1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2011 STRATO. All rights reserved.
7 #include <linux/rbtree.h>
8 #include <trace/events/btrfs.h>
13 #include "transaction.h"
14 #include "delayed-ref.h"
17 #include "tree-mod-log.h"
19 #include "accessors.h"
20 #include "extent-tree.h"
21 #include "relocation.h"
22 #include "tree-checker.h"
24 /* Just arbitrary numbers so we can be sure one of these happened. */
25 #define BACKREF_FOUND_SHARED 6
26 #define BACKREF_FOUND_NOT_SHARED 7
28 struct extent_inode_elem {
32 struct extent_inode_elem *next;
35 static int check_extent_in_eb(struct btrfs_backref_walk_ctx *ctx,
36 const struct btrfs_key *key,
37 const struct extent_buffer *eb,
38 const struct btrfs_file_extent_item *fi,
39 struct extent_inode_elem **eie)
41 const u64 data_len = btrfs_file_extent_num_bytes(eb, fi);
42 u64 offset = key->offset;
43 struct extent_inode_elem *e;
48 if (!ctx->ignore_extent_item_pos &&
49 !btrfs_file_extent_compression(eb, fi) &&
50 !btrfs_file_extent_encryption(eb, fi) &&
51 !btrfs_file_extent_other_encoding(eb, fi)) {
54 data_offset = btrfs_file_extent_offset(eb, fi);
56 if (ctx->extent_item_pos < data_offset ||
57 ctx->extent_item_pos >= data_offset + data_len)
59 offset += ctx->extent_item_pos - data_offset;
62 if (!ctx->indirect_ref_iterator || !ctx->cache_lookup)
65 cached = ctx->cache_lookup(eb->start, ctx->user_ctx, &root_ids,
70 for (int i = 0; i < root_count; i++) {
73 ret = ctx->indirect_ref_iterator(key->objectid, offset,
74 data_len, root_ids[i],
81 e = kmalloc(sizeof(*e), GFP_NOFS);
86 e->inum = key->objectid;
88 e->num_bytes = data_len;
94 static void free_inode_elem_list(struct extent_inode_elem *eie)
96 struct extent_inode_elem *eie_next;
98 for (; eie; eie = eie_next) {
104 static int find_extent_in_eb(struct btrfs_backref_walk_ctx *ctx,
105 const struct extent_buffer *eb,
106 struct extent_inode_elem **eie)
109 struct btrfs_key key;
110 struct btrfs_file_extent_item *fi;
117 * from the shared data ref, we only have the leaf but we need
118 * the key. thus, we must look into all items and see that we
119 * find one (some) with a reference to our extent item.
121 nritems = btrfs_header_nritems(eb);
122 for (slot = 0; slot < nritems; ++slot) {
123 btrfs_item_key_to_cpu(eb, &key, slot);
124 if (key.type != BTRFS_EXTENT_DATA_KEY)
126 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
127 extent_type = btrfs_file_extent_type(eb, fi);
128 if (extent_type == BTRFS_FILE_EXTENT_INLINE)
130 /* don't skip BTRFS_FILE_EXTENT_PREALLOC, we can handle that */
131 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
132 if (disk_byte != ctx->bytenr)
135 ret = check_extent_in_eb(ctx, &key, eb, fi, eie);
136 if (ret == BTRFS_ITERATE_EXTENT_INODES_STOP || ret < 0)
144 struct rb_root_cached root;
148 #define PREFTREE_INIT { .root = RB_ROOT_CACHED, .count = 0 }
151 struct preftree direct; /* BTRFS_SHARED_[DATA|BLOCK]_REF_KEY */
152 struct preftree indirect; /* BTRFS_[TREE_BLOCK|EXTENT_DATA]_REF_KEY */
153 struct preftree indirect_missing_keys;
157 * Checks for a shared extent during backref search.
159 * The share_count tracks prelim_refs (direct and indirect) having a
161 * - incremented when a ref->count transitions to >0
162 * - decremented when a ref->count transitions to <1
165 struct btrfs_backref_share_check_ctx *ctx;
166 struct btrfs_root *root;
171 * Counts number of inodes that refer to an extent (different inodes in
172 * the same root or different roots) that we could find. The sharedness
173 * check typically stops once this counter gets greater than 1, so it
174 * may not reflect the total number of inodes.
178 * The number of times we found our inode refers to the data extent we
179 * are determining the sharedness. In other words, how many file extent
180 * items we could find for our inode that point to our target data
181 * extent. The value we get here after finishing the extent sharedness
182 * check may be smaller than reality, but if it ends up being greater
183 * than 1, then we know for sure the inode has multiple file extent
184 * items that point to our inode, and we can safely assume it's useful
185 * to cache the sharedness check result.
188 bool have_delayed_delete_refs;
191 static inline int extent_is_shared(struct share_check *sc)
193 return (sc && sc->share_count > 1) ? BACKREF_FOUND_SHARED : 0;
196 static struct kmem_cache *btrfs_prelim_ref_cache;
198 int __init btrfs_prelim_ref_init(void)
200 btrfs_prelim_ref_cache = kmem_cache_create("btrfs_prelim_ref",
201 sizeof(struct prelim_ref), 0, 0, NULL);
202 if (!btrfs_prelim_ref_cache)
207 void __cold btrfs_prelim_ref_exit(void)
209 kmem_cache_destroy(btrfs_prelim_ref_cache);
212 static void free_pref(struct prelim_ref *ref)
214 kmem_cache_free(btrfs_prelim_ref_cache, ref);
218 * Return 0 when both refs are for the same block (and can be merged).
219 * A -1 return indicates ref1 is a 'lower' block than ref2, while 1
220 * indicates a 'higher' block.
222 static int prelim_ref_compare(const struct prelim_ref *ref1,
223 const struct prelim_ref *ref2)
225 if (ref1->level < ref2->level)
227 if (ref1->level > ref2->level)
229 if (ref1->root_id < ref2->root_id)
231 if (ref1->root_id > ref2->root_id)
233 if (ref1->key_for_search.type < ref2->key_for_search.type)
235 if (ref1->key_for_search.type > ref2->key_for_search.type)
237 if (ref1->key_for_search.objectid < ref2->key_for_search.objectid)
239 if (ref1->key_for_search.objectid > ref2->key_for_search.objectid)
241 if (ref1->key_for_search.offset < ref2->key_for_search.offset)
243 if (ref1->key_for_search.offset > ref2->key_for_search.offset)
245 if (ref1->parent < ref2->parent)
247 if (ref1->parent > ref2->parent)
253 static void update_share_count(struct share_check *sc, int oldcount,
254 int newcount, const struct prelim_ref *newref)
256 if ((!sc) || (oldcount == 0 && newcount < 1))
259 if (oldcount > 0 && newcount < 1)
261 else if (oldcount < 1 && newcount > 0)
264 if (newref->root_id == btrfs_root_id(sc->root) &&
265 newref->wanted_disk_byte == sc->data_bytenr &&
266 newref->key_for_search.objectid == sc->inum)
267 sc->self_ref_count += newref->count;
271 * Add @newref to the @root rbtree, merging identical refs.
273 * Callers should assume that newref has been freed after calling.
275 static void prelim_ref_insert(const struct btrfs_fs_info *fs_info,
276 struct preftree *preftree,
277 struct prelim_ref *newref,
278 struct share_check *sc)
280 struct rb_root_cached *root;
282 struct rb_node *parent = NULL;
283 struct prelim_ref *ref;
285 bool leftmost = true;
287 root = &preftree->root;
288 p = &root->rb_root.rb_node;
292 ref = rb_entry(parent, struct prelim_ref, rbnode);
293 result = prelim_ref_compare(ref, newref);
296 } else if (result > 0) {
300 /* Identical refs, merge them and free @newref */
301 struct extent_inode_elem *eie = ref->inode_list;
303 while (eie && eie->next)
307 ref->inode_list = newref->inode_list;
309 eie->next = newref->inode_list;
310 trace_btrfs_prelim_ref_merge(fs_info, ref, newref,
313 * A delayed ref can have newref->count < 0.
314 * The ref->count is updated to follow any
315 * BTRFS_[ADD|DROP]_DELAYED_REF actions.
317 update_share_count(sc, ref->count,
318 ref->count + newref->count, newref);
319 ref->count += newref->count;
325 update_share_count(sc, 0, newref->count, newref);
327 trace_btrfs_prelim_ref_insert(fs_info, newref, NULL, preftree->count);
328 rb_link_node(&newref->rbnode, parent, p);
329 rb_insert_color_cached(&newref->rbnode, root, leftmost);
333 * Release the entire tree. We don't care about internal consistency so
334 * just free everything and then reset the tree root.
336 static void prelim_release(struct preftree *preftree)
338 struct prelim_ref *ref, *next_ref;
340 rbtree_postorder_for_each_entry_safe(ref, next_ref,
341 &preftree->root.rb_root, rbnode) {
342 free_inode_elem_list(ref->inode_list);
346 preftree->root = RB_ROOT_CACHED;
351 * the rules for all callers of this function are:
352 * - obtaining the parent is the goal
353 * - if you add a key, you must know that it is a correct key
354 * - if you cannot add the parent or a correct key, then we will look into the
355 * block later to set a correct key
359 * backref type | shared | indirect | shared | indirect
360 * information | tree | tree | data | data
361 * --------------------+--------+----------+--------+----------
362 * parent logical | y | - | - | -
363 * key to resolve | - | y | y | y
364 * tree block logical | - | - | - | -
365 * root for resolving | y | y | y | y
367 * - column 1: we've the parent -> done
368 * - column 2, 3, 4: we use the key to find the parent
370 * on disk refs (inline or keyed)
371 * ==============================
372 * backref type | shared | indirect | shared | indirect
373 * information | tree | tree | data | data
374 * --------------------+--------+----------+--------+----------
375 * parent logical | y | - | y | -
376 * key to resolve | - | - | - | y
377 * tree block logical | y | y | y | y
378 * root for resolving | - | y | y | y
380 * - column 1, 3: we've the parent -> done
381 * - column 2: we take the first key from the block to find the parent
382 * (see add_missing_keys)
383 * - column 4: we use the key to find the parent
385 * additional information that's available but not required to find the parent
386 * block might help in merging entries to gain some speed.
388 static int add_prelim_ref(const struct btrfs_fs_info *fs_info,
389 struct preftree *preftree, u64 root_id,
390 const struct btrfs_key *key, int level, u64 parent,
391 u64 wanted_disk_byte, int count,
392 struct share_check *sc, gfp_t gfp_mask)
394 struct prelim_ref *ref;
396 if (root_id == BTRFS_DATA_RELOC_TREE_OBJECTID)
399 ref = kmem_cache_alloc(btrfs_prelim_ref_cache, gfp_mask);
403 ref->root_id = root_id;
405 ref->key_for_search = *key;
407 memset(&ref->key_for_search, 0, sizeof(ref->key_for_search));
409 ref->inode_list = NULL;
412 ref->parent = parent;
413 ref->wanted_disk_byte = wanted_disk_byte;
414 prelim_ref_insert(fs_info, preftree, ref, sc);
415 return extent_is_shared(sc);
418 /* direct refs use root == 0, key == NULL */
419 static int add_direct_ref(const struct btrfs_fs_info *fs_info,
420 struct preftrees *preftrees, int level, u64 parent,
421 u64 wanted_disk_byte, int count,
422 struct share_check *sc, gfp_t gfp_mask)
424 return add_prelim_ref(fs_info, &preftrees->direct, 0, NULL, level,
425 parent, wanted_disk_byte, count, sc, gfp_mask);
428 /* indirect refs use parent == 0 */
429 static int add_indirect_ref(const struct btrfs_fs_info *fs_info,
430 struct preftrees *preftrees, u64 root_id,
431 const struct btrfs_key *key, int level,
432 u64 wanted_disk_byte, int count,
433 struct share_check *sc, gfp_t gfp_mask)
435 struct preftree *tree = &preftrees->indirect;
438 tree = &preftrees->indirect_missing_keys;
439 return add_prelim_ref(fs_info, tree, root_id, key, level, 0,
440 wanted_disk_byte, count, sc, gfp_mask);
443 static int is_shared_data_backref(struct preftrees *preftrees, u64 bytenr)
445 struct rb_node **p = &preftrees->direct.root.rb_root.rb_node;
446 struct rb_node *parent = NULL;
447 struct prelim_ref *ref = NULL;
448 struct prelim_ref target = {};
451 target.parent = bytenr;
455 ref = rb_entry(parent, struct prelim_ref, rbnode);
456 result = prelim_ref_compare(ref, &target);
468 static int add_all_parents(struct btrfs_backref_walk_ctx *ctx,
469 struct btrfs_root *root, struct btrfs_path *path,
470 struct ulist *parents,
471 struct preftrees *preftrees, struct prelim_ref *ref,
476 struct extent_buffer *eb;
477 struct btrfs_key key;
478 struct btrfs_key *key_for_search = &ref->key_for_search;
479 struct btrfs_file_extent_item *fi;
480 struct extent_inode_elem *eie = NULL, *old = NULL;
482 u64 wanted_disk_byte = ref->wanted_disk_byte;
488 eb = path->nodes[level];
489 ret = ulist_add(parents, eb->start, 0, GFP_NOFS);
496 * 1. We normally enter this function with the path already pointing to
497 * the first item to check. But sometimes, we may enter it with
499 * 2. We are searching for normal backref but bytenr of this leaf
500 * matches shared data backref
501 * 3. The leaf owner is not equal to the root we are searching
503 * For these cases, go to the next leaf before we continue.
506 if (path->slots[0] >= btrfs_header_nritems(eb) ||
507 is_shared_data_backref(preftrees, eb->start) ||
508 ref->root_id != btrfs_header_owner(eb)) {
509 if (ctx->time_seq == BTRFS_SEQ_LAST)
510 ret = btrfs_next_leaf(root, path);
512 ret = btrfs_next_old_leaf(root, path, ctx->time_seq);
515 while (!ret && count < ref->count) {
517 slot = path->slots[0];
519 btrfs_item_key_to_cpu(eb, &key, slot);
521 if (key.objectid != key_for_search->objectid ||
522 key.type != BTRFS_EXTENT_DATA_KEY)
526 * We are searching for normal backref but bytenr of this leaf
527 * matches shared data backref, OR
528 * the leaf owner is not equal to the root we are searching for
531 (is_shared_data_backref(preftrees, eb->start) ||
532 ref->root_id != btrfs_header_owner(eb))) {
533 if (ctx->time_seq == BTRFS_SEQ_LAST)
534 ret = btrfs_next_leaf(root, path);
536 ret = btrfs_next_old_leaf(root, path, ctx->time_seq);
539 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
540 type = btrfs_file_extent_type(eb, fi);
541 if (type == BTRFS_FILE_EXTENT_INLINE)
543 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
544 data_offset = btrfs_file_extent_offset(eb, fi);
546 if (disk_byte == wanted_disk_byte) {
549 if (ref->key_for_search.offset == key.offset - data_offset)
553 if (!ctx->skip_inode_ref_list) {
554 ret = check_extent_in_eb(ctx, &key, eb, fi, &eie);
555 if (ret == BTRFS_ITERATE_EXTENT_INODES_STOP ||
561 ret = ulist_add_merge_ptr(parents, eb->start,
562 eie, (void **)&old, GFP_NOFS);
565 if (!ret && !ctx->skip_inode_ref_list) {
573 if (ctx->time_seq == BTRFS_SEQ_LAST)
574 ret = btrfs_next_item(root, path);
576 ret = btrfs_next_old_item(root, path, ctx->time_seq);
579 if (ret == BTRFS_ITERATE_EXTENT_INODES_STOP || ret < 0)
580 free_inode_elem_list(eie);
588 * resolve an indirect backref in the form (root_id, key, level)
589 * to a logical address
591 static int resolve_indirect_ref(struct btrfs_backref_walk_ctx *ctx,
592 struct btrfs_path *path,
593 struct preftrees *preftrees,
594 struct prelim_ref *ref, struct ulist *parents)
596 struct btrfs_root *root;
597 struct extent_buffer *eb;
600 int level = ref->level;
601 struct btrfs_key search_key = ref->key_for_search;
604 * If we're search_commit_root we could possibly be holding locks on
605 * other tree nodes. This happens when qgroups does backref walks when
606 * adding new delayed refs. To deal with this we need to look in cache
607 * for the root, and if we don't find it then we need to search the
608 * tree_root's commit root, thus the btrfs_get_fs_root_commit_root usage
611 if (path->search_commit_root)
612 root = btrfs_get_fs_root_commit_root(ctx->fs_info, path, ref->root_id);
614 root = btrfs_get_fs_root(ctx->fs_info, ref->root_id, false);
620 if (!path->search_commit_root &&
621 test_bit(BTRFS_ROOT_DELETING, &root->state)) {
626 if (btrfs_is_testing(ctx->fs_info)) {
631 if (path->search_commit_root)
632 root_level = btrfs_header_level(root->commit_root);
633 else if (ctx->time_seq == BTRFS_SEQ_LAST)
634 root_level = btrfs_header_level(root->node);
636 root_level = btrfs_old_root_level(root, ctx->time_seq);
638 if (root_level + 1 == level)
642 * We can often find data backrefs with an offset that is too large
643 * (>= LLONG_MAX, maximum allowed file offset) due to underflows when
644 * subtracting a file's offset with the data offset of its
645 * corresponding extent data item. This can happen for example in the
648 * So if we detect such case we set the search key's offset to zero to
649 * make sure we will find the matching file extent item at
650 * add_all_parents(), otherwise we will miss it because the offset
651 * taken form the backref is much larger then the offset of the file
652 * extent item. This can make us scan a very large number of file
653 * extent items, but at least it will not make us miss any.
655 * This is an ugly workaround for a behaviour that should have never
656 * existed, but it does and a fix for the clone ioctl would touch a lot
657 * of places, cause backwards incompatibility and would not fix the
658 * problem for extents cloned with older kernels.
660 if (search_key.type == BTRFS_EXTENT_DATA_KEY &&
661 search_key.offset >= LLONG_MAX)
662 search_key.offset = 0;
663 path->lowest_level = level;
664 if (ctx->time_seq == BTRFS_SEQ_LAST)
665 ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
667 ret = btrfs_search_old_slot(root, &search_key, path, ctx->time_seq);
669 btrfs_debug(ctx->fs_info,
670 "search slot in root %llu (level %d, ref count %d) returned %d for key (%llu %u %llu)",
671 ref->root_id, level, ref->count, ret,
672 ref->key_for_search.objectid, ref->key_for_search.type,
673 ref->key_for_search.offset);
677 eb = path->nodes[level];
679 if (WARN_ON(!level)) {
684 eb = path->nodes[level];
687 ret = add_all_parents(ctx, root, path, parents, preftrees, ref, level);
689 btrfs_put_root(root);
691 path->lowest_level = 0;
692 btrfs_release_path(path);
696 static struct extent_inode_elem *
697 unode_aux_to_inode_list(struct ulist_node *node)
701 return (struct extent_inode_elem *)(uintptr_t)node->aux;
704 static void free_leaf_list(struct ulist *ulist)
706 struct ulist_node *node;
707 struct ulist_iterator uiter;
709 ULIST_ITER_INIT(&uiter);
710 while ((node = ulist_next(ulist, &uiter)))
711 free_inode_elem_list(unode_aux_to_inode_list(node));
717 * We maintain three separate rbtrees: one for direct refs, one for
718 * indirect refs which have a key, and one for indirect refs which do not
719 * have a key. Each tree does merge on insertion.
721 * Once all of the references are located, we iterate over the tree of
722 * indirect refs with missing keys. An appropriate key is located and
723 * the ref is moved onto the tree for indirect refs. After all missing
724 * keys are thus located, we iterate over the indirect ref tree, resolve
725 * each reference, and then insert the resolved reference onto the
726 * direct tree (merging there too).
728 * New backrefs (i.e., for parent nodes) are added to the appropriate
729 * rbtree as they are encountered. The new backrefs are subsequently
732 static int resolve_indirect_refs(struct btrfs_backref_walk_ctx *ctx,
733 struct btrfs_path *path,
734 struct preftrees *preftrees,
735 struct share_check *sc)
739 struct ulist *parents;
740 struct ulist_node *node;
741 struct ulist_iterator uiter;
742 struct rb_node *rnode;
744 parents = ulist_alloc(GFP_NOFS);
749 * We could trade memory usage for performance here by iterating
750 * the tree, allocating new refs for each insertion, and then
751 * freeing the entire indirect tree when we're done. In some test
752 * cases, the tree can grow quite large (~200k objects).
754 while ((rnode = rb_first_cached(&preftrees->indirect.root))) {
755 struct prelim_ref *ref;
757 ref = rb_entry(rnode, struct prelim_ref, rbnode);
758 if (WARN(ref->parent,
759 "BUG: direct ref found in indirect tree")) {
764 rb_erase_cached(&ref->rbnode, &preftrees->indirect.root);
765 preftrees->indirect.count--;
767 if (ref->count == 0) {
772 if (sc && ref->root_id != btrfs_root_id(sc->root)) {
774 ret = BACKREF_FOUND_SHARED;
777 err = resolve_indirect_ref(ctx, path, preftrees, ref, parents);
779 * we can only tolerate ENOENT,otherwise,we should catch error
780 * and return directly.
782 if (err == -ENOENT) {
783 prelim_ref_insert(ctx->fs_info, &preftrees->direct, ref,
792 /* we put the first parent into the ref at hand */
793 ULIST_ITER_INIT(&uiter);
794 node = ulist_next(parents, &uiter);
795 ref->parent = node ? node->val : 0;
796 ref->inode_list = unode_aux_to_inode_list(node);
798 /* Add a prelim_ref(s) for any other parent(s). */
799 while ((node = ulist_next(parents, &uiter))) {
800 struct prelim_ref *new_ref;
802 new_ref = kmem_cache_alloc(btrfs_prelim_ref_cache,
809 memcpy(new_ref, ref, sizeof(*ref));
810 new_ref->parent = node->val;
811 new_ref->inode_list = unode_aux_to_inode_list(node);
812 prelim_ref_insert(ctx->fs_info, &preftrees->direct,
817 * Now it's a direct ref, put it in the direct tree. We must
818 * do this last because the ref could be merged/freed here.
820 prelim_ref_insert(ctx->fs_info, &preftrees->direct, ref, NULL);
822 ulist_reinit(parents);
827 * We may have inode lists attached to refs in the parents ulist, so we
828 * must free them before freeing the ulist and its refs.
830 free_leaf_list(parents);
835 * read tree blocks and add keys where required.
837 static int add_missing_keys(struct btrfs_fs_info *fs_info,
838 struct preftrees *preftrees, bool lock)
840 struct prelim_ref *ref;
841 struct extent_buffer *eb;
842 struct preftree *tree = &preftrees->indirect_missing_keys;
843 struct rb_node *node;
845 while ((node = rb_first_cached(&tree->root))) {
846 struct btrfs_tree_parent_check check = { 0 };
848 ref = rb_entry(node, struct prelim_ref, rbnode);
849 rb_erase_cached(node, &tree->root);
851 BUG_ON(ref->parent); /* should not be a direct ref */
852 BUG_ON(ref->key_for_search.type);
853 BUG_ON(!ref->wanted_disk_byte);
855 check.level = ref->level - 1;
856 check.owner_root = ref->root_id;
858 eb = read_tree_block(fs_info, ref->wanted_disk_byte, &check);
863 if (!extent_buffer_uptodate(eb)) {
865 free_extent_buffer(eb);
870 btrfs_tree_read_lock(eb);
871 if (btrfs_header_level(eb) == 0)
872 btrfs_item_key_to_cpu(eb, &ref->key_for_search, 0);
874 btrfs_node_key_to_cpu(eb, &ref->key_for_search, 0);
876 btrfs_tree_read_unlock(eb);
877 free_extent_buffer(eb);
878 prelim_ref_insert(fs_info, &preftrees->indirect, ref, NULL);
885 * add all currently queued delayed refs from this head whose seq nr is
886 * smaller or equal that seq to the list
888 static int add_delayed_refs(const struct btrfs_fs_info *fs_info,
889 struct btrfs_delayed_ref_head *head, u64 seq,
890 struct preftrees *preftrees, struct share_check *sc)
892 struct btrfs_delayed_ref_node *node;
893 struct btrfs_key key;
898 spin_lock(&head->lock);
899 for (n = rb_first_cached(&head->ref_tree); n; n = rb_next(n)) {
900 node = rb_entry(n, struct btrfs_delayed_ref_node,
905 switch (node->action) {
906 case BTRFS_ADD_DELAYED_EXTENT:
907 case BTRFS_UPDATE_DELAYED_HEAD:
910 case BTRFS_ADD_DELAYED_REF:
911 count = node->ref_mod;
913 case BTRFS_DROP_DELAYED_REF:
914 count = node->ref_mod * -1;
919 switch (node->type) {
920 case BTRFS_TREE_BLOCK_REF_KEY: {
921 /* NORMAL INDIRECT METADATA backref */
922 struct btrfs_key *key_ptr = NULL;
923 /* The owner of a tree block ref is the level. */
924 int level = btrfs_delayed_ref_owner(node);
926 if (head->extent_op && head->extent_op->update_key) {
927 btrfs_disk_key_to_cpu(&key, &head->extent_op->key);
931 ret = add_indirect_ref(fs_info, preftrees, node->ref_root,
932 key_ptr, level + 1, node->bytenr,
933 count, sc, GFP_ATOMIC);
936 case BTRFS_SHARED_BLOCK_REF_KEY: {
938 * SHARED DIRECT METADATA backref
940 * The owner of a tree block ref is the level.
942 int level = btrfs_delayed_ref_owner(node);
944 ret = add_direct_ref(fs_info, preftrees, level + 1,
945 node->parent, node->bytenr, count,
949 case BTRFS_EXTENT_DATA_REF_KEY: {
950 /* NORMAL INDIRECT DATA backref */
951 key.objectid = btrfs_delayed_ref_owner(node);
952 key.type = BTRFS_EXTENT_DATA_KEY;
953 key.offset = btrfs_delayed_ref_offset(node);
956 * If we have a share check context and a reference for
957 * another inode, we can't exit immediately. This is
958 * because even if this is a BTRFS_ADD_DELAYED_REF
959 * reference we may find next a BTRFS_DROP_DELAYED_REF
960 * which cancels out this ADD reference.
962 * If this is a DROP reference and there was no previous
963 * ADD reference, then we need to signal that when we
964 * process references from the extent tree (through
965 * add_inline_refs() and add_keyed_refs()), we should
966 * not exit early if we find a reference for another
967 * inode, because one of the delayed DROP references
968 * may cancel that reference in the extent tree.
971 sc->have_delayed_delete_refs = true;
973 ret = add_indirect_ref(fs_info, preftrees, node->ref_root,
974 &key, 0, node->bytenr, count, sc,
978 case BTRFS_SHARED_DATA_REF_KEY: {
979 /* SHARED DIRECT FULL backref */
980 ret = add_direct_ref(fs_info, preftrees, 0, node->parent,
981 node->bytenr, count, sc,
989 * We must ignore BACKREF_FOUND_SHARED until all delayed
990 * refs have been checked.
992 if (ret && (ret != BACKREF_FOUND_SHARED))
996 ret = extent_is_shared(sc);
998 spin_unlock(&head->lock);
1003 * add all inline backrefs for bytenr to the list
1005 * Returns 0 on success, <0 on error, or BACKREF_FOUND_SHARED.
1007 static int add_inline_refs(struct btrfs_backref_walk_ctx *ctx,
1008 struct btrfs_path *path,
1009 int *info_level, struct preftrees *preftrees,
1010 struct share_check *sc)
1014 struct extent_buffer *leaf;
1015 struct btrfs_key key;
1016 struct btrfs_key found_key;
1019 struct btrfs_extent_item *ei;
1024 * enumerate all inline refs
1026 leaf = path->nodes[0];
1027 slot = path->slots[0];
1029 item_size = btrfs_item_size(leaf, slot);
1030 ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
1032 if (ctx->check_extent_item) {
1033 ret = ctx->check_extent_item(ctx->bytenr, ei, leaf, ctx->user_ctx);
1038 flags = btrfs_extent_flags(leaf, ei);
1039 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1041 ptr = (unsigned long)(ei + 1);
1042 end = (unsigned long)ei + item_size;
1044 if (found_key.type == BTRFS_EXTENT_ITEM_KEY &&
1045 flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1046 struct btrfs_tree_block_info *info;
1048 info = (struct btrfs_tree_block_info *)ptr;
1049 *info_level = btrfs_tree_block_level(leaf, info);
1050 ptr += sizeof(struct btrfs_tree_block_info);
1052 } else if (found_key.type == BTRFS_METADATA_ITEM_KEY) {
1053 *info_level = found_key.offset;
1055 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
1059 struct btrfs_extent_inline_ref *iref;
1063 iref = (struct btrfs_extent_inline_ref *)ptr;
1064 type = btrfs_get_extent_inline_ref_type(leaf, iref,
1065 BTRFS_REF_TYPE_ANY);
1066 if (type == BTRFS_REF_TYPE_INVALID)
1069 offset = btrfs_extent_inline_ref_offset(leaf, iref);
1072 case BTRFS_SHARED_BLOCK_REF_KEY:
1073 ret = add_direct_ref(ctx->fs_info, preftrees,
1074 *info_level + 1, offset,
1075 ctx->bytenr, 1, NULL, GFP_NOFS);
1077 case BTRFS_SHARED_DATA_REF_KEY: {
1078 struct btrfs_shared_data_ref *sdref;
1081 sdref = (struct btrfs_shared_data_ref *)(iref + 1);
1082 count = btrfs_shared_data_ref_count(leaf, sdref);
1084 ret = add_direct_ref(ctx->fs_info, preftrees, 0, offset,
1085 ctx->bytenr, count, sc, GFP_NOFS);
1088 case BTRFS_TREE_BLOCK_REF_KEY:
1089 ret = add_indirect_ref(ctx->fs_info, preftrees, offset,
1090 NULL, *info_level + 1,
1091 ctx->bytenr, 1, NULL, GFP_NOFS);
1093 case BTRFS_EXTENT_DATA_REF_KEY: {
1094 struct btrfs_extent_data_ref *dref;
1098 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1099 count = btrfs_extent_data_ref_count(leaf, dref);
1100 key.objectid = btrfs_extent_data_ref_objectid(leaf,
1102 key.type = BTRFS_EXTENT_DATA_KEY;
1103 key.offset = btrfs_extent_data_ref_offset(leaf, dref);
1105 if (sc && key.objectid != sc->inum &&
1106 !sc->have_delayed_delete_refs) {
1107 ret = BACKREF_FOUND_SHARED;
1111 root = btrfs_extent_data_ref_root(leaf, dref);
1113 if (!ctx->skip_data_ref ||
1114 !ctx->skip_data_ref(root, key.objectid, key.offset,
1116 ret = add_indirect_ref(ctx->fs_info, preftrees,
1117 root, &key, 0, ctx->bytenr,
1118 count, sc, GFP_NOFS);
1121 case BTRFS_EXTENT_OWNER_REF_KEY:
1122 ASSERT(btrfs_fs_incompat(ctx->fs_info, SIMPLE_QUOTA));
1129 ptr += btrfs_extent_inline_ref_size(type);
1136 * add all non-inline backrefs for bytenr to the list
1138 * Returns 0 on success, <0 on error, or BACKREF_FOUND_SHARED.
1140 static int add_keyed_refs(struct btrfs_backref_walk_ctx *ctx,
1141 struct btrfs_root *extent_root,
1142 struct btrfs_path *path,
1143 int info_level, struct preftrees *preftrees,
1144 struct share_check *sc)
1146 struct btrfs_fs_info *fs_info = extent_root->fs_info;
1149 struct extent_buffer *leaf;
1150 struct btrfs_key key;
1153 ret = btrfs_next_item(extent_root, path);
1161 slot = path->slots[0];
1162 leaf = path->nodes[0];
1163 btrfs_item_key_to_cpu(leaf, &key, slot);
1165 if (key.objectid != ctx->bytenr)
1167 if (key.type < BTRFS_TREE_BLOCK_REF_KEY)
1169 if (key.type > BTRFS_SHARED_DATA_REF_KEY)
1173 case BTRFS_SHARED_BLOCK_REF_KEY:
1174 /* SHARED DIRECT METADATA backref */
1175 ret = add_direct_ref(fs_info, preftrees,
1176 info_level + 1, key.offset,
1177 ctx->bytenr, 1, NULL, GFP_NOFS);
1179 case BTRFS_SHARED_DATA_REF_KEY: {
1180 /* SHARED DIRECT FULL backref */
1181 struct btrfs_shared_data_ref *sdref;
1184 sdref = btrfs_item_ptr(leaf, slot,
1185 struct btrfs_shared_data_ref);
1186 count = btrfs_shared_data_ref_count(leaf, sdref);
1187 ret = add_direct_ref(fs_info, preftrees, 0,
1188 key.offset, ctx->bytenr, count,
1192 case BTRFS_TREE_BLOCK_REF_KEY:
1193 /* NORMAL INDIRECT METADATA backref */
1194 ret = add_indirect_ref(fs_info, preftrees, key.offset,
1195 NULL, info_level + 1, ctx->bytenr,
1198 case BTRFS_EXTENT_DATA_REF_KEY: {
1199 /* NORMAL INDIRECT DATA backref */
1200 struct btrfs_extent_data_ref *dref;
1204 dref = btrfs_item_ptr(leaf, slot,
1205 struct btrfs_extent_data_ref);
1206 count = btrfs_extent_data_ref_count(leaf, dref);
1207 key.objectid = btrfs_extent_data_ref_objectid(leaf,
1209 key.type = BTRFS_EXTENT_DATA_KEY;
1210 key.offset = btrfs_extent_data_ref_offset(leaf, dref);
1212 if (sc && key.objectid != sc->inum &&
1213 !sc->have_delayed_delete_refs) {
1214 ret = BACKREF_FOUND_SHARED;
1218 root = btrfs_extent_data_ref_root(leaf, dref);
1220 if (!ctx->skip_data_ref ||
1221 !ctx->skip_data_ref(root, key.objectid, key.offset,
1223 ret = add_indirect_ref(fs_info, preftrees, root,
1224 &key, 0, ctx->bytenr,
1225 count, sc, GFP_NOFS);
1240 * The caller has joined a transaction or is holding a read lock on the
1241 * fs_info->commit_root_sem semaphore, so no need to worry about the root's last
1242 * snapshot field changing while updating or checking the cache.
1244 static bool lookup_backref_shared_cache(struct btrfs_backref_share_check_ctx *ctx,
1245 struct btrfs_root *root,
1246 u64 bytenr, int level, bool *is_shared)
1248 const struct btrfs_fs_info *fs_info = root->fs_info;
1249 struct btrfs_backref_shared_cache_entry *entry;
1251 if (!current->journal_info)
1252 lockdep_assert_held(&fs_info->commit_root_sem);
1254 if (!ctx->use_path_cache)
1257 if (WARN_ON_ONCE(level >= BTRFS_MAX_LEVEL))
1261 * Level -1 is used for the data extent, which is not reliable to cache
1262 * because its reference count can increase or decrease without us
1263 * realizing. We cache results only for extent buffers that lead from
1264 * the root node down to the leaf with the file extent item.
1268 entry = &ctx->path_cache_entries[level];
1270 /* Unused cache entry or being used for some other extent buffer. */
1271 if (entry->bytenr != bytenr)
1275 * We cached a false result, but the last snapshot generation of the
1276 * root changed, so we now have a snapshot. Don't trust the result.
1278 if (!entry->is_shared &&
1279 entry->gen != btrfs_root_last_snapshot(&root->root_item))
1283 * If we cached a true result and the last generation used for dropping
1284 * a root changed, we can not trust the result, because the dropped root
1285 * could be a snapshot sharing this extent buffer.
1287 if (entry->is_shared &&
1288 entry->gen != btrfs_get_last_root_drop_gen(fs_info))
1291 *is_shared = entry->is_shared;
1293 * If the node at this level is shared, than all nodes below are also
1294 * shared. Currently some of the nodes below may be marked as not shared
1295 * because we have just switched from one leaf to another, and switched
1296 * also other nodes above the leaf and below the current level, so mark
1300 for (int i = 0; i < level; i++) {
1301 ctx->path_cache_entries[i].is_shared = true;
1302 ctx->path_cache_entries[i].gen = entry->gen;
1310 * The caller has joined a transaction or is holding a read lock on the
1311 * fs_info->commit_root_sem semaphore, so no need to worry about the root's last
1312 * snapshot field changing while updating or checking the cache.
1314 static void store_backref_shared_cache(struct btrfs_backref_share_check_ctx *ctx,
1315 struct btrfs_root *root,
1316 u64 bytenr, int level, bool is_shared)
1318 const struct btrfs_fs_info *fs_info = root->fs_info;
1319 struct btrfs_backref_shared_cache_entry *entry;
1322 if (!current->journal_info)
1323 lockdep_assert_held(&fs_info->commit_root_sem);
1325 if (!ctx->use_path_cache)
1328 if (WARN_ON_ONCE(level >= BTRFS_MAX_LEVEL))
1332 * Level -1 is used for the data extent, which is not reliable to cache
1333 * because its reference count can increase or decrease without us
1334 * realizing. We cache results only for extent buffers that lead from
1335 * the root node down to the leaf with the file extent item.
1340 gen = btrfs_get_last_root_drop_gen(fs_info);
1342 gen = btrfs_root_last_snapshot(&root->root_item);
1344 entry = &ctx->path_cache_entries[level];
1345 entry->bytenr = bytenr;
1346 entry->is_shared = is_shared;
1350 * If we found an extent buffer is shared, set the cache result for all
1351 * extent buffers below it to true. As nodes in the path are COWed,
1352 * their sharedness is moved to their children, and if a leaf is COWed,
1353 * then the sharedness of a data extent becomes direct, the refcount of
1354 * data extent is increased in the extent item at the extent tree.
1357 for (int i = 0; i < level; i++) {
1358 entry = &ctx->path_cache_entries[i];
1359 entry->is_shared = is_shared;
1366 * this adds all existing backrefs (inline backrefs, backrefs and delayed
1367 * refs) for the given bytenr to the refs list, merges duplicates and resolves
1368 * indirect refs to their parent bytenr.
1369 * When roots are found, they're added to the roots list
1371 * @ctx: Backref walking context object, must be not NULL.
1372 * @sc: If !NULL, then immediately return BACKREF_FOUND_SHARED when a
1373 * shared extent is detected.
1375 * Otherwise this returns 0 for success and <0 for an error.
1377 * FIXME some caching might speed things up
1379 static int find_parent_nodes(struct btrfs_backref_walk_ctx *ctx,
1380 struct share_check *sc)
1382 struct btrfs_root *root = btrfs_extent_root(ctx->fs_info, ctx->bytenr);
1383 struct btrfs_key key;
1384 struct btrfs_path *path;
1385 struct btrfs_delayed_ref_root *delayed_refs = NULL;
1386 struct btrfs_delayed_ref_head *head;
1389 struct prelim_ref *ref;
1390 struct rb_node *node;
1391 struct extent_inode_elem *eie = NULL;
1392 struct preftrees preftrees = {
1393 .direct = PREFTREE_INIT,
1394 .indirect = PREFTREE_INIT,
1395 .indirect_missing_keys = PREFTREE_INIT
1398 /* Roots ulist is not needed when using a sharedness check context. */
1400 ASSERT(ctx->roots == NULL);
1402 key.objectid = ctx->bytenr;
1403 key.offset = (u64)-1;
1404 if (btrfs_fs_incompat(ctx->fs_info, SKINNY_METADATA))
1405 key.type = BTRFS_METADATA_ITEM_KEY;
1407 key.type = BTRFS_EXTENT_ITEM_KEY;
1409 path = btrfs_alloc_path();
1413 path->search_commit_root = 1;
1414 path->skip_locking = 1;
1417 if (ctx->time_seq == BTRFS_SEQ_LAST)
1418 path->skip_locking = 1;
1423 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1428 * Key with offset -1 found, there would have to exist an extent
1429 * item with such offset, but this is out of the valid range.
1435 if (ctx->trans && likely(ctx->trans->type != __TRANS_DUMMY) &&
1436 ctx->time_seq != BTRFS_SEQ_LAST) {
1438 * We have a specific time_seq we care about and trans which
1439 * means we have the path lock, we need to grab the ref head and
1440 * lock it so we have a consistent view of the refs at the given
1443 delayed_refs = &ctx->trans->transaction->delayed_refs;
1444 spin_lock(&delayed_refs->lock);
1445 head = btrfs_find_delayed_ref_head(ctx->fs_info, delayed_refs,
1448 if (!mutex_trylock(&head->mutex)) {
1449 refcount_inc(&head->refs);
1450 spin_unlock(&delayed_refs->lock);
1452 btrfs_release_path(path);
1455 * Mutex was contended, block until it's
1456 * released and try again
1458 mutex_lock(&head->mutex);
1459 mutex_unlock(&head->mutex);
1460 btrfs_put_delayed_ref_head(head);
1463 spin_unlock(&delayed_refs->lock);
1464 ret = add_delayed_refs(ctx->fs_info, head, ctx->time_seq,
1466 mutex_unlock(&head->mutex);
1470 spin_unlock(&delayed_refs->lock);
1474 if (path->slots[0]) {
1475 struct extent_buffer *leaf;
1479 leaf = path->nodes[0];
1480 slot = path->slots[0];
1481 btrfs_item_key_to_cpu(leaf, &key, slot);
1482 if (key.objectid == ctx->bytenr &&
1483 (key.type == BTRFS_EXTENT_ITEM_KEY ||
1484 key.type == BTRFS_METADATA_ITEM_KEY)) {
1485 ret = add_inline_refs(ctx, path, &info_level,
1489 ret = add_keyed_refs(ctx, root, path, info_level,
1497 * If we have a share context and we reached here, it means the extent
1498 * is not directly shared (no multiple reference items for it),
1499 * otherwise we would have exited earlier with a return value of
1500 * BACKREF_FOUND_SHARED after processing delayed references or while
1501 * processing inline or keyed references from the extent tree.
1502 * The extent may however be indirectly shared through shared subtrees
1503 * as a result from creating snapshots, so we determine below what is
1504 * its parent node, in case we are dealing with a metadata extent, or
1505 * what's the leaf (or leaves), from a fs tree, that has a file extent
1506 * item pointing to it in case we are dealing with a data extent.
1508 ASSERT(extent_is_shared(sc) == 0);
1511 * If we are here for a data extent and we have a share_check structure
1512 * it means the data extent is not directly shared (does not have
1513 * multiple reference items), so we have to check if a path in the fs
1514 * tree (going from the root node down to the leaf that has the file
1515 * extent item pointing to the data extent) is shared, that is, if any
1516 * of the extent buffers in the path is referenced by other trees.
1518 if (sc && ctx->bytenr == sc->data_bytenr) {
1520 * If our data extent is from a generation more recent than the
1521 * last generation used to snapshot the root, then we know that
1522 * it can not be shared through subtrees, so we can skip
1523 * resolving indirect references, there's no point in
1524 * determining the extent buffers for the path from the fs tree
1525 * root node down to the leaf that has the file extent item that
1526 * points to the data extent.
1528 if (sc->data_extent_gen >
1529 btrfs_root_last_snapshot(&sc->root->root_item)) {
1530 ret = BACKREF_FOUND_NOT_SHARED;
1535 * If we are only determining if a data extent is shared or not
1536 * and the corresponding file extent item is located in the same
1537 * leaf as the previous file extent item, we can skip resolving
1538 * indirect references for a data extent, since the fs tree path
1539 * is the same (same leaf, so same path). We skip as long as the
1540 * cached result for the leaf is valid and only if there's only
1541 * one file extent item pointing to the data extent, because in
1542 * the case of multiple file extent items, they may be located
1543 * in different leaves and therefore we have multiple paths.
1545 if (sc->ctx->curr_leaf_bytenr == sc->ctx->prev_leaf_bytenr &&
1546 sc->self_ref_count == 1) {
1550 cached = lookup_backref_shared_cache(sc->ctx, sc->root,
1551 sc->ctx->curr_leaf_bytenr,
1555 ret = BACKREF_FOUND_SHARED;
1557 ret = BACKREF_FOUND_NOT_SHARED;
1563 btrfs_release_path(path);
1565 ret = add_missing_keys(ctx->fs_info, &preftrees, path->skip_locking == 0);
1569 WARN_ON(!RB_EMPTY_ROOT(&preftrees.indirect_missing_keys.root.rb_root));
1571 ret = resolve_indirect_refs(ctx, path, &preftrees, sc);
1575 WARN_ON(!RB_EMPTY_ROOT(&preftrees.indirect.root.rb_root));
1578 * This walks the tree of merged and resolved refs. Tree blocks are
1579 * read in as needed. Unique entries are added to the ulist, and
1580 * the list of found roots is updated.
1582 * We release the entire tree in one go before returning.
1584 node = rb_first_cached(&preftrees.direct.root);
1586 ref = rb_entry(node, struct prelim_ref, rbnode);
1587 node = rb_next(&ref->rbnode);
1589 * ref->count < 0 can happen here if there are delayed
1590 * refs with a node->action of BTRFS_DROP_DELAYED_REF.
1591 * prelim_ref_insert() relies on this when merging
1592 * identical refs to keep the overall count correct.
1593 * prelim_ref_insert() will merge only those refs
1594 * which compare identically. Any refs having
1595 * e.g. different offsets would not be merged,
1596 * and would retain their original ref->count < 0.
1598 if (ctx->roots && ref->count && ref->root_id && ref->parent == 0) {
1599 /* no parent == root of tree */
1600 ret = ulist_add(ctx->roots, ref->root_id, 0, GFP_NOFS);
1604 if (ref->count && ref->parent) {
1605 if (!ctx->skip_inode_ref_list && !ref->inode_list &&
1607 struct btrfs_tree_parent_check check = { 0 };
1608 struct extent_buffer *eb;
1610 check.level = ref->level;
1612 eb = read_tree_block(ctx->fs_info, ref->parent,
1618 if (!extent_buffer_uptodate(eb)) {
1619 free_extent_buffer(eb);
1624 if (!path->skip_locking)
1625 btrfs_tree_read_lock(eb);
1626 ret = find_extent_in_eb(ctx, eb, &eie);
1627 if (!path->skip_locking)
1628 btrfs_tree_read_unlock(eb);
1629 free_extent_buffer(eb);
1630 if (ret == BTRFS_ITERATE_EXTENT_INODES_STOP ||
1633 ref->inode_list = eie;
1635 * We transferred the list ownership to the ref,
1636 * so set to NULL to avoid a double free in case
1637 * an error happens after this.
1641 ret = ulist_add_merge_ptr(ctx->refs, ref->parent,
1643 (void **)&eie, GFP_NOFS);
1646 if (!ret && !ctx->skip_inode_ref_list) {
1648 * We've recorded that parent, so we must extend
1649 * its inode list here.
1651 * However if there was corruption we may not
1652 * have found an eie, return an error in this
1662 eie->next = ref->inode_list;
1666 * We have transferred the inode list ownership from
1667 * this ref to the ref we added to the 'refs' ulist.
1668 * So set this ref's inode list to NULL to avoid
1669 * use-after-free when our caller uses it or double
1670 * frees in case an error happens before we return.
1672 ref->inode_list = NULL;
1678 btrfs_free_path(path);
1680 prelim_release(&preftrees.direct);
1681 prelim_release(&preftrees.indirect);
1682 prelim_release(&preftrees.indirect_missing_keys);
1684 if (ret == BTRFS_ITERATE_EXTENT_INODES_STOP || ret < 0)
1685 free_inode_elem_list(eie);
1690 * Finds all leaves with a reference to the specified combination of
1691 * @ctx->bytenr and @ctx->extent_item_pos. The bytenr of the found leaves are
1692 * added to the ulist at @ctx->refs, and that ulist is allocated by this
1693 * function. The caller should free the ulist with free_leaf_list() if
1694 * @ctx->ignore_extent_item_pos is false, otherwise a fimple ulist_free() is
1697 * Returns 0 on success and < 0 on error. On error @ctx->refs is not allocated.
1699 int btrfs_find_all_leafs(struct btrfs_backref_walk_ctx *ctx)
1703 ASSERT(ctx->refs == NULL);
1705 ctx->refs = ulist_alloc(GFP_NOFS);
1709 ret = find_parent_nodes(ctx, NULL);
1710 if (ret == BTRFS_ITERATE_EXTENT_INODES_STOP ||
1711 (ret < 0 && ret != -ENOENT)) {
1712 free_leaf_list(ctx->refs);
1721 * Walk all backrefs for a given extent to find all roots that reference this
1722 * extent. Walking a backref means finding all extents that reference this
1723 * extent and in turn walk the backrefs of those, too. Naturally this is a
1724 * recursive process, but here it is implemented in an iterative fashion: We
1725 * find all referencing extents for the extent in question and put them on a
1726 * list. In turn, we find all referencing extents for those, further appending
1727 * to the list. The way we iterate the list allows adding more elements after
1728 * the current while iterating. The process stops when we reach the end of the
1731 * Found roots are added to @ctx->roots, which is allocated by this function if
1732 * it points to NULL, in which case the caller is responsible for freeing it
1733 * after it's not needed anymore.
1734 * This function requires @ctx->refs to be NULL, as it uses it for allocating a
1735 * ulist to do temporary work, and frees it before returning.
1737 * Returns 0 on success, < 0 on error.
1739 static int btrfs_find_all_roots_safe(struct btrfs_backref_walk_ctx *ctx)
1741 const u64 orig_bytenr = ctx->bytenr;
1742 const bool orig_skip_inode_ref_list = ctx->skip_inode_ref_list;
1743 bool roots_ulist_allocated = false;
1744 struct ulist_iterator uiter;
1747 ASSERT(ctx->refs == NULL);
1749 ctx->refs = ulist_alloc(GFP_NOFS);
1754 ctx->roots = ulist_alloc(GFP_NOFS);
1756 ulist_free(ctx->refs);
1760 roots_ulist_allocated = true;
1763 ctx->skip_inode_ref_list = true;
1765 ULIST_ITER_INIT(&uiter);
1767 struct ulist_node *node;
1769 ret = find_parent_nodes(ctx, NULL);
1770 if (ret < 0 && ret != -ENOENT) {
1771 if (roots_ulist_allocated) {
1772 ulist_free(ctx->roots);
1778 node = ulist_next(ctx->refs, &uiter);
1781 ctx->bytenr = node->val;
1785 ulist_free(ctx->refs);
1787 ctx->bytenr = orig_bytenr;
1788 ctx->skip_inode_ref_list = orig_skip_inode_ref_list;
1793 int btrfs_find_all_roots(struct btrfs_backref_walk_ctx *ctx,
1794 bool skip_commit_root_sem)
1798 if (!ctx->trans && !skip_commit_root_sem)
1799 down_read(&ctx->fs_info->commit_root_sem);
1800 ret = btrfs_find_all_roots_safe(ctx);
1801 if (!ctx->trans && !skip_commit_root_sem)
1802 up_read(&ctx->fs_info->commit_root_sem);
1806 struct btrfs_backref_share_check_ctx *btrfs_alloc_backref_share_check_ctx(void)
1808 struct btrfs_backref_share_check_ctx *ctx;
1810 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
1814 ulist_init(&ctx->refs);
1819 void btrfs_free_backref_share_ctx(struct btrfs_backref_share_check_ctx *ctx)
1824 ulist_release(&ctx->refs);
1829 * Check if a data extent is shared or not.
1831 * @inode: The inode whose extent we are checking.
1832 * @bytenr: Logical bytenr of the extent we are checking.
1833 * @extent_gen: Generation of the extent (file extent item) or 0 if it is
1835 * @ctx: A backref sharedness check context.
1837 * btrfs_is_data_extent_shared uses the backref walking code but will short
1838 * circuit as soon as it finds a root or inode that doesn't match the
1839 * one passed in. This provides a significant performance benefit for
1840 * callers (such as fiemap) which want to know whether the extent is
1841 * shared but do not need a ref count.
1843 * This attempts to attach to the running transaction in order to account for
1844 * delayed refs, but continues on even when no running transaction exists.
1846 * Return: 0 if extent is not shared, 1 if it is shared, < 0 on error.
1848 int btrfs_is_data_extent_shared(struct btrfs_inode *inode, u64 bytenr,
1850 struct btrfs_backref_share_check_ctx *ctx)
1852 struct btrfs_backref_walk_ctx walk_ctx = { 0 };
1853 struct btrfs_root *root = inode->root;
1854 struct btrfs_fs_info *fs_info = root->fs_info;
1855 struct btrfs_trans_handle *trans;
1856 struct ulist_iterator uiter;
1857 struct ulist_node *node;
1858 struct btrfs_seq_list elem = BTRFS_SEQ_LIST_INIT(elem);
1860 struct share_check shared = {
1863 .inum = btrfs_ino(inode),
1864 .data_bytenr = bytenr,
1865 .data_extent_gen = extent_gen,
1867 .self_ref_count = 0,
1868 .have_delayed_delete_refs = false,
1872 bool leaf_is_shared;
1874 for (int i = 0; i < BTRFS_BACKREF_CTX_PREV_EXTENTS_SIZE; i++) {
1875 if (ctx->prev_extents_cache[i].bytenr == bytenr)
1876 return ctx->prev_extents_cache[i].is_shared;
1879 ulist_init(&ctx->refs);
1881 trans = btrfs_join_transaction_nostart(root);
1882 if (IS_ERR(trans)) {
1883 if (PTR_ERR(trans) != -ENOENT && PTR_ERR(trans) != -EROFS) {
1884 ret = PTR_ERR(trans);
1888 down_read(&fs_info->commit_root_sem);
1890 btrfs_get_tree_mod_seq(fs_info, &elem);
1891 walk_ctx.time_seq = elem.seq;
1894 ctx->use_path_cache = true;
1897 * We may have previously determined that the current leaf is shared.
1898 * If it is, then we have a data extent that is shared due to a shared
1899 * subtree (caused by snapshotting) and we don't need to check for data
1900 * backrefs. If the leaf is not shared, then we must do backref walking
1901 * to determine if the data extent is shared through reflinks.
1903 leaf_cached = lookup_backref_shared_cache(ctx, root,
1904 ctx->curr_leaf_bytenr, 0,
1906 if (leaf_cached && leaf_is_shared) {
1911 walk_ctx.skip_inode_ref_list = true;
1912 walk_ctx.trans = trans;
1913 walk_ctx.fs_info = fs_info;
1914 walk_ctx.refs = &ctx->refs;
1916 /* -1 means we are in the bytenr of the data extent. */
1918 ULIST_ITER_INIT(&uiter);
1920 const unsigned long prev_ref_count = ctx->refs.nnodes;
1922 walk_ctx.bytenr = bytenr;
1923 ret = find_parent_nodes(&walk_ctx, &shared);
1924 if (ret == BACKREF_FOUND_SHARED ||
1925 ret == BACKREF_FOUND_NOT_SHARED) {
1926 /* If shared must return 1, otherwise return 0. */
1927 ret = (ret == BACKREF_FOUND_SHARED) ? 1 : 0;
1929 store_backref_shared_cache(ctx, root, bytenr,
1933 if (ret < 0 && ret != -ENOENT)
1938 * More than one extent buffer (bytenr) may have been added to
1939 * the ctx->refs ulist, in which case we have to check multiple
1940 * tree paths in case the first one is not shared, so we can not
1941 * use the path cache which is made for a single path. Multiple
1942 * extent buffers at the current level happen when:
1944 * 1) level -1, the data extent: If our data extent was not
1945 * directly shared (without multiple reference items), then
1946 * it might have a single reference item with a count > 1 for
1947 * the same offset, which means there are 2 (or more) file
1948 * extent items that point to the data extent - this happens
1949 * when a file extent item needs to be split and then one
1950 * item gets moved to another leaf due to a b+tree leaf split
1951 * when inserting some item. In this case the file extent
1952 * items may be located in different leaves and therefore
1953 * some of the leaves may be referenced through shared
1954 * subtrees while others are not. Since our extent buffer
1955 * cache only works for a single path (by far the most common
1956 * case and simpler to deal with), we can not use it if we
1957 * have multiple leaves (which implies multiple paths).
1959 * 2) level >= 0, a tree node/leaf: We can have a mix of direct
1960 * and indirect references on a b+tree node/leaf, so we have
1961 * to check multiple paths, and the extent buffer (the
1962 * current bytenr) may be shared or not. One example is
1963 * during relocation as we may get a shared tree block ref
1964 * (direct ref) and a non-shared tree block ref (indirect
1965 * ref) for the same node/leaf.
1967 if ((ctx->refs.nnodes - prev_ref_count) > 1)
1968 ctx->use_path_cache = false;
1971 store_backref_shared_cache(ctx, root, bytenr,
1973 node = ulist_next(&ctx->refs, &uiter);
1977 if (ctx->use_path_cache) {
1982 cached = lookup_backref_shared_cache(ctx, root, bytenr,
1985 ret = (is_shared ? 1 : 0);
1989 shared.share_count = 0;
1990 shared.have_delayed_delete_refs = false;
1995 * If the path cache is disabled, then it means at some tree level we
1996 * got multiple parents due to a mix of direct and indirect backrefs or
1997 * multiple leaves with file extent items pointing to the same data
1998 * extent. We have to invalidate the cache and cache only the sharedness
1999 * result for the levels where we got only one node/reference.
2001 if (!ctx->use_path_cache) {
2005 if (ret >= 0 && level >= 0) {
2006 bytenr = ctx->path_cache_entries[level].bytenr;
2007 ctx->use_path_cache = true;
2008 store_backref_shared_cache(ctx, root, bytenr, level, ret);
2012 for ( ; i < BTRFS_MAX_LEVEL; i++)
2013 ctx->path_cache_entries[i].bytenr = 0;
2017 * Cache the sharedness result for the data extent if we know our inode
2018 * has more than 1 file extent item that refers to the data extent.
2020 if (ret >= 0 && shared.self_ref_count > 1) {
2021 int slot = ctx->prev_extents_cache_slot;
2023 ctx->prev_extents_cache[slot].bytenr = shared.data_bytenr;
2024 ctx->prev_extents_cache[slot].is_shared = (ret == 1);
2026 slot = (slot + 1) % BTRFS_BACKREF_CTX_PREV_EXTENTS_SIZE;
2027 ctx->prev_extents_cache_slot = slot;
2032 btrfs_put_tree_mod_seq(fs_info, &elem);
2033 btrfs_end_transaction(trans);
2035 up_read(&fs_info->commit_root_sem);
2038 ulist_release(&ctx->refs);
2039 ctx->prev_leaf_bytenr = ctx->curr_leaf_bytenr;
2044 int btrfs_find_one_extref(struct btrfs_root *root, u64 inode_objectid,
2045 u64 start_off, struct btrfs_path *path,
2046 struct btrfs_inode_extref **ret_extref,
2050 struct btrfs_key key;
2051 struct btrfs_key found_key;
2052 struct btrfs_inode_extref *extref;
2053 const struct extent_buffer *leaf;
2056 key.objectid = inode_objectid;
2057 key.type = BTRFS_INODE_EXTREF_KEY;
2058 key.offset = start_off;
2060 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2065 leaf = path->nodes[0];
2066 slot = path->slots[0];
2067 if (slot >= btrfs_header_nritems(leaf)) {
2069 * If the item at offset is not found,
2070 * btrfs_search_slot will point us to the slot
2071 * where it should be inserted. In our case
2072 * that will be the slot directly before the
2073 * next INODE_REF_KEY_V2 item. In the case
2074 * that we're pointing to the last slot in a
2075 * leaf, we must move one leaf over.
2077 ret = btrfs_next_leaf(root, path);
2086 btrfs_item_key_to_cpu(leaf, &found_key, slot);
2089 * Check that we're still looking at an extended ref key for
2090 * this particular objectid. If we have different
2091 * objectid or type then there are no more to be found
2092 * in the tree and we can exit.
2095 if (found_key.objectid != inode_objectid)
2097 if (found_key.type != BTRFS_INODE_EXTREF_KEY)
2101 ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
2102 extref = (struct btrfs_inode_extref *)ptr;
2103 *ret_extref = extref;
2105 *found_off = found_key.offset;
2113 * this iterates to turn a name (from iref/extref) into a full filesystem path.
2114 * Elements of the path are separated by '/' and the path is guaranteed to be
2115 * 0-terminated. the path is only given within the current file system.
2116 * Therefore, it never starts with a '/'. the caller is responsible to provide
2117 * "size" bytes in "dest". the dest buffer will be filled backwards. finally,
2118 * the start point of the resulting string is returned. this pointer is within
2120 * in case the path buffer would overflow, the pointer is decremented further
2121 * as if output was written to the buffer, though no more output is actually
2122 * generated. that way, the caller can determine how much space would be
2123 * required for the path to fit into the buffer. in that case, the returned
2124 * value will be smaller than dest. callers must check this!
2126 char *btrfs_ref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
2127 u32 name_len, unsigned long name_off,
2128 struct extent_buffer *eb_in, u64 parent,
2129 char *dest, u32 size)
2134 s64 bytes_left = ((s64)size) - 1;
2135 struct extent_buffer *eb = eb_in;
2136 struct btrfs_key found_key;
2137 struct btrfs_inode_ref *iref;
2139 if (bytes_left >= 0)
2140 dest[bytes_left] = '\0';
2143 bytes_left -= name_len;
2144 if (bytes_left >= 0)
2145 read_extent_buffer(eb, dest + bytes_left,
2146 name_off, name_len);
2148 if (!path->skip_locking)
2149 btrfs_tree_read_unlock(eb);
2150 free_extent_buffer(eb);
2152 ret = btrfs_find_item(fs_root, path, parent, 0,
2153 BTRFS_INODE_REF_KEY, &found_key);
2159 next_inum = found_key.offset;
2161 /* regular exit ahead */
2162 if (parent == next_inum)
2165 slot = path->slots[0];
2166 eb = path->nodes[0];
2167 /* make sure we can use eb after releasing the path */
2169 path->nodes[0] = NULL;
2172 btrfs_release_path(path);
2173 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
2175 name_len = btrfs_inode_ref_name_len(eb, iref);
2176 name_off = (unsigned long)(iref + 1);
2180 if (bytes_left >= 0)
2181 dest[bytes_left] = '/';
2184 btrfs_release_path(path);
2187 return ERR_PTR(ret);
2189 return dest + bytes_left;
2193 * this makes the path point to (logical EXTENT_ITEM *)
2194 * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for
2195 * tree blocks and <0 on error.
2197 int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
2198 struct btrfs_path *path, struct btrfs_key *found_key,
2201 struct btrfs_root *extent_root = btrfs_extent_root(fs_info, logical);
2206 const struct extent_buffer *eb;
2207 struct btrfs_extent_item *ei;
2208 struct btrfs_key key;
2210 if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
2211 key.type = BTRFS_METADATA_ITEM_KEY;
2213 key.type = BTRFS_EXTENT_ITEM_KEY;
2214 key.objectid = logical;
2215 key.offset = (u64)-1;
2217 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
2222 * Key with offset -1 found, there would have to exist an extent
2223 * item with such offset, but this is out of the valid range.
2228 ret = btrfs_previous_extent_item(extent_root, path, 0);
2234 btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]);
2235 if (found_key->type == BTRFS_METADATA_ITEM_KEY)
2236 size = fs_info->nodesize;
2237 else if (found_key->type == BTRFS_EXTENT_ITEM_KEY)
2238 size = found_key->offset;
2240 if (found_key->objectid > logical ||
2241 found_key->objectid + size <= logical) {
2242 btrfs_debug(fs_info,
2243 "logical %llu is not within any extent", logical);
2247 eb = path->nodes[0];
2248 item_size = btrfs_item_size(eb, path->slots[0]);
2250 ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
2251 flags = btrfs_extent_flags(eb, ei);
2253 btrfs_debug(fs_info,
2254 "logical %llu is at position %llu within the extent (%llu EXTENT_ITEM %llu) flags %#llx size %u",
2255 logical, logical - found_key->objectid, found_key->objectid,
2256 found_key->offset, flags, item_size);
2258 WARN_ON(!flags_ret);
2260 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
2261 *flags_ret = BTRFS_EXTENT_FLAG_TREE_BLOCK;
2262 else if (flags & BTRFS_EXTENT_FLAG_DATA)
2263 *flags_ret = BTRFS_EXTENT_FLAG_DATA;
2273 * helper function to iterate extent inline refs. ptr must point to a 0 value
2274 * for the first call and may be modified. it is used to track state.
2275 * if more refs exist, 0 is returned and the next call to
2276 * get_extent_inline_ref must pass the modified ptr parameter to get the
2277 * next ref. after the last ref was processed, 1 is returned.
2278 * returns <0 on error
2280 static int get_extent_inline_ref(unsigned long *ptr,
2281 const struct extent_buffer *eb,
2282 const struct btrfs_key *key,
2283 const struct btrfs_extent_item *ei,
2285 struct btrfs_extent_inline_ref **out_eiref,
2290 struct btrfs_tree_block_info *info;
2294 flags = btrfs_extent_flags(eb, ei);
2295 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
2296 if (key->type == BTRFS_METADATA_ITEM_KEY) {
2297 /* a skinny metadata extent */
2299 (struct btrfs_extent_inline_ref *)(ei + 1);
2301 WARN_ON(key->type != BTRFS_EXTENT_ITEM_KEY);
2302 info = (struct btrfs_tree_block_info *)(ei + 1);
2304 (struct btrfs_extent_inline_ref *)(info + 1);
2307 *out_eiref = (struct btrfs_extent_inline_ref *)(ei + 1);
2309 *ptr = (unsigned long)*out_eiref;
2310 if ((unsigned long)(*ptr) >= (unsigned long)ei + item_size)
2314 end = (unsigned long)ei + item_size;
2315 *out_eiref = (struct btrfs_extent_inline_ref *)(*ptr);
2316 *out_type = btrfs_get_extent_inline_ref_type(eb, *out_eiref,
2317 BTRFS_REF_TYPE_ANY);
2318 if (*out_type == BTRFS_REF_TYPE_INVALID)
2321 *ptr += btrfs_extent_inline_ref_size(*out_type);
2322 WARN_ON(*ptr > end);
2324 return 1; /* last */
2330 * reads the tree block backref for an extent. tree level and root are returned
2331 * through out_level and out_root. ptr must point to a 0 value for the first
2332 * call and may be modified (see get_extent_inline_ref comment).
2333 * returns 0 if data was provided, 1 if there was no more data to provide or
2336 int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
2337 struct btrfs_key *key, struct btrfs_extent_item *ei,
2338 u32 item_size, u64 *out_root, u8 *out_level)
2342 struct btrfs_extent_inline_ref *eiref;
2344 if (*ptr == (unsigned long)-1)
2348 ret = get_extent_inline_ref(ptr, eb, key, ei, item_size,
2353 if (type == BTRFS_TREE_BLOCK_REF_KEY ||
2354 type == BTRFS_SHARED_BLOCK_REF_KEY)
2361 /* we can treat both ref types equally here */
2362 *out_root = btrfs_extent_inline_ref_offset(eb, eiref);
2364 if (key->type == BTRFS_EXTENT_ITEM_KEY) {
2365 struct btrfs_tree_block_info *info;
2367 info = (struct btrfs_tree_block_info *)(ei + 1);
2368 *out_level = btrfs_tree_block_level(eb, info);
2370 ASSERT(key->type == BTRFS_METADATA_ITEM_KEY);
2371 *out_level = (u8)key->offset;
2375 *ptr = (unsigned long)-1;
2380 static int iterate_leaf_refs(struct btrfs_fs_info *fs_info,
2381 struct extent_inode_elem *inode_list,
2382 u64 root, u64 extent_item_objectid,
2383 iterate_extent_inodes_t *iterate, void *ctx)
2385 struct extent_inode_elem *eie;
2388 for (eie = inode_list; eie; eie = eie->next) {
2389 btrfs_debug(fs_info,
2390 "ref for %llu resolved, key (%llu EXTEND_DATA %llu), root %llu",
2391 extent_item_objectid, eie->inum,
2393 ret = iterate(eie->inum, eie->offset, eie->num_bytes, root, ctx);
2395 btrfs_debug(fs_info,
2396 "stopping iteration for %llu due to ret=%d",
2397 extent_item_objectid, ret);
2406 * calls iterate() for every inode that references the extent identified by
2407 * the given parameters.
2408 * when the iterator function returns a non-zero value, iteration stops.
2410 int iterate_extent_inodes(struct btrfs_backref_walk_ctx *ctx,
2411 bool search_commit_root,
2412 iterate_extent_inodes_t *iterate, void *user_ctx)
2416 struct ulist_node *ref_node;
2417 struct btrfs_seq_list seq_elem = BTRFS_SEQ_LIST_INIT(seq_elem);
2418 struct ulist_iterator ref_uiter;
2420 btrfs_debug(ctx->fs_info, "resolving all inodes for extent %llu",
2423 ASSERT(ctx->trans == NULL);
2424 ASSERT(ctx->roots == NULL);
2426 if (!search_commit_root) {
2427 struct btrfs_trans_handle *trans;
2429 trans = btrfs_attach_transaction(ctx->fs_info->tree_root);
2430 if (IS_ERR(trans)) {
2431 if (PTR_ERR(trans) != -ENOENT &&
2432 PTR_ERR(trans) != -EROFS)
2433 return PTR_ERR(trans);
2440 btrfs_get_tree_mod_seq(ctx->fs_info, &seq_elem);
2441 ctx->time_seq = seq_elem.seq;
2443 down_read(&ctx->fs_info->commit_root_sem);
2446 ret = btrfs_find_all_leafs(ctx);
2452 ULIST_ITER_INIT(&ref_uiter);
2453 while (!ret && (ref_node = ulist_next(refs, &ref_uiter))) {
2454 const u64 leaf_bytenr = ref_node->val;
2455 struct ulist_node *root_node;
2456 struct ulist_iterator root_uiter;
2457 struct extent_inode_elem *inode_list;
2459 inode_list = (struct extent_inode_elem *)(uintptr_t)ref_node->aux;
2461 if (ctx->cache_lookup) {
2462 const u64 *root_ids;
2466 cached = ctx->cache_lookup(leaf_bytenr, ctx->user_ctx,
2467 &root_ids, &root_count);
2469 for (int i = 0; i < root_count; i++) {
2470 ret = iterate_leaf_refs(ctx->fs_info,
2484 ctx->roots = ulist_alloc(GFP_NOFS);
2491 ctx->bytenr = leaf_bytenr;
2492 ret = btrfs_find_all_roots_safe(ctx);
2496 if (ctx->cache_store)
2497 ctx->cache_store(leaf_bytenr, ctx->roots, ctx->user_ctx);
2499 ULIST_ITER_INIT(&root_uiter);
2500 while (!ret && (root_node = ulist_next(ctx->roots, &root_uiter))) {
2501 btrfs_debug(ctx->fs_info,
2502 "root %llu references leaf %llu, data list %#llx",
2503 root_node->val, ref_node->val,
2505 ret = iterate_leaf_refs(ctx->fs_info, inode_list,
2506 root_node->val, ctx->bytenr,
2509 ulist_reinit(ctx->roots);
2512 free_leaf_list(refs);
2515 btrfs_put_tree_mod_seq(ctx->fs_info, &seq_elem);
2516 btrfs_end_transaction(ctx->trans);
2519 up_read(&ctx->fs_info->commit_root_sem);
2522 ulist_free(ctx->roots);
2525 if (ret == BTRFS_ITERATE_EXTENT_INODES_STOP)
2531 static int build_ino_list(u64 inum, u64 offset, u64 num_bytes, u64 root, void *ctx)
2533 struct btrfs_data_container *inodes = ctx;
2534 const size_t c = 3 * sizeof(u64);
2536 if (inodes->bytes_left >= c) {
2537 inodes->bytes_left -= c;
2538 inodes->val[inodes->elem_cnt] = inum;
2539 inodes->val[inodes->elem_cnt + 1] = offset;
2540 inodes->val[inodes->elem_cnt + 2] = root;
2541 inodes->elem_cnt += 3;
2543 inodes->bytes_missing += c - inodes->bytes_left;
2544 inodes->bytes_left = 0;
2545 inodes->elem_missed += 3;
2551 int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
2552 struct btrfs_path *path,
2553 void *ctx, bool ignore_offset)
2555 struct btrfs_backref_walk_ctx walk_ctx = { 0 };
2558 struct btrfs_key found_key;
2559 int search_commit_root = path->search_commit_root;
2561 ret = extent_from_logical(fs_info, logical, path, &found_key, &flags);
2562 btrfs_release_path(path);
2565 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
2568 walk_ctx.bytenr = found_key.objectid;
2570 walk_ctx.ignore_extent_item_pos = true;
2572 walk_ctx.extent_item_pos = logical - found_key.objectid;
2573 walk_ctx.fs_info = fs_info;
2575 return iterate_extent_inodes(&walk_ctx, search_commit_root,
2576 build_ino_list, ctx);
2579 static int inode_to_path(u64 inum, u32 name_len, unsigned long name_off,
2580 struct extent_buffer *eb, struct inode_fs_paths *ipath);
2582 static int iterate_inode_refs(u64 inum, struct inode_fs_paths *ipath)
2591 struct btrfs_root *fs_root = ipath->fs_root;
2592 struct btrfs_path *path = ipath->btrfs_path;
2593 struct extent_buffer *eb;
2594 struct btrfs_inode_ref *iref;
2595 struct btrfs_key found_key;
2598 ret = btrfs_find_item(fs_root, path, inum,
2599 parent ? parent + 1 : 0, BTRFS_INODE_REF_KEY,
2605 ret = found ? 0 : -ENOENT;
2610 parent = found_key.offset;
2611 slot = path->slots[0];
2612 eb = btrfs_clone_extent_buffer(path->nodes[0]);
2617 btrfs_release_path(path);
2619 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
2621 for (cur = 0; cur < btrfs_item_size(eb, slot); cur += len) {
2622 name_len = btrfs_inode_ref_name_len(eb, iref);
2623 /* path must be released before calling iterate()! */
2624 btrfs_debug(fs_root->fs_info,
2625 "following ref at offset %u for inode %llu in tree %llu",
2626 cur, found_key.objectid,
2627 btrfs_root_id(fs_root));
2628 ret = inode_to_path(parent, name_len,
2629 (unsigned long)(iref + 1), eb, ipath);
2632 len = sizeof(*iref) + name_len;
2633 iref = (struct btrfs_inode_ref *)((char *)iref + len);
2635 free_extent_buffer(eb);
2638 btrfs_release_path(path);
2643 static int iterate_inode_extrefs(u64 inum, struct inode_fs_paths *ipath)
2650 struct btrfs_root *fs_root = ipath->fs_root;
2651 struct btrfs_path *path = ipath->btrfs_path;
2652 struct extent_buffer *eb;
2653 struct btrfs_inode_extref *extref;
2659 ret = btrfs_find_one_extref(fs_root, inum, offset, path, &extref,
2664 ret = found ? 0 : -ENOENT;
2669 slot = path->slots[0];
2670 eb = btrfs_clone_extent_buffer(path->nodes[0]);
2675 btrfs_release_path(path);
2677 item_size = btrfs_item_size(eb, slot);
2678 ptr = btrfs_item_ptr_offset(eb, slot);
2681 while (cur_offset < item_size) {
2684 extref = (struct btrfs_inode_extref *)(ptr + cur_offset);
2685 parent = btrfs_inode_extref_parent(eb, extref);
2686 name_len = btrfs_inode_extref_name_len(eb, extref);
2687 ret = inode_to_path(parent, name_len,
2688 (unsigned long)&extref->name, eb, ipath);
2692 cur_offset += btrfs_inode_extref_name_len(eb, extref);
2693 cur_offset += sizeof(*extref);
2695 free_extent_buffer(eb);
2700 btrfs_release_path(path);
2706 * returns 0 if the path could be dumped (probably truncated)
2707 * returns <0 in case of an error
2709 static int inode_to_path(u64 inum, u32 name_len, unsigned long name_off,
2710 struct extent_buffer *eb, struct inode_fs_paths *ipath)
2714 int i = ipath->fspath->elem_cnt;
2715 const int s_ptr = sizeof(char *);
2718 bytes_left = ipath->fspath->bytes_left > s_ptr ?
2719 ipath->fspath->bytes_left - s_ptr : 0;
2721 fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
2722 fspath = btrfs_ref_to_path(ipath->fs_root, ipath->btrfs_path, name_len,
2723 name_off, eb, inum, fspath_min, bytes_left);
2725 return PTR_ERR(fspath);
2727 if (fspath > fspath_min) {
2728 ipath->fspath->val[i] = (u64)(unsigned long)fspath;
2729 ++ipath->fspath->elem_cnt;
2730 ipath->fspath->bytes_left = fspath - fspath_min;
2732 ++ipath->fspath->elem_missed;
2733 ipath->fspath->bytes_missing += fspath_min - fspath;
2734 ipath->fspath->bytes_left = 0;
2741 * this dumps all file system paths to the inode into the ipath struct, provided
2742 * is has been created large enough. each path is zero-terminated and accessed
2743 * from ipath->fspath->val[i].
2744 * when it returns, there are ipath->fspath->elem_cnt number of paths available
2745 * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
2746 * number of missed paths is recorded in ipath->fspath->elem_missed, otherwise,
2747 * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
2748 * have been needed to return all paths.
2750 int paths_from_inode(u64 inum, struct inode_fs_paths *ipath)
2755 ret = iterate_inode_refs(inum, ipath);
2758 else if (ret != -ENOENT)
2761 ret = iterate_inode_extrefs(inum, ipath);
2762 if (ret == -ENOENT && found_refs)
2768 struct btrfs_data_container *init_data_container(u32 total_bytes)
2770 struct btrfs_data_container *data;
2773 alloc_bytes = max_t(size_t, total_bytes, sizeof(*data));
2774 data = kvzalloc(alloc_bytes, GFP_KERNEL);
2776 return ERR_PTR(-ENOMEM);
2778 if (total_bytes >= sizeof(*data))
2779 data->bytes_left = total_bytes - sizeof(*data);
2781 data->bytes_missing = sizeof(*data) - total_bytes;
2787 * allocates space to return multiple file system paths for an inode.
2788 * total_bytes to allocate are passed, note that space usable for actual path
2789 * information will be total_bytes - sizeof(struct inode_fs_paths).
2790 * the returned pointer must be freed with free_ipath() in the end.
2792 struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
2793 struct btrfs_path *path)
2795 struct inode_fs_paths *ifp;
2796 struct btrfs_data_container *fspath;
2798 fspath = init_data_container(total_bytes);
2800 return ERR_CAST(fspath);
2802 ifp = kmalloc(sizeof(*ifp), GFP_KERNEL);
2805 return ERR_PTR(-ENOMEM);
2808 ifp->btrfs_path = path;
2809 ifp->fspath = fspath;
2810 ifp->fs_root = fs_root;
2815 void free_ipath(struct inode_fs_paths *ipath)
2819 kvfree(ipath->fspath);
2823 struct btrfs_backref_iter *btrfs_backref_iter_alloc(struct btrfs_fs_info *fs_info)
2825 struct btrfs_backref_iter *ret;
2827 ret = kzalloc(sizeof(*ret), GFP_NOFS);
2831 ret->path = btrfs_alloc_path();
2837 /* Current backref iterator only supports iteration in commit root */
2838 ret->path->search_commit_root = 1;
2839 ret->path->skip_locking = 1;
2840 ret->fs_info = fs_info;
2845 static void btrfs_backref_iter_release(struct btrfs_backref_iter *iter)
2851 btrfs_release_path(iter->path);
2852 memset(&iter->cur_key, 0, sizeof(iter->cur_key));
2855 int btrfs_backref_iter_start(struct btrfs_backref_iter *iter, u64 bytenr)
2857 struct btrfs_fs_info *fs_info = iter->fs_info;
2858 struct btrfs_root *extent_root = btrfs_extent_root(fs_info, bytenr);
2859 struct btrfs_path *path = iter->path;
2860 struct btrfs_extent_item *ei;
2861 struct btrfs_key key;
2864 key.objectid = bytenr;
2865 key.type = BTRFS_METADATA_ITEM_KEY;
2866 key.offset = (u64)-1;
2867 iter->bytenr = bytenr;
2869 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
2874 * Key with offset -1 found, there would have to exist an extent
2875 * item with such offset, but this is out of the valid range.
2880 if (path->slots[0] == 0) {
2881 WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG));
2887 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
2888 if ((key.type != BTRFS_EXTENT_ITEM_KEY &&
2889 key.type != BTRFS_METADATA_ITEM_KEY) || key.objectid != bytenr) {
2893 memcpy(&iter->cur_key, &key, sizeof(key));
2894 iter->item_ptr = (u32)btrfs_item_ptr_offset(path->nodes[0],
2896 iter->end_ptr = (u32)(iter->item_ptr +
2897 btrfs_item_size(path->nodes[0], path->slots[0]));
2898 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
2899 struct btrfs_extent_item);
2902 * Only support iteration on tree backref yet.
2904 * This is an extra precaution for non skinny-metadata, where
2905 * EXTENT_ITEM is also used for tree blocks, that we can only use
2906 * extent flags to determine if it's a tree block.
2908 if (btrfs_extent_flags(path->nodes[0], ei) & BTRFS_EXTENT_FLAG_DATA) {
2912 iter->cur_ptr = (u32)(iter->item_ptr + sizeof(*ei));
2914 /* If there is no inline backref, go search for keyed backref */
2915 if (iter->cur_ptr >= iter->end_ptr) {
2916 ret = btrfs_next_item(extent_root, path);
2918 /* No inline nor keyed ref */
2926 btrfs_item_key_to_cpu(path->nodes[0], &iter->cur_key,
2928 if (iter->cur_key.objectid != bytenr ||
2929 (iter->cur_key.type != BTRFS_SHARED_BLOCK_REF_KEY &&
2930 iter->cur_key.type != BTRFS_TREE_BLOCK_REF_KEY)) {
2934 iter->cur_ptr = (u32)btrfs_item_ptr_offset(path->nodes[0],
2936 iter->item_ptr = iter->cur_ptr;
2937 iter->end_ptr = (u32)(iter->item_ptr + btrfs_item_size(
2938 path->nodes[0], path->slots[0]));
2943 btrfs_backref_iter_release(iter);
2947 static bool btrfs_backref_iter_is_inline_ref(struct btrfs_backref_iter *iter)
2949 if (iter->cur_key.type == BTRFS_EXTENT_ITEM_KEY ||
2950 iter->cur_key.type == BTRFS_METADATA_ITEM_KEY)
2956 * Go to the next backref item of current bytenr, can be either inlined or
2959 * Caller needs to check whether it's inline ref or not by iter->cur_key.
2961 * Return 0 if we get next backref without problem.
2962 * Return >0 if there is no extra backref for this bytenr.
2963 * Return <0 if there is something wrong happened.
2965 int btrfs_backref_iter_next(struct btrfs_backref_iter *iter)
2967 struct extent_buffer *eb = iter->path->nodes[0];
2968 struct btrfs_root *extent_root;
2969 struct btrfs_path *path = iter->path;
2970 struct btrfs_extent_inline_ref *iref;
2974 if (btrfs_backref_iter_is_inline_ref(iter)) {
2975 /* We're still inside the inline refs */
2976 ASSERT(iter->cur_ptr < iter->end_ptr);
2978 if (btrfs_backref_has_tree_block_info(iter)) {
2979 /* First tree block info */
2980 size = sizeof(struct btrfs_tree_block_info);
2982 /* Use inline ref type to determine the size */
2985 iref = (struct btrfs_extent_inline_ref *)
2986 ((unsigned long)iter->cur_ptr);
2987 type = btrfs_extent_inline_ref_type(eb, iref);
2989 size = btrfs_extent_inline_ref_size(type);
2991 iter->cur_ptr += size;
2992 if (iter->cur_ptr < iter->end_ptr)
2995 /* All inline items iterated, fall through */
2998 /* We're at keyed items, there is no inline item, go to the next one */
2999 extent_root = btrfs_extent_root(iter->fs_info, iter->bytenr);
3000 ret = btrfs_next_item(extent_root, iter->path);
3004 btrfs_item_key_to_cpu(path->nodes[0], &iter->cur_key, path->slots[0]);
3005 if (iter->cur_key.objectid != iter->bytenr ||
3006 (iter->cur_key.type != BTRFS_TREE_BLOCK_REF_KEY &&
3007 iter->cur_key.type != BTRFS_SHARED_BLOCK_REF_KEY))
3009 iter->item_ptr = (u32)btrfs_item_ptr_offset(path->nodes[0],
3011 iter->cur_ptr = iter->item_ptr;
3012 iter->end_ptr = iter->item_ptr + (u32)btrfs_item_size(path->nodes[0],
3017 void btrfs_backref_init_cache(struct btrfs_fs_info *fs_info,
3018 struct btrfs_backref_cache *cache, bool is_reloc)
3022 cache->rb_root = RB_ROOT;
3023 for (i = 0; i < BTRFS_MAX_LEVEL; i++)
3024 INIT_LIST_HEAD(&cache->pending[i]);
3025 INIT_LIST_HEAD(&cache->changed);
3026 INIT_LIST_HEAD(&cache->detached);
3027 INIT_LIST_HEAD(&cache->leaves);
3028 INIT_LIST_HEAD(&cache->pending_edge);
3029 INIT_LIST_HEAD(&cache->useless_node);
3030 cache->fs_info = fs_info;
3031 cache->is_reloc = is_reloc;
3034 struct btrfs_backref_node *btrfs_backref_alloc_node(
3035 struct btrfs_backref_cache *cache, u64 bytenr, int level)
3037 struct btrfs_backref_node *node;
3039 ASSERT(level >= 0 && level < BTRFS_MAX_LEVEL);
3040 node = kzalloc(sizeof(*node), GFP_NOFS);
3044 INIT_LIST_HEAD(&node->list);
3045 INIT_LIST_HEAD(&node->upper);
3046 INIT_LIST_HEAD(&node->lower);
3047 RB_CLEAR_NODE(&node->rb_node);
3049 node->level = level;
3050 node->bytenr = bytenr;
3055 void btrfs_backref_free_node(struct btrfs_backref_cache *cache,
3056 struct btrfs_backref_node *node)
3059 ASSERT(list_empty(&node->list));
3060 ASSERT(list_empty(&node->lower));
3061 ASSERT(node->eb == NULL);
3063 btrfs_put_root(node->root);
3068 struct btrfs_backref_edge *btrfs_backref_alloc_edge(
3069 struct btrfs_backref_cache *cache)
3071 struct btrfs_backref_edge *edge;
3073 edge = kzalloc(sizeof(*edge), GFP_NOFS);
3079 void btrfs_backref_free_edge(struct btrfs_backref_cache *cache,
3080 struct btrfs_backref_edge *edge)
3088 void btrfs_backref_unlock_node_buffer(struct btrfs_backref_node *node)
3091 btrfs_tree_unlock(node->eb);
3096 void btrfs_backref_drop_node_buffer(struct btrfs_backref_node *node)
3099 btrfs_backref_unlock_node_buffer(node);
3100 free_extent_buffer(node->eb);
3106 * Drop the backref node from cache without cleaning up its children
3109 * This can only be called on node without parent edges.
3110 * The children edges are still kept as is.
3112 void btrfs_backref_drop_node(struct btrfs_backref_cache *tree,
3113 struct btrfs_backref_node *node)
3115 ASSERT(list_empty(&node->upper));
3117 btrfs_backref_drop_node_buffer(node);
3118 list_del_init(&node->list);
3119 list_del_init(&node->lower);
3120 if (!RB_EMPTY_NODE(&node->rb_node))
3121 rb_erase(&node->rb_node, &tree->rb_root);
3122 btrfs_backref_free_node(tree, node);
3126 * Drop the backref node from cache, also cleaning up all its
3127 * upper edges and any uncached nodes in the path.
3129 * This cleanup happens bottom up, thus the node should either
3130 * be the lowest node in the cache or a detached node.
3132 void btrfs_backref_cleanup_node(struct btrfs_backref_cache *cache,
3133 struct btrfs_backref_node *node)
3135 struct btrfs_backref_node *upper;
3136 struct btrfs_backref_edge *edge;
3141 BUG_ON(!node->lowest && !node->detached);
3142 while (!list_empty(&node->upper)) {
3143 edge = list_entry(node->upper.next, struct btrfs_backref_edge,
3145 upper = edge->node[UPPER];
3146 list_del(&edge->list[LOWER]);
3147 list_del(&edge->list[UPPER]);
3148 btrfs_backref_free_edge(cache, edge);
3151 * Add the node to leaf node list if no other child block
3154 if (list_empty(&upper->lower)) {
3155 list_add_tail(&upper->lower, &cache->leaves);
3160 btrfs_backref_drop_node(cache, node);
3164 * Release all nodes/edges from current cache
3166 void btrfs_backref_release_cache(struct btrfs_backref_cache *cache)
3168 struct btrfs_backref_node *node;
3171 while (!list_empty(&cache->detached)) {
3172 node = list_entry(cache->detached.next,
3173 struct btrfs_backref_node, list);
3174 btrfs_backref_cleanup_node(cache, node);
3177 while (!list_empty(&cache->leaves)) {
3178 node = list_entry(cache->leaves.next,
3179 struct btrfs_backref_node, lower);
3180 btrfs_backref_cleanup_node(cache, node);
3183 for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
3184 while (!list_empty(&cache->pending[i])) {
3185 node = list_first_entry(&cache->pending[i],
3186 struct btrfs_backref_node,
3188 btrfs_backref_cleanup_node(cache, node);
3191 ASSERT(list_empty(&cache->pending_edge));
3192 ASSERT(list_empty(&cache->useless_node));
3193 ASSERT(list_empty(&cache->changed));
3194 ASSERT(list_empty(&cache->detached));
3195 ASSERT(RB_EMPTY_ROOT(&cache->rb_root));
3196 ASSERT(!cache->nr_nodes);
3197 ASSERT(!cache->nr_edges);
3200 void btrfs_backref_link_edge(struct btrfs_backref_edge *edge,
3201 struct btrfs_backref_node *lower,
3202 struct btrfs_backref_node *upper,
3205 ASSERT(upper && lower && upper->level == lower->level + 1);
3206 edge->node[LOWER] = lower;
3207 edge->node[UPPER] = upper;
3208 if (link_which & LINK_LOWER)
3209 list_add_tail(&edge->list[LOWER], &lower->upper);
3210 if (link_which & LINK_UPPER)
3211 list_add_tail(&edge->list[UPPER], &upper->lower);
3214 * Handle direct tree backref
3216 * Direct tree backref means, the backref item shows its parent bytenr
3217 * directly. This is for SHARED_BLOCK_REF backref (keyed or inlined).
3219 * @ref_key: The converted backref key.
3220 * For keyed backref, it's the item key.
3221 * For inlined backref, objectid is the bytenr,
3222 * type is btrfs_inline_ref_type, offset is
3223 * btrfs_inline_ref_offset.
3225 static int handle_direct_tree_backref(struct btrfs_backref_cache *cache,
3226 struct btrfs_key *ref_key,
3227 struct btrfs_backref_node *cur)
3229 struct btrfs_backref_edge *edge;
3230 struct btrfs_backref_node *upper;
3231 struct rb_node *rb_node;
3233 ASSERT(ref_key->type == BTRFS_SHARED_BLOCK_REF_KEY);
3235 /* Only reloc root uses backref pointing to itself */
3236 if (ref_key->objectid == ref_key->offset) {
3237 struct btrfs_root *root;
3239 cur->is_reloc_root = 1;
3240 /* Only reloc backref cache cares about a specific root */
3241 if (cache->is_reloc) {
3242 root = find_reloc_root(cache->fs_info, cur->bytenr);
3248 * For generic purpose backref cache, reloc root node
3251 list_add(&cur->list, &cache->useless_node);
3256 edge = btrfs_backref_alloc_edge(cache);
3260 rb_node = rb_simple_search(&cache->rb_root, ref_key->offset);
3262 /* Parent node not yet cached */
3263 upper = btrfs_backref_alloc_node(cache, ref_key->offset,
3266 btrfs_backref_free_edge(cache, edge);
3271 * Backrefs for the upper level block isn't cached, add the
3272 * block to pending list
3274 list_add_tail(&edge->list[UPPER], &cache->pending_edge);
3276 /* Parent node already cached */
3277 upper = rb_entry(rb_node, struct btrfs_backref_node, rb_node);
3278 ASSERT(upper->checked);
3279 INIT_LIST_HEAD(&edge->list[UPPER]);
3281 btrfs_backref_link_edge(edge, cur, upper, LINK_LOWER);
3286 * Handle indirect tree backref
3288 * Indirect tree backref means, we only know which tree the node belongs to.
3289 * We still need to do a tree search to find out the parents. This is for
3290 * TREE_BLOCK_REF backref (keyed or inlined).
3292 * @trans: Transaction handle.
3293 * @ref_key: The same as @ref_key in handle_direct_tree_backref()
3294 * @tree_key: The first key of this tree block.
3295 * @path: A clean (released) path, to avoid allocating path every time
3296 * the function get called.
3298 static int handle_indirect_tree_backref(struct btrfs_trans_handle *trans,
3299 struct btrfs_backref_cache *cache,
3300 struct btrfs_path *path,
3301 struct btrfs_key *ref_key,
3302 struct btrfs_key *tree_key,
3303 struct btrfs_backref_node *cur)
3305 struct btrfs_fs_info *fs_info = cache->fs_info;
3306 struct btrfs_backref_node *upper;
3307 struct btrfs_backref_node *lower;
3308 struct btrfs_backref_edge *edge;
3309 struct extent_buffer *eb;
3310 struct btrfs_root *root;
3311 struct rb_node *rb_node;
3313 bool need_check = true;
3316 root = btrfs_get_fs_root(fs_info, ref_key->offset, false);
3318 return PTR_ERR(root);
3319 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
3322 if (btrfs_root_level(&root->root_item) == cur->level) {
3324 ASSERT(btrfs_root_bytenr(&root->root_item) == cur->bytenr);
3326 * For reloc backref cache, we may ignore reloc root. But for
3327 * general purpose backref cache, we can't rely on
3328 * btrfs_should_ignore_reloc_root() as it may conflict with
3329 * current running relocation and lead to missing root.
3331 * For general purpose backref cache, reloc root detection is
3332 * completely relying on direct backref (key->offset is parent
3333 * bytenr), thus only do such check for reloc cache.
3335 if (btrfs_should_ignore_reloc_root(root) && cache->is_reloc) {
3336 btrfs_put_root(root);
3337 list_add(&cur->list, &cache->useless_node);
3344 level = cur->level + 1;
3346 /* Search the tree to find parent blocks referring to the block */
3347 path->search_commit_root = 1;
3348 path->skip_locking = 1;
3349 path->lowest_level = level;
3350 ret = btrfs_search_slot(NULL, root, tree_key, path, 0, 0);
3351 path->lowest_level = 0;
3353 btrfs_put_root(root);
3356 if (ret > 0 && path->slots[level] > 0)
3357 path->slots[level]--;
3359 eb = path->nodes[level];
3360 if (btrfs_node_blockptr(eb, path->slots[level]) != cur->bytenr) {
3362 "couldn't find block (%llu) (level %d) in tree (%llu) with key (%llu %u %llu)",
3363 cur->bytenr, level - 1, btrfs_root_id(root),
3364 tree_key->objectid, tree_key->type, tree_key->offset);
3365 btrfs_put_root(root);
3371 /* Add all nodes and edges in the path */
3372 for (; level < BTRFS_MAX_LEVEL; level++) {
3373 if (!path->nodes[level]) {
3374 ASSERT(btrfs_root_bytenr(&root->root_item) ==
3376 /* Same as previous should_ignore_reloc_root() call */
3377 if (btrfs_should_ignore_reloc_root(root) &&
3379 btrfs_put_root(root);
3380 list_add(&lower->list, &cache->useless_node);
3387 edge = btrfs_backref_alloc_edge(cache);
3389 btrfs_put_root(root);
3394 eb = path->nodes[level];
3395 rb_node = rb_simple_search(&cache->rb_root, eb->start);
3397 upper = btrfs_backref_alloc_node(cache, eb->start,
3400 btrfs_put_root(root);
3401 btrfs_backref_free_edge(cache, edge);
3405 upper->owner = btrfs_header_owner(eb);
3406 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
3410 * If we know the block isn't shared we can avoid
3411 * checking its backrefs.
3413 if (btrfs_block_can_be_shared(trans, root, eb))
3419 * Add the block to pending list if we need to check its
3420 * backrefs, we only do this once while walking up a
3421 * tree as we will catch anything else later on.
3423 if (!upper->checked && need_check) {
3425 list_add_tail(&edge->list[UPPER],
3426 &cache->pending_edge);
3430 INIT_LIST_HEAD(&edge->list[UPPER]);
3433 upper = rb_entry(rb_node, struct btrfs_backref_node,
3435 ASSERT(upper->checked);
3436 INIT_LIST_HEAD(&edge->list[UPPER]);
3438 upper->owner = btrfs_header_owner(eb);
3440 btrfs_backref_link_edge(edge, lower, upper, LINK_LOWER);
3443 btrfs_put_root(root);
3450 btrfs_release_path(path);
3455 * Add backref node @cur into @cache.
3457 * NOTE: Even if the function returned 0, @cur is not yet cached as its upper
3458 * links aren't yet bi-directional. Needs to finish such links.
3459 * Use btrfs_backref_finish_upper_links() to finish such linkage.
3461 * @trans: Transaction handle.
3462 * @path: Released path for indirect tree backref lookup
3463 * @iter: Released backref iter for extent tree search
3464 * @node_key: The first key of the tree block
3466 int btrfs_backref_add_tree_node(struct btrfs_trans_handle *trans,
3467 struct btrfs_backref_cache *cache,
3468 struct btrfs_path *path,
3469 struct btrfs_backref_iter *iter,
3470 struct btrfs_key *node_key,
3471 struct btrfs_backref_node *cur)
3473 struct btrfs_backref_edge *edge;
3474 struct btrfs_backref_node *exist;
3477 ret = btrfs_backref_iter_start(iter, cur->bytenr);
3481 * We skip the first btrfs_tree_block_info, as we don't use the key
3482 * stored in it, but fetch it from the tree block
3484 if (btrfs_backref_has_tree_block_info(iter)) {
3485 ret = btrfs_backref_iter_next(iter);
3488 /* No extra backref? This means the tree block is corrupted */
3494 WARN_ON(cur->checked);
3495 if (!list_empty(&cur->upper)) {
3497 * The backref was added previously when processing backref of
3498 * type BTRFS_TREE_BLOCK_REF_KEY
3500 ASSERT(list_is_singular(&cur->upper));
3501 edge = list_entry(cur->upper.next, struct btrfs_backref_edge,
3503 ASSERT(list_empty(&edge->list[UPPER]));
3504 exist = edge->node[UPPER];
3506 * Add the upper level block to pending list if we need check
3509 if (!exist->checked)
3510 list_add_tail(&edge->list[UPPER], &cache->pending_edge);
3515 for (; ret == 0; ret = btrfs_backref_iter_next(iter)) {
3516 struct extent_buffer *eb;
3517 struct btrfs_key key;
3521 eb = iter->path->nodes[0];
3523 key.objectid = iter->bytenr;
3524 if (btrfs_backref_iter_is_inline_ref(iter)) {
3525 struct btrfs_extent_inline_ref *iref;
3527 /* Update key for inline backref */
3528 iref = (struct btrfs_extent_inline_ref *)
3529 ((unsigned long)iter->cur_ptr);
3530 type = btrfs_get_extent_inline_ref_type(eb, iref,
3531 BTRFS_REF_TYPE_BLOCK);
3532 if (type == BTRFS_REF_TYPE_INVALID) {
3537 key.offset = btrfs_extent_inline_ref_offset(eb, iref);
3539 key.type = iter->cur_key.type;
3540 key.offset = iter->cur_key.offset;
3544 * Parent node found and matches current inline ref, no need to
3545 * rebuild this node for this inline ref
3548 ((key.type == BTRFS_TREE_BLOCK_REF_KEY &&
3549 exist->owner == key.offset) ||
3550 (key.type == BTRFS_SHARED_BLOCK_REF_KEY &&
3551 exist->bytenr == key.offset))) {
3556 /* SHARED_BLOCK_REF means key.offset is the parent bytenr */
3557 if (key.type == BTRFS_SHARED_BLOCK_REF_KEY) {
3558 ret = handle_direct_tree_backref(cache, &key, cur);
3561 } else if (key.type == BTRFS_TREE_BLOCK_REF_KEY) {
3563 * key.type == BTRFS_TREE_BLOCK_REF_KEY, inline ref
3564 * offset means the root objectid. We need to search
3565 * the tree to get its parent bytenr.
3567 ret = handle_indirect_tree_backref(trans, cache, path,
3568 &key, node_key, cur);
3573 * Unrecognized tree backref items (if it can pass tree-checker)
3581 btrfs_backref_iter_release(iter);
3586 * Finish the upwards linkage created by btrfs_backref_add_tree_node()
3588 int btrfs_backref_finish_upper_links(struct btrfs_backref_cache *cache,
3589 struct btrfs_backref_node *start)
3591 struct list_head *useless_node = &cache->useless_node;
3592 struct btrfs_backref_edge *edge;
3593 struct rb_node *rb_node;
3594 LIST_HEAD(pending_edge);
3596 ASSERT(start->checked);
3598 /* Insert this node to cache if it's not COW-only */
3599 if (!start->cowonly) {
3600 rb_node = rb_simple_insert(&cache->rb_root, start->bytenr,
3603 btrfs_backref_panic(cache->fs_info, start->bytenr,
3605 list_add_tail(&start->lower, &cache->leaves);
3609 * Use breadth first search to iterate all related edges.
3611 * The starting points are all the edges of this node
3613 list_for_each_entry(edge, &start->upper, list[LOWER])
3614 list_add_tail(&edge->list[UPPER], &pending_edge);
3616 while (!list_empty(&pending_edge)) {
3617 struct btrfs_backref_node *upper;
3618 struct btrfs_backref_node *lower;
3620 edge = list_first_entry(&pending_edge,
3621 struct btrfs_backref_edge, list[UPPER]);
3622 list_del_init(&edge->list[UPPER]);
3623 upper = edge->node[UPPER];
3624 lower = edge->node[LOWER];
3626 /* Parent is detached, no need to keep any edges */
3627 if (upper->detached) {
3628 list_del(&edge->list[LOWER]);
3629 btrfs_backref_free_edge(cache, edge);
3631 /* Lower node is orphan, queue for cleanup */
3632 if (list_empty(&lower->upper))
3633 list_add(&lower->list, useless_node);
3638 * All new nodes added in current build_backref_tree() haven't
3639 * been linked to the cache rb tree.
3640 * So if we have upper->rb_node populated, this means a cache
3641 * hit. We only need to link the edge, as @upper and all its
3642 * parents have already been linked.
3644 if (!RB_EMPTY_NODE(&upper->rb_node)) {
3645 if (upper->lowest) {
3646 list_del_init(&upper->lower);
3650 list_add_tail(&edge->list[UPPER], &upper->lower);
3654 /* Sanity check, we shouldn't have any unchecked nodes */
3655 if (!upper->checked) {
3660 /* Sanity check, COW-only node has non-COW-only parent */
3661 if (start->cowonly != upper->cowonly) {
3666 /* Only cache non-COW-only (subvolume trees) tree blocks */
3667 if (!upper->cowonly) {
3668 rb_node = rb_simple_insert(&cache->rb_root, upper->bytenr,
3671 btrfs_backref_panic(cache->fs_info,
3672 upper->bytenr, -EEXIST);
3677 list_add_tail(&edge->list[UPPER], &upper->lower);
3680 * Also queue all the parent edges of this uncached node
3681 * to finish the upper linkage
3683 list_for_each_entry(edge, &upper->upper, list[LOWER])
3684 list_add_tail(&edge->list[UPPER], &pending_edge);
3689 void btrfs_backref_error_cleanup(struct btrfs_backref_cache *cache,
3690 struct btrfs_backref_node *node)
3692 struct btrfs_backref_node *lower;
3693 struct btrfs_backref_node *upper;
3694 struct btrfs_backref_edge *edge;
3696 while (!list_empty(&cache->useless_node)) {
3697 lower = list_first_entry(&cache->useless_node,
3698 struct btrfs_backref_node, list);
3699 list_del_init(&lower->list);
3701 while (!list_empty(&cache->pending_edge)) {
3702 edge = list_first_entry(&cache->pending_edge,
3703 struct btrfs_backref_edge, list[UPPER]);
3704 list_del(&edge->list[UPPER]);
3705 list_del(&edge->list[LOWER]);
3706 lower = edge->node[LOWER];
3707 upper = edge->node[UPPER];
3708 btrfs_backref_free_edge(cache, edge);
3711 * Lower is no longer linked to any upper backref nodes and
3712 * isn't in the cache, we can free it ourselves.
3714 if (list_empty(&lower->upper) &&
3715 RB_EMPTY_NODE(&lower->rb_node))
3716 list_add(&lower->list, &cache->useless_node);
3718 if (!RB_EMPTY_NODE(&upper->rb_node))
3721 /* Add this guy's upper edges to the list to process */
3722 list_for_each_entry(edge, &upper->upper, list[LOWER])
3723 list_add_tail(&edge->list[UPPER],
3724 &cache->pending_edge);
3725 if (list_empty(&upper->upper))
3726 list_add(&upper->list, &cache->useless_node);
3729 while (!list_empty(&cache->useless_node)) {
3730 lower = list_first_entry(&cache->useless_node,
3731 struct btrfs_backref_node, list);
3732 list_del_init(&lower->list);
3735 btrfs_backref_drop_node(cache, lower);
3738 btrfs_backref_cleanup_node(cache, node);
3739 ASSERT(list_empty(&cache->useless_node) &&
3740 list_empty(&cache->pending_edge));