From mboxrd@z Thu Jan 1 00:00:00 1970 From: Avi Kivity Subject: Re: Problems with CONFIG_KVM_GUEST? Date: Sun, 02 Aug 2009 18:49:01 +0300 Message-ID: <4A75B56D.5070407@redhat.com> References: <20090801212616.GD734@mit.edu> <4A755EB3.9080502@redhat.com> <20090802153238.GC15828@mit.edu> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit To: Theodore Tso , linux-kernel@vger.kernel.org, KVM list , Marcelo Tosatti Return-path: In-Reply-To: <20090802153238.GC15828@mit.edu> Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org On 08/02/2009 06:32 PM, Theodore Tso wrote: > On Sun, Aug 02, 2009 at 12:38:59PM +0300, Avi Kivity wrote: > >> (Adding ccs) >> >> On 08/02/2009 12:26 AM, Theodore Tso wrote: >> >>> Using the XFS QA regression test suites to test ext4, I reliably get >>> lock-ups and OOPS's with I run a kernel using KVM with >>> CONFIG_KVM_GUEST configured. This is with 2.6.31-rc3 and 2.6.31-rc4. >>> >>> The problems disappear if I turn off CONFIG_KVM_GUEST; the XFS tests >>> run without problem to completion once I turn this configuration >>> option off. That's what I'll do for now, but if someone could look >>> into this, I'd appreciate it. >>> >>> >> Is this a PAE mode guest? >> > > Yes, it is. Here's the config (with CONFIG_KVM_GUEST turned off, but > otherwise what I had testing with). > > What's the host kernel version? We had problems with CONFIG_KVM_GUEST and PAE which were resolved in 2.6.30-final but perhaps failed to trickle down to whatever you're using. The upstream commit is a8cd0244 ("KVM: Make paravirt tlb flush also reload the PAE PDPTRs"). -- error compiling committee.c: too many arguments to function