Re: [PATCH v7 036/102] KVM: x86/mmu: Allow non-zero value for non-present SPTE

[Kai Huang <kai.huang@intel.com>]

On Thu, 2022-07-14 at 11:41 -0700, Isaku Yamahata wrote:
> Thanks for review. Now here is the updated version.
> 
> From f1ee540d62ba13511b2c7d3db7662e32bd263e48 Mon Sep 17 00:00:00 2001
> Message-Id: <f1ee540d62ba13511b2c7d3db7662e32bd263e48.1657823906.git.isaku.yamahata@intel.com>
> In-Reply-To: <3941849bf08a55cfbbe69b222f0fd0dac7c5ee53.1657823906.git.isaku.yamahata@intel.com>
> References: <3941849bf08a55cfbbe69b222f0fd0dac7c5ee53.1657823906.git.isaku.yamahata@intel.com>
> From: Sean Christopherson <sean.j.christopherson@intel.com>
> Date: Mon, 29 Jul 2019 19:23:46 -0700
> Subject: [PATCH 036/304] KVM: x86/mmu: Allow non-zero value for non-present
>  SPTE
> 
> TDX introduced a new ETP, Secure-EPT, in addition to the existing EPT.
> Secure-EPT maps protected guest memory, which is called private. Since
> Secure-EPT page tables is also protected, those page tables is also called
> private.  The existing EPT is often called shared EPT to distinguish from
> Secure-EPT.  And also page tables for shared EPT is also called shared.

AFAICT secure-EPT isn't quite directly related here, so I don't think you should
spend the paragraph on it.  The first paragraph should state the problem and
catch reviewer's eyeball.

> 
> TDX module enables #VE injection by setting "EPT-violation #VE" in
> secondary processor-based VM-execution controls of TD VMCS.  It also sets
> "suppress #VE" bit in Secure-EPT so that EPT violation on Secure-EPT causes
> exit to VMM.
> 
> Because guest memory is protected with TDX, VMM can't parse instructions in
> the guest memory.  Instead, MMIO hypercall is used for guest TD to pass
> necessary information to VMM.  To make unmodified device driver work, guest
> TD expects #VE on accessing shared GPA for MMIO. The #VE handler of guest
> TD converts MMIO access into MMIO hypercall.  To trigger #VE in guest TD,
> VMM needs to clear "suppress #VE" bit in shared EPT entry that corresponds
> to MMIO address.
> 
> So the execution flow related for MMIO is as follows
> 
> - TDX module sets "EPT-violation #VE" in secondary processor-based
>   VM-execution controls of TD VMCS.
> - Allocate page for shared EPT PML4E page. Shared EPT entries are
>   initialized with suppress #VE bit set.  Update the EPTP pointer.
> - TD accesses a GPA for MMIO to trigger EPT violation.  It exits to VMM with
>   EPT violation due to suppress #VE bit of EPT entries of PML4E page.
> - VMM figures out the faulted GPA is for MMIO
> - start shared EPT page table walk.
> - Allocate non-leaf EPT pages for the shared EPT.
> - Allocate leaf EPT page for the shared EPT and initialize EPT entries with
>   suppress #VE bit set.
> - VMM clears the suppress #VE bit for faulted GPA for MMIO.
>   Please notice the leaf EPT page has 512 SPTE and other 511 SPTE entries
>   need to keep "suppress #VE" bit set because GPAs for those SPTEs are not
>   known to be MMIO. (It requires further lookups.)
>   If GPA is a guest page, link the guest page from the leaf SPTE entry.
> - resume TD vcpu.
> - Guest TD gets #VE, and converts MMIO access into MMIO hypercall.
> - If the GPA maps guest memory, VMM resolves it with guest pages.

Too many details IMHO.

Also, you forgot to mention the non-zero value for non-present SPTE is not just
for MMIO, but also for shared memory.

How about below?

For TD guest, the current way to emulate MMIO doesn't work any more, as KVM is
not able to access the private memory of TD guest and do the emulation. 
Instead, TD guest expects to receive #VE when it accesses the MMIO and then it
can explicitly makes hypercall to KVM to get the expected information.

To achieve this, the TDX module always enables "EPT-violation #VE" in the VMCS
control.  And accordingly, KVM needs to configure the MMIO spte to trigger EPT
violation (instead of misconfiguration) and at the same time, also clear the
"suppress #VE" bit so the TD guest can get the #VE instead of causing actual EPT
violation to KVM.

In order for KVM to be able to have chance to set up the correct SPTE for MMIO
for TD guest, the default non-present SPTE must have the "suppress #VE" bit set
so KVM can get a real EPT violation for the first time when TD guest accesses
the MMIO.

Also, when TD guest accesses the actual shared memory, it should continue to
trigger EPT violation to the KVM instead of receiving the #VE  (the TDX module
guarantees KVM will receive EPT violation for private memory access).  This
means for the shared memory, the SPTE also must have the "suppress #VE" bit set
for the non-present SPTE.

Add support to allow a non-zero value for the non-present SPTE (i.e. when the
page table is firstly allocated, and when the SPTE is zapped) to allow setting
"suppress #VE" bit for the non-present SPTE.

Introduce a new macro SHADOW_NONPRESENT_VALUE to be the "suppress #VE" bit. 
Unconditionally set the "suppress #VE" bit (which is bit 63) for both AMD and
Intel as: 1) AMD hardware doesn't use this bit; 2) for normal VMX guest, KVM
never enables the "EPT-violation #VE" in VMCS control and "suppress #VE" bit is
ignored by hardware.

(if you want to set SHADOW_NONPRESENT_VALUE only for TDP MMU then continue to
describe, but I don't see this is done in your below patch)

> 
> SPTEs for shared EPT need suppress #VE" bit set initially when it
> is allocated or zapped, therefore non-zero non-present value for SPTE
> needs to be allowed.
> 
> TDP MMU uses REMOVED_SPTE = 0x5a0ULL as special constant to indicate the
> intermediate value to indicate one thread is operating on it and the value
> should be semi-arbitrary value.  For TDX (more exactly to use #VE), the
> value should include suppress #VE bit.  Rename REMOVED_SPTE to
> __REMOVED_SPTE and define REMOVED_SPTE as (REMOVED_SPTE | "suppress #VE")
> bit.

IMHO REMOVED_SPTE is implementation details so it's OK to not mention in
changelog.

> 
> For simplicity, "suppress #VE" bit is set unconditionally for X86_64 for
> non-present SPTE.  Because "suppress #VE" bit (bit position of 63) for
> non-present SPTE is ignored for non-TD case (AMD CPUs or Intel VMX case
> with "EPT-violation #VE" cleared), the functionality shouldn't change.
> 
> Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
> Signed-off-by: Isaku Yamahata <isaku.yamahata@intel.com>
> ---
>  arch/x86/kvm/mmu/mmu.c         | 71 ++++++++++++++++++++++++++++++++--
>  arch/x86/kvm/mmu/paging_tmpl.h |  3 +-
>  arch/x86/kvm/mmu/spte.c        |  5 ++-
>  arch/x86/kvm/mmu/spte.h        | 28 +++++++++++++-
>  arch/x86/kvm/mmu/tdp_mmu.c     | 23 +++++++----
>  5 files changed, 116 insertions(+), 14 deletions(-)
> 
> diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
> index 51306b80f47c..992f31458f94 100644
> --- a/arch/x86/kvm/mmu/mmu.c
> +++ b/arch/x86/kvm/mmu/mmu.c
> @@ -668,6 +668,55 @@ static void walk_shadow_page_lockless_end(struct kvm_vcpu *vcpu)
>  	}
>  }
>  
> +#ifdef CONFIG_X86_64
> +static inline void kvm_init_shadow_page(void *page)
> +{
> +	int ign;
> +
> +	/*
> +	 * AMD: "suppress #VE" bit is ignored
> +	 * Intel non-TD(VMX): "suppress #VE" bit is ignored because
> +	 *   EPT_VIOLATION_VE isn't set.
> +	 * guest TD: TDX module sets EPT_VIOLATION_VE
> +	 *   conventional SEPT: "suppress #VE" bit must be set to get EPT violation
> +	 *   private SEPT: "suppress #VE" bit is ignored.  CPU doesn't walk it
> +	 *
> +	 * For simplicity, unconditionally initialize SPET to set "suppress #VE".
> +	 */
> +	asm volatile ("rep stosq\n\t"
> +		      : "=c"(ign), "=D"(page)
> +		      : "a"(SHADOW_NONPRESENT_VALUE), "c"(4096/8), "D"(page)
> +		      : "memory"
> +	);
> +}
> +
> +static int mmu_topup_shadow_page_cache(struct kvm_vcpu *vcpu)
> +{
> +	struct kvm_mmu_memory_cache *mc = &vcpu->arch.mmu_shadow_page_cache;
> +	int start, end, i, r;
> +
> +	start = kvm_mmu_memory_cache_nr_free_objects(mc);
> +	r = kvm_mmu_topup_memory_cache(mc, PT64_ROOT_MAX_LEVEL);
> +
> +	/*
> +	 * Note, topup may have allocated objects even if it failed to allocate
> +	 * the minimum number of objects required to make forward progress _at
> +	 * this time_.  Initialize newly allocated objects even on failure, as
> +	 * userspace can free memory and rerun the vCPU in response to -ENOMEM.
> +	 */
> +	end = kvm_mmu_memory_cache_nr_free_objects(mc);
> +	for (i = start; i < end; i++)
> +		kvm_init_shadow_page(mc->objects[i]);
> +	return r;
> +}
> +#else
> +static int mmu_topup_shadow_page_cache(struct kvm_vcpu *vcpu)
> +{
> +	return kvm_mmu_topup_memory_cache(vcpu->arch.mmu_shadow_page_cache,
> +					  PT64_ROOT_MAX_LEVEL);
> +}
> +#endif /* CONFIG_X86_64 */
> +
>  static int mmu_topup_memory_caches(struct kvm_vcpu *vcpu, bool maybe_indirect)
>  {
>  	int r;
> @@ -677,8 +726,7 @@ static int mmu_topup_memory_caches(struct kvm_vcpu *vcpu, bool maybe_indirect)
>  				       1 + PT64_ROOT_MAX_LEVEL + PTE_PREFETCH_NUM);
>  	if (r)
>  		return r;
> -	r = kvm_mmu_topup_memory_cache(&vcpu->arch.mmu_shadow_page_cache,
> -				       PT64_ROOT_MAX_LEVEL);
> +	r = mmu_topup_shadow_page_cache(vcpu);
>  	if (r)
>  		return r;
>  	if (maybe_indirect) {
> @@ -5654,7 +5702,24 @@ int kvm_mmu_create(struct kvm_vcpu *vcpu)
>  	vcpu->arch.mmu_page_header_cache.kmem_cache = mmu_page_header_cache;
>  	vcpu->arch.mmu_page_header_cache.gfp_zero = __GFP_ZERO;
>  
> -	vcpu->arch.mmu_shadow_page_cache.gfp_zero = __GFP_ZERO;
> +	/*
> +	 * When X86_64, initial SEPT entries are initialized with
> +	 * SHADOW_NONPRESENT_VALUE.  Otherwise zeroed.  See
> +	 * mmu_topup_shadow_page_cache().
> +	 *
> +	 * Shared EPTEs need to be initialized with SUPPRESS_VE=1, otherwise
> +	 * not-present EPT violations would be reflected into the guest by
> +	 * hardware as #VE exceptions.  This is handled by initializing page
> +	 * allocations via kvm_init_shadow_page() during cache topup.
> +	 * In that case, telling the page allocation to zero-initialize the page
> +	 * would be wasted effort.
> +	 *
> +	 * The initialization is harmless for S-EPT entries because KVM's copy
> +	 * of the S-EPT isn't consumed by hardware, and because under the hood
> +	 * S-EPT entries should never #VE.
> +	 */
> +	if (!IS_ENABLED(X86_64))
> +		vcpu->arch.mmu_shadow_page_cache.gfp_zero = __GFP_ZERO;
>  
>  	vcpu->arch.mmu = &vcpu->arch.root_mmu;
>  	vcpu->arch.walk_mmu = &vcpu->arch.root_mmu;
> diff --git a/arch/x86/kvm/mmu/paging_tmpl.h b/arch/x86/kvm/mmu/paging_tmpl.h
> index fe35d8fd3276..964ec76579f0 100644
> --- a/arch/x86/kvm/mmu/paging_tmpl.h
> +++ b/arch/x86/kvm/mmu/paging_tmpl.h
> @@ -1031,7 +1031,8 @@ static int FNAME(sync_page)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp)
>  		gpa_t pte_gpa;
>  		gfn_t gfn;
>  
> -		if (!sp->spt[i])
> +		/* spt[i] has initial value of shadow page table allocation */
> +		if (sp->spt[i] != SHADOW_NONPRESENT_VALUE)
>  			continue;
>  
>  		pte_gpa = first_pte_gpa + i * sizeof(pt_element_t);
> diff --git a/arch/x86/kvm/mmu/spte.c b/arch/x86/kvm/mmu/spte.c
> index cda1851ec155..bd441458153f 100644
> --- a/arch/x86/kvm/mmu/spte.c
> +++ b/arch/x86/kvm/mmu/spte.c
> @@ -36,6 +36,9 @@ u64 __read_mostly shadow_present_mask;
>  u64 __read_mostly shadow_me_value;
>  u64 __read_mostly shadow_me_mask;
>  u64 __read_mostly shadow_acc_track_mask;
> +#ifdef CONFIG_X86_64
> +u64 __read_mostly shadow_nonpresent_value;
> +#endif
>  
>  u64 __read_mostly shadow_nonpresent_or_rsvd_mask;
>  u64 __read_mostly shadow_nonpresent_or_rsvd_lower_gfn_mask;
> @@ -360,7 +363,7 @@ void kvm_mmu_set_mmio_spte_mask(u64 mmio_value, u64 mmio_mask, u64 access_mask)
>  	 * not set any RWX bits.
>  	 */
>  	if (WARN_ON((mmio_value & mmio_mask) != mmio_value) ||
> -	    WARN_ON(mmio_value && (REMOVED_SPTE & mmio_mask) == mmio_value))
> +	    WARN_ON(mmio_value && (__REMOVED_SPTE & mmio_mask) == mmio_value))
>  		mmio_value = 0;
>  
>  	if (!mmio_value)
> diff --git a/arch/x86/kvm/mmu/spte.h b/arch/x86/kvm/mmu/spte.h
> index 0127bb6e3c7d..f5fd22f6bf5f 100644
> --- a/arch/x86/kvm/mmu/spte.h
> +++ b/arch/x86/kvm/mmu/spte.h
> @@ -140,6 +140,19 @@ static_assert(MMIO_SPTE_GEN_LOW_BITS == 8 && MMIO_SPTE_GEN_HIGH_BITS == 11);
>  
>  #define MMIO_SPTE_GEN_MASK		GENMASK_ULL(MMIO_SPTE_GEN_LOW_BITS + MMIO_SPTE_GEN_HIGH_BITS - 1, 0)
>  
> +/*
> + * non-present SPTE value for both VMX and SVM for TDP MMU.
> + * For SVM NPT, for non-present spte (bit 0 = 0), other bits are ignored.
> + * For VMX EPT, bit 63 is ignored if #VE is disabled.
> + *              bit 63 is #VE suppress if #VE is enabled.
> + */
> +#ifdef CONFIG_X86_64
> +#define SHADOW_NONPRESENT_VALUE	BIT_ULL(63)
> +static_assert(!(SHADOW_NONPRESENT_VALUE & SPTE_MMU_PRESENT_MASK));
> +#else
> +#define SHADOW_NONPRESENT_VALUE	0ULL
> +#endif
> +
>  extern u64 __read_mostly shadow_host_writable_mask;
>  extern u64 __read_mostly shadow_mmu_writable_mask;
>  extern u64 __read_mostly shadow_nx_mask;
> @@ -178,16 +191,27 @@ extern u64 __read_mostly shadow_nonpresent_or_rsvd_mask;
>   * non-present intermediate value. Other threads which encounter this value
>   * should not modify the SPTE.
>   *
> + * For X86_64 case, SHADOW_NONPRESENT_VALUE, "suppress #VE" bit, is set because
> + * "EPT violation #VE" in the secondary VM execution control may be enabled.
> + * Because TDX module sets "EPT violation #VE" for TD, "suppress #VE" bit for
> + * the conventional EPT needs to be set.
> + *
>   * Use a semi-arbitrary value that doesn't set RWX bits, i.e. is not-present on
>   * bot AMD and Intel CPUs, and doesn't set PFN bits, i.e. doesn't create a L1TF
>   * vulnerability.  Use only low bits to avoid 64-bit immediates.
>   *
>   * Only used by the TDP MMU.
>   */
> -#define REMOVED_SPTE	0x5a0ULL
> +#define __REMOVED_SPTE	0x5a0ULL
>  
>  /* Removed SPTEs must not be misconstrued as shadow present PTEs. */
> -static_assert(!(REMOVED_SPTE & SPTE_MMU_PRESENT_MASK));
> +static_assert(!(__REMOVED_SPTE & SPTE_MMU_PRESENT_MASK));
> +
> +/*
> + * See above comment around __REMOVED_SPTE.  REMOVED_SPTE is the actual
> + * intermediate value set to the removed SPET.  it sets the "suppress #VE" bit.
> + */
> +#define REMOVED_SPTE	(SHADOW_NONPRESENT_VALUE | __REMOVED_SPTE)
>  
>  static inline bool is_removed_spte(u64 spte)
>  {
> diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
> index 7b9265d67131..2ca03ec3bf52 100644
> --- a/arch/x86/kvm/mmu/tdp_mmu.c
> +++ b/arch/x86/kvm/mmu/tdp_mmu.c
> @@ -692,8 +692,16 @@ static inline int tdp_mmu_zap_spte_atomic(struct kvm *kvm,
>  	 * overwrite the special removed SPTE value. No bookkeeping is needed
>  	 * here since the SPTE is going from non-present to non-present.  Use
>  	 * the raw write helper to avoid an unnecessary check on volatile bits.
> +	 *
> +	 * Set non-present value to SHADOW_NONPRESENT_VALUE, rather than 0.
> +	 * It is because when TDX is enabled, TDX module always
> +	 * enables "EPT-violation #VE", so KVM needs to set
> +	 * "suppress #VE" bit in EPT table entries, in order to get
> +	 * real EPT violation, rather than TDVMCALL.  KVM sets
> +	 * SHADOW_NONPRESENT_VALUE (which sets "suppress #VE" bit) so it
> +	 * can be set when EPT table entries are zapped.
>  	 */
> -	__kvm_tdp_mmu_write_spte(iter->sptep, 0);
> +	__kvm_tdp_mmu_write_spte(iter->sptep, SHADOW_NONPRESENT_VALUE);

Since you also always set SHADOW_NONPRESENT_VALUE to SPTE for legacy MMU when
the page table is firstly allocated, it also makes sense to set it when SPTE is
zapped for legacy MMU.

This part is missing in this patch.

>  
>  	return 0;
>  }
> @@ -870,8 +878,8 @@ static void __tdp_mmu_zap_root(struct kvm *kvm, struct kvm_mmu_page *root,
>  			continue;
>  
>  		if (!shared)
> -			tdp_mmu_set_spte(kvm, &iter, 0);
> -		else if (tdp_mmu_set_spte_atomic(kvm, &iter, 0))
> +			tdp_mmu_set_spte(kvm, &iter, SHADOW_NONPRESENT_VALUE);
> +		else if (tdp_mmu_set_spte_atomic(kvm, &iter, SHADOW_NONPRESENT_VALUE))
>  			goto retry;
>  	}
>  }
> @@ -927,8 +935,9 @@ bool kvm_tdp_mmu_zap_sp(struct kvm *kvm, struct kvm_mmu_page *sp)
>  	if (WARN_ON_ONCE(!is_shadow_present_pte(old_spte)))
>  		return false;
>  
> -	__tdp_mmu_set_spte(kvm, kvm_mmu_page_as_id(sp), sp->ptep, old_spte, 0,
> -			   sp->gfn, sp->role.level + 1, true, true);
> +	__tdp_mmu_set_spte(kvm, kvm_mmu_page_as_id(sp), sp->ptep, old_spte,
> +			   SHADOW_NONPRESENT_VALUE, sp->gfn, sp->role.level + 1,
> +			   true, true);
>  
>  	return true;
>  }
> @@ -965,7 +974,7 @@ static bool tdp_mmu_zap_leafs(struct kvm *kvm, struct kvm_mmu_page *root,
>  		    !is_last_spte(iter.old_spte, iter.level))
>  			continue;
>  
> -		tdp_mmu_set_spte(kvm, &iter, 0);
> +		tdp_mmu_set_spte(kvm, &iter, SHADOW_NONPRESENT_VALUE);
>  		flush = true;
>  	}
>  
> @@ -1330,7 +1339,7 @@ static bool set_spte_gfn(struct kvm *kvm, struct tdp_iter *iter,
>  	 * invariant that the PFN of a present * leaf SPTE can never change.
>  	 * See __handle_changed_spte().
>  	 */
> -	tdp_mmu_set_spte(kvm, iter, 0);
> +	tdp_mmu_set_spte(kvm, iter, SHADOW_NONPRESENT_VALUE);
>  
>  	if (!pte_write(range->pte)) {
>  		new_spte = kvm_mmu_changed_pte_notifier_make_spte(iter->old_spte,
> -- 
> 2.25.1
> 
> 
>