This article proposes a novel QoS multicast routing (QSMR) protocol that leverages cross-layer design for IoT-enabled mobile ad hoc networks (MANETs) within a reconfigurable intelligent surface (RIS)-aided cell-free massive MIMO (CF-mMIMO) environment, especially when multiple active eavesdroppers (Eves) are present. The proposed QSMR protocol employs a cross-layer approach to establish a QoS multicast mesh (QMM) route, considering spectrum efficiency (SE), achievable secrecy rate (ASR), end-to-end (E2E) delay, route stability, and hop count requirements. Initially, we formulate the SE expressions for each IoT device in both uplink and downlink, and the downlink ASR expressions for users under attack. Second, we propose two optimization problems: 1) the phase-shift control (PSC) problem, aimed at maximizing the minimum uplink SE and 2) the power control (PC) problem, focused on maximizing the minimum ASR. Additionally, a deep neural network is designed to reduce the computation time required for solving the PSC and PC problems. Third, leveraging the cross-layer synergy, we propose the QSMR protocol to establish the QMM route from a source to multiple destinations by seamlessly integrating information from both the physical and network layers. Finally, the simulation results evaluate the secure performance and provide comparisons for different scenarios. Moreover, in the presence of Eves, the proposed QSMR protocol outperforms other routing protocols in terms of routing delay, control overhead, and packet delivery ratio.