On Fri, Aug 19, 2022, Like Xu wrote:
> From: Like Xu <likexu@xxxxxxxxxxx>
>
> On virtual platforms without PDCM support (e.g. AMD), #GP
> failure on MSR_IA32_PERF_CAPABILITIES is completely avoidable.
>
> Signed-off-by: Like Xu <likexu@xxxxxxxxxxx>
> ---
> x86/pmu.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/x86/pmu.c b/x86/pmu.c
> index 4eb92d8..25fafbe 100644
> --- a/x86/pmu.c
> +++ b/x86/pmu.c
> @@ -669,7 +669,8 @@ int main(int ac, char **av)
>
> check_counters();
>
> - if (rdmsr(MSR_IA32_PERF_CAPABILITIES) & PMU_CAP_FW_WRITES) {
> + if (this_cpu_has(X86_FEATURE_PDCM) &&
> + (rdmsr(MSR_IA32_PERF_CAPABILITIES) & PMU_CAP_FW_WRITES)) {
pmu_pebs.c ends up with similar code. Maybe add a helper?
static inline u64 this_cpu_perf_capabilities(void)
{
if (!this_cpu_has(X86_FEATURE_PDCM))
return 0;
rdmsr(MSR_IA32_PERF_CAPABILITIES);
}
Or even better, if we end up with a lib/pmu and something like "struct pmu_caps",
the RDMSR can be done once in the library's init routine.
