From mboxrd@z Thu Jan 1 00:00:00 1970 From: Thomas Jarosch Subject: Re: TCP connection stalls under 2.6.24.7 Date: Tue, 15 Jul 2008 18:10:42 +0200 Message-ID: <200807151810.45942.thomas.jarosch@intra2net.com> References: <47EA0DAB.7080205@securenet.de> <200807111633.42514.thomas.jarosch@intra2net.com> <200807151347.23598.thomas.jarosch@intra2net.com> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-15 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: netdev@vger.kernel.org, Patrick McHardy , Sven Riedel , Netfilter Developer Mailing List , =?iso-8859-15?q?D=E2niel_Fraga?= To: Jozsef Kadlecsik Return-path: Received: from mailbackup.webpage.t-com.de ([80.152.31.31]:53839 "EHLO fepb01-svc.webpage.t-com.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752437AbYGOQKv convert rfc822-to-8bit (ORCPT ); Tue, 15 Jul 2008 12:10:51 -0400 In-Reply-To: <200807151347.23598.thomas.jarosch@intra2net.com> Content-Disposition: inline Sender: netdev-owner@vger.kernel.org List-ID: > Luckily I'm able to reproduce the problem locally using an ADSL line = from > the same provider, so I'll now bisect the kernel from 2.6.23.17 to 2.= 6.24. After bisecting for hours, l only had ten revisions left to test. There was this commit that caught my eye: ------------------------------ commit c96fd3d461fa495400df24be3b3b66f0e0b152f9 Author: Ilpo J=E4rvinen Date: Thu Sep 20 11:36:37 2007 -0700 [TCP]: Enable SACK enhanced FRTO (RFC4138) by default ------------------------------ This change sets the value of "tcp_frto" to 2 by default. If I reset it to zero, the connection works immediately. @D=E2niel Fraga: Does disabling tcp_frto work for you, too? Disabling tcp_sack makes no difference. To summarize the situation, I had two different cases of stalling TCP connections, both connecting to busy SMTP relays servers which probably drop some packets here and t= here. I can easily reproduce the problem, so how do we go from here? Cheers, Thomas