Re: [PATCH v11 033/113] KVM: x86/mmu: Track shadow MMIO value on a per-VM basis

[Huang, Kai]

On Thu, 2023-01-12 at 08:31 -0800, isaku.yamahata@xxxxxxxxx wrote:
> From: Isaku Yamahata <isaku.yamahata@xxxxxxxxx>
> 
> TDX will use a different shadow PTE entry value for MMIO from VMX.  Add
> members to kvm_arch and track value for MMIO per-VM instead of global
> variables.  By using the per-VM EPT entry value for MMIO, the existing VMX
> logic is kept working.  Introduce a separate setter function so that guest
> TD can override later.

The guest TD itself cannot override.  It is KVM who overrides it for a TDX
guest.
 
> 
> Also require mmio spte cachcing for TDX.  Actually this is true case
			 ^
			 spell check.

> because TDX require EPT and KVM EPT allows mmio spte caching.

The second sentence doesn't make sense.  IIUC "TDX requires EPT + EPT _allows_
MMIO caching = TDX _allows_ MMIO caching", which is different from "TDX
_requires_ MMIO caching).

> 
> Signed-off-by: Sean Christopherson <sean.j.christopherson@xxxxxxxxx>
> Signed-off-by: Isaku Yamahata <isaku.yamahata@xxxxxxxxx>
> ---
>  arch/x86/include/asm/kvm_host.h |  2 ++
>  arch/x86/kvm/mmu.h              |  1 +
>  arch/x86/kvm/mmu/mmu.c          |  7 ++++---
>  arch/x86/kvm/mmu/spte.c         | 10 ++++++++--
>  arch/x86/kvm/mmu/spte.h         |  4 ++--
>  arch/x86/kvm/mmu/tdp_mmu.c      | 14 +++++++++++---
>  6 files changed, 28 insertions(+), 10 deletions(-)
> 
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index 73c987b3d2b6..807da4b95aba 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -1243,6 +1243,8 @@ struct kvm_arch {
>  	 */
>  	spinlock_t mmu_unsync_pages_lock;
>  
> +	u64 shadow_mmio_value;
> +
>  	struct list_head assigned_dev_head;
>  	struct iommu_domain *iommu_domain;
>  	bool iommu_noncoherent;
> diff --git a/arch/x86/kvm/mmu.h b/arch/x86/kvm/mmu.h
> index a45f7a96b821..50d240d52697 100644
> --- a/arch/x86/kvm/mmu.h
> +++ b/arch/x86/kvm/mmu.h
> @@ -101,6 +101,7 @@ static inline u8 kvm_get_shadow_phys_bits(void)
>  }
>  
>  void kvm_mmu_set_mmio_spte_mask(u64 mmio_value, u64 mmio_mask, u64 access_mask);
> +void kvm_mmu_set_mmio_spte_value(struct kvm *kvm, u64 mmio_value);
>  void kvm_mmu_set_me_spte_mask(u64 me_value, u64 me_mask);
>  void kvm_mmu_set_ept_masks(bool has_ad_bits, bool has_exec_only);
>  
> diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
> index 59befdfeec23..8d3d7deebdd0 100644
> --- a/arch/x86/kvm/mmu/mmu.c
> +++ b/arch/x86/kvm/mmu/mmu.c
> @@ -2450,7 +2450,7 @@ static int mmu_page_zap_pte(struct kvm *kvm, struct kvm_mmu_page *sp,
>  				return kvm_mmu_prepare_zap_page(kvm, child,
>  								invalid_list);
>  		}
> -	} else if (is_mmio_spte(pte)) {
> +	} else if (is_mmio_spte(kvm, pte)) {
>  		mmu_spte_clear_no_track(spte);
>  	}
>  	return 0;
> @@ -4119,7 +4119,7 @@ static int handle_mmio_page_fault(struct kvm_vcpu *vcpu, u64 addr, bool direct)
>  	if (WARN_ON(reserved))
>  		return -EINVAL;
>  
> -	if (is_mmio_spte(spte)) {
> +	if (is_mmio_spte(vcpu->kvm, spte)) {
>  		gfn_t gfn = get_mmio_spte_gfn(spte);
>  		unsigned int access = get_mmio_spte_access(spte);
>  
> @@ -4628,7 +4628,7 @@ static unsigned long get_cr3(struct kvm_vcpu *vcpu)
>  static bool sync_mmio_spte(struct kvm_vcpu *vcpu, u64 *sptep, gfn_t gfn,
>  			   unsigned int access)
>  {
> -	if (unlikely(is_mmio_spte(*sptep))) {
> +	if (unlikely(is_mmio_spte(vcpu->kvm, *sptep))) {
>  		if (gfn != get_mmio_spte_gfn(*sptep)) {
>  			mmu_spte_clear_no_track(sptep);
>  			return true;
> @@ -6111,6 +6111,7 @@ int kvm_mmu_init_vm(struct kvm *kvm)
>  	struct kvm_page_track_notifier_node *node = &kvm->arch.mmu_sp_tracker;
>  	int r;
>  
> +	kvm->arch.shadow_mmio_value = shadow_mmio_value;
>  	INIT_LIST_HEAD(&kvm->arch.active_mmu_pages);
>  	INIT_LIST_HEAD(&kvm->arch.zapped_obsolete_pages);
>  	INIT_LIST_HEAD(&kvm->arch.possible_nx_huge_pages);
> diff --git a/arch/x86/kvm/mmu/spte.c b/arch/x86/kvm/mmu/spte.c
> index cc0bc058fb25..a23e9205fc42 100644
> --- a/arch/x86/kvm/mmu/spte.c
> +++ b/arch/x86/kvm/mmu/spte.c
> @@ -74,10 +74,10 @@ u64 make_mmio_spte(struct kvm_vcpu *vcpu, u64 gfn, unsigned int access)
>  	u64 spte = generation_mmio_spte_mask(gen);
>  	u64 gpa = gfn << PAGE_SHIFT;
>  
> -	WARN_ON_ONCE(!shadow_mmio_value);
> +	WARN_ON_ONCE(!vcpu->kvm->arch.shadow_mmio_value);
>  
>  	access &= shadow_mmio_access_mask;
> -	spte |= shadow_mmio_value | access;
> +	spte |= vcpu->kvm->arch.shadow_mmio_value | access;
>  	spte |= gpa | shadow_nonpresent_or_rsvd_mask;
>  	spte |= (gpa & shadow_nonpresent_or_rsvd_mask)
>  		<< SHADOW_NONPRESENT_OR_RSVD_MASK_LEN;
> @@ -413,6 +413,12 @@ void kvm_mmu_set_mmio_spte_mask(u64 mmio_value, u64 mmio_mask, u64 access_mask)
>  }
>  EXPORT_SYMBOL_GPL(kvm_mmu_set_mmio_spte_mask);
>  
> +void kvm_mmu_set_mmio_spte_value(struct kvm *kvm, u64 mmio_value)
> +{
> +	kvm->arch.shadow_mmio_value = mmio_value;
> +}
> +EXPORT_SYMBOL_GPL(kvm_mmu_set_mmio_spte_value);
> +
>  void kvm_mmu_set_me_spte_mask(u64 me_value, u64 me_mask)
>  {
>  	/* shadow_me_value must be a subset of shadow_me_mask */
> diff --git a/arch/x86/kvm/mmu/spte.h b/arch/x86/kvm/mmu/spte.h
> index 471378ee9071..256395eb593f 100644
> --- a/arch/x86/kvm/mmu/spte.h
> +++ b/arch/x86/kvm/mmu/spte.h
> @@ -251,9 +251,9 @@ static inline struct kvm_mmu_page *sptep_to_sp(u64 *sptep)
>  	return to_shadow_page(__pa(sptep));
>  }
>  
> -static inline bool is_mmio_spte(u64 spte)
> +static inline bool is_mmio_spte(struct kvm *kvm, u64 spte)
>  {
> -	return (spte & shadow_mmio_mask) == shadow_mmio_value &&
> +	return (spte & shadow_mmio_mask) == kvm->arch.shadow_mmio_value &&
>  	       likely(enable_mmio_caching);
>  }
>  
> diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
> index 6111e3e9266d..dffacb7eb15a 100644
> --- a/arch/x86/kvm/mmu/tdp_mmu.c
> +++ b/arch/x86/kvm/mmu/tdp_mmu.c
> @@ -19,6 +19,14 @@ int kvm_mmu_init_tdp_mmu(struct kvm *kvm)
>  {
>  	struct workqueue_struct *wq;
>  
> +	/*
> +	 * TDs require mmio_caching to clear suppress_ve bit of SPTE for GPA
> +	 * of MMIO so that TD can convert #VE triggered by MMIO into
> +	 * TDG.VP.VMCALL<MMIO>.
> +	 */
> +	if (kvm->arch.vm_type == KVM_X86_TDX_VM && !enable_mmio_caching)
> +		return -EOPNOTSUPP;

SEV-ES does the check in hardware_setup:

void __init sev_hardware_setup(void)
{
	...
	/*
         * SEV-ES requires MMIO caching as KVM doesn't have access to the guest
         * instruction stream, i.e. can't emulate in response to a #NPF and
         * instead relies on #NPF(RSVD) being reflected into the guest as #VC
         * (the guest can then do a #VMGEXIT to request MMIO emulation).
         */
        if (!enable_mmio_caching)
                goto out;

	...
}

TDX should be done in the same way.

And IMO this chunk really doesn't belong to this patch -- I interpret this patch
as a "infrastructure patch to track shadow MMIO value on per-VM basis" (which
even should have no functional change IMHO), but this chunk is clearly doing
more than that.

> +
>  	if (!tdp_enabled || !READ_ONCE(tdp_mmu_enabled))
>  		return 0;
>  
> @@ -587,8 +595,8 @@ static void __handle_changed_spte(struct kvm *kvm, int as_id, gfn_t gfn,
>  		 * impact the guest since both the former and current SPTEs
>  		 * are nonpresent.
>  		 */
> -		if (WARN_ON(!is_mmio_spte(old_spte) &&
> -			    !is_mmio_spte(new_spte) &&
> +		if (WARN_ON(!is_mmio_spte(kvm, old_spte) &&
> +			    !is_mmio_spte(kvm, new_spte) &&
>  			    !is_removed_spte(new_spte)))
>  			pr_err("Unexpected SPTE change! Nonpresent SPTEs\n"
>  			       "should not be replaced with another,\n"
> @@ -1114,7 +1122,7 @@ static int tdp_mmu_map_handle_target_level(struct kvm_vcpu *vcpu,
>  	}
>  
>  	/* If a MMIO SPTE is installed, the MMIO will need to be emulated. */
> -	if (unlikely(is_mmio_spte(new_spte))) {
> +	if (unlikely(is_mmio_spte(vcpu->kvm, new_spte))) {
>  		vcpu->stat.pf_mmio_spte_created++;
>  		trace_mark_mmio_spte(rcu_dereference(iter->sptep), iter->gfn,
>  				     new_spte);