Re: [PATCH v10 036/108] KVM: TDX: Enable mmio spte caching always for TDX

[Huang, Kai]

On Sat, 2022-10-29 at 23:22 -0700, isaku.yamahata@xxxxxxxxx wrote:
> From: Isaku Yamahata <isaku.yamahata@xxxxxxxxx>
> 
> TDX needs to set shared spte for MMIO GFN to !SUPPRES_VE_BIT | !RWX so that
> guest TD can get #VE and then issue TDG.VP.VMCALL<MMIO>.  Enable mmio
> caching always for TDX irrelevant the module parameter enable_mmio_caching.
> 
> Signed-off-by: Isaku Yamahata <isaku.yamahata@xxxxxxxxx>
> ---
>  arch/x86/kvm/mmu/mmu.c     | 3 ++-
>  arch/x86/kvm/mmu/spte.h    | 2 +-
>  arch/x86/kvm/mmu/tdp_mmu.c | 7 +++++++
>  3 files changed, 10 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
> index 0d3fa29ccccc..9098f77cdaa4 100644
> --- a/arch/x86/kvm/mmu/mmu.c
> +++ b/arch/x86/kvm/mmu/mmu.c
> @@ -3229,7 +3229,8 @@ static int handle_abnormal_pfn(struct kvm_vcpu *vcpu, struct kvm_page_fault *fau
>  		 * and only if L1's MAXPHYADDR is inaccurate with respect to
>  		 * the hardware's).
>  		 */
> -		if (unlikely(!enable_mmio_caching) ||
> +		if (unlikely(!enable_mmio_caching &&
> +			     !kvm_gfn_shared_mask(vcpu->kvm)) ||
>  		    unlikely(fault->gfn > kvm_mmu_max_gfn()))
>  			return RET_PF_EMULATE;
>  	}
> diff --git a/arch/x86/kvm/mmu/spte.h b/arch/x86/kvm/mmu/spte.h
> index 82f0d5c08b77..fecfdcb5f321 100644
> --- a/arch/x86/kvm/mmu/spte.h
> +++ b/arch/x86/kvm/mmu/spte.h
> @@ -244,7 +244,7 @@ extern u64 __read_mostly shadow_nonpresent_or_rsvd_lower_gfn_mask;
>  static inline bool is_mmio_spte(struct kvm *kvm, u64 spte)
>  {
>  	return (spte & shadow_mmio_mask) == kvm->arch.shadow_mmio_value &&
> -	       likely(enable_mmio_caching);
> +	       likely(enable_mmio_caching || kvm_gfn_shared_mask(kvm));
>  }


Sean suggested we can change to treat "mmio_value == 0" also as MMIO caching
being enabled:

https://lore.kernel.org/lkml/244f619a4e7a1c7079830d12379872a111da418d.camel@xxxxxxxxx/
 

>  
>  static inline bool is_shadow_present_pte(u64 pte)
> diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
> index e07f14351d14..3325633b1cb5 100644
> --- a/arch/x86/kvm/mmu/tdp_mmu.c
> +++ b/arch/x86/kvm/mmu/tdp_mmu.c
> @@ -1875,6 +1875,13 @@ int kvm_tdp_mmu_get_walk(struct kvm_vcpu *vcpu, u64 addr, u64 *sptes,
>  
>  	*root_level = vcpu->arch.mmu->root_role.level;
>  
> +	/*
> +	 * mmio page fault isn't supported for protected guest because
> +	 * instructions in protected guest memory can't be parsed by VMM.
> +	 */
> +	if (WARN_ON_ONCE(kvm_gfn_shared_mask(vcpu->kvm)))
> +		return leaf;
> +

It's weird to put this here.

I think the logic is, for TDX guest (or similar protected guests),
handle_mmio_page_fault() should not be reached at all.  So I think we can just
add a WARN() against kvm_gfn_shared_mask(), or is_tdx_vm() at the very beginning
of handle_mmio_page_fault()?

>  	tdp_mmu_for_each_pte(iter, mmu, gfn, gfn + 1) {
>  		leaf = iter.level;
>  		sptes[leaf] = iter.old_spte;