Re: [PATCH V2 3/8] kvm: x86/mmu: Use kvm_mmu_invalidate_gva() in nested_ept_invalidate_addr()

[Sean Christopherson <seanjc@google.com>]

On Tue, Feb 07, 2023, Lai Jiangshan wrote:
> From: Lai Jiangshan <jiangshan.ljs@antgroup.com>
> 
> Use kvm_mmu_invalidate_gva() instead open calls to mmu->invlpg().
> 
> No functional change intended.
> 
> Signed-off-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>
> ---
>  arch/x86/kvm/mmu/mmu.c    | 1 +
>  arch/x86/kvm/vmx/nested.c | 5 ++++-
>  2 files changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
> index 8563b52b8bb7..e03cf5558773 100644
> --- a/arch/x86/kvm/mmu/mmu.c
> +++ b/arch/x86/kvm/mmu/mmu.c
> @@ -5734,6 +5734,7 @@ void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
>  			mmu->invlpg(vcpu, gva, mmu->prev_roots[i].hpa);
>  	}
>  }
> +EXPORT_SYMBOL_GPL(kvm_mmu_invalidate_gva);
>  
>  void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
>  {
> diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
> index 557b9c468734..f552f3c454b1 100644
> --- a/arch/x86/kvm/vmx/nested.c
> +++ b/arch/x86/kvm/vmx/nested.c
> @@ -358,6 +358,7 @@ static bool nested_ept_root_matches(hpa_t root_hpa, u64 root_eptp, u64 eptp)
>  static void nested_ept_invalidate_addr(struct kvm_vcpu *vcpu, gpa_t eptp,
>  				       gpa_t addr)
>  {
> +	unsigned long roots = 0;
>  	uint i;
>  	struct kvm_mmu_root_info *cached_root;
>  
> @@ -368,8 +369,10 @@ static void nested_ept_invalidate_addr(struct kvm_vcpu *vcpu, gpa_t eptp,
>  
>  		if (nested_ept_root_matches(cached_root->hpa, cached_root->pgd,
>  					    eptp))
> -			vcpu->arch.mmu->invlpg(vcpu, addr, cached_root->hpa);
> +			roots |= KVM_MMU_ROOT_PREVIOUS(i);
>  	}
> +	if (roots)
> +		kvm_mmu_invalidate_gva(vcpu, vcpu->arch.mmu, addr, roots);

This is subtly buggy.  On 32-bit kernels, the upper 32 bits of the GPA will get
dropped.  "addr" here is a gpa_t, i.e. u64, whereas kvm_mmu_invalidate_gva() takes
a gva_t, i.e. unsigned long.

FNAME(gva_to_gpa) just eats the misnomer, but I think I'd rather rename this path,
e.g. kvm_mmu_invalidate_addr()?