From mboxrd@z Thu Jan 1 00:00:00 1970 From: Benoit PAPILLAULT Subject: ip_rt_min_delay : why 2s by default? Date: Fri, 30 May 2008 12:37:06 +0200 Message-ID: <483FD8D2.1030604@free.fr> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: davem@davemloft.net To: netdev@vger.kernel.org Return-path: Received: from smtp2e.orange.fr ([80.12.242.112]:37170 "EHLO smtp2e.orange.fr" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751819AbYE3KhJ (ORCPT ); Fri, 30 May 2008 06:37:09 -0400 Sender: netdev-owner@vger.kernel.org List-ID: Hi there, I'm doing work on a IP routing protocol designed for mobility. During testing phase, I encounter some 2s delay when routing table are updated. I compare the routing table (displayed by route -n) and the routing cache (displayed by route -Cn) and there was always a delay of 2s to propagate changes from the routing table to the routing cache. Hence, packet loss occurs. Digging into the kernel source (net/ipv4/route.c), I found the ip_rt_min_delay variable which is indeed 2s. I changed it to 0 through /proc/sys/net/ipv4/route/min_delay and it seems that all my problems are gone. Questions: 1/ Is it safe to put 0 in min_delay? I'd like to make sure that it will not trigger some bugs or inconsistencies later. 2/ If it is safe, then why it's not the default value? Is there a reason for 2s or is it simply a compromise that so far nobody cares about 2s delay? Regards, Benoit PS: Test was done with a 2.6.21.6 kernel