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 "transaction.h"
23 #include "print-tree.h"
25 static int split_node(struct btrfs_trans_handle
*trans
, struct btrfs_root
26 *root
, struct btrfs_path
*path
, int level
);
27 static int split_leaf(struct btrfs_trans_handle
*trans
, struct btrfs_root
28 *root
, struct btrfs_key
*ins_key
,
29 struct btrfs_path
*path
, int data_size
, int extend
);
30 static int push_node_left(struct btrfs_trans_handle
*trans
,
31 struct btrfs_root
*root
, struct extent_buffer
*dst
,
32 struct extent_buffer
*src
, int empty
);
33 static int balance_node_right(struct btrfs_trans_handle
*trans
,
34 struct btrfs_root
*root
,
35 struct extent_buffer
*dst_buf
,
36 struct extent_buffer
*src_buf
);
37 static int del_ptr(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
38 struct btrfs_path
*path
, int level
, int slot
);
40 inline void btrfs_init_path(struct btrfs_path
*p
)
42 memset(p
, 0, sizeof(*p
));
45 struct btrfs_path
*btrfs_alloc_path(void)
47 struct btrfs_path
*path
;
48 path
= kmem_cache_alloc(btrfs_path_cachep
, GFP_NOFS
);
50 btrfs_init_path(path
);
56 void btrfs_free_path(struct btrfs_path
*p
)
58 btrfs_release_path(NULL
, p
);
59 kmem_cache_free(btrfs_path_cachep
, p
);
62 void btrfs_release_path(struct btrfs_root
*root
, struct btrfs_path
*p
)
65 for (i
= 0; i
< BTRFS_MAX_LEVEL
; i
++) {
68 free_extent_buffer(p
->nodes
[i
]);
70 memset(p
, 0, sizeof(*p
));
73 static void add_root_to_dirty_list(struct btrfs_root
*root
)
75 if (root
->track_dirty
&& list_empty(&root
->dirty_list
)) {
76 list_add(&root
->dirty_list
,
77 &root
->fs_info
->dirty_cowonly_roots
);
81 int btrfs_copy_root(struct btrfs_trans_handle
*trans
,
82 struct btrfs_root
*root
,
83 struct extent_buffer
*buf
,
84 struct extent_buffer
**cow_ret
, u64 new_root_objectid
)
86 struct extent_buffer
*cow
;
90 struct btrfs_key first_key
;
91 struct btrfs_root
*new_root
;
93 new_root
= kmalloc(sizeof(*new_root
), GFP_NOFS
);
97 memcpy(new_root
, root
, sizeof(*new_root
));
98 new_root
->root_key
.objectid
= new_root_objectid
;
100 WARN_ON(root
->ref_cows
&& trans
->transid
!=
101 root
->fs_info
->running_transaction
->transid
);
102 WARN_ON(root
->ref_cows
&& trans
->transid
!= root
->last_trans
);
104 level
= btrfs_header_level(buf
);
105 nritems
= btrfs_header_nritems(buf
);
108 btrfs_item_key_to_cpu(buf
, &first_key
, 0);
110 btrfs_node_key_to_cpu(buf
, &first_key
, 0);
112 first_key
.objectid
= 0;
114 cow
= __btrfs_alloc_free_block(trans
, new_root
, buf
->len
,
116 trans
->transid
, first_key
.objectid
,
117 level
, buf
->start
, 0);
123 copy_extent_buffer(cow
, buf
, 0, 0, cow
->len
);
124 btrfs_set_header_bytenr(cow
, cow
->start
);
125 btrfs_set_header_generation(cow
, trans
->transid
);
126 btrfs_set_header_owner(cow
, new_root_objectid
);
127 btrfs_clear_header_flag(cow
, BTRFS_HEADER_FLAG_WRITTEN
);
129 WARN_ON(btrfs_header_generation(buf
) > trans
->transid
);
130 ret
= btrfs_inc_ref(trans
, new_root
, buf
);
136 btrfs_mark_buffer_dirty(cow
);
141 int __btrfs_cow_block(struct btrfs_trans_handle
*trans
,
142 struct btrfs_root
*root
,
143 struct extent_buffer
*buf
,
144 struct extent_buffer
*parent
, int parent_slot
,
145 struct extent_buffer
**cow_ret
,
146 u64 search_start
, u64 empty_size
)
149 struct extent_buffer
*cow
;
152 int different_trans
= 0;
154 struct btrfs_key first_key
;
156 if (root
->ref_cows
) {
157 root_gen
= trans
->transid
;
161 WARN_ON(root
->ref_cows
&& trans
->transid
!=
162 root
->fs_info
->running_transaction
->transid
);
163 WARN_ON(root
->ref_cows
&& trans
->transid
!= root
->last_trans
);
165 level
= btrfs_header_level(buf
);
166 nritems
= btrfs_header_nritems(buf
);
169 btrfs_item_key_to_cpu(buf
, &first_key
, 0);
171 btrfs_node_key_to_cpu(buf
, &first_key
, 0);
173 first_key
.objectid
= 0;
175 cow
= __btrfs_alloc_free_block(trans
, root
, buf
->len
,
176 root
->root_key
.objectid
,
177 root_gen
, first_key
.objectid
, level
,
178 search_start
, empty_size
);
182 copy_extent_buffer(cow
, buf
, 0, 0, cow
->len
);
183 btrfs_set_header_bytenr(cow
, cow
->start
);
184 btrfs_set_header_generation(cow
, trans
->transid
);
185 btrfs_set_header_owner(cow
, root
->root_key
.objectid
);
186 btrfs_clear_header_flag(cow
, BTRFS_HEADER_FLAG_WRITTEN
);
188 WARN_ON(btrfs_header_generation(buf
) > trans
->transid
);
189 if (btrfs_header_generation(buf
) != trans
->transid
) {
191 ret
= btrfs_inc_ref(trans
, root
, buf
);
195 clean_tree_block(trans
, root
, buf
);
198 if (buf
== root
->node
) {
199 root_gen
= btrfs_header_generation(buf
);
201 extent_buffer_get(cow
);
202 if (buf
!= root
->commit_root
) {
203 btrfs_free_extent(trans
, root
, buf
->start
,
204 buf
->len
, root
->root_key
.objectid
,
207 free_extent_buffer(buf
);
208 add_root_to_dirty_list(root
);
210 root_gen
= btrfs_header_generation(parent
);
211 btrfs_set_node_blockptr(parent
, parent_slot
,
213 WARN_ON(trans
->transid
== 0);
214 btrfs_set_node_ptr_generation(parent
, parent_slot
,
216 btrfs_mark_buffer_dirty(parent
);
217 WARN_ON(btrfs_header_generation(parent
) != trans
->transid
);
218 btrfs_free_extent(trans
, root
, buf
->start
, buf
->len
,
219 btrfs_header_owner(parent
), root_gen
,
222 free_extent_buffer(buf
);
223 btrfs_mark_buffer_dirty(cow
);
228 int btrfs_cow_block(struct btrfs_trans_handle
*trans
,
229 struct btrfs_root
*root
, struct extent_buffer
*buf
,
230 struct extent_buffer
*parent
, int parent_slot
,
231 struct extent_buffer
**cow_ret
)
237 if (trans
->transaction
!= root
->fs_info
->running_transaction
) {
238 printk(KERN_CRIT
"trans %Lu running %Lu\n", trans
->transid
,
239 root
->fs_info
->running_transaction
->transid
);
242 if (trans
->transid
!= root
->fs_info
->generation
) {
243 printk(KERN_CRIT
"trans %Lu running %Lu\n", trans
->transid
,
244 root
->fs_info
->generation
);
248 header_trans
= btrfs_header_generation(buf
);
249 spin_lock(&root
->fs_info
->hash_lock
);
250 if (header_trans
== trans
->transid
&&
251 !btrfs_header_flag(buf
, BTRFS_HEADER_FLAG_WRITTEN
)) {
253 spin_unlock(&root
->fs_info
->hash_lock
);
256 spin_unlock(&root
->fs_info
->hash_lock
);
257 search_start
= buf
->start
& ~((u64
)(1024 * 1024 * 1024) - 1);
258 ret
= __btrfs_cow_block(trans
, root
, buf
, parent
,
259 parent_slot
, cow_ret
, search_start
, 0);
263 static int close_blocks(u64 blocknr
, u64 other
, u32 blocksize
)
265 if (blocknr
< other
&& other
- (blocknr
+ blocksize
) < 32768)
267 if (blocknr
> other
&& blocknr
- (other
+ blocksize
) < 32768)
273 * compare two keys in a memcmp fashion
275 static int comp_keys(struct btrfs_disk_key
*disk
, struct btrfs_key
*k2
)
279 btrfs_disk_key_to_cpu(&k1
, disk
);
281 if (k1
.objectid
> k2
->objectid
)
283 if (k1
.objectid
< k2
->objectid
)
285 if (k1
.type
> k2
->type
)
287 if (k1
.type
< k2
->type
)
289 if (k1
.offset
> k2
->offset
)
291 if (k1
.offset
< k2
->offset
)
297 int btrfs_realloc_node(struct btrfs_trans_handle
*trans
,
298 struct btrfs_root
*root
, struct extent_buffer
*parent
,
299 int start_slot
, int cache_only
, u64
*last_ret
,
300 struct btrfs_key
*progress
)
302 struct extent_buffer
*cur
;
303 struct extent_buffer
*tmp
;
306 u64 search_start
= *last_ret
;
316 int progress_passed
= 0;
317 struct btrfs_disk_key disk_key
;
319 parent_level
= btrfs_header_level(parent
);
320 if (cache_only
&& parent_level
!= 1)
323 if (trans
->transaction
!= root
->fs_info
->running_transaction
) {
324 printk(KERN_CRIT
"trans %Lu running %Lu\n", trans
->transid
,
325 root
->fs_info
->running_transaction
->transid
);
328 if (trans
->transid
!= root
->fs_info
->generation
) {
329 printk(KERN_CRIT
"trans %Lu running %Lu\n", trans
->transid
,
330 root
->fs_info
->generation
);
334 parent_nritems
= btrfs_header_nritems(parent
);
335 blocksize
= btrfs_level_size(root
, parent_level
- 1);
336 end_slot
= parent_nritems
;
338 if (parent_nritems
== 1)
341 for (i
= start_slot
; i
< end_slot
; i
++) {
344 if (!parent
->map_token
) {
345 map_extent_buffer(parent
,
346 btrfs_node_key_ptr_offset(i
),
347 sizeof(struct btrfs_key_ptr
),
348 &parent
->map_token
, &parent
->kaddr
,
349 &parent
->map_start
, &parent
->map_len
,
352 btrfs_node_key(parent
, &disk_key
, i
);
353 if (!progress_passed
&& comp_keys(&disk_key
, progress
) < 0)
357 blocknr
= btrfs_node_blockptr(parent
, i
);
358 gen
= btrfs_node_ptr_generation(parent
, i
);
360 last_block
= blocknr
;
363 other
= btrfs_node_blockptr(parent
, i
- 1);
364 close
= close_blocks(blocknr
, other
, blocksize
);
366 if (close
&& i
< end_slot
- 2) {
367 other
= btrfs_node_blockptr(parent
, i
+ 1);
368 close
= close_blocks(blocknr
, other
, blocksize
);
371 last_block
= blocknr
;
374 if (parent
->map_token
) {
375 unmap_extent_buffer(parent
, parent
->map_token
,
377 parent
->map_token
= NULL
;
380 cur
= btrfs_find_tree_block(root
, blocknr
, blocksize
);
382 uptodate
= btrfs_buffer_uptodate(cur
, gen
);
385 if (!cur
|| !uptodate
) {
387 free_extent_buffer(cur
);
391 cur
= read_tree_block(root
, blocknr
,
393 } else if (!uptodate
) {
394 btrfs_read_buffer(cur
, gen
);
397 if (search_start
== 0)
398 search_start
= last_block
;
400 err
= __btrfs_cow_block(trans
, root
, cur
, parent
, i
,
403 (end_slot
- i
) * blocksize
));
405 free_extent_buffer(cur
);
408 search_start
= tmp
->start
;
409 last_block
= tmp
->start
;
410 *last_ret
= search_start
;
411 if (parent_level
== 1)
412 btrfs_clear_buffer_defrag(tmp
);
413 free_extent_buffer(tmp
);
415 if (parent
->map_token
) {
416 unmap_extent_buffer(parent
, parent
->map_token
,
418 parent
->map_token
= NULL
;
424 * The leaf data grows from end-to-front in the node.
425 * this returns the address of the start of the last item,
426 * which is the stop of the leaf data stack
428 static inline unsigned int leaf_data_end(struct btrfs_root
*root
,
429 struct extent_buffer
*leaf
)
431 u32 nr
= btrfs_header_nritems(leaf
);
433 return BTRFS_LEAF_DATA_SIZE(root
);
434 return btrfs_item_offset_nr(leaf
, nr
- 1);
437 static int check_node(struct btrfs_root
*root
, struct btrfs_path
*path
,
440 struct extent_buffer
*parent
= NULL
;
441 struct extent_buffer
*node
= path
->nodes
[level
];
442 struct btrfs_disk_key parent_key
;
443 struct btrfs_disk_key node_key
;
446 struct btrfs_key cpukey
;
447 u32 nritems
= btrfs_header_nritems(node
);
449 if (path
->nodes
[level
+ 1])
450 parent
= path
->nodes
[level
+ 1];
452 slot
= path
->slots
[level
];
453 BUG_ON(nritems
== 0);
455 parent_slot
= path
->slots
[level
+ 1];
456 btrfs_node_key(parent
, &parent_key
, parent_slot
);
457 btrfs_node_key(node
, &node_key
, 0);
458 BUG_ON(memcmp(&parent_key
, &node_key
,
459 sizeof(struct btrfs_disk_key
)));
460 BUG_ON(btrfs_node_blockptr(parent
, parent_slot
) !=
461 btrfs_header_bytenr(node
));
463 BUG_ON(nritems
> BTRFS_NODEPTRS_PER_BLOCK(root
));
465 btrfs_node_key_to_cpu(node
, &cpukey
, slot
- 1);
466 btrfs_node_key(node
, &node_key
, slot
);
467 BUG_ON(comp_keys(&node_key
, &cpukey
) <= 0);
469 if (slot
< nritems
- 1) {
470 btrfs_node_key_to_cpu(node
, &cpukey
, slot
+ 1);
471 btrfs_node_key(node
, &node_key
, slot
);
472 BUG_ON(comp_keys(&node_key
, &cpukey
) >= 0);
477 static int check_leaf(struct btrfs_root
*root
, struct btrfs_path
*path
,
480 struct extent_buffer
*leaf
= path
->nodes
[level
];
481 struct extent_buffer
*parent
= NULL
;
483 struct btrfs_key cpukey
;
484 struct btrfs_disk_key parent_key
;
485 struct btrfs_disk_key leaf_key
;
486 int slot
= path
->slots
[0];
488 u32 nritems
= btrfs_header_nritems(leaf
);
490 if (path
->nodes
[level
+ 1])
491 parent
= path
->nodes
[level
+ 1];
497 parent_slot
= path
->slots
[level
+ 1];
498 btrfs_node_key(parent
, &parent_key
, parent_slot
);
499 btrfs_item_key(leaf
, &leaf_key
, 0);
501 BUG_ON(memcmp(&parent_key
, &leaf_key
,
502 sizeof(struct btrfs_disk_key
)));
503 BUG_ON(btrfs_node_blockptr(parent
, parent_slot
) !=
504 btrfs_header_bytenr(leaf
));
507 for (i
= 0; nritems
> 1 && i
< nritems
- 2; i
++) {
508 btrfs_item_key_to_cpu(leaf
, &cpukey
, i
+ 1);
509 btrfs_item_key(leaf
, &leaf_key
, i
);
510 if (comp_keys(&leaf_key
, &cpukey
) >= 0) {
511 btrfs_print_leaf(root
, leaf
);
512 printk("slot %d offset bad key\n", i
);
515 if (btrfs_item_offset_nr(leaf
, i
) !=
516 btrfs_item_end_nr(leaf
, i
+ 1)) {
517 btrfs_print_leaf(root
, leaf
);
518 printk("slot %d offset bad\n", i
);
522 if (btrfs_item_offset_nr(leaf
, i
) +
523 btrfs_item_size_nr(leaf
, i
) !=
524 BTRFS_LEAF_DATA_SIZE(root
)) {
525 btrfs_print_leaf(root
, leaf
);
526 printk("slot %d first offset bad\n", i
);
532 if (btrfs_item_size_nr(leaf
, nritems
- 1) > 4096) {
533 btrfs_print_leaf(root
, leaf
);
534 printk("slot %d bad size \n", nritems
- 1);
539 if (slot
!= 0 && slot
< nritems
- 1) {
540 btrfs_item_key(leaf
, &leaf_key
, slot
);
541 btrfs_item_key_to_cpu(leaf
, &cpukey
, slot
- 1);
542 if (comp_keys(&leaf_key
, &cpukey
) <= 0) {
543 btrfs_print_leaf(root
, leaf
);
544 printk("slot %d offset bad key\n", slot
);
547 if (btrfs_item_offset_nr(leaf
, slot
- 1) !=
548 btrfs_item_end_nr(leaf
, slot
)) {
549 btrfs_print_leaf(root
, leaf
);
550 printk("slot %d offset bad\n", slot
);
554 if (slot
< nritems
- 1) {
555 btrfs_item_key(leaf
, &leaf_key
, slot
);
556 btrfs_item_key_to_cpu(leaf
, &cpukey
, slot
+ 1);
557 BUG_ON(comp_keys(&leaf_key
, &cpukey
) >= 0);
558 if (btrfs_item_offset_nr(leaf
, slot
) !=
559 btrfs_item_end_nr(leaf
, slot
+ 1)) {
560 btrfs_print_leaf(root
, leaf
);
561 printk("slot %d offset bad\n", slot
);
565 BUG_ON(btrfs_item_offset_nr(leaf
, 0) +
566 btrfs_item_size_nr(leaf
, 0) != BTRFS_LEAF_DATA_SIZE(root
));
570 static int noinline
check_block(struct btrfs_root
*root
,
571 struct btrfs_path
*path
, int level
)
575 if (btrfs_header_level(path
->nodes
[level
]) != level
)
576 printk("warning: bad level %Lu wanted %d found %d\n",
577 path
->nodes
[level
]->start
, level
,
578 btrfs_header_level(path
->nodes
[level
]));
579 found_start
= btrfs_header_bytenr(path
->nodes
[level
]);
580 if (found_start
!= path
->nodes
[level
]->start
) {
581 printk("warning: bad bytentr %Lu found %Lu\n",
582 path
->nodes
[level
]->start
, found_start
);
585 struct extent_buffer
*buf
= path
->nodes
[level
];
587 if (memcmp_extent_buffer(buf
, root
->fs_info
->fsid
,
588 (unsigned long)btrfs_header_fsid(buf
),
590 printk("warning bad block %Lu\n", buf
->start
);
595 return check_leaf(root
, path
, level
);
596 return check_node(root
, path
, level
);
600 * search for key in the extent_buffer. The items start at offset p,
601 * and they are item_size apart. There are 'max' items in p.
603 * the slot in the array is returned via slot, and it points to
604 * the place where you would insert key if it is not found in
607 * slot may point to max if the key is bigger than all of the keys
609 static int generic_bin_search(struct extent_buffer
*eb
, unsigned long p
,
610 int item_size
, struct btrfs_key
*key
,
617 struct btrfs_disk_key
*tmp
= NULL
;
618 struct btrfs_disk_key unaligned
;
619 unsigned long offset
;
620 char *map_token
= NULL
;
622 unsigned long map_start
= 0;
623 unsigned long map_len
= 0;
627 mid
= (low
+ high
) / 2;
628 offset
= p
+ mid
* item_size
;
630 if (!map_token
|| offset
< map_start
||
631 (offset
+ sizeof(struct btrfs_disk_key
)) >
632 map_start
+ map_len
) {
634 unmap_extent_buffer(eb
, map_token
, KM_USER0
);
637 err
= map_extent_buffer(eb
, offset
,
638 sizeof(struct btrfs_disk_key
),
640 &map_start
, &map_len
, KM_USER0
);
643 tmp
= (struct btrfs_disk_key
*)(kaddr
+ offset
-
646 read_extent_buffer(eb
, &unaligned
,
647 offset
, sizeof(unaligned
));
652 tmp
= (struct btrfs_disk_key
*)(kaddr
+ offset
-
655 ret
= comp_keys(tmp
, key
);
664 unmap_extent_buffer(eb
, map_token
, KM_USER0
);
670 unmap_extent_buffer(eb
, map_token
, KM_USER0
);
675 * simple bin_search frontend that does the right thing for
678 static int bin_search(struct extent_buffer
*eb
, struct btrfs_key
*key
,
679 int level
, int *slot
)
682 return generic_bin_search(eb
,
683 offsetof(struct btrfs_leaf
, items
),
684 sizeof(struct btrfs_item
),
685 key
, btrfs_header_nritems(eb
),
688 return generic_bin_search(eb
,
689 offsetof(struct btrfs_node
, ptrs
),
690 sizeof(struct btrfs_key_ptr
),
691 key
, btrfs_header_nritems(eb
),
697 static struct extent_buffer
*read_node_slot(struct btrfs_root
*root
,
698 struct extent_buffer
*parent
, int slot
)
700 int level
= btrfs_header_level(parent
);
703 if (slot
>= btrfs_header_nritems(parent
))
708 return read_tree_block(root
, btrfs_node_blockptr(parent
, slot
),
709 btrfs_level_size(root
, level
- 1),
710 btrfs_node_ptr_generation(parent
, slot
));
713 static int balance_level(struct btrfs_trans_handle
*trans
,
714 struct btrfs_root
*root
,
715 struct btrfs_path
*path
, int level
)
717 struct extent_buffer
*right
= NULL
;
718 struct extent_buffer
*mid
;
719 struct extent_buffer
*left
= NULL
;
720 struct extent_buffer
*parent
= NULL
;
724 int orig_slot
= path
->slots
[level
];
725 int err_on_enospc
= 0;
731 mid
= path
->nodes
[level
];
732 WARN_ON(btrfs_header_generation(mid
) != trans
->transid
);
734 orig_ptr
= btrfs_node_blockptr(mid
, orig_slot
);
736 if (level
< BTRFS_MAX_LEVEL
- 1)
737 parent
= path
->nodes
[level
+ 1];
738 pslot
= path
->slots
[level
+ 1];
741 * deal with the case where there is only one pointer in the root
742 * by promoting the node below to a root
745 struct extent_buffer
*child
;
747 if (btrfs_header_nritems(mid
) != 1)
750 /* promote the child to a root */
751 child
= read_node_slot(root
, mid
, 0);
753 ret
= btrfs_cow_block(trans
, root
, child
, mid
, 0, &child
);
757 add_root_to_dirty_list(root
);
758 path
->nodes
[level
] = NULL
;
759 clean_tree_block(trans
, root
, mid
);
760 /* once for the path */
761 free_extent_buffer(mid
);
762 ret
= btrfs_free_extent(trans
, root
, mid
->start
, mid
->len
,
763 root
->root_key
.objectid
,
764 btrfs_header_generation(mid
), 0, 0, 1);
765 /* once for the root ptr */
766 free_extent_buffer(mid
);
769 if (btrfs_header_nritems(mid
) >
770 BTRFS_NODEPTRS_PER_BLOCK(root
) / 4)
773 if (btrfs_header_nritems(mid
) < 2)
776 left
= read_node_slot(root
, parent
, pslot
- 1);
778 wret
= btrfs_cow_block(trans
, root
, left
,
779 parent
, pslot
- 1, &left
);
785 right
= read_node_slot(root
, parent
, pslot
+ 1);
787 wret
= btrfs_cow_block(trans
, root
, right
,
788 parent
, pslot
+ 1, &right
);
795 /* first, try to make some room in the middle buffer */
797 orig_slot
+= btrfs_header_nritems(left
);
798 wret
= push_node_left(trans
, root
, left
, mid
, 1);
801 if (btrfs_header_nritems(mid
) < 2)
806 * then try to empty the right most buffer into the middle
809 wret
= push_node_left(trans
, root
, mid
, right
, 1);
810 if (wret
< 0 && wret
!= -ENOSPC
)
812 if (btrfs_header_nritems(right
) == 0) {
813 u64 bytenr
= right
->start
;
814 u64 generation
= btrfs_header_generation(parent
);
815 u32 blocksize
= right
->len
;
817 clean_tree_block(trans
, root
, right
);
818 free_extent_buffer(right
);
820 wret
= del_ptr(trans
, root
, path
, level
+ 1, pslot
+
824 wret
= btrfs_free_extent(trans
, root
, bytenr
,
826 btrfs_header_owner(parent
),
827 generation
, 0, 0, 1);
831 struct btrfs_disk_key right_key
;
832 btrfs_node_key(right
, &right_key
, 0);
833 btrfs_set_node_key(parent
, &right_key
, pslot
+ 1);
834 btrfs_mark_buffer_dirty(parent
);
837 if (btrfs_header_nritems(mid
) == 1) {
839 * we're not allowed to leave a node with one item in the
840 * tree during a delete. A deletion from lower in the tree
841 * could try to delete the only pointer in this node.
842 * So, pull some keys from the left.
843 * There has to be a left pointer at this point because
844 * otherwise we would have pulled some pointers from the
848 wret
= balance_node_right(trans
, root
, mid
, left
);
854 wret
= push_node_left(trans
, root
, left
, mid
, 1);
860 if (btrfs_header_nritems(mid
) == 0) {
861 /* we've managed to empty the middle node, drop it */
862 u64 root_gen
= btrfs_header_generation(parent
);
863 u64 bytenr
= mid
->start
;
864 u32 blocksize
= mid
->len
;
865 clean_tree_block(trans
, root
, mid
);
866 free_extent_buffer(mid
);
868 wret
= del_ptr(trans
, root
, path
, level
+ 1, pslot
);
871 wret
= btrfs_free_extent(trans
, root
, bytenr
, blocksize
,
872 btrfs_header_owner(parent
),
877 /* update the parent key to reflect our changes */
878 struct btrfs_disk_key mid_key
;
879 btrfs_node_key(mid
, &mid_key
, 0);
880 btrfs_set_node_key(parent
, &mid_key
, pslot
);
881 btrfs_mark_buffer_dirty(parent
);
884 /* update the path */
886 if (btrfs_header_nritems(left
) > orig_slot
) {
887 extent_buffer_get(left
);
888 path
->nodes
[level
] = left
;
889 path
->slots
[level
+ 1] -= 1;
890 path
->slots
[level
] = orig_slot
;
892 free_extent_buffer(mid
);
894 orig_slot
-= btrfs_header_nritems(left
);
895 path
->slots
[level
] = orig_slot
;
898 /* double check we haven't messed things up */
899 check_block(root
, path
, level
);
901 btrfs_node_blockptr(path
->nodes
[level
], path
->slots
[level
]))
905 free_extent_buffer(right
);
907 free_extent_buffer(left
);
911 /* returns zero if the push worked, non-zero otherwise */
912 static int noinline
push_nodes_for_insert(struct btrfs_trans_handle
*trans
,
913 struct btrfs_root
*root
,
914 struct btrfs_path
*path
, int level
)
916 struct extent_buffer
*right
= NULL
;
917 struct extent_buffer
*mid
;
918 struct extent_buffer
*left
= NULL
;
919 struct extent_buffer
*parent
= NULL
;
923 int orig_slot
= path
->slots
[level
];
929 mid
= path
->nodes
[level
];
930 WARN_ON(btrfs_header_generation(mid
) != trans
->transid
);
931 orig_ptr
= btrfs_node_blockptr(mid
, orig_slot
);
933 if (level
< BTRFS_MAX_LEVEL
- 1)
934 parent
= path
->nodes
[level
+ 1];
935 pslot
= path
->slots
[level
+ 1];
940 left
= read_node_slot(root
, parent
, pslot
- 1);
942 /* first, try to make some room in the middle buffer */
945 left_nr
= btrfs_header_nritems(left
);
946 if (left_nr
>= BTRFS_NODEPTRS_PER_BLOCK(root
) - 1) {
949 ret
= btrfs_cow_block(trans
, root
, left
, parent
,
954 wret
= push_node_left(trans
, root
,
961 struct btrfs_disk_key disk_key
;
962 orig_slot
+= left_nr
;
963 btrfs_node_key(mid
, &disk_key
, 0);
964 btrfs_set_node_key(parent
, &disk_key
, pslot
);
965 btrfs_mark_buffer_dirty(parent
);
966 if (btrfs_header_nritems(left
) > orig_slot
) {
967 path
->nodes
[level
] = left
;
968 path
->slots
[level
+ 1] -= 1;
969 path
->slots
[level
] = orig_slot
;
970 free_extent_buffer(mid
);
973 btrfs_header_nritems(left
);
974 path
->slots
[level
] = orig_slot
;
975 free_extent_buffer(left
);
979 free_extent_buffer(left
);
981 right
= read_node_slot(root
, parent
, pslot
+ 1);
984 * then try to empty the right most buffer into the middle
988 right_nr
= btrfs_header_nritems(right
);
989 if (right_nr
>= BTRFS_NODEPTRS_PER_BLOCK(root
) - 1) {
992 ret
= btrfs_cow_block(trans
, root
, right
,
998 wret
= balance_node_right(trans
, root
,
1005 struct btrfs_disk_key disk_key
;
1007 btrfs_node_key(right
, &disk_key
, 0);
1008 btrfs_set_node_key(parent
, &disk_key
, pslot
+ 1);
1009 btrfs_mark_buffer_dirty(parent
);
1011 if (btrfs_header_nritems(mid
) <= orig_slot
) {
1012 path
->nodes
[level
] = right
;
1013 path
->slots
[level
+ 1] += 1;
1014 path
->slots
[level
] = orig_slot
-
1015 btrfs_header_nritems(mid
);
1016 free_extent_buffer(mid
);
1018 free_extent_buffer(right
);
1022 free_extent_buffer(right
);
1028 * readahead one full node of leaves
1030 static void reada_for_search(struct btrfs_root
*root
, struct btrfs_path
*path
,
1031 int level
, int slot
, u64 objectid
)
1033 struct extent_buffer
*node
;
1034 struct btrfs_disk_key disk_key
;
1040 int direction
= path
->reada
;
1041 struct extent_buffer
*eb
;
1049 if (!path
->nodes
[level
])
1052 node
= path
->nodes
[level
];
1053 search
= btrfs_node_blockptr(node
, slot
);
1054 blocksize
= btrfs_level_size(root
, level
- 1);
1055 eb
= btrfs_find_tree_block(root
, search
, blocksize
);
1057 free_extent_buffer(eb
);
1061 highest_read
= search
;
1062 lowest_read
= search
;
1064 nritems
= btrfs_header_nritems(node
);
1067 if (direction
< 0) {
1071 } else if (direction
> 0) {
1076 if (path
->reada
< 0 && objectid
) {
1077 btrfs_node_key(node
, &disk_key
, nr
);
1078 if (btrfs_disk_key_objectid(&disk_key
) != objectid
)
1081 search
= btrfs_node_blockptr(node
, nr
);
1082 if ((search
>= lowest_read
&& search
<= highest_read
) ||
1083 (search
< lowest_read
&& lowest_read
- search
<= 32768) ||
1084 (search
> highest_read
&& search
- highest_read
<= 32768)) {
1085 readahead_tree_block(root
, search
, blocksize
,
1086 btrfs_node_ptr_generation(node
, nr
));
1090 if (path
->reada
< 2 && (nread
> (256 * 1024) || nscan
> 32))
1092 if(nread
> (1024 * 1024) || nscan
> 128)
1095 if (search
< lowest_read
)
1096 lowest_read
= search
;
1097 if (search
> highest_read
)
1098 highest_read
= search
;
1102 * look for key in the tree. path is filled in with nodes along the way
1103 * if key is found, we return zero and you can find the item in the leaf
1104 * level of the path (level 0)
1106 * If the key isn't found, the path points to the slot where it should
1107 * be inserted, and 1 is returned. If there are other errors during the
1108 * search a negative error number is returned.
1110 * if ins_len > 0, nodes and leaves will be split as we walk down the
1111 * tree. if ins_len < 0, nodes will be merged as we walk down the tree (if
1114 int btrfs_search_slot(struct btrfs_trans_handle
*trans
, struct btrfs_root
1115 *root
, struct btrfs_key
*key
, struct btrfs_path
*p
, int
1118 struct extent_buffer
*b
;
1122 int should_reada
= p
->reada
;
1123 u8 lowest_level
= 0;
1125 lowest_level
= p
->lowest_level
;
1126 WARN_ON(lowest_level
&& ins_len
);
1127 WARN_ON(p
->nodes
[0] != NULL
);
1128 WARN_ON(!mutex_is_locked(&root
->fs_info
->fs_mutex
));
1131 extent_buffer_get(b
);
1133 level
= btrfs_header_level(b
);
1136 wret
= btrfs_cow_block(trans
, root
, b
,
1137 p
->nodes
[level
+ 1],
1138 p
->slots
[level
+ 1],
1141 free_extent_buffer(b
);
1145 BUG_ON(!cow
&& ins_len
);
1146 if (level
!= btrfs_header_level(b
))
1148 level
= btrfs_header_level(b
);
1149 p
->nodes
[level
] = b
;
1150 ret
= check_block(root
, p
, level
);
1153 ret
= bin_search(b
, key
, level
, &slot
);
1155 if (ret
&& slot
> 0)
1157 p
->slots
[level
] = slot
;
1158 if (ins_len
> 0 && btrfs_header_nritems(b
) >=
1159 BTRFS_NODEPTRS_PER_BLOCK(root
) - 3) {
1160 int sret
= split_node(trans
, root
, p
, level
);
1164 b
= p
->nodes
[level
];
1165 slot
= p
->slots
[level
];
1166 } else if (ins_len
< 0) {
1167 int sret
= balance_level(trans
, root
, p
,
1171 b
= p
->nodes
[level
];
1173 btrfs_release_path(NULL
, p
);
1176 slot
= p
->slots
[level
];
1177 BUG_ON(btrfs_header_nritems(b
) == 1);
1179 /* this is only true while dropping a snapshot */
1180 if (level
== lowest_level
)
1184 reada_for_search(root
, p
, level
, slot
,
1187 b
= read_node_slot(root
, b
, slot
);
1189 p
->slots
[level
] = slot
;
1190 if (ins_len
> 0 && btrfs_leaf_free_space(root
, b
) <
1191 sizeof(struct btrfs_item
) + ins_len
) {
1192 int sret
= split_leaf(trans
, root
, key
,
1193 p
, ins_len
, ret
== 0);
1205 * adjust the pointers going up the tree, starting at level
1206 * making sure the right key of each node is points to 'key'.
1207 * This is used after shifting pointers to the left, so it stops
1208 * fixing up pointers when a given leaf/node is not in slot 0 of the
1211 * If this fails to write a tree block, it returns -1, but continues
1212 * fixing up the blocks in ram so the tree is consistent.
1214 static int fixup_low_keys(struct btrfs_trans_handle
*trans
,
1215 struct btrfs_root
*root
, struct btrfs_path
*path
,
1216 struct btrfs_disk_key
*key
, int level
)
1220 struct extent_buffer
*t
;
1222 for (i
= level
; i
< BTRFS_MAX_LEVEL
; i
++) {
1223 int tslot
= path
->slots
[i
];
1224 if (!path
->nodes
[i
])
1227 btrfs_set_node_key(t
, key
, tslot
);
1228 btrfs_mark_buffer_dirty(path
->nodes
[i
]);
1236 * try to push data from one node into the next node left in the
1239 * returns 0 if some ptrs were pushed left, < 0 if there was some horrible
1240 * error, and > 0 if there was no room in the left hand block.
1242 static int push_node_left(struct btrfs_trans_handle
*trans
,
1243 struct btrfs_root
*root
, struct extent_buffer
*dst
,
1244 struct extent_buffer
*src
, int empty
)
1251 src_nritems
= btrfs_header_nritems(src
);
1252 dst_nritems
= btrfs_header_nritems(dst
);
1253 push_items
= BTRFS_NODEPTRS_PER_BLOCK(root
) - dst_nritems
;
1254 WARN_ON(btrfs_header_generation(src
) != trans
->transid
);
1255 WARN_ON(btrfs_header_generation(dst
) != trans
->transid
);
1257 if (!empty
&& src_nritems
<= 8)
1260 if (push_items
<= 0) {
1265 push_items
= min(src_nritems
, push_items
);
1266 if (push_items
< src_nritems
) {
1267 /* leave at least 8 pointers in the node if
1268 * we aren't going to empty it
1270 if (src_nritems
- push_items
< 8) {
1271 if (push_items
<= 8)
1277 push_items
= min(src_nritems
- 8, push_items
);
1279 copy_extent_buffer(dst
, src
,
1280 btrfs_node_key_ptr_offset(dst_nritems
),
1281 btrfs_node_key_ptr_offset(0),
1282 push_items
* sizeof(struct btrfs_key_ptr
));
1284 if (push_items
< src_nritems
) {
1285 memmove_extent_buffer(src
, btrfs_node_key_ptr_offset(0),
1286 btrfs_node_key_ptr_offset(push_items
),
1287 (src_nritems
- push_items
) *
1288 sizeof(struct btrfs_key_ptr
));
1290 btrfs_set_header_nritems(src
, src_nritems
- push_items
);
1291 btrfs_set_header_nritems(dst
, dst_nritems
+ push_items
);
1292 btrfs_mark_buffer_dirty(src
);
1293 btrfs_mark_buffer_dirty(dst
);
1298 * try to push data from one node into the next node right in the
1301 * returns 0 if some ptrs were pushed, < 0 if there was some horrible
1302 * error, and > 0 if there was no room in the right hand block.
1304 * this will only push up to 1/2 the contents of the left node over
1306 static int balance_node_right(struct btrfs_trans_handle
*trans
,
1307 struct btrfs_root
*root
,
1308 struct extent_buffer
*dst
,
1309 struct extent_buffer
*src
)
1317 WARN_ON(btrfs_header_generation(src
) != trans
->transid
);
1318 WARN_ON(btrfs_header_generation(dst
) != trans
->transid
);
1320 src_nritems
= btrfs_header_nritems(src
);
1321 dst_nritems
= btrfs_header_nritems(dst
);
1322 push_items
= BTRFS_NODEPTRS_PER_BLOCK(root
) - dst_nritems
;
1323 if (push_items
<= 0) {
1327 if (src_nritems
< 4) {
1331 max_push
= src_nritems
/ 2 + 1;
1332 /* don't try to empty the node */
1333 if (max_push
>= src_nritems
) {
1337 if (max_push
< push_items
)
1338 push_items
= max_push
;
1340 memmove_extent_buffer(dst
, btrfs_node_key_ptr_offset(push_items
),
1341 btrfs_node_key_ptr_offset(0),
1343 sizeof(struct btrfs_key_ptr
));
1345 copy_extent_buffer(dst
, src
,
1346 btrfs_node_key_ptr_offset(0),
1347 btrfs_node_key_ptr_offset(src_nritems
- push_items
),
1348 push_items
* sizeof(struct btrfs_key_ptr
));
1350 btrfs_set_header_nritems(src
, src_nritems
- push_items
);
1351 btrfs_set_header_nritems(dst
, dst_nritems
+ push_items
);
1353 btrfs_mark_buffer_dirty(src
);
1354 btrfs_mark_buffer_dirty(dst
);
1359 * helper function to insert a new root level in the tree.
1360 * A new node is allocated, and a single item is inserted to
1361 * point to the existing root
1363 * returns zero on success or < 0 on failure.
1365 static int noinline
insert_new_root(struct btrfs_trans_handle
*trans
,
1366 struct btrfs_root
*root
,
1367 struct btrfs_path
*path
, int level
)
1371 struct extent_buffer
*lower
;
1372 struct extent_buffer
*c
;
1373 struct btrfs_disk_key lower_key
;
1375 BUG_ON(path
->nodes
[level
]);
1376 BUG_ON(path
->nodes
[level
-1] != root
->node
);
1379 root_gen
= trans
->transid
;
1383 lower
= path
->nodes
[level
-1];
1385 btrfs_item_key(lower
, &lower_key
, 0);
1387 btrfs_node_key(lower
, &lower_key
, 0);
1389 c
= __btrfs_alloc_free_block(trans
, root
, root
->nodesize
,
1390 root
->root_key
.objectid
,
1391 root_gen
, lower_key
.objectid
, level
,
1392 root
->node
->start
, 0);
1395 memset_extent_buffer(c
, 0, 0, root
->nodesize
);
1396 btrfs_set_header_nritems(c
, 1);
1397 btrfs_set_header_level(c
, level
);
1398 btrfs_set_header_bytenr(c
, c
->start
);
1399 btrfs_set_header_generation(c
, trans
->transid
);
1400 btrfs_set_header_owner(c
, root
->root_key
.objectid
);
1402 write_extent_buffer(c
, root
->fs_info
->fsid
,
1403 (unsigned long)btrfs_header_fsid(c
),
1406 write_extent_buffer(c
, root
->fs_info
->chunk_tree_uuid
,
1407 (unsigned long)btrfs_header_chunk_tree_uuid(c
),
1410 btrfs_set_node_key(c
, &lower_key
, 0);
1411 btrfs_set_node_blockptr(c
, 0, lower
->start
);
1412 lower_gen
= btrfs_header_generation(lower
);
1413 WARN_ON(lower_gen
== 0);
1415 btrfs_set_node_ptr_generation(c
, 0, lower_gen
);
1417 btrfs_mark_buffer_dirty(c
);
1419 /* the super has an extra ref to root->node */
1420 free_extent_buffer(root
->node
);
1422 add_root_to_dirty_list(root
);
1423 extent_buffer_get(c
);
1424 path
->nodes
[level
] = c
;
1425 path
->slots
[level
] = 0;
1427 if (root
->ref_cows
&& lower_gen
!= trans
->transid
) {
1428 struct btrfs_path
*back_path
= btrfs_alloc_path();
1430 ret
= btrfs_insert_extent_backref(trans
,
1431 root
->fs_info
->extent_root
,
1433 root
->root_key
.objectid
,
1434 trans
->transid
, 0, 0);
1436 btrfs_free_path(back_path
);
1442 * worker function to insert a single pointer in a node.
1443 * the node should have enough room for the pointer already
1445 * slot and level indicate where you want the key to go, and
1446 * blocknr is the block the key points to.
1448 * returns zero on success and < 0 on any error
1450 static int insert_ptr(struct btrfs_trans_handle
*trans
, struct btrfs_root
1451 *root
, struct btrfs_path
*path
, struct btrfs_disk_key
1452 *key
, u64 bytenr
, int slot
, int level
)
1454 struct extent_buffer
*lower
;
1457 BUG_ON(!path
->nodes
[level
]);
1458 lower
= path
->nodes
[level
];
1459 nritems
= btrfs_header_nritems(lower
);
1462 if (nritems
== BTRFS_NODEPTRS_PER_BLOCK(root
))
1464 if (slot
!= nritems
) {
1465 memmove_extent_buffer(lower
,
1466 btrfs_node_key_ptr_offset(slot
+ 1),
1467 btrfs_node_key_ptr_offset(slot
),
1468 (nritems
- slot
) * sizeof(struct btrfs_key_ptr
));
1470 btrfs_set_node_key(lower
, key
, slot
);
1471 btrfs_set_node_blockptr(lower
, slot
, bytenr
);
1472 WARN_ON(trans
->transid
== 0);
1473 btrfs_set_node_ptr_generation(lower
, slot
, trans
->transid
);
1474 btrfs_set_header_nritems(lower
, nritems
+ 1);
1475 btrfs_mark_buffer_dirty(lower
);
1480 * split the node at the specified level in path in two.
1481 * The path is corrected to point to the appropriate node after the split
1483 * Before splitting this tries to make some room in the node by pushing
1484 * left and right, if either one works, it returns right away.
1486 * returns 0 on success and < 0 on failure
1488 static int split_node(struct btrfs_trans_handle
*trans
, struct btrfs_root
1489 *root
, struct btrfs_path
*path
, int level
)
1492 struct extent_buffer
*c
;
1493 struct extent_buffer
*split
;
1494 struct btrfs_disk_key disk_key
;
1500 c
= path
->nodes
[level
];
1501 WARN_ON(btrfs_header_generation(c
) != trans
->transid
);
1502 if (c
== root
->node
) {
1503 /* trying to split the root, lets make a new one */
1504 ret
= insert_new_root(trans
, root
, path
, level
+ 1);
1508 ret
= push_nodes_for_insert(trans
, root
, path
, level
);
1509 c
= path
->nodes
[level
];
1510 if (!ret
&& btrfs_header_nritems(c
) <
1511 BTRFS_NODEPTRS_PER_BLOCK(root
) - 3)
1517 c_nritems
= btrfs_header_nritems(c
);
1519 root_gen
= trans
->transid
;
1523 btrfs_node_key(c
, &disk_key
, 0);
1524 split
= __btrfs_alloc_free_block(trans
, root
, root
->nodesize
,
1525 root
->root_key
.objectid
,
1527 btrfs_disk_key_objectid(&disk_key
),
1528 level
, c
->start
, 0);
1530 return PTR_ERR(split
);
1532 btrfs_set_header_flags(split
, btrfs_header_flags(c
));
1533 btrfs_set_header_level(split
, btrfs_header_level(c
));
1534 btrfs_set_header_bytenr(split
, split
->start
);
1535 btrfs_set_header_generation(split
, trans
->transid
);
1536 btrfs_set_header_owner(split
, root
->root_key
.objectid
);
1537 btrfs_set_header_flags(split
, 0);
1538 write_extent_buffer(split
, root
->fs_info
->fsid
,
1539 (unsigned long)btrfs_header_fsid(split
),
1541 write_extent_buffer(split
, root
->fs_info
->chunk_tree_uuid
,
1542 (unsigned long)btrfs_header_chunk_tree_uuid(split
),
1545 mid
= (c_nritems
+ 1) / 2;
1547 copy_extent_buffer(split
, c
,
1548 btrfs_node_key_ptr_offset(0),
1549 btrfs_node_key_ptr_offset(mid
),
1550 (c_nritems
- mid
) * sizeof(struct btrfs_key_ptr
));
1551 btrfs_set_header_nritems(split
, c_nritems
- mid
);
1552 btrfs_set_header_nritems(c
, mid
);
1555 btrfs_mark_buffer_dirty(c
);
1556 btrfs_mark_buffer_dirty(split
);
1558 btrfs_node_key(split
, &disk_key
, 0);
1559 wret
= insert_ptr(trans
, root
, path
, &disk_key
, split
->start
,
1560 path
->slots
[level
+ 1] + 1,
1565 if (path
->slots
[level
] >= mid
) {
1566 path
->slots
[level
] -= mid
;
1567 free_extent_buffer(c
);
1568 path
->nodes
[level
] = split
;
1569 path
->slots
[level
+ 1] += 1;
1571 free_extent_buffer(split
);
1577 * how many bytes are required to store the items in a leaf. start
1578 * and nr indicate which items in the leaf to check. This totals up the
1579 * space used both by the item structs and the item data
1581 static int leaf_space_used(struct extent_buffer
*l
, int start
, int nr
)
1584 int nritems
= btrfs_header_nritems(l
);
1585 int end
= min(nritems
, start
+ nr
) - 1;
1589 data_len
= btrfs_item_end_nr(l
, start
);
1590 data_len
= data_len
- btrfs_item_offset_nr(l
, end
);
1591 data_len
+= sizeof(struct btrfs_item
) * nr
;
1592 WARN_ON(data_len
< 0);
1597 * The space between the end of the leaf items and
1598 * the start of the leaf data. IOW, how much room
1599 * the leaf has left for both items and data
1601 int btrfs_leaf_free_space(struct btrfs_root
*root
, struct extent_buffer
*leaf
)
1603 int nritems
= btrfs_header_nritems(leaf
);
1605 ret
= BTRFS_LEAF_DATA_SIZE(root
) - leaf_space_used(leaf
, 0, nritems
);
1607 printk("leaf free space ret %d, leaf data size %lu, used %d nritems %d\n",
1608 ret
, (unsigned long) BTRFS_LEAF_DATA_SIZE(root
),
1609 leaf_space_used(leaf
, 0, nritems
), nritems
);
1615 * push some data in the path leaf to the right, trying to free up at
1616 * least data_size bytes. returns zero if the push worked, nonzero otherwise
1618 * returns 1 if the push failed because the other node didn't have enough
1619 * room, 0 if everything worked out and < 0 if there were major errors.
1621 static int push_leaf_right(struct btrfs_trans_handle
*trans
, struct btrfs_root
1622 *root
, struct btrfs_path
*path
, int data_size
,
1625 struct extent_buffer
*left
= path
->nodes
[0];
1626 struct extent_buffer
*right
;
1627 struct extent_buffer
*upper
;
1628 struct btrfs_disk_key disk_key
;
1634 struct btrfs_item
*item
;
1642 slot
= path
->slots
[1];
1643 if (!path
->nodes
[1]) {
1646 upper
= path
->nodes
[1];
1647 if (slot
>= btrfs_header_nritems(upper
) - 1)
1650 right
= read_node_slot(root
, upper
, slot
+ 1);
1651 free_space
= btrfs_leaf_free_space(root
, right
);
1652 if (free_space
< data_size
+ sizeof(struct btrfs_item
)) {
1653 free_extent_buffer(right
);
1657 /* cow and double check */
1658 ret
= btrfs_cow_block(trans
, root
, right
, upper
,
1661 free_extent_buffer(right
);
1664 free_space
= btrfs_leaf_free_space(root
, right
);
1665 if (free_space
< data_size
+ sizeof(struct btrfs_item
)) {
1666 free_extent_buffer(right
);
1670 left_nritems
= btrfs_header_nritems(left
);
1671 if (left_nritems
== 0) {
1672 free_extent_buffer(right
);
1681 i
= left_nritems
- 1;
1683 item
= btrfs_item_nr(left
, i
);
1685 if (path
->slots
[0] == i
)
1686 push_space
+= data_size
+ sizeof(*item
);
1688 if (!left
->map_token
) {
1689 map_extent_buffer(left
, (unsigned long)item
,
1690 sizeof(struct btrfs_item
),
1691 &left
->map_token
, &left
->kaddr
,
1692 &left
->map_start
, &left
->map_len
,
1696 this_item_size
= btrfs_item_size(left
, item
);
1697 if (this_item_size
+ sizeof(*item
) + push_space
> free_space
)
1700 push_space
+= this_item_size
+ sizeof(*item
);
1705 if (left
->map_token
) {
1706 unmap_extent_buffer(left
, left
->map_token
, KM_USER1
);
1707 left
->map_token
= NULL
;
1710 if (push_items
== 0) {
1711 free_extent_buffer(right
);
1715 if (!empty
&& push_items
== left_nritems
)
1718 /* push left to right */
1719 right_nritems
= btrfs_header_nritems(right
);
1721 push_space
= btrfs_item_end_nr(left
, left_nritems
- push_items
);
1722 push_space
-= leaf_data_end(root
, left
);
1724 /* make room in the right data area */
1725 data_end
= leaf_data_end(root
, right
);
1726 memmove_extent_buffer(right
,
1727 btrfs_leaf_data(right
) + data_end
- push_space
,
1728 btrfs_leaf_data(right
) + data_end
,
1729 BTRFS_LEAF_DATA_SIZE(root
) - data_end
);
1731 /* copy from the left data area */
1732 copy_extent_buffer(right
, left
, btrfs_leaf_data(right
) +
1733 BTRFS_LEAF_DATA_SIZE(root
) - push_space
,
1734 btrfs_leaf_data(left
) + leaf_data_end(root
, left
),
1737 memmove_extent_buffer(right
, btrfs_item_nr_offset(push_items
),
1738 btrfs_item_nr_offset(0),
1739 right_nritems
* sizeof(struct btrfs_item
));
1741 /* copy the items from left to right */
1742 copy_extent_buffer(right
, left
, btrfs_item_nr_offset(0),
1743 btrfs_item_nr_offset(left_nritems
- push_items
),
1744 push_items
* sizeof(struct btrfs_item
));
1746 /* update the item pointers */
1747 right_nritems
+= push_items
;
1748 btrfs_set_header_nritems(right
, right_nritems
);
1749 push_space
= BTRFS_LEAF_DATA_SIZE(root
);
1750 for (i
= 0; i
< right_nritems
; i
++) {
1751 item
= btrfs_item_nr(right
, i
);
1752 if (!right
->map_token
) {
1753 map_extent_buffer(right
, (unsigned long)item
,
1754 sizeof(struct btrfs_item
),
1755 &right
->map_token
, &right
->kaddr
,
1756 &right
->map_start
, &right
->map_len
,
1759 push_space
-= btrfs_item_size(right
, item
);
1760 btrfs_set_item_offset(right
, item
, push_space
);
1763 if (right
->map_token
) {
1764 unmap_extent_buffer(right
, right
->map_token
, KM_USER1
);
1765 right
->map_token
= NULL
;
1767 left_nritems
-= push_items
;
1768 btrfs_set_header_nritems(left
, left_nritems
);
1771 btrfs_mark_buffer_dirty(left
);
1772 btrfs_mark_buffer_dirty(right
);
1774 btrfs_item_key(right
, &disk_key
, 0);
1775 btrfs_set_node_key(upper
, &disk_key
, slot
+ 1);
1776 btrfs_mark_buffer_dirty(upper
);
1778 /* then fixup the leaf pointer in the path */
1779 if (path
->slots
[0] >= left_nritems
) {
1780 path
->slots
[0] -= left_nritems
;
1781 free_extent_buffer(path
->nodes
[0]);
1782 path
->nodes
[0] = right
;
1783 path
->slots
[1] += 1;
1785 free_extent_buffer(right
);
1790 * push some data in the path leaf to the left, trying to free up at
1791 * least data_size bytes. returns zero if the push worked, nonzero otherwise
1793 static int push_leaf_left(struct btrfs_trans_handle
*trans
, struct btrfs_root
1794 *root
, struct btrfs_path
*path
, int data_size
,
1797 struct btrfs_disk_key disk_key
;
1798 struct extent_buffer
*right
= path
->nodes
[0];
1799 struct extent_buffer
*left
;
1805 struct btrfs_item
*item
;
1806 u32 old_left_nritems
;
1812 u32 old_left_item_size
;
1814 slot
= path
->slots
[1];
1817 if (!path
->nodes
[1])
1820 right_nritems
= btrfs_header_nritems(right
);
1821 if (right_nritems
== 0) {
1825 left
= read_node_slot(root
, path
->nodes
[1], slot
- 1);
1826 free_space
= btrfs_leaf_free_space(root
, left
);
1827 if (free_space
< data_size
+ sizeof(struct btrfs_item
)) {
1828 free_extent_buffer(left
);
1832 /* cow and double check */
1833 ret
= btrfs_cow_block(trans
, root
, left
,
1834 path
->nodes
[1], slot
- 1, &left
);
1836 /* we hit -ENOSPC, but it isn't fatal here */
1837 free_extent_buffer(left
);
1841 free_space
= btrfs_leaf_free_space(root
, left
);
1842 if (free_space
< data_size
+ sizeof(struct btrfs_item
)) {
1843 free_extent_buffer(left
);
1850 nr
= right_nritems
- 1;
1852 for (i
= 0; i
< nr
; i
++) {
1853 item
= btrfs_item_nr(right
, i
);
1854 if (!right
->map_token
) {
1855 map_extent_buffer(right
, (unsigned long)item
,
1856 sizeof(struct btrfs_item
),
1857 &right
->map_token
, &right
->kaddr
,
1858 &right
->map_start
, &right
->map_len
,
1862 if (path
->slots
[0] == i
)
1863 push_space
+= data_size
+ sizeof(*item
);
1865 this_item_size
= btrfs_item_size(right
, item
);
1866 if (this_item_size
+ sizeof(*item
) + push_space
> free_space
)
1870 push_space
+= this_item_size
+ sizeof(*item
);
1873 if (right
->map_token
) {
1874 unmap_extent_buffer(right
, right
->map_token
, KM_USER1
);
1875 right
->map_token
= NULL
;
1878 if (push_items
== 0) {
1879 free_extent_buffer(left
);
1882 if (!empty
&& push_items
== btrfs_header_nritems(right
))
1885 /* push data from right to left */
1886 copy_extent_buffer(left
, right
,
1887 btrfs_item_nr_offset(btrfs_header_nritems(left
)),
1888 btrfs_item_nr_offset(0),
1889 push_items
* sizeof(struct btrfs_item
));
1891 push_space
= BTRFS_LEAF_DATA_SIZE(root
) -
1892 btrfs_item_offset_nr(right
, push_items
-1);
1894 copy_extent_buffer(left
, right
, btrfs_leaf_data(left
) +
1895 leaf_data_end(root
, left
) - push_space
,
1896 btrfs_leaf_data(right
) +
1897 btrfs_item_offset_nr(right
, push_items
- 1),
1899 old_left_nritems
= btrfs_header_nritems(left
);
1900 BUG_ON(old_left_nritems
< 0);
1902 old_left_item_size
= btrfs_item_offset_nr(left
, old_left_nritems
- 1);
1903 for (i
= old_left_nritems
; i
< old_left_nritems
+ push_items
; i
++) {
1906 item
= btrfs_item_nr(left
, i
);
1907 if (!left
->map_token
) {
1908 map_extent_buffer(left
, (unsigned long)item
,
1909 sizeof(struct btrfs_item
),
1910 &left
->map_token
, &left
->kaddr
,
1911 &left
->map_start
, &left
->map_len
,
1915 ioff
= btrfs_item_offset(left
, item
);
1916 btrfs_set_item_offset(left
, item
,
1917 ioff
- (BTRFS_LEAF_DATA_SIZE(root
) - old_left_item_size
));
1919 btrfs_set_header_nritems(left
, old_left_nritems
+ push_items
);
1920 if (left
->map_token
) {
1921 unmap_extent_buffer(left
, left
->map_token
, KM_USER1
);
1922 left
->map_token
= NULL
;
1925 /* fixup right node */
1926 if (push_items
> right_nritems
) {
1927 printk("push items %d nr %u\n", push_items
, right_nritems
);
1931 if (push_items
< right_nritems
) {
1932 push_space
= btrfs_item_offset_nr(right
, push_items
- 1) -
1933 leaf_data_end(root
, right
);
1934 memmove_extent_buffer(right
, btrfs_leaf_data(right
) +
1935 BTRFS_LEAF_DATA_SIZE(root
) - push_space
,
1936 btrfs_leaf_data(right
) +
1937 leaf_data_end(root
, right
), push_space
);
1939 memmove_extent_buffer(right
, btrfs_item_nr_offset(0),
1940 btrfs_item_nr_offset(push_items
),
1941 (btrfs_header_nritems(right
) - push_items
) *
1942 sizeof(struct btrfs_item
));
1944 right_nritems
-= push_items
;
1945 btrfs_set_header_nritems(right
, right_nritems
);
1946 push_space
= BTRFS_LEAF_DATA_SIZE(root
);
1947 for (i
= 0; i
< right_nritems
; i
++) {
1948 item
= btrfs_item_nr(right
, i
);
1950 if (!right
->map_token
) {
1951 map_extent_buffer(right
, (unsigned long)item
,
1952 sizeof(struct btrfs_item
),
1953 &right
->map_token
, &right
->kaddr
,
1954 &right
->map_start
, &right
->map_len
,
1958 push_space
= push_space
- btrfs_item_size(right
, item
);
1959 btrfs_set_item_offset(right
, item
, push_space
);
1961 if (right
->map_token
) {
1962 unmap_extent_buffer(right
, right
->map_token
, KM_USER1
);
1963 right
->map_token
= NULL
;
1966 btrfs_mark_buffer_dirty(left
);
1968 btrfs_mark_buffer_dirty(right
);
1970 btrfs_item_key(right
, &disk_key
, 0);
1971 wret
= fixup_low_keys(trans
, root
, path
, &disk_key
, 1);
1975 /* then fixup the leaf pointer in the path */
1976 if (path
->slots
[0] < push_items
) {
1977 path
->slots
[0] += old_left_nritems
;
1978 free_extent_buffer(path
->nodes
[0]);
1979 path
->nodes
[0] = left
;
1980 path
->slots
[1] -= 1;
1982 free_extent_buffer(left
);
1983 path
->slots
[0] -= push_items
;
1985 BUG_ON(path
->slots
[0] < 0);
1990 * split the path's leaf in two, making sure there is at least data_size
1991 * available for the resulting leaf level of the path.
1993 * returns 0 if all went well and < 0 on failure.
1995 static int split_leaf(struct btrfs_trans_handle
*trans
, struct btrfs_root
1996 *root
, struct btrfs_key
*ins_key
,
1997 struct btrfs_path
*path
, int data_size
, int extend
)
2000 struct extent_buffer
*l
;
2004 struct extent_buffer
*right
;
2005 int space_needed
= data_size
+ sizeof(struct btrfs_item
);
2012 int num_doubles
= 0;
2013 struct btrfs_disk_key disk_key
;
2016 space_needed
= data_size
;
2019 root_gen
= trans
->transid
;
2023 /* first try to make some room by pushing left and right */
2024 if (ins_key
->type
!= BTRFS_DIR_ITEM_KEY
) {
2025 wret
= push_leaf_right(trans
, root
, path
, data_size
, 0);
2030 wret
= push_leaf_left(trans
, root
, path
, data_size
, 0);
2036 /* did the pushes work? */
2037 if (btrfs_leaf_free_space(root
, l
) >= space_needed
)
2041 if (!path
->nodes
[1]) {
2042 ret
= insert_new_root(trans
, root
, path
, 1);
2049 slot
= path
->slots
[0];
2050 nritems
= btrfs_header_nritems(l
);
2051 mid
= (nritems
+ 1)/ 2;
2053 btrfs_item_key(l
, &disk_key
, 0);
2055 right
= __btrfs_alloc_free_block(trans
, root
, root
->leafsize
,
2056 root
->root_key
.objectid
,
2057 root_gen
, disk_key
.objectid
, 0,
2059 if (IS_ERR(right
)) {
2061 return PTR_ERR(right
);
2064 memset_extent_buffer(right
, 0, 0, sizeof(struct btrfs_header
));
2065 btrfs_set_header_bytenr(right
, right
->start
);
2066 btrfs_set_header_generation(right
, trans
->transid
);
2067 btrfs_set_header_owner(right
, root
->root_key
.objectid
);
2068 btrfs_set_header_level(right
, 0);
2069 write_extent_buffer(right
, root
->fs_info
->fsid
,
2070 (unsigned long)btrfs_header_fsid(right
),
2073 write_extent_buffer(right
, root
->fs_info
->chunk_tree_uuid
,
2074 (unsigned long)btrfs_header_chunk_tree_uuid(right
),
2078 leaf_space_used(l
, mid
, nritems
- mid
) + space_needed
>
2079 BTRFS_LEAF_DATA_SIZE(root
)) {
2080 if (slot
>= nritems
) {
2081 btrfs_cpu_key_to_disk(&disk_key
, ins_key
);
2082 btrfs_set_header_nritems(right
, 0);
2083 wret
= insert_ptr(trans
, root
, path
,
2084 &disk_key
, right
->start
,
2085 path
->slots
[1] + 1, 1);
2088 free_extent_buffer(path
->nodes
[0]);
2089 path
->nodes
[0] = right
;
2091 path
->slots
[1] += 1;
2092 btrfs_mark_buffer_dirty(right
);
2096 if (mid
!= nritems
&&
2097 leaf_space_used(l
, mid
, nritems
- mid
) +
2098 space_needed
> BTRFS_LEAF_DATA_SIZE(root
)) {
2103 if (leaf_space_used(l
, 0, mid
+ 1) + space_needed
>
2104 BTRFS_LEAF_DATA_SIZE(root
)) {
2105 if (!extend
&& slot
== 0) {
2106 btrfs_cpu_key_to_disk(&disk_key
, ins_key
);
2107 btrfs_set_header_nritems(right
, 0);
2108 wret
= insert_ptr(trans
, root
, path
,
2114 free_extent_buffer(path
->nodes
[0]);
2115 path
->nodes
[0] = right
;
2117 if (path
->slots
[1] == 0) {
2118 wret
= fixup_low_keys(trans
, root
,
2119 path
, &disk_key
, 1);
2123 btrfs_mark_buffer_dirty(right
);
2125 } else if (extend
&& slot
== 0) {
2129 if (mid
!= nritems
&&
2130 leaf_space_used(l
, mid
, nritems
- mid
) +
2131 space_needed
> BTRFS_LEAF_DATA_SIZE(root
)) {
2137 nritems
= nritems
- mid
;
2138 btrfs_set_header_nritems(right
, nritems
);
2139 data_copy_size
= btrfs_item_end_nr(l
, mid
) - leaf_data_end(root
, l
);
2141 copy_extent_buffer(right
, l
, btrfs_item_nr_offset(0),
2142 btrfs_item_nr_offset(mid
),
2143 nritems
* sizeof(struct btrfs_item
));
2145 copy_extent_buffer(right
, l
,
2146 btrfs_leaf_data(right
) + BTRFS_LEAF_DATA_SIZE(root
) -
2147 data_copy_size
, btrfs_leaf_data(l
) +
2148 leaf_data_end(root
, l
), data_copy_size
);
2150 rt_data_off
= BTRFS_LEAF_DATA_SIZE(root
) -
2151 btrfs_item_end_nr(l
, mid
);
2153 for (i
= 0; i
< nritems
; i
++) {
2154 struct btrfs_item
*item
= btrfs_item_nr(right
, i
);
2157 if (!right
->map_token
) {
2158 map_extent_buffer(right
, (unsigned long)item
,
2159 sizeof(struct btrfs_item
),
2160 &right
->map_token
, &right
->kaddr
,
2161 &right
->map_start
, &right
->map_len
,
2165 ioff
= btrfs_item_offset(right
, item
);
2166 btrfs_set_item_offset(right
, item
, ioff
+ rt_data_off
);
2169 if (right
->map_token
) {
2170 unmap_extent_buffer(right
, right
->map_token
, KM_USER1
);
2171 right
->map_token
= NULL
;
2174 btrfs_set_header_nritems(l
, mid
);
2176 btrfs_item_key(right
, &disk_key
, 0);
2177 wret
= insert_ptr(trans
, root
, path
, &disk_key
, right
->start
,
2178 path
->slots
[1] + 1, 1);
2182 btrfs_mark_buffer_dirty(right
);
2183 btrfs_mark_buffer_dirty(l
);
2184 BUG_ON(path
->slots
[0] != slot
);
2187 free_extent_buffer(path
->nodes
[0]);
2188 path
->nodes
[0] = right
;
2189 path
->slots
[0] -= mid
;
2190 path
->slots
[1] += 1;
2192 free_extent_buffer(right
);
2194 BUG_ON(path
->slots
[0] < 0);
2197 BUG_ON(num_doubles
!= 0);
2204 int btrfs_truncate_item(struct btrfs_trans_handle
*trans
,
2205 struct btrfs_root
*root
,
2206 struct btrfs_path
*path
,
2207 u32 new_size
, int from_end
)
2212 struct extent_buffer
*leaf
;
2213 struct btrfs_item
*item
;
2215 unsigned int data_end
;
2216 unsigned int old_data_start
;
2217 unsigned int old_size
;
2218 unsigned int size_diff
;
2221 slot_orig
= path
->slots
[0];
2222 leaf
= path
->nodes
[0];
2223 slot
= path
->slots
[0];
2225 old_size
= btrfs_item_size_nr(leaf
, slot
);
2226 if (old_size
== new_size
)
2229 nritems
= btrfs_header_nritems(leaf
);
2230 data_end
= leaf_data_end(root
, leaf
);
2232 old_data_start
= btrfs_item_offset_nr(leaf
, slot
);
2234 size_diff
= old_size
- new_size
;
2237 BUG_ON(slot
>= nritems
);
2240 * item0..itemN ... dataN.offset..dataN.size .. data0.size
2242 /* first correct the data pointers */
2243 for (i
= slot
; i
< nritems
; i
++) {
2245 item
= btrfs_item_nr(leaf
, i
);
2247 if (!leaf
->map_token
) {
2248 map_extent_buffer(leaf
, (unsigned long)item
,
2249 sizeof(struct btrfs_item
),
2250 &leaf
->map_token
, &leaf
->kaddr
,
2251 &leaf
->map_start
, &leaf
->map_len
,
2255 ioff
= btrfs_item_offset(leaf
, item
);
2256 btrfs_set_item_offset(leaf
, item
, ioff
+ size_diff
);
2259 if (leaf
->map_token
) {
2260 unmap_extent_buffer(leaf
, leaf
->map_token
, KM_USER1
);
2261 leaf
->map_token
= NULL
;
2264 /* shift the data */
2266 memmove_extent_buffer(leaf
, btrfs_leaf_data(leaf
) +
2267 data_end
+ size_diff
, btrfs_leaf_data(leaf
) +
2268 data_end
, old_data_start
+ new_size
- data_end
);
2270 struct btrfs_disk_key disk_key
;
2273 btrfs_item_key(leaf
, &disk_key
, slot
);
2275 if (btrfs_disk_key_type(&disk_key
) == BTRFS_EXTENT_DATA_KEY
) {
2277 struct btrfs_file_extent_item
*fi
;
2279 fi
= btrfs_item_ptr(leaf
, slot
,
2280 struct btrfs_file_extent_item
);
2281 fi
= (struct btrfs_file_extent_item
*)(
2282 (unsigned long)fi
- size_diff
);
2284 if (btrfs_file_extent_type(leaf
, fi
) ==
2285 BTRFS_FILE_EXTENT_INLINE
) {
2286 ptr
= btrfs_item_ptr_offset(leaf
, slot
);
2287 memmove_extent_buffer(leaf
, ptr
,
2289 offsetof(struct btrfs_file_extent_item
,
2294 memmove_extent_buffer(leaf
, btrfs_leaf_data(leaf
) +
2295 data_end
+ size_diff
, btrfs_leaf_data(leaf
) +
2296 data_end
, old_data_start
- data_end
);
2298 offset
= btrfs_disk_key_offset(&disk_key
);
2299 btrfs_set_disk_key_offset(&disk_key
, offset
+ size_diff
);
2300 btrfs_set_item_key(leaf
, &disk_key
, slot
);
2302 fixup_low_keys(trans
, root
, path
, &disk_key
, 1);
2305 item
= btrfs_item_nr(leaf
, slot
);
2306 btrfs_set_item_size(leaf
, item
, new_size
);
2307 btrfs_mark_buffer_dirty(leaf
);
2310 if (btrfs_leaf_free_space(root
, leaf
) < 0) {
2311 btrfs_print_leaf(root
, leaf
);
2317 int btrfs_extend_item(struct btrfs_trans_handle
*trans
,
2318 struct btrfs_root
*root
, struct btrfs_path
*path
,
2324 struct extent_buffer
*leaf
;
2325 struct btrfs_item
*item
;
2327 unsigned int data_end
;
2328 unsigned int old_data
;
2329 unsigned int old_size
;
2332 slot_orig
= path
->slots
[0];
2333 leaf
= path
->nodes
[0];
2335 nritems
= btrfs_header_nritems(leaf
);
2336 data_end
= leaf_data_end(root
, leaf
);
2338 if (btrfs_leaf_free_space(root
, leaf
) < data_size
) {
2339 btrfs_print_leaf(root
, leaf
);
2342 slot
= path
->slots
[0];
2343 old_data
= btrfs_item_end_nr(leaf
, slot
);
2346 if (slot
>= nritems
) {
2347 btrfs_print_leaf(root
, leaf
);
2348 printk("slot %d too large, nritems %d\n", slot
, nritems
);
2353 * item0..itemN ... dataN.offset..dataN.size .. data0.size
2355 /* first correct the data pointers */
2356 for (i
= slot
; i
< nritems
; i
++) {
2358 item
= btrfs_item_nr(leaf
, i
);
2360 if (!leaf
->map_token
) {
2361 map_extent_buffer(leaf
, (unsigned long)item
,
2362 sizeof(struct btrfs_item
),
2363 &leaf
->map_token
, &leaf
->kaddr
,
2364 &leaf
->map_start
, &leaf
->map_len
,
2367 ioff
= btrfs_item_offset(leaf
, item
);
2368 btrfs_set_item_offset(leaf
, item
, ioff
- data_size
);
2371 if (leaf
->map_token
) {
2372 unmap_extent_buffer(leaf
, leaf
->map_token
, KM_USER1
);
2373 leaf
->map_token
= NULL
;
2376 /* shift the data */
2377 memmove_extent_buffer(leaf
, btrfs_leaf_data(leaf
) +
2378 data_end
- data_size
, btrfs_leaf_data(leaf
) +
2379 data_end
, old_data
- data_end
);
2381 data_end
= old_data
;
2382 old_size
= btrfs_item_size_nr(leaf
, slot
);
2383 item
= btrfs_item_nr(leaf
, slot
);
2384 btrfs_set_item_size(leaf
, item
, old_size
+ data_size
);
2385 btrfs_mark_buffer_dirty(leaf
);
2388 if (btrfs_leaf_free_space(root
, leaf
) < 0) {
2389 btrfs_print_leaf(root
, leaf
);
2396 * Given a key and some data, insert an item into the tree.
2397 * This does all the path init required, making room in the tree if needed.
2399 int btrfs_insert_empty_items(struct btrfs_trans_handle
*trans
,
2400 struct btrfs_root
*root
,
2401 struct btrfs_path
*path
,
2402 struct btrfs_key
*cpu_key
, u32
*data_size
,
2405 struct extent_buffer
*leaf
;
2406 struct btrfs_item
*item
;
2414 unsigned int data_end
;
2415 struct btrfs_disk_key disk_key
;
2417 for (i
= 0; i
< nr
; i
++) {
2418 total_data
+= data_size
[i
];
2421 /* create a root if there isn't one */
2425 total_size
= total_data
+ (nr
- 1) * sizeof(struct btrfs_item
);
2426 ret
= btrfs_search_slot(trans
, root
, cpu_key
, path
, total_size
, 1);
2433 slot_orig
= path
->slots
[0];
2434 leaf
= path
->nodes
[0];
2436 nritems
= btrfs_header_nritems(leaf
);
2437 data_end
= leaf_data_end(root
, leaf
);
2439 if (btrfs_leaf_free_space(root
, leaf
) <
2440 sizeof(struct btrfs_item
) + total_size
) {
2441 btrfs_print_leaf(root
, leaf
);
2442 printk("not enough freespace need %u have %d\n",
2443 total_size
, btrfs_leaf_free_space(root
, leaf
));
2447 slot
= path
->slots
[0];
2450 if (slot
!= nritems
) {
2452 unsigned int old_data
= btrfs_item_end_nr(leaf
, slot
);
2454 if (old_data
< data_end
) {
2455 btrfs_print_leaf(root
, leaf
);
2456 printk("slot %d old_data %d data_end %d\n",
2457 slot
, old_data
, data_end
);
2461 * item0..itemN ... dataN.offset..dataN.size .. data0.size
2463 /* first correct the data pointers */
2464 WARN_ON(leaf
->map_token
);
2465 for (i
= slot
; i
< nritems
; i
++) {
2468 item
= btrfs_item_nr(leaf
, i
);
2469 if (!leaf
->map_token
) {
2470 map_extent_buffer(leaf
, (unsigned long)item
,
2471 sizeof(struct btrfs_item
),
2472 &leaf
->map_token
, &leaf
->kaddr
,
2473 &leaf
->map_start
, &leaf
->map_len
,
2477 ioff
= btrfs_item_offset(leaf
, item
);
2478 btrfs_set_item_offset(leaf
, item
, ioff
- total_data
);
2480 if (leaf
->map_token
) {
2481 unmap_extent_buffer(leaf
, leaf
->map_token
, KM_USER1
);
2482 leaf
->map_token
= NULL
;
2485 /* shift the items */
2486 memmove_extent_buffer(leaf
, btrfs_item_nr_offset(slot
+ nr
),
2487 btrfs_item_nr_offset(slot
),
2488 (nritems
- slot
) * sizeof(struct btrfs_item
));
2490 /* shift the data */
2491 memmove_extent_buffer(leaf
, btrfs_leaf_data(leaf
) +
2492 data_end
- total_data
, btrfs_leaf_data(leaf
) +
2493 data_end
, old_data
- data_end
);
2494 data_end
= old_data
;
2497 /* setup the item for the new data */
2498 for (i
= 0; i
< nr
; i
++) {
2499 btrfs_cpu_key_to_disk(&disk_key
, cpu_key
+ i
);
2500 btrfs_set_item_key(leaf
, &disk_key
, slot
+ i
);
2501 item
= btrfs_item_nr(leaf
, slot
+ i
);
2502 btrfs_set_item_offset(leaf
, item
, data_end
- data_size
[i
]);
2503 data_end
-= data_size
[i
];
2504 btrfs_set_item_size(leaf
, item
, data_size
[i
]);
2506 btrfs_set_header_nritems(leaf
, nritems
+ nr
);
2507 btrfs_mark_buffer_dirty(leaf
);
2511 btrfs_cpu_key_to_disk(&disk_key
, cpu_key
);
2512 ret
= fixup_low_keys(trans
, root
, path
, &disk_key
, 1);
2515 if (btrfs_leaf_free_space(root
, leaf
) < 0) {
2516 btrfs_print_leaf(root
, leaf
);
2525 * Given a key and some data, insert an item into the tree.
2526 * This does all the path init required, making room in the tree if needed.
2528 int btrfs_insert_item(struct btrfs_trans_handle
*trans
, struct btrfs_root
2529 *root
, struct btrfs_key
*cpu_key
, void *data
, u32
2533 struct btrfs_path
*path
;
2534 struct extent_buffer
*leaf
;
2537 path
= btrfs_alloc_path();
2539 ret
= btrfs_insert_empty_item(trans
, root
, path
, cpu_key
, data_size
);
2541 leaf
= path
->nodes
[0];
2542 ptr
= btrfs_item_ptr_offset(leaf
, path
->slots
[0]);
2543 write_extent_buffer(leaf
, data
, ptr
, data_size
);
2544 btrfs_mark_buffer_dirty(leaf
);
2546 btrfs_free_path(path
);
2551 * delete the pointer from a given node.
2553 * If the delete empties a node, the node is removed from the tree,
2554 * continuing all the way the root if required. The root is converted into
2555 * a leaf if all the nodes are emptied.
2557 static int del_ptr(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
2558 struct btrfs_path
*path
, int level
, int slot
)
2560 struct extent_buffer
*parent
= path
->nodes
[level
];
2565 nritems
= btrfs_header_nritems(parent
);
2566 if (slot
!= nritems
-1) {
2567 memmove_extent_buffer(parent
,
2568 btrfs_node_key_ptr_offset(slot
),
2569 btrfs_node_key_ptr_offset(slot
+ 1),
2570 sizeof(struct btrfs_key_ptr
) *
2571 (nritems
- slot
- 1));
2574 btrfs_set_header_nritems(parent
, nritems
);
2575 if (nritems
== 0 && parent
== root
->node
) {
2576 BUG_ON(btrfs_header_level(root
->node
) != 1);
2577 /* just turn the root into a leaf and break */
2578 btrfs_set_header_level(root
->node
, 0);
2579 } else if (slot
== 0) {
2580 struct btrfs_disk_key disk_key
;
2582 btrfs_node_key(parent
, &disk_key
, 0);
2583 wret
= fixup_low_keys(trans
, root
, path
, &disk_key
, level
+ 1);
2587 btrfs_mark_buffer_dirty(parent
);
2592 * delete the item at the leaf level in path. If that empties
2593 * the leaf, remove it from the tree
2595 int btrfs_del_items(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
2596 struct btrfs_path
*path
, int slot
, int nr
)
2598 struct extent_buffer
*leaf
;
2599 struct btrfs_item
*item
;
2607 leaf
= path
->nodes
[0];
2608 last_off
= btrfs_item_offset_nr(leaf
, slot
+ nr
- 1);
2610 for (i
= 0; i
< nr
; i
++)
2611 dsize
+= btrfs_item_size_nr(leaf
, slot
+ i
);
2613 nritems
= btrfs_header_nritems(leaf
);
2615 if (slot
+ nr
!= nritems
) {
2617 int data_end
= leaf_data_end(root
, leaf
);
2619 memmove_extent_buffer(leaf
, btrfs_leaf_data(leaf
) +
2621 btrfs_leaf_data(leaf
) + data_end
,
2622 last_off
- data_end
);
2624 for (i
= slot
+ nr
; i
< nritems
; i
++) {
2627 item
= btrfs_item_nr(leaf
, i
);
2628 if (!leaf
->map_token
) {
2629 map_extent_buffer(leaf
, (unsigned long)item
,
2630 sizeof(struct btrfs_item
),
2631 &leaf
->map_token
, &leaf
->kaddr
,
2632 &leaf
->map_start
, &leaf
->map_len
,
2635 ioff
= btrfs_item_offset(leaf
, item
);
2636 btrfs_set_item_offset(leaf
, item
, ioff
+ dsize
);
2639 if (leaf
->map_token
) {
2640 unmap_extent_buffer(leaf
, leaf
->map_token
, KM_USER1
);
2641 leaf
->map_token
= NULL
;
2644 memmove_extent_buffer(leaf
, btrfs_item_nr_offset(slot
),
2645 btrfs_item_nr_offset(slot
+ nr
),
2646 sizeof(struct btrfs_item
) *
2647 (nritems
- slot
- nr
));
2649 btrfs_set_header_nritems(leaf
, nritems
- nr
);
2652 /* delete the leaf if we've emptied it */
2654 if (leaf
== root
->node
) {
2655 btrfs_set_header_level(leaf
, 0);
2657 u64 root_gen
= btrfs_header_generation(path
->nodes
[1]);
2658 clean_tree_block(trans
, root
, leaf
);
2659 wret
= del_ptr(trans
, root
, path
, 1, path
->slots
[1]);
2662 wret
= btrfs_free_extent(trans
, root
,
2663 leaf
->start
, leaf
->len
,
2664 btrfs_header_owner(path
->nodes
[1]),
2670 int used
= leaf_space_used(leaf
, 0, nritems
);
2672 struct btrfs_disk_key disk_key
;
2674 btrfs_item_key(leaf
, &disk_key
, 0);
2675 wret
= fixup_low_keys(trans
, root
, path
,
2681 /* delete the leaf if it is mostly empty */
2682 if (used
< BTRFS_LEAF_DATA_SIZE(root
) / 4) {
2683 /* push_leaf_left fixes the path.
2684 * make sure the path still points to our leaf
2685 * for possible call to del_ptr below
2687 slot
= path
->slots
[1];
2688 extent_buffer_get(leaf
);
2690 wret
= push_leaf_left(trans
, root
, path
, 1, 1);
2691 if (wret
< 0 && wret
!= -ENOSPC
)
2694 if (path
->nodes
[0] == leaf
&&
2695 btrfs_header_nritems(leaf
)) {
2696 wret
= push_leaf_right(trans
, root
, path
, 1, 1);
2697 if (wret
< 0 && wret
!= -ENOSPC
)
2701 if (btrfs_header_nritems(leaf
) == 0) {
2703 u64 bytenr
= leaf
->start
;
2704 u32 blocksize
= leaf
->len
;
2706 root_gen
= btrfs_header_generation(
2709 clean_tree_block(trans
, root
, leaf
);
2711 wret
= del_ptr(trans
, root
, path
, 1, slot
);
2715 free_extent_buffer(leaf
);
2716 wret
= btrfs_free_extent(trans
, root
, bytenr
,
2718 btrfs_header_owner(path
->nodes
[1]),
2723 btrfs_mark_buffer_dirty(leaf
);
2724 free_extent_buffer(leaf
);
2727 btrfs_mark_buffer_dirty(leaf
);
2734 * walk up the tree as far as required to find the previous leaf.
2735 * returns 0 if it found something or 1 if there are no lesser leaves.
2736 * returns < 0 on io errors.
2738 int btrfs_prev_leaf(struct btrfs_root
*root
, struct btrfs_path
*path
)
2742 struct extent_buffer
*c
;
2743 struct extent_buffer
*next
= NULL
;
2745 while(level
< BTRFS_MAX_LEVEL
) {
2746 if (!path
->nodes
[level
])
2749 slot
= path
->slots
[level
];
2750 c
= path
->nodes
[level
];
2753 if (level
== BTRFS_MAX_LEVEL
)
2760 free_extent_buffer(next
);
2762 next
= read_node_slot(root
, c
, slot
);
2765 path
->slots
[level
] = slot
;
2768 c
= path
->nodes
[level
];
2769 free_extent_buffer(c
);
2770 slot
= btrfs_header_nritems(next
);
2773 path
->nodes
[level
] = next
;
2774 path
->slots
[level
] = slot
;
2777 next
= read_node_slot(root
, next
, slot
);
2783 * walk up the tree as far as required to find the next leaf.
2784 * returns 0 if it found something or 1 if there are no greater leaves.
2785 * returns < 0 on io errors.
2787 int btrfs_next_leaf(struct btrfs_root
*root
, struct btrfs_path
*path
)
2791 struct extent_buffer
*c
;
2792 struct extent_buffer
*next
= NULL
;
2794 while(level
< BTRFS_MAX_LEVEL
) {
2795 if (!path
->nodes
[level
])
2798 slot
= path
->slots
[level
] + 1;
2799 c
= path
->nodes
[level
];
2800 if (slot
>= btrfs_header_nritems(c
)) {
2802 if (level
== BTRFS_MAX_LEVEL
)
2808 free_extent_buffer(next
);
2811 reada_for_search(root
, path
, level
, slot
, 0);
2813 next
= read_node_slot(root
, c
, slot
);
2816 path
->slots
[level
] = slot
;
2819 c
= path
->nodes
[level
];
2820 free_extent_buffer(c
);
2821 path
->nodes
[level
] = next
;
2822 path
->slots
[level
] = 0;
2826 reada_for_search(root
, path
, level
, 0, 0);
2827 next
= read_node_slot(root
, next
, 0);
2832 int btrfs_previous_item(struct btrfs_root
*root
,
2833 struct btrfs_path
*path
, u64 min_objectid
,
2836 struct btrfs_key found_key
;
2837 struct extent_buffer
*leaf
;
2841 if (path
->slots
[0] == 0) {
2842 ret
= btrfs_prev_leaf(root
, path
);
2848 leaf
= path
->nodes
[0];
2849 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2850 if (found_key
.type
== type
)