Re: [PATCH 05/15] KVM: VMX: Add document to state that write to uret msr should always be intercepted

From: Sean Christopherson <seanjc@google.com>
To: Paolo Bonzini <pbonzini@redhat.com>
Cc: Lai Jiangshan <jiangshanlai@gmail.com>,
	linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
	Lai Jiangshan <laijs@linux.alibaba.com>,
	Vitaly Kuznetsov <vkuznets@redhat.com>,
	Wanpeng Li <wanpengli@tencent.com>,
	Jim Mattson <jmattson@google.com>, Joerg Roedel <joro@8bytes.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	x86@kernel.org, "H. Peter Anvin" <hpa@zytor.com>
Subject: Re: [PATCH 05/15] KVM: VMX: Add document to state that write to uret msr should always be intercepted
Date: Tue, 7 Dec 2021 20:38:07 +0000	[thread overview]
Message-ID: <Ya/GL0zyobfM1rUF@google.com> (raw)
In-Reply-To: <226fc242-ae46-3214-4e01-dbfdf5f7c0fb@redhat.com>

On Thu, Nov 18, 2021, Paolo Bonzini wrote:
> On 11/18/21 12:08, Lai Jiangshan wrote:
> > From: Lai Jiangshan <laijs@linux.alibaba.com>
> > 
> > And adds a corresponding sanity check code.
> > 
> > Signed-off-by: Lai Jiangshan <laijs@linux.alibaba.com>
> > ---
> >   arch/x86/kvm/vmx/vmx.c | 10 +++++++++-
> >   1 file changed, 9 insertions(+), 1 deletion(-)
> > 
> > diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
> > index e8a41fdc3c4d..cd081219b668 100644
> > --- a/arch/x86/kvm/vmx/vmx.c
> > +++ b/arch/x86/kvm/vmx/vmx.c
> > @@ -3703,13 +3703,21 @@ void vmx_disable_intercept_for_msr(struct kvm_vcpu *vcpu, u32 msr, int type)
> >   	if (!cpu_has_vmx_msr_bitmap())
> >   		return;
> > +	/*
> > +	 * Write to uret msr should always be intercepted due to the mechanism
> > +	 * must know the current value.  Santity check to avoid any inadvertent
> > +	 * mistake in coding.
> > +	 */
> > +	if (WARN_ON_ONCE(vmx_find_uret_msr(vmx, msr) && (type & MSR_TYPE_W)))
> > +		return;
> > +
> 
> I'm not sure about this one, it's relatively expensive to call
> vmx_find_uret_msr.
> 
> User-return MSRs and disable-intercept MSRs are almost the opposite: uret is
> for MSRs that the host (not even the processor) never uses,
> disable-intercept is for MSRs that the guest reads/writes often.  As such it
> seems almost impossible that they overlap.

And they aren't fundamentally mutually exclusive, e.g. KVM could pass-through an
MSR and then do RDMSR in vmx_prepare_switch_to_host() to refresh the uret data
with the current (guest) value.  It'd be silly, but it would work.