qemu-s390x
[Top][All Lists]
Advanced

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

Re: [PATCH v1 2/9] s390x: toplogy: adding drawers and books to smp parsi


From: Daniel P . Berrangé
Subject: Re: [PATCH v1 2/9] s390x: toplogy: adding drawers and books to smp parsing
Date: Tue, 20 Jul 2021 10:19:43 +0100
User-agent: Mutt/2.0.7 (2021-05-04)

On Tue, Jul 20, 2021 at 10:46:31AM +0200, Pierre Morel wrote:
> 
> 
> On 7/20/21 10:20 AM, Cornelia Huck wrote:
> > On Tue, Jul 20 2021, Pierre Morel <pmorel@linux.ibm.com> wrote:
> > 
> > > On 7/19/21 5:50 PM, Cornelia Huck wrote:
> > > > On Fri, Jul 16 2021, Daniel P. Berrangé <berrange@redhat.com> wrote:
> > > > > Is the book/drawer thing architecture specific, or is it machine
> > > > > type / CPU specific. ie do /all/ the s390x machine types / CPUS
> > > > > QEMU support the book/drawer concept, or only a subset.
> > > > 
> > > > Should not be by machine type, but might be by cpu model (e.g. older
> > > > hardware lacking the needed support for exposing this to the guest.) IBM
> > > > folks, please correct me if I'm wrong.
> > > 
> > > 
> > > Looks correct to me this is an information indicated by a facility
> > > introduced with Z10 if I do not make an error.
> > 
> > Hm. Would that become a problem if we made availability of parameters
> > dependent upon a value in the machine (see the other thread I cc:ed you
> > on?)
> > 
> 
> Why?
> The parameter can always be there, it is just that with older cpu model we
> will not report the topology information to the guest.

I mostly see this as an error reporting problem, perhaps with an optional
capabilitty reporting facility.

eg if someone requests 'z9'  or older together with books > 1 or
drawers > 1, it would be ideal if QEMU reports an error message,
rather than silently ignoring it and not reporting this info to
the guest.

Taking it further, it some app queries CPU models  via QMP, it
could be desirable if the CPU model data returned indicates whether
drawers/books > 1 are conceptually available.  Likewise for dies
on x86, since I cna't imagine guest OS doing sensible things
with an i486 CPU and dies=2

Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|




reply via email to

[Prev in Thread] Current Thread [Next in Thread]