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>
22 #include "print-tree.h"
23 #include "transaction.h"
25 static int finish_current_insert(struct btrfs_trans_handle
*trans
, struct
26 btrfs_root
*extent_root
);
27 static int del_pending_extents(struct btrfs_trans_handle
*trans
, struct
28 btrfs_root
*extent_root
);
30 static int cache_block_group(struct btrfs_root
*root
,
31 struct btrfs_block_group_cache
*block_group
)
33 struct btrfs_path
*path
;
36 struct btrfs_leaf
*leaf
;
37 struct radix_tree_root
*extent_radix
;
45 root
= root
->fs_info
->extent_root
;
46 extent_radix
= &root
->fs_info
->extent_map_radix
;
48 if (block_group
->cached
)
50 if (block_group
->data
)
52 path
= btrfs_alloc_path();
56 key
.objectid
= block_group
->key
.objectid
;
59 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
60 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
63 if (ret
&& path
->slots
[0] > 0)
65 limit
= block_group
->key
.objectid
+ block_group
->key
.offset
;
67 leaf
= btrfs_buffer_leaf(path
->nodes
[0]);
68 slot
= path
->slots
[0];
69 if (slot
>= btrfs_header_nritems(&leaf
->header
)) {
70 ret
= btrfs_next_leaf(root
, path
);
77 hole_size
= block_group
->key
.objectid
+
78 block_group
->key
.offset
- last
;
80 last
= block_group
->key
.objectid
;
81 hole_size
= block_group
->key
.offset
;
83 for (i
= 0; i
< hole_size
; i
++) {
84 set_radix_bit(extent_radix
,
90 btrfs_disk_key_to_cpu(&key
, &leaf
->items
[slot
].key
);
91 if (key
.objectid
>= block_group
->key
.objectid
+
92 block_group
->key
.offset
) {
94 hole_size
= block_group
->key
.objectid
+
95 block_group
->key
.offset
- last
;
97 last
= block_group
->key
.objectid
;
98 hole_size
= block_group
->key
.offset
;
100 for (i
= 0; i
< hole_size
; i
++) {
101 set_radix_bit(extent_radix
, last
+ i
);
105 if (btrfs_key_type(&key
) == BTRFS_EXTENT_ITEM_KEY
) {
107 last
= key
.objectid
+ key
.offset
;
110 hole_size
= key
.objectid
- last
;
111 for (i
= 0; i
< hole_size
; i
++) {
112 set_radix_bit(extent_radix
, last
+ i
);
114 last
= key
.objectid
+ key
.offset
;
120 block_group
->cached
= 1;
122 btrfs_free_path(path
);
126 struct btrfs_block_group_cache
*btrfs_lookup_block_group(struct
130 struct btrfs_block_group_cache
*block_group
;
133 ret
= radix_tree_gang_lookup(&info
->block_group_radix
,
134 (void **)&block_group
,
137 if (block_group
->key
.objectid
<= blocknr
&& blocknr
<=
138 block_group
->key
.objectid
+ block_group
->key
.offset
)
141 ret
= radix_tree_gang_lookup(&info
->block_group_data_radix
,
142 (void **)&block_group
,
145 if (block_group
->key
.objectid
<= blocknr
&& blocknr
<=
146 block_group
->key
.objectid
+ block_group
->key
.offset
)
152 static u64
leaf_range(struct btrfs_root
*root
)
154 u64 size
= BTRFS_LEAF_DATA_SIZE(root
);
155 do_div(size
, sizeof(struct btrfs_extent_item
) +
156 sizeof(struct btrfs_item
));
160 static u64
find_search_start(struct btrfs_root
*root
,
161 struct btrfs_block_group_cache
**cache_ret
,
162 u64 search_start
, int num
)
164 unsigned long gang
[8];
166 struct btrfs_block_group_cache
*cache
= *cache_ret
;
167 u64 last
= max(search_start
, cache
->key
.objectid
);
172 ret
= cache_block_group(root
, cache
);
176 ret
= find_first_radix_bit(&root
->fs_info
->extent_map_radix
,
177 gang
, last
, ARRAY_SIZE(gang
));
180 last
= gang
[ret
-1] + 1;
182 if (ret
!= ARRAY_SIZE(gang
)) {
185 if (gang
[ret
-1] - gang
[0] > leaf_range(root
)) {
189 if (gang
[0] >= cache
->key
.objectid
+ cache
->key
.offset
) {
195 return max(cache
->last_alloc
, search_start
);
198 cache
= btrfs_lookup_block_group(root
->fs_info
,
199 last
+ cache
->key
.offset
- 1);
201 return max((*cache_ret
)->last_alloc
, search_start
);
203 cache
= btrfs_find_block_group(root
, cache
,
204 last
+ cache
->key
.offset
- 1, 0, 0);
209 static u64
div_factor(u64 num
, int factor
)
216 struct btrfs_block_group_cache
*btrfs_find_block_group(struct btrfs_root
*root
,
217 struct btrfs_block_group_cache
218 *hint
, u64 search_start
,
221 struct btrfs_block_group_cache
*cache
[8];
222 struct btrfs_block_group_cache
*found_group
= NULL
;
223 struct btrfs_fs_info
*info
= root
->fs_info
;
224 struct radix_tree_root
*radix
;
225 struct radix_tree_root
*swap_radix
;
239 radix
= &info
->block_group_data_radix
;
240 swap_radix
= &info
->block_group_radix
;
242 radix
= &info
->block_group_radix
;
243 swap_radix
= &info
->block_group_data_radix
;
247 struct btrfs_block_group_cache
*shint
;
248 shint
= btrfs_lookup_block_group(info
, search_start
);
249 if (shint
->data
== data
) {
250 used
= btrfs_block_group_used(&shint
->item
);
251 if (used
+ shint
->pinned
<
252 div_factor(shint
->key
.offset
, factor
)) {
257 if (hint
&& hint
->data
== data
) {
258 used
= btrfs_block_group_used(&hint
->item
);
259 if (used
+ hint
->pinned
<
260 div_factor(hint
->key
.offset
, factor
)) {
263 if (used
>= div_factor(hint
->key
.offset
, 8)) {
264 radix_tree_tag_clear(radix
,
266 hint
->key
.offset
- 1,
267 BTRFS_BLOCK_GROUP_AVAIL
);
269 last
= hint
->key
.offset
* 3;
270 if (hint
->key
.objectid
>= last
)
271 last
= max(search_start
+ hint
->key
.offset
- 1,
272 hint
->key
.objectid
- last
);
274 last
= hint
->key
.objectid
+ hint
->key
.offset
;
278 hint_last
= max(hint
->key
.objectid
, search_start
);
280 hint_last
= search_start
;
285 ret
= radix_tree_gang_lookup_tag(radix
, (void **)cache
,
286 last
, ARRAY_SIZE(cache
),
287 BTRFS_BLOCK_GROUP_AVAIL
);
290 for (i
= 0; i
< ret
; i
++) {
291 last
= cache
[i
]->key
.objectid
+
292 cache
[i
]->key
.offset
;
293 used
= btrfs_block_group_used(&cache
[i
]->item
);
294 if (used
+ cache
[i
]->pinned
<
295 div_factor(cache
[i
]->key
.offset
, factor
)) {
296 found_group
= cache
[i
];
299 if (used
>= div_factor(cache
[i
]->key
.offset
, 8)) {
300 radix_tree_tag_clear(radix
,
301 cache
[i
]->key
.objectid
+
302 cache
[i
]->key
.offset
- 1,
303 BTRFS_BLOCK_GROUP_AVAIL
);
311 ret
= radix_tree_gang_lookup(radix
, (void **)cache
,
312 last
, ARRAY_SIZE(cache
));
315 for (i
= 0; i
< ret
; i
++) {
316 last
= cache
[i
]->key
.objectid
+
317 cache
[i
]->key
.offset
;
318 used
= btrfs_block_group_used(&cache
[i
]->item
);
319 if (used
+ cache
[i
]->pinned
< cache
[i
]->key
.offset
) {
320 found_group
= cache
[i
];
323 if (used
>= cache
[i
]->key
.offset
) {
324 radix_tree_tag_clear(radix
,
325 cache
[i
]->key
.objectid
+
326 cache
[i
]->key
.offset
- 1,
327 BTRFS_BLOCK_GROUP_AVAIL
);
338 struct radix_tree_root
*tmp
= radix
;
346 ret
= radix_tree_gang_lookup(radix
,
347 (void **)&found_group
, 0, 1);
349 ret
= radix_tree_gang_lookup(swap_radix
,
350 (void **)&found_group
,
359 int btrfs_inc_extent_ref(struct btrfs_trans_handle
*trans
,
360 struct btrfs_root
*root
,
361 u64 blocknr
, u64 num_blocks
)
363 struct btrfs_path
*path
;
365 struct btrfs_key key
;
366 struct btrfs_leaf
*l
;
367 struct btrfs_extent_item
*item
;
370 path
= btrfs_alloc_path();
374 key
.objectid
= blocknr
;
376 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
377 key
.offset
= num_blocks
;
378 ret
= btrfs_search_slot(trans
, root
->fs_info
->extent_root
, &key
, path
,
386 l
= btrfs_buffer_leaf(path
->nodes
[0]);
387 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
388 refs
= btrfs_extent_refs(item
);
389 btrfs_set_extent_refs(item
, refs
+ 1);
390 btrfs_mark_buffer_dirty(path
->nodes
[0]);
392 btrfs_release_path(root
->fs_info
->extent_root
, path
);
393 btrfs_free_path(path
);
394 finish_current_insert(trans
, root
->fs_info
->extent_root
);
395 del_pending_extents(trans
, root
->fs_info
->extent_root
);
399 static int lookup_extent_ref(struct btrfs_trans_handle
*trans
,
400 struct btrfs_root
*root
, u64 blocknr
,
401 u64 num_blocks
, u32
*refs
)
403 struct btrfs_path
*path
;
405 struct btrfs_key key
;
406 struct btrfs_leaf
*l
;
407 struct btrfs_extent_item
*item
;
409 path
= btrfs_alloc_path();
410 key
.objectid
= blocknr
;
411 key
.offset
= num_blocks
;
413 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
414 ret
= btrfs_search_slot(trans
, root
->fs_info
->extent_root
, &key
, path
,
420 l
= btrfs_buffer_leaf(path
->nodes
[0]);
421 item
= btrfs_item_ptr(l
, path
->slots
[0], struct btrfs_extent_item
);
422 *refs
= btrfs_extent_refs(item
);
424 btrfs_free_path(path
);
428 int btrfs_inc_root_ref(struct btrfs_trans_handle
*trans
,
429 struct btrfs_root
*root
)
431 return btrfs_inc_extent_ref(trans
, root
, bh_blocknr(root
->node
), 1);
434 int btrfs_inc_ref(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
435 struct buffer_head
*buf
)
438 struct btrfs_node
*buf_node
;
439 struct btrfs_leaf
*buf_leaf
;
440 struct btrfs_disk_key
*key
;
441 struct btrfs_file_extent_item
*fi
;
450 buf_node
= btrfs_buffer_node(buf
);
451 leaf
= btrfs_is_leaf(buf_node
);
452 buf_leaf
= btrfs_buffer_leaf(buf
);
453 for (i
= 0; i
< btrfs_header_nritems(&buf_node
->header
); i
++) {
456 key
= &buf_leaf
->items
[i
].key
;
457 if (btrfs_disk_key_type(key
) != BTRFS_EXTENT_DATA_KEY
)
459 fi
= btrfs_item_ptr(buf_leaf
, i
,
460 struct btrfs_file_extent_item
);
461 if (btrfs_file_extent_type(fi
) ==
462 BTRFS_FILE_EXTENT_INLINE
)
464 disk_blocknr
= btrfs_file_extent_disk_blocknr(fi
);
465 if (disk_blocknr
== 0)
467 ret
= btrfs_inc_extent_ref(trans
, root
, disk_blocknr
,
468 btrfs_file_extent_disk_num_blocks(fi
));
474 blocknr
= btrfs_node_blockptr(buf_node
, i
);
475 ret
= btrfs_inc_extent_ref(trans
, root
, blocknr
, 1);
485 for (i
=0; i
< faili
; i
++) {
488 key
= &buf_leaf
->items
[i
].key
;
489 if (btrfs_disk_key_type(key
) != BTRFS_EXTENT_DATA_KEY
)
491 fi
= btrfs_item_ptr(buf_leaf
, i
,
492 struct btrfs_file_extent_item
);
493 if (btrfs_file_extent_type(fi
) ==
494 BTRFS_FILE_EXTENT_INLINE
)
496 disk_blocknr
= btrfs_file_extent_disk_blocknr(fi
);
497 if (disk_blocknr
== 0)
499 err
= btrfs_free_extent(trans
, root
, disk_blocknr
,
500 btrfs_file_extent_disk_num_blocks(fi
), 0);
503 blocknr
= btrfs_node_blockptr(buf_node
, i
);
504 err
= btrfs_free_extent(trans
, root
, blocknr
, 1, 0);
511 static int write_one_cache_group(struct btrfs_trans_handle
*trans
,
512 struct btrfs_root
*root
,
513 struct btrfs_path
*path
,
514 struct btrfs_block_group_cache
*cache
)
518 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
519 struct btrfs_block_group_item
*bi
;
521 ret
= btrfs_search_slot(trans
, extent_root
, &cache
->key
, path
, 0, 1);
525 bi
= btrfs_item_ptr(btrfs_buffer_leaf(path
->nodes
[0]), path
->slots
[0],
526 struct btrfs_block_group_item
);
527 memcpy(bi
, &cache
->item
, sizeof(*bi
));
528 btrfs_mark_buffer_dirty(path
->nodes
[0]);
529 btrfs_release_path(extent_root
, path
);
531 finish_current_insert(trans
, extent_root
);
532 pending_ret
= del_pending_extents(trans
, extent_root
);
538 cache
->last_alloc
= cache
->first_free
;
543 static int write_dirty_block_radix(struct btrfs_trans_handle
*trans
,
544 struct btrfs_root
*root
,
545 struct radix_tree_root
*radix
)
547 struct btrfs_block_group_cache
*cache
[8];
552 struct btrfs_path
*path
;
553 unsigned long off
= 0;
555 path
= btrfs_alloc_path();
560 ret
= radix_tree_gang_lookup_tag(radix
, (void **)cache
,
561 off
, ARRAY_SIZE(cache
),
562 BTRFS_BLOCK_GROUP_DIRTY
);
565 for (i
= 0; i
< ret
; i
++) {
566 err
= write_one_cache_group(trans
, root
,
569 * if we fail to write the cache group, we want
570 * to keep it marked dirty in hopes that a later
575 off
= cache
[i
]->key
.objectid
+
576 cache
[i
]->key
.offset
;
580 radix_tree_tag_clear(radix
, cache
[i
]->key
.objectid
+
581 cache
[i
]->key
.offset
- 1,
582 BTRFS_BLOCK_GROUP_DIRTY
);
585 btrfs_free_path(path
);
589 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle
*trans
,
590 struct btrfs_root
*root
)
594 ret
= write_dirty_block_radix(trans
, root
,
595 &root
->fs_info
->block_group_radix
);
596 ret2
= write_dirty_block_radix(trans
, root
,
597 &root
->fs_info
->block_group_data_radix
);
605 static int update_block_group(struct btrfs_trans_handle
*trans
,
606 struct btrfs_root
*root
,
607 u64 blocknr
, u64 num
, int alloc
, int mark_free
,
610 struct btrfs_block_group_cache
*cache
;
611 struct btrfs_fs_info
*info
= root
->fs_info
;
619 cache
= btrfs_lookup_block_group(info
, blocknr
);
623 block_in_group
= blocknr
- cache
->key
.objectid
;
624 WARN_ON(block_in_group
> cache
->key
.offset
);
625 radix_tree_tag_set(cache
->radix
, cache
->key
.objectid
+
626 cache
->key
.offset
- 1,
627 BTRFS_BLOCK_GROUP_DIRTY
);
629 old_val
= btrfs_block_group_used(&cache
->item
);
630 num
= min(total
, cache
->key
.offset
- block_in_group
);
632 if (blocknr
> cache
->last_alloc
)
633 cache
->last_alloc
= blocknr
;
635 for (i
= 0; i
< num
; i
++) {
636 clear_radix_bit(&info
->extent_map_radix
,
640 if (cache
->data
!= data
&&
641 old_val
< (cache
->key
.offset
>> 1)) {
643 radix_tree_delete(cache
->radix
,
644 cache
->key
.objectid
+
645 cache
->key
.offset
- 1);
649 &info
->block_group_data_radix
;
651 BTRFS_BLOCK_GROUP_DATA
;
653 cache
->radix
= &info
->block_group_radix
;
655 ~BTRFS_BLOCK_GROUP_DATA
;
657 ret
= radix_tree_insert(cache
->radix
,
658 cache
->key
.objectid
+
659 cache
->key
.offset
- 1,
665 if (blocknr
< cache
->first_free
)
666 cache
->first_free
= blocknr
;
667 if (!cache
->data
&& mark_free
) {
668 for (i
= 0; i
< num
; i
++) {
669 set_radix_bit(&info
->extent_map_radix
,
673 if (old_val
< (cache
->key
.offset
>> 1) &&
674 old_val
+ num
>= (cache
->key
.offset
>> 1)) {
675 radix_tree_tag_set(cache
->radix
,
676 cache
->key
.objectid
+
677 cache
->key
.offset
- 1,
678 BTRFS_BLOCK_GROUP_AVAIL
);
681 btrfs_set_block_group_used(&cache
->item
, old_val
);
688 int btrfs_copy_pinned(struct btrfs_root
*root
, struct radix_tree_root
*copy
)
690 unsigned long gang
[8];
692 struct radix_tree_root
*pinned_radix
= &root
->fs_info
->pinned_radix
;
697 ret
= find_first_radix_bit(pinned_radix
, gang
, last
,
701 for (i
= 0 ; i
< ret
; i
++) {
702 set_radix_bit(copy
, gang
[i
]);
706 ret
= find_first_radix_bit(&root
->fs_info
->extent_ins_radix
, gang
, 0,
712 int btrfs_finish_extent_commit(struct btrfs_trans_handle
*trans
,
713 struct btrfs_root
*root
,
714 struct radix_tree_root
*unpin_radix
)
716 unsigned long gang
[8];
717 struct btrfs_block_group_cache
*block_group
;
721 struct radix_tree_root
*pinned_radix
= &root
->fs_info
->pinned_radix
;
722 struct radix_tree_root
*extent_radix
= &root
->fs_info
->extent_map_radix
;
725 ret
= find_first_radix_bit(unpin_radix
, gang
, 0,
731 for (i
= 0; i
< ret
; i
++) {
732 clear_radix_bit(pinned_radix
, gang
[i
]);
733 clear_radix_bit(unpin_radix
, gang
[i
]);
734 block_group
= btrfs_lookup_block_group(root
->fs_info
,
737 WARN_ON(block_group
->pinned
== 0);
738 block_group
->pinned
--;
739 if (gang
[i
] < block_group
->last_alloc
)
740 block_group
->last_alloc
= gang
[i
];
741 if (!block_group
->data
)
742 set_radix_bit(extent_radix
, gang
[i
]);
749 static int finish_current_insert(struct btrfs_trans_handle
*trans
, struct
750 btrfs_root
*extent_root
)
752 struct btrfs_key ins
;
753 struct btrfs_extent_item extent_item
;
757 unsigned long gang
[8];
758 struct btrfs_fs_info
*info
= extent_root
->fs_info
;
760 btrfs_set_extent_refs(&extent_item
, 1);
763 btrfs_set_key_type(&ins
, BTRFS_EXTENT_ITEM_KEY
);
764 btrfs_set_extent_owner(&extent_item
, extent_root
->root_key
.objectid
);
767 ret
= find_first_radix_bit(&info
->extent_ins_radix
, gang
, 0,
772 for (i
= 0; i
< ret
; i
++) {
773 ins
.objectid
= gang
[i
];
774 err
= btrfs_insert_item(trans
, extent_root
, &ins
,
776 sizeof(extent_item
));
777 clear_radix_bit(&info
->extent_ins_radix
, gang
[i
]);
784 static int pin_down_block(struct btrfs_root
*root
, u64 blocknr
, int pending
)
787 struct btrfs_header
*header
;
788 struct buffer_head
*bh
;
791 bh
= btrfs_find_tree_block(root
, blocknr
);
793 if (buffer_uptodate(bh
)) {
795 root
->fs_info
->running_transaction
->transid
;
796 header
= btrfs_buffer_header(bh
);
797 if (btrfs_header_generation(header
) ==
799 btrfs_block_release(root
, bh
);
803 btrfs_block_release(root
, bh
);
805 err
= set_radix_bit(&root
->fs_info
->pinned_radix
, blocknr
);
807 struct btrfs_block_group_cache
*cache
;
808 cache
= btrfs_lookup_block_group(root
->fs_info
,
814 err
= set_radix_bit(&root
->fs_info
->pending_del_radix
, blocknr
);
821 * remove an extent from the root, returns 0 on success
823 static int __free_extent(struct btrfs_trans_handle
*trans
, struct btrfs_root
824 *root
, u64 blocknr
, u64 num_blocks
, int pin
,
827 struct btrfs_path
*path
;
828 struct btrfs_key key
;
829 struct btrfs_fs_info
*info
= root
->fs_info
;
830 struct btrfs_root
*extent_root
= info
->extent_root
;
832 struct btrfs_extent_item
*ei
;
835 key
.objectid
= blocknr
;
837 btrfs_set_key_type(&key
, BTRFS_EXTENT_ITEM_KEY
);
838 key
.offset
= num_blocks
;
840 path
= btrfs_alloc_path();
844 ret
= btrfs_search_slot(trans
, extent_root
, &key
, path
, -1, 1);
848 ei
= btrfs_item_ptr(btrfs_buffer_leaf(path
->nodes
[0]), path
->slots
[0],
849 struct btrfs_extent_item
);
850 BUG_ON(ei
->refs
== 0);
851 refs
= btrfs_extent_refs(ei
) - 1;
852 btrfs_set_extent_refs(ei
, refs
);
853 btrfs_mark_buffer_dirty(path
->nodes
[0]);
855 u64 super_blocks_used
;
858 ret
= pin_down_block(root
, blocknr
, 0);
862 super_blocks_used
= btrfs_super_blocks_used(&info
->super_copy
);
863 btrfs_set_super_blocks_used(&info
->super_copy
,
864 super_blocks_used
- num_blocks
);
865 ret
= btrfs_del_item(trans
, extent_root
, path
);
869 ret
= update_block_group(trans
, root
, blocknr
, num_blocks
, 0,
873 btrfs_free_path(path
);
874 finish_current_insert(trans
, extent_root
);
879 * find all the blocks marked as pending in the radix tree and remove
880 * them from the extent map
882 static int del_pending_extents(struct btrfs_trans_handle
*trans
, struct
883 btrfs_root
*extent_root
)
888 unsigned long gang
[4];
890 struct radix_tree_root
*pending_radix
;
891 struct radix_tree_root
*pinned_radix
;
892 struct btrfs_block_group_cache
*cache
;
894 pending_radix
= &extent_root
->fs_info
->pending_del_radix
;
895 pinned_radix
= &extent_root
->fs_info
->pinned_radix
;
898 ret
= find_first_radix_bit(pending_radix
, gang
, 0,
902 for (i
= 0; i
< ret
; i
++) {
903 wret
= set_radix_bit(pinned_radix
, gang
[i
]);
906 btrfs_lookup_block_group(extent_root
->fs_info
,
912 printk(KERN_CRIT
"set_radix_bit, err %d\n",
916 wret
= clear_radix_bit(pending_radix
, gang
[i
]);
918 wret
= __free_extent(trans
, extent_root
,
928 * remove an extent from the root, returns 0 on success
930 int btrfs_free_extent(struct btrfs_trans_handle
*trans
, struct btrfs_root
931 *root
, u64 blocknr
, u64 num_blocks
, int pin
)
933 struct btrfs_root
*extent_root
= root
->fs_info
->extent_root
;
937 if (root
== extent_root
) {
938 pin_down_block(root
, blocknr
, 1);
941 ret
= __free_extent(trans
, root
, blocknr
, num_blocks
, pin
, pin
== 0);
942 pending_ret
= del_pending_extents(trans
, root
->fs_info
->extent_root
);
943 return ret
? ret
: pending_ret
;
947 * walks the btree of allocated extents and find a hole of a given size.
948 * The key ins is changed to record the hole:
949 * ins->objectid == block start
950 * ins->flags = BTRFS_EXTENT_ITEM_KEY
951 * ins->offset == number of blocks
952 * Any available blocks before search_start are skipped.
954 static int find_free_extent(struct btrfs_trans_handle
*trans
, struct btrfs_root
955 *orig_root
, u64 num_blocks
, u64 empty_size
,
956 u64 search_start
, u64 search_end
, u64 hint_block
,
957 struct btrfs_key
*ins
, u64 exclude_start
,
958 u64 exclude_nr
, int data
)
960 struct btrfs_path
*path
;
961 struct btrfs_key key
;
967 u64 orig_search_start
= search_start
;
969 struct btrfs_leaf
*l
;
970 struct btrfs_root
* root
= orig_root
->fs_info
->extent_root
;
971 struct btrfs_fs_info
*info
= root
->fs_info
;
972 int total_needed
= num_blocks
;
974 struct btrfs_block_group_cache
*block_group
;
979 WARN_ON(num_blocks
< 1);
981 btrfs_set_key_type(ins
, BTRFS_EXTENT_ITEM_KEY
);
983 level
= btrfs_header_level(btrfs_buffer_header(root
->node
));
984 if (search_end
== (u64
)-1)
985 search_end
= btrfs_super_total_blocks(&info
->super_copy
);
987 block_group
= btrfs_lookup_block_group(info
, hint_block
);
988 block_group
= btrfs_find_block_group(root
, block_group
,
989 hint_block
, data
, 1);
991 block_group
= btrfs_find_block_group(root
,
992 trans
->block_group
, 0,
996 total_needed
+= empty_size
;
997 path
= btrfs_alloc_path();
1000 if (!block_group
->data
)
1001 search_start
= find_search_start(root
, &block_group
,
1002 search_start
, total_needed
);
1003 else if (!full_scan
)
1004 search_start
= max(block_group
->last_alloc
, search_start
);
1006 btrfs_init_path(path
);
1007 ins
->objectid
= search_start
;
1011 ret
= btrfs_search_slot(trans
, root
, ins
, path
, 0, 0);
1015 if (path
->slots
[0] > 0) {
1019 l
= btrfs_buffer_leaf(path
->nodes
[0]);
1020 btrfs_disk_key_to_cpu(&key
, &l
->items
[path
->slots
[0]].key
);
1022 * a rare case, go back one key if we hit a block group item
1023 * instead of an extent item
1025 if (btrfs_key_type(&key
) != BTRFS_EXTENT_ITEM_KEY
&&
1026 key
.objectid
+ key
.offset
>= search_start
) {
1027 ins
->objectid
= key
.objectid
;
1028 ins
->offset
= key
.offset
- 1;
1029 btrfs_release_path(root
, path
);
1030 ret
= btrfs_search_slot(trans
, root
, ins
, path
, 0, 0);
1034 if (path
->slots
[0] > 0) {
1040 l
= btrfs_buffer_leaf(path
->nodes
[0]);
1041 slot
= path
->slots
[0];
1042 if (slot
>= btrfs_header_nritems(&l
->header
)) {
1044 limit
= last_block
+
1045 (block_group
->key
.offset
>> 1);
1047 limit
= search_start
+
1048 (block_group
->key
.offset
>> 1);
1049 ret
= btrfs_next_leaf(root
, path
);
1055 ins
->objectid
= search_start
;
1056 ins
->offset
= search_end
- search_start
;
1060 ins
->objectid
= last_block
> search_start
?
1061 last_block
: search_start
;
1062 ins
->offset
= search_end
- ins
->objectid
;
1066 btrfs_disk_key_to_cpu(&key
, &l
->items
[slot
].key
);
1067 if (key
.objectid
>= search_start
&& key
.objectid
> last_block
&&
1069 if (last_block
< search_start
)
1070 last_block
= search_start
;
1071 hole_size
= key
.objectid
- last_block
;
1072 if (hole_size
>= num_blocks
) {
1073 ins
->objectid
= last_block
;
1074 ins
->offset
= hole_size
;
1079 if (btrfs_key_type(&key
) != BTRFS_EXTENT_ITEM_KEY
)
1083 last_block
= key
.objectid
+ key
.offset
;
1084 if (!full_scan
&& last_block
>= block_group
->key
.objectid
+
1085 block_group
->key
.offset
) {
1086 btrfs_release_path(root
, path
);
1087 search_start
= block_group
->key
.objectid
+
1088 block_group
->key
.offset
* 2;
1096 /* we have to make sure we didn't find an extent that has already
1097 * been allocated by the map tree or the original allocation
1099 btrfs_release_path(root
, path
);
1100 BUG_ON(ins
->objectid
< search_start
);
1102 if (ins
->objectid
+ num_blocks
>= search_end
) {
1107 search_start
= orig_search_start
;
1110 total_needed
-= empty_size
;
1116 for (test_block
= ins
->objectid
;
1117 test_block
< ins
->objectid
+ num_blocks
; test_block
++) {
1118 if (test_radix_bit(&info
->pinned_radix
, test_block
) ||
1119 test_radix_bit(&info
->extent_ins_radix
, test_block
)) {
1120 search_start
= test_block
+ 1;
1124 if (exclude_nr
> 0 && (ins
->objectid
+ num_blocks
> exclude_start
&&
1125 ins
->objectid
< exclude_start
+ exclude_nr
)) {
1126 search_start
= exclude_start
+ exclude_nr
;
1130 block_group
= btrfs_lookup_block_group(info
, ins
->objectid
);
1132 trans
->block_group
= block_group
;
1134 ins
->offset
= num_blocks
;
1135 btrfs_free_path(path
);
1139 if (search_start
+ num_blocks
>= search_end
) {
1140 search_start
= orig_search_start
;
1147 total_needed
-= empty_size
;
1152 block_group
= btrfs_lookup_block_group(info
, search_start
);
1155 block_group
= btrfs_find_block_group(root
, block_group
,
1156 search_start
, data
, 0);
1160 btrfs_release_path(root
, path
);
1161 btrfs_free_path(path
);
1165 * finds a free extent and does all the dirty work required for allocation
1166 * returns the key for the extent through ins, and a tree buffer for
1167 * the first block of the extent through buf.
1169 * returns 0 if everything worked, non-zero otherwise.
1171 int btrfs_alloc_extent(struct btrfs_trans_handle
*trans
,
1172 struct btrfs_root
*root
, u64 owner
,
1173 u64 num_blocks
, u64 empty_size
, u64 hint_block
,
1174 u64 search_end
, struct btrfs_key
*ins
, int data
)
1178 u64 super_blocks_used
;
1179 u64 search_start
= 0;
1180 struct btrfs_fs_info
*info
= root
->fs_info
;
1181 struct btrfs_root
*extent_root
= info
->extent_root
;
1182 struct btrfs_extent_item extent_item
;
1184 btrfs_set_extent_refs(&extent_item
, 1);
1185 btrfs_set_extent_owner(&extent_item
, owner
);
1187 WARN_ON(num_blocks
< 1);
1188 ret
= find_free_extent(trans
, root
, num_blocks
, empty_size
,
1189 search_start
, search_end
, hint_block
, ins
,
1190 trans
->alloc_exclude_start
,
1191 trans
->alloc_exclude_nr
, data
);
1196 super_blocks_used
= btrfs_super_blocks_used(&info
->super_copy
);
1197 btrfs_set_super_blocks_used(&info
->super_copy
, super_blocks_used
+
1200 if (root
== extent_root
) {
1201 BUG_ON(num_blocks
!= 1);
1202 set_radix_bit(&root
->fs_info
->extent_ins_radix
, ins
->objectid
);
1206 WARN_ON(trans
->alloc_exclude_nr
);
1207 trans
->alloc_exclude_start
= ins
->objectid
;
1208 trans
->alloc_exclude_nr
= ins
->offset
;
1209 ret
= btrfs_insert_item(trans
, extent_root
, ins
, &extent_item
,
1210 sizeof(extent_item
));
1212 trans
->alloc_exclude_start
= 0;
1213 trans
->alloc_exclude_nr
= 0;
1216 finish_current_insert(trans
, extent_root
);
1217 pending_ret
= del_pending_extents(trans
, extent_root
);
1226 ret
= update_block_group(trans
, root
, ins
->objectid
, ins
->offset
, 1, 0,
1233 * helper function to allocate a block for a given tree
1234 * returns the tree buffer or NULL.
1236 struct buffer_head
*btrfs_alloc_free_block(struct btrfs_trans_handle
*trans
,
1237 struct btrfs_root
*root
, u64 hint
,
1240 struct btrfs_key ins
;
1242 struct buffer_head
*buf
;
1244 ret
= btrfs_alloc_extent(trans
, root
, root
->root_key
.objectid
,
1245 1, empty_size
, hint
,
1246 (unsigned long)-1, &ins
, 0);
1249 return ERR_PTR(ret
);
1251 buf
= btrfs_find_create_tree_block(root
, ins
.objectid
);
1253 btrfs_free_extent(trans
, root
, ins
.objectid
, 1, 0);
1254 return ERR_PTR(-ENOMEM
);
1256 WARN_ON(buffer_dirty(buf
));
1257 set_buffer_uptodate(buf
);
1258 set_buffer_checked(buf
);
1259 set_radix_bit(&trans
->transaction
->dirty_pages
, buf
->b_page
->index
);
1263 static int drop_leaf_ref(struct btrfs_trans_handle
*trans
,
1264 struct btrfs_root
*root
, struct buffer_head
*cur
)
1266 struct btrfs_disk_key
*key
;
1267 struct btrfs_leaf
*leaf
;
1268 struct btrfs_file_extent_item
*fi
;
1273 BUG_ON(!btrfs_is_leaf(btrfs_buffer_node(cur
)));
1274 leaf
= btrfs_buffer_leaf(cur
);
1275 nritems
= btrfs_header_nritems(&leaf
->header
);
1276 for (i
= 0; i
< nritems
; i
++) {
1278 key
= &leaf
->items
[i
].key
;
1279 if (btrfs_disk_key_type(key
) != BTRFS_EXTENT_DATA_KEY
)
1281 fi
= btrfs_item_ptr(leaf
, i
, struct btrfs_file_extent_item
);
1282 if (btrfs_file_extent_type(fi
) == BTRFS_FILE_EXTENT_INLINE
)
1285 * FIXME make sure to insert a trans record that
1286 * repeats the snapshot del on crash
1288 disk_blocknr
= btrfs_file_extent_disk_blocknr(fi
);
1289 if (disk_blocknr
== 0)
1291 ret
= btrfs_free_extent(trans
, root
, disk_blocknr
,
1292 btrfs_file_extent_disk_num_blocks(fi
),
1299 static void reada_walk_down(struct btrfs_root
*root
,
1300 struct btrfs_node
*node
)
1308 nritems
= btrfs_header_nritems(&node
->header
);
1309 for (i
= 0; i
< nritems
; i
++) {
1310 blocknr
= btrfs_node_blockptr(node
, i
);
1311 ret
= lookup_extent_ref(NULL
, root
, blocknr
, 1, &refs
);
1315 ret
= readahead_tree_block(root
, blocknr
);
1322 * helper function for drop_snapshot, this walks down the tree dropping ref
1323 * counts as it goes.
1325 static int walk_down_tree(struct btrfs_trans_handle
*trans
, struct btrfs_root
1326 *root
, struct btrfs_path
*path
, int *level
)
1328 struct buffer_head
*next
;
1329 struct buffer_head
*cur
;
1334 WARN_ON(*level
< 0);
1335 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
1336 ret
= lookup_extent_ref(trans
, root
, bh_blocknr(path
->nodes
[*level
]),
1343 * walk down to the last node level and free all the leaves
1345 while(*level
>= 0) {
1346 WARN_ON(*level
< 0);
1347 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
1348 cur
= path
->nodes
[*level
];
1350 if (*level
> 0 && path
->slots
[*level
] == 0)
1351 reada_walk_down(root
, btrfs_buffer_node(cur
));
1353 if (btrfs_header_level(btrfs_buffer_header(cur
)) != *level
)
1356 if (path
->slots
[*level
] >=
1357 btrfs_header_nritems(btrfs_buffer_header(cur
)))
1360 ret
= drop_leaf_ref(trans
, root
, cur
);
1364 blocknr
= btrfs_node_blockptr(btrfs_buffer_node(cur
),
1365 path
->slots
[*level
]);
1366 ret
= lookup_extent_ref(trans
, root
, blocknr
, 1, &refs
);
1369 path
->slots
[*level
]++;
1370 ret
= btrfs_free_extent(trans
, root
, blocknr
, 1, 1);
1374 next
= read_tree_block(root
, blocknr
);
1375 WARN_ON(*level
<= 0);
1376 if (path
->nodes
[*level
-1])
1377 btrfs_block_release(root
, path
->nodes
[*level
-1]);
1378 path
->nodes
[*level
-1] = next
;
1379 *level
= btrfs_header_level(btrfs_buffer_header(next
));
1380 path
->slots
[*level
] = 0;
1383 WARN_ON(*level
< 0);
1384 WARN_ON(*level
>= BTRFS_MAX_LEVEL
);
1385 ret
= btrfs_free_extent(trans
, root
,
1386 bh_blocknr(path
->nodes
[*level
]), 1, 1);
1387 btrfs_block_release(root
, path
->nodes
[*level
]);
1388 path
->nodes
[*level
] = NULL
;
1395 * helper for dropping snapshots. This walks back up the tree in the path
1396 * to find the first node higher up where we haven't yet gone through
1399 static int walk_up_tree(struct btrfs_trans_handle
*trans
, struct btrfs_root
1400 *root
, struct btrfs_path
*path
, int *level
)
1405 struct btrfs_root_item
*root_item
= &root
->root_item
;
1407 for(i
= *level
; i
< BTRFS_MAX_LEVEL
- 1 && path
->nodes
[i
]; i
++) {
1408 slot
= path
->slots
[i
];
1409 if (slot
< btrfs_header_nritems(
1410 btrfs_buffer_header(path
->nodes
[i
])) - 1) {
1411 struct btrfs_node
*node
;
1412 node
= btrfs_buffer_node(path
->nodes
[i
]);
1415 WARN_ON(*level
== 0);
1416 memcpy(&root_item
->drop_progress
,
1417 &node
->ptrs
[path
->slots
[i
]].key
,
1418 sizeof(root_item
->drop_progress
));
1419 root_item
->drop_level
= i
;
1422 ret
= btrfs_free_extent(trans
, root
,
1423 bh_blocknr(path
->nodes
[*level
]),
1426 btrfs_block_release(root
, path
->nodes
[*level
]);
1427 path
->nodes
[*level
] = NULL
;
1435 * drop the reference count on the tree rooted at 'snap'. This traverses
1436 * the tree freeing any blocks that have a ref count of zero after being
1439 int btrfs_drop_snapshot(struct btrfs_trans_handle
*trans
, struct btrfs_root
1445 struct btrfs_path
*path
;
1449 struct btrfs_root_item
*root_item
= &root
->root_item
;
1451 path
= btrfs_alloc_path();
1454 level
= btrfs_header_level(btrfs_buffer_header(root
->node
));
1456 if (btrfs_disk_key_objectid(&root_item
->drop_progress
) == 0) {
1457 path
->nodes
[level
] = root
->node
;
1458 path
->slots
[level
] = 0;
1460 struct btrfs_key key
;
1461 struct btrfs_disk_key
*found_key
;
1462 struct btrfs_node
*node
;
1464 btrfs_disk_key_to_cpu(&key
, &root_item
->drop_progress
);
1465 level
= root_item
->drop_level
;
1466 path
->lowest_level
= level
;
1467 wret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1472 node
= btrfs_buffer_node(path
->nodes
[level
]);
1473 found_key
= &node
->ptrs
[path
->slots
[level
]].key
;
1474 WARN_ON(memcmp(found_key
, &root_item
->drop_progress
,
1475 sizeof(*found_key
)));
1478 wret
= walk_down_tree(trans
, root
, path
, &level
);
1484 wret
= walk_up_tree(trans
, root
, path
, &level
);
1490 if (num_walks
> 2) {
1496 for (i
= 0; i
<= orig_level
; i
++) {
1497 if (path
->nodes
[i
]) {
1498 btrfs_block_release(root
, path
->nodes
[i
]);
1503 btrfs_free_path(path
);
1507 static int free_block_group_radix(struct radix_tree_root
*radix
)
1510 struct btrfs_block_group_cache
*cache
[8];
1514 ret
= radix_tree_gang_lookup(radix
, (void **)cache
, 0,
1518 for (i
= 0; i
< ret
; i
++) {
1519 radix_tree_delete(radix
, cache
[i
]->key
.objectid
+
1520 cache
[i
]->key
.offset
- 1);
1527 int btrfs_free_block_groups(struct btrfs_fs_info
*info
)
1531 unsigned long gang
[16];
1534 ret
= free_block_group_radix(&info
->block_group_radix
);
1535 ret2
= free_block_group_radix(&info
->block_group_data_radix
);
1542 ret
= find_first_radix_bit(&info
->extent_map_radix
,
1543 gang
, 0, ARRAY_SIZE(gang
));
1546 for (i
= 0; i
< ret
; i
++) {
1547 clear_radix_bit(&info
->extent_map_radix
, gang
[i
]);
1553 int btrfs_read_block_groups(struct btrfs_root
*root
)
1555 struct btrfs_path
*path
;
1558 struct btrfs_block_group_item
*bi
;
1559 struct btrfs_block_group_cache
*cache
;
1560 struct btrfs_fs_info
*info
= root
->fs_info
;
1561 struct radix_tree_root
*radix
;
1562 struct btrfs_key key
;
1563 struct btrfs_key found_key
;
1564 struct btrfs_leaf
*leaf
;
1565 u64 group_size_blocks
;
1568 group_size_blocks
= BTRFS_BLOCK_GROUP_SIZE
>>
1569 root
->fs_info
->sb
->s_blocksize_bits
;
1570 root
= info
->extent_root
;
1572 key
.offset
= group_size_blocks
;
1574 btrfs_set_key_type(&key
, BTRFS_BLOCK_GROUP_ITEM_KEY
);
1576 path
= btrfs_alloc_path();
1581 ret
= btrfs_search_slot(NULL
, info
->extent_root
,
1587 leaf
= btrfs_buffer_leaf(path
->nodes
[0]);
1588 btrfs_disk_key_to_cpu(&found_key
,
1589 &leaf
->items
[path
->slots
[0]].key
);
1590 cache
= kmalloc(sizeof(*cache
), GFP_NOFS
);
1596 bi
= btrfs_item_ptr(leaf
, path
->slots
[0],
1597 struct btrfs_block_group_item
);
1598 if (bi
->flags
& BTRFS_BLOCK_GROUP_DATA
) {
1599 radix
= &info
->block_group_data_radix
;
1602 radix
= &info
->block_group_radix
;
1606 memcpy(&cache
->item
, bi
, sizeof(*bi
));
1607 memcpy(&cache
->key
, &found_key
, sizeof(found_key
));
1608 cache
->last_alloc
= cache
->key
.objectid
;
1609 cache
->first_free
= cache
->key
.objectid
;
1613 cache
->radix
= radix
;
1615 key
.objectid
= found_key
.objectid
+ found_key
.offset
;
1616 btrfs_release_path(root
, path
);
1617 ret
= radix_tree_insert(radix
, found_key
.objectid
+
1618 found_key
.offset
- 1,
1621 used
= btrfs_block_group_used(bi
);
1622 if (used
< div_factor(key
.offset
, 8)) {
1623 radix_tree_tag_set(radix
, found_key
.objectid
+
1624 found_key
.offset
- 1,
1625 BTRFS_BLOCK_GROUP_AVAIL
);
1628 btrfs_super_total_blocks(&info
->super_copy
))
1632 btrfs_free_path(path
);