/kind bug

What happened:

During an update of the cinder-csi-nodeplugin DaemonSet one Pod of the cinder-csi-nodeplugin could not be started again (CreateContainerError) because on that Node there was a Pod running as well that contained double quotes in its mountPath.

A kubectl describe revealed following:

  Warning  Failed     11s                kubelet, kube-node03  Error: failed to generate container "3a0bd914c862e613de04175150aaee6e645983bdbbc9f0f7ecce6af5e4060d20" spec: parsing '/kubelet/pods/a236ebe7-40b7-48b5-8458-ff1321d0ee67/volumes/kubernetes.io~configmap/content/"bar"' failed: unable to unquote root field: invalid syntax

The above error comes directly from containerd in my case. See: https://github.com/containerd/containerd/blob/v1.3.4/mount/mountinfo_linux.go#L90

From a Kubernetes validation point of view using double quotes seems to be allowed, as the resource that uses it can be created without any problems. The cinder-csi-nodeplugin pod on that Node can only be started again when the Pod that references the double quote mount is being deleted.

What you expected to happen:

That the cinder-csi-nodeplugin handles this error more gracefully.

How to reproduce it:

• Have the cinder-csi-nodeplugin DaemonSet running
• Deploy follwing test resources:

  apiVersion: v1
  kind: ConfigMap
  metadata:
    name: content
  data:
    test: |-
      foo
  ---
  apiVersion: v1
  kind: Pod
  metadata:
    name: busybox
    labels:
      app: busybox
  spec:
    containers:
    - image: busybox
      command:
        - /bin/sh
        - -c
        - "tail -f /dev/null"
      imagePullPolicy: IfNotPresent
      name: busybox
      volumeMounts:
      - name: content
        mountPath: /tmp
      - name: content
        mountPath: '/tmp/"bar"'
        subPath: '"bar"'
    restartPolicy: Always
    volumes:
    - name: content
      configMap:
        name: content

• Delete the cinder-csi-nodeplugin Pod on the Node where the above Pod has been started
• After deletion the cinder-csi-nodeplugin tries to start again but remains in CreateContainerError state with the unquote error event mentioned above

Anything else we need to know?:

The cinder-csi-nodeplugin Pods is not affected while it is running when a resource is deployed with a double quote in its mountPath. Only when the cinder-csi-nodeplugin Pod is being deleted/updated the error appears.
The issue was observed in version v1.17.0 but could also be reproduced on version v.1.16.0.

Environment:

• openstack-cloud-controller-manager(or other related binary) version: v1.17.0