile
;
1731 if (root
->ref_cows
) {
1732 if (!(data
& BTRFS_BLOCK_GROUP_METADATA
)) {
1733 ret
= do_chunk_alloc(trans
, root
->fs_info
->extent_root
,
1735 BTRFS_BLOCK_GROUP_METADATA
|
1736 (info
->metadata_alloc_profile
&
1737 info
->avail_metadata_alloc_bits
));
1740 ret
= do_chunk_alloc(trans
, root
->fs_info
->extent_root
,
1741 num_bytes
+ 2 * 1024 * 1024, data
);
1745 new_hint
= max(hint_byte
, root
->fs_info
->alloc_start
);
1746 if (new_hint
< btrfs_super_total_bytes(&info
->super_copy
))
1747 hint_byte
= new_hint
;
1749 WARN_ON(num_bytes
< root
->sectorsize
);
1750 ret
= find_free_extent(trans
, root
, num_bytes
, empty_size
,
1751 search_start
, search_end
, hint_byte
, ins
,
1752 trans
->alloc_exclude_start
,
1753 trans
->alloc_exclude_nr
, data
);
1758 /* block accounting for super block */
1759 super_used
= btrfs_super_bytes_used(&info
->super_copy
);
1760 btrfs_set_super_bytes_used(&info
->super_copy
, super_used
+ num_bytes
);
1762 /* block accounting for root item */
1763 root_used
= btrfs_root_used(&root
->root_item
);
1764 btrfs_set_root_used(&root
->root_item
, root_used
+ num_bytes
);
1766 clear_extent_dirty(&root
->fs_info
->free_space_cache
,
1767 ins
->objectid
, ins
->objectid
+ ins
->offset
- 1,
1770 if (root
== extent_root
) {
1771 set_extent_bits(&root
->fs_info
->extent_ins
, ins
->objectid
,
1772 ins
->objectid
+ ins
->offset
- 1,
1773 EXTENT_LOCKED
, GFP_NOFS
);
1777 WARN_ON(trans
->alloc_exclude_nr
);
1778 trans
->alloc_exclude_start
= ins
->objectid
;
1779 trans
->alloc_exclude_nr
= ins
->offset
;
1781 memcpy(&keys
[0], ins
, sizeof(*ins
));
1782 keys
[1].offset
= hash_extent_ref(root_objectid
, ref_generation
,
1783 owner
, owner_offset
);
1784 keys
[1].objectid
= ins
->objectid
;
1785 keys
[1].type
= BTRFS_EXTENT_REF_KEY
;
1786 sizes
[0] = sizeof(*extent_item
);
1787 sizes
[1] = sizeof(*ref
);
1789 path
= btrfs_alloc_path();
1792 ret
= btrfs_insert_empty_items(trans
, extent_root
, path
, keys
,
1796 extent_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
1797 struct btrfs_extent_item
);
1798 btrfs_set_extent_refs(path
->nodes
[0], extent_item
, 1);
1799 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0] + 1,
1800 struct btrfs_extent_ref
);
1802 btrfs_set_ref_root(path
->nodes
[0], ref
, root_objectid
);
1803 btrfs_set_ref_generation(path
->nodes
[0], ref
, ref_generation
);
1804 btrfs_set_ref_objectid(path
->nodes
[0], ref
, owner
);
1805 btrfs_set_ref_offset(path
->nodes
[0], ref
, owner_offset
);
1807 btrfs_mark_buffer_dirty(path
->nodes
[0]);
1809 trans
->alloc_exclude_start
= 0;
1810 trans
->alloc_exclude_nr
= 0;
1811 btrfs_free_path(path
);
1812 finish_current_insert(trans
, extent_root
);
1813 pending_ret
= del_pending_extents(trans
, extent_root
);
1823 ret
= update_block_group(trans
, root
, ins
->objectid
, ins
->offset
, 1, 0);
1825 printk("update block group failed for %Lu %Lu\n",
1826 ins
->objectid
, ins
->offset
);
1833 * helper function to allocate a block for a given tree
1834 * returns the tree buffer or NULL.
1836 struct extent_buffer
*btrfs_alloc_free_block(struct btrfs_trans_handle
*trans
,
1837 struct btrfs_root
*root
,
1839 u64 root_objectid
, u64 hint
,
1845 ref_generation
= trans
->transid
;
1850 return __btrfs_alloc_free_block(trans
, root
, blocksize
, root_objectid
,
1851 ref_generation
, 0, 0, hint
, empty_size
);
1855 * helper function to allocate a block for a given tree
1856 * returns the tree buffer or NULL.
1858 struct extent_buffer
*__btrfs_alloc_free_block(struct btrfs_trans_handle
*trans
,
1859 struct btrfs_root
*root
,
1868 struct btrfs_key ins
;
1870 struct extent_buffer
*buf
;
1872 ret
= btrfs_alloc_extent(trans
, root
, blocksize
,
1873 root_objectid
, ref_generation
,
1874 level
, first_objectid
, empty_size
, hint
,
1878 return ERR_PTR(ret
);
1880 buf
= btrfs_find_create_tree_block(root
, ins
.objectid
, blocksize
);
1882 btrfs_free_extent(trans
, root
, ins
.objectid
, blocksize
,
1883 root
->root_key
.objectid
, ref_generation
,
1885 return ERR_PTR(-ENOMEM
);
1887 btrfs_set_header_generation(buf
, trans
->transid
);
1888 clean_tree_block(trans
, root
, buf
);
1889 wait_on_tree_block_writeback(root
, buf
);
1890 btrfs_set_buffer_uptodate(buf
);
1892 if (PageDirty(buf
->first_page
)) {
1893 printk("page %lu dirty\n", buf
->first_page
->index
);
1897 set_extent_dirty(&trans
->transaction
->dirty_pages
, buf
->start
,
1898 buf
->start
+ buf
->len
- 1, GFP_NOFS
);
1899 if (!btrfs_test_opt(root
, SSD
))
1900 btrfs_set_buffer_defrag(buf
);
1901 trans
->blocks_used
++;
1905 static int noinline
drop_leaf_ref(struct btrfs_trans_handle
*trans
,
1906 struct btrfs_root
*root
,
1907 struct extent_buffer
*leaf
)
1910 u64 leaf_generation
;
1911 struct btrfs_key key
;
1912 struct btrfs_file_extent_item
*fi
;
1917 BUG_ON(!btrfs_is_leaf(leaf
));
1918 nritems
= btrfs_header_nritems(leaf
);
1919 leaf_owner
= btrfs_header_owner(leaf
);
1920 leaf_generation
= btrfs_header_generation(leaf
);
1922 for (i
= 0; i
< nritems
; i
++) {
1925 btrfs_item_key_to_cpu(leaf
, &key
, i
);
1926 if (btrfs_key_type(&key
) != BTRFS_EXTENT_DATA_KEY
)
1928 fi
= btrfs_item_ptr(leaf
, i
, struct btrfs_file_extent_item
);
1929 if (btrfs_file_extent_type(leaf
, fi
) ==
1930 BTRFS_FILE_EXTENT_INLINE
)
1933 * FIXME make sure to insert a trans record that
1934 * repeats the snapshot del on crash
1936 disk_bytenr
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
1937 if (disk_bytenr
== 0)
1939 ret
= btrfs_free_extent(trans
, root
, disk_bytenr
,
1940 btrfs_file_extent_disk_num_bytes(leaf
, fi
),
1941 leaf_owner
, leaf_generation
,
1942 key
.objectid
, key
.offset
, 0);
1948 static void noinline
reada_walk_down(struct btrfs_root
*root
,
1949 struct extent_buffer
*node
,
1962 nritems
= btrfs_header_nritems(node
);
1963 level
= btrfs_header_level(node
);
1967 for (i
= slot
; i
< nritems
&& skipped
< 32; i
++) {
1968 bytenr
= btrfs_node_blockptr(node
, i
);
1969 if (last
&& ((bytenr
> last
&& bytenr
- last
> 32 * 1024) ||
1970 (last
> bytenr
&& last
- bytenr
> 32 * 1024))) {
1974 blocksize
= btrfs_level_size(root
, level
- 1);
1976 ret
= lookup_extent_ref(NULL
, root
, bytenr
,
1984 mutex_unlock(&root
->fs_info
->fs_mutex
);
1985 ret
= readahead_tree_block(root
, bytenr
, blocksize
);
1986 last
= bytenr
+ blocksize
;
1988 mutex_lock(&root
->fs_info
->fs_mutex
);
1995 * helper function for drop_snapshot, this walks down the tree dropping ref
1996 * counts as it goes.
1998 static int noinline
walk_down_tree(struct btrfs_trans_handle
*trans
,
1999 struct btrfs_root
*root
,
2000 struct btrfs_path
*path
, int *level
)
2005 struct extent_buffer
*next
;
2006 struct extent_buffer
*cur
;
2007 struct extent_buffer
*parent
;
2012 WARN_ON(*level
< 0);
2013 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
2014 ret
= lookup_extent_ref(trans
, root
,
2015 path
->nodes
[*level
]->start
,
2016 path
->nodes
[*level
]->len
, &refs
);
2022 * walk down to the last node level and free all the leaves
2024 while(*level
>= 0) {
2025 WARN_ON(*level
< 0);
2026 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
2027 cur
= path
->nodes
[*level
];
2029 if (btrfs_header_level(cur
) != *level
)
2032 if (path
->slots
[*level
] >=
2033 btrfs_header_nritems(cur
))
2036 ret
= drop_leaf_ref(trans
, root
, cur
);
2040 bytenr
= btrfs_node_blockptr(cur
, path
->slots
[*level
]);
2041 blocksize
= btrfs_level_size(root
, *level
- 1);
2042 ret
= lookup_extent_ref(trans
, root
, bytenr
, blocksize
, &refs
);
2045 parent
= path
->nodes
[*level
];
2046 root_owner
= btrfs_header_owner(parent
);
2047 root_gen
= btrfs_header_generation(parent
);
2048 path
->slots
[*level
]++;
2049 ret
= btrfs_free_extent(trans
, root
, bytenr
,
2050 blocksize
, root_owner
,
2055 next
= btrfs_find_tree_block(root
, bytenr
, blocksize
);
2056 if (!next
|| !btrfs_buffer_uptodate(next
)) {
2057 free_extent_buffer(next
);
2058 reada_walk_down(root
, cur
, path
->slots
[*level
]);
2060 mutex_unlock(&root
->fs_info
->fs_mutex
);
2061 next
= read_tree_block(root
, bytenr
, blocksize
);
2062 mutex_lock(&root
->fs_info
->fs_mutex
);
2064 /* we've dropped the lock, double check */
2065 ret
= lookup_extent_ref(trans
, root
, bytenr
,
2069 parent
= path
->nodes
[*level
];
2070 root_owner
= btrfs_header_owner(parent
);
2071 root_gen
= btrfs_header_generation(parent
);
2073 path
->slots
[*level
]++;
2074 free_extent_buffer(next
);
2075 ret
= btrfs_free_extent(trans
, root
, bytenr
,
2083 btrfs_verify_block_csum(root
, next
);
2085 WARN_ON(*level
<= 0);
2086 if (path
->nodes
[*level
-1])
2087 free_extent_buffer(path
->nodes
[*level
-1]);
2088 path
->nodes
[*level
-1] = next
;
2089 *level
= btrfs_header_level(next
);
2090 path
->slots
[*level
] = 0;
2093 WARN_ON(*level
< 0);
2094 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
2096 if (path
->nodes
[*level
] == root
->node
) {
2097 root_owner
= root
->root_key
.objectid
;
2098 parent
= path
->nodes
[*level
];
2100 parent
= path
->nodes
[*level
+ 1];
2101 root_owner
= btrfs_header_owner(parent
);
2104 root_gen
= btrfs_header_generation(parent
);
2105 ret
= btrfs_free_extent(trans
, root
, path
->nodes
[*level
]->start
,
2106 path
->nodes
[*level
]->len
,
2107 root_owner
, root_gen
, 0, 0, 1);
2108 free_extent_buffer(path
->nodes
[*level
]);
2109 path
->nodes
[*level
] = NULL
;
2116 * helper for dropping snapshots. This walks back up the tree in the path
2117 * to find the first node higher up where we haven't yet gone through
2120 static int noinline
walk_up_tree(struct btrfs_trans_handle
*trans
,
2121 struct btrfs_root
*root
,
2122 struct btrfs_path
*path
, int *level
)
2126 struct btrfs_root_item
*root_item
= &root
->root_item
;
2131 for(i
= *level
; i
< BTRFS_MAX_LEVEL
- 1 && path
->nodes
[i
]; i
++) {
2132 slot
= path
->slots
[i
];
2133 if (slot
< btrfs_header_nritems(path
->nodes
[i
]) - 1) {
2134 struct extent_buffer
*node
;
2135 struct btrfs_disk_key disk_key
;
2136 node
= path
->nodes
[i
];
2139 WARN_ON(*level
== 0);
2140 btrfs_node_key(node
, &disk_key
, path
->slots
[i
]);
2141 memcpy(&root_item
->drop_progress
,
2142 &disk_key
, sizeof(disk_key
));
2143 root_item
->drop_level
= i
;
2146 if (path
->nodes
[*level
] == root
->node
) {
2147 root_owner
= root
->root_key
.objectid
;
2149 btrfs_header_generation(path
->nodes
[*level
]);
2151 struct extent_buffer
*node
;
2152 node
= path
->nodes
[*level
+ 1];
2153 root_owner
= btrfs_header_owner(node
);
2154 root_gen
= btrfs_header_generation(node
);
2156 ret
= btrfs_free_extent(trans
, root
,
2157 path
->nodes
[*level
]->start
,
2158 path
->nodes
[*level
]->len
,
2159 root_owner
, root_gen
, 0, 0, 1);
2161 free_extent_buffer(path
->nodes
[*level
]);
2162 path
->nodes
[*level
] = NULL
;
2170 * drop the reference count on the tree rooted at 'snap'. This traverses
2171 * the tree freeing any blocks that have a ref count of zero after being
2174 int btrfs_drop_snapshot(struct btrfs_trans_handle
*trans
, struct btrfs_root
2180 struct btrfs_path
*path
;
2183 struct btrfs_root_item
*root_item
= &root
->root_item
;
2185 path
= btrfs_alloc_path();
2188 level
= btrfs_header_level(root
->node
);
2190 if (btrfs_disk_key_objectid(&root_item
->drop_progress
) == 0) {
2191 path
->nodes
[level
] = root
->node
;
2192 extent_buffer_get(root
->node
);
2193 path
->slots
[level
] = 0;
2195 struct btrfs_key key
;
2196 struct btrfs_disk_key found_key
;
2197 struct extent_buffer
*node
;
2199 btrfs_disk_key_to_cpu(&key
, &root_item
->drop_progress
);
2200 level
= root_item
->drop_level
;
2201 path
->lowest_level
= level
;
2202 wret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2207 node
= path
->nodes
[level
];
2208 btrfs_node_key(node
, &found_key
, path
->slots
[level
]);
2209 WARN_ON(memcmp(&found_key
, &root_item
->drop_progress
,
2210 sizeof(found_key
)));
2213 wret
= walk_down_tree(trans
, root
, path
, &level
);
2219 wret
= walk_up_tree(trans
, root
, path
, &level
);
2227 for (i
= 0; i
<= orig_level
; i
++) {
2228 if (path
->nodes
[i
]) {
2229 free_extent_buffer(path
->nodes
[i
]);
2230 path
->nodes
[i
] = NULL
;
2234 btrfs_free_path(path
);
2238 int btrfs_free_block_groups(struct btrfs_fs_info
*info
)
2245 ret
= find_first_extent_bit(&info
->block_group_cache
, 0,
2246 &start
, &end
, (unsigned int)-1);
2249 ret
= get_state_private(&info
->block_group_cache
, start
, &ptr
);
2251 kfree((void *)(unsigned long)ptr
);
2252 clear_extent_bits(&info
->block_group_cache
, start
,
2253 end
, (unsigned int)-1, GFP_NOFS
);
2256 ret
= find_first_extent_bit(&info
->free_space_cache
, 0,
2257 &start
, &end
, EXTENT_DIRTY
);
2260 clear_extent_dirty(&info
->free_space_cache
, start
,
2266 static int noinline
relocate_inode_pages(struct inode
*inode
, u64 start
,
2272 u64 existing_delalloc
;
2273 unsigned long last_index
;
2276 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2277 struct file_ra_state
*ra
;
2279 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
2281 mutex_lock(&inode
->i_mutex
);
2282 i
= start
>> PAGE_CACHE_SHIFT
;
2283 last_index
= (start
+ len
- 1) >> PAGE_CACHE_SHIFT
;
2285 file_ra_state_init(ra
, inode
->i_mapping
);
2286 btrfs_force_ra(inode
->i_mapping
, ra
, NULL
, i
, last_index
);
2289 for (; i
<= last_index
; i
++) {
2290 page
= grab_cache_page(inode
->i_mapping
, i
);
2293 if (!PageUptodate(page
)) {
2294 btrfs_readpage(NULL
, page
);
2296 if (!PageUptodate(page
)) {
2298 page_cache_release(page
);
2302 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
2303 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
2305 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
2307 delalloc_start
= page_start
;
2308 existing_delalloc
= count_range_bits(io_tree
,
2309 &delalloc_start
, page_end
,
2310 PAGE_CACHE_SIZE
, EXTENT_DELALLOC
);
2312 set_extent_delalloc(io_tree
, page_start
,
2313 page_end
, GFP_NOFS
);
2315 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
2316 set_page_dirty(page
);
2318 page_cache_release(page
);
2322 mutex_unlock(&inode
->i_mutex
);
2327 * note, this releases the path
2329 static int noinline
relocate_one_reference(struct btrfs_root
*extent_root
,
2330 struct btrfs_path
*path
,
2331 struct btrfs_key
*extent_key
)
2333 struct inode
*inode
;
2334 struct btrfs_root
*found_root
;
2335 struct btrfs_key
*root_location
;
2336 struct btrfs_extent_ref
*ref
;
2343 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
2344 struct btrfs_extent_ref
);
2345 ref_root
= btrfs_ref_root(path
->nodes
[0], ref
);
2346 ref_gen
= btrfs_ref_generation(path
->nodes
[0], ref
);
2347 ref_objectid
= btrfs_ref_objectid(path
->nodes
[0], ref
);
2348 ref_offset
= btrfs_ref_offset(path
->nodes
[0], ref
);
2349 btrfs_release_path(extent_root
, path
);
2351 root_location
= kmalloc(sizeof(*root_location
), GFP_NOFS
);
2352 root_location
->objectid
= ref_root
;
2354 root_location
->offset
= 0;
2356 root_location
->offset
= (u64
)-1;
2357 root_location
->type
= BTRFS_ROOT_ITEM_KEY
;
2359 found_root
= btrfs_read_fs_root_no_name(extent_root
->fs_info
,
2361 BUG_ON(!found_root
);
2362 kfree(root_location
);
2364 if (ref_objectid
>= BTRFS_FIRST_FREE_OBJECTID
) {
2365 mutex_unlock(&extent_root
->fs_info
->fs_mutex
);
2366 inode
= btrfs_iget_locked(extent_root
->fs_info
->sb
,
2367 ref_objectid
, found_root
);
2368 if (inode
->i_state
& I_NEW
) {
2369 /* the inode and parent dir are two different roots */
2370 BTRFS_I(inode
)->root
= found_root
;
2371 BTRFS_I(inode
)->location
.objectid
= ref_objectid
;
2372 BTRFS_I(inode
)->location
.type
= BTRFS_INODE_ITEM_KEY
;
2373 BTRFS_I(inode
)->location
.offset
= 0;
2374 btrfs_read_locked_inode(inode
);
2375 unlock_new_inode(inode
);
2378 /* this can happen if the reference is not against
2379 * the latest version of the tree root
2381 if (is_bad_inode(inode
)) {
2382 mutex_lock(&extent_root
->fs_info
->fs_mutex
);
2385 relocate_inode_pages(inode
, ref_offset
, extent_key
->offset
);
2386 /* FIXME, data=ordered will help get rid of this */
2387 filemap_fdatawrite(inode
->i_mapping
);
2389 mutex_lock(&extent_root
->fs_info
->fs_mutex
);
2391 struct btrfs_trans_handle
*trans
;
2392 struct btrfs_key found_key
;
2393 struct extent_buffer
*eb
;
2397 trans
= btrfs_start_transaction(found_root
, 1);
2398 eb
= read_tree_block(found_root
, extent_key
->objectid
,
2399 extent_key
->offset
);
2400 level
= btrfs_header_level(eb
);
2403 btrfs_item_key_to_cpu(eb
, &found_key
, 0);
2405 btrfs_node_key_to_cpu(eb
, &found_key
, 0);
2407 free_extent_buffer(eb
);
2409 path
->lowest_level
= level
;
2411 ret
= btrfs_search_slot(trans
, found_root
, &found_key
, path
,
2413 path
->lowest_level
= 0;
2414 for (i
= level
; i
< BTRFS_MAX_LEVEL
; i
++) {
2415 if (!path
->nodes
[i
])
2417 free_extent_buffer(path
->nodes
[i
]);
2418 path
->nodes
[i
] = NULL
;
2420 btrfs_release_path(found_root
, path
);
2421 btrfs_end_transaction(trans
, found_root
);
2428 static int noinline
relocate_one_extent(struct btrfs_root
*extent_root
,
2429 struct btrfs_path
*path
,
2430 struct btrfs_key
*extent_key
)
2432 struct btrfs_key key
;
2433 struct btrfs_key found_key
;
2434 struct extent_buffer
*leaf
;
2439 key
.objectid
= extent_key
->objectid
;
2440 key
.type
= BTRFS_EXTENT_REF_KEY
;
2444 ret
= btrfs_search_slot(NULL
, extent_root
, &key
, path
, 0, 0);
2450 leaf
= path
->nodes
[0];
2451 nritems
= btrfs_header_nritems(leaf
);
2452 if (path
->slots
[0] == nritems
)
2455 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2456 if (found_key
.objectid
!= extent_key
->objectid
)
2459 if (found_key
.type
!= BTRFS_EXTENT_REF_KEY
)
2462 key
.offset
= found_key
.offset
+ 1;
2463 item_size
= btrfs_item_size_nr(leaf
, path
->slots
[0]);
2465 ret
= relocate_one_reference(extent_root
, path
, extent_key
);
2471 btrfs_release_path(extent_root
, path
);
2475 int btrfs_shrink_extent_tree(struct btrfs_root
*root
, u64 new_size
)
2477 struct btrfs_trans_handle
*trans
;
2478 struct btrfs_root
*tree_root
= root
->fs_info
->tree_root
;
2479 struct btrfs_path
*path
;
2482 struct btrfs_fs_info
*info
= root
->fs_info
;
2483 struct extent_io_tree
*block_group_cache
;
2484 struct btrfs_key key
;
2485 struct btrfs_key found_key
;
2486 struct extent_buffer
*leaf
;
2491 btrfs_set_super_total_bytes(&info
->super_copy
, new_size
);
2492 clear_extent_dirty(&info
->free_space_cache
, new_size
, (u64
)-1,
2494 block_group_cache
= &info
->block_group_cache
;
2495 path
= btrfs_alloc_path();
2496 root
= root
->fs_info
->extent_root
;
2501 key
.objectid
= new_size
;
2504 cur_byte
= key
.objectid
;
2506 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2510 ret
= btrfs_previous_item(root
, path
, 0, BTRFS_EXTENT_ITEM_KEY
);
2514 leaf
= path
->nodes
[0];
2515 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2516 if (found_key
.objectid
+ found_key
.offset
> new_size
) {
2517 cur_byte
= found_key
.objectid
;
2518 key
.objectid
= cur_byte
;
2521 btrfs_release_path(root
, path
);
2524 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2528 leaf
= path
->nodes
[0];
2529 nritems
= btrfs_header_nritems(leaf
);
2531 if (path
->slots
[0] >= nritems
) {
2532 ret
= btrfs_next_leaf(root
, path
);
2539 leaf
= path
->nodes
[0];
2540 nritems
= btrfs_header_nritems(leaf
);
2543 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2545 if (progress
&& need_resched()) {
2546 memcpy(&key
, &found_key
, sizeof(key
));
2547 mutex_unlock(&root
->fs_info
->fs_mutex
);
2549 mutex_lock(&root
->fs_info
->fs_mutex
);
2550 btrfs_release_path(root
, path
);
2551 btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2557 if (btrfs_key_type(&found_key
) != BTRFS_EXTENT_ITEM_KEY
||
2558 found_key
.objectid
+ found_key
.offset
<= cur_byte
) {
2564 cur_byte
= found_key
.objectid
+ found_key
.offset
;
2565 key
.objectid
= cur_byte
;
2566 btrfs_release_path(root
, path
);
2567 ret
= relocate_one_extent(root
, path
, &found_key
);
2570 btrfs_release_path(root
, path
);
2572 if (total_found
> 0) {
2573 trans
= btrfs_start_transaction(tree_root
, 1);
2574 btrfs_commit_transaction(trans
, tree_root
);
2576 mutex_unlock(&root
->fs_info
->fs_mutex
);
2577 btrfs_clean_old_snapshots(tree_root
);
2578 mutex_lock(&root
->fs_info
->fs_mutex
);
2580 trans
= btrfs_start_transaction(tree_root
, 1);
2581 btrfs_commit_transaction(trans
, tree_root
);
2585 trans
= btrfs_start_transaction(root
, 1);
2586 key
.objectid
= new_size
;
2592 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
2596 leaf
= path
->nodes
[0];
2597 nritems
= btrfs_header_nritems(leaf
);
2599 if (path
->slots
[0] >= nritems
) {
2600 ret
= btrfs_next_leaf(root
, path
);
2607 leaf
= path
->nodes
[0];
2608 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2611 * btrfs_next_leaf doesn't cow buffers, we have to
2612 * do the search again
2614 memcpy(&key
, &found_key
, sizeof(key
));
2615 btrfs_release_path(root
, path
);
2619 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2620 if (btrfs_key_type(&found_key
) != BTRFS_BLOCK_GROUP_ITEM_KEY
) {
2621 printk("shrinker found key %Lu %u %Lu\n",
2622 found_key
.objectid
, found_key
.type
,
2627 ret
= get_state_private(&info
->block_group_cache
,
2628 found_key
.objectid
, &ptr
);
2630 kfree((void *)(unsigned long)ptr
);
2632 clear_extent_bits(&info
->block_group_cache
, found_key
.objectid
,
2633 found_key
.objectid
+ found_key
.offset
- 1,
2634 (unsigned int)-1, GFP_NOFS
);
2636 key
.objectid
= found_key
.objectid
+ 1;
2637 btrfs_del_item(trans
, root
, path
);
2638 btrfs_release_path(root
, path
);
2640 if (need_resched()) {
2641 mutex_unlock(&root
->fs_info
->fs_mutex
);
2643 mutex_lock(&root
->fs_info
->fs_mutex
);
2646 clear_extent_dirty(&info
->free_space_cache
, new_size
, (u64
)-1,
2648 btrfs_commit_transaction(trans
, root
);
2650 btrfs_free_path(path
);
2654 int btrfs_grow_extent_tree(struct btrfs_trans_handle
*trans
,
2655 struct btrfs_root
*root
, u64 new_size
)
2657 btrfs_set_super_total_bytes(&root
->fs_info
->super_copy
, new_size
);
2661 int find_first_block_group(struct btrfs_root
*root
, struct btrfs_path
*path
,
2662 struct btrfs_key
*key
)
2665 struct btrfs_key found_key
;
2666 struct extent_buffer
*leaf
;
2669 ret
= btrfs_search_slot(NULL
, root
, key
, path
, 0, 0);
2673 slot
= path
->slots
[0];
2674 leaf
= path
->nodes
[0];
2675 if (slot
>= btrfs_header_nritems(leaf
)) {
2676 ret
= btrfs_next_leaf(root
, path
);
2683 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
2685 if (found_key
.objectid
>= key
->objectid
&&
2686 found_key
.type
== BTRFS_BLOCK_GROUP_ITEM_KEY
)
2695 int btrfs_read_block_groups(struct btrfs_root
*root
)
2697 struct btrfs_path
*path
;
2700 struct btrfs_block_group_cache
*cache
;
2701 struct btrfs_fs_info
*info
= root
->fs_info
;
2702 struct btrfs_space_info
*space_info
;
2703 struct extent_io_tree
*block_group_cache
;
2704 struct btrfs_key key
;
2705 struct btrfs_key found_key
;
2706 struct extent_buffer
*leaf
;
2708 block_group_cache
= &info
->block_group_cache
;
2709 root
= info
->extent_root
;
2712 btrfs_set_key_type(&key
, BTRFS_BLOCK_GROUP_ITEM_KEY
);
2713 path
= btrfs_alloc_path();
2718 ret
= find_first_block_group(root
, path
, &key
);
2726 leaf
= path
->nodes
[0];
2727 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2728 cache
= kmalloc(sizeof(*cache
), GFP_NOFS
);
2734 read_extent_buffer(leaf
, &cache
->item
,
2735 btrfs_item_ptr_offset(leaf
, path
->slots
[0]),
2736 sizeof(cache
->item
));
2737 memcpy(&cache
->key
, &found_key
, sizeof(found_key
));
2741 key
.objectid
= found_key
.objectid
+ found_key
.offset
;
2742 btrfs_release_path(root
, path
);
2743 cache
->flags
= btrfs_block_group_flags(&cache
->item
);
2745 if (cache
->flags
& BTRFS_BLOCK_GROUP_DATA
) {
2746 bit
= BLOCK_GROUP_DATA
;
2747 } else if (cache
->flags
& BTRFS_BLOCK_GROUP_SYSTEM
) {
2748 bit
= BLOCK_GROUP_SYSTEM
;
2749 } else if (cache
->flags
& BTRFS_BLOCK_GROUP_METADATA
) {
2750 bit
= BLOCK_GROUP_METADATA
;
2752 set_avail_alloc_bits(info
, cache
->flags
);
2754 ret
= update_space_info(info
, cache
->flags
, found_key
.offset
,
2755 btrfs_block_group_used(&cache
->item
),
2758 cache
->space_info
= space_info
;
2760 /* use EXTENT_LOCKED to prevent merging */
2761 set_extent_bits(block_group_cache
, found_key
.objectid
,
2762 found_key
.objectid
+ found_key
.offset
- 1,
2763 bit
| EXTENT_LOCKED
, GFP_NOFS
);
2764 set_state_private(block_group_cache
, found_key
.objectid
,
2765 (unsigned long)cache
);
2768 btrfs_super_total_bytes(&info
->super_copy
))
2773 btrfs_free_path(path
);
2777 int btrfs_make_block_group(struct btrfs_trans_handle
*trans
,
2778 struct btrfs_root
*root
, u64 bytes_used
,
2779 u64 type
, u64 chunk_tree
, u64 chunk_objectid
,
2784 struct btrfs_root
*extent_root
;
2785 struct btrfs_block_group_cache
*cache
;
2786 struct extent_io_tree
*block_group_cache
;
2788 extent_root
= root
->fs_info
->extent_root
;
2789 block_group_cache
= &root
->fs_info
->block_group_cache
;
2791 cache
= kmalloc(sizeof(*cache
), GFP_NOFS
);
2793 cache
->key
.objectid
= chunk_objectid
;
2794 cache
->key
.offset
= size
;
2797 btrfs_set_key_type(&cache
->key
, BTRFS_BLOCK_GROUP_ITEM_KEY
);
2798 memset(&cache
->item
, 0, sizeof(cache
->item
));
2799 btrfs_set_block_group_used(&cache
->item
, bytes_used
);
2800 btrfs_set_block_group_chunk_tree(&cache
->item
, chunk_tree
);
2801 btrfs_set_block_group_chunk_objectid(&cache
->item
, chunk_objectid
);
2802 cache
->flags
= type
;
2803 btrfs_set_block_group_flags(&cache
->item
, type
);
2805 ret
= update_space_info(root
->fs_info
, cache
->flags
, size
, bytes_used
,
2806 &cache
->space_info
);
2809 bit
= block_group_state_bits(type
);
2810 set_extent_bits(block_group_cache
, chunk_objectid
,
2811 chunk_objectid
+ size
- 1,
2812 bit
| EXTENT_LOCKED
, GFP_NOFS
);
2813 set_state_private(block_group_cache
, chunk_objectid
,
2814 (unsigned long)cache
);
2816 ret
= btrfs_insert_item(trans
, extent_root
, &cache
->key
, &cache
->item
,
2817 sizeof(cache
->item
));
2820 finish_current_insert(trans
, extent_root
);
2821 ret
= del_pending_extents(trans
, extent_root
);
2823 set_avail_alloc_bits(extent_root
->fs_info
, type
);