This is a clone of issue OCPBUGS-39226. The following is the description of the original issue:
—
Description of problem:

Version-Release number of selected component (if applicable):

How reproducible: Always

Repro Steps:

Add: "bridge=br0:enpf0,enpf2 ip=br0:dhcp" to dracut cmdline. Make sure either enpf0/enpf2 is the primary network of the cluster subnet.

The linux bridge can be configured to add a virtual switch between one or many ports. This can be done by a simple machine config that adds:
"bridge=br0:enpf0,enpf2 ip=br0:dhcp"
to the the kernel command line options which will be processed by dracut.

The use case of adding such a virtual bridge for simple IEEE802.1 switching is to support PCIe devices that act as co-processors in a baremetal server. For example:
-------- ---------------------

-------- ---------------------
This co-processor could be a "DPU" network interface card. Thus the co-processor can be part of the same underlay network as the cluster and pods can be scheduled on the Host and the Co-processor. This allows for pods to be offloaded to the co-processor for scaling workloads.

Actual results:

ovs-configuration service fails.

Expected results:

ovs-configuration service passes with the bridge interface added to the ovs bridge.