On an inverse linear programming problem