Re: [PATCH v5 00/27] IB/Verbs: IB Management Helpers

["Dave Goodell (dgoodell)" <dgoodell-FYB4Gu1CFyUAvxtiuMwx3w@public.gmane.org>]

On Apr 21, 2015, at 6:36 PM, Liran Liss <liranl-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org> wrote:

> An ib_dev (or a port of) should be distinguished by 3 qualifiers:
> - The link layer:
> -- Ethernet (shared by iWARP, USNIC, and ROCE)
> -- Infiniband
> 
> - The transport (*)
> -- IBTA transport (shared by IB and ROCE)
> -- iWARP transport
> -- USNIC transport

I haven't been following this discussion as closely as I could have, but I want to clarify something about usNIC.  There are two "transports" used by usNIC:

1. The legacy RDMA_TRANSPORT_USNIC type, which indicates usNIC traffic will be Ethernet frames with Ethertype==0x8915 (like RoCE) but containing a custom, non-IBTA-sanctioned header format instead of a full GRH.  This "transport" is still supported by the usnic_verbs kernel driver but in practice is no longer in use.  For this "transport" there isn't really any clear L3/L4 header to point to.

2. The current RDMA_TRANSPORT_USNIC_UDP type, which indicates usNIC traffic will be standard UDP/IP/Ethernet packets.

> (*) Transport means both:
> - The L4 wire protocols (e.g., BTH+ headers of IBTA, optionally encapsulated by UDP in ROCEv2, or the iWARP stack)
> - The transport semantics (for example, there are slight semantic differences between IBTA and iWARP)

No usNIC hardware or software currently performs any hardware offload of RDMA features (only UD is supported), so there is no usNIC equivalent of the BTH+ to discuss right now.

-Dave

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html