Skip to contentSkip to conversation

KVM: x86: Fix the condition of #PF interception caused by MKTME

PATCH Applied Tao Su 3 participants 5 messages Mar 19, 2024 – Apr 10, 2024

Subsystems (3)
lore
Share & tools
lore

Tao Su

author
#0 lore raw

PATCH 1/1 KVM: x86: Fix the condition of #PF interception caused by MKTME

Intel MKTME repurposes several high bits of physical address as 'keyID', so boot_cpu_data.x86_phys_bits doesn't hold physical address bits reported by CPUID anymore.

If guest.MAXPHYADDR < host.MAXPHYADDR, the bit field of ‘keyID’ belongs to reserved bits in guest’s view, so intercepting #PF to fix error code is necessary, just replace boot_cpu_data.x86_phys_bits with kvm_get_shadow_phys_bits() to fix.

---
 arch/x86/kvm/vmx/vmx.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
arch/x86/kvm/vmx/vmx.h +3 -1
index 65786dbe7d60..79b1757df74a 100644
@@ -15,6 +15,7 @@
 #include "vmx_ops.h"
 #include "../cpuid.h"
 #include "run_flags.h"
+#include "../mmu.h"
 
 #define MSR_TYPE_R	1
 #define MSR_TYPE_W	2
@@ -719,7 +720,8 @@ static inline bool vmx_need_pf_intercept(struct kvm_vcpu *vcpu)
 	if (!enable_ept)
 		return true;
 
-	return allow_smaller_maxphyaddr && cpuid_maxphyaddr(vcpu) < boot_cpu_data.x86_phys_bits;
+	return allow_smaller_maxphyaddr &&
+		cpuid_maxphyaddr(vcpu) < kvm_get_shadow_phys_bits();
 }
 
 static inline bool is_unrestricted_guest(struct kvm_vcpu *vcpu)

base-commit: b3603fcb79b1036acae10602bffc4855a4b9af80

Signature · 1 line
2.34.1
maintainer Applied
#1 lore raw
Quoted 10 lines · Tao Su

Intel MKTME repurposes several high bits of physical address as 'keyID', so boot_cpu_data.x86_phys_bits doesn't hold physical address bits reported by CPUID anymore.

If guest.MAXPHYADDR < host.MAXPHYADDR, the bit field of ‘keyID’ belongs to reserved bits in guest’s view, so intercepting #PF to fix error code is necessary, just replace boot_cpu_data.x86_phys_bits with kvm_get_shadow_phys_bits() to fix.

[...]

Applied to kvm-x86 fixes, with a massaged shortlog/changelog.

Note, I don't love using kvm_get_shadow_phys_bits(), but only because doing CPUID every time is so pointlessly suboptimal. I have a series to clean up all of the related code, which I'll hopefully post later this week.

But I didn't see any reason to hold up this fix, as I really hope no one is using allow_smaller_maxphyaddr in a nested VM with EPT enabled, which is the only case where CPUID is likely to have a meaningful impact (due to causing a VM-Exit).

[1/1] KVM: x86: Fix the condition of #PF interception caused by MKTME

      https://github.com/kvm-x86/linux/commit/7f2817ef52a1

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

#2 lore raw
Quoted 10 lines · Sean Christopherson
Quoted 9 lines · Tao Su

Intel MKTME repurposes several high bits of physical address as 'keyID', so boot_cpu_data.x86_phys_bits doesn't hold physical address bits reported by CPUID anymore.

If guest.MAXPHYADDR < host.MAXPHYADDR, the bit field of ‘keyID’ belongs to reserved bits in guest’s view, so intercepting #PF to fix error code is necessary, just replace boot_cpu_data.x86_phys_bits with kvm_get_shadow_phys_bits() to fix.

Quoted 5 lines
Quoted 1 line

[...]

Applied to kvm-x86 fixes, with a massaged shortlog/changelog.

Note, I don't love using kvm_get_shadow_phys_bits(),

cannot agree more.

Quoted 3 lines

but only because doing
CPUID every time is so pointlessly suboptimal. I have a series to clean up all of the related code, which I'll hopefully post later this week.

Look forward to it.

Quoted 10 lines

But I didn't see any reason to hold up this fix, as I really hope no one is using allow_smaller_maxphyaddr in a nested VM with EPT enabled, which is the only case where CPUID is likely to have a meaningful impact (due to causing a VM-Exit).

[1/1] KVM: x86: Fix the condition of #PF interception caused by MKTME

       https://github.com/kvm-x86/linux/commit/7f2817ef52a1

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

#3 lore raw
In reply to Tao Su
Quoted 27 lines · Tao Su

Intel MKTME repurposes several high bits of physical address as 'keyID', so boot_cpu_data.x86_phys_bits doesn't hold physical address bits reported by CPUID anymore.

If guest.MAXPHYADDR < host.MAXPHYADDR, the bit field of ‘keyID’ belongs to reserved bits in guest’s view, so intercepting #PF to fix error code is necessary, just replace boot_cpu_data.x86_phys_bits with kvm_get_shadow_phys_bits() to fix.

---
 arch/x86/kvm/vmx/vmx.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
arch/x86/kvm/vmx/vmx.h +1 -0
index 65786dbe7d60..79b1757df74a 100644
@@ -15,6 +15,7 @@
 #include "vmx_ops.h"
 #include "../cpuid.h"
 #include "run_flags.h"
+#include "../mmu.h"
 
 #define MSR_TYPE_R	1
 #define MSR_TYPE_W	2

arch/x86/kvm/vmx/vmx.h static inline bool vmx_need_pf_intercept(struct kvm_vcpu *vcpu) · line 724

Quoted 6 lines
arch/x86/kvm/vmx/vmx.h +2 -1
@@ -719,7 +720,8 @@ static inline bool vmx_need_pf_intercept(struct kvm_vcpu *vcpu)
 	if (!enable_ept)
 		return true;
 
-	return allow_smaller_maxphyaddr && cpuid_maxphyaddr(vcpu) < boot_cpu_data.x86_phys_bits;
+	return allow_smaller_maxphyaddr &&
+		cpuid_maxphyaddr(vcpu) < kvm_get_shadow_phys_bits();

For posterity, because I had a brief moment where I thought we done messed up:

No change is needed in the reporting of MAXPHYADDR in KVM_GET_SUPPORTED_CPUID, as reporting boot_cpu_data.x86_phys_bits as MAXPHYADDR when TDP is disabled is ok because KVM always intercepts #PF when TDP is disabled, and KVM already reports the full/raw MAXPHYADDR when TDP is enabled.

author
#4 lore raw
Quoted 37 lines · Sean Christopherson
Quoted 27 lines · Tao Su

Intel MKTME repurposes several high bits of physical address as 'keyID', so boot_cpu_data.x86_phys_bits doesn't hold physical address bits reported by CPUID anymore.

If guest.MAXPHYADDR < host.MAXPHYADDR, the bit field of ‘keyID’ belongs to reserved bits in guest’s view, so intercepting #PF to fix error code is necessary, just replace boot_cpu_data.x86_phys_bits with kvm_get_shadow_phys_bits() to fix.

---
 arch/x86/kvm/vmx/vmx.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
arch/x86/kvm/vmx/vmx.h +1 -0
index 65786dbe7d60..79b1757df74a 100644
@@ -15,6 +15,7 @@
 #include "vmx_ops.h"
 #include "../cpuid.h"
 #include "run_flags.h"
+#include "../mmu.h"
 
 #define MSR_TYPE_R	1
 #define MSR_TYPE_W	2

arch/x86/kvm/vmx/vmx.h static inline bool vmx_need_pf_intercept(struct kvm_vcpu *vcpu) · line 724

Quoted 6 lines
arch/x86/kvm/vmx/vmx.h +2 -1
@@ -719,7 +720,8 @@ static inline bool vmx_need_pf_intercept(struct kvm_vcpu *vcpu)
 	if (!enable_ept)
 		return true;
 
-	return allow_smaller_maxphyaddr && cpuid_maxphyaddr(vcpu) < boot_cpu_data.x86_phys_bits;
+	return allow_smaller_maxphyaddr &&
+		cpuid_maxphyaddr(vcpu) < kvm_get_shadow_phys_bits();

For posterity, because I had a brief moment where I thought we done messed up:

Quoted 4 lines

No change is needed in the reporting of MAXPHYADDR in KVM_GET_SUPPORTED_CPUID, as reporting boot_cpu_data.x86_phys_bits as MAXPHYADDR when TDP is disabled is ok because KVM always intercepts #PF when TDP is disabled, and KVM already reports the full/raw MAXPHYADDR when TDP is enabled.

You are right, but userspace can fully control guest.MAXPHYADDR when TDP is enabled. Please see the unit-test[*], I think this issue could show up earlier if phys-bits is set larger, such as 46-bits.

[*] https://gitlab.com/kvm-unit-tests/kvm-unit-tests/-/blob/master/x86/unittests.cfg?ref_type=heads#L156