spi-topcliff-pch: Fix issue for transmitting over 4KByte
[zen-stable.git] / fs / btrfs / backref.c
blob56136d9046f099b702d77e95cdf160ce8af85e2c
1 /*
2 * Copyright (C) 2011 STRATO. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include "ctree.h"
20 #include "disk-io.h"
21 #include "backref.h"
22 #include "ulist.h"
23 #include "transaction.h"
24 #include "delayed-ref.h"
27 * this structure records all encountered refs on the way up to the root
29 struct __prelim_ref {
30 struct list_head list;
31 u64 root_id;
32 struct btrfs_key key;
33 int level;
34 int count;
35 u64 parent;
36 u64 wanted_disk_byte;
39 static int __add_prelim_ref(struct list_head *head, u64 root_id,
40 struct btrfs_key *key, int level, u64 parent,
41 u64 wanted_disk_byte, int count)
43 struct __prelim_ref *ref;
45 /* in case we're adding delayed refs, we're holding the refs spinlock */
46 ref = kmalloc(sizeof(*ref), GFP_ATOMIC);
47 if (!ref)
48 return -ENOMEM;
50 ref->root_id = root_id;
51 if (key)
52 ref->key = *key;
53 else
54 memset(&ref->key, 0, sizeof(ref->key));
56 ref->level = level;
57 ref->count = count;
58 ref->parent = parent;
59 ref->wanted_disk_byte = wanted_disk_byte;
60 list_add_tail(&ref->list, head);
62 return 0;
65 static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
66 struct ulist *parents,
67 struct extent_buffer *eb, int level,
68 u64 wanted_objectid, u64 wanted_disk_byte)
70 int ret;
71 int slot;
72 struct btrfs_file_extent_item *fi;
73 struct btrfs_key key;
74 u64 disk_byte;
76 add_parent:
77 ret = ulist_add(parents, eb->start, 0, GFP_NOFS);
78 if (ret < 0)
79 return ret;
81 if (level != 0)
82 return 0;
85 * if the current leaf is full with EXTENT_DATA items, we must
86 * check the next one if that holds a reference as well.
87 * ref->count cannot be used to skip this check.
88 * repeat this until we don't find any additional EXTENT_DATA items.
90 while (1) {
91 ret = btrfs_next_leaf(root, path);
92 if (ret < 0)
93 return ret;
94 if (ret)
95 return 0;
97 eb = path->nodes[0];
98 for (slot = 0; slot < btrfs_header_nritems(eb); ++slot) {
99 btrfs_item_key_to_cpu(eb, &key, slot);
100 if (key.objectid != wanted_objectid ||
101 key.type != BTRFS_EXTENT_DATA_KEY)
102 return 0;
103 fi = btrfs_item_ptr(eb, slot,
104 struct btrfs_file_extent_item);
105 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
106 if (disk_byte == wanted_disk_byte)
107 goto add_parent;
111 return 0;
115 * resolve an indirect backref in the form (root_id, key, level)
116 * to a logical address
118 static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
119 int search_commit_root,
120 struct __prelim_ref *ref,
121 struct ulist *parents)
123 struct btrfs_path *path;
124 struct btrfs_root *root;
125 struct btrfs_key root_key;
126 struct btrfs_key key = {0};
127 struct extent_buffer *eb;
128 int ret = 0;
129 int root_level;
130 int level = ref->level;
132 path = btrfs_alloc_path();
133 if (!path)
134 return -ENOMEM;
135 path->search_commit_root = !!search_commit_root;
137 root_key.objectid = ref->root_id;
138 root_key.type = BTRFS_ROOT_ITEM_KEY;
139 root_key.offset = (u64)-1;
140 root = btrfs_read_fs_root_no_name(fs_info, &root_key);
141 if (IS_ERR(root)) {
142 ret = PTR_ERR(root);
143 goto out;
146 rcu_read_lock();
147 root_level = btrfs_header_level(root->node);
148 rcu_read_unlock();
150 if (root_level + 1 == level)
151 goto out;
153 path->lowest_level = level;
154 ret = btrfs_search_slot(NULL, root, &ref->key, path, 0, 0);
155 pr_debug("search slot in root %llu (level %d, ref count %d) returned "
156 "%d for key (%llu %u %llu)\n",
157 (unsigned long long)ref->root_id, level, ref->count, ret,
158 (unsigned long long)ref->key.objectid, ref->key.type,
159 (unsigned long long)ref->key.offset);
160 if (ret < 0)
161 goto out;
163 eb = path->nodes[level];
164 if (!eb) {
165 WARN_ON(1);
166 ret = 1;
167 goto out;
170 if (level == 0) {
171 if (ret == 1 && path->slots[0] >= btrfs_header_nritems(eb)) {
172 ret = btrfs_next_leaf(root, path);
173 if (ret)
174 goto out;
175 eb = path->nodes[0];
178 btrfs_item_key_to_cpu(eb, &key, path->slots[0]);
181 /* the last two parameters will only be used for level == 0 */
182 ret = add_all_parents(root, path, parents, eb, level, key.objectid,
183 ref->wanted_disk_byte);
184 out:
185 btrfs_free_path(path);
186 return ret;
190 * resolve all indirect backrefs from the list
192 static int __resolve_indirect_refs(struct btrfs_fs_info *fs_info,
193 int search_commit_root,
194 struct list_head *head)
196 int err;
197 int ret = 0;
198 struct __prelim_ref *ref;
199 struct __prelim_ref *ref_safe;
200 struct __prelim_ref *new_ref;
201 struct ulist *parents;
202 struct ulist_node *node;
204 parents = ulist_alloc(GFP_NOFS);
205 if (!parents)
206 return -ENOMEM;
209 * _safe allows us to insert directly after the current item without
210 * iterating over the newly inserted items.
211 * we're also allowed to re-assign ref during iteration.
213 list_for_each_entry_safe(ref, ref_safe, head, list) {
214 if (ref->parent) /* already direct */
215 continue;
216 if (ref->count == 0)
217 continue;
218 err = __resolve_indirect_ref(fs_info, search_commit_root,
219 ref, parents);
220 if (err) {
221 if (ret == 0)
222 ret = err;
223 continue;
226 /* we put the first parent into the ref at hand */
227 node = ulist_next(parents, NULL);
228 ref->parent = node ? node->val : 0;
230 /* additional parents require new refs being added here */
231 while ((node = ulist_next(parents, node))) {
232 new_ref = kmalloc(sizeof(*new_ref), GFP_NOFS);
233 if (!new_ref) {
234 ret = -ENOMEM;
235 break;
237 memcpy(new_ref, ref, sizeof(*ref));
238 new_ref->parent = node->val;
239 list_add(&new_ref->list, &ref->list);
241 ulist_reinit(parents);
244 ulist_free(parents);
245 return ret;
249 * merge two lists of backrefs and adjust counts accordingly
251 * mode = 1: merge identical keys, if key is set
252 * mode = 2: merge identical parents
254 static int __merge_refs(struct list_head *head, int mode)
256 struct list_head *pos1;
258 list_for_each(pos1, head) {
259 struct list_head *n2;
260 struct list_head *pos2;
261 struct __prelim_ref *ref1;
263 ref1 = list_entry(pos1, struct __prelim_ref, list);
265 if (mode == 1 && ref1->key.type == 0)
266 continue;
267 for (pos2 = pos1->next, n2 = pos2->next; pos2 != head;
268 pos2 = n2, n2 = pos2->next) {
269 struct __prelim_ref *ref2;
271 ref2 = list_entry(pos2, struct __prelim_ref, list);
273 if (mode == 1) {
274 if (memcmp(&ref1->key, &ref2->key,
275 sizeof(ref1->key)) ||
276 ref1->level != ref2->level ||
277 ref1->root_id != ref2->root_id)
278 continue;
279 ref1->count += ref2->count;
280 } else {
281 if (ref1->parent != ref2->parent)
282 continue;
283 ref1->count += ref2->count;
285 list_del(&ref2->list);
286 kfree(ref2);
290 return 0;
294 * add all currently queued delayed refs from this head whose seq nr is
295 * smaller or equal that seq to the list
297 static int __add_delayed_refs(struct btrfs_delayed_ref_head *head, u64 seq,
298 struct btrfs_key *info_key,
299 struct list_head *prefs)
301 struct btrfs_delayed_extent_op *extent_op = head->extent_op;
302 struct rb_node *n = &head->node.rb_node;
303 int sgn;
304 int ret = 0;
306 if (extent_op && extent_op->update_key)
307 btrfs_disk_key_to_cpu(info_key, &extent_op->key);
309 while ((n = rb_prev(n))) {
310 struct btrfs_delayed_ref_node *node;
311 node = rb_entry(n, struct btrfs_delayed_ref_node,
312 rb_node);
313 if (node->bytenr != head->node.bytenr)
314 break;
315 WARN_ON(node->is_head);
317 if (node->seq > seq)
318 continue;
320 switch (node->action) {
321 case BTRFS_ADD_DELAYED_EXTENT:
322 case BTRFS_UPDATE_DELAYED_HEAD:
323 WARN_ON(1);
324 continue;
325 case BTRFS_ADD_DELAYED_REF:
326 sgn = 1;
327 break;
328 case BTRFS_DROP_DELAYED_REF:
329 sgn = -1;
330 break;
331 default:
332 BUG_ON(1);
334 switch (node->type) {
335 case BTRFS_TREE_BLOCK_REF_KEY: {
336 struct btrfs_delayed_tree_ref *ref;
338 ref = btrfs_delayed_node_to_tree_ref(node);
339 ret = __add_prelim_ref(prefs, ref->root, info_key,
340 ref->level + 1, 0, node->bytenr,
341 node->ref_mod * sgn);
342 break;
344 case BTRFS_SHARED_BLOCK_REF_KEY: {
345 struct btrfs_delayed_tree_ref *ref;
347 ref = btrfs_delayed_node_to_tree_ref(node);
348 ret = __add_prelim_ref(prefs, ref->root, info_key,
349 ref->level + 1, ref->parent,
350 node->bytenr,
351 node->ref_mod * sgn);
352 break;
354 case BTRFS_EXTENT_DATA_REF_KEY: {
355 struct btrfs_delayed_data_ref *ref;
356 struct btrfs_key key;
358 ref = btrfs_delayed_node_to_data_ref(node);
360 key.objectid = ref->objectid;
361 key.type = BTRFS_EXTENT_DATA_KEY;
362 key.offset = ref->offset;
363 ret = __add_prelim_ref(prefs, ref->root, &key, 0, 0,
364 node->bytenr,
365 node->ref_mod * sgn);
366 break;
368 case BTRFS_SHARED_DATA_REF_KEY: {
369 struct btrfs_delayed_data_ref *ref;
370 struct btrfs_key key;
372 ref = btrfs_delayed_node_to_data_ref(node);
374 key.objectid = ref->objectid;
375 key.type = BTRFS_EXTENT_DATA_KEY;
376 key.offset = ref->offset;
377 ret = __add_prelim_ref(prefs, ref->root, &key, 0,
378 ref->parent, node->bytenr,
379 node->ref_mod * sgn);
380 break;
382 default:
383 WARN_ON(1);
385 BUG_ON(ret);
388 return 0;
392 * add all inline backrefs for bytenr to the list
394 static int __add_inline_refs(struct btrfs_fs_info *fs_info,
395 struct btrfs_path *path, u64 bytenr,
396 struct btrfs_key *info_key, int *info_level,
397 struct list_head *prefs)
399 int ret = 0;
400 int slot;
401 struct extent_buffer *leaf;
402 struct btrfs_key key;
403 unsigned long ptr;
404 unsigned long end;
405 struct btrfs_extent_item *ei;
406 u64 flags;
407 u64 item_size;
410 * enumerate all inline refs
412 leaf = path->nodes[0];
413 slot = path->slots[0] - 1;
415 item_size = btrfs_item_size_nr(leaf, slot);
416 BUG_ON(item_size < sizeof(*ei));
418 ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
419 flags = btrfs_extent_flags(leaf, ei);
421 ptr = (unsigned long)(ei + 1);
422 end = (unsigned long)ei + item_size;
424 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
425 struct btrfs_tree_block_info *info;
426 struct btrfs_disk_key disk_key;
428 info = (struct btrfs_tree_block_info *)ptr;
429 *info_level = btrfs_tree_block_level(leaf, info);
430 btrfs_tree_block_key(leaf, info, &disk_key);
431 btrfs_disk_key_to_cpu(info_key, &disk_key);
432 ptr += sizeof(struct btrfs_tree_block_info);
433 BUG_ON(ptr > end);
434 } else {
435 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
438 while (ptr < end) {
439 struct btrfs_extent_inline_ref *iref;
440 u64 offset;
441 int type;
443 iref = (struct btrfs_extent_inline_ref *)ptr;
444 type = btrfs_extent_inline_ref_type(leaf, iref);
445 offset = btrfs_extent_inline_ref_offset(leaf, iref);
447 switch (type) {
448 case BTRFS_SHARED_BLOCK_REF_KEY:
449 ret = __add_prelim_ref(prefs, 0, info_key,
450 *info_level + 1, offset,
451 bytenr, 1);
452 break;
453 case BTRFS_SHARED_DATA_REF_KEY: {
454 struct btrfs_shared_data_ref *sdref;
455 int count;
457 sdref = (struct btrfs_shared_data_ref *)(iref + 1);
458 count = btrfs_shared_data_ref_count(leaf, sdref);
459 ret = __add_prelim_ref(prefs, 0, NULL, 0, offset,
460 bytenr, count);
461 break;
463 case BTRFS_TREE_BLOCK_REF_KEY:
464 ret = __add_prelim_ref(prefs, offset, info_key,
465 *info_level + 1, 0, bytenr, 1);
466 break;
467 case BTRFS_EXTENT_DATA_REF_KEY: {
468 struct btrfs_extent_data_ref *dref;
469 int count;
470 u64 root;
472 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
473 count = btrfs_extent_data_ref_count(leaf, dref);
474 key.objectid = btrfs_extent_data_ref_objectid(leaf,
475 dref);
476 key.type = BTRFS_EXTENT_DATA_KEY;
477 key.offset = btrfs_extent_data_ref_offset(leaf, dref);
478 root = btrfs_extent_data_ref_root(leaf, dref);
479 ret = __add_prelim_ref(prefs, root, &key, 0, 0, bytenr,
480 count);
481 break;
483 default:
484 WARN_ON(1);
486 BUG_ON(ret);
487 ptr += btrfs_extent_inline_ref_size(type);
490 return 0;
494 * add all non-inline backrefs for bytenr to the list
496 static int __add_keyed_refs(struct btrfs_fs_info *fs_info,
497 struct btrfs_path *path, u64 bytenr,
498 struct btrfs_key *info_key, int info_level,
499 struct list_head *prefs)
501 struct btrfs_root *extent_root = fs_info->extent_root;
502 int ret;
503 int slot;
504 struct extent_buffer *leaf;
505 struct btrfs_key key;
507 while (1) {
508 ret = btrfs_next_item(extent_root, path);
509 if (ret < 0)
510 break;
511 if (ret) {
512 ret = 0;
513 break;
516 slot = path->slots[0];
517 leaf = path->nodes[0];
518 btrfs_item_key_to_cpu(leaf, &key, slot);
520 if (key.objectid != bytenr)
521 break;
522 if (key.type < BTRFS_TREE_BLOCK_REF_KEY)
523 continue;
524 if (key.type > BTRFS_SHARED_DATA_REF_KEY)
525 break;
527 switch (key.type) {
528 case BTRFS_SHARED_BLOCK_REF_KEY:
529 ret = __add_prelim_ref(prefs, 0, info_key,
530 info_level + 1, key.offset,
531 bytenr, 1);
532 break;
533 case BTRFS_SHARED_DATA_REF_KEY: {
534 struct btrfs_shared_data_ref *sdref;
535 int count;
537 sdref = btrfs_item_ptr(leaf, slot,
538 struct btrfs_shared_data_ref);
539 count = btrfs_shared_data_ref_count(leaf, sdref);
540 ret = __add_prelim_ref(prefs, 0, NULL, 0, key.offset,
541 bytenr, count);
542 break;
544 case BTRFS_TREE_BLOCK_REF_KEY:
545 ret = __add_prelim_ref(prefs, key.offset, info_key,
546 info_level + 1, 0, bytenr, 1);
547 break;
548 case BTRFS_EXTENT_DATA_REF_KEY: {
549 struct btrfs_extent_data_ref *dref;
550 int count;
551 u64 root;
553 dref = btrfs_item_ptr(leaf, slot,
554 struct btrfs_extent_data_ref);
555 count = btrfs_extent_data_ref_count(leaf, dref);
556 key.objectid = btrfs_extent_data_ref_objectid(leaf,
557 dref);
558 key.type = BTRFS_EXTENT_DATA_KEY;
559 key.offset = btrfs_extent_data_ref_offset(leaf, dref);
560 root = btrfs_extent_data_ref_root(leaf, dref);
561 ret = __add_prelim_ref(prefs, root, &key, 0, 0,
562 bytenr, count);
563 break;
565 default:
566 WARN_ON(1);
568 BUG_ON(ret);
571 return ret;
575 * this adds all existing backrefs (inline backrefs, backrefs and delayed
576 * refs) for the given bytenr to the refs list, merges duplicates and resolves
577 * indirect refs to their parent bytenr.
578 * When roots are found, they're added to the roots list
580 * FIXME some caching might speed things up
582 static int find_parent_nodes(struct btrfs_trans_handle *trans,
583 struct btrfs_fs_info *fs_info, u64 bytenr,
584 u64 seq, struct ulist *refs, struct ulist *roots)
586 struct btrfs_key key;
587 struct btrfs_path *path;
588 struct btrfs_key info_key = { 0 };
589 struct btrfs_delayed_ref_root *delayed_refs = NULL;
590 struct btrfs_delayed_ref_head *head;
591 int info_level = 0;
592 int ret;
593 int search_commit_root = (trans == BTRFS_BACKREF_SEARCH_COMMIT_ROOT);
594 struct list_head prefs_delayed;
595 struct list_head prefs;
596 struct __prelim_ref *ref;
598 INIT_LIST_HEAD(&prefs);
599 INIT_LIST_HEAD(&prefs_delayed);
601 key.objectid = bytenr;
602 key.type = BTRFS_EXTENT_ITEM_KEY;
603 key.offset = (u64)-1;
605 path = btrfs_alloc_path();
606 if (!path)
607 return -ENOMEM;
608 path->search_commit_root = !!search_commit_root;
611 * grab both a lock on the path and a lock on the delayed ref head.
612 * We need both to get a consistent picture of how the refs look
613 * at a specified point in time
615 again:
616 head = NULL;
618 ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
619 if (ret < 0)
620 goto out;
621 BUG_ON(ret == 0);
623 if (trans != BTRFS_BACKREF_SEARCH_COMMIT_ROOT) {
625 * look if there are updates for this ref queued and lock the
626 * head
628 delayed_refs = &trans->transaction->delayed_refs;
629 spin_lock(&delayed_refs->lock);
630 head = btrfs_find_delayed_ref_head(trans, bytenr);
631 if (head) {
632 if (!mutex_trylock(&head->mutex)) {
633 atomic_inc(&head->node.refs);
634 spin_unlock(&delayed_refs->lock);
636 btrfs_release_path(path);
639 * Mutex was contended, block until it's
640 * released and try again
642 mutex_lock(&head->mutex);
643 mutex_unlock(&head->mutex);
644 btrfs_put_delayed_ref(&head->node);
645 goto again;
647 ret = __add_delayed_refs(head, seq, &info_key,
648 &prefs_delayed);
649 if (ret) {
650 spin_unlock(&delayed_refs->lock);
651 goto out;
654 spin_unlock(&delayed_refs->lock);
657 if (path->slots[0]) {
658 struct extent_buffer *leaf;
659 int slot;
661 leaf = path->nodes[0];
662 slot = path->slots[0] - 1;
663 btrfs_item_key_to_cpu(leaf, &key, slot);
664 if (key.objectid == bytenr &&
665 key.type == BTRFS_EXTENT_ITEM_KEY) {
666 ret = __add_inline_refs(fs_info, path, bytenr,
667 &info_key, &info_level, &prefs);
668 if (ret)
669 goto out;
670 ret = __add_keyed_refs(fs_info, path, bytenr, &info_key,
671 info_level, &prefs);
672 if (ret)
673 goto out;
676 btrfs_release_path(path);
679 * when adding the delayed refs above, the info_key might not have
680 * been known yet. Go over the list and replace the missing keys
682 list_for_each_entry(ref, &prefs_delayed, list) {
683 if ((ref->key.offset | ref->key.type | ref->key.objectid) == 0)
684 memcpy(&ref->key, &info_key, sizeof(ref->key));
686 list_splice_init(&prefs_delayed, &prefs);
688 ret = __merge_refs(&prefs, 1);
689 if (ret)
690 goto out;
692 ret = __resolve_indirect_refs(fs_info, search_commit_root, &prefs);
693 if (ret)
694 goto out;
696 ret = __merge_refs(&prefs, 2);
697 if (ret)
698 goto out;
700 while (!list_empty(&prefs)) {
701 ref = list_first_entry(&prefs, struct __prelim_ref, list);
702 list_del(&ref->list);
703 if (ref->count < 0)
704 WARN_ON(1);
705 if (ref->count && ref->root_id && ref->parent == 0) {
706 /* no parent == root of tree */
707 ret = ulist_add(roots, ref->root_id, 0, GFP_NOFS);
708 BUG_ON(ret < 0);
710 if (ref->count && ref->parent) {
711 ret = ulist_add(refs, ref->parent, 0, GFP_NOFS);
712 BUG_ON(ret < 0);
714 kfree(ref);
717 out:
718 if (head)
719 mutex_unlock(&head->mutex);
720 btrfs_free_path(path);
721 while (!list_empty(&prefs)) {
722 ref = list_first_entry(&prefs, struct __prelim_ref, list);
723 list_del(&ref->list);
724 kfree(ref);
726 while (!list_empty(&prefs_delayed)) {
727 ref = list_first_entry(&prefs_delayed, struct __prelim_ref,
728 list);
729 list_del(&ref->list);
730 kfree(ref);
733 return ret;
737 * Finds all leafs with a reference to the specified combination of bytenr and
738 * offset. key_list_head will point to a list of corresponding keys (caller must
739 * free each list element). The leafs will be stored in the leafs ulist, which
740 * must be freed with ulist_free.
742 * returns 0 on success, <0 on error
744 static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
745 struct btrfs_fs_info *fs_info, u64 bytenr,
746 u64 num_bytes, u64 seq, struct ulist **leafs)
748 struct ulist *tmp;
749 int ret;
751 tmp = ulist_alloc(GFP_NOFS);
752 if (!tmp)
753 return -ENOMEM;
754 *leafs = ulist_alloc(GFP_NOFS);
755 if (!*leafs) {
756 ulist_free(tmp);
757 return -ENOMEM;
760 ret = find_parent_nodes(trans, fs_info, bytenr, seq, *leafs, tmp);
761 ulist_free(tmp);
763 if (ret < 0 && ret != -ENOENT) {
764 ulist_free(*leafs);
765 return ret;
768 return 0;
772 * walk all backrefs for a given extent to find all roots that reference this
773 * extent. Walking a backref means finding all extents that reference this
774 * extent and in turn walk the backrefs of those, too. Naturally this is a
775 * recursive process, but here it is implemented in an iterative fashion: We
776 * find all referencing extents for the extent in question and put them on a
777 * list. In turn, we find all referencing extents for those, further appending
778 * to the list. The way we iterate the list allows adding more elements after
779 * the current while iterating. The process stops when we reach the end of the
780 * list. Found roots are added to the roots list.
782 * returns 0 on success, < 0 on error.
784 int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
785 struct btrfs_fs_info *fs_info, u64 bytenr,
786 u64 num_bytes, u64 seq, struct ulist **roots)
788 struct ulist *tmp;
789 struct ulist_node *node = NULL;
790 int ret;
792 tmp = ulist_alloc(GFP_NOFS);
793 if (!tmp)
794 return -ENOMEM;
795 *roots = ulist_alloc(GFP_NOFS);
796 if (!*roots) {
797 ulist_free(tmp);
798 return -ENOMEM;
801 while (1) {
802 ret = find_parent_nodes(trans, fs_info, bytenr, seq,
803 tmp, *roots);
804 if (ret < 0 && ret != -ENOENT) {
805 ulist_free(tmp);
806 ulist_free(*roots);
807 return ret;
809 node = ulist_next(tmp, node);
810 if (!node)
811 break;
812 bytenr = node->val;
815 ulist_free(tmp);
816 return 0;
820 static int __inode_info(u64 inum, u64 ioff, u8 key_type,
821 struct btrfs_root *fs_root, struct btrfs_path *path,
822 struct btrfs_key *found_key)
824 int ret;
825 struct btrfs_key key;
826 struct extent_buffer *eb;
828 key.type = key_type;
829 key.objectid = inum;
830 key.offset = ioff;
832 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
833 if (ret < 0)
834 return ret;
836 eb = path->nodes[0];
837 if (ret && path->slots[0] >= btrfs_header_nritems(eb)) {
838 ret = btrfs_next_leaf(fs_root, path);
839 if (ret)
840 return ret;
841 eb = path->nodes[0];
844 btrfs_item_key_to_cpu(eb, found_key, path->slots[0]);
845 if (found_key->type != key.type || found_key->objectid != key.objectid)
846 return 1;
848 return 0;
852 * this makes the path point to (inum INODE_ITEM ioff)
854 int inode_item_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
855 struct btrfs_path *path)
857 struct btrfs_key key;
858 return __inode_info(inum, ioff, BTRFS_INODE_ITEM_KEY, fs_root, path,
859 &key);
862 static int inode_ref_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
863 struct btrfs_path *path,
864 struct btrfs_key *found_key)
866 return __inode_info(inum, ioff, BTRFS_INODE_REF_KEY, fs_root, path,
867 found_key);
871 * this iterates to turn a btrfs_inode_ref into a full filesystem path. elements
872 * of the path are separated by '/' and the path is guaranteed to be
873 * 0-terminated. the path is only given within the current file system.
874 * Therefore, it never starts with a '/'. the caller is responsible to provide
875 * "size" bytes in "dest". the dest buffer will be filled backwards. finally,
876 * the start point of the resulting string is returned. this pointer is within
877 * dest, normally.
878 * in case the path buffer would overflow, the pointer is decremented further
879 * as if output was written to the buffer, though no more output is actually
880 * generated. that way, the caller can determine how much space would be
881 * required for the path to fit into the buffer. in that case, the returned
882 * value will be smaller than dest. callers must check this!
884 static char *iref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
885 struct btrfs_inode_ref *iref,
886 struct extent_buffer *eb_in, u64 parent,
887 char *dest, u32 size)
889 u32 len;
890 int slot;
891 u64 next_inum;
892 int ret;
893 s64 bytes_left = size - 1;
894 struct extent_buffer *eb = eb_in;
895 struct btrfs_key found_key;
897 if (bytes_left >= 0)
898 dest[bytes_left] = '\0';
900 while (1) {
901 len = btrfs_inode_ref_name_len(eb, iref);
902 bytes_left -= len;
903 if (bytes_left >= 0)
904 read_extent_buffer(eb, dest + bytes_left,
905 (unsigned long)(iref + 1), len);
906 if (eb != eb_in)
907 free_extent_buffer(eb);
908 ret = inode_ref_info(parent, 0, fs_root, path, &found_key);
909 if (ret > 0)
910 ret = -ENOENT;
911 if (ret)
912 break;
913 next_inum = found_key.offset;
915 /* regular exit ahead */
916 if (parent == next_inum)
917 break;
919 slot = path->slots[0];
920 eb = path->nodes[0];
921 /* make sure we can use eb after releasing the path */
922 if (eb != eb_in)
923 atomic_inc(&eb->refs);
924 btrfs_release_path(path);
926 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
927 parent = next_inum;
928 --bytes_left;
929 if (bytes_left >= 0)
930 dest[bytes_left] = '/';
933 btrfs_release_path(path);
935 if (ret)
936 return ERR_PTR(ret);
938 return dest + bytes_left;
942 * this makes the path point to (logical EXTENT_ITEM *)
943 * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for
944 * tree blocks and <0 on error.
946 int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
947 struct btrfs_path *path, struct btrfs_key *found_key)
949 int ret;
950 u64 flags;
951 u32 item_size;
952 struct extent_buffer *eb;
953 struct btrfs_extent_item *ei;
954 struct btrfs_key key;
956 key.type = BTRFS_EXTENT_ITEM_KEY;
957 key.objectid = logical;
958 key.offset = (u64)-1;
960 ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
961 if (ret < 0)
962 return ret;
963 ret = btrfs_previous_item(fs_info->extent_root, path,
964 0, BTRFS_EXTENT_ITEM_KEY);
965 if (ret < 0)
966 return ret;
968 btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]);
969 if (found_key->type != BTRFS_EXTENT_ITEM_KEY ||
970 found_key->objectid > logical ||
971 found_key->objectid + found_key->offset <= logical) {
972 pr_debug("logical %llu is not within any extent\n",
973 (unsigned long long)logical);
974 return -ENOENT;
977 eb = path->nodes[0];
978 item_size = btrfs_item_size_nr(eb, path->slots[0]);
979 BUG_ON(item_size < sizeof(*ei));
981 ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
982 flags = btrfs_extent_flags(eb, ei);
984 pr_debug("logical %llu is at position %llu within the extent (%llu "
985 "EXTENT_ITEM %llu) flags %#llx size %u\n",
986 (unsigned long long)logical,
987 (unsigned long long)(logical - found_key->objectid),
988 (unsigned long long)found_key->objectid,
989 (unsigned long long)found_key->offset,
990 (unsigned long long)flags, item_size);
991 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
992 return BTRFS_EXTENT_FLAG_TREE_BLOCK;
993 if (flags & BTRFS_EXTENT_FLAG_DATA)
994 return BTRFS_EXTENT_FLAG_DATA;
996 return -EIO;
1000 * helper function to iterate extent inline refs. ptr must point to a 0 value
1001 * for the first call and may be modified. it is used to track state.
1002 * if more refs exist, 0 is returned and the next call to
1003 * __get_extent_inline_ref must pass the modified ptr parameter to get the
1004 * next ref. after the last ref was processed, 1 is returned.
1005 * returns <0 on error
1007 static int __get_extent_inline_ref(unsigned long *ptr, struct extent_buffer *eb,
1008 struct btrfs_extent_item *ei, u32 item_size,
1009 struct btrfs_extent_inline_ref **out_eiref,
1010 int *out_type)
1012 unsigned long end;
1013 u64 flags;
1014 struct btrfs_tree_block_info *info;
1016 if (!*ptr) {
1017 /* first call */
1018 flags = btrfs_extent_flags(eb, ei);
1019 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1020 info = (struct btrfs_tree_block_info *)(ei + 1);
1021 *out_eiref =
1022 (struct btrfs_extent_inline_ref *)(info + 1);
1023 } else {
1024 *out_eiref = (struct btrfs_extent_inline_ref *)(ei + 1);
1026 *ptr = (unsigned long)*out_eiref;
1027 if ((void *)*ptr >= (void *)ei + item_size)
1028 return -ENOENT;
1031 end = (unsigned long)ei + item_size;
1032 *out_eiref = (struct btrfs_extent_inline_ref *)*ptr;
1033 *out_type = btrfs_extent_inline_ref_type(eb, *out_eiref);
1035 *ptr += btrfs_extent_inline_ref_size(*out_type);
1036 WARN_ON(*ptr > end);
1037 if (*ptr == end)
1038 return 1; /* last */
1040 return 0;
1044 * reads the tree block backref for an extent. tree level and root are returned
1045 * through out_level and out_root. ptr must point to a 0 value for the first
1046 * call and may be modified (see __get_extent_inline_ref comment).
1047 * returns 0 if data was provided, 1 if there was no more data to provide or
1048 * <0 on error.
1050 int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
1051 struct btrfs_extent_item *ei, u32 item_size,
1052 u64 *out_root, u8 *out_level)
1054 int ret;
1055 int type;
1056 struct btrfs_tree_block_info *info;
1057 struct btrfs_extent_inline_ref *eiref;
1059 if (*ptr == (unsigned long)-1)
1060 return 1;
1062 while (1) {
1063 ret = __get_extent_inline_ref(ptr, eb, ei, item_size,
1064 &eiref, &type);
1065 if (ret < 0)
1066 return ret;
1068 if (type == BTRFS_TREE_BLOCK_REF_KEY ||
1069 type == BTRFS_SHARED_BLOCK_REF_KEY)
1070 break;
1072 if (ret == 1)
1073 return 1;
1076 /* we can treat both ref types equally here */
1077 info = (struct btrfs_tree_block_info *)(ei + 1);
1078 *out_root = btrfs_extent_inline_ref_offset(eb, eiref);
1079 *out_level = btrfs_tree_block_level(eb, info);
1081 if (ret == 1)
1082 *ptr = (unsigned long)-1;
1084 return 0;
1087 static int iterate_leaf_refs(struct btrfs_fs_info *fs_info, u64 logical,
1088 u64 orig_extent_item_objectid,
1089 u64 extent_item_pos, u64 root,
1090 iterate_extent_inodes_t *iterate, void *ctx)
1092 u64 disk_byte;
1093 struct btrfs_key key;
1094 struct btrfs_file_extent_item *fi;
1095 struct extent_buffer *eb;
1096 int slot;
1097 int nritems;
1098 int ret = 0;
1099 int extent_type;
1100 u64 data_offset;
1101 u64 data_len;
1103 eb = read_tree_block(fs_info->tree_root, logical,
1104 fs_info->tree_root->leafsize, 0);
1105 if (!eb)
1106 return -EIO;
1109 * from the shared data ref, we only have the leaf but we need
1110 * the key. thus, we must look into all items and see that we
1111 * find one (some) with a reference to our extent item.
1113 nritems = btrfs_header_nritems(eb);
1114 for (slot = 0; slot < nritems; ++slot) {
1115 btrfs_item_key_to_cpu(eb, &key, slot);
1116 if (key.type != BTRFS_EXTENT_DATA_KEY)
1117 continue;
1118 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
1119 extent_type = btrfs_file_extent_type(eb, fi);
1120 if (extent_type == BTRFS_FILE_EXTENT_INLINE)
1121 continue;
1122 /* don't skip BTRFS_FILE_EXTENT_PREALLOC, we can handle that */
1123 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
1124 if (disk_byte != orig_extent_item_objectid)
1125 continue;
1127 data_offset = btrfs_file_extent_offset(eb, fi);
1128 data_len = btrfs_file_extent_num_bytes(eb, fi);
1130 if (extent_item_pos < data_offset ||
1131 extent_item_pos >= data_offset + data_len)
1132 continue;
1134 pr_debug("ref for %llu resolved, key (%llu EXTEND_DATA %llu), "
1135 "root %llu\n", orig_extent_item_objectid,
1136 key.objectid, key.offset, root);
1137 ret = iterate(key.objectid,
1138 key.offset + (extent_item_pos - data_offset),
1139 root, ctx);
1140 if (ret) {
1141 pr_debug("stopping iteration because ret=%d\n", ret);
1142 break;
1146 free_extent_buffer(eb);
1148 return ret;
1152 * calls iterate() for every inode that references the extent identified by
1153 * the given parameters.
1154 * when the iterator function returns a non-zero value, iteration stops.
1156 int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
1157 u64 extent_item_objectid, u64 extent_item_pos,
1158 int search_commit_root,
1159 iterate_extent_inodes_t *iterate, void *ctx)
1161 int ret;
1162 struct list_head data_refs = LIST_HEAD_INIT(data_refs);
1163 struct list_head shared_refs = LIST_HEAD_INIT(shared_refs);
1164 struct btrfs_trans_handle *trans;
1165 struct ulist *refs = NULL;
1166 struct ulist *roots = NULL;
1167 struct ulist_node *ref_node = NULL;
1168 struct ulist_node *root_node = NULL;
1169 struct seq_list seq_elem;
1170 struct btrfs_delayed_ref_root *delayed_refs = NULL;
1172 pr_debug("resolving all inodes for extent %llu\n",
1173 extent_item_objectid);
1175 if (search_commit_root) {
1176 trans = BTRFS_BACKREF_SEARCH_COMMIT_ROOT;
1177 } else {
1178 trans = btrfs_join_transaction(fs_info->extent_root);
1179 if (IS_ERR(trans))
1180 return PTR_ERR(trans);
1182 delayed_refs = &trans->transaction->delayed_refs;
1183 spin_lock(&delayed_refs->lock);
1184 btrfs_get_delayed_seq(delayed_refs, &seq_elem);
1185 spin_unlock(&delayed_refs->lock);
1188 ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
1189 extent_item_pos, seq_elem.seq,
1190 &refs);
1192 if (ret)
1193 goto out;
1195 while (!ret && (ref_node = ulist_next(refs, ref_node))) {
1196 ret = btrfs_find_all_roots(trans, fs_info, ref_node->val, -1,
1197 seq_elem.seq, &roots);
1198 if (ret)
1199 break;
1200 while (!ret && (root_node = ulist_next(roots, root_node))) {
1201 pr_debug("root %llu references leaf %llu\n",
1202 root_node->val, ref_node->val);
1203 ret = iterate_leaf_refs(fs_info, ref_node->val,
1204 extent_item_objectid,
1205 extent_item_pos, root_node->val,
1206 iterate, ctx);
1210 ulist_free(refs);
1211 ulist_free(roots);
1212 out:
1213 if (!search_commit_root) {
1214 btrfs_put_delayed_seq(delayed_refs, &seq_elem);
1215 btrfs_end_transaction(trans, fs_info->extent_root);
1218 return ret;
1221 int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
1222 struct btrfs_path *path,
1223 iterate_extent_inodes_t *iterate, void *ctx)
1225 int ret;
1226 u64 extent_item_pos;
1227 struct btrfs_key found_key;
1228 int search_commit_root = path->search_commit_root;
1230 ret = extent_from_logical(fs_info, logical, path,
1231 &found_key);
1232 btrfs_release_path(path);
1233 if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1234 ret = -EINVAL;
1235 if (ret < 0)
1236 return ret;
1238 extent_item_pos = logical - found_key.objectid;
1239 ret = iterate_extent_inodes(fs_info, found_key.objectid,
1240 extent_item_pos, search_commit_root,
1241 iterate, ctx);
1243 return ret;
1246 static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
1247 struct btrfs_path *path,
1248 iterate_irefs_t *iterate, void *ctx)
1250 int ret;
1251 int slot;
1252 u32 cur;
1253 u32 len;
1254 u32 name_len;
1255 u64 parent = 0;
1256 int found = 0;
1257 struct extent_buffer *eb;
1258 struct btrfs_item *item;
1259 struct btrfs_inode_ref *iref;
1260 struct btrfs_key found_key;
1262 while (1) {
1263 ret = inode_ref_info(inum, parent ? parent+1 : 0, fs_root, path,
1264 &found_key);
1265 if (ret < 0)
1266 break;
1267 if (ret) {
1268 ret = found ? 0 : -ENOENT;
1269 break;
1271 ++found;
1273 parent = found_key.offset;
1274 slot = path->slots[0];
1275 eb = path->nodes[0];
1276 /* make sure we can use eb after releasing the path */
1277 atomic_inc(&eb->refs);
1278 btrfs_release_path(path);
1280 item = btrfs_item_nr(eb, slot);
1281 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
1283 for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) {
1284 name_len = btrfs_inode_ref_name_len(eb, iref);
1285 /* path must be released before calling iterate()! */
1286 pr_debug("following ref at offset %u for inode %llu in "
1287 "tree %llu\n", cur,
1288 (unsigned long long)found_key.objectid,
1289 (unsigned long long)fs_root->objectid);
1290 ret = iterate(parent, iref, eb, ctx);
1291 if (ret) {
1292 free_extent_buffer(eb);
1293 break;
1295 len = sizeof(*iref) + name_len;
1296 iref = (struct btrfs_inode_ref *)((char *)iref + len);
1298 free_extent_buffer(eb);
1301 btrfs_release_path(path);
1303 return ret;
1307 * returns 0 if the path could be dumped (probably truncated)
1308 * returns <0 in case of an error
1310 static int inode_to_path(u64 inum, struct btrfs_inode_ref *iref,
1311 struct extent_buffer *eb, void *ctx)
1313 struct inode_fs_paths *ipath = ctx;
1314 char *fspath;
1315 char *fspath_min;
1316 int i = ipath->fspath->elem_cnt;
1317 const int s_ptr = sizeof(char *);
1318 u32 bytes_left;
1320 bytes_left = ipath->fspath->bytes_left > s_ptr ?
1321 ipath->fspath->bytes_left - s_ptr : 0;
1323 fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
1324 fspath = iref_to_path(ipath->fs_root, ipath->btrfs_path, iref, eb,
1325 inum, fspath_min, bytes_left);
1326 if (IS_ERR(fspath))
1327 return PTR_ERR(fspath);
1329 if (fspath > fspath_min) {
1330 pr_debug("path resolved: %s\n", fspath);
1331 ipath->fspath->val[i] = (u64)(unsigned long)fspath;
1332 ++ipath->fspath->elem_cnt;
1333 ipath->fspath->bytes_left = fspath - fspath_min;
1334 } else {
1335 pr_debug("missed path, not enough space. missing bytes: %lu, "
1336 "constructed so far: %s\n",
1337 (unsigned long)(fspath_min - fspath), fspath_min);
1338 ++ipath->fspath->elem_missed;
1339 ipath->fspath->bytes_missing += fspath_min - fspath;
1340 ipath->fspath->bytes_left = 0;
1343 return 0;
1347 * this dumps all file system paths to the inode into the ipath struct, provided
1348 * is has been created large enough. each path is zero-terminated and accessed
1349 * from ipath->fspath->val[i].
1350 * when it returns, there are ipath->fspath->elem_cnt number of paths available
1351 * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
1352 * number of missed paths in recored in ipath->fspath->elem_missed, otherwise,
1353 * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
1354 * have been needed to return all paths.
1356 int paths_from_inode(u64 inum, struct inode_fs_paths *ipath)
1358 return iterate_irefs(inum, ipath->fs_root, ipath->btrfs_path,
1359 inode_to_path, ipath);
1363 * allocates space to return multiple file system paths for an inode.
1364 * total_bytes to allocate are passed, note that space usable for actual path
1365 * information will be total_bytes - sizeof(struct inode_fs_paths).
1366 * the returned pointer must be freed with free_ipath() in the end.
1368 struct btrfs_data_container *init_data_container(u32 total_bytes)
1370 struct btrfs_data_container *data;
1371 size_t alloc_bytes;
1373 alloc_bytes = max_t(size_t, total_bytes, sizeof(*data));
1374 data = kmalloc(alloc_bytes, GFP_NOFS);
1375 if (!data)
1376 return ERR_PTR(-ENOMEM);
1378 if (total_bytes >= sizeof(*data)) {
1379 data->bytes_left = total_bytes - sizeof(*data);
1380 data->bytes_missing = 0;
1381 } else {
1382 data->bytes_missing = sizeof(*data) - total_bytes;
1383 data->bytes_left = 0;
1386 data->elem_cnt = 0;
1387 data->elem_missed = 0;
1389 return data;
1393 * allocates space to return multiple file system paths for an inode.
1394 * total_bytes to allocate are passed, note that space usable for actual path
1395 * information will be total_bytes - sizeof(struct inode_fs_paths).
1396 * the returned pointer must be freed with free_ipath() in the end.
1398 struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
1399 struct btrfs_path *path)
1401 struct inode_fs_paths *ifp;
1402 struct btrfs_data_container *fspath;
1404 fspath = init_data_container(total_bytes);
1405 if (IS_ERR(fspath))
1406 return (void *)fspath;
1408 ifp = kmalloc(sizeof(*ifp), GFP_NOFS);
1409 if (!ifp) {
1410 kfree(fspath);
1411 return ERR_PTR(-ENOMEM);
1414 ifp->btrfs_path = path;
1415 ifp->fspath = fspath;
1416 ifp->fs_root = fs_root;
1418 return ifp;
1421 void free_ipath(struct inode_fs_paths *ipath)
1423 kfree(ipath);