Re: [RFC PATCH 12/18] KVM: x86/mmu: Introduce shadow_ux_mask

[Sean Christopherson <seanjc@google.com>]

On Thu, Mar 13, 2025, Jon Kohler wrote:
> @@ -28,6 +28,7 @@ u64 __read_mostly shadow_host_writable_mask;
>  u64 __read_mostly shadow_mmu_writable_mask;
>  u64 __read_mostly shadow_nx_mask;
>  u64 __read_mostly shadow_x_mask; /* mutual exclusive with nx_mask */
> +u64 __read_mostly shadow_ux_mask;
>  u64 __read_mostly shadow_user_mask;
>  u64 __read_mostly shadow_accessed_mask;
>  u64 __read_mostly shadow_dirty_mask;
> @@ -313,8 +314,14 @@ u64 make_huge_page_split_spte(struct kvm *kvm, u64 huge_spte,
>  		 * the page executable as the NX hugepage mitigation no longer
>  		 * applies.
>  		 */
> -		if ((role.access & ACC_EXEC_MASK) && is_nx_huge_page_enabled(kvm))
> +		if ((role.access & ACC_EXEC_MASK) && is_nx_huge_page_enabled(kvm)) {

This is wrong, and probably so is every other chunk of KVM that looks at
ACC_EXEC_MASK.  E.g. if a guest hugepage is executable for user but not supervisor,
KVM will fail to make the small child user-executable.

The bug in make_spte() is even worse, because KVM would let an MBEC-aware guest
trigger the iTLB multi-hit #MC.

>  			child_spte = make_spte_executable(child_spte);
> +			// TODO: For LKML: switch to vcpu->arch.pt_guest_exec_control? up
> +			// for suggestions on how best to toggle this.

No, it belongs in the role.

> +			if (enable_pt_guest_exec_control &&
> +			    role.access & ACC_USER_EXEC_MASK)
> +				child_spte |= shadow_ux_mask;
> +		}
>  	}
>  
>  	return child_spte;
> @@ -326,7 +333,7 @@ u64 make_nonleaf_spte(u64 *child_pt, bool ad_disabled)
>  	u64 spte = SPTE_MMU_PRESENT_MASK;
>  
>  	spte |= __pa(child_pt) | shadow_present_mask | PT_WRITABLE_MASK |
> -		shadow_user_mask | shadow_x_mask | shadow_me_value;
> +		shadow_user_mask | shadow_x_mask | shadow_ux_mask | shadow_me_value;
>  
>  	if (ad_disabled)
>  		spte |= SPTE_TDP_AD_DISABLED;
> @@ -420,7 +427,8 @@ void kvm_mmu_set_me_spte_mask(u64 me_value, u64 me_mask)
>  }
>  EXPORT_SYMBOL_GPL(kvm_mmu_set_me_spte_mask);
>  
> -void kvm_mmu_set_ept_masks(bool has_ad_bits, bool has_exec_only)
> +void kvm_mmu_set_ept_masks(bool has_ad_bits, bool has_exec_only,
> +			   bool has_guest_exec_ctrl)
>  {
>  	shadow_user_mask	= VMX_EPT_READABLE_MASK;
>  	shadow_accessed_mask	= has_ad_bits ? VMX_EPT_ACCESS_BIT : 0ull;
> @@ -428,8 +436,14 @@ void kvm_mmu_set_ept_masks(bool has_ad_bits, bool has_exec_only)
>  	shadow_nx_mask		= 0ull;
>  	shadow_x_mask		= VMX_EPT_EXECUTABLE_MASK;
>  	/* VMX_EPT_SUPPRESS_VE_BIT is needed for W or X violation. */
> +	// For LKML Review:
> +	// Do we need to modify shadow_present_mask in the MBEC case?

No, because MBEC bifurcates X, it doesn't change whether or not an EPTE can be
X without being R.  From the SDM:

  1. If the “mode-based execute control for EPT” VM-execution control is 1,
     setting bit 0 indicates also that software may also configure EPT
     paging-structure entries in which bits 1:0 are both clear and in which bit 10
     is set (indicating a translation that can be used to fetch instructions from a
     supervisor-mode linear address or a user-mode linear address).

>  	shadow_present_mask	=
>  		(has_exec_only ? 0ull : VMX_EPT_READABLE_MASK) | VMX_EPT_SUPPRESS_VE_BIT;
> +
> +	shadow_ux_mask		=
> +		has_guest_exec_ctrl ? VMX_EPT_USER_EXECUTABLE_MASK : 0ull;

This is EPT specific code, just call this what it is:

	shadow_ux_mask		= has_mbec ? VMX_EPT_USER_EXECUTABLE_MASK : 0ull;
> +
>  	/*
>  	 * EPT overrides the host MTRRs, and so KVM must program the desired
>  	 * memtype directly into the SPTEs.  Note, this mask is just the mask
> @@ -484,6 +498,7 @@ void kvm_mmu_reset_all_pte_masks(void)
>  	shadow_dirty_mask	= PT_DIRTY_MASK;
>  	shadow_nx_mask		= PT64_NX_MASK;
>  	shadow_x_mask		= 0;
> +	shadow_ux_mask		= 0;
>  	shadow_present_mask	= PT_PRESENT_MASK;
>  
>  	/*
> diff --git a/arch/x86/kvm/mmu/spte.h b/arch/x86/kvm/mmu/spte.h
> index d9e22133b6d0..dc2f0dc9c46e 100644
> --- a/arch/x86/kvm/mmu/spte.h
> +++ b/arch/x86/kvm/mmu/spte.h
> @@ -171,6 +171,7 @@ extern u64 __read_mostly shadow_mmu_writable_mask;
>  extern u64 __read_mostly shadow_nx_mask;
>  extern u64 __read_mostly shadow_x_mask; /* mutual exclusive with nx_mask */
>  extern u64 __read_mostly shadow_user_mask;
> +extern u64 __read_mostly shadow_ux_mask;
>  extern u64 __read_mostly shadow_accessed_mask;
>  extern u64 __read_mostly shadow_dirty_mask;
>  extern u64 __read_mostly shadow_mmio_value;
> diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
> index 0aadfa924045..d16e3f170258 100644
> --- a/arch/x86/kvm/vmx/vmx.c
> +++ b/arch/x86/kvm/vmx/vmx.c
> @@ -8544,7 +8544,8 @@ __init int vmx_hardware_setup(void)
>  
>  	if (enable_ept)
>  		kvm_mmu_set_ept_masks(enable_ept_ad_bits,
> -				      cpu_has_vmx_ept_execute_only());
> +				      cpu_has_vmx_ept_execute_only(),
> +				      enable_pt_guest_exec_control);

Without the module param, just cpu_has_vmx_mbec().