Re: [PULL 06/35] hw/acpi: refactor acpi hp modules so that targets can just use what they need

[Ani Sinha]

On Thu, 21 Jul 2022, Dr. David Alan Gilbert wrote:

> * Ani Sinha (ani@anisinha.ca) wrote:
> >
> >
> > On Wed, 20 Jul 2022, Peter Maydell wrote:
> >
> > > On Wed, 20 Jul 2022 at 19:37, Ani Sinha <ani@anisinha.ca> wrote:
> > > >
> > > >
> > > >
> > > > On Tue, 19 Jul 2022, Peter Maydell wrote:
> > > >
> > > > > On Sat, 4 Sept 2021 at 22:36, Michael S. Tsirkin <mst@redhat.com> 
> > > > > wrote:
> > > > > How is this intended to work? The obvious fix from my point
> > > > > of view would just be to say "piix4.c requires pcihp.c"
> > > > > and cause CONFIG_ACPI_PIIX4 to pull in CONFIG_ACPI_PCIHP,
> > > > > but that seems like it would be rather undoing the point
> > > > > of this change.
> > > >
> > > > Yes. From the commit log and the vague recollection I have about this
> > > > change :
> > > >
> > > > > For example, mips only needs support for PIIX4 and does not
> > > > > need acpi pci hotplug support or cpu hotplug support or memory hotplug
> > > > support
> > > > > etc
> > > >
> > > > So does malta really need acpi hotplug? If not, then the stubbing out of
> > > > the vmstate struct is correct.
> > >
> > > It's not, because the vmstate struct is actually used when you
> > > savevm/loadvm a malta machine. If the malta shouldn't have
> > > acpi hotplug then we need to arrange for the hotplug code
> > > to be avoided at an earlier point, not just stub in the
> > > vmstate struct field.
> >
> > yes I think that would be more appropriate fix, I agree. Since mst added
> > that vmstate, maybe he can comment on this.
>
> Can't we just change the stub to be a valid vmstate structure?

This would be only a short term solution until we can re-org the code so
that Malta does not use acpi hotplug anymore.

>
> Dave
> (I coincidentally found this today because I'd been cc'd on
> https://gitlab.com/qemu-project/qemu/-/issues/995 a few months back
> and only just noticed)
>