Our testpmd_DPDK application drops RX packets, even at very low line rates.  After digging in a bit deeper, it looks like when spawned via crun the process is getting cpu.cfs_quota_us = 60000 whereas with runc it is getting the expected value of cpu.cfs_quota_us = -1  - This is the same pod definition and the same OCP install, the only thing we're doing is toggling crun vs runc. 

OpenShift 4.14.0-rc.4

100% of the time

1. Deploy OpenShift with crun and the telco DU PerformanceProfile and SRIOV configured
2. Deploy the pod as specified in the attached 'dpdk-testpmd.yaml' which has 'cpu-load-balancing.crio.io: disable' and 'cpu-quota.crio.io: disable'
3. Check the cpu.cfs_quota_us for the resulting process for example: 'cat /sys/fs/cgroup/cpu,cpuacct//kubepods.slice/kubepods-pod51723ec0_d2f3_484f_941b_f9fc53de17b7.slice/crio-e1a05a6f6d566ebc68948742cde62d591584f9befe232f85510b1bd06f35ed24.scope/[container/]cpu.cfs_quota_us'

60000

-1

- Switching back to runc solves the problem- quota and balancing are acting as expected there.
- We also noticed load-balancing is not set correctly with crun
- We're working on a simpler reproducer that doesn't require SRIOV or TestPMD