2 * Copyright (C) 2011 STRATO. 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/vmalloc.h>
24 #include "transaction.h"
25 #include "delayed-ref.h"
28 struct extent_inode_elem
{
31 struct extent_inode_elem
*next
;
34 static int check_extent_in_eb(struct btrfs_key
*key
, struct extent_buffer
*eb
,
35 struct btrfs_file_extent_item
*fi
,
37 struct extent_inode_elem
**eie
)
40 struct extent_inode_elem
*e
;
42 if (!btrfs_file_extent_compression(eb
, fi
) &&
43 !btrfs_file_extent_encryption(eb
, fi
) &&
44 !btrfs_file_extent_other_encoding(eb
, fi
)) {
48 data_offset
= btrfs_file_extent_offset(eb
, fi
);
49 data_len
= btrfs_file_extent_num_bytes(eb
, fi
);
51 if (extent_item_pos
< data_offset
||
52 extent_item_pos
>= data_offset
+ data_len
)
54 offset
= extent_item_pos
- data_offset
;
57 e
= kmalloc(sizeof(*e
), GFP_NOFS
);
62 e
->inum
= key
->objectid
;
63 e
->offset
= key
->offset
+ offset
;
69 static int find_extent_in_eb(struct extent_buffer
*eb
, u64 wanted_disk_byte
,
71 struct extent_inode_elem
**eie
)
75 struct btrfs_file_extent_item
*fi
;
82 * from the shared data ref, we only have the leaf but we need
83 * the key. thus, we must look into all items and see that we
84 * find one (some) with a reference to our extent item.
86 nritems
= btrfs_header_nritems(eb
);
87 for (slot
= 0; slot
< nritems
; ++slot
) {
88 btrfs_item_key_to_cpu(eb
, &key
, slot
);
89 if (key
.type
!= BTRFS_EXTENT_DATA_KEY
)
91 fi
= btrfs_item_ptr(eb
, slot
, struct btrfs_file_extent_item
);
92 extent_type
= btrfs_file_extent_type(eb
, fi
);
93 if (extent_type
== BTRFS_FILE_EXTENT_INLINE
)
95 /* don't skip BTRFS_FILE_EXTENT_PREALLOC, we can handle that */
96 disk_byte
= btrfs_file_extent_disk_bytenr(eb
, fi
);
97 if (disk_byte
!= wanted_disk_byte
)
100 ret
= check_extent_in_eb(&key
, eb
, fi
, extent_item_pos
, eie
);
109 * this structure records all encountered refs on the way up to the root
111 struct __prelim_ref
{
112 struct list_head list
;
114 struct btrfs_key key_for_search
;
117 struct extent_inode_elem
*inode_list
;
119 u64 wanted_disk_byte
;
122 static struct kmem_cache
*btrfs_prelim_ref_cache
;
124 int __init
btrfs_prelim_ref_init(void)
126 btrfs_prelim_ref_cache
= kmem_cache_create("btrfs_prelim_ref",
127 sizeof(struct __prelim_ref
),
129 SLAB_RECLAIM_ACCOUNT
| SLAB_MEM_SPREAD
,
131 if (!btrfs_prelim_ref_cache
)
136 void btrfs_prelim_ref_exit(void)
138 if (btrfs_prelim_ref_cache
)
139 kmem_cache_destroy(btrfs_prelim_ref_cache
);
143 * the rules for all callers of this function are:
144 * - obtaining the parent is the goal
145 * - if you add a key, you must know that it is a correct key
146 * - if you cannot add the parent or a correct key, then we will look into the
147 * block later to set a correct key
151 * backref type | shared | indirect | shared | indirect
152 * information | tree | tree | data | data
153 * --------------------+--------+----------+--------+----------
154 * parent logical | y | - | - | -
155 * key to resolve | - | y | y | y
156 * tree block logical | - | - | - | -
157 * root for resolving | y | y | y | y
159 * - column 1: we've the parent -> done
160 * - column 2, 3, 4: we use the key to find the parent
162 * on disk refs (inline or keyed)
163 * ==============================
164 * backref type | shared | indirect | shared | indirect
165 * information | tree | tree | data | data
166 * --------------------+--------+----------+--------+----------
167 * parent logical | y | - | y | -
168 * key to resolve | - | - | - | y
169 * tree block logical | y | y | y | y
170 * root for resolving | - | y | y | y
172 * - column 1, 3: we've the parent -> done
173 * - column 2: we take the first key from the block to find the parent
174 * (see __add_missing_keys)
175 * - column 4: we use the key to find the parent
177 * additional information that's available but not required to find the parent
178 * block might help in merging entries to gain some speed.
181 static int __add_prelim_ref(struct list_head
*head
, u64 root_id
,
182 struct btrfs_key
*key
, int level
,
183 u64 parent
, u64 wanted_disk_byte
, int count
,
186 struct __prelim_ref
*ref
;
188 if (root_id
== BTRFS_DATA_RELOC_TREE_OBJECTID
)
191 ref
= kmem_cache_alloc(btrfs_prelim_ref_cache
, gfp_mask
);
195 ref
->root_id
= root_id
;
197 ref
->key_for_search
= *key
;
199 memset(&ref
->key_for_search
, 0, sizeof(ref
->key_for_search
));
201 ref
->inode_list
= NULL
;
204 ref
->parent
= parent
;
205 ref
->wanted_disk_byte
= wanted_disk_byte
;
206 list_add_tail(&ref
->list
, head
);
211 static int add_all_parents(struct btrfs_root
*root
, struct btrfs_path
*path
,
212 struct ulist
*parents
, struct __prelim_ref
*ref
,
213 int level
, u64 time_seq
, const u64
*extent_item_pos
)
217 struct extent_buffer
*eb
;
218 struct btrfs_key key
;
219 struct btrfs_key
*key_for_search
= &ref
->key_for_search
;
220 struct btrfs_file_extent_item
*fi
;
221 struct extent_inode_elem
*eie
= NULL
, *old
= NULL
;
223 u64 wanted_disk_byte
= ref
->wanted_disk_byte
;
227 eb
= path
->nodes
[level
];
228 ret
= ulist_add(parents
, eb
->start
, 0, GFP_NOFS
);
235 * We normally enter this function with the path already pointing to
236 * the first item to check. But sometimes, we may enter it with
237 * slot==nritems. In that case, go to the next leaf before we continue.
239 if (path
->slots
[0] >= btrfs_header_nritems(path
->nodes
[0]))
240 ret
= btrfs_next_old_leaf(root
, path
, time_seq
);
242 while (!ret
&& count
< ref
->count
) {
244 slot
= path
->slots
[0];
246 btrfs_item_key_to_cpu(eb
, &key
, slot
);
248 if (key
.objectid
!= key_for_search
->objectid
||
249 key
.type
!= BTRFS_EXTENT_DATA_KEY
)
252 fi
= btrfs_item_ptr(eb
, slot
, struct btrfs_file_extent_item
);
253 disk_byte
= btrfs_file_extent_disk_bytenr(eb
, fi
);
255 if (disk_byte
== wanted_disk_byte
) {
259 if (extent_item_pos
) {
260 ret
= check_extent_in_eb(&key
, eb
, fi
,
268 ret
= ulist_add_merge(parents
, eb
->start
,
270 (u64
*)&old
, GFP_NOFS
);
273 if (!ret
&& extent_item_pos
) {
280 ret
= btrfs_next_old_item(root
, path
, time_seq
);
289 * resolve an indirect backref in the form (root_id, key, level)
290 * to a logical address
292 static int __resolve_indirect_ref(struct btrfs_fs_info
*fs_info
,
293 struct btrfs_path
*path
, u64 time_seq
,
294 struct __prelim_ref
*ref
,
295 struct ulist
*parents
,
296 const u64
*extent_item_pos
)
298 struct btrfs_root
*root
;
299 struct btrfs_key root_key
;
300 struct extent_buffer
*eb
;
303 int level
= ref
->level
;
306 root_key
.objectid
= ref
->root_id
;
307 root_key
.type
= BTRFS_ROOT_ITEM_KEY
;
308 root_key
.offset
= (u64
)-1;
310 index
= srcu_read_lock(&fs_info
->subvol_srcu
);
312 root
= btrfs_read_fs_root_no_name(fs_info
, &root_key
);
314 srcu_read_unlock(&fs_info
->subvol_srcu
, index
);
319 root_level
= btrfs_old_root_level(root
, time_seq
);
321 if (root_level
+ 1 == level
) {
322 srcu_read_unlock(&fs_info
->subvol_srcu
, index
);
326 path
->lowest_level
= level
;
327 ret
= btrfs_search_old_slot(root
, &ref
->key_for_search
, path
, time_seq
);
329 /* root node has been locked, we can release @subvol_srcu safely here */
330 srcu_read_unlock(&fs_info
->subvol_srcu
, index
);
332 pr_debug("search slot in root %llu (level %d, ref count %d) returned "
333 "%d for key (%llu %u %llu)\n",
334 ref
->root_id
, level
, ref
->count
, ret
,
335 ref
->key_for_search
.objectid
, ref
->key_for_search
.type
,
336 ref
->key_for_search
.offset
);
340 eb
= path
->nodes
[level
];
342 if (WARN_ON(!level
)) {
347 eb
= path
->nodes
[level
];
350 ret
= add_all_parents(root
, path
, parents
, ref
, level
, time_seq
,
353 path
->lowest_level
= 0;
354 btrfs_release_path(path
);
359 * resolve all indirect backrefs from the list
361 static int __resolve_indirect_refs(struct btrfs_fs_info
*fs_info
,
362 struct btrfs_path
*path
, u64 time_seq
,
363 struct list_head
*head
,
364 const u64
*extent_item_pos
)
368 struct __prelim_ref
*ref
;
369 struct __prelim_ref
*ref_safe
;
370 struct __prelim_ref
*new_ref
;
371 struct ulist
*parents
;
372 struct ulist_node
*node
;
373 struct ulist_iterator uiter
;
375 parents
= ulist_alloc(GFP_NOFS
);
380 * _safe allows us to insert directly after the current item without
381 * iterating over the newly inserted items.
382 * we're also allowed to re-assign ref during iteration.
384 list_for_each_entry_safe(ref
, ref_safe
, head
, list
) {
385 if (ref
->parent
) /* already direct */
389 err
= __resolve_indirect_ref(fs_info
, path
, time_seq
, ref
,
390 parents
, extent_item_pos
);
392 * we can only tolerate ENOENT,otherwise,we should catch error
393 * and return directly.
395 if (err
== -ENOENT
) {
402 /* we put the first parent into the ref at hand */
403 ULIST_ITER_INIT(&uiter
);
404 node
= ulist_next(parents
, &uiter
);
405 ref
->parent
= node
? node
->val
: 0;
406 ref
->inode_list
= node
?
407 (struct extent_inode_elem
*)(uintptr_t)node
->aux
: NULL
;
409 /* additional parents require new refs being added here */
410 while ((node
= ulist_next(parents
, &uiter
))) {
411 new_ref
= kmem_cache_alloc(btrfs_prelim_ref_cache
,
417 memcpy(new_ref
, ref
, sizeof(*ref
));
418 new_ref
->parent
= node
->val
;
419 new_ref
->inode_list
= (struct extent_inode_elem
*)
420 (uintptr_t)node
->aux
;
421 list_add(&new_ref
->list
, &ref
->list
);
423 ulist_reinit(parents
);
430 static inline int ref_for_same_block(struct __prelim_ref
*ref1
,
431 struct __prelim_ref
*ref2
)
433 if (ref1
->level
!= ref2
->level
)
435 if (ref1
->root_id
!= ref2
->root_id
)
437 if (ref1
->key_for_search
.type
!= ref2
->key_for_search
.type
)
439 if (ref1
->key_for_search
.objectid
!= ref2
->key_for_search
.objectid
)
441 if (ref1
->key_for_search
.offset
!= ref2
->key_for_search
.offset
)
443 if (ref1
->parent
!= ref2
->parent
)
450 * read tree blocks and add keys where required.
452 static int __add_missing_keys(struct btrfs_fs_info
*fs_info
,
453 struct list_head
*head
)
455 struct list_head
*pos
;
456 struct extent_buffer
*eb
;
458 list_for_each(pos
, head
) {
459 struct __prelim_ref
*ref
;
460 ref
= list_entry(pos
, struct __prelim_ref
, list
);
464 if (ref
->key_for_search
.type
)
466 BUG_ON(!ref
->wanted_disk_byte
);
467 eb
= read_tree_block(fs_info
->tree_root
, ref
->wanted_disk_byte
,
468 fs_info
->tree_root
->leafsize
, 0);
469 if (!eb
|| !extent_buffer_uptodate(eb
)) {
470 free_extent_buffer(eb
);
473 btrfs_tree_read_lock(eb
);
474 if (btrfs_header_level(eb
) == 0)
475 btrfs_item_key_to_cpu(eb
, &ref
->key_for_search
, 0);
477 btrfs_node_key_to_cpu(eb
, &ref
->key_for_search
, 0);
478 btrfs_tree_read_unlock(eb
);
479 free_extent_buffer(eb
);
485 * merge two lists of backrefs and adjust counts accordingly
487 * mode = 1: merge identical keys, if key is set
488 * FIXME: if we add more keys in __add_prelim_ref, we can merge more here.
489 * additionally, we could even add a key range for the blocks we
490 * looked into to merge even more (-> replace unresolved refs by those
492 * mode = 2: merge identical parents
494 static void __merge_refs(struct list_head
*head
, int mode
)
496 struct list_head
*pos1
;
498 list_for_each(pos1
, head
) {
499 struct list_head
*n2
;
500 struct list_head
*pos2
;
501 struct __prelim_ref
*ref1
;
503 ref1
= list_entry(pos1
, struct __prelim_ref
, list
);
505 for (pos2
= pos1
->next
, n2
= pos2
->next
; pos2
!= head
;
506 pos2
= n2
, n2
= pos2
->next
) {
507 struct __prelim_ref
*ref2
;
508 struct __prelim_ref
*xchg
;
509 struct extent_inode_elem
*eie
;
511 ref2
= list_entry(pos2
, struct __prelim_ref
, list
);
514 if (!ref_for_same_block(ref1
, ref2
))
516 if (!ref1
->parent
&& ref2
->parent
) {
522 if (ref1
->parent
!= ref2
->parent
)
526 eie
= ref1
->inode_list
;
527 while (eie
&& eie
->next
)
530 eie
->next
= ref2
->inode_list
;
532 ref1
->inode_list
= ref2
->inode_list
;
533 ref1
->count
+= ref2
->count
;
535 list_del(&ref2
->list
);
536 kmem_cache_free(btrfs_prelim_ref_cache
, ref2
);
543 * add all currently queued delayed refs from this head whose seq nr is
544 * smaller or equal that seq to the list
546 static int __add_delayed_refs(struct btrfs_delayed_ref_head
*head
, u64 seq
,
547 struct list_head
*prefs
)
549 struct btrfs_delayed_extent_op
*extent_op
= head
->extent_op
;
550 struct rb_node
*n
= &head
->node
.rb_node
;
551 struct btrfs_key key
;
552 struct btrfs_key op_key
= {0};
556 if (extent_op
&& extent_op
->update_key
)
557 btrfs_disk_key_to_cpu(&op_key
, &extent_op
->key
);
559 spin_lock(&head
->lock
);
560 n
= rb_first(&head
->ref_root
);
562 struct btrfs_delayed_ref_node
*node
;
563 node
= rb_entry(n
, struct btrfs_delayed_ref_node
,
569 switch (node
->action
) {
570 case BTRFS_ADD_DELAYED_EXTENT
:
571 case BTRFS_UPDATE_DELAYED_HEAD
:
574 case BTRFS_ADD_DELAYED_REF
:
577 case BTRFS_DROP_DELAYED_REF
:
583 switch (node
->type
) {
584 case BTRFS_TREE_BLOCK_REF_KEY
: {
585 struct btrfs_delayed_tree_ref
*ref
;
587 ref
= btrfs_delayed_node_to_tree_ref(node
);
588 ret
= __add_prelim_ref(prefs
, ref
->root
, &op_key
,
589 ref
->level
+ 1, 0, node
->bytenr
,
590 node
->ref_mod
* sgn
, GFP_ATOMIC
);
593 case BTRFS_SHARED_BLOCK_REF_KEY
: {
594 struct btrfs_delayed_tree_ref
*ref
;
596 ref
= btrfs_delayed_node_to_tree_ref(node
);
597 ret
= __add_prelim_ref(prefs
, ref
->root
, NULL
,
598 ref
->level
+ 1, ref
->parent
,
600 node
->ref_mod
* sgn
, GFP_ATOMIC
);
603 case BTRFS_EXTENT_DATA_REF_KEY
: {
604 struct btrfs_delayed_data_ref
*ref
;
605 ref
= btrfs_delayed_node_to_data_ref(node
);
607 key
.objectid
= ref
->objectid
;
608 key
.type
= BTRFS_EXTENT_DATA_KEY
;
609 key
.offset
= ref
->offset
;
610 ret
= __add_prelim_ref(prefs
, ref
->root
, &key
, 0, 0,
612 node
->ref_mod
* sgn
, GFP_ATOMIC
);
615 case BTRFS_SHARED_DATA_REF_KEY
: {
616 struct btrfs_delayed_data_ref
*ref
;
618 ref
= btrfs_delayed_node_to_data_ref(node
);
620 key
.objectid
= ref
->objectid
;
621 key
.type
= BTRFS_EXTENT_DATA_KEY
;
622 key
.offset
= ref
->offset
;
623 ret
= __add_prelim_ref(prefs
, ref
->root
, &key
, 0,
624 ref
->parent
, node
->bytenr
,
625 node
->ref_mod
* sgn
, GFP_ATOMIC
);
634 spin_unlock(&head
->lock
);
639 * add all inline backrefs for bytenr to the list
641 static int __add_inline_refs(struct btrfs_fs_info
*fs_info
,
642 struct btrfs_path
*path
, u64 bytenr
,
643 int *info_level
, struct list_head
*prefs
)
647 struct extent_buffer
*leaf
;
648 struct btrfs_key key
;
649 struct btrfs_key found_key
;
652 struct btrfs_extent_item
*ei
;
657 * enumerate all inline refs
659 leaf
= path
->nodes
[0];
660 slot
= path
->slots
[0];
662 item_size
= btrfs_item_size_nr(leaf
, slot
);
663 BUG_ON(item_size
< sizeof(*ei
));
665 ei
= btrfs_item_ptr(leaf
, slot
, struct btrfs_extent_item
);
666 flags
= btrfs_extent_flags(leaf
, ei
);
667 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
669 ptr
= (unsigned long)(ei
+ 1);
670 end
= (unsigned long)ei
+ item_size
;
672 if (found_key
.type
== BTRFS_EXTENT_ITEM_KEY
&&
673 flags
& BTRFS_EXTENT_FLAG_TREE_BLOCK
) {
674 struct btrfs_tree_block_info
*info
;
676 info
= (struct btrfs_tree_block_info
*)ptr
;
677 *info_level
= btrfs_tree_block_level(leaf
, info
);
678 ptr
+= sizeof(struct btrfs_tree_block_info
);
680 } else if (found_key
.type
== BTRFS_METADATA_ITEM_KEY
) {
681 *info_level
= found_key
.offset
;
683 BUG_ON(!(flags
& BTRFS_EXTENT_FLAG_DATA
));
687 struct btrfs_extent_inline_ref
*iref
;
691 iref
= (struct btrfs_extent_inline_ref
*)ptr
;
692 type
= btrfs_extent_inline_ref_type(leaf
, iref
);
693 offset
= btrfs_extent_inline_ref_offset(leaf
, iref
);
696 case BTRFS_SHARED_BLOCK_REF_KEY
:
697 ret
= __add_prelim_ref(prefs
, 0, NULL
,
698 *info_level
+ 1, offset
,
699 bytenr
, 1, GFP_NOFS
);
701 case BTRFS_SHARED_DATA_REF_KEY
: {
702 struct btrfs_shared_data_ref
*sdref
;
705 sdref
= (struct btrfs_shared_data_ref
*)(iref
+ 1);
706 count
= btrfs_shared_data_ref_count(leaf
, sdref
);
707 ret
= __add_prelim_ref(prefs
, 0, NULL
, 0, offset
,
708 bytenr
, count
, GFP_NOFS
);
711 case BTRFS_TREE_BLOCK_REF_KEY
:
712 ret
= __add_prelim_ref(prefs
, offset
, NULL
,
714 bytenr
, 1, GFP_NOFS
);
716 case BTRFS_EXTENT_DATA_REF_KEY
: {
717 struct btrfs_extent_data_ref
*dref
;
721 dref
= (struct btrfs_extent_data_ref
*)(&iref
->offset
);
722 count
= btrfs_extent_data_ref_count(leaf
, dref
);
723 key
.objectid
= btrfs_extent_data_ref_objectid(leaf
,
725 key
.type
= BTRFS_EXTENT_DATA_KEY
;
726 key
.offset
= btrfs_extent_data_ref_offset(leaf
, dref
);
727 root
= btrfs_extent_data_ref_root(leaf
, dref
);
728 ret
= __add_prelim_ref(prefs
, root
, &key
, 0, 0,
729 bytenr
, count
, GFP_NOFS
);
737 ptr
+= btrfs_extent_inline_ref_size(type
);
744 * add all non-inline backrefs for bytenr to the list
746 static int __add_keyed_refs(struct btrfs_fs_info
*fs_info
,
747 struct btrfs_path
*path
, u64 bytenr
,
748 int info_level
, struct list_head
*prefs
)
750 struct btrfs_root
*extent_root
= fs_info
->extent_root
;
753 struct extent_buffer
*leaf
;
754 struct btrfs_key key
;
757 ret
= btrfs_next_item(extent_root
, path
);
765 slot
= path
->slots
[0];
766 leaf
= path
->nodes
[0];
767 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
769 if (key
.objectid
!= bytenr
)
771 if (key
.type
< BTRFS_TREE_BLOCK_REF_KEY
)
773 if (key
.type
> BTRFS_SHARED_DATA_REF_KEY
)
777 case BTRFS_SHARED_BLOCK_REF_KEY
:
778 ret
= __add_prelim_ref(prefs
, 0, NULL
,
779 info_level
+ 1, key
.offset
,
780 bytenr
, 1, GFP_NOFS
);
782 case BTRFS_SHARED_DATA_REF_KEY
: {
783 struct btrfs_shared_data_ref
*sdref
;
786 sdref
= btrfs_item_ptr(leaf
, slot
,
787 struct btrfs_shared_data_ref
);
788 count
= btrfs_shared_data_ref_count(leaf
, sdref
);
789 ret
= __add_prelim_ref(prefs
, 0, NULL
, 0, key
.offset
,
790 bytenr
, count
, GFP_NOFS
);
793 case BTRFS_TREE_BLOCK_REF_KEY
:
794 ret
= __add_prelim_ref(prefs
, key
.offset
, NULL
,
796 bytenr
, 1, GFP_NOFS
);
798 case BTRFS_EXTENT_DATA_REF_KEY
: {
799 struct btrfs_extent_data_ref
*dref
;
803 dref
= btrfs_item_ptr(leaf
, slot
,
804 struct btrfs_extent_data_ref
);
805 count
= btrfs_extent_data_ref_count(leaf
, dref
);
806 key
.objectid
= btrfs_extent_data_ref_objectid(leaf
,
808 key
.type
= BTRFS_EXTENT_DATA_KEY
;
809 key
.offset
= btrfs_extent_data_ref_offset(leaf
, dref
);
810 root
= btrfs_extent_data_ref_root(leaf
, dref
);
811 ret
= __add_prelim_ref(prefs
, root
, &key
, 0, 0,
812 bytenr
, count
, GFP_NOFS
);
827 * this adds all existing backrefs (inline backrefs, backrefs and delayed
828 * refs) for the given bytenr to the refs list, merges duplicates and resolves
829 * indirect refs to their parent bytenr.
830 * When roots are found, they're added to the roots list
832 * FIXME some caching might speed things up
834 static int find_parent_nodes(struct btrfs_trans_handle
*trans
,
835 struct btrfs_fs_info
*fs_info
, u64 bytenr
,
836 u64 time_seq
, struct ulist
*refs
,
837 struct ulist
*roots
, const u64
*extent_item_pos
)
839 struct btrfs_key key
;
840 struct btrfs_path
*path
;
841 struct btrfs_delayed_ref_root
*delayed_refs
= NULL
;
842 struct btrfs_delayed_ref_head
*head
;
845 struct list_head prefs_delayed
;
846 struct list_head prefs
;
847 struct __prelim_ref
*ref
;
849 INIT_LIST_HEAD(&prefs
);
850 INIT_LIST_HEAD(&prefs_delayed
);
852 key
.objectid
= bytenr
;
853 key
.offset
= (u64
)-1;
854 if (btrfs_fs_incompat(fs_info
, SKINNY_METADATA
))
855 key
.type
= BTRFS_METADATA_ITEM_KEY
;
857 key
.type
= BTRFS_EXTENT_ITEM_KEY
;
859 path
= btrfs_alloc_path();
863 path
->search_commit_root
= 1;
866 * grab both a lock on the path and a lock on the delayed ref head.
867 * We need both to get a consistent picture of how the refs look
868 * at a specified point in time
873 ret
= btrfs_search_slot(trans
, fs_info
->extent_root
, &key
, path
, 0, 0);
880 * look if there are updates for this ref queued and lock the
883 delayed_refs
= &trans
->transaction
->delayed_refs
;
884 spin_lock(&delayed_refs
->lock
);
885 head
= btrfs_find_delayed_ref_head(trans
, bytenr
);
887 if (!mutex_trylock(&head
->mutex
)) {
888 atomic_inc(&head
->node
.refs
);
889 spin_unlock(&delayed_refs
->lock
);
891 btrfs_release_path(path
);
894 * Mutex was contended, block until it's
895 * released and try again
897 mutex_lock(&head
->mutex
);
898 mutex_unlock(&head
->mutex
);
899 btrfs_put_delayed_ref(&head
->node
);
902 spin_unlock(&delayed_refs
->lock
);
903 ret
= __add_delayed_refs(head
, time_seq
,
905 mutex_unlock(&head
->mutex
);
909 spin_unlock(&delayed_refs
->lock
);
913 if (path
->slots
[0]) {
914 struct extent_buffer
*leaf
;
918 leaf
= path
->nodes
[0];
919 slot
= path
->slots
[0];
920 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
921 if (key
.objectid
== bytenr
&&
922 (key
.type
== BTRFS_EXTENT_ITEM_KEY
||
923 key
.type
== BTRFS_METADATA_ITEM_KEY
)) {
924 ret
= __add_inline_refs(fs_info
, path
, bytenr
,
925 &info_level
, &prefs
);
928 ret
= __add_keyed_refs(fs_info
, path
, bytenr
,
934 btrfs_release_path(path
);
936 list_splice_init(&prefs_delayed
, &prefs
);
938 ret
= __add_missing_keys(fs_info
, &prefs
);
942 __merge_refs(&prefs
, 1);
944 ret
= __resolve_indirect_refs(fs_info
, path
, time_seq
, &prefs
,
949 __merge_refs(&prefs
, 2);
951 while (!list_empty(&prefs
)) {
952 ref
= list_first_entry(&prefs
, struct __prelim_ref
, list
);
953 WARN_ON(ref
->count
< 0);
954 if (ref
->count
&& ref
->root_id
&& ref
->parent
== 0) {
955 /* no parent == root of tree */
956 ret
= ulist_add(roots
, ref
->root_id
, 0, GFP_NOFS
);
960 if (ref
->count
&& ref
->parent
) {
961 struct extent_inode_elem
*eie
= NULL
;
962 if (extent_item_pos
&& !ref
->inode_list
) {
964 struct extent_buffer
*eb
;
965 bsz
= btrfs_level_size(fs_info
->extent_root
,
967 eb
= read_tree_block(fs_info
->extent_root
,
968 ref
->parent
, bsz
, 0);
969 if (!eb
|| !extent_buffer_uptodate(eb
)) {
970 free_extent_buffer(eb
);
974 ret
= find_extent_in_eb(eb
, bytenr
,
975 *extent_item_pos
, &eie
);
976 free_extent_buffer(eb
);
979 ref
->inode_list
= eie
;
981 ret
= ulist_add_merge(refs
, ref
->parent
,
982 (uintptr_t)ref
->inode_list
,
983 (u64
*)&eie
, GFP_NOFS
);
986 if (!ret
&& extent_item_pos
) {
988 * we've recorded that parent, so we must extend
989 * its inode list here
994 eie
->next
= ref
->inode_list
;
997 list_del(&ref
->list
);
998 kmem_cache_free(btrfs_prelim_ref_cache
, ref
);
1002 btrfs_free_path(path
);
1003 while (!list_empty(&prefs
)) {
1004 ref
= list_first_entry(&prefs
, struct __prelim_ref
, list
);
1005 list_del(&ref
->list
);
1006 kmem_cache_free(btrfs_prelim_ref_cache
, ref
);
1008 while (!list_empty(&prefs_delayed
)) {
1009 ref
= list_first_entry(&prefs_delayed
, struct __prelim_ref
,
1011 list_del(&ref
->list
);
1012 kmem_cache_free(btrfs_prelim_ref_cache
, ref
);
1018 static void free_leaf_list(struct ulist
*blocks
)
1020 struct ulist_node
*node
= NULL
;
1021 struct extent_inode_elem
*eie
;
1022 struct extent_inode_elem
*eie_next
;
1023 struct ulist_iterator uiter
;
1025 ULIST_ITER_INIT(&uiter
);
1026 while ((node
= ulist_next(blocks
, &uiter
))) {
1029 eie
= (struct extent_inode_elem
*)(uintptr_t)node
->aux
;
1030 for (; eie
; eie
= eie_next
) {
1031 eie_next
= eie
->next
;
1041 * Finds all leafs with a reference to the specified combination of bytenr and
1042 * offset. key_list_head will point to a list of corresponding keys (caller must
1043 * free each list element). The leafs will be stored in the leafs ulist, which
1044 * must be freed with ulist_free.
1046 * returns 0 on success, <0 on error
1048 static int btrfs_find_all_leafs(struct btrfs_trans_handle
*trans
,
1049 struct btrfs_fs_info
*fs_info
, u64 bytenr
,
1050 u64 time_seq
, struct ulist
**leafs
,
1051 const u64
*extent_item_pos
)
1056 tmp
= ulist_alloc(GFP_NOFS
);
1059 *leafs
= ulist_alloc(GFP_NOFS
);
1065 ret
= find_parent_nodes(trans
, fs_info
, bytenr
,
1066 time_seq
, *leafs
, tmp
, extent_item_pos
);
1069 if (ret
< 0 && ret
!= -ENOENT
) {
1070 free_leaf_list(*leafs
);
1078 * walk all backrefs for a given extent to find all roots that reference this
1079 * extent. Walking a backref means finding all extents that reference this
1080 * extent and in turn walk the backrefs of those, too. Naturally this is a
1081 * recursive process, but here it is implemented in an iterative fashion: We
1082 * find all referencing extents for the extent in question and put them on a
1083 * list. In turn, we find all referencing extents for those, further appending
1084 * to the list. The way we iterate the list allows adding more elements after
1085 * the current while iterating. The process stops when we reach the end of the
1086 * list. Found roots are added to the roots list.
1088 * returns 0 on success, < 0 on error.
1090 int btrfs_find_all_roots(struct btrfs_trans_handle
*trans
,
1091 struct btrfs_fs_info
*fs_info
, u64 bytenr
,
1092 u64 time_seq
, struct ulist
**roots
)
1095 struct ulist_node
*node
= NULL
;
1096 struct ulist_iterator uiter
;
1099 tmp
= ulist_alloc(GFP_NOFS
);
1102 *roots
= ulist_alloc(GFP_NOFS
);
1108 ULIST_ITER_INIT(&uiter
);
1110 ret
= find_parent_nodes(trans
, fs_info
, bytenr
,
1111 time_seq
, tmp
, *roots
, NULL
);
1112 if (ret
< 0 && ret
!= -ENOENT
) {
1117 node
= ulist_next(tmp
, &uiter
);
1129 * this makes the path point to (inum INODE_ITEM ioff)
1131 int inode_item_info(u64 inum
, u64 ioff
, struct btrfs_root
*fs_root
,
1132 struct btrfs_path
*path
)
1134 struct btrfs_key key
;
1135 return btrfs_find_item(fs_root
, path
, inum
, ioff
,
1136 BTRFS_INODE_ITEM_KEY
, &key
);
1139 static int inode_ref_info(u64 inum
, u64 ioff
, struct btrfs_root
*fs_root
,
1140 struct btrfs_path
*path
,
1141 struct btrfs_key
*found_key
)
1143 return btrfs_find_item(fs_root
, path
, inum
, ioff
,
1144 BTRFS_INODE_REF_KEY
, found_key
);
1147 int btrfs_find_one_extref(struct btrfs_root
*root
, u64 inode_objectid
,
1148 u64 start_off
, struct btrfs_path
*path
,
1149 struct btrfs_inode_extref
**ret_extref
,
1153 struct btrfs_key key
;
1154 struct btrfs_key found_key
;
1155 struct btrfs_inode_extref
*extref
;
1156 struct extent_buffer
*leaf
;
1159 key
.objectid
= inode_objectid
;
1160 btrfs_set_key_type(&key
, BTRFS_INODE_EXTREF_KEY
);
1161 key
.offset
= start_off
;
1163 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1168 leaf
= path
->nodes
[0];
1169 slot
= path
->slots
[0];
1170 if (slot
>= btrfs_header_nritems(leaf
)) {
1172 * If the item at offset is not found,
1173 * btrfs_search_slot will point us to the slot
1174 * where it should be inserted. In our case
1175 * that will be the slot directly before the
1176 * next INODE_REF_KEY_V2 item. In the case
1177 * that we're pointing to the last slot in a
1178 * leaf, we must move one leaf over.
1180 ret
= btrfs_next_leaf(root
, path
);
1189 btrfs_item_key_to_cpu(leaf
, &found_key
, slot
);
1192 * Check that we're still looking at an extended ref key for
1193 * this particular objectid. If we have different
1194 * objectid or type then there are no more to be found
1195 * in the tree and we can exit.
1198 if (found_key
.objectid
!= inode_objectid
)
1200 if (btrfs_key_type(&found_key
) != BTRFS_INODE_EXTREF_KEY
)
1204 ptr
= btrfs_item_ptr_offset(leaf
, path
->slots
[0]);
1205 extref
= (struct btrfs_inode_extref
*)ptr
;
1206 *ret_extref
= extref
;
1208 *found_off
= found_key
.offset
;
1216 * this iterates to turn a name (from iref/extref) into a full filesystem path.
1217 * Elements of the path are separated by '/' and the path is guaranteed to be
1218 * 0-terminated. the path is only given within the current file system.
1219 * Therefore, it never starts with a '/'. the caller is responsible to provide
1220 * "size" bytes in "dest". the dest buffer will be filled backwards. finally,
1221 * the start point of the resulting string is returned. this pointer is within
1223 * in case the path buffer would overflow, the pointer is decremented further
1224 * as if output was written to the buffer, though no more output is actually
1225 * generated. that way, the caller can determine how much space would be
1226 * required for the path to fit into the buffer. in that case, the returned
1227 * value will be smaller than dest. callers must check this!
1229 char *btrfs_ref_to_path(struct btrfs_root
*fs_root
, struct btrfs_path
*path
,
1230 u32 name_len
, unsigned long name_off
,
1231 struct extent_buffer
*eb_in
, u64 parent
,
1232 char *dest
, u32 size
)
1237 s64 bytes_left
= ((s64
)size
) - 1;
1238 struct extent_buffer
*eb
= eb_in
;
1239 struct btrfs_key found_key
;
1240 int leave_spinning
= path
->leave_spinning
;
1241 struct btrfs_inode_ref
*iref
;
1243 if (bytes_left
>= 0)
1244 dest
[bytes_left
] = '\0';
1246 path
->leave_spinning
= 1;
1248 bytes_left
-= name_len
;
1249 if (bytes_left
>= 0)
1250 read_extent_buffer(eb
, dest
+ bytes_left
,
1251 name_off
, name_len
);
1253 btrfs_tree_read_unlock_blocking(eb
);
1254 free_extent_buffer(eb
);
1256 ret
= inode_ref_info(parent
, 0, fs_root
, path
, &found_key
);
1262 next_inum
= found_key
.offset
;
1264 /* regular exit ahead */
1265 if (parent
== next_inum
)
1268 slot
= path
->slots
[0];
1269 eb
= path
->nodes
[0];
1270 /* make sure we can use eb after releasing the path */
1272 atomic_inc(&eb
->refs
);
1273 btrfs_tree_read_lock(eb
);
1274 btrfs_set_lock_blocking_rw(eb
, BTRFS_READ_LOCK
);
1276 btrfs_release_path(path
);
1277 iref
= btrfs_item_ptr(eb
, slot
, struct btrfs_inode_ref
);
1279 name_len
= btrfs_inode_ref_name_len(eb
, iref
);
1280 name_off
= (unsigned long)(iref
+ 1);
1284 if (bytes_left
>= 0)
1285 dest
[bytes_left
] = '/';
1288 btrfs_release_path(path
);
1289 path
->leave_spinning
= leave_spinning
;
1292 return ERR_PTR(ret
);
1294 return dest
+ bytes_left
;
1298 * this makes the path point to (logical EXTENT_ITEM *)
1299 * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for
1300 * tree blocks and <0 on error.
1302 int extent_from_logical(struct btrfs_fs_info
*fs_info
, u64 logical
,
1303 struct btrfs_path
*path
, struct btrfs_key
*found_key
,
1310 struct extent_buffer
*eb
;
1311 struct btrfs_extent_item
*ei
;
1312 struct btrfs_key key
;
1314 if (btrfs_fs_incompat(fs_info
, SKINNY_METADATA
))
1315 key
.type
= BTRFS_METADATA_ITEM_KEY
;
1317 key
.type
= BTRFS_EXTENT_ITEM_KEY
;
1318 key
.objectid
= logical
;
1319 key
.offset
= (u64
)-1;
1321 ret
= btrfs_search_slot(NULL
, fs_info
->extent_root
, &key
, path
, 0, 0);
1327 if (path
->slots
[0] == 0) {
1328 btrfs_set_path_blocking(path
);
1329 ret
= btrfs_prev_leaf(fs_info
->extent_root
, path
);
1332 pr_debug("logical %llu is not within "
1333 "any extent\n", logical
);
1341 nritems
= btrfs_header_nritems(path
->nodes
[0]);
1343 pr_debug("logical %llu is not within any extent\n",
1347 if (path
->slots
[0] == nritems
)
1350 btrfs_item_key_to_cpu(path
->nodes
[0], found_key
,
1352 if (found_key
->type
== BTRFS_EXTENT_ITEM_KEY
||
1353 found_key
->type
== BTRFS_METADATA_ITEM_KEY
)
1357 if (found_key
->type
== BTRFS_METADATA_ITEM_KEY
)
1358 size
= fs_info
->extent_root
->leafsize
;
1359 else if (found_key
->type
== BTRFS_EXTENT_ITEM_KEY
)
1360 size
= found_key
->offset
;
1362 if (found_key
->objectid
> logical
||
1363 found_key
->objectid
+ size
<= logical
) {
1364 pr_debug("logical %llu is not within any extent\n", logical
);
1368 eb
= path
->nodes
[0];
1369 item_size
= btrfs_item_size_nr(eb
, path
->slots
[0]);
1370 BUG_ON(item_size
< sizeof(*ei
));
1372 ei
= btrfs_item_ptr(eb
, path
->slots
[0], struct btrfs_extent_item
);
1373 flags
= btrfs_extent_flags(eb
, ei
);
1375 pr_debug("logical %llu is at position %llu within the extent (%llu "
1376 "EXTENT_ITEM %llu) flags %#llx size %u\n",
1377 logical
, logical
- found_key
->objectid
, found_key
->objectid
,
1378 found_key
->offset
, flags
, item_size
);
1380 WARN_ON(!flags_ret
);
1382 if (flags
& BTRFS_EXTENT_FLAG_TREE_BLOCK
)
1383 *flags_ret
= BTRFS_EXTENT_FLAG_TREE_BLOCK
;
1384 else if (flags
& BTRFS_EXTENT_FLAG_DATA
)
1385 *flags_ret
= BTRFS_EXTENT_FLAG_DATA
;
1395 * helper function to iterate extent inline refs. ptr must point to a 0 value
1396 * for the first call and may be modified. it is used to track state.
1397 * if more refs exist, 0 is returned and the next call to
1398 * __get_extent_inline_ref must pass the modified ptr parameter to get the
1399 * next ref. after the last ref was processed, 1 is returned.
1400 * returns <0 on error
1402 static int __get_extent_inline_ref(unsigned long *ptr
, struct extent_buffer
*eb
,
1403 struct btrfs_extent_item
*ei
, u32 item_size
,
1404 struct btrfs_extent_inline_ref
**out_eiref
,
1409 struct btrfs_tree_block_info
*info
;
1413 flags
= btrfs_extent_flags(eb
, ei
);
1414 if (flags
& BTRFS_EXTENT_FLAG_TREE_BLOCK
) {
1415 info
= (struct btrfs_tree_block_info
*)(ei
+ 1);
1417 (struct btrfs_extent_inline_ref
*)(info
+ 1);
1419 *out_eiref
= (struct btrfs_extent_inline_ref
*)(ei
+ 1);
1421 *ptr
= (unsigned long)*out_eiref
;
1422 if ((void *)*ptr
>= (void *)ei
+ item_size
)
1426 end
= (unsigned long)ei
+ item_size
;
1427 *out_eiref
= (struct btrfs_extent_inline_ref
*)*ptr
;
1428 *out_type
= btrfs_extent_inline_ref_type(eb
, *out_eiref
);
1430 *ptr
+= btrfs_extent_inline_ref_size(*out_type
);
1431 WARN_ON(*ptr
> end
);
1433 return 1; /* last */
1439 * reads the tree block backref for an extent. tree level and root are returned
1440 * through out_level and out_root. ptr must point to a 0 value for the first
1441 * call and may be modified (see __get_extent_inline_ref comment).
1442 * returns 0 if data was provided, 1 if there was no more data to provide or
1445 int tree_backref_for_extent(unsigned long *ptr
, struct extent_buffer
*eb
,
1446 struct btrfs_extent_item
*ei
, u32 item_size
,
1447 u64
*out_root
, u8
*out_level
)
1451 struct btrfs_tree_block_info
*info
;
1452 struct btrfs_extent_inline_ref
*eiref
;
1454 if (*ptr
== (unsigned long)-1)
1458 ret
= __get_extent_inline_ref(ptr
, eb
, ei
, item_size
,
1463 if (type
== BTRFS_TREE_BLOCK_REF_KEY
||
1464 type
== BTRFS_SHARED_BLOCK_REF_KEY
)
1471 /* we can treat both ref types equally here */
1472 info
= (struct btrfs_tree_block_info
*)(ei
+ 1);
1473 *out_root
= btrfs_extent_inline_ref_offset(eb
, eiref
);
1474 *out_level
= btrfs_tree_block_level(eb
, info
);
1477 *ptr
= (unsigned long)-1;
1482 static int iterate_leaf_refs(struct extent_inode_elem
*inode_list
,
1483 u64 root
, u64 extent_item_objectid
,
1484 iterate_extent_inodes_t
*iterate
, void *ctx
)
1486 struct extent_inode_elem
*eie
;
1489 for (eie
= inode_list
; eie
; eie
= eie
->next
) {
1490 pr_debug("ref for %llu resolved, key (%llu EXTEND_DATA %llu), "
1491 "root %llu\n", extent_item_objectid
,
1492 eie
->inum
, eie
->offset
, root
);
1493 ret
= iterate(eie
->inum
, eie
->offset
, root
, ctx
);
1495 pr_debug("stopping iteration for %llu due to ret=%d\n",
1496 extent_item_objectid
, ret
);
1505 * calls iterate() for every inode that references the extent identified by
1506 * the given parameters.
1507 * when the iterator function returns a non-zero value, iteration stops.
1509 int iterate_extent_inodes(struct btrfs_fs_info
*fs_info
,
1510 u64 extent_item_objectid
, u64 extent_item_pos
,
1511 int search_commit_root
,
1512 iterate_extent_inodes_t
*iterate
, void *ctx
)
1515 struct btrfs_trans_handle
*trans
= NULL
;
1516 struct ulist
*refs
= NULL
;
1517 struct ulist
*roots
= NULL
;
1518 struct ulist_node
*ref_node
= NULL
;
1519 struct ulist_node
*root_node
= NULL
;
1520 struct seq_list tree_mod_seq_elem
= {};
1521 struct ulist_iterator ref_uiter
;
1522 struct ulist_iterator root_uiter
;
1524 pr_debug("resolving all inodes for extent %llu\n",
1525 extent_item_objectid
);
1527 if (!search_commit_root
) {
1528 trans
= btrfs_join_transaction(fs_info
->extent_root
);
1530 return PTR_ERR(trans
);
1531 btrfs_get_tree_mod_seq(fs_info
, &tree_mod_seq_elem
);
1534 ret
= btrfs_find_all_leafs(trans
, fs_info
, extent_item_objectid
,
1535 tree_mod_seq_elem
.seq
, &refs
,
1540 ULIST_ITER_INIT(&ref_uiter
);
1541 while (!ret
&& (ref_node
= ulist_next(refs
, &ref_uiter
))) {
1542 ret
= btrfs_find_all_roots(trans
, fs_info
, ref_node
->val
,
1543 tree_mod_seq_elem
.seq
, &roots
);
1546 ULIST_ITER_INIT(&root_uiter
);
1547 while (!ret
&& (root_node
= ulist_next(roots
, &root_uiter
))) {
1548 pr_debug("root %llu references leaf %llu, data list "
1549 "%#llx\n", root_node
->val
, ref_node
->val
,
1551 ret
= iterate_leaf_refs((struct extent_inode_elem
*)
1552 (uintptr_t)ref_node
->aux
,
1554 extent_item_objectid
,
1560 free_leaf_list(refs
);
1562 if (!search_commit_root
) {
1563 btrfs_put_tree_mod_seq(fs_info
, &tree_mod_seq_elem
);
1564 btrfs_end_transaction(trans
, fs_info
->extent_root
);
1570 int iterate_inodes_from_logical(u64 logical
, struct btrfs_fs_info
*fs_info
,
1571 struct btrfs_path
*path
,
1572 iterate_extent_inodes_t
*iterate
, void *ctx
)
1575 u64 extent_item_pos
;
1577 struct btrfs_key found_key
;
1578 int search_commit_root
= path
->search_commit_root
;
1580 ret
= extent_from_logical(fs_info
, logical
, path
, &found_key
, &flags
);
1581 btrfs_release_path(path
);
1584 if (flags
& BTRFS_EXTENT_FLAG_TREE_BLOCK
)
1587 extent_item_pos
= logical
- found_key
.objectid
;
1588 ret
= iterate_extent_inodes(fs_info
, found_key
.objectid
,
1589 extent_item_pos
, search_commit_root
,
1595 typedef int (iterate_irefs_t
)(u64 parent
, u32 name_len
, unsigned long name_off
,
1596 struct extent_buffer
*eb
, void *ctx
);
1598 static int iterate_inode_refs(u64 inum
, struct btrfs_root
*fs_root
,
1599 struct btrfs_path
*path
,
1600 iterate_irefs_t
*iterate
, void *ctx
)
1609 struct extent_buffer
*eb
;
1610 struct btrfs_item
*item
;
1611 struct btrfs_inode_ref
*iref
;
1612 struct btrfs_key found_key
;
1615 ret
= inode_ref_info(inum
, parent
? parent
+1 : 0, fs_root
, path
,
1620 ret
= found
? 0 : -ENOENT
;
1625 parent
= found_key
.offset
;
1626 slot
= path
->slots
[0];
1627 eb
= btrfs_clone_extent_buffer(path
->nodes
[0]);
1632 extent_buffer_get(eb
);
1633 btrfs_tree_read_lock(eb
);
1634 btrfs_set_lock_blocking_rw(eb
, BTRFS_READ_LOCK
);
1635 btrfs_release_path(path
);
1637 item
= btrfs_item_nr(slot
);
1638 iref
= btrfs_item_ptr(eb
, slot
, struct btrfs_inode_ref
);
1640 for (cur
= 0; cur
< btrfs_item_size(eb
, item
); cur
+= len
) {
1641 name_len
= btrfs_inode_ref_name_len(eb
, iref
);
1642 /* path must be released before calling iterate()! */
1643 pr_debug("following ref at offset %u for inode %llu in "
1644 "tree %llu\n", cur
, found_key
.objectid
,
1646 ret
= iterate(parent
, name_len
,
1647 (unsigned long)(iref
+ 1), eb
, ctx
);
1650 len
= sizeof(*iref
) + name_len
;
1651 iref
= (struct btrfs_inode_ref
*)((char *)iref
+ len
);
1653 btrfs_tree_read_unlock_blocking(eb
);
1654 free_extent_buffer(eb
);
1657 btrfs_release_path(path
);
1662 static int iterate_inode_extrefs(u64 inum
, struct btrfs_root
*fs_root
,
1663 struct btrfs_path
*path
,
1664 iterate_irefs_t
*iterate
, void *ctx
)
1671 struct extent_buffer
*eb
;
1672 struct btrfs_inode_extref
*extref
;
1673 struct extent_buffer
*leaf
;
1679 ret
= btrfs_find_one_extref(fs_root
, inum
, offset
, path
, &extref
,
1684 ret
= found
? 0 : -ENOENT
;
1689 slot
= path
->slots
[0];
1690 eb
= btrfs_clone_extent_buffer(path
->nodes
[0]);
1695 extent_buffer_get(eb
);
1697 btrfs_tree_read_lock(eb
);
1698 btrfs_set_lock_blocking_rw(eb
, BTRFS_READ_LOCK
);
1699 btrfs_release_path(path
);
1701 leaf
= path
->nodes
[0];
1702 item_size
= btrfs_item_size_nr(leaf
, slot
);
1703 ptr
= btrfs_item_ptr_offset(leaf
, slot
);
1706 while (cur_offset
< item_size
) {
1709 extref
= (struct btrfs_inode_extref
*)(ptr
+ cur_offset
);
1710 parent
= btrfs_inode_extref_parent(eb
, extref
);
1711 name_len
= btrfs_inode_extref_name_len(eb
, extref
);
1712 ret
= iterate(parent
, name_len
,
1713 (unsigned long)&extref
->name
, eb
, ctx
);
1717 cur_offset
+= btrfs_inode_extref_name_len(leaf
, extref
);
1718 cur_offset
+= sizeof(*extref
);
1720 btrfs_tree_read_unlock_blocking(eb
);
1721 free_extent_buffer(eb
);
1726 btrfs_release_path(path
);
1731 static int iterate_irefs(u64 inum
, struct btrfs_root
*fs_root
,
1732 struct btrfs_path
*path
, iterate_irefs_t
*iterate
,
1738 ret
= iterate_inode_refs(inum
, fs_root
, path
, iterate
, ctx
);
1741 else if (ret
!= -ENOENT
)
1744 ret
= iterate_inode_extrefs(inum
, fs_root
, path
, iterate
, ctx
);
1745 if (ret
== -ENOENT
&& found_refs
)
1752 * returns 0 if the path could be dumped (probably truncated)
1753 * returns <0 in case of an error
1755 static int inode_to_path(u64 inum
, u32 name_len
, unsigned long name_off
,
1756 struct extent_buffer
*eb
, void *ctx
)
1758 struct inode_fs_paths
*ipath
= ctx
;
1761 int i
= ipath
->fspath
->elem_cnt
;
1762 const int s_ptr
= sizeof(char *);
1765 bytes_left
= ipath
->fspath
->bytes_left
> s_ptr
?
1766 ipath
->fspath
->bytes_left
- s_ptr
: 0;
1768 fspath_min
= (char *)ipath
->fspath
->val
+ (i
+ 1) * s_ptr
;
1769 fspath
= btrfs_ref_to_path(ipath
->fs_root
, ipath
->btrfs_path
, name_len
,
1770 name_off
, eb
, inum
, fspath_min
, bytes_left
);
1772 return PTR_ERR(fspath
);
1774 if (fspath
> fspath_min
) {
1775 ipath
->fspath
->val
[i
] = (u64
)(unsigned long)fspath
;
1776 ++ipath
->fspath
->elem_cnt
;
1777 ipath
->fspath
->bytes_left
= fspath
- fspath_min
;
1779 ++ipath
->fspath
->elem_missed
;
1780 ipath
->fspath
->bytes_missing
+= fspath_min
- fspath
;
1781 ipath
->fspath
->bytes_left
= 0;
1788 * this dumps all file system paths to the inode into the ipath struct, provided
1789 * is has been created large enough. each path is zero-terminated and accessed
1790 * from ipath->fspath->val[i].
1791 * when it returns, there are ipath->fspath->elem_cnt number of paths available
1792 * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
1793 * number of missed paths in recored in ipath->fspath->elem_missed, otherwise,
1794 * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
1795 * have been needed to return all paths.
1797 int paths_from_inode(u64 inum
, struct inode_fs_paths
*ipath
)
1799 return iterate_irefs(inum
, ipath
->fs_root
, ipath
->btrfs_path
,
1800 inode_to_path
, ipath
);
1803 struct btrfs_data_container
*init_data_container(u32 total_bytes
)
1805 struct btrfs_data_container
*data
;
1808 alloc_bytes
= max_t(size_t, total_bytes
, sizeof(*data
));
1809 data
= vmalloc(alloc_bytes
);
1811 return ERR_PTR(-ENOMEM
);
1813 if (total_bytes
>= sizeof(*data
)) {
1814 data
->bytes_left
= total_bytes
- sizeof(*data
);
1815 data
->bytes_missing
= 0;
1817 data
->bytes_missing
= sizeof(*data
) - total_bytes
;
1818 data
->bytes_left
= 0;
1822 data
->elem_missed
= 0;
1828 * allocates space to return multiple file system paths for an inode.
1829 * total_bytes to allocate are passed, note that space usable for actual path
1830 * information will be total_bytes - sizeof(struct inode_fs_paths).
1831 * the returned pointer must be freed with free_ipath() in the end.
1833 struct inode_fs_paths
*init_ipath(s32 total_bytes
, struct btrfs_root
*fs_root
,
1834 struct btrfs_path
*path
)
1836 struct inode_fs_paths
*ifp
;
1837 struct btrfs_data_container
*fspath
;
1839 fspath
= init_data_container(total_bytes
);
1841 return (void *)fspath
;
1843 ifp
= kmalloc(sizeof(*ifp
), GFP_NOFS
);
1846 return ERR_PTR(-ENOMEM
);
1849 ifp
->btrfs_path
= path
;
1850 ifp
->fspath
= fspath
;
1851 ifp
->fs_root
= fs_root
;
1856 void free_ipath(struct inode_fs_paths
*ipath
)
1860 vfree(ipath
->fspath
);