KVM emulates the ARCH_CAPABILITIES on x86 for both Intel and AMD
cpus, although the IA32_ARCH_CAPABILITIES MSR is an Intel-specific
MSR and it makes no sense to emulate it on AMD.

As a consequence, VMs created on AMD with qemu -cpu host and using
KVM will advertise the ARCH_CAPABILITIES feature and provide the
IA32_ARCH_CAPABILITIES MSR. This can cause issues (like Windows BSOD)
as the guest OS might not expect this MSR to exist on such cpus (the
AMD documentation specifies that ARCH_CAPABILITIES feature and MSR
are not defined on the AMD architecture).

A fix was proposed in KVM code, however KVM maintainers don't want to
change this behavior that exists for 6+ years and suggest changes to be
done in QEMU instead.  Therefore, hide the bit from "-cpu host":
migration of -cpu host guests is only possible between identical host
kernel and QEMU versions, therefore this is not a problematic breakage.

If a future AMD machine does include the MSR, that would re-expose the
Windows guest bug; but it would not be KVM/QEMU's problem at that
point, as we'd be following a genuine physical CPU impl.

Reported-by: Alexandre Chartre <alexandre.chartre@oracle.com>
Suggested-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 target/i386/kvm/kvm.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/target/i386/kvm/kvm.c b/target/i386/kvm/kvm.c
index e8c8be09bae..369626f8c8d 100644
--- a/target/i386/kvm/kvm.c
+++ b/target/i386/kvm/kvm.c
@@ -503,8 +503,12 @@ uint32_t kvm_arch_get_supported_cpuid(KVMState *s, uint32_t function,
          * Linux v4.17-v4.20 incorrectly return ARCH_CAPABILITIES on SVM hosts.
          * We can detect the bug by checking if MSR_IA32_ARCH_CAPABILITIES is
          * returned by KVM_GET_MSR_INDEX_LIST.
+         *
+         * But also, because Windows does not like ARCH_CAPABILITIES on AMD
+         * mcahines at all, do not show the fake ARCH_CAPABILITIES MSR that
+         * KVM sets up.
          */
-        if (!has_msr_arch_capabs) {
+        if (!has_msr_arch_capabs || !(edx & CPUID_7_0_EDX_ARCH_CAPABILITIES)) {
             ret &= ~CPUID_7_0_EDX_ARCH_CAPABILITIES;
         }
     } else if (function == 7 && index == 1 && reg == R_EAX) {
-- 
2.50.1

On Thu, Jul 17, 2025 at 12:34:48PM +0200, Paolo Bonzini wrote:
> Date: Thu, 17 Jul 2025 12:34:48 +0200
> From: Paolo Bonzini <pbonzini@redhat.com>
> Subject: [PATCH] target/i386: do not expose ARCH_CAPABILITIES on AMD CPU
> X-Mailer: git-send-email 2.50.1
> 
> KVM emulates the ARCH_CAPABILITIES on x86 for both Intel and AMD
> cpus, although the IA32_ARCH_CAPABILITIES MSR is an Intel-specific
> MSR and it makes no sense to emulate it on AMD.
> 
> As a consequence, VMs created on AMD with qemu -cpu host and using
> KVM will advertise the ARCH_CAPABILITIES feature and provide the
> IA32_ARCH_CAPABILITIES MSR. This can cause issues (like Windows BSOD)
> as the guest OS might not expect this MSR to exist on such cpus (the
> AMD documentation specifies that ARCH_CAPABILITIES feature and MSR
> are not defined on the AMD architecture).
> 
> A fix was proposed in KVM code, however KVM maintainers don't want to
> change this behavior that exists for 6+ years and suggest changes to be
> done in QEMU instead.  Therefore, hide the bit from "-cpu host":
> migration of -cpu host guests is only possible between identical host
> kernel and QEMU versions, therefore this is not a problematic breakage.
> 
> If a future AMD machine does include the MSR, that would re-expose the
> Windows guest bug; but it would not be KVM/QEMU's problem at that
> point, as we'd be following a genuine physical CPU impl.

Make sense.

> Reported-by: Alexandre Chartre <alexandre.chartre@oracle.com>
> Suggested-by: Daniel P. Berrangé <berrange@redhat.com>
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  target/i386/kvm/kvm.c | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)

Good to see this fix. Late but,

Reviewed-by: Zhao Liu <zhao1.liu@intel.com>

On 7/17/2025 6:34 PM, Paolo Bonzini wrote:
> KVM emulates the ARCH_CAPABILITIES on x86 for both Intel and AMD
> cpus, although the IA32_ARCH_CAPABILITIES MSR is an Intel-specific
> MSR and it makes no sense to emulate it on AMD.
> 
> As a consequence, VMs created on AMD with qemu -cpu host and using
> KVM will advertise the ARCH_CAPABILITIES feature and provide the
> IA32_ARCH_CAPABILITIES MSR. This can cause issues (like Windows BSOD)
> as the guest OS might not expect this MSR to exist on such cpus (the
> AMD documentation specifies that ARCH_CAPABILITIES feature and MSR
> are not defined on the AMD architecture).
> 
> A fix was proposed in KVM code, however KVM maintainers don't want to
> change this behavior that exists for 6+ years and suggest changes to be
> done in QEMU instead.  Therefore, hide the bit from "-cpu host":
> migration of -cpu host guests is only possible between identical host
> kernel and QEMU versions, therefore this is not a problematic breakage.
> 
> If a future AMD machine does include the MSR, that would re-expose the
> Windows guest bug; but it would not be KVM/QEMU's problem at that
> point, as we'd be following a genuine physical CPU impl.
> 
> Reported-by: Alexandre Chartre <alexandre.chartre@oracle.com>
> Suggested-by: Daniel P. Berrangé <berrange@redhat.com>
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>

This looks reasonable to me.

Though it will break the usecase of people boot VM with "-cpu 
AMD-model,+arch_capabilities", like in [1], I think it's OK break it and 
force people to not do such thing any more.

[1] https://gitlab.com/qemu-project/qemu/-/issues/3001

> ---
>   target/i386/kvm/kvm.c | 6 +++++-
>   1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/target/i386/kvm/kvm.c b/target/i386/kvm/kvm.c
> index e8c8be09bae..369626f8c8d 100644
> --- a/target/i386/kvm/kvm.c
> +++ b/target/i386/kvm/kvm.c
> @@ -503,8 +503,12 @@ uint32_t kvm_arch_get_supported_cpuid(KVMState *s, uint32_t function,
>            * Linux v4.17-v4.20 incorrectly return ARCH_CAPABILITIES on SVM hosts.
>            * We can detect the bug by checking if MSR_IA32_ARCH_CAPABILITIES is
>            * returned by KVM_GET_MSR_INDEX_LIST.
> +         *
> +         * But also, because Windows does not like ARCH_CAPABILITIES on AMD
> +         * mcahines at all, do not show the fake ARCH_CAPABILITIES MSR that
> +         * KVM sets up.
>            */
> -        if (!has_msr_arch_capabs) {
> +        if (!has_msr_arch_capabs || !(edx & CPUID_7_0_EDX_ARCH_CAPABILITIES)) {
>               ret &= ~CPUID_7_0_EDX_ARCH_CAPABILITIES;
>           }
>       } else if (function == 7 && index == 1 && reg == R_EAX) {