Update vfio-manager daemonset to use k8s-driver-manager image

assets/state-vfio-manager/0600_daemonset.yaml → assets/state-vfio-manager/0500_daemonset.yaml

@@ -62,20 +62,19 @@ spec:
           imagePullPolicy: IfNotPresent
           command: ["/bin/sh", "-c"]
           args:
-            - /bin/vfio-manage.sh bind --all && sleep inf
+            - vfio-manage bind --all && while true; do sleep 86400; done
           resources:
             limits:
               memory: 200Mi
             requests:
               cpu: 100m
               memory: 200Mi
           volumeMounts:
-          - name: nvidia-vfio-manager
-            readOnly: true
-            mountPath: /bin/vfio-manage.sh
-            subPath: vfio-manage.sh
           - name: host-sys
             mountPath: /sys
+          - name: host-lib-modules
+            mountPath: /lib/modules
+            readOnly: true
           - name: host-root
             mountPath: /host
           securityContext:
@@ -85,17 +84,17 @@ spec:
           lifecycle:
             preStop:
               exec:
-                command: ["/bin/sh", "-c", "/bin/vfio-manage.sh unbind --all"]
+                command: ["vfio-manage unbind --all"]
       terminationGracePeriodSeconds: 30
       volumes:
-        - name: nvidia-vfio-manager
-          configMap:
-            name: nvidia-vfio-manager
-            defaultMode: 448
         - name: host-sys
           hostPath:
             path: /sys
             type: Directory
+        - name: host-lib-modules
+          hostPath:
+            path: /lib/modules
+            type: Directory
         - name: run-nvidia
           hostPath:
             path: /run/nvidia

bundle/manifests/gpu-operator-certified.clusterserviceversion.yaml

@@ -224,8 +224,9 @@ spec:
       image: ghcr.io/nvidia/gpu-operator:main-latest
     - name: k8s-driver-manager-image
       image: nvcr.io/nvidia/cloud-native/k8s-driver-manager@sha256:c549346eb993fda62e9bf665aabaacc88abc06b0b24e69635427d4d71c2d5ed4
+    # TODO: update to a k8s-driver-manager image from nvcr.io once the next version is published
     - name: vfio-manager-image
-      image: nvcr.io/nvidia/cuda@sha256:d19fe621624c4eb6ac931b8558daa3ecc0c3f07f1e2a52e0267e083d22dceade
+      image: ghcr.io/nvidia/k8s-driver-manager:0d30fa9c
     - name: sandbox-device-plugin-image
       image: nvcr.io/nvidia/kubevirt-gpu-device-plugin@sha256:119de9a331a47203858b99901f44d0c4a8052961b4e60327f4b100d0ab8c9df0
     - name: vgpu-device-manager-image
@@ -940,8 +941,9 @@ spec:
                     value: "nvcr.io/nvidia/cloud-native/k8s-mig-manager@sha256:8e0803d2f29776cd4cc0501381a20a0b04b2da507a794d66a15894c57beaadb5"
                   - name: "CUDA_BASE_IMAGE"
                     value: "nvcr.io/nvidia/cuda@sha256:d19fe621624c4eb6ac931b8558daa3ecc0c3f07f1e2a52e0267e083d22dceade"
+                  # TODO: update to a k8s-driver-manager image from nvcr.io once the next version is published
                   - name: "VFIO_MANAGER_IMAGE"
-                    value: "nvcr.io/nvidia/cuda@sha256:d19fe621624c4eb6ac931b8558daa3ecc0c3f07f1e2a52e0267e083d22dceade"
+                    value: "ghcr.io/nvidia/k8s-driver-manager:0d30fa9c"
                   - name: "SANDBOX_DEVICE_PLUGIN_IMAGE"
                     value: "nvcr.io/nvidia/kubevirt-gpu-device-plugin@sha256:119de9a331a47203858b99901f44d0c4a8052961b4e60327f4b100d0ab8c9df0"
                   - name: "VGPU_DEVICE_MANAGER_IMAGE"

deployments/gpu-operator/values.yaml

@@ -445,9 +445,10 @@ vgpuDeviceManager:
 
 vfioManager:
   enabled: true
-  repository: nvcr.io/nvidia
-  image: cuda
-  version: 13.0.1-base-ubi9
+  # TODO: update to a k8s-driver-manager image from nvcr.io once the next version is published
+  repository: ghcr.io/nvidia
+  image: k8s-driver-manager
+  version: 0d30fa9c
   imagePullPolicy: IfNotPresent
   imagePullSecrets: []
   env: []