In this paper, we consider a single-cell multi-user orthogonal frequency division multiple access (OFDMA) network with one unmanned aerial vehicle (UAV), which works as an amplify-and-forward relay to improve the quality-of-service (QoS) of the user equipments (UEs) in the cell edge. Aiming to improve the throughput while guaranteeing the user fairness, we jointly optimize the communication mode, subchannel allocation, power allocation, and UAV trajectory, which is an NP-hard problem. To design the UAV trajectory and resource allocation efficiently, we first decompose the problem into three subproblems, i.e., mode selection and subchannel allocation, trajectory optimization, and power allocation, and then solve these subproblems iteratively. Simulation results show that the proposed algorithm outperforms the random algorithm and the cellular scheme.