Re: [PATCH v4 01/21] KVM: SVM: Drop counter MSR checks in amd_is_valid_msr()
From: Yosry Ahmed
Date: Tue Nov 17 2026 - 19:28:37 EST
On Mon, Nov 16, 2026 at 01:02:49PM -0600, Chen, Zide wrote:
>
>
> On 11/14/2026 12:49 PM, Yosry Ahmed wrote:
> > On Thu, Nov 12, 2026 at 01:32:07PM -0700, Zide Chen wrote:
> >> amd_msr_idx_to_pmc() accurately validates individual counter MSRs by
> >> checking whether the corresponding counter is supported for the vCPU.
> >> Validating counter MSRs again in amd_is_valid_msr() is redundant.
> >>
> >> Furthermore, validating counter MSRs using broad checks like
> >> "pmu->version > 0" or "PERFCTR_CORE" is inaccurate and misleading. It
> >> falsely validates unexposed counters, which then fail in
> >> amd_pmu_{get,set}_msr(), injecting an errant #GP instead of taking
> >> KVM's fallback (read 0, write ignored) and breaking migration for
> >> unadvertised counters.
> >>
> >> Finally, the range check in the default case is dead code because AMD
> >> CPUs currently support at most 6 Core Perf counters. Even if future AMD
> >> CPUs support more than 6, counter MSRs would still be sufficiently
> >> validated by amd_msr_idx_to_pmc().
> >>
> >> Drop the counter cases and dead range check from amd_is_valid_msr(),
> >> leaving counter validation solely to amd_msr_idx_to_pmc().
> >>
> >> Fixes: 4a2771895ca6 ("KVM: x86/svm/pmu: Add AMD PerfMonV2 support")
> >> Reported-by: Jim Mattson <jmattson@xxxxxxxxxx>
> >> Signed-off-by: Zide Chen <zide.chen@xxxxxxxxx>
> >
> > Should we also drop the get_gp_pmc(.., MSR_IA32_PERFCTR0),
> > get_gp_pmc(.., MSR_P6_EVNTSEL0), and get_fixed_pmc() calls from the
> > default case in intel_is_valid_msr()?
> >
> > They are also covered by intel_msr_idx_to_pmc() IIUC.
> It's a shame that I didn't take one step further to check the Intel code.
FWIW, for the patch itself:
Reviewed-by: Yosry Ahmed <yosry@xxxxxxxxxx>
>
> I completely agree that intel_is_valid_msr() should be simplified:
Looks good, I assume that would be a separate patch.
>
> diff --git a/arch/x86/kvm/vmx/pmu_intel.c b/arch/x86/kvm/vmx/pmu_intel.c
> index 4824e7eb98d3..77ce713101a4 100644
> --- a/arch/x86/kvm/vmx/pmu_intel.c
> +++ b/arch/x86/kvm/vmx/pmu_intel.c
> @@ -229,10 +229,7 @@ static bool intel_is_valid_msr(struct kvm_vcpu
> *vcpu, u32 msr)
> ((perf_capabilities & PERF_CAP_PEBS_FORMAT) > 3);
> break;
> default:
> - ret = get_gp_pmc(pmu, msr, MSR_IA32_PERFCTR0) ||
> - get_gp_pmc(pmu, msr, MSR_P6_EVNTSEL0) ||
> - get_fixed_pmc(pmu, msr) || get_fw_gp_pmc(pmu,
> msr) ||
> - intel_pmu_is_valid_lbr_msr(vcpu, msr);
> + ret = intel_pmu_is_valid_lbr_msr(vcpu, msr);
> break;
> }
>
> Additionally, intel_msr_idx_to_pmc() currently misses get_fw_gp_pmc(),
> which leaves FW counter writes untracked in kvm_pmu_mark_pmc_in_use().
> Moving get_fw_gp_pmc() into intel_msr_idx_to_pmc() fixes that bug and
> allows dropping all counter checks from intel_is_valid_msr():
>
> @@ -247,6 +244,7 @@ static struct kvm_pmc *intel_msr_idx_to_pmc(struct
> kvm_vcpu *vcpu, u32 msr)
> pmc = get_fixed_pmc(pmu, msr);
> pmc = pmc ? pmc : get_gp_pmc(pmu, msr, MSR_P6_EVNTSEL0);
> pmc = pmc ? pmc : get_gp_pmc(pmu, msr, MSR_IA32_PERFCTR0);
> + pmc = pmc ? pmc : get_fw_gp_pmc(pmu, msr);
>
> return pmc;
> }
>
> This seems a little bit too complicated in intel_msr_idx_to_pmc(),
> especially once PMU v6 is introduced and more v6 specific MSRs are
> available. We may do some cleanup (not in this series though)
Yeah we can clean this up later.
>
> - Fold get_fw_gp_pmc() into get_gp_pmc() and hide the base MSR from the
> API signature to keep the callers clean.
> - Move these helpers from pmu.h to pmu_intel.c, since AMD already uses
> its own helpers.
>
> @@ -245,10 +294,9 @@ static struct kvm_pmc *intel_msr_idx_to_pmc(struct
> kvm_vcpu *vcpu, u32 msr)
> struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
> struct kvm_pmc *pmc;
>
> - pmc = get_fixed_pmc(pmu, msr);
> - pmc = pmc ? pmc : get_gp_pmc(pmu, msr, MSR_P6_EVNTSEL0);
> - pmc = pmc ? pmc : get_gp_pmc(pmu, msr, MSR_IA32_PERFCTR0);
> - pmc = pmc ? pmc : get_fw_gp_pmc(pmu, msr);
> + pmc = intel_get_fixed_pmc(pmu, msr);
> + pmc = pmc ? pmc : intel_get_gp_eventsel_pmc(pmu, msr);
> + pmc = pmc ? pmc : intel_get_gp_counter_pmc(pmu, msr);
>
> return pmc;
> }
>
> @@ -419,18 +467,17 @@ static int intel_pmu_get_msr(struct kvm_vcpu
> *vcpu, struct msr_data *msr_info)
> msr_info->data = pmu->pebs_data_cfg;
> break;
> default:
> - if ((pmc = get_gp_pmc(pmu, msr, MSR_IA32_PERFCTR0)) ||
> - (pmc = get_gp_pmc(pmu, msr, MSR_IA32_PMC0))) {
> + if ((pmc = intel_get_gp_counter_pmc(pmu, msr))) {
> u64 val = pmc_read_counter(pmc);
> msr_info->data =
> val & pmu->counter_bitmask[KVM_PMC_GP];
>