Re: [PATCH v3 6/9] KVM: vmx/pmu: Add MSR_ARCH_LBR_CTL emulation for Arch LBR

From: Sean Christopherson <seanjc@google.com>
To: "Xu, Like" <like.xu@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>,
	Paolo Bonzini <pbonzini@redhat.com>,
	Vitaly Kuznetsov <vkuznets@redhat.com>,
	Wanpeng Li <wanpengli@tencent.com>,
	Jim Mattson <jmattson@google.com>, Joerg Roedel <joro@8bytes.org>,
	Kan Liang <kan.liang@linux.intel.com>,
	Dave Hansen <dave.hansen@intel.com>,
	wei.w.wang@intel.com, Borislav Petkov <bp@alien8.de>,
	kvm@vger.kernel.org, x86@kernel.org,
	linux-kernel@vger.kernel.org, Like Xu <like.xu@linux.intel.com>
Subject: Re: [PATCH v3 6/9] KVM: vmx/pmu: Add MSR_ARCH_LBR_CTL emulation for Arch LBR
Date: Thu, 4 Mar 2021 08:25:56 -0800	[thread overview]
Message-ID: <YEEKFEJyLYSplm7w@google.com> (raw)
In-Reply-To: <88d2481d-5435-1ffb-dc98-5534c446bd52@intel.com>

On Thu, Mar 04, 2021, Xu, Like wrote:
> On 2021/3/4 1:19, Sean Christopherson wrote:
> > > @@ -4463,6 +4470,8 @@ static void vmx_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event)
> > >   		vmcs_writel(GUEST_SYSENTER_ESP, 0);
> > >   		vmcs_writel(GUEST_SYSENTER_EIP, 0);
> > >   		vmcs_write64(GUEST_IA32_DEBUGCTL, 0);
> > > +		if (cpu_has_vmx_arch_lbr())
> > > +			vmcs_write64(GUEST_IA32_LBR_CTL, 0);
> > Not that any guest is likely to care, but is the MSR cleared on INIT?  The SDM
> > has specific language for warm reset, but I can't find anything for INIT.
> > 
> >    On a warm reset, all LBR MSRs, including IA32_LBR_DEPTH, have their values
> >    preserved. However, IA32_LBR_CTL.LBREn is cleared to 0, disabling LBRs. If a
> >    warm reset is triggered while the processor is in C6, also known as warm init,
> >    all LBR MSRs will be reset to their initial values.
> 
> I was told that the reset behavior of GUEST_IA32_LBR_CTL
> would be the same as the GUEST_IA32_DEBUGCTL (true for INIT as well).

Yes, and DEBUGCTL is preserved on INIT.

	if (!init_event) {
		vmcs_write32(GUEST_SYSENTER_CS, 0);
		vmcs_writel(GUEST_SYSENTER_ESP, 0);
		vmcs_writel(GUEST_SYSENTER_EIP, 0);
		vmcs_write64(GUEST_IA32_DEBUGCTL, 0);
	}

Table 22-10 in the SDM:

	All Other MSRs | Pwr up or Reset: | INIT:
			 Undefined          Unchanged

If IA32_LBR_DEPTH is weirdly exempt, it needs to be documented.  I doubt that's
the case though.

> It looks we have not strictly distinguished the guest's power concept C*.
> Do we have two trap paths for "warm reset" and "warm init" ?

No.  Despite the name .vcpu_reset, KVM doesn't even have a RESET path, userspace
is responsible for modelling RESET.