[PATCH v3 7/8] KVM: x86: nSVM: Handle restore of legacy nested state

Jim Mattson posted 8 patches 3 days, 15 hours ago
[PATCH v3 7/8] KVM: x86: nSVM: Handle restore of legacy nested state
Posted by Jim Mattson 3 days, 15 hours ago
When nested NPT is enabled and KVM_SET_NESTED_STATE is used to restore an
old checkpoint (without a valid gPAT), the current IA32_PAT value must be
used as L2's gPAT.

The current IA32_PAT value may be restored by KVM_SET_MSRS after
KVM_SET_NESTED_STATE. Furthermore, there may be a KVM_GET_NESTED_STATE
before the first KVM_RUN.

Introduce a new boolean, svm->nested.legacy_gpat_semantics. When set, hPAT
updates are also applied to gPAT, preserving the old behavior where L2
shared L1's PAT. svm_vcpu_pre_run() clears this boolean at the first
KVM_RUN.

Signed-off-by: Jim Mattson <jmattson@google.com>
---
 arch/x86/kvm/svm/nested.c | 11 ++++++++---
 arch/x86/kvm/svm/svm.c    |  2 ++
 arch/x86/kvm/svm/svm.h    |  9 +++++++++
 3 files changed, 19 insertions(+), 3 deletions(-)

diff --git a/arch/x86/kvm/svm/nested.c b/arch/x86/kvm/svm/nested.c
index 3f512fb630db..a7d6fc1382a7 100644
--- a/arch/x86/kvm/svm/nested.c
+++ b/arch/x86/kvm/svm/nested.c
@@ -1936,9 +1936,14 @@ static int svm_set_nested_state(struct kvm_vcpu *vcpu,
 	if (ret)
 		goto out_free;
 
-	if (nested_npt_enabled(svm) &&
-	    (kvm_state->hdr.svm.flags & KVM_STATE_SVM_VALID_GPAT))
-		svm_set_gpat(svm, kvm_state->hdr.svm.gpat);
+	if (nested_npt_enabled(svm)) {
+		if (kvm_state->hdr.svm.flags & KVM_STATE_SVM_VALID_GPAT) {
+			svm_set_gpat(svm, kvm_state->hdr.svm.gpat);
+		} else {
+			svm_set_gpat(svm, vcpu->arch.pat);
+			svm->nested.legacy_gpat_semantics = true;
+		}
+	}
 
 	svm->nested.force_msr_bitmap_recalc = true;
 
diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
index b62c32c3942d..a6a44deec82b 100644
--- a/arch/x86/kvm/svm/svm.c
+++ b/arch/x86/kvm/svm/svm.c
@@ -4239,6 +4239,8 @@ static int svm_vcpu_pre_run(struct kvm_vcpu *vcpu)
 	if (to_kvm_sev_info(vcpu->kvm)->need_init)
 		return -EINVAL;
 
+	to_svm(vcpu)->nested.legacy_gpat_semantics = false;
+
 	return 1;
 }
 
diff --git a/arch/x86/kvm/svm/svm.h b/arch/x86/kvm/svm/svm.h
index a0e94a2c51a1..a559cd45c8a9 100644
--- a/arch/x86/kvm/svm/svm.h
+++ b/arch/x86/kvm/svm/svm.h
@@ -221,6 +221,13 @@ struct svm_nested_state {
 	 * on its side.
 	 */
 	bool force_msr_bitmap_recalc;
+
+	/*
+	 * Indicates that a legacy nested state was restored (without a
+	 * valid gPAT). In this mode, updates to hPAT are also applied to
+	 * gPAT, preserving the old behavior where L2 shared L1's PAT.
+	 */
+	bool legacy_gpat_semantics;
 };
 
 struct vcpu_sev_es_state {
@@ -604,6 +611,8 @@ static inline void svm_set_hpat(struct vcpu_svm *svm, u64 data)
 		if (is_guest_mode(&svm->vcpu) && !nested_npt_enabled(svm))
 			svm_set_vmcb_gpat(svm->nested.vmcb02.ptr, data);
 	}
+	if (svm->nested.legacy_gpat_semantics)
+		svm_set_gpat(svm, data);
 }
 
 static inline bool nested_vnmi_enabled(struct vcpu_svm *svm)
-- 
2.53.0.rc2.204.g2597b5adb4-goog
Re: [PATCH v3 7/8] KVM: x86: nSVM: Handle restore of legacy nested state
Posted by Sean Christopherson 2 days, 17 hours ago
On Thu, Feb 05, 2026, Jim Mattson wrote:
> When nested NPT is enabled and KVM_SET_NESTED_STATE is used to restore an
> old checkpoint (without a valid gPAT), the current IA32_PAT value must be
> used as L2's gPAT.
> 
> The current IA32_PAT value may be restored by KVM_SET_MSRS after
> KVM_SET_NESTED_STATE. Furthermore, there may be a KVM_GET_NESTED_STATE
> before the first KVM_RUN.
> 
> Introduce a new boolean, svm->nested.legacy_gpat_semantics. When set, hPAT
> updates are also applied to gPAT, preserving the old behavior where L2
> shared L1's PAT. svm_vcpu_pre_run() clears this boolean at the first
> KVM_RUN.

State this last point as a command and explain why.  E.g. I think this is why?

  Clear legacy_gpat_semantics on KVM_RUN so that the legacy semantics are
  scoped to a single restore (inasmuch as possible).  E.g. to support
  restoring a snapshot from an old KVM, and then later restoring a snapshot
  from a new KVM.
Re: [PATCH v3 7/8] KVM: x86: nSVM: Handle restore of legacy nested state
Posted by Jim Mattson 2 days, 14 hours ago
On Fri, Feb 6, 2026 at 11:18 AM Sean Christopherson <seanjc@google.com> wrote:
>
> On Thu, Feb 05, 2026, Jim Mattson wrote:
> > When nested NPT is enabled and KVM_SET_NESTED_STATE is used to restore an
> > old checkpoint (without a valid gPAT), the current IA32_PAT value must be
> > used as L2's gPAT.
> >
> > The current IA32_PAT value may be restored by KVM_SET_MSRS after
> > KVM_SET_NESTED_STATE. Furthermore, there may be a KVM_GET_NESTED_STATE
> > before the first KVM_RUN.
> >
> > Introduce a new boolean, svm->nested.legacy_gpat_semantics. When set, hPAT
> > updates are also applied to gPAT, preserving the old behavior where L2
> > shared L1's PAT. svm_vcpu_pre_run() clears this boolean at the first
> > KVM_RUN.
>
> State this last point as a command and explain why.  E.g. I think this is why?
>
>   Clear legacy_gpat_semantics on KVM_RUN so that the legacy semantics are
>   scoped to a single restore (inasmuch as possible).  E.g. to support
>   restoring a snapshot from an old KVM, and then later restoring a snapshot
>   from a new KVM.

Actually, it's just because I'd like to return to normal behavior once
the non-atomic restore operation is complete. I'll add that note.