From f617e2fd52484fb74236a597d0f9068ec7d9d2dd Mon Sep 17 00:00:00 2001
From: Jan Schmidt <list.btrfs@jan-o-sch.net>
Date: Thu, 14 Jun 2012 16:10:13 +0200
Subject: Btrfs: remove obsolete btrfs_next_leaf call from
 __resolve_indirect_ref

When resolving indirect refs, we used to call btrfs_next_leaf in case we
didn't find an exact match. While we should find exact matches most of the
time, in case we don't, we must continue searching. Treating those matches
differently depending on the level we're searching doesn't make sense.

Even worse, we might end up searching for a key larger than the largest, in
which case there is no next_leaf and subsequent jobs would fail. This commit
drops the bogous lines.

Signed-off-by: Jan Schmidt <list.btrfs@jan-o-sch.net>
---
 fs/btrfs/backref.c | 10 +---------
 1 file changed, 1 insertion(+), 9 deletions(-)

(limited to 'fs/btrfs/backref.c')

diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index 3f75895c919..579131de1b3 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -294,16 +294,8 @@ static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
 		goto out;
 	}
 
-	if (level == 0) {
-		if (ret == 1 && path->slots[0] >= btrfs_header_nritems(eb)) {
-			ret = btrfs_next_leaf(root, path);
-			if (ret)
-				goto out;
-			eb = path->nodes[0];
-		}
-
+	if (level == 0)
 		btrfs_item_key_to_cpu(eb, &key, path->slots[0]);
-	}
 
 	ret = add_all_parents(root, path, parents, level, &key,
 				ref->wanted_disk_byte, extent_item_pos);
-- 
cgit v1.2.3-70-g09d2


From 3d7806eca43e73a9721d2e09369200ed93036bd0 Mon Sep 17 00:00:00 2001
From: Jan Schmidt <list.btrfs@jan-o-sch.net>
Date: Mon, 11 Jun 2012 08:29:29 +0200
Subject: Btrfs: add btrfs_next_old_leaf

To make sense of the tree mod log, the backref walker not only needs
btrfs_search_old_slot, but it also called btrfs_next_leaf, which in turn was
calling btrfs_search_slot. This obviously didn't give the correct result.

This commit adds btrfs_next_old_leaf, a drop-in replacement for
btrfs_next_leaf with a time_seq parameter. If it is zero, it behaves exactly
like btrfs_next_leaf. If it is non-zero, it will use btrfs_search_old_slot
with this time_seq parameter.

Signed-off-by: Jan Schmidt <list.btrfs@jan-o-sch.net>
---
 fs/btrfs/backref.c |  7 ++++---
 fs/btrfs/ctree.c   | 11 ++++++++++-
 fs/btrfs/ctree.h   |  2 ++
 3 files changed, 16 insertions(+), 4 deletions(-)

(limited to 'fs/btrfs/backref.c')

diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index 579131de1b3..8f7d1237b7a 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -179,7 +179,8 @@ static int __add_prelim_ref(struct list_head *head, u64 root_id,
 
 static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
 				struct ulist *parents, int level,
-				struct btrfs_key *key, u64 wanted_disk_byte,
+				struct btrfs_key *key, u64 time_seq,
+				u64 wanted_disk_byte,
 				const u64 *extent_item_pos)
 {
 	int ret;
@@ -212,7 +213,7 @@ add_parent:
 	 */
 	while (1) {
 		eie = NULL;
-		ret = btrfs_next_leaf(root, path);
+		ret = btrfs_next_old_leaf(root, path, time_seq);
 		if (ret < 0)
 			return ret;
 		if (ret)
@@ -297,7 +298,7 @@ static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
 	if (level == 0)
 		btrfs_item_key_to_cpu(eb, &key, path->slots[0]);
 
-	ret = add_all_parents(root, path, parents, level, &key,
+	ret = add_all_parents(root, path, parents, level, &key, time_seq,
 				ref->wanted_disk_byte, extent_item_pos);
 out:
 	btrfs_free_path(path);
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 50d7c99ddce..cb76b2a1b90 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -5016,6 +5016,12 @@ next:
  * returns < 0 on io errors.
  */
 int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
+{
+	return btrfs_next_old_leaf(root, path, 0);
+}
+
+int btrfs_next_old_leaf(struct btrfs_root *root, struct btrfs_path *path,
+			u64 time_seq)
 {
 	int slot;
 	int level;
@@ -5041,7 +5047,10 @@ again:
 	path->keep_locks = 1;
 	path->leave_spinning = 1;
 
-	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (time_seq)
+		ret = btrfs_search_old_slot(root, &key, path, time_seq);
+	else
+		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 	path->keep_locks = 0;
 
 	if (ret < 0)
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 0151ca1ac65..db15e9e7b91 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -2753,6 +2753,8 @@ static inline int btrfs_insert_empty_item(struct btrfs_trans_handle *trans,
 }
 
 int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path);
+int btrfs_next_old_leaf(struct btrfs_root *root, struct btrfs_path *path,
+			u64 time_seq);
 static inline int btrfs_next_item(struct btrfs_root *root, struct btrfs_path *p)
 {
 	++p->slots[0];
-- 
cgit v1.2.3-70-g09d2


From 69bca40d41c613927b150c5392505f1894fe3010 Mon Sep 17 00:00:00 2001
From: Alexander Block <ablock84@googlemail.com>
Date: Tue, 19 Jun 2012 07:42:26 -0600
Subject: Btrfs: don't assume to be on the correct extent in add_all_parents

add_all_parents did assume that path is already at a correct extent data
item, which may not be true in case of data extents that were partly
rewritten and splitted.

We need to check if we're on a matching extent for every item and only
for the ones after the first. The loop is changed to do this now.

This patch also fixes a bug introduced with commit 3b127fd8 "Btrfs:
remove obsolete btrfs_next_leaf call from __resolve_indirect_ref".
The removal of next_leaf did sometimes result in slot==nritems when
the above described case happens, and thus resulting in invalid values
(e.g. wanted_obejctid) in add_all_parents (leading to missed backrefs
or even crashes).

Signed-off-by: Alexander Block <ablock84@googlemail.com>
Signed-off-by: Jan Schmidt <list.btrfs@jan-o-sch.net>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
---
 fs/btrfs/backref.c | 94 ++++++++++++++++++++++++++++++------------------------
 1 file changed, 52 insertions(+), 42 deletions(-)

(limited to 'fs/btrfs/backref.c')

diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index 8f7d1237b7a..7301cdb4b2c 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -179,61 +179,74 @@ static int __add_prelim_ref(struct list_head *head, u64 root_id,
 
 static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
 				struct ulist *parents, int level,
-				struct btrfs_key *key, u64 time_seq,
+				struct btrfs_key *key_for_search, u64 time_seq,
 				u64 wanted_disk_byte,
 				const u64 *extent_item_pos)
 {
-	int ret;
-	int slot = path->slots[level];
-	struct extent_buffer *eb = path->nodes[level];
+	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;
-	u64 wanted_objectid = key->objectid;
 
-add_parent:
-	if (level == 0 && extent_item_pos) {
-		fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
-		ret = check_extent_in_eb(key, eb, fi, *extent_item_pos, &eie);
+	if (level != 0) {
+		eb = path->nodes[level];
+		ret = ulist_add(parents, eb->start, 0, GFP_NOFS);
 		if (ret < 0)
 			return ret;
-	}
-	ret = ulist_add(parents, eb->start, (unsigned long)eie, GFP_NOFS);
-	if (ret < 0)
-		return ret;
-
-	if (level != 0)
 		return 0;
+	}
 
 	/*
-	 * if the current leaf is full with EXTENT_DATA items, we must
-	 * check the next one if that holds a reference as well.
-	 * ref->count cannot be used to skip this check.
-	 * repeat this until we don't find any additional EXTENT_DATA items.
+	 * 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.
 	 */
-	while (1) {
-		eie = NULL;
+	if (path->slots[0] >= btrfs_header_nritems(path->nodes[0]))
 		ret = btrfs_next_old_leaf(root, path, time_seq);
-		if (ret < 0)
-			return ret;
-		if (ret)
-			return 0;
 
+	while (!ret) {
 		eb = path->nodes[0];
-		for (slot = 0; slot < btrfs_header_nritems(eb); ++slot) {
-			btrfs_item_key_to_cpu(eb, key, slot);
-			if (key->objectid != wanted_objectid ||
-			    key->type != BTRFS_EXTENT_DATA_KEY)
-				return 0;
-			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)
-				goto add_parent;
+		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);
 	}
 
-	return 0;
+	if (ret > 0)
+		ret = 0;
+	return ret;
 }
 
 /*
@@ -250,7 +263,6 @@ static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
 	struct btrfs_path *path;
 	struct btrfs_root *root;
 	struct btrfs_key root_key;
-	struct btrfs_key key = {0};
 	struct extent_buffer *eb;
 	int ret = 0;
 	int root_level;
@@ -295,11 +307,9 @@ static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
 		goto out;
 	}
 
-	if (level == 0)
-		btrfs_item_key_to_cpu(eb, &key, path->slots[0]);
-
-	ret = add_all_parents(root, path, parents, level, &key, time_seq,
-				ref->wanted_disk_byte, extent_item_pos);
+	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;
-- 
cgit v1.2.3-70-g09d2


From 9345457f4a539a40056431aeb6f068750857472f Mon Sep 17 00:00:00 2001
From: Jan Schmidt <list.btrfs@jan-o-sch.net>
Date: Wed, 27 Jun 2012 15:23:09 +0200
Subject: Btrfs: support root level changes in __resolve_indirect_ref

With the tree mod log, we can have a tree that's two levels high, but
btrfs_search_old_slot may still return a path with the tree root at level
one instead. __resolve_indirect_ref must care for this and accept parents in
a lower level than expected.

Signed-off-by: Jan Schmidt <list.btrfs@jan-o-sch.net>
---
 fs/btrfs/backref.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)

(limited to 'fs/btrfs/backref.c')

diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index 7301cdb4b2c..cf0df904347 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -301,10 +301,14 @@ static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
 		goto out;
 
 	eb = path->nodes[level];
-	if (!eb) {
-		WARN_ON(1);
-		ret = 1;
-		goto out;
+	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,
-- 
cgit v1.2.3-70-g09d2


From 155725c9c051a343be5e555bf943da827e6cf721 Mon Sep 17 00:00:00 2001
From: Jan Schmidt <list.btrfs@jan-o-sch.net>
Date: Fri, 22 Jun 2012 14:01:00 +0200
Subject: Btrfs: leave critical region in btrfs_find_all_roots as soon as
 possible

When delayed refs exist, btrfs_find_all_roots used to hold the delayed ref
mutex way longer than actually required. We ought to drop it immediately
after we're done collecting all the delayed refs.

Signed-off-by: Jan Schmidt <list.btrfs@jan-o-sch.net>
---
 fs/btrfs/backref.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

(limited to 'fs/btrfs/backref.c')

diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index cf0df904347..a383c18e74e 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -839,6 +839,7 @@ again:
 			}
 			ret = __add_delayed_refs(head, delayed_ref_seq,
 						 &prefs_delayed);
+			mutex_unlock(&head->mutex);
 			if (ret) {
 				spin_unlock(&delayed_refs->lock);
 				goto out;
@@ -932,8 +933,6 @@ again:
 	}
 
 out:
-	if (head)
-		mutex_unlock(&head->mutex);
 	btrfs_free_path(path);
 	while (!list_empty(&prefs)) {
 		ref = list_first_entry(&prefs, struct __prelim_ref, list);
-- 
cgit v1.2.3-70-g09d2


From 097b8a7c9e48e2cb50fd0eb9315791921beaf484 Mon Sep 17 00:00:00 2001
From: Jan Schmidt <list.btrfs@jan-o-sch.net>
Date: Thu, 21 Jun 2012 11:08:04 +0200
Subject: Btrfs: join tree mod log code with the code holding back delayed refs

We've got two mechanisms both required for reliable backref resolving (tree
mod log and holding back delayed refs). You cannot make use of one without
the other. So instead of requiring the user of this mechanism to setup both
correctly, we join them into a single interface.

Additionally, we stop inserting non-blockers into fs_info->tree_mod_seq_list
as we did before, which was of no value.

Signed-off-by: Jan Schmidt <list.btrfs@jan-o-sch.net>
---
 fs/btrfs/backref.c     |  30 ++----
 fs/btrfs/backref.h     |   3 +-
 fs/btrfs/ctree.c       | 275 ++++++++++++++++++++++++++++++-------------------
 fs/btrfs/ctree.h       |  31 +++---
 fs/btrfs/delayed-ref.c |  44 ++++----
 fs/btrfs/delayed-ref.h |  49 +--------
 fs/btrfs/disk-io.c     |   2 +
 fs/btrfs/extent-tree.c |  21 ++--
 fs/btrfs/transaction.c |   4 -
 9 files changed, 240 insertions(+), 219 deletions(-)

(limited to 'fs/btrfs/backref.c')

diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index a383c18e74e..7d80ddd8f54 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -773,9 +773,8 @@ static int __add_keyed_refs(struct btrfs_fs_info *fs_info,
  */
 static int find_parent_nodes(struct btrfs_trans_handle *trans,
 			     struct btrfs_fs_info *fs_info, u64 bytenr,
-			     u64 delayed_ref_seq, u64 time_seq,
-			     struct ulist *refs, struct ulist *roots,
-			     const u64 *extent_item_pos)
+			     u64 time_seq, struct ulist *refs,
+			     struct ulist *roots, const u64 *extent_item_pos)
 {
 	struct btrfs_key key;
 	struct btrfs_path *path;
@@ -837,7 +836,7 @@ again:
 				btrfs_put_delayed_ref(&head->node);
 				goto again;
 			}
-			ret = __add_delayed_refs(head, delayed_ref_seq,
+			ret = __add_delayed_refs(head, time_seq,
 						 &prefs_delayed);
 			mutex_unlock(&head->mutex);
 			if (ret) {
@@ -981,8 +980,7 @@ static void free_leaf_list(struct ulist *blocks)
  */
 static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
 				struct btrfs_fs_info *fs_info, u64 bytenr,
-				u64 delayed_ref_seq, u64 time_seq,
-				struct ulist **leafs,
+				u64 time_seq, struct ulist **leafs,
 				const u64 *extent_item_pos)
 {
 	struct ulist *tmp;
@@ -997,7 +995,7 @@ static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
 		return -ENOMEM;
 	}
 
-	ret = find_parent_nodes(trans, fs_info, bytenr, delayed_ref_seq,
+	ret = find_parent_nodes(trans, fs_info, bytenr,
 				time_seq, *leafs, tmp, extent_item_pos);
 	ulist_free(tmp);
 
@@ -1024,8 +1022,7 @@ static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
  */
 int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
 				struct btrfs_fs_info *fs_info, u64 bytenr,
-				u64 delayed_ref_seq, u64 time_seq,
-				struct ulist **roots)
+				u64 time_seq, struct ulist **roots)
 {
 	struct ulist *tmp;
 	struct ulist_node *node = NULL;
@@ -1043,7 +1040,7 @@ int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
 
 	ULIST_ITER_INIT(&uiter);
 	while (1) {
-		ret = find_parent_nodes(trans, fs_info, bytenr, delayed_ref_seq,
+		ret = find_parent_nodes(trans, fs_info, bytenr,
 					time_seq, tmp, *roots, NULL);
 		if (ret < 0 && ret != -ENOENT) {
 			ulist_free(tmp);
@@ -1376,11 +1373,9 @@ int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
 	struct ulist *roots = NULL;
 	struct ulist_node *ref_node = NULL;
 	struct ulist_node *root_node = NULL;
-	struct seq_list seq_elem = {};
 	struct seq_list tree_mod_seq_elem = {};
 	struct ulist_iterator ref_uiter;
 	struct ulist_iterator root_uiter;
-	struct btrfs_delayed_ref_root *delayed_refs = NULL;
 
 	pr_debug("resolving all inodes for extent %llu\n",
 			extent_item_objectid);
@@ -1391,16 +1386,11 @@ int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
 		trans = btrfs_join_transaction(fs_info->extent_root);
 		if (IS_ERR(trans))
 			return PTR_ERR(trans);
-
-		delayed_refs = &trans->transaction->delayed_refs;
-		spin_lock(&delayed_refs->lock);
-		btrfs_get_delayed_seq(delayed_refs, &seq_elem);
-		spin_unlock(&delayed_refs->lock);
 		btrfs_get_tree_mod_seq(fs_info, &tree_mod_seq_elem);
 	}
 
 	ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
-				   seq_elem.seq, tree_mod_seq_elem.seq, &refs,
+				   tree_mod_seq_elem.seq, &refs,
 				   &extent_item_pos);
 	if (ret)
 		goto out;
@@ -1408,8 +1398,7 @@ int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
 	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,
-						seq_elem.seq,
-						tree_mod_seq_elem.seq, &roots);
+					   tree_mod_seq_elem.seq, &roots);
 		if (ret)
 			break;
 		ULIST_ITER_INIT(&root_uiter);
@@ -1431,7 +1420,6 @@ int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
 out:
 	if (!search_commit_root) {
 		btrfs_put_tree_mod_seq(fs_info, &tree_mod_seq_elem);
-		btrfs_put_delayed_seq(delayed_refs, &seq_elem);
 		btrfs_end_transaction(trans, fs_info->extent_root);
 	}
 
diff --git a/fs/btrfs/backref.h b/fs/btrfs/backref.h
index c18d8ac7b79..3a1ad3e2dcb 100644
--- a/fs/btrfs/backref.h
+++ b/fs/btrfs/backref.h
@@ -58,8 +58,7 @@ int paths_from_inode(u64 inum, struct inode_fs_paths *ipath);
 
 int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
 				struct btrfs_fs_info *fs_info, u64 bytenr,
-				u64 delayed_ref_seq, u64 time_seq,
-				struct ulist **roots);
+				u64 time_seq, struct ulist **roots);
 
 struct btrfs_data_container *init_data_container(u32 total_bytes);
 struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 67fe46fdee6..bef68ab3220 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -321,7 +321,7 @@ struct tree_mod_root {
 struct tree_mod_elem {
 	struct rb_node node;
 	u64 index;		/* shifted logical */
-	struct seq_list elem;
+	u64 seq;
 	enum mod_log_op op;
 
 	/* this is used for MOD_LOG_KEY_* and MOD_LOG_MOVE_KEYS operations */
@@ -341,20 +341,50 @@ struct tree_mod_elem {
 	struct tree_mod_root old_root;
 };
 
-static inline void
-__get_tree_mod_seq(struct btrfs_fs_info *fs_info, struct seq_list *elem)
+static inline void tree_mod_log_read_lock(struct btrfs_fs_info *fs_info)
 {
-	elem->seq = atomic_inc_return(&fs_info->tree_mod_seq);
-	list_add_tail(&elem->list, &fs_info->tree_mod_seq_list);
+	read_lock(&fs_info->tree_mod_log_lock);
 }
 
-void btrfs_get_tree_mod_seq(struct btrfs_fs_info *fs_info,
-			    struct seq_list *elem)
+static inline void tree_mod_log_read_unlock(struct btrfs_fs_info *fs_info)
+{
+	read_unlock(&fs_info->tree_mod_log_lock);
+}
+
+static inline void tree_mod_log_write_lock(struct btrfs_fs_info *fs_info)
+{
+	write_lock(&fs_info->tree_mod_log_lock);
+}
+
+static inline void tree_mod_log_write_unlock(struct btrfs_fs_info *fs_info)
 {
-	elem->flags = 1;
+	write_unlock(&fs_info->tree_mod_log_lock);
+}
+
+/*
+ * This adds a new blocker to the tree mod log's blocker list if the @elem
+ * passed does not already have a sequence number set. So when a caller expects
+ * to record tree modifications, it should ensure to set elem->seq to zero
+ * before calling btrfs_get_tree_mod_seq.
+ * Returns a fresh, unused tree log modification sequence number, even if no new
+ * blocker was added.
+ */
+u64 btrfs_get_tree_mod_seq(struct btrfs_fs_info *fs_info,
+			   struct seq_list *elem)
+{
+	u64 seq;
+
+	tree_mod_log_write_lock(fs_info);
 	spin_lock(&fs_info->tree_mod_seq_lock);
-	__get_tree_mod_seq(fs_info, elem);
+	if (!elem->seq) {
+		elem->seq = btrfs_inc_tree_mod_seq(fs_info);
+		list_add_tail(&elem->list, &fs_info->tree_mod_seq_list);
+	}
+	seq = btrfs_inc_tree_mod_seq(fs_info);
 	spin_unlock(&fs_info->tree_mod_seq_lock);
+	tree_mod_log_write_unlock(fs_info);
+
+	return seq;
 }
 
 void btrfs_put_tree_mod_seq(struct btrfs_fs_info *fs_info,
@@ -371,41 +401,46 @@ void btrfs_put_tree_mod_seq(struct btrfs_fs_info *fs_info,
 	if (!seq_putting)
 		return;
 
-	BUG_ON(!(elem->flags & 1));
 	spin_lock(&fs_info->tree_mod_seq_lock);
 	list_del(&elem->list);
+	elem->seq = 0;
 
 	list_for_each_entry(cur_elem, &fs_info->tree_mod_seq_list, list) {
-		if ((cur_elem->flags & 1) && cur_elem->seq < min_seq) {
+		if (cur_elem->seq < min_seq) {
 			if (seq_putting > cur_elem->seq) {
 				/*
 				 * blocker with lower sequence number exists, we
 				 * cannot remove anything from the log
 				 */
-				goto out;
+				spin_unlock(&fs_info->tree_mod_seq_lock);
+				return;
 			}
 			min_seq = cur_elem->seq;
 		}
 	}
+	spin_unlock(&fs_info->tree_mod_seq_lock);
+
+	/*
+	 * we removed the lowest blocker from the blocker list, so there may be
+	 * more processible delayed refs.
+	 */
+	wake_up(&fs_info->tree_mod_seq_wait);
 
 	/*
 	 * anything that's lower than the lowest existing (read: blocked)
 	 * sequence number can be removed from the tree.
 	 */
-	write_lock(&fs_info->tree_mod_log_lock);
+	tree_mod_log_write_lock(fs_info);
 	tm_root = &fs_info->tree_mod_log;
 	for (node = rb_first(tm_root); node; node = next) {
 		next = rb_next(node);
 		tm = container_of(node, struct tree_mod_elem, node);
-		if (tm->elem.seq > min_seq)
+		if (tm->seq > min_seq)
 			continue;
 		rb_erase(node, tm_root);
-		list_del(&tm->elem.list);
 		kfree(tm);
 	}
-	write_unlock(&fs_info->tree_mod_log_lock);
-out:
-	spin_unlock(&fs_info->tree_mod_seq_lock);
+	tree_mod_log_write_unlock(fs_info);
 }
 
 /*
@@ -423,11 +458,9 @@ __tree_mod_log_insert(struct btrfs_fs_info *fs_info, struct tree_mod_elem *tm)
 	struct rb_node **new;
 	struct rb_node *parent = NULL;
 	struct tree_mod_elem *cur;
-	int ret = 0;
 
-	BUG_ON(!tm || !tm->elem.seq);
+	BUG_ON(!tm || !tm->seq);
 
-	write_lock(&fs_info->tree_mod_log_lock);
 	tm_root = &fs_info->tree_mod_log;
 	new = &tm_root->rb_node;
 	while (*new) {
@@ -437,88 +470,81 @@ __tree_mod_log_insert(struct btrfs_fs_info *fs_info, struct tree_mod_elem *tm)
 			new = &((*new)->rb_left);
 		else if (cur->index > tm->index)
 			new = &((*new)->rb_right);
-		else if (cur->elem.seq < tm->elem.seq)
+		else if (cur->seq < tm->seq)
 			new = &((*new)->rb_left);
-		else if (cur->elem.seq > tm->elem.seq)
+		else if (cur->seq > tm->seq)
 			new = &((*new)->rb_right);
 		else {
 			kfree(tm);
-			ret = -EEXIST;
-			goto unlock;
+			return -EEXIST;
 		}
 	}
 
 	rb_link_node(&tm->node, parent, new);
 	rb_insert_color(&tm->node, tm_root);
-unlock:
-	write_unlock(&fs_info->tree_mod_log_lock);
-	return ret;
+	return 0;
 }
 
+/*
+ * Determines if logging can be omitted. Returns 1 if it can. Otherwise, it
+ * returns zero with the tree_mod_log_lock acquired. The caller must hold
+ * this until all tree mod log insertions are recorded in the rb tree and then
+ * call tree_mod_log_write_unlock() to release.
+ */
 static inline int tree_mod_dont_log(struct btrfs_fs_info *fs_info,
 				    struct extent_buffer *eb) {
 	smp_mb();
 	if (list_empty(&(fs_info)->tree_mod_seq_list))
 		return 1;
-	if (!eb)
-		return 0;
-	if (btrfs_header_level(eb) == 0)
+	if (eb && btrfs_header_level(eb) == 0)
 		return 1;
+
+	tree_mod_log_write_lock(fs_info);
+	if (list_empty(&fs_info->tree_mod_seq_list)) {
+		/*
+		 * someone emptied the list while we were waiting for the lock.
+		 * we must not add to the list when no blocker exists.
+		 */
+		tree_mod_log_write_unlock(fs_info);
+		return 1;
+	}
+
 	return 0;
 }
 
 /*
- * This allocates memory and gets a tree modification sequence number when
- * needed.
+ * This allocates memory and gets a tree modification sequence number.
  *
- * Returns 0 when no sequence number is needed, < 0 on error.
- * Returns 1 when a sequence number was added. In this case,
- * fs_info->tree_mod_seq_lock was acquired and must be released by the caller
- * after inserting into the rb tree.
+ * Returns <0 on error.
+ * Returns >0 (the added sequence number) on success.
  */
 static inline int tree_mod_alloc(struct btrfs_fs_info *fs_info, gfp_t flags,
 				 struct tree_mod_elem **tm_ret)
 {
 	struct tree_mod_elem *tm;
-	int seq;
 
-	if (tree_mod_dont_log(fs_info, NULL))
-		return 0;
-
-	tm = *tm_ret = kzalloc(sizeof(*tm), flags);
+	/*
+	 * once we switch from spin locks to something different, we should
+	 * honor the flags parameter here.
+	 */
+	tm = *tm_ret = kzalloc(sizeof(*tm), GFP_ATOMIC);
 	if (!tm)
 		return -ENOMEM;
 
-	tm->elem.flags = 0;
-	spin_lock(&fs_info->tree_mod_seq_lock);
-	if (list_empty(&fs_info->tree_mod_seq_list)) {
-		/*
-		 * someone emptied the list while we were waiting for the lock.
-		 * we must not add to the list, because no blocker exists. items
-		 * are removed from the list only when the existing blocker is
-		 * removed from the list.
-		 */
-		kfree(tm);
-		seq = 0;
-		spin_unlock(&fs_info->tree_mod_seq_lock);
-	} else {
-		__get_tree_mod_seq(fs_info, &tm->elem);
-		seq = tm->elem.seq;
-	}
-
-	return seq;
+	tm->seq = btrfs_inc_tree_mod_seq(fs_info);
+	return tm->seq;
 }
 
-static noinline int
-tree_mod_log_insert_key_mask(struct btrfs_fs_info *fs_info,
-			     struct extent_buffer *eb, int slot,
-			     enum mod_log_op op, gfp_t flags)
+static inline int
+__tree_mod_log_insert_key(struct btrfs_fs_info *fs_info,
+			  struct extent_buffer *eb, int slot,
+			  enum mod_log_op op, gfp_t flags)
 {
-	struct tree_mod_elem *tm;
 	int ret;
+	struct tree_mod_elem *tm;
 
 	ret = tree_mod_alloc(fs_info, flags, &tm);
-	if (ret <= 0)
+	if (ret < 0)
 		return ret;
 
 	tm->index = eb->start >> PAGE_CACHE_SHIFT;
@@ -530,8 +556,22 @@ tree_mod_log_insert_key_mask(struct btrfs_fs_info *fs_info,
 	tm->slot = slot;
 	tm->generation = btrfs_node_ptr_generation(eb, slot);
 
-	ret = __tree_mod_log_insert(fs_info, tm);
-	spin_unlock(&fs_info->tree_mod_seq_lock);
+	return __tree_mod_log_insert(fs_info, tm);
+}
+
+static noinline int
+tree_mod_log_insert_key_mask(struct btrfs_fs_info *fs_info,
+			     struct extent_buffer *eb, int slot,
+			     enum mod_log_op op, gfp_t flags)
+{
+	int ret;
+
+	if (tree_mod_dont_log(fs_info, eb))
+		return 0;
+
+	ret = __tree_mod_log_insert_key(fs_info, eb, slot, op, flags);
+
+	tree_mod_log_write_unlock(fs_info);
 	return ret;
 }
 
@@ -542,6 +582,14 @@ tree_mod_log_insert_key(struct btrfs_fs_info *fs_info, struct extent_buffer *eb,
 	return tree_mod_log_insert_key_mask(fs_info, eb, slot, op, GFP_NOFS);
 }
 
+static noinline int
+tree_mod_log_insert_key_locked(struct btrfs_fs_info *fs_info,
+			     struct extent_buffer *eb, int slot,
+			     enum mod_log_op op)
+{
+	return __tree_mod_log_insert_key(fs_info, eb, slot, op, GFP_NOFS);
+}
+
 static noinline int
 tree_mod_log_insert_move(struct btrfs_fs_info *fs_info,
 			 struct extent_buffer *eb, int dst_slot, int src_slot,
@@ -555,14 +603,14 @@ tree_mod_log_insert_move(struct btrfs_fs_info *fs_info,
 		return 0;
 
 	for (i = 0; i + dst_slot < src_slot && i < nr_items; i++) {
-		ret = tree_mod_log_insert_key(fs_info, eb, i + dst_slot,
+		ret = tree_mod_log_insert_key_locked(fs_info, eb, i + dst_slot,
 					      MOD_LOG_KEY_REMOVE_WHILE_MOVING);
 		BUG_ON(ret < 0);
 	}
 
 	ret = tree_mod_alloc(fs_info, flags, &tm);
-	if (ret <= 0)
-		return ret;
+	if (ret < 0)
+		goto out;
 
 	tm->index = eb->start >> PAGE_CACHE_SHIFT;
 	tm->slot = src_slot;
@@ -571,10 +619,26 @@ tree_mod_log_insert_move(struct btrfs_fs_info *fs_info,
 	tm->op = MOD_LOG_MOVE_KEYS;
 
 	ret = __tree_mod_log_insert(fs_info, tm);
-	spin_unlock(&fs_info->tree_mod_seq_lock);
+out:
+	tree_mod_log_write_unlock(fs_info);
 	return ret;
 }
 
+static inline void
+__tree_mod_log_free_eb(struct btrfs_fs_info *fs_info, struct extent_buffer *eb)
+{
+	int i;
+	u32 nritems;
+	int ret;
+
+	nritems = btrfs_header_nritems(eb);
+	for (i = nritems - 1; i >= 0; i--) {
+		ret = tree_mod_log_insert_key_locked(fs_info, eb, i,
+					      MOD_LOG_KEY_REMOVE_WHILE_FREEING);
+		BUG_ON(ret < 0);
+	}
+}
+
 static noinline int
 tree_mod_log_insert_root(struct btrfs_fs_info *fs_info,
 			 struct extent_buffer *old_root,
@@ -583,9 +647,14 @@ tree_mod_log_insert_root(struct btrfs_fs_info *fs_info,
 	struct tree_mod_elem *tm;
 	int ret;
 
+	if (tree_mod_dont_log(fs_info, NULL))
+		return 0;
+
+	__tree_mod_log_free_eb(fs_info, old_root);
+
 	ret = tree_mod_alloc(fs_info, flags, &tm);
-	if (ret <= 0)
-		return ret;
+	if (ret < 0)
+		goto out;
 
 	tm->index = new_root->start >> PAGE_CACHE_SHIFT;
 	tm->old_root.logical = old_root->start;
@@ -594,7 +663,8 @@ tree_mod_log_insert_root(struct btrfs_fs_info *fs_info,
 	tm->op = MOD_LOG_ROOT_REPLACE;
 
 	ret = __tree_mod_log_insert(fs_info, tm);
-	spin_unlock(&fs_info->tree_mod_seq_lock);
+out:
+	tree_mod_log_write_unlock(fs_info);
 	return ret;
 }
 
@@ -608,7 +678,7 @@ __tree_mod_log_search(struct btrfs_fs_info *fs_info, u64 start, u64 min_seq,
 	struct tree_mod_elem *found = NULL;
 	u64 index = start >> PAGE_CACHE_SHIFT;
 
-	read_lock(&fs_info->tree_mod_log_lock);
+	tree_mod_log_read_lock(fs_info);
 	tm_root = &fs_info->tree_mod_log;
 	node = tm_root->rb_node;
 	while (node) {
@@ -617,18 +687,18 @@ __tree_mod_log_search(struct btrfs_fs_info *fs_info, u64 start, u64 min_seq,
 			node = node->rb_left;
 		} else if (cur->index > index) {
 			node = node->rb_right;
-		} else if (cur->elem.seq < min_seq) {
+		} else if (cur->seq < min_seq) {
 			node = node->rb_left;
 		} else if (!smallest) {
 			/* we want the node with the highest seq */
 			if (found)
-				BUG_ON(found->elem.seq > cur->elem.seq);
+				BUG_ON(found->seq > cur->seq);
 			found = cur;
 			node = node->rb_left;
-		} else if (cur->elem.seq > min_seq) {
+		} else if (cur->seq > min_seq) {
 			/* we want the node with the smallest seq */
 			if (found)
-				BUG_ON(found->elem.seq < cur->elem.seq);
+				BUG_ON(found->seq < cur->seq);
 			found = cur;
 			node = node->rb_right;
 		} else {
@@ -636,7 +706,7 @@ __tree_mod_log_search(struct btrfs_fs_info *fs_info, u64 start, u64 min_seq,
 			break;
 		}
 	}
-	read_unlock(&fs_info->tree_mod_log_lock);
+	tree_mod_log_read_unlock(fs_info);
 
 	return found;
 }
@@ -664,7 +734,7 @@ tree_mod_log_search(struct btrfs_fs_info *fs_info, u64 start, u64 min_seq)
 	return __tree_mod_log_search(fs_info, start, min_seq, 0);
 }
 
-static inline void
+static noinline void
 tree_mod_log_eb_copy(struct btrfs_fs_info *fs_info, struct extent_buffer *dst,
 		     struct extent_buffer *src, unsigned long dst_offset,
 		     unsigned long src_offset, int nr_items)
@@ -675,18 +745,23 @@ tree_mod_log_eb_copy(struct btrfs_fs_info *fs_info, struct extent_buffer *dst,
 	if (tree_mod_dont_log(fs_info, NULL))
 		return;
 
-	if (btrfs_header_level(dst) == 0 && btrfs_header_level(src) == 0)
+	if (btrfs_header_level(dst) == 0 && btrfs_header_level(src) == 0) {
+		tree_mod_log_write_unlock(fs_info);
 		return;
+	}
 
-	/* speed this up by single seq for all operations? */
 	for (i = 0; i < nr_items; i++) {
-		ret = tree_mod_log_insert_key(fs_info, src, i + src_offset,
-					      MOD_LOG_KEY_REMOVE);
+		ret = tree_mod_log_insert_key_locked(fs_info, src,
+						     i + src_offset,
+						     MOD_LOG_KEY_REMOVE);
 		BUG_ON(ret < 0);
-		ret = tree_mod_log_insert_key(fs_info, dst, i + dst_offset,
-					      MOD_LOG_KEY_ADD);
+		ret = tree_mod_log_insert_key_locked(fs_info, dst,
+						     i + dst_offset,
+						     MOD_LOG_KEY_ADD);
 		BUG_ON(ret < 0);
 	}
+
+	tree_mod_log_write_unlock(fs_info);
 }
 
 static inline void
@@ -699,7 +774,7 @@ tree_mod_log_eb_move(struct btrfs_fs_info *fs_info, struct extent_buffer *dst,
 	BUG_ON(ret < 0);
 }
 
-static inline void
+static noinline void
 tree_mod_log_set_node_key(struct btrfs_fs_info *fs_info,
 			  struct extent_buffer *eb,
 			  struct btrfs_disk_key *disk_key, int slot, int atomic)
@@ -712,30 +787,22 @@ tree_mod_log_set_node_key(struct btrfs_fs_info *fs_info,
 	BUG_ON(ret < 0);
 }
 
-static void tree_mod_log_free_eb(struct btrfs_fs_info *fs_info,
-				 struct extent_buffer *eb)
+static noinline void
+tree_mod_log_free_eb(struct btrfs_fs_info *fs_info, struct extent_buffer *eb)
 {
-	int i;
-	int ret;
-	u32 nritems;
-
 	if (tree_mod_dont_log(fs_info, eb))
 		return;
 
-	nritems = btrfs_header_nritems(eb);
-	for (i = nritems - 1; i >= 0; i--) {
-		ret = tree_mod_log_insert_key(fs_info, eb, i,
-					      MOD_LOG_KEY_REMOVE_WHILE_FREEING);
-		BUG_ON(ret < 0);
-	}
+	__tree_mod_log_free_eb(fs_info, eb);
+
+	tree_mod_log_write_unlock(fs_info);
 }
 
-static inline void
+static noinline void
 tree_mod_log_set_root_pointer(struct btrfs_root *root,
 			      struct extent_buffer *new_root_node)
 {
 	int ret;
-	tree_mod_log_free_eb(root->fs_info, root->node);
 	ret = tree_mod_log_insert_root(root->fs_info, root->node,
 				       new_root_node, GFP_NOFS);
 	BUG_ON(ret < 0);
@@ -1069,7 +1136,7 @@ __tree_mod_log_rewind(struct extent_buffer *eb, u64 time_seq,
 	unsigned long p_size = sizeof(struct btrfs_key_ptr);
 
 	n = btrfs_header_nritems(eb);
-	while (tm && tm->elem.seq >= time_seq) {
+	while (tm && tm->seq >= time_seq) {
 		/*
 		 * all the operations are recorded with the operator used for
 		 * the modification. as we're going backwards, we do the
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 84ac723f58f..8f8dc46f44e 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1030,6 +1030,13 @@ struct btrfs_block_group_cache {
 	struct list_head cluster_list;
 };
 
+/* delayed seq elem */
+struct seq_list {
+	struct list_head list;
+	u64 seq;
+};
+
+/* fs_info */
 struct reloc_control;
 struct btrfs_device;
 struct btrfs_fs_devices;
@@ -1144,6 +1151,8 @@ struct btrfs_fs_info {
 	spinlock_t tree_mod_seq_lock;
 	atomic_t tree_mod_seq;
 	struct list_head tree_mod_seq_list;
+	struct seq_list tree_mod_seq_elem;
+	wait_queue_head_t tree_mod_seq_wait;
 
 	/* this protects tree_mod_log */
 	rwlock_t tree_mod_log_lock;
@@ -2798,6 +2807,16 @@ static inline void free_fs_info(struct btrfs_fs_info *fs_info)
 	kfree(fs_info);
 }
 
+/* tree mod log functions from ctree.c */
+u64 btrfs_get_tree_mod_seq(struct btrfs_fs_info *fs_info,
+			   struct seq_list *elem);
+void btrfs_put_tree_mod_seq(struct btrfs_fs_info *fs_info,
+			    struct seq_list *elem);
+static inline u64 btrfs_inc_tree_mod_seq(struct btrfs_fs_info *fs_info)
+{
+	return atomic_inc_return(&fs_info->tree_mod_seq);
+}
+
 /* root-item.c */
 int btrfs_find_root_ref(struct btrfs_root *tree_root,
 			struct btrfs_path *path,
@@ -3157,18 +3176,6 @@ void btrfs_reada_detach(void *handle);
 int btree_readahead_hook(struct btrfs_root *root, struct extent_buffer *eb,
 			 u64 start, int err);
 
-/* delayed seq elem */
-struct seq_list {
-	struct list_head list;
-	u64 seq;
-	u32 flags;
-};
-
-void btrfs_get_tree_mod_seq(struct btrfs_fs_info *fs_info,
-			    struct seq_list *elem);
-void btrfs_put_tree_mod_seq(struct btrfs_fs_info *fs_info,
-			    struct seq_list *elem);
-
 static inline int is_fstree(u64 rootid)
 {
 	if (rootid == BTRFS_FS_TREE_OBJECTID ||
diff --git a/fs/btrfs/delayed-ref.c b/fs/btrfs/delayed-ref.c
index 13ae7b04790..21a75771763 100644
--- a/fs/btrfs/delayed-ref.c
+++ b/fs/btrfs/delayed-ref.c
@@ -233,22 +233,26 @@ int btrfs_delayed_ref_lock(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
-int btrfs_check_delayed_seq(struct btrfs_delayed_ref_root *delayed_refs,
+int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info,
+			    struct btrfs_delayed_ref_root *delayed_refs,
 			    u64 seq)
 {
 	struct seq_list *elem;
-
-	assert_spin_locked(&delayed_refs->lock);
-	if (list_empty(&delayed_refs->seq_head))
-		return 0;
-
-	elem = list_first_entry(&delayed_refs->seq_head, struct seq_list, list);
-	if (seq >= elem->seq) {
-		pr_debug("holding back delayed_ref %llu, lowest is %llu (%p)\n",
-			 seq, elem->seq, delayed_refs);
-		return 1;
+	int ret = 0;
+
+	spin_lock(&fs_info->tree_mod_seq_lock);
+	if (!list_empty(&fs_info->tree_mod_seq_list)) {
+		elem = list_first_entry(&fs_info->tree_mod_seq_list,
+					struct seq_list, list);
+		if (seq >= elem->seq) {
+			pr_debug("holding back delayed_ref %llu, lowest is "
+				 "%llu (%p)\n", seq, elem->seq, delayed_refs);
+			ret = 1;
+		}
 	}
-	return 0;
+
+	spin_unlock(&fs_info->tree_mod_seq_lock);
+	return ret;
 }
 
 int btrfs_find_ref_cluster(struct btrfs_trans_handle *trans,
@@ -526,7 +530,7 @@ static noinline void add_delayed_tree_ref(struct btrfs_fs_info *fs_info,
 	ref->in_tree = 1;
 
 	if (is_fstree(ref_root))
-		seq = inc_delayed_seq(delayed_refs);
+		seq = btrfs_inc_tree_mod_seq(fs_info);
 	ref->seq = seq;
 
 	full_ref = btrfs_delayed_node_to_tree_ref(ref);
@@ -585,7 +589,7 @@ static noinline void add_delayed_data_ref(struct btrfs_fs_info *fs_info,
 	ref->in_tree = 1;
 
 	if (is_fstree(ref_root))
-		seq = inc_delayed_seq(delayed_refs);
+		seq = btrfs_inc_tree_mod_seq(fs_info);
 	ref->seq = seq;
 
 	full_ref = btrfs_delayed_node_to_data_ref(ref);
@@ -659,8 +663,8 @@ int btrfs_add_delayed_tree_ref(struct btrfs_fs_info *fs_info,
 				   num_bytes, parent, ref_root, level, action,
 				   for_cow);
 	if (!is_fstree(ref_root) &&
-	    waitqueue_active(&delayed_refs->seq_wait))
-		wake_up(&delayed_refs->seq_wait);
+	    waitqueue_active(&fs_info->tree_mod_seq_wait))
+		wake_up(&fs_info->tree_mod_seq_wait);
 	spin_unlock(&delayed_refs->lock);
 
 	return 0;
@@ -708,8 +712,8 @@ int btrfs_add_delayed_data_ref(struct btrfs_fs_info *fs_info,
 				   num_bytes, parent, ref_root, owner, offset,
 				   action, for_cow);
 	if (!is_fstree(ref_root) &&
-	    waitqueue_active(&delayed_refs->seq_wait))
-		wake_up(&delayed_refs->seq_wait);
+	    waitqueue_active(&fs_info->tree_mod_seq_wait))
+		wake_up(&fs_info->tree_mod_seq_wait);
 	spin_unlock(&delayed_refs->lock);
 
 	return 0;
@@ -736,8 +740,8 @@ int btrfs_add_delayed_extent_op(struct btrfs_fs_info *fs_info,
 				   num_bytes, BTRFS_UPDATE_DELAYED_HEAD,
 				   extent_op->is_data);
 
-	if (waitqueue_active(&delayed_refs->seq_wait))
-		wake_up(&delayed_refs->seq_wait);
+	if (waitqueue_active(&fs_info->tree_mod_seq_wait))
+		wake_up(&fs_info->tree_mod_seq_wait);
 	spin_unlock(&delayed_refs->lock);
 	return 0;
 }
diff --git a/fs/btrfs/delayed-ref.h b/fs/btrfs/delayed-ref.h
index 413927fb995..2b5cb27f986 100644
--- a/fs/btrfs/delayed-ref.h
+++ b/fs/btrfs/delayed-ref.h
@@ -139,26 +139,6 @@ struct btrfs_delayed_ref_root {
 	int flushing;
 
 	u64 run_delayed_start;
-
-	/*
-	 * seq number of delayed refs. We need to know if a backref was being
-	 * added before the currently processed ref or afterwards.
-	 */
-	u64 seq;
-
-	/*
-	 * seq_list holds a list of all seq numbers that are currently being
-	 * added to the list. While walking backrefs (btrfs_find_all_roots,
-	 * qgroups), which might take some time, no newer ref must be processed,
-	 * as it might influence the outcome of the walk.
-	 */
-	struct list_head seq_head;
-
-	/*
-	 * when the only refs we have in the list must not be processed, we want
-	 * to wait for more refs to show up or for the end of backref walking.
-	 */
-	wait_queue_head_t seq_wait;
 };
 
 static inline void btrfs_put_delayed_ref(struct btrfs_delayed_ref_node *ref)
@@ -195,33 +175,8 @@ int btrfs_delayed_ref_lock(struct btrfs_trans_handle *trans,
 int btrfs_find_ref_cluster(struct btrfs_trans_handle *trans,
 			   struct list_head *cluster, u64 search_start);
 
-static inline u64 inc_delayed_seq(struct btrfs_delayed_ref_root *delayed_refs)
-{
-	assert_spin_locked(&delayed_refs->lock);
-	++delayed_refs->seq;
-	return delayed_refs->seq;
-}
-
-static inline void
-btrfs_get_delayed_seq(struct btrfs_delayed_ref_root *delayed_refs,
-		      struct seq_list *elem)
-{
-	assert_spin_locked(&delayed_refs->lock);
-	elem->seq = delayed_refs->seq;
-	list_add_tail(&elem->list, &delayed_refs->seq_head);
-}
-
-static inline void
-btrfs_put_delayed_seq(struct btrfs_delayed_ref_root *delayed_refs,
-		      struct seq_list *elem)
-{
-	spin_lock(&delayed_refs->lock);
-	list_del(&elem->list);
-	wake_up(&delayed_refs->seq_wait);
-	spin_unlock(&delayed_refs->lock);
-}
-
-int btrfs_check_delayed_seq(struct btrfs_delayed_ref_root *delayed_refs,
+int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info,
+			    struct btrfs_delayed_ref_root *delayed_refs,
 			    u64 seq);
 
 /*
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 8cc47103a32..19a39e10d6f 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1944,6 +1944,8 @@ int open_ctree(struct super_block *sb,
 	fs_info->free_chunk_space = 0;
 	fs_info->tree_mod_log = RB_ROOT;
 
+	init_waitqueue_head(&fs_info->tree_mod_seq_wait);
+
 	/* readahead state */
 	INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
 	spin_lock_init(&fs_info->reada_lock);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 6e1d36702ff..94ce79f76e5 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -2217,6 +2217,7 @@ static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
 	struct btrfs_delayed_ref_node *ref;
 	struct btrfs_delayed_ref_head *locked_ref = NULL;
 	struct btrfs_delayed_extent_op *extent_op;
+	struct btrfs_fs_info *fs_info = root->fs_info;
 	int ret;
 	int count = 0;
 	int must_insert_reserved = 0;
@@ -2255,7 +2256,7 @@ static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
 		ref = select_delayed_ref(locked_ref);
 
 		if (ref && ref->seq &&
-		    btrfs_check_delayed_seq(delayed_refs, ref->seq)) {
+		    btrfs_check_delayed_seq(fs_info, delayed_refs, ref->seq)) {
 			/*
 			 * there are still refs with lower seq numbers in the
 			 * process of being added. Don't run this ref yet.
@@ -2337,7 +2338,7 @@ static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
 		}
 
 next:
-		do_chunk_alloc(trans, root->fs_info->extent_root,
+		do_chunk_alloc(trans, fs_info->extent_root,
 			       2 * 1024 * 1024,
 			       btrfs_get_alloc_profile(root, 0),
 			       CHUNK_ALLOC_NO_FORCE);
@@ -2347,18 +2348,19 @@ next:
 	return count;
 }
 
-static void wait_for_more_refs(struct btrfs_delayed_ref_root *delayed_refs,
+static void wait_for_more_refs(struct btrfs_fs_info *fs_info,
+			       struct btrfs_delayed_ref_root *delayed_refs,
 			       unsigned long num_refs,
 			       struct list_head *first_seq)
 {
 	spin_unlock(&delayed_refs->lock);
 	pr_debug("waiting for more refs (num %ld, first %p)\n",
 		 num_refs, first_seq);
-	wait_event(delayed_refs->seq_wait,
+	wait_event(fs_info->tree_mod_seq_wait,
 		   num_refs != delayed_refs->num_entries ||
-		   delayed_refs->seq_head.next != first_seq);
+		   fs_info->tree_mod_seq_list.next != first_seq);
 	pr_debug("done waiting for more refs (num %ld, first %p)\n",
-		 delayed_refs->num_entries, delayed_refs->seq_head.next);
+		 delayed_refs->num_entries, fs_info->tree_mod_seq_list.next);
 	spin_lock(&delayed_refs->lock);
 }
 
@@ -2403,6 +2405,7 @@ int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
 again:
 	consider_waiting = 0;
 	spin_lock(&delayed_refs->lock);
+
 	if (count == 0) {
 		count = delayed_refs->num_entries * 2;
 		run_most = 1;
@@ -2437,7 +2440,7 @@ again:
 				num_refs = delayed_refs->num_entries;
 				first_seq = root->fs_info->tree_mod_seq_list.next;
 			} else {
-				wait_for_more_refs(delayed_refs,
+				wait_for_more_refs(root->fs_info, delayed_refs,
 						   num_refs, first_seq);
 				/*
 				 * after waiting, things have changed. we
@@ -5190,8 +5193,8 @@ static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
 	rb_erase(&head->node.rb_node, &delayed_refs->root);
 
 	delayed_refs->num_entries--;
-	if (waitqueue_active(&delayed_refs->seq_wait))
-		wake_up(&delayed_refs->seq_wait);
+	if (waitqueue_active(&root->fs_info->tree_mod_seq_wait))
+		wake_up(&root->fs_info->tree_mod_seq_wait);
 
 	/*
 	 * we don't take a ref on the node because we're removing it from the
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index b72b068183e..621c8dc48fb 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -38,7 +38,6 @@ void put_transaction(struct btrfs_transaction *transaction)
 	if (atomic_dec_and_test(&transaction->use_count)) {
 		BUG_ON(!list_empty(&transaction->list));
 		WARN_ON(transaction->delayed_refs.root.rb_node);
-		WARN_ON(!list_empty(&transaction->delayed_refs.seq_head));
 		memset(transaction, 0, sizeof(*transaction));
 		kmem_cache_free(btrfs_transaction_cachep, transaction);
 	}
@@ -126,7 +125,6 @@ loop:
 	cur_trans->delayed_refs.num_heads = 0;
 	cur_trans->delayed_refs.flushing = 0;
 	cur_trans->delayed_refs.run_delayed_start = 0;
-	cur_trans->delayed_refs.seq = 1;
 
 	/*
 	 * although the tree mod log is per file system and not per transaction,
@@ -145,10 +143,8 @@ loop:
 	}
 	atomic_set(&fs_info->tree_mod_seq, 0);
 
-	init_waitqueue_head(&cur_trans->delayed_refs.seq_wait);
 	spin_lock_init(&cur_trans->commit_lock);
 	spin_lock_init(&cur_trans->delayed_refs.lock);
-	INIT_LIST_HEAD(&cur_trans->delayed_refs.seq_head);
 
 	INIT_LIST_HEAD(&cur_trans->pending_snapshots);
 	list_add_tail(&cur_trans->list, &fs_info->trans_list);
-- 
cgit v1.2.3-70-g09d2


From 91cb916ca26feb99c78c131a1643af3d10fefd96 Mon Sep 17 00:00:00 2001
From: Alexander Block <ablock84@googlemail.com>
Date: Sun, 3 Jun 2012 14:23:23 +0200
Subject: Btrfs: make iref_to_path non static

Make iref_to_path non static (needed in send) and rename
it to btrfs_iref_to_path

Signed-off-by: Alexander Block <ablock84@googlemail.com>
---
 fs/btrfs/backref.c | 10 +++++-----
 fs/btrfs/backref.h |  4 ++++
 2 files changed, 9 insertions(+), 5 deletions(-)

(limited to 'fs/btrfs/backref.c')

diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index a383c18e74e..e99fe0e31da 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -1125,10 +1125,10 @@ 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;
@@ -1543,7 +1543,7 @@ static int inode_to_path(u64 inum, struct btrfs_inode_ref *iref,
 					ipath->fspath->bytes_left - s_ptr : 0;
 
 	fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
-	fspath = iref_to_path(ipath->fs_root, ipath->btrfs_path, iref, eb,
+	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);
diff --git a/fs/btrfs/backref.h b/fs/btrfs/backref.h
index c18d8ac7b79..1a765792fbf 100644
--- a/fs/btrfs/backref.h
+++ b/fs/btrfs/backref.h
@@ -21,6 +21,7 @@
 
 #include "ioctl.h"
 #include "ulist.h"
+#include "extent_io.h"
 
 #define BTRFS_BACKREF_SEARCH_COMMIT_ROOT ((struct btrfs_trans_handle *)0)
 
@@ -60,6 +61,9 @@ int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
 				struct btrfs_fs_info *fs_info, u64 bytenr,
 				u64 delayed_ref_seq, u64 time_seq,
 				struct ulist **roots);
+char *btrfs_iref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
+			 struct btrfs_inode_ref *iref, struct extent_buffer *eb,
+			 u64 parent, char *dest, u32 size);
 
 struct btrfs_data_container *init_data_container(u32 total_bytes);
 struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
-- 
cgit v1.2.3-70-g09d2


From 3627bf4503b504077332c13496cb1bd54713bcbb Mon Sep 17 00:00:00 2001
From: Stefan Behrens <sbehrens@giantdisaster.de>
Date: Wed, 1 Aug 2012 04:28:01 -0600
Subject: Btrfs: fix that error value is changed by mistake

In iterate_inodes_from_logical() the error result from
extent_from_logical() is patched by mistake. Typically ENOENT is
patched to EINVAL because (-ENOENT & BTRFS_EXTENT_FLAG_TREE_BLOCK)
evaluates to true.

Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
---
 fs/btrfs/backref.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

(limited to 'fs/btrfs/backref.c')

diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index a256f3b2a84..ff6475f409d 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -1438,10 +1438,10 @@ int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
 	ret = extent_from_logical(fs_info, logical, path,
 					&found_key);
 	btrfs_release_path(path);
-	if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
-		ret = -EINVAL;
 	if (ret < 0)
 		return ret;
+	if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
+		return -EINVAL;
 
 	extent_item_pos = logical - found_key.objectid;
 	ret = iterate_extent_inodes(fs_info, found_key.objectid,
-- 
cgit v1.2.3-70-g09d2