We consider a multi-way decode-and-forward (DF) relaying network with very large antenna arrays at the relay station. In this system, each user and the relay operate in half-duplex and time-division duplexing (TDD) modes. To exchange information among all users, we propose a new transmission protocol which combines massive multiple-input multiple-output (MIMO) technology with linear processing, self-interference cancelation, and successive cancelation decoding. Our proposed transmission protocol reduces the number of time-slots for data exchange among users by approximately 2 times, compared to the conventional data transmission protocol. For this new topology, we derive a very tight approximation of the spectral efficiency in closed-form assuming perfect channel state information (CSI). Then, CSI acquisition method at the relay and the users is provided and analysed. We show via numerical simulations, the performance gap between imperfect and perfect CSI cases is small. The closed-form expression of the spectral efficiency enables us to design two power allocation schemes. In the first power allocation scheme, we choose the transmit powers at the users and the relay to maximize the sum spectral efficiency, subject to a given quality-of-service requirement for each user. In the second power allocation scheme, the objective is the energy efficiency taking into account the hardware power consumption. Both power allocation schemes can be efficiently executed by iteratively solving a sequence of convex problems. Numerical results verify the effectiveness of the proposed transmission protocol and the power allocation schemes compared to the state-of-the-art.