Btrfs: fix that error value is changed by mistake
[linux-3.10.git] / fs / btrfs / backref.c
index 2351df0..ff6475f 100644 (file)
 #include "ctree.h"
 #include "disk-io.h"
 #include "backref.h"
+#include "ulist.h"
+#include "transaction.h"
+#include "delayed-ref.h"
+#include "locking.h"
 
-struct __data_ref {
-       struct list_head list;
+struct extent_inode_elem {
        u64 inum;
-       u64 root;
-       u64 extent_data_item_offset;
+       u64 offset;
+       struct extent_inode_elem *next;
 };
 
-struct __shared_ref {
-       struct list_head list;
+static int check_extent_in_eb(struct btrfs_key *key, struct extent_buffer *eb,
+                               struct btrfs_file_extent_item *fi,
+                               u64 extent_item_pos,
+                               struct extent_inode_elem **eie)
+{
+       u64 data_offset;
+       u64 data_len;
+       struct extent_inode_elem *e;
+
+       data_offset = btrfs_file_extent_offset(eb, fi);
+       data_len = btrfs_file_extent_num_bytes(eb, fi);
+
+       if (extent_item_pos < data_offset ||
+           extent_item_pos >= data_offset + data_len)
+               return 1;
+
+       e = kmalloc(sizeof(*e), GFP_NOFS);
+       if (!e)
+               return -ENOMEM;
+
+       e->next = *eie;
+       e->inum = key->objectid;
+       e->offset = key->offset + (extent_item_pos - data_offset);
+       *eie = e;
+
+       return 0;
+}
+
+static int find_extent_in_eb(struct extent_buffer *eb, u64 wanted_disk_byte,
+                               u64 extent_item_pos,
+                               struct extent_inode_elem **eie)
+{
        u64 disk_byte;
+       struct btrfs_key key;
+       struct btrfs_file_extent_item *fi;
+       int slot;
+       int nritems;
+       int extent_type;
+       int ret;
+
+       /*
+        * from the shared data ref, we only have the leaf but we need
+        * the key. thus, we must look into all items and see that we
+        * find one (some) with a reference to our extent item.
+        */
+       nritems = btrfs_header_nritems(eb);
+       for (slot = 0; slot < nritems; ++slot) {
+               btrfs_item_key_to_cpu(eb, &key, slot);
+               if (key.type != BTRFS_EXTENT_DATA_KEY)
+                       continue;
+               fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
+               extent_type = btrfs_file_extent_type(eb, fi);
+               if (extent_type == BTRFS_FILE_EXTENT_INLINE)
+                       continue;
+               /* don't skip BTRFS_FILE_EXTENT_PREALLOC, we can handle that */
+               disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
+               if (disk_byte != wanted_disk_byte)
+                       continue;
+
+               ret = check_extent_in_eb(&key, eb, fi, extent_item_pos, eie);
+               if (ret < 0)
+                       return ret;
+       }
+
+       return 0;
+}
+
+/*
+ * this structure records all encountered refs on the way up to the root
+ */
+struct __prelim_ref {
+       struct list_head list;
+       u64 root_id;
+       struct btrfs_key key_for_search;
+       int level;
+       int count;
+       struct extent_inode_elem *inode_list;
+       u64 parent;
+       u64 wanted_disk_byte;
 };
 
+/*
+ * the rules for all callers of this function are:
+ * - obtaining the parent is the goal
+ * - if you add a key, you must know that it is a correct key
+ * - if you cannot add the parent or a correct key, then we will look into the
+ *   block later to set a correct key
+ *
+ * delayed refs
+ * ============
+ *        backref type | shared | indirect | shared | indirect
+ * information         |   tree |     tree |   data |     data
+ * --------------------+--------+----------+--------+----------
+ *      parent logical |    y   |     -    |    -   |     -
+ *      key to resolve |    -   |     y    |    y   |     y
+ *  tree block logical |    -   |     -    |    -   |     -
+ *  root for resolving |    y   |     y    |    y   |     y
+ *
+ * - column 1:       we've the parent -> done
+ * - column 2, 3, 4: we use the key to find the parent
+ *
+ * on disk refs (inline or keyed)
+ * ==============================
+ *        backref type | shared | indirect | shared | indirect
+ * information         |   tree |     tree |   data |     data
+ * --------------------+--------+----------+--------+----------
+ *      parent logical |    y   |     -    |    y   |     -
+ *      key to resolve |    -   |     -    |    -   |     y
+ *  tree block logical |    y   |     y    |    y   |     y
+ *  root for resolving |    -   |     y    |    y   |     y
+ *
+ * - column 1, 3: we've the parent -> done
+ * - column 2:    we take the first key from the block to find the parent
+ *                (see __add_missing_keys)
+ * - column 4:    we use the key to find the parent
+ *
+ * additional information that's available but not required to find the parent
+ * block might help in merging entries to gain some speed.
+ */
+
+static int __add_prelim_ref(struct list_head *head, u64 root_id,
+                           struct btrfs_key *key, int level,
+                           u64 parent, u64 wanted_disk_byte, int count)
+{
+       struct __prelim_ref *ref;
+
+       /* in case we're adding delayed refs, we're holding the refs spinlock */
+       ref = kmalloc(sizeof(*ref), GFP_ATOMIC);
+       if (!ref)
+               return -ENOMEM;
+
+       ref->root_id = root_id;
+       if (key)
+               ref->key_for_search = *key;
+       else
+               memset(&ref->key_for_search, 0, sizeof(ref->key_for_search));
+
+       ref->inode_list = NULL;
+       ref->level = level;
+       ref->count = count;
+       ref->parent = parent;
+       ref->wanted_disk_byte = wanted_disk_byte;
+       list_add_tail(&ref->list, head);
+
+       return 0;
+}
+
+static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
+                               struct ulist *parents, int level,
+                               struct btrfs_key *key_for_search, u64 time_seq,
+                               u64 wanted_disk_byte,
+                               const u64 *extent_item_pos)
+{
+       int ret = 0;
+       int slot;
+       struct extent_buffer *eb;
+       struct btrfs_key key;
+       struct btrfs_file_extent_item *fi;
+       struct extent_inode_elem *eie = NULL;
+       u64 disk_byte;
+
+       if (level != 0) {
+               eb = path->nodes[level];
+               ret = ulist_add(parents, eb->start, 0, GFP_NOFS);
+               if (ret < 0)
+                       return ret;
+               return 0;
+       }
+
+       /*
+        * We normally enter this function with the path already pointing to
+        * the first item to check. But sometimes, we may enter it with
+        * slot==nritems. In that case, go to the next leaf before we continue.
+        */
+       if (path->slots[0] >= btrfs_header_nritems(path->nodes[0]))
+               ret = btrfs_next_old_leaf(root, path, time_seq);
+
+       while (!ret) {
+               eb = path->nodes[0];
+               slot = path->slots[0];
+
+               btrfs_item_key_to_cpu(eb, &key, slot);
+
+               if (key.objectid != key_for_search->objectid ||
+                   key.type != BTRFS_EXTENT_DATA_KEY)
+                       break;
+
+               fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
+               disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
+
+               if (disk_byte == wanted_disk_byte) {
+                       eie = NULL;
+                       if (extent_item_pos) {
+                               ret = check_extent_in_eb(&key, eb, fi,
+                                               *extent_item_pos,
+                                               &eie);
+                               if (ret < 0)
+                                       break;
+                       }
+                       if (!ret) {
+                               ret = ulist_add(parents, eb->start,
+                                               (unsigned long)eie, GFP_NOFS);
+                               if (ret < 0)
+                                       break;
+                               if (!extent_item_pos) {
+                                       ret = btrfs_next_old_leaf(root, path,
+                                                       time_seq);
+                                       continue;
+                               }
+                       }
+               }
+               ret = btrfs_next_old_item(root, path, time_seq);
+       }
+
+       if (ret > 0)
+               ret = 0;
+       return ret;
+}
+
+/*
+ * resolve an indirect backref in the form (root_id, key, level)
+ * to a logical address
+ */
+static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
+                                       int search_commit_root,
+                                       u64 time_seq,
+                                       struct __prelim_ref *ref,
+                                       struct ulist *parents,
+                                       const u64 *extent_item_pos)
+{
+       struct btrfs_path *path;
+       struct btrfs_root *root;
+       struct btrfs_key root_key;
+       struct extent_buffer *eb;
+       int ret = 0;
+       int root_level;
+       int level = ref->level;
+
+       path = btrfs_alloc_path();
+       if (!path)
+               return -ENOMEM;
+       path->search_commit_root = !!search_commit_root;
+
+       root_key.objectid = ref->root_id;
+       root_key.type = BTRFS_ROOT_ITEM_KEY;
+       root_key.offset = (u64)-1;
+       root = btrfs_read_fs_root_no_name(fs_info, &root_key);
+       if (IS_ERR(root)) {
+               ret = PTR_ERR(root);
+               goto out;
+       }
+
+       rcu_read_lock();
+       root_level = btrfs_header_level(root->node);
+       rcu_read_unlock();
+
+       if (root_level + 1 == level)
+               goto out;
+
+       path->lowest_level = level;
+       ret = btrfs_search_old_slot(root, &ref->key_for_search, path, time_seq);
+       pr_debug("search slot in root %llu (level %d, ref count %d) returned "
+                "%d for key (%llu %u %llu)\n",
+                (unsigned long long)ref->root_id, level, ref->count, ret,
+                (unsigned long long)ref->key_for_search.objectid,
+                ref->key_for_search.type,
+                (unsigned long long)ref->key_for_search.offset);
+       if (ret < 0)
+               goto out;
+
+       eb = path->nodes[level];
+       while (!eb) {
+               if (!level) {
+                       WARN_ON(1);
+                       ret = 1;
+                       goto out;
+               }
+               level--;
+               eb = path->nodes[level];
+       }
+
+       ret = add_all_parents(root, path, parents, level, &ref->key_for_search,
+                               time_seq, ref->wanted_disk_byte,
+                               extent_item_pos);
+out:
+       btrfs_free_path(path);
+       return ret;
+}
+
+/*
+ * resolve all indirect backrefs from the list
+ */
+static int __resolve_indirect_refs(struct btrfs_fs_info *fs_info,
+                                  int search_commit_root, u64 time_seq,
+                                  struct list_head *head,
+                                  const u64 *extent_item_pos)
+{
+       int err;
+       int ret = 0;
+       struct __prelim_ref *ref;
+       struct __prelim_ref *ref_safe;
+       struct __prelim_ref *new_ref;
+       struct ulist *parents;
+       struct ulist_node *node;
+       struct ulist_iterator uiter;
+
+       parents = ulist_alloc(GFP_NOFS);
+       if (!parents)
+               return -ENOMEM;
+
+       /*
+        * _safe allows us to insert directly after the current item without
+        * iterating over the newly inserted items.
+        * we're also allowed to re-assign ref during iteration.
+        */
+       list_for_each_entry_safe(ref, ref_safe, head, list) {
+               if (ref->parent)        /* already direct */
+                       continue;
+               if (ref->count == 0)
+                       continue;
+               err = __resolve_indirect_ref(fs_info, search_commit_root,
+                                            time_seq, ref, parents,
+                                            extent_item_pos);
+               if (err) {
+                       if (ret == 0)
+                               ret = err;
+                       continue;
+               }
+
+               /* we put the first parent into the ref at hand */
+               ULIST_ITER_INIT(&uiter);
+               node = ulist_next(parents, &uiter);
+               ref->parent = node ? node->val : 0;
+               ref->inode_list =
+                       node ? (struct extent_inode_elem *)node->aux : 0;
+
+               /* additional parents require new refs being added here */
+               while ((node = ulist_next(parents, &uiter))) {
+                       new_ref = kmalloc(sizeof(*new_ref), GFP_NOFS);
+                       if (!new_ref) {
+                               ret = -ENOMEM;
+                               break;
+                       }
+                       memcpy(new_ref, ref, sizeof(*ref));
+                       new_ref->parent = node->val;
+                       new_ref->inode_list =
+                                       (struct extent_inode_elem *)node->aux;
+                       list_add(&new_ref->list, &ref->list);
+               }
+               ulist_reinit(parents);
+       }
+
+       ulist_free(parents);
+       return ret;
+}
+
+static inline int ref_for_same_block(struct __prelim_ref *ref1,
+                                    struct __prelim_ref *ref2)
+{
+       if (ref1->level != ref2->level)
+               return 0;
+       if (ref1->root_id != ref2->root_id)
+               return 0;
+       if (ref1->key_for_search.type != ref2->key_for_search.type)
+               return 0;
+       if (ref1->key_for_search.objectid != ref2->key_for_search.objectid)
+               return 0;
+       if (ref1->key_for_search.offset != ref2->key_for_search.offset)
+               return 0;
+       if (ref1->parent != ref2->parent)
+               return 0;
+
+       return 1;
+}
+
+/*
+ * read tree blocks and add keys where required.
+ */
+static int __add_missing_keys(struct btrfs_fs_info *fs_info,
+                             struct list_head *head)
+{
+       struct list_head *pos;
+       struct extent_buffer *eb;
+
+       list_for_each(pos, head) {
+               struct __prelim_ref *ref;
+               ref = list_entry(pos, struct __prelim_ref, list);
+
+               if (ref->parent)
+                       continue;
+               if (ref->key_for_search.type)
+                       continue;
+               BUG_ON(!ref->wanted_disk_byte);
+               eb = read_tree_block(fs_info->tree_root, ref->wanted_disk_byte,
+                                    fs_info->tree_root->leafsize, 0);
+               BUG_ON(!eb);
+               btrfs_tree_read_lock(eb);
+               if (btrfs_header_level(eb) == 0)
+                       btrfs_item_key_to_cpu(eb, &ref->key_for_search, 0);
+               else
+                       btrfs_node_key_to_cpu(eb, &ref->key_for_search, 0);
+               btrfs_tree_read_unlock(eb);
+               free_extent_buffer(eb);
+       }
+       return 0;
+}
+
+/*
+ * merge two lists of backrefs and adjust counts accordingly
+ *
+ * mode = 1: merge identical keys, if key is set
+ *    FIXME: if we add more keys in __add_prelim_ref, we can merge more here.
+ *           additionally, we could even add a key range for the blocks we
+ *           looked into to merge even more (-> replace unresolved refs by those
+ *           having a parent).
+ * mode = 2: merge identical parents
+ */
+static int __merge_refs(struct list_head *head, int mode)
+{
+       struct list_head *pos1;
+
+       list_for_each(pos1, head) {
+               struct list_head *n2;
+               struct list_head *pos2;
+               struct __prelim_ref *ref1;
+
+               ref1 = list_entry(pos1, struct __prelim_ref, list);
+
+               for (pos2 = pos1->next, n2 = pos2->next; pos2 != head;
+                    pos2 = n2, n2 = pos2->next) {
+                       struct __prelim_ref *ref2;
+                       struct __prelim_ref *xchg;
+
+                       ref2 = list_entry(pos2, struct __prelim_ref, list);
+
+                       if (mode == 1) {
+                               if (!ref_for_same_block(ref1, ref2))
+                                       continue;
+                               if (!ref1->parent && ref2->parent) {
+                                       xchg = ref1;
+                                       ref1 = ref2;
+                                       ref2 = xchg;
+                               }
+                               ref1->count += ref2->count;
+                       } else {
+                               if (ref1->parent != ref2->parent)
+                                       continue;
+                               ref1->count += ref2->count;
+                       }
+                       list_del(&ref2->list);
+                       kfree(ref2);
+               }
+
+       }
+       return 0;
+}
+
+/*
+ * add all currently queued delayed refs from this head whose seq nr is
+ * smaller or equal that seq to the list
+ */
+static int __add_delayed_refs(struct btrfs_delayed_ref_head *head, u64 seq,
+                             struct list_head *prefs)
+{
+       struct btrfs_delayed_extent_op *extent_op = head->extent_op;
+       struct rb_node *n = &head->node.rb_node;
+       struct btrfs_key key;
+       struct btrfs_key op_key = {0};
+       int sgn;
+       int ret = 0;
+
+       if (extent_op && extent_op->update_key)
+               btrfs_disk_key_to_cpu(&op_key, &extent_op->key);
+
+       while ((n = rb_prev(n))) {
+               struct btrfs_delayed_ref_node *node;
+               node = rb_entry(n, struct btrfs_delayed_ref_node,
+                               rb_node);
+               if (node->bytenr != head->node.bytenr)
+                       break;
+               WARN_ON(node->is_head);
+
+               if (node->seq > seq)
+                       continue;
+
+               switch (node->action) {
+               case BTRFS_ADD_DELAYED_EXTENT:
+               case BTRFS_UPDATE_DELAYED_HEAD:
+                       WARN_ON(1);
+                       continue;
+               case BTRFS_ADD_DELAYED_REF:
+                       sgn = 1;
+                       break;
+               case BTRFS_DROP_DELAYED_REF:
+                       sgn = -1;
+                       break;
+               default:
+                       BUG_ON(1);
+               }
+               switch (node->type) {
+               case BTRFS_TREE_BLOCK_REF_KEY: {
+                       struct btrfs_delayed_tree_ref *ref;
+
+                       ref = btrfs_delayed_node_to_tree_ref(node);
+                       ret = __add_prelim_ref(prefs, ref->root, &op_key,
+                                              ref->level + 1, 0, node->bytenr,
+                                              node->ref_mod * sgn);
+                       break;
+               }
+               case BTRFS_SHARED_BLOCK_REF_KEY: {
+                       struct btrfs_delayed_tree_ref *ref;
+
+                       ref = btrfs_delayed_node_to_tree_ref(node);
+                       ret = __add_prelim_ref(prefs, ref->root, NULL,
+                                              ref->level + 1, ref->parent,
+                                              node->bytenr,
+                                              node->ref_mod * sgn);
+                       break;
+               }
+               case BTRFS_EXTENT_DATA_REF_KEY: {
+                       struct btrfs_delayed_data_ref *ref;
+                       ref = btrfs_delayed_node_to_data_ref(node);
+
+                       key.objectid = ref->objectid;
+                       key.type = BTRFS_EXTENT_DATA_KEY;
+                       key.offset = ref->offset;
+                       ret = __add_prelim_ref(prefs, ref->root, &key, 0, 0,
+                                              node->bytenr,
+                                              node->ref_mod * sgn);
+                       break;
+               }
+               case BTRFS_SHARED_DATA_REF_KEY: {
+                       struct btrfs_delayed_data_ref *ref;
+
+                       ref = btrfs_delayed_node_to_data_ref(node);
+
+                       key.objectid = ref->objectid;
+                       key.type = BTRFS_EXTENT_DATA_KEY;
+                       key.offset = ref->offset;
+                       ret = __add_prelim_ref(prefs, ref->root, &key, 0,
+                                              ref->parent, node->bytenr,
+                                              node->ref_mod * sgn);
+                       break;
+               }
+               default:
+                       WARN_ON(1);
+               }
+               BUG_ON(ret);
+       }
+
+       return 0;
+}
+
+/*
+ * add all inline backrefs for bytenr to the list
+ */
+static int __add_inline_refs(struct btrfs_fs_info *fs_info,
+                            struct btrfs_path *path, u64 bytenr,
+                            int *info_level, struct list_head *prefs)
+{
+       int ret = 0;
+       int slot;
+       struct extent_buffer *leaf;
+       struct btrfs_key key;
+       unsigned long ptr;
+       unsigned long end;
+       struct btrfs_extent_item *ei;
+       u64 flags;
+       u64 item_size;
+
+       /*
+        * enumerate all inline refs
+        */
+       leaf = path->nodes[0];
+       slot = path->slots[0];
+
+       item_size = btrfs_item_size_nr(leaf, slot);
+       BUG_ON(item_size < sizeof(*ei));
+
+       ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
+       flags = btrfs_extent_flags(leaf, ei);
+
+       ptr = (unsigned long)(ei + 1);
+       end = (unsigned long)ei + item_size;
+
+       if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
+               struct btrfs_tree_block_info *info;
+
+               info = (struct btrfs_tree_block_info *)ptr;
+               *info_level = btrfs_tree_block_level(leaf, info);
+               ptr += sizeof(struct btrfs_tree_block_info);
+               BUG_ON(ptr > end);
+       } else {
+               BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
+       }
+
+       while (ptr < end) {
+               struct btrfs_extent_inline_ref *iref;
+               u64 offset;
+               int type;
+
+               iref = (struct btrfs_extent_inline_ref *)ptr;
+               type = btrfs_extent_inline_ref_type(leaf, iref);
+               offset = btrfs_extent_inline_ref_offset(leaf, iref);
+
+               switch (type) {
+               case BTRFS_SHARED_BLOCK_REF_KEY:
+                       ret = __add_prelim_ref(prefs, 0, NULL,
+                                               *info_level + 1, offset,
+                                               bytenr, 1);
+                       break;
+               case BTRFS_SHARED_DATA_REF_KEY: {
+                       struct btrfs_shared_data_ref *sdref;
+                       int count;
+
+                       sdref = (struct btrfs_shared_data_ref *)(iref + 1);
+                       count = btrfs_shared_data_ref_count(leaf, sdref);
+                       ret = __add_prelim_ref(prefs, 0, NULL, 0, offset,
+                                              bytenr, count);
+                       break;
+               }
+               case BTRFS_TREE_BLOCK_REF_KEY:
+                       ret = __add_prelim_ref(prefs, offset, NULL,
+                                              *info_level + 1, 0,
+                                              bytenr, 1);
+                       break;
+               case BTRFS_EXTENT_DATA_REF_KEY: {
+                       struct btrfs_extent_data_ref *dref;
+                       int count;
+                       u64 root;
+
+                       dref = (struct btrfs_extent_data_ref *)(&iref->offset);
+                       count = btrfs_extent_data_ref_count(leaf, dref);
+                       key.objectid = btrfs_extent_data_ref_objectid(leaf,
+                                                                     dref);
+                       key.type = BTRFS_EXTENT_DATA_KEY;
+                       key.offset = btrfs_extent_data_ref_offset(leaf, dref);
+                       root = btrfs_extent_data_ref_root(leaf, dref);
+                       ret = __add_prelim_ref(prefs, root, &key, 0, 0,
+                                              bytenr, count);
+                       break;
+               }
+               default:
+                       WARN_ON(1);
+               }
+               BUG_ON(ret);
+               ptr += btrfs_extent_inline_ref_size(type);
+       }
+
+       return 0;
+}
+
+/*
+ * add all non-inline backrefs for bytenr to the list
+ */
+static int __add_keyed_refs(struct btrfs_fs_info *fs_info,
+                           struct btrfs_path *path, u64 bytenr,
+                           int info_level, struct list_head *prefs)
+{
+       struct btrfs_root *extent_root = fs_info->extent_root;
+       int ret;
+       int slot;
+       struct extent_buffer *leaf;
+       struct btrfs_key key;
+
+       while (1) {
+               ret = btrfs_next_item(extent_root, path);
+               if (ret < 0)
+                       break;
+               if (ret) {
+                       ret = 0;
+                       break;
+               }
+
+               slot = path->slots[0];
+               leaf = path->nodes[0];
+               btrfs_item_key_to_cpu(leaf, &key, slot);
+
+               if (key.objectid != bytenr)
+                       break;
+               if (key.type < BTRFS_TREE_BLOCK_REF_KEY)
+                       continue;
+               if (key.type > BTRFS_SHARED_DATA_REF_KEY)
+                       break;
+
+               switch (key.type) {
+               case BTRFS_SHARED_BLOCK_REF_KEY:
+                       ret = __add_prelim_ref(prefs, 0, NULL,
+                                               info_level + 1, key.offset,
+                                               bytenr, 1);
+                       break;
+               case BTRFS_SHARED_DATA_REF_KEY: {
+                       struct btrfs_shared_data_ref *sdref;
+                       int count;
+
+                       sdref = btrfs_item_ptr(leaf, slot,
+                                             struct btrfs_shared_data_ref);
+                       count = btrfs_shared_data_ref_count(leaf, sdref);
+                       ret = __add_prelim_ref(prefs, 0, NULL, 0, key.offset,
+                                               bytenr, count);
+                       break;
+               }
+               case BTRFS_TREE_BLOCK_REF_KEY:
+                       ret = __add_prelim_ref(prefs, key.offset, NULL,
+                                              info_level + 1, 0,
+                                              bytenr, 1);
+                       break;
+               case BTRFS_EXTENT_DATA_REF_KEY: {
+                       struct btrfs_extent_data_ref *dref;
+                       int count;
+                       u64 root;
+
+                       dref = btrfs_item_ptr(leaf, slot,
+                                             struct btrfs_extent_data_ref);
+                       count = btrfs_extent_data_ref_count(leaf, dref);
+                       key.objectid = btrfs_extent_data_ref_objectid(leaf,
+                                                                     dref);
+                       key.type = BTRFS_EXTENT_DATA_KEY;
+                       key.offset = btrfs_extent_data_ref_offset(leaf, dref);
+                       root = btrfs_extent_data_ref_root(leaf, dref);
+                       ret = __add_prelim_ref(prefs, root, &key, 0, 0,
+                                              bytenr, count);
+                       break;
+               }
+               default:
+                       WARN_ON(1);
+               }
+               BUG_ON(ret);
+       }
+
+       return ret;
+}
+
+/*
+ * this adds all existing backrefs (inline backrefs, backrefs and delayed
+ * refs) for the given bytenr to the refs list, merges duplicates and resolves
+ * indirect refs to their parent bytenr.
+ * When roots are found, they're added to the roots list
+ *
+ * FIXME some caching might speed things up
+ */
+static int find_parent_nodes(struct btrfs_trans_handle *trans,
+                            struct btrfs_fs_info *fs_info, u64 bytenr,
+                            u64 time_seq, struct ulist *refs,
+                            struct ulist *roots, const u64 *extent_item_pos)
+{
+       struct btrfs_key key;
+       struct btrfs_path *path;
+       struct btrfs_delayed_ref_root *delayed_refs = NULL;
+       struct btrfs_delayed_ref_head *head;
+       int info_level = 0;
+       int ret;
+       int search_commit_root = (trans == BTRFS_BACKREF_SEARCH_COMMIT_ROOT);
+       struct list_head prefs_delayed;
+       struct list_head prefs;
+       struct __prelim_ref *ref;
+
+       INIT_LIST_HEAD(&prefs);
+       INIT_LIST_HEAD(&prefs_delayed);
+
+       key.objectid = bytenr;
+       key.type = BTRFS_EXTENT_ITEM_KEY;
+       key.offset = (u64)-1;
+
+       path = btrfs_alloc_path();
+       if (!path)
+               return -ENOMEM;
+       path->search_commit_root = !!search_commit_root;
+
+       /*
+        * grab both a lock on the path and a lock on the delayed ref head.
+        * We need both to get a consistent picture of how the refs look
+        * at a specified point in time
+        */
+again:
+       head = NULL;
+
+       ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
+       if (ret < 0)
+               goto out;
+       BUG_ON(ret == 0);
+
+       if (trans != BTRFS_BACKREF_SEARCH_COMMIT_ROOT) {
+               /*
+                * look if there are updates for this ref queued and lock the
+                * head
+                */
+               delayed_refs = &trans->transaction->delayed_refs;
+               spin_lock(&delayed_refs->lock);
+               head = btrfs_find_delayed_ref_head(trans, bytenr);
+               if (head) {
+                       if (!mutex_trylock(&head->mutex)) {
+                               atomic_inc(&head->node.refs);
+                               spin_unlock(&delayed_refs->lock);
+
+                               btrfs_release_path(path);
+
+                               /*
+                                * Mutex was contended, block until it's
+                                * released and try again
+                                */
+                               mutex_lock(&head->mutex);
+                               mutex_unlock(&head->mutex);
+                               btrfs_put_delayed_ref(&head->node);
+                               goto again;
+                       }
+                       ret = __add_delayed_refs(head, time_seq,
+                                                &prefs_delayed);
+                       mutex_unlock(&head->mutex);
+                       if (ret) {
+                               spin_unlock(&delayed_refs->lock);
+                               goto out;
+                       }
+               }
+               spin_unlock(&delayed_refs->lock);
+       }
+
+       if (path->slots[0]) {
+               struct extent_buffer *leaf;
+               int slot;
+
+               path->slots[0]--;
+               leaf = path->nodes[0];
+               slot = path->slots[0];
+               btrfs_item_key_to_cpu(leaf, &key, slot);
+               if (key.objectid == bytenr &&
+                   key.type == BTRFS_EXTENT_ITEM_KEY) {
+                       ret = __add_inline_refs(fs_info, path, bytenr,
+                                               &info_level, &prefs);
+                       if (ret)
+                               goto out;
+                       ret = __add_keyed_refs(fs_info, path, bytenr,
+                                              info_level, &prefs);
+                       if (ret)
+                               goto out;
+               }
+       }
+       btrfs_release_path(path);
+
+       list_splice_init(&prefs_delayed, &prefs);
+
+       ret = __add_missing_keys(fs_info, &prefs);
+       if (ret)
+               goto out;
+
+       ret = __merge_refs(&prefs, 1);
+       if (ret)
+               goto out;
+
+       ret = __resolve_indirect_refs(fs_info, search_commit_root, time_seq,
+                                     &prefs, extent_item_pos);
+       if (ret)
+               goto out;
+
+       ret = __merge_refs(&prefs, 2);
+       if (ret)
+               goto out;
+
+       while (!list_empty(&prefs)) {
+               ref = list_first_entry(&prefs, struct __prelim_ref, list);
+               list_del(&ref->list);
+               if (ref->count < 0)
+                       WARN_ON(1);
+               if (ref->count && ref->root_id && ref->parent == 0) {
+                       /* no parent == root of tree */
+                       ret = ulist_add(roots, ref->root_id, 0, GFP_NOFS);
+                       BUG_ON(ret < 0);
+               }
+               if (ref->count && ref->parent) {
+                       struct extent_inode_elem *eie = NULL;
+                       if (extent_item_pos && !ref->inode_list) {
+                               u32 bsz;
+                               struct extent_buffer *eb;
+                               bsz = btrfs_level_size(fs_info->extent_root,
+                                                       info_level);
+                               eb = read_tree_block(fs_info->extent_root,
+                                                          ref->parent, bsz, 0);
+                               BUG_ON(!eb);
+                               ret = find_extent_in_eb(eb, bytenr,
+                                                       *extent_item_pos, &eie);
+                               ref->inode_list = eie;
+                               free_extent_buffer(eb);
+                       }
+                       ret = ulist_add_merge(refs, ref->parent,
+                                             (unsigned long)ref->inode_list,
+                                             (unsigned long *)&eie, GFP_NOFS);
+                       if (!ret && extent_item_pos) {
+                               /*
+                                * we've recorded that parent, so we must extend
+                                * its inode list here
+                                */
+                               BUG_ON(!eie);
+                               while (eie->next)
+                                       eie = eie->next;
+                               eie->next = ref->inode_list;
+                       }
+                       BUG_ON(ret < 0);
+               }
+               kfree(ref);
+       }
+
+out:
+       btrfs_free_path(path);
+       while (!list_empty(&prefs)) {
+               ref = list_first_entry(&prefs, struct __prelim_ref, list);
+               list_del(&ref->list);
+               kfree(ref);
+       }
+       while (!list_empty(&prefs_delayed)) {
+               ref = list_first_entry(&prefs_delayed, struct __prelim_ref,
+                                      list);
+               list_del(&ref->list);
+               kfree(ref);
+       }
+
+       return ret;
+}
+
+static void free_leaf_list(struct ulist *blocks)
+{
+       struct ulist_node *node = NULL;
+       struct extent_inode_elem *eie;
+       struct extent_inode_elem *eie_next;
+       struct ulist_iterator uiter;
+
+       ULIST_ITER_INIT(&uiter);
+       while ((node = ulist_next(blocks, &uiter))) {
+               if (!node->aux)
+                       continue;
+               eie = (struct extent_inode_elem *)node->aux;
+               for (; eie; eie = eie_next) {
+                       eie_next = eie->next;
+                       kfree(eie);
+               }
+               node->aux = 0;
+       }
+
+       ulist_free(blocks);
+}
+
+/*
+ * Finds all leafs with a reference to the specified combination of bytenr and
+ * offset. key_list_head will point to a list of corresponding keys (caller must
+ * free each list element). The leafs will be stored in the leafs ulist, which
+ * must be freed with ulist_free.
+ *
+ * returns 0 on success, <0 on error
+ */
+static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
+                               struct btrfs_fs_info *fs_info, u64 bytenr,
+                               u64 time_seq, struct ulist **leafs,
+                               const u64 *extent_item_pos)
+{
+       struct ulist *tmp;
+       int ret;
+
+       tmp = ulist_alloc(GFP_NOFS);
+       if (!tmp)
+               return -ENOMEM;
+       *leafs = ulist_alloc(GFP_NOFS);
+       if (!*leafs) {
+               ulist_free(tmp);
+               return -ENOMEM;
+       }
+
+       ret = find_parent_nodes(trans, fs_info, bytenr,
+                               time_seq, *leafs, tmp, extent_item_pos);
+       ulist_free(tmp);
+
+       if (ret < 0 && ret != -ENOENT) {
+               free_leaf_list(*leafs);
+               return ret;
+       }
+
+       return 0;
+}
+
+/*
+ * walk all backrefs for a given extent to find all roots that reference this
+ * extent. Walking a backref means finding all extents that reference this
+ * extent and in turn walk the backrefs of those, too. Naturally this is a
+ * recursive process, but here it is implemented in an iterative fashion: We
+ * find all referencing extents for the extent in question and put them on a
+ * list. In turn, we find all referencing extents for those, further appending
+ * to the list. The way we iterate the list allows adding more elements after
+ * the current while iterating. The process stops when we reach the end of the
+ * list. Found roots are added to the roots list.
+ *
+ * returns 0 on success, < 0 on error.
+ */
+int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
+                               struct btrfs_fs_info *fs_info, u64 bytenr,
+                               u64 time_seq, struct ulist **roots)
+{
+       struct ulist *tmp;
+       struct ulist_node *node = NULL;
+       struct ulist_iterator uiter;
+       int ret;
+
+       tmp = ulist_alloc(GFP_NOFS);
+       if (!tmp)
+               return -ENOMEM;
+       *roots = ulist_alloc(GFP_NOFS);
+       if (!*roots) {
+               ulist_free(tmp);
+               return -ENOMEM;
+       }
+
+       ULIST_ITER_INIT(&uiter);
+       while (1) {
+               ret = find_parent_nodes(trans, fs_info, bytenr,
+                                       time_seq, tmp, *roots, NULL);
+               if (ret < 0 && ret != -ENOENT) {
+                       ulist_free(tmp);
+                       ulist_free(*roots);
+                       return ret;
+               }
+               node = ulist_next(tmp, &uiter);
+               if (!node)
+                       break;
+               bytenr = node->val;
+       }
+
+       ulist_free(tmp);
+       return 0;
+}
+
+
 static int __inode_info(u64 inum, u64 ioff, u8 key_type,
                        struct btrfs_root *fs_root, struct btrfs_path *path,
                        struct btrfs_key *found_key)
@@ -96,10 +1122,10 @@ static int inode_ref_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
  * required for the path to fit into the buffer. in that case, the returned
  * value will be smaller than dest. callers must check this!
  */
-static char *iref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
-                               struct btrfs_inode_ref *iref,
-                               struct extent_buffer *eb_in, u64 parent,
-                               char *dest, u32 size)
+char *btrfs_iref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
+                        struct btrfs_inode_ref *iref,
+                        struct extent_buffer *eb_in, u64 parent,
+                        char *dest, u32 size)
 {
        u32 len;
        int slot;
@@ -108,19 +1134,25 @@ static char *iref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
        s64 bytes_left = size - 1;
        struct extent_buffer *eb = eb_in;
        struct btrfs_key found_key;
+       int leave_spinning = path->leave_spinning;
 
        if (bytes_left >= 0)
                dest[bytes_left] = '\0';
 
+       path->leave_spinning = 1;
        while (1) {
                len = btrfs_inode_ref_name_len(eb, iref);
                bytes_left -= len;
                if (bytes_left >= 0)
                        read_extent_buffer(eb, dest + bytes_left,
                                                (unsigned long)(iref + 1), len);
-               if (eb != eb_in)
+               if (eb != eb_in) {
+                       btrfs_tree_read_unlock_blocking(eb);
                        free_extent_buffer(eb);
+               }
                ret = inode_ref_info(parent, 0, fs_root, path, &found_key);
+               if (ret > 0)
+                       ret = -ENOENT;
                if (ret)
                        break;
                next_inum = found_key.offset;
@@ -132,8 +1164,11 @@ static char *iref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
                slot = path->slots[0];
                eb = path->nodes[0];
                /* make sure we can use eb after releasing the path */
-               if (eb != eb_in)
+               if (eb != eb_in) {
                        atomic_inc(&eb->refs);
+                       btrfs_tree_read_lock(eb);
+                       btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
+               }
                btrfs_release_path(path);
 
                iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
@@ -144,6 +1179,7 @@ static char *iref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
        }
 
        btrfs_release_path(path);
+       path->leave_spinning = leave_spinning;
 
        if (ret)
                return ERR_PTR(ret);
@@ -181,8 +1217,11 @@ int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
        btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]);
        if (found_key->type != BTRFS_EXTENT_ITEM_KEY ||
            found_key->objectid > logical ||
-           found_key->objectid + found_key->offset <= logical)
+           found_key->objectid + found_key->offset <= logical) {
+               pr_debug("logical %llu is not within any extent\n",
+                        (unsigned long long)logical);
                return -ENOENT;
+       }
 
        eb = path->nodes[0];
        item_size = btrfs_item_size_nr(eb, path->slots[0]);
@@ -191,6 +1230,13 @@ int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
        ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
        flags = btrfs_extent_flags(eb, ei);
 
+       pr_debug("logical %llu is at position %llu within the extent (%llu "
+                "EXTENT_ITEM %llu) flags %#llx size %u\n",
+                (unsigned long long)logical,
+                (unsigned long long)(logical - found_key->objectid),
+                (unsigned long long)found_key->objectid,
+                (unsigned long long)found_key->offset,
+                (unsigned long long)flags, item_size);
        if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
                return BTRFS_EXTENT_FLAG_TREE_BLOCK;
        if (flags & BTRFS_EXTENT_FLAG_DATA)
@@ -287,295 +1333,94 @@ int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
        return 0;
 }
 
-static int __data_list_add(struct list_head *head, u64 inum,
-                               u64 extent_data_item_offset, u64 root)
-{
-       struct __data_ref *ref;
-
-       ref = kmalloc(sizeof(*ref), GFP_NOFS);
-       if (!ref)
-               return -ENOMEM;
-
-       ref->inum = inum;
-       ref->extent_data_item_offset = extent_data_item_offset;
-       ref->root = root;
-       list_add_tail(&ref->list, head);
-
-       return 0;
-}
-
-static int __data_list_add_eb(struct list_head *head, struct extent_buffer *eb,
-                               struct btrfs_extent_data_ref *dref)
-{
-       return __data_list_add(head, btrfs_extent_data_ref_objectid(eb, dref),
-                               btrfs_extent_data_ref_offset(eb, dref),
-                               btrfs_extent_data_ref_root(eb, dref));
-}
-
-static int __shared_list_add(struct list_head *head, u64 disk_byte)
-{
-       struct __shared_ref *ref;
-
-       ref = kmalloc(sizeof(*ref), GFP_NOFS);
-       if (!ref)
-               return -ENOMEM;
-
-       ref->disk_byte = disk_byte;
-       list_add_tail(&ref->list, head);
-
-       return 0;
-}
-
-static int __iter_shared_inline_ref_inodes(struct btrfs_fs_info *fs_info,
-                                          u64 logical, u64 inum,
-                                          u64 extent_data_item_offset,
-                                          u64 extent_offset,
-                                          struct btrfs_path *path,
-                                          struct list_head *data_refs,
-                                          iterate_extent_inodes_t *iterate,
-                                          void *ctx)
-{
-       u64 ref_root;
-       u32 item_size;
-       struct btrfs_key key;
-       struct extent_buffer *eb;
-       struct btrfs_extent_item *ei;
-       struct btrfs_extent_inline_ref *eiref;
-       struct __data_ref *ref;
-       int ret;
-       int type;
-       int last;
-       unsigned long ptr = 0;
-
-       WARN_ON(!list_empty(data_refs));
-       ret = extent_from_logical(fs_info, logical, path, &key);
-       if (ret & BTRFS_EXTENT_FLAG_DATA)
-               ret = -EIO;
-       if (ret < 0)
-               goto out;
-
-       eb = path->nodes[0];
-       ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
-       item_size = btrfs_item_size_nr(eb, path->slots[0]);
-
-       ret = 0;
-       ref_root = 0;
-       /*
-        * as done in iterate_extent_inodes, we first build a list of refs to
-        * iterate, then free the path and then iterate them to avoid deadlocks.
-        */
-       do {
-               last = __get_extent_inline_ref(&ptr, eb, ei, item_size,
-                                               &eiref, &type);
-               if (last < 0) {
-                       ret = last;
-                       goto out;
-               }
-               if (type == BTRFS_TREE_BLOCK_REF_KEY ||
-                   type == BTRFS_SHARED_BLOCK_REF_KEY) {
-                       ref_root = btrfs_extent_inline_ref_offset(eb, eiref);
-                       ret = __data_list_add(data_refs, inum,
-                                               extent_data_item_offset,
-                                               ref_root);
-               }
-       } while (!ret && !last);
-
-       btrfs_release_path(path);
-
-       if (ref_root == 0) {
-               printk(KERN_ERR "btrfs: failed to find tree block ref "
-                       "for shared data backref %llu\n", logical);
-               WARN_ON(1);
-               ret = -EIO;
-       }
-
-out:
-       while (!list_empty(data_refs)) {
-               ref = list_first_entry(data_refs, struct __data_ref, list);
-               list_del(&ref->list);
-               if (!ret)
-                       ret = iterate(ref->inum, extent_offset +
-                                       ref->extent_data_item_offset,
-                                       ref->root, ctx);
-               kfree(ref);
-       }
-
-       return ret;
-}
-
-static int __iter_shared_inline_ref(struct btrfs_fs_info *fs_info,
-                                   u64 logical, u64 orig_extent_item_objectid,
-                                   u64 extent_offset, struct btrfs_path *path,
-                                   struct list_head *data_refs,
-                                   iterate_extent_inodes_t *iterate,
-                                   void *ctx)
+static int iterate_leaf_refs(struct extent_inode_elem *inode_list,
+                               u64 root, u64 extent_item_objectid,
+                               iterate_extent_inodes_t *iterate, void *ctx)
 {
-       u64 disk_byte;
-       struct btrfs_key key;
-       struct btrfs_file_extent_item *fi;
-       struct extent_buffer *eb;
-       int slot;
-       int nritems;
-       int ret;
-       int found = 0;
-
-       eb = read_tree_block(fs_info->tree_root, logical,
-                               fs_info->tree_root->leafsize, 0);
-       if (!eb)
-               return -EIO;
-
-       /*
-        * from the shared data ref, we only have the leaf but we need
-        * the key. thus, we must look into all items and see that we
-        * find one (some) with a reference to our extent item.
-        */
-       nritems = btrfs_header_nritems(eb);
-       for (slot = 0; slot < nritems; ++slot) {
-               btrfs_item_key_to_cpu(eb, &key, slot);
-               if (key.type != BTRFS_EXTENT_DATA_KEY)
-                       continue;
-               fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
-               if (!fi) {
-                       free_extent_buffer(eb);
-                       return -EIO;
-               }
-               disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
-               if (disk_byte != orig_extent_item_objectid) {
-                       if (found)
-                               break;
-                       else
-                               continue;
-               }
-               ++found;
-               ret = __iter_shared_inline_ref_inodes(fs_info, logical,
-                                                       key.objectid,
-                                                       key.offset,
-                                                       extent_offset, path,
-                                                       data_refs,
-                                                       iterate, ctx);
-               if (ret)
+       struct extent_inode_elem *eie;
+       int ret = 0;
+
+       for (eie = inode_list; eie; eie = eie->next) {
+               pr_debug("ref for %llu resolved, key (%llu EXTEND_DATA %llu), "
+                        "root %llu\n", extent_item_objectid,
+                        eie->inum, eie->offset, root);
+               ret = iterate(eie->inum, eie->offset, root, ctx);
+               if (ret) {
+                       pr_debug("stopping iteration for %llu due to ret=%d\n",
+                                extent_item_objectid, ret);
                        break;
+               }
        }
 
-       if (!found) {
-               printk(KERN_ERR "btrfs: failed to follow shared data backref "
-                       "to parent %llu\n", logical);
-               WARN_ON(1);
-               ret = -EIO;
-       }
-
-       free_extent_buffer(eb);
        return ret;
 }
 
 /*
  * calls iterate() for every inode that references the extent identified by
- * the given parameters. will use the path given as a parameter and return it
- * released.
+ * the given parameters.
  * when the iterator function returns a non-zero value, iteration stops.
  */
 int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
-                               struct btrfs_path *path,
-                               u64 extent_item_objectid,
-                               u64 extent_offset,
+                               u64 extent_item_objectid, u64 extent_item_pos,
+                               int search_commit_root,
                                iterate_extent_inodes_t *iterate, void *ctx)
 {
-       unsigned long ptr = 0;
-       int last;
        int ret;
-       int type;
-       u64 logical;
-       u32 item_size;
-       struct btrfs_extent_inline_ref *eiref;
-       struct btrfs_extent_data_ref *dref;
-       struct extent_buffer *eb;
-       struct btrfs_extent_item *ei;
-       struct btrfs_key key;
        struct list_head data_refs = LIST_HEAD_INIT(data_refs);
        struct list_head shared_refs = LIST_HEAD_INIT(shared_refs);
-       struct __data_ref *ref_d;
-       struct __shared_ref *ref_s;
-
-       eb = path->nodes[0];
-       ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
-       item_size = btrfs_item_size_nr(eb, path->slots[0]);
-
-       /* first we iterate the inline refs, ... */
-       do {
-               last = __get_extent_inline_ref(&ptr, eb, ei, item_size,
-                                               &eiref, &type);
-               if (last == -ENOENT) {
-                       ret = 0;
-                       break;
-               }
-               if (last < 0) {
-                       ret = last;
-                       break;
-               }
+       struct btrfs_trans_handle *trans;
+       struct ulist *refs = NULL;
+       struct ulist *roots = NULL;
+       struct ulist_node *ref_node = NULL;
+       struct ulist_node *root_node = NULL;
+       struct seq_list tree_mod_seq_elem = {};
+       struct ulist_iterator ref_uiter;
+       struct ulist_iterator root_uiter;
+
+       pr_debug("resolving all inodes for extent %llu\n",
+                       extent_item_objectid);
+
+       if (search_commit_root) {
+               trans = BTRFS_BACKREF_SEARCH_COMMIT_ROOT;
+       } else {
+               trans = btrfs_join_transaction(fs_info->extent_root);
+               if (IS_ERR(trans))
+                       return PTR_ERR(trans);
+               btrfs_get_tree_mod_seq(fs_info, &tree_mod_seq_elem);
+       }
 
-               if (type == BTRFS_EXTENT_DATA_REF_KEY) {
-                       dref = (struct btrfs_extent_data_ref *)(&eiref->offset);
-                       ret = __data_list_add_eb(&data_refs, eb, dref);
-               } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
-                       logical = btrfs_extent_inline_ref_offset(eb, eiref);
-                       ret = __shared_list_add(&shared_refs, logical);
-               }
-       } while (!ret && !last);
+       ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
+                                  tree_mod_seq_elem.seq, &refs,
+                                  &extent_item_pos);
+       if (ret)
+               goto out;
 
-       /* ... then we proceed to in-tree references and ... */
-       while (!ret) {
-               ++path->slots[0];
-               if (path->slots[0] > btrfs_header_nritems(eb)) {
-                       ret = btrfs_next_leaf(fs_info->extent_root, path);
-                       if (ret) {
-                               if (ret == 1)
-                                       ret = 0; /* we're done */
-                               break;
-                       }
-                       eb = path->nodes[0];
-               }
-               btrfs_item_key_to_cpu(eb, &key, path->slots[0]);
-               if (key.objectid != extent_item_objectid)
+       ULIST_ITER_INIT(&ref_uiter);
+       while (!ret && (ref_node = ulist_next(refs, &ref_uiter))) {
+               ret = btrfs_find_all_roots(trans, fs_info, ref_node->val,
+                                          tree_mod_seq_elem.seq, &roots);
+               if (ret)
                        break;
-               if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
-                       dref = btrfs_item_ptr(eb, path->slots[0],
-                                               struct btrfs_extent_data_ref);
-                       ret = __data_list_add_eb(&data_refs, eb, dref);
-               } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
-                       ret = __shared_list_add(&shared_refs, key.offset);
+               ULIST_ITER_INIT(&root_uiter);
+               while (!ret && (root_node = ulist_next(roots, &root_uiter))) {
+                       pr_debug("root %llu references leaf %llu, data list "
+                                "%#lx\n", root_node->val, ref_node->val,
+                                ref_node->aux);
+                       ret = iterate_leaf_refs(
+                               (struct extent_inode_elem *)ref_node->aux,
+                               root_node->val, extent_item_objectid,
+                               iterate, ctx);
                }
+               ulist_free(roots);
+               roots = NULL;
        }
 
-       btrfs_release_path(path);
-
-       /*
-        * ... only at the very end we can process the refs we found. this is
-        * because the iterator function we call is allowed to make tree lookups
-        * and we have to avoid deadlocks. additionally, we need more tree
-        * lookups ourselves for shared data refs.
-        */
-       while (!list_empty(&data_refs)) {
-               ref_d = list_first_entry(&data_refs, struct __data_ref, list);
-               list_del(&ref_d->list);
-               if (!ret)
-                       ret = iterate(ref_d->inum, extent_offset +
-                                       ref_d->extent_data_item_offset,
-                                       ref_d->root, ctx);
-               kfree(ref_d);
-       }
-
-       while (!list_empty(&shared_refs)) {
-               ref_s = list_first_entry(&shared_refs, struct __shared_ref,
-                                       list);
-               list_del(&ref_s->list);
-               if (!ret)
-                       ret = __iter_shared_inline_ref(fs_info,
-                                                       ref_s->disk_byte,
-                                                       extent_item_objectid,
-                                                       extent_offset, path,
-                                                       &data_refs,
-                                                       iterate, ctx);
-               kfree(ref_s);
+       free_leaf_list(refs);
+       ulist_free(roots);
+out:
+       if (!search_commit_root) {
+               btrfs_put_tree_mod_seq(fs_info, &tree_mod_seq_elem);
+               btrfs_end_transaction(trans, fs_info->extent_root);
        }
 
        return ret;
@@ -586,19 +1431,22 @@ int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
                                iterate_extent_inodes_t *iterate, void *ctx)
 {
        int ret;
-       u64 offset;
+       u64 extent_item_pos;
        struct btrfs_key found_key;
+       int search_commit_root = path->search_commit_root;
 
        ret = extent_from_logical(fs_info, logical, path,
                                        &found_key);
-       if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
-               ret = -EINVAL;
+       btrfs_release_path(path);
        if (ret < 0)
                return ret;
+       if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
+               return -EINVAL;
 
-       offset = logical - found_key.objectid;
-       ret = iterate_extent_inodes(fs_info, path, found_key.objectid,
-                                       offset, iterate, ctx);
+       extent_item_pos = logical - found_key.objectid;
+       ret = iterate_extent_inodes(fs_info, found_key.objectid,
+                                       extent_item_pos, search_commit_root,
+                                       iterate, ctx);
 
        return ret;
 }
@@ -607,7 +1455,7 @@ static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
                                struct btrfs_path *path,
                                iterate_irefs_t *iterate, void *ctx)
 {
-       int ret;
+       int ret = 0;
        int slot;
        u32 cur;
        u32 len;
@@ -619,7 +1467,8 @@ static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
        struct btrfs_inode_ref *iref;
        struct btrfs_key found_key;
 
-       while (1) {
+       while (!ret) {
+               path->leave_spinning = 1;
                ret = inode_ref_info(inum, parent ? parent+1 : 0, fs_root, path,
                                        &found_key);
                if (ret < 0)
@@ -635,6 +1484,8 @@ static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
                eb = path->nodes[0];
                /* make sure we can use eb after releasing the path */
                atomic_inc(&eb->refs);
+               btrfs_tree_read_lock(eb);
+               btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
                btrfs_release_path(path);
 
                item = btrfs_item_nr(eb, slot);
@@ -643,14 +1494,17 @@ static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
                for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) {
                        name_len = btrfs_inode_ref_name_len(eb, iref);
                        /* path must be released before calling iterate()! */
+                       pr_debug("following ref at offset %u for inode %llu in "
+                                "tree %llu\n", cur,
+                                (unsigned long long)found_key.objectid,
+                                (unsigned long long)fs_root->objectid);
                        ret = iterate(parent, iref, eb, ctx);
-                       if (ret) {
-                               free_extent_buffer(eb);
+                       if (ret)
                                break;
-                       }
                        len = sizeof(*iref) + name_len;
                        iref = (struct btrfs_inode_ref *)((char *)iref + len);
                }
+               btrfs_tree_read_unlock_blocking(eb);
                free_extent_buffer(eb);
        }
 
@@ -676,17 +1530,21 @@ static int inode_to_path(u64 inum, struct btrfs_inode_ref *iref,
        bytes_left = ipath->fspath->bytes_left > s_ptr ?
                                        ipath->fspath->bytes_left - s_ptr : 0;
 
-       fspath_min = (char *)ipath->fspath->str + (i + 1) * s_ptr;
-       fspath = iref_to_path(ipath->fs_root, ipath->btrfs_path, iref, eb,
+       fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
+       fspath = btrfs_iref_to_path(ipath->fs_root, ipath->btrfs_path, iref, eb,
                                inum, fspath_min, bytes_left);
        if (IS_ERR(fspath))
                return PTR_ERR(fspath);
 
        if (fspath > fspath_min) {
-               ipath->fspath->str[i] = fspath;
+               pr_debug("path resolved: %s\n", fspath);
+               ipath->fspath->val[i] = (u64)(unsigned long)fspath;
                ++ipath->fspath->elem_cnt;
                ipath->fspath->bytes_left = fspath - fspath_min;
        } else {
+               pr_debug("missed path, not enough space. missing bytes: %lu, "
+                        "constructed so far: %s\n",
+                        (unsigned long)(fspath_min - fspath), fspath_min);
                ++ipath->fspath->elem_missed;
                ipath->fspath->bytes_missing += fspath_min - fspath;
                ipath->fspath->bytes_left = 0;
@@ -698,9 +1556,9 @@ static int inode_to_path(u64 inum, struct btrfs_inode_ref *iref,
 /*
  * this dumps all file system paths to the inode into the ipath struct, provided
  * is has been created large enough. each path is zero-terminated and accessed
- * from ipath->fspath->str[i].
+ * from ipath->fspath->val[i].
  * when it returns, there are ipath->fspath->elem_cnt number of paths available
- * in ipath->fspath->str[]. when the allocated space wasn't sufficient, the
+ * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
  * number of missed paths in recored in ipath->fspath->elem_missed, otherwise,
  * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
  * have been needed to return all paths.
@@ -711,12 +1569,6 @@ int paths_from_inode(u64 inum, struct inode_fs_paths *ipath)
                                inode_to_path, ipath);
 }
 
-/*
- * allocates space to return multiple file system paths for an inode.
- * total_bytes to allocate are passed, note that space usable for actual path
- * information will be total_bytes - sizeof(struct inode_fs_paths).
- * the returned pointer must be freed with free_ipath() in the end.
- */
 struct btrfs_data_container *init_data_container(u32 total_bytes)
 {
        struct btrfs_data_container *data;
@@ -772,5 +1624,8 @@ struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
 
 void free_ipath(struct inode_fs_paths *ipath)
 {
+       if (!ipath)
+               return;
+       kfree(ipath->fspath);
        kfree(ipath);
 }