|
|
|
|
|
|
|
|
|
|
xen-devel
Re: [Xen-devel] dom0 pvops crash
On Tue, Feb 02, 2010 at 10:31:00AM +0000, Ian Campbell wrote:
> On Tue, 2010-02-02 at 10:23 +0000, Pasi Kärkkäinen wrote:
> > On Mon, Jan 25, 2010 at 06:26:26PM +0000, Ian Campbell wrote:
> > > On Mon, 2010-01-25 at 18:18 +0000, Ian Pratt wrote:
> > > > > Here's the hacky fix from last summer:
> > > > > http://lists.xensource.com/archives/html/xen-devel/2009-06/msg00519.html
> > > > >
> > > > > On my testbox I can trigger that bug just by compiling a kernel in
> > > > > dom0.
> > > > >
> > > > > CONFIG_HIGHPTE=n is stable for me, while CONFIG_HIGHPTE=y most
> > > > > probably
> > > > > makes dom0 crash during a kernel compile.
> > > > >
> > > > > (dom0 and Xen are 32bit PAE)
> > > >
> > > > Even if it worked reliably, CONFIG_HIGHPTE performance would be
> > > > miserable: It takes an operation that is always going to be slow on
> > > > any virtualized system (updating PTEs), and pounds the cr*p out of
> > > > it...
> > >
> > > Unfortunately distros tend to enable it on native and with pvops there
> > > is often no separate Xen kernel build where it can be disabled.
> > >
> >
> > I just noticed the default Fedora 12 32bit PAE kernel has CONFIG_HIGHPTE=y.
> > Running simple "make -j2 bzImage && make -j2 modules" in the guest crashes
> > the kernel.
> >
> > Some kind of workaround should be made..
>
> Can you try the patch from
> <1264613173.16526.97718.camel@xxxxxxxxxxxxxxxxxxxxxx>? (earlier in this
> thread).
>
Yeah, sure. I can test that after a couple of days.. some travelling to do now.
> It's a hack in the form given there but I'm planning to clean it up so
> it would be useful to know if the basic concept is sound.
>
Yup.
-- Pasi
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
|
|
|
|
|