In this letter, we consider the power control problem in cognitive radio (CR) networks when both primary user (PU) location and wireless channel are unknown. Prior work in power control assumes perfect knowledge of PU and CR locations, which is not practical due to localization errors and node mobility. We assume the distance estimation error in CR–PU links to model location uncertainties and derive the distribution of channel gain with distance-dependent path loss and shadowing. We then proceed to develop an optimization framework for CR power control, which maximizes the CR data rate under PU interference power constraint. Simulation results showing the CR data rate and interference probability to the PUs are presented to demonstrate the superior performance of the proposed algorithm compared with reference schemes.