By imposing additional constraints in the downlink beamforming optimization. more general Quality of Service (QoS) measures than the average Signal to Interference and Noise Ratio (SINR) alone, can be introduced. Herein a rapidly converging algorithm solving the downlink beamforming problem with additional indefinite quadratic constraints on the beamforming vector is presented. The proposed algorithm is significantly faster than the previously proposed solution, which involves semidefinite programming. Also, the algorithm is easy to implement, since it only involves eigenvalue problems.