Jason Gunthorpe <jguntho...@obsidianresearch.com> wrote: > Can you talk abit about compatibility please? What happens when nodes > with this patch are on the same network as nodes without it?
The CM on the passive side would send a reject with the reason being "invalid gid" so this will not go unnoticed. > Does this patch remove the encoding of the VLAN from the GID? YES, and I explained in argument #1 why the vlan being there doesn't work in many environments, in other words, its something that needs to be fix, and this series addresses that. > How is the destination MAC derived now? as it was before, using address resolution, e.g ARPs sent by the RDMA-CM. > There is a RoCE standard, it doesn't say much, but how the MAC and GRH > GID are related/derived really should be specified... > > Not sure about copying the IP/IPv6 address from the interface into the > HW, there has always been pressure to keep verbs separate from the net > stack.. At the very least patch #2 should have its change log updated > to actually reflect what is in the patch. Sure, I'll see what needs to be better explained in the change-log. Note that the inbox RoCE implementation is tightly coupled to net-devices, e.g the GID table population is based on netevents of related netdevices. Or. -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html