Re: [RFC] KVM: SVM: do not drop VMCB CPL to 0 if SS is not present

From: Paolo Bonzini <pbonzini@redhat.com>
To: Roman Penyaev <roman.penyaev@profitbricks.com>,
	Mikhail Sennikovskii <mikhail.sennikovskii@profitbricks.com>,
	Gleb Natapov <gleb@kernel.org>,
	kvm@vger.kernel.org, linux-kernel@vger.kernel.org,
	Andy Lutomirski <luto@amacapital.net>
Subject: Re: [RFC] KVM: SVM: do not drop VMCB CPL to 0 if SS is not present
Date: Tue, 30 May 2017 17:13:02 +0200	[thread overview]
Message-ID: <b4f1b0c0-0db7-e399-b28b-d7f0ceada2d8@redhat.com> (raw)
In-Reply-To: <CAJrWOzD6Xq==b-zYCDdFLgSRMPM-NkNuTSDFEtX=7MreT45i7Q@mail.gmail.com>

On 19/05/2017 18:14, Roman Penyaev wrote:
> 
> 1. Simple one, KVM SVM side, which makes sure that CPL is not updated
>    if segment is unusable:
> 
>    --- a/arch/x86/kvm/svm.c
>    +++ b/arch/x86/kvm/svm.c
>    @@ -1549,7 +1549,7 @@ static void svm_set_segment(struct kvm_vcpu *vcpu,
>             * forces SS.DPL to 3 on sysret, so we ignore that case; fixing it
>             * would entail passing the CPL to userspace and back.
>             */
>    -       if (seg == VCPU_SREG_SS)
>    +       if (seg == VCPU_SREG_SS && !var->unusable)
>                    svm->vmcb->save.cpl = (s->attrib >>
> SVM_SELECTOR_DPL_SHIFT) & 3;

Based on the discussion between you and Andy, my understanding is that
it would not be enough to ensure that the attributes are preserved
across a roundtrip through KVM_GET_SEGMENT and KVM_SET_SEGMENT.  We need
a workaround in the hypervisor if we don't want to pass the CPL to
userspace and back.

Maybe if 1) in 64-bit mode 2) SS.P=0 3) SS selector != 0, then the CPL
can be taken from SS.RPL?

Thanks,

Paolo