Skip to contentSkip to conversation

KVM: SEV: Add support for the ALLOWED_SEV_FEATURES feature

PATCH Applied Kim Phillips 3 participants 5 messages Mar 10, 2025 – Apr 25, 2025

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

v5 cover KVM: SEV: Add support for the ALLOWED_SEV_FEATURES feature

AMD EPYC 5th generation processors have introduced a feature that allows the hypervisor to control the SEV_FEATURES that are set for, or by, a guest [1]. ALLOWED_SEV_FEATURES can be used by the hypervisor to enforce that SEV-ES and SEV-SNP guests cannot enable features that the hypervisor does not want to be enabled.

Patch 1/2 adds support to detect the feature.

Patch 2/2 configures the ALLOWED_SEV_FEATURES field in the VMCB according to the features the hypervisor supports.

Tested SNP by setting random feature bits to the sev_features assignment in wakeup_cpu_via_vmgexit() (but not its ghcb_set_rax).

Tested SEV-ES by manipulating the save->sev_features assignment in sev_es_sync_vmsa(). Note that SEV-ES "allows" operation only works on features available in SEV-ES, i.e., it ignores SNP-only features. Zen5 SEV-ES features are DEBUG_SWAP, PREVENT_HOST_IBS, VMGEXIT_PARAMETER, PMC_VIRTUALIZATION, and IBS_VIRTUALIZATION.

Based on x86-kvm/next.

[1] Section 15.36.20 "Allowed SEV Features", AMD64 Architecture

    Programmer's Manual, Pub. 24593 Rev. 3.42 - March 2024:
    https://bugzilla.kernel.org/attachment.cgi?id=306250

v5:
- Add reviews-by
- Add the two new vmcb fields to dump_vmcb() output (Pankaj)
- Address comment by Tom and put single assignment in sev_es_init_vmcb

v4: https://lore.kernel.org/kvm/[email protected]/
- Revert the user-opt in (Sean, sorry for the misunderstanding)
- this basically undoes v3 uAPI changes and makes the feature always-on, if available
- rebased on top of x86-kvm/next

v3: https://lore.kernel.org/kvm/[email protected]/
- Assign allowed_sev_features based on user-provided vmsa_features mask (Sean)
- Users now have to explicitly opt-in with a qemu "allowed-sev-features=on" switch.
- Rebased on top of 6.14-rc1 and reworked authorship chain (tglx)

v2: https://lore.kernel.org/lkml/[email protected]/
- Added some SEV_FEATURES require to be explicitly allowed by ALLOWED_SEV_FEATURES wording (Sean).
- Added Nikunj's Reviewed-by.

v1: https://lore.kernel.org/lkml/[email protected]/

Kim Phillips (1):
KVM: SEV: Configure "ALLOWED_SEV_FEATURES" VMCB Field

Kishon Vijay Abraham I (1):
x86/cpufeatures: Add "Allowed SEV Features" Feature

 arch/x86/include/asm/cpufeatures.h | 1 +
 arch/x86/include/asm/svm.h         | 7 ++++++-
 arch/x86/kvm/svm/sev.c             | 5 +++++
 arch/x86/kvm/svm/svm.c             | 2 ++
 4 files changed, 14 insertions(+), 1 deletion(-)

base-commit: c9ea48bb6ee6b28bbc956c1e8af98044618fed5e

Signature · 1 line
2.43.0
#1 lore raw

PATCH v5 1/2 x86/cpufeatures: Add "Allowed SEV Features" Feature

From: Kishon Vijay Abraham I <[email protected]>

Add CPU feature detection for "Allowed SEV Features" to allow the Hypervisor to enforce that SEV-ES and SEV-SNP guest VMs cannot enable features (via SEV_FEATURES) that the Hypervisor does not support or wish to be enabled.

---
 arch/x86/include/asm/cpufeatures.h | 1 +
 1 file changed, 1 insertion(+)
arch/x86/include/asm/cpufeatures.h +1 -0
index 8f8aaf94dc00..6a12c8c48bd2 100644
@@ -454,6 +454,7 @@
 #define X86_FEATURE_DEBUG_SWAP		(19*32+14) /* "debug_swap" SEV-ES full debug state swap support */
 #define X86_FEATURE_RMPREAD		(19*32+21) /* RMPREAD instruction */
 #define X86_FEATURE_SEGMENTED_RMP	(19*32+23) /* Segmented RMP support */
+#define X86_FEATURE_ALLOWED_SEV_FEATURES (19*32+27) /* Allowed SEV Features */
 #define X86_FEATURE_SVSM		(19*32+28) /* "svsm" SVSM present */
 #define X86_FEATURE_HV_INUSE_WR_ALLOWED	(19*32+30) /* Allow Write to in-use hypervisor-owned pages */
 
Signature · 1 line
2.43.0
#2 lore raw

PATCH v5 2/2 KVM: SEV: Configure "ALLOWED_SEV_FEATURES" VMCB Field

In reply to Kim Phillips

AMD EPYC 5th generation processors have introduced a feature that allows the hypervisor to control the SEV_FEATURES that are set for, or by, a guest [1]. ALLOWED_SEV_FEATURES can be used by the hypervisor to enforce that SEV-ES and SEV-SNP guests cannot enable features that the hypervisor does not want to be enabled.

Always enable ALLOWED_SEV_FEATURES. A VMRUN will fail if any non-reserved bits are 1 in SEV_FEATURES but are 0 in ALLOWED_SEV_FEATURES.

Some SEV_FEATURES - currently PmcVirtualization and SecureAvic (see Appendix B, Table B-4) - require an opt-in via ALLOWED_SEV_FEATURES, i.e. are off-by-default, whereas all other features are effectively on-by-default, but still honor ALLOWED_SEV_FEATURES.

[1] Section 15.36.20 "Allowed SEV Features", AMD64 Architecture

    Programmer's Manual, Pub. 24593 Rev. 3.42 - March 2024:
    https://bugzilla.kernel.org/attachment.cgi?id=306250
---
 arch/x86/include/asm/svm.h | 7 ++++++-
 arch/x86/kvm/svm/sev.c     | 5 +++++
 arch/x86/kvm/svm/svm.c     | 2 ++
 3 files changed, 13 insertions(+), 1 deletion(-)
arch/x86/include/asm/svm.h +6 -1
index 9b7fa99ae951..b382fd251e5b 100644
@@ -159,7 +159,10 @@ struct __attribute__ ((__packed__)) vmcb_control_area {
 	u64 avic_physical_id;	/* Offset 0xf8 */
 	u8 reserved_7[8];
 	u64 vmsa_pa;		/* Used for an SEV-ES guest */
-	u8 reserved_8[720];
+	u8 reserved_8[40];
+	u64 allowed_sev_features;	/* Offset 0x138 */
+	u64 guest_sev_features;		/* Offset 0x140 */
+	u8 reserved_9[664];
 	/*
 	 * Offset 0x3e0, 32 bytes reserved
 	 * for use by hypervisor/software.
@@ -291,6 +294,8 @@ static_assert((X2AVIC_MAX_PHYSICAL_ID & AVIC_PHYSICAL_MAX_INDEX_MASK) == X2AVIC_
 #define SVM_SEV_FEAT_ALTERNATE_INJECTION		BIT(4)
 #define SVM_SEV_FEAT_DEBUG_SWAP				BIT(5)
 
+#define VMCB_ALLOWED_SEV_FEATURES_VALID			BIT_ULL(63)
+
 struct vmcb_seg {
 	u16 selector;
 	u16 attrib;
arch/x86/kvm/svm/sev.c +5 -0
index 0bc708ee2788..f9ec139901ef 100644
@@ -4449,6 +4449,7 @@ void sev_vcpu_after_set_cpuid(struct vcpu_svm *svm)
 
 static void sev_es_init_vmcb(struct vcpu_svm *svm)
 {
+	struct kvm_sev_info *sev = to_kvm_sev_info(svm->vcpu.kvm);
 	struct vmcb *vmcb = svm->vmcb01.ptr;
 	struct kvm_vcpu *vcpu = &svm->vcpu;
 
@@ -4464,6 +4465,10 @@ static void sev_es_init_vmcb(struct vcpu_svm *svm)
 	if (svm->sev_es.vmsa && !svm->sev_es.snp_has_guest_vmsa)
 		svm->vmcb->control.vmsa_pa = __pa(svm->sev_es.vmsa);
 
+	if (cpu_feature_enabled(X86_FEATURE_ALLOWED_SEV_FEATURES))
+		svm->vmcb->control.allowed_sev_features = sev->vmsa_features |
+							  VMCB_ALLOWED_SEV_FEATURES_VALID;
+
 	/* Can't intercept CR register access, HV can't modify CR registers */
 	svm_clr_intercept(svm, INTERCEPT_CR0_READ);
 	svm_clr_intercept(svm, INTERCEPT_CR4_READ);
arch/x86/kvm/svm/svm.c +2 -0
index 8abeab91d329..bff6e9c34586 100644
@@ -3435,6 +3435,8 @@ static void dump_vmcb(struct kvm_vcpu *vcpu)
 	pr_err("%-20s%016llx\n", "avic_logical_id:", control->avic_logical_id);
 	pr_err("%-20s%016llx\n", "avic_physical_id:", control->avic_physical_id);
 	pr_err("%-20s%016llx\n", "vmsa_pa:", control->vmsa_pa);
+	pr_err("%-20s%016llx\n", "allowed_sev_features:", control->allowed_sev_features);
+	pr_err("%-20s%016llx\n", "guest_sev_features:", control->guest_sev_features);
 	pr_err("VMCB State Save Area:\n");
 	pr_err("%-5s s: %04x a: %04x l: %08x b: %016llx\n",
 	       "es:",
Signature · 1 line
2.43.0
Reviewed-by on v5 2/2
#3 lore raw

Re: [PATCH v5 2/2] KVM: SEV: Configure "ALLOWED_SEV_FEATURES" VMCB Field

Quoted 15 lines · Kim Phillips

AMD EPYC 5th generation processors have introduced a feature that allows the hypervisor to control the SEV_FEATURES that are set for, or by, a guest [1]. ALLOWED_SEV_FEATURES can be used by the hypervisor to enforce that SEV-ES and SEV-SNP guests cannot enable features that the hypervisor does not want to be enabled.

Always enable ALLOWED_SEV_FEATURES. A VMRUN will fail if any non-reserved bits are 1 in SEV_FEATURES but are 0 in ALLOWED_SEV_FEATURES.

Some SEV_FEATURES - currently PmcVirtualization and SecureAvic (see Appendix B, Table B-4) - require an opt-in via ALLOWED_SEV_FEATURES, i.e. are off-by-default, whereas all other features are effectively on-by-default, but still honor ALLOWED_SEV_FEATURES.

Quoted 8 lines

[1] Section 15.36.20 "Allowed SEV Features", AMD64 Architecture

    Programmer's Manual, Pub. 24593 Rev. 3.42 - March 2024:
    https://bugzilla.kernel.org/attachment.cgi?id=306250
Quoted 67 lines
---
 arch/x86/include/asm/svm.h | 7 ++++++-
 arch/x86/kvm/svm/sev.c     | 5 +++++
 arch/x86/kvm/svm/svm.c     | 2 ++
 3 files changed, 13 insertions(+), 1 deletion(-)
arch/x86/include/asm/svm.h +6 -1
index 9b7fa99ae951..b382fd251e5b 100644
@@ -159,7 +159,10 @@ struct __attribute__ ((__packed__)) vmcb_control_area {
 	u64 avic_physical_id;	/* Offset 0xf8 */
 	u8 reserved_7[8];
 	u64 vmsa_pa;		/* Used for an SEV-ES guest */
-	u8 reserved_8[720];
+	u8 reserved_8[40];
+	u64 allowed_sev_features;	/* Offset 0x138 */
+	u64 guest_sev_features;		/* Offset 0x140 */
+	u8 reserved_9[664];
 	/*
 	 * Offset 0x3e0, 32 bytes reserved
 	 * for use by hypervisor/software.
@@ -291,6 +294,8 @@ static_assert((X2AVIC_MAX_PHYSICAL_ID & AVIC_PHYSICAL_MAX_INDEX_MASK) == X2AVIC_
 #define SVM_SEV_FEAT_ALTERNATE_INJECTION		BIT(4)
 #define SVM_SEV_FEAT_DEBUG_SWAP				BIT(5)
 
+#define VMCB_ALLOWED_SEV_FEATURES_VALID			BIT_ULL(63)
+
 struct vmcb_seg {
 	u16 selector;
 	u16 attrib;
arch/x86/kvm/svm/sev.c +5 -0
index 0bc708ee2788..f9ec139901ef 100644
@@ -4449,6 +4449,7 @@ void sev_vcpu_after_set_cpuid(struct vcpu_svm *svm)
 
 static void sev_es_init_vmcb(struct vcpu_svm *svm)
 {
+	struct kvm_sev_info *sev = to_kvm_sev_info(svm->vcpu.kvm);
 	struct vmcb *vmcb = svm->vmcb01.ptr;
 	struct kvm_vcpu *vcpu = &svm->vcpu;
 
@@ -4464,6 +4465,10 @@ static void sev_es_init_vmcb(struct vcpu_svm *svm)
 	if (svm->sev_es.vmsa && !svm->sev_es.snp_has_guest_vmsa)
 		svm->vmcb->control.vmsa_pa = __pa(svm->sev_es.vmsa);
 
+	if (cpu_feature_enabled(X86_FEATURE_ALLOWED_SEV_FEATURES))
+		svm->vmcb->control.allowed_sev_features = sev->vmsa_features |
+							  VMCB_ALLOWED_SEV_FEATURES_VALID;
+
 	/* Can't intercept CR register access, HV can't modify CR registers */
 	svm_clr_intercept(svm, INTERCEPT_CR0_READ);
 	svm_clr_intercept(svm, INTERCEPT_CR4_READ);
arch/x86/kvm/svm/svm.c +2 -0
index 8abeab91d329..bff6e9c34586 100644
@@ -3435,6 +3435,8 @@ static void dump_vmcb(struct kvm_vcpu *vcpu)
 	pr_err("%-20s%016llx\n", "avic_logical_id:", control->avic_logical_id);
 	pr_err("%-20s%016llx\n", "avic_physical_id:", control->avic_physical_id);
 	pr_err("%-20s%016llx\n", "vmsa_pa:", control->vmsa_pa);
+	pr_err("%-20s%016llx\n", "allowed_sev_features:", control->allowed_sev_features);
+	pr_err("%-20s%016llx\n", "guest_sev_features:", control->guest_sev_features);
 	pr_err("VMCB State Save Area:\n");
 	pr_err("%-5s s: %04x a: %04x l: %08x b: %016llx\n",
 	       "es:",
maintainer Applied
#4 lore raw
In reply to Kim Phillips
Quoted 9 lines · Kim Phillips

AMD EPYC 5th generation processors have introduced a feature that allows the hypervisor to control the SEV_FEATURES that are set for, or by, a guest [1]. ALLOWED_SEV_FEATURES can be used by the hypervisor to enforce that SEV-ES and SEV-SNP guests cannot enable features that the hypervisor does not want to be enabled.

Patch 1/2 adds support to detect the feature.

[...]

Applied to kvm-x86 svm, thanks!

[1/2] x86/cpufeatures: Add "Allowed SEV Features" Feature

      https://github.com/kvm-x86/linux/commit/67e672eb2e89
[2/2] KVM: SEV: Configure "ALLOWED_SEV_FEATURES" VMCB Field
      https://github.com/kvm-x86/linux/commit/51c4b387555d

--
https://github.com/kvm-x86/linux/tree/next