From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ben Greear Subject: Re: [PATCH 1/6] ipv6: use timer pending Date: Tue, 22 Jul 2008 17:35:30 -0700 Message-ID: <48867CD2.4080209@candelatech.com> References: <20080721192834.221648982@vyatta.com> <20080721192919.971385103@vyatta.com> <488518BF.1090207@candelatech.com> <20080722120453.114615dc@extreme> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org, Patrick McHardy To: Stephen Hemminger Return-path: Received: from mail.candelatech.com ([208.74.158.172]:44105 "EHLO ns3.lanforge.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751571AbYGWAfe (ORCPT ); Tue, 22 Jul 2008 20:35:34 -0400 In-Reply-To: <20080722120453.114615dc@extreme> Sender: netdev-owner@vger.kernel.org List-ID: Stephen Hemminger wrote: > On Mon, 21 Jul 2008 16:16:15 -0700 > Ben Greear wrote: > >> Stephen Hemminger wrote: >>> This fixes the bridge reference count problem and cleanups ipv6 FIB timer management. >>> Don't use expires field, because it is not a proper way to test, >>> instead use timer_pending(). >> I patched this into my 2.6.25, and it does not seem to help >> my refcount problem. I did not apply any of your other >> patches as it seemed they were mostly cleanups. >> >> This is not to say that this patch is wrong...but >> it seems we'll have to keep looking for other problems >> as well. > > I agree this patch is not a direct fix for the the problem. What happened > was that I was retesting on latest 2.6 net kernel and adding tracing > and checking. In the process, the problem went away. Either > a Heisenbug or some other part of the configuration changed. It turns out my problem was due to a patch I wrote to ensure netlink notify of ipv6 address adds (regardless of DAD completion). http://lists.openwall.net/netdev/2008/06/18/158 I was using ipv6_ifa_notify instead of inet6_ifa_notify. Patrick found this, so he gets all the credit. Of course, since my patch is not in the main tree, whatever refcount problems you guys are seeing must be something different. I'll keep the debug code in my tree and if I managed to reproduce it again, will be sure to try to track it down. Thanks, Ben -- Ben Greear Candela Technologies Inc http://www.candelatech.com