On 6/21/2024 12:16 AM, Sean Christopherson wrote:
> On Thu, Jun 20, 2024, Dapeng Mi wrote:
>> The existing macro, KVM_INTEL_PMC_MAX_GENERIC, ambiguously represents the
>> maximum supported General Purpose (GP) counter number for both Intel and
>> AMD platforms. This could lead to issues if AMD begins to support more GP
>> counters than Intel.
>>
>> To resolve this, a new platform-independent macro, KVM_PMC_MAX_GENERIC,
>> is introduced to represent the maximum GP counter number across all x86
>> platforms.
>>
>> No logic changes are introduced in this patch.
>>
>> Signed-off-by: Dapeng Mi <dapeng1.mi@xxxxxxxxxxxxxxx>
>> ---
>> arch/x86/include/asm/kvm_host.h | 9 +++++----
>> arch/x86/kvm/svm/pmu.c | 2 +-
>> arch/x86/kvm/vmx/pmu_intel.c | 2 ++
>> 3 files changed, 8 insertions(+), 5 deletions(-)
>>
>> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
>> index 57440bda4dc4..18137be6504a 100644
>> --- a/arch/x86/include/asm/kvm_host.h
>> +++ b/arch/x86/include/asm/kvm_host.h
>> @@ -534,11 +534,12 @@ struct kvm_pmc {
>>
>> /* More counters may conflict with other existing Architectural MSRs */
>> #define KVM_INTEL_PMC_MAX_GENERIC 8
>> -#define MSR_ARCH_PERFMON_PERFCTR_MAX (MSR_ARCH_PERFMON_PERFCTR0 + KVM_INTEL_PMC_MAX_GENERIC - 1)
>> -#define MSR_ARCH_PERFMON_EVENTSEL_MAX (MSR_ARCH_PERFMON_EVENTSEL0 + KVM_INTEL_PMC_MAX_GENERIC - 1)
>> +#define KVM_AMD_PMC_MAX_GENERIC 6
>> +#define KVM_PMC_MAX_GENERIC KVM_INTEL_PMC_MAX_GENERIC
> Since we're changing the macro, maybe take the opportunity to use a better name?
> E.g. KVM_MAX_NR_GP_COUNTERS? And then in a follow-up patch, give fixed counters
> the same treatment, e.g. KVM_MAX_NR_FIXED_COUNTERS. Or maybe KVM_MAX_NR_GP_PMCS
> and KVM_MAX_NR_FIXED_PMCS?
Yeah, would change to KVM_MAX_NR_GP_COUNTERS and KVM_MAX_NR_FIXED_COUNTERS
in next version.
>
>> +#define MSR_ARCH_PERFMON_PERFCTR_MAX (MSR_ARCH_PERFMON_PERFCTR0 + KVM_PMC_MAX_GENERIC - 1)
>> +#define MSR_ARCH_PERFMON_EVENTSEL_MAX (MSR_ARCH_PERFMON_EVENTSEL0 + KVM_PMC_MAX_GENERIC - 1)
> And I'm very, very tempted to say we should simply delete these two, along with
> MSR_ARCH_PERFMON_FIXED_CTR_MAX, and just open code the "end" MSR in the one user.
> Especially since "KVM" doesn't appear anyone in the name, i.e. because the names
> misrepresent KVM's semi-arbitrary max as the *architectural* max.
Agree. MSR_ARCH_PERFMON_PERFCTR_MAX indeed brings some misleading which
make users think it's a HW's limitation instead of KVM's limitation.
>
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 6ad19d913d31..547dfe40d017 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -7432,17 +7432,20 @@ static void kvm_probe_msr_to_save(u32 msr_index)
> intel_pt_validate_hw_cap(PT_CAP_num_address_ranges) * 2))
> return;
> break;
> - case MSR_ARCH_PERFMON_PERFCTR0 ... MSR_ARCH_PERFMON_PERFCTR_MAX:
> + case MSR_ARCH_PERFMON_PERFCTR0 ...
> + MSR_ARCH_PERFMON_PERFCTR0 + KVM_MAX_NR_GP_COUNTERS - 1:
> if (msr_index - MSR_ARCH_PERFMON_PERFCTR0 >=
> kvm_pmu_cap.num_counters_gp)
> return;
> break;
> - case MSR_ARCH_PERFMON_EVENTSEL0 ... MSR_ARCH_PERFMON_EVENTSEL_MAX:
> + case MSR_ARCH_PERFMON_EVENTSEL0 ...
> + MSR_ARCH_PERFMON_EVENTSEL0 + KVM_MAX_NR_GP_COUNTERS - 1:
> if (msr_index - MSR_ARCH_PERFMON_EVENTSEL0 >=
> kvm_pmu_cap.num_counters_gp)
> return;
> break;
> - case MSR_ARCH_PERFMON_FIXED_CTR0 ... MSR_ARCH_PERFMON_FIXED_CTR_MAX:
> + case MSR_ARCH_PERFMON_FIXED_CTR0 ...
> + MSR_ARCH_PERFMON_FIXED_CTR0 + KVM_MAR_NR_FIXED_COUNTERS - 1:
> if (msr_index - MSR_ARCH_PERFMON_FIXED_CTR0 >=
> kvm_pmu_cap.num_counters_fixed)
> return;
Thanks.
>
>> #define KVM_PMC_MAX_FIXED 3
>> #define MSR_ARCH_PERFMON_FIXED_CTR_MAX (MSR_ARCH_PERFMON_FIXED_CTR0 + KVM_PMC_MAX_FIXED - 1)
>> -#define KVM_AMD_PMC_MAX_GENERIC 6
>>
>> struct kvm_pmu {
>> u8 version;
>> @@ -554,7 +555,7 @@ struct kvm_pmu {
>> u64 global_status_rsvd;
>> u64 reserved_bits;
>> u64 raw_event_mask;
>> - struct kvm_pmc gp_counters[KVM_INTEL_PMC_MAX_GENERIC];
>> + struct kvm_pmc gp_counters[KVM_PMC_MAX_GENERIC];
>> struct kvm_pmc fixed_counters[KVM_PMC_MAX_FIXED];
>>
>> /*
>> diff --git a/arch/x86/kvm/svm/pmu.c b/arch/x86/kvm/svm/pmu.c
>> index 6e908bdc3310..2fca247798eb 100644
>> --- a/arch/x86/kvm/svm/pmu.c
>> +++ b/arch/x86/kvm/svm/pmu.c
>> @@ -218,7 +218,7 @@ static void amd_pmu_init(struct kvm_vcpu *vcpu)
>> int i;
>>
>> BUILD_BUG_ON(KVM_AMD_PMC_MAX_GENERIC > AMD64_NUM_COUNTERS_CORE);
>> - BUILD_BUG_ON(KVM_AMD_PMC_MAX_GENERIC > INTEL_PMC_MAX_GENERIC);
>> + BUILD_BUG_ON(KVM_AMD_PMC_MAX_GENERIC > KVM_PMC_MAX_GENERIC);
>>
>> for (i = 0; i < KVM_AMD_PMC_MAX_GENERIC ; i++) {
>> pmu->gp_counters[i].type = KVM_PMC_GP;
>> diff --git a/arch/x86/kvm/vmx/pmu_intel.c b/arch/x86/kvm/vmx/pmu_intel.c
>> index fb5cbd6cbeff..a4b0bee04596 100644
>> --- a/arch/x86/kvm/vmx/pmu_intel.c
>> +++ b/arch/x86/kvm/vmx/pmu_intel.c
>> @@ -570,6 +570,8 @@ static void intel_pmu_init(struct kvm_vcpu *vcpu)
>> struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
>> struct lbr_desc *lbr_desc = vcpu_to_lbr_desc(vcpu);
>>
>> + BUILD_BUG_ON(KVM_INTEL_PMC_MAX_GENERIC > KVM_PMC_MAX_GENERIC);
> Rather than BUILD_BUG_ON() for both Intel and AMD, can't we just do?
>
> #define KVM_MAX_NR_GP_COUNTERS max(KVM_INTEL_PMC_MAX_GENERIC, KVM_AMD_PMC_MAX_GENERIC)
Actually I tried this, but compiler would report the below error since
KVM_PMC_MAX_GENERIC would used to define the array
gp_counters[KVM_PMC_MAX_GENERIC];
./include/linux/minmax.h:48:50: error: braced-group within expression
allowed only inside a function
>
>> +
>> for (i = 0; i < KVM_INTEL_PMC_MAX_GENERIC; i++) {
>> pmu->gp_counters[i].type = KVM_PMC_GP;
>> pmu->gp_counters[i].vcpu = vcpu;
>> --
>> 2.34.1
>>
