Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/signal
[linux-3.10.git] / fs / btrfs / backref.c
1 /*
2  * Copyright (C) 2011 STRATO.  All rights reserved.
3  *
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.
7  *
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.
12  *
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.
17  */
18
19 #include "ctree.h"
20 #include "disk-io.h"
21 #include "backref.h"
22 #include "ulist.h"
23 #include "transaction.h"
24 #include "delayed-ref.h"
25 #include "locking.h"
26
27 struct extent_inode_elem {
28         u64 inum;
29         u64 offset;
30         struct extent_inode_elem *next;
31 };
32
33 static int check_extent_in_eb(struct btrfs_key *key, struct extent_buffer *eb,
34                                 struct btrfs_file_extent_item *fi,
35                                 u64 extent_item_pos,
36                                 struct extent_inode_elem **eie)
37 {
38         u64 data_offset;
39         u64 data_len;
40         struct extent_inode_elem *e;
41
42         data_offset = btrfs_file_extent_offset(eb, fi);
43         data_len = btrfs_file_extent_num_bytes(eb, fi);
44
45         if (extent_item_pos < data_offset ||
46             extent_item_pos >= data_offset + data_len)
47                 return 1;
48
49         e = kmalloc(sizeof(*e), GFP_NOFS);
50         if (!e)
51                 return -ENOMEM;
52
53         e->next = *eie;
54         e->inum = key->objectid;
55         e->offset = key->offset + (extent_item_pos - data_offset);
56         *eie = e;
57
58         return 0;
59 }
60
61 static int find_extent_in_eb(struct extent_buffer *eb, u64 wanted_disk_byte,
62                                 u64 extent_item_pos,
63                                 struct extent_inode_elem **eie)
64 {
65         u64 disk_byte;
66         struct btrfs_key key;
67         struct btrfs_file_extent_item *fi;
68         int slot;
69         int nritems;
70         int extent_type;
71         int ret;
72
73         /*
74          * from the shared data ref, we only have the leaf but we need
75          * the key. thus, we must look into all items and see that we
76          * find one (some) with a reference to our extent item.
77          */
78         nritems = btrfs_header_nritems(eb);
79         for (slot = 0; slot < nritems; ++slot) {
80                 btrfs_item_key_to_cpu(eb, &key, slot);
81                 if (key.type != BTRFS_EXTENT_DATA_KEY)
82                         continue;
83                 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
84                 extent_type = btrfs_file_extent_type(eb, fi);
85                 if (extent_type == BTRFS_FILE_EXTENT_INLINE)
86                         continue;
87                 /* don't skip BTRFS_FILE_EXTENT_PREALLOC, we can handle that */
88                 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
89                 if (disk_byte != wanted_disk_byte)
90                         continue;
91
92                 ret = check_extent_in_eb(&key, eb, fi, extent_item_pos, eie);
93                 if (ret < 0)
94                         return ret;
95         }
96
97         return 0;
98 }
99
100 /*
101  * this structure records all encountered refs on the way up to the root
102  */
103 struct __prelim_ref {
104         struct list_head list;
105         u64 root_id;
106         struct btrfs_key key_for_search;
107         int level;
108         int count;
109         struct extent_inode_elem *inode_list;
110         u64 parent;
111         u64 wanted_disk_byte;
112 };
113
114 /*
115  * the rules for all callers of this function are:
116  * - obtaining the parent is the goal
117  * - if you add a key, you must know that it is a correct key
118  * - if you cannot add the parent or a correct key, then we will look into the
119  *   block later to set a correct key
120  *
121  * delayed refs
122  * ============
123  *        backref type | shared | indirect | shared | indirect
124  * information         |   tree |     tree |   data |     data
125  * --------------------+--------+----------+--------+----------
126  *      parent logical |    y   |     -    |    -   |     -
127  *      key to resolve |    -   |     y    |    y   |     y
128  *  tree block logical |    -   |     -    |    -   |     -
129  *  root for resolving |    y   |     y    |    y   |     y
130  *
131  * - column 1:       we've the parent -> done
132  * - column 2, 3, 4: we use the key to find the parent
133  *
134  * on disk refs (inline or keyed)
135  * ==============================
136  *        backref type | shared | indirect | shared | indirect
137  * information         |   tree |     tree |   data |     data
138  * --------------------+--------+----------+--------+----------
139  *      parent logical |    y   |     -    |    y   |     -
140  *      key to resolve |    -   |     -    |    -   |     y
141  *  tree block logical |    y   |     y    |    y   |     y
142  *  root for resolving |    -   |     y    |    y   |     y
143  *
144  * - column 1, 3: we've the parent -> done
145  * - column 2:    we take the first key from the block to find the parent
146  *                (see __add_missing_keys)
147  * - column 4:    we use the key to find the parent
148  *
149  * additional information that's available but not required to find the parent
150  * block might help in merging entries to gain some speed.
151  */
152
153 static int __add_prelim_ref(struct list_head *head, u64 root_id,
154                             struct btrfs_key *key, int level,
155                             u64 parent, u64 wanted_disk_byte, int count)
156 {
157         struct __prelim_ref *ref;
158
159         /* in case we're adding delayed refs, we're holding the refs spinlock */
160         ref = kmalloc(sizeof(*ref), GFP_ATOMIC);
161         if (!ref)
162                 return -ENOMEM;
163
164         ref->root_id = root_id;
165         if (key)
166                 ref->key_for_search = *key;
167         else
168                 memset(&ref->key_for_search, 0, sizeof(ref->key_for_search));
169
170         ref->inode_list = NULL;
171         ref->level = level;
172         ref->count = count;
173         ref->parent = parent;
174         ref->wanted_disk_byte = wanted_disk_byte;
175         list_add_tail(&ref->list, head);
176
177         return 0;
178 }
179
180 static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
181                                 struct ulist *parents, int level,
182                                 struct btrfs_key *key, u64 wanted_disk_byte,
183                                 const u64 *extent_item_pos)
184 {
185         int ret;
186         int slot = path->slots[level];
187         struct extent_buffer *eb = path->nodes[level];
188         struct btrfs_file_extent_item *fi;
189         struct extent_inode_elem *eie = NULL;
190         u64 disk_byte;
191         u64 wanted_objectid = key->objectid;
192
193 add_parent:
194         if (level == 0 && extent_item_pos) {
195                 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
196                 ret = check_extent_in_eb(key, eb, fi, *extent_item_pos, &eie);
197                 if (ret < 0)
198                         return ret;
199         }
200         ret = ulist_add(parents, eb->start, (unsigned long)eie, GFP_NOFS);
201         if (ret < 0)
202                 return ret;
203
204         if (level != 0)
205                 return 0;
206
207         /*
208          * if the current leaf is full with EXTENT_DATA items, we must
209          * check the next one if that holds a reference as well.
210          * ref->count cannot be used to skip this check.
211          * repeat this until we don't find any additional EXTENT_DATA items.
212          */
213         while (1) {
214                 eie = NULL;
215                 ret = btrfs_next_leaf(root, path);
216                 if (ret < 0)
217                         return ret;
218                 if (ret)
219                         return 0;
220
221                 eb = path->nodes[0];
222                 for (slot = 0; slot < btrfs_header_nritems(eb); ++slot) {
223                         btrfs_item_key_to_cpu(eb, key, slot);
224                         if (key->objectid != wanted_objectid ||
225                             key->type != BTRFS_EXTENT_DATA_KEY)
226                                 return 0;
227                         fi = btrfs_item_ptr(eb, slot,
228                                                 struct btrfs_file_extent_item);
229                         disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
230                         if (disk_byte == wanted_disk_byte)
231                                 goto add_parent;
232                 }
233         }
234
235         return 0;
236 }
237
238 /*
239  * resolve an indirect backref in the form (root_id, key, level)
240  * to a logical address
241  */
242 static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
243                                         int search_commit_root,
244                                         u64 time_seq,
245                                         struct __prelim_ref *ref,
246                                         struct ulist *parents,
247                                         const u64 *extent_item_pos)
248 {
249         struct btrfs_path *path;
250         struct btrfs_root *root;
251         struct btrfs_key root_key;
252         struct btrfs_key key = {0};
253         struct extent_buffer *eb;
254         int ret = 0;
255         int root_level;
256         int level = ref->level;
257
258         path = btrfs_alloc_path();
259         if (!path)
260                 return -ENOMEM;
261         path->search_commit_root = !!search_commit_root;
262
263         root_key.objectid = ref->root_id;
264         root_key.type = BTRFS_ROOT_ITEM_KEY;
265         root_key.offset = (u64)-1;
266         root = btrfs_read_fs_root_no_name(fs_info, &root_key);
267         if (IS_ERR(root)) {
268                 ret = PTR_ERR(root);
269                 goto out;
270         }
271
272         rcu_read_lock();
273         root_level = btrfs_header_level(root->node);
274         rcu_read_unlock();
275
276         if (root_level + 1 == level)
277                 goto out;
278
279         path->lowest_level = level;
280         ret = btrfs_search_old_slot(root, &ref->key_for_search, path, time_seq);
281         pr_debug("search slot in root %llu (level %d, ref count %d) returned "
282                  "%d for key (%llu %u %llu)\n",
283                  (unsigned long long)ref->root_id, level, ref->count, ret,
284                  (unsigned long long)ref->key_for_search.objectid,
285                  ref->key_for_search.type,
286                  (unsigned long long)ref->key_for_search.offset);
287         if (ret < 0)
288                 goto out;
289
290         eb = path->nodes[level];
291         if (!eb) {
292                 WARN_ON(1);
293                 ret = 1;
294                 goto out;
295         }
296
297         if (level == 0) {
298                 if (ret == 1 && path->slots[0] >= btrfs_header_nritems(eb)) {
299                         ret = btrfs_next_leaf(root, path);
300                         if (ret)
301                                 goto out;
302                         eb = path->nodes[0];
303                 }
304
305                 btrfs_item_key_to_cpu(eb, &key, path->slots[0]);
306         }
307
308         ret = add_all_parents(root, path, parents, level, &key,
309                                 ref->wanted_disk_byte, extent_item_pos);
310 out:
311         btrfs_free_path(path);
312         return ret;
313 }
314
315 /*
316  * resolve all indirect backrefs from the list
317  */
318 static int __resolve_indirect_refs(struct btrfs_fs_info *fs_info,
319                                    int search_commit_root, u64 time_seq,
320                                    struct list_head *head,
321                                    const u64 *extent_item_pos)
322 {
323         int err;
324         int ret = 0;
325         struct __prelim_ref *ref;
326         struct __prelim_ref *ref_safe;
327         struct __prelim_ref *new_ref;
328         struct ulist *parents;
329         struct ulist_node *node;
330         struct ulist_iterator uiter;
331
332         parents = ulist_alloc(GFP_NOFS);
333         if (!parents)
334                 return -ENOMEM;
335
336         /*
337          * _safe allows us to insert directly after the current item without
338          * iterating over the newly inserted items.
339          * we're also allowed to re-assign ref during iteration.
340          */
341         list_for_each_entry_safe(ref, ref_safe, head, list) {
342                 if (ref->parent)        /* already direct */
343                         continue;
344                 if (ref->count == 0)
345                         continue;
346                 err = __resolve_indirect_ref(fs_info, search_commit_root,
347                                              time_seq, ref, parents,
348                                              extent_item_pos);
349                 if (err) {
350                         if (ret == 0)
351                                 ret = err;
352                         continue;
353                 }
354
355                 /* we put the first parent into the ref at hand */
356                 ULIST_ITER_INIT(&uiter);
357                 node = ulist_next(parents, &uiter);
358                 ref->parent = node ? node->val : 0;
359                 ref->inode_list =
360                         node ? (struct extent_inode_elem *)node->aux : 0;
361
362                 /* additional parents require new refs being added here */
363                 while ((node = ulist_next(parents, &uiter))) {
364                         new_ref = kmalloc(sizeof(*new_ref), GFP_NOFS);
365                         if (!new_ref) {
366                                 ret = -ENOMEM;
367                                 break;
368                         }
369                         memcpy(new_ref, ref, sizeof(*ref));
370                         new_ref->parent = node->val;
371                         new_ref->inode_list =
372                                         (struct extent_inode_elem *)node->aux;
373                         list_add(&new_ref->list, &ref->list);
374                 }
375                 ulist_reinit(parents);
376         }
377
378         ulist_free(parents);
379         return ret;
380 }
381
382 static inline int ref_for_same_block(struct __prelim_ref *ref1,
383                                      struct __prelim_ref *ref2)
384 {
385         if (ref1->level != ref2->level)
386                 return 0;
387         if (ref1->root_id != ref2->root_id)
388                 return 0;
389         if (ref1->key_for_search.type != ref2->key_for_search.type)
390                 return 0;
391         if (ref1->key_for_search.objectid != ref2->key_for_search.objectid)
392                 return 0;
393         if (ref1->key_for_search.offset != ref2->key_for_search.offset)
394                 return 0;
395         if (ref1->parent != ref2->parent)
396                 return 0;
397
398         return 1;
399 }
400
401 /*
402  * read tree blocks and add keys where required.
403  */
404 static int __add_missing_keys(struct btrfs_fs_info *fs_info,
405                               struct list_head *head)
406 {
407         struct list_head *pos;
408         struct extent_buffer *eb;
409
410         list_for_each(pos, head) {
411                 struct __prelim_ref *ref;
412                 ref = list_entry(pos, struct __prelim_ref, list);
413
414                 if (ref->parent)
415                         continue;
416                 if (ref->key_for_search.type)
417                         continue;
418                 BUG_ON(!ref->wanted_disk_byte);
419                 eb = read_tree_block(fs_info->tree_root, ref->wanted_disk_byte,
420                                      fs_info->tree_root->leafsize, 0);
421                 BUG_ON(!eb);
422                 btrfs_tree_read_lock(eb);
423                 if (btrfs_header_level(eb) == 0)
424                         btrfs_item_key_to_cpu(eb, &ref->key_for_search, 0);
425                 else
426                         btrfs_node_key_to_cpu(eb, &ref->key_for_search, 0);
427                 btrfs_tree_read_unlock(eb);
428                 free_extent_buffer(eb);
429         }
430         return 0;
431 }
432
433 /*
434  * merge two lists of backrefs and adjust counts accordingly
435  *
436  * mode = 1: merge identical keys, if key is set
437  *    FIXME: if we add more keys in __add_prelim_ref, we can merge more here.
438  *           additionally, we could even add a key range for the blocks we
439  *           looked into to merge even more (-> replace unresolved refs by those
440  *           having a parent).
441  * mode = 2: merge identical parents
442  */
443 static int __merge_refs(struct list_head *head, int mode)
444 {
445         struct list_head *pos1;
446
447         list_for_each(pos1, head) {
448                 struct list_head *n2;
449                 struct list_head *pos2;
450                 struct __prelim_ref *ref1;
451
452                 ref1 = list_entry(pos1, struct __prelim_ref, list);
453
454                 for (pos2 = pos1->next, n2 = pos2->next; pos2 != head;
455                      pos2 = n2, n2 = pos2->next) {
456                         struct __prelim_ref *ref2;
457                         struct __prelim_ref *xchg;
458
459                         ref2 = list_entry(pos2, struct __prelim_ref, list);
460
461                         if (mode == 1) {
462                                 if (!ref_for_same_block(ref1, ref2))
463                                         continue;
464                                 if (!ref1->parent && ref2->parent) {
465                                         xchg = ref1;
466                                         ref1 = ref2;
467                                         ref2 = xchg;
468                                 }
469                                 ref1->count += ref2->count;
470                         } else {
471                                 if (ref1->parent != ref2->parent)
472                                         continue;
473                                 ref1->count += ref2->count;
474                         }
475                         list_del(&ref2->list);
476                         kfree(ref2);
477                 }
478
479         }
480         return 0;
481 }
482
483 /*
484  * add all currently queued delayed refs from this head whose seq nr is
485  * smaller or equal that seq to the list
486  */
487 static int __add_delayed_refs(struct btrfs_delayed_ref_head *head, u64 seq,
488                               struct list_head *prefs)
489 {
490         struct btrfs_delayed_extent_op *extent_op = head->extent_op;
491         struct rb_node *n = &head->node.rb_node;
492         struct btrfs_key key;
493         struct btrfs_key op_key = {0};
494         int sgn;
495         int ret = 0;
496
497         if (extent_op && extent_op->update_key)
498                 btrfs_disk_key_to_cpu(&op_key, &extent_op->key);
499
500         while ((n = rb_prev(n))) {
501                 struct btrfs_delayed_ref_node *node;
502                 node = rb_entry(n, struct btrfs_delayed_ref_node,
503                                 rb_node);
504                 if (node->bytenr != head->node.bytenr)
505                         break;
506                 WARN_ON(node->is_head);
507
508                 if (node->seq > seq)
509                         continue;
510
511                 switch (node->action) {
512                 case BTRFS_ADD_DELAYED_EXTENT:
513                 case BTRFS_UPDATE_DELAYED_HEAD:
514                         WARN_ON(1);
515                         continue;
516                 case BTRFS_ADD_DELAYED_REF:
517                         sgn = 1;
518                         break;
519                 case BTRFS_DROP_DELAYED_REF:
520                         sgn = -1;
521                         break;
522                 default:
523                         BUG_ON(1);
524                 }
525                 switch (node->type) {
526                 case BTRFS_TREE_BLOCK_REF_KEY: {
527                         struct btrfs_delayed_tree_ref *ref;
528
529                         ref = btrfs_delayed_node_to_tree_ref(node);
530                         ret = __add_prelim_ref(prefs, ref->root, &op_key,
531                                                ref->level + 1, 0, node->bytenr,
532                                                node->ref_mod * sgn);
533                         break;
534                 }
535                 case BTRFS_SHARED_BLOCK_REF_KEY: {
536                         struct btrfs_delayed_tree_ref *ref;
537
538                         ref = btrfs_delayed_node_to_tree_ref(node);
539                         ret = __add_prelim_ref(prefs, ref->root, NULL,
540                                                ref->level + 1, ref->parent,
541                                                node->bytenr,
542                                                node->ref_mod * sgn);
543                         break;
544                 }
545                 case BTRFS_EXTENT_DATA_REF_KEY: {
546                         struct btrfs_delayed_data_ref *ref;
547                         ref = btrfs_delayed_node_to_data_ref(node);
548
549                         key.objectid = ref->objectid;
550                         key.type = BTRFS_EXTENT_DATA_KEY;
551                         key.offset = ref->offset;
552                         ret = __add_prelim_ref(prefs, ref->root, &key, 0, 0,
553                                                node->bytenr,
554                                                node->ref_mod * sgn);
555                         break;
556                 }
557                 case BTRFS_SHARED_DATA_REF_KEY: {
558                         struct btrfs_delayed_data_ref *ref;
559
560                         ref = btrfs_delayed_node_to_data_ref(node);
561
562                         key.objectid = ref->objectid;
563                         key.type = BTRFS_EXTENT_DATA_KEY;
564                         key.offset = ref->offset;
565                         ret = __add_prelim_ref(prefs, ref->root, &key, 0,
566                                                ref->parent, node->bytenr,
567                                                node->ref_mod * sgn);
568                         break;
569                 }
570                 default:
571                         WARN_ON(1);
572                 }
573                 BUG_ON(ret);
574         }
575
576         return 0;
577 }
578
579 /*
580  * add all inline backrefs for bytenr to the list
581  */
582 static int __add_inline_refs(struct btrfs_fs_info *fs_info,
583                              struct btrfs_path *path, u64 bytenr,
584                              int *info_level, struct list_head *prefs)
585 {
586         int ret = 0;
587         int slot;
588         struct extent_buffer *leaf;
589         struct btrfs_key key;
590         unsigned long ptr;
591         unsigned long end;
592         struct btrfs_extent_item *ei;
593         u64 flags;
594         u64 item_size;
595
596         /*
597          * enumerate all inline refs
598          */
599         leaf = path->nodes[0];
600         slot = path->slots[0];
601
602         item_size = btrfs_item_size_nr(leaf, slot);
603         BUG_ON(item_size < sizeof(*ei));
604
605         ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
606         flags = btrfs_extent_flags(leaf, ei);
607
608         ptr = (unsigned long)(ei + 1);
609         end = (unsigned long)ei + item_size;
610
611         if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
612                 struct btrfs_tree_block_info *info;
613
614                 info = (struct btrfs_tree_block_info *)ptr;
615                 *info_level = btrfs_tree_block_level(leaf, info);
616                 ptr += sizeof(struct btrfs_tree_block_info);
617                 BUG_ON(ptr > end);
618         } else {
619                 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
620         }
621
622         while (ptr < end) {
623                 struct btrfs_extent_inline_ref *iref;
624                 u64 offset;
625                 int type;
626
627                 iref = (struct btrfs_extent_inline_ref *)ptr;
628                 type = btrfs_extent_inline_ref_type(leaf, iref);
629                 offset = btrfs_extent_inline_ref_offset(leaf, iref);
630
631                 switch (type) {
632                 case BTRFS_SHARED_BLOCK_REF_KEY:
633                         ret = __add_prelim_ref(prefs, 0, NULL,
634                                                 *info_level + 1, offset,
635                                                 bytenr, 1);
636                         break;
637                 case BTRFS_SHARED_DATA_REF_KEY: {
638                         struct btrfs_shared_data_ref *sdref;
639                         int count;
640
641                         sdref = (struct btrfs_shared_data_ref *)(iref + 1);
642                         count = btrfs_shared_data_ref_count(leaf, sdref);
643                         ret = __add_prelim_ref(prefs, 0, NULL, 0, offset,
644                                                bytenr, count);
645                         break;
646                 }
647                 case BTRFS_TREE_BLOCK_REF_KEY:
648                         ret = __add_prelim_ref(prefs, offset, NULL,
649                                                *info_level + 1, 0,
650                                                bytenr, 1);
651                         break;
652                 case BTRFS_EXTENT_DATA_REF_KEY: {
653                         struct btrfs_extent_data_ref *dref;
654                         int count;
655                         u64 root;
656
657                         dref = (struct btrfs_extent_data_ref *)(&iref->offset);
658                         count = btrfs_extent_data_ref_count(leaf, dref);
659                         key.objectid = btrfs_extent_data_ref_objectid(leaf,
660                                                                       dref);
661                         key.type = BTRFS_EXTENT_DATA_KEY;
662                         key.offset = btrfs_extent_data_ref_offset(leaf, dref);
663                         root = btrfs_extent_data_ref_root(leaf, dref);
664                         ret = __add_prelim_ref(prefs, root, &key, 0, 0,
665                                                bytenr, count);
666                         break;
667                 }
668                 default:
669                         WARN_ON(1);
670                 }
671                 BUG_ON(ret);
672                 ptr += btrfs_extent_inline_ref_size(type);
673         }
674
675         return 0;
676 }
677
678 /*
679  * add all non-inline backrefs for bytenr to the list
680  */
681 static int __add_keyed_refs(struct btrfs_fs_info *fs_info,
682                             struct btrfs_path *path, u64 bytenr,
683                             int info_level, struct list_head *prefs)
684 {
685         struct btrfs_root *extent_root = fs_info->extent_root;
686         int ret;
687         int slot;
688         struct extent_buffer *leaf;
689         struct btrfs_key key;
690
691         while (1) {
692                 ret = btrfs_next_item(extent_root, path);
693                 if (ret < 0)
694                         break;
695                 if (ret) {
696                         ret = 0;
697                         break;
698                 }
699
700                 slot = path->slots[0];
701                 leaf = path->nodes[0];
702                 btrfs_item_key_to_cpu(leaf, &key, slot);
703
704                 if (key.objectid != bytenr)
705                         break;
706                 if (key.type < BTRFS_TREE_BLOCK_REF_KEY)
707                         continue;
708                 if (key.type > BTRFS_SHARED_DATA_REF_KEY)
709                         break;
710
711                 switch (key.type) {
712                 case BTRFS_SHARED_BLOCK_REF_KEY:
713                         ret = __add_prelim_ref(prefs, 0, NULL,
714                                                 info_level + 1, key.offset,
715                                                 bytenr, 1);
716                         break;
717                 case BTRFS_SHARED_DATA_REF_KEY: {
718                         struct btrfs_shared_data_ref *sdref;
719                         int count;
720
721                         sdref = btrfs_item_ptr(leaf, slot,
722                                               struct btrfs_shared_data_ref);
723                         count = btrfs_shared_data_ref_count(leaf, sdref);
724                         ret = __add_prelim_ref(prefs, 0, NULL, 0, key.offset,
725                                                 bytenr, count);
726                         break;
727                 }
728                 case BTRFS_TREE_BLOCK_REF_KEY:
729                         ret = __add_prelim_ref(prefs, key.offset, NULL,
730                                                info_level + 1, 0,
731                                                bytenr, 1);
732                         break;
733                 case BTRFS_EXTENT_DATA_REF_KEY: {
734                         struct btrfs_extent_data_ref *dref;
735                         int count;
736                         u64 root;
737
738                         dref = btrfs_item_ptr(leaf, slot,
739                                               struct btrfs_extent_data_ref);
740                         count = btrfs_extent_data_ref_count(leaf, dref);
741                         key.objectid = btrfs_extent_data_ref_objectid(leaf,
742                                                                       dref);
743                         key.type = BTRFS_EXTENT_DATA_KEY;
744                         key.offset = btrfs_extent_data_ref_offset(leaf, dref);
745                         root = btrfs_extent_data_ref_root(leaf, dref);
746                         ret = __add_prelim_ref(prefs, root, &key, 0, 0,
747                                                bytenr, count);
748                         break;
749                 }
750                 default:
751                         WARN_ON(1);
752                 }
753                 BUG_ON(ret);
754         }
755
756         return ret;
757 }
758
759 /*
760  * this adds all existing backrefs (inline backrefs, backrefs and delayed
761  * refs) for the given bytenr to the refs list, merges duplicates and resolves
762  * indirect refs to their parent bytenr.
763  * When roots are found, they're added to the roots list
764  *
765  * FIXME some caching might speed things up
766  */
767 static int find_parent_nodes(struct btrfs_trans_handle *trans,
768                              struct btrfs_fs_info *fs_info, u64 bytenr,
769                              u64 delayed_ref_seq, u64 time_seq,
770                              struct ulist *refs, struct ulist *roots,
771                              const u64 *extent_item_pos)
772 {
773         struct btrfs_key key;
774         struct btrfs_path *path;
775         struct btrfs_delayed_ref_root *delayed_refs = NULL;
776         struct btrfs_delayed_ref_head *head;
777         int info_level = 0;
778         int ret;
779         int search_commit_root = (trans == BTRFS_BACKREF_SEARCH_COMMIT_ROOT);
780         struct list_head prefs_delayed;
781         struct list_head prefs;
782         struct __prelim_ref *ref;
783
784         INIT_LIST_HEAD(&prefs);
785         INIT_LIST_HEAD(&prefs_delayed);
786
787         key.objectid = bytenr;
788         key.type = BTRFS_EXTENT_ITEM_KEY;
789         key.offset = (u64)-1;
790
791         path = btrfs_alloc_path();
792         if (!path)
793                 return -ENOMEM;
794         path->search_commit_root = !!search_commit_root;
795
796         /*
797          * grab both a lock on the path and a lock on the delayed ref head.
798          * We need both to get a consistent picture of how the refs look
799          * at a specified point in time
800          */
801 again:
802         head = NULL;
803
804         ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
805         if (ret < 0)
806                 goto out;
807         BUG_ON(ret == 0);
808
809         if (trans != BTRFS_BACKREF_SEARCH_COMMIT_ROOT) {
810                 /*
811                  * look if there are updates for this ref queued and lock the
812                  * head
813                  */
814                 delayed_refs = &trans->transaction->delayed_refs;
815                 spin_lock(&delayed_refs->lock);
816                 head = btrfs_find_delayed_ref_head(trans, bytenr);
817                 if (head) {
818                         if (!mutex_trylock(&head->mutex)) {
819                                 atomic_inc(&head->node.refs);
820                                 spin_unlock(&delayed_refs->lock);
821
822                                 btrfs_release_path(path);
823
824                                 /*
825                                  * Mutex was contended, block until it's
826                                  * released and try again
827                                  */
828                                 mutex_lock(&head->mutex);
829                                 mutex_unlock(&head->mutex);
830                                 btrfs_put_delayed_ref(&head->node);
831                                 goto again;
832                         }
833                         ret = __add_delayed_refs(head, delayed_ref_seq,
834                                                  &prefs_delayed);
835                         if (ret) {
836                                 spin_unlock(&delayed_refs->lock);
837                                 goto out;
838                         }
839                 }
840                 spin_unlock(&delayed_refs->lock);
841         }
842
843         if (path->slots[0]) {
844                 struct extent_buffer *leaf;
845                 int slot;
846
847                 path->slots[0]--;
848                 leaf = path->nodes[0];
849                 slot = path->slots[0];
850                 btrfs_item_key_to_cpu(leaf, &key, slot);
851                 if (key.objectid == bytenr &&
852                     key.type == BTRFS_EXTENT_ITEM_KEY) {
853                         ret = __add_inline_refs(fs_info, path, bytenr,
854                                                 &info_level, &prefs);
855                         if (ret)
856                                 goto out;
857                         ret = __add_keyed_refs(fs_info, path, bytenr,
858                                                info_level, &prefs);
859                         if (ret)
860                                 goto out;
861                 }
862         }
863         btrfs_release_path(path);
864
865         list_splice_init(&prefs_delayed, &prefs);
866
867         ret = __add_missing_keys(fs_info, &prefs);
868         if (ret)
869                 goto out;
870
871         ret = __merge_refs(&prefs, 1);
872         if (ret)
873                 goto out;
874
875         ret = __resolve_indirect_refs(fs_info, search_commit_root, time_seq,
876                                       &prefs, extent_item_pos);
877         if (ret)
878                 goto out;
879
880         ret = __merge_refs(&prefs, 2);
881         if (ret)
882                 goto out;
883
884         while (!list_empty(&prefs)) {
885                 ref = list_first_entry(&prefs, struct __prelim_ref, list);
886                 list_del(&ref->list);
887                 if (ref->count < 0)
888                         WARN_ON(1);
889                 if (ref->count && ref->root_id && ref->parent == 0) {
890                         /* no parent == root of tree */
891                         ret = ulist_add(roots, ref->root_id, 0, GFP_NOFS);
892                         BUG_ON(ret < 0);
893                 }
894                 if (ref->count && ref->parent) {
895                         struct extent_inode_elem *eie = NULL;
896                         if (extent_item_pos && !ref->inode_list) {
897                                 u32 bsz;
898                                 struct extent_buffer *eb;
899                                 bsz = btrfs_level_size(fs_info->extent_root,
900                                                         info_level);
901                                 eb = read_tree_block(fs_info->extent_root,
902                                                            ref->parent, bsz, 0);
903                                 BUG_ON(!eb);
904                                 ret = find_extent_in_eb(eb, bytenr,
905                                                         *extent_item_pos, &eie);
906                                 ref->inode_list = eie;
907                                 free_extent_buffer(eb);
908                         }
909                         ret = ulist_add_merge(refs, ref->parent,
910                                               (unsigned long)ref->inode_list,
911                                               (unsigned long *)&eie, GFP_NOFS);
912                         if (!ret && extent_item_pos) {
913                                 /*
914                                  * we've recorded that parent, so we must extend
915                                  * its inode list here
916                                  */
917                                 BUG_ON(!eie);
918                                 while (eie->next)
919                                         eie = eie->next;
920                                 eie->next = ref->inode_list;
921                         }
922                         BUG_ON(ret < 0);
923                 }
924                 kfree(ref);
925         }
926
927 out:
928         if (head)
929                 mutex_unlock(&head->mutex);
930         btrfs_free_path(path);
931         while (!list_empty(&prefs)) {
932                 ref = list_first_entry(&prefs, struct __prelim_ref, list);
933                 list_del(&ref->list);
934                 kfree(ref);
935         }
936         while (!list_empty(&prefs_delayed)) {
937                 ref = list_first_entry(&prefs_delayed, struct __prelim_ref,
938                                        list);
939                 list_del(&ref->list);
940                 kfree(ref);
941         }
942
943         return ret;
944 }
945
946 static void free_leaf_list(struct ulist *blocks)
947 {
948         struct ulist_node *node = NULL;
949         struct extent_inode_elem *eie;
950         struct extent_inode_elem *eie_next;
951         struct ulist_iterator uiter;
952
953         ULIST_ITER_INIT(&uiter);
954         while ((node = ulist_next(blocks, &uiter))) {
955                 if (!node->aux)
956                         continue;
957                 eie = (struct extent_inode_elem *)node->aux;
958                 for (; eie; eie = eie_next) {
959                         eie_next = eie->next;
960                         kfree(eie);
961                 }
962                 node->aux = 0;
963         }
964
965         ulist_free(blocks);
966 }
967
968 /*
969  * Finds all leafs with a reference to the specified combination of bytenr and
970  * offset. key_list_head will point to a list of corresponding keys (caller must
971  * free each list element). The leafs will be stored in the leafs ulist, which
972  * must be freed with ulist_free.
973  *
974  * returns 0 on success, <0 on error
975  */
976 static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
977                                 struct btrfs_fs_info *fs_info, u64 bytenr,
978                                 u64 delayed_ref_seq, u64 time_seq,
979                                 struct ulist **leafs,
980                                 const u64 *extent_item_pos)
981 {
982         struct ulist *tmp;
983         int ret;
984
985         tmp = ulist_alloc(GFP_NOFS);
986         if (!tmp)
987                 return -ENOMEM;
988         *leafs = ulist_alloc(GFP_NOFS);
989         if (!*leafs) {
990                 ulist_free(tmp);
991                 return -ENOMEM;
992         }
993
994         ret = find_parent_nodes(trans, fs_info, bytenr, delayed_ref_seq,
995                                 time_seq, *leafs, tmp, extent_item_pos);
996         ulist_free(tmp);
997
998         if (ret < 0 && ret != -ENOENT) {
999                 free_leaf_list(*leafs);
1000                 return ret;
1001         }
1002
1003         return 0;
1004 }
1005
1006 /*
1007  * walk all backrefs for a given extent to find all roots that reference this
1008  * extent. Walking a backref means finding all extents that reference this
1009  * extent and in turn walk the backrefs of those, too. Naturally this is a
1010  * recursive process, but here it is implemented in an iterative fashion: We
1011  * find all referencing extents for the extent in question and put them on a
1012  * list. In turn, we find all referencing extents for those, further appending
1013  * to the list. The way we iterate the list allows adding more elements after
1014  * the current while iterating. The process stops when we reach the end of the
1015  * list. Found roots are added to the roots list.
1016  *
1017  * returns 0 on success, < 0 on error.
1018  */
1019 int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
1020                                 struct btrfs_fs_info *fs_info, u64 bytenr,
1021                                 u64 delayed_ref_seq, u64 time_seq,
1022                                 struct ulist **roots)
1023 {
1024         struct ulist *tmp;
1025         struct ulist_node *node = NULL;
1026         struct ulist_iterator uiter;
1027         int ret;
1028
1029         tmp = ulist_alloc(GFP_NOFS);
1030         if (!tmp)
1031                 return -ENOMEM;
1032         *roots = ulist_alloc(GFP_NOFS);
1033         if (!*roots) {
1034                 ulist_free(tmp);
1035                 return -ENOMEM;
1036         }
1037
1038         ULIST_ITER_INIT(&uiter);
1039         while (1) {
1040                 ret = find_parent_nodes(trans, fs_info, bytenr, delayed_ref_seq,
1041                                         time_seq, tmp, *roots, NULL);
1042                 if (ret < 0 && ret != -ENOENT) {
1043                         ulist_free(tmp);
1044                         ulist_free(*roots);
1045                         return ret;
1046                 }
1047                 node = ulist_next(tmp, &uiter);
1048                 if (!node)
1049                         break;
1050                 bytenr = node->val;
1051         }
1052
1053         ulist_free(tmp);
1054         return 0;
1055 }
1056
1057
1058 static int __inode_info(u64 inum, u64 ioff, u8 key_type,
1059                         struct btrfs_root *fs_root, struct btrfs_path *path,
1060                         struct btrfs_key *found_key)
1061 {
1062         int ret;
1063         struct btrfs_key key;
1064         struct extent_buffer *eb;
1065
1066         key.type = key_type;
1067         key.objectid = inum;
1068         key.offset = ioff;
1069
1070         ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1071         if (ret < 0)
1072                 return ret;
1073
1074         eb = path->nodes[0];
1075         if (ret && path->slots[0] >= btrfs_header_nritems(eb)) {
1076                 ret = btrfs_next_leaf(fs_root, path);
1077                 if (ret)
1078                         return ret;
1079                 eb = path->nodes[0];
1080         }
1081
1082         btrfs_item_key_to_cpu(eb, found_key, path->slots[0]);
1083         if (found_key->type != key.type || found_key->objectid != key.objectid)
1084                 return 1;
1085
1086         return 0;
1087 }
1088
1089 /*
1090  * this makes the path point to (inum INODE_ITEM ioff)
1091  */
1092 int inode_item_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
1093                         struct btrfs_path *path)
1094 {
1095         struct btrfs_key key;
1096         return __inode_info(inum, ioff, BTRFS_INODE_ITEM_KEY, fs_root, path,
1097                                 &key);
1098 }
1099
1100 static int inode_ref_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
1101                                 struct btrfs_path *path,
1102                                 struct btrfs_key *found_key)
1103 {
1104         return __inode_info(inum, ioff, BTRFS_INODE_REF_KEY, fs_root, path,
1105                                 found_key);
1106 }
1107
1108 /*
1109  * this iterates to turn a btrfs_inode_ref into a full filesystem path. elements
1110  * of the path are separated by '/' and the path is guaranteed to be
1111  * 0-terminated. the path is only given within the current file system.
1112  * Therefore, it never starts with a '/'. the caller is responsible to provide
1113  * "size" bytes in "dest". the dest buffer will be filled backwards. finally,
1114  * the start point of the resulting string is returned. this pointer is within
1115  * dest, normally.
1116  * in case the path buffer would overflow, the pointer is decremented further
1117  * as if output was written to the buffer, though no more output is actually
1118  * generated. that way, the caller can determine how much space would be
1119  * required for the path to fit into the buffer. in that case, the returned
1120  * value will be smaller than dest. callers must check this!
1121  */
1122 static char *iref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
1123                                 struct btrfs_inode_ref *iref,
1124                                 struct extent_buffer *eb_in, u64 parent,
1125                                 char *dest, u32 size)
1126 {
1127         u32 len;
1128         int slot;
1129         u64 next_inum;
1130         int ret;
1131         s64 bytes_left = size - 1;
1132         struct extent_buffer *eb = eb_in;
1133         struct btrfs_key found_key;
1134         int leave_spinning = path->leave_spinning;
1135
1136         if (bytes_left >= 0)
1137                 dest[bytes_left] = '\0';
1138
1139         path->leave_spinning = 1;
1140         while (1) {
1141                 len = btrfs_inode_ref_name_len(eb, iref);
1142                 bytes_left -= len;
1143                 if (bytes_left >= 0)
1144                         read_extent_buffer(eb, dest + bytes_left,
1145                                                 (unsigned long)(iref + 1), len);
1146                 if (eb != eb_in) {
1147                         btrfs_tree_read_unlock_blocking(eb);
1148                         free_extent_buffer(eb);
1149                 }
1150                 ret = inode_ref_info(parent, 0, fs_root, path, &found_key);
1151                 if (ret > 0)
1152                         ret = -ENOENT;
1153                 if (ret)
1154                         break;
1155                 next_inum = found_key.offset;
1156
1157                 /* regular exit ahead */
1158                 if (parent == next_inum)
1159                         break;
1160
1161                 slot = path->slots[0];
1162                 eb = path->nodes[0];
1163                 /* make sure we can use eb after releasing the path */
1164                 if (eb != eb_in) {
1165                         atomic_inc(&eb->refs);
1166                         btrfs_tree_read_lock(eb);
1167                         btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1168                 }
1169                 btrfs_release_path(path);
1170
1171                 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
1172                 parent = next_inum;
1173                 --bytes_left;
1174                 if (bytes_left >= 0)
1175                         dest[bytes_left] = '/';
1176         }
1177
1178         btrfs_release_path(path);
1179         path->leave_spinning = leave_spinning;
1180
1181         if (ret)
1182                 return ERR_PTR(ret);
1183
1184         return dest + bytes_left;
1185 }
1186
1187 /*
1188  * this makes the path point to (logical EXTENT_ITEM *)
1189  * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for
1190  * tree blocks and <0 on error.
1191  */
1192 int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
1193                         struct btrfs_path *path, struct btrfs_key *found_key)
1194 {
1195         int ret;
1196         u64 flags;
1197         u32 item_size;
1198         struct extent_buffer *eb;
1199         struct btrfs_extent_item *ei;
1200         struct btrfs_key key;
1201
1202         key.type = BTRFS_EXTENT_ITEM_KEY;
1203         key.objectid = logical;
1204         key.offset = (u64)-1;
1205
1206         ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
1207         if (ret < 0)
1208                 return ret;
1209         ret = btrfs_previous_item(fs_info->extent_root, path,
1210                                         0, BTRFS_EXTENT_ITEM_KEY);
1211         if (ret < 0)
1212                 return ret;
1213
1214         btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]);
1215         if (found_key->type != BTRFS_EXTENT_ITEM_KEY ||
1216             found_key->objectid > logical ||
1217             found_key->objectid + found_key->offset <= logical) {
1218                 pr_debug("logical %llu is not within any extent\n",
1219                          (unsigned long long)logical);
1220                 return -ENOENT;
1221         }
1222
1223         eb = path->nodes[0];
1224         item_size = btrfs_item_size_nr(eb, path->slots[0]);
1225         BUG_ON(item_size < sizeof(*ei));
1226
1227         ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
1228         flags = btrfs_extent_flags(eb, ei);
1229
1230         pr_debug("logical %llu is at position %llu within the extent (%llu "
1231                  "EXTENT_ITEM %llu) flags %#llx size %u\n",
1232                  (unsigned long long)logical,
1233                  (unsigned long long)(logical - found_key->objectid),
1234                  (unsigned long long)found_key->objectid,
1235                  (unsigned long long)found_key->offset,
1236                  (unsigned long long)flags, item_size);
1237         if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1238                 return BTRFS_EXTENT_FLAG_TREE_BLOCK;
1239         if (flags & BTRFS_EXTENT_FLAG_DATA)
1240                 return BTRFS_EXTENT_FLAG_DATA;
1241
1242         return -EIO;
1243 }
1244
1245 /*
1246  * helper function to iterate extent inline refs. ptr must point to a 0 value
1247  * for the first call and may be modified. it is used to track state.
1248  * if more refs exist, 0 is returned and the next call to
1249  * __get_extent_inline_ref must pass the modified ptr parameter to get the
1250  * next ref. after the last ref was processed, 1 is returned.
1251  * returns <0 on error
1252  */
1253 static int __get_extent_inline_ref(unsigned long *ptr, struct extent_buffer *eb,
1254                                 struct btrfs_extent_item *ei, u32 item_size,
1255                                 struct btrfs_extent_inline_ref **out_eiref,
1256                                 int *out_type)
1257 {
1258         unsigned long end;
1259         u64 flags;
1260         struct btrfs_tree_block_info *info;
1261
1262         if (!*ptr) {
1263                 /* first call */
1264                 flags = btrfs_extent_flags(eb, ei);
1265                 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1266                         info = (struct btrfs_tree_block_info *)(ei + 1);
1267                         *out_eiref =
1268                                 (struct btrfs_extent_inline_ref *)(info + 1);
1269                 } else {
1270                         *out_eiref = (struct btrfs_extent_inline_ref *)(ei + 1);
1271                 }
1272                 *ptr = (unsigned long)*out_eiref;
1273                 if ((void *)*ptr >= (void *)ei + item_size)
1274                         return -ENOENT;
1275         }
1276
1277         end = (unsigned long)ei + item_size;
1278         *out_eiref = (struct btrfs_extent_inline_ref *)*ptr;
1279         *out_type = btrfs_extent_inline_ref_type(eb, *out_eiref);
1280
1281         *ptr += btrfs_extent_inline_ref_size(*out_type);
1282         WARN_ON(*ptr > end);
1283         if (*ptr == end)
1284                 return 1; /* last */
1285
1286         return 0;
1287 }
1288
1289 /*
1290  * reads the tree block backref for an extent. tree level and root are returned
1291  * through out_level and out_root. ptr must point to a 0 value for the first
1292  * call and may be modified (see __get_extent_inline_ref comment).
1293  * returns 0 if data was provided, 1 if there was no more data to provide or
1294  * <0 on error.
1295  */
1296 int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
1297                                 struct btrfs_extent_item *ei, u32 item_size,
1298                                 u64 *out_root, u8 *out_level)
1299 {
1300         int ret;
1301         int type;
1302         struct btrfs_tree_block_info *info;
1303         struct btrfs_extent_inline_ref *eiref;
1304
1305         if (*ptr == (unsigned long)-1)
1306                 return 1;
1307
1308         while (1) {
1309                 ret = __get_extent_inline_ref(ptr, eb, ei, item_size,
1310                                                 &eiref, &type);
1311                 if (ret < 0)
1312                         return ret;
1313
1314                 if (type == BTRFS_TREE_BLOCK_REF_KEY ||
1315                     type == BTRFS_SHARED_BLOCK_REF_KEY)
1316                         break;
1317
1318                 if (ret == 1)
1319                         return 1;
1320         }
1321
1322         /* we can treat both ref types equally here */
1323         info = (struct btrfs_tree_block_info *)(ei + 1);
1324         *out_root = btrfs_extent_inline_ref_offset(eb, eiref);
1325         *out_level = btrfs_tree_block_level(eb, info);
1326
1327         if (ret == 1)
1328                 *ptr = (unsigned long)-1;
1329
1330         return 0;
1331 }
1332
1333 static int iterate_leaf_refs(struct extent_inode_elem *inode_list,
1334                                 u64 root, u64 extent_item_objectid,
1335                                 iterate_extent_inodes_t *iterate, void *ctx)
1336 {
1337         struct extent_inode_elem *eie;
1338         int ret = 0;
1339
1340         for (eie = inode_list; eie; eie = eie->next) {
1341                 pr_debug("ref for %llu resolved, key (%llu EXTEND_DATA %llu), "
1342                          "root %llu\n", extent_item_objectid,
1343                          eie->inum, eie->offset, root);
1344                 ret = iterate(eie->inum, eie->offset, root, ctx);
1345                 if (ret) {
1346                         pr_debug("stopping iteration for %llu due to ret=%d\n",
1347                                  extent_item_objectid, ret);
1348                         break;
1349                 }
1350         }
1351
1352         return ret;
1353 }
1354
1355 /*
1356  * calls iterate() for every inode that references the extent identified by
1357  * the given parameters.
1358  * when the iterator function returns a non-zero value, iteration stops.
1359  */
1360 int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
1361                                 u64 extent_item_objectid, u64 extent_item_pos,
1362                                 int search_commit_root,
1363                                 iterate_extent_inodes_t *iterate, void *ctx)
1364 {
1365         int ret;
1366         struct list_head data_refs = LIST_HEAD_INIT(data_refs);
1367         struct list_head shared_refs = LIST_HEAD_INIT(shared_refs);
1368         struct btrfs_trans_handle *trans;
1369         struct ulist *refs = NULL;
1370         struct ulist *roots = NULL;
1371         struct ulist_node *ref_node = NULL;
1372         struct ulist_node *root_node = NULL;
1373         struct seq_list seq_elem = {};
1374         struct seq_list tree_mod_seq_elem = {};
1375         struct ulist_iterator ref_uiter;
1376         struct ulist_iterator root_uiter;
1377         struct btrfs_delayed_ref_root *delayed_refs = NULL;
1378
1379         pr_debug("resolving all inodes for extent %llu\n",
1380                         extent_item_objectid);
1381
1382         if (search_commit_root) {
1383                 trans = BTRFS_BACKREF_SEARCH_COMMIT_ROOT;
1384         } else {
1385                 trans = btrfs_join_transaction(fs_info->extent_root);
1386                 if (IS_ERR(trans))
1387                         return PTR_ERR(trans);
1388
1389                 delayed_refs = &trans->transaction->delayed_refs;
1390                 spin_lock(&delayed_refs->lock);
1391                 btrfs_get_delayed_seq(delayed_refs, &seq_elem);
1392                 spin_unlock(&delayed_refs->lock);
1393                 btrfs_get_tree_mod_seq(fs_info, &tree_mod_seq_elem);
1394         }
1395
1396         ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
1397                                    seq_elem.seq, tree_mod_seq_elem.seq, &refs,
1398                                    &extent_item_pos);
1399         if (ret)
1400                 goto out;
1401
1402         ULIST_ITER_INIT(&ref_uiter);
1403         while (!ret && (ref_node = ulist_next(refs, &ref_uiter))) {
1404                 ret = btrfs_find_all_roots(trans, fs_info, ref_node->val,
1405                                                 seq_elem.seq,
1406                                                 tree_mod_seq_elem.seq, &roots);
1407                 if (ret)
1408                         break;
1409                 ULIST_ITER_INIT(&root_uiter);
1410                 while (!ret && (root_node = ulist_next(roots, &root_uiter))) {
1411                         pr_debug("root %llu references leaf %llu, data list "
1412                                  "%#lx\n", root_node->val, ref_node->val,
1413                                  ref_node->aux);
1414                         ret = iterate_leaf_refs(
1415                                 (struct extent_inode_elem *)ref_node->aux,
1416                                 root_node->val, extent_item_objectid,
1417                                 iterate, ctx);
1418                 }
1419                 ulist_free(roots);
1420                 roots = NULL;
1421         }
1422
1423         free_leaf_list(refs);
1424         ulist_free(roots);
1425 out:
1426         if (!search_commit_root) {
1427                 btrfs_put_tree_mod_seq(fs_info, &tree_mod_seq_elem);
1428                 btrfs_put_delayed_seq(delayed_refs, &seq_elem);
1429                 btrfs_end_transaction(trans, fs_info->extent_root);
1430         }
1431
1432         return ret;
1433 }
1434
1435 int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
1436                                 struct btrfs_path *path,
1437                                 iterate_extent_inodes_t *iterate, void *ctx)
1438 {
1439         int ret;
1440         u64 extent_item_pos;
1441         struct btrfs_key found_key;
1442         int search_commit_root = path->search_commit_root;
1443
1444         ret = extent_from_logical(fs_info, logical, path,
1445                                         &found_key);
1446         btrfs_release_path(path);
1447         if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1448                 ret = -EINVAL;
1449         if (ret < 0)
1450                 return ret;
1451
1452         extent_item_pos = logical - found_key.objectid;
1453         ret = iterate_extent_inodes(fs_info, found_key.objectid,
1454                                         extent_item_pos, search_commit_root,
1455                                         iterate, ctx);
1456
1457         return ret;
1458 }
1459
1460 static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
1461                                 struct btrfs_path *path,
1462                                 iterate_irefs_t *iterate, void *ctx)
1463 {
1464         int ret = 0;
1465         int slot;
1466         u32 cur;
1467         u32 len;
1468         u32 name_len;
1469         u64 parent = 0;
1470         int found = 0;
1471         struct extent_buffer *eb;
1472         struct btrfs_item *item;
1473         struct btrfs_inode_ref *iref;
1474         struct btrfs_key found_key;
1475
1476         while (!ret) {
1477                 path->leave_spinning = 1;
1478                 ret = inode_ref_info(inum, parent ? parent+1 : 0, fs_root, path,
1479                                         &found_key);
1480                 if (ret < 0)
1481                         break;
1482                 if (ret) {
1483                         ret = found ? 0 : -ENOENT;
1484                         break;
1485                 }
1486                 ++found;
1487
1488                 parent = found_key.offset;
1489                 slot = path->slots[0];
1490                 eb = path->nodes[0];
1491                 /* make sure we can use eb after releasing the path */
1492                 atomic_inc(&eb->refs);
1493                 btrfs_tree_read_lock(eb);
1494                 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1495                 btrfs_release_path(path);
1496
1497                 item = btrfs_item_nr(eb, slot);
1498                 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
1499
1500                 for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) {
1501                         name_len = btrfs_inode_ref_name_len(eb, iref);
1502                         /* path must be released before calling iterate()! */
1503                         pr_debug("following ref at offset %u for inode %llu in "
1504                                  "tree %llu\n", cur,
1505                                  (unsigned long long)found_key.objectid,
1506                                  (unsigned long long)fs_root->objectid);
1507                         ret = iterate(parent, iref, eb, ctx);
1508                         if (ret)
1509                                 break;
1510                         len = sizeof(*iref) + name_len;
1511                         iref = (struct btrfs_inode_ref *)((char *)iref + len);
1512                 }
1513                 btrfs_tree_read_unlock_blocking(eb);
1514                 free_extent_buffer(eb);
1515         }
1516
1517         btrfs_release_path(path);
1518
1519         return ret;
1520 }
1521
1522 /*
1523  * returns 0 if the path could be dumped (probably truncated)
1524  * returns <0 in case of an error
1525  */
1526 static int inode_to_path(u64 inum, struct btrfs_inode_ref *iref,
1527                                 struct extent_buffer *eb, void *ctx)
1528 {
1529         struct inode_fs_paths *ipath = ctx;
1530         char *fspath;
1531         char *fspath_min;
1532         int i = ipath->fspath->elem_cnt;
1533         const int s_ptr = sizeof(char *);
1534         u32 bytes_left;
1535
1536         bytes_left = ipath->fspath->bytes_left > s_ptr ?
1537                                         ipath->fspath->bytes_left - s_ptr : 0;
1538
1539         fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
1540         fspath = iref_to_path(ipath->fs_root, ipath->btrfs_path, iref, eb,
1541                                 inum, fspath_min, bytes_left);
1542         if (IS_ERR(fspath))
1543                 return PTR_ERR(fspath);
1544
1545         if (fspath > fspath_min) {
1546                 pr_debug("path resolved: %s\n", fspath);
1547                 ipath->fspath->val[i] = (u64)(unsigned long)fspath;
1548                 ++ipath->fspath->elem_cnt;
1549                 ipath->fspath->bytes_left = fspath - fspath_min;
1550         } else {
1551                 pr_debug("missed path, not enough space. missing bytes: %lu, "
1552                          "constructed so far: %s\n",
1553                          (unsigned long)(fspath_min - fspath), fspath_min);
1554                 ++ipath->fspath->elem_missed;
1555                 ipath->fspath->bytes_missing += fspath_min - fspath;
1556                 ipath->fspath->bytes_left = 0;
1557         }
1558
1559         return 0;
1560 }
1561
1562 /*
1563  * this dumps all file system paths to the inode into the ipath struct, provided
1564  * is has been created large enough. each path is zero-terminated and accessed
1565  * from ipath->fspath->val[i].
1566  * when it returns, there are ipath->fspath->elem_cnt number of paths available
1567  * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
1568  * number of missed paths in recored in ipath->fspath->elem_missed, otherwise,
1569  * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
1570  * have been needed to return all paths.
1571  */
1572 int paths_from_inode(u64 inum, struct inode_fs_paths *ipath)
1573 {
1574         return iterate_irefs(inum, ipath->fs_root, ipath->btrfs_path,
1575                                 inode_to_path, ipath);
1576 }
1577
1578 struct btrfs_data_container *init_data_container(u32 total_bytes)
1579 {
1580         struct btrfs_data_container *data;
1581         size_t alloc_bytes;
1582
1583         alloc_bytes = max_t(size_t, total_bytes, sizeof(*data));
1584         data = kmalloc(alloc_bytes, GFP_NOFS);
1585         if (!data)
1586                 return ERR_PTR(-ENOMEM);
1587
1588         if (total_bytes >= sizeof(*data)) {
1589                 data->bytes_left = total_bytes - sizeof(*data);
1590                 data->bytes_missing = 0;
1591         } else {
1592                 data->bytes_missing = sizeof(*data) - total_bytes;
1593                 data->bytes_left = 0;
1594         }
1595
1596         data->elem_cnt = 0;
1597         data->elem_missed = 0;
1598
1599         return data;
1600 }
1601
1602 /*
1603  * allocates space to return multiple file system paths for an inode.
1604  * total_bytes to allocate are passed, note that space usable for actual path
1605  * information will be total_bytes - sizeof(struct inode_fs_paths).
1606  * the returned pointer must be freed with free_ipath() in the end.
1607  */
1608 struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
1609                                         struct btrfs_path *path)
1610 {
1611         struct inode_fs_paths *ifp;
1612         struct btrfs_data_container *fspath;
1613
1614         fspath = init_data_container(total_bytes);
1615         if (IS_ERR(fspath))
1616                 return (void *)fspath;
1617
1618         ifp = kmalloc(sizeof(*ifp), GFP_NOFS);
1619         if (!ifp) {
1620                 kfree(fspath);
1621                 return ERR_PTR(-ENOMEM);
1622         }
1623
1624         ifp->btrfs_path = path;
1625         ifp->fspath = fspath;
1626         ifp->fs_root = fs_root;
1627
1628         return ifp;
1629 }
1630
1631 void free_ipath(struct inode_fs_paths *ipath)
1632 {
1633         if (!ipath)
1634                 return;
1635         kfree(ipath->fspath);
1636         kfree(ipath);
1637 }