Integrate vendor VFIO vGPUs into the instance lifecycle

[yummybomb]

Summary

Top layer of the vendor VFIO vGPU stack (generalize-vgpu-device #322 ← hypervisor-liveness #363 ← vendor-vfio-backend #364 ← this). The backend itself is in #364; this layer wires it into the instance lifecycle and adds the release guards that make reusable-VF assignments safe:

• Live-instance claim guard on release — an assignment goes stale when its release succeeds but the metadata save does not (or start fails between the release and its first save). Since vendor VFIO reuses the same VF path across assignments, honoring a stale claim could clear a VF out from under the live instance that owns it now. Every release consults live instance metadata (liveness verified via Harden hypervisor process liveness checks #363's socket-ownership check) and drops stale metadata without touching the device.
• Assignment durability — assignments are tagged with the owning instance ID, persisted before booting a started instance without carrying over the previous VMM identity, and retained in metadata when rollback release fails in create/start so later release paths can still find the device.
• Fail-closed startup reconciliation — the protected set of live VFs is built from ListInstancesForReconcile, which fails on any unreadable metadata instead of silently skipping it. Recent assignments receive bounded startup protection when their PID is absent or stale; when the inventory is unavailable, vendor VFIO reconciliation is skipped entirely while mdev reconciliation still runs.
• Hypervisor scope — hypervisor selection remains a caller policy, preserving the framework layer's existing mdev behavior. The production rollout continues to place vGPU instances on QEMU; the Cloud Hypervisor vendor-VFIO limitation (upstream cloud-hypervisor#7572) is documented in lib/devices/GPU.md.
• Failure reporting — a failed create or start whose vGPU release also failed surfaces as vgpu_cleanup_pending with retained/unretained guidance; the retention record keeps identity fields (name, image, hypervisor) so it lists as a recognizable, deletable instance. The instance API exposes gpu.device_path, the assignment identity on vendor VFIO hosts where mdev_uuid is empty.

Testing

• go build ./..., go vet ./... clean
• go test -race passes for lib/instances targeted suites, lib/devices, lib/resources, lib/builds (TestCreateInstanceWithNetwork needs image pulls + iptables and fails in this environment on the unmodified base as well)

---

Note

High Risk
Changes core GPU assignment, release, and startup reconciliation with fail-closed behavior when metadata is ambiguous—mistakes could release devices in use or leave orphaned VF assignments.

Overview
Wires vendor VFIO vGPUs into create/start/stop/delete and startup reconciliation, replacing the previous hard block on that framework in CreateVGPU.

Lifecycle and safety: Assignments are persisted with GPUAssignedAt, cleared hypervisor identity on start before re-assigning, and retained in metadata when rollback cannot destroy the device (cleanupFailedCreate, start rollback). Failed create/start surfaces vgpu_cleanup_pending (500) with retained vs reconcile guidance; errors.As on VGPUCleanupPendingError is checked before wrapped causes. Vendor VFIO release scans instance metadata for live claimants on the same sysfs path (with startup grace when PID is missing/stale); mdev releases skip that scan. Startup reconcileVGPUs builds a protected path set from ListInstancesForReconcile (strict metadata load) and reschedules while grace remains.

API: Instance gpu responses add optional device_path alongside profile/mdev UUID.

Tests: Broad coverage for retention, claim guards, reconcile protection, and API error mapping; integration vGPU tests no longer skip vendor VFIO.

^{Reviewed by Cursor Bugbot for commit 3cc1cfd. Bugbot is set up for automated code reviews on this repo. Configure here.}

[yummybomb]

Trimmed over-engineering after an architectural review against the parent task (host kernel 6.8 / Ubuntu 24.04 support). Most of the removed weight came from earlier review-round additions rather than the original design: