From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jarek Poplawski Subject: Re: 2.6.29.1: unregister_netdevice problem Date: Sat, 16 May 2009 09:06:29 +0200 Message-ID: <4A0E65F5.1040906@gmail.com> References: <49F76BF6.4040709@gmail.com> <20090429054510.GA8267@night.netis.ru> <20090429090809.GA2995@ami.dom.local> <20090508062640.GA4555@night.netis.ru> <20090508104628.GA6287@ff.dom.local> <20090510073559.GA8929@night.netis.ru> <20090510124641.GA2978@ami.dom.local> <20090515071946.GA15669@night.netis.ru> <20090515080649.GA5469@ff.dom.local> <20090515085414.GA27744@night.netis.ru> <20090515101209.GA6367@ff.dom.local> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Cc: "Alexander V. Lukyanov" , linux-kernel@vger.kernel.org, netdev@vger.kernel.org To: unlisted-recipients:; (no To-header on input) Return-path: Received: from mail-bw0-f174.google.com ([209.85.218.174]:61470 "EHLO mail-bw0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752475AbZEPHHu (ORCPT ); Sat, 16 May 2009 03:07:50 -0400 In-Reply-To: <20090515101209.GA6367@ff.dom.local> Sender: netdev-owner@vger.kernel.org List-ID: Jarek Poplawski wrote, On 05/15/2009 12:12 PM: > On Fri, May 15, 2009 at 12:54:14PM +0400, Alexander V. Lukyanov wrote: >> On Fri, May 15, 2009 at 08:06:49AM +0000, Jarek Poplawski wrote: >>> On Fri, May 15, 2009 at 11:19:46AM +0400, Alexander V. Lukyanov wrote: >>> ... >>>> Looks like 2.6.28.10 does not have this refcnt problem. Also, after I have >>>> reversed net/ipv4/route.c changes from 2.6.29, the problem does not occur either. >>> Very nice work! It should be a piece of cake now - there were not much >>> changes. I'll look at it, as usual ;-), but, of course if it's >>> possible, it would be great to try "in the meantime" with reverted the >>> biggest of them append below. >> I suspect the problem occurs when cand==rthi. > > Looks like good catch! (But there could be more than this.) Then, of > course it would be interesting to try first some fix (like nulling > rthi after rt_free(cand), I guess). On the other hand, since this looks like quite obvious bug (even if it doesn't fix your problem), feel free to send here a patch fixing it any way you like without waiting for the final test results (Cc-ing authors of the offending patch, I hope). Jarek P.