From mboxrd@z Thu Jan 1 00:00:00 1970 From: shemminger@oldman.hamilton.local Subject: [PATCH 0/3] TCP timestamp hardening (update) Date: Wed, 25 Jul 2007 11:44:35 +0100 Message-ID: <20070725104434.998037763@oldman.hamilton.local> Cc: netdev@vger.kernel.org To: David Miller Return-path: Received: from smtp2.linux-foundation.org ([207.189.120.14]:36663 "EHLO smtp2.linux-foundation.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1758803AbXGYLB3 (ORCPT ); Wed, 25 Jul 2007 07:01:29 -0400 Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org These patches deal with issues brought up by Gavin McCullagh about reactions of Cubic and HTCP to hostile receivers that return bogus timestamp options. If the receiver crafts a timestamp that is larger than the original, then some of the congestion control algorithms maybe come unfair. The solution in these patches is to only use local values to measure RTT for congestion control. The timestamp is still used as described in RFC's to measure RTT used for retransmit timer. Thank you to Sangtae Ha for testing, these, see: http://netsrv.csc.ncsu.edu/net-2.6.22/stephen_lowres/ He also found some pre-existing problems with TCP-LP that might be related to NAPI on the receiver. This should go into 2.6.23. But not into the stable kernel since the risk of causing regression is greater than the possible risk exposure. --