On Fri, 26 Feb 2016 15:42:29 -0800, Tom Herbert wrote: > Agreed, and I don't see why there even needs to be modes. VXLAN-GPE > can carry arbitrary protocols with a next-header field. For Ethernet, > MPLS, IPv4, and IPv6 it should just be a simple mapping of the next > header to Ethertype for purposes of processing the payload.
That's exactly what this patchset does, Tom. The mapping is done in vxlan_parse_gpe_hdr and vxlan_build_gpe_hdr. Ethernet is special, though. It needs to be a standalone mode, otherwise frames encapsulated including an Ethernet header wouldn't be processed and there would be no way to send such packets - the only distinction the driver can use is skb->protocol and that won't become ETH_P_TEB magically. Jiri
