[PATCH] KVM: SVM: Fix svm the vmexit error_code of WRMSR

[PATCH] KVM: SVM: Fix svm the vmexit error_code of WRMSR

[Haiwei Li]

 From 1f755f75dfd73ad7cabb0e0f43e9993dd9f69120 Mon Sep 17 00:00:00 2001
From: Haiwei Li <lihaiwei@tencent.com>
Date: Mon, 2 Mar 2020 19:19:59 +0800
Subject: [PATCH] KVM: SVM: Fix svm the vmexit error_code of WRMSR

In svm, exit_code of write_msr is not EXIT_REASON_MSR_WRITE which
belongs to vmx.

According to amd manual, SVM_EXIT_MSR(7ch) is the exit_code of VMEXIT_MSR
due to RDMSR or WRMSR access to protected MSR. Additionally, the processor
indicates in the VMCB's EXITINFO1 whether a RDMSR(EXITINFO1=0) or
WRMSR(EXITINFO1=1) was intercepted.

Signed-off-by: Haiwei Li <lihaiwei@tencent.com>
---
  arch/x86/kvm/svm.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
index fd3fc9f..ef71755 100644
--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -6296,7 +6296,8 @@ static void svm_handle_exit_irqoff(struct kvm_vcpu
*vcpu,
         enum exit_fastpath_completion *exit_fastpath)
  {
         if (!is_guest_mode(vcpu) &&
-               to_svm(vcpu)->vmcb->control.exit_code ==
EXIT_REASON_MSR_WRITE)
+               (to_svm(vcpu)->vmcb->control.exit_code == SVM_EXIT_MSR) &&
+               (to_svm(vcpu)->vmcb->control.exit_info_1 & 1))
                 *exit_fastpath = handle_fastpath_set_msr_irqoff(vcpu);
  }

--
1.8.3.1

Re: [PATCH] KVM: SVM: Fix svm the vmexit error_code of WRMSR

[Vitaly Kuznetsov]

Haiwei Li <lihaiwei.kernel@gmail.com> writes:

>  From 1f755f75dfd73ad7cabb0e0f43e9993dd9f69120 Mon Sep 17 00:00:00 2001
> From: Haiwei Li <lihaiwei@tencent.com>
> Date: Mon, 2 Mar 2020 19:19:59 +0800
> Subject: [PATCH] KVM: SVM: Fix svm the vmexit error_code of WRMSR
>
> In svm, exit_code of write_msr is not EXIT_REASON_MSR_WRITE which
> belongs to vmx.

EXIT_REASON_MSR_WRITE is '32', in SVM this corresponds to
SVM_EXIT_READ_DR0. There were issues I guess. Or did you only detect
that the fastpath is not working?

>
> According to amd manual, SVM_EXIT_MSR(7ch) is the exit_code of VMEXIT_MSR
> due to RDMSR or WRMSR access to protected MSR. Additionally, the processor
> indicates in the VMCB's EXITINFO1 whether a RDMSR(EXITINFO1=0) or
> WRMSR(EXITINFO1=1) was intercepted.
>
> Signed-off-by: Haiwei Li <lihaiwei@tencent.com>

Fixes: 1e9e2622a149 ("KVM: VMX: FIXED+PHYSICAL mode single target IPI fastpath")

> ---
>   arch/x86/kvm/svm.c | 3 ++-
>   1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
> index fd3fc9f..ef71755 100644
> --- a/arch/x86/kvm/svm.c
> +++ b/arch/x86/kvm/svm.c
> @@ -6296,7 +6296,8 @@ static void svm_handle_exit_irqoff(struct kvm_vcpu
> *vcpu,
>          enum exit_fastpath_completion *exit_fastpath)
>   {
>          if (!is_guest_mode(vcpu) &&
> -               to_svm(vcpu)->vmcb->control.exit_code ==
> EXIT_REASON_MSR_WRITE)

There is an extra newline here (in case it's not just me).

> +               (to_svm(vcpu)->vmcb->control.exit_code == SVM_EXIT_MSR) &&
> +               (to_svm(vcpu)->vmcb->control.exit_info_1 & 1))

Could we add defines for '1' and '0', like
SVM_EXITINFO_MSR_WRITE/SVM_EXITINFO_MSR_READ maybe?

>                  *exit_fastpath = handle_fastpath_set_msr_irqoff(vcpu);
>   }
>
> --
> 1.8.3.1
>

-- 
Vitaly

Re: [PATCH] KVM: SVM: Fix svm the vmexit error_code of WRMSR

[Paolo Bonzini]

On 02/03/20 13:54, Vitaly Kuznetsov wrote:
>>          enum exit_fastpath_completion *exit_fastpath)
>>   {
>>          if (!is_guest_mode(vcpu) &&
>> -               to_svm(vcpu)->vmcb->control.exit_code ==
>> EXIT_REASON_MSR_WRITE)
> There is an extra newline here (in case it's not just me).

Yes, the whole patch has broken newlines.  I fixed it up and applied.

>> +               (to_svm(vcpu)->vmcb->control.exit_code == SVM_EXIT_MSR) &&
>> +               (to_svm(vcpu)->vmcb->control.exit_info_1 & 1))
>
> Could we add defines for '1' and '0', like
> SVM_EXITINFO_MSR_WRITE/SVM_EXITINFO_MSR_READ maybe?

We can eliminate "& 1" completely since that's what msr_interception does.

Paolo