From mboxrd@z Thu Jan 1 00:00:00 1970 From: Or Gerlitz Subject: Re: [PATCH V5 7/8] IB/uverbs: Resolve Ethernet L2 addresses when modifying QP Date: Wed, 20 Nov 2013 12:08:11 +0200 Message-ID: <528C8A0B.3030700@mellanox.com> References: <1384381792-2023-1-git-send-email-ogerlitz@mellanox.com> <1384381792-2023-8-git-send-email-ogerlitz@mellanox.com> Mime-Version: 1.0 Content-Type: text/plain; charset="ISO-8859-1"; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: Sender: linux-rdma-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Somnath Kotur , Devesh Sharma , Or Gerlitz , Roland Dreier Cc: "linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org" , monis , matanb , Tzahi Oved , Moni Shoua List-Id: linux-rdma@vger.kernel.org On 20/11/2013 09:15, Somnath Kotur wrote: >> However, in the approach to move address resolution to core verbs i.e. >> >ib_modify_qp() and ib_create_ah(), vendor driver will have freedom to >> >resolve l2 addresses in its own way. > This I am not sure if it's a good idea for each vendor driver to implement L2 address resolution in it's own way? Not sure if that was the intent behind Roland's statement ? I agree with Somnath, I don't see the point in putting L3 --> L2 address resolution within vendor drivers. This will create huge code duplication, and more problems. Again, with the proposed patches all kernel ULPs that are applicable to RoCE are covered and hence there's no address resolution for kernel session. To comply with non-modified user space applications/libraries V5 added a code to do address resolution and Roland just pointed out it may makes more sense to put that small code piece in the core verbs modify_qp function and not in the uverbs call. Or. -- 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