With the increasing demands for spectral and energy efficiency, distributed multiple-input multiple-output (MIMO) networks assisted with reconfigurable intelligent surfaces (RISs) have attracted considerable attention. In this paper, we investigate the joint optimization of access point (AP) scheduling and precoding for RIS-aided distributed MIMO networks. To make this challenging problem more tractable, we decouple the joint optimization into two hierarchical subproblems, which eventually formulate a two-timescale scheme. To further reduce the computational complexity of joint precoding, we propose a hierarchical deep reinforcement learning (HDRL) framework to maximize the system sum spectral efficiency (SE), which leads to a distributed deployment with centralized training. Simulation results show that the proposed framework yields significant reduction in the computational complexity with slight performance loss, and strong generalizability against varying system parameters.