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;
MIME-Version: 1.0
References: <1625825111-6604-1-git-send-email-weijiang.yang@intel.com> <1625825111-6604-6-git-send-email-weijiang.yang@intel.com>
In-Reply-To: <1625825111-6604-6-git-send-email-weijiang.yang@intel.com>
From:   Jim Mattson <jmattson@google.com>
Date:   Fri, 9 Jul 2021 14:55:35 -0700
Message-ID: <CALMp9eR8mbVXS5E6sB7TwEocytpWcG_6w-ijmfxAd4ciHPtfmw@mail.gmail.com>
Subject: Re: [PATCH v5 05/13] KVM: vmx/pmu: Emulate MSR_ARCH_LBR_CTL for guest
 Arch LBR
To:     Yang Weijiang <weijiang.yang@intel.com>
Cc:     pbonzini@redhat.com, seanjc@google.com, vkuznets@redhat.com,
        wei.w.wang@intel.com, like.xu.linux@gmail.com, kvm@vger.kernel.org,
        linux-kernel@vger.kernel.org, Like Xu <like.xu@linux.intel.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Fri, Jul 9, 2021 at 2:51 AM Yang Weijiang <weijiang.yang@intel.com> wrot=
e:
>
> From: Like Xu <like.xu@linux.intel.com>
>
> Arch LBRs are enabled by setting MSR_ARCH_LBR_CTL.LBREn to 1. A new guest
> state field named "Guest IA32_LBR_CTL" is added to enhance guest LBR usag=
e.
> When guest Arch LBR is enabled, a guest LBR event will be created like th=
e
> model-specific LBR does. Clear guest LBR enable bit on host PMI handling =
so
> guest can see expected config.
>
> On processors that support Arch LBR, MSR_IA32_DEBUGCTLMSR[bit 0] has no
> meaning. It can be written to 0 or 1, but reads will always return 0.
> Like IA32_DEBUGCTL, IA32_ARCH_LBR_CTL msr is also reserved on INIT.

I suspect you mean "preserved" rather than "reserved."

> Signed-off-by: Like Xu <like.xu@linux.intel.com>
> Signed-off-by: Yang Weijiang <weijiang.yang@intel.com>
> ---
>  arch/x86/events/intel/lbr.c      |  2 --
>  arch/x86/include/asm/msr-index.h |  1 +
>  arch/x86/include/asm/vmx.h       |  2 ++
>  arch/x86/kvm/vmx/pmu_intel.c     | 31 ++++++++++++++++++++++++++-----
>  arch/x86/kvm/vmx/vmx.c           |  9 +++++++++
>  5 files changed, 38 insertions(+), 7 deletions(-)
>

> diff --git a/arch/x86/kvm/vmx/pmu_intel.c b/arch/x86/kvm/vmx/pmu_intel.c
> index da68f0e74702..4500c564c63a 100644
> --- a/arch/x86/kvm/vmx/pmu_intel.c
> +++ b/arch/x86/kvm/vmx/pmu_intel.c
> @@ -19,6 +19,11 @@
>  #include "pmu.h"
>
>  #define MSR_PMC_FULL_WIDTH_BIT      (MSR_IA32_PMC0 - MSR_IA32_PERFCTR0)
> +/*
> + * Regardless of the Arch LBR or legacy LBR, when the LBR_EN bit 0 of th=
e
> + * corresponding control MSR is set to 1, LBR recording will be enabled.
> + */

Is this comment misplaced? It doesn't seem to have anything to do with
the macro being defined below.

> @@ -458,6 +467,14 @@ static int intel_pmu_set_msr(struct kvm_vcpu *vcpu, =
struct msr_data *msr_info)
>                 lbr_desc->records.nr =3D data;
>                 lbr_desc->arch_lbr_reset =3D true;
>                 return 0;
> +       case MSR_ARCH_LBR_CTL:
> +               if (data & ~KVM_ARCH_LBR_CTL_MASK)

Is a static mask sufficient? Per the Intel=C2=AE Architecture Instruction
Set Extensions and Future Features Programming Reference, some of
these bits may not be supported on all microarchitectures. See Table
7-8. CPUID Leaf 01CH Enumeration of Architectural LBR Capabilities.

> +                       break;
> +               vmcs_write64(GUEST_IA32_LBR_CTL, data);
> +               if (intel_pmu_lbr_is_enabled(vcpu) && !lbr_desc->event &&
> +                   (data & ARCH_LBR_CTL_LBREN))
> +                       intel_pmu_create_guest_lbr_event(vcpu);

Nothing has to be done when the LBREN bit goes from 1 to 0?

> +               return 0;
>         default:
>                 if ((pmc =3D get_gp_pmc(pmu, msr, MSR_IA32_PERFCTR0)) ||
>                     (pmc =3D get_gp_pmc(pmu, msr, MSR_IA32_PMC0))) {

Per the Intel=C2=AE Architecture Instruction Set Extensions and Future
Features Programming Reference, "IA32_LBR_CTL.LBREn is saved and
cleared on #SMI, and restored on RSM." I don't see that happening
anywhere. That manual also says, "On a warm reset...IA32_LBR_CTL.LBREn
is cleared to 0, disabling LBRs." I don't see that happening either.

I have a question about section 7.1.4.4 in that manual. It says, "On a
debug breakpoint event (#DB), IA32_LBR_CTL.LBREn is cleared." When,
exactly, does that happen? In particular, if kvm synthesizes such an
event (for example, in kvm_vcpu_do_singlestep), does
IA32_LBR_CTL.LBREn automatically get cleared (after loading the guest
IA32_LBR_CTL value from the VMCS)? Or does kvm need to explicitly
clear that bit in the VMCS before injecting the #DB?