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
)
233 if (!block_group_bits(cache
, data
)) {
234 printk("block group bits don't match %Lu %d\n", cache
->flags
, data
);
240 cache
= btrfs_lookup_block_group(root
->fs_info
, search_start
);
242 printk("Unable to find block group for %Lu\n", search_start
);
248 last
= cache
->key
.objectid
+ cache
->key
.offset
;
250 cache
= btrfs_lookup_block_group(root
->fs_info
, last
);
251 if (!cache
|| cache
->key
.objectid
>= total_fs_bytes
) {
260 if (cache_miss
&& !cache
->cached
) {
261 cache_block_group(root
, cache
);
263 cache
= btrfs_lookup_block_group(root
->fs_info
, last
);
265 cache
= btrfs_find_block_group(root
, cache
, last
, data
, 0);
273 static u64
div_factor(u64 num
, int factor
)
282 static int block_group_state_bits(u64 flags
)
285 if (flags
& BTRFS_BLOCK_GROUP_DATA
)
286 bits
|= BLOCK_GROUP_DATA
;
287 if (flags
& BTRFS_BLOCK_GROUP_METADATA
)
288 bits
|= BLOCK_GROUP_METADATA
;
289 if (flags
& BTRFS_BLOCK_GROUP_SYSTEM
)
290 bits
|= BLOCK_GROUP_SYSTEM
;
294 struct btrfs_block_group_cache
*btrfs_find_block_group(struct btrfs_root
*root
,
295 struct btrfs_block_group_cache
296 *hint
, u64 search_start
,
299 struct btrfs_block_group_cache
*cache
;
300 struct extent_io_tree
*block_group_cache
;
301 struct btrfs_block_group_cache
*found_group
= NULL
;
302 struct btrfs_fs_info
*info
= root
->fs_info
;
316 block_group_cache
= &info
->block_group_cache
;
317 total_fs_bytes
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
322 bit
= block_group_state_bits(data
);
324 if (search_start
&& search_start
< total_fs_bytes
) {
325 struct btrfs_block_group_cache
*shint
;
326 shint
= btrfs_lookup_block_group(info
, search_start
);
327 if (shint
&& block_group_bits(shint
, data
)) {
328 used
= btrfs_block_group_used(&shint
->item
);
329 if (used
+ shint
->pinned
<
330 div_factor(shint
->key
.offset
, factor
)) {
335 if (hint
&& block_group_bits(hint
, data
) &&
336 hint
->key
.objectid
< total_fs_bytes
) {
337 used
= btrfs_block_group_used(&hint
->item
);
338 if (used
+ hint
->pinned
<
339 div_factor(hint
->key
.offset
, factor
)) {
342 last
= hint
->key
.objectid
+ hint
->key
.offset
;
346 hint_last
= max(hint
->key
.objectid
, search_start
);
348 hint_last
= search_start
;
350 if (hint_last
>= total_fs_bytes
)
351 hint_last
= search_start
;
356 ret
= find_first_extent_bit(block_group_cache
, last
,
361 ret
= get_state_private(block_group_cache
, start
, &ptr
);
365 cache
= (struct btrfs_block_group_cache
*)(unsigned long)ptr
;
366 last
= cache
->key
.objectid
+ cache
->key
.offset
;
367 used
= btrfs_block_group_used(&cache
->item
);
369 if (cache
->key
.objectid
> total_fs_bytes
)
372 if (block_group_bits(cache
, data
)) {
374 free_check
= cache
->key
.offset
;
376 free_check
= div_factor(cache
->key
.offset
,
379 if (used
+ cache
->pinned
< free_check
) {
395 static u64
hash_extent_ref(u64 root_objectid
, u64 ref_generation
,
396 u64 owner
, u64 owner_offset
)
398 u32 high_crc
= ~(u32
)0;
399 u32 low_crc
= ~(u32
)0;
402 lenum
= cpu_to_le64(root_objectid
);
403 high_crc
= crc32c(high_crc
, &lenum
, sizeof(lenum
));
404 lenum
= cpu_to_le64(ref_generation
);
405 low_crc
= crc32c(low_crc
, &lenum
, sizeof(lenum
));
406 if (owner
>= BTRFS_FIRST_FREE_OBJECTID
) {
407 lenum
= cpu_to_le64(owner
);
408 low_crc
= crc32c(low_crc
, &lenum
, sizeof(lenum
));
409 lenum
= cpu_to_le64(owner_offset
);
410 low_crc
= crc32c(low_crc
, &lenum
, sizeof(lenum
));
412 return ((u64
)high_crc
<< 32) | (u64
)low_crc
;
415 static int match_extent_ref(struct extent_buffer
*leaf
,
416 struct btrfs_extent_ref
*disk_ref
,
417 struct btrfs_extent_ref
*cpu_ref
)
422 if (cpu_ref
->objectid
)
423 len
= sizeof(*cpu_ref
);
425 len
= 2 * sizeof(u64
);
426 ret
= memcmp_extent_buffer(leaf
, cpu_ref
, (unsigned long)disk_ref
,
431 static int noinline
lookup_extent_backref(struct btrfs_trans_handle
*trans
,
432 struct btrfs_root
*root
,
433 struct btrfs_path
*path
, u64 bytenr
,
435 u64 ref_generation
, u64 owner
,
436 u64 owner_offset
, int del
)
439 struct btrfs_key key
;
440 struct btrfs_key found_key
;
441 struct btrfs_extent_ref ref
;
442 struct extent_buffer
*leaf
;
443 struct btrfs_extent_ref
*disk_ref
;
447 btrfs_set_stack_ref_root(&ref
, root_objectid
);
448 btrfs_set_stack_ref_generation(&ref
, ref_generation
);
449 btrfs_set_stack_ref_objectid(&ref
, owner
);
450 btrfs_set_stack_ref_offset(&ref
, owner_offset
);
452 hash
= hash_extent_ref(root_objectid
, ref_generation
, owner
,
455 key
.objectid
= bytenr
;
456 key
.type
= BTRFS_EXTENT_REF_KEY
;
459 ret
= btrfs_search_slot(trans
, root
, &key
, path
,
463 leaf
= path
->nodes
[0];
465 u32 nritems
= btrfs_header_nritems(leaf
);
466 if (path
->slots
[0] >= nritems
) {
467 ret2
= btrfs_next_leaf(root
, path
);
470 leaf
= path
->nodes
[0];
472 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
473 if (found_key
.objectid
!= bytenr
||
474 found_key
.type
!= BTRFS_EXTENT_REF_KEY
)
476 key
.offset
= found_key
.offset
;
478 btrfs_release_path(root
, path
);
482 disk_ref
= btrfs_item_ptr(path
->nodes
[0],
484 struct btrfs_extent_ref
);
485 if (match_extent_ref(path
->nodes
[0], disk_ref
, &ref
)) {
489 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
490 key
.offset
= found_key
.offset
+ 1;
491 btrfs_release_path(root
, path
);
498 * Back reference rules. Back refs have three main goals:
500 * 1) differentiate between all holders of references to an extent so that
501 * when a reference is dropped we can make sure it was a valid reference
502 * before freeing the extent.
504 * 2) Provide enough information to quickly find the holders of an extent
505 * if we notice a given block is corrupted or bad.
507 * 3) Make it easy to migrate blocks for FS shrinking or storage pool
508 * maintenance. This is actually the same as #2, but with a slightly
509 * different use case.
511 * File extents can be referenced by:
513 * - multiple snapshots, subvolumes, or different generations in one subvol
514 * - different files inside a single subvolume (in theory, not implemented yet)
515 * - different offsets inside a file (bookend extents in file.c)
517 * The extent ref structure has fields for:
519 * - Objectid of the subvolume root
520 * - Generation number of the tree holding the reference
521 * - objectid of the file holding the reference
522 * - offset in the file corresponding to the key holding the reference
524 * When a file extent is allocated the fields are filled in:
525 * (root_key.objectid, trans->transid, inode objectid, offset in file)
527 * When a leaf is cow'd new references are added for every file extent found
528 * in the leaf. It looks the same as the create case, but trans->transid
529 * will be different when the block is cow'd.
531 * (root_key.objectid, trans->transid, inode objectid, offset in file)
533 * When a file extent is removed either during snapshot deletion or file
534 * truncation, the corresponding back reference is found
537 * (btrfs_header_owner(leaf), btrfs_header_generation(leaf),
538 * inode objectid, offset in file)
540 * Btree extents can be referenced by:
542 * - Different subvolumes
543 * - Different generations of the same subvolume
545 * Storing sufficient information for a full reverse mapping of a btree
546 * block would require storing the lowest key of the block in the backref,
547 * and it would require updating that lowest key either before write out or
548 * every time it changed. Instead, the objectid of the lowest key is stored
549 * along with the level of the tree block. This provides a hint
550 * about where in the btree the block can be found. Searches through the
551 * btree only need to look for a pointer to that block, so they stop one
552 * level higher than the level recorded in the backref.
554 * Some btrees do not do reference counting on their extents. These
555 * include the extent tree and the tree of tree roots. Backrefs for these
556 * trees always have a generation of zero.
558 * When a tree block is created, back references are inserted:
560 * (root->root_key.objectid, trans->transid or zero, level, lowest_key_objectid)
562 * When a tree block is cow'd in a reference counted root,
563 * new back references are added for all the blocks it points to.
564 * These are of the form (trans->transid will have increased since creation):
566 * (root->root_key.objectid, trans->transid, level, lowest_key_objectid)
568 * Because the lowest_key_objectid and the level are just hints
569 * they are not used when backrefs are deleted. When a backref is deleted:
571 * if backref was for a tree root:
572 * root_objectid = root->root_key.objectid
574 * root_objectid = btrfs_header_owner(parent)
576 * (root_objectid, btrfs_header_generation(parent) or zero, 0, 0)
578 * Back Reference Key hashing:
580 * Back references have four fields, each 64 bits long. Unfortunately,
581 * This is hashed into a single 64 bit number and placed into the key offset.
582 * The key objectid corresponds to the first byte in the extent, and the
583 * key type is set to BTRFS_EXTENT_REF_KEY
585 int btrfs_insert_extent_backref(struct btrfs_trans_handle
*trans
,
586 struct btrfs_root
*root
,
587 struct btrfs_path
*path
, u64 bytenr
,
588 u64 root_objectid
, u64 ref_generation
,
589 u64 owner
, u64 owner_offset
)
592 struct btrfs_key key
;
593 struct btrfs_extent_ref ref
;
594 struct btrfs_extent_ref
*disk_ref
;
597 btrfs_set_stack_ref_root(&ref
, root_objectid
);
598 btrfs_set_stack_ref_generation(&ref
, ref_generation
);
599 btrfs_set_stack_ref_objectid(&ref
, owner
);
600 btrfs_set_stack_ref_offset(&ref
, owner_offset
);
602 hash
= hash_extent_ref(root_objectid
, ref_generation
, owner
,
605 key
.objectid
= bytenr
;
606 key
.type
= BTRFS_EXTENT_REF_KEY
;
608 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
, sizeof(ref
));
609 while (ret
== -EEXIST
) {
610 disk_ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
611 struct btrfs_extent_ref
);
612 if (match_extent_ref(path
->nodes
[0], disk_ref
, &ref
))
615 btrfs_release_path(root
, path
);
616 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
621 disk_ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
622 struct btrfs_extent_ref
);
623 write_extent_buffer(path
->nodes
[0], &ref
, (unsigned long)disk_ref
,
625 btrfs_mark_buffer_dirty(path
->nodes
[0]);
627 btrfs_release_path(root
, path
);
631 int btrfs_inc_extent_ref(struct btrfs_trans_handle
*trans
,
632 struct btrfs_root
*root
,
633 u64 bytenr
, u64 num_bytes
,
634 u64 root_objectid
, u64 ref_generation
,
635 u64 owner
, u64 owner_offset
)
637 struct btrfs_path
*path
;
639 struct btrfs_key key
;
640 struct extent_buffer
*l
;
641 struct btrfs_extent_item
*item
;
644 WARN_ON(num_bytes
< root
->sectorsize
);
645 path
= btrfs_alloc_path();
650 key
.objectid
= bytenr
;
651 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
652 key
.offset
= num_bytes
;
653 ret
= btrfs_search_slot(trans
, root
->fs_info
->extent_root
, &key
, path
,
662 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
663 refs
= btrfs_extent_refs(l
, item
);
664 btrfs_set_extent_refs(l
, item
, refs
+ 1);
665 btrfs_mark_buffer_dirty(path
->nodes
[0]);
667 btrfs_release_path(root
->fs_info
->extent_root
, path
);
670 ret
= btrfs_insert_extent_backref(trans
, root
->fs_info
->extent_root
,
671 path
, bytenr
, root_objectid
,
672 ref_generation
, owner
, owner_offset
);
674 finish_current_insert(trans
, root
->fs_info
->extent_root
);
675 del_pending_extents(trans
, root
->fs_info
->extent_root
);
677 btrfs_free_path(path
);
681 int btrfs_extent_post_op(struct btrfs_trans_handle
*trans
,
682 struct btrfs_root
*root
)
684 finish_current_insert(trans
, root
->fs_info
->extent_root
);
685 del_pending_extents(trans
, root
->fs_info
->extent_root
);
689 static int lookup_extent_ref(struct btrfs_trans_handle
*trans
,
690 struct btrfs_root
*root
, u64 bytenr
,
691 u64 num_bytes
, u32
*refs
)
693 struct btrfs_path
*path
;
695 struct btrfs_key key
;
696 struct extent_buffer
*l
;
697 struct btrfs_extent_item
*item
;
699 WARN_ON(num_bytes
< root
->sectorsize
);
700 path
= btrfs_alloc_path();
702 key
.objectid
= bytenr
;
703 key
.offset
= num_bytes
;
704 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
705 ret
= btrfs_search_slot(trans
, root
->fs_info
->extent_root
, &key
, path
,
710 btrfs_print_leaf(root
, path
->nodes
[0]);
711 printk("failed to find block number %Lu\n", bytenr
);
715 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
716 *refs
= btrfs_extent_refs(l
, item
);
718 btrfs_free_path(path
);
722 u32
btrfs_count_snapshots_in_path(struct btrfs_root
*root
,
723 struct btrfs_path
*count_path
,
726 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
727 struct btrfs_path
*path
;
730 u64 root_objectid
= root
->root_key
.objectid
;
735 struct btrfs_key key
;
736 struct btrfs_key found_key
;
737 struct extent_buffer
*l
;
738 struct btrfs_extent_item
*item
;
739 struct btrfs_extent_ref
*ref_item
;
742 path
= btrfs_alloc_path();
745 bytenr
= first_extent
;
747 bytenr
= count_path
->nodes
[level
]->start
;
750 key
.objectid
= bytenr
;
753 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
754 ret
= btrfs_search_slot(NULL
, extent_root
, &key
, path
, 0, 0);
760 btrfs_item_key_to_cpu(l
, &found_key
, path
->slots
[0]);
762 if (found_key
.objectid
!= bytenr
||
763 found_key
.type
!= BTRFS_EXTENT_ITEM_KEY
) {
767 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
770 nritems
= btrfs_header_nritems(l
);
771 if (path
->slots
[0] >= nritems
) {
772 ret
= btrfs_next_leaf(extent_root
, path
);
777 btrfs_item_key_to_cpu(l
, &found_key
, path
->slots
[0]);
778 if (found_key
.objectid
!= bytenr
)
781 if (found_key
.type
!= BTRFS_EXTENT_REF_KEY
) {
787 ref_item
= btrfs_item_ptr(l
, path
->slots
[0],
788 struct btrfs_extent_ref
);
789 found_objectid
= btrfs_ref_root(l
, ref_item
);
791 if (found_objectid
!= root_objectid
) {
798 if (cur_count
== 0) {
802 if (level
>= 0 && root
->node
== count_path
->nodes
[level
])
805 btrfs_release_path(root
, path
);
809 btrfs_free_path(path
);
812 int btrfs_inc_root_ref(struct btrfs_trans_handle
*trans
,
813 struct btrfs_root
*root
, u64 owner_objectid
)
819 struct btrfs_disk_key disk_key
;
821 level
= btrfs_header_level(root
->node
);
822 generation
= trans
->transid
;
823 nritems
= btrfs_header_nritems(root
->node
);
826 btrfs_item_key(root
->node
, &disk_key
, 0);
828 btrfs_node_key(root
->node
, &disk_key
, 0);
829 key_objectid
= btrfs_disk_key_objectid(&disk_key
);
833 return btrfs_inc_extent_ref(trans
, root
, root
->node
->start
,
834 root
->node
->len
, owner_objectid
,
835 generation
, level
, key_objectid
);
838 int btrfs_inc_ref(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
839 struct extent_buffer
*buf
)
843 struct btrfs_key key
;
844 struct btrfs_file_extent_item
*fi
;
853 level
= btrfs_header_level(buf
);
854 nritems
= btrfs_header_nritems(buf
);
855 for (i
= 0; i
< nritems
; i
++) {
858 btrfs_item_key_to_cpu(buf
, &key
, i
);
859 if (btrfs_key_type(&key
) != BTRFS_EXTENT_DATA_KEY
)
861 fi
= btrfs_item_ptr(buf
, i
,
862 struct btrfs_file_extent_item
);
863 if (btrfs_file_extent_type(buf
, fi
) ==
864 BTRFS_FILE_EXTENT_INLINE
)
866 disk_bytenr
= btrfs_file_extent_disk_bytenr(buf
, fi
);
867 if (disk_bytenr
== 0)
869 ret
= btrfs_inc_extent_ref(trans
, root
, disk_bytenr
,
870 btrfs_file_extent_disk_num_bytes(buf
, fi
),
871 root
->root_key
.objectid
, trans
->transid
,
872 key
.objectid
, key
.offset
);
878 bytenr
= btrfs_node_blockptr(buf
, i
);
879 btrfs_node_key_to_cpu(buf
, &key
, i
);
880 ret
= btrfs_inc_extent_ref(trans
, root
, bytenr
,
881 btrfs_level_size(root
, level
- 1),
882 root
->root_key
.objectid
,
884 level
- 1, key
.objectid
);
895 for (i
=0; i
< faili
; i
++) {
898 btrfs_item_key_to_cpu(buf
, &key
, i
);
899 if (btrfs_key_type(&key
) != BTRFS_EXTENT_DATA_KEY
)
901 fi
= btrfs_item_ptr(buf
, i
,
902 struct btrfs_file_extent_item
);
903 if (btrfs_file_extent_type(buf
, fi
) ==
904 BTRFS_FILE_EXTENT_INLINE
)
906 disk_bytenr
= btrfs_file_extent_disk_bytenr(buf
, fi
);
907 if (disk_bytenr
== 0)
909 err
= btrfs_free_extent(trans
, root
, disk_bytenr
,
910 btrfs_file_extent_disk_num_bytes(buf
,
914 bytenr
= btrfs_node_blockptr(buf
, i
);
915 err
= btrfs_free_extent(trans
, root
, bytenr
,
916 btrfs_level_size(root
, level
- 1), 0);
924 static int write_one_cache_group(struct btrfs_trans_handle
*trans
,
925 struct btrfs_root
*root
,
926 struct btrfs_path
*path
,
927 struct btrfs_block_group_cache
*cache
)
931 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
933 struct extent_buffer
*leaf
;
935 ret
= btrfs_search_slot(trans
, extent_root
, &cache
->key
, path
, 0, 1);
940 leaf
= path
->nodes
[0];
941 bi
= btrfs_item_ptr_offset(leaf
, path
->slots
[0]);
942 write_extent_buffer(leaf
, &cache
->item
, bi
, sizeof(cache
->item
));
943 btrfs_mark_buffer_dirty(leaf
);
944 btrfs_release_path(extent_root
, path
);
946 finish_current_insert(trans
, extent_root
);
947 pending_ret
= del_pending_extents(trans
, extent_root
);
956 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle
*trans
,
957 struct btrfs_root
*root
)
959 struct extent_io_tree
*block_group_cache
;
960 struct btrfs_block_group_cache
*cache
;
964 struct btrfs_path
*path
;
970 block_group_cache
= &root
->fs_info
->block_group_cache
;
971 path
= btrfs_alloc_path();
976 ret
= find_first_extent_bit(block_group_cache
, last
,
977 &start
, &end
, BLOCK_GROUP_DIRTY
);
982 ret
= get_state_private(block_group_cache
, start
, &ptr
);
986 cache
= (struct btrfs_block_group_cache
*)(unsigned long)ptr
;
987 err
= write_one_cache_group(trans
, root
,
990 * if we fail to write the cache group, we want
991 * to keep it marked dirty in hopes that a later
998 clear_extent_bits(block_group_cache
, start
, end
,
999 BLOCK_GROUP_DIRTY
, GFP_NOFS
);
1001 btrfs_free_path(path
);
1005 static struct btrfs_space_info
*__find_space_info(struct btrfs_fs_info
*info
,
1008 struct list_head
*head
= &info
->space_info
;
1009 struct list_head
*cur
;
1010 struct btrfs_space_info
*found
;
1011 list_for_each(cur
, head
) {
1012 found
= list_entry(cur
, struct btrfs_space_info
, list
);
1013 if (found
->flags
== flags
)
1020 static int update_space_info(struct btrfs_fs_info
*info
, u64 flags
,
1021 u64 total_bytes
, u64 bytes_used
,
1022 struct btrfs_space_info
**space_info
)
1024 struct btrfs_space_info
*found
;
1026 found
= __find_space_info(info
, flags
);
1028 found
->total_bytes
+= total_bytes
;
1029 found
->bytes_used
+= bytes_used
;
1030 WARN_ON(found
->total_bytes
< found
->bytes_used
);
1031 *space_info
= found
;
1034 found
= kmalloc(sizeof(*found
), GFP_NOFS
);
1038 list_add(&found
->list
, &info
->space_info
);
1039 found
->flags
= flags
;
1040 found
->total_bytes
= total_bytes
;
1041 found
->bytes_used
= bytes_used
;
1042 found
->bytes_pinned
= 0;
1044 *space_info
= found
;
1048 static void set_avail_alloc_bits(struct btrfs_fs_info
*fs_info
, u64 flags
)
1050 u64 extra_flags
= flags
& (BTRFS_BLOCK_GROUP_RAID0
|
1051 BTRFS_BLOCK_GROUP_RAID1
|
1052 BTRFS_BLOCK_GROUP_DUP
);
1054 if (flags
& BTRFS_BLOCK_GROUP_DATA
)
1055 fs_info
->avail_data_alloc_bits
|= extra_flags
;
1056 if (flags
& BTRFS_BLOCK_GROUP_METADATA
)
1057 fs_info
->avail_metadata_alloc_bits
|= extra_flags
;
1058 if (flags
& BTRFS_BLOCK_GROUP_SYSTEM
)
1059 fs_info
->avail_system_alloc_bits
|= extra_flags
;
1063 static int do_chunk_alloc(struct btrfs_trans_handle
*trans
,
1064 struct btrfs_root
*extent_root
, u64 alloc_bytes
,
1067 struct btrfs_space_info
*space_info
;
1073 space_info
= __find_space_info(extent_root
->fs_info
, flags
);
1075 ret
= update_space_info(extent_root
->fs_info
, flags
,
1079 BUG_ON(!space_info
);
1081 if (space_info
->full
)
1084 thresh
= div_factor(space_info
->total_bytes
, 6);
1085 if ((space_info
->bytes_used
+ space_info
->bytes_pinned
+ alloc_bytes
) <
1089 ret
= btrfs_alloc_chunk(trans
, extent_root
, &start
, &num_bytes
, flags
);
1090 if (ret
== -ENOSPC
) {
1091 printk("space info full %Lu\n", flags
);
1092 space_info
->full
= 1;
1098 ret
= btrfs_make_block_group(trans
, extent_root
, 0, flags
,
1099 extent_root
->fs_info
->chunk_root
->root_key
.objectid
,
1106 static int update_block_group(struct btrfs_trans_handle
*trans
,
1107 struct btrfs_root
*root
,
1108 u64 bytenr
, u64 num_bytes
, int alloc
,
1111 struct btrfs_block_group_cache
*cache
;
1112 struct btrfs_fs_info
*info
= root
->fs_info
;
1113 u64 total
= num_bytes
;
1120 cache
= btrfs_lookup_block_group(info
, bytenr
);
1124 byte_in_group
= bytenr
- cache
->key
.objectid
;
1125 WARN_ON(byte_in_group
> cache
->key
.offset
);
1126 start
= cache
->key
.objectid
;
1127 end
= start
+ cache
->key
.offset
- 1;
1128 set_extent_bits(&info
->block_group_cache
, start
, end
,
1129 BLOCK_GROUP_DIRTY
, GFP_NOFS
);
1131 old_val
= btrfs_block_group_used(&cache
->item
);
1132 num_bytes
= min(total
, cache
->key
.offset
- byte_in_group
);
1134 old_val
+= num_bytes
;
1135 cache
->space_info
->bytes_used
+= num_bytes
;
1137 old_val
-= num_bytes
;
1138 cache
->space_info
->bytes_used
-= num_bytes
;
1140 set_extent_dirty(&info
->free_space_cache
,
1141 bytenr
, bytenr
+ num_bytes
- 1,
1145 btrfs_set_block_group_used(&cache
->item
, old_val
);
1147 bytenr
+= num_bytes
;
1152 static int update_pinned_extents(struct btrfs_root
*root
,
1153 u64 bytenr
, u64 num
, int pin
)
1156 struct btrfs_block_group_cache
*cache
;
1157 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1160 set_extent_dirty(&fs_info
->pinned_extents
,
1161 bytenr
, bytenr
+ num
- 1, GFP_NOFS
);
1163 clear_extent_dirty(&fs_info
->pinned_extents
,
1164 bytenr
, bytenr
+ num
- 1, GFP_NOFS
);
1167 cache
= btrfs_lookup_block_group(fs_info
, bytenr
);
1169 len
= min(num
, cache
->key
.offset
-
1170 (bytenr
- cache
->key
.objectid
));
1172 cache
->pinned
+= len
;
1173 cache
->space_info
->bytes_pinned
+= len
;
1174 fs_info
->total_pinned
+= len
;
1176 cache
->pinned
-= len
;
1177 cache
->space_info
->bytes_pinned
-= len
;
1178 fs_info
->total_pinned
-= len
;
1186 int btrfs_copy_pinned(struct btrfs_root
*root
, struct extent_io_tree
*copy
)
1191 struct extent_io_tree
*pinned_extents
= &root
->fs_info
->pinned_extents
;
1195 ret
= find_first_extent_bit(pinned_extents
, last
,
1196 &start
, &end
, EXTENT_DIRTY
);
1199 set_extent_dirty(copy
, start
, end
, GFP_NOFS
);
1205 int btrfs_finish_extent_commit(struct btrfs_trans_handle
*trans
,
1206 struct btrfs_root
*root
,
1207 struct extent_io_tree
*unpin
)
1212 struct extent_io_tree
*free_space_cache
;
1213 free_space_cache
= &root
->fs_info
->free_space_cache
;
1216 ret
= find_first_extent_bit(unpin
, 0, &start
, &end
,
1220 update_pinned_extents(root
, start
, end
+ 1 - start
, 0);
1221 clear_extent_dirty(unpin
, start
, end
, GFP_NOFS
);
1222 set_extent_dirty(free_space_cache
, start
, end
, GFP_NOFS
);
1227 static int finish_current_insert(struct btrfs_trans_handle
*trans
,
1228 struct btrfs_root
*extent_root
)
1232 struct btrfs_fs_info
*info
= extent_root
->fs_info
;
1233 struct extent_buffer
*eb
;
1234 struct btrfs_path
*path
;
1235 struct btrfs_key ins
;
1236 struct btrfs_disk_key first
;
1237 struct btrfs_extent_item extent_item
;
1242 btrfs_set_stack_extent_refs(&extent_item
, 1);
1243 btrfs_set_key_type(&ins
, BTRFS_EXTENT_ITEM_KEY
);
1244 path
= btrfs_alloc_path();
1247 ret
= find_first_extent_bit(&info
->extent_ins
, 0, &start
,
1248 &end
, EXTENT_LOCKED
);
1252 ins
.objectid
= start
;
1253 ins
.offset
= end
+ 1 - start
;
1254 err
= btrfs_insert_item(trans
, extent_root
, &ins
,
1255 &extent_item
, sizeof(extent_item
));
1256 clear_extent_bits(&info
->extent_ins
, start
, end
, EXTENT_LOCKED
,
1258 eb
= read_tree_block(extent_root
, ins
.objectid
, ins
.offset
);
1259 level
= btrfs_header_level(eb
);
1261 btrfs_item_key(eb
, &first
, 0);
1263 btrfs_node_key(eb
, &first
, 0);
1265 err
= btrfs_insert_extent_backref(trans
, extent_root
, path
,
1266 start
, extent_root
->root_key
.objectid
,
1268 btrfs_disk_key_objectid(&first
));
1270 free_extent_buffer(eb
);
1272 btrfs_free_path(path
);
1276 static int pin_down_bytes(struct btrfs_root
*root
, u64 bytenr
, u32 num_bytes
,
1280 struct extent_buffer
*buf
;
1283 buf
= btrfs_find_tree_block(root
, bytenr
, num_bytes
);
1285 if (btrfs_buffer_uptodate(buf
)) {
1287 root
->fs_info
->running_transaction
->transid
;
1288 u64 header_transid
=
1289 btrfs_header_generation(buf
);
1290 if (header_transid
== transid
&&
1291 !btrfs_header_flag(buf
,
1292 BTRFS_HEADER_FLAG_WRITTEN
)) {
1293 clean_tree_block(NULL
, root
, buf
);
1294 free_extent_buffer(buf
);
1298 free_extent_buffer(buf
);
1300 update_pinned_extents(root
, bytenr
, num_bytes
, 1);
1302 set_extent_bits(&root
->fs_info
->pending_del
,
1303 bytenr
, bytenr
+ num_bytes
- 1,
1304 EXTENT_LOCKED
, GFP_NOFS
);
1311 * remove an extent from the root, returns 0 on success
1313 static int __free_extent(struct btrfs_trans_handle
*trans
, struct btrfs_root
1314 *root
, u64 bytenr
, u64 num_bytes
,
1315 u64 root_objectid
, u64 ref_generation
,
1316 u64 owner_objectid
, u64 owner_offset
, int pin
,
1319 struct btrfs_path
*path
;
1320 struct btrfs_key key
;
1321 struct btrfs_fs_info
*info
= root
->fs_info
;
1322 struct btrfs_root
*extent_root
= info
->extent_root
;
1323 struct extent_buffer
*leaf
;
1325 int extent_slot
= 0;
1326 int found_extent
= 0;
1328 struct btrfs_extent_item
*ei
;
1331 key
.objectid
= bytenr
;
1332 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
1333 key
.offset
= num_bytes
;
1334 path
= btrfs_alloc_path();
1339 ret
= lookup_extent_backref(trans
, extent_root
, path
,
1340 bytenr
, root_objectid
,
1342 owner_objectid
, owner_offset
, 1);
1344 struct btrfs_key found_key
;
1345 extent_slot
= path
->slots
[0];
1346 while(extent_slot
> 0) {
1348 btrfs_item_key_to_cpu(path
->nodes
[0], &found_key
,
1350 if (found_key
.objectid
!= bytenr
)
1352 if (found_key
.type
== BTRFS_EXTENT_ITEM_KEY
&&
1353 found_key
.offset
== num_bytes
) {
1357 if (path
->slots
[0] - extent_slot
> 5)
1361 ret
= btrfs_del_item(trans
, extent_root
, path
);
1363 btrfs_print_leaf(extent_root
, path
->nodes
[0]);
1365 printk("Unable to find ref byte nr %Lu root %Lu "
1366 " gen %Lu owner %Lu offset %Lu\n", bytenr
,
1367 root_objectid
, ref_generation
, owner_objectid
,
1370 if (!found_extent
) {
1371 btrfs_release_path(extent_root
, path
);
1372 ret
= btrfs_search_slot(trans
, extent_root
, &key
, path
, -1, 1);
1376 extent_slot
= path
->slots
[0];
1379 leaf
= path
->nodes
[0];
1380 ei
= btrfs_item_ptr(leaf
, extent_slot
,
1381 struct btrfs_extent_item
);
1382 refs
= btrfs_extent_refs(leaf
, ei
);
1385 btrfs_set_extent_refs(leaf
, ei
, refs
);
1387 btrfs_mark_buffer_dirty(leaf
);
1389 if (refs
== 0 && found_extent
&& path
->slots
[0] == extent_slot
+ 1) {
1390 /* if the back ref and the extent are next to each other
1391 * they get deleted below in one shot
1393 path
->slots
[0] = extent_slot
;
1395 } else if (found_extent
) {
1396 /* otherwise delete the extent back ref */
1397 ret
= btrfs_del_item(trans
, extent_root
, path
);
1399 /* if refs are 0, we need to setup the path for deletion */
1401 btrfs_release_path(extent_root
, path
);
1402 ret
= btrfs_search_slot(trans
, extent_root
, &key
, path
,
1415 ret
= pin_down_bytes(root
, bytenr
, num_bytes
, 0);
1421 /* block accounting for super block */
1422 super_used
= btrfs_super_bytes_used(&info
->super_copy
);
1423 btrfs_set_super_bytes_used(&info
->super_copy
,
1424 super_used
- num_bytes
);
1426 /* block accounting for root item */
1427 root_used
= btrfs_root_used(&root
->root_item
);
1428 btrfs_set_root_used(&root
->root_item
,
1429 root_used
- num_bytes
);
1430 ret
= btrfs_del_items(trans
, extent_root
, path
, path
->slots
[0],
1435 ret
= update_block_group(trans
, root
, bytenr
, num_bytes
, 0,
1439 btrfs_free_path(path
);
1440 finish_current_insert(trans
, extent_root
);
1445 * find all the blocks marked as pending in the radix tree and remove
1446 * them from the extent map
1448 static int del_pending_extents(struct btrfs_trans_handle
*trans
, struct
1449 btrfs_root
*extent_root
)
1455 struct extent_io_tree
*pending_del
;
1456 struct extent_io_tree
*pinned_extents
;
1458 pending_del
= &extent_root
->fs_info
->pending_del
;
1459 pinned_extents
= &extent_root
->fs_info
->pinned_extents
;
1462 ret
= find_first_extent_bit(pending_del
, 0, &start
, &end
,
1466 update_pinned_extents(extent_root
, start
, end
+ 1 - start
, 1);
1467 clear_extent_bits(pending_del
, start
, end
, EXTENT_LOCKED
,
1469 ret
= __free_extent(trans
, extent_root
,
1470 start
, end
+ 1 - start
,
1471 extent_root
->root_key
.objectid
,
1480 * remove an extent from the root, returns 0 on success
1482 int btrfs_free_extent(struct btrfs_trans_handle
*trans
, struct btrfs_root
1483 *root
, u64 bytenr
, u64 num_bytes
,
1484 u64 root_objectid
, u64 ref_generation
,
1485 u64 owner_objectid
, u64 owner_offset
, int pin
)
1487 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
1491 WARN_ON(num_bytes
< root
->sectorsize
);
1492 if (!root
->ref_cows
)
1495 if (root
== extent_root
) {
1496 pin_down_bytes(root
, bytenr
, num_bytes
, 1);
1499 ret
= __free_extent(trans
, root
, bytenr
, num_bytes
, root_objectid
,
1500 ref_generation
, owner_objectid
, owner_offset
,
1502 pending_ret
= del_pending_extents(trans
, root
->fs_info
->extent_root
);
1503 return ret
? ret
: pending_ret
;
1506 static u64
stripe_align(struct btrfs_root
*root
, u64 val
)
1508 u64 mask
= ((u64
)root
->stripesize
- 1);
1509 u64 ret
= (val
+ mask
) & ~mask
;
1514 * walks the btree of allocated extents and find a hole of a given size.
1515 * The key ins is changed to record the hole:
1516 * ins->objectid == block start
1517 * ins->flags = BTRFS_EXTENT_ITEM_KEY
1518 * ins->offset == number of blocks
1519 * Any available blocks before search_start are skipped.
1521 static int noinline
find_free_extent(struct btrfs_trans_handle
*trans
,
1522 struct btrfs_root
*orig_root
,
1523 u64 num_bytes
, u64 empty_size
,
1524 u64 search_start
, u64 search_end
,
1525 u64 hint_byte
, struct btrfs_key
*ins
,
1526 u64 exclude_start
, u64 exclude_nr
,
1530 u64 orig_search_start
= search_start
;
1531 struct btrfs_root
* root
= orig_root
->fs_info
->extent_root
;
1532 struct btrfs_fs_info
*info
= root
->fs_info
;
1533 u64 total_needed
= num_bytes
;
1534 u64
*last_ptr
= NULL
;
1535 struct btrfs_block_group_cache
*block_group
;
1538 int empty_cluster
= 2 * 1024 * 1024;
1540 WARN_ON(num_bytes
< root
->sectorsize
);
1541 btrfs_set_key_type(ins
, BTRFS_EXTENT_ITEM_KEY
);
1543 if (data
& BTRFS_BLOCK_GROUP_METADATA
) {
1544 last_ptr
= &root
->fs_info
->last_alloc
;
1545 empty_cluster
= 256 * 1024;
1548 if ((data
& BTRFS_BLOCK_GROUP_DATA
) && btrfs_test_opt(root
, SSD
)) {
1549 last_ptr
= &root
->fs_info
->last_data_alloc
;
1554 hint_byte
= *last_ptr
;
1556 empty_size
+= empty_cluster
;
1560 if (search_end
== (u64
)-1)
1561 search_end
= btrfs_super_total_bytes(&info
->super_copy
);
1564 block_group
= btrfs_lookup_block_group(info
, hint_byte
);
1566 hint_byte
= search_start
;
1567 block_group
= btrfs_find_block_group(root
, block_group
,
1568 hint_byte
, data
, 1);
1569 if (last_ptr
&& *last_ptr
== 0 && block_group
)
1570 hint_byte
= block_group
->key
.objectid
;
1572 block_group
= btrfs_find_block_group(root
,
1574 search_start
, data
, 1);
1576 search_start
= max(search_start
, hint_byte
);
1578 total_needed
+= empty_size
;
1582 block_group
= btrfs_lookup_block_group(info
, search_start
);
1584 block_group
= btrfs_lookup_block_group(info
,
1587 ret
= find_search_start(root
, &block_group
, &search_start
,
1588 total_needed
, data
);
1589 if (ret
== -ENOSPC
&& last_ptr
&& *last_ptr
) {
1591 block_group
= btrfs_lookup_block_group(info
,
1593 search_start
= orig_search_start
;
1594 ret
= find_search_start(root
, &block_group
, &search_start
,
1595 total_needed
, data
);
1602 if (last_ptr
&& *last_ptr
&& search_start
!= *last_ptr
) {
1605 empty_size
+= empty_cluster
;
1606 total_needed
+= empty_size
;
1608 block_group
= btrfs_lookup_block_group(info
,
1610 search_start
= orig_search_start
;
1611 ret
= find_search_start(root
, &block_group
,
1612 &search_start
, total_needed
, data
);
1619 search_start
= stripe_align(root
, search_start
);
1620 ins
->objectid
= search_start
;
1621 ins
->offset
= num_bytes
;
1623 if (ins
->objectid
+ num_bytes
>= search_end
)
1626 if (ins
->objectid
+ num_bytes
>
1627 block_group
->key
.objectid
+ block_group
->key
.offset
) {
1628 search_start
= block_group
->key
.objectid
+
1629 block_group
->key
.offset
;
1633 if (test_range_bit(&info
->extent_ins
, ins
->objectid
,
1634 ins
->objectid
+ num_bytes
-1, EXTENT_LOCKED
, 0)) {
1635 search_start
= ins
->objectid
+ num_bytes
;
1639 if (test_range_bit(&info
->pinned_extents
, ins
->objectid
,
1640 ins
->objectid
+ num_bytes
-1, EXTENT_DIRTY
, 0)) {
1641 search_start
= ins
->objectid
+ num_bytes
;
1645 if (exclude_nr
> 0 && (ins
->objectid
+ num_bytes
> exclude_start
&&
1646 ins
->objectid
< exclude_start
+ exclude_nr
)) {
1647 search_start
= exclude_start
+ exclude_nr
;
1651 if (!(data
& BTRFS_BLOCK_GROUP_DATA
)) {
1652 block_group
= btrfs_lookup_block_group(info
, ins
->objectid
);
1654 trans
->block_group
= block_group
;
1656 ins
->offset
= num_bytes
;
1658 *last_ptr
= ins
->objectid
+ ins
->offset
;
1660 btrfs_super_total_bytes(&root
->fs_info
->super_copy
)) {
1667 if (search_start
+ num_bytes
>= search_end
) {
1669 search_start
= orig_search_start
;
1676 total_needed
-= empty_size
;
1681 block_group
= btrfs_lookup_block_group(info
, search_start
);
1683 block_group
= btrfs_find_block_group(root
, block_group
,
1684 search_start
, data
, 0);
1691 * finds a free extent and does all the dirty work required for allocation
1692 * returns the key for the extent through ins, and a tree buffer for
1693 * the first block of the extent through buf.
1695 * returns 0 if everything worked, non-zero otherwise.
1697 int btrfs_alloc_extent(struct btrfs_trans_handle
*trans
,
1698 struct btrfs_root
*root
,
1699 u64 num_bytes
, u64 root_objectid
, u64 ref_generation
,
1700 u64 owner
, u64 owner_offset
,
1701 u64 empty_size
, u64 hint_byte
,
1702 u64 search_end
, struct btrfs_key
*ins
, int data
)
1708 u64 search_start
= 0;
1712 struct btrfs_fs_info
*info
= root
->fs_info
;
1713 struct btrfs_root
*extent_root
= info
->extent_root
;
1714 struct btrfs_extent_item
*extent_item
;
1715 struct btrfs_extent_ref
*ref
;
1716 struct btrfs_path
*path
;
1717 struct btrfs_key keys
[2];
1720 alloc_profile
= info
->avail_data_alloc_bits
&
1721 info
->data_alloc_profile
;
1722 data
= BTRFS_BLOCK_GROUP_DATA
| alloc_profile
;
1723 } else if (root
== root
->fs_info
->chunk_root
) {
1724 alloc_profile
= info
->avail_system_alloc_bits
&
1725 info
->system_alloc_profile
;
1726 data
= BTRFS_BLOCK_GROUP_SYSTEM
| alloc_profile
;
1728 alloc_profile
= info
->avail_metadata_alloc_bits
&
1729 info
->metadata_alloc_profile
;
1730 data
= BTRFS_BLOCK_GROUP_METADATA
| alloc_profile
;
1733 if (root
->ref_cows
) {
1734 if (!(data
& BTRFS_BLOCK_GROUP_METADATA
)) {
1735 ret
= do_chunk_alloc(trans
, root
->fs_info
->extent_root
,
1737 BTRFS_BLOCK_GROUP_METADATA
|
1738 (info
->metadata_alloc_profile
&
1739 info
->avail_metadata_alloc_bits
));
1742 ret
= do_chunk_alloc(trans
, root
->fs_info
->extent_root
,
1743 num_bytes
+ 2 * 1024 * 1024, data
);
1747 new_hint
= max(hint_byte
, root
->fs_info
->alloc_start
);
1748 if (new_hint
< btrfs_super_total_bytes(&info
->super_copy
))
1749 hint_byte
= new_hint
;
1751 WARN_ON(num_bytes
< root
->sectorsize
);
1752 ret
= find_free_extent(trans
, root
, num_bytes
, empty_size
,
1753 search_start
, search_end
, hint_byte
, ins
,
1754 trans
->alloc_exclude_start
,
1755 trans
->alloc_exclude_nr
, data
);
1760 /* block accounting for super block */
1761 super_used
= btrfs_super_bytes_used(&info
->super_copy
);
1762 btrfs_set_super_bytes_used(&info
->super_copy
, super_used
+ num_bytes
);
1764 /* block accounting for root item */
1765 root_used
= btrfs_root_used(&root
->root_item
);
1766 btrfs_set_root_used(&root
->root_item
, root_used
+ num_bytes
);
1768 clear_extent_dirty(&root
->fs_info
->free_space_cache
,
1769 ins
->objectid
, ins
->objectid
+ ins
->offset
- 1,
1772 if (root
== extent_root
) {
1773 set_extent_bits(&root
->fs_info
->extent_ins
, ins
->objectid
,
1774 ins
->objectid
+ ins
->offset
- 1,
1775 EXTENT_LOCKED
, GFP_NOFS
);
1779 WARN_ON(trans
->alloc_exclude_nr
);
1780 trans
->alloc_exclude_start
= ins
->objectid
;
1781 trans
->alloc_exclude_nr
= ins
->offset
;
1783 memcpy(&keys
[0], ins
, sizeof(*ins
));
1784 keys
[1].offset
= hash_extent_ref(root_objectid
, ref_generation
,
1785 owner
, owner_offset
);
1786 keys
[1].objectid
= ins
->objectid
;
1787 keys
[1].type
= BTRFS_EXTENT_REF_KEY
;
1788 sizes
[0] = sizeof(*extent_item
);
1789 sizes
[1] = sizeof(*ref
);
1791 path
= btrfs_alloc_path();
1794 ret
= btrfs_insert_empty_items(trans
, extent_root
, path
, keys
,
1798 extent_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
1799 struct btrfs_extent_item
);
1800 btrfs_set_extent_refs(path
->nodes
[0], extent_item
, 1);
1801 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0] + 1,
1802 struct btrfs_extent_ref
);
1804 btrfs_set_ref_root(path
->nodes
[0], ref
, root_objectid
);
1805 btrfs_set_ref_generation(path
->nodes
[0], ref
, ref_generation
);
1806 btrfs_set_ref_objectid(path
->nodes
[0], ref
, owner
);
1807 btrfs_set_ref_offset(path
->nodes
[0], ref
, owner_offset
);
1809 btrfs_mark_buffer_dirty(path
->nodes
[0]);
1811 trans
->alloc_exclude_start
= 0;
1812 trans
->alloc_exclude_nr
= 0;
1813 btrfs_free_path(path
);
1814 finish_current_insert(trans
, extent_root
);
1815 pending_ret
= del_pending_extents(trans
, extent_root
);
1825 ret
= update_block_group(trans
, root
, ins
->objectid
, ins
->offset
, 1, 0);
1827 printk("update block group failed for %Lu %Lu\n",
1828 ins
->objectid
, ins
->offset
);
1835 * helper function to allocate a block for a given tree
1836 * returns the tree buffer or NULL.
1838 struct extent_buffer
*btrfs_alloc_free_block(struct btrfs_trans_handle
*trans
,
1839 struct btrfs_root
*root
,
1841 u64 root_objectid
, u64 hint
,
1847 ref_generation
= trans
->transid
;
1852 return __btrfs_alloc_free_block(trans
, root
, blocksize
, root_objectid
,
1853 ref_generation
, 0, 0, hint
, empty_size
);
1857 * helper function to allocate a block for a given tree
1858 * returns the tree buffer or NULL.
1860 struct extent_buffer
*__btrfs_alloc_free_block(struct btrfs_trans_handle
*trans
,
1861 struct btrfs_root
*root
,
1870 struct btrfs_key ins
;
1872 struct extent_buffer
*buf
;
1874 ret
= btrfs_alloc_extent(trans
, root
, blocksize
,
1875 root_objectid
, ref_generation
,
1876 level
, first_objectid
, empty_size
, hint
,
1880 return ERR_PTR(ret
);
1882 buf
= btrfs_find_create_tree_block(root
, ins
.objectid
, blocksize
);
1884 btrfs_free_extent(trans
, root
, ins
.objectid
, blocksize
,
1885 root
->root_key
.objectid
, ref_generation
,
1887 return ERR_PTR(-ENOMEM
);
1889 btrfs_set_header_generation(buf
, trans
->transid
);
1890 clean_tree_block(trans
, root
, buf
);
1891 wait_on_tree_block_writeback(root
, buf
);
1892 btrfs_set_buffer_uptodate(buf
);
1894 if (PageDirty(buf
->first_page
)) {
1895 printk("page %lu dirty\n", buf
->first_page
->index
);
1899 set_extent_dirty(&trans
->transaction
->dirty_pages
, buf
->start
,
1900 buf
->start
+ buf
->len
- 1, GFP_NOFS
);
1901 if (!btrfs_test_opt(root
, SSD
))
1902 btrfs_set_buffer_defrag(buf
);
1903 trans
->blocks_used
++;
1907 static int noinline
drop_leaf_ref(struct btrfs_trans_handle
*trans
,
1908 struct btrfs_root
*root
,
1909 struct extent_buffer
*leaf
)
1912 u64 leaf_generation
;
1913 struct btrfs_key key
;
1914 struct btrfs_file_extent_item
*fi
;
1919 BUG_ON(!btrfs_is_leaf(leaf
));
1920 nritems
= btrfs_header_nritems(leaf
);
1921 leaf_owner
= btrfs_header_owner(leaf
);
1922 leaf_generation
= btrfs_header_generation(leaf
);
1924 for (i
= 0; i
< nritems
; i
++) {
1927 btrfs_item_key_to_cpu(leaf
, &key
, i
);
1928 if (btrfs_key_type(&key
) != BTRFS_EXTENT_DATA_KEY
)
1930 fi
= btrfs_item_ptr(leaf
, i
, struct btrfs_file_extent_item
);
1931 if (btrfs_file_extent_type(leaf
, fi
) ==
1932 BTRFS_FILE_EXTENT_INLINE
)
1935 * FIXME make sure to insert a trans record that
1936 * repeats the snapshot del on crash
1938 disk_bytenr
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
1939 if (disk_bytenr
== 0)
1941 ret
= btrfs_free_extent(trans
, root
, disk_bytenr
,
1942 btrfs_file_extent_disk_num_bytes(leaf
, fi
),
1943 leaf_owner
, leaf_generation
,
1944 key
.objectid
, key
.offset
, 0);
1950 static void noinline
reada_walk_down(struct btrfs_root
*root
,
1951 struct extent_buffer
*node
,
1964 nritems
= btrfs_header_nritems(node
);
1965 level
= btrfs_header_level(node
);
1969 for (i
= slot
; i
< nritems
&& skipped
< 32; i
++) {
1970 bytenr
= btrfs_node_blockptr(node
, i
);
1971 if (last
&& ((bytenr
> last
&& bytenr
- last
> 32 * 1024) ||
1972 (last
> bytenr
&& last
- bytenr
> 32 * 1024))) {
1976 blocksize
= btrfs_level_size(root
, level
- 1);
1978 ret
= lookup_extent_ref(NULL
, root
, bytenr
,
1986 mutex_unlock(&root
->fs_info
->fs_mutex
);
1987 ret
= readahead_tree_block(root
, bytenr
, blocksize
);
1988 last
= bytenr
+ blocksize
;
1990 mutex_lock(&root
->fs_info
->fs_mutex
);
1997 * helper function for drop_snapshot, this walks down the tree dropping ref
1998 * counts as it goes.
2000 static int noinline
walk_down_tree(struct btrfs_trans_handle
*trans
,
2001 struct btrfs_root
*root
,
2002 struct btrfs_path
*path
, int *level
)
2007 struct extent_buffer
*next
;
2008 struct extent_buffer
*cur
;
2009 struct extent_buffer
*parent
;
2014 WARN_ON(*level
< 0);
2015 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
2016 ret
= lookup_extent_ref(trans
, root
,
2017 path
->nodes
[*level
]->start
,
2018 path
->nodes
[*level
]->len
, &refs
);
2024 * walk down to the last node level and free all the leaves
2026 while(*level
>= 0) {
2027 WARN_ON(*level
< 0);
2028 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
2029 cur
= path
->nodes
[*level
];
2031 if (btrfs_header_level(cur
) != *level
)
2034 if (path
->slots
[*level
] >=
2035 btrfs_header_nritems(cur
))
2038 ret
= drop_leaf_ref(trans
, root
, cur
);
2042 bytenr
= btrfs_node_blockptr(cur
, path
->slots
[*level
]);
2043 blocksize
= btrfs_level_size(root
, *level
- 1);
2044 ret
= lookup_extent_ref(trans
, root
, bytenr
, blocksize
, &refs
);
2047 parent
= path
->nodes
[*level
];
2048 root_owner
= btrfs_header_owner(parent
);
2049 root_gen
= btrfs_header_generation(parent
);
2050 path
->slots
[*level
]++;
2051 ret
= btrfs_free_extent(trans
, root
, bytenr
,
2052 blocksize
, root_owner
,
2057 next
= btrfs_find_tree_block(root
, bytenr
, blocksize
);
2058 if (!next
|| !btrfs_buffer_uptodate(next
)) {
2059 free_extent_buffer(next
);
2060 reada_walk_down(root
, cur
, path
->slots
[*level
]);
2062 mutex_unlock(&root
->fs_info
->fs_mutex
);
2063 next
= read_tree_block(root
, bytenr
, blocksize
);
2064 mutex_lock(&root
->fs_info
->fs_mutex
);
2066 /* we've dropped the lock, double check */
2067 ret
= lookup_extent_ref(trans
, root
, bytenr
,
2071 parent
= path
->nodes
[*level
];
2072 root_owner
= btrfs_header_owner(parent
);
2073 root_gen
= btrfs_header_generation(parent
);
2075 path
->slots
[*level
]++;
2076 free_extent_buffer(next
);
2077 ret
= btrfs_free_extent(trans
, root
, bytenr
,
2085 btrfs_verify_block_csum(root
, next
);
2087 WARN_ON(*level
<= 0);
2088 if (path
->nodes
[*level
-1])
2089 free_extent_buffer(path
->nodes
[*level
-1]);
2090 path
->nodes
[*level
-1] = next
;
2091 *level
= btrfs_header_level(next
);
2092 path
->slots
[*level
] = 0;
2095 WARN_ON(*level
< 0);
2096 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
2098 if (path
->nodes
[*level
] == root
->node
) {
2099 root_owner
= root
->root_key
.objectid
;
2100 parent
= path
->nodes
[*level
];
2102 parent
= path
->nodes
[*level
+ 1];
2103 root_owner
= btrfs_header_owner(parent
);
2106 root_gen
= btrfs_header_generation(parent
);
2107 ret
= btrfs_free_extent(trans
, root
, path
->nodes
[*level
]->start
,
2108 path
->nodes
[*level
]->len
,
2109 root_owner
, root_gen
, 0, 0, 1);
2110 free_extent_buffer(path
->nodes
[*level
]);
2111 path
->nodes
[*level
] = NULL
;
2118 * helper for dropping snapshots. This walks back up the tree in the path
2119 * to find the first node higher up where we haven't yet gone through
2122 static int noinline
walk_up_tree(struct btrfs_trans_handle
*trans
,
2123 struct btrfs_root
*root
,
2124 struct btrfs_path
*path
, int *level
)
2128 struct btrfs_root_item
*root_item
= &root
->root_item
;
2133 for(i
= *level
; i
< BTRFS_MAX_LEVEL
- 1 && path
->nodes
[i
]; i
++) {
2134 slot
= path
->slots
[i
];
2135 if (slot
< btrfs_header_nritems(path
->nodes
[i
]) - 1) {
2136 struct extent_buffer
*node
;
2137 struct btrfs_disk_key disk_key
;
2138 node
= path
->nodes
[i
];
2141 WARN_ON(*level
== 0);
2142 btrfs_node_key(node
, &disk_key
, path
->slots
[i
]);
2143 memcpy(&root_item
->drop_progress
,
2144 &disk_key
, sizeof(disk_key
));
2145 root_item
->drop_level
= i
;
2148 if (path
->nodes
[*level
] == root
->node
) {
2149 root_owner
= root
->root_key
.objectid
;
2151 btrfs_header_generation(path
->nodes
[*level
]);
2153 struct extent_buffer
*node
;
2154 node
= path
->nodes
[*level
+ 1];
2155 root_owner
= btrfs_header_owner(node
);
2156 root_gen
= btrfs_header_generation(node
);
2158 ret
= btrfs_free_extent(trans
, root
,
2159 path
->nodes
[*level
]->start
,
2160 path
->nodes
[*level
]->len
,
2161 root_owner
, root_gen
, 0, 0, 1);
2163 free_extent_buffer(path
->nodes
[*level
]);
2164 path
->nodes
[*level
] = NULL
;
2172 * drop the reference count on the tree rooted at 'snap'. This traverses
2173 * the tree freeing any blocks that have a ref count of zero after being
2176 int btrfs_drop_snapshot(struct btrfs_trans_handle
*trans
, struct btrfs_root
2182 struct btrfs_path
*path
;
2185 struct btrfs_root_item
*root_item
= &root
->root_item
;
2187 path
= btrfs_alloc_path();
2190 level
= btrfs_header_level(root
->node
);
2192 if (btrfs_disk_key_objectid(&root_item
->drop_progress
) == 0) {
2193 path
->nodes
[level
] = root
->node
;
2194 extent_buffer_get(root
->node
);
2195 path
->slots
[level
] = 0;
2197 struct btrfs_key key
;
2198 struct btrfs_disk_key found_key
;
2199 struct extent_buffer
*node
;
2201 btrfs_disk_key_to_cpu(&key
, &root_item
->drop_progress
);
2202 level
= root_item
->drop_level
;
2203 path
->lowest_level
= level
;
2204 wret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2209 node
= path
->nodes
[level
];
2210 btrfs_node_key(node
, &found_key
, path
->slots
[level
]);
2211 WARN_ON(memcmp(&found_key
, &root_item
->drop_progress
,
2212 sizeof(found_key
)));
2215 wret
= walk_down_tree(trans
, root
, path
, &level
);
2221 wret
= walk_up_tree(trans
, root
, path
, &level
);
2229 for (i
= 0; i
<= orig_level
; i
++) {
2230 if (path
->nodes
[i
]) {
2231 free_extent_buffer(path
->nodes
[i
]);
2232 path
->nodes
[i
] = NULL
;
2236 btrfs_free_path(path
);
2240 int btrfs_free_block_groups(struct btrfs_fs_info
*info
)
2247 ret
= find_first_extent_bit(&info
->block_group_cache
, 0,
2248 &start
, &end
, (unsigned int)-1);
2251 ret
= get_state_private(&info
->block_group_cache
, start
, &ptr
);
2253 kfree((void *)(unsigned long)ptr
);
2254 clear_extent_bits(&info
->block_group_cache
, start
,
2255 end
, (unsigned int)-1, GFP_NOFS
);
2258 ret
= find_first_extent_bit(&info
->free_space_cache
, 0,
2259 &start
, &end
, EXTENT_DIRTY
);
2262 clear_extent_dirty(&info
->free_space_cache
, start
,
2268 static int noinline
relocate_inode_pages(struct inode
*inode
, u64 start
,
2274 u64 existing_delalloc
;
2275 unsigned long last_index
;
2278 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2279 struct file_ra_state
*ra
;
2281 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
2283 mutex_lock(&inode
->i_mutex
);
2284 i
= start
>> PAGE_CACHE_SHIFT
;
2285 last_index
= (start
+ len
- 1) >> PAGE_CACHE_SHIFT
;
2287 file_ra_state_init(ra
, inode
->i_mapping
);
2288 btrfs_force_ra(inode
->i_mapping
, ra
, NULL
, i
, last_index
);
2291 for (; i
<= last_index
; i
++) {
2292 page
= grab_cache_page(inode
->i_mapping
, i
);
2295 if (!PageUptodate(page
)) {
2296 btrfs_readpage(NULL
, page
);
2298 if (!PageUptodate(page
)) {
2300 page_cache_release(page
);
2304 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
2305 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
2307 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
2309 delalloc_start
= page_start
;
2310 existing_delalloc
= count_range_bits(io_tree
,
2311 &delalloc_start
, page_end
,
2312 PAGE_CACHE_SIZE
, EXTENT_DELALLOC
);
2314 set_extent_delalloc(io_tree
, page_start
,
2315 page_end
, GFP_NOFS
);
2317 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
2318 set_page_dirty(page
);
2320 page_cache_release(page
);
2324 mutex_unlock(&inode
->i_mutex
);
2329 * note, this releases the path
2331 static int noinline
relocate_one_reference(struct btrfs_root
*extent_root
,
2332 struct btrfs_path
*path
,
2333 struct btrfs_key
*extent_key
)
2335 struct inode
*inode
;
2336 struct btrfs_root
*found_root
;
2337 struct btrfs_key
*root_location
;
2338 struct btrfs_extent_ref
*ref
;
2345 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
2346 struct btrfs_extent_ref
);
2347 ref_root
= btrfs_ref_root(path
->nodes
[0], ref
);
2348 ref_gen
= btrfs_ref_generation(path
->nodes
[0], ref
);
2349 ref_objectid
= btrfs_ref_objectid(path
->nodes
[0], ref
);
2350 ref_offset
= btrfs_ref_offset(path
->nodes
[0], ref
);
2351 btrfs_release_path(extent_root
, path
);
2353 root_location
= kmalloc(sizeof(*root_location
), GFP_NOFS
);
2354 root_location
->objectid
= ref_root
;
2356 root_location
->offset
= 0;
2358 root_location
->offset
= (u64
)-1;
2359 root_location
->type
= BTRFS_ROOT_ITEM_KEY
;
2361 found_root
= btrfs_read_fs_root_no_name(extent_root
->fs_info
,
2363 BUG_ON(!found_root
);
2364 kfree(root_location
);
2366 if (ref_objectid
>= BTRFS_FIRST_FREE_OBJECTID
) {
2367 mutex_unlock(&extent_root
->fs_info
->fs_mutex
);
2368 inode
= btrfs_iget_locked(extent_root
->fs_info
->sb
,
2369 ref_objectid
, found_root
);
2370 if (inode
->i_state
& I_NEW
) {
2371 /* the inode and parent dir are two different roots */
2372 BTRFS_I(inode
)->root
= found_root
;
2373 BTRFS_I(inode
)->location
.objectid
= ref_objectid
;
2374 BTRFS_I(inode
)->location
.type
= BTRFS_INODE_ITEM_KEY
;
2375 BTRFS_I(inode
)->location
.offset
= 0;
2376 btrfs_read_locked_inode(inode
);
2377 unlock_new_inode(inode
);
2380 /* this can happen if the reference is not against
2381 * the latest version of the tree root
2383 if (is_bad_inode(inode
)) {
2384 mutex_lock(&extent_root
->fs_info
->fs_mutex
);
2387 relocate_inode_pages(inode
, ref_offset
, extent_key
->offset
);
2388 /* FIXME, data=ordered will help get rid of this */
2389 filemap_fdatawrite(inode
->i_mapping
);
2391 mutex_lock(&extent_root
->fs_info
->fs_mutex
);
2393 struct btrfs_trans_handle
*trans
;
2394 struct btrfs_key found_key
;
2395 struct extent_buffer
*eb
;
2399 trans
= btrfs_start_transaction(found_root
, 1);
2400 eb
= read_tree_block(found_root
, extent_key
->objectid
,
2401 extent_key
->offset
);
2402 level
= btrfs_header_level(eb
);
2405 btrfs_item_key_to_cpu(eb
, &found_key
, 0);
2407 btrfs_node_key_to_cpu(eb
, &found_key
, 0);
2409 free_extent_buffer(eb
);
2411 path
->lowest_level
= level
;
2413 ret
= btrfs_search_slot(trans
, found_root
, &found_key
, path
,
2415 path
->lowest_level
= 0;
2416 for (i
= level
; i
< BTRFS_MAX_LEVEL
; i
++) {
2417 if (!path
->nodes
[i
])
2419 free_extent_buffer(path
->nodes
[i
]);
2420 path
->nodes
[i
] = NULL
;
2422 btrfs_release_path(found_root
, path
);
2423 btrfs_end_transaction(trans
, found_root
);
2430 static int noinline
relocate_one_extent(struct btrfs_root
*extent_root
,
2431 struct btrfs_path
*path
,
2432 struct btrfs_key
*extent_key
)
2434 struct btrfs_key key
;
2435 struct btrfs_key found_key
;
2436 struct extent_buffer
*leaf
;
2441 key
.objectid
= extent_key
->objectid
;
2442 key
.type
= BTRFS_EXTENT_REF_KEY
;
2446 ret
= btrfs_search_slot(NULL
, extent_root
, &key
, path
, 0, 0);
2452 leaf
= path
->nodes
[0];
2453 nritems
= btrfs_header_nritems(leaf
);
2454 if (path
->slots
[0] == nritems
)
2457 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2458 if (found_key
.objectid
!= extent_key
->objectid
)
2461 if (found_key
.type
!= BTRFS_EXTENT_REF_KEY
)
2464 key
.offset
= found_key
.offset
+ 1;
2465 item_size
= btrfs_item_size_nr(leaf
, path
->slots
[0]);
2467 ret
= relocate_one_reference(extent_root
, path
, extent_key
);
2473 btrfs_release_path(extent_root
, path
);
2477 int btrfs_shrink_extent_tree(struct btrfs_root
*root
, u64 new_size
)
2479 struct btrfs_trans_handle
*trans
;
2480 struct btrfs_root
*tree_root
= root
->fs_info
->tree_root
;
2481 struct btrfs_path
*path
;
2484 struct btrfs_fs_info
*info
= root
->fs_info
;
2485 struct extent_io_tree
*block_group_cache
;
2486 struct btrfs_key key
;
2487 struct btrfs_key found_key
;
2488 struct extent_buffer
*leaf
;
2493 btrfs_set_super_total_bytes(&info
->super_copy
, new_size
);
2494 clear_extent_dirty(&info
->free_space_cache
, new_size
, (u64
)-1,
2496 block_group_cache
= &info
->block_group_cache
;
2497 path
= btrfs_alloc_path();
2498 root
= root
->fs_info
->extent_root
;
2503 key
.objectid
= new_size
;
2506 cur_byte
= key
.objectid
;
2508 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2512 ret
= btrfs_previous_item(root
, path
, 0, BTRFS_EXTENT_ITEM_KEY
);
2516 leaf
= path
->nodes
[0];
2517 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2518 if (found_key
.objectid
+ found_key
.offset
> new_size
) {
2519 cur_byte
= found_key
.objectid
;
2520 key
.objectid
= cur_byte
;
2523 btrfs_release_path(root
, path
);
2526 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2530 leaf
= path
->nodes
[0];
2531 nritems
= btrfs_header_nritems(leaf
);
2533 if (path
->slots
[0] >= nritems
) {
2534 ret
= btrfs_next_leaf(root
, path
);
2541 leaf
= path
->nodes
[0];
2542 nritems
= btrfs_header_nritems(leaf
);
2545 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2547 if (progress
&& need_resched()) {
2548 memcpy(&key
, &found_key
, sizeof(key
));
2549 mutex_unlock(&root
->fs_info
->fs_mutex
);
2551 mutex_lock(&root
->fs_info
->fs_mutex
);
2552 btrfs_release_path(root
, path
);
2553 btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2559 if (btrfs_key_type(&found_key
) != BTRFS_EXTENT_ITEM_KEY
||
2560 found_key
.objectid
+ found_key
.offset
<= cur_byte
) {
2566 cur_byte
= found_key
.objectid
+ found_key
.offset
;
2567 key
.objectid
= cur_byte
;
2568 btrfs_release_path(root
, path
);
2569 ret
= relocate_one_extent(root
, path
, &found_key
);
2572 btrfs_release_path(root
, path
);
2574 if (total_found
> 0) {
2575 trans
= btrfs_start_transaction(tree_root
, 1);
2576 btrfs_commit_transaction(trans
, tree_root
);
2578 mutex_unlock(&root
->fs_info
->fs_mutex
);
2579 btrfs_clean_old_snapshots(tree_root
);
2580 mutex_lock(&root
->fs_info
->fs_mutex
);
2582 trans
= btrfs_start_transaction(tree_root
, 1);
2583 btrfs_commit_transaction(trans
, tree_root
);
2587 trans
= btrfs_start_transaction(root
, 1);
2588 key
.objectid
= new_size
;
2594 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
2598 leaf
= path
->nodes
[0];
2599 nritems
= btrfs_header_nritems(leaf
);
2601 if (path
->slots
[0] >= nritems
) {
2602 ret
= btrfs_next_leaf(root
, path
);
2609 leaf
= path
->nodes
[0];
2610 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2613 * btrfs_next_leaf doesn't cow buffers, we have to
2614 * do the search again
2616 memcpy(&key
, &found_key
, sizeof(key
));
2617 btrfs_release_path(root
, path
);
2621 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2622 if (btrfs_key_type(&found_key
) != BTRFS_BLOCK_GROUP_ITEM_KEY
) {
2623 printk("shrinker found key %Lu %u %Lu\n",
2624 found_key
.objectid
, found_key
.type
,
2629 ret
= get_state_private(&info
->block_group_cache
,
2630 found_key
.objectid
, &ptr
);
2632 kfree((void *)(unsigned long)ptr
);
2634 clear_extent_bits(&info
->block_group_cache
, found_key
.objectid
,
2635 found_key
.objectid
+ found_key
.offset
- 1,
2636 (unsigned int)-1, GFP_NOFS
);
2638 key
.objectid
= found_key
.objectid
+ 1;
2639 btrfs_del_item(trans
, root
, path
);
2640 btrfs_release_path(root
, path
);
2642 if (need_resched()) {
2643 mutex_unlock(&root
->fs_info
->fs_mutex
);
2645 mutex_lock(&root
->fs_info
->fs_mutex
);
2648 clear_extent_dirty(&info
->free_space_cache
, new_size
, (u64
)-1,
2650 btrfs_commit_transaction(trans
, root
);
2652 btrfs_free_path(path
);
2656 int btrfs_grow_extent_tree(struct btrfs_trans_handle
*trans
,
2657 struct btrfs_root
*root
, u64 new_size
)
2659 btrfs_set_super_total_bytes(&root
->fs_info
->super_copy
, new_size
);
2663 int find_first_block_group(struct btrfs_root
*root
, struct btrfs_path
*path
,
2664 struct btrfs_key
*key
)
2667 struct btrfs_key found_key
;
2668 struct extent_buffer
*leaf
;
2671 ret
= btrfs_search_slot(NULL
, root
, key
, path
, 0, 0);
2675 slot
= path
->slots
[0];
2676 leaf
= path
->nodes
[0];
2677 if (slot
>= btrfs_header_nritems(leaf
)) {
2678 ret
= btrfs_next_leaf(root
, path
);
2685 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
2687 if (found_key
.objectid
>= key
->objectid
&&
2688 found_key
.type
== BTRFS_BLOCK_GROUP_ITEM_KEY
)
2697 int btrfs_read_block_groups(struct btrfs_root
*root
)
2699 struct btrfs_path
*path
;
2702 struct btrfs_block_group_cache
*cache
;
2703 struct btrfs_fs_info
*info
= root
->fs_info
;
2704 struct btrfs_space_info
*space_info
;
2705 struct extent_io_tree
*block_group_cache
;
2706 struct btrfs_key key
;
2707 struct btrfs_key found_key
;
2708 struct extent_buffer
*leaf
;
2710 block_group_cache
= &info
->block_group_cache
;
2711 root
= info
->extent_root
;
2714 btrfs_set_key_type(&key
, BTRFS_BLOCK_GROUP_ITEM_KEY
);
2715 path
= btrfs_alloc_path();
2720 ret
= find_first_block_group(root
, path
, &key
);
2728 leaf
= path
->nodes
[0];
2729 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2730 cache
= kmalloc(sizeof(*cache
), GFP_NOFS
);
2736 read_extent_buffer(leaf
, &cache
->item
,
2737 btrfs_item_ptr_offset(leaf
, path
->slots
[0]),
2738 sizeof(cache
->item
));
2739 memcpy(&cache
->key
, &found_key
, sizeof(found_key
));
2743 key
.objectid
= found_key
.objectid
+ found_key
.offset
;
2744 btrfs_release_path(root
, path
);
2745 cache
->flags
= btrfs_block_group_flags(&cache
->item
);
2747 if (cache
->flags
& BTRFS_BLOCK_GROUP_DATA
) {
2748 bit
= BLOCK_GROUP_DATA
;
2749 } else if (cache
->flags
& BTRFS_BLOCK_GROUP_SYSTEM
) {
2750 bit
= BLOCK_GROUP_SYSTEM
;
2751 } else if (cache
->flags
& BTRFS_BLOCK_GROUP_METADATA
) {
2752 bit
= BLOCK_GROUP_METADATA
;
2754 set_avail_alloc_bits(info
, cache
->flags
);
2756 ret
= update_space_info(info
, cache
->flags
, found_key
.offset
,
2757 btrfs_block_group_used(&cache
->item
),
2760 cache
->space_info
= space_info
;
2762 /* use EXTENT_LOCKED to prevent merging */
2763 set_extent_bits(block_group_cache
, found_key
.objectid
,
2764 found_key
.objectid
+ found_key
.offset
- 1,
2765 bit
| EXTENT_LOCKED
, GFP_NOFS
);
2766 set_state_private(block_group_cache
, found_key
.objectid
,
2767 (unsigned long)cache
);
2770 btrfs_super_total_bytes(&info
->super_copy
))
2775 btrfs_free_path(path
);
2779 int btrfs_make_block_group(struct btrfs_trans_handle
*trans
,
2780 struct btrfs_root
*root
, u64 bytes_used
,
2781 u64 type
, u64 chunk_tree
, u64 chunk_objectid
,
2786 struct btrfs_root
*extent_root
;
2787 struct btrfs_block_group_cache
*cache
;
2788 struct extent_io_tree
*block_group_cache
;
2790 extent_root
= root
->fs_info
->extent_root
;
2791 block_group_cache
= &root
->fs_info
->block_group_cache
;
2793 cache
= kmalloc(sizeof(*cache
), GFP_NOFS
);
2795 cache
->key
.objectid
= chunk_objectid
;
2796 cache
->key
.offset
= size
;
2799 btrfs_set_key_type(&cache
->key
, BTRFS_BLOCK_GROUP_ITEM_KEY
);
2800 memset(&cache
->item
, 0, sizeof(cache
->item
));
2801 btrfs_set_block_group_used(&cache
->item
, bytes_used
);
2802 btrfs_set_block_group_chunk_tree(&cache
->item
, chunk_tree
);
2803 btrfs_set_block_group_chunk_objectid(&cache
->item
, chunk_objectid
);
2804 cache
->flags
= type
;
2805 btrfs_set_block_group_flags(&cache
->item
, type
);
2807 ret
= update_space_info(root
->fs_info
, cache
->flags
, size
, bytes_used
,
2808 &cache
->space_info
);
2811 bit
= block_group_state_bits(type
);
2812 set_extent_bits(block_group_cache
, chunk_objectid
,
2813 chunk_objectid
+ size
- 1,
2814 bit
| EXTENT_LOCKED
, GFP_NOFS
);
2815 set_state_private(block_group_cache
, chunk_objectid
,
2816 (unsigned long)cache
);
2818 ret
= btrfs_insert_item(trans
, extent_root
, &cache
->key
, &cache
->item
,
2819 sizeof(cache
->item
));
2822 finish_current_insert(trans
, extent_root
);
2823 ret
= del_pending_extents(trans
, extent_root
);
2825 set_avail_alloc_bits(extent_root
->fs_info
, type
);