RISC-V Summit

Today’s embedded industries exhibit a strong trend towards consolidation. Virtualization is the de facto technology used to achieve this goal, since (i) it provides strong isolation between partitions and (ii) eases the porting and integration engineering costs. However, a single hypervisor might not be suited for the heterogeneous requirements of different partitions. This results in the selection of feature-rich hypervisors with large TCBs, typically consisting of the system’s single point of failure. In this talk, we share our experience while implementing the latest RISC-V (i) Hypervisor and (ii) enhanced PMP (Smepmp) extensions on the RISC-V CVA6 core, and we explain how these extensions can be synergically leveraged to implement a novel heterogeneous partitioning and virtualization architecture with enhanced security and safety guarantees. In this architecture, a hardened implementation of OpenSBI domains is used to partition system resources (i.e., memory, peripherals, interrupts), enabling the deployment of distinct hypervisors per partition, according to its requirements. We end with an illustration of possible use cases and a discussion on how to orchestrate such systems.