Re: [Xen-devel] PAE PV guest kernel regression

[Andrew Jones <drjones@xxxxxxxxxx>]

----- Original Message -----
> >>> On 06.06.12 at 10:54, Andrew Jones <drjones@xxxxxxxxxx> wrote:
> 
> > https://bugzilla.redhat.com/show_bug.cgi?id=829016
> > 
> > was opened yesterday. Maybe the following commit
> > is the culprit?
> > 
> > commit cb8095bba6d24118135a5683a956f4f4fb5f17bb
> > Author: Jan Beulich <JBeulich@xxxxxxxx>
> > Date:   Fri Jan 20 16:22:04 2012 +0000
> > 
> >     x86: atomic64 assembly improvements
> 
> Hardly - that change didn't even touch atomic64_read_cx8()
> or any of its constraints.
> 
> Further more, this
> 
> *pdpt = 0000000023be6027 *pde = 00000000037c2067 *pte =
> 8000000023bdb061
> Oops: 0003 [#1] SMP
> 
> tells us that this was a write-protection fault, yet the hypervisor
> failed to emulate this (hence the hypervisor log would likely help).
> After quite some time spent looking through the upstream 3.0.4
> sources (and my local object files) I can't, however, spot where
> the call to atomic64_read() is actually located in the functions in
> question.
> 

Ah, I didn't check to see what Fedora pulled in on top of 3.4. Had I
done that I would have immediately suspected a different patch instead
(mm-pmd_read_atomic-fix-32bit-PAE-pmd-walk-vs-pmd_populate-SMP-race-condition.patch,
upstream commit 26c191788f18). We've already encountered one problem
with this patch for RHEL6 and fixed it. The patch F17 has, however, is
already the "fixed" version. Now the difference between RHEL6 and F17
though is that F17 has CONFIG_TRANSPARENT_HUGEPAGE=y for 32b guests,
but RHEL6 does not. So now with this patch F17 is calling
atomic64_read() from pmd_none_or_trans_huge_or_clear_bad().

Drew

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel