Abstract vGPU devices behind a framework dispatch

[yummybomb]

Summary

Bottom layer of the vendor VFIO vGPU stack (this ← #322 ← #363 ← #364 ← #321). Behavior-preserving refactor only — no lifecycle semantics change in this layer.

Kernel 6.8 hosts assign NVIDIA vGPUs through a vendor-specific VFIO interface instead of mdev, so the mdev-shaped seams get generalized before the new backend lands above:

• VGPUDevice / VGPUFramework abstraction — mdev moves behind a framework dispatch (CreateVGPU / DestroyVGPU / DiscoverVGPU), with assignments described by a framework + device path instead of a bare mdev UUID.
• Hypervisor selection — vGPU instances require QEMU; the policy moves to callers instead of being implied by mdev.
• Metadata — VF allocation tracked with a single field; instance metadata carries GPUFramework / GPUDevicePath alongside the mdev UUID.
• QEMU args — vGPU attaches via sysfsdev, and the dead mdev branch is dropped from PCI passthrough args.

Testing

• go build ./..., go vet clean
• go test -race ./lib/devices/ ./lib/hypervisor/... and targeted lib/instances suites pass (TestSocketCacheKeyChangesWhenSocketIsRecreated and the network/image-dependent instances tests fail identically on the unmodified stack head in this environment)

---

Note

Medium Risk
Changes GPU allocation, teardown, and how VMs attach vGPU devices across instance lifecycle and QEMU/CH config, though current behavior still routes through mdev on Linux.

Overview
Introduces a framework-agnostic vGPU layer so instance lifecycle and hypervisors no longer talk to mdev directly. Linux still creates/destroys mdev via CreateVGPU / DestroyVGPU; macOS stubs return the existing unsupported error.

Devices & metadata: Adds VGPUDevice, VGPUFramework, and renames VF has_mdev → allocated. Instance metadata now stores GPUFramework and GPUDevicePath (with GPUMdevUUID kept for mdev); helpers in vgpu.go centralize set/clear/release and legacy path resolution from UUID-only records.

VM attachment: hypervisor.VMConfig gains VGPUDevicePath, separate from whole-GPU PCIDevices. QEMU always attaches vGPU with vfio-pci,sysfsdev=… from that field and no longer treats mdev paths inside the PCI passthrough loop. Cloud Hypervisor appends the vGPU sysfs path to its device list the same way.

Lifecycle: Create, start, stop, and delete call the new APIs and releaseStoredVGPU instead of inline CreateMdev / DestroyMdev.

^{Reviewed by Cursor Bugbot for commit 4499016. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit f677357. Configure here.}