Btrfs: Allow tree blocks larger than the page size

Signed-off-by: Chris Mason <chris.mason@oracle.com>

fs/btrfs/ctree.c

@@ -83,16 +83,15 @@ static int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 
 	WARN_ON(root->ref_cows && trans->transid != root->last_trans);
 
-	cow = btrfs_alloc_free_block(trans, root, search_start, empty_size);
+	cow = btrfs_alloc_free_block(trans, root, buf->len,
+				     search_start, empty_size);
 	if (IS_ERR(cow))
 		return PTR_ERR(cow);
 
 	cow->alloc_addr = (unsigned long)__builtin_return_address(0);
-	if (buf->len != root->sectorsize || cow->len != root->sectorsize)
-		WARN_ON(1);
 
 	copy_extent_buffer(cow, buf, 0, 0, cow->len);
-	btrfs_set_header_blocknr(cow, extent_buffer_blocknr(cow));
+	btrfs_set_header_bytenr(cow, cow->start);
 	btrfs_set_header_generation(cow, trans->transid);
 	btrfs_set_header_owner(cow, root->root_key.objectid);
 
@@ -110,16 +109,16 @@ static int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 		root->node = cow;
 		extent_buffer_get(cow);
 		if (buf != root->commit_root) {
-			btrfs_free_extent(trans, root,
-					  extent_buffer_blocknr(buf), 1, 1);
+			btrfs_free_extent(trans, root, buf->start,
+					  buf->len, 1);
 		}
 		free_extent_buffer(buf);
 	} else {
 		btrfs_set_node_blockptr(parent, parent_slot,
-					extent_buffer_blocknr(cow));
+					cow->start);
 		btrfs_mark_buffer_dirty(parent);
 		WARN_ON(btrfs_header_generation(parent) != trans->transid);
-		btrfs_free_extent(trans, root, extent_buffer_blocknr(buf),1,1);
+		btrfs_free_extent(trans, root, buf->start, buf->len, 1);
 	}
 	free_extent_buffer(buf);
 	btrfs_mark_buffer_dirty(cow);
@@ -149,13 +148,14 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		return 0;
 	}
 
-	search_start = extent_buffer_blocknr(buf) & ~((u64)65535);
+	search_start = buf->start & ~((u64)BTRFS_BLOCK_GROUP_SIZE - 1);
 	ret = __btrfs_cow_block(trans, root, buf, parent,
 				 parent_slot, cow_ret, search_start, 0);
 	(*cow_ret)->alloc_addr = (unsigned long)__builtin_return_address(0);
 	return ret;
 }
 
+#if 0
 static int close_blocks(u64 blocknr, u64 other)
 {
 	if (blocknr < other && other - blocknr < 8)
@@ -165,7 +165,6 @@ static int close_blocks(u64 blocknr, u64 other)
 	return 0;
 }
 
-#if 0
 static int should_defrag_leaf(struct extent_buffer *eb)
 {
 	return 0;
@@ -355,7 +354,7 @@ static int check_node(struct btrfs_root *root, struct btrfs_path *path,
 		BUG_ON(memcmp(&parent_key, &node_key,
 			      sizeof(struct btrfs_disk_key)));
 		BUG_ON(btrfs_node_blockptr(parent, parent_slot) !=
-		       btrfs_header_blocknr(node));
+		       btrfs_header_bytenr(node));
 	}
 	BUG_ON(nritems > BTRFS_NODEPTRS_PER_BLOCK(root));
 	if (slot != 0) {
@@ -398,7 +397,7 @@ static int check_leaf(struct btrfs_root *root, struct btrfs_path *path,
 		BUG_ON(memcmp(&parent_key, &leaf_key,
 		       sizeof(struct btrfs_disk_key)));
 		BUG_ON(btrfs_node_blockptr(parent, parent_slot) !=
-		       btrfs_header_blocknr(leaf));
+		       btrfs_header_bytenr(leaf));
 	}
 #if 0
 	for (i = 0; nritems > 1 && i < nritems - 2; i++) {
@@ -467,14 +466,16 @@ static int check_leaf(struct btrfs_root *root, struct btrfs_path *path,
 static int check_block(struct btrfs_root *root, struct btrfs_path *path,
 			int level)
 {
+#if 0
 	struct extent_buffer *buf = path->nodes[level];
 
 	if (memcmp_extent_buffer(buf, root->fs_info->fsid,
 				 (unsigned long)btrfs_header_fsid(buf),
 				 BTRFS_FSID_SIZE)) {
 		printk("warning bad block %Lu\n", buf->start);
-		BUG();
+		return 1;
 	}
+#endif
 	if (level == 0)
 		return check_leaf(root, path, level);
 	return check_node(root, path, level);
@@ -585,7 +586,8 @@ static struct extent_buffer *read_node_slot(struct btrfs_root *root,
 		return NULL;
 	if (slot >= btrfs_header_nritems(parent))
 		return NULL;
-	return read_tree_block(root, btrfs_node_blockptr(parent, slot));
+	return read_tree_block(root, btrfs_node_blockptr(parent, slot),
+		       btrfs_level_size(root, btrfs_header_level(parent) - 1));
 }
 
 static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
@@ -618,7 +620,6 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	 */
 	if (!parent) {
 		struct extent_buffer *child;
-		u64 blocknr = extent_buffer_blocknr(mid);
 
 		if (btrfs_header_nritems(mid) != 1)
 			return 0;
@@ -632,9 +633,10 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 		wait_on_tree_block_writeback(root, mid);
 		/* once for the path */
 		free_extent_buffer(mid);
+		ret = btrfs_free_extent(trans, root, mid->start, mid->len, 1);
 		/* once for the root ptr */
 		free_extent_buffer(mid);
-		return btrfs_free_extent(trans, root, blocknr, 1, 1);
+		return ret;
 	}
 	if (btrfs_header_nritems(mid) >
 	    BTRFS_NODEPTRS_PER_BLOCK(root) / 4)
@@ -680,7 +682,9 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 		if (wret < 0 && wret != -ENOSPC)
 			ret = wret;
 		if (btrfs_header_nritems(right) == 0) {
-			u64 blocknr = extent_buffer_blocknr(right);
+			u64 bytenr = right->start;
+			u32 blocksize = right->len;
+
 			clean_tree_block(trans, root, right);
 			wait_on_tree_block_writeback(root, right);
 			free_extent_buffer(right);
@@ -689,7 +693,8 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 				       1);
 			if (wret)
 				ret = wret;
-			wret = btrfs_free_extent(trans, root, blocknr, 1, 1);
+			wret = btrfs_free_extent(trans, root, bytenr,
+						 blocksize, 1);
 			if (wret)
 				ret = wret;
 		} else {
@@ -719,7 +724,8 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	}
 	if (btrfs_header_nritems(mid) == 0) {
 		/* we've managed to empty the middle node, drop it */
-		u64 blocknr = extent_buffer_blocknr(mid);
+		u64 bytenr = mid->start;
+		u32 blocksize = mid->len;
 		clean_tree_block(trans, root, mid);
 		wait_on_tree_block_writeback(root, mid);
 		free_extent_buffer(mid);
@@ -727,7 +733,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 		wret = del_ptr(trans, root, path, level + 1, pslot);
 		if (wret)
 			ret = wret;
-		wret = btrfs_free_extent(trans, root, blocknr, 1, 1);
+		wret = btrfs_free_extent(trans, root, bytenr, blocksize, 1);
 		if (wret)
 			ret = wret;
 	} else {
@@ -830,7 +836,6 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 				path->slots[level] = orig_slot;
 				free_extent_buffer(left);
 			}
-			check_node(root, path, level);
 			return 0;
 		}
 		free_extent_buffer(left);
@@ -874,12 +879,10 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 			} else {
 				free_extent_buffer(right);
 			}
-			check_node(root, path, level);
 			return 0;
 		}
 		free_extent_buffer(right);
 	}
-	check_node(root, path, level);
 	return 1;
 }
 
@@ -889,19 +892,23 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 			     int level, int slot)
 {
+	return;
+#if 0
 	struct extent_buffer *node;
 	int i;
 	u32 nritems;
-	u64 blocknr;
+	u64 bytenr;
 	u64 search;
 	u64 cluster_start;
 	int ret;
 	int nread = 0;
 	int direction = path->reada;
+	int level;
 	struct radix_tree_root found;
 	unsigned long gang[8];
 	struct extent_buffer *eb;
 
+
 	if (level == 0)
 		return;
 
@@ -918,8 +925,9 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 
 	init_bit_radix(&found);
 	nritems = btrfs_header_nritems(node);
+	level = btrfs_header_level(node) - 1;
 	for (i = slot; i < nritems; i++) {
-		blocknr = btrfs_node_blockptr(node, i);
+		bytenr = btrfs_node_blockptr(node, i);
 		set_radix_bit(&found, blocknr);
 	}
 	if (direction > 0) {
@@ -944,6 +952,7 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 			}
 		}
 	}
+#endif
 }
 /*
  * look for key in the tree.  path is filled in with nodes along the way
@@ -963,7 +972,7 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
 		      ins_len, int cow)
 {
 	struct extent_buffer *b;
-	u64 blocknr;
+	u64 bytenr;
 	int slot;
 	int ret;
 	int level;
@@ -1027,10 +1036,11 @@ again:
 			/* this is only true while dropping a snapshot */
 			if (level == lowest_level)
 				break;
-			blocknr = btrfs_node_blockptr(b, slot);
+			bytenr = btrfs_node_blockptr(b, slot);
 			if (should_reada)
 				reada_for_search(root, p, level, slot);
-			b = read_tree_block(root, btrfs_node_blockptr(b, slot));
+			b = read_tree_block(root, bytenr,
+					    btrfs_level_size(root, level - 1));
 		} else {
 			p->slots[level] = slot;
 			if (ins_len > 0 && btrfs_leaf_free_space(root, b) <
@@ -1193,14 +1203,14 @@ static int insert_new_root(struct btrfs_trans_handle *trans,
 	BUG_ON(path->nodes[level]);
 	BUG_ON(path->nodes[level-1] != root->node);
 
-	c = btrfs_alloc_free_block(trans, root,
-				   extent_buffer_blocknr(root->node), 0);
+	c = btrfs_alloc_free_block(trans, root, root->nodesize,
+				   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);
-	btrfs_set_header_blocknr(c, extent_buffer_blocknr(c));
+	btrfs_set_header_bytenr(c, c->start);
 	btrfs_set_header_generation(c, trans->transid);
 	btrfs_set_header_owner(c, root->root_key.objectid);
 	lower = path->nodes[level-1];
@@ -1213,7 +1223,7 @@ static int insert_new_root(struct btrfs_trans_handle *trans,
 	else
 		btrfs_node_key(lower, &lower_key, 0);
 	btrfs_set_node_key(c, &lower_key, 0);
-	btrfs_set_node_blockptr(c, 0, extent_buffer_blocknr(lower));
+	btrfs_set_node_blockptr(c, 0, lower->start);
 
 	btrfs_mark_buffer_dirty(c);
 
@@ -1237,7 +1247,7 @@ static int insert_new_root(struct btrfs_trans_handle *trans,
  */
 static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_path *path, struct btrfs_disk_key
-		      *key, u64 blocknr, int slot, int level)
+		      *key, u64 bytenr, int slot, int level)
 {
 	struct extent_buffer *lower;
 	int nritems;
@@ -1256,10 +1266,9 @@ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root
 			      (nritems - slot) * sizeof(struct btrfs_key_ptr));
 	}
 	btrfs_set_node_key(lower, key, slot);
-	btrfs_set_node_blockptr(lower, slot, blocknr);
+	btrfs_set_node_blockptr(lower, slot, bytenr);
 	btrfs_set_header_nritems(lower, nritems + 1);
 	btrfs_mark_buffer_dirty(lower);
-	check_node(root, path, level);
 	return 0;
 }
 
@@ -1300,14 +1309,14 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 	}
 
 	c_nritems = btrfs_header_nritems(c);
-	split = btrfs_alloc_free_block(trans, root,
-				       extent_buffer_blocknr(c), 0);
+	split = btrfs_alloc_free_block(trans, root, root->nodesize,
+				       c->start, 0);
 	if (IS_ERR(split))
 		return PTR_ERR(split);
 
 	btrfs_set_header_flags(split, btrfs_header_flags(c));
 	btrfs_set_header_level(split, btrfs_header_level(c));
-	btrfs_set_header_blocknr(split, extent_buffer_blocknr(split));
+	btrfs_set_header_bytenr(split, split->start);
 	btrfs_set_header_generation(split, trans->transid);
 	btrfs_set_header_owner(split, root->root_key.objectid);
 	write_extent_buffer(split, root->fs_info->fsid,
@@ -1328,8 +1337,7 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_mark_buffer_dirty(split);
 
 	btrfs_node_key(split, &disk_key, 0);
-	wret = insert_ptr(trans, root, path, &disk_key,
-			  extent_buffer_blocknr(split),
+	wret = insert_ptr(trans, root, path, &disk_key, split->start,
 			  path->slots[level + 1] + 1,
 			  level + 1);
 	if (wret)
@@ -1407,6 +1415,7 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 	u32 left_nritems;
 	u32 right_nritems;
 	u32 data_end;
+	u32 this_item_size;
 	int ret;
 
 	slot = path->slots[1];
@@ -1417,7 +1426,8 @@ 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));
+	right = read_tree_block(root, btrfs_node_blockptr(upper, slot + 1),
+				root->leafsize);
 	free_space = btrfs_leaf_free_space(root, right);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
 		free_extent_buffer(right);
@@ -1445,13 +1455,27 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	for (i = left_nritems - 1; i >= 1; i--) {
 		item = btrfs_item_nr(left, i);
+
 		if (path->slots[0] == i)
 			push_space += data_size + sizeof(*item);
-		if (btrfs_item_size(left, item) + sizeof(*item) + push_space >
-		    free_space)
+
+		if (!left->map_token) {
+			map_extent_buffer(left, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&left->map_token, &left->kaddr,
+					&left->map_start, &left->map_len,
+					KM_USER1);
+		}
+
+		this_item_size = btrfs_item_size(left, item);
+		if (this_item_size + sizeof(*item) + push_space > free_space)
 			break;
 		push_items++;
-		push_space += btrfs_item_size(left, item) + sizeof(*item);
+		push_space += this_item_size + sizeof(*item);
+	}
+	if (left->map_token) {
+		unmap_extent_buffer(left, left->map_token, KM_USER1);
+		left->map_token = NULL;
 	}
 
 	if (push_items == 0) {
@@ -1493,11 +1517,23 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 	right_nritems += push_items;
 	btrfs_set_header_nritems(right, right_nritems);
 	push_space = BTRFS_LEAF_DATA_SIZE(root);
+
 	for (i = 0; i < right_nritems; i++) {
 		item = btrfs_item_nr(right, i);
-		btrfs_set_item_offset(right, item, push_space -
-				      btrfs_item_size(right, item));
-		push_space = btrfs_item_offset(right, item);
+		if (!right->map_token) {
+			map_extent_buffer(right, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&right->map_token, &right->kaddr,
+					&right->map_start, &right->map_len,
+					KM_USER1);
+		}
+		push_space -= btrfs_item_size(right, item);
+		btrfs_set_item_offset(right, item, push_space);
+	}
+
+	if (right->map_token) {
+		unmap_extent_buffer(right, right->map_token, KM_USER1);
+		right->map_token = NULL;
 	}
 	left_nritems -= push_items;
 	btrfs_set_header_nritems(left, left_nritems);
@@ -1518,8 +1554,6 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 	} else {
 		free_extent_buffer(right);
 	}
-	if (path->nodes[1])
-		check_node(root, path, 1);
 	return 0;
 }
 /*
@@ -1542,6 +1576,8 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 	u32 right_nritems;
 	int ret = 0;
 	int wret;
+	u32 this_item_size;
+	u32 old_left_item_size;
 
 	slot = path->slots[1];
 	if (slot == 0)
@@ -1550,7 +1586,7 @@ 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));
+			       slot - 1), root->leafsize);
 	free_space = btrfs_leaf_free_space(root, left);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
 		free_extent_buffer(left);
@@ -1579,14 +1615,30 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	for (i = 0; i < right_nritems - 1; i++) {
 		item = btrfs_item_nr(right, i);
+		if (!right->map_token) {
+			map_extent_buffer(right, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&right->map_token, &right->kaddr,
+					&right->map_start, &right->map_len,
+					KM_USER1);
+		}
+
 		if (path->slots[0] == i)
 			push_space += data_size + sizeof(*item);
-		if (btrfs_item_size(right, item) + sizeof(*item) + push_space >
-		    free_space)
+
+		this_item_size = btrfs_item_size(right, item);
+		if (this_item_size + sizeof(*item) + push_space > free_space)
 			break;
+
 		push_items++;
-		push_space += btrfs_item_size(right, item) + sizeof(*item);
+		push_space += this_item_size + sizeof(*item);
+	}
+
+	if (right->map_token) {
+		unmap_extent_buffer(right, right->map_token, KM_USER1);
+		right->map_token = NULL;
 	}
+
 	if (push_items == 0) {
 		free_extent_buffer(left);
 		return 1;
@@ -1611,15 +1663,28 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 	old_left_nritems = btrfs_header_nritems(left);
 	BUG_ON(old_left_nritems < 0);
 
+	old_left_item_size = btrfs_item_offset_nr(left, old_left_nritems - 1);
 	for (i = old_left_nritems; i < old_left_nritems + push_items; i++) {
 		u32 ioff;
+
 		item = btrfs_item_nr(left, i);
+		if (!left->map_token) {
+			map_extent_buffer(left, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&left->map_token, &left->kaddr,
+					&left->map_start, &left->map_len,
+					KM_USER1);
+		}
+
 		ioff = btrfs_item_offset(left, item);
 		btrfs_set_item_offset(left, item,
-		      ioff - (BTRFS_LEAF_DATA_SIZE(root) -
-		      btrfs_item_offset_nr(left, old_left_nritems - 1)));
+		      ioff - (BTRFS_LEAF_DATA_SIZE(root) - old_left_item_size));
 	}
 	btrfs_set_header_nritems(left, old_left_nritems + push_items);
+	if (left->map_token) {
+		unmap_extent_buffer(left, left->map_token, KM_USER1);
+		left->map_token = NULL;
+	}
 
 	/* fixup right node */
 	push_space = btrfs_item_offset_nr(right, push_items - 1) -
@@ -1640,9 +1705,21 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	for (i = 0; i < right_nritems; i++) {
 		item = btrfs_item_nr(right, i);
-		btrfs_set_item_offset(right, item, push_space -
-				      btrfs_item_size(right, item));
-		push_space = btrfs_item_offset(right, item);
+
+		if (!right->map_token) {
+			map_extent_buffer(right, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&right->map_token, &right->kaddr,
+					&right->map_start, &right->map_len,
+					KM_USER1);
+		}
+
+		push_space = push_space - btrfs_item_size(right, item);
+		btrfs_set_item_offset(right, item, push_space);
+	}
+	if (right->map_token) {
+		unmap_extent_buffer(right, right->map_token, KM_USER1);
+		right->map_token = NULL;
 	}
 
 	btrfs_mark_buffer_dirty(left);
@@ -1664,8 +1741,6 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 		path->slots[0] -= push_items;
 	}
 	BUG_ON(path->slots[0] < 0);
-	if (path->nodes[1])
-		check_node(root, path, 1);
 	return ret;
 }
 
@@ -1718,13 +1793,13 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 	nritems = btrfs_header_nritems(l);
 	mid = (nritems + 1)/ 2;
 
-	right = btrfs_alloc_free_block(trans, root,
-					      extent_buffer_blocknr(l), 0);
+	right = btrfs_alloc_free_block(trans, root, root->leafsize,
+				       l->start, 0);
 	if (IS_ERR(right))
 		return PTR_ERR(right);
 
 	memset_extent_buffer(right, 0, 0, sizeof(struct btrfs_header));
-	btrfs_set_header_blocknr(right, extent_buffer_blocknr(right));
+	btrfs_set_header_bytenr(right, right->start);
 	btrfs_set_header_generation(right, trans->transid);
 	btrfs_set_header_owner(right, root->root_key.objectid);
 	btrfs_set_header_level(right, 0);
@@ -1740,8 +1815,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 				btrfs_cpu_key_to_disk(&disk_key, ins_key);
 				btrfs_set_header_nritems(right, 0);
 				wret = insert_ptr(trans, root, path,
-						  &disk_key,
-						  extent_buffer_blocknr(right),
+						  &disk_key, right->start,
 						  path->slots[1] + 1, 1);
 				if (wret)
 					ret = wret;
@@ -1762,7 +1836,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 				btrfs_set_header_nritems(right, 0);
 				wret = insert_ptr(trans, root, path,
 						  &disk_key,
-						  extent_buffer_blocknr(right),
+						  right->start,
 						  path->slots[1], 1);
 				if (wret)
 					ret = wret;
@@ -1799,15 +1873,30 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	for (i = 0; i < nritems; i++) {
 		struct btrfs_item *item = btrfs_item_nr(right, i);
-		u32 ioff = btrfs_item_offset(right, item);
+		u32 ioff;
+
+		if (!right->map_token) {
+			map_extent_buffer(right, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&right->map_token, &right->kaddr,
+					&right->map_start, &right->map_len,
+					KM_USER1);
+		}
+
+		ioff = btrfs_item_offset(right, item);
 		btrfs_set_item_offset(right, item, ioff + rt_data_off);
 	}
 
+	if (right->map_token) {
+		unmap_extent_buffer(right, right->map_token, KM_USER1);
+		right->map_token = NULL;
+	}
+
 	btrfs_set_header_nritems(l, mid);
 	ret = 0;
 	btrfs_item_key(right, &disk_key, 0);
-	wret = insert_ptr(trans, root, path, &disk_key,
-			  extent_buffer_blocknr(right), path->slots[1] + 1, 1);
+	wret = insert_ptr(trans, root, path, &disk_key, right->start,
+			  path->slots[1] + 1, 1);
 	if (wret)
 		ret = wret;
 
@@ -1824,19 +1913,17 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 		free_extent_buffer(right);
 
 	BUG_ON(path->slots[0] < 0);
-	check_node(root, path, 1);
-	check_leaf(root, path, 0);
 
 	if (!double_split)
 		return ret;
 
-	right = btrfs_alloc_free_block(trans, root,
-				       extent_buffer_blocknr(l), 0);
+	right = btrfs_alloc_free_block(trans, root, root->leafsize,
+				       l->start, 0);
 	if (IS_ERR(right))
 		return PTR_ERR(right);
 
 	memset_extent_buffer(right, 0, 0, sizeof(struct btrfs_header));
-	btrfs_set_header_blocknr(right, extent_buffer_blocknr(right));
+	btrfs_set_header_bytenr(right, right->start);
 	btrfs_set_header_generation(right, trans->transid);
 	btrfs_set_header_owner(right, root->root_key.objectid);
 	btrfs_set_header_level(right, 0);
@@ -1847,8 +1934,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_cpu_key_to_disk(&disk_key, ins_key);
 	btrfs_set_header_nritems(right, 0);
 	wret = insert_ptr(trans, root, path,
-			  &disk_key,
-			  extent_buffer_blocknr(right),
+			  &disk_key, right->start,
 			  path->slots[1], 1);
 	if (wret)
 		ret = wret;
@@ -1860,8 +1946,6 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 	free_extent_buffer(path->nodes[0]);
 	path->nodes[0] = right;
 	path->slots[0] = 0;
-	check_node(root, path, 1);
-	check_leaf(root, path, 0);
 	return ret;
 }
 
@@ -1904,9 +1988,24 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans,
 	for (i = slot; i < nritems; i++) {
 		u32 ioff;
 		item = btrfs_item_nr(leaf, i);
+
+		if (!leaf->map_token) {
+			map_extent_buffer(leaf, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&leaf->map_token, &leaf->kaddr,
+					&leaf->map_start, &leaf->map_len,
+					KM_USER1);
+		}
+
 		ioff = btrfs_item_offset(leaf, item);
 		btrfs_set_item_offset(leaf, item, ioff + size_diff);
 	}
+
+	if (leaf->map_token) {
+		unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
+		leaf->map_token = NULL;
+	}
+
 	/* shift the data */
 	memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) +
 		      data_end + size_diff, btrfs_leaf_data(leaf) +
@@ -1921,7 +2020,6 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans,
 		btrfs_print_leaf(root, leaf);
 		BUG();
 	}
-	check_leaf(root, path, 0);
 	return ret;
 }
 
@@ -1963,10 +2061,23 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans,
 	for (i = slot; i < nritems; i++) {
 		u32 ioff;
 		item = btrfs_item_nr(leaf, i);
+
+		if (!leaf->map_token) {
+			map_extent_buffer(leaf, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&leaf->map_token, &leaf->kaddr,
+					&leaf->map_start, &leaf->map_len,
+					KM_USER1);
+		}
 		ioff = btrfs_item_offset(leaf, item);
 		btrfs_set_item_offset(leaf, item, ioff - data_size);
 	}
 
+	if (leaf->map_token) {
+		unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
+		leaf->map_token = NULL;
+	}
+
 	/* shift the data */
 	memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) +
 		      data_end - data_size, btrfs_leaf_data(leaf) +
@@ -1983,7 +2094,6 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans,
 		btrfs_print_leaf(root, leaf);
 		BUG();
 	}
-	check_leaf(root, path, 0);
 	return ret;
 }
 
@@ -2046,12 +2156,26 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans,
 		 * item0..itemN ... dataN.offset..dataN.size .. data0.size
 		 */
 		/* first correct the data pointers */
+		WARN_ON(leaf->map_token);
 		for (i = slot; i < nritems; i++) {
 			u32 ioff;
+
 			item = btrfs_item_nr(leaf, i);
+			if (!leaf->map_token) {
+				map_extent_buffer(leaf, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&leaf->map_token, &leaf->kaddr,
+					&leaf->map_start, &leaf->map_len,
+					KM_USER1);
+			}
+
 			ioff = btrfs_item_offset(leaf, item);
 			btrfs_set_item_offset(leaf, item, ioff - data_size);
 		}
+		if (leaf->map_token) {
+			unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
+			leaf->map_token = NULL;
+		}
 
 		/* shift the items */
 		memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + 1),
@@ -2081,7 +2205,6 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans,
 		btrfs_print_leaf(root, leaf);
 		BUG();
 	}
-	check_leaf(root, path, 0);
 out:
 	return ret;
 }
@@ -2186,10 +2309,24 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 
 		for (i = slot + 1; i < nritems; i++) {
 			u32 ioff;
+
 			item = btrfs_item_nr(leaf, i);
+			if (!leaf->map_token) {
+				map_extent_buffer(leaf, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&leaf->map_token, &leaf->kaddr,
+					&leaf->map_start, &leaf->map_len,
+					KM_USER1);
+			}
 			ioff = btrfs_item_offset(leaf, item);
 			btrfs_set_item_offset(leaf, item, ioff + dsize);
 		}
+
+		if (leaf->map_token) {
+			unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
+			leaf->map_token = NULL;
+		}
+
 		memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot),
 			      btrfs_item_nr_offset(slot + 1),
 			      sizeof(struct btrfs_item) *
@@ -2209,8 +2346,7 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			if (wret)
 				ret = wret;
 			wret = btrfs_free_extent(trans, root,
-						 extent_buffer_blocknr(leaf),
-						 1, 1);
+						 leaf->start, leaf->len, 1);
 			if (wret)
 				ret = wret;
 		}
@@ -2247,7 +2383,8 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			}
 
 			if (btrfs_header_nritems(leaf) == 0) {
-				u64 blocknr = extent_buffer_blocknr(leaf);
+				u64 bytenr = leaf->start;
+				u32 blocksize = leaf->len;
 
 				clean_tree_block(trans, root, leaf);
 				wait_on_tree_block_writeback(root, leaf);
@@ -2257,8 +2394,8 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 					ret = wret;
 
 				free_extent_buffer(leaf);
-				wret = btrfs_free_extent(trans, root, blocknr,
-							 1, 1);
+				wret = btrfs_free_extent(trans, root, bytenr,
+							 blocksize, 1);
 				if (wret)
 					ret = wret;
 			} else {
@@ -2281,7 +2418,7 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 {
 	int slot;
 	int level = 1;
-	u64 blocknr;
+	u64 bytenr;
 	struct extent_buffer *c;
 	struct extent_buffer *next = NULL;
 
@@ -2296,14 +2433,15 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 			continue;
 		}
 
-		blocknr = btrfs_node_blockptr(c, slot);
+		bytenr = btrfs_node_blockptr(c, slot);
 		if (next)
 			free_extent_buffer(next);
 
 		if (path->reada)
 			reada_for_search(root, path, level, slot);
 
-		next = read_tree_block(root, blocknr);
+		next = read_tree_block(root, bytenr,
+				       btrfs_level_size(root, level -1));
 		break;
 	}
 	path->slots[level] = slot;
@@ -2317,7 +2455,8 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 			break;
 		if (path->reada)
 			reada_for_search(root, path, level, 0);
-		next = read_tree_block(root, btrfs_node_blockptr(next, 0));
+		next = read_tree_block(root, btrfs_node_blockptr(next, 0),
+				       btrfs_level_size(root, level - 1));
 	}
 	return 0;
 }

fs/btrfs/ctree.h

@@ -96,7 +96,7 @@ struct btrfs_key {
 struct btrfs_header {
 	u8 csum[BTRFS_CSUM_SIZE];
 	u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
-	__le64 blocknr; /* which block this node is supposed to live in */
+	__le64 bytenr; /* which block this node is supposed to live in */
 	__le64 generation;
 	__le64 owner;
 	__le32 nritems;
@@ -122,16 +122,17 @@ struct btrfs_super_block {
 	u8 csum[BTRFS_CSUM_SIZE];
 	/* the first 3 fields must match struct btrfs_header */
 	u8 fsid[16];    /* FS specific uuid */
-	__le64 blocknr; /* this block number */
+	__le64 bytenr; /* this block number */
 	__le64 magic;
 	__le64 generation;
 	__le64 root;
-	__le64 total_blocks;
-	__le64 blocks_used;
+	__le64 total_bytes;
+	__le64 bytes_used;
 	__le64 root_dir_objectid;
 	__le32 sectorsize;
 	__le32 nodesize;
 	__le32 leafsize;
+	u8 root_level;
 } __attribute__ ((__packed__));
 
 /*
@@ -231,13 +232,14 @@ struct btrfs_dir_item {
 struct btrfs_root_item {
 	struct btrfs_inode_item inode;
 	__le64 root_dirid;
-	__le64 blocknr;
-	__le64 block_limit;
-	__le64 blocks_used;
+	__le64 bytenr;
+	__le64 byte_limit;
+	__le64 bytes_used;
 	__le32 flags;
 	__le32 refs;
 	struct btrfs_disk_key drop_progress;
 	u8 drop_level;
+	u8 level;
 } __attribute__ ((__packed__));
 
 #define BTRFS_FILE_EXTENT_REG 0
@@ -250,8 +252,8 @@ struct btrfs_file_extent_item {
 	 * disk space consumed by the extent, checksum blocks are included
 	 * in these numbers
 	 */
-	__le64 disk_blocknr;
-	__le64 disk_num_blocks;
+	__le64 disk_bytenr;
+	__le64 disk_num_bytes;
 	/*
 	 * the logical offset in file blocks (no csums)
 	 * this extent record is for.  This allows a file extent to point
@@ -263,7 +265,7 @@ struct btrfs_file_extent_item {
 	/*
 	 * the logical number of file blocks (no csums included)
 	 */
-	__le64 num_blocks;
+	__le64 num_bytes;
 } __attribute__ ((__packed__));
 
 struct btrfs_csum_item {
@@ -429,6 +431,7 @@ static inline u##bits btrfs_##name(struct extent_buffer *eb,		\
 	int err;							\
 	char *map_token;						\
 	char *kaddr;							\
+	int unmap_on_exit = (eb->map_token == NULL);			\
 	unsigned long map_start;					\
 	unsigned long map_len;						\
 	unsigned long offset = (unsigned long)s +			\
@@ -436,12 +439,13 @@ static inline u##bits btrfs_##name(struct extent_buffer *eb,		\
 	err = map_extent_buffer(eb, offset,				\
 			        sizeof(((type *)0)->member),		\
 				&map_token, &kaddr,			\
-				&map_start, &map_len, KM_USER0);	\
+				&map_start, &map_len, KM_USER1);	\
 	if (!err) {							\
 		__le##bits *tmp = (__le##bits *)(kaddr + offset -	\
 					       map_start);		\
 		u##bits res = le##bits##_to_cpu(*tmp);			\
-		unmap_extent_buffer(eb, map_token, KM_USER0);		\
+		if (unmap_on_exit)					\
+			unmap_extent_buffer(eb, map_token, KM_USER1);	\
 		return res;						\
 	} else {							\
 		__le##bits res;						\
@@ -457,17 +461,19 @@ static inline void btrfs_set_##name(struct extent_buffer *eb,		\
 	char *kaddr;							\
 	unsigned long map_start;					\
 	unsigned long map_len;						\
+	int unmap_on_exit = (eb->map_token == NULL);			\
 	unsigned long offset = (unsigned long)s +			\
 				offsetof(type, member);			\
 	err = map_extent_buffer(eb, offset,				\
 			        sizeof(((type *)0)->member),		\
 				&map_token, &kaddr,			\
-				&map_start, &map_len, KM_USER0);	\
+				&map_start, &map_len, KM_USER1);	\
 	if (!err) {							\
 		__le##bits *tmp = (__le##bits *)(kaddr + offset -	\
 					       map_start);		\
 		*tmp = cpu_to_le##bits(val);				\
-		unmap_extent_buffer(eb, map_token, KM_USER0);		\
+		if (unmap_on_exit)					\
+			unmap_extent_buffer(eb, map_token, KM_USER1);	\
 	} else {							\
 		val = cpu_to_le##bits(val);				\
 		write_eb_member(eb, s, type, member, &val);		\
@@ -483,15 +489,17 @@ static inline u##bits btrfs_##name(struct extent_buffer *eb)		\
 	unsigned long map_start;					\
 	unsigned long map_len;						\
 	unsigned long offset = offsetof(type, member);			\
+	int unmap_on_exit = (eb->map_token == NULL);			\
 	err = map_extent_buffer(eb, offset,				\
 			        sizeof(((type *)0)->member),		\
 				&map_token, &kaddr,			\
-				&map_start, &map_len, KM_USER0);	\
+				&map_start, &map_len, KM_USER1);	\
 	if (!err) {							\
 		__le##bits *tmp = (__le##bits *)(kaddr + offset -	\
 					       map_start);		\
 		u##bits res = le##bits##_to_cpu(*tmp);			\
-		unmap_extent_buffer(eb, map_token, KM_USER0);		\
+		if (unmap_on_exit)					\
+			unmap_extent_buffer(eb, map_token, KM_USER1);	\
 		return res;						\
 	} else {							\
 		__le##bits res;						\
@@ -508,15 +516,17 @@ static inline void btrfs_set_##name(struct extent_buffer *eb,		\
 	unsigned long map_start;					\
 	unsigned long map_len;						\
 	unsigned long offset = offsetof(type, member);			\
+	int unmap_on_exit = (eb->map_token == NULL);			\
 	err = map_extent_buffer(eb, offset,				\
 			        sizeof(((type *)0)->member),		\
 				&map_token, &kaddr,			\
-				&map_start, &map_len, KM_USER0);	\
+				&map_start, &map_len, KM_USER1);	\
 	if (!err) {							\
 		__le##bits *tmp = (__le##bits *)(kaddr + offset -	\
 					       map_start);		\
 		*tmp = cpu_to_le##bits(val);				\
-		unmap_extent_buffer(eb, map_token, KM_USER0);		\
+		if (unmap_on_exit)					\
+			unmap_extent_buffer(eb, map_token, KM_USER1);	\
 	} else {							\
 		val = cpu_to_le##bits(val);				\
 		write_eb_member(eb, NULL, type, member, &val);		\
@@ -769,7 +779,7 @@ static inline void btrfs_set_key_type(struct btrfs_key *key, u8 val)
 }
 
 /* struct btrfs_header */
-BTRFS_SETGET_HEADER_FUNCS(header_blocknr, struct btrfs_header, blocknr, 64);
+BTRFS_SETGET_HEADER_FUNCS(header_bytenr, struct btrfs_header, bytenr, 64);
 BTRFS_SETGET_HEADER_FUNCS(header_generation, struct btrfs_header,
 			  generation, 64);
 BTRFS_SETGET_HEADER_FUNCS(header_owner, struct btrfs_header, owner, 64);
@@ -817,24 +827,28 @@ static inline int btrfs_is_leaf(struct extent_buffer *eb)
 
 /* struct btrfs_root_item */
 BTRFS_SETGET_FUNCS(disk_root_refs, struct btrfs_root_item, refs, 32);
-BTRFS_SETGET_FUNCS(disk_root_blocknr, struct btrfs_root_item, blocknr, 64);
+BTRFS_SETGET_FUNCS(disk_root_bytenr, struct btrfs_root_item, bytenr, 64);
+BTRFS_SETGET_FUNCS(disk_root_level, struct btrfs_root_item, level, 8);
 
-BTRFS_SETGET_STACK_FUNCS(root_blocknr, struct btrfs_root_item, blocknr, 64);
+BTRFS_SETGET_STACK_FUNCS(root_bytenr, struct btrfs_root_item, bytenr, 64);
+BTRFS_SETGET_STACK_FUNCS(root_level, struct btrfs_root_item, level, 8);
 BTRFS_SETGET_STACK_FUNCS(root_dirid, struct btrfs_root_item, root_dirid, 64);
 BTRFS_SETGET_STACK_FUNCS(root_refs, struct btrfs_root_item, refs, 32);
 BTRFS_SETGET_STACK_FUNCS(root_flags, struct btrfs_root_item, flags, 32);
-BTRFS_SETGET_STACK_FUNCS(root_used, struct btrfs_root_item, blocks_used, 64);
-BTRFS_SETGET_STACK_FUNCS(root_limit, struct btrfs_root_item, block_limit, 64);
+BTRFS_SETGET_STACK_FUNCS(root_used, struct btrfs_root_item, bytes_used, 64);
+BTRFS_SETGET_STACK_FUNCS(root_limit, struct btrfs_root_item, byte_limit, 64);
 
 /* struct btrfs_super_block */
-BTRFS_SETGET_STACK_FUNCS(super_blocknr, struct btrfs_super_block, blocknr, 64);
+BTRFS_SETGET_STACK_FUNCS(super_bytenr, struct btrfs_super_block, bytenr, 64);
 BTRFS_SETGET_STACK_FUNCS(super_generation, struct btrfs_super_block,
 			 generation, 64);
 BTRFS_SETGET_STACK_FUNCS(super_root, struct btrfs_super_block, root, 64);
-BTRFS_SETGET_STACK_FUNCS(super_total_blocks, struct btrfs_super_block,
-		   total_blocks, 64);
-BTRFS_SETGET_STACK_FUNCS(super_blocks_used, struct btrfs_super_block,
-		   blocks_used, 64);
+BTRFS_SETGET_STACK_FUNCS(super_root_level, struct btrfs_super_block,
+			 root_level, 8);
+BTRFS_SETGET_STACK_FUNCS(super_total_bytes, struct btrfs_super_block,
+			 total_bytes, 64);
+BTRFS_SETGET_STACK_FUNCS(super_bytes_used, struct btrfs_super_block,
+			 bytes_used, 64);
 BTRFS_SETGET_STACK_FUNCS(super_sectorsize, struct btrfs_super_block,
 			 sectorsize, 32);
 BTRFS_SETGET_STACK_FUNCS(super_nodesize, struct btrfs_super_block,
@@ -856,33 +870,33 @@ static inline unsigned long btrfs_file_extent_inline_start(struct
 						   btrfs_file_extent_item *e)
 {
 	unsigned long offset = (unsigned long)e;
-	offset += offsetof(struct btrfs_file_extent_item, disk_blocknr);
+	offset += offsetof(struct btrfs_file_extent_item, disk_bytenr);
 	return offset;
 }
 
 static inline u32 btrfs_file_extent_calc_inline_size(u32 datasize)
 {
-	return offsetof(struct btrfs_file_extent_item, disk_blocknr) + datasize;
+	return offsetof(struct btrfs_file_extent_item, disk_bytenr) + datasize;
 }
 
 static inline u32 btrfs_file_extent_inline_len(struct extent_buffer *eb,
 					       struct btrfs_item *e)
 {
 	unsigned long offset;
-	offset = offsetof(struct btrfs_file_extent_item, disk_blocknr);
+	offset = offsetof(struct btrfs_file_extent_item, disk_bytenr);
 	return btrfs_item_size(eb, e) - offset;
 }
 
-BTRFS_SETGET_FUNCS(file_extent_disk_blocknr, struct btrfs_file_extent_item,
-		   disk_blocknr, 64);
+BTRFS_SETGET_FUNCS(file_extent_disk_bytenr, struct btrfs_file_extent_item,
+		   disk_bytenr, 64);
 BTRFS_SETGET_FUNCS(file_extent_generation, struct btrfs_file_extent_item,
 		   generation, 64);
-BTRFS_SETGET_FUNCS(file_extent_disk_num_blocks, struct btrfs_file_extent_item,
-		   disk_num_blocks, 64);
+BTRFS_SETGET_FUNCS(file_extent_disk_num_bytes, struct btrfs_file_extent_item,
+		   disk_num_bytes, 64);
 BTRFS_SETGET_FUNCS(file_extent_offset, struct btrfs_file_extent_item,
 		  offset, 64);
-BTRFS_SETGET_FUNCS(file_extent_num_blocks, struct btrfs_file_extent_item,
-		   num_blocks, 64);
+BTRFS_SETGET_FUNCS(file_extent_num_bytes, struct btrfs_file_extent_item,
+		   num_bytes, 64);
 
 static inline struct btrfs_root *btrfs_sb(struct super_block *sb)
 {
@@ -906,6 +920,12 @@ static inline int btrfs_set_root_name(struct btrfs_root *root,
 	return 0;
 }
 
+static inline u32 btrfs_level_size(struct btrfs_root *root, int level) {
+	if (level == 0)
+		return root->leafsize;
+	return root->nodesize;
+}
+
 /* helper function to cast into the data area of the leaf. */
 #define btrfs_item_ptr(leaf, slot, type) \
 	((type *)(btrfs_leaf_data(leaf) + \
@@ -927,7 +947,7 @@ int btrfs_extent_post_op(struct btrfs_trans_handle *trans,
 int btrfs_copy_pinned(struct btrfs_root *root, struct extent_map_tree *copy);
 struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
 							 btrfs_fs_info *info,
-							 u64 blocknr);
+							 u64 bytenr);
 struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
 						 struct btrfs_block_group_cache
 						 *hint, u64 search_start,
@@ -935,22 +955,22 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
 int btrfs_inc_root_ref(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root);
 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
-					    struct btrfs_root *root, u64 hint,
-					    u64 empty_size);
+					    struct btrfs_root *root, u32 size,
+					    u64 hint, u64 empty_size);
 int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root, u64 owner,
-		       u64 num_blocks, u64 empty_size, u64 search_start,
+		       u64 num_bytes, u64 empty_size, u64 search_start,
 		       u64 search_end, struct btrfs_key *ins, int data);
 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		  struct extent_buffer *buf);
 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, u64 blocknr, u64 num_blocks, int pin);
+		      *root, u64 bytenr, u64 num_bytes, int pin);
 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
 			       struct btrfs_root *root,
 			       struct extent_map_tree *unpin);
 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root,
-				u64 blocknr, u64 num_blocks);
+				u64 bytenr, u64 num_bytes);
 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
 				    struct btrfs_root *root);
 int btrfs_free_block_groups(struct btrfs_fs_info *info);
@@ -1040,12 +1060,12 @@ int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root
 int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 			       struct btrfs_root *root,
 			       u64 objectid, u64 pos, u64 offset,
-			       u64 disk_num_blocks,
-			       u64 num_blocks);
+			       u64 disk_num_bytes,
+			       u64 num_bytes);
 int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root,
 			     struct btrfs_path *path, u64 objectid,
-			     u64 blocknr, int mod);
+			     u64 bytenr, int mod);
 int btrfs_csum_file_block(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root,
 			  u64 objectid, u64 offset,

fs/btrfs/disk-io.c

@@ -28,6 +28,7 @@
 #include "disk-io.h"
 #include "transaction.h"
 #include "btrfs_inode.h"
+#include "print-tree.h"
 
 #if 0
 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
@@ -43,26 +44,25 @@ static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
 #endif
 
 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
-					    u64 blocknr)
+					    u64 bytenr, u32 blocksize)
 {
 	struct inode *btree_inode = root->fs_info->btree_inode;
 	struct extent_buffer *eb;
 	eb = find_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
-				   blocknr * root->sectorsize,
-				   root->sectorsize, GFP_NOFS);
+				bytenr, blocksize, GFP_NOFS);
 	if (eb)
 		eb->alloc_addr = (unsigned long)__builtin_return_address(0);
 	return eb;
 }
 
 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
-						 u64 blocknr)
+						 u64 bytenr, u32 blocksize)
 {
 	struct inode *btree_inode = root->fs_info->btree_inode;
 	struct extent_buffer *eb;
+
 	eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
-				   blocknr * root->sectorsize,
-				   root->sectorsize, GFP_NOFS);
+				 bytenr, blocksize, GFP_NOFS);
 	eb->alloc_addr = (unsigned long)__builtin_return_address(0);
 	return eb;
 }
@@ -208,13 +208,13 @@ static struct address_space_operations btree_aops = {
 	.sync_page	= block_sync_page,
 };
 
-int readahead_tree_block(struct btrfs_root *root, u64 blocknr)
+int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
 {
 	struct extent_buffer *buf = NULL;
 	struct inode *btree_inode = root->fs_info->btree_inode;
 	int ret = 0;
 
-	buf = btrfs_find_create_tree_block(root, blocknr);
+	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
 	if (!buf)
 		return 0;
 	read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
@@ -223,12 +223,13 @@ int readahead_tree_block(struct btrfs_root *root, u64 blocknr)
 	return ret;
 }
 
-struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr)
+struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
+				      u32 blocksize)
 {
 	struct extent_buffer *buf = NULL;
 	struct inode *btree_inode = root->fs_info->btree_inode;
 
-	buf = btrfs_find_create_tree_block(root, blocknr);
+	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
 	if (!buf)
 		return NULL;
 	read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
@@ -261,7 +262,7 @@ int set_tree_block_dirty(struct btrfs_root *root, struct extent_buffer *buf)
 	return 0;
 }
 
-static int __setup_root(int blocksize,
+static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 			struct btrfs_root *root,
 			struct btrfs_fs_info *fs_info,
 			u64 objectid)
@@ -269,9 +270,9 @@ static int __setup_root(int blocksize,
 	root->node = NULL;
 	root->inode = NULL;
 	root->commit_root = NULL;
-	root->sectorsize = blocksize;
-	root->nodesize = blocksize;
-	root->leafsize = blocksize;
+	root->sectorsize = sectorsize;
+	root->nodesize = nodesize;
+	root->leafsize = leafsize;
 	root->ref_cows = 0;
 	root->fs_info = fs_info;
 	root->objectid = objectid;
@@ -291,21 +292,23 @@ static int __setup_root(int blocksize,
 	return 0;
 }
 
-static int find_and_setup_root(int blocksize,
-			       struct btrfs_root *tree_root,
+static int find_and_setup_root(struct btrfs_root *tree_root,
 			       struct btrfs_fs_info *fs_info,
 			       u64 objectid,
 			       struct btrfs_root *root)
 {
 	int ret;
+	u32 blocksize;
 
-	__setup_root(blocksize, root, fs_info, objectid);
+	__setup_root(tree_root->nodesize, tree_root->leafsize,
+		     tree_root->sectorsize, root, fs_info, objectid);
 	ret = btrfs_find_last_root(tree_root, objectid,
 				   &root->root_item, &root->root_key);
 	BUG_ON(ret);
 
-	root->node = read_tree_block(root,
-				     btrfs_root_blocknr(&root->root_item));
+	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
+	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
+				     blocksize);
 	BUG_ON(!root->node);
 	return 0;
 }
@@ -318,14 +321,14 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
 	struct btrfs_path *path;
 	struct extent_buffer *l;
 	u64 highest_inode;
+	u32 blocksize;
 	int ret = 0;
 
 	root = kzalloc(sizeof(*root), GFP_NOFS);
 	if (!root)
 		return ERR_PTR(-ENOMEM);
 	if (location->offset == (u64)-1) {
-		ret = find_and_setup_root(fs_info->sb->s_blocksize,
-					  fs_info->tree_root, fs_info,
+		ret = find_and_setup_root(tree_root, fs_info,
 					  location->objectid, root);
 		if (ret) {
 			kfree(root);
@@ -334,7 +337,8 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
 		goto insert;
 	}
 
-	__setup_root(fs_info->sb->s_blocksize, root, fs_info,
+	__setup_root(tree_root->nodesize, tree_root->leafsize,
+		     tree_root->sectorsize, root, fs_info,
 		     location->objectid);
 
 	path = btrfs_alloc_path();
@@ -357,8 +361,9 @@ out:
 		kfree(root);
 		return ERR_PTR(ret);
 	}
-	root->node = read_tree_block(root,
-				     btrfs_root_blocknr(&root->root_item));
+	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
+	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
+				     blocksize);
 	BUG_ON(!root->node);
 insert:
 	root->ref_cows = 1;
@@ -418,6 +423,10 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 
 struct btrfs_root *open_ctree(struct super_block *sb)
 {
+	u32 sectorsize;
+	u32 nodesize;
+	u32 leafsize;
+	u32 blocksize;
 	struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
 						 GFP_NOFS);
 	struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
@@ -474,12 +483,12 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	mutex_init(&fs_info->trans_mutex);
 	mutex_init(&fs_info->fs_mutex);
 
-	__setup_root(sb->s_blocksize, tree_root,
+	__setup_root(512, 512, 512, tree_root,
 		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
 
 	fs_info->sb_buffer = read_tree_block(tree_root,
-					     BTRFS_SUPER_INFO_OFFSET /
-					     sb->s_blocksize);
+					     BTRFS_SUPER_INFO_OFFSET,
+					     512);
 
 	if (!fs_info->sb_buffer)
 		goto fail_iput;
@@ -494,9 +503,15 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	if (!btrfs_super_root(disk_super))
 		goto fail_sb_buffer;
 
+	nodesize = btrfs_super_nodesize(disk_super);
+	leafsize = btrfs_super_leafsize(disk_super);
+	sectorsize = btrfs_super_sectorsize(disk_super);
+	tree_root->nodesize = nodesize;
+	tree_root->leafsize = leafsize;
+	tree_root->sectorsize = sectorsize;
+
 	i_size_write(fs_info->btree_inode,
-		     btrfs_super_total_blocks(disk_super) <<
-		     fs_info->btree_inode->i_blkbits);
+		     btrfs_super_total_bytes(disk_super));
 
 
 	if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
@@ -504,13 +519,22 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 		printk("btrfs: valid FS not found on %s\n", sb->s_id);
 		goto fail_sb_buffer;
 	}
+	blocksize = btrfs_level_size(tree_root,
+				     btrfs_super_root_level(disk_super));
 	tree_root->node = read_tree_block(tree_root,
-					  btrfs_super_root(disk_super));
+					  btrfs_super_root(disk_super),
+					  blocksize);
 	if (!tree_root->node)
 		goto fail_sb_buffer;
 
+#if 0
+	btrfs_print_leaf(tree_root, tree_root->node);
+	err = -EIO;
+	goto fail_tree_root;
+#endif
 	mutex_lock(&fs_info->fs_mutex);
-	ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
+
+	ret = find_and_setup_root(tree_root, fs_info,
 				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
 	if (ret) {
 		mutex_unlock(&fs_info->fs_mutex);
@@ -611,11 +635,11 @@ int close_ctree(struct btrfs_root *root)
 		free_extent_buffer(fs_info->tree_root->node);
 
 	free_extent_buffer(fs_info->sb_buffer);
-	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
-	iput(fs_info->btree_inode);
 
 	btrfs_free_block_groups(root->fs_info);
 	del_fs_roots(fs_info);
+	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
+	iput(fs_info->btree_inode);
 	kfree(fs_info->extent_root);
 	kfree(fs_info->tree_root);
 	return 0;
@@ -642,7 +666,7 @@ void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
 
 	if (transid != root->fs_info->generation) {
 		printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
-			(unsigned long long)extent_buffer_blocknr(buf),
+			(unsigned long long)buf->start,
 			transid, root->fs_info->generation);
 		WARN_ON(1);
 	}

fs/btrfs/disk-io.h

@@ -21,10 +21,11 @@
 
 #define BTRFS_SUPER_INFO_OFFSET (16 * 1024)
 
-struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr);
-int readahead_tree_block(struct btrfs_root *root, u64 blocknr);
+struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
+				      u32 blocksize);
+int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize);
 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
-						   u64 blocknr);
+						   u64 bytenr, u32 blocksize);
 int clean_tree_block(struct btrfs_trans_handle *trans,
 		     struct btrfs_root *root, struct extent_buffer *buf);
 struct btrfs_root *open_ctree(struct super_block *sb);
@@ -32,7 +33,7 @@ int close_ctree(struct btrfs_root *root);
 int write_ctree_super(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root);
 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
-					    u64 blocknr);
+					    u64 bytenr, u32 blocksize);
 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 				      struct btrfs_key *location,
 				      const char *name, int namelen);

fs/btrfs/extent-tree.c

@@ -129,7 +129,7 @@ err:
 
 struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
 							 btrfs_fs_info *info,
-							 u64 blocknr)
+							 u64 bytenr)
 {
 	struct extent_map_tree *block_group_cache;
 	struct btrfs_block_group_cache *block_group = NULL;
@@ -140,7 +140,7 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
 
 	block_group_cache = &info->block_group_cache;
 	ret = find_first_extent_bit(block_group_cache,
-				    blocknr, &start, &end,
+				    bytenr, &start, &end,
 				    BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA);
 	if (ret) {
 		return NULL;
@@ -152,7 +152,7 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
 	block_group = (struct btrfs_block_group_cache *)ptr;
 
 
-	if (block_group->key.objectid <= blocknr && blocknr <=
+	if (block_group->key.objectid <= bytenr && bytenr <=
 	    block_group->key.objectid + block_group->key.offset)
 		return block_group;
 
@@ -315,7 +315,7 @@ found:
 
 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root,
-				u64 blocknr, u64 num_blocks)
+				u64 bytenr, u64 num_bytes)
 {
 	struct btrfs_path *path;
 	int ret;
@@ -324,13 +324,14 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 	struct btrfs_extent_item *item;
 	u32 refs;
 
+	WARN_ON(num_bytes < root->sectorsize);
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
 
-	key.objectid = blocknr;
+	key.objectid = bytenr;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
-	key.offset = num_blocks;
+	key.offset = num_bytes;
 	ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
 				0, 1);
 	if (ret < 0)
@@ -361,8 +362,8 @@ int btrfs_extent_post_op(struct btrfs_trans_handle *trans,
 }
 
 static int lookup_extent_ref(struct btrfs_trans_handle *trans,
-			     struct btrfs_root *root, u64 blocknr,
-			     u64 num_blocks, u32 *refs)
+			     struct btrfs_root *root, u64 bytenr,
+			     u64 num_bytes, u32 *refs)
 {
 	struct btrfs_path *path;
 	int ret;
@@ -370,9 +371,10 @@ static int lookup_extent_ref(struct btrfs_trans_handle *trans,
 	struct extent_buffer *l;
 	struct btrfs_extent_item *item;
 
+	WARN_ON(num_bytes < root->sectorsize);
 	path = btrfs_alloc_path();
-	key.objectid = blocknr;
-	key.offset = num_blocks;
+	key.objectid = bytenr;
+	key.offset = num_bytes;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 	ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
 				0, 0);
@@ -380,7 +382,7 @@ static int lookup_extent_ref(struct btrfs_trans_handle *trans,
 		goto out;
 	if (ret != 0) {
 		btrfs_print_leaf(root, path->nodes[0]);
-		printk("failed to find block number %Lu\n", blocknr);
+		printk("failed to find block number %Lu\n", bytenr);
 		BUG();
 	}
 	l = path->nodes[0];
@@ -394,19 +396,19 @@ out:
 int btrfs_inc_root_ref(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root)
 {
-	return btrfs_inc_extent_ref(trans, root,
-				    extent_buffer_blocknr(root->node), 1);
+	return btrfs_inc_extent_ref(trans, root, root->node->start,
+				    root->node->len);
 }
 
 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		  struct extent_buffer *buf)
 {
-	u64 blocknr;
+	u64 bytenr;
 	u32 nritems;
 	struct btrfs_key key;
 	struct btrfs_file_extent_item *fi;
 	int i;
-	int leaf;
+	int level;
 	int ret;
 	int faili;
 	int err;
@@ -414,11 +416,11 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	if (!root->ref_cows)
 		return 0;
 
-	leaf = btrfs_is_leaf(buf);
+	level = btrfs_header_level(buf);
 	nritems = btrfs_header_nritems(buf);
 	for (i = 0; i < nritems; i++) {
-		if (leaf) {
-			u64 disk_blocknr;
+		if (level == 0) {
+			u64 disk_bytenr;
 			btrfs_item_key_to_cpu(buf, &key, i);
 			if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
 				continue;
@@ -427,18 +429,19 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			if (btrfs_file_extent_type(buf, fi) ==
 			    BTRFS_FILE_EXTENT_INLINE)
 				continue;
-			disk_blocknr = btrfs_file_extent_disk_blocknr(buf, fi);
-			if (disk_blocknr == 0)
+			disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
+			if (disk_bytenr == 0)
 				continue;
-			ret = btrfs_inc_extent_ref(trans, root, disk_blocknr,
-				    btrfs_file_extent_disk_num_blocks(buf, fi));
+			ret = btrfs_inc_extent_ref(trans, root, disk_bytenr,
+				    btrfs_file_extent_disk_num_bytes(buf, fi));
 			if (ret) {
 				faili = i;
 				goto fail;
 			}
 		} else {
-			blocknr = btrfs_node_blockptr(buf, i);
-			ret = btrfs_inc_extent_ref(trans, root, blocknr, 1);
+			bytenr = btrfs_node_blockptr(buf, i);
+			ret = btrfs_inc_extent_ref(trans, root, bytenr,
+					   btrfs_level_size(root, level - 1));
 			if (ret) {
 				faili = i;
 				goto fail;
@@ -449,8 +452,8 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 fail:
 	WARN_ON(1);
 	for (i =0; i < faili; i++) {
-		if (leaf) {
-			u64 disk_blocknr;
+		if (level == 0) {
+			u64 disk_bytenr;
 			btrfs_item_key_to_cpu(buf, &key, i);
 			if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
 				continue;
@@ -459,16 +462,17 @@ fail:
 			if (btrfs_file_extent_type(buf, fi) ==
 			    BTRFS_FILE_EXTENT_INLINE)
 				continue;
-			disk_blocknr = btrfs_file_extent_disk_blocknr(buf, fi);
-			if (disk_blocknr == 0)
+			disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
+			if (disk_bytenr == 0)
 				continue;
-			err = btrfs_free_extent(trans, root, disk_blocknr,
-				    btrfs_file_extent_disk_num_blocks(buf,
+			err = btrfs_free_extent(trans, root, disk_bytenr,
+				    btrfs_file_extent_disk_num_bytes(buf,
 								      fi), 0);
 			BUG_ON(err);
 		} else {
-			blocknr = btrfs_node_blockptr(buf, i);
-			err = btrfs_free_extent(trans, root, blocknr, 1, 0);
+			bytenr = btrfs_node_blockptr(buf, i);
+			err = btrfs_free_extent(trans, root, bytenr,
+					btrfs_level_size(root, level - 1), 0);
 			BUG_ON(err);
 		}
 	}
@@ -558,31 +562,31 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
 
 static int update_block_group(struct btrfs_trans_handle *trans,
 			      struct btrfs_root *root,
-			      u64 blocknr, u64 num, int alloc, int mark_free,
-			      int data)
+			      u64 bytenr, u64 num_bytes, int alloc,
+			      int mark_free, int data)
 {
 	struct btrfs_block_group_cache *cache;
 	struct btrfs_fs_info *info = root->fs_info;
-	u64 total = num;
+	u64 total = num_bytes;
 	u64 old_val;
-	u64 block_in_group;
+	u64 byte_in_group;
 	u64 start;
 	u64 end;
 
 	while(total) {
-		cache = btrfs_lookup_block_group(info, blocknr);
+		cache = btrfs_lookup_block_group(info, bytenr);
 		if (!cache) {
 			return -1;
 		}
-		block_in_group = blocknr - cache->key.objectid;
-		WARN_ON(block_in_group > cache->key.offset);
+		byte_in_group = bytenr - cache->key.objectid;
+		WARN_ON(byte_in_group > cache->key.offset);
 		start = cache->key.objectid;
 		end = start + cache->key.offset - 1;
 		set_extent_bits(&info->block_group_cache, start, end,
 				BLOCK_GROUP_DIRTY, GFP_NOFS);
 
 		old_val = btrfs_block_group_used(&cache->item);
-		num = min(total, cache->key.offset - block_in_group);
+		num_bytes = min(total, cache->key.offset - byte_in_group);
 		if (alloc) {
 			if (cache->data != data &&
 			    old_val < (cache->key.offset >> 1)) {
@@ -608,18 +612,18 @@ static int update_block_group(struct btrfs_trans_handle *trans,
 						start, end, bit_to_set,
 						GFP_NOFS);
 			}
-			old_val += num;
+			old_val += num_bytes;
 		} else {
-			old_val -= num;
+			old_val -= num_bytes;
 			if (mark_free) {
 				set_extent_dirty(&info->free_space_cache,
-						 blocknr, blocknr + num - 1,
+						 bytenr, bytenr + num_bytes - 1,
 						 GFP_NOFS);
 			}
 		}
 		btrfs_set_block_group_used(&cache->item, old_val);
-		total -= num;
-		blocknr += num;
+		total -= num_bytes;
+		bytenr += num_bytes;
 	}
 	return 0;
 }
@@ -701,13 +705,14 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct
 	return 0;
 }
 
-static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending)
+static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes,
+			  int pending)
 {
 	int err = 0;
 	struct extent_buffer *buf;
 
 	if (!pending) {
-		buf = btrfs_find_tree_block(root, blocknr);
+		buf = btrfs_find_tree_block(root, bytenr, num_bytes);
 		if (buf) {
 			if (btrfs_buffer_uptodate(buf)) {
 				u64 transid =
@@ -720,10 +725,11 @@ static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending)
 			free_extent_buffer(buf);
 		}
 		set_extent_dirty(&root->fs_info->pinned_extents,
-				 blocknr, blocknr, GFP_NOFS);
+				 bytenr, bytenr + num_bytes - 1, GFP_NOFS);
 	} else {
 		set_extent_bits(&root->fs_info->pending_del,
-				  blocknr, blocknr, EXTENT_LOCKED, GFP_NOFS);
+				bytenr, bytenr + num_bytes - 1,
+				EXTENT_LOCKED, GFP_NOFS);
 	}
 	BUG_ON(err < 0);
 	return 0;
@@ -733,7 +739,7 @@ static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending)
  * remove an extent from the root, returns 0 on success
  */
 static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
-			 *root, u64 blocknr, u64 num_blocks, int pin,
+			 *root, u64 bytenr, u64 num_bytes, int pin,
 			 int mark_free)
 {
 	struct btrfs_path *path;
@@ -745,9 +751,9 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	struct btrfs_extent_item *ei;
 	u32 refs;
 
-	key.objectid = blocknr;
+	key.objectid = bytenr;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
-	key.offset = num_blocks;
+	key.offset = num_bytes;
 
 	path = btrfs_alloc_path();
 	if (!path)
@@ -768,28 +774,29 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_mark_buffer_dirty(leaf);
 
 	if (refs == 0) {
-		u64 super_blocks_used, root_blocks_used;
+		u64 super_used;
+		u64 root_used;
 
 		if (pin) {
-			ret = pin_down_block(root, blocknr, 0);
+			ret = pin_down_bytes(root, bytenr, num_bytes, 0);
 			BUG_ON(ret);
 		}
 
 		/* block accounting for super block */
-		super_blocks_used = btrfs_super_blocks_used(&info->super_copy);
-		btrfs_set_super_blocks_used(&info->super_copy,
-					    super_blocks_used - num_blocks);
+		super_used = btrfs_super_bytes_used(&info->super_copy);
+		btrfs_set_super_bytes_used(&info->super_copy,
+					   super_used - num_bytes);
 
 		/* block accounting for root item */
-		root_blocks_used = btrfs_root_used(&root->root_item);
+		root_used = btrfs_root_used(&root->root_item);
 		btrfs_set_root_used(&root->root_item,
-					   root_blocks_used - num_blocks);
+					   root_used - num_bytes);
 
 		ret = btrfs_del_item(trans, extent_root, path);
 		if (ret) {
 			return ret;
 		}
-		ret = update_block_group(trans, root, blocknr, num_blocks, 0,
+		ret = update_block_group(trans, root, bytenr, num_bytes, 0,
 					 mark_free, 0);
 		BUG_ON(ret);
 	}
@@ -836,17 +843,18 @@ static int del_pending_extents(struct btrfs_trans_handle *trans, struct
  * remove an extent from the root, returns 0 on success
  */
 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, u64 blocknr, u64 num_blocks, int pin)
+		      *root, u64 bytenr, u64 num_bytes, int pin)
 {
 	struct btrfs_root *extent_root = root->fs_info->extent_root;
 	int pending_ret;
 	int ret;
 
+	WARN_ON(num_bytes < root->sectorsize);
 	if (root == extent_root) {
-		pin_down_block(root, blocknr, 1);
+		pin_down_bytes(root, bytenr, num_bytes, 1);
 		return 0;
 	}
-	ret = __free_extent(trans, root, blocknr, num_blocks, pin, pin == 0);
+	ret = __free_extent(trans, root, bytenr, num_bytes, pin, pin == 0);
 	pending_ret = del_pending_extents(trans, root->fs_info->extent_root);
 	return ret ? ret : pending_ret;
 }
@@ -860,8 +868,8 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
  * Any available blocks before search_start are skipped.
  */
 static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
-			    *orig_root, u64 num_blocks, u64 empty_size,
-			    u64 search_start, u64 search_end, u64 hint_block,
+			    *orig_root, u64 num_bytes, u64 empty_size,
+			    u64 search_start, u64 search_end, u64 hint_byte,
 			    struct btrfs_key *ins, u64 exclude_start,
 			    u64 exclude_nr, int data)
 {
@@ -870,30 +878,29 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	int ret;
 	u64 hole_size = 0;
 	int slot = 0;
-	u64 last_block = 0;
+	u64 last_byte = 0;
 	u64 orig_search_start = search_start;
 	int start_found;
 	struct extent_buffer *l;
 	struct btrfs_root * root = orig_root->fs_info->extent_root;
 	struct btrfs_fs_info *info = root->fs_info;
-	int total_needed = num_blocks;
+	u64 total_needed = num_bytes;
 	int level;
 	struct btrfs_block_group_cache *block_group;
 	int full_scan = 0;
 	int wrapped = 0;
-	u64 cached_search_start = 0;
 
-	WARN_ON(num_blocks < 1);
+	WARN_ON(num_bytes < root->sectorsize);
 	btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
 
 	level = btrfs_header_level(root->node);
 
 	if (search_end == (u64)-1)
-		search_end = btrfs_super_total_blocks(&info->super_copy);
-	if (hint_block) {
-		block_group = btrfs_lookup_block_group(info, hint_block);
+		search_end = btrfs_super_total_bytes(&info->super_copy);
+	if (hint_byte) {
+		block_group = btrfs_lookup_block_group(info, hint_byte);
 		block_group = btrfs_find_block_group(root, block_group,
-						     hint_block, data, 1);
+						     hint_byte, data, 1);
 	} else {
 		block_group = btrfs_find_block_group(root,
 						     trans->block_group, 0,
@@ -906,7 +913,6 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 check_failed:
 	search_start = find_search_start(root, &block_group,
 					 search_start, total_needed, data);
-	cached_search_start = search_start;
 
 	btrfs_init_path(path);
 	ins->objectid = search_start;
@@ -958,27 +964,27 @@ check_failed:
 				start_found = 1;
 				goto check_pending;
 			}
-			ins->objectid = last_block > search_start ?
-					last_block : search_start;
+			ins->objectid = last_byte > search_start ?
+					last_byte : search_start;
 			ins->offset = search_end - ins->objectid;
 			goto check_pending;
 		}
 		btrfs_item_key_to_cpu(l, &key, slot);
 
-		if (key.objectid >= search_start && key.objectid > last_block &&
+		if (key.objectid >= search_start && key.objectid > last_byte &&
 		    start_found) {
-			if (last_block < search_start)
-				last_block = search_start;
-			hole_size = key.objectid - last_block;
-			if (hole_size >= num_blocks) {
-				ins->objectid = last_block;
+			if (last_byte < search_start)
+				last_byte = search_start;
+			hole_size = key.objectid - last_byte;
+			if (hole_size >= num_bytes) {
+				ins->objectid = last_byte;
 				ins->offset = hole_size;
 				goto check_pending;
 			}
 		}
 		if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY) {
 			if (!start_found) {
-				last_block = key.objectid;
+				last_byte = key.objectid;
 				start_found = 1;
 			}
 			goto next;
@@ -986,9 +992,9 @@ check_failed:
 
 
 		start_found = 1;
-		last_block = key.objectid + key.offset;
+		last_byte = key.objectid + key.offset;
 
-		if (!full_scan && last_block >= block_group->key.objectid +
+		if (!full_scan && last_byte >= block_group->key.objectid +
 		    block_group->key.offset) {
 			btrfs_release_path(root, path);
 			search_start = block_group->key.objectid +
@@ -1006,20 +1012,20 @@ check_pending:
 	btrfs_release_path(root, path);
 	BUG_ON(ins->objectid < search_start);
 
-	if (ins->objectid + num_blocks >= search_end)
+	if (ins->objectid + num_bytes >= search_end)
 		goto enospc;
 
 	if (test_range_bit(&info->extent_ins, ins->objectid,
-			   ins->objectid + num_blocks -1, EXTENT_LOCKED, 0)) {
-		search_start = ins->objectid + num_blocks;
+			   ins->objectid + num_bytes -1, EXTENT_LOCKED, 0)) {
+		search_start = ins->objectid + num_bytes;
 		goto new_group;
 	}
 	if (test_range_bit(&info->pinned_extents, ins->objectid,
-			   ins->objectid + num_blocks -1, EXTENT_DIRTY, 0)) {
-		search_start = ins->objectid + num_blocks;
+			   ins->objectid + num_bytes -1, EXTENT_DIRTY, 0)) {
+		search_start = ins->objectid + num_bytes;
 		goto new_group;
 	}
-	if (exclude_nr > 0 && (ins->objectid + num_blocks > exclude_start &&
+	if (exclude_nr > 0 && (ins->objectid + num_bytes > exclude_start &&
 	    ins->objectid < exclude_start + exclude_nr)) {
 		search_start = exclude_start + exclude_nr;
 		goto new_group;
@@ -1029,12 +1035,12 @@ check_pending:
 		if (block_group)
 			trans->block_group = block_group;
 	}
-	ins->offset = num_blocks;
+	ins->offset = num_bytes;
 	btrfs_free_path(path);
 	return 0;
 
 new_group:
-	if (search_start + num_blocks >= search_end) {
+	if (search_start + num_bytes >= search_end) {
 enospc:
 		search_start = orig_search_start;
 		if (full_scan) {
@@ -1069,12 +1075,12 @@ error:
  */
 int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root, u64 owner,
-		       u64 num_blocks, u64 empty_size, u64 hint_block,
+		       u64 num_bytes, u64 empty_size, u64 hint_byte,
 		       u64 search_end, struct btrfs_key *ins, int data)
 {
 	int ret;
 	int pending_ret;
-	u64 super_blocks_used, root_blocks_used;
+	u64 super_used, root_used;
 	u64 search_start = 0;
 	struct btrfs_fs_info *info = root->fs_info;
 	struct btrfs_root *extent_root = info->extent_root;
@@ -1083,9 +1089,9 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 	btrfs_set_stack_extent_refs(&extent_item, 1);
 	btrfs_set_stack_extent_owner(&extent_item, owner);
 
-	WARN_ON(num_blocks < 1);
-	ret = find_free_extent(trans, root, num_blocks, empty_size,
-			       search_start, search_end, hint_block, ins,
+	WARN_ON(num_bytes < root->sectorsize);
+	ret = find_free_extent(trans, root, num_bytes, empty_size,
+			       search_start, search_end, hint_byte, ins,
 			       trans->alloc_exclude_start,
 			       trans->alloc_exclude_nr, data);
 	BUG_ON(ret);
@@ -1093,21 +1099,18 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 		return ret;
 
 	/* block accounting for super block */
-	super_blocks_used = btrfs_super_blocks_used(&info->super_copy);
-	btrfs_set_super_blocks_used(&info->super_copy, super_blocks_used +
-				    num_blocks);
+	super_used = btrfs_super_bytes_used(&info->super_copy);
+	btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes);
 
 	/* block accounting for root item */
-	root_blocks_used = btrfs_root_used(&root->root_item);
-	btrfs_set_root_used(&root->root_item, root_blocks_used +
-				   num_blocks);
+	root_used = btrfs_root_used(&root->root_item);
+	btrfs_set_root_used(&root->root_item, root_used + num_bytes);
 
 	clear_extent_dirty(&root->fs_info->free_space_cache,
 			   ins->objectid, ins->objectid + ins->offset - 1,
 			   GFP_NOFS);
 
 	if (root == extent_root) {
-		BUG_ON(num_blocks != 1);
 		set_extent_bits(&root->fs_info->extent_ins, ins->objectid,
 				ins->objectid + ins->offset - 1,
 				EXTENT_LOCKED, GFP_NOFS);
@@ -1146,7 +1149,8 @@ update_block:
  * returns the tree buffer or NULL.
  */
 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
-					     struct btrfs_root *root, u64 hint,
+					     struct btrfs_root *root,
+					     u32 blocksize, u64 hint,
 					     u64 empty_size)
 {
 	struct btrfs_key ins;
@@ -1154,14 +1158,15 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 	struct extent_buffer *buf;
 
 	ret = btrfs_alloc_extent(trans, root, root->root_key.objectid,
-				 1, empty_size, hint, (u64)-1, &ins, 0);
+				 blocksize, empty_size, hint,
+				 (u64)-1, &ins, 0);
 	if (ret) {
 		BUG_ON(ret > 0);
 		return ERR_PTR(ret);
 	}
-	buf = btrfs_find_create_tree_block(root, ins.objectid);
+	buf = btrfs_find_create_tree_block(root, ins.objectid, blocksize);
 	if (!buf) {
-		btrfs_free_extent(trans, root, ins.objectid, 1, 0);
+		btrfs_free_extent(trans, root, ins.objectid, blocksize, 0);
 		return ERR_PTR(-ENOMEM);
 	}
 	btrfs_set_buffer_uptodate(buf);
@@ -1191,7 +1196,7 @@ static int drop_leaf_ref(struct btrfs_trans_handle *trans,
 	BUG_ON(!btrfs_is_leaf(leaf));
 	nritems = btrfs_header_nritems(leaf);
 	for (i = 0; i < nritems; i++) {
-		u64 disk_blocknr;
+		u64 disk_bytenr;
 
 		btrfs_item_key_to_cpu(leaf, &key, i);
 		if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
@@ -1204,11 +1209,11 @@ static int drop_leaf_ref(struct btrfs_trans_handle *trans,
 		 * FIXME make sure to insert a trans record that
 		 * repeats the snapshot del on crash
 		 */
-		disk_blocknr = btrfs_file_extent_disk_blocknr(leaf, fi);
-		if (disk_blocknr == 0)
+		disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+		if (disk_bytenr == 0)
 			continue;
-		ret = btrfs_free_extent(trans, root, disk_blocknr,
-				btrfs_file_extent_disk_num_blocks(leaf, fi), 0);
+		ret = btrfs_free_extent(trans, root, disk_bytenr,
+				btrfs_file_extent_disk_num_bytes(leaf, fi), 0);
 		BUG_ON(ret);
 	}
 	return 0;
@@ -1219,19 +1224,23 @@ static void reada_walk_down(struct btrfs_root *root,
 {
 	int i;
 	u32 nritems;
-	u64 blocknr;
+	u64 bytenr;
 	int ret;
 	u32 refs;
+	int level;
+	u32 blocksize;
 
 	nritems = btrfs_header_nritems(node);
+	level = btrfs_header_level(node);
 	for (i = 0; i < nritems; i++) {
-		blocknr = btrfs_node_blockptr(node, i);
-		ret = lookup_extent_ref(NULL, root, blocknr, 1, &refs);
+		bytenr = btrfs_node_blockptr(node, i);
+		blocksize = btrfs_level_size(root, level - 1);
+		ret = lookup_extent_ref(NULL, root, bytenr, blocksize, &refs);
 		BUG_ON(ret);
 		if (refs != 1)
 			continue;
 		mutex_unlock(&root->fs_info->fs_mutex);
-		ret = readahead_tree_block(root, blocknr);
+		ret = readahead_tree_block(root, bytenr, blocksize);
 		cond_resched();
 		mutex_lock(&root->fs_info->fs_mutex);
 		if (ret)
@@ -1248,15 +1257,16 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 {
 	struct extent_buffer *next;
 	struct extent_buffer *cur;
-	u64 blocknr;
+	u64 bytenr;
+	u32 blocksize;
 	int ret;
 	u32 refs;
 
 	WARN_ON(*level < 0);
 	WARN_ON(*level >= BTRFS_MAX_LEVEL);
 	ret = lookup_extent_ref(trans, root,
-				extent_buffer_blocknr(path->nodes[*level]),
-				1, &refs);
+				path->nodes[*level]->start,
+				path->nodes[*level]->len, &refs);
 	BUG_ON(ret);
 	if (refs > 1)
 		goto out;
@@ -1283,30 +1293,33 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 			BUG_ON(ret);
 			break;
 		}
-		blocknr = btrfs_node_blockptr(cur, path->slots[*level]);
-		ret = lookup_extent_ref(trans, root, blocknr, 1, &refs);
+		bytenr = btrfs_node_blockptr(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) {
 			path->slots[*level]++;
-			ret = btrfs_free_extent(trans, root, blocknr, 1, 1);
+			ret = btrfs_free_extent(trans, root, bytenr,
+						blocksize, 1);
 			BUG_ON(ret);
 			continue;
 		}
-		next = btrfs_find_tree_block(root, blocknr);
+		next = btrfs_find_tree_block(root, bytenr, blocksize);
 		if (!next || !btrfs_buffer_uptodate(next)) {
 			free_extent_buffer(next);
 			mutex_unlock(&root->fs_info->fs_mutex);
-			next = read_tree_block(root, blocknr);
+			next = read_tree_block(root, bytenr, blocksize);
 			mutex_lock(&root->fs_info->fs_mutex);
 
 			/* we dropped the lock, check one more time */
-			ret = lookup_extent_ref(trans, root, blocknr, 1, &refs);
+			ret = lookup_extent_ref(trans, root, bytenr,
+						blocksize, &refs);
 			BUG_ON(ret);
 			if (refs != 1) {
 				path->slots[*level]++;
 				free_extent_buffer(next);
 				ret = btrfs_free_extent(trans, root,
-							blocknr, 1, 1);
+							bytenr, blocksize, 1);
 				BUG_ON(ret);
 				continue;
 			}
@@ -1321,8 +1334,8 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 out:
 	WARN_ON(*level < 0);
 	WARN_ON(*level >= BTRFS_MAX_LEVEL);
-	ret = btrfs_free_extent(trans, root,
-			extent_buffer_blocknr(path->nodes[*level]), 1, 1);
+	ret = btrfs_free_extent(trans, root, path->nodes[*level]->start,
+				path->nodes[*level]->len, 1);
 	free_extent_buffer(path->nodes[*level]);
 	path->nodes[*level] = NULL;
 	*level += 1;
@@ -1359,8 +1372,8 @@ static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 			return 0;
 		} else {
 			ret = btrfs_free_extent(trans, root,
-				    extent_buffer_blocknr(path->nodes[*level]),
-				    1, 1);
+						path->nodes[*level]->start,
+						path->nodes[*level]->len, 1);
 			BUG_ON(ret);
 			free_extent_buffer(path->nodes[*level]);
 			path->nodes[*level] = NULL;
@@ -1476,16 +1489,12 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 	struct btrfs_key key;
 	struct btrfs_key found_key;
 	struct extent_buffer *leaf;
-	u64 group_size_blocks;
 
 	block_group_cache = &info->block_group_cache;
 
-	group_size_blocks = BTRFS_BLOCK_GROUP_SIZE >>
-		info->sb->s_blocksize_bits;
-
 	root = info->extent_root;
 	key.objectid = 0;
-	key.offset = group_size_blocks;
+	key.offset = BTRFS_BLOCK_GROUP_SIZE;
 	btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
 
 	path = btrfs_alloc_path();
@@ -1532,7 +1541,7 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 				  (u64)cache);
 
 		if (key.objectid >=
-		    btrfs_super_total_blocks(&info->super_copy))
+		    btrfs_super_total_bytes(&info->super_copy))
 			break;
 	}
 

fs/btrfs/extent_map.c

@@ -1963,18 +1963,27 @@ static inline struct page *extent_buffer_page(struct extent_buffer *eb, int i)
 	struct page *p;
 	if (i == 0)
 		return eb->first_page;
+
 	i += eb->start >> PAGE_CACHE_SHIFT;
+	if (eb->last_page && eb->last_page->index == i)
+		return eb->last_page;
+
 	p = find_get_page(eb->first_page->mapping, i);
 	page_cache_release(p);
+	eb->last_page = p;
 	return p;
 }
 
+static inline unsigned long num_extent_pages(u64 start, u64 len)
+{
+	return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) -
+		(start >> PAGE_CACHE_SHIFT);
+}
 struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree,
 					  u64 start, unsigned long len,
 					  gfp_t mask)
 {
-	unsigned long num_pages = ((start + len - 1) >> PAGE_CACHE_SHIFT) -
-				  (start >> PAGE_CACHE_SHIFT) + 1;
+	unsigned long num_pages = num_extent_pages(start, len);
 	unsigned long i;
 	unsigned long index = start >> PAGE_CACHE_SHIFT;
 	struct extent_buffer *eb;
@@ -1986,7 +1995,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree,
 	if (!eb || IS_ERR(eb))
 		return NULL;
 
-	eb->alloc_addr = __builtin_return_address(0);
+	eb->alloc_addr = (unsigned long)__builtin_return_address(0);
 	eb->start = start;
 	eb->len = len;
 	atomic_set(&eb->refs, 1);
@@ -1994,6 +2003,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree,
 	for (i = 0; i < num_pages; i++, index++) {
 		p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM);
 		if (!p) {
+			WARN_ON(1);
 			/* make sure the free only frees the pages we've
 			 * grabbed a reference on
 			 */
@@ -2021,8 +2031,7 @@ struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree,
 					 u64 start, unsigned long len,
 					  gfp_t mask)
 {
-	unsigned long num_pages = ((start + len - 1) >> PAGE_CACHE_SHIFT) -
-				  (start >> PAGE_CACHE_SHIFT) + 1;
+	unsigned long num_pages = num_extent_pages(start, len);
 	unsigned long i;
 	unsigned long index = start >> PAGE_CACHE_SHIFT;
 	struct extent_buffer *eb;
@@ -2033,7 +2042,7 @@ struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree,
 	if (!eb || IS_ERR(eb))
 		return NULL;
 
-	eb->alloc_addr = __builtin_return_address(0);
+	eb->alloc_addr = (unsigned long)__builtin_return_address(0);
 	eb->start = start;
 	eb->len = len;
 	atomic_set(&eb->refs, 1);
@@ -2070,8 +2079,7 @@ void free_extent_buffer(struct extent_buffer *eb)
 	if (!atomic_dec_and_test(&eb->refs))
 		return;
 
-	num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
-		(eb->start >> PAGE_CACHE_SHIFT) + 1;
+	num_pages = num_extent_pages(eb->start, eb->len);
 
 	if (eb->first_page)
 		page_cache_release(eb->first_page);
@@ -2094,8 +2102,7 @@ int clear_extent_buffer_dirty(struct extent_map_tree *tree,
 	u64 end = start + eb->len - 1;
 
 	set = clear_extent_dirty(tree, start, end, GFP_NOFS);
-	num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
-		(eb->start >> PAGE_CACHE_SHIFT) + 1;
+	num_pages = num_extent_pages(eb->start, eb->len);
 
 	for (i = 0; i < num_pages; i++) {
 		page = extent_buffer_page(eb, i);
@@ -2145,8 +2152,7 @@ int set_extent_buffer_uptodate(struct extent_map_tree *tree,
 	struct page *page;
 	unsigned long num_pages;
 
-	num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
-		(eb->start >> PAGE_CACHE_SHIFT) + 1;
+	num_pages = num_extent_pages(eb->start, eb->len);
 
 	set_extent_uptodate(tree, eb->start, eb->start + eb->len - 1,
 			    GFP_NOFS);
@@ -2191,8 +2197,7 @@ int read_extent_buffer_pages(struct extent_map_tree *tree,
 		return 0;
 	}
 
-	num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
-		(eb->start >> PAGE_CACHE_SHIFT) + 1;
+	num_pages = num_extent_pages(eb->start, eb->len);
 	for (i = 0; i < num_pages; i++) {
 		page = extent_buffer_page(eb, i);
 		if (PageUptodate(page)) {
@@ -2267,14 +2272,14 @@ void read_extent_buffer(struct extent_buffer *eb, void *dstv,
 }
 EXPORT_SYMBOL(read_extent_buffer);
 
-int map_extent_buffer(struct extent_buffer *eb, unsigned long start,
-		      unsigned long min_len,
-		      char **token, char **map,
-		      unsigned long *map_start,
-		      unsigned long *map_len, int km)
+static int __map_extent_buffer(struct extent_buffer *eb, unsigned long start,
+			       unsigned long min_len, char **token, char **map,
+			       unsigned long *map_start,
+			       unsigned long *map_len, int km)
 {
 	size_t offset = start & (PAGE_CACHE_SIZE - 1);
 	char *kaddr;
+	struct page *p;
 	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
 	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
 	unsigned long end_i = (start_offset + start + min_len) >>
@@ -2283,21 +2288,59 @@ int map_extent_buffer(struct extent_buffer *eb, unsigned long start,
 	if (i != end_i)
 		return -EINVAL;
 
-	WARN_ON(start > eb->len);
+	if (start >= eb->len) {
+		printk("bad start in map eb start %Lu len %lu caller start %lu min %lu\n", eb->start, eb->len, start, min_len);
+		WARN_ON(1);
+	}
 
 	if (i == 0) {
 		offset = start_offset;
 		*map_start = 0;
 	} else {
+		offset = 0;
 		*map_start = (i << PAGE_CACHE_SHIFT) - start_offset;
 	}
 
-	kaddr = kmap_atomic(extent_buffer_page(eb, i), km);
+	p = extent_buffer_page(eb, i);
+	WARN_ON(!PageUptodate(p));
+	kaddr = kmap_atomic(p, km);
 	*token = kaddr;
 	*map = kaddr + offset;
 	*map_len = PAGE_CACHE_SIZE - offset;
 	return 0;
 }
+
+int map_extent_buffer(struct extent_buffer *eb, unsigned long start,
+		      unsigned long min_len,
+		      char **token, char **map,
+		      unsigned long *map_start,
+		      unsigned long *map_len, int km)
+{
+	int err;
+	int save = 0;
+	if (eb->map_token) {
+		if (start >= eb->map_start &&
+		    start + min_len <= eb->map_start + eb->map_len) {
+			*token = eb->map_token;
+			*map = eb->kaddr;
+			*map_start = eb->map_start;
+			*map_len = eb->map_len;
+			return 0;
+		}
+		unmap_extent_buffer(eb, eb->map_token, km);
+		eb->map_token = NULL;
+		save = 1;
+	}
+	err = __map_extent_buffer(eb, start, min_len, token, map,
+				   map_start, map_len, km);
+	if (!err && save) {
+		eb->map_token = *token;
+		eb->kaddr = *map;
+		eb->map_start = *map_start;
+		eb->map_len = *map_len;
+	}
+	return err;
+}
 EXPORT_SYMBOL(map_extent_buffer);
 
 void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km)
@@ -2574,7 +2617,6 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
 			((unsigned long)PAGE_CACHE_SIZE - 1);
 		src_off_in_page = src_end &
 			((unsigned long)PAGE_CACHE_SIZE - 1);
-
 		if (src_i == 0)
 			src_off_in_page += start_offset;
 		if (dst_i == 0)
@@ -2582,14 +2624,13 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
 
 		cur = min(len, src_off_in_page + 1);
 		cur = min(cur, dst_off_in_page + 1);
-
 		move_pages(extent_buffer_page(dst, dst_i),
 			   extent_buffer_page(dst, src_i),
 			   dst_off_in_page - cur + 1,
 			   src_off_in_page - cur + 1, cur);
 
-		dst_end -= cur - 1;
-		src_end -= cur - 1;
+		dst_end -= cur;
+		src_end -= cur;
 		len -= cur;
 	}
 }

fs/btrfs/extent_map.h

@@ -70,7 +70,12 @@ struct extent_buffer {
 	struct list_head list;
 	struct list_head leak_list;
 	struct page *first_page;
+	struct page *last_page;
 	unsigned long alloc_addr;
+	char *map_token;
+	char *kaddr;
+	unsigned long map_start;
+	unsigned long map_len;
 };
 
 typedef struct extent_map *(get_extent_t)(struct inode *inode,
@@ -147,11 +152,6 @@ static inline void extent_buffer_get(struct extent_buffer *eb)
 	atomic_inc(&eb->refs);
 }
 
-static inline u64 extent_buffer_blocknr(struct extent_buffer *eb)
-{
-	return eb->start / 4096;
-}
-
 int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
 			  unsigned long start,
 			  unsigned long len);

fs/btrfs/file-item.c

@@ -27,8 +27,8 @@
 int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 			       struct btrfs_root *root,
 			       u64 objectid, u64 pos,
-			       u64 offset, u64 disk_num_blocks,
-			       u64 num_blocks)
+			       u64 offset, u64 disk_num_bytes,
+			       u64 num_bytes)
 {
 	int ret = 0;
 	struct btrfs_file_extent_item *item;
@@ -50,10 +50,10 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 	leaf = path->nodes[0];
 	item = btrfs_item_ptr(leaf, path->slots[0],
 			      struct btrfs_file_extent_item);
-	btrfs_set_file_extent_disk_blocknr(leaf, item, offset);
-	btrfs_set_file_extent_disk_num_blocks(leaf, item, disk_num_blocks);
+	btrfs_set_file_extent_disk_bytenr(leaf, item, offset);
+	btrfs_set_file_extent_disk_num_bytes(leaf, item, disk_num_bytes);
 	btrfs_set_file_extent_offset(leaf, item, 0);
-	btrfs_set_file_extent_num_blocks(leaf, item, num_blocks);
+	btrfs_set_file_extent_num_bytes(leaf, item, num_bytes);
 	btrfs_set_file_extent_generation(leaf, item, trans->transid);
 	btrfs_set_file_extent_type(leaf, item, BTRFS_FILE_EXTENT_REG);
 	btrfs_mark_buffer_dirty(leaf);

fs/btrfs/file.c

@@ -120,9 +120,9 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans,
 	btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE);
 	ptr = btrfs_file_extent_inline_start(ei);
 
-	kaddr = kmap_atomic(page, KM_USER0);
+	kaddr = kmap_atomic(page, KM_USER1);
 	write_extent_buffer(leaf, kaddr + page_offset, ptr, size);
-	kunmap_atomic(kaddr, KM_USER0);
+	kunmap_atomic(kaddr, KM_USER1);
 	btrfs_mark_buffer_dirty(leaf);
 fail:
 	btrfs_free_path(path);
@@ -142,11 +142,12 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 	struct inode *inode = file->f_path.dentry->d_inode;
 	struct extent_map *em;
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
-	u64 hint_block;
-	u64 num_blocks;
+	u64 hint_byte;
+	u64 num_bytes;
 	u64 start_pos;
 	u64 end_of_last_block;
 	u64 end_pos = pos + write_bytes;
+	u32 inline_size;
 	loff_t isize = i_size_read(inode);
 
 	em = alloc_extent_map(GFP_NOFS);
@@ -156,11 +157,12 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 	em->bdev = inode->i_sb->s_bdev;
 
 	start_pos = pos & ~((u64)root->sectorsize - 1);
-	num_blocks = (write_bytes + pos - start_pos + root->sectorsize - 1) >>
-			inode->i_blkbits;
+	num_bytes = (write_bytes + pos - start_pos +
+		    root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
 
 	down_read(&BTRFS_I(inode)->root->snap_sem);
-	end_of_last_block = start_pos + (num_blocks << inode->i_blkbits) - 1;
+	end_of_last_block = start_pos + num_bytes - 1;
+
 	lock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS);
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
@@ -169,8 +171,8 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 		goto out_unlock;
 	}
 	btrfs_set_trans_block_group(trans, inode);
-	inode->i_blocks += num_blocks << 3;
-	hint_block = 0;
+	inode->i_blocks += num_bytes >> 9;
+	hint_byte = 0;
 
 	if ((end_of_last_block & 4095) == 0) {
 		printk("strange end of last %Lu %zu %Lu\n", start_pos, write_bytes, end_of_last_block);
@@ -191,11 +193,10 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 			err = btrfs_drop_extents(trans, root, inode,
 						 last_pos_in_file,
 						 last_pos_in_file + hole_size,
-						 &hint_block);
+						 &hint_byte);
 			if (err)
 				goto failed;
 
-			hole_size >>= inode->i_blkbits;
 			err = btrfs_insert_file_extent(trans, root,
 						       inode->i_ino,
 						       last_pos_in_file,
@@ -209,8 +210,10 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 	 * either allocate an extent for the new bytes or setup the key
 	 * to show we are doing inline data in the extent
 	 */
+	inline_size = end_pos - start_pos;
 	if (isize >= PAGE_CACHE_SIZE || pos + write_bytes < inode->i_size ||
-	    pos + write_bytes - start_pos > BTRFS_MAX_INLINE_DATA_SIZE(root)) {
+	    inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root) ||
+	    inline_size >= PAGE_CACHE_SIZE) {
 		u64 last_end;
 		for (i = 0; i < num_pages; i++) {
 			struct page *p = pages[i];
@@ -224,10 +227,9 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 	} else {
 		struct page *p = pages[0];
 		/* step one, delete the existing extents in this range */
-		/* FIXME blocksize != pagesize */
 		err = btrfs_drop_extents(trans, root, inode, start_pos,
 			 (pos + write_bytes + root->sectorsize -1) &
-			 ~((u64)root->sectorsize - 1), &hint_block);
+			 ~((u64)root->sectorsize - 1), &hint_byte);
 		if (err)
 			goto failed;
 
@@ -283,7 +285,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end)
  */
 int btrfs_drop_extents(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root, struct inode *inode,
-		       u64 start, u64 end, u64 *hint_block)
+		       u64 start, u64 end, u64 *hint_byte)
 {
 	int ret;
 	struct btrfs_key key;
@@ -346,8 +348,7 @@ next_slot:
 			found_type = btrfs_file_extent_type(leaf, extent);
 			if (found_type == BTRFS_FILE_EXTENT_REG) {
 				extent_end = key.offset +
-				 (btrfs_file_extent_num_blocks(leaf, extent) <<
-					 inode->i_blkbits);
+				     btrfs_file_extent_num_bytes(leaf, extent);
 				found_extent = 1;
 			} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
 				struct btrfs_item *item;
@@ -386,17 +387,17 @@ next_slot:
 
 		if (end < extent_end && end >= key.offset) {
 			if (found_extent) {
-				u64 disk_blocknr =
-				    btrfs_file_extent_disk_blocknr(leaf,extent);
-				u64 disk_num_blocks =
-				    btrfs_file_extent_disk_num_blocks(leaf,
+				u64 disk_bytenr =
+				    btrfs_file_extent_disk_bytenr(leaf, extent);
+				u64 disk_num_bytes =
+				    btrfs_file_extent_disk_num_bytes(leaf,
 								      extent);
 				read_extent_buffer(leaf, &old,
 						   (unsigned long)extent,
 						   sizeof(old));
-				if (disk_blocknr != 0) {
+				if (disk_bytenr != 0) {
 					ret = btrfs_inc_extent_ref(trans, root,
-					         disk_blocknr, disk_num_blocks);
+					         disk_bytenr, disk_num_bytes);
 					BUG_ON(ret);
 				}
 			}
@@ -410,21 +411,19 @@ next_slot:
 			keep = 1;
 			WARN_ON(start & (root->sectorsize - 1));
 			if (found_extent) {
-				new_num = (start - key.offset) >>
-					inode->i_blkbits;
-				old_num = btrfs_file_extent_num_blocks(leaf,
-								       extent);
-				*hint_block =
-					btrfs_file_extent_disk_blocknr(leaf,
-								       extent);
-				if (btrfs_file_extent_disk_blocknr(leaf,
-								   extent)) {
+				new_num = start - key.offset;
+				old_num = btrfs_file_extent_num_bytes(leaf,
+								      extent);
+				*hint_byte =
+					btrfs_file_extent_disk_bytenr(leaf,
+								      extent);
+				if (btrfs_file_extent_disk_bytenr(leaf,
+								  extent)) {
 					inode->i_blocks -=
-						(old_num - new_num) << 3;
+						(old_num - new_num) >> 9;
 				}
-				btrfs_set_file_extent_num_blocks(leaf,
-								 extent,
-								 new_num);
+				btrfs_set_file_extent_num_bytes(leaf, extent,
+								new_num);
 				btrfs_mark_buffer_dirty(leaf);
 			} else {
 				WARN_ON(1);
@@ -432,33 +431,32 @@ next_slot:
 		}
 		/* delete the entire extent */
 		if (!keep) {
-			u64 disk_blocknr = 0;
-			u64 disk_num_blocks = 0;
-			u64 extent_num_blocks = 0;
+			u64 disk_bytenr = 0;
+			u64 disk_num_bytes = 0;
+			u64 extent_num_bytes = 0;
 			if (found_extent) {
-				disk_blocknr =
-				      btrfs_file_extent_disk_blocknr(leaf,
+				disk_bytenr =
+				      btrfs_file_extent_disk_bytenr(leaf,
 								     extent);
-				disk_num_blocks =
-				      btrfs_file_extent_disk_num_blocks(leaf,
-									extent);
-				extent_num_blocks =
-				      btrfs_file_extent_num_blocks(leaf,
-								   extent);
-				*hint_block =
-					btrfs_file_extent_disk_blocknr(leaf,
+				disk_num_bytes =
+				      btrfs_file_extent_disk_num_bytes(leaf,
 								       extent);
+				extent_num_bytes =
+				      btrfs_file_extent_num_bytes(leaf, extent);
+				*hint_byte =
+					btrfs_file_extent_disk_bytenr(leaf,
+								      extent);
 			}
 			ret = btrfs_del_item(trans, root, path);
 			/* TODO update progress marker and return */
 			BUG_ON(ret);
 			btrfs_release_path(root, path);
 			extent = NULL;
-			if (found_extent && disk_blocknr != 0) {
-				inode->i_blocks -= extent_num_blocks << 3;
+			if (found_extent && disk_bytenr != 0) {
+				inode->i_blocks -= extent_num_bytes >> 9;
 				ret = btrfs_free_extent(trans, root,
-							disk_blocknr,
-							disk_num_blocks, 0);
+							disk_bytenr,
+							disk_num_bytes, 0);
 			}
 
 			BUG_ON(ret);
@@ -491,20 +489,19 @@ next_slot:
 					    (unsigned long)extent, sizeof(old));
 
 			btrfs_set_file_extent_offset(leaf, extent,
-				    le64_to_cpu(old.offset) +
-				    ((end - key.offset) >> inode->i_blkbits));
-			WARN_ON(le64_to_cpu(old.num_blocks) <
-				(extent_end - end) >> inode->i_blkbits);
-			btrfs_set_file_extent_num_blocks(leaf, extent,
-				    (extent_end - end) >> inode->i_blkbits);
-
+				    le64_to_cpu(old.offset) + end - key.offset);
+			WARN_ON(le64_to_cpu(old.num_bytes) <
+				(extent_end - end));
+			btrfs_set_file_extent_num_bytes(leaf, extent,
+							extent_end - end);
 			btrfs_set_file_extent_type(leaf, extent,
 						   BTRFS_FILE_EXTENT_REG);
+
 			btrfs_mark_buffer_dirty(path->nodes[0]);
-			if (le64_to_cpu(old.disk_blocknr) != 0) {
+			if (le64_to_cpu(old.disk_bytenr) != 0) {
 				inode->i_blocks +=
-				      btrfs_file_extent_num_blocks(leaf,
-								   extent) << 3;
+				      btrfs_file_extent_num_bytes(leaf,
+								  extent) >> 9;
 			}
 			ret = 0;
 			goto out;
@@ -531,12 +528,9 @@ static int prepare_pages(struct btrfs_root *root,
 	unsigned long index = pos >> PAGE_CACHE_SHIFT;
 	struct inode *inode = file->f_path.dentry->d_inode;
 	int err = 0;
-	u64 num_blocks;
 	u64 start_pos;
 
 	start_pos = pos & ~((u64)root->sectorsize - 1);
-	num_blocks = (write_bytes + pos - start_pos + root->sectorsize - 1) >>
-			inode->i_blkbits;
 
 	memset(pages, 0, num_pages * sizeof(struct page *));
 

fs/btrfs/inode.c

@@ -77,19 +77,19 @@ static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
 	struct btrfs_trans_handle *trans;
 	struct btrfs_key ins;
 	u64 alloc_hint = 0;
-	u64 num_blocks;
+	u64 num_bytes;
 	int ret;
-	u64 blocksize = 1 << inode->i_blkbits;
+	u64 blocksize = root->sectorsize;
 
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
 	btrfs_set_trans_block_group(trans, inode);
 	BUG_ON(!trans);
-	num_blocks = (end - start + blocksize) & ~(blocksize - 1);
+	num_bytes = (end - start + blocksize) & ~(blocksize - 1);
 	ret = btrfs_drop_extents(trans, root, inode,
-				 start, start + num_blocks, &alloc_hint);
-	num_blocks = num_blocks >> inode->i_blkbits;
-	ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_blocks, 0,
+				 start, start + num_bytes, &alloc_hint);
+
+	ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_bytes, 0,
 				 alloc_hint, (u64)-1, &ins, 1);
 	if (ret) {
 		WARN_ON(1);
@@ -186,7 +186,8 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
 zeroit:
 	printk("btrfs csum failed ino %lu off %llu\n",
 	       page->mapping->host->i_ino, (unsigned long long)start);
-	memset(kaddr + offset, 1, end - start + 1); flush_dcache_page(page);
+	memset(kaddr + offset, 1, end - start + 1);
+	flush_dcache_page(page);
 	kunmap_atomic(kaddr, KM_IRQ0);
 	return 0;
 }
@@ -547,7 +548,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 	struct extent_buffer *leaf;
 	struct btrfs_file_extent_item *fi;
 	u64 extent_start = 0;
-	u64 extent_num_blocks = 0;
+	u64 extent_num_bytes = 0;
 	u64 item_end = 0;
 	int found_extent;
 	int del_item;
@@ -593,8 +594,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 			if (btrfs_file_extent_type(leaf, fi) !=
 			    BTRFS_FILE_EXTENT_INLINE) {
 				item_end +=
-				    btrfs_file_extent_num_blocks(leaf, fi) <<
-				    inode->i_blkbits;
+				    btrfs_file_extent_num_bytes(leaf, fi);
 			}
 		}
 		if (found_type == BTRFS_CSUM_ITEM_KEY) {
@@ -626,28 +626,27 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 			   btrfs_file_extent_type(leaf, fi) !=
 			   BTRFS_FILE_EXTENT_INLINE) {
 			u64 num_dec;
-			extent_start = btrfs_file_extent_disk_blocknr(leaf, fi);
+			extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
 			if (!del_item) {
-				u64 orig_num_blocks =
-					btrfs_file_extent_num_blocks(leaf, fi);
-				extent_num_blocks = inode->i_size -
+				u64 orig_num_bytes =
+					btrfs_file_extent_num_bytes(leaf, fi);
+				extent_num_bytes = inode->i_size -
 					found_key.offset + root->sectorsize - 1;
-				extent_num_blocks >>= inode->i_blkbits;
-				btrfs_set_file_extent_num_blocks(leaf, fi,
-							 extent_num_blocks);
-				num_dec = (orig_num_blocks -
-					   extent_num_blocks) << 3;
+				btrfs_set_file_extent_num_bytes(leaf, fi,
+							 extent_num_bytes);
+				num_dec = (orig_num_bytes -
+					   extent_num_bytes) >> 9;
 				if (extent_start != 0) {
 					inode->i_blocks -= num_dec;
 				}
 				btrfs_mark_buffer_dirty(leaf);
 			} else {
-				extent_num_blocks =
-					btrfs_file_extent_disk_num_blocks(leaf,
-									  fi);
+				extent_num_bytes =
+					btrfs_file_extent_disk_num_bytes(leaf,
+									 fi);
 				/* FIXME blocksize != 4096 */
-				num_dec = btrfs_file_extent_num_blocks(leaf,
-								       fi) << 3;
+				num_dec = btrfs_file_extent_num_bytes(leaf,
+								       fi) >> 9;
 				if (extent_start != 0) {
 					found_extent = 1;
 					inode->i_blocks -= num_dec;
@@ -664,7 +663,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 		btrfs_release_path(root, path);
 		if (found_extent) {
 			ret = btrfs_free_extent(trans, root, extent_start,
-						extent_num_blocks, 0);
+						extent_num_bytes, 0);
 			BUG_ON(ret);
 		}
 	}
@@ -709,7 +708,8 @@ static int btrfs_cow_one_page(struct inode *inode, struct page *page,
 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 {
 	struct inode *inode = mapping->host;
-	unsigned blocksize = 1 << inode->i_blkbits;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u32 blocksize = root->sectorsize;
 	pgoff_t index = from >> PAGE_CACHE_SHIFT;
 	unsigned offset = from & (PAGE_CACHE_SIZE-1);
 	struct page *page;
@@ -719,7 +719,7 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 	if ((offset & (blocksize - 1)) == 0)
 		goto out;
 
-	down_read(&BTRFS_I(inode)->root->snap_sem);
+	down_read(&root->snap_sem);
 	ret = -ENOMEM;
 	page = grab_cache_page(mapping, index);
 	if (!page)
@@ -778,8 +778,6 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
 		err = btrfs_drop_extents(trans, root, inode,
 					 pos, pos + hole_size, &alloc_hint);
 
-		hole_size >>= inode->i_blkbits;
-
 		err = btrfs_insert_file_extent(trans, root, inode->i_ino,
 					       pos, 0, 0, hole_size);
 		btrfs_end_transaction(trans, root);
@@ -1490,7 +1488,7 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
 {
 	int ret;
 	int err = 0;
-	u64 blocknr;
+	u64 bytenr;
 	u64 extent_start = 0;
 	u64 extent_end = 0;
 	u64 objectid = inode->i_ino;
@@ -1540,10 +1538,6 @@ again:
 	leaf = path->nodes[0];
 	item = btrfs_item_ptr(leaf, path->slots[0],
 			      struct btrfs_file_extent_item);
-
-	blocknr = btrfs_file_extent_disk_blocknr(leaf, item);
-	blocknr += btrfs_file_extent_offset(leaf, item);
-
 	/* are we inside the extent that was found? */
 	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 	found_type = btrfs_key_type(&found_key);
@@ -1556,8 +1550,7 @@ again:
 	extent_start = found_key.offset;
 	if (found_type == BTRFS_FILE_EXTENT_REG) {
 		extent_end = extent_start +
-		       (btrfs_file_extent_num_blocks(leaf, item) <<
-			inode->i_blkbits);
+		       btrfs_file_extent_num_bytes(leaf, item);
 		err = 0;
 		if (start < extent_start || start >= extent_end) {
 			em->start = start;
@@ -1570,17 +1563,18 @@ again:
 			}
 			goto not_found_em;
 		}
-		if (btrfs_file_extent_disk_blocknr(leaf, item) == 0) {
+		bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
+		if (bytenr == 0) {
 			em->start = extent_start;
 			em->end = extent_end - 1;
 			em->block_start = EXTENT_MAP_HOLE;
 			em->block_end = EXTENT_MAP_HOLE;
 			goto insert;
 		}
-		em->block_start = blocknr << inode->i_blkbits;
+		bytenr += btrfs_file_extent_offset(leaf, item);
+		em->block_start = bytenr;
 		em->block_end = em->block_start +
-			(btrfs_file_extent_num_blocks(leaf, item) <<
-			 inode->i_blkbits) - 1;
+			btrfs_file_extent_num_bytes(leaf, item) - 1;
 		em->start = extent_start;
 		em->end = extent_end - 1;
 		goto insert;
@@ -1592,7 +1586,8 @@ again:
 		size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
 						    path->slots[0]));
 
-		extent_end = extent_start | ((u64)root->sectorsize - 1);
+		extent_end = (extent_start + size) |
+			((u64)root->sectorsize - 1);
 		if (start < extent_start || start >= extent_end) {
 			em->start = start;
 			if (start < extent_start) {
@@ -1617,8 +1612,10 @@ again:
 		ptr = btrfs_file_extent_inline_start(item);
 		map = kmap(page);
 		read_extent_buffer(leaf, map + page_offset, ptr, size);
+		/*
 		memset(map + page_offset + size, 0,
 		       root->sectorsize - (page_offset + size));
+		       */
 		flush_dcache_page(page);
 		kunmap(page);
 		set_extent_uptodate(em_tree, extent_start,
@@ -1836,13 +1833,13 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 	trans = btrfs_start_transaction(root, 1);
 	BUG_ON(!trans);
 
-	leaf = btrfs_alloc_free_block(trans, root, 0, 0);
+	leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, 0);
 	if (IS_ERR(leaf))
 		return PTR_ERR(leaf);
 
 	btrfs_set_header_nritems(leaf, 0);
 	btrfs_set_header_level(leaf, 0);
-	btrfs_set_header_blocknr(leaf, extent_buffer_blocknr(leaf));
+	btrfs_set_header_bytenr(leaf, leaf->start);
 	btrfs_set_header_generation(leaf, trans->transid);
 	btrfs_set_header_owner(leaf, root->root_key.objectid);
 	write_extent_buffer(leaf, root->fs_info->fsid,
@@ -1858,7 +1855,8 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 	inode_item->nblocks = cpu_to_le64(1);
 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
 
-	btrfs_set_root_blocknr(&root_item, extent_buffer_blocknr(leaf));
+	btrfs_set_root_bytenr(&root_item, leaf->start);
+	btrfs_set_root_level(&root_item, 0);
 	btrfs_set_root_refs(&root_item, 1);
 	btrfs_set_root_used(&root_item, 0);
 
@@ -1971,8 +1969,8 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
 
 	btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
-	btrfs_set_root_blocknr(&new_root_item,
-			       extent_buffer_blocknr(root->node));
+	btrfs_set_root_bytenr(&new_root_item, root->node->start);
+	btrfs_set_root_level(&new_root_item, btrfs_header_level(root->node));
 
 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 				&new_root_item);

fs/btrfs/print-tree.c

@@ -36,7 +36,7 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 	u32 type;
 
 	printk("leaf %llu total ptrs %d free space %d\n",
-		(unsigned long long)btrfs_header_blocknr(l), nr,
+		(unsigned long long)btrfs_header_bytenr(l), nr,
 		btrfs_leaf_free_space(root, l));
 	for (i = 0 ; i < nr ; i++) {
 		item = btrfs_item_nr(l, i);
@@ -65,8 +65,8 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 			break;
 		case BTRFS_ROOT_ITEM_KEY:
 			ri = btrfs_item_ptr(l, i, struct btrfs_root_item);
-			printk("\t\troot data blocknr %llu refs %u\n",
-				(unsigned long long)btrfs_disk_root_blocknr(l, ri),
+			printk("\t\troot data bytenr %llu refs %u\n",
+				(unsigned long long)btrfs_disk_root_bytenr(l, ri),
 				btrfs_disk_root_refs(l, ri));
 			break;
 		case BTRFS_EXTENT_ITEM_KEY:
@@ -84,12 +84,12 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 			           btrfs_file_extent_inline_len(l, item));
 				break;
 			}
-			printk("\t\textent data disk block %llu nr %llu\n",
-			       (unsigned long long)btrfs_file_extent_disk_blocknr(l, fi),
-			       (unsigned long long)btrfs_file_extent_disk_num_blocks(l, fi));
+			printk("\t\textent data disk bytenr %llu nr %llu\n",
+			       (unsigned long long)btrfs_file_extent_disk_bytenr(l, fi),
+			       (unsigned long long)btrfs_file_extent_disk_num_bytes(l, fi));
 			printk("\t\textent data offset %llu nr %llu\n",
 			  (unsigned long long)btrfs_file_extent_offset(l, fi),
-			  (unsigned long long)btrfs_file_extent_num_blocks(l, fi));
+			  (unsigned long long)btrfs_file_extent_num_bytes(l, fi));
 			break;
 		case BTRFS_BLOCK_GROUP_ITEM_KEY:
 			bi = btrfs_item_ptr(l, i,
@@ -106,16 +106,18 @@ void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *c)
 	int i;
 	u32 nr;
 	struct btrfs_key key;
+	int level;
 
 	if (!c)
 		return;
 	nr = btrfs_header_nritems(c);
-	if (btrfs_is_leaf(c)) {
+	level = btrfs_header_level(c);
+	if (level == 0) {
 		btrfs_print_leaf(root, c);
 		return;
 	}
 	printk("node %llu level %d total ptrs %d free spc %u\n",
-	       (unsigned long long)btrfs_header_blocknr(c),
+	       (unsigned long long)btrfs_header_bytenr(c),
 	       btrfs_header_level(c), nr,
 	       (u32)BTRFS_NODEPTRS_PER_BLOCK(root) - nr);
 	for (i = 0; i < nr; i++) {
@@ -129,7 +131,8 @@ void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *c)
 	}
 	for (i = 0; i < nr; i++) {
 		struct extent_buffer *next = read_tree_block(root,
-						btrfs_node_blockptr(c, i));
+					btrfs_node_blockptr(c, i),
+					btrfs_level_size(root, level - 1));
 		if (btrfs_is_leaf(next) &&
 		    btrfs_header_level(c) != 1)
 			BUG();

fs/btrfs/super.c

@@ -303,10 +303,12 @@ static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 {
 	struct btrfs_root *root = btrfs_sb(dentry->d_sb);
 	struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
+	int bits = dentry->d_sb->s_blocksize_bits;
 
 	buf->f_namelen = BTRFS_NAME_LEN;
-	buf->f_blocks = btrfs_super_total_blocks(disk_super);
-	buf->f_bfree = buf->f_blocks - btrfs_super_blocks_used(disk_super);
+	buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
+	buf->f_bfree = buf->f_blocks -
+		(btrfs_super_bytes_used(disk_super) >> bits);
 	buf->f_bavail = buf->f_bfree;
 	buf->f_bsize = dentry->d_sb->s_blocksize;
 	buf->f_type = BTRFS_SUPER_MAGIC;

fs/btrfs/sysfs.c

@@ -42,14 +42,15 @@ static ssize_t root_block_limit_show(struct btrfs_root *root, char *buf)
 
 static ssize_t super_blocks_used_show(struct btrfs_fs_info *fs, char *buf)
 {
+
 	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_super_blocks_used(&fs->super_copy));
+		(unsigned long long)btrfs_super_bytes_used(&fs->super_copy));
 }
 
 static ssize_t super_total_blocks_show(struct btrfs_fs_info *fs, char *buf)
 {
 	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_super_total_blocks(&fs->super_copy));
+		(unsigned long long)btrfs_super_total_bytes(&fs->super_copy));
 }
 
 static ssize_t super_blocksize_show(struct btrfs_fs_info *fs, char *buf)

fs/btrfs/transaction.c

@@ -205,12 +205,13 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
 
 	btrfs_write_dirty_block_groups(trans, extent_root);
 	while(1) {
-		old_extent_block = btrfs_root_blocknr(&extent_root->root_item);
-		if (old_extent_block ==
-		    extent_buffer_blocknr(extent_root->node))
+		old_extent_block = btrfs_root_bytenr(&extent_root->root_item);
+		if (old_extent_block == extent_root->node->start)
 			break;
-		btrfs_set_root_blocknr(&extent_root->root_item,
-			       extent_buffer_blocknr(extent_root->node));
+		btrfs_set_root_bytenr(&extent_root->root_item,
+				      extent_root->node->start);
+		btrfs_set_root_level(&extent_root->root_item,
+				     btrfs_header_level(extent_root->node));
 		ret = btrfs_update_root(trans, tree_root,
 					&extent_root->root_key,
 					&extent_root->root_item);
@@ -284,8 +285,8 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans,
 				     (unsigned long)root->root_key.objectid,
 				     BTRFS_ROOT_TRANS_TAG);
 			if (root->commit_root == root->node) {
-				WARN_ON(extent_buffer_blocknr(root->node) !=
-					btrfs_root_blocknr(&root->root_item));
+				WARN_ON(root->node->start !=
+					btrfs_root_bytenr(&root->root_item));
 				free_extent_buffer(root->commit_root);
 				root->commit_root = NULL;
 
@@ -314,8 +315,10 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans,
 			root->commit_root = NULL;
 
 			root->root_key.offset = root->fs_info->generation;
-			btrfs_set_root_blocknr(&root->root_item,
-				       extent_buffer_blocknr(root->node));
+			btrfs_set_root_bytenr(&root->root_item,
+					      root->node->start);
+			btrfs_set_root_level(&root->root_item,
+					     btrfs_header_level(root->node));
 			err = btrfs_insert_root(trans, root->fs_info->tree_root,
 						&root->root_key,
 						&root->root_item);
@@ -407,8 +410,8 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 	struct dirty_root *dirty;
 	struct btrfs_trans_handle *trans;
 	unsigned long nr;
-	u64 num_blocks;
-	u64 blocks_used;
+	u64 num_bytes;
+	u64 bytes_used;
 	int ret = 0;
 	int err;
 
@@ -419,7 +422,7 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 		dirty = list_entry(list->next, struct dirty_root, list);
 		list_del_init(&dirty->list);
 
-		num_blocks = btrfs_root_used(&dirty->root->root_item);
+		num_bytes = btrfs_root_used(&dirty->root->root_item);
 		root = dirty->latest_root;
 
 		while(1) {
@@ -446,12 +449,12 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 		}
 		BUG_ON(ret);
 
-		num_blocks -= btrfs_root_used(&dirty->root->root_item);
-		blocks_used = btrfs_root_used(&root->root_item);
-		if (num_blocks) {
+		num_bytes -= btrfs_root_used(&dirty->root->root_item);
+		bytes_used = btrfs_root_used(&root->root_item);
+		if (num_bytes) {
 			record_root_in_trans(root);
 			btrfs_set_root_used(&root->root_item,
-						   blocks_used - num_blocks);
+					    bytes_used - num_bytes);
 		}
 		ret = btrfs_del_root(trans, tree_root, &dirty->root->root_key);
 		if (ret) {
@@ -560,7 +563,9 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	btrfs_set_super_generation(&root->fs_info->super_copy,
 				   cur_trans->transid);
 	btrfs_set_super_root(&root->fs_info->super_copy,
-		     extent_buffer_blocknr(root->fs_info->tree_root->node));
+			     root->fs_info->tree_root->node->start);
+	btrfs_set_super_root_level(&root->fs_info->super_copy,
+			   btrfs_header_level(root->fs_info->tree_root->node));
 
 	write_extent_buffer(root->fs_info->sb_buffer,
 			    &root->fs_info->super_copy, 0,

fs/btrfs/tree-defrag.c

@@ -27,13 +27,15 @@ static void reada_defrag(struct btrfs_root *root,
 {
 	int i;
 	u32 nritems;
-	u64 blocknr;
+	u64 bytenr;
+	u32 blocksize;
 	int ret;
 
+	blocksize = btrfs_level_size(root, btrfs_header_level(node) - 1);
 	nritems = btrfs_header_nritems(node);
 	for (i = 0; i < nritems; i++) {
-		blocknr = btrfs_node_blockptr(node, i);
-		ret = readahead_tree_block(root, blocknr);
+		bytenr = btrfs_node_blockptr(node, i);
+		ret = readahead_tree_block(root, bytenr, blocksize);
 		if (ret)
 			break;
 	}
@@ -46,7 +48,7 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 {
 	struct extent_buffer *next;
 	struct extent_buffer *cur;
-	u64 blocknr;
+	u64 bytenr;
 	int ret = 0;
 	int is_extent = 0;
 
@@ -80,10 +82,11 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 
 			break;
 		}
-		blocknr = btrfs_node_blockptr(cur, path->slots[*level]);
+		bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
 
 		if (cache_only) {
-			next = btrfs_find_tree_block(root, blocknr);
+			next = btrfs_find_tree_block(root, bytenr,
+					   btrfs_level_size(root, *level - 1));
 			/* FIXME, test for defrag */
 			if (!next || !btrfs_buffer_uptodate(next)) {
 				free_extent_buffer(next);
@@ -91,7 +94,8 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 				continue;
 			}
 		} else {
-			next = read_tree_block(root, blocknr);
+			next = read_tree_block(root, bytenr,
+				       btrfs_level_size(root, *level - 1));
 		}
 		ret = btrfs_cow_block(trans, root, next, path->nodes[*level],
 				      path->slots[*level], &next);