> From: linux-rdma-ow...@vger.kernel.org [mailto:linux-rdma- > ow...@vger.kernel.org] On Behalf Of Doug Ledford
> In particular, Liran piped up with this comment: > > "Also, I don't want to do any route resolution on the Rx path. A UD QP > completion just reports the details of the packet it received. > > Conceptually, an incoming packet may not even match an SGID index at all. > Maybe, responses should be sent from another device. This should not be > decided that the point that a packet was received." > > The part that bothers me about this is that this statement makes sense when > just thinking about the spec, as you say. However, once you consider > namespaces, security implications make this statement spec compliant, but > still unacceptable. The spec itself is silent on namespaces. But, you guys > wanted, and you got, namespace support. > Since that's beyond spec, and carries security requirements, I think it's > fair to > say that from now on, the Linux kernel RDMA stack can no longer *just* be > spec compliant. There are additional concerns that must always be > addressed with new changes, and those are the namespace constraint > preservation concerns. > Hi Doug, Currently, there is no namespace support for RoCE, so the RoCEv2 patches have *nothing* to do with this. That said, the RoCE specification does not contradict or inhibit any future implementation for namespaces. The CMA will get the <VLAN, IP> from ib_wc and resolve to a netdev (or sgid_index->netdev, whatever) and process the request accordingly. We can have endless theoretical discussions on features that are not even implemented yet (e.g., RoCE namespace support) each time we add a minor straightforward, *spec-compliant* change that *all* RoCE vendors adhere to. If someone wishes to introduce a new concept, API refactoring proposal, or similar for community review, please do so with a different RFC. This is hindering progress of the whole RDMA stack development! For example, the posted SoftRoCE patches are waiting just for this. The RoCEv2 patches have been posted upstream for review for months (!) now. I simply cannot understand why this is lagging for so long; let's start to get the wheels rolling. --Liran
