Channel state information (CSI) is important to reap the full benefits of millimeter wave (mmWave) massive multiple-input multiple-output (MIMO) systems. The traditional channel estimation methods using pilot frames (PF) lead to excessive overhead. To reduce the demand for PF, data frames (DF) can be adopted for joint channel estimation and data recovery. However, the computational complexity of the DF-based methods is prohibitively high. To reduce the computational complexity, we propose a joint channel estimation and data recovery (JCD) method assisted by a small number of PF for mmWave massive MIMO systems. The proposed method has two stages. In Stage 1, differing from the traditional PF-based methods, the proposed PF-assisted method is utilized to capture the angle of arrival (AoA) of principal components (PC) of channels. In Stage 2, JCD is designed for parallel implementation based on the multi-user decoupling strategy. The theoretical analysis demonstrates that the PF-assisted JCD method can achieve equivalent performance to the Bayesian-optimal DF-based method, while greatly reducing the computational complexity. Simulation results are also presented to validate the analytical results.