Loading...
Note: File does not exist in v3.1.
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
31struct extent_inode_elem {
32 u64 inum;
33 u64 offset;
34 struct extent_inode_elem *next;
35};
36
37static 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
72static 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
82static 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 */
124struct __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
135static struct kmem_cache *btrfs_prelim_ref_cache;
136
137int __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
149void btrfs_prelim_ref_exit(void)
150{
151 kmem_cache_destroy(btrfs_prelim_ref_cache);
152}
153
154/*
155 * the rules for all callers of this function are:
156 * - obtaining the parent is the goal
157 * - if you add a key, you must know that it is a correct key
158 * - if you cannot add the parent or a correct key, then we will look into the
159 * block later to set a correct key
160 *
161 * delayed refs
162 * ============
163 * backref type | shared | indirect | shared | indirect
164 * information | tree | tree | data | data
165 * --------------------+--------+----------+--------+----------
166 * parent logical | y | - | - | -
167 * key to resolve | - | y | y | y
168 * tree block logical | - | - | - | -
169 * root for resolving | y | y | y | y
170 *
171 * - column 1: we've the parent -> done
172 * - column 2, 3, 4: we use the key to find the parent
173 *
174 * on disk refs (inline or keyed)
175 * ==============================
176 * backref type | shared | indirect | shared | indirect
177 * information | tree | tree | data | data
178 * --------------------+--------+----------+--------+----------
179 * parent logical | y | - | y | -
180 * key to resolve | - | - | - | y
181 * tree block logical | y | y | y | y
182 * root for resolving | - | y | y | y
183 *
184 * - column 1, 3: we've the parent -> done
185 * - column 2: we take the first key from the block to find the parent
186 * (see __add_missing_keys)
187 * - column 4: we use the key to find the parent
188 *
189 * additional information that's available but not required to find the parent
190 * block might help in merging entries to gain some speed.
191 */
192
193static int __add_prelim_ref(struct list_head *head, u64 root_id,
194 struct btrfs_key *key, int level,
195 u64 parent, u64 wanted_disk_byte, int count,
196 gfp_t gfp_mask)
197{
198 struct __prelim_ref *ref;
199
200 if (root_id == BTRFS_DATA_RELOC_TREE_OBJECTID)
201 return 0;
202
203 ref = kmem_cache_alloc(btrfs_prelim_ref_cache, gfp_mask);
204 if (!ref)
205 return -ENOMEM;
206
207 ref->root_id = root_id;
208 if (key) {
209 ref->key_for_search = *key;
210 /*
211 * We can often find data backrefs with an offset that is too
212 * large (>= LLONG_MAX, maximum allowed file offset) due to
213 * underflows when subtracting a file's offset with the data
214 * offset of its corresponding extent data item. This can
215 * happen for example in the clone ioctl.
216 * So if we detect such case we set the search key's offset to
217 * zero to make sure we will find the matching file extent item
218 * at add_all_parents(), otherwise we will miss it because the
219 * offset taken form the backref is much larger then the offset
220 * of the file extent item. This can make us scan a very large
221 * number of file extent items, but at least it will not make
222 * us miss any.
223 * This is an ugly workaround for a behaviour that should have
224 * never existed, but it does and a fix for the clone ioctl
225 * would touch a lot of places, cause backwards incompatibility
226 * and would not fix the problem for extents cloned with older
227 * kernels.
228 */
229 if (ref->key_for_search.type == BTRFS_EXTENT_DATA_KEY &&
230 ref->key_for_search.offset >= LLONG_MAX)
231 ref->key_for_search.offset = 0;
232 } else {
233 memset(&ref->key_for_search, 0, sizeof(ref->key_for_search));
234 }
235
236 ref->inode_list = NULL;
237 ref->level = level;
238 ref->count = count;
239 ref->parent = parent;
240 ref->wanted_disk_byte = wanted_disk_byte;
241 list_add_tail(&ref->list, head);
242
243 return 0;
244}
245
246static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
247 struct ulist *parents, struct __prelim_ref *ref,
248 int level, u64 time_seq, const u64 *extent_item_pos,
249 u64 total_refs)
250{
251 int ret = 0;
252 int slot;
253 struct extent_buffer *eb;
254 struct btrfs_key key;
255 struct btrfs_key *key_for_search = &ref->key_for_search;
256 struct btrfs_file_extent_item *fi;
257 struct extent_inode_elem *eie = NULL, *old = NULL;
258 u64 disk_byte;
259 u64 wanted_disk_byte = ref->wanted_disk_byte;
260 u64 count = 0;
261
262 if (level != 0) {
263 eb = path->nodes[level];
264 ret = ulist_add(parents, eb->start, 0, GFP_NOFS);
265 if (ret < 0)
266 return ret;
267 return 0;
268 }
269
270 /*
271 * We normally enter this function with the path already pointing to
272 * the first item to check. But sometimes, we may enter it with
273 * slot==nritems. In that case, go to the next leaf before we continue.
274 */
275 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
276 if (time_seq == (u64)-1)
277 ret = btrfs_next_leaf(root, path);
278 else
279 ret = btrfs_next_old_leaf(root, path, time_seq);
280 }
281
282 while (!ret && count < total_refs) {
283 eb = path->nodes[0];
284 slot = path->slots[0];
285
286 btrfs_item_key_to_cpu(eb, &key, slot);
287
288 if (key.objectid != key_for_search->objectid ||
289 key.type != BTRFS_EXTENT_DATA_KEY)
290 break;
291
292 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
293 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
294
295 if (disk_byte == wanted_disk_byte) {
296 eie = NULL;
297 old = NULL;
298 count++;
299 if (extent_item_pos) {
300 ret = check_extent_in_eb(&key, eb, fi,
301 *extent_item_pos,
302 &eie);
303 if (ret < 0)
304 break;
305 }
306 if (ret > 0)
307 goto next;
308 ret = ulist_add_merge_ptr(parents, eb->start,
309 eie, (void **)&old, GFP_NOFS);
310 if (ret < 0)
311 break;
312 if (!ret && extent_item_pos) {
313 while (old->next)
314 old = old->next;
315 old->next = eie;
316 }
317 eie = NULL;
318 }
319next:
320 if (time_seq == (u64)-1)
321 ret = btrfs_next_item(root, path);
322 else
323 ret = btrfs_next_old_item(root, path, time_seq);
324 }
325
326 if (ret > 0)
327 ret = 0;
328 else if (ret < 0)
329 free_inode_elem_list(eie);
330 return ret;
331}
332
333/*
334 * resolve an indirect backref in the form (root_id, key, level)
335 * to a logical address
336 */
337static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
338 struct btrfs_path *path, u64 time_seq,
339 struct __prelim_ref *ref,
340 struct ulist *parents,
341 const u64 *extent_item_pos, u64 total_refs)
342{
343 struct btrfs_root *root;
344 struct btrfs_key root_key;
345 struct extent_buffer *eb;
346 int ret = 0;
347 int root_level;
348 int level = ref->level;
349 int index;
350
351 root_key.objectid = ref->root_id;
352 root_key.type = BTRFS_ROOT_ITEM_KEY;
353 root_key.offset = (u64)-1;
354
355 index = srcu_read_lock(&fs_info->subvol_srcu);
356
357 root = btrfs_get_fs_root(fs_info, &root_key, false);
358 if (IS_ERR(root)) {
359 srcu_read_unlock(&fs_info->subvol_srcu, index);
360 ret = PTR_ERR(root);
361 goto out;
362 }
363
364 if (btrfs_test_is_dummy_root(root)) {
365 srcu_read_unlock(&fs_info->subvol_srcu, index);
366 ret = -ENOENT;
367 goto out;
368 }
369
370 if (path->search_commit_root)
371 root_level = btrfs_header_level(root->commit_root);
372 else if (time_seq == (u64)-1)
373 root_level = btrfs_header_level(root->node);
374 else
375 root_level = btrfs_old_root_level(root, time_seq);
376
377 if (root_level + 1 == level) {
378 srcu_read_unlock(&fs_info->subvol_srcu, index);
379 goto out;
380 }
381
382 path->lowest_level = level;
383 if (time_seq == (u64)-1)
384 ret = btrfs_search_slot(NULL, root, &ref->key_for_search, path,
385 0, 0);
386 else
387 ret = btrfs_search_old_slot(root, &ref->key_for_search, path,
388 time_seq);
389
390 /* root node has been locked, we can release @subvol_srcu safely here */
391 srcu_read_unlock(&fs_info->subvol_srcu, index);
392
393 pr_debug("search slot in root %llu (level %d, ref count %d) returned "
394 "%d for key (%llu %u %llu)\n",
395 ref->root_id, level, ref->count, ret,
396 ref->key_for_search.objectid, ref->key_for_search.type,
397 ref->key_for_search.offset);
398 if (ret < 0)
399 goto out;
400
401 eb = path->nodes[level];
402 while (!eb) {
403 if (WARN_ON(!level)) {
404 ret = 1;
405 goto out;
406 }
407 level--;
408 eb = path->nodes[level];
409 }
410
411 ret = add_all_parents(root, path, parents, ref, level, time_seq,
412 extent_item_pos, total_refs);
413out:
414 path->lowest_level = 0;
415 btrfs_release_path(path);
416 return ret;
417}
418
419/*
420 * resolve all indirect backrefs from the list
421 */
422static int __resolve_indirect_refs(struct btrfs_fs_info *fs_info,
423 struct btrfs_path *path, u64 time_seq,
424 struct list_head *head,
425 const u64 *extent_item_pos, u64 total_refs,
426 u64 root_objectid)
427{
428 int err;
429 int ret = 0;
430 struct __prelim_ref *ref;
431 struct __prelim_ref *ref_safe;
432 struct __prelim_ref *new_ref;
433 struct ulist *parents;
434 struct ulist_node *node;
435 struct ulist_iterator uiter;
436
437 parents = ulist_alloc(GFP_NOFS);
438 if (!parents)
439 return -ENOMEM;
440
441 /*
442 * _safe allows us to insert directly after the current item without
443 * iterating over the newly inserted items.
444 * we're also allowed to re-assign ref during iteration.
445 */
446 list_for_each_entry_safe(ref, ref_safe, head, list) {
447 if (ref->parent) /* already direct */
448 continue;
449 if (ref->count == 0)
450 continue;
451 if (root_objectid && ref->root_id != root_objectid) {
452 ret = BACKREF_FOUND_SHARED;
453 goto out;
454 }
455 err = __resolve_indirect_ref(fs_info, path, time_seq, ref,
456 parents, extent_item_pos,
457 total_refs);
458 /*
459 * we can only tolerate ENOENT,otherwise,we should catch error
460 * and return directly.
461 */
462 if (err == -ENOENT) {
463 continue;
464 } else if (err) {
465 ret = err;
466 goto out;
467 }
468
469 /* we put the first parent into the ref at hand */
470 ULIST_ITER_INIT(&uiter);
471 node = ulist_next(parents, &uiter);
472 ref->parent = node ? node->val : 0;
473 ref->inode_list = node ?
474 (struct extent_inode_elem *)(uintptr_t)node->aux : NULL;
475
476 /* additional parents require new refs being added here */
477 while ((node = ulist_next(parents, &uiter))) {
478 new_ref = kmem_cache_alloc(btrfs_prelim_ref_cache,
479 GFP_NOFS);
480 if (!new_ref) {
481 ret = -ENOMEM;
482 goto out;
483 }
484 memcpy(new_ref, ref, sizeof(*ref));
485 new_ref->parent = node->val;
486 new_ref->inode_list = (struct extent_inode_elem *)
487 (uintptr_t)node->aux;
488 list_add(&new_ref->list, &ref->list);
489 }
490 ulist_reinit(parents);
491 }
492out:
493 ulist_free(parents);
494 return ret;
495}
496
497static inline int ref_for_same_block(struct __prelim_ref *ref1,
498 struct __prelim_ref *ref2)
499{
500 if (ref1->level != ref2->level)
501 return 0;
502 if (ref1->root_id != ref2->root_id)
503 return 0;
504 if (ref1->key_for_search.type != ref2->key_for_search.type)
505 return 0;
506 if (ref1->key_for_search.objectid != ref2->key_for_search.objectid)
507 return 0;
508 if (ref1->key_for_search.offset != ref2->key_for_search.offset)
509 return 0;
510 if (ref1->parent != ref2->parent)
511 return 0;
512
513 return 1;
514}
515
516/*
517 * read tree blocks and add keys where required.
518 */
519static int __add_missing_keys(struct btrfs_fs_info *fs_info,
520 struct list_head *head)
521{
522 struct __prelim_ref *ref;
523 struct extent_buffer *eb;
524
525 list_for_each_entry(ref, head, list) {
526 if (ref->parent)
527 continue;
528 if (ref->key_for_search.type)
529 continue;
530 BUG_ON(!ref->wanted_disk_byte);
531 eb = read_tree_block(fs_info->tree_root, ref->wanted_disk_byte,
532 0);
533 if (IS_ERR(eb)) {
534 return PTR_ERR(eb);
535 } else if (!extent_buffer_uptodate(eb)) {
536 free_extent_buffer(eb);
537 return -EIO;
538 }
539 btrfs_tree_read_lock(eb);
540 if (btrfs_header_level(eb) == 0)
541 btrfs_item_key_to_cpu(eb, &ref->key_for_search, 0);
542 else
543 btrfs_node_key_to_cpu(eb, &ref->key_for_search, 0);
544 btrfs_tree_read_unlock(eb);
545 free_extent_buffer(eb);
546 }
547 return 0;
548}
549
550/*
551 * merge backrefs and adjust counts accordingly
552 *
553 * mode = 1: merge identical keys, if key is set
554 * FIXME: if we add more keys in __add_prelim_ref, we can merge more here.
555 * additionally, we could even add a key range for the blocks we
556 * looked into to merge even more (-> replace unresolved refs by those
557 * having a parent).
558 * mode = 2: merge identical parents
559 */
560static void __merge_refs(struct list_head *head, int mode)
561{
562 struct __prelim_ref *pos1;
563
564 list_for_each_entry(pos1, head, list) {
565 struct __prelim_ref *pos2 = pos1, *tmp;
566
567 list_for_each_entry_safe_continue(pos2, tmp, head, list) {
568 struct __prelim_ref *ref1 = pos1, *ref2 = pos2;
569 struct extent_inode_elem *eie;
570
571 if (!ref_for_same_block(ref1, ref2))
572 continue;
573 if (mode == 1) {
574 if (!ref1->parent && ref2->parent)
575 swap(ref1, ref2);
576 } else {
577 if (ref1->parent != ref2->parent)
578 continue;
579 }
580
581 eie = ref1->inode_list;
582 while (eie && eie->next)
583 eie = eie->next;
584 if (eie)
585 eie->next = ref2->inode_list;
586 else
587 ref1->inode_list = ref2->inode_list;
588 ref1->count += ref2->count;
589
590 list_del(&ref2->list);
591 kmem_cache_free(btrfs_prelim_ref_cache, ref2);
592 }
593
594 }
595}
596
597/*
598 * add all currently queued delayed refs from this head whose seq nr is
599 * smaller or equal that seq to the list
600 */
601static int __add_delayed_refs(struct btrfs_delayed_ref_head *head, u64 seq,
602 struct list_head *prefs, u64 *total_refs,
603 u64 inum)
604{
605 struct btrfs_delayed_ref_node *node;
606 struct btrfs_delayed_extent_op *extent_op = head->extent_op;
607 struct btrfs_key key;
608 struct btrfs_key op_key = {0};
609 int sgn;
610 int ret = 0;
611
612 if (extent_op && extent_op->update_key)
613 btrfs_disk_key_to_cpu(&op_key, &extent_op->key);
614
615 spin_lock(&head->lock);
616 list_for_each_entry(node, &head->ref_list, list) {
617 if (node->seq > seq)
618 continue;
619
620 switch (node->action) {
621 case BTRFS_ADD_DELAYED_EXTENT:
622 case BTRFS_UPDATE_DELAYED_HEAD:
623 WARN_ON(1);
624 continue;
625 case BTRFS_ADD_DELAYED_REF:
626 sgn = 1;
627 break;
628 case BTRFS_DROP_DELAYED_REF:
629 sgn = -1;
630 break;
631 default:
632 BUG_ON(1);
633 }
634 *total_refs += (node->ref_mod * sgn);
635 switch (node->type) {
636 case BTRFS_TREE_BLOCK_REF_KEY: {
637 struct btrfs_delayed_tree_ref *ref;
638
639 ref = btrfs_delayed_node_to_tree_ref(node);
640 ret = __add_prelim_ref(prefs, ref->root, &op_key,
641 ref->level + 1, 0, node->bytenr,
642 node->ref_mod * sgn, GFP_ATOMIC);
643 break;
644 }
645 case BTRFS_SHARED_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, 0, NULL,
650 ref->level + 1, ref->parent,
651 node->bytenr,
652 node->ref_mod * sgn, GFP_ATOMIC);
653 break;
654 }
655 case BTRFS_EXTENT_DATA_REF_KEY: {
656 struct btrfs_delayed_data_ref *ref;
657 ref = btrfs_delayed_node_to_data_ref(node);
658
659 key.objectid = ref->objectid;
660 key.type = BTRFS_EXTENT_DATA_KEY;
661 key.offset = ref->offset;
662
663 /*
664 * Found a inum that doesn't match our known inum, we
665 * know it's shared.
666 */
667 if (inum && ref->objectid != inum) {
668 ret = BACKREF_FOUND_SHARED;
669 break;
670 }
671
672 ret = __add_prelim_ref(prefs, ref->root, &key, 0, 0,
673 node->bytenr,
674 node->ref_mod * sgn, GFP_ATOMIC);
675 break;
676 }
677 case BTRFS_SHARED_DATA_REF_KEY: {
678 struct btrfs_delayed_data_ref *ref;
679
680 ref = btrfs_delayed_node_to_data_ref(node);
681 ret = __add_prelim_ref(prefs, 0, NULL, 0,
682 ref->parent, node->bytenr,
683 node->ref_mod * sgn, GFP_ATOMIC);
684 break;
685 }
686 default:
687 WARN_ON(1);
688 }
689 if (ret)
690 break;
691 }
692 spin_unlock(&head->lock);
693 return ret;
694}
695
696/*
697 * add all inline backrefs for bytenr to the list
698 */
699static int __add_inline_refs(struct btrfs_fs_info *fs_info,
700 struct btrfs_path *path, u64 bytenr,
701 int *info_level, struct list_head *prefs,
702 u64 *total_refs, u64 inum)
703{
704 int ret = 0;
705 int slot;
706 struct extent_buffer *leaf;
707 struct btrfs_key key;
708 struct btrfs_key found_key;
709 unsigned long ptr;
710 unsigned long end;
711 struct btrfs_extent_item *ei;
712 u64 flags;
713 u64 item_size;
714
715 /*
716 * enumerate all inline refs
717 */
718 leaf = path->nodes[0];
719 slot = path->slots[0];
720
721 item_size = btrfs_item_size_nr(leaf, slot);
722 BUG_ON(item_size < sizeof(*ei));
723
724 ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
725 flags = btrfs_extent_flags(leaf, ei);
726 *total_refs += btrfs_extent_refs(leaf, ei);
727 btrfs_item_key_to_cpu(leaf, &found_key, slot);
728
729 ptr = (unsigned long)(ei + 1);
730 end = (unsigned long)ei + item_size;
731
732 if (found_key.type == BTRFS_EXTENT_ITEM_KEY &&
733 flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
734 struct btrfs_tree_block_info *info;
735
736 info = (struct btrfs_tree_block_info *)ptr;
737 *info_level = btrfs_tree_block_level(leaf, info);
738 ptr += sizeof(struct btrfs_tree_block_info);
739 BUG_ON(ptr > end);
740 } else if (found_key.type == BTRFS_METADATA_ITEM_KEY) {
741 *info_level = found_key.offset;
742 } else {
743 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
744 }
745
746 while (ptr < end) {
747 struct btrfs_extent_inline_ref *iref;
748 u64 offset;
749 int type;
750
751 iref = (struct btrfs_extent_inline_ref *)ptr;
752 type = btrfs_extent_inline_ref_type(leaf, iref);
753 offset = btrfs_extent_inline_ref_offset(leaf, iref);
754
755 switch (type) {
756 case BTRFS_SHARED_BLOCK_REF_KEY:
757 ret = __add_prelim_ref(prefs, 0, NULL,
758 *info_level + 1, offset,
759 bytenr, 1, GFP_NOFS);
760 break;
761 case BTRFS_SHARED_DATA_REF_KEY: {
762 struct btrfs_shared_data_ref *sdref;
763 int count;
764
765 sdref = (struct btrfs_shared_data_ref *)(iref + 1);
766 count = btrfs_shared_data_ref_count(leaf, sdref);
767 ret = __add_prelim_ref(prefs, 0, NULL, 0, offset,
768 bytenr, count, GFP_NOFS);
769 break;
770 }
771 case BTRFS_TREE_BLOCK_REF_KEY:
772 ret = __add_prelim_ref(prefs, offset, NULL,
773 *info_level + 1, 0,
774 bytenr, 1, GFP_NOFS);
775 break;
776 case BTRFS_EXTENT_DATA_REF_KEY: {
777 struct btrfs_extent_data_ref *dref;
778 int count;
779 u64 root;
780
781 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
782 count = btrfs_extent_data_ref_count(leaf, dref);
783 key.objectid = btrfs_extent_data_ref_objectid(leaf,
784 dref);
785 key.type = BTRFS_EXTENT_DATA_KEY;
786 key.offset = btrfs_extent_data_ref_offset(leaf, dref);
787
788 if (inum && key.objectid != inum) {
789 ret = BACKREF_FOUND_SHARED;
790 break;
791 }
792
793 root = btrfs_extent_data_ref_root(leaf, dref);
794 ret = __add_prelim_ref(prefs, root, &key, 0, 0,
795 bytenr, count, GFP_NOFS);
796 break;
797 }
798 default:
799 WARN_ON(1);
800 }
801 if (ret)
802 return ret;
803 ptr += btrfs_extent_inline_ref_size(type);
804 }
805
806 return 0;
807}
808
809/*
810 * add all non-inline backrefs for bytenr to the list
811 */
812static int __add_keyed_refs(struct btrfs_fs_info *fs_info,
813 struct btrfs_path *path, u64 bytenr,
814 int info_level, struct list_head *prefs, u64 inum)
815{
816 struct btrfs_root *extent_root = fs_info->extent_root;
817 int ret;
818 int slot;
819 struct extent_buffer *leaf;
820 struct btrfs_key key;
821
822 while (1) {
823 ret = btrfs_next_item(extent_root, path);
824 if (ret < 0)
825 break;
826 if (ret) {
827 ret = 0;
828 break;
829 }
830
831 slot = path->slots[0];
832 leaf = path->nodes[0];
833 btrfs_item_key_to_cpu(leaf, &key, slot);
834
835 if (key.objectid != bytenr)
836 break;
837 if (key.type < BTRFS_TREE_BLOCK_REF_KEY)
838 continue;
839 if (key.type > BTRFS_SHARED_DATA_REF_KEY)
840 break;
841
842 switch (key.type) {
843 case BTRFS_SHARED_BLOCK_REF_KEY:
844 ret = __add_prelim_ref(prefs, 0, NULL,
845 info_level + 1, key.offset,
846 bytenr, 1, GFP_NOFS);
847 break;
848 case BTRFS_SHARED_DATA_REF_KEY: {
849 struct btrfs_shared_data_ref *sdref;
850 int count;
851
852 sdref = btrfs_item_ptr(leaf, slot,
853 struct btrfs_shared_data_ref);
854 count = btrfs_shared_data_ref_count(leaf, sdref);
855 ret = __add_prelim_ref(prefs, 0, NULL, 0, key.offset,
856 bytenr, count, GFP_NOFS);
857 break;
858 }
859 case BTRFS_TREE_BLOCK_REF_KEY:
860 ret = __add_prelim_ref(prefs, key.offset, NULL,
861 info_level + 1, 0,
862 bytenr, 1, GFP_NOFS);
863 break;
864 case BTRFS_EXTENT_DATA_REF_KEY: {
865 struct btrfs_extent_data_ref *dref;
866 int count;
867 u64 root;
868
869 dref = btrfs_item_ptr(leaf, slot,
870 struct btrfs_extent_data_ref);
871 count = btrfs_extent_data_ref_count(leaf, dref);
872 key.objectid = btrfs_extent_data_ref_objectid(leaf,
873 dref);
874 key.type = BTRFS_EXTENT_DATA_KEY;
875 key.offset = btrfs_extent_data_ref_offset(leaf, dref);
876
877 if (inum && key.objectid != inum) {
878 ret = BACKREF_FOUND_SHARED;
879 break;
880 }
881
882 root = btrfs_extent_data_ref_root(leaf, dref);
883 ret = __add_prelim_ref(prefs, root, &key, 0, 0,
884 bytenr, count, GFP_NOFS);
885 break;
886 }
887 default:
888 WARN_ON(1);
889 }
890 if (ret)
891 return ret;
892
893 }
894
895 return ret;
896}
897
898/*
899 * this adds all existing backrefs (inline backrefs, backrefs and delayed
900 * refs) for the given bytenr to the refs list, merges duplicates and resolves
901 * indirect refs to their parent bytenr.
902 * When roots are found, they're added to the roots list
903 *
904 * NOTE: This can return values > 0
905 *
906 * If time_seq is set to (u64)-1, it will not search delayed_refs, and behave
907 * much like trans == NULL case, the difference only lies in it will not
908 * commit root.
909 * The special case is for qgroup to search roots in commit_transaction().
910 *
911 * FIXME some caching might speed things up
912 */
913static int find_parent_nodes(struct btrfs_trans_handle *trans,
914 struct btrfs_fs_info *fs_info, u64 bytenr,
915 u64 time_seq, struct ulist *refs,
916 struct ulist *roots, const u64 *extent_item_pos,
917 u64 root_objectid, u64 inum)
918{
919 struct btrfs_key key;
920 struct btrfs_path *path;
921 struct btrfs_delayed_ref_root *delayed_refs = NULL;
922 struct btrfs_delayed_ref_head *head;
923 int info_level = 0;
924 int ret;
925 struct list_head prefs_delayed;
926 struct list_head prefs;
927 struct __prelim_ref *ref;
928 struct extent_inode_elem *eie = NULL;
929 u64 total_refs = 0;
930
931 INIT_LIST_HEAD(&prefs);
932 INIT_LIST_HEAD(&prefs_delayed);
933
934 key.objectid = bytenr;
935 key.offset = (u64)-1;
936 if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
937 key.type = BTRFS_METADATA_ITEM_KEY;
938 else
939 key.type = BTRFS_EXTENT_ITEM_KEY;
940
941 path = btrfs_alloc_path();
942 if (!path)
943 return -ENOMEM;
944 if (!trans) {
945 path->search_commit_root = 1;
946 path->skip_locking = 1;
947 }
948
949 if (time_seq == (u64)-1)
950 path->skip_locking = 1;
951
952 /*
953 * grab both a lock on the path and a lock on the delayed ref head.
954 * We need both to get a consistent picture of how the refs look
955 * at a specified point in time
956 */
957again:
958 head = NULL;
959
960 ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
961 if (ret < 0)
962 goto out;
963 BUG_ON(ret == 0);
964
965#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
966 if (trans && likely(trans->type != __TRANS_DUMMY) &&
967 time_seq != (u64)-1) {
968#else
969 if (trans && time_seq != (u64)-1) {
970#endif
971 /*
972 * look if there are updates for this ref queued and lock the
973 * head
974 */
975 delayed_refs = &trans->transaction->delayed_refs;
976 spin_lock(&delayed_refs->lock);
977 head = btrfs_find_delayed_ref_head(trans, bytenr);
978 if (head) {
979 if (!mutex_trylock(&head->mutex)) {
980 atomic_inc(&head->node.refs);
981 spin_unlock(&delayed_refs->lock);
982
983 btrfs_release_path(path);
984
985 /*
986 * Mutex was contended, block until it's
987 * released and try again
988 */
989 mutex_lock(&head->mutex);
990 mutex_unlock(&head->mutex);
991 btrfs_put_delayed_ref(&head->node);
992 goto again;
993 }
994 spin_unlock(&delayed_refs->lock);
995 ret = __add_delayed_refs(head, time_seq,
996 &prefs_delayed, &total_refs,
997 inum);
998 mutex_unlock(&head->mutex);
999 if (ret)
1000 goto out;
1001 } else {
1002 spin_unlock(&delayed_refs->lock);
1003 }
1004 }
1005
1006 if (path->slots[0]) {
1007 struct extent_buffer *leaf;
1008 int slot;
1009
1010 path->slots[0]--;
1011 leaf = path->nodes[0];
1012 slot = path->slots[0];
1013 btrfs_item_key_to_cpu(leaf, &key, slot);
1014 if (key.objectid == bytenr &&
1015 (key.type == BTRFS_EXTENT_ITEM_KEY ||
1016 key.type == BTRFS_METADATA_ITEM_KEY)) {
1017 ret = __add_inline_refs(fs_info, path, bytenr,
1018 &info_level, &prefs,
1019 &total_refs, inum);
1020 if (ret)
1021 goto out;
1022 ret = __add_keyed_refs(fs_info, path, bytenr,
1023 info_level, &prefs, inum);
1024 if (ret)
1025 goto out;
1026 }
1027 }
1028 btrfs_release_path(path);
1029
1030 list_splice_init(&prefs_delayed, &prefs);
1031
1032 ret = __add_missing_keys(fs_info, &prefs);
1033 if (ret)
1034 goto out;
1035
1036 __merge_refs(&prefs, 1);
1037
1038 ret = __resolve_indirect_refs(fs_info, path, time_seq, &prefs,
1039 extent_item_pos, total_refs,
1040 root_objectid);
1041 if (ret)
1042 goto out;
1043
1044 __merge_refs(&prefs, 2);
1045
1046 while (!list_empty(&prefs)) {
1047 ref = list_first_entry(&prefs, struct __prelim_ref, list);
1048 WARN_ON(ref->count < 0);
1049 if (roots && ref->count && ref->root_id && ref->parent == 0) {
1050 if (root_objectid && ref->root_id != root_objectid) {
1051 ret = BACKREF_FOUND_SHARED;
1052 goto out;
1053 }
1054
1055 /* no parent == root of tree */
1056 ret = ulist_add(roots, ref->root_id, 0, GFP_NOFS);
1057 if (ret < 0)
1058 goto out;
1059 }
1060 if (ref->count && ref->parent) {
1061 if (extent_item_pos && !ref->inode_list &&
1062 ref->level == 0) {
1063 struct extent_buffer *eb;
1064
1065 eb = read_tree_block(fs_info->extent_root,
1066 ref->parent, 0);
1067 if (IS_ERR(eb)) {
1068 ret = PTR_ERR(eb);
1069 goto out;
1070 } else if (!extent_buffer_uptodate(eb)) {
1071 free_extent_buffer(eb);
1072 ret = -EIO;
1073 goto out;
1074 }
1075 btrfs_tree_read_lock(eb);
1076 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1077 ret = find_extent_in_eb(eb, bytenr,
1078 *extent_item_pos, &eie);
1079 btrfs_tree_read_unlock_blocking(eb);
1080 free_extent_buffer(eb);
1081 if (ret < 0)
1082 goto out;
1083 ref->inode_list = eie;
1084 }
1085 ret = ulist_add_merge_ptr(refs, ref->parent,
1086 ref->inode_list,
1087 (void **)&eie, GFP_NOFS);
1088 if (ret < 0)
1089 goto out;
1090 if (!ret && extent_item_pos) {
1091 /*
1092 * we've recorded that parent, so we must extend
1093 * its inode list here
1094 */
1095 BUG_ON(!eie);
1096 while (eie->next)
1097 eie = eie->next;
1098 eie->next = ref->inode_list;
1099 }
1100 eie = NULL;
1101 }
1102 list_del(&ref->list);
1103 kmem_cache_free(btrfs_prelim_ref_cache, ref);
1104 }
1105
1106out:
1107 btrfs_free_path(path);
1108 while (!list_empty(&prefs)) {
1109 ref = list_first_entry(&prefs, struct __prelim_ref, list);
1110 list_del(&ref->list);
1111 kmem_cache_free(btrfs_prelim_ref_cache, ref);
1112 }
1113 while (!list_empty(&prefs_delayed)) {
1114 ref = list_first_entry(&prefs_delayed, struct __prelim_ref,
1115 list);
1116 list_del(&ref->list);
1117 kmem_cache_free(btrfs_prelim_ref_cache, ref);
1118 }
1119 if (ret < 0)
1120 free_inode_elem_list(eie);
1121 return ret;
1122}
1123
1124static void free_leaf_list(struct ulist *blocks)
1125{
1126 struct ulist_node *node = NULL;
1127 struct extent_inode_elem *eie;
1128 struct ulist_iterator uiter;
1129
1130 ULIST_ITER_INIT(&uiter);
1131 while ((node = ulist_next(blocks, &uiter))) {
1132 if (!node->aux)
1133 continue;
1134 eie = (struct extent_inode_elem *)(uintptr_t)node->aux;
1135 free_inode_elem_list(eie);
1136 node->aux = 0;
1137 }
1138
1139 ulist_free(blocks);
1140}
1141
1142/*
1143 * Finds all leafs with a reference to the specified combination of bytenr and
1144 * offset. key_list_head will point to a list of corresponding keys (caller must
1145 * free each list element). The leafs will be stored in the leafs ulist, which
1146 * must be freed with ulist_free.
1147 *
1148 * returns 0 on success, <0 on error
1149 */
1150static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
1151 struct btrfs_fs_info *fs_info, u64 bytenr,
1152 u64 time_seq, struct ulist **leafs,
1153 const u64 *extent_item_pos)
1154{
1155 int ret;
1156
1157 *leafs = ulist_alloc(GFP_NOFS);
1158 if (!*leafs)
1159 return -ENOMEM;
1160
1161 ret = find_parent_nodes(trans, fs_info, bytenr,
1162 time_seq, *leafs, NULL, extent_item_pos, 0, 0);
1163 if (ret < 0 && ret != -ENOENT) {
1164 free_leaf_list(*leafs);
1165 return ret;
1166 }
1167
1168 return 0;
1169}
1170
1171/*
1172 * walk all backrefs for a given extent to find all roots that reference this
1173 * extent. Walking a backref means finding all extents that reference this
1174 * extent and in turn walk the backrefs of those, too. Naturally this is a
1175 * recursive process, but here it is implemented in an iterative fashion: We
1176 * find all referencing extents for the extent in question and put them on a
1177 * list. In turn, we find all referencing extents for those, further appending
1178 * to the list. The way we iterate the list allows adding more elements after
1179 * the current while iterating. The process stops when we reach the end of the
1180 * list. Found roots are added to the roots list.
1181 *
1182 * returns 0 on success, < 0 on error.
1183 */
1184static int __btrfs_find_all_roots(struct btrfs_trans_handle *trans,
1185 struct btrfs_fs_info *fs_info, u64 bytenr,
1186 u64 time_seq, struct ulist **roots)
1187{
1188 struct ulist *tmp;
1189 struct ulist_node *node = NULL;
1190 struct ulist_iterator uiter;
1191 int ret;
1192
1193 tmp = ulist_alloc(GFP_NOFS);
1194 if (!tmp)
1195 return -ENOMEM;
1196 *roots = ulist_alloc(GFP_NOFS);
1197 if (!*roots) {
1198 ulist_free(tmp);
1199 return -ENOMEM;
1200 }
1201
1202 ULIST_ITER_INIT(&uiter);
1203 while (1) {
1204 ret = find_parent_nodes(trans, fs_info, bytenr,
1205 time_seq, tmp, *roots, NULL, 0, 0);
1206 if (ret < 0 && ret != -ENOENT) {
1207 ulist_free(tmp);
1208 ulist_free(*roots);
1209 return ret;
1210 }
1211 node = ulist_next(tmp, &uiter);
1212 if (!node)
1213 break;
1214 bytenr = node->val;
1215 cond_resched();
1216 }
1217
1218 ulist_free(tmp);
1219 return 0;
1220}
1221
1222int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
1223 struct btrfs_fs_info *fs_info, u64 bytenr,
1224 u64 time_seq, struct ulist **roots)
1225{
1226 int ret;
1227
1228 if (!trans)
1229 down_read(&fs_info->commit_root_sem);
1230 ret = __btrfs_find_all_roots(trans, fs_info, bytenr, time_seq, roots);
1231 if (!trans)
1232 up_read(&fs_info->commit_root_sem);
1233 return ret;
1234}
1235
1236/**
1237 * btrfs_check_shared - tell us whether an extent is shared
1238 *
1239 * @trans: optional trans handle
1240 *
1241 * btrfs_check_shared uses the backref walking code but will short
1242 * circuit as soon as it finds a root or inode that doesn't match the
1243 * one passed in. This provides a significant performance benefit for
1244 * callers (such as fiemap) which want to know whether the extent is
1245 * shared but do not need a ref count.
1246 *
1247 * Return: 0 if extent is not shared, 1 if it is shared, < 0 on error.
1248 */
1249int btrfs_check_shared(struct btrfs_trans_handle *trans,
1250 struct btrfs_fs_info *fs_info, u64 root_objectid,
1251 u64 inum, u64 bytenr)
1252{
1253 struct ulist *tmp = NULL;
1254 struct ulist *roots = NULL;
1255 struct ulist_iterator uiter;
1256 struct ulist_node *node;
1257 struct seq_list elem = SEQ_LIST_INIT(elem);
1258 int ret = 0;
1259
1260 tmp = ulist_alloc(GFP_NOFS);
1261 roots = ulist_alloc(GFP_NOFS);
1262 if (!tmp || !roots) {
1263 ulist_free(tmp);
1264 ulist_free(roots);
1265 return -ENOMEM;
1266 }
1267
1268 if (trans)
1269 btrfs_get_tree_mod_seq(fs_info, &elem);
1270 else
1271 down_read(&fs_info->commit_root_sem);
1272 ULIST_ITER_INIT(&uiter);
1273 while (1) {
1274 ret = find_parent_nodes(trans, fs_info, bytenr, elem.seq, tmp,
1275 roots, NULL, root_objectid, inum);
1276 if (ret == BACKREF_FOUND_SHARED) {
1277 /* this is the only condition under which we return 1 */
1278 ret = 1;
1279 break;
1280 }
1281 if (ret < 0 && ret != -ENOENT)
1282 break;
1283 ret = 0;
1284 node = ulist_next(tmp, &uiter);
1285 if (!node)
1286 break;
1287 bytenr = node->val;
1288 cond_resched();
1289 }
1290 if (trans)
1291 btrfs_put_tree_mod_seq(fs_info, &elem);
1292 else
1293 up_read(&fs_info->commit_root_sem);
1294 ulist_free(tmp);
1295 ulist_free(roots);
1296 return ret;
1297}
1298
1299int btrfs_find_one_extref(struct btrfs_root *root, u64 inode_objectid,
1300 u64 start_off, struct btrfs_path *path,
1301 struct btrfs_inode_extref **ret_extref,
1302 u64 *found_off)
1303{
1304 int ret, slot;
1305 struct btrfs_key key;
1306 struct btrfs_key found_key;
1307 struct btrfs_inode_extref *extref;
1308 struct extent_buffer *leaf;
1309 unsigned long ptr;
1310
1311 key.objectid = inode_objectid;
1312 key.type = BTRFS_INODE_EXTREF_KEY;
1313 key.offset = start_off;
1314
1315 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1316 if (ret < 0)
1317 return ret;
1318
1319 while (1) {
1320 leaf = path->nodes[0];
1321 slot = path->slots[0];
1322 if (slot >= btrfs_header_nritems(leaf)) {
1323 /*
1324 * If the item at offset is not found,
1325 * btrfs_search_slot will point us to the slot
1326 * where it should be inserted. In our case
1327 * that will be the slot directly before the
1328 * next INODE_REF_KEY_V2 item. In the case
1329 * that we're pointing to the last slot in a
1330 * leaf, we must move one leaf over.
1331 */
1332 ret = btrfs_next_leaf(root, path);
1333 if (ret) {
1334 if (ret >= 1)
1335 ret = -ENOENT;
1336 break;
1337 }
1338 continue;
1339 }
1340
1341 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1342
1343 /*
1344 * Check that we're still looking at an extended ref key for
1345 * this particular objectid. If we have different
1346 * objectid or type then there are no more to be found
1347 * in the tree and we can exit.
1348 */
1349 ret = -ENOENT;
1350 if (found_key.objectid != inode_objectid)
1351 break;
1352 if (found_key.type != BTRFS_INODE_EXTREF_KEY)
1353 break;
1354
1355 ret = 0;
1356 ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
1357 extref = (struct btrfs_inode_extref *)ptr;
1358 *ret_extref = extref;
1359 if (found_off)
1360 *found_off = found_key.offset;
1361 break;
1362 }
1363
1364 return ret;
1365}
1366
1367/*
1368 * this iterates to turn a name (from iref/extref) into a full filesystem path.
1369 * Elements of the path are separated by '/' and the path is guaranteed to be
1370 * 0-terminated. the path is only given within the current file system.
1371 * Therefore, it never starts with a '/'. the caller is responsible to provide
1372 * "size" bytes in "dest". the dest buffer will be filled backwards. finally,
1373 * the start point of the resulting string is returned. this pointer is within
1374 * dest, normally.
1375 * in case the path buffer would overflow, the pointer is decremented further
1376 * as if output was written to the buffer, though no more output is actually
1377 * generated. that way, the caller can determine how much space would be
1378 * required for the path to fit into the buffer. in that case, the returned
1379 * value will be smaller than dest. callers must check this!
1380 */
1381char *btrfs_ref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
1382 u32 name_len, unsigned long name_off,
1383 struct extent_buffer *eb_in, u64 parent,
1384 char *dest, u32 size)
1385{
1386 int slot;
1387 u64 next_inum;
1388 int ret;
1389 s64 bytes_left = ((s64)size) - 1;
1390 struct extent_buffer *eb = eb_in;
1391 struct btrfs_key found_key;
1392 int leave_spinning = path->leave_spinning;
1393 struct btrfs_inode_ref *iref;
1394
1395 if (bytes_left >= 0)
1396 dest[bytes_left] = '\0';
1397
1398 path->leave_spinning = 1;
1399 while (1) {
1400 bytes_left -= name_len;
1401 if (bytes_left >= 0)
1402 read_extent_buffer(eb, dest + bytes_left,
1403 name_off, name_len);
1404 if (eb != eb_in) {
1405 if (!path->skip_locking)
1406 btrfs_tree_read_unlock_blocking(eb);
1407 free_extent_buffer(eb);
1408 }
1409 ret = btrfs_find_item(fs_root, path, parent, 0,
1410 BTRFS_INODE_REF_KEY, &found_key);
1411 if (ret > 0)
1412 ret = -ENOENT;
1413 if (ret)
1414 break;
1415
1416 next_inum = found_key.offset;
1417
1418 /* regular exit ahead */
1419 if (parent == next_inum)
1420 break;
1421
1422 slot = path->slots[0];
1423 eb = path->nodes[0];
1424 /* make sure we can use eb after releasing the path */
1425 if (eb != eb_in) {
1426 if (!path->skip_locking)
1427 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1428 path->nodes[0] = NULL;
1429 path->locks[0] = 0;
1430 }
1431 btrfs_release_path(path);
1432 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
1433
1434 name_len = btrfs_inode_ref_name_len(eb, iref);
1435 name_off = (unsigned long)(iref + 1);
1436
1437 parent = next_inum;
1438 --bytes_left;
1439 if (bytes_left >= 0)
1440 dest[bytes_left] = '/';
1441 }
1442
1443 btrfs_release_path(path);
1444 path->leave_spinning = leave_spinning;
1445
1446 if (ret)
1447 return ERR_PTR(ret);
1448
1449 return dest + bytes_left;
1450}
1451
1452/*
1453 * this makes the path point to (logical EXTENT_ITEM *)
1454 * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for
1455 * tree blocks and <0 on error.
1456 */
1457int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
1458 struct btrfs_path *path, struct btrfs_key *found_key,
1459 u64 *flags_ret)
1460{
1461 int ret;
1462 u64 flags;
1463 u64 size = 0;
1464 u32 item_size;
1465 struct extent_buffer *eb;
1466 struct btrfs_extent_item *ei;
1467 struct btrfs_key key;
1468
1469 if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
1470 key.type = BTRFS_METADATA_ITEM_KEY;
1471 else
1472 key.type = BTRFS_EXTENT_ITEM_KEY;
1473 key.objectid = logical;
1474 key.offset = (u64)-1;
1475
1476 ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
1477 if (ret < 0)
1478 return ret;
1479
1480 ret = btrfs_previous_extent_item(fs_info->extent_root, path, 0);
1481 if (ret) {
1482 if (ret > 0)
1483 ret = -ENOENT;
1484 return ret;
1485 }
1486 btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]);
1487 if (found_key->type == BTRFS_METADATA_ITEM_KEY)
1488 size = fs_info->extent_root->nodesize;
1489 else if (found_key->type == BTRFS_EXTENT_ITEM_KEY)
1490 size = found_key->offset;
1491
1492 if (found_key->objectid > logical ||
1493 found_key->objectid + size <= logical) {
1494 pr_debug("logical %llu is not within any extent\n", logical);
1495 return -ENOENT;
1496 }
1497
1498 eb = path->nodes[0];
1499 item_size = btrfs_item_size_nr(eb, path->slots[0]);
1500 BUG_ON(item_size < sizeof(*ei));
1501
1502 ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
1503 flags = btrfs_extent_flags(eb, ei);
1504
1505 pr_debug("logical %llu is at position %llu within the extent (%llu "
1506 "EXTENT_ITEM %llu) flags %#llx size %u\n",
1507 logical, logical - found_key->objectid, found_key->objectid,
1508 found_key->offset, flags, item_size);
1509
1510 WARN_ON(!flags_ret);
1511 if (flags_ret) {
1512 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1513 *flags_ret = BTRFS_EXTENT_FLAG_TREE_BLOCK;
1514 else if (flags & BTRFS_EXTENT_FLAG_DATA)
1515 *flags_ret = BTRFS_EXTENT_FLAG_DATA;
1516 else
1517 BUG_ON(1);
1518 return 0;
1519 }
1520
1521 return -EIO;
1522}
1523
1524/*
1525 * helper function to iterate extent inline refs. ptr must point to a 0 value
1526 * for the first call and may be modified. it is used to track state.
1527 * if more refs exist, 0 is returned and the next call to
1528 * __get_extent_inline_ref must pass the modified ptr parameter to get the
1529 * next ref. after the last ref was processed, 1 is returned.
1530 * returns <0 on error
1531 */
1532static int __get_extent_inline_ref(unsigned long *ptr, struct extent_buffer *eb,
1533 struct btrfs_key *key,
1534 struct btrfs_extent_item *ei, u32 item_size,
1535 struct btrfs_extent_inline_ref **out_eiref,
1536 int *out_type)
1537{
1538 unsigned long end;
1539 u64 flags;
1540 struct btrfs_tree_block_info *info;
1541
1542 if (!*ptr) {
1543 /* first call */
1544 flags = btrfs_extent_flags(eb, ei);
1545 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1546 if (key->type == BTRFS_METADATA_ITEM_KEY) {
1547 /* a skinny metadata extent */
1548 *out_eiref =
1549 (struct btrfs_extent_inline_ref *)(ei + 1);
1550 } else {
1551 WARN_ON(key->type != BTRFS_EXTENT_ITEM_KEY);
1552 info = (struct btrfs_tree_block_info *)(ei + 1);
1553 *out_eiref =
1554 (struct btrfs_extent_inline_ref *)(info + 1);
1555 }
1556 } else {
1557 *out_eiref = (struct btrfs_extent_inline_ref *)(ei + 1);
1558 }
1559 *ptr = (unsigned long)*out_eiref;
1560 if ((unsigned long)(*ptr) >= (unsigned long)ei + item_size)
1561 return -ENOENT;
1562 }
1563
1564 end = (unsigned long)ei + item_size;
1565 *out_eiref = (struct btrfs_extent_inline_ref *)(*ptr);
1566 *out_type = btrfs_extent_inline_ref_type(eb, *out_eiref);
1567
1568 *ptr += btrfs_extent_inline_ref_size(*out_type);
1569 WARN_ON(*ptr > end);
1570 if (*ptr == end)
1571 return 1; /* last */
1572
1573 return 0;
1574}
1575
1576/*
1577 * reads the tree block backref for an extent. tree level and root are returned
1578 * through out_level and out_root. ptr must point to a 0 value for the first
1579 * call and may be modified (see __get_extent_inline_ref comment).
1580 * returns 0 if data was provided, 1 if there was no more data to provide or
1581 * <0 on error.
1582 */
1583int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
1584 struct btrfs_key *key, struct btrfs_extent_item *ei,
1585 u32 item_size, u64 *out_root, u8 *out_level)
1586{
1587 int ret;
1588 int type;
1589 struct btrfs_extent_inline_ref *eiref;
1590
1591 if (*ptr == (unsigned long)-1)
1592 return 1;
1593
1594 while (1) {
1595 ret = __get_extent_inline_ref(ptr, eb, key, ei, item_size,
1596 &eiref, &type);
1597 if (ret < 0)
1598 return ret;
1599
1600 if (type == BTRFS_TREE_BLOCK_REF_KEY ||
1601 type == BTRFS_SHARED_BLOCK_REF_KEY)
1602 break;
1603
1604 if (ret == 1)
1605 return 1;
1606 }
1607
1608 /* we can treat both ref types equally here */
1609 *out_root = btrfs_extent_inline_ref_offset(eb, eiref);
1610
1611 if (key->type == BTRFS_EXTENT_ITEM_KEY) {
1612 struct btrfs_tree_block_info *info;
1613
1614 info = (struct btrfs_tree_block_info *)(ei + 1);
1615 *out_level = btrfs_tree_block_level(eb, info);
1616 } else {
1617 ASSERT(key->type == BTRFS_METADATA_ITEM_KEY);
1618 *out_level = (u8)key->offset;
1619 }
1620
1621 if (ret == 1)
1622 *ptr = (unsigned long)-1;
1623
1624 return 0;
1625}
1626
1627static int iterate_leaf_refs(struct extent_inode_elem *inode_list,
1628 u64 root, u64 extent_item_objectid,
1629 iterate_extent_inodes_t *iterate, void *ctx)
1630{
1631 struct extent_inode_elem *eie;
1632 int ret = 0;
1633
1634 for (eie = inode_list; eie; eie = eie->next) {
1635 pr_debug("ref for %llu resolved, key (%llu EXTEND_DATA %llu), "
1636 "root %llu\n", extent_item_objectid,
1637 eie->inum, eie->offset, root);
1638 ret = iterate(eie->inum, eie->offset, root, ctx);
1639 if (ret) {
1640 pr_debug("stopping iteration for %llu due to ret=%d\n",
1641 extent_item_objectid, ret);
1642 break;
1643 }
1644 }
1645
1646 return ret;
1647}
1648
1649/*
1650 * calls iterate() for every inode that references the extent identified by
1651 * the given parameters.
1652 * when the iterator function returns a non-zero value, iteration stops.
1653 */
1654int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
1655 u64 extent_item_objectid, u64 extent_item_pos,
1656 int search_commit_root,
1657 iterate_extent_inodes_t *iterate, void *ctx)
1658{
1659 int ret;
1660 struct btrfs_trans_handle *trans = NULL;
1661 struct ulist *refs = NULL;
1662 struct ulist *roots = NULL;
1663 struct ulist_node *ref_node = NULL;
1664 struct ulist_node *root_node = NULL;
1665 struct seq_list tree_mod_seq_elem = SEQ_LIST_INIT(tree_mod_seq_elem);
1666 struct ulist_iterator ref_uiter;
1667 struct ulist_iterator root_uiter;
1668
1669 pr_debug("resolving all inodes for extent %llu\n",
1670 extent_item_objectid);
1671
1672 if (!search_commit_root) {
1673 trans = btrfs_join_transaction(fs_info->extent_root);
1674 if (IS_ERR(trans))
1675 return PTR_ERR(trans);
1676 btrfs_get_tree_mod_seq(fs_info, &tree_mod_seq_elem);
1677 } else {
1678 down_read(&fs_info->commit_root_sem);
1679 }
1680
1681 ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
1682 tree_mod_seq_elem.seq, &refs,
1683 &extent_item_pos);
1684 if (ret)
1685 goto out;
1686
1687 ULIST_ITER_INIT(&ref_uiter);
1688 while (!ret && (ref_node = ulist_next(refs, &ref_uiter))) {
1689 ret = __btrfs_find_all_roots(trans, fs_info, ref_node->val,
1690 tree_mod_seq_elem.seq, &roots);
1691 if (ret)
1692 break;
1693 ULIST_ITER_INIT(&root_uiter);
1694 while (!ret && (root_node = ulist_next(roots, &root_uiter))) {
1695 pr_debug("root %llu references leaf %llu, data list "
1696 "%#llx\n", root_node->val, ref_node->val,
1697 ref_node->aux);
1698 ret = iterate_leaf_refs((struct extent_inode_elem *)
1699 (uintptr_t)ref_node->aux,
1700 root_node->val,
1701 extent_item_objectid,
1702 iterate, ctx);
1703 }
1704 ulist_free(roots);
1705 }
1706
1707 free_leaf_list(refs);
1708out:
1709 if (!search_commit_root) {
1710 btrfs_put_tree_mod_seq(fs_info, &tree_mod_seq_elem);
1711 btrfs_end_transaction(trans, fs_info->extent_root);
1712 } else {
1713 up_read(&fs_info->commit_root_sem);
1714 }
1715
1716 return ret;
1717}
1718
1719int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
1720 struct btrfs_path *path,
1721 iterate_extent_inodes_t *iterate, void *ctx)
1722{
1723 int ret;
1724 u64 extent_item_pos;
1725 u64 flags = 0;
1726 struct btrfs_key found_key;
1727 int search_commit_root = path->search_commit_root;
1728
1729 ret = extent_from_logical(fs_info, logical, path, &found_key, &flags);
1730 btrfs_release_path(path);
1731 if (ret < 0)
1732 return ret;
1733 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1734 return -EINVAL;
1735
1736 extent_item_pos = logical - found_key.objectid;
1737 ret = iterate_extent_inodes(fs_info, found_key.objectid,
1738 extent_item_pos, search_commit_root,
1739 iterate, ctx);
1740
1741 return ret;
1742}
1743
1744typedef int (iterate_irefs_t)(u64 parent, u32 name_len, unsigned long name_off,
1745 struct extent_buffer *eb, void *ctx);
1746
1747static int iterate_inode_refs(u64 inum, struct btrfs_root *fs_root,
1748 struct btrfs_path *path,
1749 iterate_irefs_t *iterate, void *ctx)
1750{
1751 int ret = 0;
1752 int slot;
1753 u32 cur;
1754 u32 len;
1755 u32 name_len;
1756 u64 parent = 0;
1757 int found = 0;
1758 struct extent_buffer *eb;
1759 struct btrfs_item *item;
1760 struct btrfs_inode_ref *iref;
1761 struct btrfs_key found_key;
1762
1763 while (!ret) {
1764 ret = btrfs_find_item(fs_root, path, inum,
1765 parent ? parent + 1 : 0, BTRFS_INODE_REF_KEY,
1766 &found_key);
1767
1768 if (ret < 0)
1769 break;
1770 if (ret) {
1771 ret = found ? 0 : -ENOENT;
1772 break;
1773 }
1774 ++found;
1775
1776 parent = found_key.offset;
1777 slot = path->slots[0];
1778 eb = btrfs_clone_extent_buffer(path->nodes[0]);
1779 if (!eb) {
1780 ret = -ENOMEM;
1781 break;
1782 }
1783 extent_buffer_get(eb);
1784 btrfs_tree_read_lock(eb);
1785 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1786 btrfs_release_path(path);
1787
1788 item = btrfs_item_nr(slot);
1789 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
1790
1791 for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) {
1792 name_len = btrfs_inode_ref_name_len(eb, iref);
1793 /* path must be released before calling iterate()! */
1794 pr_debug("following ref at offset %u for inode %llu in "
1795 "tree %llu\n", cur, found_key.objectid,
1796 fs_root->objectid);
1797 ret = iterate(parent, name_len,
1798 (unsigned long)(iref + 1), eb, ctx);
1799 if (ret)
1800 break;
1801 len = sizeof(*iref) + name_len;
1802 iref = (struct btrfs_inode_ref *)((char *)iref + len);
1803 }
1804 btrfs_tree_read_unlock_blocking(eb);
1805 free_extent_buffer(eb);
1806 }
1807
1808 btrfs_release_path(path);
1809
1810 return ret;
1811}
1812
1813static int iterate_inode_extrefs(u64 inum, struct btrfs_root *fs_root,
1814 struct btrfs_path *path,
1815 iterate_irefs_t *iterate, void *ctx)
1816{
1817 int ret;
1818 int slot;
1819 u64 offset = 0;
1820 u64 parent;
1821 int found = 0;
1822 struct extent_buffer *eb;
1823 struct btrfs_inode_extref *extref;
1824 u32 item_size;
1825 u32 cur_offset;
1826 unsigned long ptr;
1827
1828 while (1) {
1829 ret = btrfs_find_one_extref(fs_root, inum, offset, path, &extref,
1830 &offset);
1831 if (ret < 0)
1832 break;
1833 if (ret) {
1834 ret = found ? 0 : -ENOENT;
1835 break;
1836 }
1837 ++found;
1838
1839 slot = path->slots[0];
1840 eb = btrfs_clone_extent_buffer(path->nodes[0]);
1841 if (!eb) {
1842 ret = -ENOMEM;
1843 break;
1844 }
1845 extent_buffer_get(eb);
1846
1847 btrfs_tree_read_lock(eb);
1848 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1849 btrfs_release_path(path);
1850
1851 item_size = btrfs_item_size_nr(eb, slot);
1852 ptr = btrfs_item_ptr_offset(eb, slot);
1853 cur_offset = 0;
1854
1855 while (cur_offset < item_size) {
1856 u32 name_len;
1857
1858 extref = (struct btrfs_inode_extref *)(ptr + cur_offset);
1859 parent = btrfs_inode_extref_parent(eb, extref);
1860 name_len = btrfs_inode_extref_name_len(eb, extref);
1861 ret = iterate(parent, name_len,
1862 (unsigned long)&extref->name, eb, ctx);
1863 if (ret)
1864 break;
1865
1866 cur_offset += btrfs_inode_extref_name_len(eb, extref);
1867 cur_offset += sizeof(*extref);
1868 }
1869 btrfs_tree_read_unlock_blocking(eb);
1870 free_extent_buffer(eb);
1871
1872 offset++;
1873 }
1874
1875 btrfs_release_path(path);
1876
1877 return ret;
1878}
1879
1880static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
1881 struct btrfs_path *path, iterate_irefs_t *iterate,
1882 void *ctx)
1883{
1884 int ret;
1885 int found_refs = 0;
1886
1887 ret = iterate_inode_refs(inum, fs_root, path, iterate, ctx);
1888 if (!ret)
1889 ++found_refs;
1890 else if (ret != -ENOENT)
1891 return ret;
1892
1893 ret = iterate_inode_extrefs(inum, fs_root, path, iterate, ctx);
1894 if (ret == -ENOENT && found_refs)
1895 return 0;
1896
1897 return ret;
1898}
1899
1900/*
1901 * returns 0 if the path could be dumped (probably truncated)
1902 * returns <0 in case of an error
1903 */
1904static int inode_to_path(u64 inum, u32 name_len, unsigned long name_off,
1905 struct extent_buffer *eb, void *ctx)
1906{
1907 struct inode_fs_paths *ipath = ctx;
1908 char *fspath;
1909 char *fspath_min;
1910 int i = ipath->fspath->elem_cnt;
1911 const int s_ptr = sizeof(char *);
1912 u32 bytes_left;
1913
1914 bytes_left = ipath->fspath->bytes_left > s_ptr ?
1915 ipath->fspath->bytes_left - s_ptr : 0;
1916
1917 fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
1918 fspath = btrfs_ref_to_path(ipath->fs_root, ipath->btrfs_path, name_len,
1919 name_off, eb, inum, fspath_min, bytes_left);
1920 if (IS_ERR(fspath))
1921 return PTR_ERR(fspath);
1922
1923 if (fspath > fspath_min) {
1924 ipath->fspath->val[i] = (u64)(unsigned long)fspath;
1925 ++ipath->fspath->elem_cnt;
1926 ipath->fspath->bytes_left = fspath - fspath_min;
1927 } else {
1928 ++ipath->fspath->elem_missed;
1929 ipath->fspath->bytes_missing += fspath_min - fspath;
1930 ipath->fspath->bytes_left = 0;
1931 }
1932
1933 return 0;
1934}
1935
1936/*
1937 * this dumps all file system paths to the inode into the ipath struct, provided
1938 * is has been created large enough. each path is zero-terminated and accessed
1939 * from ipath->fspath->val[i].
1940 * when it returns, there are ipath->fspath->elem_cnt number of paths available
1941 * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
1942 * number of missed paths in recored in ipath->fspath->elem_missed, otherwise,
1943 * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
1944 * have been needed to return all paths.
1945 */
1946int paths_from_inode(u64 inum, struct inode_fs_paths *ipath)
1947{
1948 return iterate_irefs(inum, ipath->fs_root, ipath->btrfs_path,
1949 inode_to_path, ipath);
1950}
1951
1952struct btrfs_data_container *init_data_container(u32 total_bytes)
1953{
1954 struct btrfs_data_container *data;
1955 size_t alloc_bytes;
1956
1957 alloc_bytes = max_t(size_t, total_bytes, sizeof(*data));
1958 data = vmalloc(alloc_bytes);
1959 if (!data)
1960 return ERR_PTR(-ENOMEM);
1961
1962 if (total_bytes >= sizeof(*data)) {
1963 data->bytes_left = total_bytes - sizeof(*data);
1964 data->bytes_missing = 0;
1965 } else {
1966 data->bytes_missing = sizeof(*data) - total_bytes;
1967 data->bytes_left = 0;
1968 }
1969
1970 data->elem_cnt = 0;
1971 data->elem_missed = 0;
1972
1973 return data;
1974}
1975
1976/*
1977 * allocates space to return multiple file system paths for an inode.
1978 * total_bytes to allocate are passed, note that space usable for actual path
1979 * information will be total_bytes - sizeof(struct inode_fs_paths).
1980 * the returned pointer must be freed with free_ipath() in the end.
1981 */
1982struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
1983 struct btrfs_path *path)
1984{
1985 struct inode_fs_paths *ifp;
1986 struct btrfs_data_container *fspath;
1987
1988 fspath = init_data_container(total_bytes);
1989 if (IS_ERR(fspath))
1990 return (void *)fspath;
1991
1992 ifp = kmalloc(sizeof(*ifp), GFP_NOFS);
1993 if (!ifp) {
1994 kfree(fspath);
1995 return ERR_PTR(-ENOMEM);
1996 }
1997
1998 ifp->btrfs_path = path;
1999 ifp->fspath = fspath;
2000 ifp->fs_root = fs_root;
2001
2002 return ifp;
2003}
2004
2005void free_ipath(struct inode_fs_paths *ipath)
2006{
2007 if (!ipath)
2008 return;
2009 vfree(ipath->fspath);
2010 kfree(ipath);
2011}