On the most general linear theory of gravitation