?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
dmaengine: imx-sdma: Move message level to debug
[linux-2.6/btrfs-unstable.git] / fs / btrfs / backref.c
blobecbc63d3143e78d53a9ab0e3dd2091686e51d620
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 <linux/vmalloc.h>
20 #include "ctree.h"
21 #include "disk-io.h"
22 #include "backref.h"
23 #include "ulist.h"
24 #include "transaction.h"
25 #include "delayed-ref.h"
26 #include "locking.h"
27
28 /* Just an arbitrary number so we can be sure this happened */
29 #define BACKREF_FOUND_SHARED 6
30
31 struct extent_inode_elem {
32 u64 inum;
33 u64 offset;
34 struct extent_inode_elem *next;
35 };
36
37 static int check_extent_in_eb(struct btrfs_key *key, struct extent_buffer *eb,
38 struct btrfs_file_extent_item *fi,
39 u64 extent_item_pos,
40 struct extent_inode_elem **eie)
41 {
42 u64 offset = 0;
43 struct extent_inode_elem *e;
44
45 if (!btrfs_file_extent_compression(eb, fi) &&
46 !btrfs_file_extent_encryption(eb, fi) &&
47 !btrfs_file_extent_other_encoding(eb, fi)) {
48 u64 data_offset;
49 u64 data_len;
50
51 data_offset = btrfs_file_extent_offset(eb, fi);
52 data_len = btrfs_file_extent_num_bytes(eb, fi);
53
54 if (extent_item_pos < data_offset ||
55 extent_item_pos >= data_offset + data_len)
56 return 1;
57 offset = extent_item_pos - data_offset;
58 }
59
60 e = kmalloc(sizeof(*e), GFP_NOFS);
61 if (!e)
62 return -ENOMEM;
63
64 e->next = *eie;
65 e->inum = key->objectid;
66 e->offset = key->offset + offset;
67 *eie = e;
68
69 return 0;
70 }
71
72 static void free_inode_elem_list(struct extent_inode_elem *eie)
73 {
74 struct extent_inode_elem *eie_next;
75
76 for (; eie; eie = eie_next) {
77 eie_next = eie->next;
78 kfree(eie);
79 }
80 }
81
82 static int find_extent_in_eb(struct extent_buffer *eb, u64 wanted_disk_byte,
83 u64 extent_item_pos,
84 struct extent_inode_elem **eie)
85 {
86 u64 disk_byte;
87 struct btrfs_key key;
88 struct btrfs_file_extent_item *fi;
89 int slot;
90 int nritems;
91 int extent_type;
92 int ret;
93
94 /*
95 * from the shared data ref, we only have the leaf but we need
96 * the key. thus, we must look into all items and see that we
97 * find one (some) with a reference to our extent item.
98 */
99 nritems = btrfs_header_nritems(eb);
100 for (slot = 0; slot < nritems; ++slot) {
101 btrfs_item_key_to_cpu(eb, &key, slot);
102 if (key.type != BTRFS_EXTENT_DATA_KEY)
103 continue;
104 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
105 extent_type = btrfs_file_extent_type(eb, fi);
106 if (extent_type == BTRFS_FILE_EXTENT_INLINE)
107 continue;
108 /* don't skip BTRFS_FILE_EXTENT_PREALLOC, we can handle that */
109 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
110 if (disk_byte != wanted_disk_byte)
111 continue;
112
113 ret = check_extent_in_eb(&key, eb, fi, extent_item_pos, eie);
114 if (ret < 0)
115 return ret;
116 }
117
118 return 0;
119 }
120
121 /*
122 * this structure records all encountered refs on the way up to the root
123 */
124 struct __prelim_ref {
125 struct list_head list;
126 u64 root_id;
127 struct btrfs_key key_for_search;
128 int level;
129 int count;
130 struct extent_inode_elem *inode_list;
131 u64 parent;
132 u64 wanted_disk_byte;
133 };
134
135 static struct kmem_cache *btrfs_prelim_ref_cache;
136
137 int __init btrfs_prelim_ref_init(void)
138 {
139 btrfs_prelim_ref_cache = kmem_cache_create("btrfs_prelim_ref",
140 sizeof(struct __prelim_ref),
141 0,
142 SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
143 NULL);
144 if (!btrfs_prelim_ref_cache)
145 return -ENOMEM;
146 return 0;
147 }
148
149 void btrfs_prelim_ref_exit(void)
150 {
151 if (btrfs_prelim_ref_cache)
152 kmem_cache_destroy(btrfs_prelim_ref_cache);
153 }
154
155 /*
156 * the rules for all callers of this function are:
157 * - obtaining the parent is the goal
158 * - if you add a key, you must know that it is a correct key
159 * - if you cannot add the parent or a correct key, then we will look into the
160 * block later to set a correct key
161 *
162 * delayed refs
163 * ============
164 * backref type | shared | indirect | shared | indirect
165 * information | tree | tree | data | data
166 * --------------------+--------+----------+--------+----------
167 * parent logical | y | - | - | -
168 * key to resolve | - | y | y | y
169 * tree block logical | - | - | - | -
170 * root for resolving | y | y | y | y
171 *
172 * - column 1: we've the parent -> done
173 * - column 2, 3, 4: we use the key to find the parent
174 *
175 * on disk refs (inline or keyed)
176 * ==============================
177 * backref type | shared | indirect | shared | indirect
178 * information | tree | tree | data | data
179 * --------------------+--------+----------+--------+----------
180 * parent logical | y | - | y | -
181 * key to resolve | - | - | - | y
182 * tree block logical | y | y | y | y
183 * root for resolving | - | y | y | y
184 *
185 * - column 1, 3: we've the parent -> done
186 * - column 2: we take the first key from the block to find the parent
187 * (see __add_missing_keys)
188 * - column 4: we use the key to find the parent
189 *
190 * additional information that's available but not required to find the parent
191 * block might help in merging entries to gain some speed.
192 */
193
194 static int __add_prelim_ref(struct list_head *head, u64 root_id,
195 struct btrfs_key *key, int level,
196 u64 parent, u64 wanted_disk_byte, int count,
197 gfp_t gfp_mask)
198 {
199 struct __prelim_ref *ref;
200
201 if (root_id == BTRFS_DATA_RELOC_TREE_OBJECTID)
202 return 0;
203
204 ref = kmem_cache_alloc(btrfs_prelim_ref_cache, gfp_mask);
205 if (!ref)
206 return -ENOMEM;
207
208 ref->root_id = root_id;
209 if (key) {
210 ref->key_for_search = *key;
211 /*
212 * We can often find data backrefs with an offset that is too
213 * large (>= LLONG_MAX, maximum allowed file offset) due to
214 * underflows when subtracting a file's offset with the data
215 * offset of its corresponding extent data item. This can
216 * happen for example in the clone ioctl.
217 * So if we detect such case we set the search key's offset to
218 * zero to make sure we will find the matching file extent item
219 * at add_all_parents(), otherwise we will miss it because the
220 * offset taken form the backref is much larger then the offset
221 * of the file extent item. This can make us scan a very large
222 * number of file extent items, but at least it will not make
223 * us miss any.
224 * This is an ugly workaround for a behaviour that should have
225 * never existed, but it does and a fix for the clone ioctl
226 * would touch a lot of places, cause backwards incompatibility
227 * and would not fix the problem for extents cloned with older
228 * kernels.
229 */
230 if (ref->key_for_search.type == BTRFS_EXTENT_DATA_KEY &&
231 ref->key_for_search.offset >= LLONG_MAX)
232 ref->key_for_search.offset = 0;
233 } else {
234 memset(&ref->key_for_search, 0, sizeof(ref->key_for_search));
235 }
236
237 ref->inode_list = NULL;
238 ref->level = level;
239 ref->count = count;
240 ref->parent = parent;
241 ref->wanted_disk_byte = wanted_disk_byte;
242 list_add_tail(&ref->list, head);
243
244 return 0;
245 }
246
247 static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
248 struct ulist *parents, struct __prelim_ref *ref,
249 int level, u64 time_seq, const u64 *extent_item_pos,
250 u64 total_refs)
251 {
252 int ret = 0;
253 int slot;
254 struct extent_buffer *eb;
255 struct btrfs_key key;
256 struct btrfs_key *key_for_search = &ref->key_for_search;
257 struct btrfs_file_extent_item *fi;
258 struct extent_inode_elem *eie = NULL, *old = NULL;
259 u64 disk_byte;
260 u64 wanted_disk_byte = ref->wanted_disk_byte;
261 u64 count = 0;
262
263 if (level != 0) {
264 eb = path->nodes[level];
265 ret = ulist_add(parents, eb->start, 0, GFP_NOFS);
266 if (ret < 0)
267 return ret;
268 return 0;
269 }
270
271 /*
272 * We normally enter this function with the path already pointing to
273 * the first item to check. But sometimes, we may enter it with
274 * slot==nritems. In that case, go to the next leaf before we continue.
275 */
276 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
277 if (time_seq == (u64)-1)
278 ret = btrfs_next_leaf(root, path);
279 else
280 ret = btrfs_next_old_leaf(root, path, time_seq);
281 }
282
283 while (!ret && count < total_refs) {
284 eb = path->nodes[0];
285 slot = path->slots[0];
286
287 btrfs_item_key_to_cpu(eb, &key, slot);
288
289 if (key.objectid != key_for_search->objectid ||
290 key.type != BTRFS_EXTENT_DATA_KEY)
291 break;
292
293 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
294 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
295
296 if (disk_byte == wanted_disk_byte) {
297 eie = NULL;
298 old = NULL;
299 count++;
300 if (extent_item_pos) {
301 ret = check_extent_in_eb(&key, eb, fi,
302 *extent_item_pos,
303 &eie);
304 if (ret < 0)
305 break;
306 }
307 if (ret > 0)
308 goto next;
309 ret = ulist_add_merge_ptr(parents, eb->start,
310 eie, (void **)&old, GFP_NOFS);
311 if (ret < 0)
312 break;
313 if (!ret && extent_item_pos) {
314 while (old->next)
315 old = old->next;
316 old->next = eie;
317 }
318 eie = NULL;
319 }
320 next:
321 if (time_seq == (u64)-1)
322 ret = btrfs_next_item(root, path);
323 else
324 ret = btrfs_next_old_item(root, path, time_seq);
325 }
326
327 if (ret > 0)
328 ret = 0;
329 else if (ret < 0)
330 free_inode_elem_list(eie);
331 return ret;
332 }
333
334 /*
335 * resolve an indirect backref in the form (root_id, key, level)
336 * to a logical address
337 */
338 static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
339 struct btrfs_path *path, u64 time_seq,
340 struct __prelim_ref *ref,
341 struct ulist *parents,
342 const u64 *extent_item_pos, u64 total_refs)
343 {
344 struct btrfs_root *root;
345 struct btrfs_key root_key;
346 struct extent_buffer *eb;
347 int ret = 0;
348 int root_level;
349 int level = ref->level;
350 int index;
351
352 root_key.objectid = ref->root_id;
353 root_key.type = BTRFS_ROOT_ITEM_KEY;
354 root_key.offset = (u64)-1;
355
356 index = srcu_read_lock(&fs_info->subvol_srcu);
357
358 root = btrfs_read_fs_root_no_name(fs_info, &root_key);
359 if (IS_ERR(root)) {
360 srcu_read_unlock(&fs_info->subvol_srcu, index);
361 ret = PTR_ERR(root);
362 goto out;
363 }
364
365 if (path->search_commit_root)
366 root_level = btrfs_header_level(root->commit_root);
367 else if (time_seq == (u64)-1)
368 root_level = btrfs_header_level(root->node);
369 else
370 root_level = btrfs_old_root_level(root, time_seq);
371
372 if (root_level + 1 == level) {
373 srcu_read_unlock(&fs_info->subvol_srcu, index);
374 goto out;
375 }
376
377 path->lowest_level = level;
378 if (time_seq == (u64)-1)
379 ret = btrfs_search_slot(NULL, root, &ref->key_for_search, path,
380 0, 0);
381 else
382 ret = btrfs_search_old_slot(root, &ref->key_for_search, path,
383 time_seq);
384
385 /* root node has been locked, we can release @subvol_srcu safely here */
386 srcu_read_unlock(&fs_info->subvol_srcu, index);
387
388 pr_debug("search slot in root %llu (level %d, ref count %d) returned "
389 "%d for key (%llu %u %llu)\n",
390 ref->root_id, level, ref->count, ret,
391 ref->key_for_search.objectid, ref->key_for_search.type,
392 ref->key_for_search.offset);
393 if (ret < 0)
394 goto out;
395
396 eb = path->nodes[level];
397 while (!eb) {
398 if (WARN_ON(!level)) {
399 ret = 1;
400 goto out;
401 }
402 level--;
403 eb = path->nodes[level];
404 }
405
406 ret = add_all_parents(root, path, parents, ref, level, time_seq,
407 extent_item_pos, total_refs);
408 out:
409 path->lowest_level = 0;
410 btrfs_release_path(path);
411 return ret;
412 }
413
414 /*
415 * resolve all indirect backrefs from the list
416 */
417 static int __resolve_indirect_refs(struct btrfs_fs_info *fs_info,
418 struct btrfs_path *path, u64 time_seq,
419 struct list_head *head,
420 const u64 *extent_item_pos, u64 total_refs,
421 u64 root_objectid)
422 {
423 int err;
424 int ret = 0;
425 struct __prelim_ref *ref;
426 struct __prelim_ref *ref_safe;
427 struct __prelim_ref *new_ref;
428 struct ulist *parents;
429 struct ulist_node *node;
430 struct ulist_iterator uiter;
431
432 parents = ulist_alloc(GFP_NOFS);
433 if (!parents)
434 return -ENOMEM;
435
436 /*
437 * _safe allows us to insert directly after the current item without
438 * iterating over the newly inserted items.
439 * we're also allowed to re-assign ref during iteration.
440 */
441 list_for_each_entry_safe(ref, ref_safe, head, list) {
442 if (ref->parent) /* already direct */
443 continue;
444 if (ref->count == 0)
445 continue;
446 if (root_objectid && ref->root_id != root_objectid) {
447 ret = BACKREF_FOUND_SHARED;
448 goto out;
449 }
450 err = __resolve_indirect_ref(fs_info, path, time_seq, ref,
451 parents, extent_item_pos,
452 total_refs);
453 /*
454 * we can only tolerate ENOENT,otherwise,we should catch error
455 * and return directly.
456 */
457 if (err == -ENOENT) {
458 continue;
459 } else if (err) {
460 ret = err;
461 goto out;
462 }
463
464 /* we put the first parent into the ref at hand */
465 ULIST_ITER_INIT(&uiter);
466 node = ulist_next(parents, &uiter);
467 ref->parent = node ? node->val : 0;
468 ref->inode_list = node ?
469 (struct extent_inode_elem *)(uintptr_t)node->aux : NULL;
470
471 /* additional parents require new refs being added here */
472 while ((node = ulist_next(parents, &uiter))) {
473 new_ref = kmem_cache_alloc(btrfs_prelim_ref_cache,
474 GFP_NOFS);
475 if (!new_ref) {
476 ret = -ENOMEM;
477 goto out;
478 }
479 memcpy(new_ref, ref, sizeof(*ref));
480 new_ref->parent = node->val;
481 new_ref->inode_list = (struct extent_inode_elem *)
482 (uintptr_t)node->aux;
483 list_add(&new_ref->list, &ref->list);
484 }
485 ulist_reinit(parents);
486 }
487 out:
488 ulist_free(parents);
489 return ret;
490 }
491
492 static inline int ref_for_same_block(struct __prelim_ref *ref1,
493 struct __prelim_ref *ref2)
494 {
495 if (ref1->level != ref2->level)
496 return 0;
497 if (ref1->root_id != ref2->root_id)
498 return 0;
499 if (ref1->key_for_search.type != ref2->key_for_search.type)
500 return 0;
501 if (ref1->key_for_search.objectid != ref2->key_for_search.objectid)
502 return 0;
503 if (ref1->key_for_search.offset != ref2->key_for_search.offset)
504 return 0;
505 if (ref1->parent != ref2->parent)
506 return 0;
507
508 return 1;
509 }
510
511 /*
512 * read tree blocks and add keys where required.
513 */
514 static int __add_missing_keys(struct btrfs_fs_info *fs_info,
515 struct list_head *head)
516 {
517 struct list_head *pos;
518 struct extent_buffer *eb;
519
520 list_for_each(pos, head) {
521 struct __prelim_ref *ref;
522 ref = list_entry(pos, struct __prelim_ref, list);
523
524 if (ref->parent)
525 continue;
526 if (ref->key_for_search.type)
527 continue;
528 BUG_ON(!ref->wanted_disk_byte);
529 eb = read_tree_block(fs_info->tree_root, ref->wanted_disk_byte,
530 0);
531 if (IS_ERR(eb)) {
532 return PTR_ERR(eb);
533 } else if (!extent_buffer_uptodate(eb)) {
534 free_extent_buffer(eb);
535 return -EIO;
536 }
537 btrfs_tree_read_lock(eb);
538 if (btrfs_header_level(eb) == 0)
539 btrfs_item_key_to_cpu(eb, &ref->key_for_search, 0);
540 else
541 btrfs_node_key_to_cpu(eb, &ref->key_for_search, 0);
542 btrfs_tree_read_unlock(eb);
543 free_extent_buffer(eb);
544 }
545 return 0;
546 }
547
548 /*
549 * merge backrefs and adjust counts accordingly
550 *
551 * mode = 1: merge identical keys, if key is set
552 * FIXME: if we add more keys in __add_prelim_ref, we can merge more here.
553 * additionally, we could even add a key range for the blocks we
554 * looked into to merge even more (-> replace unresolved refs by those
555 * having a parent).
556 * mode = 2: merge identical parents
557 */
558 static void __merge_refs(struct list_head *head, int mode)
559 {
560 struct list_head *pos1;
561
562 list_for_each(pos1, head) {
563 struct list_head *n2;
564 struct list_head *pos2;
565 struct __prelim_ref *ref1;
566
567 ref1 = list_entry(pos1, struct __prelim_ref, list);
568
569 for (pos2 = pos1->next, n2 = pos2->next; pos2 != head;
570 pos2 = n2, n2 = pos2->next) {
571 struct __prelim_ref *ref2;
572 struct __prelim_ref *xchg;
573 struct extent_inode_elem *eie;
574
575 ref2 = list_entry(pos2, struct __prelim_ref, list);
576
577 if (!ref_for_same_block(ref1, ref2))
578 continue;
579 if (mode == 1) {
580 if (!ref1->parent && ref2->parent) {
581 xchg = ref1;
582 ref1 = ref2;
583 ref2 = xchg;
584 }
585 } else {
586 if (ref1->parent != ref2->parent)
587 continue;
588 }
589
590 eie = ref1->inode_list;
591 while (eie && eie->next)
592 eie = eie->next;
593 if (eie)
594 eie->next = ref2->inode_list;
595 else
596 ref1->inode_list = ref2->inode_list;
597 ref1->count += ref2->count;
598
599 list_del(&ref2->list);
600 kmem_cache_free(btrfs_prelim_ref_cache, ref2);
601 }
602
603 }
604 }
605
606 /*
607 * add all currently queued delayed refs from this head whose seq nr is
608 * smaller or equal that seq to the list
609 */
610 static int __add_delayed_refs(struct btrfs_delayed_ref_head *head, u64 seq,
611 struct list_head *prefs, u64 *total_refs,
612 u64 inum)
613 {
614 struct btrfs_delayed_ref_node *node;
615 struct btrfs_delayed_extent_op *extent_op = head->extent_op;
616 struct btrfs_key key;
617 struct btrfs_key op_key = {0};
618 int sgn;
619 int ret = 0;
620
621 if (extent_op && extent_op->update_key)
622 btrfs_disk_key_to_cpu(&op_key, &extent_op->key);
623
624 spin_lock(&head->lock);
625 list_for_each_entry(node, &head->ref_list, list) {
626 if (node->seq > seq)
627 continue;
628
629 switch (node->action) {
630 case BTRFS_ADD_DELAYED_EXTENT:
631 case BTRFS_UPDATE_DELAYED_HEAD:
632 WARN_ON(1);
633 continue;
634 case BTRFS_ADD_DELAYED_REF:
635 sgn = 1;
636 break;
637 case BTRFS_DROP_DELAYED_REF:
638 sgn = -1;
639 break;
640 default:
641 BUG_ON(1);
642 }
643 *total_refs += (node->ref_mod * sgn);
644 switch (node->type) {
645 case BTRFS_TREE_BLOCK_REF_KEY: {
646 struct btrfs_delayed_tree_ref *ref;
647
648 ref = btrfs_delayed_node_to_tree_ref(node);
649 ret = __add_prelim_ref(prefs, ref->root, &op_key,
650 ref->level + 1, 0, node->bytenr,
651 node->ref_mod * sgn, GFP_ATOMIC);
652 break;
653 }
654 case BTRFS_SHARED_BLOCK_REF_KEY: {
655 struct btrfs_delayed_tree_ref *ref;
656
657 ref = btrfs_delayed_node_to_tree_ref(node);
658 ret = __add_prelim_ref(prefs, 0, NULL,
659 ref->level + 1, ref->parent,
660 node->bytenr,
661 node->ref_mod * sgn, GFP_ATOMIC);
662 break;
663 }
664 case BTRFS_EXTENT_DATA_REF_KEY: {
665 struct btrfs_delayed_data_ref *ref;
666 ref = btrfs_delayed_node_to_data_ref(node);
667
668 key.objectid = ref->objectid;
669 key.type = BTRFS_EXTENT_DATA_KEY;
670 key.offset = ref->offset;
671
672 /*
673 * Found a inum that doesn't match our known inum, we
674 * know it's shared.
675 */
676 if (inum && ref->objectid != inum) {
677 ret = BACKREF_FOUND_SHARED;
678 break;
679 }
680
681 ret = __add_prelim_ref(prefs, ref->root, &key, 0, 0,
682 node->bytenr,
683 node->ref_mod * sgn, GFP_ATOMIC);
684 break;
685 }
686 case BTRFS_SHARED_DATA_REF_KEY: {
687 struct btrfs_delayed_data_ref *ref;
688
689 ref = btrfs_delayed_node_to_data_ref(node);
690 ret = __add_prelim_ref(prefs, 0, NULL, 0,
691 ref->parent, node->bytenr,
692 node->ref_mod * sgn, GFP_ATOMIC);
693 break;
694 }
695 default:
696 WARN_ON(1);
697 }
698 if (ret)
699 break;
700 }
701 spin_unlock(&head->lock);
702 return ret;
703 }
704
705 /*
706 * add all inline backrefs for bytenr to the list
707 */
708 static int __add_inline_refs(struct btrfs_fs_info *fs_info,
709 struct btrfs_path *path, u64 bytenr,
710 int *info_level, struct list_head *prefs,
711 u64 *total_refs, u64 inum)
712 {
713 int ret = 0;
714 int slot;
715 struct extent_buffer *leaf;
716 struct btrfs_key key;
717 struct btrfs_key found_key;
718 unsigned long ptr;
719 unsigned long end;
720 struct btrfs_extent_item *ei;
721 u64 flags;
722 u64 item_size;
723
724 /*
725 * enumerate all inline refs
726 */
727 leaf = path->nodes[0];
728 slot = path->slots[0];
729
730 item_size = btrfs_item_size_nr(leaf, slot);
731 BUG_ON(item_size < sizeof(*ei));
732
733 ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
734 flags = btrfs_extent_flags(leaf, ei);
735 *total_refs += btrfs_extent_refs(leaf, ei);
736 btrfs_item_key_to_cpu(leaf, &found_key, slot);
737
738 ptr = (unsigned long)(ei + 1);
739 end = (unsigned long)ei + item_size;
740
741 if (found_key.type == BTRFS_EXTENT_ITEM_KEY &&
742 flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
743 struct btrfs_tree_block_info *info;
744
745 info = (struct btrfs_tree_block_info *)ptr;
746 *info_level = btrfs_tree_block_level(leaf, info);
747 ptr += sizeof(struct btrfs_tree_block_info);
748 BUG_ON(ptr > end);
749 } else if (found_key.type == BTRFS_METADATA_ITEM_KEY) {
750 *info_level = found_key.offset;
751 } else {
752 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
753 }
754
755 while (ptr < end) {
756 struct btrfs_extent_inline_ref *iref;
757 u64 offset;
758 int type;
759
760 iref = (struct btrfs_extent_inline_ref *)ptr;
761 type = btrfs_extent_inline_ref_type(leaf, iref);
762 offset = btrfs_extent_inline_ref_offset(leaf, iref);
763
764 switch (type) {
765 case BTRFS_SHARED_BLOCK_REF_KEY:
766 ret = __add_prelim_ref(prefs, 0, NULL,
767 *info_level + 1, offset,
768 bytenr, 1, GFP_NOFS);
769 break;
770 case BTRFS_SHARED_DATA_REF_KEY: {
771 struct btrfs_shared_data_ref *sdref;
772 int count;
773
774 sdref = (struct btrfs_shared_data_ref *)(iref + 1);
775 count = btrfs_shared_data_ref_count(leaf, sdref);
776 ret = __add_prelim_ref(prefs, 0, NULL, 0, offset,
777 bytenr, count, GFP_NOFS);
778 break;
779 }
780 case BTRFS_TREE_BLOCK_REF_KEY:
781 ret = __add_prelim_ref(prefs, offset, NULL,
782 *info_level + 1, 0,
783 bytenr, 1, GFP_NOFS);
784 break;
785 case BTRFS_EXTENT_DATA_REF_KEY: {
786 struct btrfs_extent_data_ref *dref;
787 int count;
788 u64 root;
789
790 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
791 count = btrfs_extent_data_ref_count(leaf, dref);
792 key.objectid = btrfs_extent_data_ref_objectid(leaf,
793 dref);
794 key.type = BTRFS_EXTENT_DATA_KEY;
795 key.offset = btrfs_extent_data_ref_offset(leaf, dref);
796
797 if (inum && key.objectid != inum) {
798 ret = BACKREF_FOUND_SHARED;
799 break;
800 }
801
802 root = btrfs_extent_data_ref_root(leaf, dref);
803 ret = __add_prelim_ref(prefs, root, &key, 0, 0,
804 bytenr, count, GFP_NOFS);
805 break;
806 }
807 default:
808 WARN_ON(1);
809 }
810 if (ret)
811 return ret;
812 ptr += btrfs_extent_inline_ref_size(type);
813 }
814
815 return 0;
816 }
817
818 /*
819 * add all non-inline backrefs for bytenr to the list
820 */
821 static int __add_keyed_refs(struct btrfs_fs_info *fs_info,
822 struct btrfs_path *path, u64 bytenr,
823 int info_level, struct list_head *prefs, u64 inum)
824 {
825 struct btrfs_root *extent_root = fs_info->extent_root;
826 int ret;
827 int slot;
828 struct extent_buffer *leaf;
829 struct btrfs_key key;
830
831 while (1) {
832 ret = btrfs_next_item(extent_root, path);
833 if (ret < 0)
834 break;
835 if (ret) {
836 ret = 0;
837 break;
838 }
839
840 slot = path->slots[0];
841 leaf = path->nodes[0];
842 btrfs_item_key_to_cpu(leaf, &key, slot);
843
844 if (key.objectid != bytenr)
845 break;
846 if (key.type < BTRFS_TREE_BLOCK_REF_KEY)
847 continue;
848 if (key.type > BTRFS_SHARED_DATA_REF_KEY)
849 break;
850
851 switch (key.type) {
852 case BTRFS_SHARED_BLOCK_REF_KEY:
853 ret = __add_prelim_ref(prefs, 0, NULL,
854 info_level + 1, key.offset,
855 bytenr, 1, GFP_NOFS);
856 break;
857 case BTRFS_SHARED_DATA_REF_KEY: {
858 struct btrfs_shared_data_ref *sdref;
859 int count;
860
861 sdref = btrfs_item_ptr(leaf, slot,
862 struct btrfs_shared_data_ref);
863 count = btrfs_shared_data_ref_count(leaf, sdref);
864 ret = __add_prelim_ref(prefs, 0, NULL, 0, key.offset,
865 bytenr, count, GFP_NOFS);
866 break;
867 }
868 case BTRFS_TREE_BLOCK_REF_KEY:
869 ret = __add_prelim_ref(prefs, key.offset, NULL,
870 info_level + 1, 0,
871 bytenr, 1, GFP_NOFS);
872 break;
873 case BTRFS_EXTENT_DATA_REF_KEY: {
874 struct btrfs_extent_data_ref *dref;
875 int count;
876 u64 root;
877
878 dref = btrfs_item_ptr(leaf, slot,
879 struct btrfs_extent_data_ref);
880 count = btrfs_extent_data_ref_count(leaf, dref);
881 key.objectid = btrfs_extent_data_ref_objectid(leaf,
882 dref);
883 key.type = BTRFS_EXTENT_DATA_KEY;
884 key.offset = btrfs_extent_data_ref_offset(leaf, dref);
885
886 if (inum && key.objectid != inum) {
887 ret = BACKREF_FOUND_SHARED;
888 break;
889 }
890
891 root = btrfs_extent_data_ref_root(leaf, dref);
892 ret = __add_prelim_ref(prefs, root, &key, 0, 0,
893 bytenr, count, GFP_NOFS);
894 break;
895 }
896 default:
897 WARN_ON(1);
898 }
899 if (ret)
900 return ret;
901
902 }
903
904 return ret;
905 }
906
907 /*
908 * this adds all existing backrefs (inline backrefs, backrefs and delayed
909 * refs) for the given bytenr to the refs list, merges duplicates and resolves
910 * indirect refs to their parent bytenr.
911 * When roots are found, they're added to the roots list
912 *
913 * NOTE: This can return values > 0
914 *
915 * If time_seq is set to (u64)-1, it will not search delayed_refs, and behave
916 * much like trans == NULL case, the difference only lies in it will not
917 * commit root.
918 * The special case is for qgroup to search roots in commit_transaction().
919 *
920 * FIXME some caching might speed things up
921 */
922 static int find_parent_nodes(struct btrfs_trans_handle *trans,
923 struct btrfs_fs_info *fs_info, u64 bytenr,
924 u64 time_seq, struct ulist *refs,
925 struct ulist *roots, const u64 *extent_item_pos,
926 u64 root_objectid, u64 inum)
927 {
928 struct btrfs_key key;
929 struct btrfs_path *path;
930 struct btrfs_delayed_ref_root *delayed_refs = NULL;
931 struct btrfs_delayed_ref_head *head;
932 int info_level = 0;
933 int ret;
934 struct list_head prefs_delayed;
935 struct list_head prefs;
936 struct __prelim_ref *ref;
937 struct extent_inode_elem *eie = NULL;
938 u64 total_refs = 0;
939
940 INIT_LIST_HEAD(&prefs);
941 INIT_LIST_HEAD(&prefs_delayed);
942
943 key.objectid = bytenr;
944 key.offset = (u64)-1;
945 if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
946 key.type = BTRFS_METADATA_ITEM_KEY;
947 else
948 key.type = BTRFS_EXTENT_ITEM_KEY;
949
950 path = btrfs_alloc_path();
951 if (!path)
952 return -ENOMEM;
953 if (!trans) {
954 path->search_commit_root = 1;
955 path->skip_locking = 1;
956 }
957
958 if (time_seq == (u64)-1)
959 path->skip_locking = 1;
960
961 /*
962 * grab both a lock on the path and a lock on the delayed ref head.
963 * We need both to get a consistent picture of how the refs look
964 * at a specified point in time
965 */
966 again:
967 head = NULL;
968
969 ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
970 if (ret < 0)
971 goto out;
972 BUG_ON(ret == 0);
973
974 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
975 if (trans && likely(trans->type != __TRANS_DUMMY) &&
976 time_seq != (u64)-1) {
977 #else
978 if (trans && time_seq != (u64)-1) {
979 #endif
980 /*
981 * look if there are updates for this ref queued and lock the
982 * head
983 */
984 delayed_refs = &trans->transaction->delayed_refs;
985 spin_lock(&delayed_refs->lock);
986 head = btrfs_find_delayed_ref_head(trans, bytenr);
987 if (head) {
988 if (!mutex_trylock(&head->mutex)) {
989 atomic_inc(&head->node.refs);
990 spin_unlock(&delayed_refs->lock);
991
992 btrfs_release_path(path);
993
994 /*
995 * Mutex was contended, block until it's
996 * released and try again
997 */
998 mutex_lock(&head->mutex);
999 mutex_unlock(&head->mutex);
1000 btrfs_put_delayed_ref(&head->node);
1001 goto again;
1002 }
1003 spin_unlock(&delayed_refs->lock);
1004 ret = __add_delayed_refs(head, time_seq,
1005 &prefs_delayed, &total_refs,
1006 inum);
1007 mutex_unlock(&head->mutex);
1008 if (ret)
1009 goto out;
1010 } else {
1011 spin_unlock(&delayed_refs->lock);
1012 }
1013 }
1014
1015 if (path->slots[0]) {
1016 struct extent_buffer *leaf;
1017 int slot;
1018
1019 path->slots[0]--;
1020 leaf = path->nodes[0];
1021 slot = path->slots[0];
1022 btrfs_item_key_to_cpu(leaf, &key, slot);
1023 if (key.objectid == bytenr &&
1024 (key.type == BTRFS_EXTENT_ITEM_KEY ||
1025 key.type == BTRFS_METADATA_ITEM_KEY)) {
1026 ret = __add_inline_refs(fs_info, path, bytenr,
1027 &info_level, &prefs,
1028 &total_refs, inum);
1029 if (ret)
1030 goto out;
1031 ret = __add_keyed_refs(fs_info, path, bytenr,
1032 info_level, &prefs, inum);
1033 if (ret)
1034 goto out;
1035 }
1036 }
1037 btrfs_release_path(path);
1038
1039 list_splice_init(&prefs_delayed, &prefs);
1040
1041 ret = __add_missing_keys(fs_info, &prefs);
1042 if (ret)
1043 goto out;
1044
1045 __merge_refs(&prefs, 1);
1046
1047 ret = __resolve_indirect_refs(fs_info, path, time_seq, &prefs,
1048 extent_item_pos, total_refs,
1049 root_objectid);
1050 if (ret)
1051 goto out;
1052
1053 __merge_refs(&prefs, 2);
1054
1055 while (!list_empty(&prefs)) {
1056 ref = list_first_entry(&prefs, struct __prelim_ref, list);
1057 WARN_ON(ref->count < 0);
1058 if (roots && ref->count && ref->root_id && ref->parent == 0) {
1059 if (root_objectid && ref->root_id != root_objectid) {
1060 ret = BACKREF_FOUND_SHARED;
1061 goto out;
1062 }
1063
1064 /* no parent == root of tree */
1065 ret = ulist_add(roots, ref->root_id, 0, GFP_NOFS);
1066 if (ret < 0)
1067 goto out;
1068 }
1069 if (ref->count && ref->parent) {
1070 if (extent_item_pos && !ref->inode_list &&
1071 ref->level == 0) {
1072 struct extent_buffer *eb;
1073
1074 eb = read_tree_block(fs_info->extent_root,
1075 ref->parent, 0);
1076 if (IS_ERR(eb)) {
1077 ret = PTR_ERR(eb);
1078 goto out;
1079 } else if (!extent_buffer_uptodate(eb)) {
1080 free_extent_buffer(eb);
1081 ret = -EIO;
1082 goto out;
1083 }
1084 btrfs_tree_read_lock(eb);
1085 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1086 ret = find_extent_in_eb(eb, bytenr,
1087 *extent_item_pos, &eie);
1088 btrfs_tree_read_unlock_blocking(eb);
1089 free_extent_buffer(eb);
1090 if (ret < 0)
1091 goto out;
1092 ref->inode_list = eie;
1093 }
1094 ret = ulist_add_merge_ptr(refs, ref->parent,
1095 ref->inode_list,
1096 (void **)&eie, GFP_NOFS);
1097 if (ret < 0)
1098 goto out;
1099 if (!ret && extent_item_pos) {
1100 /*
1101 * we've recorded that parent, so we must extend
1102 * its inode list here
1103 */
1104 BUG_ON(!eie);
1105 while (eie->next)
1106 eie = eie->next;
1107 eie->next = ref->inode_list;
1108 }
1109 eie = NULL;
1110 }
1111 list_del(&ref->list);
1112 kmem_cache_free(btrfs_prelim_ref_cache, ref);
1113 }
1114
1115 out:
1116 btrfs_free_path(path);
1117 while (!list_empty(&prefs)) {
1118 ref = list_first_entry(&prefs, struct __prelim_ref, list);
1119 list_del(&ref->list);
1120 kmem_cache_free(btrfs_prelim_ref_cache, ref);
1121 }
1122 while (!list_empty(&prefs_delayed)) {
1123 ref = list_first_entry(&prefs_delayed, struct __prelim_ref,
1124 list);
1125 list_del(&ref->list);
1126 kmem_cache_free(btrfs_prelim_ref_cache, ref);
1127 }
1128 if (ret < 0)
1129 free_inode_elem_list(eie);
1130 return ret;
1131 }
1132
1133 static void free_leaf_list(struct ulist *blocks)
1134 {
1135 struct ulist_node *node = NULL;
1136 struct extent_inode_elem *eie;
1137 struct ulist_iterator uiter;
1138
1139 ULIST_ITER_INIT(&uiter);
1140 while ((node = ulist_next(blocks, &uiter))) {
1141 if (!node->aux)
1142 continue;
1143 eie = (struct extent_inode_elem *)(uintptr_t)node->aux;
1144 free_inode_elem_list(eie);
1145 node->aux = 0;
1146 }
1147
1148 ulist_free(blocks);
1149 }
1150
1151 /*
1152 * Finds all leafs with a reference to the specified combination of bytenr and
1153 * offset. key_list_head will point to a list of corresponding keys (caller must
1154 * free each list element). The leafs will be stored in the leafs ulist, which
1155 * must be freed with ulist_free.
1156 *
1157 * returns 0 on success, <0 on error
1158 */
1159 static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
1160 struct btrfs_fs_info *fs_info, u64 bytenr,
1161 u64 time_seq, struct ulist **leafs,
1162 const u64 *extent_item_pos)
1163 {
1164 int ret;
1165
1166 *leafs = ulist_alloc(GFP_NOFS);
1167 if (!*leafs)
1168 return -ENOMEM;
1169
1170 ret = find_parent_nodes(trans, fs_info, bytenr,
1171 time_seq, *leafs, NULL, extent_item_pos, 0, 0);
1172 if (ret < 0 && ret != -ENOENT) {
1173 free_leaf_list(*leafs);
1174 return ret;
1175 }
1176
1177 return 0;
1178 }
1179
1180 /*
1181 * walk all backrefs for a given extent to find all roots that reference this
1182 * extent. Walking a backref means finding all extents that reference this
1183 * extent and in turn walk the backrefs of those, too. Naturally this is a
1184 * recursive process, but here it is implemented in an iterative fashion: We
1185 * find all referencing extents for the extent in question and put them on a
1186 * list. In turn, we find all referencing extents for those, further appending
1187 * to the list. The way we iterate the list allows adding more elements after
1188 * the current while iterating. The process stops when we reach the end of the
1189 * list. Found roots are added to the roots list.
1190 *
1191 * returns 0 on success, < 0 on error.
1192 */
1193 static int __btrfs_find_all_roots(struct btrfs_trans_handle *trans,
1194 struct btrfs_fs_info *fs_info, u64 bytenr,
1195 u64 time_seq, struct ulist **roots)
1196 {
1197 struct ulist *tmp;
1198 struct ulist_node *node = NULL;
1199 struct ulist_iterator uiter;
1200 int ret;
1201
1202 tmp = ulist_alloc(GFP_NOFS);
1203 if (!tmp)
1204 return -ENOMEM;
1205 *roots = ulist_alloc(GFP_NOFS);
1206 if (!*roots) {
1207 ulist_free(tmp);
1208 return -ENOMEM;
1209 }
1210
1211 ULIST_ITER_INIT(&uiter);
1212 while (1) {
1213 ret = find_parent_nodes(trans, fs_info, bytenr,
1214 time_seq, tmp, *roots, NULL, 0, 0);
1215 if (ret < 0 && ret != -ENOENT) {
1216 ulist_free(tmp);
1217 ulist_free(*roots);
1218 return ret;
1219 }
1220 node = ulist_next(tmp, &uiter);
1221 if (!node)
1222 break;
1223 bytenr = node->val;
1224 cond_resched();
1225 }
1226
1227 ulist_free(tmp);
1228 return 0;
1229 }
1230
1231 int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
1232 struct btrfs_fs_info *fs_info, u64 bytenr,
1233 u64 time_seq, struct ulist **roots)
1234 {
1235 int ret;
1236
1237 if (!trans)
1238 down_read(&fs_info->commit_root_sem);
1239 ret = __btrfs_find_all_roots(trans, fs_info, bytenr, time_seq, roots);
1240 if (!trans)
1241 up_read(&fs_info->commit_root_sem);
1242 return ret;
1243 }
1244
1245 /**
1246 * btrfs_check_shared - tell us whether an extent is shared
1247 *
1248 * @trans: optional trans handle
1249 *
1250 * btrfs_check_shared uses the backref walking code but will short
1251 * circuit as soon as it finds a root or inode that doesn't match the
1252 * one passed in. This provides a significant performance benefit for
1253 * callers (such as fiemap) which want to know whether the extent is
1254 * shared but do not need a ref count.
1255 *
1256 * Return: 0 if extent is not shared, 1 if it is shared, < 0 on error.
1257 */
1258 int btrfs_check_shared(struct btrfs_trans_handle *trans,
1259 struct btrfs_fs_info *fs_info, u64 root_objectid,
1260 u64 inum, u64 bytenr)
1261 {
1262 struct ulist *tmp = NULL;
1263 struct ulist *roots = NULL;
1264 struct ulist_iterator uiter;
1265 struct ulist_node *node;
1266 struct seq_list elem = SEQ_LIST_INIT(elem);
1267 int ret = 0;
1268
1269 tmp = ulist_alloc(GFP_NOFS);
1270 roots = ulist_alloc(GFP_NOFS);
1271 if (!tmp || !roots) {
1272 ulist_free(tmp);
1273 ulist_free(roots);
1274 return -ENOMEM;
1275 }
1276
1277 if (trans)
1278 btrfs_get_tree_mod_seq(fs_info, &elem);
1279 else
1280 down_read(&fs_info->commit_root_sem);
1281 ULIST_ITER_INIT(&uiter);
1282 while (1) {
1283 ret = find_parent_nodes(trans, fs_info, bytenr, elem.seq, tmp,
1284 roots, NULL, root_objectid, inum);
1285 if (ret == BACKREF_FOUND_SHARED) {
1286 /* this is the only condition under which we return 1 */
1287 ret = 1;
1288 break;
1289 }
1290 if (ret < 0 && ret != -ENOENT)
1291 break;
1292 ret = 0;
1293 node = ulist_next(tmp, &uiter);
1294 if (!node)
1295 break;
1296 bytenr = node->val;
1297 cond_resched();
1298 }
1299 if (trans)
1300 btrfs_put_tree_mod_seq(fs_info, &elem);
1301 else
1302 up_read(&fs_info->commit_root_sem);
1303 ulist_free(tmp);
1304 ulist_free(roots);
1305 return ret;
1306 }
1307
1308 int btrfs_find_one_extref(struct btrfs_root *root, u64 inode_objectid,
1309 u64 start_off, struct btrfs_path *path,
1310 struct btrfs_inode_extref **ret_extref,
1311 u64 *found_off)
1312 {
1313 int ret, slot;
1314 struct btrfs_key key;
1315 struct btrfs_key found_key;
1316 struct btrfs_inode_extref *extref;
1317 struct extent_buffer *leaf;
1318 unsigned long ptr;
1319
1320 key.objectid = inode_objectid;
1321 key.type = BTRFS_INODE_EXTREF_KEY;
1322 key.offset = start_off;
1323
1324 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1325 if (ret < 0)
1326 return ret;
1327
1328 while (1) {
1329 leaf = path->nodes[0];
1330 slot = path->slots[0];
1331 if (slot >= btrfs_header_nritems(leaf)) {
1332 /*
1333 * If the item at offset is not found,
1334 * btrfs_search_slot will point us to the slot
1335 * where it should be inserted. In our case
1336 * that will be the slot directly before the
1337 * next INODE_REF_KEY_V2 item. In the case
1338 * that we're pointing to the last slot in a
1339 * leaf, we must move one leaf over.
1340 */
1341 ret = btrfs_next_leaf(root, path);
1342 if (ret) {
1343 if (ret >= 1)
1344 ret = -ENOENT;
1345 break;
1346 }
1347 continue;
1348 }
1349
1350 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1351
1352 /*
1353 * Check that we're still looking at an extended ref key for
1354 * this particular objectid. If we have different
1355 * objectid or type then there are no more to be found
1356 * in the tree and we can exit.
1357 */
1358 ret = -ENOENT;
1359 if (found_key.objectid != inode_objectid)
1360 break;
1361 if (found_key.type != BTRFS_INODE_EXTREF_KEY)
1362 break;
1363
1364 ret = 0;
1365 ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
1366 extref = (struct btrfs_inode_extref *)ptr;
1367 *ret_extref = extref;
1368 if (found_off)
1369 *found_off = found_key.offset;
1370 break;
1371 }
1372
1373 return ret;
1374 }
1375
1376 /*
1377 * this iterates to turn a name (from iref/extref) into a full filesystem path.
1378 * Elements of the path are separated by '/' and the path is guaranteed to be
1379 * 0-terminated. the path is only given within the current file system.
1380 * Therefore, it never starts with a '/'. the caller is responsible to provide
1381 * "size" bytes in "dest". the dest buffer will be filled backwards. finally,
1382 * the start point of the resulting string is returned. this pointer is within
1383 * dest, normally.
1384 * in case the path buffer would overflow, the pointer is decremented further
1385 * as if output was written to the buffer, though no more output is actually
1386 * generated. that way, the caller can determine how much space would be
1387 * required for the path to fit into the buffer. in that case, the returned
1388 * value will be smaller than dest. callers must check this!
1389 */
1390 char *btrfs_ref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
1391 u32 name_len, unsigned long name_off,
1392 struct extent_buffer *eb_in, u64 parent,
1393 char *dest, u32 size)
1394 {
1395 int slot;
1396 u64 next_inum;
1397 int ret;
1398 s64 bytes_left = ((s64)size) - 1;
1399 struct extent_buffer *eb = eb_in;
1400 struct btrfs_key found_key;
1401 int leave_spinning = path->leave_spinning;
1402 struct btrfs_inode_ref *iref;
1403
1404 if (bytes_left >= 0)
1405 dest[bytes_left] = '\0';
1406
1407 path->leave_spinning = 1;
1408 while (1) {
1409 bytes_left -= name_len;
1410 if (bytes_left >= 0)
1411 read_extent_buffer(eb, dest + bytes_left,
1412 name_off, name_len);
1413 if (eb != eb_in) {
1414 btrfs_tree_read_unlock_blocking(eb);
1415 free_extent_buffer(eb);
1416 }
1417 ret = btrfs_find_item(fs_root, path, parent, 0,
1418 BTRFS_INODE_REF_KEY, &found_key);
1419 if (ret > 0)
1420 ret = -ENOENT;
1421 if (ret)
1422 break;
1423
1424 next_inum = found_key.offset;
1425
1426 /* regular exit ahead */
1427 if (parent == next_inum)
1428 break;
1429
1430 slot = path->slots[0];
1431 eb = path->nodes[0];
1432 /* make sure we can use eb after releasing the path */
1433 if (eb != eb_in) {
1434 atomic_inc(&eb->refs);
1435 btrfs_tree_read_lock(eb);
1436 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1437 }
1438 btrfs_release_path(path);
1439 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
1440
1441 name_len = btrfs_inode_ref_name_len(eb, iref);
1442 name_off = (unsigned long)(iref + 1);
1443
1444 parent = next_inum;
1445 --bytes_left;
1446 if (bytes_left >= 0)
1447 dest[bytes_left] = '/';
1448 }
1449
1450 btrfs_release_path(path);
1451 path->leave_spinning = leave_spinning;
1452
1453 if (ret)
1454 return ERR_PTR(ret);
1455
1456 return dest + bytes_left;
1457 }
1458
1459 /*
1460 * this makes the path point to (logical EXTENT_ITEM *)
1461 * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for
1462 * tree blocks and <0 on error.
1463 */
1464 int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
1465 struct btrfs_path *path, struct btrfs_key *found_key,
1466 u64 *flags_ret)
1467 {
1468 int ret;
1469 u64 flags;
1470 u64 size = 0;
1471 u32 item_size;
1472 struct extent_buffer *eb;
1473 struct btrfs_extent_item *ei;
1474 struct btrfs_key key;
1475
1476 if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
1477 key.type = BTRFS_METADATA_ITEM_KEY;
1478 else
1479 key.type = BTRFS_EXTENT_ITEM_KEY;
1480 key.objectid = logical;
1481 key.offset = (u64)-1;
1482
1483 ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
1484 if (ret < 0)
1485 return ret;
1486
1487 ret = btrfs_previous_extent_item(fs_info->extent_root, path, 0);
1488 if (ret) {
1489 if (ret > 0)
1490 ret = -ENOENT;
1491 return ret;
1492 }
1493 btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]);
1494 if (found_key->type == BTRFS_METADATA_ITEM_KEY)
1495 size = fs_info->extent_root->nodesize;
1496 else if (found_key->type == BTRFS_EXTENT_ITEM_KEY)
1497 size = found_key->offset;
1498
1499 if (found_key->objectid > logical ||
1500 found_key->objectid + size <= logical) {
1501 pr_debug("logical %llu is not within any extent\n", logical);
1502 return -ENOENT;
1503 }
1504
1505 eb = path->nodes[0];
1506 item_size = btrfs_item_size_nr(eb, path->slots[0]);
1507 BUG_ON(item_size < sizeof(*ei));
1508
1509 ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
1510 flags = btrfs_extent_flags(eb, ei);
1511
1512 pr_debug("logical %llu is at position %llu within the extent (%llu "
1513 "EXTENT_ITEM %llu) flags %#llx size %u\n",
1514 logical, logical - found_key->objectid, found_key->objectid,
1515 found_key->offset, flags, item_size);
1516
1517 WARN_ON(!flags_ret);
1518 if (flags_ret) {
1519 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1520 *flags_ret = BTRFS_EXTENT_FLAG_TREE_BLOCK;
1521 else if (flags & BTRFS_EXTENT_FLAG_DATA)
1522 *flags_ret = BTRFS_EXTENT_FLAG_DATA;
1523 else
1524 BUG_ON(1);
1525 return 0;
1526 }
1527
1528 return -EIO;
1529 }
1530
1531 /*
1532 * helper function to iterate extent inline refs. ptr must point to a 0 value
1533 * for the first call and may be modified. it is used to track state.
1534 * if more refs exist, 0 is returned and the next call to
1535 * __get_extent_inline_ref must pass the modified ptr parameter to get the
1536 * next ref. after the last ref was processed, 1 is returned.
1537 * returns <0 on error
1538 */
1539 static int __get_extent_inline_ref(unsigned long *ptr, struct extent_buffer *eb,
1540 struct btrfs_key *key,
1541 struct btrfs_extent_item *ei, u32 item_size,
1542 struct btrfs_extent_inline_ref **out_eiref,
1543 int *out_type)
1544 {
1545 unsigned long end;
1546 u64 flags;
1547 struct btrfs_tree_block_info *info;
1548
1549 if (!*ptr) {
1550 /* first call */
1551 flags = btrfs_extent_flags(eb, ei);
1552 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1553 if (key->type == BTRFS_METADATA_ITEM_KEY) {
1554 /* a skinny metadata extent */
1555 *out_eiref =
1556 (struct btrfs_extent_inline_ref *)(ei + 1);
1557 } else {
1558 WARN_ON(key->type != BTRFS_EXTENT_ITEM_KEY);
1559 info = (struct btrfs_tree_block_info *)(ei + 1);
1560 *out_eiref =
1561 (struct btrfs_extent_inline_ref *)(info + 1);
1562 }
1563 } else {
1564 *out_eiref = (struct btrfs_extent_inline_ref *)(ei + 1);
1565 }
1566 *ptr = (unsigned long)*out_eiref;
1567 if ((unsigned long)(*ptr) >= (unsigned long)ei + item_size)
1568 return -ENOENT;
1569 }
1570
1571 end = (unsigned long)ei + item_size;
1572 *out_eiref = (struct btrfs_extent_inline_ref *)(*ptr);
1573 *out_type = btrfs_extent_inline_ref_type(eb, *out_eiref);
1574
1575 *ptr += btrfs_extent_inline_ref_size(*out_type);
1576 WARN_ON(*ptr > end);
1577 if (*ptr == end)
1578 return 1; /* last */
1579
1580 return 0;
1581 }
1582
1583 /*
1584 * reads the tree block backref for an extent. tree level and root are returned
1585 * through out_level and out_root. ptr must point to a 0 value for the first
1586 * call and may be modified (see __get_extent_inline_ref comment).
1587 * returns 0 if data was provided, 1 if there was no more data to provide or
1588 * <0 on error.
1589 */
1590 int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
1591 struct btrfs_key *key, struct btrfs_extent_item *ei,
1592 u32 item_size, u64 *out_root, u8 *out_level)
1593 {
1594 int ret;
1595 int type;
1596 struct btrfs_extent_inline_ref *eiref;
1597
1598 if (*ptr == (unsigned long)-1)
1599 return 1;
1600
1601 while (1) {
1602 ret = __get_extent_inline_ref(ptr, eb, key, ei, item_size,
1603 &eiref, &type);
1604 if (ret < 0)
1605 return ret;
1606
1607 if (type == BTRFS_TREE_BLOCK_REF_KEY ||
1608 type == BTRFS_SHARED_BLOCK_REF_KEY)
1609 break;
1610
1611 if (ret == 1)
1612 return 1;
1613 }
1614
1615 /* we can treat both ref types equally here */
1616 *out_root = btrfs_extent_inline_ref_offset(eb, eiref);
1617
1618 if (key->type == BTRFS_EXTENT_ITEM_KEY) {
1619 struct btrfs_tree_block_info *info;
1620
1621 info = (struct btrfs_tree_block_info *)(ei + 1);
1622 *out_level = btrfs_tree_block_level(eb, info);
1623 } else {
1624 ASSERT(key->type == BTRFS_METADATA_ITEM_KEY);
1625 *out_level = (u8)key->offset;
1626 }
1627
1628 if (ret == 1)
1629 *ptr = (unsigned long)-1;
1630
1631 return 0;
1632 }
1633
1634 static int iterate_leaf_refs(struct extent_inode_elem *inode_list,
1635 u64 root, u64 extent_item_objectid,
1636 iterate_extent_inodes_t *iterate, void *ctx)
1637 {
1638 struct extent_inode_elem *eie;
1639 int ret = 0;
1640
1641 for (eie = inode_list; eie; eie = eie->next) {
1642 pr_debug("ref for %llu resolved, key (%llu EXTEND_DATA %llu), "
1643 "root %llu\n", extent_item_objectid,
1644 eie->inum, eie->offset, root);
1645 ret = iterate(eie->inum, eie->offset, root, ctx);
1646 if (ret) {
1647 pr_debug("stopping iteration for %llu due to ret=%d\n",
1648 extent_item_objectid, ret);
1649 break;
1650 }
1651 }
1652
1653 return ret;
1654 }
1655
1656 /*
1657 * calls iterate() for every inode that references the extent identified by
1658 * the given parameters.
1659 * when the iterator function returns a non-zero value, iteration stops.
1660 */
1661 int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
1662 u64 extent_item_objectid, u64 extent_item_pos,
1663 int search_commit_root,
1664 iterate_extent_inodes_t *iterate, void *ctx)
1665 {
1666 int ret;
1667 struct btrfs_trans_handle *trans = NULL;
1668 struct ulist *refs = NULL;
1669 struct ulist *roots = NULL;
1670 struct ulist_node *ref_node = NULL;
1671 struct ulist_node *root_node = NULL;
1672 struct seq_list tree_mod_seq_elem = SEQ_LIST_INIT(tree_mod_seq_elem);
1673 struct ulist_iterator ref_uiter;
1674 struct ulist_iterator root_uiter;
1675
1676 pr_debug("resolving all inodes for extent %llu\n",
1677 extent_item_objectid);
1678
1679 if (!search_commit_root) {
1680 trans = btrfs_join_transaction(fs_info->extent_root);
1681 if (IS_ERR(trans))
1682 return PTR_ERR(trans);
1683 btrfs_get_tree_mod_seq(fs_info, &tree_mod_seq_elem);
1684 } else {
1685 down_read(&fs_info->commit_root_sem);
1686 }
1687
1688 ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
1689 tree_mod_seq_elem.seq, &refs,
1690 &extent_item_pos);
1691 if (ret)
1692 goto out;
1693
1694 ULIST_ITER_INIT(&ref_uiter);
1695 while (!ret && (ref_node = ulist_next(refs, &ref_uiter))) {
1696 ret = __btrfs_find_all_roots(trans, fs_info, ref_node->val,
1697 tree_mod_seq_elem.seq, &roots);
1698 if (ret)
1699 break;
1700 ULIST_ITER_INIT(&root_uiter);
1701 while (!ret && (root_node = ulist_next(roots, &root_uiter))) {
1702 pr_debug("root %llu references leaf %llu, data list "
1703 "%#llx\n", root_node->val, ref_node->val,
1704 ref_node->aux);
1705 ret = iterate_leaf_refs((struct extent_inode_elem *)
1706 (uintptr_t)ref_node->aux,
1707 root_node->val,
1708 extent_item_objectid,
1709 iterate, ctx);
1710 }
1711 ulist_free(roots);
1712 }
1713
1714 free_leaf_list(refs);
1715 out:
1716 if (!search_commit_root) {
1717 btrfs_put_tree_mod_seq(fs_info, &tree_mod_seq_elem);
1718 btrfs_end_transaction(trans, fs_info->extent_root);
1719 } else {
1720 up_read(&fs_info->commit_root_sem);
1721 }
1722
1723 return ret;
1724 }
1725
1726 int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
1727 struct btrfs_path *path,
1728 iterate_extent_inodes_t *iterate, void *ctx)
1729 {
1730 int ret;
1731 u64 extent_item_pos;
1732 u64 flags = 0;
1733 struct btrfs_key found_key;
1734 int search_commit_root = path->search_commit_root;
1735
1736 ret = extent_from_logical(fs_info, logical, path, &found_key, &flags);
1737 btrfs_release_path(path);
1738 if (ret < 0)
1739 return ret;
1740 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1741 return -EINVAL;
1742
1743 extent_item_pos = logical - found_key.objectid;
1744 ret = iterate_extent_inodes(fs_info, found_key.objectid,
1745 extent_item_pos, search_commit_root,
1746 iterate, ctx);
1747
1748 return ret;
1749 }
1750
1751 typedef int (iterate_irefs_t)(u64 parent, u32 name_len, unsigned long name_off,
1752 struct extent_buffer *eb, void *ctx);
1753
1754 static int iterate_inode_refs(u64 inum, struct btrfs_root *fs_root,
1755 struct btrfs_path *path,
1756 iterate_irefs_t *iterate, void *ctx)
1757 {
1758 int ret = 0;
1759 int slot;
1760 u32 cur;
1761 u32 len;
1762 u32 name_len;
1763 u64 parent = 0;
1764 int found = 0;
1765 struct extent_buffer *eb;
1766 struct btrfs_item *item;
1767 struct btrfs_inode_ref *iref;
1768 struct btrfs_key found_key;
1769
1770 while (!ret) {
1771 ret = btrfs_find_item(fs_root, path, inum,
1772 parent ? parent + 1 : 0, BTRFS_INODE_REF_KEY,
1773 &found_key);
1774
1775 if (ret < 0)
1776 break;
1777 if (ret) {
1778 ret = found ? 0 : -ENOENT;
1779 break;
1780 }
1781 ++found;
1782
1783 parent = found_key.offset;
1784 slot = path->slots[0];
1785 eb = btrfs_clone_extent_buffer(path->nodes[0]);
1786 if (!eb) {
1787 ret = -ENOMEM;
1788 break;
1789 }
1790 extent_buffer_get(eb);
1791 btrfs_tree_read_lock(eb);
1792 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1793 btrfs_release_path(path);
1794
1795 item = btrfs_item_nr(slot);
1796 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
1797
1798 for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) {
1799 name_len = btrfs_inode_ref_name_len(eb, iref);
1800 /* path must be released before calling iterate()! */
1801 pr_debug("following ref at offset %u for inode %llu in "
1802 "tree %llu\n", cur, found_key.objectid,
1803 fs_root->objectid);
1804 ret = iterate(parent, name_len,
1805 (unsigned long)(iref + 1), eb, ctx);
1806 if (ret)
1807 break;
1808 len = sizeof(*iref) + name_len;
1809 iref = (struct btrfs_inode_ref *)((char *)iref + len);
1810 }
1811 btrfs_tree_read_unlock_blocking(eb);
1812 free_extent_buffer(eb);
1813 }
1814
1815 btrfs_release_path(path);
1816
1817 return ret;
1818 }
1819
1820 static int iterate_inode_extrefs(u64 inum, struct btrfs_root *fs_root,
1821 struct btrfs_path *path,
1822 iterate_irefs_t *iterate, void *ctx)
1823 {
1824 int ret;
1825 int slot;
1826 u64 offset = 0;
1827 u64 parent;
1828 int found = 0;
1829 struct extent_buffer *eb;
1830 struct btrfs_inode_extref *extref;
1831 struct extent_buffer *leaf;
1832 u32 item_size;
1833 u32 cur_offset;
1834 unsigned long ptr;
1835
1836 while (1) {
1837 ret = btrfs_find_one_extref(fs_root, inum, offset, path, &extref,
1838 &offset);
1839 if (ret < 0)
1840 break;
1841 if (ret) {
1842 ret = found ? 0 : -ENOENT;
1843 break;
1844 }
1845 ++found;
1846
1847 slot = path->slots[0];
1848 eb = btrfs_clone_extent_buffer(path->nodes[0]);
1849 if (!eb) {
1850 ret = -ENOMEM;
1851 break;
1852 }
1853 extent_buffer_get(eb);
1854
1855 btrfs_tree_read_lock(eb);
1856 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1857 btrfs_release_path(path);
1858
1859 leaf = path->nodes[0];
1860 item_size = btrfs_item_size_nr(leaf, slot);
1861 ptr = btrfs_item_ptr_offset(leaf, slot);
1862 cur_offset = 0;
1863
1864 while (cur_offset < item_size) {
1865 u32 name_len;
1866
1867 extref = (struct btrfs_inode_extref *)(ptr + cur_offset);
1868 parent = btrfs_inode_extref_parent(eb, extref);
1869 name_len = btrfs_inode_extref_name_len(eb, extref);
1870 ret = iterate(parent, name_len,
1871 (unsigned long)&extref->name, eb, ctx);
1872 if (ret)
1873 break;
1874
1875 cur_offset += btrfs_inode_extref_name_len(leaf, extref);
1876 cur_offset += sizeof(*extref);
1877 }
1878 btrfs_tree_read_unlock_blocking(eb);
1879 free_extent_buffer(eb);
1880
1881 offset++;
1882 }
1883
1884 btrfs_release_path(path);
1885
1886 return ret;
1887 }
1888
1889 static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
1890 struct btrfs_path *path, iterate_irefs_t *iterate,
1891 void *ctx)
1892 {
1893 int ret;
1894 int found_refs = 0;
1895
1896 ret = iterate_inode_refs(inum, fs_root, path, iterate, ctx);
1897 if (!ret)
1898 ++found_refs;
1899 else if (ret != -ENOENT)
1900 return ret;
1901
1902 ret = iterate_inode_extrefs(inum, fs_root, path, iterate, ctx);
1903 if (ret == -ENOENT && found_refs)
1904 return 0;
1905
1906 return ret;
1907 }
1908
1909 /*
1910 * returns 0 if the path could be dumped (probably truncated)
1911 * returns <0 in case of an error
1912 */
1913 static int inode_to_path(u64 inum, u32 name_len, unsigned long name_off,
1914 struct extent_buffer *eb, void *ctx)
1915 {
1916 struct inode_fs_paths *ipath = ctx;
1917 char *fspath;
1918 char *fspath_min;
1919 int i = ipath->fspath->elem_cnt;
1920 const int s_ptr = sizeof(char *);
1921 u32 bytes_left;
1922
1923 bytes_left = ipath->fspath->bytes_left > s_ptr ?
1924 ipath->fspath->bytes_left - s_ptr : 0;
1925
1926 fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
1927 fspath = btrfs_ref_to_path(ipath->fs_root, ipath->btrfs_path, name_len,
1928 name_off, eb, inum, fspath_min, bytes_left);
1929 if (IS_ERR(fspath))
1930 return PTR_ERR(fspath);
1931
1932 if (fspath > fspath_min) {
1933 ipath->fspath->val[i] = (u64)(unsigned long)fspath;
1934 ++ipath->fspath->elem_cnt;
1935 ipath->fspath->bytes_left = fspath - fspath_min;
1936 } else {
1937 ++ipath->fspath->elem_missed;
1938 ipath->fspath->bytes_missing += fspath_min - fspath;
1939 ipath->fspath->bytes_left = 0;
1940 }
1941
1942 return 0;
1943 }
1944
1945 /*
1946 * this dumps all file system paths to the inode into the ipath struct, provided
1947 * is has been created large enough. each path is zero-terminated and accessed
1948 * from ipath->fspath->val[i].
1949 * when it returns, there are ipath->fspath->elem_cnt number of paths available
1950 * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
1951 * number of missed paths in recored in ipath->fspath->elem_missed, otherwise,
1952 * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
1953 * have been needed to return all paths.
1954 */
1955 int paths_from_inode(u64 inum, struct inode_fs_paths *ipath)
1956 {
1957 return iterate_irefs(inum, ipath->fs_root, ipath->btrfs_path,
1958 inode_to_path, ipath);
1959 }
1960
1961 struct btrfs_data_container *init_data_container(u32 total_bytes)
1962 {
1963 struct btrfs_data_container *data;
1964 size_t alloc_bytes;
1965
1966 alloc_bytes = max_t(size_t, total_bytes, sizeof(*data));
1967 data = vmalloc(alloc_bytes);
1968 if (!data)
1969 return ERR_PTR(-ENOMEM);
1970
1971 if (total_bytes >= sizeof(*data)) {
1972 data->bytes_left = total_bytes - sizeof(*data);
1973 data->bytes_missing = 0;
1974 } else {
1975 data->bytes_missing = sizeof(*data) - total_bytes;
1976 data->bytes_left = 0;
1977 }
1978
1979 data->elem_cnt = 0;
1980 data->elem_missed = 0;
1981
1982 return data;
1983 }
1984
1985 /*
1986 * allocates space to return multiple file system paths for an inode.
1987 * total_bytes to allocate are passed, note that space usable for actual path
1988 * information will be total_bytes - sizeof(struct inode_fs_paths).
1989 * the returned pointer must be freed with free_ipath() in the end.
1990 */
1991 struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
1992 struct btrfs_path *path)
1993 {
1994 struct inode_fs_paths *ifp;
1995 struct btrfs_data_container *fspath;
1996
1997 fspath = init_data_container(total_bytes);
1998 if (IS_ERR(fspath))
1999 return (void *)fspath;
2000
2001 ifp = kmalloc(sizeof(*ifp), GFP_NOFS);
2002 if (!ifp) {
2003 kfree(fspath);
2004 return ERR_PTR(-ENOMEM);
2005 }
2006
2007 ifp->btrfs_path = path;
2008 ifp->fspath = fspath;
2009 ifp->fs_root = fs_root;
2010
2011 return ifp;
2012 }
2013
2014 void free_ipath(struct inode_fs_paths *ipath)
2015 {
2016 if (!ipath)
2017 return;
2018 vfree(ipath->fspath);
2019 kfree(ipath);
2020 }
(deprecated) Btrfs devel tree