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.
20 #include "transaction.h"
21 #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
);
38 inline void btrfs_init_path(struct btrfs_path
*p
)
40 memset(p
, 0, sizeof(*p
));
43 struct btrfs_path
*btrfs_alloc_path(void)
45 struct btrfs_path
*path
;
46 path
= kzalloc(sizeof(struct btrfs_path
), GFP_NOFS
);
50 void btrfs_free_path(struct btrfs_path
*p
)
54 btrfs_release_path(p
);
58 void btrfs_release_path(struct btrfs_path
*p
)
61 for (i
= 0; i
< BTRFS_MAX_LEVEL
; i
++) {
64 free_extent_buffer(p
->nodes
[i
]);
66 memset(p
, 0, sizeof(*p
));
69 void add_root_to_dirty_list(struct btrfs_root
*root
)
71 if (root
->track_dirty
&& list_empty(&root
->dirty_list
)) {
72 list_add(&root
->dirty_list
,
73 &root
->fs_info
->dirty_cowonly_roots
);
77 int btrfs_copy_root(struct btrfs_trans_handle
*trans
,
78 struct btrfs_root
*root
,
79 struct extent_buffer
*buf
,
80 struct extent_buffer
**cow_ret
, u64 new_root_objectid
)
82 struct extent_buffer
*cow
;
85 struct btrfs_root
*new_root
;
86 struct btrfs_disk_key disk_key
;
88 new_root
= kmalloc(sizeof(*new_root
), GFP_NOFS
);
92 memcpy(new_root
, root
, sizeof(*new_root
));
93 new_root
->root_key
.objectid
= new_root_objectid
;
95 WARN_ON(root
->ref_cows
&& trans
->transid
!=
96 root
->fs_info
->running_transaction
->transid
);
97 WARN_ON(root
->ref_cows
&& trans
->transid
!= root
->last_trans
);
99 level
= btrfs_header_level(buf
);
101 btrfs_item_key(buf
, &disk_key
, 0);
103 btrfs_node_key(buf
, &disk_key
, 0);
104 cow
= btrfs_alloc_free_block(trans
, new_root
, buf
->len
,
105 new_root_objectid
, &disk_key
,
106 level
, buf
->start
, 0);
112 copy_extent_buffer(cow
, buf
, 0, 0, cow
->len
);
113 btrfs_set_header_bytenr(cow
, cow
->start
);
114 btrfs_set_header_generation(cow
, trans
->transid
);
115 btrfs_set_header_backref_rev(cow
, BTRFS_MIXED_BACKREF_REV
);
116 btrfs_clear_header_flag(cow
, BTRFS_HEADER_FLAG_WRITTEN
|
117 BTRFS_HEADER_FLAG_RELOC
);
118 if (new_root_objectid
== BTRFS_TREE_RELOC_OBJECTID
)
119 btrfs_set_header_flag(cow
, BTRFS_HEADER_FLAG_RELOC
);
121 btrfs_set_header_owner(cow
, new_root_objectid
);
123 write_extent_buffer(cow
, root
->fs_info
->fsid
,
124 btrfs_header_fsid(), BTRFS_FSID_SIZE
);
126 WARN_ON(btrfs_header_generation(buf
) > trans
->transid
);
127 ret
= btrfs_inc_ref(trans
, new_root
, cow
, 0);
133 btrfs_mark_buffer_dirty(cow
);
139 * check if the tree block can be shared by multiple trees
141 static int btrfs_block_can_be_shared(struct btrfs_root
*root
,
142 struct extent_buffer
*buf
)
145 * Tree blocks not in refernece counted trees and tree roots
146 * are never shared. If a block was allocated after the last
147 * snapshot and the block was not allocated by tree relocation,
148 * we know the block is not shared.
150 if (root
->ref_cows
&&
151 buf
!= root
->node
&& buf
!= root
->commit_root
&&
152 (btrfs_header_generation(buf
) <=
153 btrfs_root_last_snapshot(&root
->root_item
) ||
154 btrfs_header_flag(buf
, BTRFS_HEADER_FLAG_RELOC
)))
156 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
157 if (root
->ref_cows
&&
158 btrfs_header_backref_rev(buf
) < BTRFS_MIXED_BACKREF_REV
)
164 static noinline
int update_ref_for_cow(struct btrfs_trans_handle
*trans
,
165 struct btrfs_root
*root
,
166 struct extent_buffer
*buf
,
167 struct extent_buffer
*cow
)
176 * Backrefs update rules:
178 * Always use full backrefs for extent pointers in tree block
179 * allocated by tree relocation.
181 * If a shared tree block is no longer referenced by its owner
182 * tree (btrfs_header_owner(buf) == root->root_key.objectid),
183 * use full backrefs for extent pointers in tree block.
185 * If a tree block is been relocating
186 * (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID),
187 * use full backrefs for extent pointers in tree block.
188 * The reason for this is some operations (such as drop tree)
189 * are only allowed for blocks use full backrefs.
192 if (btrfs_block_can_be_shared(root
, buf
)) {
193 ret
= btrfs_lookup_extent_info(trans
, root
, buf
->start
,
194 btrfs_header_level(buf
), 1,
200 if (root
->root_key
.objectid
== BTRFS_TREE_RELOC_OBJECTID
||
201 btrfs_header_backref_rev(buf
) < BTRFS_MIXED_BACKREF_REV
)
202 flags
= BTRFS_BLOCK_FLAG_FULL_BACKREF
;
207 owner
= btrfs_header_owner(buf
);
208 BUG_ON(!(flags
& BTRFS_BLOCK_FLAG_FULL_BACKREF
) &&
209 owner
== BTRFS_TREE_RELOC_OBJECTID
);
212 if ((owner
== root
->root_key
.objectid
||
213 root
->root_key
.objectid
== BTRFS_TREE_RELOC_OBJECTID
) &&
214 !(flags
& BTRFS_BLOCK_FLAG_FULL_BACKREF
)) {
215 ret
= btrfs_inc_ref(trans
, root
, buf
, 1);
218 if (root
->root_key
.objectid
==
219 BTRFS_TREE_RELOC_OBJECTID
) {
220 ret
= btrfs_dec_ref(trans
, root
, buf
, 0);
222 ret
= btrfs_inc_ref(trans
, root
, cow
, 1);
225 new_flags
|= BTRFS_BLOCK_FLAG_FULL_BACKREF
;
228 if (root
->root_key
.objectid
==
229 BTRFS_TREE_RELOC_OBJECTID
)
230 ret
= btrfs_inc_ref(trans
, root
, cow
, 1);
232 ret
= btrfs_inc_ref(trans
, root
, cow
, 0);
235 if (new_flags
!= 0) {
236 ret
= btrfs_set_block_flags(trans
, root
, buf
->start
,
237 btrfs_header_level(buf
),
242 if (flags
& BTRFS_BLOCK_FLAG_FULL_BACKREF
) {
243 if (root
->root_key
.objectid
==
244 BTRFS_TREE_RELOC_OBJECTID
)
245 ret
= btrfs_inc_ref(trans
, root
, cow
, 1);
247 ret
= btrfs_inc_ref(trans
, root
, cow
, 0);
249 ret
= btrfs_dec_ref(trans
, root
, buf
, 1);
252 clean_tree_block(trans
, root
, buf
);
257 int __btrfs_cow_block(struct btrfs_trans_handle
*trans
,
258 struct btrfs_root
*root
,
259 struct extent_buffer
*buf
,
260 struct extent_buffer
*parent
, int parent_slot
,
261 struct extent_buffer
**cow_ret
,
262 u64 search_start
, u64 empty_size
)
264 struct extent_buffer
*cow
;
265 struct btrfs_disk_key disk_key
;
268 WARN_ON(root
->ref_cows
&& trans
->transid
!=
269 root
->fs_info
->running_transaction
->transid
);
270 WARN_ON(root
->ref_cows
&& trans
->transid
!= root
->last_trans
);
272 level
= btrfs_header_level(buf
);
275 btrfs_item_key(buf
, &disk_key
, 0);
277 btrfs_node_key(buf
, &disk_key
, 0);
279 cow
= btrfs_alloc_free_block(trans
, root
, buf
->len
,
280 root
->root_key
.objectid
, &disk_key
,
281 level
, search_start
, empty_size
);
285 copy_extent_buffer(cow
, buf
, 0, 0, cow
->len
);
286 btrfs_set_header_bytenr(cow
, cow
->start
);
287 btrfs_set_header_generation(cow
, trans
->transid
);
288 btrfs_set_header_backref_rev(cow
, BTRFS_MIXED_BACKREF_REV
);
289 btrfs_clear_header_flag(cow
, BTRFS_HEADER_FLAG_WRITTEN
|
290 BTRFS_HEADER_FLAG_RELOC
);
291 if (root
->root_key
.objectid
== BTRFS_TREE_RELOC_OBJECTID
)
292 btrfs_set_header_flag(cow
, BTRFS_HEADER_FLAG_RELOC
);
294 btrfs_set_header_owner(cow
, root
->root_key
.objectid
);
296 write_extent_buffer(cow
, root
->fs_info
->fsid
,
297 btrfs_header_fsid(), BTRFS_FSID_SIZE
);
299 WARN_ON(!(buf
->flags
& EXTENT_BAD_TRANSID
) &&
300 btrfs_header_generation(buf
) > trans
->transid
);
302 update_ref_for_cow(trans
, root
, buf
, cow
);
304 if (buf
== root
->node
) {
306 extent_buffer_get(cow
);
308 btrfs_free_extent(trans
, root
, buf
->start
, buf
->len
,
309 0, root
->root_key
.objectid
, level
, 0);
310 free_extent_buffer(buf
);
311 add_root_to_dirty_list(root
);
313 btrfs_set_node_blockptr(parent
, parent_slot
,
315 WARN_ON(trans
->transid
== 0);
316 btrfs_set_node_ptr_generation(parent
, parent_slot
,
318 btrfs_mark_buffer_dirty(parent
);
319 WARN_ON(btrfs_header_generation(parent
) != trans
->transid
);
321 btrfs_free_extent(trans
, root
, buf
->start
, buf
->len
,
322 0, root
->root_key
.objectid
, level
, 1);
324 if (!list_empty(&buf
->recow
)) {
325 list_del_init(&buf
->recow
);
326 free_extent_buffer(buf
);
328 free_extent_buffer(buf
);
329 btrfs_mark_buffer_dirty(cow
);
334 static inline int should_cow_block(struct btrfs_trans_handle
*trans
,
335 struct btrfs_root
*root
,
336 struct extent_buffer
*buf
)
338 if (btrfs_header_generation(buf
) == trans
->transid
&&
339 !btrfs_header_flag(buf
, BTRFS_HEADER_FLAG_WRITTEN
) &&
340 !(root
->root_key
.objectid
!= BTRFS_TREE_RELOC_OBJECTID
&&
341 btrfs_header_flag(buf
, BTRFS_HEADER_FLAG_RELOC
)))
346 int btrfs_cow_block(struct btrfs_trans_handle
*trans
,
347 struct btrfs_root
*root
, struct extent_buffer
*buf
,
348 struct extent_buffer
*parent
, int parent_slot
,
349 struct extent_buffer
**cow_ret
)
354 if (trans->transaction != root->fs_info->running_transaction) {
355 printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid,
356 root->fs_info->running_transaction->transid);
360 if (trans
->transid
!= root
->fs_info
->generation
) {
361 printk(KERN_CRIT
"trans %llu running %llu\n",
362 (unsigned long long)trans
->transid
,
363 (unsigned long long)root
->fs_info
->generation
);
366 if (!should_cow_block(trans
, root
, buf
)) {
371 search_start
= buf
->start
& ~((u64
)(1024 * 1024 * 1024) - 1);
372 ret
= __btrfs_cow_block(trans
, root
, buf
, parent
,
373 parent_slot
, cow_ret
, search_start
, 0);
377 int btrfs_comp_cpu_keys(struct btrfs_key
*k1
, struct btrfs_key
*k2
)
379 if (k1
->objectid
> k2
->objectid
)
381 if (k1
->objectid
< k2
->objectid
)
383 if (k1
->type
> k2
->type
)
385 if (k1
->type
< k2
->type
)
387 if (k1
->offset
> k2
->offset
)
389 if (k1
->offset
< k2
->offset
)
395 * compare two keys in a memcmp fashion
397 static int btrfs_comp_keys(struct btrfs_disk_key
*disk
, struct btrfs_key
*k2
)
401 btrfs_disk_key_to_cpu(&k1
, disk
);
402 return btrfs_comp_cpu_keys(&k1
, k2
);
406 * The leaf data grows from end-to-front in the node.
407 * this returns the address of the start of the last item,
408 * which is the stop of the leaf data stack
410 static inline unsigned int leaf_data_end(struct btrfs_root
*root
,
411 struct extent_buffer
*leaf
)
413 u32 nr
= btrfs_header_nritems(leaf
);
415 return BTRFS_LEAF_DATA_SIZE(root
);
416 return btrfs_item_offset_nr(leaf
, nr
- 1);
419 enum btrfs_tree_block_status
420 btrfs_check_node(struct btrfs_root
*root
, struct btrfs_disk_key
*parent_key
,
421 struct extent_buffer
*buf
)
424 struct btrfs_key cpukey
;
425 struct btrfs_disk_key key
;
426 u32 nritems
= btrfs_header_nritems(buf
);
427 enum btrfs_tree_block_status ret
= BTRFS_TREE_BLOCK_INVALID_NRITEMS
;
429 if (nritems
== 0 || nritems
> BTRFS_NODEPTRS_PER_BLOCK(root
))
432 ret
= BTRFS_TREE_BLOCK_INVALID_PARENT_KEY
;
433 if (parent_key
&& parent_key
->type
) {
434 btrfs_node_key(buf
, &key
, 0);
435 if (memcmp(parent_key
, &key
, sizeof(key
)))
438 ret
= BTRFS_TREE_BLOCK_BAD_KEY_ORDER
;
439 for (i
= 0; nritems
> 1 && i
< nritems
- 2; i
++) {
440 btrfs_node_key(buf
, &key
, i
);
441 btrfs_node_key_to_cpu(buf
, &cpukey
, i
+ 1);
442 if (btrfs_comp_keys(&key
, &cpukey
) >= 0)
445 return BTRFS_TREE_BLOCK_CLEAN
;
447 if (btrfs_header_owner(buf
) == BTRFS_EXTENT_TREE_OBJECTID
) {
449 btrfs_disk_key_to_cpu(&cpukey
, parent_key
);
451 btrfs_node_key_to_cpu(buf
, &cpukey
, 0);
452 btrfs_add_corrupt_extent_record(root
->fs_info
, &cpukey
,
453 buf
->start
, buf
->len
,
454 btrfs_header_level(buf
));
459 enum btrfs_tree_block_status
460 btrfs_check_leaf(struct btrfs_root
*root
, struct btrfs_disk_key
*parent_key
,
461 struct extent_buffer
*buf
)
464 struct btrfs_key cpukey
;
465 struct btrfs_disk_key key
;
466 u32 nritems
= btrfs_header_nritems(buf
);
467 enum btrfs_tree_block_status ret
= BTRFS_TREE_BLOCK_INVALID_NRITEMS
;
469 if (nritems
* sizeof(struct btrfs_item
) > buf
->len
) {
470 fprintf(stderr
, "invalid number of items %llu\n",
471 (unsigned long long)buf
->start
);
475 if (btrfs_header_level(buf
) != 0) {
476 ret
= BTRFS_TREE_BLOCK_INVALID_LEVEL
;
477 fprintf(stderr
, "leaf is not a leaf %llu\n",
478 (unsigned long long)btrfs_header_bytenr(buf
));
481 if (btrfs_leaf_free_space(root
, buf
) < 0) {
482 ret
= BTRFS_TREE_BLOCK_INVALID_FREE_SPACE
;
483 fprintf(stderr
, "leaf free space incorrect %llu %d\n",
484 (unsigned long long)btrfs_header_bytenr(buf
),
485 btrfs_leaf_free_space(root
, buf
));
490 return BTRFS_TREE_BLOCK_CLEAN
;
492 btrfs_item_key(buf
, &key
, 0);
493 if (parent_key
&& parent_key
->type
&&
494 memcmp(parent_key
, &key
, sizeof(key
))) {
495 ret
= BTRFS_TREE_BLOCK_INVALID_PARENT_KEY
;
496 fprintf(stderr
, "leaf parent key incorrect %llu\n",
497 (unsigned long long)btrfs_header_bytenr(buf
));
500 for (i
= 0; nritems
> 1 && i
< nritems
- 1; i
++) {
501 btrfs_item_key(buf
, &key
, i
);
502 btrfs_item_key_to_cpu(buf
, &cpukey
, i
+ 1);
503 if (btrfs_comp_keys(&key
, &cpukey
) >= 0) {
504 ret
= BTRFS_TREE_BLOCK_BAD_KEY_ORDER
;
505 fprintf(stderr
, "bad key ordering %d %d\n", i
, i
+1);
508 if (btrfs_item_offset_nr(buf
, i
) !=
509 btrfs_item_end_nr(buf
, i
+ 1)) {
510 ret
= BTRFS_TREE_BLOCK_INVALID_OFFSETS
;
511 fprintf(stderr
, "incorrect offsets %u %u\n",
512 btrfs_item_offset_nr(buf
, i
),
513 btrfs_item_end_nr(buf
, i
+ 1));
516 if (i
== 0 && btrfs_item_end_nr(buf
, i
) !=
517 BTRFS_LEAF_DATA_SIZE(root
)) {
518 ret
= BTRFS_TREE_BLOCK_INVALID_OFFSETS
;
519 fprintf(stderr
, "bad item end %u wanted %u\n",
520 btrfs_item_end_nr(buf
, i
),
521 (unsigned)BTRFS_LEAF_DATA_SIZE(root
));
526 for (i
= 0; i
< nritems
; i
++) {
527 if (btrfs_item_end_nr(buf
, i
) > BTRFS_LEAF_DATA_SIZE(root
)) {
528 btrfs_item_key(buf
, &key
, 0);
529 btrfs_print_key(&key
);
531 ret
= BTRFS_TREE_BLOCK_INVALID_OFFSETS
;
532 fprintf(stderr
, "slot end outside of leaf %llu > %llu\n",
533 (unsigned long long)btrfs_item_end_nr(buf
, i
),
534 (unsigned long long)BTRFS_LEAF_DATA_SIZE(root
));
539 return BTRFS_TREE_BLOCK_CLEAN
;
541 if (btrfs_header_owner(buf
) == BTRFS_EXTENT_TREE_OBJECTID
) {
543 btrfs_disk_key_to_cpu(&cpukey
, parent_key
);
545 btrfs_item_key_to_cpu(buf
, &cpukey
, 0);
547 btrfs_add_corrupt_extent_record(root
->fs_info
, &cpukey
,
548 buf
->start
, buf
->len
, 0);
553 static int noinline
check_block(struct btrfs_root
*root
,
554 struct btrfs_path
*path
, int level
)
556 struct btrfs_disk_key key
;
557 struct btrfs_disk_key
*key_ptr
= NULL
;
558 struct extent_buffer
*parent
;
559 enum btrfs_tree_block_status ret
;
561 if (path
->skip_check_block
)
563 if (path
->nodes
[level
+ 1]) {
564 parent
= path
->nodes
[level
+ 1];
565 btrfs_node_key(parent
, &key
, path
->slots
[level
+ 1]);
569 ret
= btrfs_check_leaf(root
, key_ptr
, path
->nodes
[0]);
571 ret
= btrfs_check_node(root
, key_ptr
, path
->nodes
[level
]);
572 if (ret
== BTRFS_TREE_BLOCK_CLEAN
)
578 * search for key in the extent_buffer. The items start at offset p,
579 * and they are item_size apart. There are 'max' items in p.
581 * the slot in the array is returned via slot, and it points to
582 * the place where you would insert key if it is not found in
585 * slot may point to max if the key is bigger than all of the keys
587 static int generic_bin_search(struct extent_buffer
*eb
, unsigned long p
,
588 int item_size
, struct btrfs_key
*key
,
595 unsigned long offset
;
596 struct btrfs_disk_key
*tmp
;
599 mid
= (low
+ high
) / 2;
600 offset
= p
+ mid
* item_size
;
602 tmp
= (struct btrfs_disk_key
*)(eb
->data
+ offset
);
603 ret
= btrfs_comp_keys(tmp
, key
);
619 * simple bin_search frontend that does the right thing for
622 static int bin_search(struct extent_buffer
*eb
, struct btrfs_key
*key
,
623 int level
, int *slot
)
626 return generic_bin_search(eb
,
627 offsetof(struct btrfs_leaf
, items
),
628 sizeof(struct btrfs_item
),
629 key
, btrfs_header_nritems(eb
),
632 return generic_bin_search(eb
,
633 offsetof(struct btrfs_node
, ptrs
),
634 sizeof(struct btrfs_key_ptr
),
635 key
, btrfs_header_nritems(eb
),
639 struct extent_buffer
*read_node_slot(struct btrfs_root
*root
,
640 struct extent_buffer
*parent
, int slot
)
642 int level
= btrfs_header_level(parent
);
645 if (slot
>= btrfs_header_nritems(parent
))
651 return read_tree_block(root
, btrfs_node_blockptr(parent
, slot
),
652 btrfs_level_size(root
, level
- 1),
653 btrfs_node_ptr_generation(parent
, slot
));
656 static int balance_level(struct btrfs_trans_handle
*trans
,
657 struct btrfs_root
*root
,
658 struct btrfs_path
*path
, int level
)
660 struct extent_buffer
*right
= NULL
;
661 struct extent_buffer
*mid
;
662 struct extent_buffer
*left
= NULL
;
663 struct extent_buffer
*parent
= NULL
;
667 int orig_slot
= path
->slots
[level
];
673 mid
= path
->nodes
[level
];
674 WARN_ON(btrfs_header_generation(mid
) != trans
->transid
);
676 orig_ptr
= btrfs_node_blockptr(mid
, orig_slot
);
678 if (level
< BTRFS_MAX_LEVEL
- 1) {
679 parent
= path
->nodes
[level
+ 1];
680 pslot
= path
->slots
[level
+ 1];
684 * deal with the case where there is only one pointer in the root
685 * by promoting the node below to a root
688 struct extent_buffer
*child
;
690 if (btrfs_header_nritems(mid
) != 1)
693 /* promote the child to a root */
694 child
= read_node_slot(root
, mid
, 0);
695 BUG_ON(!extent_buffer_uptodate(child
));
696 ret
= btrfs_cow_block(trans
, root
, child
, mid
, 0, &child
);
700 add_root_to_dirty_list(root
);
701 path
->nodes
[level
] = NULL
;
702 clean_tree_block(trans
, root
, mid
);
703 wait_on_tree_block_writeback(root
, mid
);
704 /* once for the path */
705 free_extent_buffer(mid
);
707 ret
= btrfs_free_extent(trans
, root
, mid
->start
, mid
->len
,
708 0, root
->root_key
.objectid
,
710 /* once for the root ptr */
711 free_extent_buffer(mid
);
714 if (btrfs_header_nritems(mid
) >
715 BTRFS_NODEPTRS_PER_BLOCK(root
) / 4)
718 left
= read_node_slot(root
, parent
, pslot
- 1);
719 if (extent_buffer_uptodate(left
)) {
720 wret
= btrfs_cow_block(trans
, root
, left
,
721 parent
, pslot
- 1, &left
);
727 right
= read_node_slot(root
, parent
, pslot
+ 1);
728 if (extent_buffer_uptodate(right
)) {
729 wret
= btrfs_cow_block(trans
, root
, right
,
730 parent
, pslot
+ 1, &right
);
737 /* first, try to make some room in the middle buffer */
739 orig_slot
+= btrfs_header_nritems(left
);
740 wret
= push_node_left(trans
, root
, left
, mid
, 1);
746 * then try to empty the right most buffer into the middle
749 wret
= push_node_left(trans
, root
, mid
, right
, 1);
750 if (wret
< 0 && wret
!= -ENOSPC
)
752 if (btrfs_header_nritems(right
) == 0) {
753 u64 bytenr
= right
->start
;
754 u32 blocksize
= right
->len
;
756 clean_tree_block(trans
, root
, right
);
757 wait_on_tree_block_writeback(root
, right
);
758 free_extent_buffer(right
);
760 wret
= btrfs_del_ptr(trans
, root
, path
,
761 level
+ 1, pslot
+ 1);
764 wret
= btrfs_free_extent(trans
, root
, bytenr
,
766 root
->root_key
.objectid
,
771 struct btrfs_disk_key right_key
;
772 btrfs_node_key(right
, &right_key
, 0);
773 btrfs_set_node_key(parent
, &right_key
, pslot
+ 1);
774 btrfs_mark_buffer_dirty(parent
);
777 if (btrfs_header_nritems(mid
) == 1) {
779 * we're not allowed to leave a node with one item in the
780 * tree during a delete. A deletion from lower in the tree
781 * could try to delete the only pointer in this node.
782 * So, pull some keys from the left.
783 * There has to be a left pointer at this point because
784 * otherwise we would have pulled some pointers from the
788 wret
= balance_node_right(trans
, root
, mid
, left
);
794 wret
= push_node_left(trans
, root
, left
, mid
, 1);
800 if (btrfs_header_nritems(mid
) == 0) {
801 /* we've managed to empty the middle node, drop it */
802 u64 bytenr
= mid
->start
;
803 u32 blocksize
= mid
->len
;
804 clean_tree_block(trans
, root
, mid
);
805 wait_on_tree_block_writeback(root
, mid
);
806 free_extent_buffer(mid
);
808 wret
= btrfs_del_ptr(trans
, root
, path
, level
+ 1, pslot
);
811 wret
= btrfs_free_extent(trans
, root
, bytenr
, blocksize
,
812 0, root
->root_key
.objectid
,
817 /* update the parent key to reflect our changes */
818 struct btrfs_disk_key mid_key
;
819 btrfs_node_key(mid
, &mid_key
, 0);
820 btrfs_set_node_key(parent
, &mid_key
, pslot
);
821 btrfs_mark_buffer_dirty(parent
);
824 /* update the path */
826 if (btrfs_header_nritems(left
) > orig_slot
) {
827 extent_buffer_get(left
);
828 path
->nodes
[level
] = left
;
829 path
->slots
[level
+ 1] -= 1;
830 path
->slots
[level
] = orig_slot
;
832 free_extent_buffer(mid
);
834 orig_slot
-= btrfs_header_nritems(left
);
835 path
->slots
[level
] = orig_slot
;
838 /* double check we haven't messed things up */
839 check_block(root
, path
, level
);
841 btrfs_node_blockptr(path
->nodes
[level
], path
->slots
[level
]))
845 free_extent_buffer(right
);
847 free_extent_buffer(left
);
851 /* returns zero if the push worked, non-zero otherwise */
852 static int noinline
push_nodes_for_insert(struct btrfs_trans_handle
*trans
,
853 struct btrfs_root
*root
,
854 struct btrfs_path
*path
, int level
)
856 struct extent_buffer
*right
= NULL
;
857 struct extent_buffer
*mid
;
858 struct extent_buffer
*left
= NULL
;
859 struct extent_buffer
*parent
= NULL
;
863 int orig_slot
= path
->slots
[level
];
868 mid
= path
->nodes
[level
];
869 WARN_ON(btrfs_header_generation(mid
) != trans
->transid
);
871 if (level
< BTRFS_MAX_LEVEL
- 1) {
872 parent
= path
->nodes
[level
+ 1];
873 pslot
= path
->slots
[level
+ 1];
879 left
= read_node_slot(root
, parent
, pslot
- 1);
881 /* first, try to make some room in the middle buffer */
882 if (extent_buffer_uptodate(left
)) {
884 left_nr
= btrfs_header_nritems(left
);
885 if (left_nr
>= BTRFS_NODEPTRS_PER_BLOCK(root
) - 1) {
888 ret
= btrfs_cow_block(trans
, root
, left
, parent
,
893 wret
= push_node_left(trans
, root
,
900 struct btrfs_disk_key disk_key
;
901 orig_slot
+= left_nr
;
902 btrfs_node_key(mid
, &disk_key
, 0);
903 btrfs_set_node_key(parent
, &disk_key
, pslot
);
904 btrfs_mark_buffer_dirty(parent
);
905 if (btrfs_header_nritems(left
) > orig_slot
) {
906 path
->nodes
[level
] = left
;
907 path
->slots
[level
+ 1] -= 1;
908 path
->slots
[level
] = orig_slot
;
909 free_extent_buffer(mid
);
912 btrfs_header_nritems(left
);
913 path
->slots
[level
] = orig_slot
;
914 free_extent_buffer(left
);
918 free_extent_buffer(left
);
920 right
= read_node_slot(root
, parent
, pslot
+ 1);
923 * then try to empty the right most buffer into the middle
925 if (extent_buffer_uptodate(right
)) {
927 right_nr
= btrfs_header_nritems(right
);
928 if (right_nr
>= BTRFS_NODEPTRS_PER_BLOCK(root
) - 1) {
931 ret
= btrfs_cow_block(trans
, root
, right
,
937 wret
= balance_node_right(trans
, root
,
944 struct btrfs_disk_key disk_key
;
946 btrfs_node_key(right
, &disk_key
, 0);
947 btrfs_set_node_key(parent
, &disk_key
, pslot
+ 1);
948 btrfs_mark_buffer_dirty(parent
);
950 if (btrfs_header_nritems(mid
) <= orig_slot
) {
951 path
->nodes
[level
] = right
;
952 path
->slots
[level
+ 1] += 1;
953 path
->slots
[level
] = orig_slot
-
954 btrfs_header_nritems(mid
);
955 free_extent_buffer(mid
);
957 free_extent_buffer(right
);
961 free_extent_buffer(right
);
967 * readahead one full node of leaves
969 void reada_for_search(struct btrfs_root
*root
, struct btrfs_path
*path
,
970 int level
, int slot
, u64 objectid
)
972 struct extent_buffer
*node
;
973 struct btrfs_disk_key disk_key
;
979 int direction
= path
->reada
;
980 struct extent_buffer
*eb
;
988 if (!path
->nodes
[level
])
991 node
= path
->nodes
[level
];
992 search
= btrfs_node_blockptr(node
, slot
);
993 blocksize
= btrfs_level_size(root
, level
- 1);
994 eb
= btrfs_find_tree_block(root
, search
, blocksize
);
996 free_extent_buffer(eb
);
1000 highest_read
= search
;
1001 lowest_read
= search
;
1003 nritems
= btrfs_header_nritems(node
);
1006 if (direction
< 0) {
1010 } else if (direction
> 0) {
1015 if (path
->reada
< 0 && objectid
) {
1016 btrfs_node_key(node
, &disk_key
, nr
);
1017 if (btrfs_disk_key_objectid(&disk_key
) != objectid
)
1020 search
= btrfs_node_blockptr(node
, nr
);
1021 if ((search
>= lowest_read
&& search
<= highest_read
) ||
1022 (search
< lowest_read
&& lowest_read
- search
<= 32768) ||
1023 (search
> highest_read
&& search
- highest_read
<= 32768)) {
1024 readahead_tree_block(root
, search
, blocksize
,
1025 btrfs_node_ptr_generation(node
, nr
));
1029 if (path
->reada
< 2 && (nread
> (256 * 1024) || nscan
> 32))
1031 if(nread
> (1024 * 1024) || nscan
> 128)
1034 if (search
< lowest_read
)
1035 lowest_read
= search
;
1036 if (search
> highest_read
)
1037 highest_read
= search
;
1041 int btrfs_find_item(struct btrfs_root
*fs_root
, struct btrfs_path
*found_path
,
1042 u64 iobjectid
, u64 ioff
, u8 key_type
,
1043 struct btrfs_key
*found_key
)
1046 struct btrfs_key key
;
1047 struct extent_buffer
*eb
;
1048 struct btrfs_path
*path
;
1050 key
.type
= key_type
;
1051 key
.objectid
= iobjectid
;
1054 if (found_path
== NULL
) {
1055 path
= btrfs_alloc_path();
1061 ret
= btrfs_search_slot(NULL
, fs_root
, &key
, path
, 0, 0);
1062 if ((ret
< 0) || (found_key
== NULL
))
1065 eb
= path
->nodes
[0];
1066 if (ret
&& path
->slots
[0] >= btrfs_header_nritems(eb
)) {
1067 ret
= btrfs_next_leaf(fs_root
, path
);
1070 eb
= path
->nodes
[0];
1073 btrfs_item_key_to_cpu(eb
, found_key
, path
->slots
[0]);
1074 if (found_key
->type
!= key
.type
||
1075 found_key
->objectid
!= key
.objectid
) {
1081 if (path
!= found_path
)
1082 btrfs_free_path(path
);
1087 * look for key in the tree. path is filled in with nodes along the way
1088 * if key is found, we return zero and you can find the item in the leaf
1089 * level of the path (level 0)
1091 * If the key isn't found, the path points to the slot where it should
1092 * be inserted, and 1 is returned. If there are other errors during the
1093 * search a negative error number is returned.
1095 * if ins_len > 0, nodes and leaves will be split as we walk down the
1096 * tree. if ins_len < 0, nodes will be merged as we walk down the tree (if
1099 int btrfs_search_slot(struct btrfs_trans_handle
*trans
, struct btrfs_root
1100 *root
, struct btrfs_key
*key
, struct btrfs_path
*p
, int
1103 struct extent_buffer
*b
;
1107 int should_reada
= p
->reada
;
1108 u8 lowest_level
= 0;
1110 lowest_level
= p
->lowest_level
;
1111 WARN_ON(lowest_level
&& ins_len
> 0);
1112 WARN_ON(p
->nodes
[0] != NULL
);
1114 WARN_ON(!mutex_is_locked(&root->fs_info->fs_mutex));
1118 extent_buffer_get(b
);
1120 level
= btrfs_header_level(b
);
1123 wret
= btrfs_cow_block(trans
, root
, b
,
1124 p
->nodes
[level
+ 1],
1125 p
->slots
[level
+ 1],
1128 free_extent_buffer(b
);
1132 BUG_ON(!cow
&& ins_len
);
1133 if (level
!= btrfs_header_level(b
))
1135 level
= btrfs_header_level(b
);
1136 p
->nodes
[level
] = b
;
1137 ret
= check_block(root
, p
, level
);
1140 ret
= bin_search(b
, key
, level
, &slot
);
1142 if (ret
&& slot
> 0)
1144 p
->slots
[level
] = slot
;
1145 if ((p
->search_for_split
|| ins_len
> 0) &&
1146 btrfs_header_nritems(b
) >=
1147 BTRFS_NODEPTRS_PER_BLOCK(root
) - 3) {
1148 int sret
= split_node(trans
, root
, p
, level
);
1152 b
= p
->nodes
[level
];
1153 slot
= p
->slots
[level
];
1154 } else if (ins_len
< 0) {
1155 int sret
= balance_level(trans
, root
, p
,
1159 b
= p
->nodes
[level
];
1161 btrfs_release_path(p
);
1164 slot
= p
->slots
[level
];
1165 BUG_ON(btrfs_header_nritems(b
) == 1);
1167 /* this is only true while dropping a snapshot */
1168 if (level
== lowest_level
)
1172 reada_for_search(root
, p
, level
, slot
,
1175 b
= read_node_slot(root
, b
, slot
);
1176 if (!extent_buffer_uptodate(b
))
1179 p
->slots
[level
] = slot
;
1181 ins_len
> btrfs_leaf_free_space(root
, b
)) {
1182 int sret
= split_leaf(trans
, root
, key
,
1183 p
, ins_len
, ret
== 0);
1195 * adjust the pointers going up the tree, starting at level
1196 * making sure the right key of each node is points to 'key'.
1197 * This is used after shifting pointers to the left, so it stops
1198 * fixing up pointers when a given leaf/node is not in slot 0 of the
1201 void btrfs_fixup_low_keys(struct btrfs_root
*root
, struct btrfs_path
*path
,
1202 struct btrfs_disk_key
*key
, int level
)
1205 struct extent_buffer
*t
;
1207 for (i
= level
; i
< BTRFS_MAX_LEVEL
; i
++) {
1208 int tslot
= path
->slots
[i
];
1209 if (!path
->nodes
[i
])
1212 btrfs_set_node_key(t
, key
, tslot
);
1213 btrfs_mark_buffer_dirty(path
->nodes
[i
]);
1222 * This function isn't completely safe. It's the caller's responsibility
1223 * that the new key won't break the order
1225 int btrfs_set_item_key_safe(struct btrfs_root
*root
, struct btrfs_path
*path
,
1226 struct btrfs_key
*new_key
)
1228 struct btrfs_disk_key disk_key
;
1229 struct extent_buffer
*eb
;
1232 eb
= path
->nodes
[0];
1233 slot
= path
->slots
[0];
1235 btrfs_item_key(eb
, &disk_key
, slot
- 1);
1236 if (btrfs_comp_keys(&disk_key
, new_key
) >= 0)
1239 if (slot
< btrfs_header_nritems(eb
) - 1) {
1240 btrfs_item_key(eb
, &disk_key
, slot
+ 1);
1241 if (btrfs_comp_keys(&disk_key
, new_key
) <= 0)
1245 btrfs_cpu_key_to_disk(&disk_key
, new_key
);
1246 btrfs_set_item_key(eb
, &disk_key
, slot
);
1247 btrfs_mark_buffer_dirty(eb
);
1249 btrfs_fixup_low_keys(root
, path
, &disk_key
, 1);
1254 * update an item key without the safety checks. This is meant to be called by
1257 void btrfs_set_item_key_unsafe(struct btrfs_root
*root
,
1258 struct btrfs_path
*path
,
1259 struct btrfs_key
*new_key
)
1261 struct btrfs_disk_key disk_key
;
1262 struct extent_buffer
*eb
;
1265 eb
= path
->nodes
[0];
1266 slot
= path
->slots
[0];
1268 btrfs_cpu_key_to_disk(&disk_key
, new_key
);
1269 btrfs_set_item_key(eb
, &disk_key
, slot
);
1270 btrfs_mark_buffer_dirty(eb
);
1272 btrfs_fixup_low_keys(root
, path
, &disk_key
, 1);
1276 * try to push data from one node into the next node left in the
1279 * returns 0 if some ptrs were pushed left, < 0 if there was some horrible
1280 * error, and > 0 if there was no room in the left hand block.
1282 static int push_node_left(struct btrfs_trans_handle
*trans
,
1283 struct btrfs_root
*root
, struct extent_buffer
*dst
,
1284 struct extent_buffer
*src
, int empty
)
1291 src_nritems
= btrfs_header_nritems(src
);
1292 dst_nritems
= btrfs_header_nritems(dst
);
1293 push_items
= BTRFS_NODEPTRS_PER_BLOCK(root
) - dst_nritems
;
1294 WARN_ON(btrfs_header_generation(src
) != trans
->transid
);
1295 WARN_ON(btrfs_header_generation(dst
) != trans
->transid
);
1297 if (!empty
&& src_nritems
<= 8)
1300 if (push_items
<= 0) {
1305 push_items
= min(src_nritems
, push_items
);
1306 if (push_items
< src_nritems
) {
1307 /* leave at least 8 pointers in the node if
1308 * we aren't going to empty it
1310 if (src_nritems
- push_items
< 8) {
1311 if (push_items
<= 8)
1317 push_items
= min(src_nritems
- 8, push_items
);
1319 copy_extent_buffer(dst
, src
,
1320 btrfs_node_key_ptr_offset(dst_nritems
),
1321 btrfs_node_key_ptr_offset(0),
1322 push_items
* sizeof(struct btrfs_key_ptr
));
1324 if (push_items
< src_nritems
) {
1325 memmove_extent_buffer(src
, btrfs_node_key_ptr_offset(0),
1326 btrfs_node_key_ptr_offset(push_items
),
1327 (src_nritems
- push_items
) *
1328 sizeof(struct btrfs_key_ptr
));
1330 btrfs_set_header_nritems(src
, src_nritems
- push_items
);
1331 btrfs_set_header_nritems(dst
, dst_nritems
+ push_items
);
1332 btrfs_mark_buffer_dirty(src
);
1333 btrfs_mark_buffer_dirty(dst
);
1339 * try to push data from one node into the next node right in the
1342 * returns 0 if some ptrs were pushed, < 0 if there was some horrible
1343 * error, and > 0 if there was no room in the right hand block.
1345 * this will only push up to 1/2 the contents of the left node over
1347 static int balance_node_right(struct btrfs_trans_handle
*trans
,
1348 struct btrfs_root
*root
,
1349 struct extent_buffer
*dst
,
1350 struct extent_buffer
*src
)
1358 WARN_ON(btrfs_header_generation(src
) != trans
->transid
);
1359 WARN_ON(btrfs_header_generation(dst
) != trans
->transid
);
1361 src_nritems
= btrfs_header_nritems(src
);
1362 dst_nritems
= btrfs_header_nritems(dst
);
1363 push_items
= BTRFS_NODEPTRS_PER_BLOCK(root
) - dst_nritems
;
1364 if (push_items
<= 0) {
1368 if (src_nritems
< 4) {
1372 max_push
= src_nritems
/ 2 + 1;
1373 /* don't try to empty the node */
1374 if (max_push
>= src_nritems
) {
1378 if (max_push
< push_items
)
1379 push_items
= max_push
;
1381 memmove_extent_buffer(dst
, btrfs_node_key_ptr_offset(push_items
),
1382 btrfs_node_key_ptr_offset(0),
1384 sizeof(struct btrfs_key_ptr
));
1386 copy_extent_buffer(dst
, src
,
1387 btrfs_node_key_ptr_offset(0),
1388 btrfs_node_key_ptr_offset(src_nritems
- push_items
),
1389 push_items
* sizeof(struct btrfs_key_ptr
));
1391 btrfs_set_header_nritems(src
, src_nritems
- push_items
);
1392 btrfs_set_header_nritems(dst
, dst_nritems
+ push_items
);
1394 btrfs_mark_buffer_dirty(src
);
1395 btrfs_mark_buffer_dirty(dst
);
1401 * helper function to insert a new root level in the tree.
1402 * A new node is allocated, and a single item is inserted to
1403 * point to the existing root
1405 * returns zero on success or < 0 on failure.
1407 static int noinline
insert_new_root(struct btrfs_trans_handle
*trans
,
1408 struct btrfs_root
*root
,
1409 struct btrfs_path
*path
, int level
)
1412 struct extent_buffer
*lower
;
1413 struct extent_buffer
*c
;
1414 struct extent_buffer
*old
;
1415 struct btrfs_disk_key lower_key
;
1417 BUG_ON(path
->nodes
[level
]);
1418 BUG_ON(path
->nodes
[level
-1] != root
->node
);
1420 lower
= path
->nodes
[level
-1];
1422 btrfs_item_key(lower
, &lower_key
, 0);
1424 btrfs_node_key(lower
, &lower_key
, 0);
1426 c
= btrfs_alloc_free_block(trans
, root
, root
->nodesize
,
1427 root
->root_key
.objectid
, &lower_key
,
1428 level
, root
->node
->start
, 0);
1433 memset_extent_buffer(c
, 0, 0, sizeof(struct btrfs_header
));
1434 btrfs_set_header_nritems(c
, 1);
1435 btrfs_set_header_level(c
, level
);
1436 btrfs_set_header_bytenr(c
, c
->start
);
1437 btrfs_set_header_generation(c
, trans
->transid
);
1438 btrfs_set_header_backref_rev(c
, BTRFS_MIXED_BACKREF_REV
);
1439 btrfs_set_header_owner(c
, root
->root_key
.objectid
);
1441 write_extent_buffer(c
, root
->fs_info
->fsid
,
1442 btrfs_header_fsid(), BTRFS_FSID_SIZE
);
1444 write_extent_buffer(c
, root
->fs_info
->chunk_tree_uuid
,
1445 btrfs_header_chunk_tree_uuid(c
),
1448 btrfs_set_node_key(c
, &lower_key
, 0);
1449 btrfs_set_node_blockptr(c
, 0, lower
->start
);
1450 lower_gen
= btrfs_header_generation(lower
);
1451 WARN_ON(lower_gen
!= trans
->transid
);
1453 btrfs_set_node_ptr_generation(c
, 0, lower_gen
);
1455 btrfs_mark_buffer_dirty(c
);
1460 /* the super has an extra ref to root->node */
1461 free_extent_buffer(old
);
1463 add_root_to_dirty_list(root
);
1464 extent_buffer_get(c
);
1465 path
->nodes
[level
] = c
;
1466 path
->slots
[level
] = 0;
1471 * worker function to insert a single pointer in a node.
1472 * the node should have enough room for the pointer already
1474 * slot and level indicate where you want the key to go, and
1475 * blocknr is the block the key points to.
1477 * returns zero on success and < 0 on any error
1479 static int insert_ptr(struct btrfs_trans_handle
*trans
, struct btrfs_root
1480 *root
, struct btrfs_path
*path
, struct btrfs_disk_key
1481 *key
, u64 bytenr
, int slot
, int level
)
1483 struct extent_buffer
*lower
;
1486 BUG_ON(!path
->nodes
[level
]);
1487 lower
= path
->nodes
[level
];
1488 nritems
= btrfs_header_nritems(lower
);
1491 if (nritems
== BTRFS_NODEPTRS_PER_BLOCK(root
))
1493 if (slot
!= nritems
) {
1494 memmove_extent_buffer(lower
,
1495 btrfs_node_key_ptr_offset(slot
+ 1),
1496 btrfs_node_key_ptr_offset(slot
),
1497 (nritems
- slot
) * sizeof(struct btrfs_key_ptr
));
1499 btrfs_set_node_key(lower
, key
, slot
);
1500 btrfs_set_node_blockptr(lower
, slot
, bytenr
);
1501 WARN_ON(trans
->transid
== 0);
1502 btrfs_set_node_ptr_generation(lower
, slot
, trans
->transid
);
1503 btrfs_set_header_nritems(lower
, nritems
+ 1);
1504 btrfs_mark_buffer_dirty(lower
);
1509 * split the node at the specified level in path in two.
1510 * The path is corrected to point to the appropriate node after the split
1512 * Before splitting this tries to make some room in the node by pushing
1513 * left and right, if either one works, it returns right away.
1515 * returns 0 on success and < 0 on failure
1517 static int split_node(struct btrfs_trans_handle
*trans
, struct btrfs_root
1518 *root
, struct btrfs_path
*path
, int level
)
1520 struct extent_buffer
*c
;
1521 struct extent_buffer
*split
;
1522 struct btrfs_disk_key disk_key
;
1528 c
= path
->nodes
[level
];
1529 WARN_ON(btrfs_header_generation(c
) != trans
->transid
);
1530 if (c
== root
->node
) {
1531 /* trying to split the root, lets make a new one */
1532 ret
= insert_new_root(trans
, root
, path
, level
+ 1);
1536 ret
= push_nodes_for_insert(trans
, root
, path
, level
);
1537 c
= path
->nodes
[level
];
1538 if (!ret
&& btrfs_header_nritems(c
) <
1539 BTRFS_NODEPTRS_PER_BLOCK(root
) - 3)
1545 c_nritems
= btrfs_header_nritems(c
);
1546 mid
= (c_nritems
+ 1) / 2;
1547 btrfs_node_key(c
, &disk_key
, mid
);
1549 split
= btrfs_alloc_free_block(trans
, root
, root
->nodesize
,
1550 root
->root_key
.objectid
,
1551 &disk_key
, level
, c
->start
, 0);
1553 return PTR_ERR(split
);
1555 memset_extent_buffer(split
, 0, 0, sizeof(struct btrfs_header
));
1556 btrfs_set_header_level(split
, btrfs_header_level(c
));
1557 btrfs_set_header_bytenr(split
, split
->start
);
1558 btrfs_set_header_generation(split
, trans
->transid
);
1559 btrfs_set_header_backref_rev(split
, BTRFS_MIXED_BACKREF_REV
);
1560 btrfs_set_header_owner(split
, root
->root_key
.objectid
);
1561 write_extent_buffer(split
, root
->fs_info
->fsid
,
1562 btrfs_header_fsid(), BTRFS_FSID_SIZE
);
1563 write_extent_buffer(split
, root
->fs_info
->chunk_tree_uuid
,
1564 btrfs_header_chunk_tree_uuid(split
),
1568 copy_extent_buffer(split
, c
,
1569 btrfs_node_key_ptr_offset(0),
1570 btrfs_node_key_ptr_offset(mid
),
1571 (c_nritems
- mid
) * sizeof(struct btrfs_key_ptr
));
1572 btrfs_set_header_nritems(split
, c_nritems
- mid
);
1573 btrfs_set_header_nritems(c
, mid
);
1576 btrfs_mark_buffer_dirty(c
);
1577 btrfs_mark_buffer_dirty(split
);
1579 wret
= insert_ptr(trans
, root
, path
, &disk_key
, split
->start
,
1580 path
->slots
[level
+ 1] + 1,
1585 if (path
->slots
[level
] >= mid
) {
1586 path
->slots
[level
] -= mid
;
1587 free_extent_buffer(c
);
1588 path
->nodes
[level
] = split
;
1589 path
->slots
[level
+ 1] += 1;
1591 free_extent_buffer(split
);
1597 * how many bytes are required to store the items in a leaf. start
1598 * and nr indicate which items in the leaf to check. This totals up the
1599 * space used both by the item structs and the item data
1601 static int leaf_space_used(struct extent_buffer
*l
, int start
, int nr
)
1604 int nritems
= btrfs_header_nritems(l
);
1605 int end
= min(nritems
, start
+ nr
) - 1;
1609 data_len
= btrfs_item_end_nr(l
, start
);
1610 data_len
= data_len
- btrfs_item_offset_nr(l
, end
);
1611 data_len
+= sizeof(struct btrfs_item
) * nr
;
1612 WARN_ON(data_len
< 0);
1617 * The space between the end of the leaf items and
1618 * the start of the leaf data. IOW, how much room
1619 * the leaf has left for both items and data
1621 int btrfs_leaf_free_space(struct btrfs_root
*root
, struct extent_buffer
*leaf
)
1623 u32 nodesize
= (root
? BTRFS_LEAF_DATA_SIZE(root
) : leaf
->len
);
1624 int nritems
= btrfs_header_nritems(leaf
);
1626 ret
= nodesize
- leaf_space_used(leaf
, 0, nritems
);
1628 printk("leaf free space ret %d, leaf data size %u, used %d nritems %d\n",
1629 ret
, nodesize
, leaf_space_used(leaf
, 0, nritems
),
1636 * push some data in the path leaf to the right, trying to free up at
1637 * least data_size bytes. returns zero if the push worked, nonzero otherwise
1639 * returns 1 if the push failed because the other node didn't have enough
1640 * room, 0 if everything worked out and < 0 if there were major errors.
1642 static int push_leaf_right(struct btrfs_trans_handle
*trans
, struct btrfs_root
1643 *root
, struct btrfs_path
*path
, int data_size
,
1646 struct extent_buffer
*left
= path
->nodes
[0];
1647 struct extent_buffer
*right
;
1648 struct extent_buffer
*upper
;
1649 struct btrfs_disk_key disk_key
;
1655 struct btrfs_item
*item
;
1663 slot
= path
->slots
[1];
1664 if (!path
->nodes
[1]) {
1667 upper
= path
->nodes
[1];
1668 if (slot
>= btrfs_header_nritems(upper
) - 1)
1671 right
= read_node_slot(root
, upper
, slot
+ 1);
1672 if (!extent_buffer_uptodate(right
)) {
1674 return PTR_ERR(right
);
1677 free_space
= btrfs_leaf_free_space(root
, right
);
1678 if (free_space
< data_size
) {
1679 free_extent_buffer(right
);
1683 /* cow and double check */
1684 ret
= btrfs_cow_block(trans
, root
, right
, upper
,
1687 free_extent_buffer(right
);
1690 free_space
= btrfs_leaf_free_space(root
, right
);
1691 if (free_space
< data_size
) {
1692 free_extent_buffer(right
);
1696 left_nritems
= btrfs_header_nritems(left
);
1697 if (left_nritems
== 0) {
1698 free_extent_buffer(right
);
1707 i
= left_nritems
- 1;
1709 item
= btrfs_item_nr(i
);
1711 if (path
->slots
[0] == i
)
1712 push_space
+= data_size
+ sizeof(*item
);
1714 this_item_size
= btrfs_item_size(left
, item
);
1715 if (this_item_size
+ sizeof(*item
) + push_space
> free_space
)
1718 push_space
+= this_item_size
+ sizeof(*item
);
1724 if (push_items
== 0) {
1725 free_extent_buffer(right
);
1729 if (!empty
&& push_items
== left_nritems
)
1732 /* push left to right */
1733 right_nritems
= btrfs_header_nritems(right
);
1735 push_space
= btrfs_item_end_nr(left
, left_nritems
- push_items
);
1736 push_space
-= leaf_data_end(root
, left
);
1738 /* make room in the right data area */
1739 data_end
= leaf_data_end(root
, right
);
1740 memmove_extent_buffer(right
,
1741 btrfs_leaf_data(right
) + data_end
- push_space
,
1742 btrfs_leaf_data(right
) + data_end
,
1743 BTRFS_LEAF_DATA_SIZE(root
) - data_end
);
1745 /* copy from the left data area */
1746 copy_extent_buffer(right
, left
, btrfs_leaf_data(right
) +
1747 BTRFS_LEAF_DATA_SIZE(root
) - push_space
,
1748 btrfs_leaf_data(left
) + leaf_data_end(root
, left
),
1751 memmove_extent_buffer(right
, btrfs_item_nr_offset(push_items
),
1752 btrfs_item_nr_offset(0),
1753 right_nritems
* sizeof(struct btrfs_item
));
1755 /* copy the items from left to right */
1756 copy_extent_buffer(right
, left
, btrfs_item_nr_offset(0),
1757 btrfs_item_nr_offset(left_nritems
- push_items
),
1758 push_items
* sizeof(struct btrfs_item
));
1760 /* update the item pointers */
1761 right_nritems
+= push_items
;
1762 btrfs_set_header_nritems(right
, right_nritems
);
1763 push_space
= BTRFS_LEAF_DATA_SIZE(root
);
1764 for (i
= 0; i
< right_nritems
; i
++) {
1765 item
= btrfs_item_nr(i
);
1766 push_space
-= btrfs_item_size(right
, item
);
1767 btrfs_set_item_offset(right
, item
, push_space
);
1770 left_nritems
-= push_items
;
1771 btrfs_set_header_nritems(left
, left_nritems
);
1774 btrfs_mark_buffer_dirty(left
);
1775 btrfs_mark_buffer_dirty(right
);
1777 btrfs_item_key(right
, &disk_key
, 0);
1778 btrfs_set_node_key(upper
, &disk_key
, slot
+ 1);
1779 btrfs_mark_buffer_dirty(upper
);
1781 /* then fixup the leaf pointer in the path */
1782 if (path
->slots
[0] >= left_nritems
) {
1783 path
->slots
[0] -= left_nritems
;
1784 free_extent_buffer(path
->nodes
[0]);
1785 path
->nodes
[0] = right
;
1786 path
->slots
[1] += 1;
1788 free_extent_buffer(right
);
1793 * push some data in the path leaf to the left, trying to free up at
1794 * least data_size bytes. returns zero if the push worked, nonzero otherwise
1796 static int push_leaf_left(struct btrfs_trans_handle
*trans
, struct btrfs_root
1797 *root
, struct btrfs_path
*path
, int data_size
,
1800 struct btrfs_disk_key disk_key
;
1801 struct extent_buffer
*right
= path
->nodes
[0];
1802 struct extent_buffer
*left
;
1808 struct btrfs_item
*item
;
1809 u32 old_left_nritems
;
1814 u32 old_left_item_size
;
1816 slot
= path
->slots
[1];
1819 if (!path
->nodes
[1])
1822 right_nritems
= btrfs_header_nritems(right
);
1823 if (right_nritems
== 0) {
1827 left
= read_node_slot(root
, path
->nodes
[1], slot
- 1);
1828 free_space
= btrfs_leaf_free_space(root
, left
);
1829 if (free_space
< data_size
) {
1830 free_extent_buffer(left
);
1834 /* cow and double check */
1835 ret
= btrfs_cow_block(trans
, root
, left
,
1836 path
->nodes
[1], slot
- 1, &left
);
1838 /* we hit -ENOSPC, but it isn't fatal here */
1839 free_extent_buffer(left
);
1843 free_space
= btrfs_leaf_free_space(root
, left
);
1844 if (free_space
< data_size
) {
1845 free_extent_buffer(left
);
1852 nr
= right_nritems
- 1;
1854 for (i
= 0; i
< nr
; i
++) {
1855 item
= btrfs_item_nr(i
);
1857 if (path
->slots
[0] == i
)
1858 push_space
+= data_size
+ sizeof(*item
);
1860 this_item_size
= btrfs_item_size(right
, item
);
1861 if (this_item_size
+ sizeof(*item
) + push_space
> free_space
)
1865 push_space
+= this_item_size
+ sizeof(*item
);
1868 if (push_items
== 0) {
1869 free_extent_buffer(left
);
1872 if (!empty
&& push_items
== btrfs_header_nritems(right
))
1875 /* push data from right to left */
1876 copy_extent_buffer(left
, right
,
1877 btrfs_item_nr_offset(btrfs_header_nritems(left
)),
1878 btrfs_item_nr_offset(0),
1879 push_items
* sizeof(struct btrfs_item
));
1881 push_space
= BTRFS_LEAF_DATA_SIZE(root
) -
1882 btrfs_item_offset_nr(right
, push_items
-1);
1884 copy_extent_buffer(left
, right
, btrfs_leaf_data(left
) +
1885 leaf_data_end(root
, left
) - push_space
,
1886 btrfs_leaf_data(right
) +
1887 btrfs_item_offset_nr(right
, push_items
- 1),
1889 old_left_nritems
= btrfs_header_nritems(left
);
1890 BUG_ON(old_left_nritems
== 0);
1892 old_left_item_size
= btrfs_item_offset_nr(left
, old_left_nritems
- 1);
1893 for (i
= old_left_nritems
; i
< old_left_nritems
+ push_items
; i
++) {
1896 item
= btrfs_item_nr(i
);
1897 ioff
= btrfs_item_offset(left
, item
);
1898 btrfs_set_item_offset(left
, item
,
1899 ioff
- (BTRFS_LEAF_DATA_SIZE(root
) - old_left_item_size
));
1901 btrfs_set_header_nritems(left
, old_left_nritems
+ push_items
);
1903 /* fixup right node */
1904 if (push_items
> right_nritems
) {
1905 printk("push items %d nr %u\n", push_items
, right_nritems
);
1909 if (push_items
< right_nritems
) {
1910 push_space
= btrfs_item_offset_nr(right
, push_items
- 1) -
1911 leaf_data_end(root
, right
);
1912 memmove_extent_buffer(right
, btrfs_leaf_data(right
) +
1913 BTRFS_LEAF_DATA_SIZE(root
) - push_space
,
1914 btrfs_leaf_data(right
) +
1915 leaf_data_end(root
, right
), push_space
);
1917 memmove_extent_buffer(right
, btrfs_item_nr_offset(0),
1918 btrfs_item_nr_offset(push_items
),
1919 (btrfs_header_nritems(right
) - push_items
) *
1920 sizeof(struct btrfs_item
));
1922 right_nritems
-= push_items
;
1923 btrfs_set_header_nritems(right
, right_nritems
);
1924 push_space
= BTRFS_LEAF_DATA_SIZE(root
);
1925 for (i
= 0; i
< right_nritems
; i
++) {
1926 item
= btrfs_item_nr(i
);
1927 push_space
= push_space
- btrfs_item_size(right
, item
);
1928 btrfs_set_item_offset(right
, item
, push_space
);
1931 btrfs_mark_buffer_dirty(left
);
1933 btrfs_mark_buffer_dirty(right
);
1935 btrfs_item_key(right
, &disk_key
, 0);
1936 btrfs_fixup_low_keys(root
, path
, &disk_key
, 1);
1938 /* then fixup the leaf pointer in the path */
1939 if (path
->slots
[0] < push_items
) {
1940 path
->slots
[0] += old_left_nritems
;
1941 free_extent_buffer(path
->nodes
[0]);
1942 path
->nodes
[0] = left
;
1943 path
->slots
[1] -= 1;
1945 free_extent_buffer(left
);
1946 path
->slots
[0] -= push_items
;
1948 BUG_ON(path
->slots
[0] < 0);
1953 * split the path's leaf in two, making sure there is at least data_size
1954 * available for the resulting leaf level of the path.
1956 * returns 0 if all went well and < 0 on failure.
1958 static noinline
int copy_for_split(struct btrfs_trans_handle
*trans
,
1959 struct btrfs_root
*root
,
1960 struct btrfs_path
*path
,
1961 struct extent_buffer
*l
,
1962 struct extent_buffer
*right
,
1963 int slot
, int mid
, int nritems
)
1970 struct btrfs_disk_key disk_key
;
1972 nritems
= nritems
- mid
;
1973 btrfs_set_header_nritems(right
, nritems
);
1974 data_copy_size
= btrfs_item_end_nr(l
, mid
) - leaf_data_end(root
, l
);
1976 copy_extent_buffer(right
, l
, btrfs_item_nr_offset(0),
1977 btrfs_item_nr_offset(mid
),
1978 nritems
* sizeof(struct btrfs_item
));
1980 copy_extent_buffer(right
, l
,
1981 btrfs_leaf_data(right
) + BTRFS_LEAF_DATA_SIZE(root
) -
1982 data_copy_size
, btrfs_leaf_data(l
) +
1983 leaf_data_end(root
, l
), data_copy_size
);
1985 rt_data_off
= BTRFS_LEAF_DATA_SIZE(root
) -
1986 btrfs_item_end_nr(l
, mid
);
1988 for (i
= 0; i
< nritems
; i
++) {
1989 struct btrfs_item
*item
= btrfs_item_nr(i
);
1990 u32 ioff
= btrfs_item_offset(right
, item
);
1991 btrfs_set_item_offset(right
, item
, ioff
+ rt_data_off
);
1994 btrfs_set_header_nritems(l
, mid
);
1996 btrfs_item_key(right
, &disk_key
, 0);
1997 wret
= insert_ptr(trans
, root
, path
, &disk_key
, right
->start
,
1998 path
->slots
[1] + 1, 1);
2002 btrfs_mark_buffer_dirty(right
);
2003 btrfs_mark_buffer_dirty(l
);
2004 BUG_ON(path
->slots
[0] != slot
);
2007 free_extent_buffer(path
->nodes
[0]);
2008 path
->nodes
[0] = right
;
2009 path
->slots
[0] -= mid
;
2010 path
->slots
[1] += 1;
2012 free_extent_buffer(right
);
2015 BUG_ON(path
->slots
[0] < 0);
2021 * split the path's leaf in two, making sure there is at least data_size
2022 * available for the resulting leaf level of the path.
2024 * returns 0 if all went well and < 0 on failure.
2026 static noinline
int split_leaf(struct btrfs_trans_handle
*trans
,
2027 struct btrfs_root
*root
,
2028 struct btrfs_key
*ins_key
,
2029 struct btrfs_path
*path
, int data_size
,
2032 struct btrfs_disk_key disk_key
;
2033 struct extent_buffer
*l
;
2037 struct extent_buffer
*right
;
2041 int num_doubles
= 0;
2044 slot
= path
->slots
[0];
2045 if (extend
&& data_size
+ btrfs_item_size_nr(l
, slot
) +
2046 sizeof(struct btrfs_item
) > BTRFS_LEAF_DATA_SIZE(root
))
2049 /* first try to make some room by pushing left and right */
2050 if (data_size
&& ins_key
->type
!= BTRFS_DIR_ITEM_KEY
) {
2051 wret
= push_leaf_right(trans
, root
, path
, data_size
, 0);
2055 wret
= push_leaf_left(trans
, root
, path
, data_size
, 0);
2061 /* did the pushes work? */
2062 if (btrfs_leaf_free_space(root
, l
) >= data_size
)
2066 if (!path
->nodes
[1]) {
2067 ret
= insert_new_root(trans
, root
, path
, 1);
2074 slot
= path
->slots
[0];
2075 nritems
= btrfs_header_nritems(l
);
2076 mid
= (nritems
+ 1) / 2;
2080 leaf_space_used(l
, mid
, nritems
- mid
) + data_size
>
2081 BTRFS_LEAF_DATA_SIZE(root
)) {
2082 if (slot
>= nritems
) {
2086 if (mid
!= nritems
&&
2087 leaf_space_used(l
, mid
, nritems
- mid
) +
2088 data_size
> BTRFS_LEAF_DATA_SIZE(root
)) {
2094 if (leaf_space_used(l
, 0, mid
) + data_size
>
2095 BTRFS_LEAF_DATA_SIZE(root
)) {
2096 if (!extend
&& data_size
&& slot
== 0) {
2098 } else if ((extend
|| !data_size
) && slot
== 0) {
2102 if (mid
!= nritems
&&
2103 leaf_space_used(l
, mid
, nritems
- mid
) +
2104 data_size
> BTRFS_LEAF_DATA_SIZE(root
)) {
2112 btrfs_cpu_key_to_disk(&disk_key
, ins_key
);
2114 btrfs_item_key(l
, &disk_key
, mid
);
2116 right
= btrfs_alloc_free_block(trans
, root
, root
->leafsize
,
2117 root
->root_key
.objectid
,
2118 &disk_key
, 0, l
->start
, 0);
2119 if (IS_ERR(right
)) {
2121 return PTR_ERR(right
);
2124 memset_extent_buffer(right
, 0, 0, sizeof(struct btrfs_header
));
2125 btrfs_set_header_bytenr(right
, right
->start
);
2126 btrfs_set_header_generation(right
, trans
->transid
);
2127 btrfs_set_header_backref_rev(right
, BTRFS_MIXED_BACKREF_REV
);
2128 btrfs_set_header_owner(right
, root
->root_key
.objectid
);
2129 btrfs_set_header_level(right
, 0);
2130 write_extent_buffer(right
, root
->fs_info
->fsid
,
2131 btrfs_header_fsid(), BTRFS_FSID_SIZE
);
2133 write_extent_buffer(right
, root
->fs_info
->chunk_tree_uuid
,
2134 btrfs_header_chunk_tree_uuid(right
),
2139 btrfs_set_header_nritems(right
, 0);
2140 wret
= insert_ptr(trans
, root
, path
,
2141 &disk_key
, right
->start
,
2142 path
->slots
[1] + 1, 1);
2146 free_extent_buffer(path
->nodes
[0]);
2147 path
->nodes
[0] = right
;
2149 path
->slots
[1] += 1;
2151 btrfs_set_header_nritems(right
, 0);
2152 wret
= insert_ptr(trans
, root
, path
,
2158 free_extent_buffer(path
->nodes
[0]);
2159 path
->nodes
[0] = right
;
2161 if (path
->slots
[1] == 0) {
2162 btrfs_fixup_low_keys(root
, path
,
2166 btrfs_mark_buffer_dirty(right
);
2170 ret
= copy_for_split(trans
, root
, path
, l
, right
, slot
, mid
, nritems
);
2174 BUG_ON(num_doubles
!= 0);
2183 * This function splits a single item into two items,
2184 * giving 'new_key' to the new item and splitting the
2185 * old one at split_offset (from the start of the item).
2187 * The path may be released by this operation. After
2188 * the split, the path is pointing to the old item. The
2189 * new item is going to be in the same node as the old one.
2191 * Note, the item being split must be smaller enough to live alone on
2192 * a tree block with room for one extra struct btrfs_item
2194 * This allows us to split the item in place, keeping a lock on the
2195 * leaf the entire time.
2197 int btrfs_split_item(struct btrfs_trans_handle
*trans
,
2198 struct btrfs_root
*root
,
2199 struct btrfs_path
*path
,
2200 struct btrfs_key
*new_key
,
2201 unsigned long split_offset
)
2204 struct extent_buffer
*leaf
;
2205 struct btrfs_key orig_key
;
2206 struct btrfs_item
*item
;
2207 struct btrfs_item
*new_item
;
2212 struct btrfs_disk_key disk_key
;
2215 leaf
= path
->nodes
[0];
2216 btrfs_item_key_to_cpu(leaf
, &orig_key
, path
->slots
[0]);
2217 if (btrfs_leaf_free_space(root
, leaf
) >= sizeof(struct btrfs_item
))
2220 item_size
= btrfs_item_size_nr(leaf
, path
->slots
[0]);
2221 btrfs_release_path(path
);
2223 path
->search_for_split
= 1;
2225 ret
= btrfs_search_slot(trans
, root
, &orig_key
, path
, 0, 1);
2226 path
->search_for_split
= 0;
2228 /* if our item isn't there or got smaller, return now */
2229 if (ret
!= 0 || item_size
!= btrfs_item_size_nr(path
->nodes
[0],
2234 ret
= split_leaf(trans
, root
, &orig_key
, path
, 0, 0);
2237 BUG_ON(btrfs_leaf_free_space(root
, leaf
) < sizeof(struct btrfs_item
));
2238 leaf
= path
->nodes
[0];
2241 item
= btrfs_item_nr(path
->slots
[0]);
2242 orig_offset
= btrfs_item_offset(leaf
, item
);
2243 item_size
= btrfs_item_size(leaf
, item
);
2246 buf
= kmalloc(item_size
, GFP_NOFS
);
2248 read_extent_buffer(leaf
, buf
, btrfs_item_ptr_offset(leaf
,
2249 path
->slots
[0]), item_size
);
2250 slot
= path
->slots
[0] + 1;
2251 leaf
= path
->nodes
[0];
2253 nritems
= btrfs_header_nritems(leaf
);
2255 if (slot
!= nritems
) {
2256 /* shift the items */
2257 memmove_extent_buffer(leaf
, btrfs_item_nr_offset(slot
+ 1),
2258 btrfs_item_nr_offset(slot
),
2259 (nritems
- slot
) * sizeof(struct btrfs_item
));
2263 btrfs_cpu_key_to_disk(&disk_key
, new_key
);
2264 btrfs_set_item_key(leaf
, &disk_key
, slot
);
2266 new_item
= btrfs_item_nr(slot
);
2268 btrfs_set_item_offset(leaf
, new_item
, orig_offset
);
2269 btrfs_set_item_size(leaf
, new_item
, item_size
- split_offset
);
2271 btrfs_set_item_offset(leaf
, item
,
2272 orig_offset
+ item_size
- split_offset
);
2273 btrfs_set_item_size(leaf
, item
, split_offset
);
2275 btrfs_set_header_nritems(leaf
, nritems
+ 1);
2277 /* write the data for the start of the original item */
2278 write_extent_buffer(leaf
, buf
,
2279 btrfs_item_ptr_offset(leaf
, path
->slots
[0]),
2282 /* write the data for the new item */
2283 write_extent_buffer(leaf
, buf
+ split_offset
,
2284 btrfs_item_ptr_offset(leaf
, slot
),
2285 item_size
- split_offset
);
2286 btrfs_mark_buffer_dirty(leaf
);
2289 if (btrfs_leaf_free_space(root
, leaf
) < 0) {
2290 btrfs_print_leaf(root
, leaf
);
2297 int btrfs_truncate_item(struct btrfs_trans_handle
*trans
,
2298 struct btrfs_root
*root
,
2299 struct btrfs_path
*path
,
2300 u32 new_size
, int from_end
)
2304 struct extent_buffer
*leaf
;
2305 struct btrfs_item
*item
;
2307 unsigned int data_end
;
2308 unsigned int old_data_start
;
2309 unsigned int old_size
;
2310 unsigned int size_diff
;
2313 leaf
= path
->nodes
[0];
2314 slot
= path
->slots
[0];
2316 old_size
= btrfs_item_size_nr(leaf
, slot
);
2317 if (old_size
== new_size
)
2320 nritems
= btrfs_header_nritems(leaf
);
2321 data_end
= leaf_data_end(root
, leaf
);
2323 old_data_start
= btrfs_item_offset_nr(leaf
, slot
);
2325 size_diff
= old_size
- new_size
;
2328 BUG_ON(slot
>= nritems
);
2331 * item0..itemN ... dataN.offset..dataN.size .. data0.size
2333 /* first correct the data pointers */
2334 for (i
= slot
; i
< nritems
; i
++) {
2336 item
= btrfs_item_nr(i
);
2337 ioff
= btrfs_item_offset(leaf
, item
);
2338 btrfs_set_item_offset(leaf
, item
, ioff
+ size_diff
);
2341 /* shift the data */
2343 memmove_extent_buffer(leaf
, btrfs_leaf_data(leaf
) +
2344 data_end
+ size_diff
, btrfs_leaf_data(leaf
) +
2345 data_end
, old_data_start
+ new_size
- data_end
);
2347 struct btrfs_disk_key disk_key
;
2350 btrfs_item_key(leaf
, &disk_key
, slot
);
2352 if (btrfs_disk_key_type(&disk_key
) == BTRFS_EXTENT_DATA_KEY
) {
2354 struct btrfs_file_extent_item
*fi
;
2356 fi
= btrfs_item_ptr(leaf
, slot
,
2357 struct btrfs_file_extent_item
);
2358 fi
= (struct btrfs_file_extent_item
*)(
2359 (unsigned long)fi
- size_diff
);
2361 if (btrfs_file_extent_type(leaf
, fi
) ==
2362 BTRFS_FILE_EXTENT_INLINE
) {
2363 ptr
= btrfs_item_ptr_offset(leaf
, slot
);
2364 memmove_extent_buffer(leaf
, ptr
,
2366 offsetof(struct btrfs_file_extent_item
,
2371 memmove_extent_buffer(leaf
, btrfs_leaf_data(leaf
) +
2372 data_end
+ size_diff
, btrfs_leaf_data(leaf
) +
2373 data_end
, old_data_start
- data_end
);
2375 offset
= btrfs_disk_key_offset(&disk_key
);
2376 btrfs_set_disk_key_offset(&disk_key
, offset
+ size_diff
);
2377 btrfs_set_item_key(leaf
, &disk_key
, slot
);
2379 btrfs_fixup_low_keys(root
, path
, &disk_key
, 1);
2382 item
= btrfs_item_nr(slot
);
2383 btrfs_set_item_size(leaf
, item
, new_size
);
2384 btrfs_mark_buffer_dirty(leaf
);
2387 if (btrfs_leaf_free_space(root
, leaf
) < 0) {
2388 btrfs_print_leaf(root
, leaf
);
2394 int btrfs_extend_item(struct btrfs_trans_handle
*trans
,
2395 struct btrfs_root
*root
, struct btrfs_path
*path
,
2400 struct extent_buffer
*leaf
;
2401 struct btrfs_item
*item
;
2403 unsigned int data_end
;
2404 unsigned int old_data
;
2405 unsigned int old_size
;
2408 leaf
= path
->nodes
[0];
2410 nritems
= btrfs_header_nritems(leaf
);
2411 data_end
= leaf_data_end(root
, leaf
);
2413 if (btrfs_leaf_free_space(root
, leaf
) < data_size
) {
2414 btrfs_print_leaf(root
, leaf
);
2417 slot
= path
->slots
[0];
2418 old_data
= btrfs_item_end_nr(leaf
, slot
);
2421 if (slot
>= nritems
) {
2422 btrfs_print_leaf(root
, leaf
);
2423 printk("slot %d too large, nritems %d\n", slot
, nritems
);
2428 * item0..itemN ... dataN.offset..dataN.size .. data0.size
2430 /* first correct the data pointers */
2431 for (i
= slot
; i
< nritems
; i
++) {
2433 item
= btrfs_item_nr(i
);
2434 ioff
= btrfs_item_offset(leaf
, item
);
2435 btrfs_set_item_offset(leaf
, item
, ioff
- data_size
);
2438 /* shift the data */
2439 memmove_extent_buffer(leaf
, btrfs_leaf_data(leaf
) +
2440 data_end
- data_size
, btrfs_leaf_data(leaf
) +
2441 data_end
, old_data
- data_end
);
2443 data_end
= old_data
;
2444 old_size
= btrfs_item_size_nr(leaf
, slot
);
2445 item
= btrfs_item_nr(slot
);
2446 btrfs_set_item_size(leaf
, item
, old_size
+ data_size
);
2447 btrfs_mark_buffer_dirty(leaf
);
2450 if (btrfs_leaf_free_space(root
, leaf
) < 0) {
2451 btrfs_print_leaf(root
, leaf
);
2458 * Given a key and some data, insert an item into the tree.
2459 * This does all the path init required, making room in the tree if needed.
2461 int btrfs_insert_empty_items(struct btrfs_trans_handle
*trans
,
2462 struct btrfs_root
*root
,
2463 struct btrfs_path
*path
,
2464 struct btrfs_key
*cpu_key
, u32
*data_size
,
2467 struct extent_buffer
*leaf
;
2468 struct btrfs_item
*item
;
2475 unsigned int data_end
;
2476 struct btrfs_disk_key disk_key
;
2478 for (i
= 0; i
< nr
; i
++) {
2479 total_data
+= data_size
[i
];
2482 /* create a root if there isn't one */
2486 total_size
= total_data
+ nr
* sizeof(struct btrfs_item
);
2487 ret
= btrfs_search_slot(trans
, root
, cpu_key
, path
, total_size
, 1);
2494 leaf
= path
->nodes
[0];
2496 nritems
= btrfs_header_nritems(leaf
);
2497 data_end
= leaf_data_end(root
, leaf
);
2499 if (btrfs_leaf_free_space(root
, leaf
) < total_size
) {
2500 btrfs_print_leaf(root
, leaf
);
2501 printk("not enough freespace need %u have %d\n",
2502 total_size
, btrfs_leaf_free_space(root
, leaf
));
2506 slot
= path
->slots
[0];
2509 if (slot
!= nritems
) {
2510 unsigned int old_data
= btrfs_item_end_nr(leaf
, slot
);
2512 if (old_data
< data_end
) {
2513 btrfs_print_leaf(root
, leaf
);
2514 printk("slot %d old_data %d data_end %d\n",
2515 slot
, old_data
, data_end
);
2519 * item0..itemN ... dataN.offset..dataN.size .. data0.size
2521 /* first correct the data pointers */
2522 for (i
= slot
; i
< nritems
; i
++) {
2525 item
= btrfs_item_nr(i
);
2526 ioff
= btrfs_item_offset(leaf
, item
);
2527 btrfs_set_item_offset(leaf
, item
, ioff
- total_data
);
2530 /* shift the items */
2531 memmove_extent_buffer(leaf
, btrfs_item_nr_offset(slot
+ nr
),
2532 btrfs_item_nr_offset(slot
),
2533 (nritems
- slot
) * sizeof(struct btrfs_item
));
2535 /* shift the data */
2536 memmove_extent_buffer(leaf
, btrfs_leaf_data(leaf
) +
2537 data_end
- total_data
, btrfs_leaf_data(leaf
) +
2538 data_end
, old_data
- data_end
);
2539 data_end
= old_data
;
2542 /* setup the item for the new data */
2543 for (i
= 0; i
< nr
; i
++) {
2544 btrfs_cpu_key_to_disk(&disk_key
, cpu_key
+ i
);
2545 btrfs_set_item_key(leaf
, &disk_key
, slot
+ i
);
2546 item
= btrfs_item_nr(slot
+ i
);
2547 btrfs_set_item_offset(leaf
, item
, data_end
- data_size
[i
]);
2548 data_end
-= data_size
[i
];
2549 btrfs_set_item_size(leaf
, item
, data_size
[i
]);
2551 btrfs_set_header_nritems(leaf
, nritems
+ nr
);
2552 btrfs_mark_buffer_dirty(leaf
);
2556 btrfs_cpu_key_to_disk(&disk_key
, cpu_key
);
2557 btrfs_fixup_low_keys(root
, path
, &disk_key
, 1);
2560 if (btrfs_leaf_free_space(root
, leaf
) < 0) {
2561 btrfs_print_leaf(root
, leaf
);
2570 * Given a key and some data, insert an item into the tree.
2571 * This does all the path init required, making room in the tree if needed.
2573 int btrfs_insert_item(struct btrfs_trans_handle
*trans
, struct btrfs_root
2574 *root
, struct btrfs_key
*cpu_key
, void *data
, u32
2578 struct btrfs_path
*path
;
2579 struct extent_buffer
*leaf
;
2582 path
= btrfs_alloc_path();
2584 ret
= btrfs_insert_empty_item(trans
, root
, path
, cpu_key
, data_size
);
2586 leaf
= path
->nodes
[0];
2587 ptr
= btrfs_item_ptr_offset(leaf
, path
->slots
[0]);
2588 write_extent_buffer(leaf
, data
, ptr
, data_size
);
2589 btrfs_mark_buffer_dirty(leaf
);
2591 btrfs_free_path(path
);
2596 * delete the pointer from a given node.
2598 * If the delete empties a node, the node is removed from the tree,
2599 * continuing all the way the root if required. The root is converted into
2600 * a leaf if all the nodes are emptied.
2602 int btrfs_del_ptr(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
2603 struct btrfs_path
*path
, int level
, int slot
)
2605 struct extent_buffer
*parent
= path
->nodes
[level
];
2609 nritems
= btrfs_header_nritems(parent
);
2610 if (slot
!= nritems
-1) {
2611 memmove_extent_buffer(parent
,
2612 btrfs_node_key_ptr_offset(slot
),
2613 btrfs_node_key_ptr_offset(slot
+ 1),
2614 sizeof(struct btrfs_key_ptr
) *
2615 (nritems
- slot
- 1));
2618 btrfs_set_header_nritems(parent
, nritems
);
2619 if (nritems
== 0 && parent
== root
->node
) {
2620 BUG_ON(btrfs_header_level(root
->node
) != 1);
2621 /* just turn the root into a leaf and break */
2622 btrfs_set_header_level(root
->node
, 0);
2623 } else if (slot
== 0) {
2624 struct btrfs_disk_key disk_key
;
2626 btrfs_node_key(parent
, &disk_key
, 0);
2627 btrfs_fixup_low_keys(root
, path
, &disk_key
, level
+ 1);
2629 btrfs_mark_buffer_dirty(parent
);
2634 * a helper function to delete the leaf pointed to by path->slots[1] and
2637 * This deletes the pointer in path->nodes[1] and frees the leaf
2638 * block extent. zero is returned if it all worked out, < 0 otherwise.
2640 * The path must have already been setup for deleting the leaf, including
2641 * all the proper balancing. path->nodes[1] must be locked.
2643 static noinline
int btrfs_del_leaf(struct btrfs_trans_handle
*trans
,
2644 struct btrfs_root
*root
,
2645 struct btrfs_path
*path
,
2646 struct extent_buffer
*leaf
)
2650 WARN_ON(btrfs_header_generation(leaf
) != trans
->transid
);
2651 ret
= btrfs_del_ptr(trans
, root
, path
, 1, path
->slots
[1]);
2655 ret
= btrfs_free_extent(trans
, root
, leaf
->start
, leaf
->len
,
2656 0, root
->root_key
.objectid
, 0, 0);
2661 * delete the item at the leaf level in path. If that empties
2662 * the leaf, remove it from the tree
2664 int btrfs_del_items(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
2665 struct btrfs_path
*path
, int slot
, int nr
)
2667 struct extent_buffer
*leaf
;
2668 struct btrfs_item
*item
;
2676 leaf
= path
->nodes
[0];
2677 last_off
= btrfs_item_offset_nr(leaf
, slot
+ nr
- 1);
2679 for (i
= 0; i
< nr
; i
++)
2680 dsize
+= btrfs_item_size_nr(leaf
, slot
+ i
);
2682 nritems
= btrfs_header_nritems(leaf
);
2684 if (slot
+ nr
!= nritems
) {
2685 int data_end
= leaf_data_end(root
, leaf
);
2687 memmove_extent_buffer(leaf
, btrfs_leaf_data(leaf
) +
2689 btrfs_leaf_data(leaf
) + data_end
,
2690 last_off
- data_end
);
2692 for (i
= slot
+ nr
; i
< nritems
; i
++) {
2695 item
= btrfs_item_nr(i
);
2696 ioff
= btrfs_item_offset(leaf
, item
);
2697 btrfs_set_item_offset(leaf
, item
, ioff
+ dsize
);
2700 memmove_extent_buffer(leaf
, btrfs_item_nr_offset(slot
),
2701 btrfs_item_nr_offset(slot
+ nr
),
2702 sizeof(struct btrfs_item
) *
2703 (nritems
- slot
- nr
));
2705 btrfs_set_header_nritems(leaf
, nritems
- nr
);
2708 /* delete the leaf if we've emptied it */
2710 if (leaf
== root
->node
) {
2711 btrfs_set_header_level(leaf
, 0);
2713 clean_tree_block(trans
, root
, leaf
);
2714 wait_on_tree_block_writeback(root
, leaf
);
2716 wret
= btrfs_del_leaf(trans
, root
, path
, leaf
);
2722 int used
= leaf_space_used(leaf
, 0, nritems
);
2724 struct btrfs_disk_key disk_key
;
2726 btrfs_item_key(leaf
, &disk_key
, 0);
2727 btrfs_fixup_low_keys(root
, path
, &disk_key
, 1);
2730 /* delete the leaf if it is mostly empty */
2731 if (used
< BTRFS_LEAF_DATA_SIZE(root
) / 4) {
2732 /* push_leaf_left fixes the path.
2733 * make sure the path still points to our leaf
2734 * for possible call to del_ptr below
2736 slot
= path
->slots
[1];
2737 extent_buffer_get(leaf
);
2739 wret
= push_leaf_left(trans
, root
, path
, 1, 1);
2740 if (wret
< 0 && wret
!= -ENOSPC
)
2743 if (path
->nodes
[0] == leaf
&&
2744 btrfs_header_nritems(leaf
)) {
2745 wret
= push_leaf_right(trans
, root
, path
, 1, 1);
2746 if (wret
< 0 && wret
!= -ENOSPC
)
2750 if (btrfs_header_nritems(leaf
) == 0) {
2751 clean_tree_block(trans
, root
, leaf
);
2752 wait_on_tree_block_writeback(root
, leaf
);
2754 path
->slots
[1] = slot
;
2755 ret
= btrfs_del_leaf(trans
, root
, path
, leaf
);
2757 free_extent_buffer(leaf
);
2760 btrfs_mark_buffer_dirty(leaf
);
2761 free_extent_buffer(leaf
);
2764 btrfs_mark_buffer_dirty(leaf
);
2771 * walk up the tree as far as required to find the previous leaf.
2772 * returns 0 if it found something or 1 if there are no lesser leaves.
2773 * returns < 0 on io errors.
2775 int btrfs_prev_leaf(struct btrfs_root
*root
, struct btrfs_path
*path
)
2779 struct extent_buffer
*c
;
2780 struct extent_buffer
*next
= NULL
;
2782 while(level
< BTRFS_MAX_LEVEL
) {
2783 if (!path
->nodes
[level
])
2786 slot
= path
->slots
[level
];
2787 c
= path
->nodes
[level
];
2790 if (level
== BTRFS_MAX_LEVEL
)
2796 next
= read_node_slot(root
, c
, slot
);
2797 if (!extent_buffer_uptodate(next
)) {
2799 return PTR_ERR(next
);
2804 path
->slots
[level
] = slot
;
2807 c
= path
->nodes
[level
];
2808 free_extent_buffer(c
);
2809 slot
= btrfs_header_nritems(next
);
2812 path
->nodes
[level
] = next
;
2813 path
->slots
[level
] = slot
;
2816 next
= read_node_slot(root
, next
, slot
);
2817 if (!extent_buffer_uptodate(next
)) {
2819 return PTR_ERR(next
);
2827 * walk up the tree as far as required to find the next leaf.
2828 * returns 0 if it found something or 1 if there are no greater leaves.
2829 * returns < 0 on io errors.
2831 int btrfs_next_leaf(struct btrfs_root
*root
, struct btrfs_path
*path
)
2835 struct extent_buffer
*c
;
2836 struct extent_buffer
*next
= NULL
;
2838 while(level
< BTRFS_MAX_LEVEL
) {
2839 if (!path
->nodes
[level
])
2842 slot
= path
->slots
[level
] + 1;
2843 c
= path
->nodes
[level
];
2844 if (slot
>= btrfs_header_nritems(c
)) {
2846 if (level
== BTRFS_MAX_LEVEL
)
2852 reada_for_search(root
, path
, level
, slot
, 0);
2854 next
= read_node_slot(root
, c
, slot
);
2855 if (!extent_buffer_uptodate(next
))
2859 path
->slots
[level
] = slot
;
2862 c
= path
->nodes
[level
];
2863 free_extent_buffer(c
);
2864 path
->nodes
[level
] = next
;
2865 path
->slots
[level
] = 0;
2869 reada_for_search(root
, path
, level
, 0, 0);
2870 next
= read_node_slot(root
, next
, 0);
2871 if (!extent_buffer_uptodate(next
))
2877 int btrfs_previous_item(struct btrfs_root
*root
,
2878 struct btrfs_path
*path
, u64 min_objectid
,
2881 struct btrfs_key found_key
;
2882 struct extent_buffer
*leaf
;
2886 if (path
->slots
[0] == 0) {
2887 ret
= btrfs_prev_leaf(root
, path
);
2893 leaf
= path
->nodes
[0];
2894 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2895 if (found_key
.type
== type
)
2902 * search in extent tree to find a previous Metadata/Data extent item with
2905 * returns 0 if something is found, 1 if nothing was found and < 0 on error
2907 int btrfs_previous_extent_item(struct btrfs_root
*root
,
2908 struct btrfs_path
*path
, u64 min_objectid
)
2910 struct btrfs_key found_key
;
2911 struct extent_buffer
*leaf
;
2916 if (path
->slots
[0] == 0) {
2917 ret
= btrfs_prev_leaf(root
, path
);
2923 leaf
= path
->nodes
[0];
2924 nritems
= btrfs_header_nritems(leaf
);
2927 if (path
->slots
[0] == nritems
)
2930 btrfs_item_key_to_cpu(leaf
, &found_key
, path
->slots
[0]);
2931 if (found_key
.objectid
< min_objectid
)
2933 if (found_key
.type
== BTRFS_EXTENT_ITEM_KEY
||
2934 found_key
.type
== BTRFS_METADATA_ITEM_KEY
)
2936 if (found_key
.objectid
== min_objectid
&&
2937 found_key
.type
< BTRFS_EXTENT_ITEM_KEY
)