From mboxrd@z Thu Jan 1 00:00:00 1970 From: Chris Leech Subject: Question about supporting non-IP offloads (FCoE) Date: Wed, 24 Sep 2008 13:55:26 -0700 Message-ID: <48DAA93E.5010102@intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit To: netdev@vger.kernel.org Return-path: Received: from mga01.intel.com ([192.55.52.88]:59246 "EHLO mga01.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752254AbYIXUz1 (ORCPT ); Wed, 24 Sep 2008 16:55:27 -0400 Sender: netdev-owner@vger.kernel.org List-ID: I'm looking at adding support for stateless offloads to the Fibre Channel over Ethernet stack at open-fcoe.org that roughly match currently supported TCP offloads: Fibre Channel CRC offload on Tx and Rx, and sequencing offload (GSO) on Tx. >>From the net_device all I see the need for is two feature flags bits for the transmit path, one for FC-CRC and one for FCoE GSO. In the skb, the FCoE needs match up well with the existing CRC and GSO fields. I would like to reuse the ip_summed bits rather than add to the skb, but I suspect this will be the most controversial part of the FCoE offloads. Are there any objections to the idea of adding checks for FCoE protocol to the dev_queue_xmit/dev_can_checksum/can_checksum_protocol path? Obviously skb_checksum_help isn't going to work as a software fallback and a protocol check would need to be added there as well, or a per-protocol helper like is done for gso_segment. - Chris