Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Vitaly Kuznetsov <vkuznets@redhat.com>
To:     Liran Alon <liran.alon@oracle.com>
Cc:     kvm@vger.kernel.org, linux-kernel@vger.kernel.org,
        Paolo Bonzini <pbonzini@redhat.com>,
        Radim =?utf-8?B?S3LEjW3DocWZ?= <rkrcmar@redhat.com>
Subject: Re: [PATCH] x86/kvm/nVMCS: fix VMCLEAR when Enlightened VMCS is in use
In-Reply-To: <69274969-E2BE-442C-B2D2-0AF94338C31B@oracle.com>
References: <20190624133028.3710-1-vkuznets@redhat.com> <CEFF2A14-611A-417C-BC0A-8814862F26C6@oracle.com> <87r27jdq68.fsf@vitty.brq.redhat.com> <69274969-E2BE-442C-B2D2-0AF94338C31B@oracle.com>
Date:   Wed, 26 Jun 2019 11:39:38 +0200
Message-ID: <87k1d8d6sl.fsf@vitty.brq.redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Liran Alon <liran.alon@oracle.com> writes:

>> On 24 Jun 2019, at 17:16, Vitaly Kuznetsov <vkuznets@redhat.com> wrote:
>> 
>> 
>> That said I'm ok with dropping nested_release_evmcs() for consistency
>> but we can't just drop 'if (vmptr == vmx->nested.hv_evmcs_vmptr)’.
>
> Right. I meant that we can just change code to:
>
> /* Add relevant comment here as this is not trivial why we do this */
> If (likely(!vmx->nested.enlightened_vmcs_enabled) ||
>     nested_enlightened_vmentry(vcpu, &evmptr)) {
>
>     if (vmptr == vmx->nested.current_vmptr)
>         nested_release_vmcs12(vcpu);
>
>     kvm_vcpu_write_guest(…);
> }
>

The change, to my surprise, resulted in a set of L2 guest crashes. After
some debugging I figured out that clean fields is to blame: after
Windows does VMCLEAR it doesn't maintain clean field data before the
next VMLAUNCH - and nested_vmx_handle_enlightened_vmptrld() does nothing
in case evmcs_vmptr stays unchanged (so VMLAUNCH follows VMCLEAR on the
same vCPU). We apparently need to invalidate clean fields data on every
VMLAUCH. This is fix of its own, I'll do more testing and send v2.

-- 
Vitaly