VirtIO–SCMI Based Controlled SoC Resource Access for Non-Root VMs #231
Description
Current hvisor passthrough works for MMIO/IRQ/IOMMU, but non-root VMs on SoCs (ARM/RISC-V) often fail device init due to lack of controlled access to platform resources (clock, reset, power).
Proposed Feature:
- Implement a VirtIO–SCMI forwarding mechanism: non-root VMs request platform resources via VirtIO, routed through hvisor to root Linux.
- Keep hvisor lightweight: only forward messages and enforce isolation.
- Root Linux handles actual resource operations via CCF, Reset Controller, genpd.
- Provide non-root VMs with secure, portable access to platform resources to enable proper device initialization.
Benefits:
- Improve device passthrough usability in embedded/edge SoCs.
- Keep hypervisor TCB minimal.
- Standardized, maintainable solution leveraging SCMI & VirtIO.
Notes / Considerations:
- Focus on end-to-end prototype first (RK3588 or similar SoC).
- Evaluate functional correctness, security isolation, and overhead.