2 * Copyright (C) 2011 STRATO. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
23 #include "transaction.h"
24 #include "delayed-ref.h"
27 * this structure records all encountered refs on the way up to the root
30 struct list_head list;
39 static int __add_prelim_ref(struct list_head *head, u64 root_id,
40 struct btrfs_key *key, int level, u64 parent,
41 u64 wanted_disk_byte, int count)
43 struct __prelim_ref *ref;
45 /* in case we're adding delayed refs, we're holding the refs spinlock */
46 ref = kmalloc(sizeof(*ref), GFP_ATOMIC);
50 ref->root_id = root_id;
54 memset(&ref->key, 0, sizeof(ref->key));
59 ref->wanted_disk_byte = wanted_disk_byte;
60 list_add_tail(&ref->list, head);
65 static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
66 struct ulist *parents,
67 struct extent_buffer *eb, int level,
68 u64 wanted_objectid, u64 wanted_disk_byte)
72 struct btrfs_file_extent_item *fi;
77 ret = ulist_add(parents, eb->start, 0, GFP_NOFS);
85 * if the current leaf is full with EXTENT_DATA items, we must
86 * check the next one if that holds a reference as well.
87 * ref->count cannot be used to skip this check.
88 * repeat this until we don't find any additional EXTENT_DATA items.
91 ret = btrfs_next_leaf(root, path);
98 for (slot = 0; slot < btrfs_header_nritems(eb); ++slot) {
99 btrfs_item_key_to_cpu(eb, &key, slot);
100 if (key.objectid != wanted_objectid ||
101 key.type != BTRFS_EXTENT_DATA_KEY)
103 fi = btrfs_item_ptr(eb, slot,
104 struct btrfs_file_extent_item);
105 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
106 if (disk_byte == wanted_disk_byte)
115 * resolve an indirect backref in the form (root_id, key, level)
116 * to a logical address
118 static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
119 struct __prelim_ref *ref,
120 struct ulist *parents)
122 struct btrfs_path *path;
123 struct btrfs_root *root;
124 struct btrfs_key root_key;
125 struct btrfs_key key = {0};
126 struct extent_buffer *eb;
129 int level = ref->level;
131 path = btrfs_alloc_path();
135 root_key.objectid = ref->root_id;
136 root_key.type = BTRFS_ROOT_ITEM_KEY;
137 root_key.offset = (u64)-1;
138 root = btrfs_read_fs_root_no_name(fs_info, &root_key);
145 root_level = btrfs_header_level(root->node);
148 if (root_level + 1 == level)
151 path->lowest_level = level;
152 ret = btrfs_search_slot(NULL, root, &ref->key, path, 0, 0);
153 pr_debug("search slot in root %llu (level %d, ref count %d) returned "
154 "%d for key (%llu %u %llu)\n",
155 (unsigned long long)ref->root_id, level, ref->count, ret,
156 (unsigned long long)ref->key.objectid, ref->key.type,
157 (unsigned long long)ref->key.offset);
161 eb = path->nodes[level];
169 if (ret == 1 && path->slots[0] >= btrfs_header_nritems(eb)) {
170 ret = btrfs_next_leaf(root, path);
176 btrfs_item_key_to_cpu(eb, &key, path->slots[0]);
179 /* the last two parameters will only be used for level == 0 */
180 ret = add_all_parents(root, path, parents, eb, level, key.objectid,
181 ref->wanted_disk_byte);
183 btrfs_free_path(path);
188 * resolve all indirect backrefs from the list
190 static int __resolve_indirect_refs(struct btrfs_fs_info *fs_info,
191 struct list_head *head)
195 struct __prelim_ref *ref;
196 struct __prelim_ref *ref_safe;
197 struct __prelim_ref *new_ref;
198 struct ulist *parents;
199 struct ulist_node *node;
201 parents = ulist_alloc(GFP_NOFS);
206 * _safe allows us to insert directly after the current item without
207 * iterating over the newly inserted items.
208 * we're also allowed to re-assign ref during iteration.
210 list_for_each_entry_safe(ref, ref_safe, head, list) {
211 if (ref->parent) /* already direct */
215 err = __resolve_indirect_ref(fs_info, ref, parents);
222 /* we put the first parent into the ref at hand */
223 node = ulist_next(parents, NULL);
224 ref->parent = node ? node->val : 0;
226 /* additional parents require new refs being added here */
227 while ((node = ulist_next(parents, node))) {
228 new_ref = kmalloc(sizeof(*new_ref), GFP_NOFS);
233 memcpy(new_ref, ref, sizeof(*ref));
234 new_ref->parent = node->val;
235 list_add(&new_ref->list, &ref->list);
237 ulist_reinit(parents);
245 * merge two lists of backrefs and adjust counts accordingly
247 * mode = 1: merge identical keys, if key is set
248 * mode = 2: merge identical parents
250 static int __merge_refs(struct list_head *head, int mode)
252 struct list_head *pos1;
254 list_for_each(pos1, head) {
255 struct list_head *n2;
256 struct list_head *pos2;
257 struct __prelim_ref *ref1;
259 ref1 = list_entry(pos1, struct __prelim_ref, list);
261 if (mode == 1 && ref1->key.type == 0)
263 for (pos2 = pos1->next, n2 = pos2->next; pos2 != head;
264 pos2 = n2, n2 = pos2->next) {
265 struct __prelim_ref *ref2;
267 ref2 = list_entry(pos2, struct __prelim_ref, list);
270 if (memcmp(&ref1->key, &ref2->key,
271 sizeof(ref1->key)) ||
272 ref1->level != ref2->level ||
273 ref1->root_id != ref2->root_id)
275 ref1->count += ref2->count;
277 if (ref1->parent != ref2->parent)
279 ref1->count += ref2->count;
281 list_del(&ref2->list);
290 * add all currently queued delayed refs from this head whose seq nr is
291 * smaller or equal that seq to the list
293 static int __add_delayed_refs(struct btrfs_delayed_ref_head *head, u64 seq,
294 struct btrfs_key *info_key,
295 struct list_head *prefs)
297 struct btrfs_delayed_extent_op *extent_op = head->extent_op;
298 struct rb_node *n = &head->node.rb_node;
302 if (extent_op && extent_op->update_key)
303 btrfs_disk_key_to_cpu(info_key, &extent_op->key);
305 while ((n = rb_prev(n))) {
306 struct btrfs_delayed_ref_node *node;
307 node = rb_entry(n, struct btrfs_delayed_ref_node,
309 if (node->bytenr != head->node.bytenr)
311 WARN_ON(node->is_head);
316 switch (node->action) {
317 case BTRFS_ADD_DELAYED_EXTENT:
318 case BTRFS_UPDATE_DELAYED_HEAD:
321 case BTRFS_ADD_DELAYED_REF:
324 case BTRFS_DROP_DELAYED_REF:
330 switch (node->type) {
331 case BTRFS_TREE_BLOCK_REF_KEY: {
332 struct btrfs_delayed_tree_ref *ref;
334 ref = btrfs_delayed_node_to_tree_ref(node);
335 ret = __add_prelim_ref(prefs, ref->root, info_key,
336 ref->level + 1, 0, node->bytenr,
337 node->ref_mod * sgn);
340 case BTRFS_SHARED_BLOCK_REF_KEY: {
341 struct btrfs_delayed_tree_ref *ref;
343 ref = btrfs_delayed_node_to_tree_ref(node);
344 ret = __add_prelim_ref(prefs, ref->root, info_key,
345 ref->level + 1, ref->parent,
347 node->ref_mod * sgn);
350 case BTRFS_EXTENT_DATA_REF_KEY: {
351 struct btrfs_delayed_data_ref *ref;
352 struct btrfs_key key;
354 ref = btrfs_delayed_node_to_data_ref(node);
356 key.objectid = ref->objectid;
357 key.type = BTRFS_EXTENT_DATA_KEY;
358 key.offset = ref->offset;
359 ret = __add_prelim_ref(prefs, ref->root, &key, 0, 0,
361 node->ref_mod * sgn);
364 case BTRFS_SHARED_DATA_REF_KEY: {
365 struct btrfs_delayed_data_ref *ref;
366 struct btrfs_key key;
368 ref = btrfs_delayed_node_to_data_ref(node);
370 key.objectid = ref->objectid;
371 key.type = BTRFS_EXTENT_DATA_KEY;
372 key.offset = ref->offset;
373 ret = __add_prelim_ref(prefs, ref->root, &key, 0,
374 ref->parent, node->bytenr,
375 node->ref_mod * sgn);
388 * add all inline backrefs for bytenr to the list
390 static int __add_inline_refs(struct btrfs_fs_info *fs_info,
391 struct btrfs_path *path, u64 bytenr,
392 struct btrfs_key *info_key, int *info_level,
393 struct list_head *prefs)
397 struct extent_buffer *leaf;
398 struct btrfs_key key;
401 struct btrfs_extent_item *ei;
406 * enumerate all inline refs
408 leaf = path->nodes[0];
409 slot = path->slots[0] - 1;
411 item_size = btrfs_item_size_nr(leaf, slot);
412 BUG_ON(item_size < sizeof(*ei));
414 ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
415 flags = btrfs_extent_flags(leaf, ei);
417 ptr = (unsigned long)(ei + 1);
418 end = (unsigned long)ei + item_size;
420 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
421 struct btrfs_tree_block_info *info;
422 struct btrfs_disk_key disk_key;
424 info = (struct btrfs_tree_block_info *)ptr;
425 *info_level = btrfs_tree_block_level(leaf, info);
426 btrfs_tree_block_key(leaf, info, &disk_key);
427 btrfs_disk_key_to_cpu(info_key, &disk_key);
428 ptr += sizeof(struct btrfs_tree_block_info);
431 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
435 struct btrfs_extent_inline_ref *iref;
439 iref = (struct btrfs_extent_inline_ref *)ptr;
440 type = btrfs_extent_inline_ref_type(leaf, iref);
441 offset = btrfs_extent_inline_ref_offset(leaf, iref);
444 case BTRFS_SHARED_BLOCK_REF_KEY:
445 ret = __add_prelim_ref(prefs, 0, info_key,
446 *info_level + 1, offset,
449 case BTRFS_SHARED_DATA_REF_KEY: {
450 struct btrfs_shared_data_ref *sdref;
453 sdref = (struct btrfs_shared_data_ref *)(iref + 1);
454 count = btrfs_shared_data_ref_count(leaf, sdref);
455 ret = __add_prelim_ref(prefs, 0, NULL, 0, offset,
459 case BTRFS_TREE_BLOCK_REF_KEY:
460 ret = __add_prelim_ref(prefs, offset, info_key,
461 *info_level + 1, 0, bytenr, 1);
463 case BTRFS_EXTENT_DATA_REF_KEY: {
464 struct btrfs_extent_data_ref *dref;
468 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
469 count = btrfs_extent_data_ref_count(leaf, dref);
470 key.objectid = btrfs_extent_data_ref_objectid(leaf,
472 key.type = BTRFS_EXTENT_DATA_KEY;
473 key.offset = btrfs_extent_data_ref_offset(leaf, dref);
474 root = btrfs_extent_data_ref_root(leaf, dref);
475 ret = __add_prelim_ref(prefs, root, &key, 0, 0, bytenr,
483 ptr += btrfs_extent_inline_ref_size(type);
490 * add all non-inline backrefs for bytenr to the list
492 static int __add_keyed_refs(struct btrfs_fs_info *fs_info,
493 struct btrfs_path *path, u64 bytenr,
494 struct btrfs_key *info_key, int info_level,
495 struct list_head *prefs)
497 struct btrfs_root *extent_root = fs_info->extent_root;
500 struct extent_buffer *leaf;
501 struct btrfs_key key;
504 ret = btrfs_next_item(extent_root, path);
512 slot = path->slots[0];
513 leaf = path->nodes[0];
514 btrfs_item_key_to_cpu(leaf, &key, slot);
516 if (key.objectid != bytenr)
518 if (key.type < BTRFS_TREE_BLOCK_REF_KEY)
520 if (key.type > BTRFS_SHARED_DATA_REF_KEY)
524 case BTRFS_SHARED_BLOCK_REF_KEY:
525 ret = __add_prelim_ref(prefs, 0, info_key,
526 info_level + 1, key.offset,
529 case BTRFS_SHARED_DATA_REF_KEY: {
530 struct btrfs_shared_data_ref *sdref;
533 sdref = btrfs_item_ptr(leaf, slot,
534 struct btrfs_shared_data_ref);
535 count = btrfs_shared_data_ref_count(leaf, sdref);
536 ret = __add_prelim_ref(prefs, 0, NULL, 0, key.offset,
540 case BTRFS_TREE_BLOCK_REF_KEY:
541 ret = __add_prelim_ref(prefs, key.offset, info_key,
542 info_level + 1, 0, bytenr, 1);
544 case BTRFS_EXTENT_DATA_REF_KEY: {
545 struct btrfs_extent_data_ref *dref;
549 dref = btrfs_item_ptr(leaf, slot,
550 struct btrfs_extent_data_ref);
551 count = btrfs_extent_data_ref_count(leaf, dref);
552 key.objectid = btrfs_extent_data_ref_objectid(leaf,
554 key.type = BTRFS_EXTENT_DATA_KEY;
555 key.offset = btrfs_extent_data_ref_offset(leaf, dref);
556 root = btrfs_extent_data_ref_root(leaf, dref);
557 ret = __add_prelim_ref(prefs, root, &key, 0, 0,
571 * this adds all existing backrefs (inline backrefs, backrefs and delayed
572 * refs) for the given bytenr to the refs list, merges duplicates and resolves
573 * indirect refs to their parent bytenr.
574 * When roots are found, they're added to the roots list
576 * FIXME some caching might speed things up
578 static int find_parent_nodes(struct btrfs_trans_handle *trans,
579 struct btrfs_fs_info *fs_info, u64 bytenr,
580 u64 seq, struct ulist *refs, struct ulist *roots)
582 struct btrfs_key key;
583 struct btrfs_path *path;
584 struct btrfs_key info_key = { 0 };
585 struct btrfs_delayed_ref_root *delayed_refs = NULL;
586 struct btrfs_delayed_ref_head *head = NULL;
589 struct list_head prefs_delayed;
590 struct list_head prefs;
591 struct __prelim_ref *ref;
593 INIT_LIST_HEAD(&prefs);
594 INIT_LIST_HEAD(&prefs_delayed);
596 key.objectid = bytenr;
597 key.type = BTRFS_EXTENT_ITEM_KEY;
598 key.offset = (u64)-1;
600 path = btrfs_alloc_path();
605 * grab both a lock on the path and a lock on the delayed ref head.
606 * We need both to get a consistent picture of how the refs look
607 * at a specified point in time
610 ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
616 * look if there are updates for this ref queued and lock the head
618 delayed_refs = &trans->transaction->delayed_refs;
619 spin_lock(&delayed_refs->lock);
620 head = btrfs_find_delayed_ref_head(trans, bytenr);
622 if (!mutex_trylock(&head->mutex)) {
623 atomic_inc(&head->node.refs);
624 spin_unlock(&delayed_refs->lock);
626 btrfs_release_path(path);
629 * Mutex was contended, block until it's
630 * released and try again
632 mutex_lock(&head->mutex);
633 mutex_unlock(&head->mutex);
634 btrfs_put_delayed_ref(&head->node);
637 ret = __add_delayed_refs(head, seq, &info_key, &prefs_delayed);
641 spin_unlock(&delayed_refs->lock);
643 if (path->slots[0]) {
644 struct extent_buffer *leaf;
647 leaf = path->nodes[0];
648 slot = path->slots[0] - 1;
649 btrfs_item_key_to_cpu(leaf, &key, slot);
650 if (key.objectid == bytenr &&
651 key.type == BTRFS_EXTENT_ITEM_KEY) {
652 ret = __add_inline_refs(fs_info, path, bytenr,
653 &info_key, &info_level, &prefs);
656 ret = __add_keyed_refs(fs_info, path, bytenr, &info_key,
662 btrfs_release_path(path);
665 * when adding the delayed refs above, the info_key might not have
666 * been known yet. Go over the list and replace the missing keys
668 list_for_each_entry(ref, &prefs_delayed, list) {
669 if ((ref->key.offset | ref->key.type | ref->key.objectid) == 0)
670 memcpy(&ref->key, &info_key, sizeof(ref->key));
672 list_splice_init(&prefs_delayed, &prefs);
674 ret = __merge_refs(&prefs, 1);
678 ret = __resolve_indirect_refs(fs_info, &prefs);
682 ret = __merge_refs(&prefs, 2);
686 while (!list_empty(&prefs)) {
687 ref = list_first_entry(&prefs, struct __prelim_ref, list);
688 list_del(&ref->list);
691 if (ref->count && ref->root_id && ref->parent == 0) {
692 /* no parent == root of tree */
693 ret = ulist_add(roots, ref->root_id, 0, GFP_NOFS);
696 if (ref->count && ref->parent) {
697 ret = ulist_add(refs, ref->parent, 0, GFP_NOFS);
705 mutex_unlock(&head->mutex);
706 btrfs_free_path(path);
707 while (!list_empty(&prefs)) {
708 ref = list_first_entry(&prefs, struct __prelim_ref, list);
709 list_del(&ref->list);
712 while (!list_empty(&prefs_delayed)) {
713 ref = list_first_entry(&prefs_delayed, struct __prelim_ref,
715 list_del(&ref->list);
723 * Finds all leafs with a reference to the specified combination of bytenr and
724 * offset. key_list_head will point to a list of corresponding keys (caller must
725 * free each list element). The leafs will be stored in the leafs ulist, which
726 * must be freed with ulist_free.
728 * returns 0 on success, <0 on error
730 static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
731 struct btrfs_fs_info *fs_info, u64 bytenr,
732 u64 num_bytes, u64 seq, struct ulist **leafs)
737 tmp = ulist_alloc(GFP_NOFS);
740 *leafs = ulist_alloc(GFP_NOFS);
746 ret = find_parent_nodes(trans, fs_info, bytenr, seq, *leafs, tmp);
749 if (ret < 0 && ret != -ENOENT) {
758 * walk all backrefs for a given extent to find all roots that reference this
759 * extent. Walking a backref means finding all extents that reference this
760 * extent and in turn walk the backrefs of those, too. Naturally this is a
761 * recursive process, but here it is implemented in an iterative fashion: We
762 * find all referencing extents for the extent in question and put them on a
763 * list. In turn, we find all referencing extents for those, further appending
764 * to the list. The way we iterate the list allows adding more elements after
765 * the current while iterating. The process stops when we reach the end of the
766 * list. Found roots are added to the roots list.
768 * returns 0 on success, < 0 on error.
770 int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
771 struct btrfs_fs_info *fs_info, u64 bytenr,
772 u64 num_bytes, u64 seq, struct ulist **roots)
775 struct ulist_node *node = NULL;
778 tmp = ulist_alloc(GFP_NOFS);
781 *roots = ulist_alloc(GFP_NOFS);
788 ret = find_parent_nodes(trans, fs_info, bytenr, seq,
790 if (ret < 0 && ret != -ENOENT) {
795 node = ulist_next(tmp, node);
806 static int __inode_info(u64 inum, u64 ioff, u8 key_type,
807 struct btrfs_root *fs_root, struct btrfs_path *path,
808 struct btrfs_key *found_key)
811 struct btrfs_key key;
812 struct extent_buffer *eb;
818 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
823 if (ret && path->slots[0] >= btrfs_header_nritems(eb)) {
824 ret = btrfs_next_leaf(fs_root, path);
830 btrfs_item_key_to_cpu(eb, found_key, path->slots[0]);
831 if (found_key->type != key.type || found_key->objectid != key.objectid)
838 * this makes the path point to (inum INODE_ITEM ioff)
840 int inode_item_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
841 struct btrfs_path *path)
843 struct btrfs_key key;
844 return __inode_info(inum, ioff, BTRFS_INODE_ITEM_KEY, fs_root, path,
848 static int inode_ref_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
849 struct btrfs_path *path,
850 struct btrfs_key *found_key)
852 return __inode_info(inum, ioff, BTRFS_INODE_REF_KEY, fs_root, path,
857 * this iterates to turn a btrfs_inode_ref into a full filesystem path. elements
858 * of the path are separated by '/' and the path is guaranteed to be
859 * 0-terminated. the path is only given within the current file system.
860 * Therefore, it never starts with a '/'. the caller is responsible to provide
861 * "size" bytes in "dest". the dest buffer will be filled backwards. finally,
862 * the start point of the resulting string is returned. this pointer is within
864 * in case the path buffer would overflow, the pointer is decremented further
865 * as if output was written to the buffer, though no more output is actually
866 * generated. that way, the caller can determine how much space would be
867 * required for the path to fit into the buffer. in that case, the returned
868 * value will be smaller than dest. callers must check this!
870 static char *iref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
871 struct btrfs_inode_ref *iref,
872 struct extent_buffer *eb_in, u64 parent,
873 char *dest, u32 size)
879 s64 bytes_left = size - 1;
880 struct extent_buffer *eb = eb_in;
881 struct btrfs_key found_key;
884 dest[bytes_left] = '\0';
887 len = btrfs_inode_ref_name_len(eb, iref);
890 read_extent_buffer(eb, dest + bytes_left,
891 (unsigned long)(iref + 1), len);
893 free_extent_buffer(eb);
894 ret = inode_ref_info(parent, 0, fs_root, path, &found_key);
899 next_inum = found_key.offset;
901 /* regular exit ahead */
902 if (parent == next_inum)
905 slot = path->slots[0];
907 /* make sure we can use eb after releasing the path */
909 atomic_inc(&eb->refs);
910 btrfs_release_path(path);
912 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
916 dest[bytes_left] = '/';
919 btrfs_release_path(path);
924 return dest + bytes_left;
928 * this makes the path point to (logical EXTENT_ITEM *)
929 * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for
930 * tree blocks and <0 on error.
932 int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
933 struct btrfs_path *path, struct btrfs_key *found_key)
938 struct extent_buffer *eb;
939 struct btrfs_extent_item *ei;
940 struct btrfs_key key;
942 key.type = BTRFS_EXTENT_ITEM_KEY;
943 key.objectid = logical;
944 key.offset = (u64)-1;
946 ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
949 ret = btrfs_previous_item(fs_info->extent_root, path,
950 0, BTRFS_EXTENT_ITEM_KEY);
954 btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]);
955 if (found_key->type != BTRFS_EXTENT_ITEM_KEY ||
956 found_key->objectid > logical ||
957 found_key->objectid + found_key->offset <= logical) {
958 pr_debug("logical %llu is not within any extent\n",
959 (unsigned long long)logical);
964 item_size = btrfs_item_size_nr(eb, path->slots[0]);
965 BUG_ON(item_size < sizeof(*ei));
967 ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
968 flags = btrfs_extent_flags(eb, ei);
970 pr_debug("logical %llu is at position %llu within the extent (%llu "
971 "EXTENT_ITEM %llu) flags %#llx size %u\n",
972 (unsigned long long)logical,
973 (unsigned long long)(logical - found_key->objectid),
974 (unsigned long long)found_key->objectid,
975 (unsigned long long)found_key->offset,
976 (unsigned long long)flags, item_size);
977 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
978 return BTRFS_EXTENT_FLAG_TREE_BLOCK;
979 if (flags & BTRFS_EXTENT_FLAG_DATA)
980 return BTRFS_EXTENT_FLAG_DATA;
986 * helper function to iterate extent inline refs. ptr must point to a 0 value
987 * for the first call and may be modified. it is used to track state.
988 * if more refs exist, 0 is returned and the next call to
989 * __get_extent_inline_ref must pass the modified ptr parameter to get the
990 * next ref. after the last ref was processed, 1 is returned.
991 * returns <0 on error
993 static int __get_extent_inline_ref(unsigned long *ptr, struct extent_buffer *eb,
994 struct btrfs_extent_item *ei, u32 item_size,
995 struct btrfs_extent_inline_ref **out_eiref,
1000 struct btrfs_tree_block_info *info;
1004 flags = btrfs_extent_flags(eb, ei);
1005 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1006 info = (struct btrfs_tree_block_info *)(ei + 1);
1008 (struct btrfs_extent_inline_ref *)(info + 1);
1010 *out_eiref = (struct btrfs_extent_inline_ref *)(ei + 1);
1012 *ptr = (unsigned long)*out_eiref;
1013 if ((void *)*ptr >= (void *)ei + item_size)
1017 end = (unsigned long)ei + item_size;
1018 *out_eiref = (struct btrfs_extent_inline_ref *)*ptr;
1019 *out_type = btrfs_extent_inline_ref_type(eb, *out_eiref);
1021 *ptr += btrfs_extent_inline_ref_size(*out_type);
1022 WARN_ON(*ptr > end);
1024 return 1; /* last */
1030 * reads the tree block backref for an extent. tree level and root are returned
1031 * through out_level and out_root. ptr must point to a 0 value for the first
1032 * call and may be modified (see __get_extent_inline_ref comment).
1033 * returns 0 if data was provided, 1 if there was no more data to provide or
1036 int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
1037 struct btrfs_extent_item *ei, u32 item_size,
1038 u64 *out_root, u8 *out_level)
1042 struct btrfs_tree_block_info *info;
1043 struct btrfs_extent_inline_ref *eiref;
1045 if (*ptr == (unsigned long)-1)
1049 ret = __get_extent_inline_ref(ptr, eb, ei, item_size,
1054 if (type == BTRFS_TREE_BLOCK_REF_KEY ||
1055 type == BTRFS_SHARED_BLOCK_REF_KEY)
1062 /* we can treat both ref types equally here */
1063 info = (struct btrfs_tree_block_info *)(ei + 1);
1064 *out_root = btrfs_extent_inline_ref_offset(eb, eiref);
1065 *out_level = btrfs_tree_block_level(eb, info);
1068 *ptr = (unsigned long)-1;
1073 static int iterate_leaf_refs(struct btrfs_fs_info *fs_info,
1074 struct btrfs_path *path, u64 logical,
1075 u64 orig_extent_item_objectid,
1076 u64 extent_item_pos, u64 root,
1077 iterate_extent_inodes_t *iterate, void *ctx)
1080 struct btrfs_key key;
1081 struct btrfs_file_extent_item *fi;
1082 struct extent_buffer *eb;
1090 eb = read_tree_block(fs_info->tree_root, logical,
1091 fs_info->tree_root->leafsize, 0);
1096 * from the shared data ref, we only have the leaf but we need
1097 * the key. thus, we must look into all items and see that we
1098 * find one (some) with a reference to our extent item.
1100 nritems = btrfs_header_nritems(eb);
1101 for (slot = 0; slot < nritems; ++slot) {
1102 btrfs_item_key_to_cpu(eb, &key, slot);
1103 if (key.type != BTRFS_EXTENT_DATA_KEY)
1105 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
1106 extent_type = btrfs_file_extent_type(eb, fi);
1107 if (extent_type == BTRFS_FILE_EXTENT_INLINE)
1109 /* don't skip BTRFS_FILE_EXTENT_PREALLOC, we can handle that */
1110 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
1111 if (disk_byte != orig_extent_item_objectid)
1114 data_offset = btrfs_file_extent_offset(eb, fi);
1115 data_len = btrfs_file_extent_num_bytes(eb, fi);
1117 if (extent_item_pos < data_offset ||
1118 extent_item_pos >= data_offset + data_len)
1121 pr_debug("ref for %llu resolved, key (%llu EXTEND_DATA %llu), "
1122 "root %llu\n", orig_extent_item_objectid,
1123 key.objectid, key.offset, root);
1124 ret = iterate(key.objectid,
1125 key.offset + (extent_item_pos - data_offset),
1128 pr_debug("stopping iteration because ret=%d\n", ret);
1133 free_extent_buffer(eb);
1139 * calls iterate() for every inode that references the extent identified by
1140 * the given parameters.
1141 * when the iterator function returns a non-zero value, iteration stops.
1142 * path is guaranteed to be in released state when iterate() is called.
1144 int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
1145 struct btrfs_path *path,
1146 u64 extent_item_objectid, u64 extent_item_pos,
1147 iterate_extent_inodes_t *iterate, void *ctx)
1150 struct list_head data_refs = LIST_HEAD_INIT(data_refs);
1151 struct list_head shared_refs = LIST_HEAD_INIT(shared_refs);
1152 struct btrfs_trans_handle *trans;
1154 struct ulist *roots;
1155 struct ulist_node *ref_node = NULL;
1156 struct ulist_node *root_node = NULL;
1157 struct seq_list seq_elem;
1158 struct btrfs_delayed_ref_root *delayed_refs;
1160 trans = btrfs_join_transaction(fs_info->extent_root);
1162 return PTR_ERR(trans);
1164 pr_debug("resolving all inodes for extent %llu\n",
1165 extent_item_objectid);
1167 delayed_refs = &trans->transaction->delayed_refs;
1168 spin_lock(&delayed_refs->lock);
1169 btrfs_get_delayed_seq(delayed_refs, &seq_elem);
1170 spin_unlock(&delayed_refs->lock);
1172 ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
1173 extent_item_pos, seq_elem.seq,
1179 while (!ret && (ref_node = ulist_next(refs, ref_node))) {
1180 ret = btrfs_find_all_roots(trans, fs_info, ref_node->val, -1,
1181 seq_elem.seq, &roots);
1184 while (!ret && (root_node = ulist_next(roots, root_node))) {
1185 pr_debug("root %llu references leaf %llu\n",
1186 root_node->val, ref_node->val);
1187 ret = iterate_leaf_refs(fs_info, path, ref_node->val,
1188 extent_item_objectid,
1189 extent_item_pos, root_node->val,
1197 btrfs_put_delayed_seq(delayed_refs, &seq_elem);
1198 btrfs_end_transaction(trans, fs_info->extent_root);
1202 int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
1203 struct btrfs_path *path,
1204 iterate_extent_inodes_t *iterate, void *ctx)
1207 u64 extent_item_pos;
1208 struct btrfs_key found_key;
1210 ret = extent_from_logical(fs_info, logical, path,
1212 btrfs_release_path(path);
1213 if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1218 extent_item_pos = logical - found_key.objectid;
1219 ret = iterate_extent_inodes(fs_info, path, found_key.objectid,
1220 extent_item_pos, iterate, ctx);
1225 static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
1226 struct btrfs_path *path,
1227 iterate_irefs_t *iterate, void *ctx)
1236 struct extent_buffer *eb;
1237 struct btrfs_item *item;
1238 struct btrfs_inode_ref *iref;
1239 struct btrfs_key found_key;
1242 ret = inode_ref_info(inum, parent ? parent+1 : 0, fs_root, path,
1247 ret = found ? 0 : -ENOENT;
1252 parent = found_key.offset;
1253 slot = path->slots[0];
1254 eb = path->nodes[0];
1255 /* make sure we can use eb after releasing the path */
1256 atomic_inc(&eb->refs);
1257 btrfs_release_path(path);
1259 item = btrfs_item_nr(eb, slot);
1260 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
1262 for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) {
1263 name_len = btrfs_inode_ref_name_len(eb, iref);
1264 /* path must be released before calling iterate()! */
1265 pr_debug("following ref at offset %u for inode %llu in "
1267 (unsigned long long)found_key.objectid,
1268 (unsigned long long)fs_root->objectid);
1269 ret = iterate(parent, iref, eb, ctx);
1271 free_extent_buffer(eb);
1274 len = sizeof(*iref) + name_len;
1275 iref = (struct btrfs_inode_ref *)((char *)iref + len);
1277 free_extent_buffer(eb);
1280 btrfs_release_path(path);
1286 * returns 0 if the path could be dumped (probably truncated)
1287 * returns <0 in case of an error
1289 static int inode_to_path(u64 inum, struct btrfs_inode_ref *iref,
1290 struct extent_buffer *eb, void *ctx)
1292 struct inode_fs_paths *ipath = ctx;
1295 int i = ipath->fspath->elem_cnt;
1296 const int s_ptr = sizeof(char *);
1299 bytes_left = ipath->fspath->bytes_left > s_ptr ?
1300 ipath->fspath->bytes_left - s_ptr : 0;
1302 fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
1303 fspath = iref_to_path(ipath->fs_root, ipath->btrfs_path, iref, eb,
1304 inum, fspath_min, bytes_left);
1306 return PTR_ERR(fspath);
1308 if (fspath > fspath_min) {
1309 pr_debug("path resolved: %s\n", fspath);
1310 ipath->fspath->val[i] = (u64)(unsigned long)fspath;
1311 ++ipath->fspath->elem_cnt;
1312 ipath->fspath->bytes_left = fspath - fspath_min;
1314 pr_debug("missed path, not enough space. missing bytes: %lu, "
1315 "constructed so far: %s\n",
1316 (unsigned long)(fspath_min - fspath), fspath_min);
1317 ++ipath->fspath->elem_missed;
1318 ipath->fspath->bytes_missing += fspath_min - fspath;
1319 ipath->fspath->bytes_left = 0;
1326 * this dumps all file system paths to the inode into the ipath struct, provided
1327 * is has been created large enough. each path is zero-terminated and accessed
1328 * from ipath->fspath->val[i].
1329 * when it returns, there are ipath->fspath->elem_cnt number of paths available
1330 * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
1331 * number of missed paths in recored in ipath->fspath->elem_missed, otherwise,
1332 * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
1333 * have been needed to return all paths.
1335 int paths_from_inode(u64 inum, struct inode_fs_paths *ipath)
1337 return iterate_irefs(inum, ipath->fs_root, ipath->btrfs_path,
1338 inode_to_path, ipath);
1342 * allocates space to return multiple file system paths for an inode.
1343 * total_bytes to allocate are passed, note that space usable for actual path
1344 * information will be total_bytes - sizeof(struct inode_fs_paths).
1345 * the returned pointer must be freed with free_ipath() in the end.
1347 struct btrfs_data_container *init_data_container(u32 total_bytes)
1349 struct btrfs_data_container *data;
1352 alloc_bytes = max_t(size_t, total_bytes, sizeof(*data));
1353 data = kmalloc(alloc_bytes, GFP_NOFS);
1355 return ERR_PTR(-ENOMEM);
1357 if (total_bytes >= sizeof(*data)) {
1358 data->bytes_left = total_bytes - sizeof(*data);
1359 data->bytes_missing = 0;
1361 data->bytes_missing = sizeof(*data) - total_bytes;
1362 data->bytes_left = 0;
1366 data->elem_missed = 0;
1372 * allocates space to return multiple file system paths for an inode.
1373 * total_bytes to allocate are passed, note that space usable for actual path
1374 * information will be total_bytes - sizeof(struct inode_fs_paths).
1375 * the returned pointer must be freed with free_ipath() in the end.
1377 struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
1378 struct btrfs_path *path)
1380 struct inode_fs_paths *ifp;
1381 struct btrfs_data_container *fspath;
1383 fspath = init_data_container(total_bytes);
1385 return (void *)fspath;
1387 ifp = kmalloc(sizeof(*ifp), GFP_NOFS);
1390 return ERR_PTR(-ENOMEM);
1393 ifp->btrfs_path = path;
1394 ifp->fspath = fspath;
1395 ifp->fs_root = fs_root;
1400 void free_ipath(struct inode_fs_paths *ipath)