[PATCH v5 33/44] KVM: x86/pmu: Bypass perf checks when emulating mediated PMU counter accesses

[Sandipan Das]

On 07-08-2025 01:26, Sean Christopherson wrote:
> From: Dapeng Mi <dapeng1.mi at linux.intel.com>
> 
> When emulating a PMC counter read or write for a mediated PMU, bypass the
> perf checks and emulated_counter logic as the counters aren't proxied
> through perf, i.e. pmc->counter always holds the guest's up-to-date value,
> and thus there's no need to defer emulated overflow checks.
> 
> Suggested-by: Sean Christopherson <seanjc at google.com>
> Signed-off-by: Dapeng Mi <dapeng1.mi at linux.intel.com>
> Co-developed-by: Mingwei Zhang <mizhang at google.com>
> Signed-off-by: Mingwei Zhang <mizhang at google.com>
> [sean: split from event filtering change, write shortlog+changelog]
> Signed-off-by: Sean Christopherson <seanjc at google.com>
> ---
>  arch/x86/kvm/pmu.c | 5 +++++
>  arch/x86/kvm/pmu.h | 3 +++
>  2 files changed, 8 insertions(+)
> 
> diff --git a/arch/x86/kvm/pmu.c b/arch/x86/kvm/pmu.c
> index 817ef852bdf9..082d2905882b 100644
> --- a/arch/x86/kvm/pmu.c
> +++ b/arch/x86/kvm/pmu.c
> @@ -377,6 +377,11 @@ static void pmc_update_sample_period(struct kvm_pmc *pmc)
>  
>  void pmc_write_counter(struct kvm_pmc *pmc, u64 val)
>  {
> +	if (kvm_vcpu_has_mediated_pmu(pmc->vcpu)) {
> +		pmc->counter = val & pmc_bitmask(pmc);
> +		return;
> +	}
> +
>  	/*
>  	 * Drop any unconsumed accumulated counts, the WRMSR is a write, not a
>  	 * read-modify-write.  Adjust the counter value so that its value is
> diff --git a/arch/x86/kvm/pmu.h b/arch/x86/kvm/pmu.h
> index 51963a3a167a..1c9d26d60a60 100644
> --- a/arch/x86/kvm/pmu.h
> +++ b/arch/x86/kvm/pmu.h
> @@ -111,6 +111,9 @@ static inline u64 pmc_read_counter(struct kvm_pmc *pmc)
>  {
>  	u64 counter, enabled, running;
>  
> +	if (kvm_vcpu_has_mediated_pmu(pmc->vcpu))
> +		return pmc->counter & pmc_bitmask(pmc);
> +
>  	counter = pmc->counter + pmc->emulated_counter;
>  
>  	if (pmc->perf_event && !pmc->is_paused)

Reviewed-by: Sandipan Das <sandipan.das at amd.com>