From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757152AbYDBOg2 (ORCPT ); Wed, 2 Apr 2008 10:36:28 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1754912AbYDBOgU (ORCPT ); Wed, 2 Apr 2008 10:36:20 -0400 Received: from mx1.redhat.com ([66.187.233.31]:41441 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753812AbYDBOgU (ORCPT ); Wed, 2 Apr 2008 10:36:20 -0400 Subject: Re: [PATCH 0/3] xen: Fix oops when mapping/unmapping addr above 1Gb From: Mark McLoughlin Reply-To: Mark McLoughlin To: Jeremy Fitzhardinge Cc: Ingo Molnar , linux-kernel@vger.kernel.org, xen-devel@lists.xensource.com, Thomas Gleixner , "H. Peter Anvin" In-Reply-To: <47F0FB5F.2000908@goop.org> References: <1206722297-11763-1-git-send-email-markmc@redhat.com> <20080331131442.GF14636@elte.hu> <47F0FB5F.2000908@goop.org> Content-Type: text/plain Date: Wed, 02 Apr 2008 15:35:43 +0100 Message-Id: <1207146943.11911.1.camel@muff> Mime-Version: 1.0 X-Mailer: Evolution 2.12.3 (2.12.3-3.fc8) Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Mon, 2008-03-31 at 07:55 -0700, Jeremy Fitzhardinge wrote: > Ingo Molnar wrote: > > * Mark McLoughlin wrote: > > > >> In Fedora rawhide we're currently seeing oopses whenever a > >> processes dlopen()s a library which prelink() has previously relocated > >> above 1Gb. Some more details at: > >> > >> https://bugzilla.redhat.com/436453 > >> > >> The issue seems to be that during xen_pgd_pin(), we only pin > >> PTE pages, but xen's release_pmd() tries to unpin PMD pages too. > >> > >> The following series of patches is against the latest x86 > >> tree. > >> > > > > thanks, i've picked them up to get them tested - but it would be nice to > > have an ack from Jeremy as well. This seems like v2.6.25 material as > > well, right? > > Didn't I reply? Anyway, the patch looks good. It's definitely 2.6.25 > material; its fallout from the changes I did to dynamically > allocate/deallocate pmds rather than always do them with the pgd. Okay, here are the patches rebased to 2.6.25-rc8. I've reproduced the issue with plain -rc8 and confirmed the patches fix it. Cheers, Mark.