Skip to contentSkip to conversation

Enable Shadow Stack Virtualization for SVM

PATCH In review John Allen 2 participants 12 messages Sep 8, 2025 – Sep 15, 2025

Subsystems (3)
lore
Share & tools
lore

John Allen

author
#0 lore raw

v4 cover Enable Shadow Stack Virtualization for SVM

AMD Zen3 and newer processors support shadow stack, a feature designed to protect against ROP (return-oriented programming) attacks in which an attacker manipulates return addresses on the call stack in order to execute arbitrary code. To prevent this, shadow stacks can be allocated that are only used by control transfer and return instructions. When a CALL instruction is issued, it writes the return address to both the program stack and the shadow stack. When the subsequent RET instruction is issued, it pops the return address from both stacks and compares them. If the addresses don't match, a control-protection exception is raised.

Shadow stack and a related feature, Indirect Branch Tracking (IBT), are collectively referred to as Control-flow Enforcement Technology (CET). However, current AMD processors only support shadow stack and not IBT.

This series adds support for shadow stack in SVM guests and builds upon the support added in the CET guest support patch series [1]. Additional patches are required to support shadow stack enabled guests in qemu [2].

[1]: CET guest support patches (v13)
https://lore.kernel.org/all/[email protected]/

[2]: CET qemu patches
https://lore.kernel.org/all/[email protected]/

[3]: Previous SVM support patches (v3)
https://lore.kernel.org/all/[email protected]/

---

RFC v2:
- Rebased on v3 of the Intel CET virtualization series, dropping the patch that moved cet_is_msr_accessible to common code as that has been pulled into the Intel series.
- Minor change removing curly brackets around if statement introduced

    in patch 6/6.

RFC v3:
- Rebased on v5 of the Intel CET virtualization series.
- Add patch changing the name of vmplX_ssp SEV-ES save area fields to plX_ssp.
- Merge this series intended for KVM with the separate guest kernel

    patch (now patch 7/8).

- Update MSR passthrough code to conditionally pass through shadow stack MSRS based on both host and guest support.
- Don't save PL0_SSP, PL1_SSP, and PL2_SSP MSRs on SEV-ES VMRUN as these are currently unused.
v1:
- Remove RFC tag from series
- Rebase on v6 of the Intel CET virtualization series
- Use KVM-governed feature to track SHSTK for SVM
v2:
- Add new patch renaming boot_*msr to raw_*msr. Utilize raw_rdmsr when reading XSS on SEV-ES cpuid instructions.
- Omit unnecessary patch for saving shadow stack msrs on SEV-ES VMRUN
- Omit passing through of XSS for SEV-ES as support has already been properly implemented in a26b7cd22546 ("KVM: SEV: Do not intercept accesses to MSR_IA32_XSS for SEV-ES guests")
v3:
- Rebased on v11 of the Intel CET Virtualization series.
- Split guest kernel patches into a separate series as these are independent of this series and are needed to support non-KVM hypervisors.
v4:
- Rebased on v13 of the Intel CET Virtualization series.
- Add SEV-ES save area fields to dump_vmcb.
- Don't pass through MSR_IA32_INT_SSP_TAB.
- Don't remove clearing of IBT capability.

John Allen (5):
KVM: x86: SVM: Emulate reads and writes to shadow stack MSRs
KVM: x86: SVM: Update dump_vmcb with shadow stack save area additions
KVM: x86: SVM: Pass through shadow stack MSRs
KVM: SVM: Add MSR_IA32_XSS to the GHCB for hypervisor kernel
KVM: SVM: Enable shadow stack virtualization for SVM

 arch/x86/kvm/svm/sev.c |  5 +++++
 arch/x86/kvm/svm/svm.c | 43 +++++++++++++++++++++++++++++++++++++++---
 arch/x86/kvm/svm/svm.h |  1 +
 3 files changed, 46 insertions(+), 3 deletions(-)
Signature · 1 line
2.47.3
author
#1 lore raw

PATCH v4 1/5 KVM: x86: SVM: Emulate reads and writes to shadow stack MSRs

Emulate shadow stack MSR access by reading and writing to the corresponding fields in the VMCB.

---
 arch/x86/kvm/svm/svm.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
arch/x86/kvm/svm/svm.c +18 -0
index e4af4907c7d8..fee60f3378e1 100644
@@ -2767,6 +2767,15 @@ static int svm_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 		if (guest_cpuid_is_intel_compatible(vcpu))
 			msr_info->data |= (u64)svm->sysenter_esp_hi << 32;
 		break;
+	case MSR_IA32_S_CET:
+		msr_info->data = svm->vmcb->save.s_cet;
+		break;
+	case MSR_IA32_INT_SSP_TAB:
+		msr_info->data = svm->vmcb->save.isst_addr;
+		break;
+	case MSR_KVM_INTERNAL_GUEST_SSP:
+		msr_info->data = svm->vmcb->save.ssp;
+		break;
 	case MSR_TSC_AUX:
 		msr_info->data = svm->tsc_aux;
 		break;
@@ -2999,6 +3008,15 @@ static int svm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr)
 		svm->vmcb01.ptr->save.sysenter_esp = (u32)data;
 		svm->sysenter_esp_hi = guest_cpuid_is_intel_compatible(vcpu) ? (data >> 32) : 0;
 		break;
+	case MSR_IA32_S_CET:
+		svm->vmcb->save.s_cet = data;
+		break;
+	case MSR_IA32_INT_SSP_TAB:
+		svm->vmcb->save.isst_addr = data;
+		break;
+	case MSR_KVM_INTERNAL_GUEST_SSP:
+		svm->vmcb->save.ssp = data;
+		break;
 	case MSR_TSC_AUX:
 		/*
 		 * TSC_AUX is always virtualized for SEV-ES guests when the
Signature · 1 line
2.47.3
#2 lore raw

Re: [PATCH v4 1/5] KVM: x86: SVM: Emulate reads and writes to shadow stack MSRs

Quoted 29 lines · John Allen

Emulate shadow stack MSR access by reading and writing to the corresponding fields in the VMCB.

---
 arch/x86/kvm/svm/svm.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
arch/x86/kvm/svm/svm.c +9 -0
index e4af4907c7d8..fee60f3378e1 100644
@@ -2767,6 +2767,15 @@ static int svm_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 		if (guest_cpuid_is_intel_compatible(vcpu))
 			msr_info->data |= (u64)svm->sysenter_esp_hi << 32;
 		break;
+	case MSR_IA32_S_CET:
+		msr_info->data = svm->vmcb->save.s_cet;
+		break;
+	case MSR_IA32_INT_SSP_TAB:
+		msr_info->data = svm->vmcb->save.isst_addr;
+		break;
+	case MSR_KVM_INTERNAL_GUEST_SSP:
+		msr_info->data = svm->vmcb->save.ssp;
+		break;
 	case MSR_TSC_AUX:
 		msr_info->data = svm->tsc_aux;
 		break;

arch/x86/kvm/svm/svm.c static int svm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr) · line 3012

Quoted 5 lines
arch/x86/kvm/svm/svm.c +2 -0
@@ -2999,6 +3008,15 @@ static int svm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr)
 		svm->vmcb01.ptr->save.sysenter_esp = (u32)data;
 		svm->sysenter_esp_hi = guest_cpuid_is_intel_compatible(vcpu) ? (data >> 32) : 0;
 		break;
+	case MSR_IA32_S_CET:
+		svm->vmcb->save.s_cet = data;

These writes should mark VMCB_CET (the dirty/clean flag) dirty, and obviously KVM should mark VMCB_CET clean along with everything else on #VMEXIT.

Quoted 13 lines
arch/x86/kvm/svm/svm.c +7 -0
+		break;
+	case MSR_IA32_INT_SSP_TAB:
+		svm->vmcb->save.isst_addr = data;
+		break;
+	case MSR_KVM_INTERNAL_GUEST_SSP:
+		svm->vmcb->save.ssp = data;
+		break;
 	case MSR_TSC_AUX:
 		/*
 		 * TSC_AUX is always virtualized for SEV-ES guests when the
Signature · 1 line
2.47.3
author
#3 lore raw

PATCH v4 2/5 KVM: x86: SVM: Update dump_vmcb with shadow stack save area additions

In reply to John Allen

Add shadow stack VMCB fields to dump_vmcb. PL0_SSP, PL1_SSP, PL2_SSP, PL3_SSP, and U_CET are part of the SEV-ES save area and are encrypted, but can be decrypted and dumped if the guest policy allows debugging.

---
v4:
- Dump shstk fields in sev-es save area.

---
 arch/x86/kvm/svm/svm.c | 11 +++++++++++
 1 file changed, 11 insertions(+)
arch/x86/kvm/svm/svm.c +11 -0
index fee60f3378e1..aee1bb8c01d0 100644
@@ -3407,6 +3407,10 @@ static void dump_vmcb(struct kvm_vcpu *vcpu)
 	       "rip:", save->rip, "rflags:", save->rflags);
 	pr_err("%-15s %016llx %-13s %016llx\n",
 	       "rsp:", save->rsp, "rax:", save->rax);
+	pr_err("%-15s %016llx %-13s %016llx\n",
+	       "s_cet:", save->s_cet, "ssp:", save->ssp);
+	pr_err("%-15s %016llx\n",
+	       "isst_addr:", save->isst_addr);
 	pr_err("%-15s %016llx %-13s %016llx\n",
 	       "star:", save01->star, "lstar:", save01->lstar);
 	pr_err("%-15s %016llx %-13s %016llx\n",
@@ -3431,6 +3435,13 @@ static void dump_vmcb(struct kvm_vcpu *vcpu)
 		pr_err("%-15s %016llx\n",
 		       "sev_features", vmsa->sev_features);
 
+		pr_err("%-15s %016llx %-13s %016llx\n",
+		       "pl0_ssp:", vmsa->pl0_ssp, "pl1_ssp:", vmsa->pl1_ssp);
+		pr_err("%-15s %016llx %-13s %016llx\n",
+		       "pl2_ssp:", vmsa->pl2_ssp, "pl3_ssp:", vmsa->pl3_ssp);
+		pr_err("%-15s %016llx\n",
+		       "u_cet:", vmsa->u_cet);
+
 		pr_err("%-15s %016llx %-13s %016llx\n",
 		       "rax:", vmsa->rax, "rbx:", vmsa->rbx);
 		pr_err("%-15s %016llx %-13s %016llx\n",
Signature · 1 line
2.47.3
author
#4 lore raw

PATCH v4 3/5 KVM: x86: SVM: Pass through shadow stack MSRs

In reply to John Allen

Pass through XSAVE managed CET MSRs on SVM when KVM supports shadow stack. These cannot be intercepted without also intercepting XSAVE which would likely cause unacceptable performance overhead. MSR_IA32_INT_SSP_TAB is not managed by XSAVE, so it is intercepted.

---
v4:
- Don't pass through MSR_IA32_INT_SSP_TAB

---
 arch/x86/kvm/svm/svm.c | 11 +++++++++++
 1 file changed, 11 insertions(+)
arch/x86/kvm/svm/svm.c +11 -0
index aee1bb8c01d0..b18573b530aa 100644
@@ -844,6 +844,17 @@ static void svm_recalc_msr_intercepts(struct kvm_vcpu *vcpu)
 		svm_disable_intercept_for_msr(vcpu, MSR_IA32_MPERF, MSR_TYPE_R);
 	}
 
+	if (kvm_cpu_cap_has(X86_FEATURE_SHSTK)) {
+		bool shstk_enabled = guest_cpu_cap_has(vcpu, X86_FEATURE_SHSTK);
+
+		svm_set_intercept_for_msr(vcpu, MSR_IA32_U_CET, MSR_TYPE_RW, !shstk_enabled);
+		svm_set_intercept_for_msr(vcpu, MSR_IA32_S_CET, MSR_TYPE_RW, !shstk_enabled);
+		svm_set_intercept_for_msr(vcpu, MSR_IA32_PL0_SSP, MSR_TYPE_RW, !shstk_enabled);
+		svm_set_intercept_for_msr(vcpu, MSR_IA32_PL1_SSP, MSR_TYPE_RW, !shstk_enabled);
+		svm_set_intercept_for_msr(vcpu, MSR_IA32_PL2_SSP, MSR_TYPE_RW, !shstk_enabled);
+		svm_set_intercept_for_msr(vcpu, MSR_IA32_PL3_SSP, MSR_TYPE_RW, !shstk_enabled);
+	}
+
 	if (sev_es_guest(vcpu->kvm))
 		sev_es_recalc_msr_intercepts(vcpu);
 
Signature · 1 line
2.47.3
author
#5 lore raw

PATCH v4 4/5 KVM: SVM: Add MSR_IA32_XSS to the GHCB for hypervisor kernel

In reply to John Allen

When a guest issues a cpuid instruction for Fn0000000D_x0B_{x00,x01}, KVM will be intercepting the CPUID instruction and will need to access the guest MSR_IA32_XSS value. For SEV-ES, the XSS value is encrypted and needs to be included in the GHCB to be visible to the hypervisor.

---
v2:
- Omit passing through XSS as this has already been properly implemented in a26b7cd22546 ("KVM: SEV: Do not intercept accesses to MSR_IA32_XSS for SEV-ES guests")
v3:
- Move guest kernel GHCB_ACCESSORS definition to new series.
v4:
- Change logic structure to be more intuitive.

---
 arch/x86/kvm/svm/sev.c | 5 +++++
 arch/x86/kvm/svm/svm.h | 1 +
 2 files changed, 6 insertions(+)
arch/x86/kvm/svm/sev.c +5 -0
index f4381878a9e5..33c42dd853b3 100644
@@ -3310,6 +3310,11 @@ static void sev_es_sync_from_ghcb(struct vcpu_svm *svm)
 		vcpu->arch.cpuid_dynamic_bits_dirty = true;
 	}
 
+	if (kvm_ghcb_xss_is_valid(svm)) {
+		vcpu->arch.ia32_xss = ghcb_get_xss(ghcb);
+		vcpu->arch.cpuid_dynamic_bits_dirty = true;
+	}
+
 	/* Copy the GHCB exit information into the VMCB fields */
 	exit_code = ghcb_get_sw_exit_code(ghcb);
 	control->exit_code = lower_32_bits(exit_code);
arch/x86/kvm/svm/svm.h +1 -0
index 3c7f208b7935..552c58b050f1 100644
@@ -935,5 +935,6 @@ DEFINE_KVM_GHCB_ACCESSORS(sw_exit_info_1)
 DEFINE_KVM_GHCB_ACCESSORS(sw_exit_info_2)
 DEFINE_KVM_GHCB_ACCESSORS(sw_scratch)
 DEFINE_KVM_GHCB_ACCESSORS(xcr0)
+DEFINE_KVM_GHCB_ACCESSORS(xss)
 
 #endif
Signature · 1 line
2.47.3
#6 lore raw

Re: [PATCH v4 4/5] KVM: SVM: Add MSR_IA32_XSS to the GHCB for hypervisor kernel

Quoted 5 lines · John Allen

When a guest issues a cpuid instruction for Fn0000000D_x0B_{x00,x01}, KVM will be intercepting the CPUID instruction and will need to access the guest MSR_IA32_XSS value. For SEV-ES, the XSS value is encrypted and needs to be included in the GHCB to be visible to the hypervisor.

Quoted 8 lines

---
v2:
- Omit passing through XSS as this has already been properly implemented in a26b7cd22546 ("KVM: SEV: Do not intercept accesses to MSR_IA32_XSS for SEV-ES guests")
v3:
- Move guest kernel GHCB_ACCESSORS definition to new series.

Except that broke _this_ series.

arch/x86/kvm/svm/sev.c: In function ‘sev_es_sync_from_ghcb’:
arch/x86/kvm/svm/sev.c:3293:39: error: implicit declaration of function ‘ghcb_get_xss’;
did you mean ‘ghcb_get_rsi’? [-Wimplicit-function-declaration]

 3293 |                 vcpu->arch.ia32_xss = ghcb_get_xss(ghcb);
      |                                       ^~~~~~~~~~~~
      |                                       ghcb_get_rsi

AR drivers/base/built-in.a
AR drivers/built-in.a

Quoted 12 lines

v4:
- Change logic structure to be more intuitive.

---
 arch/x86/kvm/svm/sev.c | 5 +++++
 arch/x86/kvm/svm/svm.h | 1 +
 2 files changed, 6 insertions(+)
arch/x86/kvm/svm/sev.c +0 -0
index f4381878a9e5..33c42dd853b3 100644

arch/x86/kvm/svm/sev.c static void sev_es_sync_from_ghcb(struct vcpu_svm *svm) · line 3314

Quoted 5 lines
arch/x86/kvm/svm/sev.c +2 -0
@@ -3310,6 +3310,11 @@ static void sev_es_sync_from_ghcb(struct vcpu_svm *svm)
 		vcpu->arch.cpuid_dynamic_bits_dirty = true;
 	}
 
+	if (kvm_ghcb_xss_is_valid(svm)) {
+		vcpu->arch.ia32_xss = ghcb_get_xss(ghcb);

Honestly, I think the ghcb_get_xxx() helpers do more harm than good. For set() and if_valid(), I'm totally on board with a wrapper. For get(), unless we WARN on trying to read an invalid field, I just don't see the point. Ugh, and we _can't_ WARN, at least not in KVM, because of the whole TOCTOU mess.

Case in point, this and the xcr0 check can elide setting cpuid_dynamic_bits_dirty if XCR0/XSS isn't actually changing, but then this

	if (kvm_ghcb_xcr0_is_valid(svm) && vcpu->arch.xcr0 != ghcb_get_xcr0(ghcb)) {
		vcpu->arch.xcr0 = ghcb_get_xcr0(ghcb);
		vcpu->arch.cpuid_dynamic_bits_dirty = true;
	}

looks wonky unless the reader knows that ghcb_get_xcr0() is just reading a struct field, which obviously isn't terribly difficult to figure out, but the macros make it more than a bit annoying.

Argh, even worse, that check is technically subject to a TOCTOU bug as well. It just doesn't matter in practice because the guest can only hose it self, e.g. by swizzling XCR0/XSS. But it's still flawed.

And for both XCR0/XSS, KVM lets the guest throw garbage into vcpu->arch.xcr0 and now vcpu->arch.xss. Maybe that's not problematic in practice, but I'd rather not find out the hard way.

Lastly, open coding the write to cpuid_dynamic_bits_dirty and vcpu->arch.xcr0 is just gross.

So to avoid a rather pointless dependency for CET, which I'm trying my darndest to land in 6.18, I'm going to put together a separate fixup patch and replace this patchh to end up with code that does:

	if (kvm_ghcb_xcr0_is_valid(svm)
		__kvm_set_xcr(vcpu, 0, kvm_ghcb_get_xcr0(ghcb));

	if (kvm_ghcb_xss_is_valid(svm))
		__kvm_emulate_msr_write(vcpu, MSR_IA32_XSS, kvm_ghcb_get_xss(ghcb));
author
#7 lore raw

Re: [PATCH v4 4/5] KVM: SVM: Add MSR_IA32_XSS to the GHCB for hypervisor kernel

Quoted 17 lines · Sean Christopherson
Quoted 5 lines · John Allen

When a guest issues a cpuid instruction for Fn0000000D_x0B_{x00,x01}, KVM will be intercepting the CPUID instruction and will need to access the guest MSR_IA32_XSS value. For SEV-ES, the XSS value is encrypted and needs to be included in the GHCB to be visible to the hypervisor.

Quoted 8 lines

---
v2:
- Omit passing through XSS as this has already been properly implemented in a26b7cd22546 ("KVM: SEV: Do not intercept accesses to MSR_IA32_XSS for SEV-ES guests")
v3:
- Move guest kernel GHCB_ACCESSORS definition to new series.

Except that broke _this_ series.

Quoted 8 lines

arch/x86/kvm/svm/sev.c: In function ‘sev_es_sync_from_ghcb’:
arch/x86/kvm/svm/sev.c:3293:39: error: implicit declaration of function ‘ghcb_get_xss’;
did you mean ‘ghcb_get_rsi’? [-Wimplicit-function-declaration]

 3293 |                 vcpu->arch.ia32_xss = ghcb_get_xss(ghcb);
      |                                       ^~~~~~~~~~~~
      |                                       ghcb_get_rsi

AR drivers/base/built-in.a
AR drivers/built-in.a

Apologies, that series should be considered a prerequisite for this series. I pulled the guest kernel patch into a separate series since it doesn't depend on the main series and we ideally would want it to be pulled in ASAP rather than wait on the rest of the series since it enables linux guests running on non-KVM hypervisors.

Thanks,
John

author
#8 lore raw

PATCH v4 5/5 KVM: SVM: Enable shadow stack virtualization for SVM

In reply to John Allen

Remove the explicit clearing of shadow stack CPU capabilities.

---
v4:
- Don't remove clearing of IBT feature.

---
 arch/x86/kvm/svm/svm.c | 3 ---
 1 file changed, 3 deletions(-)
arch/x86/kvm/svm/svm.c +0 -3
index b18573b530aa..304531d6c8b0 100644
@@ -5262,10 +5262,7 @@ static __init void svm_set_cpu_caps(void)
 	kvm_set_cpu_caps();
 
 	kvm_caps.supported_perf_cap = 0;
-	kvm_caps.supported_xss = 0;
 
-	/* KVM doesn't yet support CET virtualization for SVM. */
-	kvm_cpu_cap_clear(X86_FEATURE_SHSTK);
 	kvm_cpu_cap_clear(X86_FEATURE_IBT);
 
 	/* CPUID 0x80000001 and 0x8000000A (SVM features) */
Signature · 1 line
2.47.3
#9 lore raw
In reply to John Allen
Quoted 1 line · John Allen

This series adds support for shadow stack in SVM guests

                  ^
                  |
                some

I mean, who cares about nested, right?

Sorry for being snippy, but I am more than a bit peeved that we're effectively on revision 6 of this series, and apparently no one has thought to do even basic tested of nested SVM. And I'm even more grumpy that writing tests continues to be low priority in general, which is especially concerning for such a large, complex feature.

Adding support for nested was easy enough (famous last words), but I really wish I could get back the ~hour I spent figuring out what was missing...

author
#10 lore raw
Quoted 11 lines · Sean Christopherson
Quoted 1 line · John Allen

This series adds support for shadow stack in SVM guests

                  ^
                  |
                some

I mean, who cares about nested, right?

Sorry for being snippy, but I am more than a bit peeved that we're effectively on revision 6 of this series, and apparently no one has thought to do even basic tested of nested SVM.

Hi Sean,

I have been testing nested with this feature (or so I thought). Can you explain what you did to test and what wasn't working?

Apologies, and thanks for taking the time to look into the problem.

Thanks,
John

#11 lore raw
Quoted 9 lines · John Allen
Quoted 8 lines · Sean Christopherson
Quoted 1 line · John Allen

This series adds support for shadow stack in SVM guests

                  ^
                  |
                some

I mean, who cares about nested, right?

Quoted 7 lines
Quoted 3 lines

Sorry for being snippy, but I am more than a bit peeved that we're effectively on revision 6 of this series, and apparently no one has thought to do even basic tested of nested SVM.

Hi Sean,

I have been testing nested with this feature (or so I thought).

The issue here is that Linux only supports shadow stacks at CPL3, i.e. only exercises MSR_IA32_U_CET, and for whatever reason the KVM-Unit-Test only tests MSR_IA32_U_CET too (and is stupidly not compatible with AMD due to requiring SHSTK *and* IBT). So just running those in nested won't provide any coverage for S_CET.

Quoted 1 line

Can you explain what you did to test and what wasn't working?

Read/write MSR_IA32_S_CET to a non-zero value from L2 by running the proposed selftest[*] in L1. Because KVM doesn't propagate S_CET to/from vmcb12, the writes from L2 are effectively lost.

An ever better way to cover this would be a selftest or KUT test to explicitly read/write MSRs in L2, and/or fill vmcs12/vmcb12 from L1 and verify L2 sees the desired value.

https://lore.kernel.org/all/[email protected]

Quoted 1 line

Apologies, and thanks for taking the time to look into the problem.

No worries, I didn't intend to single you out, I was essentially just yelling at everyone involved :-)