Re: [PATCH] KVM: arm64: Set/unset vGIC v4 forwarding if direct IRQs are supported

[Oliver Upton <oliver.upton@linux.dev>]

On Tue, Jul 29, 2025 at 09:56:12AM -0700, Raghavendra Rao Ananta wrote:
> On Tue, Jul 29, 2025 at 7:37 AM Oliver Upton <oliver.upton@linux.dev> wrote:
> >
> > Hi Raghu,
> >
> > Thanks for reporting this so quickly :)
> >
> > On Mon, Jul 28, 2025 at 10:37:10PM +0000, Raghavendra Rao Ananta wrote:
> > > diff --git a/arch/arm64/kvm/vgic/vgic-v4.c b/arch/arm64/kvm/vgic/vgic-v4.c
> > > index e7e284d47a77..873a190bcff7 100644
> > > --- a/arch/arm64/kvm/vgic/vgic-v4.c
> > > +++ b/arch/arm64/kvm/vgic/vgic-v4.c
> > > @@ -433,7 +433,7 @@ int kvm_vgic_v4_set_forwarding(struct kvm *kvm, int virq,
> > >       unsigned long flags;
> > >       int ret = 0;
> > >
> > > -     if (!vgic_supports_direct_msis(kvm))
> > > +     if (!vgic_supports_direct_irqs(kvm))
> > >               return 0;
> > >
> > >       /*
> > > @@ -533,7 +533,7 @@ int kvm_vgic_v4_unset_forwarding(struct kvm *kvm, int host_irq)
> > >       unsigned long flags;
> > >       int ret = 0;
> > >
> > > -     if (!vgic_supports_direct_msis(kvm))
> > > +     if (!vgic_supports_direct_irqs(kvm))
> > >               return 0;
> >
> > I'm not sure this is what we want, since a precondition of actually
> > doing vLPI injection is the guest having an ITS. Could you try the
> > following?
> >
> > Thanks,
> > Oliver
> >
> > diff --git a/arch/arm64/kvm/vgic/vgic-mmio-v3.c b/arch/arm64/kvm/vgic/vgic-mmio-v3.c
> > index a3ef185209e9..70d50c77e5dc 100644
> > --- a/arch/arm64/kvm/vgic/vgic-mmio-v3.c
> > +++ b/arch/arm64/kvm/vgic/vgic-mmio-v3.c
> > @@ -50,6 +50,14 @@ bool vgic_has_its(struct kvm *kvm)
> >
> >  bool vgic_supports_direct_msis(struct kvm *kvm)
> >  {
> > +       /*
> > +        * Deliberately conflate vLPI and vSGI support on GICv4.1 hardware,
> > +        * indirectly allowing userspace to control whether or not vPEs are
> > +        * allocated for the VM.
> > +        */
> > +       if (system_supports_direct_sgis() && !vgic_supports_direct_sgis(kvm))
> > +               return false;
> > +
> >         return kvm_vgic_global_state.has_gicv4 && vgic_has_its(kvm);
> >  }
> >
> > diff --git a/arch/arm64/kvm/vgic/vgic.h b/arch/arm64/kvm/vgic/vgic.h
> > index 1384a04c0784..de1c1d3261c3 100644
> > --- a/arch/arm64/kvm/vgic/vgic.h
> > +++ b/arch/arm64/kvm/vgic/vgic.h
> > @@ -396,15 +396,7 @@ bool vgic_supports_direct_sgis(struct kvm *kvm);
> >
> >  static inline bool vgic_supports_direct_irqs(struct kvm *kvm)
> >  {
> > -       /*
> > -        * Deliberately conflate vLPI and vSGI support on GICv4.1 hardware,
> > -        * indirectly allowing userspace to control whether or not vPEs are
> > -        * allocated for the VM.
> > -        */
> > -       if (system_supports_direct_sgis())
> > -               return vgic_supports_direct_sgis(kvm);
> > -
> > -       return vgic_supports_direct_msis(kvm);
> > +       return vgic_supports_direct_msis(kvm) || vgic_supports_direct_sgis(kvm);
> >  }
> >
> >  int vgic_v4_init(struct kvm *kvm);
> 
> Yes, the diff seems fine (tested as well). Would you be pushing a v2
> or do you want me to (on your behalf)?

Go ahead and respin this diff, thanks!

Oliver