mirror of
https://github.com/kata-containers/kata-containers.git
synced 2026-07-16 17:04:14 +00:00
62a398e56c
Add qemu-nvidia-cpu and qemu-nvidia-cpu-runtime-rs to the shared hypervisor allow-lists and to the docker, nerdctl and free-runner k8s matrices, so the CPU-only NVIDIA classes get the same smoke coverage as the other qemu variants on x86_64. Add run-k8s-tests-on-nvidia-cpu-arm64.yaml and wire it into ci.yaml: it runs the k8s test suite for qemu-nvidia-cpu and qemu-nvidia-cpu-runtime-rs on real arm64 (GH200) hardware, giving the CPU-only classes aarch64 coverage without disturbing the existing GPU arm64 CI. The NVIDIA NUMA bats and the nv-test allow-list learn about the CPU-only classes too, so they can be run by hand on a multi-node host, but the arm64 CI deliberately runs only the standard k8s suite: the GH200 it uses exposes a single CPU-bearing NUMA node, so the multi-NUMA tests would merely skip. While here, add an is_verity_enabled_runtime_class helper (the CPU-only classes boot the verity-backed nvidia base image without being confidential) and an is_nvidia_hypervisor helper, and use them in k8s-measured-rootfs.bats and the kata-deploy Helm arch wiring. Signed-off-by: Fabiano Fidêncio <ffidencio@nvidia.com> Assisted-by: Cursor <cursoragent@cursor.com>