Return-Path: Received: from smtp1.wiktel.com ([69.89.207.151]:33554 "EHLO smtp1.wiktel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750927AbdJCAJy (ORCPT ); Mon, 2 Oct 2017 20:09:54 -0400 Subject: Re: PROBLEM: NFS Client Ignores TCP Resets To: Olga Kornievskaia Cc: NeilBrown , Trond Myklebust , Anna Schumaker , linux-nfs References: <56BFE55D.1010509@wiktel.com> <87twjjpcl8.fsf@notabene.neil.brown.name> <57062C53.9080102@wiktel.com> <87wpo9lydp.fsf@notabene.neil.brown.name> From: Richard Laager Message-ID: <20de1b7e-efca-e2e1-e9e0-e9aff5a1a1af@wiktel.com> Date: Mon, 2 Oct 2017 19:09:53 -0500 MIME-Version: 1.0 In-Reply-To: Content-Type: text/plain; charset=utf-8 Sender: linux-nfs-owner@vger.kernel.org List-ID: On 10/02/2017 05:03 PM, Olga Kornievskaia wrote: > On Mon, Oct 2, 2017 at 3:48 PM, Richard Laager wrote: >> On 10/02/2017 02:45 PM, Olga Kornievskaia wrote: >>> Did turning off tcp checksum offloading take care of the problem by any chance? >> >> I don't think so, but I'm not 100% sure. I keep meaning to get back to >> working on this, but since UDP is working and it's somewhat disruptive >> to test, I haven't. > > From the network trace that you've provided shouldn't SEQ# should have > been 545 instead of 465. If server is sending wrong values in the RST > then client would drop it? I don't know enough about TCP to be sure. But the RST has Ack: 465. I think that's normal? -- Richard