f2fs: fix to unlock node page when it was truncated

If the node page was truncated, its block address became zero.
This means that we don't need to write the node page, but have to unlock
NODE_WRITE, decrease the number of dirty node pages, and then unlock_page
before returning the f2fs_write_node_page with zero.

Reviewed-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>

fs/f2fs/node.c

@@ -1106,15 +1106,15 @@ static int f2fs_write_node_page(struct page *page,
 
 	/* This page is already truncated */
 	if (ni.blk_addr == NULL_ADDR)
-		return 0;
+		goto out;
 
 	set_page_writeback(page);
 
 	/* insert node offset */
 	write_node_page(sbi, page, nid, ni.blk_addr, &new_addr);
 	set_node_addr(sbi, &ni, new_addr);
+out:
 	dec_page_count(sbi, F2FS_DIRTY_NODES);
-
 	mutex_unlock_op(sbi, NODE_WRITE);
 	unlock_page(page);
 	return 0;