From mboxrd@z Thu Jan  1 00:00:00 1970
From: Paolo Bonzini <pbonzini@redhat.com>
Subject: Re: KVM: nested VMX: disable perf cpuid reporting
Date: Fri, 19 Sep 2014 15:28:50 +0200
Message-ID: <541C2F92.8090706@redhat.com>
References: <20140918212457.GA9227@amt.cnet>
Mime-Version: 1.0
Content-Type: text/plain; charset=windows-1252
Content-Transfer-Encoding: 7bit
Cc: Gleb Natapov <gleb@kernel.org>, Bandan Das <bsd@redhat.com>,
	Nadav Amit <namit@cs.technion.ac.il>
To: Marcelo Tosatti <mtosatti@redhat.com>,
	kvm-devel <kvm@vger.kernel.org>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mx1.redhat.com ([209.132.183.28]:13440 "EHLO mx1.redhat.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1755831AbaISN3C (ORCPT <rfc822;kvm@vger.kernel.org>);
	Fri, 19 Sep 2014 09:29:02 -0400
In-Reply-To: <20140918212457.GA9227@amt.cnet>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

Il 18/09/2014 23:24, Marcelo Tosatti ha scritto:
> 
> Initilization of L2 guest with -cpu host, on L1 guest with -cpu host 
> triggers:
> 
> (qemu) KVM: entry failed, hardware error 0x7
> ...
> nested_vmx_run: VMCS MSR_{LOAD,STORE} unsupported
> 
> Nested VMX MSR load/store support is not sufficient to 
> allow perf for L2 guest.
> 
> Until properly fixed, trap CPUID and disable function 0xA.
> 
> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
> 
> diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
> index e28d798..976e3a5 100644
> --- a/arch/x86/kvm/cpuid.c
> +++ b/arch/x86/kvm/cpuid.c
> @@ -774,6 +774,12 @@ void kvm_cpuid(struct kvm_vcpu *vcpu, u32 *eax, u32 *ebx, u32 *ecx, u32 *edx)
>  	if (!best)
>  		best = check_cpuid_limit(vcpu, function, index);
>  
> +	/*
> +	 * Perfmon not yet supported for L2 guest.
> +	 */
> +	if (is_guest_mode(vcpu) && function == 0xa)
> +		best = NULL;
> +
>  	if (best) {
>  		*eax = best->eax;
>  		*ebx = best->ebx;
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index e8f08e9..2ab5047 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -6986,6 +6986,8 @@ static bool nested_vmx_exit_handled(struct kvm_vcpu *vcpu)
>  	case EXIT_REASON_TASK_SWITCH:
>  		return 1;
>  	case EXIT_REASON_CPUID:
> +		if (kvm_register_read(vcpu, VCPU_REGS_RAX) == 0xa)
> +			return 0;
>  		return 1;
>  	case EXIT_REASON_HLT:
>  		return nested_cpu_has(vmcs12, CPU_BASED_HLT_EXITING);
> 

Nice solution. :)

Paolo