From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jesse Gross Subject: Re: [PATCH v1 1/6] net: Generalize udp based tunnel offload Date: Mon, 7 Dec 2015 08:20:23 -0800 Message-ID: References: <20151205.130313.1952213347830357234.davem@davemloft.net> <20151205.172742.1214683922564543621.davem@davemloft.net> <5664F6BF.4060909@cumulusnetworks.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Cc: Alexander Duyck , Tom Herbert , David Miller , Hannes Frederic Sowa , John Linville , Anjali Singhai Jain , Netdev , Kiran Patil To: David Ahern Return-path: Received: from mail.kernel.org ([198.145.29.136]:54225 "EHLO mail.kernel.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753967AbbLGQUq (ORCPT ); Mon, 7 Dec 2015 11:20:46 -0500 Received: from mail.kernel.org (localhost [127.0.0.1]) by mail.kernel.org (Postfix) with ESMTP id B9470204B0 for ; Mon, 7 Dec 2015 16:20:44 +0000 (UTC) Received: from mail-vk0-f47.google.com (mail-vk0-f47.google.com [209.85.213.47]) (using TLSv1.2 with cipher AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPSA id CD4E2203A4 for ; Mon, 7 Dec 2015 16:20:43 +0000 (UTC) Received: by vkha189 with SMTP id a189so105104505vkh.2 for ; Mon, 07 Dec 2015 08:20:43 -0800 (PST) In-Reply-To: <5664F6BF.4060909@cumulusnetworks.com> Sender: netdev-owner@vger.kernel.org List-ID: On Sun, Dec 6, 2015 at 7:02 PM, David Ahern wrote: > On 12/6/15 6:20 PM, Alexander Duyck wrote: >> >> That works for Linux to Linux, but what about the cases where you have >> a non-Linux endpoint on the other end such as something like a Cisco >> switch? > > > Why does is matter what kind of switch the NIC is connected to? I think Cisco was just an example, not anything particular about their switches. But there are two general problems: * Some protocols, like VXLAN, recommend that the UDP checksum be zero so this is what pretty much everyone implements. As a result, independent of the merits of using the checksum, most non-Linux endpoints won't support it. * The reason why this recommendation exists in the first place is that most ASIC based switches can't compute/verify UDP checksums. They slice off the headers and only run that through the chip's core memory, so the rest of the packet isn't available to compute a checksum over.