How udosteinberg/nova works
Project Type and Purpose
• NOVA is a microhypervisor combining microkernel and hypervisor functionality.
• It provides a minimal trusted computing base (TCB) for user‐mode servers and guest VMs, implements capability‐based authorization, scheduling, IPC, and basic virtualization on ARMv8‐A and x86_64 platforms.
File Structure Analysis
• Top‐level:
• cmd/ — utility code (digest, CLI tools)
• inc/ — architecture‐independent and arch‐specific headers (aarch64, x86_64)
• src/ — implementation files, split into arch‐specific subdirs and common code
• doc/ — interface specification (boot protocol, API)
• Makefile, .gitlab-ci.yml — build and CI configuration
• Architecture modules: each arch subdirectory under inc/ and src/ contains low-level code for CPU bring-up, interrupt controllers (GIC, APIC), system tables (ACPI, FDT), memory management, virtualization extensions (VMX, SVM, PSCI).
README Insights
• Describes target platforms (ARMv8-A ACPI/FDT, x86 VT-x/VT-d, AMD-V)
• Explains capability model, multi‐server design, build variants (CFP, TXT), boot process (via UEFI or multiboot).
Main Components
a. Hardware Layer
• CPU(s) with virtualization extensions (EL2 on ARM, VMX/SVM on x86)
• Memory bus, interrupt controller (GIC/APIC), platform firmware (ACPI tables or FDT)
b. NOVA Microhypervisor Core (EL2 / Ring ‑1)
• Bootloader/U-EFI entry
• Kernel entry (arch-specific bootstrap)
• Memory Management (paging, EPT/NPT)
• Scheduler (temporal separation)
• Capability Manager (spatial separation)
• Interrupt handling (GIC driver / APIC driver)
• Hypercall interface
• ACPI/FDT parser
c. User‐Mode Servers (EL1 / Ring 0)
• Device drivers, protocol stacks, policy servers
• IPC channels to core and other servers
d. Guest Operating Systems (EL1 / Ring 1)
• Unmodified VMs running on top of the hypervisor
• Virtual devices mediated by the microhypervisor
e. Tools and Build System
• Cross‐toolchains (PREFIX_aarch64, PREFIX_x86_64) and build targets (Makefile)
• cmd/ utilities for image creation and signing
Relationships and Interactions
• Core ↔ Firmware tables (ACPI/FDT) for device discovery
• Core ↔ Memory → Guest VMs (EPT/NPT mappings)
• Core ↔ Interrupt Controller → dispatch to VMs or servers
• Core ↔ User Servers via IPC (capability‐protected endpoints)
• Guest VMs ↔ Virtual Devices mediated by Core
• cmd tools → build images → deployed to hardware/QEMU
Architectural Patterns and Technologies
• Microkernel pattern: separation of minimal core vs. user‐mode servers
• Capability‐based access control: fine‐grained resource rights
• Virtualization extensions: VMX/EPT on x86, PSCI on ARM
• Multi‐architecture code sharing: common core + arch plugins
• Build variants: CFP (CET IBT/SSS), TXT measured launch
Detailed Capture of Architectural Elements
• Memory regions: hypervisor, user‐servers, guest RAM
• Scheduling domain and capability manager as distinct core services
• Virtual device channels: VM → hypervisor → physical device via server