In this paper, we present SURF, a distributed channel selection strategy for efficient data dissemination in multi-hop cognitive radio ad-hoc networks (CRNs). SURF classifies the available channels on the basis of primary radio unoccupancy and the number of cognitive radio neighbors using the channels. Through extensive NS-2 simulations, we compare the performance of SURF with three related approaches. Simulation results confirm that SURF is effective in selecting the best channels for efficient communication and for highest dissemination reachability in multi-hop CRNs.