[libvirt] [PATCH] qemu: Permit PCI-free aarch64 mach-virt guests

[Andrea Bolognani]

On Fri, 2016-06-17 at 18:38 +0200, Andrea Bolognani wrote:
> On Fri, 2016-06-17 at 12:01 -0400, Laine Stump wrote:
> > On 06/17/2016 11:46 AM, Martin Kletzander wrote:
> > > On Fri, Jun 17, 2016 at 11:36:05AM -0400, Laine Stump wrote:
> > > > On 06/17/2016 08:43 AM, Andrea Bolognani wrote:
> > > > >             * other than the pcie-root. This is so that there will be 
> > > > > hot-pluggable
> > > > > -         * PCI slots available
> > > > > +         * PCI slots available.
> > > > > +         *
> > > > > +         * We skip this step for aarch64 mach-virt guests, where we 
> > > > > want to
> > > > > +         * be able to have a pure virtio-mmio topology
> > > > >             */
> > > > >            if (virDomainControllerFind(def, 
> > > > > VIR_DOMAIN_CONTROLLER_TYPE_PCI, 1) < 0 &&
> > > > > +            !qemuDomainMachineIsVirt(def) &&
> > > >  
> > > > You're assuming that the only virt* machinetypes will be aarch64, which
> > > > may be reasonable now, but not in the future (periodically someone from
> > > > qemu will mention the idea of a "virt" machinetype for x86, which is
> > > > legacy-free and accepts only virtio devices). Wouldn't a more specific
> > > > comparison be better here (and in the other places in this patch)?
> > >  
> > > Just my $.02 here, but since our qemuDomainMachineIsVirt() is made
> > > specifically for aarch64 arches, I think the right thing to do would be
> > > just add architecture check into that function as using it throughout
> > > the codebase ought to actually be what all the callers want.
> >  
> > Sure, a single function would be great. Its name should reflect that it 
> > is for *aarch64* virt machines though (in anticipation of other arches 
> > getting a virt machinetype).
> 
> I think we can stick with qemuDomainMachineIsVirt() for now,
> and create per-arch variants if and when another architecture
> grows its own virt machine type.
> 
> But I'm totally for moving the architecture checks inside
> the function; in fact, I'd like to do the same for pSeries
> guests. I'll work on it next week.

I've now implemented the suggestion.

  https://www.redhat.com/archives/libvir-list/2016-June/msg01635.html

-- Andrea Bolognani / Red Hat / Virtualization