X-Git-Url: http://drtracing.org/?a=blobdiff_plain;f=fs%2Fbtrfs%2Fctree.c;h=bbf9bf3740664e6e3d521377d1cff0de4afa9e87;hb=7d9eb12c8739e7dc80c78c6b3596f912ecd8f941;hp=8fa92a2d9819714489d4ee70931d2d58bbecfa15;hpb=8f662a76c6af8eb367fa519e9bb9766040d9cea8;p=deliverable%2Flinux.git

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 8fa92a2d9819..bbf9bf374066 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);
@@ -29,7 +30,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 		      struct btrfs_path *path, int data_size, int extend);
 static int push_node_left(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root, struct extent_buffer *dst,
-			  struct extent_buffer *src);
+			  struct extent_buffer *src, int empty);
 static int balance_node_right(struct btrfs_trans_handle *trans,
 			      struct btrfs_root *root,
 			      struct extent_buffer *dst_buf,
@@ -62,12 +63,57 @@ void btrfs_free_path(struct btrfs_path *p)
 void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
 {
 	int i;
+
 	for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
+		p->slots[i] = 0;
 		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]);
+		p->nodes[i] = NULL;
+	}
+}
+
+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)) {
+		list_add(&root->dirty_list,
+			 &root->fs_info->dirty_cowonly_roots);
 	}
-	memset(p, 0, sizeof(*p));
 }
 
 int btrfs_copy_root(struct btrfs_trans_handle *trans,
@@ -103,7 +149,7 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans,
 	} 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);
@@ -116,6 +162,7 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans,
 	btrfs_set_header_bytenr(cow, cow->start);
 	btrfs_set_header_generation(cow, trans->transid);
 	btrfs_set_header_owner(cow, new_root_objectid);
+	btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN);
 
 	WARN_ON(btrfs_header_generation(buf) > trans->transid);
 	ret = btrfs_inc_ref(trans, new_root, buf);
@@ -142,14 +189,19 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 	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 {
 		root_gen = 0;
 	}
-
 	WARN_ON(root->ref_cows && trans->transid !=
 		root->fs_info->running_transaction->transid);
 	WARN_ON(root->ref_cows && trans->transid != root->last_trans);
@@ -164,7 +216,7 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 	} 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);
@@ -175,6 +227,7 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 	btrfs_set_header_bytenr(cow, cow->start);
 	btrfs_set_header_generation(cow, trans->transid);
 	btrfs_set_header_owner(cow, root->root_key.objectid);
+	btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN);
 
 	WARN_ON(btrfs_header_generation(buf) > trans->transid);
 	if (btrfs_header_generation(buf) != trans->transid) {
@@ -187,15 +240,21 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 	}
 
 	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,
 					  root_gen, 0, 0, 1);
 		}
 		free_extent_buffer(buf);
+		add_root_to_dirty_list(root);
 	} else {
 		root_gen = btrfs_header_generation(parent);
 		btrfs_set_node_blockptr(parent, parent_slot,
@@ -209,6 +268,8 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 				  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;
@@ -221,7 +282,9 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		    struct extent_buffer **cow_ret)
 {
 	u64 search_start;
+	u64 header_trans;
 	int ret;
+
 	if (trans->transaction != root->fs_info->running_transaction) {
 		printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid,
 		       root->fs_info->running_transaction->transid);
@@ -232,12 +295,17 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		       root->fs_info->generation);
 		WARN_ON(1);
 	}
-	if (btrfs_header_generation(buf) == trans->transid) {
+
+	header_trans = btrfs_header_generation(buf);
+	spin_lock(&root->fs_info->hash_lock);
+	if (header_trans == trans->transid &&
+	    !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
 		*cow_ret = buf;
+		spin_unlock(&root->fs_info->hash_lock);
 		return 0;
 	}
-
-	search_start = buf->start & ~((u64)BTRFS_BLOCK_GROUP_SIZE - 1);
+	spin_unlock(&root->fs_info->hash_lock);
+	search_start = buf->start & ~((u64)(1024 * 1024 * 1024) - 1);
 	ret = __btrfs_cow_block(trans, root, buf, parent,
 				 parent_slot, cow_ret, search_start, 0);
 	return ret;
@@ -283,8 +351,8 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 		       struct btrfs_key *progress)
 {
 	struct extent_buffer *cur;
-	struct extent_buffer *tmp;
 	u64 blocknr;
+	u64 gen;
 	u64 search_start = *last_ret;
 	u64 last_block = 0;
 	u64 other;
@@ -337,6 +405,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 
 		progress_passed = 1;
 		blocknr = btrfs_node_blockptr(parent, i);
+		gen = btrfs_node_ptr_generation(parent, i);
 		if (last_block == 0)
 			last_block = blocknr;
 
@@ -344,7 +413,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 			other = btrfs_node_blockptr(parent, i - 1);
 			close = close_blocks(blocknr, other, blocksize);
 		}
-		if (close && i < end_slot - 2) {
+		if (!close && i < end_slot - 2) {
 			other = btrfs_node_blockptr(parent, i + 1);
 			close = close_blocks(blocknr, other, blocksize);
 		}
@@ -360,7 +429,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 
 		cur = btrfs_find_tree_block(root, blocknr, blocksize);
 		if (cur)
-			uptodate = btrfs_buffer_uptodate(cur);
+			uptodate = btrfs_buffer_uptodate(cur, gen);
 		else
 			uptodate = 0;
 		if (!cur || !uptodate) {
@@ -370,28 +439,29 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 			}
 			if (!cur) {
 				cur = read_tree_block(root, blocknr,
-							 blocksize);
+							 blocksize, gen);
 			} else if (!uptodate) {
-				btrfs_read_buffer(cur);
+				btrfs_read_buffer(cur, gen);
 			}
 		}
 		if (search_start == 0)
 			search_start = last_block;
 
+		btrfs_tree_lock(cur);
 		err = __btrfs_cow_block(trans, root, cur, parent, i,
-					&tmp, search_start,
+					&cur, search_start,
 					min(16 * blocksize,
 					    (end_slot - i) * blocksize));
 		if (err) {
+			btrfs_tree_unlock(cur);
 			free_extent_buffer(cur);
 			break;
 		}
-		search_start = tmp->start;
-		last_block = tmp->start;
+		search_start = cur->start;
+		last_block = cur->start;
 		*last_ret = search_start;
-		if (parent_level == 1)
-			btrfs_clear_buffer_defrag(tmp);
-		free_extent_buffer(tmp);
+		btrfs_tree_unlock(cur);
+		free_extent_buffer(cur);
 	}
 	if (parent->map_token) {
 		unmap_extent_buffer(parent, parent->map_token,
@@ -548,10 +618,20 @@ static int check_leaf(struct btrfs_root *root, struct btrfs_path *path,
 	return 0;
 }
 
-static int check_block(struct btrfs_root *root, struct btrfs_path *path,
-			int level)
+static int noinline check_block(struct btrfs_root *root,
+				struct btrfs_path *path, int level)
 {
+	u64 found_start;
 	return 0;
+	if (btrfs_header_level(path->nodes[level]) != level)
+	    printk("warning: bad level %Lu wanted %d found %d\n",
+		   path->nodes[level]->start, level,
+		   btrfs_header_level(path->nodes[level]));
+	found_start = btrfs_header_bytenr(path->nodes[level]);
+	if (found_start != path->nodes[level]->start) {
+	    printk("warning: bad bytentr %Lu found %Lu\n",
+		   path->nodes[level]->start, found_start);
+	}
 #if 0
 	struct extent_buffer *buf = path->nodes[level];
 
@@ -668,16 +748,22 @@ static int bin_search(struct extent_buffer *eb, struct btrfs_key *key,
 static struct extent_buffer *read_node_slot(struct btrfs_root *root,
 				   struct extent_buffer *parent, int slot)
 {
+	int level = btrfs_header_level(parent);
 	if (slot < 0)
 		return NULL;
 	if (slot >= btrfs_header_nritems(parent))
 		return NULL;
+
+	BUG_ON(level == 0);
+
 	return read_tree_block(root, btrfs_node_blockptr(parent, slot),
-		       btrfs_level_size(root, btrfs_header_level(parent) - 1));
+		       btrfs_level_size(root, level - 1),
+		       btrfs_node_ptr_generation(parent, slot));
 }
 
-static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
-			 *root, struct btrfs_path *path, int level)
+static int balance_level(struct btrfs_trans_handle *trans,
+			 struct btrfs_root *root,
+			 struct btrfs_path *path, int level)
 {
 	struct extent_buffer *right = NULL;
 	struct extent_buffer *mid;
@@ -694,6 +780,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 		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);
@@ -714,11 +801,21 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 
 		/* 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);
-		wait_on_tree_block_writeback(root, mid);
+		btrfs_tree_unlock(mid);
 		/* once for the path */
 		free_extent_buffer(mid);
 		ret = btrfs_free_extent(trans, root, mid->start, mid->len,
@@ -737,6 +834,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	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) {
@@ -746,6 +844,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	}
 	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) {
@@ -757,7 +856,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	/* first, try to make some room in the middle buffer */
 	if (left) {
 		orig_slot += btrfs_header_nritems(left);
-		wret = push_node_left(trans, root, left, mid);
+		wret = push_node_left(trans, root, left, mid, 1);
 		if (wret < 0)
 			ret = wret;
 		if (btrfs_header_nritems(mid) < 2)
@@ -768,7 +867,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	 * then try to empty the right most buffer into the middle
 	 */
 	if (right) {
-		wret = push_node_left(trans, root, mid, right);
+		wret = push_node_left(trans, root, mid, right, 1);
 		if (wret < 0 && wret != -ENOSPC)
 			ret = wret;
 		if (btrfs_header_nritems(right) == 0) {
@@ -777,7 +876,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 			u32 blocksize = right->len;
 
 			clean_tree_block(trans, root, right);
-			wait_on_tree_block_writeback(root, right);
+			btrfs_tree_unlock(right);
 			free_extent_buffer(right);
 			right = NULL;
 			wret = del_ptr(trans, root, path, level + 1, pslot +
@@ -813,6 +912,11 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 			ret = wret;
 			goto enospc;
 		}
+		if (wret == 1) {
+			wret = push_node_left(trans, root, left, mid, 1);
+			if (wret < 0)
+				ret = wret;
+		}
 		BUG_ON(wret == 1);
 	}
 	if (btrfs_header_nritems(mid) == 0) {
@@ -820,8 +924,9 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 		u64 root_gen = btrfs_header_generation(parent);
 		u64 bytenr = mid->start;
 		u32 blocksize = mid->len;
+
 		clean_tree_block(trans, root, mid);
-		wait_on_tree_block_writeback(root, mid);
+		btrfs_tree_unlock(mid);
 		free_extent_buffer(mid);
 		mid = NULL;
 		wret = del_ptr(trans, root, path, level + 1, pslot);
@@ -844,11 +949,14 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	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;
@@ -860,17 +968,22 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	    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;
 }
 
 /* returns zero if the push worked, non-zero otherwise */
-static int push_nodes_for_insert(struct btrfs_trans_handle *trans,
-				struct btrfs_root *root,
-				struct btrfs_path *path, int level)
+static int noinline push_nodes_for_insert(struct btrfs_trans_handle *trans,
+					  struct btrfs_root *root,
+					  struct btrfs_path *path, int level)
 {
 	struct extent_buffer *right = NULL;
 	struct extent_buffer *mid;
@@ -901,6 +1014,8 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 	/* 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;
@@ -911,7 +1026,7 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 				wret = 1;
 			else {
 				wret = push_node_left(trans, root,
-						      left, mid);
+						      left, mid, 0);
 			}
 		}
 		if (wret < 0)
@@ -926,24 +1041,28 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 				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;
@@ -972,12 +1091,15 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 				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;
@@ -1009,6 +1131,7 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 		return;
 
 	node = path->nodes[level];
+
 	search = btrfs_node_blockptr(node, slot);
 	blocksize = btrfs_level_size(root, level - 1);
 	eb = btrfs_find_tree_block(root, search, blocksize);
@@ -1041,7 +1164,8 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 		if ((search >= lowest_read && search <= highest_read) ||
 		    (search < lowest_read && lowest_read - search <= 32768) ||
 		    (search > highest_read && search - highest_read <= 32768)) {
-			readahead_tree_block(root, search, blocksize);
+			readahead_tree_block(root, search, blocksize,
+				     btrfs_node_ptr_generation(node, nr));
 			nread += blocksize;
 		}
 		nscan++;
@@ -1056,6 +1180,43 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 			highest_read = search;
 	}
 }
+
+static void unlock_up(struct btrfs_path *path, int level, int lowest_unlock)
+{
+	int i;
+	int skip_level = level;
+	int no_skips = 0;
+	struct extent_buffer *t;
+
+	for (i = level; i < BTRFS_MAX_LEVEL; i++) {
+		if (!path->nodes[i])
+			break;
+		if (!path->locks[i])
+			break;
+		if (!no_skips && path->slots[i] == 0) {
+			skip_level = i + 1;
+			continue;
+		}
+		if (!no_skips && path->keep_locks) {
+			u32 nritems;
+			t = path->nodes[i];
+			nritems = btrfs_header_nritems(t);
+			if (nritems < 1 || path->slots[i] >= nritems - 1) {
+				skip_level = i + 1;
+				continue;
+			}
+		}
+		if (skip_level < i && i >= lowest_unlock)
+			no_skips = 1;
+
+		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
@@ -1074,21 +1235,30 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
 		      ins_len, int cow)
 {
 	struct extent_buffer *b;
-	u64 bytenr;
-	u64 ptr_gen;
+	struct extent_buffer *tmp;
 	int slot;
 	int ret;
 	int level;
 	int should_reada = p->reada;
+	int lowest_unlock = 1;
+	int blocksize;
 	u8 lowest_level = 0;
+	u64 blocknr;
+	u64 gen;
 
 	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(cow && root == root->fs_info->extent_root &&
+		!mutex_is_locked(&root->fs_info->alloc_mutex));
+	if (ins_len < 0)
+		lowest_unlock = 2;
 again:
-	b = root->node;
-	extent_buffer_get(b);
+	if (p->skip_locking)
+		b = btrfs_root_node(root);
+	else
+		b = btrfs_lock_root_node(root);
+
 	while (b) {
 		level = btrfs_header_level(b);
 		if (cow) {
@@ -1107,16 +1277,19 @@ again:
 			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)
 				slot -= 1;
 			p->slots[level] = slot;
 			if (ins_len > 0 && btrfs_header_nritems(b) >=
-			    BTRFS_NODEPTRS_PER_BLOCK(root) - 1) {
+			    BTRFS_NODEPTRS_PER_BLOCK(root) - 3) {
 				int sret = split_node(trans, root, p, level);
 				BUG_ON(sret > 0);
 				if (sret)
@@ -1136,23 +1309,52 @@ again:
 				slot = p->slots[level];
 				BUG_ON(btrfs_header_nritems(b) == 1);
 			}
+			unlock_up(p, level, lowest_unlock);
+
 			/* this is only true while dropping a snapshot */
-			if (level == lowest_level)
+			if (level == lowest_level) {
 				break;
-			bytenr = btrfs_node_blockptr(b, slot);
-			ptr_gen = btrfs_node_ptr_generation(b, slot);
-			if (should_reada)
-				reada_for_search(root, p, level, slot,
-						 key->objectid);
-			b = read_tree_block(root, bytenr,
-					    btrfs_level_size(root, level - 1));
-			if (ptr_gen != btrfs_header_generation(b)) {
-				printk("block %llu bad gen wanted %llu "
-				       "found %llu\n",
-			        (unsigned long long)b->start,
-				(unsigned long long)ptr_gen,
-			        (unsigned long long)btrfs_header_generation(b));
 			}
+
+			blocknr = btrfs_node_blockptr(b, slot);
+			gen = btrfs_node_ptr_generation(b, slot);
+			blocksize = btrfs_level_size(root, level - 1);
+
+			tmp = btrfs_find_tree_block(root, blocknr, blocksize);
+			if (tmp && btrfs_buffer_uptodate(tmp, gen)) {
+				b = tmp;
+			} else {
+				/*
+				 * reduce lock contention at high levels
+				 * of the btree by dropping locks before
+				 * we read.
+				 */
+				if (level > 1) {
+					btrfs_release_path(NULL, p);
+					if (tmp)
+						free_extent_buffer(tmp);
+					if (should_reada)
+						reada_for_search(root, p,
+								 level, slot,
+								 key->objectid);
+
+					tmp = read_tree_block(root, blocknr,
+							 blocksize, gen);
+					if (tmp)
+						free_extent_buffer(tmp);
+					goto again;
+				} else {
+					if (tmp)
+						free_extent_buffer(tmp);
+					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);
 		} else {
 			p->slots[level] = slot;
 			if (ins_len > 0 && btrfs_leaf_free_space(root, b) <
@@ -1163,6 +1365,7 @@ again:
 				if (sret)
 					return sret;
 			}
+			unlock_up(p, level, lowest_unlock);
 			return ret;
 		}
 	}
@@ -1207,9 +1410,9 @@ static int fixup_low_keys(struct btrfs_trans_handle *trans,
  * returns 0 if some ptrs were pushed left, < 0 if there was some horrible
  * error, and > 0 if there was no room in the left hand block.
  */
-static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root
-			  *root, struct extent_buffer *dst,
-			  struct extent_buffer *src)
+static int push_node_left(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root, struct extent_buffer *dst,
+			  struct extent_buffer *src, int empty)
 {
 	int push_items = 0;
 	int src_nritems;
@@ -1222,12 +1425,27 @@ static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root
 	WARN_ON(btrfs_header_generation(src) != trans->transid);
 	WARN_ON(btrfs_header_generation(dst) != trans->transid);
 
+	if (!empty && src_nritems <= 8)
+		return 1;
+
 	if (push_items <= 0) {
 		return 1;
 	}
 
-	if (src_nritems < push_items)
-		push_items = src_nritems;
+	if (empty) {
+		push_items = min(src_nritems, push_items);
+		if (push_items < src_nritems) {
+			/* leave at least 8 pointers in the node if
+			 * we aren't going to empty it
+			 */
+			if (src_nritems - push_items < 8) {
+				if (push_items <= 8)
+					return 1;
+				push_items -= 8;
+			}
+		}
+	} else
+		push_items = min(src_nritems - 8, push_items);
 
 	copy_extent_buffer(dst, src,
 			   btrfs_node_key_ptr_offset(dst_nritems),
@@ -1273,13 +1491,19 @@ static int balance_node_right(struct btrfs_trans_handle *trans,
 	src_nritems = btrfs_header_nritems(src);
 	dst_nritems = btrfs_header_nritems(dst);
 	push_items = BTRFS_NODEPTRS_PER_BLOCK(root) - dst_nritems;
-	if (push_items <= 0)
+	if (push_items <= 0) {
+		return 1;
+	}
+
+	if (src_nritems < 4) {
 		return 1;
+	}
 
 	max_push = src_nritems / 2 + 1;
 	/* don't try to empty the node */
-	if (max_push >= src_nritems)
+	if (max_push >= src_nritems) {
 		return 1;
+	}
 
 	if (max_push < push_items)
 		push_items = max_push;
@@ -1309,7 +1533,7 @@ static int balance_node_right(struct btrfs_trans_handle *trans,
  *
  * returns zero on success or < 0 on failure.
  */
-static int insert_new_root(struct btrfs_trans_handle *trans,
+static int noinline insert_new_root(struct btrfs_trans_handle *trans,
 			   struct btrfs_root *root,
 			   struct btrfs_path *path, int level)
 {
@@ -1317,6 +1541,7 @@ static int insert_new_root(struct btrfs_trans_handle *trans,
 	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]);
@@ -1333,12 +1558,13 @@ static int insert_new_root(struct btrfs_trans_handle *trans,
 	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);
@@ -1349,6 +1575,11 @@ static int insert_new_root(struct btrfs_trans_handle *trans,
 	write_extent_buffer(c, root->fs_info->fsid,
 			    (unsigned long)btrfs_header_fsid(c),
 			    BTRFS_FSID_SIZE);
+
+	write_extent_buffer(c, root->fs_info->chunk_tree_uuid,
+			    (unsigned long)btrfs_header_chunk_tree_uuid(c),
+			    BTRFS_UUID_SIZE);
+
 	btrfs_set_node_key(c, &lower_key, 0);
 	btrfs_set_node_blockptr(c, 0, lower->start);
 	lower_gen = btrfs_header_generation(lower);
@@ -1358,22 +1589,31 @@ static int insert_new_root(struct btrfs_trans_handle *trans,
 
 	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;
@@ -1449,7 +1689,7 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 		ret = push_nodes_for_insert(trans, root, path, level);
 		c = path->nodes[level];
 		if (!ret && btrfs_header_nritems(c) <
-		    BTRFS_NODEPTRS_PER_BLOCK(root) - 1)
+		    BTRFS_NODEPTRS_PER_BLOCK(root) - 3)
 			return 0;
 		if (ret < 0)
 			return ret;
@@ -1462,7 +1702,7 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 		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),
@@ -1475,9 +1715,13 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_set_header_bytenr(split, split->start);
 	btrfs_set_header_generation(split, trans->transid);
 	btrfs_set_header_owner(split, root->root_key.objectid);
+	btrfs_set_header_flags(split, 0);
 	write_extent_buffer(split, root->fs_info->fsid,
 			    (unsigned long)btrfs_header_fsid(split),
 			    BTRFS_FSID_SIZE);
+	write_extent_buffer(split, root->fs_info->chunk_tree_uuid,
+			    (unsigned long)btrfs_header_chunk_tree_uuid(split),
+			    BTRFS_UUID_SIZE);
 
 	mid = (c_nritems + 1) / 2;
 
@@ -1501,10 +1745,12 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	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;
@@ -1584,32 +1830,27 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (slot >= btrfs_header_nritems(upper) - 1)
 		return 1;
 
-	right = read_tree_block(root, btrfs_node_blockptr(upper, slot + 1),
-				root->leafsize);
+	WARN_ON(!btrfs_tree_locked(path->nodes[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;
@@ -1645,10 +1886,8 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 		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);
@@ -1716,14 +1955,24 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 	/* 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
@@ -1760,12 +2009,14 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 		return 1;
 	}
 
-	left = read_tree_block(root, btrfs_node_blockptr(path->nodes[1],
-			       slot - 1), root->leafsize);
+	WARN_ON(!btrfs_tree_locked(path->nodes[1]));
+
+	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 */
@@ -1773,14 +2024,14 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 			      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)
@@ -1815,8 +2066,8 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 	}
 
 	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);
@@ -1914,15 +2165,23 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 	/* 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;
 }
 
 /*
@@ -1991,12 +2250,14 @@ again:
 
 	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);
-	if (IS_ERR(right))
+	if (IS_ERR(right)) {
+		BUG_ON(1);
 		return PTR_ERR(right);
+	}
 
 	memset_extent_buffer(right, 0, 0, sizeof(struct btrfs_header));
 	btrfs_set_header_bytenr(right, right->start);
@@ -2006,6 +2267,10 @@ again:
 	write_extent_buffer(right, root->fs_info->fsid,
 			    (unsigned long)btrfs_header_fsid(right),
 			    BTRFS_FSID_SIZE);
+
+	write_extent_buffer(right, root->fs_info->chunk_tree_uuid,
+			    (unsigned long)btrfs_header_chunk_tree_uuid(right),
+			    BTRFS_UUID_SIZE);
 	if (mid <= slot) {
 		if (nritems == 1 ||
 		    leaf_space_used(l, mid, nritems - mid) + space_needed >
@@ -2018,10 +2283,13 @@ again:
 						  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;
 				path->slots[1] += 1;
+				btrfs_mark_buffer_dirty(right);
 				return ret;
 			}
 			mid = slot;
@@ -2043,6 +2311,7 @@ again:
 						  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;
@@ -2052,6 +2321,7 @@ again:
 					if (wret)
 						ret = wret;
 				}
+				btrfs_mark_buffer_dirty(right);
 				return ret;
 			} else if (extend && slot == 0) {
 				mid = 1;
@@ -2115,12 +2385,15 @@ again:
 	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);
 
@@ -2327,27 +2600,30 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans,
  * Given a key and some data, insert an item into the tree.
  * This does all the path init required, making room in the tree if needed.
  */
-int btrfs_insert_empty_item(struct btrfs_trans_handle *trans,
+int btrfs_insert_empty_items(struct btrfs_trans_handle *trans,
 			    struct btrfs_root *root,
 			    struct btrfs_path *path,
-			    struct btrfs_key *cpu_key, u32 data_size)
+			    struct btrfs_key *cpu_key, u32 *data_size,
+			    int nr)
 {
 	struct extent_buffer *leaf;
 	struct btrfs_item *item;
 	int ret = 0;
 	int slot;
 	int slot_orig;
+	int i;
 	u32 nritems;
+	u32 total_size = 0;
+	u32 total_data = 0;
 	unsigned int data_end;
 	struct btrfs_disk_key disk_key;
 
-	btrfs_cpu_key_to_disk(&disk_key, cpu_key);
-
-	/* create a root if there isn't one */
-	if (!root->node)
-		BUG();
+	for (i = 0; i < nr; i++) {
+		total_data += data_size[i];
+	}
 
-	ret = btrfs_search_slot(trans, root, cpu_key, path, data_size, 1);
+	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) {
 		return -EEXIST;
 	}
@@ -2361,10 +2637,10 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans,
 	data_end = leaf_data_end(root, leaf);
 
 	if (btrfs_leaf_free_space(root, leaf) <
-	    sizeof(struct btrfs_item) + data_size) {
+	    sizeof(struct btrfs_item) + total_size) {
 		btrfs_print_leaf(root, leaf);
 		printk("not enough freespace need %u have %d\n",
-		       data_size, btrfs_leaf_free_space(root, leaf));
+		       total_size, btrfs_leaf_free_space(root, leaf));
 		BUG();
 	}
 
@@ -2399,7 +2675,7 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans,
 			}
 
 			ioff = btrfs_item_offset(leaf, item);
-			btrfs_set_item_offset(leaf, item, ioff - data_size);
+			btrfs_set_item_offset(leaf, item, ioff - total_data);
 		}
 		if (leaf->map_token) {
 			unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
@@ -2407,28 +2683,34 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans,
 		}
 
 		/* shift the items */
-		memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + 1),
+		memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + nr),
 			      btrfs_item_nr_offset(slot),
 			      (nritems - slot) * sizeof(struct btrfs_item));
 
 		/* shift the data */
 		memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) +
-			      data_end - data_size, btrfs_leaf_data(leaf) +
+			      data_end - total_data, btrfs_leaf_data(leaf) +
 			      data_end, old_data - data_end);
 		data_end = old_data;
 	}
 
 	/* setup the item for the new data */
-	btrfs_set_item_key(leaf, &disk_key, slot);
-	item = btrfs_item_nr(leaf, slot);
-	btrfs_set_item_offset(leaf, item, data_end - data_size);
-	btrfs_set_item_size(leaf, item, data_size);
-	btrfs_set_header_nritems(leaf, nritems + 1);
+	for (i = 0; i < nr; i++) {
+		btrfs_cpu_key_to_disk(&disk_key, cpu_key + i);
+		btrfs_set_item_key(leaf, &disk_key, slot + i);
+		item = btrfs_item_nr(leaf, slot + i);
+		btrfs_set_item_offset(leaf, item, data_end - data_size[i]);
+		data_end -= data_size[i];
+		btrfs_set_item_size(leaf, item, data_size[i]);
+	}
+	btrfs_set_header_nritems(leaf, nritems + nr);
 	btrfs_mark_buffer_dirty(leaf);
 
 	ret = 0;
-	if (slot == 0)
+	if (slot == 0) {
+		btrfs_cpu_key_to_disk(&disk_key, cpu_key);
 		ret = fixup_low_keys(trans, root, path, &disk_key, 1);
+	}
 
 	if (btrfs_leaf_free_space(root, leaf) < 0) {
 		btrfs_print_leaf(root, leaf);
@@ -2509,34 +2791,36 @@ static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
  * delete the item at the leaf level in path.  If that empties
  * the leaf, remove it from the tree
  */
-int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		   struct btrfs_path *path)
+int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		    struct btrfs_path *path, int slot, int nr)
 {
-	int slot;
 	struct extent_buffer *leaf;
 	struct btrfs_item *item;
-	int doff;
-	int dsize;
+	int last_off;
+	int dsize = 0;
 	int ret = 0;
 	int wret;
+	int i;
 	u32 nritems;
 
 	leaf = path->nodes[0];
-	slot = path->slots[0];
-	doff = btrfs_item_offset_nr(leaf, slot);
-	dsize = btrfs_item_size_nr(leaf, slot);
+	last_off = btrfs_item_offset_nr(leaf, slot + nr - 1);
+
+	for (i = 0; i < nr; i++)
+		dsize += btrfs_item_size_nr(leaf, slot + i);
+
 	nritems = btrfs_header_nritems(leaf);
 
-	if (slot != nritems - 1) {
+	if (slot + nr != nritems) {
 		int i;
 		int data_end = leaf_data_end(root, leaf);
 
 		memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) +
 			      data_end + dsize,
 			      btrfs_leaf_data(leaf) + data_end,
-			      doff - data_end);
+			      last_off - data_end);
 
-		for (i = slot + 1; i < nritems; i++) {
+		for (i = slot + nr; i < nritems; i++) {
 			u32 ioff;
 
 			item = btrfs_item_nr(leaf, i);
@@ -2557,12 +2841,12 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		}
 
 		memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot),
-			      btrfs_item_nr_offset(slot + 1),
+			      btrfs_item_nr_offset(slot + nr),
 			      sizeof(struct btrfs_item) *
-			      (nritems - slot - 1));
+			      (nritems - slot - nr));
 	}
-	btrfs_set_header_nritems(leaf, nritems - 1);
-	nritems--;
+	btrfs_set_header_nritems(leaf, nritems - nr);
+	nritems -= nr;
 
 	/* delete the leaf if we've emptied it */
 	if (nritems == 0) {
@@ -2570,8 +2854,6 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			btrfs_set_header_level(leaf, 0);
 		} else {
 			u64 root_gen = btrfs_header_generation(path->nodes[1]);
-			clean_tree_block(trans, root, leaf);
-			wait_on_tree_block_writeback(root, leaf);
 			wret = del_ptr(trans, root, path, 1, path->slots[1]);
 			if (wret)
 				ret = wret;
@@ -2595,7 +2877,7 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		}
 
 		/* delete the leaf if it is mostly empty */
-		if (used < BTRFS_LEAF_DATA_SIZE(root) / 3) {
+		if (used < BTRFS_LEAF_DATA_SIZE(root) / 4) {
 			/* push_leaf_left fixes the path.
 			 * make sure the path still points to our leaf
 			 * for possible call to del_ptr below
@@ -2603,13 +2885,13 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			slot = path->slots[1];
 			extent_buffer_get(leaf);
 
-			wret = push_leaf_right(trans, root, path, 1, 1);
+			wret = push_leaf_left(trans, root, path, 1, 1);
 			if (wret < 0 && wret != -ENOSPC)
 				ret = wret;
 
 			if (path->nodes[0] == leaf &&
 			    btrfs_header_nritems(leaf)) {
-				wret = push_leaf_left(trans, root, path, 1, 1);
+				wret = push_leaf_right(trans, root, path, 1, 1);
 				if (wret < 0 && wret != -ENOSPC)
 					ret = wret;
 			}
@@ -2622,9 +2904,6 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 				root_gen = btrfs_header_generation(
 							   path->nodes[1]);
 
-				clean_tree_block(trans, root, leaf);
-				wait_on_tree_block_writeback(root, leaf);
-
 				wret = del_ptr(trans, root, path, 1, slot);
 				if (wret)
 					ret = wret;
@@ -2637,7 +2916,13 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 				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 {
@@ -2648,60 +2933,220 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 }
 
 /*
- * 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)
 {
-	u64 bytenr;
+	struct btrfs_key key;
+	struct btrfs_disk_key found_key;
+	int ret;
+
+	btrfs_item_key_to_cpu(path->nodes[0], &key, 0);
+
+	if (key.offset > 0)
+		key.offset--;
+	else if (key.type > 0)
+		key.type--;
+	else if (key.objectid > 0)
+		key.objectid--;
+	else
+		return 1;
+
+	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;
+}
+
+/*
+ * A helper function to walk down the tree starting at min_key, and looking
+ * for nodes or leaves that are either in cache or have a minimum
+ * transaction id.  This is used by the btree defrag code, but could
+ * also be used to search for blocks that have changed since a given
+ * transaction id.
+ *
+ * This does not cow, but it does stuff the starting key it finds back
+ * into min_key, so you can call btrfs_search_slot with cow=1 on the
+ * key and get a writable path.
+ *
+ * This does lock as it descends, and path->keep_locks should be set
+ * to 1 by the caller.
+ *
+ * This honors path->lowest_level to prevent descent past a given level
+ * of the tree.
+ *
+ * returns zero if something useful was found, < 0 on error and 1 if there
+ * was nothing in the tree that matched the search criteria.
+ */
+int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key,
+			 struct btrfs_path *path, int cache_only,
+			 u64 min_trans)
+{
+	struct extent_buffer *cur;
+	struct btrfs_key found_key;
+	int slot;
+	u32 nritems;
+	int level;
+	int ret = 1;
+
+again:
+	cur = btrfs_lock_root_node(root);
+	level = btrfs_header_level(cur);
+	path->nodes[level] = cur;
+	path->locks[level] = 1;
+
+	if (btrfs_header_generation(cur) < min_trans) {
+		ret = 1;
+		goto out;
+	}
+	while(1) {
+		nritems = btrfs_header_nritems(cur);
+		level = btrfs_header_level(cur);
+		bin_search(cur, min_key, level, &slot);
+
+		/* at level = 0, we're done, setup the path and exit */
+		if (level == 0) {
+			ret = 0;
+			path->slots[level] = slot;
+			btrfs_item_key_to_cpu(cur, &found_key, slot);
+			goto out;
+		}
+		/*
+		 * check this node pointer against the cache_only and
+		 * min_trans parameters.  If it isn't in cache or is too
+		 * old, skip to the next one.
+		 */
+		while(slot < nritems) {
+			u64 blockptr;
+			u64 gen;
+			struct extent_buffer *tmp;
+			blockptr = btrfs_node_blockptr(cur, slot);
+			gen = btrfs_node_ptr_generation(cur, slot);
+			if (gen < min_trans) {
+				slot++;
+				continue;
+			}
+			if (!cache_only)
+				break;
+
+			tmp = btrfs_find_tree_block(root, blockptr,
+					    btrfs_level_size(root, level - 1));
+
+			if (tmp && btrfs_buffer_uptodate(tmp, gen)) {
+				free_extent_buffer(tmp);
+				break;
+			}
+			if (tmp)
+				free_extent_buffer(tmp);
+			slot++;
+		}
+		/*
+		 * we didn't find a candidate key in this node, walk forward
+		 * and find another one
+		 */
+		if (slot >= nritems) {
+			ret = btrfs_find_next_key(root, path, min_key, level,
+						  cache_only, min_trans);
+			if (ret == 0) {
+				btrfs_release_path(root, path);
+				goto again;
+			} else {
+				goto out;
+			}
+		}
+		/* save our key for returning back */
+		btrfs_node_key_to_cpu(cur, &found_key, slot);
+		path->slots[level] = slot;
+		if (level == path->lowest_level) {
+			ret = 0;
+			unlock_up(path, level, 1);
+			goto out;
+		}
+		cur = read_node_slot(root, cur, slot);
+
+		btrfs_tree_lock(cur);
+		path->locks[level - 1] = 1;
+		path->nodes[level - 1] = cur;
+		unlock_up(path, level, 1);
+	}
+out:
+	if (ret == 0)
+		memcpy(min_key, &found_key, sizeof(found_key));
+	return ret;
+}
+
+/*
+ * this is similar to btrfs_next_leaf, but does not try to preserve
+ * and fixup the path.  It looks for and returns the next key in the
+ * tree based on the current path and the cache_only and min_trans
+ * parameters.
+ *
+ * 0 is returned if another key is found, < 0 if there are any errors
+ * and 1 is returned if there are no higher keys in the tree
+ *
+ * path->keep_locks should be set to 1 on the search made before
+ * calling this function.
+ */
+int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path,
+			struct btrfs_key *key, int lowest_level,
+			int cache_only, u64 min_trans)
+{
+	int level = lowest_level;
 	int slot;
-	int level = 1;
 	struct extent_buffer *c;
-	struct extent_buffer *next = NULL;
 
 	while(level < BTRFS_MAX_LEVEL) {
 		if (!path->nodes[level])
 			return 1;
 
-		slot = path->slots[level];
+		slot = path->slots[level] + 1;
 		c = path->nodes[level];
-		if (slot == 0) {
+next:
+		if (slot >= btrfs_header_nritems(c)) {
 			level++;
-			if (level == BTRFS_MAX_LEVEL)
+			if (level == BTRFS_MAX_LEVEL) {
 				return 1;
+			}
 			continue;
 		}
-		slot--;
-
-		bytenr = btrfs_node_blockptr(c, slot);
-		if (next)
-			free_extent_buffer(next);
+		if (level == 0)
+			btrfs_item_key_to_cpu(c, key, slot);
+		else {
+			u64 blockptr = btrfs_node_blockptr(c, slot);
+			u64 gen = btrfs_node_ptr_generation(c, slot);
 
-		next = read_tree_block(root, bytenr,
-				       btrfs_level_size(root, level - 1));
-		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_tree_block(root, btrfs_node_blockptr(next, slot),
-				       btrfs_level_size(root, level - 1));
+			if (cache_only) {
+				struct extent_buffer *cur;
+				cur = btrfs_find_tree_block(root, blockptr,
+					    btrfs_level_size(root, level - 1));
+				if (!cur || !btrfs_buffer_uptodate(cur, gen)) {
+					slot++;
+					if (cur)
+						free_extent_buffer(cur);
+					goto next;
+				}
+				free_extent_buffer(cur);
+			}
+			if (gen < min_trans) {
+				slot++;
+				goto next;
+			}
+			btrfs_node_key_to_cpu(c, key, slot);
+		}
+		return 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.
  */
@@ -2709,9 +3154,38 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 {
 	int slot;
 	int level = 1;
-	u64 bytenr;
 	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);
+
+	btrfs_release_path(root, path);
+	path->keep_locks = 1;
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	path->keep_locks = 0;
+
+	if (ret < 0)
+		return ret;
+
+	nritems = btrfs_header_nritems(path->nodes[0]);
+	/*
+	 * by releasing the path above we dropped all our locks.  A balance
+	 * could have added more items next to the key that used to be
+	 * at the very end of the block.  So, check again here and
+	 * advance the path if there are now more items available.
+	 */
+	if (nritems > 0 && path->slots[0] < nritems - 1) {
+		path->slots[0]++;
+		goto done;
+	}
 
 	while(level < BTRFS_MAX_LEVEL) {
 		if (!path->nodes[level])
@@ -2721,35 +3195,80 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 		c = path->nodes[level];
 		if (slot >= btrfs_header_nritems(c)) {
 			level++;
-			if (level == BTRFS_MAX_LEVEL)
+			if (level == BTRFS_MAX_LEVEL) {
 				return 1;
+			}
 			continue;
 		}
 
-		bytenr = btrfs_node_blockptr(c, slot);
-		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_tree_block(root, bytenr,
-				       btrfs_level_size(root, level -1));
+		next = read_node_slot(root, c, slot);
+		if (!path->skip_locking) {
+			WARN_ON(!btrfs_tree_locked(c));
+			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;
+		if (!path->skip_locking)
+			path->locks[level] = 1;
 		if (!level)
 			break;
-		if (path->reada)
-			reada_for_search(root, path, level, 0, 0);
-		next = read_tree_block(root, btrfs_node_blockptr(next, 0),
-				       btrfs_level_size(root, level - 1));
+		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) {
+			WARN_ON(!btrfs_tree_locked(path->nodes[level]));
+			btrfs_tree_lock(next);
+		}
 	}
+done:
+	unlock_up(path, 0, 1);
 	return 0;
 }
+
+/*
+ * this uses btrfs_prev_leaf to walk backwards in the tree, and keeps
+ * searching until it gets past min_objectid or finds an item of 'type'
+ *
+ * returns 0 if something is found, 1 if nothing was found and < 0 on error
+ */
+int btrfs_previous_item(struct btrfs_root *root,
+			struct btrfs_path *path, u64 min_objectid,
+			int type)
+{
+	struct btrfs_key found_key;
+	struct extent_buffer *leaf;
+	int ret;
+
+	while(1) {
+		if (path->slots[0] == 0) {
+			ret = btrfs_prev_leaf(root, path);
+			if (ret != 0)
+				return ret;
+		} else {
+			path->slots[0]--;
+		}
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+		if (found_key.type == type)
+			return 0;
+	}
+	return 1;
+}
+