podvm: Add nvidia gpu support for mkosi image

WIP

Signed-off-by: Pradipta Banerjee <[email protected]>

Author: bpradipt

src/cloud-api-adaptor/podvm-mkosi/mkosi.postinst

@@ -11,7 +11,7 @@ mv "${BUILDROOT}/etc/issue.d" "${BUILDROOT}/usr/lib/issue.d" || true
   echo "IMAGE_ID=\"podvm\""
   echo "IMAGE_VERSION=\"${IMAGE_VERSION-v0.0.0}\""
   echo "VARIANT_ID=\"${VARIANT_ID}\""
-} >> "${BUILDROOT}/etc/os-release"
+} >>"${BUILDROOT}/etc/os-release"
 
 # remove unused units of legacy image
 rm -f "${BUILDROOT}"/etc/systemd/system/{run-image,run-kata\x2dcontainers}.mount
@@ -25,3 +25,8 @@ ln -s /dev/null "${BUILDROOT}/etc/systemd/system/[email protected]"
 ln -s /dev/null "${BUILDROOT}/etc/systemd/system/systemd-pcrphase-initrd.service"
 ln -s /dev/null "${BUILDROOT}/etc/systemd/system/systemd-pcrphase-sysinit.service"
 ln -s /dev/null "${BUILDROOT}/etc/systemd/system/systemd-pcrphase.service"
+
+# Update NVIDIA config
+#
+echo -e "blacklist nouveau\nblacklist nova_core" >/etc/modprobe.d/blacklist_nv_alt.conf
+#

src/cloud-api-adaptor/podvm-mkosi/mkosi.presets/system/mkosi.conf.d/fedora-gpu.conf

@@ -0,0 +1,9 @@
+[Match]
+Distribution=fedora
+
+[Content]
+Packages=
+nvidia-container-toolkit
+# 570.172.08 or newer is required due to: https://github.com/NVIDIA/open-gpu-kernel-modules/issues/893
+nvidia-driver-cuda-570.172.08
+kmod-nvidia-open-dkms-570.172.08

src/cloud-api-adaptor/podvm-mkosi/mkosi.skeleton/etc/yum.repos.d/cuda-fedora.repo

@@ -0,0 +1,6 @@
+[cuda-fedora41-x86_64]
+name=cuda-fedora41-x86_64
+baseurl=https://developer.download.nvidia.com/compute/cuda/repos/fedora41/x86_64
+enabled=1
+gpgcheck=1
+gpgkey=https://developer.download.nvidia.com/compute/cuda/repos/fedora41/x86_64/D42D0685.pub

src/cloud-api-adaptor/podvm-mkosi/mkosi.skeleton/usr/lib/systemd/system-preset/30-coco.preset

@@ -10,3 +10,5 @@ enable setup-nat-for-imds.service
 enable gen-issue.service
 enable image-env.service
 enable scratch-storage.service
+
+enable nvidia-cdi.service

src/cloud-api-adaptor/podvm/files/etc/systemd/system/multi-user.target.wants/nvidia-cdi.service

@@ -0,0 +1 @@
+../nvidia-cdi.service

src/cloud-api-adaptor/podvm/files/etc/systemd/system/nvidia-cdi.service

@@ -0,0 +1,11 @@
+[Unit]
+Description=Generate NVIDIA CDI Configuration
+Before=kata-agent.service
+
+[Service]
+Type=oneshot
+ExecStart=/usr/local/bin/generate-nvidia-cdi.sh
+RemainAfterExit=true
+
+[Install]
+WantedBy=multi-user.target

src/cloud-api-adaptor/podvm/files/usr/local/bin/generate-nvidia-cdi.sh

@@ -0,0 +1,10 @@
+#!/bin/bash
+
+#load drivers
+nvidia-ctk -d system create-device-nodes --control-devices --load-kernel-modules
+
+nvidia-persistenced
+# set confidential compute to ready state
+nvidia-smi conf-compute -srs 1
+# Generate NVIDIA CDI configuration
+nvidia-ctk cdi generate --output=/var/run/cdi/nvidia.yaml > /var/log/nvidia-cdi-gen.log 2>&1