diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 7ed6b39..0e2dcc7 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -6,7 +6,7 @@
 	   hash.o file-item.o inode-item.o inode-map.o disk-io.o \
 	   transaction.o bit-radix.o inode.o file.o tree-defrag.o \
 	   extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \
-	   extent_io.o volumes.o async-thread.o ioctl.o
+	   extent_io.o volumes.o async-thread.o ioctl.o locking.o
 
 btrfs-$(CONFIG_FS_POSIX_ACL)	+= acl.o
 else
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 1d404bd..75625c6 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -21,6 +21,7 @@
 #include "disk-io.h"
 #include "transaction.h"
 #include "print-tree.h"
+#include "locking.h"
 
 static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_path *path, int level);
@@ -64,12 +65,47 @@
 	int i;
 	for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
 		if (!p->nodes[i])
-			break;
+			continue;
+		if (p->locks[i]) {
+			btrfs_tree_unlock(p->nodes[i]);
+			p->locks[i] = 0;
+		}
 		free_extent_buffer(p->nodes[i]);
 	}
 	memset(p, 0, sizeof(*p));
 }
 
+struct extent_buffer *btrfs_root_node(struct btrfs_root *root)
+{
+	struct extent_buffer *eb;
+	spin_lock(&root->node_lock);
+	eb = root->node;
+	extent_buffer_get(eb);
+	spin_unlock(&root->node_lock);
+	return eb;
+}
+
+struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root)
+{
+	struct extent_buffer *eb;
+
+	while(1) {
+		eb = btrfs_root_node(root);
+		btrfs_tree_lock(eb);
+
+		spin_lock(&root->node_lock);
+		if (eb == root->node) {
+			spin_unlock(&root->node_lock);
+			break;
+		}
+		spin_unlock(&root->node_lock);
+
+		btrfs_tree_unlock(eb);
+		free_extent_buffer(eb);
+	}
+	return eb;
+}
+
 static void add_root_to_dirty_list(struct btrfs_root *root)
 {
 	if (root->track_dirty && list_empty(&root->dirty_list)) {
@@ -111,7 +147,7 @@
 	} else {
 		first_key.objectid = 0;
 	}
-	cow = __btrfs_alloc_free_block(trans, new_root, buf->len,
+	cow = btrfs_alloc_free_block(trans, new_root, buf->len,
 				       new_root_objectid,
 				       trans->transid, first_key.objectid,
 				       level, buf->start, 0);
@@ -151,8 +187,14 @@
 	int ret = 0;
 	int different_trans = 0;
 	int level;
+	int unlock_orig = 0;
 	struct btrfs_key first_key;
 
+	if (*cow_ret == buf)
+		unlock_orig = 1;
+
+	WARN_ON(!btrfs_tree_locked(buf));
+
 	if (root->ref_cows) {
 		root_gen = trans->transid;
 	} else {
@@ -172,7 +214,7 @@
 	} else {
 		first_key.objectid = 0;
 	}
-	cow = __btrfs_alloc_free_block(trans, root, buf->len,
+	cow = btrfs_alloc_free_block(trans, root, buf->len,
 				     root->root_key.objectid,
 				     root_gen, first_key.objectid, level,
 				     search_start, empty_size);
@@ -196,9 +238,14 @@
 	}
 
 	if (buf == root->node) {
+		WARN_ON(parent && parent != buf);
 		root_gen = btrfs_header_generation(buf);
+
+		spin_lock(&root->node_lock);
 		root->node = cow;
 		extent_buffer_get(cow);
+		spin_unlock(&root->node_lock);
+
 		if (buf != root->commit_root) {
 			btrfs_free_extent(trans, root, buf->start,
 					  buf->len, root->root_key.objectid,
@@ -219,6 +266,8 @@
 				  btrfs_header_owner(parent), root_gen,
 				  0, 0, 1);
 	}
+	if (unlock_orig)
+		btrfs_tree_unlock(buf);
 	free_extent_buffer(buf);
 	btrfs_mark_buffer_dirty(cow);
 	*cow_ret = cow;
@@ -316,6 +365,9 @@
 	int progress_passed = 0;
 	struct btrfs_disk_key disk_key;
 
+	/* FIXME this code needs locking */
+	return 0;
+
 	parent_level = btrfs_header_level(parent);
 	if (cache_only && parent_level != 1)
 		return 0;
@@ -729,6 +781,7 @@
 		return 0;
 
 	mid = path->nodes[level];
+	WARN_ON(!path->locks[level]);
 	WARN_ON(btrfs_header_generation(mid) != trans->transid);
 
 	orig_ptr = btrfs_node_blockptr(mid, orig_slot);
@@ -749,14 +802,21 @@
 
 		/* promote the child to a root */
 		child = read_node_slot(root, mid, 0);
+		btrfs_tree_lock(child);
 		BUG_ON(!child);
 		ret = btrfs_cow_block(trans, root, child, mid, 0, &child);
 		BUG_ON(ret);
 
+		spin_lock(&root->node_lock);
 		root->node = child;
+		spin_unlock(&root->node_lock);
+
 		add_root_to_dirty_list(root);
+		btrfs_tree_unlock(child);
+		path->locks[level] = 0;
 		path->nodes[level] = NULL;
 		clean_tree_block(trans, root, mid);
+		btrfs_tree_unlock(mid);
 		/* once for the path */
 		free_extent_buffer(mid);
 		ret = btrfs_free_extent(trans, root, mid->start, mid->len,
@@ -775,6 +835,7 @@
 
 	left = read_node_slot(root, parent, pslot - 1);
 	if (left) {
+		btrfs_tree_lock(left);
 		wret = btrfs_cow_block(trans, root, left,
 				       parent, pslot - 1, &left);
 		if (wret) {
@@ -784,6 +845,7 @@
 	}
 	right = read_node_slot(root, parent, pslot + 1);
 	if (right) {
+		btrfs_tree_lock(right);
 		wret = btrfs_cow_block(trans, root, right,
 				       parent, pslot + 1, &right);
 		if (wret) {
@@ -815,6 +877,7 @@
 			u32 blocksize = right->len;
 
 			clean_tree_block(trans, root, right);
+			btrfs_tree_unlock(right);
 			free_extent_buffer(right);
 			right = NULL;
 			wret = del_ptr(trans, root, path, level + 1, pslot +
@@ -862,7 +925,9 @@
 		u64 root_gen = btrfs_header_generation(parent);
 		u64 bytenr = mid->start;
 		u32 blocksize = mid->len;
+
 		clean_tree_block(trans, root, mid);
+		btrfs_tree_unlock(mid);
 		free_extent_buffer(mid);
 		mid = NULL;
 		wret = del_ptr(trans, root, path, level + 1, pslot);
@@ -885,11 +950,14 @@
 	if (left) {
 		if (btrfs_header_nritems(left) > orig_slot) {
 			extent_buffer_get(left);
+			/* left was locked after cow */
 			path->nodes[level] = left;
 			path->slots[level + 1] -= 1;
 			path->slots[level] = orig_slot;
-			if (mid)
+			if (mid) {
+				btrfs_tree_unlock(mid);
 				free_extent_buffer(mid);
+			}
 		} else {
 			orig_slot -= btrfs_header_nritems(left);
 			path->slots[level] = orig_slot;
@@ -901,10 +969,15 @@
 	    btrfs_node_blockptr(path->nodes[level], path->slots[level]))
 		BUG();
 enospc:
-	if (right)
+	if (right) {
+		btrfs_tree_unlock(right);
 		free_extent_buffer(right);
-	if (left)
+	}
+	if (left) {
+		if (path->nodes[level] != left)
+			btrfs_tree_unlock(left);
 		free_extent_buffer(left);
+	}
 	return ret;
 }
 
@@ -942,6 +1015,8 @@
 	/* first, try to make some room in the middle buffer */
 	if (left) {
 		u32 left_nr;
+
+		btrfs_tree_lock(left);
 		left_nr = btrfs_header_nritems(left);
 		if (left_nr >= BTRFS_NODEPTRS_PER_BLOCK(root) - 1) {
 			wret = 1;
@@ -967,24 +1042,28 @@
 				path->nodes[level] = left;
 				path->slots[level + 1] -= 1;
 				path->slots[level] = orig_slot;
+				btrfs_tree_unlock(mid);
 				free_extent_buffer(mid);
 			} else {
 				orig_slot -=
 					btrfs_header_nritems(left);
 				path->slots[level] = orig_slot;
+				btrfs_tree_unlock(left);
 				free_extent_buffer(left);
 			}
 			return 0;
 		}
+		btrfs_tree_unlock(left);
 		free_extent_buffer(left);
 	}
-	right= read_node_slot(root, parent, pslot + 1);
+	right = read_node_slot(root, parent, pslot + 1);
 
 	/*
 	 * then try to empty the right most buffer into the middle
 	 */
 	if (right) {
 		u32 right_nr;
+		btrfs_tree_lock(right);
 		right_nr = btrfs_header_nritems(right);
 		if (right_nr >= BTRFS_NODEPTRS_PER_BLOCK(root) - 1) {
 			wret = 1;
@@ -1013,12 +1092,15 @@
 				path->slots[level + 1] += 1;
 				path->slots[level] = orig_slot -
 					btrfs_header_nritems(mid);
+				btrfs_tree_unlock(mid);
 				free_extent_buffer(mid);
 			} else {
+				btrfs_tree_unlock(right);
 				free_extent_buffer(right);
 			}
 			return 0;
 		}
+		btrfs_tree_unlock(right);
 		free_extent_buffer(right);
 	}
 	return 1;
@@ -1050,6 +1132,8 @@
 		return;
 
 	node = path->nodes[level];
+	WARN_ON(!path->skip_locking && !btrfs_tree_locked(node));
+
 	search = btrfs_node_blockptr(node, slot);
 	blocksize = btrfs_level_size(root, level - 1);
 	eb = btrfs_find_tree_block(root, search, blocksize);
@@ -1098,6 +1182,39 @@
 			highest_read = search;
 	}
 }
+
+static void unlock_up(struct btrfs_path *path, int level, int lowest_unlock)
+{
+	int i;
+	int skip_level = level;
+	struct extent_buffer *t;
+
+	for (i = level; i < BTRFS_MAX_LEVEL; i++) {
+		if (!path->nodes[i])
+			break;
+		if (!path->locks[i])
+			break;
+		if (path->slots[i] == 0) {
+			skip_level = i + 1;
+			continue;
+		}
+		if (path->keep_locks) {
+			u32 nritems;
+			t = path->nodes[i];
+			nritems = btrfs_header_nritems(t);
+			if (path->slots[i] >= nritems - 1) {
+				skip_level = i + 1;
+				continue;
+			}
+		}
+		t = path->nodes[i];
+		if (i >= lowest_unlock && i > skip_level && path->locks[i]) {
+			btrfs_tree_unlock(t);
+			path->locks[i] = 0;
+		}
+	}
+}
+
 /*
  * look for key in the tree.  path is filled in with nodes along the way
  * if key is found, we return zero and you can find the item in the leaf
@@ -1120,15 +1237,27 @@
 	int ret;
 	int level;
 	int should_reada = p->reada;
+	int lowest_unlock = 1;
 	u8 lowest_level = 0;
 
 	lowest_level = p->lowest_level;
 	WARN_ON(lowest_level && ins_len);
 	WARN_ON(p->nodes[0] != NULL);
-	WARN_ON(!mutex_is_locked(&root->fs_info->fs_mutex));
+	// WARN_ON(!mutex_is_locked(&root->fs_info->fs_mutex));
+	WARN_ON(root == root->fs_info->extent_root &&
+		!mutex_is_locked(&root->fs_info->alloc_mutex));
+	WARN_ON(root == root->fs_info->chunk_root &&
+		!mutex_is_locked(&root->fs_info->chunk_mutex));
+	WARN_ON(root == root->fs_info->dev_root &&
+		!mutex_is_locked(&root->fs_info->chunk_mutex));
+	if (ins_len < 0)
+		lowest_unlock = 2;
 again:
-	b = root->node;
-	extent_buffer_get(b);
+	if (!p->skip_locking)
+		b = btrfs_lock_root_node(root);
+	else
+		b = btrfs_root_node(root);
+
 	while (b) {
 		level = btrfs_header_level(b);
 		if (cow) {
@@ -1147,9 +1276,12 @@
 			WARN_ON(1);
 		level = btrfs_header_level(b);
 		p->nodes[level] = b;
+		if (!p->skip_locking)
+			p->locks[level] = 1;
 		ret = check_block(root, p, level);
 		if (ret)
 			return -1;
+
 		ret = bin_search(b, key, level, &slot);
 		if (level != 0) {
 			if (ret && slot > 0)
@@ -1177,14 +1309,19 @@
 				BUG_ON(btrfs_header_nritems(b) == 1);
 			}
 			/* this is only true while dropping a snapshot */
-			if (level == lowest_level)
+			if (level == lowest_level) {
+				unlock_up(p, level, lowest_unlock);
 				break;
+			}
 
 			if (should_reada)
 				reada_for_search(root, p, level, slot,
 						 key->objectid);
 
 			b = read_node_slot(root, b, slot);
+			if (!p->skip_locking)
+				btrfs_tree_lock(b);
+			unlock_up(p, level, lowest_unlock);
 		} else {
 			p->slots[level] = slot;
 			if (ins_len > 0 && btrfs_leaf_free_space(root, b) <
@@ -1195,6 +1332,7 @@
 				if (sret)
 					return sret;
 			}
+			unlock_up(p, level, lowest_unlock);
 			return ret;
 		}
 	}
@@ -1225,6 +1363,13 @@
 			break;
 		t = path->nodes[i];
 		btrfs_set_node_key(t, key, tslot);
+		if (!btrfs_tree_locked(path->nodes[i])) {
+			int ii;
+printk("fixup without lock on level %d\n", btrfs_header_level(path->nodes[i]));
+			for (ii = 0; ii < BTRFS_MAX_LEVEL; ii++) {
+printk("level %d slot %d\n", ii, path->slots[ii]);
+			}
+		}
 		btrfs_mark_buffer_dirty(path->nodes[i]);
 		if (tslot != 0)
 			break;
@@ -1370,6 +1515,7 @@
 	u64 lower_gen;
 	struct extent_buffer *lower;
 	struct extent_buffer *c;
+	struct extent_buffer *old;
 	struct btrfs_disk_key lower_key;
 
 	BUG_ON(path->nodes[level]);
@@ -1386,12 +1532,13 @@
 	else
 		btrfs_node_key(lower, &lower_key, 0);
 
-	c = __btrfs_alloc_free_block(trans, root, root->nodesize,
+	c = btrfs_alloc_free_block(trans, root, root->nodesize,
 				   root->root_key.objectid,
 				   root_gen, lower_key.objectid, level,
 				   root->node->start, 0);
 	if (IS_ERR(c))
 		return PTR_ERR(c);
+
 	memset_extent_buffer(c, 0, 0, root->nodesize);
 	btrfs_set_header_nritems(c, 1);
 	btrfs_set_header_level(c, level);
@@ -1416,23 +1563,31 @@
 
 	btrfs_mark_buffer_dirty(c);
 
-	/* the super has an extra ref to root->node */
-	free_extent_buffer(root->node);
+	spin_lock(&root->node_lock);
+	old = root->node;
 	root->node = c;
+	spin_unlock(&root->node_lock);
+
+	/* the super has an extra ref to root->node */
+	free_extent_buffer(old);
+
 	add_root_to_dirty_list(root);
 	extent_buffer_get(c);
 	path->nodes[level] = c;
+	path->locks[level] = 1;
 	path->slots[level] = 0;
 
 	if (root->ref_cows && lower_gen != trans->transid) {
 		struct btrfs_path *back_path = btrfs_alloc_path();
 		int ret;
+		mutex_lock(&root->fs_info->alloc_mutex);
 		ret = btrfs_insert_extent_backref(trans,
 						  root->fs_info->extent_root,
 						  path, lower->start,
 						  root->root_key.objectid,
 						  trans->transid, 0, 0);
 		BUG_ON(ret);
+		mutex_unlock(&root->fs_info->alloc_mutex);
 		btrfs_free_path(back_path);
 	}
 	return 0;
@@ -1521,7 +1676,7 @@
 		root_gen = 0;
 
 	btrfs_node_key(c, &disk_key, 0);
-	split = __btrfs_alloc_free_block(trans, root, root->nodesize,
+	split = btrfs_alloc_free_block(trans, root, root->nodesize,
 					 root->root_key.objectid,
 					 root_gen,
 					 btrfs_disk_key_objectid(&disk_key),
@@ -1564,10 +1719,12 @@
 
 	if (path->slots[level] >= mid) {
 		path->slots[level] -= mid;
+		btrfs_tree_unlock(c);
 		free_extent_buffer(c);
 		path->nodes[level] = split;
 		path->slots[level + 1] += 1;
 	} else {
+		btrfs_tree_unlock(split);
 		free_extent_buffer(split);
 	}
 	return ret;
@@ -1648,30 +1805,24 @@
 		return 1;
 
 	right = read_node_slot(root, upper, slot + 1);
+	btrfs_tree_lock(right);
 	free_space = btrfs_leaf_free_space(root, right);
-	if (free_space < data_size + sizeof(struct btrfs_item)) {
-		free_extent_buffer(right);
-		return 1;
-	}
+	if (free_space < data_size + sizeof(struct btrfs_item))
+		goto out_unlock;
 
 	/* cow and double check */
 	ret = btrfs_cow_block(trans, root, right, upper,
 			      slot + 1, &right);
-	if (ret) {
-		free_extent_buffer(right);
-		return 1;
-	}
+	if (ret)
+		goto out_unlock;
+
 	free_space = btrfs_leaf_free_space(root, right);
-	if (free_space < data_size + sizeof(struct btrfs_item)) {
-		free_extent_buffer(right);
-		return 1;
-	}
+	if (free_space < data_size + sizeof(struct btrfs_item))
+		goto out_unlock;
 
 	left_nritems = btrfs_header_nritems(left);
-	if (left_nritems == 0) {
-		free_extent_buffer(right);
-		return 1;
-	}
+	if (left_nritems == 0)
+		goto out_unlock;
 
 	if (empty)
 		nr = 0;
@@ -1707,10 +1858,8 @@
 		left->map_token = NULL;
 	}
 
-	if (push_items == 0) {
-		free_extent_buffer(right);
-		return 1;
-	}
+	if (push_items == 0)
+		goto out_unlock;
 
 	if (!empty && push_items == left_nritems)
 		WARN_ON(1);
@@ -1778,14 +1927,24 @@
 	/* then fixup the leaf pointer in the path */
 	if (path->slots[0] >= left_nritems) {
 		path->slots[0] -= left_nritems;
+		if (btrfs_header_nritems(path->nodes[0]) == 0)
+			clean_tree_block(trans, root, path->nodes[0]);
+		btrfs_tree_unlock(path->nodes[0]);
 		free_extent_buffer(path->nodes[0]);
 		path->nodes[0] = right;
 		path->slots[1] += 1;
 	} else {
+		btrfs_tree_unlock(right);
 		free_extent_buffer(right);
 	}
 	return 0;
+
+out_unlock:
+	btrfs_tree_unlock(right);
+	free_extent_buffer(right);
+	return 1;
 }
+
 /*
  * push some data in the path leaf to the left, trying to free up at
  * least data_size bytes.  returns zero if the push worked, nonzero otherwise
@@ -1823,10 +1982,11 @@
 	}
 
 	left = read_node_slot(root, path->nodes[1], slot - 1);
+	btrfs_tree_lock(left);
 	free_space = btrfs_leaf_free_space(root, left);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
-		free_extent_buffer(left);
-		return 1;
+		ret = 1;
+		goto out;
 	}
 
 	/* cow and double check */
@@ -1834,14 +1994,14 @@
 			      path->nodes[1], slot - 1, &left);
 	if (ret) {
 		/* we hit -ENOSPC, but it isn't fatal here */
-		free_extent_buffer(left);
-		return 1;
+		ret = 1;
+		goto out;
 	}
 
 	free_space = btrfs_leaf_free_space(root, left);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
-		free_extent_buffer(left);
-		return 1;
+		ret = 1;
+		goto out;
 	}
 
 	if (empty)
@@ -1876,8 +2036,8 @@
 	}
 
 	if (push_items == 0) {
-		free_extent_buffer(left);
-		return 1;
+		ret = 1;
+		goto out;
 	}
 	if (!empty && push_items == btrfs_header_nritems(right))
 		WARN_ON(1);
@@ -1975,15 +2135,23 @@
 	/* then fixup the leaf pointer in the path */
 	if (path->slots[0] < push_items) {
 		path->slots[0] += old_left_nritems;
+		if (btrfs_header_nritems(path->nodes[0]) == 0)
+			clean_tree_block(trans, root, path->nodes[0]);
+		btrfs_tree_unlock(path->nodes[0]);
 		free_extent_buffer(path->nodes[0]);
 		path->nodes[0] = left;
 		path->slots[1] -= 1;
 	} else {
+		btrfs_tree_unlock(left);
 		free_extent_buffer(left);
 		path->slots[0] -= push_items;
 	}
 	BUG_ON(path->slots[0] < 0);
 	return ret;
+out:
+	btrfs_tree_unlock(left);
+	free_extent_buffer(left);
+	return ret;
 }
 
 /*
@@ -2052,7 +2220,7 @@
 
 	btrfs_item_key(l, &disk_key, 0);
 
-	right = __btrfs_alloc_free_block(trans, root, root->leafsize,
+	right = btrfs_alloc_free_block(trans, root, root->leafsize,
 					 root->root_key.objectid,
 					 root_gen, disk_key.objectid, 0,
 					 l->start, 0);
@@ -2085,6 +2253,8 @@
 						  path->slots[1] + 1, 1);
 				if (wret)
 					ret = wret;
+
+				btrfs_tree_unlock(path->nodes[0]);
 				free_extent_buffer(path->nodes[0]);
 				path->nodes[0] = right;
 				path->slots[0] = 0;
@@ -2111,6 +2281,7 @@
 						  path->slots[1], 1);
 				if (wret)
 					ret = wret;
+				btrfs_tree_unlock(path->nodes[0]);
 				free_extent_buffer(path->nodes[0]);
 				path->nodes[0] = right;
 				path->slots[0] = 0;
@@ -2184,12 +2355,15 @@
 	BUG_ON(path->slots[0] != slot);
 
 	if (mid <= slot) {
+		btrfs_tree_unlock(path->nodes[0]);
 		free_extent_buffer(path->nodes[0]);
 		path->nodes[0] = right;
 		path->slots[0] -= mid;
 		path->slots[1] += 1;
-	} else
+	} else {
+		btrfs_tree_unlock(right);
 		free_extent_buffer(right);
+	}
 
 	BUG_ON(path->slots[0] < 0);
 
@@ -2418,10 +2592,6 @@
 		total_data += data_size[i];
 	}
 
-	/* create a root if there isn't one */
-	if (!root->node)
-		BUG();
-
 	total_size = total_data + (nr - 1) * sizeof(struct btrfs_item);
 	ret = btrfs_search_slot(trans, root, cpu_key, path, total_size, 1);
 	if (ret == 0) {
@@ -2516,7 +2686,6 @@
 		btrfs_print_leaf(root, leaf);
 		BUG();
 	}
-
 out:
 	return ret;
 }
@@ -2655,7 +2824,6 @@
 			btrfs_set_header_level(leaf, 0);
 		} else {
 			u64 root_gen = btrfs_header_generation(path->nodes[1]);
-			clean_tree_block(trans, root, leaf);
 			wret = del_ptr(trans, root, path, 1, path->slots[1]);
 			if (wret)
 				ret = wret;
@@ -2706,8 +2874,6 @@
 				root_gen = btrfs_header_generation(
 							   path->nodes[1]);
 
-				clean_tree_block(trans, root, leaf);
-
 				wret = del_ptr(trans, root, path, 1, slot);
 				if (wret)
 					ret = wret;
@@ -2720,7 +2886,13 @@
 				if (wret)
 					ret = wret;
 			} else {
-				btrfs_mark_buffer_dirty(leaf);
+				/* if we're still in the path, make sure
+				 * we're dirty.  Otherwise, one of the
+				 * push_leaf functions must have already
+				 * dirtied this buffer
+				 */
+				if (path->nodes[0] == leaf)
+					btrfs_mark_buffer_dirty(leaf);
 				free_extent_buffer(leaf);
 			}
 		} else {
@@ -2731,56 +2903,40 @@
 }
 
 /*
- * walk up the tree as far as required to find the previous leaf.
+ * search the tree again to find a leaf with lesser keys
  * returns 0 if it found something or 1 if there are no lesser leaves.
  * returns < 0 on io errors.
  */
 int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path)
 {
-	int slot;
-	int level = 1;
-	struct extent_buffer *c;
-	struct extent_buffer *next = NULL;
+	struct btrfs_key key;
+	struct btrfs_disk_key found_key;
+	int ret;
 
-	while(level < BTRFS_MAX_LEVEL) {
-		if (!path->nodes[level])
-			return 1;
+	btrfs_item_key_to_cpu(path->nodes[0], &key, 0);
 
-		slot = path->slots[level];
-		c = path->nodes[level];
-		if (slot == 0) {
-			level++;
-			if (level == BTRFS_MAX_LEVEL)
-				return 1;
-			continue;
-		}
-		slot--;
+	if (key.offset > 0)
+		key.offset--;
+	else if (key.type > 0)
+		key.type--;
+	else if (key.objectid > 0)
+		key.objectid--;
+	else
+		return 1;
 
-		if (next)
-			free_extent_buffer(next);
-
-		next = read_node_slot(root, c, slot);
-		break;
-	}
-	path->slots[level] = slot;
-	while(1) {
-		level--;
-		c = path->nodes[level];
-		free_extent_buffer(c);
-		slot = btrfs_header_nritems(next);
-		if (slot != 0)
-			slot--;
-		path->nodes[level] = next;
-		path->slots[level] = slot;
-		if (!level)
-			break;
-		next = read_node_slot(root, next, slot);
-	}
-	return 0;
+	btrfs_release_path(root, path);
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		return ret;
+	btrfs_item_key(path->nodes[0], &found_key, 0);
+	ret = comp_keys(&found_key, &key);
+	if (ret < 0)
+		return 0;
+	return 1;
 }
 
 /*
- * walk up the tree as far as required to find the next leaf.
+ * search the tree again to find a leaf with greater keys
  * returns 0 if it found something or 1 if there are no greater leaves.
  * returns < 0 on io errors.
  */
@@ -2790,6 +2946,28 @@
 	int level = 1;
 	struct extent_buffer *c;
 	struct extent_buffer *next = NULL;
+	struct btrfs_key key;
+	u32 nritems;
+	int ret;
+
+	nritems = btrfs_header_nritems(path->nodes[0]);
+	if (nritems == 0) {
+		return 1;
+	}
+
+	btrfs_item_key_to_cpu(path->nodes[0], &key, nritems - 1);
+
+	path->keep_locks = 1;
+	btrfs_release_path(root, path);
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	path->keep_locks = 0;
+
+	if (ret < 0)
+		return ret;
+
+	if (path->slots[0] < nritems - 1) {
+		goto done;
+	}
 
 	while(level < BTRFS_MAX_LEVEL) {
 		if (!path->nodes[level])
@@ -2799,33 +2977,45 @@
 		c = path->nodes[level];
 		if (slot >= btrfs_header_nritems(c)) {
 			level++;
-			if (level == BTRFS_MAX_LEVEL)
+			if (level == BTRFS_MAX_LEVEL) {
 				return 1;
+			}
 			continue;
 		}
 
-		if (next)
+		if (next) {
+			btrfs_tree_unlock(next);
 			free_extent_buffer(next);
+		}
 
-		if (path->reada)
+		if (level == 1 && path->locks[1] && path->reada)
 			reada_for_search(root, path, level, slot, 0);
 
 		next = read_node_slot(root, c, slot);
+		if (!path->skip_locking)
+			btrfs_tree_lock(next);
 		break;
 	}
 	path->slots[level] = slot;
 	while(1) {
 		level--;
 		c = path->nodes[level];
+		if (path->locks[level])
+			btrfs_tree_unlock(c);
 		free_extent_buffer(c);
 		path->nodes[level] = next;
 		path->slots[level] = 0;
+		path->locks[level] = 1;
 		if (!level)
 			break;
-		if (path->reada)
-			reada_for_search(root, path, level, 0, 0);
+		if (level == 1 && path->locks[1] && path->reada)
+			reada_for_search(root, path, level, slot, 0);
 		next = read_node_slot(root, next, 0);
+		if (!path->skip_locking)
+			btrfs_tree_lock(next);
 	}
+done:
+	unlock_up(path, 0, 1);
 	return 0;
 }
 
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index dcea9d7..50891b3 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -330,8 +330,13 @@
 struct btrfs_path {
 	struct extent_buffer *nodes[BTRFS_MAX_LEVEL];
 	int slots[BTRFS_MAX_LEVEL];
+	/* if there is real range locking, this locks field will change */
+	int locks[BTRFS_MAX_LEVEL];
 	int reada;
+	/* keep some upper locks as we walk down */
+	int keep_locks;
 	int lowest_level;
+	int skip_locking;
 };
 
 /*
@@ -515,6 +520,8 @@
 	spinlock_t hash_lock;
 	struct mutex trans_mutex;
 	struct mutex fs_mutex;
+	struct mutex alloc_mutex;
+	struct mutex chunk_mutex;
 	struct list_head trans_list;
 	struct list_head hashers;
 	struct list_head dead_roots;
@@ -576,6 +583,10 @@
  */
 struct btrfs_root {
 	struct extent_buffer *node;
+
+	/* the node lock is held while changing the node pointer */
+	spinlock_t node_lock;
+
 	struct extent_buffer *commit_root;
 	struct btrfs_root_item root_item;
 	struct btrfs_key root_key;
@@ -1353,13 +1364,7 @@
 						 struct btrfs_block_group_cache
 						 *hint, u64 search_start,
 						 int data, int owner);
-int btrfs_inc_root_ref(struct btrfs_trans_handle *trans,
-		       struct btrfs_root *root, u64 owner_objectid);
 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
-					    struct btrfs_root *root, u32 size,
-					    u64 root_objectid,
-					    u64 hint, u64 empty_size);
-struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 					     struct btrfs_root *root,
 					     u32 blocksize,
 					     u64 root_objectid,
@@ -1368,8 +1373,6 @@
 					     int level,
 					     u64 hint,
 					     u64 empty_size);
-int btrfs_grow_extent_tree(struct btrfs_trans_handle *trans,
-			   struct btrfs_root *root, u64 new_size);
 int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size);
 int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans,
 				 struct btrfs_root *root,
@@ -1409,6 +1412,10 @@
 int btrfs_previous_item(struct btrfs_root *root,
 			struct btrfs_path *path, u64 min_objectid,
 			int type);
+
+struct extent_buffer *btrfs_root_node(struct btrfs_root *root);
+struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root);
+
 int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		    struct btrfs_root *root, struct extent_buffer *buf,
 		    struct extent_buffer *parent, int parent_slot,
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index e5c758e..fe40bdd 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -32,6 +32,7 @@
 #include "volumes.h"
 #include "print-tree.h"
 #include "async-thread.h"
+#include "locking.h"
 
 #if 0
 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
@@ -681,9 +682,11 @@
 {
 	struct inode *btree_inode = root->fs_info->btree_inode;
 	if (btrfs_header_generation(buf) ==
-	    root->fs_info->running_transaction->transid)
+	    root->fs_info->running_transaction->transid) {
+		WARN_ON(!btrfs_tree_locked(buf));
 		clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
 					  buf);
+	}
 	return 0;
 }
 
@@ -720,6 +723,7 @@
 	root->in_sysfs = 0;
 
 	INIT_LIST_HEAD(&root->dirty_list);
+	spin_lock_init(&root->node_lock);
 	memset(&root->root_key, 0, sizeof(root->root_key));
 	memset(&root->root_item, 0, sizeof(root->root_item));
 	memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
@@ -1196,6 +1200,8 @@
 
 	mutex_init(&fs_info->trans_mutex);
 	mutex_init(&fs_info->fs_mutex);
+	mutex_init(&fs_info->alloc_mutex);
+	mutex_init(&fs_info->chunk_mutex);
 
 #if 0
 	ret = add_hasher(fs_info, "crc32c");
@@ -1274,7 +1280,9 @@
 
 	mutex_lock(&fs_info->fs_mutex);
 
+	mutex_lock(&fs_info->chunk_mutex);
 	ret = btrfs_read_sys_array(tree_root);
+	mutex_unlock(&fs_info->chunk_mutex);
 	if (ret) {
 		printk("btrfs: failed to read the system array on %s\n",
 		       sb->s_id);
@@ -1296,7 +1304,9 @@
 	         (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
 		 BTRFS_UUID_SIZE);
 
+	mutex_lock(&fs_info->chunk_mutex);
 	ret = btrfs_read_chunk_tree(chunk_root);
+	mutex_unlock(&fs_info->chunk_mutex);
 	BUG_ON(ret);
 
 	btrfs_close_extra_devices(fs_devices);
@@ -1654,6 +1664,7 @@
 	u64 transid = btrfs_header_generation(buf);
 	struct inode *btree_inode = root->fs_info->btree_inode;
 
+	WARN_ON(!btrfs_tree_locked(buf));
 	if (transid != root->fs_info->generation) {
 		printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
 			(unsigned long long)buf->start,
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 41a6346..7e40c51 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -25,6 +25,7 @@
 #include "print-tree.h"
 #include "transaction.h"
 #include "volumes.h"
+#include "locking.h"
 
 #define BLOCK_GROUP_DATA     EXTENT_WRITEBACK
 #define BLOCK_GROUP_METADATA EXTENT_UPTODATE
@@ -36,7 +37,28 @@
 				 btrfs_root *extent_root);
 static int del_pending_extents(struct btrfs_trans_handle *trans, struct
 			       btrfs_root *extent_root);
+static struct btrfs_block_group_cache *
+__btrfs_find_block_group(struct btrfs_root *root,
+			 struct btrfs_block_group_cache *hint,
+			 u64 search_start, int data, int owner);
 
+void maybe_lock_mutex(struct btrfs_root *root)
+{
+	if (root != root->fs_info->extent_root &&
+	    root != root->fs_info->chunk_root &&
+	    root != root->fs_info->dev_root) {
+		mutex_lock(&root->fs_info->alloc_mutex);
+	}
+}
+
+void maybe_unlock_mutex(struct btrfs_root *root)
+{
+	if (root != root->fs_info->extent_root &&
+	    root != root->fs_info->chunk_root &&
+	    root != root->fs_info->dev_root) {
+		mutex_unlock(&root->fs_info->alloc_mutex);
+	}
+}
 
 static int cache_block_group(struct btrfs_root *root,
 			     struct btrfs_block_group_cache *block_group)
@@ -66,6 +88,7 @@
 		return -ENOMEM;
 
 	path->reada = 2;
+	path->skip_locking = 1;
 	first_free = block_group->key.objectid;
 	key.objectid = block_group->key.objectid;
 	key.offset = 0;
@@ -290,7 +313,7 @@
 		cache = btrfs_lookup_first_block_group(root->fs_info, last);
 	}
 	cache_miss = 0;
-	cache = btrfs_find_block_group(root, cache, last, data, 0);
+	cache = __btrfs_find_block_group(root, cache, last, data, 0);
 	if (!cache)
 		goto no_cache;
 	*cache_ret = cache;
@@ -318,10 +341,10 @@
 	return bits;
 }
 
-struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
-						 struct btrfs_block_group_cache
-						 *hint, u64 search_start,
-						 int data, int owner)
+static struct btrfs_block_group_cache *
+__btrfs_find_block_group(struct btrfs_root *root,
+			 struct btrfs_block_group_cache *hint,
+			 u64 search_start, int data, int owner)
 {
 	struct btrfs_block_group_cache *cache;
 	struct extent_io_tree *block_group_cache;
@@ -411,6 +434,18 @@
 	return found_group;
 }
 
+struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
+						 struct btrfs_block_group_cache
+						 *hint, u64 search_start,
+						 int data, int owner)
+{
+
+	struct btrfs_block_group_cache *ret;
+	mutex_lock(&root->fs_info->alloc_mutex);
+	ret = __btrfs_find_block_group(root, hint, search_start, data, owner);
+	mutex_unlock(&root->fs_info->alloc_mutex);
+	return ret;
+}
 static u64 hash_extent_ref(u64 root_objectid, u64 ref_generation,
 			   u64 owner, u64 owner_offset)
 {
@@ -646,7 +681,7 @@
 	return ret;
 }
 
-int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
+static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root,
 				u64 bytenr, u64 num_bytes,
 				u64 root_objectid, u64 ref_generation,
@@ -696,6 +731,22 @@
 	return 0;
 }
 
+int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
+				struct btrfs_root *root,
+				u64 bytenr, u64 num_bytes,
+				u64 root_objectid, u64 ref_generation,
+				u64 owner, u64 owner_offset)
+{
+	int ret;
+
+	mutex_lock(&root->fs_info->alloc_mutex);
+	ret = __btrfs_inc_extent_ref(trans, root, bytenr, num_bytes,
+				     root_objectid, ref_generation,
+				     owner, owner_offset);
+	mutex_unlock(&root->fs_info->alloc_mutex);
+	return ret;
+}
+
 int btrfs_extent_post_op(struct btrfs_trans_handle *trans,
 			 struct btrfs_root *root)
 {
@@ -760,6 +811,10 @@
 	struct btrfs_extent_ref *ref_item;
 	int level = -1;
 
+	/* FIXME, needs locking */
+	BUG();
+
+	mutex_lock(&root->fs_info->alloc_mutex);
 	path = btrfs_alloc_path();
 again:
 	if (level == -1)
@@ -854,33 +909,9 @@
 
 out:
 	btrfs_free_path(path);
+	mutex_unlock(&root->fs_info->alloc_mutex);
 	return total_count;
 }
-int btrfs_inc_root_ref(struct btrfs_trans_handle *trans,
-		       struct btrfs_root *root, u64 owner_objectid)
-{
-	u64 generation;
-	u64 key_objectid;
-	u64 level;
-	u32 nritems;
-	struct btrfs_disk_key disk_key;
-
-	level = btrfs_header_level(root->node);
-	generation = trans->transid;
-	nritems = btrfs_header_nritems(root->node);
-	if (nritems > 0) {
-		if (level == 0)
-			btrfs_item_key(root->node, &disk_key, 0);
-		else
-			btrfs_node_key(root->node, &disk_key, 0);
-		key_objectid = btrfs_disk_key_objectid(&disk_key);
-	} else {
-		key_objectid = 0;
-	}
-	return btrfs_inc_extent_ref(trans, root, root->node->start,
-				    root->node->len, owner_objectid,
-				    generation, level, key_objectid);
-}
 
 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		  struct extent_buffer *buf)
@@ -897,6 +928,7 @@
 	if (!root->ref_cows)
 		return 0;
 
+	mutex_lock(&root->fs_info->alloc_mutex);
 	level = btrfs_header_level(buf);
 	nritems = btrfs_header_nritems(buf);
 	for (i = 0; i < nritems; i++) {
@@ -913,7 +945,7 @@
 			disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
 			if (disk_bytenr == 0)
 				continue;
-			ret = btrfs_inc_extent_ref(trans, root, disk_bytenr,
+			ret = __btrfs_inc_extent_ref(trans, root, disk_bytenr,
 				    btrfs_file_extent_disk_num_bytes(buf, fi),
 				    root->root_key.objectid, trans->transid,
 				    key.objectid, key.offset);
@@ -924,7 +956,7 @@
 		} else {
 			bytenr = btrfs_node_blockptr(buf, i);
 			btrfs_node_key_to_cpu(buf, &key, i);
-			ret = btrfs_inc_extent_ref(trans, root, bytenr,
+			ret = __btrfs_inc_extent_ref(trans, root, bytenr,
 					   btrfs_level_size(root, level - 1),
 					   root->root_key.objectid,
 					   trans->transid,
@@ -935,6 +967,7 @@
 			}
 		}
 	}
+	mutex_unlock(&root->fs_info->alloc_mutex);
 	return 0;
 fail:
 	WARN_ON(1);
@@ -965,6 +998,7 @@
 		}
 	}
 #endif
+	mutex_unlock(&root->fs_info->alloc_mutex);
 	return ret;
 }
 
@@ -1019,6 +1053,7 @@
 	if (!path)
 		return -ENOMEM;
 
+	mutex_lock(&root->fs_info->alloc_mutex);
 	while(1) {
 		ret = find_first_extent_bit(block_group_cache, last,
 					    &start, &end, BLOCK_GROUP_DIRTY);
@@ -1045,6 +1080,7 @@
 				  BLOCK_GROUP_DIRTY, GFP_NOFS);
 	}
 	btrfs_free_path(path);
+	mutex_unlock(&root->fs_info->alloc_mutex);
 	return werr;
 }
 
@@ -1162,26 +1198,28 @@
 		space_info->force_alloc = 0;
 	}
 	if (space_info->full)
-		return 0;
+		goto out;
 
 	thresh = div_factor(space_info->total_bytes, 6);
 	if (!force &&
 	   (space_info->bytes_used + space_info->bytes_pinned + alloc_bytes) <
 	    thresh)
-		return 0;
+		goto out;
 
+	mutex_lock(&extent_root->fs_info->chunk_mutex);
 	ret = btrfs_alloc_chunk(trans, extent_root, &start, &num_bytes, flags);
 	if (ret == -ENOSPC) {
 printk("space info full %Lu\n", flags);
 		space_info->full = 1;
-		return 0;
+		goto out;
 	}
 	BUG_ON(ret);
 
 	ret = btrfs_make_block_group(trans, extent_root, 0, flags,
 		     BTRFS_FIRST_CHUNK_TREE_OBJECTID, start, num_bytes);
 	BUG_ON(ret);
-
+	mutex_unlock(&extent_root->fs_info->chunk_mutex);
+out:
 	return 0;
 }
 
@@ -1318,6 +1356,7 @@
 	struct extent_io_tree *free_space_cache;
 	free_space_cache = &root->fs_info->free_space_cache;
 
+	mutex_lock(&root->fs_info->alloc_mutex);
 	while(1) {
 		ret = find_first_extent_bit(unpin, 0, &start, &end,
 					    EXTENT_DIRTY);
@@ -1327,6 +1366,7 @@
 		clear_extent_dirty(unpin, start, end, GFP_NOFS);
 		set_extent_dirty(free_space_cache, start, end, GFP_NOFS);
 	}
+	mutex_unlock(&root->fs_info->alloc_mutex);
 	return 0;
 }
 
@@ -1363,18 +1403,24 @@
 				  GFP_NOFS);
 		eb = read_tree_block(extent_root, ins.objectid, ins.offset,
 				     trans->transid);
+		btrfs_tree_lock(eb);
 		level = btrfs_header_level(eb);
 		if (level == 0) {
 			btrfs_item_key(eb, &first, 0);
 		} else {
 			btrfs_node_key(eb, &first, 0);
 		}
+		btrfs_tree_unlock(eb);
+		free_extent_buffer(eb);
+		/*
+		 * the first key is just a hint, so the race we've created
+		 * against reading it is fine
+		 */
 		err = btrfs_insert_extent_backref(trans, extent_root, path,
 					  start, extent_root->root_key.objectid,
 					  0, level,
 					  btrfs_disk_key_objectid(&first));
 		BUG_ON(err);
-		free_extent_buffer(eb);
 	}
 	btrfs_free_path(path);
 	return 0;
@@ -1384,12 +1430,14 @@
 			  int pending)
 {
 	int err = 0;
-	struct extent_buffer *buf;
 
 	if (!pending) {
+#if 0
+		struct extent_buffer *buf;
 		buf = btrfs_find_tree_block(root, bytenr, num_bytes);
 		if (buf) {
-			if (btrfs_buffer_uptodate(buf, 0)) {
+			if (!btrfs_try_tree_lock(buf) &&
+			    btrfs_buffer_uptodate(buf, 0)) {
 				u64 transid =
 				    root->fs_info->running_transaction->transid;
 				u64 header_transid =
@@ -1398,12 +1446,15 @@
 				    !btrfs_header_flag(buf,
 					       BTRFS_HEADER_FLAG_WRITTEN)) {
 					clean_tree_block(NULL, root, buf);
+					btrfs_tree_unlock(buf);
 					free_extent_buffer(buf);
 					return 1;
 				}
+				btrfs_tree_unlock(buf);
 			}
 			free_extent_buffer(buf);
 		}
+#endif
 		update_pinned_extents(root, bytenr, num_bytes, 1);
 	} else {
 		set_extent_bits(&root->fs_info->pending_del,
@@ -1586,10 +1637,11 @@
 /*
  * remove an extent from the root, returns 0 on success
  */
-int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, u64 bytenr, u64 num_bytes,
-		      u64 root_objectid, u64 ref_generation,
-		      u64 owner_objectid, u64 owner_offset, int pin)
+static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root, u64 bytenr,
+			       u64 num_bytes, u64 root_objectid,
+			       u64 ref_generation, u64 owner_objectid,
+			       u64 owner_offset, int pin)
 {
 	struct btrfs_root *extent_root = root->fs_info->extent_root;
 	int pending_ret;
@@ -1610,6 +1662,22 @@
 	return ret ? ret : pending_ret;
 }
 
+int btrfs_free_extent(struct btrfs_trans_handle *trans,
+		      struct btrfs_root *root, u64 bytenr,
+		      u64 num_bytes, u64 root_objectid,
+		      u64 ref_generation, u64 owner_objectid,
+		      u64 owner_offset, int pin)
+{
+	int ret;
+
+	maybe_lock_mutex(root);
+	ret = __btrfs_free_extent(trans, root, bytenr, num_bytes,
+				  root_objectid, ref_generation,
+				  owner_objectid, owner_offset, pin);
+	maybe_unlock_mutex(root);
+	return ret;
+}
+
 static u64 stripe_align(struct btrfs_root *root, u64 val)
 {
 	u64 mask = ((u64)root->stripesize - 1);
@@ -1679,12 +1747,12 @@
 		block_group = btrfs_lookup_first_block_group(info, hint_byte);
 		if (!block_group)
 			hint_byte = search_start;
-		block_group = btrfs_find_block_group(root, block_group,
+		block_group = __btrfs_find_block_group(root, block_group,
 						     hint_byte, data, 1);
 		if (last_ptr && *last_ptr == 0 && block_group)
 			hint_byte = block_group->key.objectid;
 	} else {
-		block_group = btrfs_find_block_group(root,
+		block_group = __btrfs_find_block_group(root,
 						     trans->block_group,
 						     search_start, data, 1);
 	}
@@ -1806,7 +1874,7 @@
 	}
 	block_group = btrfs_lookup_first_block_group(info, search_start);
 	cond_resched();
-	block_group = btrfs_find_block_group(root, block_group,
+	block_group = __btrfs_find_block_group(root, block_group,
 					     search_start, data, 0);
 	goto check_failed;
 
@@ -1843,6 +1911,8 @@
 	struct btrfs_path *path;
 	struct btrfs_key keys[2];
 
+	maybe_lock_mutex(root);
+
 	if (data) {
 		alloc_profile = info->avail_data_alloc_bits &
 			        info->data_alloc_profile;
@@ -1892,9 +1962,10 @@
 	if (ret) {
 		printk("allocation failed flags %Lu\n", data);
 	}
-	BUG_ON(ret);
-	if (ret)
-		return ret;
+	if (ret) {
+		BUG();
+		goto out;
+	}
 
 	/* block accounting for super block */
 	super_used = btrfs_super_bytes_used(&info->super_copy);
@@ -1953,11 +2024,11 @@
 	finish_current_insert(trans, extent_root);
 	pending_ret = del_pending_extents(trans, extent_root);
 
-	if (ret) {
-		return ret;
-	}
+	if (ret)
+		goto out;
 	if (pending_ret) {
-		return pending_ret;
+		ret = pending_ret;
+		goto out;
 	}
 
 update_block:
@@ -1967,9 +2038,10 @@
 		       ins->objectid, ins->offset);
 		BUG();
 	}
-	return 0;
+out:
+	maybe_unlock_mutex(root);
+	return ret;
 }
-
 /*
  * helper function to allocate a block for a given tree
  * returns the tree buffer or NULL.
@@ -1977,28 +2049,6 @@
 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 					     struct btrfs_root *root,
 					     u32 blocksize,
-					     u64 root_objectid, u64 hint,
-					     u64 empty_size)
-{
-	u64 ref_generation;
-
-	if (root->ref_cows)
-		ref_generation = trans->transid;
-	else
-		ref_generation = 0;
-
-
-	return __btrfs_alloc_free_block(trans, root, blocksize, root_objectid,
-					ref_generation, 0, 0, hint, empty_size);
-}
-
-/*
- * helper function to allocate a block for a given tree
- * returns the tree buffer or NULL.
- */
-struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
-					     struct btrfs_root *root,
-					     u32 blocksize,
 					     u64 root_objectid,
 					     u64 ref_generation,
 					     u64 first_objectid,
@@ -2026,6 +2076,7 @@
 		return ERR_PTR(-ENOMEM);
 	}
 	btrfs_set_header_generation(buf, trans->transid);
+	btrfs_tree_lock(buf);
 	clean_tree_block(trans, root, buf);
 	btrfs_set_buffer_uptodate(buf);
 
@@ -2076,7 +2127,7 @@
 		disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
 		if (disk_bytenr == 0)
 			continue;
-		ret = btrfs_free_extent(trans, root, disk_bytenr,
+		ret = __btrfs_free_extent(trans, root, disk_bytenr,
 				btrfs_file_extent_disk_num_bytes(leaf, fi),
 				leaf_owner, leaf_generation,
 				key.objectid, key.offset, 0);
@@ -2151,6 +2202,8 @@
 	int ret;
 	u32 refs;
 
+	mutex_lock(&root->fs_info->alloc_mutex);
+
 	WARN_ON(*level < 0);
 	WARN_ON(*level >= BTRFS_MAX_LEVEL);
 	ret = lookup_extent_ref(trans, root,
@@ -2182,6 +2235,7 @@
 		bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
 		ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
 		blocksize = btrfs_level_size(root, *level - 1);
+
 		ret = lookup_extent_ref(trans, root, bytenr, blocksize, &refs);
 		BUG_ON(ret);
 		if (refs != 1) {
@@ -2189,7 +2243,7 @@
 			root_owner = btrfs_header_owner(parent);
 			root_gen = btrfs_header_generation(parent);
 			path->slots[*level]++;
-			ret = btrfs_free_extent(trans, root, bytenr,
+			ret = __btrfs_free_extent(trans, root, bytenr,
 						blocksize, root_owner,
 						root_gen, 0, 0, 1);
 			BUG_ON(ret);
@@ -2201,9 +2255,11 @@
 			reada_walk_down(root, cur, path->slots[*level]);
 
 			mutex_unlock(&root->fs_info->fs_mutex);
+			mutex_unlock(&root->fs_info->alloc_mutex);
 			next = read_tree_block(root, bytenr, blocksize,
 					       ptr_gen);
 			mutex_lock(&root->fs_info->fs_mutex);
+			mutex_lock(&root->fs_info->alloc_mutex);
 
 			/* we've dropped the lock, double check */
 			ret = lookup_extent_ref(trans, root, bytenr,
@@ -2216,7 +2272,7 @@
 
 				path->slots[*level]++;
 				free_extent_buffer(next);
-				ret = btrfs_free_extent(trans, root, bytenr,
+				ret = __btrfs_free_extent(trans, root, bytenr,
 							blocksize,
 							root_owner,
 							root_gen, 0, 0, 1);
@@ -2244,13 +2300,14 @@
 	}
 
 	root_gen = btrfs_header_generation(parent);
-	ret = btrfs_free_extent(trans, root, path->nodes[*level]->start,
+	ret = __btrfs_free_extent(trans, root, path->nodes[*level]->start,
 				path->nodes[*level]->len,
 				root_owner, root_gen, 0, 0, 1);
 	free_extent_buffer(path->nodes[*level]);
 	path->nodes[*level] = NULL;
 	*level += 1;
 	BUG_ON(ret);
+	mutex_unlock(&root->fs_info->alloc_mutex);
 	return 0;
 }
 
@@ -2350,6 +2407,12 @@
 		btrfs_node_key(node, &found_key, path->slots[level]);
 		WARN_ON(memcmp(&found_key, &root_item->drop_progress,
 			       sizeof(found_key)));
+		for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
+			if (path->nodes[i] && path->locks[i]) {
+				path->locks[i] = 0;
+				btrfs_tree_unlock(path->nodes[i]);
+			}
+		}
 	}
 	while(1) {
 		wret = walk_down_tree(trans, root, path, &level);
@@ -2383,6 +2446,8 @@
 	u64 end;
 	u64 ptr;
 	int ret;
+
+	mutex_lock(&info->alloc_mutex);
 	while(1) {
 		ret = find_first_extent_bit(&info->block_group_cache, 0,
 					    &start, &end, (unsigned int)-1);
@@ -2402,6 +2467,7 @@
 		clear_extent_dirty(&info->free_space_cache, start,
 				   end, GFP_NOFS);
 	}
+	mutex_unlock(&info->alloc_mutex);
 	return 0;
 }
 
@@ -2678,6 +2744,7 @@
 
 		eb = read_tree_block(found_root, extent_key->objectid,
 				     extent_key->offset, 0);
+		btrfs_tree_lock(eb);
 		level = btrfs_header_level(eb);
 
 		if (level == 0)
@@ -2685,6 +2752,7 @@
 		else
 			btrfs_node_key_to_cpu(eb, &found_key, 0);
 
+		btrfs_tree_unlock(eb);
 		free_extent_buffer(eb);
 
 		ret = find_root_for_ref(extent_root, path, &found_key,
@@ -2888,6 +2956,7 @@
 	int ret;
 	int progress;
 
+	mutex_lock(&root->fs_info->alloc_mutex);
 	shrink_block_group = btrfs_lookup_block_group(root->fs_info,
 						      shrink_start);
 	BUG_ON(!shrink_block_group);
@@ -3044,20 +3113,22 @@
 			   (unsigned int)-1, GFP_NOFS);
 out:
 	btrfs_free_path(path);
+	mutex_unlock(&root->fs_info->alloc_mutex);
 	return ret;
 }
 
 int find_first_block_group(struct btrfs_root *root, struct btrfs_path *path,
 			   struct btrfs_key *key)
 {
-	int ret;
+	int ret = 0;
 	struct btrfs_key found_key;
 	struct extent_buffer *leaf;
 	int slot;
 
 	ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
 	if (ret < 0)
-		return ret;
+		goto out;
+
 	while(1) {
 		slot = path->slots[0];
 		leaf = path->nodes[0];
@@ -3066,18 +3137,20 @@
 			if (ret == 0)
 				continue;
 			if (ret < 0)
-				goto error;
+				goto out;
 			break;
 		}
 		btrfs_item_key_to_cpu(leaf, &found_key, slot);
 
 		if (found_key.objectid >= key->objectid &&
-		    found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY)
-			return 0;
+		    found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) {
+			ret = 0;
+			goto out;
+		}
 		path->slots[0]++;
 	}
 	ret = -ENOENT;
-error:
+out:
 	return ret;
 }
 
@@ -3103,6 +3176,7 @@
 	if (!path)
 		return -ENOMEM;
 
+	mutex_lock(&root->fs_info->alloc_mutex);
 	while(1) {
 		ret = find_first_block_group(root, path, &key);
 		if (ret > 0) {
@@ -3158,6 +3232,7 @@
 	ret = 0;
 error:
 	btrfs_free_path(path);
+	mutex_unlock(&root->fs_info->alloc_mutex);
 	return ret;
 }
 
@@ -3205,5 +3280,6 @@
 	ret = del_pending_extents(trans, extent_root);
 	BUG_ON(ret);
 	set_avail_alloc_bits(extent_root->fs_info, type);
+
 	return 0;
 }
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 17c508a..bd15cdc 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -2889,7 +2889,6 @@
 
 	for (i = 0; i < num_pages; i++) {
 		page = extent_buffer_page(eb, i);
-		lock_page(page);
 		if (i == 0)
 			set_page_extent_head(page, eb->len);
 		else
@@ -2907,7 +2906,6 @@
 			end  = start + PAGE_CACHE_SIZE - 1;
 			if (test_range_bit(tree, start, end,
 					   EXTENT_DIRTY, 0)) {
-				unlock_page(page);
 				continue;
 			}
 		}
@@ -2919,7 +2917,6 @@
 						PAGECACHE_TAG_DIRTY);
 		}
 		read_unlock_irq(&page->mapping->tree_lock);
-		unlock_page(page);
 	}
 	return 0;
 }
@@ -2948,17 +2945,12 @@
 		 * on us if the page isn't already dirty.
 		 */
 		if (i == 0) {
-			lock_page(page);
 			set_page_extent_head(page, eb->len);
 		} else if (PagePrivate(page) &&
 			   page->private != EXTENT_PAGE_PRIVATE) {
-			lock_page(page);
 			set_page_extent_mapped(page);
-			unlock_page(page);
 		}
 		__set_page_dirty_nobuffers(extent_buffer_page(eb, i));
-		if (i == 0)
-			unlock_page(page);
 	}
 	return set_extent_dirty(tree, eb->start,
 				eb->start + eb->len - 1, GFP_NOFS);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 0c79346..61bd895 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -115,6 +115,7 @@
 	trans = btrfs_start_transaction(root, 1);
 	BUG_ON(!trans);
 	btrfs_set_trans_block_group(trans, inode);
+	mutex_unlock(&root->fs_info->fs_mutex);
 
 	num_bytes = (end - start + blocksize) & ~(blocksize - 1);
 	num_bytes = max(blocksize,  num_bytes);
@@ -159,6 +160,7 @@
 	btrfs_add_ordered_inode(inode);
 	btrfs_update_inode(trans, root, inode);
 out:
+	mutex_lock(&root->fs_info->fs_mutex);
 	btrfs_end_transaction(trans, root);
 	return ret;
 }
@@ -349,10 +351,12 @@
 
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
+	mutex_unlock(&root->fs_info->fs_mutex);
 
 	btrfs_set_trans_block_group(trans, inode);
 	btrfs_csum_file_blocks(trans, root, inode, bio, sums);
 
+	mutex_lock(&root->fs_info->fs_mutex);
 	ret = btrfs_end_transaction(trans, root);
 	BUG_ON(ret);
 	mutex_unlock(&root->fs_info->fs_mutex);
@@ -807,6 +811,7 @@
 		goto err;
 	}
 	ret = btrfs_delete_one_dir_name(trans, root, path, di);
+	btrfs_release_path(root, path);
 
 	dentry->d_inode->i_ctime = dir->i_ctime;
 	ret = btrfs_del_inode_ref(trans, root, name, name_len,
@@ -881,8 +886,9 @@
 	struct btrfs_trans_handle *trans;
 	unsigned long nr = 0;
 
-	if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
+	if (inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
 		return -ENOTEMPTY;
+	}
 
 	mutex_lock(&root->fs_info->fs_mutex);
 	ret = btrfs_check_free_space(root, 1, 1);
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 6fb4558..3fbf74e 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -43,6 +43,7 @@
 #include "ioctl.h"
 #include "print-tree.h"
 #include "volumes.h"
+#include "locking.h"
 
 
 
@@ -75,9 +76,9 @@
 	if (ret)
 		goto fail;
 
-	leaf = __btrfs_alloc_free_block(trans, root, root->leafsize,
-					objectid, trans->transid, 0, 0,
-					0, 0);
+	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
+				      objectid, trans->transid, 0, 0,
+				      0, 0);
 	if (IS_ERR(leaf))
 		return PTR_ERR(leaf);
 
@@ -108,6 +109,7 @@
 	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
 	root_item.drop_level = 0;
 
+	btrfs_tree_unlock(leaf);
 	free_extent_buffer(leaf);
 	leaf = NULL;
 
diff --git a/fs/btrfs/locking.c b/fs/btrfs/locking.c
new file mode 100644
index 0000000..80813a3
--- /dev/null
+++ b/fs/btrfs/locking.c
@@ -0,0 +1,50 @@
+/*
+ * Copyright (C) 2008 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+#include <linux/sched.h>
+#include <linux/gfp.h>
+#include <linux/pagemap.h>
+#include <linux/spinlock.h>
+#include <linux/page-flags.h>
+#include <linux/bug.h>
+#include "ctree.h"
+#include "extent_io.h"
+#include "locking.h"
+
+int btrfs_tree_lock(struct extent_buffer *eb)
+{
+	lock_page(eb->first_page);
+	return 0;
+}
+
+int btrfs_try_tree_lock(struct extent_buffer *eb)
+{
+	return TestSetPageLocked(eb->first_page);
+}
+
+int btrfs_tree_unlock(struct extent_buffer *eb)
+{
+	WARN_ON(!PageLocked(eb->first_page));
+	unlock_page(eb->first_page);
+	return 0;
+}
+
+int btrfs_tree_locked(struct extent_buffer *eb)
+{
+	return PageLocked(eb->first_page);
+}
+
diff --git a/fs/btrfs/locking.h b/fs/btrfs/locking.h
new file mode 100644
index 0000000..2dab96d
--- /dev/null
+++ b/fs/btrfs/locking.h
@@ -0,0 +1,26 @@
+/*
+ * Copyright (C) 2008 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef __BTRFS_LOCKING_
+#define __BTRFS_LOCKING_
+
+int btrfs_tree_lock(struct extent_buffer *eb);
+int btrfs_tree_unlock(struct extent_buffer *eb);
+int btrfs_tree_locked(struct extent_buffer *eb);
+int btrfs_try_tree_lock(struct extent_buffer *eb);
+#endif
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index f04684f..1ed433a 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -23,6 +23,7 @@
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
+#include "locking.h"
 
 static int total_trans = 0;
 extern struct kmem_cache *btrfs_trans_handle_cachep;
@@ -96,8 +97,7 @@
 			radix_tree_tag_set(&root->fs_info->fs_roots_radix,
 				   (unsigned long)root->root_key.objectid,
 				   BTRFS_ROOT_DEFRAG_TAG);
-			root->commit_root = root->node;
-			extent_buffer_get(root->node);
+			root->commit_root = btrfs_root_node(root);
 		} else {
 			WARN_ON(1);
 		}
@@ -559,6 +559,7 @@
 	struct btrfs_root *tree_root = fs_info->tree_root;
 	struct btrfs_root *root = pending->root;
 	struct extent_buffer *tmp;
+	struct extent_buffer *old;
 	int ret;
 	int namelen;
 	u64 objectid;
@@ -578,16 +579,18 @@
 	key.offset = 1;
 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
 
-	extent_buffer_get(root->node);
-	btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
-	free_extent_buffer(tmp);
+	old = btrfs_lock_root_node(root);
+	btrfs_cow_block(trans, root, old, NULL, 0, &old);
 
-	btrfs_copy_root(trans, root, root->node, &tmp, objectid);
+	btrfs_copy_root(trans, root, old, &tmp, objectid);
+	btrfs_tree_unlock(old);
+	free_extent_buffer(old);
 
 	btrfs_set_root_bytenr(new_root_item, tmp->start);
 	btrfs_set_root_level(new_root_item, btrfs_header_level(tmp));
 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 				new_root_item);
+	btrfs_tree_unlock(tmp);
 	free_extent_buffer(tmp);
 	if (ret)
 		goto fail;
diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c
index 155961c..fab851d 100644
--- a/fs/btrfs/tree-defrag.c
+++ b/fs/btrfs/tree-defrag.c
@@ -181,6 +181,8 @@
 	if (root->fs_info->extent_root == root)
 		is_extent = 1;
 
+	goto out;
+
 	if (root->ref_cows == 0 && !is_extent)
 		goto out;