When the customer applied a sriov policy that configures VFs on the PF that is also used as the primary networking interface and the MTU that is set on the VF, that will also affect the PF (max of the MTU of all the VFs for that PF will be used at the PF level). If the max MTU of the VFs is lower than the MTU of the cluster, network connectivity will be disrupted. We should not allow the user to set a bad MTU when applying policy.

Since it's not immediately obvious how to detect cluster MTU outside of OCP, we will opt to only allow to increase MTU from what's the default on the PF. This way, we remain in a working state. This is in contrast to what is allowed today, i.e. if the MTU _was_ 9k, and the customer sets 8k, we allow it while we know that this is actually risky.

4.12 but older too

always

1. create SRIOV policy with VFs on PF that is used for primary network but use a lower MTU what's on SDN PF
2. Wait for SRIOV to be applied

connectivity will be disrupted

Sriov policy should be rejected

https://access.redhat.com/support/cases/#/case/03552513