Red Hat Enterprise Virtualisation RHEV issue with Cisco UCS 5108 tag VLAN 0 ?

[Dieselboy]

Long story short, we use Red Hat Virtualisation. We've upgraded our system and the hypervisors now run RHEL7. We notice that one group of VMs which are running on a specific VLAN lose their network connectivity. The affected VMs are RHEL and without a gnome desktop. Gnome desktop resolves the network issue but I don't know anymore as to why. The RHEL systems affected are with <2.6.500 kernels.
Anyway, packet captures on hypervisor shows the broken packets being tagged with VLAN 0.

Traffic comes in from the network and leaves UCS towards the hypervisor untagged / native so the hypervisor can see these packets. For some reason the RHEL7 hypervisor gets these untagged and places 802.1Q VLAN0 on them.

I've never seen VLAN 0 before in my 11 years of doing networks. Is VLAN0 even a legal tag? I understood even VLAN 1 should mean no 802.1Q tag.

Red Hat KB article surrounding this: https://access.redhat.com/solutions/1617343

Cisco Bug: https://bst.cloudapps.cisco.com/bugsearch/bug/CSCuu29425

Cisco state it should be fixed in RHEL 7.2. It's not fixed in RHEL7.2 hypervisor appliance ISO.

Because of how this is broken I reconfigured UCS to tag that VLAN towards the hypervisor so that it (the correct VLAN tag) must be honored. This does indeed fix the issue.

Red Hat are blaming Cisco
Cisco are blaming Red Hat it seems.
I'm going with Cisco on this one

[wintermute000]

TIL VLAN0 is reserved for QOS functionality and is expected to be rewritten by the access switch. Woah!

http://networkengineering.stackexchange.com/questions/24404/vlan-0-1-and-4095-are-reserved-what-are-they-reserved-for

[Dieselboy]

Nice, thanks for the link. In that link there's an IEEE quote, basically what you said.

So it is valid then

[wintermute000]

TBH I've never seen VLAN0 in prod either and I wouldn't hold my breath for Cisco to give you a fix

[Dieselboy]

The cisco bug link for this problem on the link above says the issue is fixed but doesn't give a version. The description says it's fixed in RHEL7.2. So basically saying it's Red Hat's problem