Re: [PATCH v3 13/17] KVM: x86/tdp_mmu: Support mirror root for TDP MMU

["Edgecombe, Rick P" <rick.p.edgecombe@intel.com>]

On Thu, 2024-07-04 at 16:51 +0800, Yan Zhao wrote:
> On Wed, Jun 19, 2024 at 03:36:10PM -0700, Rick Edgecombe wrote:
> > diff --git a/arch/x86/kvm/mmu/tdp_mmu.h b/arch/x86/kvm/mmu/tdp_mmu.h
> > index b887c225ff24..2903f03a34be 100644
> > --- a/arch/x86/kvm/mmu/tdp_mmu.h
> > +++ b/arch/x86/kvm/mmu/tdp_mmu.h
> > @@ -10,7 +10,7 @@
> >   void kvm_mmu_init_tdp_mmu(struct kvm *kvm);
> >   void kvm_mmu_uninit_tdp_mmu(struct kvm *kvm);
> >   
> > -void kvm_tdp_mmu_alloc_root(struct kvm_vcpu *vcpu);
> > +void kvm_tdp_mmu_alloc_root(struct kvm_vcpu *vcpu, bool private);
> >   
> >   __must_check static inline bool kvm_tdp_mmu_get_root(struct kvm_mmu_page
> > *root)
> This function name is very similar to below tdp_mmu_get_root().

That is true.

> 
> > +static inline struct kvm_mmu_page *tdp_mmu_get_root_for_fault(struct
> > kvm_vcpu *vcpu,
> > +                                                             struct
> > kvm_page_fault *fault)
> > +{
> > +       if (unlikely(!kvm_is_addr_direct(vcpu->kvm, fault->addr)))
> > +               return root_to_sp(vcpu->arch.mmu->mirror_root_hpa);
> > +
> > +       return root_to_sp(vcpu->arch.mmu->root.hpa);
> > +}
> > +
> > +static inline struct kvm_mmu_page *tdp_mmu_get_root(struct kvm_vcpu *vcpu,
> > +                                                   enum
> > kvm_tdp_mmu_root_types type)
> > +{
> > +       if (unlikely(type == KVM_MIRROR_ROOTS))
> > +               return root_to_sp(vcpu->arch.mmu->mirror_root_hpa);
> > +
> > +       return root_to_sp(vcpu->arch.mmu->root.hpa);
> > +}
> No need to put the two functions in tdp_mmu.h as they are not called in mmu.c.

I think it is nice to have the root/enum logic above close together.

> 
> Could we move them to tdp_mmu.c and rename them to something like
> tdp_mmu_type_to_root() and tdp_mmu_fault_to_root() ?

tdp_mmu_get_root_for_fault() was proposed by Paolo, and tdp_mmu_get_root() was
discussed without comment. Not to say there is anything wrong with the names
proposed, but I think this is wading into bikeshedding territory at this stage.