We study distributed power control for cognitive radio in this paper. The cognitive radio environment of multiusers is viewed as a non-cooperative game. In the condition of incomplete information, the efficient power control for two users can be realized by finite repeated gaming. If the repeated times M satisfy certain term, the two users can cooperate in the preceding phase of the repeated gaming, thus reaching the Pareto stable equilibrium point of the power allocation. Only when the repeated game is nearly over, users will adopt competitive power allocation strategy.