Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Vitaly Kuznetsov <vkuznets@redhat.com>
To:     Sean Christopherson <seanjc@google.com>
Cc:     kvm@vger.kernel.org, Paolo Bonzini <pbonzini@redhat.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Jim Mattson <jmattson@google.com>,
        Oliver Upton <oupton@google.com>, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] KVM: x86: Check the right feature bit for
 MSR_KVM_ASYNC_PF_ACK access
In-Reply-To: <YPmopoGY4hwuVHAp@google.com>
References: <20210722123018.260035-1-vkuznets@redhat.com>
 <YPmopoGY4hwuVHAp@google.com>
Date:   Fri, 23 Jul 2021 08:58:33 +0200
Message-ID: <87lf5x7bza.fsf@vitty.brq.redhat.com>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

Sean Christopherson <seanjc@google.com> writes:

> On Thu, Jul 22, 2021, Vitaly Kuznetsov wrote:
>> MSR_KVM_ASYNC_PF_ACK MSR is part of interrupt based asynchronous page fault
>> interface and not the original (deprecated) KVM_FEATURE_ASYNC_PF. This is
>> stated in Documentation/virt/kvm/msr.rst.
>> 
>> Fixes: 66570e966dd9 ("kvm: x86: only provide PV features if enabled in guest's CPUID")
>> Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
>> ---
>>  arch/x86/kvm/x86.c | 4 ++--
>>  1 file changed, 2 insertions(+), 2 deletions(-)
>> 
>> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
>> index d715ae9f9108..88ff7a1af198 100644
>> --- a/arch/x86/kvm/x86.c
>> +++ b/arch/x86/kvm/x86.c
>> @@ -3406,7 +3406,7 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
>>  			return 1;
>>  		break;
>>  	case MSR_KVM_ASYNC_PF_ACK:
>> -		if (!guest_pv_has(vcpu, KVM_FEATURE_ASYNC_PF))
>> +		if (!guest_pv_has(vcpu, KVM_FEATURE_ASYNC_PF_INT))
>
> Do we want to require both, or do we want to let userspace be stupid?
>

It's OK to be stupid :-)

Thinking more about it, I'd suggest we go the other way around: allow
access to MSR_KVM_ASYNC_PF_EN when either KVM_FEATURE_ASYNC_PF or
KVM_FEATURE_ASYNC_PF_INT are present. This will allow to eventually
deprecate KVM_FEATURE_ASYNC_PF completely and switch to
KVM_FEATURE_ASYNC_PF_INT exclusively.

>>  			return 1;
>>  		if (data & 0x1) {
>>  			vcpu->arch.apf.pageready_pending = false;
>> @@ -3745,7 +3745,7 @@ int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
>>  		msr_info->data = vcpu->arch.apf.msr_int_val;
>>  		break;
>>  	case MSR_KVM_ASYNC_PF_ACK:
>> -		if (!guest_pv_has(vcpu, KVM_FEATURE_ASYNC_PF))
>> +		if (!guest_pv_has(vcpu, KVM_FEATURE_ASYNC_PF_INT))
>>  			return 1;
>>  
>>  		msr_info->data = 0;
>> -- 
>> 2.31.1
>> 
>

-- 
Vitaly