Unfortunately my remark on this patch was only for the irq0 override thing,
which is x86-specific. I think other architectures still need userspace
irqchip in some cases.

Paolo

Il mar 22 set 2020, 19:29 Eduardo Habkost <ehabkost@redhat.com> ha scritto:

> On Tue, Sep 22, 2020 at 07:17:24PM +0200, Paolo Bonzini wrote:
> > On 22/09/20 19:03, Eduardo Habkost wrote:
> > >>  I can remove kernel_irqchip=off once
> > >> apic irq0 override is assumed.
> > > [4 years later]
> > >
> > > Can we remove it now?  I couldn't find out if we can assume
> > > kvm_has_gsi_routing() is true everywhere, or just on a few
> > > architectures.
> >
> > Yes, we can.  Other architectures don't matter, what counts is that we
> > can require kvm_has_gsi_routing() == true on x86.
>
> Do we have other architectures where we can assume that?  I'm in
> the mood for deleting some code today.
>
> --
> Eduardo
>
>