Re: about the guest(Redhat6.3) shows white screen, but the suse, ubuntu is ok

From: Matt Wilson <msw@amazon.com>
To: Dario Faggioli <dario.faggioli@citrix.com>
Cc: Elena Ufimtseva <ufimtseva@gmail.com>,
	butine@zju.edu.cn, xen-devel@lists.xen.org
Subject: Re: about the guest(Redhat6.3) shows white screen, but the suse, ubuntu is ok
Date: Mon, 22 Jul 2013 23:26:05 -0700	[thread overview]
Message-ID: <20130723062602.GB17505@u109add4315675089e695.ant.amazon.com> (raw)
In-Reply-To: <1374492730.3312.10.camel@Abyss>

On Mon, Jul 22, 2013 at 12:32:10PM +0100, Dario Faggioli wrote:
> On dom, 2013-07-21 at 22:41 +0800, butine@zju.edu.cn wrote:
> > hello,Dario,
> > 
> Hi,
> 
> > I have implemented the HVM guest NUMA.
> >
> Oh, really? Well, that would be really nice indeed. I'm on vacation till
> Wednesday, so I'll take a look at you patch on Thursday (or even
> later :-( ).
> 
> I'm Cc-ing Matt, which was also looking into it. I'm also Cc-ing Elena
> which (if I understood Butine's work correctly), is looking into pretty
> much the same thing, although for PV.

> One thing I can already tell you is that this is a really big patch, and
> it'd be really really uncomfortable to review it like this... Could you
> perhaps split it into smaller patches and send a proper series?
> 
> Take a look at other patch series submissions on the xen-devel mailing
> list...

Hi,

I had a quick look at your patch. There's quite a lot of code there!
My NUMA support for HVM guests patch is a lot smaller. I've been
meaning to find some time to add some support in libxl before posting
it. Perhaps I should post my series (which is based in part on patches
posted a while ago on xen-devel) as a RFC so you can see how it works.

> > After I created the guest(Redhat6.3),it would show white screen when
> > guest started.But the suse,ubuntu is ok.why??
> > 
> Mmm.. No idea. :-(
> 
> Actually, I'm not even sure I understood what happens, to properly try
> to help in debugging any issue, I think we'd require much more
> information and some logs.

In looking at your patches, you don't seem to adjust the initial APIC
ID of the virtual CPUs. The APIC ID needs to be changed so that cpu
topology enumeration via cpuid works correctly. I've seen Linux panic
early when the ACPI tables say that a NUMA topology is expected but
the CPU topology via cpuid doesn't match.

--msw

> For instance, have a look here:
> http://blog.xen.org/index.php/2013/06/04/reporting-a-bug-against-the-xen-hypervisor/
> 
> Nice to see that you're working on this though, and thanks for sharing
> the code! :-)
> 
> Thanks and Regards,
> Dario
> 