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"
28 #define BLOCK_GROUP_DATA EXTENT_WRITEBACK
29 #define BLOCK_GROUP_METADATA EXTENT_UPTODATE
30 #define BLOCK_GROUP_DIRTY EXTENT_DIRTY
32 static int finish_current_insert(struct btrfs_trans_handle
*trans
, struct
33 btrfs_root
*extent_root
);
34 static int del_pending_extents(struct btrfs_trans_handle
*trans
, struct
35 btrfs_root
*extent_root
);
36 static int find_previous_extent(struct btrfs_root
*root
,
37 struct btrfs_path
*path
)
39 struct btrfs_key found_key
;
40 struct extent_buffer
*leaf
;
44 if (path
->slots
[0] == 0) {
45 ret
= btrfs_prev_leaf(root
, path
);
51 leaf
= path
->nodes
[0];
52 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
53 if (found_key
.type
== BTRFS_EXTENT_ITEM_KEY
)
59 static int cache_block_group(struct btrfs_root
*root
,
60 struct btrfs_block_group_cache
*block_group
)
62 struct btrfs_path
*path
;
65 struct extent_buffer
*leaf
;
66 struct extent_io_tree
*free_space_cache
;
76 root
= root
->fs_info
->extent_root
;
77 free_space_cache
= &root
->fs_info
->free_space_cache
;
79 if (block_group
->cached
)
82 path
= btrfs_alloc_path();
87 first_free
= block_group
->key
.objectid
;
88 key
.objectid
= block_group
->key
.objectid
;
90 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
91 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
94 ret
= find_previous_extent(root
, path
);
98 leaf
= path
->nodes
[0];
99 btrfs_item_key_to_cpu(leaf
, &key
, path
->slots
[0]);
100 if (key
.objectid
+ key
.offset
> first_free
)
101 first_free
= key
.objectid
+ key
.offset
;
104 leaf
= path
->nodes
[0];
105 slot
= path
->slots
[0];
106 if (slot
>= btrfs_header_nritems(leaf
)) {
107 ret
= btrfs_next_leaf(root
, path
);
116 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
117 if (key
.objectid
< block_group
->key
.objectid
) {
120 if (key
.objectid
>= block_group
->key
.objectid
+
121 block_group
->key
.offset
) {
125 if (btrfs_key_type(&key
) == BTRFS_EXTENT_ITEM_KEY
) {
130 if (key
.objectid
> last
) {
131 hole_size
= key
.objectid
- last
;
132 set_extent_dirty(free_space_cache
, last
,
133 last
+ hole_size
- 1,
136 last
= key
.objectid
+ key
.offset
;
144 if (block_group
->key
.objectid
+
145 block_group
->key
.offset
> last
) {
146 hole_size
= block_group
->key
.objectid
+
147 block_group
->key
.offset
- last
;
148 set_extent_dirty(free_space_cache
, last
,
149 last
+ hole_size
- 1, GFP_NOFS
);
151 block_group
->cached
= 1;
153 btrfs_free_path(path
);
157 struct btrfs_block_group_cache
*btrfs_lookup_block_group(struct
161 struct extent_io_tree
*block_group_cache
;
162 struct btrfs_block_group_cache
*block_group
= NULL
;
168 block_group_cache
= &info
->block_group_cache
;
169 ret
= find_first_extent_bit(block_group_cache
,
170 bytenr
, &start
, &end
,
171 BLOCK_GROUP_DATA
| BLOCK_GROUP_METADATA
);
175 ret
= get_state_private(block_group_cache
, start
, &ptr
);
179 block_group
= (struct btrfs_block_group_cache
*)(unsigned long)ptr
;
180 if (block_group
->key
.objectid
<= bytenr
&& bytenr
<
181 block_group
->key
.objectid
+ block_group
->key
.offset
)
185 static u64 noinline
find_search_start(struct btrfs_root
*root
,
186 struct btrfs_block_group_cache
**cache_ret
,
187 u64 search_start
, int num
, int data
)
190 struct btrfs_block_group_cache
*cache
= *cache_ret
;
191 struct extent_io_tree
*free_space_cache
;
192 struct extent_state
*state
;
202 total_fs_bytes
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
203 free_space_cache
= &root
->fs_info
->free_space_cache
;
206 ret
= cache_block_group(root
, cache
);
210 last
= max(search_start
, cache
->key
.objectid
);
212 spin_lock_irq(&free_space_cache
->lock
);
213 state
= find_first_extent_bit_state(free_space_cache
, last
, EXTENT_DIRTY
);
218 spin_unlock_irq(&free_space_cache
->lock
);
222 start
= max(last
, state
->start
);
223 last
= state
->end
+ 1;
224 if (last
- start
< num
) {
225 if (last
== cache
->key
.objectid
+ cache
->key
.offset
)
228 state
= extent_state_next(state
);
229 } while(state
&& !(state
->state
& EXTENT_DIRTY
));
232 spin_unlock_irq(&free_space_cache
->lock
);
233 if (data
!= BTRFS_BLOCK_GROUP_MIXED
&&
234 start
+ num
> cache
->key
.objectid
+ cache
->key
.offset
)
236 if (start
+ num
> total_fs_bytes
)
241 cache
= btrfs_lookup_block_group(root
->fs_info
, search_start
);
243 printk("Unable to find block group for %Lu\n",
251 last
= cache
->key
.objectid
+ cache
->key
.offset
;
253 cache
= btrfs_lookup_block_group(root
->fs_info
, last
);
254 if (!cache
|| cache
->key
.objectid
>= total_fs_bytes
) {
259 data
= BTRFS_BLOCK_GROUP_MIXED
;
264 if (cache_miss
&& !cache
->cached
) {
265 cache_block_group(root
, cache
);
267 cache
= btrfs_lookup_block_group(root
->fs_info
, last
);
269 cache
= btrfs_find_block_group(root
, cache
, last
, data
, 0);
277 static u64
div_factor(u64 num
, int factor
)
286 struct btrfs_block_group_cache
*btrfs_find_block_group(struct btrfs_root
*root
,
287 struct btrfs_block_group_cache
288 *hint
, u64 search_start
,
291 struct btrfs_block_group_cache
*cache
;
292 struct extent_io_tree
*block_group_cache
;
293 struct btrfs_block_group_cache
*found_group
= NULL
;
294 struct btrfs_fs_info
*info
= root
->fs_info
;
309 block_group_cache
= &info
->block_group_cache
;
310 total_fs_bytes
= btrfs_super_total_bytes(&root
->fs_info
->super_copy
);
315 if (data
== BTRFS_BLOCK_GROUP_MIXED
) {
316 bit
= BLOCK_GROUP_DATA
| BLOCK_GROUP_METADATA
;
319 bit
= BLOCK_GROUP_DATA
;
321 bit
= BLOCK_GROUP_METADATA
;
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
&& (shint
->data
== data
||
327 shint
->data
== BTRFS_BLOCK_GROUP_MIXED
)) {
328 used
= btrfs_block_group_used(&shint
->item
);
329 if (used
+ shint
->pinned
<
330 div_factor(shint
->key
.offset
, factor
)) {
335 if (hint
&& hint
->key
.objectid
< total_fs_bytes
&&
336 (hint
->data
== data
|| hint
->data
== BTRFS_BLOCK_GROUP_MIXED
)) {
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
)
373 free_check
= cache
->key
.offset
;
375 free_check
= div_factor(cache
->key
.offset
, factor
);
376 if (used
+ cache
->pinned
< free_check
) {
389 bit
= BLOCK_GROUP_DATA
| BLOCK_GROUP_METADATA
;
397 static u64
hash_extent_ref(u64 root_objectid
, u64 ref_generation
,
398 u64 owner
, u64 owner_offset
)
400 u32 high_crc
= ~(u32
)0;
401 u32 low_crc
= ~(u32
)0;
404 lenum
= cpu_to_le64(root_objectid
);
405 high_crc
= crc32c(high_crc
, &lenum
, sizeof(lenum
));
406 lenum
= cpu_to_le64(ref_generation
);
407 low_crc
= crc32c(low_crc
, &lenum
, sizeof(lenum
));
408 if (owner
>= BTRFS_FIRST_FREE_OBJECTID
) {
409 lenum
= cpu_to_le64(owner
);
410 low_crc
= crc32c(low_crc
, &lenum
, sizeof(lenum
));
411 lenum
= cpu_to_le64(owner_offset
);
412 low_crc
= crc32c(low_crc
, &lenum
, sizeof(lenum
));
414 return ((u64
)high_crc
<< 32) | (u64
)low_crc
;
417 static int match_extent_ref(struct extent_buffer
*leaf
,
418 struct btrfs_extent_ref
*disk_ref
,
419 struct btrfs_extent_ref
*cpu_ref
)
424 if (cpu_ref
->objectid
)
425 len
= sizeof(*cpu_ref
);
427 len
= 2 * sizeof(u64
);
428 ret
= memcmp_extent_buffer(leaf
, cpu_ref
, (unsigned long)disk_ref
,
433 static int noinline
lookup_extent_backref(struct btrfs_trans_handle
*trans
,
434 struct btrfs_root
*root
,
435 struct btrfs_path
*path
, u64 bytenr
,
437 u64 ref_generation
, u64 owner
,
438 u64 owner_offset
, int del
)
441 struct btrfs_key key
;
442 struct btrfs_key found_key
;
443 struct btrfs_extent_ref ref
;
444 struct extent_buffer
*leaf
;
445 struct btrfs_extent_ref
*disk_ref
;
449 btrfs_set_stack_ref_root(&ref
, root_objectid
);
450 btrfs_set_stack_ref_generation(&ref
, ref_generation
);
451 btrfs_set_stack_ref_objectid(&ref
, owner
);
452 btrfs_set_stack_ref_offset(&ref
, owner_offset
);
454 hash
= hash_extent_ref(root_objectid
, ref_generation
, owner
,
457 key
.objectid
= bytenr
;
458 key
.type
= BTRFS_EXTENT_REF_KEY
;
461 ret
= btrfs_search_slot(trans
, root
, &key
, path
,
465 leaf
= path
->nodes
[0];
467 u32 nritems
= btrfs_header_nritems(leaf
);
468 if (path
->slots
[0] >= nritems
) {
469 ret2
= btrfs_next_leaf(root
, path
);
472 leaf
= path
->nodes
[0];
474 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
475 if (found_key
.objectid
!= bytenr
||
476 found_key
.type
!= BTRFS_EXTENT_REF_KEY
)
478 key
.offset
= found_key
.offset
;
480 btrfs_release_path(root
, path
);
484 disk_ref
= btrfs_item_ptr(path
->nodes
[0],
486 struct btrfs_extent_ref
);
487 if (match_extent_ref(path
->nodes
[0], disk_ref
, &ref
)) {
491 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
492 key
.offset
= found_key
.offset
+ 1;
493 btrfs_release_path(root
, path
);
500 * Back reference rules. Back refs have three main goals:
502 * 1) differentiate between all holders of references to an extent so that
503 * when a reference is dropped we can make sure it was a valid reference
504 * before freeing the extent.
506 * 2) Provide enough information to quickly find the holders of an extent
507 * if we notice a given block is corrupted or bad.
509 * 3) Make it easy to migrate blocks for FS shrinking or storage pool
510 * maintenance. This is actually the same as #2, but with a slightly
511 * different use case.
513 * File extents can be referenced by:
515 * - multiple snapshots, subvolumes, or different generations in one subvol
516 * - different files inside a single subvolume (in theory, not implemented yet)
517 * - different offsets inside a file (bookend extents in file.c)
519 * The extent ref structure has fields for:
521 * - Objectid of the subvolume root
522 * - Generation number of the tree holding the reference
523 * - objectid of the file holding the reference
524 * - offset in the file corresponding to the key holding the reference
526 * When a file extent is allocated the fields are filled in:
527 * (root_key.objectid, trans->transid, inode objectid, offset in file)
529 * When a leaf is cow'd new references are added for every file extent found
530 * in the leaf. It looks the same as the create case, but trans->transid
531 * will be different when the block is cow'd.
533 * (root_key.objectid, trans->transid, inode objectid, offset in file)
535 * When a file extent is removed either during snapshot deletion or file
536 * truncation, the corresponding back reference is found
539 * (btrfs_header_owner(leaf), btrfs_header_generation(leaf),
540 * inode objectid, offset in file)
542 * Btree extents can be referenced by:
544 * - Different subvolumes
545 * - Different generations of the same subvolume
547 * Storing sufficient information for a full reverse mapping of a btree
548 * block would require storing the lowest key of the block in the backref,
549 * and it would require updating that lowest key either before write out or
550 * every time it changed. Instead, the objectid of the lowest key is stored
551 * along with the level of the tree block. This provides a hint
552 * about where in the btree the block can be found. Searches through the
553 * btree only need to look for a pointer to that block, so they stop one
554 * level higher than the level recorded in the backref.
556 * Some btrees do not do reference counting on their extents. These
557 * include the extent tree and the tree of tree roots. Backrefs for these
558 * trees always have a generation of zero.
560 * When a tree block is created, back references are inserted:
562 * (root->root_key.objectid, trans->transid or zero, level, lowest_key_objectid)
564 * When a tree block is cow'd in a reference counted root,
565 * new back references are added for all the blocks it points to.
566 * These are of the form (trans->transid will have increased since creation):
568 * (root->root_key.objectid, trans->transid, level, lowest_key_objectid)
570 * Because the lowest_key_objectid and the level are just hints
571 * they are not used when backrefs are deleted. When a backref is deleted:
573 * if backref was for a tree root:
574 * root_objectid = root->root_key.objectid
576 * root_objectid = btrfs_header_owner(parent)
578 * (root_objectid, btrfs_header_generation(parent) or zero, 0, 0)
580 * Back Reference Key hashing:
582 * Back references have four fields, each 64 bits long. Unfortunately,
583 * This is hashed into a single 64 bit number and placed into the key offset.
584 * The key objectid corresponds to the first byte in the extent, and the
585 * key type is set to BTRFS_EXTENT_REF_KEY
587 int btrfs_insert_extent_backref(struct btrfs_trans_handle
*trans
,
588 struct btrfs_root
*root
,
589 struct btrfs_path
*path
, u64 bytenr
,
590 u64 root_objectid
, u64 ref_generation
,
591 u64 owner
, u64 owner_offset
)
594 struct btrfs_key key
;
595 struct btrfs_extent_ref ref
;
596 struct btrfs_extent_ref
*disk_ref
;
599 btrfs_set_stack_ref_root(&ref
, root_objectid
);
600 btrfs_set_stack_ref_generation(&ref
, ref_generation
);
601 btrfs_set_stack_ref_objectid(&ref
, owner
);
602 btrfs_set_stack_ref_offset(&ref
, owner_offset
);
604 hash
= hash_extent_ref(root_objectid
, ref_generation
, owner
,
607 key
.objectid
= bytenr
;
608 key
.type
= BTRFS_EXTENT_REF_KEY
;
610 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
, sizeof(ref
));
611 while (ret
== -EEXIST
) {
612 disk_ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
613 struct btrfs_extent_ref
);
614 if (match_extent_ref(path
->nodes
[0], disk_ref
, &ref
))
617 btrfs_release_path(root
, path
);
618 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
623 disk_ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
624 struct btrfs_extent_ref
);
625 write_extent_buffer(path
->nodes
[0], &ref
, (unsigned long)disk_ref
,
627 btrfs_mark_buffer_dirty(path
->nodes
[0]);
629 btrfs_release_path(root
, path
);
633 int btrfs_inc_extent_ref(struct btrfs_trans_handle
*trans
,
634 struct btrfs_root
*root
,
635 u64 bytenr
, u64 num_bytes
,
636 u64 root_objectid
, u64 ref_generation
,
637 u64 owner
, u64 owner_offset
)
639 struct btrfs_path
*path
;
641 struct btrfs_key key
;
642 struct extent_buffer
*l
;
643 struct btrfs_extent_item
*item
;
646 WARN_ON(num_bytes
< root
->sectorsize
);
647 path
= btrfs_alloc_path();
652 key
.objectid
= bytenr
;
653 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
654 key
.offset
= num_bytes
;
655 ret
= btrfs_search_slot(trans
, root
->fs_info
->extent_root
, &key
, path
,
664 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
665 refs
= btrfs_extent_refs(l
, item
);
666 btrfs_set_extent_refs(l
, item
, refs
+ 1);
667 btrfs_mark_buffer_dirty(path
->nodes
[0]);
669 btrfs_release_path(root
->fs_info
->extent_root
, path
);
672 ret
= btrfs_insert_extent_backref(trans
, root
->fs_info
->extent_root
,
673 path
, bytenr
, root_objectid
,
674 ref_generation
, owner
, owner_offset
);
676 finish_current_insert(trans
, root
->fs_info
->extent_root
);
677 del_pending_extents(trans
, root
->fs_info
->extent_root
);
679 btrfs_free_path(path
);
683 int btrfs_extent_post_op(struct btrfs_trans_handle
*trans
,
684 struct btrfs_root
*root
)
686 finish_current_insert(trans
, root
->fs_info
->extent_root
);
687 del_pending_extents(trans
, root
->fs_info
->extent_root
);
691 static int lookup_extent_ref(struct btrfs_trans_handle
*trans
,
692 struct btrfs_root
*root
, u64 bytenr
,
693 u64 num_bytes
, u32
*refs
)
695 struct btrfs_path
*path
;
697 struct btrfs_key key
;
698 struct extent_buffer
*l
;
699 struct btrfs_extent_item
*item
;
701 WARN_ON(num_bytes
< root
->sectorsize
);
702 path
= btrfs_alloc_path();
704 key
.objectid
= bytenr
;
705 key
.offset
= num_bytes
;
706 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
707 ret
= btrfs_search_slot(trans
, root
->fs_info
->extent_root
, &key
, path
,
712 btrfs_print_leaf(root
, path
->nodes
[0]);
713 printk("failed to find block number %Lu\n", bytenr
);
717 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
718 *refs
= btrfs_extent_refs(l
, item
);
720 btrfs_free_path(path
);
724 u32
btrfs_count_snapshots_in_path(struct btrfs_root
*root
,
725 struct btrfs_path
*count_path
,
728 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
729 struct btrfs_path
*path
;
732 u64 root_objectid
= root
->root_key
.objectid
;
737 struct btrfs_key key
;
738 struct btrfs_key found_key
;
739 struct extent_buffer
*l
;
740 struct btrfs_extent_item
*item
;
741 struct btrfs_extent_ref
*ref_item
;
744 path
= btrfs_alloc_path();
747 bytenr
= first_extent
;
749 bytenr
= count_path
->nodes
[level
]->start
;
752 key
.objectid
= bytenr
;
755 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
756 ret
= btrfs_search_slot(NULL
, extent_root
, &key
, path
, 0, 0);
762 btrfs_item_key_to_cpu(l
, &found_key
, path
->slots
[0]);
764 if (found_key
.objectid
!= bytenr
||
765 found_key
.type
!= BTRFS_EXTENT_ITEM_KEY
) {
769 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
772 nritems
= btrfs_header_nritems(l
);
773 if (path
->slots
[0] >= nritems
) {
774 ret
= btrfs_next_leaf(extent_root
, path
);
779 btrfs_item_key_to_cpu(l
, &found_key
, path
->slots
[0]);
780 if (found_key
.objectid
!= bytenr
)
783 if (found_key
.type
!= BTRFS_EXTENT_REF_KEY
) {
789 ref_item
= btrfs_item_ptr(l
, path
->slots
[0],
790 struct btrfs_extent_ref
);
791 found_objectid
= btrfs_ref_root(l
, ref_item
);
793 if (found_objectid
!= root_objectid
) {
800 if (cur_count
== 0) {
804 if (level
>= 0 && root
->node
== count_path
->nodes
[level
])
807 btrfs_release_path(root
, path
);
811 btrfs_free_path(path
);
814 int btrfs_inc_root_ref(struct btrfs_trans_handle
*trans
,
815 struct btrfs_root
*root
, u64 owner_objectid
)
821 struct btrfs_disk_key disk_key
;
823 level
= btrfs_header_level(root
->node
);
824 generation
= trans
->transid
;
825 nritems
= btrfs_header_nritems(root
->node
);
828 btrfs_item_key(root
->node
, &disk_key
, 0);
830 btrfs_node_key(root
->node
, &disk_key
, 0);
831 key_objectid
= btrfs_disk_key_objectid(&disk_key
);
835 return btrfs_inc_extent_ref(trans
, root
, root
->node
->start
,
836 root
->node
->len
, owner_objectid
,
837 generation
, level
, key_objectid
);
840 int btrfs_inc_ref(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
841 struct extent_buffer
*buf
)
845 struct btrfs_key key
;
846 struct btrfs_file_extent_item
*fi
;
855 level
= btrfs_header_level(buf
);
856 nritems
= btrfs_header_nritems(buf
);
857 for (i
= 0; i
< nritems
; i
++) {
860 btrfs_item_key_to_cpu(buf
, &key
, i
);
861 if (btrfs_key_type(&key
) != BTRFS_EXTENT_DATA_KEY
)
863 fi
= btrfs_item_ptr(buf
, i
,
864 struct btrfs_file_extent_item
);
865 if (btrfs_file_extent_type(buf
, fi
) ==
866 BTRFS_FILE_EXTENT_INLINE
)
868 disk_bytenr
= btrfs_file_extent_disk_bytenr(buf
, fi
);
869 if (disk_bytenr
== 0)
871 ret
= btrfs_inc_extent_ref(trans
, root
, disk_bytenr
,
872 btrfs_file_extent_disk_num_bytes(buf
, fi
),
873 root
->root_key
.objectid
, trans
->transid
,
874 key
.objectid
, key
.offset
);
880 bytenr
= btrfs_node_blockptr(buf
, i
);
881 btrfs_node_key_to_cpu(buf
, &key
, i
);
882 ret
= btrfs_inc_extent_ref(trans
, root
, bytenr
,
883 btrfs_level_size(root
, level
- 1),
884 root
->root_key
.objectid
,
886 level
- 1, key
.objectid
);
897 for (i
=0; i
< faili
; i
++) {
900 btrfs_item_key_to_cpu(buf
, &key
, i
);
901 if (btrfs_key_type(&key
) != BTRFS_EXTENT_DATA_KEY
)
903 fi
= btrfs_item_ptr(buf
, i
,
904 struct btrfs_file_extent_item
);
905 if (btrfs_file_extent_type(buf
, fi
) ==
906 BTRFS_FILE_EXTENT_INLINE
)
908 disk_bytenr
= btrfs_file_extent_disk_bytenr(buf
, fi
);
909 if (disk_bytenr
== 0)
911 err
= btrfs_free_extent(trans
, root
, disk_bytenr
,
912 btrfs_file_extent_disk_num_bytes(buf
,
916 bytenr
= btrfs_node_blockptr(buf
, i
);
917 err
= btrfs_free_extent(trans
, root
, bytenr
,
918 btrfs_level_size(root
, level
- 1), 0);
926 static int write_one_cache_group(struct btrfs_trans_handle
*trans
,
927 struct btrfs_root
*root
,
928 struct btrfs_path
*path
,
929 struct btrfs_block_group_cache
*cache
)
933 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
935 struct extent_buffer
*leaf
;
937 ret
= btrfs_search_slot(trans
, extent_root
, &cache
->key
, path
, 0, 1);
942 leaf
= path
->nodes
[0];
943 bi
= btrfs_item_ptr_offset(leaf
, path
->slots
[0]);
944 write_extent_buffer(leaf
, &cache
->item
, bi
, sizeof(cache
->item
));
945 btrfs_mark_buffer_dirty(leaf
);
946 btrfs_release_path(extent_root
, path
);
948 finish_current_insert(trans
, extent_root
);
949 pending_ret
= del_pending_extents(trans
, extent_root
);
958 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle
*trans
,
959 struct btrfs_root
*root
)
961 struct extent_io_tree
*block_group_cache
;
962 struct btrfs_block_group_cache
*cache
;
966 struct btrfs_path
*path
;
972 block_group_cache
= &root
->fs_info
->block_group_cache
;
973 path
= btrfs_alloc_path();
978 ret
= find_first_extent_bit(block_group_cache
, last
,
979 &start
, &end
, BLOCK_GROUP_DIRTY
);
984 ret
= get_state_private(block_group_cache
, start
, &ptr
);
988 cache
= (struct btrfs_block_group_cache
*)(unsigned long)ptr
;
989 err
= write_one_cache_group(trans
, root
,
992 * if we fail to write the cache group, we want
993 * to keep it marked dirty in hopes that a later
1000 clear_extent_bits(block_group_cache
, start
, end
,
1001 BLOCK_GROUP_DIRTY
, GFP_NOFS
);
1003 btrfs_free_path(path
);
1007 static int update_block_group(struct btrfs_trans_handle
*trans
,
1008 struct btrfs_root
*root
,
1009 u64 bytenr
, u64 num_bytes
, int alloc
,
1010 int mark_free
, int data
)
1012 struct btrfs_block_group_cache
*cache
;
1013 struct btrfs_fs_info
*info
= root
->fs_info
;
1014 u64 total
= num_bytes
;
1021 cache
= btrfs_lookup_block_group(info
, bytenr
);
1025 byte_in_group
= bytenr
- cache
->key
.objectid
;
1026 WARN_ON(byte_in_group
> cache
->key
.offset
);
1027 start
= cache
->key
.objectid
;
1028 end
= start
+ cache
->key
.offset
- 1;
1029 set_extent_bits(&info
->block_group_cache
, start
, end
,
1030 BLOCK_GROUP_DIRTY
, GFP_NOFS
);
1032 old_val
= btrfs_block_group_used(&cache
->item
);
1033 num_bytes
= min(total
, cache
->key
.offset
- byte_in_group
);
1035 if (cache
->data
!= data
&&
1036 old_val
< (cache
->key
.offset
>> 1)) {
1041 bit_to_clear
= BLOCK_GROUP_METADATA
;
1042 bit_to_set
= BLOCK_GROUP_DATA
;
1043 cache
->item
.flags
&=
1044 ~BTRFS_BLOCK_GROUP_MIXED
;
1045 cache
->item
.flags
|=
1046 BTRFS_BLOCK_GROUP_DATA
;
1048 bit_to_clear
= BLOCK_GROUP_DATA
;
1049 bit_to_set
= BLOCK_GROUP_METADATA
;
1050 cache
->item
.flags
&=
1051 ~BTRFS_BLOCK_GROUP_MIXED
;
1052 cache
->item
.flags
&=
1053 ~BTRFS_BLOCK_GROUP_DATA
;
1055 clear_extent_bits(&info
->block_group_cache
,
1056 start
, end
, bit_to_clear
,
1058 set_extent_bits(&info
->block_group_cache
,
1059 start
, end
, bit_to_set
,
1061 } else if (cache
->data
!= data
&&
1062 cache
->data
!= BTRFS_BLOCK_GROUP_MIXED
) {
1063 cache
->data
= BTRFS_BLOCK_GROUP_MIXED
;
1064 set_extent_bits(&info
->block_group_cache
,
1067 BLOCK_GROUP_METADATA
,
1070 old_val
+= num_bytes
;
1072 old_val
-= num_bytes
;
1074 set_extent_dirty(&info
->free_space_cache
,
1075 bytenr
, bytenr
+ num_bytes
- 1,
1079 btrfs_set_block_group_used(&cache
->item
, old_val
);
1081 bytenr
+= num_bytes
;
1085 static int update_pinned_extents(struct btrfs_root
*root
,
1086 u64 bytenr
, u64 num
, int pin
)
1089 struct btrfs_block_group_cache
*cache
;
1090 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1093 set_extent_dirty(&fs_info
->pinned_extents
,
1094 bytenr
, bytenr
+ num
- 1, GFP_NOFS
);
1096 clear_extent_dirty(&fs_info
->pinned_extents
,
1097 bytenr
, bytenr
+ num
- 1, GFP_NOFS
);
1100 cache
= btrfs_lookup_block_group(fs_info
, bytenr
);
1102 len
= min(num
, cache
->key
.offset
-
1103 (bytenr
- cache
->key
.objectid
));
1105 cache
->pinned
+= len
;
1106 fs_info
->total_pinned
+= len
;
1108 cache
->pinned
-= len
;
1109 fs_info
->total_pinned
-= len
;
1117 int btrfs_copy_pinned(struct btrfs_root
*root
, struct extent_io_tree
*copy
)
1122 struct extent_io_tree
*pinned_extents
= &root
->fs_info
->pinned_extents
;
1126 ret
= find_first_extent_bit(pinned_extents
, last
,
1127 &start
, &end
, EXTENT_DIRTY
);
1130 set_extent_dirty(copy
, start
, end
, GFP_NOFS
);
1136 int btrfs_finish_extent_commit(struct btrfs_trans_handle
*trans
,
1137 struct btrfs_root
*root
,
1138 struct extent_io_tree
*unpin
)
1143 struct extent_io_tree
*free_space_cache
;
1144 free_space_cache
= &root
->fs_info
->free_space_cache
;
1147 ret
= find_first_extent_bit(unpin
, 0, &start
, &end
,
1151 update_pinned_extents(root
, start
, end
+ 1 - start
, 0);
1152 clear_extent_dirty(unpin
, start
, end
, GFP_NOFS
);
1153 set_extent_dirty(free_space_cache
, start
, end
, GFP_NOFS
);
1158 static int finish_current_insert(struct btrfs_trans_handle
*trans
,
1159 struct btrfs_root
*extent_root
)
1163 struct btrfs_fs_info
*info
= extent_root
->fs_info
;
1164 struct extent_buffer
*eb
;
1165 struct btrfs_path
*path
;
1166 struct btrfs_key ins
;
1167 struct btrfs_disk_key first
;
1168 struct btrfs_extent_item extent_item
;
1173 btrfs_set_stack_extent_refs(&extent_item
, 1);
1174 btrfs_set_key_type(&ins
, BTRFS_EXTENT_ITEM_KEY
);
1175 path
= btrfs_alloc_path();
1178 ret
= find_first_extent_bit(&info
->extent_ins
, 0, &start
,
1179 &end
, EXTENT_LOCKED
);
1183 ins
.objectid
= start
;
1184 ins
.offset
= end
+ 1 - start
;
1185 err
= btrfs_insert_item(trans
, extent_root
, &ins
,
1186 &extent_item
, sizeof(extent_item
));
1187 clear_extent_bits(&info
->extent_ins
, start
, end
, EXTENT_LOCKED
,
1189 eb
= read_tree_block(extent_root
, ins
.objectid
, ins
.offset
);
1190 level
= btrfs_header_level(eb
);
1192 btrfs_item_key(eb
, &first
, 0);
1194 btrfs_node_key(eb
, &first
, 0);
1196 err
= btrfs_insert_extent_backref(trans
, extent_root
, path
,
1197 start
, extent_root
->root_key
.objectid
,
1199 btrfs_disk_key_objectid(&first
));
1201 free_extent_buffer(eb
);
1203 btrfs_free_path(path
);
1207 static int pin_down_bytes(struct btrfs_root
*root
, u64 bytenr
, u32 num_bytes
,
1211 struct extent_buffer
*buf
;
1214 buf
= btrfs_find_tree_block(root
, bytenr
, num_bytes
);
1216 if (btrfs_buffer_uptodate(buf
)) {
1218 root
->fs_info
->running_transaction
->transid
;
1219 u64 header_transid
=
1220 btrfs_header_generation(buf
);
1221 if (header_transid
== transid
) {
1222 clean_tree_block(NULL
, root
, buf
);
1223 free_extent_buffer(buf
);
1227 free_extent_buffer(buf
);
1229 update_pinned_extents(root
, bytenr
, num_bytes
, 1);
1231 set_extent_bits(&root
->fs_info
->pending_del
,
1232 bytenr
, bytenr
+ num_bytes
- 1,
1233 EXTENT_LOCKED
, GFP_NOFS
);
1240 * remove an extent from the root, returns 0 on success
1242 static int __free_extent(struct btrfs_trans_handle
*trans
, struct btrfs_root
1243 *root
, u64 bytenr
, u64 num_bytes
,
1244 u64 root_objectid
, u64 ref_generation
,
1245 u64 owner_objectid
, u64 owner_offset
, int pin
,
1248 struct btrfs_path
*path
;
1249 struct btrfs_key key
;
1250 struct btrfs_fs_info
*info
= root
->fs_info
;
1251 struct btrfs_root
*extent_root
= info
->extent_root
;
1252 struct extent_buffer
*leaf
;
1254 struct btrfs_extent_item
*ei
;
1257 key
.objectid
= bytenr
;
1258 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
1259 key
.offset
= num_bytes
;
1260 path
= btrfs_alloc_path();
1265 ret
= lookup_extent_backref(trans
, extent_root
, path
,
1266 bytenr
, root_objectid
,
1268 owner_objectid
, owner_offset
, 1);
1270 ret
= btrfs_del_item(trans
, extent_root
, path
);
1272 btrfs_print_leaf(extent_root
, path
->nodes
[0]);
1274 printk("Unable to find ref byte nr %Lu root %Lu "
1275 " gen %Lu owner %Lu offset %Lu\n", bytenr
,
1276 root_objectid
, ref_generation
, owner_objectid
,
1279 btrfs_release_path(extent_root
, path
);
1280 ret
= btrfs_search_slot(trans
, extent_root
, &key
, path
, -1, 1);
1285 leaf
= path
->nodes
[0];
1286 ei
= btrfs_item_ptr(leaf
, path
->slots
[0],
1287 struct btrfs_extent_item
);
1288 refs
= btrfs_extent_refs(leaf
, ei
);
1291 btrfs_set_extent_refs(leaf
, ei
, refs
);
1292 btrfs_mark_buffer_dirty(leaf
);
1299 ret
= pin_down_bytes(root
, bytenr
, num_bytes
, 0);
1305 /* block accounting for super block */
1306 super_used
= btrfs_super_bytes_used(&info
->super_copy
);
1307 btrfs_set_super_bytes_used(&info
->super_copy
,
1308 super_used
- num_bytes
);
1310 /* block accounting for root item */
1311 root_used
= btrfs_root_used(&root
->root_item
);
1312 btrfs_set_root_used(&root
->root_item
,
1313 root_used
- num_bytes
);
1315 ret
= btrfs_del_item(trans
, extent_root
, path
);
1319 ret
= update_block_group(trans
, root
, bytenr
, num_bytes
, 0,
1323 btrfs_free_path(path
);
1324 finish_current_insert(trans
, extent_root
);
1329 * find all the blocks marked as pending in the radix tree and remove
1330 * them from the extent map
1332 static int del_pending_extents(struct btrfs_trans_handle
*trans
, struct
1333 btrfs_root
*extent_root
)
1339 struct extent_io_tree
*pending_del
;
1340 struct extent_io_tree
*pinned_extents
;
1342 pending_del
= &extent_root
->fs_info
->pending_del
;
1343 pinned_extents
= &extent_root
->fs_info
->pinned_extents
;
1346 ret
= find_first_extent_bit(pending_del
, 0, &start
, &end
,
1350 update_pinned_extents(extent_root
, start
, end
+ 1 - start
, 1);
1351 clear_extent_bits(pending_del
, start
, end
, EXTENT_LOCKED
,
1353 ret
= __free_extent(trans
, extent_root
,
1354 start
, end
+ 1 - start
,
1355 extent_root
->root_key
.objectid
,
1364 * remove an extent from the root, returns 0 on success
1366 int btrfs_free_extent(struct btrfs_trans_handle
*trans
, struct btrfs_root
1367 *root
, u64 bytenr
, u64 num_bytes
,
1368 u64 root_objectid
, u64 ref_generation
,
1369 u64 owner_objectid
, u64 owner_offset
, int pin
)
1371 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
1375 WARN_ON(num_bytes
< root
->sectorsize
);
1376 if (!root
->ref_cows
)
1379 if (root
== extent_root
) {
1380 pin_down_bytes(root
, bytenr
, num_bytes
, 1);
1383 ret
= __free_extent(trans
, root
, bytenr
, num_bytes
, root_objectid
,
1384 ref_generation
, owner_objectid
, owner_offset
,
1386 pending_ret
= del_pending_extents(trans
, root
->fs_info
->extent_root
);
1387 return ret
? ret
: pending_ret
;
1390 static u64
stripe_align(struct btrfs_root
*root
, u64 val
)
1392 u64 mask
= ((u64
)root
->stripesize
- 1);
1393 u64 ret
= (val
+ mask
) & ~mask
;
1398 * walks the btree of allocated extents and find a hole of a given size.
1399 * The key ins is changed to record the hole:
1400 * ins->objectid == block start
1401 * ins->flags = BTRFS_EXTENT_ITEM_KEY
1402 * ins->offset == number of blocks
1403 * Any available blocks before search_start are skipped.
1405 static int noinline
find_free_extent(struct btrfs_trans_handle
*trans
,
1406 struct btrfs_root
*orig_root
,
1407 u64 num_bytes
, u64 empty_size
,
1408 u64 search_start
, u64 search_end
,
1409 u64 hint_byte
, struct btrfs_key
*ins
,
1410 u64 exclude_start
, u64 exclude_nr
,
1413 struct btrfs_path
*path
;
1414 struct btrfs_key key
;
1420 u64
*last_ptr
= NULL
;
1421 u64 orig_search_start
= search_start
;
1423 struct extent_buffer
*l
;
1424 struct btrfs_root
* root
= orig_root
->fs_info
->extent_root
;
1425 struct btrfs_fs_info
*info
= root
->fs_info
;
1426 u64 total_needed
= num_bytes
;
1428 struct btrfs_block_group_cache
*block_group
;
1434 WARN_ON(num_bytes
< root
->sectorsize
);
1435 btrfs_set_key_type(ins
, BTRFS_EXTENT_ITEM_KEY
);
1437 level
= btrfs_header_level(root
->node
);
1439 if (num_bytes
>= 32 * 1024 * 1024 && hint_byte
) {
1440 data
= BTRFS_BLOCK_GROUP_MIXED
;
1444 last_ptr
= &root
->fs_info
->last_alloc
;
1445 empty_cluster
= 128 * 1024;
1448 if (data
&& btrfs_test_opt(root
, SSD
)) {
1449 last_ptr
= &root
->fs_info
->last_data_alloc
;
1450 empty_cluster
= 2 * 1024 * 1024;
1455 hint_byte
= *last_ptr
;
1457 hint_byte
= hint_byte
&
1458 ~((u64
)BTRFS_BLOCK_GROUP_SIZE
- 1);
1459 empty_size
+= empty_cluster
;
1461 search_start
= max(search_start
, hint_byte
);
1464 search_end
= min(search_end
,
1465 btrfs_super_total_bytes(&info
->super_copy
));
1467 block_group
= btrfs_lookup_block_group(info
, hint_byte
);
1469 hint_byte
= search_start
;
1470 block_group
= btrfs_find_block_group(root
, block_group
,
1471 hint_byte
, data
, 1);
1473 block_group
= btrfs_find_block_group(root
,
1475 search_start
, data
, 1);
1478 total_needed
+= empty_size
;
1479 path
= btrfs_alloc_path();
1482 block_group
= btrfs_lookup_block_group(info
, search_start
);
1484 block_group
= btrfs_lookup_block_group(info
,
1487 search_start
= find_search_start(root
, &block_group
, search_start
,
1488 total_needed
, data
);
1490 if (last_ptr
&& *last_ptr
&& search_start
!= *last_ptr
) {
1493 empty_size
+= empty_cluster
;
1494 total_needed
+= empty_size
;
1496 search_start
= find_search_start(root
, &block_group
,
1497 search_start
, total_needed
,
1501 search_start
= stripe_align(root
, search_start
);
1502 cached_start
= search_start
;
1503 btrfs_init_path(path
);
1504 ins
->objectid
= search_start
;
1509 ret
= btrfs_search_slot(trans
, root
, ins
, path
, 0, 0);
1512 ret
= find_previous_extent(root
, path
);
1516 btrfs_item_key_to_cpu(l
, &key
, path
->slots
[0]);
1519 slot
= path
->slots
[0];
1520 if (slot
>= btrfs_header_nritems(l
)) {
1521 ret
= btrfs_next_leaf(root
, path
);
1527 search_start
= max(search_start
,
1528 block_group
->key
.objectid
);
1530 aligned
= stripe_align(root
, search_start
);
1531 ins
->objectid
= aligned
;
1532 if (aligned
>= search_end
) {
1536 ins
->offset
= search_end
- aligned
;
1540 ins
->objectid
= stripe_align(root
,
1541 last_byte
> search_start
?
1542 last_byte
: search_start
);
1543 if (search_end
<= ins
->objectid
) {
1547 ins
->offset
= search_end
- ins
->objectid
;
1548 BUG_ON(ins
->objectid
>= search_end
);
1551 btrfs_item_key_to_cpu(l
, &key
, slot
);
1553 if (key
.objectid
>= search_start
&& key
.objectid
> last_byte
&&
1555 if (last_byte
< search_start
)
1556 last_byte
= search_start
;
1557 aligned
= stripe_align(root
, last_byte
);
1558 hole_size
= key
.objectid
- aligned
;
1559 if (key
.objectid
> aligned
&& hole_size
>= num_bytes
) {
1560 ins
->objectid
= aligned
;
1561 ins
->offset
= hole_size
;
1565 if (btrfs_key_type(&key
) != BTRFS_EXTENT_ITEM_KEY
) {
1566 if (!start_found
&& btrfs_key_type(&key
) ==
1567 BTRFS_BLOCK_GROUP_ITEM_KEY
) {
1568 last_byte
= key
.objectid
;
1576 last_byte
= key
.objectid
+ key
.offset
;
1578 if (!full_scan
&& data
!= BTRFS_BLOCK_GROUP_MIXED
&&
1579 last_byte
>= block_group
->key
.objectid
+
1580 block_group
->key
.offset
) {
1581 btrfs_release_path(root
, path
);
1582 search_start
= block_group
->key
.objectid
+
1583 block_group
->key
.offset
;
1591 /* we have to make sure we didn't find an extent that has already
1592 * been allocated by the map tree or the original allocation
1594 btrfs_release_path(root
, path
);
1595 BUG_ON(ins
->objectid
< search_start
);
1597 if (ins
->objectid
+ num_bytes
>= search_end
)
1599 if (!full_scan
&& data
!= BTRFS_BLOCK_GROUP_MIXED
&&
1600 ins
->objectid
+ num_bytes
> block_group
->
1601 key
.objectid
+ block_group
->key
.offset
) {
1602 search_start
= block_group
->key
.objectid
+
1603 block_group
->key
.offset
;
1606 if (test_range_bit(&info
->extent_ins
, ins
->objectid
,
1607 ins
->objectid
+ num_bytes
-1, EXTENT_LOCKED
, 0)) {
1608 search_start
= ins
->objectid
+ num_bytes
;
1611 if (test_range_bit(&info
->pinned_extents
, ins
->objectid
,
1612 ins
->objectid
+ num_bytes
-1, EXTENT_DIRTY
, 0)) {
1613 search_start
= ins
->objectid
+ num_bytes
;
1616 if (exclude_nr
> 0 && (ins
->objectid
+ num_bytes
> exclude_start
&&
1617 ins
->objectid
< exclude_start
+ exclude_nr
)) {
1618 search_start
= exclude_start
+ exclude_nr
;
1622 block_group
= btrfs_lookup_block_group(info
, ins
->objectid
);
1624 trans
->block_group
= block_group
;
1626 ins
->offset
= num_bytes
;
1627 btrfs_free_path(path
);
1629 *last_ptr
= ins
->objectid
+ ins
->offset
;
1631 btrfs_super_total_bytes(&root
->fs_info
->super_copy
)) {
1638 if (search_start
+ num_bytes
>= search_end
) {
1640 search_start
= orig_search_start
;
1647 total_needed
-= empty_size
;
1649 data
= BTRFS_BLOCK_GROUP_MIXED
;
1653 block_group
= btrfs_lookup_block_group(info
, search_start
);
1655 block_group
= btrfs_find_block_group(root
, block_group
,
1656 search_start
, data
, 0);
1660 btrfs_release_path(root
, path
);
1661 btrfs_free_path(path
);
1665 * finds a free extent and does all the dirty work required for allocation
1666 * returns the key for the extent through ins, and a tree buffer for
1667 * the first block of the extent through buf.
1669 * returns 0 if everything worked, non-zero otherwise.
1671 int btrfs_alloc_extent(struct btrfs_trans_handle
*trans
,
1672 struct btrfs_root
*root
,
1673 u64 num_bytes
, u64 root_objectid
, u64 ref_generation
,
1674 u64 owner
, u64 owner_offset
,
1675 u64 empty_size
, u64 hint_byte
,
1676 u64 search_end
, struct btrfs_key
*ins
, int data
)
1682 u64 search_start
= 0;
1685 struct btrfs_fs_info
*info
= root
->fs_info
;
1686 struct btrfs_root
*extent_root
= info
->extent_root
;
1687 struct btrfs_extent_item
*extent_item
;
1688 struct btrfs_extent_ref
*ref
;
1689 struct btrfs_path
*path
;
1690 struct btrfs_key keys
[2];
1692 new_hint
= max(hint_byte
, root
->fs_info
->alloc_start
);
1693 if (new_hint
< btrfs_super_total_bytes(&info
->super_copy
))
1694 hint_byte
= new_hint
;
1696 WARN_ON(num_bytes
< root
->sectorsize
);
1697 ret
= find_free_extent(trans
, root
, num_bytes
, empty_size
,
1698 search_start
, search_end
, hint_byte
, ins
,
1699 trans
->alloc_exclude_start
,
1700 trans
->alloc_exclude_nr
, data
);
1705 /* block accounting for super block */
1706 super_used
= btrfs_super_bytes_used(&info
->super_copy
);
1707 btrfs_set_super_bytes_used(&info
->super_copy
, super_used
+ num_bytes
);
1709 /* block accounting for root item */
1710 root_used
= btrfs_root_used(&root
->root_item
);
1711 btrfs_set_root_used(&root
->root_item
, root_used
+ num_bytes
);
1713 clear_extent_dirty(&root
->fs_info
->free_space_cache
,
1714 ins
->objectid
, ins
->objectid
+ ins
->offset
- 1,
1717 if (root
== extent_root
) {
1718 set_extent_bits(&root
->fs_info
->extent_ins
, ins
->objectid
,
1719 ins
->objectid
+ ins
->offset
- 1,
1720 EXTENT_LOCKED
, GFP_NOFS
);
1725 WARN_ON(trans
->alloc_exclude_nr
);
1726 trans
->alloc_exclude_start
= ins
->objectid
;
1727 trans
->alloc_exclude_nr
= ins
->offset
;
1729 memcpy(&keys
[0], ins
, sizeof(*ins
));
1730 keys
[1].offset
= hash_extent_ref(root_objectid
, ref_generation
,
1731 owner
, owner_offset
);
1732 keys
[1].objectid
= ins
->objectid
;
1733 keys
[1].type
= BTRFS_EXTENT_REF_KEY
;
1734 sizes
[0] = sizeof(*extent_item
);
1735 sizes
[1] = sizeof(*ref
);
1737 path
= btrfs_alloc_path();
1740 ret
= btrfs_insert_empty_items(trans
, extent_root
, path
, keys
,
1744 extent_item
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
1745 struct btrfs_extent_item
);
1746 btrfs_set_extent_refs(path
->nodes
[0], extent_item
, 1);
1747 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0] + 1,
1748 struct btrfs_extent_ref
);
1750 btrfs_set_ref_root(path
->nodes
[0], ref
, root_objectid
);
1751 btrfs_set_ref_generation(path
->nodes
[0], ref
, ref_generation
);
1752 btrfs_set_ref_objectid(path
->nodes
[0], ref
, owner
);
1753 btrfs_set_ref_offset(path
->nodes
[0], ref
, owner_offset
);
1755 btrfs_mark_buffer_dirty(path
->nodes
[0]);
1757 trans
->alloc_exclude_start
= 0;
1758 trans
->alloc_exclude_nr
= 0;
1759 btrfs_free_path(path
);
1760 finish_current_insert(trans
, extent_root
);
1761 pending_ret
= del_pending_extents(trans
, extent_root
);
1771 ret
= update_block_group(trans
, root
, ins
->objectid
, ins
->offset
, 1, 0,
1774 printk("update block group failed for %Lu %Lu\n",
1775 ins
->objectid
, ins
->offset
);
1782 * helper function to allocate a block for a given tree
1783 * returns the tree buffer or NULL.
1785 struct extent_buffer
*btrfs_alloc_free_block(struct btrfs_trans_handle
*trans
,
1786 struct btrfs_root
*root
,
1788 u64 root_objectid
, u64 hint
,
1794 ref_generation
= trans
->transid
;
1799 return __btrfs_alloc_free_block(trans
, root
, blocksize
, root_objectid
,
1800 ref_generation
, 0, 0, hint
, empty_size
);
1804 * helper function to allocate a block for a given tree
1805 * returns the tree buffer or NULL.
1807 struct extent_buffer
*__btrfs_alloc_free_block(struct btrfs_trans_handle
*trans
,
1808 struct btrfs_root
*root
,
1817 struct btrfs_key ins
;
1819 struct extent_buffer
*buf
;
1821 ret
= btrfs_alloc_extent(trans
, root
, blocksize
,
1822 root_objectid
, ref_generation
,
1823 level
, first_objectid
, empty_size
, hint
,
1827 return ERR_PTR(ret
);
1829 buf
= btrfs_find_create_tree_block(root
, ins
.objectid
, blocksize
);
1831 btrfs_free_extent(trans
, root
, ins
.objectid
, blocksize
,
1832 root
->root_key
.objectid
, ref_generation
,
1834 return ERR_PTR(-ENOMEM
);
1836 btrfs_set_header_generation(buf
, trans
->transid
);
1837 clean_tree_block(trans
, root
, buf
);
1838 wait_on_tree_block_writeback(root
, buf
);
1839 btrfs_set_buffer_uptodate(buf
);
1841 if (PageDirty(buf
->first_page
)) {
1842 printk("page %lu dirty\n", buf
->first_page
->index
);
1846 set_extent_dirty(&trans
->transaction
->dirty_pages
, buf
->start
,
1847 buf
->start
+ buf
->len
- 1, GFP_NOFS
);
1848 set_extent_bits(&BTRFS_I(root
->fs_info
->btree_inode
)->io_tree
,
1849 buf
->start
, buf
->start
+ buf
->len
- 1,
1850 EXTENT_CSUM
, GFP_NOFS
);
1851 buf
->flags
|= EXTENT_CSUM
;
1852 if (!btrfs_test_opt(root
, SSD
))
1853 btrfs_set_buffer_defrag(buf
);
1854 trans
->blocks_used
++;
1858 static int noinline
drop_leaf_ref(struct btrfs_trans_handle
*trans
,
1859 struct btrfs_root
*root
,
1860 struct extent_buffer
*leaf
)
1863 u64 leaf_generation
;
1864 struct btrfs_key key
;
1865 struct btrfs_file_extent_item
*fi
;
1870 BUG_ON(!btrfs_is_leaf(leaf
));
1871 nritems
= btrfs_header_nritems(leaf
);
1872 leaf_owner
= btrfs_header_owner(leaf
);
1873 leaf_generation
= btrfs_header_generation(leaf
);
1875 for (i
= 0; i
< nritems
; i
++) {
1878 btrfs_item_key_to_cpu(leaf
, &key
, i
);
1879 if (btrfs_key_type(&key
) != BTRFS_EXTENT_DATA_KEY
)
1881 fi
= btrfs_item_ptr(leaf
, i
, struct btrfs_file_extent_item
);
1882 if (btrfs_file_extent_type(leaf
, fi
) ==
1883 BTRFS_FILE_EXTENT_INLINE
)
1886 * FIXME make sure to insert a trans record that
1887 * repeats the snapshot del on crash
1889 disk_bytenr
= btrfs_file_extent_disk_bytenr(leaf
, fi
);
1890 if (disk_bytenr
== 0)
1892 ret
= btrfs_free_extent(trans
, root
, disk_bytenr
,
1893 btrfs_file_extent_disk_num_bytes(leaf
, fi
),
1894 leaf_owner
, leaf_generation
,
1895 key
.objectid
, key
.offset
, 0);
1901 static void noinline
reada_walk_down(struct btrfs_root
*root
,
1902 struct extent_buffer
*node
,
1915 nritems
= btrfs_header_nritems(node
);
1916 level
= btrfs_header_level(node
);
1920 for (i
= slot
; i
< nritems
&& skipped
< 32; i
++) {
1921 bytenr
= btrfs_node_blockptr(node
, i
);
1922 if (last
&& ((bytenr
> last
&& bytenr
- last
> 32 * 1024) ||
1923 (last
> bytenr
&& last
- bytenr
> 32 * 1024))) {
1927 blocksize
= btrfs_level_size(root
, level
- 1);
1929 ret
= lookup_extent_ref(NULL
, root
, bytenr
,
1937 mutex_unlock(&root
->fs_info
->fs_mutex
);
1938 ret
= readahead_tree_block(root
, bytenr
, blocksize
);
1939 last
= bytenr
+ blocksize
;
1941 mutex_lock(&root
->fs_info
->fs_mutex
);
1948 * helper function for drop_snapshot, this walks down the tree dropping ref
1949 * counts as it goes.
1951 static int noinline
walk_down_tree(struct btrfs_trans_handle
*trans
,
1952 struct btrfs_root
*root
,
1953 struct btrfs_path
*path
, int *level
)
1958 struct extent_buffer
*next
;
1959 struct extent_buffer
*cur
;
1960 struct extent_buffer
*parent
;
1965 WARN_ON(*level
< 0);
1966 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
1967 ret
= lookup_extent_ref(trans
, root
,
1968 path
->nodes
[*level
]->start
,
1969 path
->nodes
[*level
]->len
, &refs
);
1975 * walk down to the last node level and free all the leaves
1977 while(*level
>= 0) {
1978 WARN_ON(*level
< 0);
1979 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
1980 cur
= path
->nodes
[*level
];
1982 if (btrfs_header_level(cur
) != *level
)
1985 if (path
->slots
[*level
] >=
1986 btrfs_header_nritems(cur
))
1989 ret
= drop_leaf_ref(trans
, root
, cur
);
1993 bytenr
= btrfs_node_blockptr(cur
, path
->slots
[*level
]);
1994 blocksize
= btrfs_level_size(root
, *level
- 1);
1995 ret
= lookup_extent_ref(trans
, root
, bytenr
, blocksize
, &refs
);
1998 parent
= path
->nodes
[*level
];
1999 root_owner
= btrfs_header_owner(parent
);
2000 root_gen
= btrfs_header_generation(parent
);
2001 path
->slots
[*level
]++;
2002 ret
= btrfs_free_extent(trans
, root
, bytenr
,
2003 blocksize
, root_owner
,
2008 next
= btrfs_find_tree_block(root
, bytenr
, blocksize
);
2009 if (!next
|| !btrfs_buffer_uptodate(next
)) {
2010 free_extent_buffer(next
);
2011 reada_walk_down(root
, cur
, path
->slots
[*level
]);
2012 mutex_unlock(&root
->fs_info
->fs_mutex
);
2013 next
= read_tree_block(root
, bytenr
, blocksize
);
2014 mutex_lock(&root
->fs_info
->fs_mutex
);
2016 /* we dropped the lock, check one more time */
2017 ret
= lookup_extent_ref(trans
, root
, bytenr
,
2021 parent
= path
->nodes
[*level
];
2022 root_owner
= btrfs_header_owner(parent
);
2023 root_gen
= btrfs_header_generation(parent
);
2025 path
->slots
[*level
]++;
2026 free_extent_buffer(next
);
2027 ret
= btrfs_free_extent(trans
, root
, bytenr
,
2035 WARN_ON(*level
<= 0);
2036 if (path
->nodes
[*level
-1])
2037 free_extent_buffer(path
->nodes
[*level
-1]);
2038 path
->nodes
[*level
-1] = next
;
2039 *level
= btrfs_header_level(next
);
2040 path
->slots
[*level
] = 0;
2043 WARN_ON(*level
< 0);
2044 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
2046 if (path
->nodes
[*level
] == root
->node
) {
2047 root_owner
= root
->root_key
.objectid
;
2048 parent
= path
->nodes
[*level
];
2050 parent
= path
->nodes
[*level
+ 1];
2051 root_owner
= btrfs_header_owner(parent
);
2054 root_gen
= btrfs_header_generation(parent
);
2055 ret
= btrfs_free_extent(trans
, root
, path
->nodes
[*level
]->start
,
2056 path
->nodes
[*level
]->len
,
2057 root_owner
, root_gen
, 0, 0, 1);
2058 free_extent_buffer(path
->nodes
[*level
]);
2059 path
->nodes
[*level
] = NULL
;
2066 * helper for dropping snapshots. This walks back up the tree in the path
2067 * to find the first node higher up where we haven't yet gone through
2070 static int noinline
walk_up_tree(struct btrfs_trans_handle
*trans
,
2071 struct btrfs_root
*root
,
2072 struct btrfs_path
*path
, int *level
)
2076 struct btrfs_root_item
*root_item
= &root
->root_item
;
2081 for(i
= *level
; i
< BTRFS_MAX_LEVEL
- 1 && path
->nodes
[i
]; i
++) {
2082 slot
= path
->slots
[i
];
2083 if (slot
< btrfs_header_nritems(path
->nodes
[i
]) - 1) {
2084 struct extent_buffer
*node
;
2085 struct btrfs_disk_key disk_key
;
2086 node
= path
->nodes
[i
];
2089 WARN_ON(*level
== 0);
2090 btrfs_node_key(node
, &disk_key
, path
->slots
[i
]);
2091 memcpy(&root_item
->drop_progress
,
2092 &disk_key
, sizeof(disk_key
));
2093 root_item
->drop_level
= i
;
2096 if (path
->nodes
[*level
] == root
->node
) {
2097 root_owner
= root
->root_key
.objectid
;
2099 btrfs_header_generation(path
->nodes
[*level
]);
2101 struct extent_buffer
*node
;
2102 node
= path
->nodes
[*level
+ 1];
2103 root_owner
= btrfs_header_owner(node
);
2104 root_gen
= btrfs_header_generation(node
);
2106 ret
= btrfs_free_extent(trans
, root
,
2107 path
->nodes
[*level
]->start
,
2108 path
->nodes
[*level
]->len
,
2109 root_owner
, root_gen
, 0, 0, 1);
2111 free_extent_buffer(path
->nodes
[*level
]);
2112 path
->nodes
[*level
] = NULL
;
2120 * drop the reference count on the tree rooted at 'snap'. This traverses
2121 * the tree freeing any blocks that have a ref count of zero after being
2124 int btrfs_drop_snapshot(struct btrfs_trans_handle
*trans
, struct btrfs_root
2130 struct btrfs_path
*path
;
2133 struct btrfs_root_item
*root_item
= &root
->root_item
;
2135 path
= btrfs_alloc_path();
2138 level
= btrfs_header_level(root
->node
);
2140 if (btrfs_disk_key_objectid(&root_item
->drop_progress
) == 0) {
2141 path
->nodes
[level
] = root
->node
;
2142 extent_buffer_get(root
->node
);
2143 path
->slots
[level
] = 0;
2145 struct btrfs_key key
;
2146 struct btrfs_disk_key found_key
;
2147 struct extent_buffer
*node
;
2149 btrfs_disk_key_to_cpu(&key
, &root_item
->drop_progress
);
2150 level
= root_item
->drop_level
;
2151 path
->lowest_level
= level
;
2152 wret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2157 node
= path
->nodes
[level
];
2158 btrfs_node_key(node
, &found_key
, path
->slots
[level
]);
2159 WARN_ON(memcmp(&found_key
, &root_item
->drop_progress
,
2160 sizeof(found_key
)));
2163 wret
= walk_down_tree(trans
, root
, path
, &level
);
2169 wret
= walk_up_tree(trans
, root
, path
, &level
);
2177 for (i
= 0; i
<= orig_level
; i
++) {
2178 if (path
->nodes
[i
]) {
2179 free_extent_buffer(path
->nodes
[i
]);
2180 path
->nodes
[i
] = NULL
;
2184 btrfs_free_path(path
);
2188 int btrfs_free_block_groups(struct btrfs_fs_info
*info
)
2195 ret
= find_first_extent_bit(&info
->block_group_cache
, 0,
2196 &start
, &end
, (unsigned int)-1);
2199 ret
= get_state_private(&info
->block_group_cache
, start
, &ptr
);
2201 kfree((void *)(unsigned long)ptr
);
2202 clear_extent_bits(&info
->block_group_cache
, start
,
2203 end
, (unsigned int)-1, GFP_NOFS
);
2206 ret
= find_first_extent_bit(&info
->free_space_cache
, 0,
2207 &start
, &end
, EXTENT_DIRTY
);
2210 clear_extent_dirty(&info
->free_space_cache
, start
,
2216 static int noinline
relocate_inode_pages(struct inode
*inode
, u64 start
,
2222 u64 existing_delalloc
;
2223 unsigned long last_index
;
2226 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
2227 struct file_ra_state
*ra
;
2229 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
2231 mutex_lock(&inode
->i_mutex
);
2232 i
= start
>> PAGE_CACHE_SHIFT
;
2233 last_index
= (start
+ len
- 1) >> PAGE_CACHE_SHIFT
;
2235 file_ra_state_init(ra
, inode
->i_mapping
);
2236 btrfs_force_ra(inode
->i_mapping
, ra
, NULL
, i
, last_index
);
2239 for (; i
<= last_index
; i
++) {
2240 page
= grab_cache_page(inode
->i_mapping
, i
);
2243 if (!PageUptodate(page
)) {
2244 btrfs_readpage(NULL
, page
);
2246 if (!PageUptodate(page
)) {
2248 page_cache_release(page
);
2252 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
2253 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
2255 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
2257 delalloc_start
= page_start
;
2258 existing_delalloc
= count_range_bits(io_tree
,
2259 &delalloc_start
, page_end
,
2260 PAGE_CACHE_SIZE
, EXTENT_DELALLOC
);
2262 set_extent_delalloc(io_tree
, page_start
,
2263 page_end
, GFP_NOFS
);
2265 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
2266 set_page_dirty(page
);
2268 page_cache_release(page
);
2272 mutex_unlock(&inode
->i_mutex
);
2277 * note, this releases the path
2279 static int noinline
relocate_one_reference(struct btrfs_root
*extent_root
,
2280 struct btrfs_path
*path
,
2281 struct btrfs_key
*extent_key
)
2283 struct inode
*inode
;
2284 struct btrfs_root
*found_root
;
2285 struct btrfs_key
*root_location
;
2286 struct btrfs_extent_ref
*ref
;
2293 ref
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
2294 struct btrfs_extent_ref
);
2295 ref_root
= btrfs_ref_root(path
->nodes
[0], ref
);
2296 ref_gen
= btrfs_ref_generation(path
->nodes
[0], ref
);
2297 ref_objectid
= btrfs_ref_objectid(path
->nodes
[0], ref
);
2298 ref_offset
= btrfs_ref_offset(path
->nodes
[0], ref
);
2299 btrfs_release_path(extent_root
, path
);
2301 root_location
= kmalloc(sizeof(*root_location
), GFP_NOFS
);
2302 root_location
->objectid
= ref_root
;
2304 root_location
->offset
= 0;
2306 root_location
->offset
= (u64
)-1;
2307 root_location
->type
= BTRFS_ROOT_ITEM_KEY
;
2309 found_root
= btrfs_read_fs_root_no_name(extent_root
->fs_info
,
2311 BUG_ON(!found_root
);
2312 kfree(root_location
);
2314 if (ref_objectid
>= BTRFS_FIRST_FREE_OBJECTID
) {
2315 mutex_unlock(&extent_root
->fs_info
->fs_mutex
);
2316 inode
= btrfs_iget_locked(extent_root
->fs_info
->sb
,
2317 ref_objectid
, found_root
);
2318 if (inode
->i_state
& I_NEW
) {
2319 /* the inode and parent dir are two different roots */
2320 BTRFS_I(inode
)->root
= found_root
;
2321 BTRFS_I(inode
)->location
.objectid
= ref_objectid
;
2322 BTRFS_I(inode
)->location
.type
= BTRFS_INODE_ITEM_KEY
;
2323 BTRFS_I(inode
)->location
.offset
= 0;
2324 btrfs_read_locked_inode(inode
);
2325 unlock_new_inode(inode
);
2328 /* this can happen if the reference is not against
2329 * the latest version of the tree root
2331 if (is_bad_inode(inode
)) {
2332 mutex_lock(&extent_root
->fs_info
->fs_mutex
);
2335 relocate_inode_pages(inode
, ref_offset
, extent_key
->offset
);
2336 /* FIXME, data=ordered will help get rid of this */
2337 filemap_fdatawrite(inode
->i_mapping
);
2339 mutex_lock(&extent_root
->fs_info
->fs_mutex
);
2341 struct btrfs_trans_handle
*trans
;
2342 struct btrfs_key found_key
;
2343 struct extent_buffer
*eb
;
2347 trans
= btrfs_start_transaction(found_root
, 1);
2348 eb
= read_tree_block(found_root
, extent_key
->objectid
,
2349 extent_key
->offset
);
2350 level
= btrfs_header_level(eb
);
2353 btrfs_item_key_to_cpu(eb
, &found_key
, 0);
2355 btrfs_node_key_to_cpu(eb
, &found_key
, 0);
2357 free_extent_buffer(eb
);
2359 path
->lowest_level
= level
;
2361 ret
= btrfs_search_slot(trans
, found_root
, &found_key
, path
,
2363 path
->lowest_level
= 0;
2364 for (i
= level
; i
< BTRFS_MAX_LEVEL
; i
++) {
2365 if (!path
->nodes
[i
])
2367 free_extent_buffer(path
->nodes
[i
]);
2368 path
->nodes
[i
] = NULL
;
2370 btrfs_release_path(found_root
, path
);
2371 btrfs_end_transaction(trans
, found_root
);
2378 static int noinline
relocate_one_extent(struct btrfs_root
*extent_root
,
2379 struct btrfs_path
*path
,
2380 struct btrfs_key
*extent_key
)
2382 struct btrfs_key key
;
2383 struct btrfs_key found_key
;
2384 struct extent_buffer
*leaf
;
2389 key
.objectid
= extent_key
->objectid
;
2390 key
.type
= BTRFS_EXTENT_REF_KEY
;
2394 ret
= btrfs_search_slot(NULL
, extent_root
, &key
, path
, 0, 0);
2400 leaf
= path
->nodes
[0];
2401 nritems
= btrfs_header_nritems(leaf
);
2402 if (path
->slots
[0] == nritems
)
2405 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2406 if (found_key
.objectid
!= extent_key
->objectid
)
2409 if (found_key
.type
!= BTRFS_EXTENT_REF_KEY
)
2412 key
.offset
= found_key
.offset
+ 1;
2413 item_size
= btrfs_item_size_nr(leaf
, path
->slots
[0]);
2415 ret
= relocate_one_reference(extent_root
, path
, extent_key
);
2421 btrfs_release_path(extent_root
, path
);
2425 int btrfs_shrink_extent_tree(struct btrfs_root
*root
, u64 new_size
)
2427 struct btrfs_trans_handle
*trans
;
2428 struct btrfs_root
*tree_root
= root
->fs_info
->tree_root
;
2429 struct btrfs_path
*path
;
2432 struct btrfs_fs_info
*info
= root
->fs_info
;
2433 struct extent_io_tree
*block_group_cache
;
2434 struct btrfs_key key
;
2435 struct btrfs_key found_key
;
2436 struct extent_buffer
*leaf
;
2441 btrfs_set_super_total_bytes(&info
->super_copy
, new_size
);
2442 clear_extent_dirty(&info
->free_space_cache
, new_size
, (u64
)-1,
2444 block_group_cache
= &info
->block_group_cache
;
2445 path
= btrfs_alloc_path();
2446 root
= root
->fs_info
->extent_root
;
2451 key
.objectid
= new_size
;
2454 cur_byte
= key
.objectid
;
2456 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2460 ret
= find_previous_extent(root
, path
);
2464 leaf
= path
->nodes
[0];
2465 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2466 if (found_key
.objectid
+ found_key
.offset
> new_size
) {
2467 cur_byte
= found_key
.objectid
;
2468 key
.objectid
= cur_byte
;
2471 btrfs_release_path(root
, path
);
2474 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2478 leaf
= path
->nodes
[0];
2479 nritems
= btrfs_header_nritems(leaf
);
2481 if (path
->slots
[0] >= nritems
) {
2482 ret
= btrfs_next_leaf(root
, path
);
2489 leaf
= path
->nodes
[0];
2490 nritems
= btrfs_header_nritems(leaf
);
2493 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2495 if (progress
&& need_resched()) {
2496 memcpy(&key
, &found_key
, sizeof(key
));
2497 mutex_unlock(&root
->fs_info
->fs_mutex
);
2499 mutex_lock(&root
->fs_info
->fs_mutex
);
2500 btrfs_release_path(root
, path
);
2501 btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2507 if (btrfs_key_type(&found_key
) != BTRFS_EXTENT_ITEM_KEY
||
2508 found_key
.objectid
+ found_key
.offset
<= cur_byte
) {
2514 cur_byte
= found_key
.objectid
+ found_key
.offset
;
2515 key
.objectid
= cur_byte
;
2516 btrfs_release_path(root
, path
);
2517 ret
= relocate_one_extent(root
, path
, &found_key
);
2520 btrfs_release_path(root
, path
);
2522 if (total_found
> 0) {
2523 trans
= btrfs_start_transaction(tree_root
, 1);
2524 btrfs_commit_transaction(trans
, tree_root
);
2526 mutex_unlock(&root
->fs_info
->fs_mutex
);
2527 btrfs_clean_old_snapshots(tree_root
);
2528 mutex_lock(&root
->fs_info
->fs_mutex
);
2530 trans
= btrfs_start_transaction(tree_root
, 1);
2531 btrfs_commit_transaction(trans
, tree_root
);
2535 trans
= btrfs_start_transaction(root
, 1);
2536 key
.objectid
= new_size
;
2542 ret
= btrfs_search_slot(trans
, root
, &key
, path
, -1, 1);
2546 leaf
= path
->nodes
[0];
2547 nritems
= btrfs_header_nritems(leaf
);
2549 if (path
->slots
[0] >= nritems
) {
2550 ret
= btrfs_next_leaf(root
, path
);
2557 leaf
= path
->nodes
[0];
2558 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2561 * btrfs_next_leaf doesn't cow buffers, we have to
2562 * do the search again
2564 memcpy(&key
, &found_key
, sizeof(key
));
2565 btrfs_release_path(root
, path
);
2569 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2570 if (btrfs_key_type(&found_key
) != BTRFS_BLOCK_GROUP_ITEM_KEY
) {
2571 printk("shrinker found key %Lu %u %Lu\n",
2572 found_key
.objectid
, found_key
.type
,
2577 ret
= get_state_private(&info
->block_group_cache
,
2578 found_key
.objectid
, &ptr
);
2580 kfree((void *)(unsigned long)ptr
);
2582 clear_extent_bits(&info
->block_group_cache
, found_key
.objectid
,
2583 found_key
.objectid
+ found_key
.offset
- 1,
2584 (unsigned int)-1, GFP_NOFS
);
2586 key
.objectid
= found_key
.objectid
+ 1;
2587 btrfs_del_item(trans
, root
, path
);
2588 btrfs_release_path(root
, path
);
2590 if (need_resched()) {
2591 mutex_unlock(&root
->fs_info
->fs_mutex
);
2593 mutex_lock(&root
->fs_info
->fs_mutex
);
2596 clear_extent_dirty(&info
->free_space_cache
, new_size
, (u64
)-1,
2598 btrfs_commit_transaction(trans
, root
);
2600 btrfs_free_path(path
);
2604 int btrfs_grow_extent_tree(struct btrfs_trans_handle
*trans
,
2605 struct btrfs_root
*root
, u64 new_size
)
2607 struct btrfs_path
*path
;
2612 struct btrfs_block_group_cache
*cache
;
2613 struct btrfs_block_group_item
*item
;
2614 struct btrfs_fs_info
*info
= root
->fs_info
;
2615 struct extent_io_tree
*block_group_cache
;
2616 struct btrfs_key key
;
2617 struct extent_buffer
*leaf
;
2621 old_size
= btrfs_super_total_bytes(&info
->super_copy
);
2622 block_group_cache
= &info
->block_group_cache
;
2624 root
= info
->extent_root
;
2626 cache
= btrfs_lookup_block_group(root
->fs_info
, old_size
- 1);
2628 cur_byte
= cache
->key
.objectid
+ cache
->key
.offset
;
2629 if (cur_byte
>= new_size
)
2632 key
.offset
= BTRFS_BLOCK_GROUP_SIZE
;
2633 btrfs_set_key_type(&key
, BTRFS_BLOCK_GROUP_ITEM_KEY
);
2635 path
= btrfs_alloc_path();
2639 while(cur_byte
< new_size
) {
2640 key
.objectid
= cur_byte
;
2641 ret
= btrfs_insert_empty_item(trans
, root
, path
, &key
,
2642 sizeof(struct btrfs_block_group_item
));
2644 leaf
= path
->nodes
[0];
2645 item
= btrfs_item_ptr(leaf
, path
->slots
[0],
2646 struct btrfs_block_group_item
);
2648 btrfs_set_disk_block_group_used(leaf
, item
, 0);
2649 div_long_long_rem(nr
, 3, &rem
);
2651 btrfs_set_disk_block_group_flags(leaf
, item
,
2652 BTRFS_BLOCK_GROUP_DATA
);
2654 btrfs_set_disk_block_group_flags(leaf
, item
, 0);
2658 cache
= kmalloc(sizeof(*cache
), GFP_NOFS
);
2661 read_extent_buffer(leaf
, &cache
->item
, (unsigned long)item
,
2662 sizeof(cache
->item
));
2664 memcpy(&cache
->key
, &key
, sizeof(key
));
2667 cur_byte
= key
.objectid
+ key
.offset
;
2668 btrfs_release_path(root
, path
);
2670 if (cache
->item
.flags
& BTRFS_BLOCK_GROUP_DATA
) {
2671 bit
= BLOCK_GROUP_DATA
;
2672 cache
->data
= BTRFS_BLOCK_GROUP_DATA
;
2674 bit
= BLOCK_GROUP_METADATA
;
2678 /* use EXTENT_LOCKED to prevent merging */
2679 set_extent_bits(block_group_cache
, key
.objectid
,
2680 key
.objectid
+ key
.offset
- 1,
2681 bit
| EXTENT_LOCKED
, GFP_NOFS
);
2682 set_state_private(block_group_cache
, key
.objectid
,
2683 (unsigned long)cache
);
2685 btrfs_free_path(path
);
2687 btrfs_set_super_total_bytes(&info
->super_copy
, new_size
);
2691 int btrfs_read_block_groups(struct btrfs_root
*root
)
2693 struct btrfs_path
*path
;
2697 struct btrfs_block_group_cache
*cache
;
2698 struct btrfs_fs_info
*info
= root
->fs_info
;
2699 struct extent_io_tree
*block_group_cache
;
2700 struct btrfs_key key
;
2701 struct btrfs_key found_key
;
2702 struct extent_buffer
*leaf
;
2704 block_group_cache
= &info
->block_group_cache
;
2706 root
= info
->extent_root
;
2708 key
.offset
= BTRFS_BLOCK_GROUP_SIZE
;
2709 btrfs_set_key_type(&key
, BTRFS_BLOCK_GROUP_ITEM_KEY
);
2711 path
= btrfs_alloc_path();
2716 ret
= btrfs_search_slot(NULL
, info
->extent_root
,
2722 leaf
= path
->nodes
[0];
2723 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2724 cache
= kmalloc(sizeof(*cache
), GFP_NOFS
);
2730 read_extent_buffer(leaf
, &cache
->item
,
2731 btrfs_item_ptr_offset(leaf
, path
->slots
[0]),
2732 sizeof(cache
->item
));
2733 memcpy(&cache
->key
, &found_key
, sizeof(found_key
));
2736 key
.objectid
= found_key
.objectid
+ found_key
.offset
;
2737 btrfs_release_path(root
, path
);
2739 if (cache
->item
.flags
& BTRFS_BLOCK_GROUP_MIXED
) {
2740 bit
= BLOCK_GROUP_DATA
| BLOCK_GROUP_METADATA
;
2741 cache
->data
= BTRFS_BLOCK_GROUP_MIXED
;
2742 } else if (cache
->item
.flags
& BTRFS_BLOCK_GROUP_DATA
) {
2743 bit
= BLOCK_GROUP_DATA
;
2744 cache
->data
= BTRFS_BLOCK_GROUP_DATA
;
2746 bit
= BLOCK_GROUP_METADATA
;
2750 /* use EXTENT_LOCKED to prevent merging */
2751 set_extent_bits(block_group_cache
, found_key
.objectid
,
2752 found_key
.objectid
+ found_key
.offset
- 1,
2753 bit
| EXTENT_LOCKED
, GFP_NOFS
);
2754 set_state_private(block_group_cache
, found_key
.objectid
,
2755 (unsigned long)cache
);
2758 btrfs_super_total_bytes(&info
->super_copy
))
2762 btrfs_free_path(path
);