From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Sat, 1 Sep 2001 14:08:47 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Sat, 1 Sep 2001 14:08:38 -0400 Received: from minus.inr.ac.ru ([193.233.7.97]:36612 "HELO ms2.inr.ac.ru") by vger.kernel.org with SMTP id ; Sat, 1 Sep 2001 14:08:27 -0400 From: kuznet@ms2.inr.ac.ru Message-Id: <200109011808.WAA19782@ms2.inr.ac.ru> Subject: Re: Excessive TCP retransmits over lossless, high latency link To: lk@tantalophile.demon.co.uk (Jamie Lokier) Date: Sat, 1 Sep 2001 22:08:24 +0400 (MSK DST) Cc: davem@redhat.com, ak@muc.de, linux-kernel@vger.kernel.org In-Reply-To: <20010901181729.A2204@thefinal.cern.ch> from "Jamie Lokier" at Sep 1, 1 06:17:29 pm X-Mailer: ELM [version 2.4 PL24] MIME-Version: 1.0 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Hello! > The interesting thing is that there isn't any evidence of packet loss. Why did you disable bith sacks and timestamps? Exactly to get maximal damage from long delay link? > Is there some /proc/sys setting to fix this, a kernel patch, or is it > perhaps fixed in a newer kernel already? No patches to block send required ACKs exist of course. :-) All the problem is at sender, it mispredicts rtt. What OS is sender? If it is linux too, try to use default configuration not playing with /proc/sys/net/tcp_*, especially with timestamps and sacks and the situation should rectify. Also, please, send full (binary!) tcpdump from SYN and to FIN. Andi says right thing, but I am still puzzled why rtt is miscalculated it should be estimated correctly. Well, and if sender is not linux... no ideas. Alexey