Re: [PATCH v3 31/31] KVM: MMU: make spte an in-out argument in make_spte

[Paolo Bonzini]

On 29/09/21 01:20, David Matlack wrote:
> >   bool make_spte(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
> >   	       struct kvm_memory_slot *slot,
> >   	       unsigned int pte_access, gfn_t gfn, kvm_pfn_t pfn,
> > -	       u64 old_spte, bool speculative, bool can_unsync,
> > -	       bool host_writable, u64 *new_spte)
> > +	       bool speculative, bool can_unsync,
> > +	       bool host_writable, u64 *sptep)
> I'd prefer a different name since `sptep` has specific meaning
> throughout the mmu code. (It's the address of the spte in the page
> table.)
>
> Case in point, I was going to suggest we can get rid of struct
> kvm_mmu_page since it can be derived from the sptep and then realized
> how wrong that was:).
>
> Instead of receiving the new spte as a parameter what do you think about
> changing make_spte to return the new spte? I think that would make the
> code more readable (but won't reduce the number of arguments because
> you'd have to add wrprot).

You have a point.  I've dropped this patch for now.

Paolo