This paper considers the containment control problem for general linear multi-agent systems which have communication topologies with non-cooperative edges.It admits multiple leaders which are dynamically changing via interactions between themselves and the neighboring leaders.A distributed control protocol is designed with the nearest neighbor rule.By using the linear matrix inequality technique and the linear system control theory,it is shown that the followers' state trajectories can converge to the convex hull formed by each leader's trajectory as well as its symmetric trajectory,and the containment control problem for linear multi-agent system over signed graph can be achieved.Simulation results are performed to verify the correctness of the conclusion.