Re: [PATCH v6 08/33] KVM: arm64: Use accessor functions for core regs

[Steffen Eiden <seiden@linux.ibm.com>]

On Tue, Aug 18, 2026 at 01:54:07PM +0100, Marc Zyngier wrote:
> On Tue, 18 Aug 2026 10:29:24 +0100,
> Steffen Eiden <seiden@linux.ibm.com> wrote:
> > 
> > On Thu, Aug 13, 2026 at 10:56:37AM +0100, Marc Zyngier wrote:
> > > On Wed, 12 Aug 2026 16:36:04 +0100,
> > > Steffen Eiden <seiden@linux.ibm.com> wrote:
> > > >
> 
> [...]
> 
> > > > diff --git a/arch/arm64/kvm/guest.c b/arch/arm64/kvm/guest.c
> > > > index b01d6622b872..773f6c8e5026 100644
> > > > --- a/arch/arm64/kvm/guest.c
> > > > +++ b/arch/arm64/kvm/guest.c
> > > > @@ -134,19 +134,19 @@ static void *core_reg_addr(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
> > > >  	     KVM_REG_ARM_CORE_REG(regs.regs[30]):
> > > >  		off -= KVM_REG_ARM_CORE_REG(regs.mregs[0]);
> > > >  		off /= 2;
> > > > -		return &vcpu->arch.ctxt.regs.regs[off];
> > > > +		return &vcpu_gp_regs(vcpu)[off];
> > > 
> > > Doesn't this belong to the previous patch?
> > > 
> > 
> > My thoughts about positioning this change in this patch:
> > 
> > The previous patch changes the semantics of vcpu_gp_regs and just
> > touched the part where the old semantic of the function was used.
> > This is mainly done by not using vcpu_gp_regs where it was used before
> > as it cannot be used to access e.g. pstate anymore.
> > 
> > This patch then introduces and *uses* new accessor functions for
> > abstractions (and uses the vcpu_gp_regs abstraction where it was not
> > used before). 
> > -> The patch changes direct accesses to using accessor functions.
> > 
> > However, I have no strong feelings for this. If you want this in the
> > previous patch, I can do that.
> 
> I think it'd make more sense. You can capture this in the commit
> message to make it clear.

Just that we are on the same page: you want me to move the change to the
previous patch?

	Steffen