In this paper, the leader-following consensus problem for multi-agent systems is considered. Each agent is assumed to be modeled by a linear multi-input system. A novel (generalized homogeneous) consensus control protocol is designed under the assumption that there are some uncertainties in the dynamic of the leader. Some conditions (LMIs) are derived to select the control parameters in order to ensure the input-to-state stability and global finite-time stability of the consensus errors with desired homogeneity degrees. Some simulations are performed to illustrate the obtained results.