[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] upgrade to 2.6.18.8 ?


  • To: "Eduardo Habkost" <ehabkost@xxxxxxxxxx>
  • From: "Michael Marineau" <mike@xxxxxxxxxxxx>
  • Date: Wed, 10 Oct 2007 13:32:49 -0700
  • Cc: Keir Fraser <keir@xxxxxxxxxxxxx>, xen-devel@xxxxxxxxxxxxxxxxxxx, Teck Choon Giam <giamteckchoon@xxxxxxxxx>, Stefan Berger <stefanb@xxxxxxxxxx>
  • Delivery-date: Wed, 10 Oct 2007 13:33:29 -0700
  • Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=beta; h=received:message-id:date:from:sender:to:subject:cc:in-reply-to:mime-version:content-type:references:x-google-sender-auth; b=k0WPFPd26HZ7xrr8X+CXTMyaYhK78iXnSI9ne4Dgh4VwsF6IVkSyoG7CFUgE4HB2TJoIozldCHxXOSE7u3pUCxc2pwYMus3Xlwxey9/2tkoCD+wRviD6qjUCjn4BoOwZc95WqC4uNLo2R4DXjRHevO1bHmOSeFQTgGxCVTfYqRs=
  • List-id: Xen developer discussion <xen-devel.lists.xensource.com>

On 10/3/07, Eduardo Habkost <ehabkost@xxxxxxxxxx> wrote:
> On Wed, Oct 03, 2007 at 07:41:36PM +0800, Teck Choon Giam wrote:
> > On 10/3/07, Stefan Berger <stefanb@xxxxxxxxxx> wrote:
> > >
> > > I was wondering whether we could apply the 2.6.18.8 patch to the linux 
> > > tree?
> > > Assuming that that patch makes 2.6.18 better...
> > >
> > > It should be fairly simple. Only two hunks dont apply (both in
> > > net/core/skbuff.c) and those changes seem to already have been applied
> > > anyway.
> >
> > I have been running 2.6.18.8 with Xen-3.1.0/Xen-3.1.1-rc1/Xen3.1.1-rc2
> > and the patch series I just exclude
> > net-gso-6-linear-segmentation.patch.  The only problem I encountered
> > is anything larger than 512MB in domainU might have such error:
> >
> > Sep 27 01:46:43 XXX kernel: ------------[ cut here ]------------
> > Sep 27 01:46:43 XXX kernel: kernel BUG at
> > /usr/src/xen-3.1.1-rc1-src/linux-2.6.18.8-xen/arch/i386/mm/hypervisor.c:167!
>
>
> Do you have CONFIG_HIGHPTE enabled? CONFIG_HIGHPTE have caused problems
> on 2.6.20 and 2.6.21 rebases of Xen 3.1 on Fedora, on pgd_pin() and
> other page table updating hypercalls, because the kernel somehow was
> trying to map page table entries as writeable.
>
> 'xm dmesg' should give more information on the cause of this Oops.
>
>
> >
> > I didn't try on 2.6.18 on domainU more than 512MB though so can't
> > really comment on whether this is just happening to kernel 2.6.18.8.
>
> I haven't reproduced my CONFIG_HIGHPTE problems on 2.6.18-xen-3.1.0,
> also. I don't know if it was a problem of our Xen rebase or the problem
> is present on the 2.6.18 xenbits tree.
>
> Enabling CONFIG_HIGHPTE under Xen is not recommended, anyway. Even if
> it was working, performance will probably suffer.
>
> --
> Eduardo

Wow, this appears (at least in my quick test today) to be the cause of
the oops I keep hitting, reported in the thread titled "Hunting down
an oops in Xen 3.1.0's 2.6.18 kernel". I would *highly* recommend the
attached trivial patch to disable HIGHPTE to avoid this.

-- 
Michael Marineau
Oregon State University
mike@xxxxxxxxxxxx

Attachment: disable-highpte.patch
Description: Binary data

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.