From mboxrd@z Thu Jan 1 00:00:00 1970 From: Josh Durgin Subject: Re: reproductible kernel oops with kernel 3.2 inside kvm Date: Thu, 17 May 2012 15:21:55 -0700 Message-ID: <4FB57A03.70309@inktank.com> References: <4FA290C1.9080808@univ-nantes.fr> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from mail-pz0-f46.google.com ([209.85.210.46]:47281 "EHLO mail-pz0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S967208Ab2EQWV6 (ORCPT ); Thu, 17 May 2012 18:21:58 -0400 Received: by dady13 with SMTP id y13so2993615dad.19 for ; Thu, 17 May 2012 15:21:57 -0700 (PDT) In-Reply-To: <4FA290C1.9080808@univ-nantes.fr> Sender: ceph-devel-owner@vger.kernel.org List-ID: To: Yann Dupont Cc: ceph-devel Hi Yann, Sorry for the late response. On 05/03/2012 07:05 AM, Yann Dupont wrote: > Hello. I'm stress testing ceph since some time now, with quite good > results. I really like ceph and will probably use in in some > pre-production services. > > Anyway I've seen some bugs. > > One of them is instability if the kernel is running inside KVM, leading > to a very fast (and reproductible) kernel oops. On bare metal this > particular oops doesn't happen. > > The kernel oops itself involve ceph, but it could be a real bug in kvm too. > > The host machine is runnning 3.2.2 > kvm is quite ancien (0.14) > guest OS is ubuntu 12.04 with his standard kernel. Retried with custom > 3.2 kernel with the same problem. I'm not sure how many people are using the kernel client within kvm, but I haven't seen this problem before. Since it's in d_prune, it's probably Ceph related, but perhaps kvm makes a race condition trigger more often in your environment. I filed http://tracker.newdream.net/issues/2444 to track this.