Description

Steps to reproduce the issue:

1. Prow CRIO periodic jobs

Describe the results you received:
Kubernetes NodeConformance and NodeFeatures jobs failing with latest CRIO

Describe the results you expected:
NodeConformance and NodeFeatures jobs were running fine before 7826110 was merged. That change bumps github.com/syndtr/gocapability v0.0.0-20180916011248-d98352740cb2 to github.com/syndtr/gocapability v0.0.0-20200815063812-42c35b437635.

This makes the prow test in upstream k8s fail with following error,

Jan 11 06:53:40.801: INFO: At 2021-01-11 06:48:43 +0000 UTC - event for pod-host-path-test: {kubelet tmp-node-e2e-ea1796c0-fedora-coreos-33-20201214-3-1-gcp-x86-64} Failed: Error: container create failed: time="2021-01-11T06:48:43Z" level=error msg="container_linux.go:370: starting container process caused: unknown capability \"CAP_PERFMON\""

It's nicely documented by Sebastiaan van Stijn at moby/moby#41563 on why we should keep the github.com/syndtr/gocapability pinned at v0.0.0-20180916011248-d98352740cb2 until the new runc is released.

I will send the PR to fix it.

Additional information you deem important (e.g. issue happens only occasionally):

Output of crio --version:

(paste your output here)

Additional environment details (AWS, VirtualBox, physical, etc.):