Re: [PATCH v2 1/2] cpufreq: intel_pstate: Fix scaling for hybrid-capable CPUs not reporting hybrid

[srinivas pandruvada <srinivas.pandruvada@linux.intel.com>]

On Fri, 2026-05-08 at 14:30 +0800, Henry Tseng wrote:
> Commit 9b18d536b124 ("cpufreq: intel_pstate: Use CPPC to get scaling
> factors") restructured hwp_get_cpu_scaling() so that, when the CPU
> model is registered in intel_hybrid_scaling_factor[] and
> hybrid_get_cpu_type() does not return INTEL_CPU_TYPE_CORE, the
> function early-returns core_get_scaling() (100000) instead of
> falling through to intel_pstate_cppc_get_scaling().
> 
> This regresses the behavior previously addressed by
> commit 0fcfc9e51990 ("cpufreq: intel_pstate: Fix scaling for
> hybrid-capable systems with disabled E-cores").  On hybrid-capable
> processors not reporting X86_FEATURE_HYBRID_CPU (no E-cores
> enumerated), hybrid_get_cpu_type() returns 0.  Before 9b18d536b124,
> such CPUs fell through to intel_pstate_cppc_get_scaling(), which
> returned the registered hybrid_scaling_factor.  After 9b18d536b124,
> for models registered in the table the early return takes
> core_get_scaling() before reaching that fallback, so the registered
> factor is not used.
> 
> Fix this by returning hybrid_scaling_factor directly for CPU models
> registered in intel_hybrid_scaling_factor[] that do not report
> X86_FEATURE_HYBRID_CPU, before checking hybrid_get_cpu_type().
> 
> With that fix, non-hybrid CPUs in the table now reach
> intel_pstate_hybrid_hwp_adjust() with scaling != perf_ctl_scaling.
> Since commit 5313ec4a215a ("cpufreq: intel_pstate: Improve printing
> of
> debug messages") sets hwp_is_hybrid = true unconditionally inside
> that
> function, set hwp_is_hybrid based on X86_FEATURE_HYBRID_CPU instead,
> so the flag is not set on non-hybrid systems.
> 
> Fixes: 9b18d536b124 ("cpufreq: intel_pstate: Use CPPC to get scaling
> factors")
> Signed-off-by: Henry Tseng <henrytseng@qnap.com>
> ---
>  drivers/cpufreq/intel_pstate.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/cpufreq/intel_pstate.c
> b/drivers/cpufreq/intel_pstate.c
> index 1292da53e5fc..d39592e86570 100644
> --- a/drivers/cpufreq/intel_pstate.c
> +++ b/drivers/cpufreq/intel_pstate.c
> @@ -585,7 +585,7 @@ static void intel_pstate_hybrid_hwp_adjust(struct
> cpudata *cpu)
>  	if (scaling == perf_ctl_scaling)
>  		return;
>  
> -	hwp_is_hybrid = true;
> +	hwp_is_hybrid = cpu_feature_enabled(X86_FEATURE_HYBRID_CPU);
>  
>  	cpu->pstate.turbo_freq = rounddown(cpu->pstate.turbo_pstate
> * scaling,
>  					   perf_ctl_scaling);
> @@ -2275,6 +2275,9 @@ static int knl_get_turbo_pstate(int cpu)
>  static int hwp_get_cpu_scaling(int cpu)
>  {
>  	if (hybrid_scaling_factor) {
> +		if (!cpu_feature_enabled(X86_FEATURE_HYBRID_CPU))
> +			return hybrid_scaling_factor;

It breaks Alder Lake systems with only E-cores and no P-cores. So need
to think about this. So this change didn't restore the behavior of
0fcfc9e51990.
So need to think about this.

Thanks,
Srinivas


> +
>  		/*
>  		 * Return the hybrid scaling factor for P-cores and
> use the
>  		 * default core scaling for E-cores.