What runs here
A type-1 hypervisor controls hardware resources beneath guest operating systems. On this Cortex-A-class application processor, it creates virtual machines (VMs) for workloads with different timing, safety and security needs. A VM might run Linux, QNX or Android; the actual combination depends on the vehicle architecture.
| Component | Role |
|---|---|
| Hypervisor core | Creates VMs and controls privileged access to processor resources. |
| Resource configuration | Assigns cores or processor time, memory, interrupts and devices to each VM. |
| Memory and DMA isolation | Restricts memory access by guest processors and DMA-capable devices. |
| Virtual I/O support | Connects guest drivers to shared devices through controlled backends. |
| Lifecycle and fault handling | Coordinates VM startup, shutdown, fault reporting and permitted recovery. |
Some device backends and management functions run in privileged service VMs. Their placement matters: a service VM can become a dependency of every guest that uses its devices.
Why it looks like this
An HPC combines substantial computing capacity with shared memory, accelerators and high-speed I/O. Virtualisation allows workloads needing different operating systems to share that silicon. Consolidation can reduce separate hardware, but introduces integration, validation and resource-management costs.
Hardware support makes this practical. Arm stage-2 address translation lets the hypervisor restrict the physical memory accessible to a guest. An SMMU, where implemented and correctly connected, can restrict device DMA. Processor memory protection alone does not constrain every device access. Arm’s virtualisation guide explains these mechanisms.
Safety requires more than separate address spaces. Dedicated cores can simplify scheduling analysis, yet caches, memory controllers, interconnects, interrupts, power management and accelerators may remain shared. Contention can delay a safety-related workload even when no memory boundary is crossed.
The grid’s QM–ASIL B scope describes the intended workloads; it does not automatically qualify the hypervisor or complete system. Safety requirements determine which mechanisms and dependencies need assurance. Software architectural design and verification belong within the development process described by ISO 26262-6:2018.
Different update frequencies also motivate VM boundaries. Frequently updated HMI software can have a separate deployment lifecycle from a safety-related service, provided shared interfaces, resource assumptions and platform dependencies remain compatible.
How it connects
Above this layer, each guest OS schedules its applications and hosts middleware such as AUTOSAR Adaptive, SOME/IP or DDS. The hypervisor provides virtual hardware and isolation mechanisms; it does not implement the application service architecture.
Below it, the hypervisor uses processor virtualisation extensions, interrupt-controller facilities, memory translation and available device-isolation hardware. Trusted boot firmware establishes the initial execution chain and configuration. A trusted execution environment has a different role and does not replace VM partitioning.
Devices may be assigned directly to a guest or accessed through a backend. VirtIO standardises interfaces between guest drivers and virtual devices; it does not itself establish safety assurance or bounded latency. The OASIS VirtIO specification defines these interfaces.
VMs communicate through explicitly configured shared memory, virtual networking or other controlled channels. Traffic to other vehicle nodes passes through assigned or mediated network interfaces. Safety-related communication still needs appropriate integrity, freshness and timeout mechanisms. Communication with the safety island likewise needs a defined protocol, supervision and fault response.
Compared with other node classes
- HPC safety island: Physically separate MCU cores provide a distinct execution boundary, although shared power, interconnects or peripherals can still create dependencies.
- Zone controller and MCU-based ECU: A memory protection unit (MPU) typically protects tasks or OS applications within an RTOS. This differs from the application processor’s MMU-based virtual machines and separate guest kernels.
- Smart node / LIN slave: This grid allocates no virtualisation layer; small firmware typically runs directly on the controller or a compact scheduler.
Best practices
- Define ownership of every core, memory region, interrupt and device, including DMA paths and reset controls.
- Establish timing budgets and measure interference under concurrent CPU, memory, network and accelerator load.
- Keep privileged code and service dependencies small; include required backends in safety and security analyses.
- Specify communication permissions, buffer limits and failure behaviour at every VM boundary.
- Validate guest restart, backend failure and device reset without assuming neighbouring VMs remain unaffected.
- Treat hypervisor configuration, guest images and shared interfaces as a versioned platform; reassess isolation assumptions after updates.
Common pitfalls
- Treating separate VMs or dedicated cores as proof of freedom from interference.
- Passing through a device without checking DMA isolation, interrupt routing and reset scope.
- Assuming a qualified hypervisor automatically qualifies guest software or the integrated platform.
- Allowing a shared driver VM to become an unanalysed source of delay or common failure.