Hardware Configuration

All configuration files for reference target devices are in modules/reference/hardware, and the machine table that turns them into build targets is targets/laptop/machines.nix.

x86_64 Platforms

Intel laptops — the generic image

Intel laptops share one image. It passes display, network and audio through by PCI class rather than per-device ID, matches input devices by udev property, and carries per-board quirks (IOMMU ACS ids, suspend mode, webcam and fingerprint VID:PIDs) in union lists that one image applies to the whole fleet. There is no per-model target, and a new Intel laptop does not need one.

• Targets: intel-laptop-debug / intel-laptop-release
• Installers: intel-laptop-debug-installer / intel-laptop-debug-netboot-installer
• A/B update image: intel-laptop-debug-sysupdate / intel-laptop-release-sysupdate
• Verified on: Lenovo ThinkPad X1 Carbon, Dell Latitude 7330 Rugged, System76 Darter Pro
• Architecture: x86_64

Two feature axes vary the image; a target name gains one suffix per selected axis. The memory axes are mutually exclusive.

Axis	Suffix	Effect
low-mem	-low-mem	For machines with 16 GB or less: shrinks the flatpak app VM
minimal-mem	-minimal-mem	Tablet-class units: shrinks the GUI VM to just under 2 GB
storeDisk	-storeDisk	Puts the VM Nix store on disk instead of in memory

So intel-laptop-low-mem-debug is the Dell Latitude 7330 build, and intel-laptop-low-mem-storeDisk-release combines both.

Intel laptop, extended application set

• Targets: intel-laptop-extras-debug / intel-laptop-extras-release
• Description: The same image with the mvp-user-trial-extras product profile
• Architecture: x86_64

Boards the generic image cannot describe

These keep their own target because they are not Intel laptops: AMD IOMMU, desktop chassis, or a discrete GPU whose audio function must be paired with the display device by hand rather than swept up by PCI class matching.

Alienware m18 R2

• Targets: alienware-m18-R2-debug / alienware-m18-R2-release
• Why separate: NVIDIA dGPU passed as a static multi-function pair, out-of-tree NIC driver in net-vm, host-owned panel backlight, and suspend disabled because the board does not resume
• Architecture: x86_64

Lenovo ThinkPad T14 AMD Gen 5

• Targets: lenovo-t14-amd-gen5-debug / lenovo-t14-amd-gen5-release
• Why separate: amd_iommu, amdgpu, AMD ACP audio, and a simpledrm renderer because iGPU passthrough does not work on this board
• Architecture: x86_64

Demo Tower MK1

• Targets: demo-tower-mk1-debug / demo-tower-mk1-release
• Why separate: AMD desktop; discrete Quadro passed through as a four-function group that includes its own xHCI controller
• Architecture: x86_64

Tower 5080

• Targets: tower-5080-debug / tower-5080-release
• Why separate: Intel desktop; discrete RTX 5080 needing the open NVIDIA drivers and NVIDIA modules in the guest initrd
• Architecture: x86_64

Virtualization Platforms

VM Debug

• Targets: vm-debug / vm-debug-nogui / vm-release
• Description: Virtual machine configurations for testing and development
• Architecture: x86_64

VMware

• Target: vmware-debug
• Description: VMware-optimized configuration
• Architecture: x86_64

AArch64 (ARM64) Platforms

NVIDIA Jetson Platform (Primary AArch64)

NVIDIA Jetson AGX Orin (32GB)

• Native Targets: nvidia-jetson-orin-agx-debug / nvidia-jetson-orin-agx-release
• Cross-compiled: nvidia-jetson-orin-agx-debug-from-x86_64 / nvidia-jetson-orin-agx-release-from-x86_64
• No Demo Apps: nvidia-jetson-orin-agx-debug-nodemoapps / nvidia-jetson-orin-agx-release-nodemoapps
• Cross + No Demo: nvidia-jetson-orin-agx-debug-nodemoapps-from-x86_64 / nvidia-jetson-orin-agx-release-nodemoapps-from-x86_64
• Architecture: AArch64
• Status: Primary AArch64 reference platform

NVIDIA Jetson AGX Orin (64GB)

• Native Targets: nvidia-jetson-orin-agx64-debug / nvidia-jetson-orin-agx64-release
• Cross-compiled: nvidia-jetson-orin-agx64-debug-from-x86_64 / nvidia-jetson-orin-agx64-release-from-x86_64
• No Demo Apps: nvidia-jetson-orin-agx64-debug-nodemoapps / nvidia-jetson-orin-agx64-release-nodemoapps
• Cross + No Demo: nvidia-jetson-orin-agx64-debug-nodemoapps-from-x86_64 / nvidia-jetson-orin-agx64-release-nodemoapps-from-x86_64
• Architecture: AArch64

NVIDIA Jetson Orin NX

• Native Targets: nvidia-jetson-orin-nx-debug / nvidia-jetson-orin-nx-release
• Cross-compiled: nvidia-jetson-orin-nx-debug-from-x86_64 / nvidia-jetson-orin-nx-release-from-x86_64
• No Demo Apps: nvidia-jetson-orin-nx-debug-nodemoapps / nvidia-jetson-orin-nx-release-nodemoapps
• Cross + No Demo: nvidia-jetson-orin-nx-debug-nodemoapps-from-x86_64 / nvidia-jetson-orin-nx-release-nodemoapps-from-x86_64
• Architecture: AArch64

NXP Platform

NXP i.MX 8M Plus EVK

• Native Targets: nxp-imx8mp-evk-debug / nxp-imx8mp-evk-release
• Cross-compiled: nxp-imx8mp-evk-debug-from-x86_64 / nxp-imx8mp-evk-release-from-x86_64
• Architecture: AArch64
• Use Case: Industrial and embedded applications

Utility Configurations

Hardware Scanning

• Target: laptop-hw-scan
• Description: Hardware compatibility scanning utility
• Architecture: x86_64

Build Recommendations

Primary Targets for Development

• x86_64: intel-laptop-debug (the generic laptop image, and the one the hardware rigs run)
• AArch64: nvidia-jetson-orin-agx-debug-from-x86_64 (primary ARM64 platform)

Cross-Compilation

All AArch64 targets have corresponding -from-x86_64 variants for cross-compilation from x86_64 development machines.

Installer Images

Every laptop target has an -installer variant (a bootable ISO carrying the image) and a -netboot-installer variant (PXE, fetching the image at install time).

Adding New Hardware

An Intel laptop

Usually nothing to add — build intel-laptop-debug and flash it. If something on the board does not come up, add the one quirk it needs to the list that already covers it, rather than creating a target:

Symptom	Where the quirk goes
A device shares an IOMMU group with another and cannot be split	ghaf.hardware.passthrough.pciAcsOverride.ids in modules/reference/hardware/flake-module.nix
Suspend does not resume, but s2idle works	ghaf.services.power-manager.suspend.s2idleModels in modules/reference/hardware/intel-laptop/extra-config-host.nix — a DMI "<board_vendor> <board_name>" string
The internal webcam or fingerprint reader is not passed through	modules/reference/passthrough/known-devices.nix — a VID:PID entry
A vendor hotkey block does not reach the GUI VM	modules/hardware/passthrough/evdev-rules.nix — a platform pathTag
A host driver claims a device before vfio-pci	the host module_blacklist in modules/reference/hardware/intel-laptop/intel-laptop.nix

The unit needs less memory than the fleet default? Use an existing axis (low-mem, minimal-mem) rather than a new target.

A board the generic image cannot describe

AMD hosts, desktops, and discrete GPUs still need their own module — the generic image is Intel-only, and PCI class matching cannot pair a dGPU with its audio function.

1. Run laptop-hw-scan on the machine to dump its PCI, USB and DMI data.
2. Add a module directory under modules/reference/hardware and register it as hardware-<name> in that directory’s flake-module.nix. Use tower-5080 as the reference for a discrete-GPU board.
3. Add one stanza to targets/laptop/machines.nix. That file is data only — no ghaf.*, no lib calls, no conditionals; everything else belongs in the module. This is enforced by checks.laptop-table-is-data.