From mboxrd@z Thu Jan 1 00:00:00 1970 From: Vincent ETIENNE Subject: Re: [PROBLEM] Bonding driver in linux-2.6.21-rc6-mm1 Date: Sun, 29 Apr 2007 00:04:31 +0200 Message-ID: <200704290004.31698.ve@vetienne.net> References: <200704262058.33087.ve@vetienne.net> <200704282237.42753.ve@vetienne.net> <20070428143102.f39299b5.akpm@linux-foundation.org> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: Stephen Hemminger , netdev@vger.kernel.org To: Andrew Morton Return-path: Received: from mail1.vetienne.net ([87.98.219.23]:2262 "EHLO mail1.vetienne.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1160997AbXD1WEd convert rfc822-to-8bit (ORCPT ); Sat, 28 Apr 2007 18:04:33 -0400 In-Reply-To: <20070428143102.f39299b5.akpm@linux-foundation.org> Content-Disposition: inline Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org Le Saturday 28 April 2007 23:31:02 Andrew Morton, vous avez =E9crit=A0: > > > Is there anything i can do to get you more precise information ? > > I guess if you could provide us with your .config and a step-by-step = recipe > which developers should use to reproduce the problem then we should b= e able > to fix this pretty easily when someone finds the time to do so. =46or the config, i have uploaded it to=20 http://mail1.vetienne.net/linux/config-2.6.21-rc7-mm2.=20 =46or reproducing the problem, it 's simple (for me at least) : - setup a bonding interface ( IP 192.168.1.5/255.255.255.0 so nothing= =20 fancy ) bond (mode 6 : alb ) on two NIC card (in my case E1000 and TG3= ).=20 Don't know if it's important or not but the two cards are connected to = the=20 same cheap 100Mb switch ( So no gigabit ) - Reboot : when the bond interface went up ( after the two physical s= laves )=20 the bug is triggered ( for the moment always, but i have only rebooted = 2 or 3=20 times with 2.6.21-rc7-mm2 ). - You could also play with networks cable and put down/up one interfac= e or=20 another during normal work : it would often triggered the same trace (= not=20 always although ). You don't need to have load on the network. The inte= rface=20 that went down or up doesn't seem to be important as far as i can see. - No lockup of the kernel, network still work fine after the problem