Re: [PATCH] cpufreq: amd_pstate: Fix initial highest_perf value

From: "Rafael J. Wysocki" <rafael@kernel.org>
To: Perry Yuan <Perry.Yuan@amd.com>
Cc: Rafael Wysocki <rafael.j.wysocki@intel.com>,
	Huang Rui <ray.huang@amd.com>,
	Viresh Kumar <viresh.kumar@linaro.org>,
	Deepak Sharma <Deepak.Sharma@amd.com>,
	Mario Limonciello <Mario.Limonciello@amd.com>,
	Nathan Fontenot <Nathan.Fontenot@amd.com>,
	Alex Deucher <Alexander.Deucher@amd.com>,
	"Su, Jinzhou (Joe)" <Jinzhou.Su@amd.com>,
	Shimmer.Huang@amd.com, "Du, Xiaojian" <Xiaojian.Du@amd.com>,
	"Meng, Li (Jassmine)" <Li.Meng@amd.com>,
	Linux PM <linux-pm@vger.kernel.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] cpufreq: amd_pstate: Fix initial highest_perf value
Date: Thu, 25 Aug 2022 13:58:55 +0200	[thread overview]
Message-ID: <CAJZ5v0gdHvLLPrvj7cq2ki6xDH=X6qpXBfyVTFNa5wNqY7Sh5g@mail.gmail.com> (raw)
In-Reply-To: <20220814164421.330891-1-Perry.Yuan@amd.com>

On Sun, Aug 14, 2022 at 6:57 PM Perry Yuan <Perry.Yuan@amd.com> wrote:
>
> To avoid some new AMD processors use wrong highest perf when amd pstate
> driver loaded, this fix will query the highest perf from MSR register
> MSR_AMD_CPPC_CAP1 and cppc_acpi interface firstly, then compare with the
> highest perf value got by calling amd_get_highest_perf() function.
>
> The lower value will be the correct highest perf we need to use.
> Otherwise the CPU max MHz will be incorrect if the
> amd_get_highest_perf() did not cover the new process family and model ID.
>
> Like this lscpu info, the max frequency is incorrect.
>
> Vendor ID:               AuthenticAMD
>     Socket(s):           1
>     Stepping:            2
>     CPU max MHz:         5410.0000
>     CPU min MHz:         400.0000
>     BogoMIPS:            5600.54
>
> Fixes: 3743d55b289c2 (x86, sched: Fix the AMD CPPC maximum performance value on certain AMD Ryzen generations)
> Signed-off-by: Perry Yuan <Perry.Yuan@amd.com>

This needs to be reviewed by the AMD P-state driver maintainer.

> ---
>  drivers/cpufreq/amd-pstate.c | 14 ++++++++++++--
>  1 file changed, 12 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/cpufreq/amd-pstate.c b/drivers/cpufreq/amd-pstate.c
> index 9f4375f7ab46..f778ab37e436 100644
> --- a/drivers/cpufreq/amd-pstate.c
> +++ b/drivers/cpufreq/amd-pstate.c
> @@ -354,6 +354,7 @@ static inline int amd_pstate_enable(bool enable)
>  static int pstate_init_perf(struct amd_cpudata *cpudata)
>  {
>         u64 cap1;
> +       u32 highest_perf;
>
>         int ret = rdmsrl_safe_on_cpu(cpudata->cpu, MSR_AMD_CPPC_CAP1,
>                                      &cap1);
> @@ -365,7 +366,11 @@ static int pstate_init_perf(struct amd_cpudata *cpudata)
>          *
>          * CPPC entry doesn't indicate the highest performance in some ASICs.
>          */
> -       WRITE_ONCE(cpudata->highest_perf, amd_get_highest_perf());
> +       highest_perf = amd_get_highest_perf();
> +       if (highest_perf > AMD_CPPC_HIGHEST_PERF(cap1))
> +               highest_perf = AMD_CPPC_HIGHEST_PERF(cap1);
> +
> +       WRITE_ONCE(cpudata->highest_perf, highest_perf);
>
>         WRITE_ONCE(cpudata->nominal_perf, AMD_CPPC_NOMINAL_PERF(cap1));
>         WRITE_ONCE(cpudata->lowest_nonlinear_perf, AMD_CPPC_LOWNONLIN_PERF(cap1));
> @@ -377,12 +382,17 @@ static int pstate_init_perf(struct amd_cpudata *cpudata)
>  static int cppc_init_perf(struct amd_cpudata *cpudata)
>  {
>         struct cppc_perf_caps cppc_perf;
> +       u32 highest_perf;
>
>         int ret = cppc_get_perf_caps(cpudata->cpu, &cppc_perf);
>         if (ret)
>                 return ret;
>
> -       WRITE_ONCE(cpudata->highest_perf, amd_get_highest_perf());
> +       highest_perf = amd_get_highest_perf();
> +       if (highest_perf > cppc_perf.highest_perf)
> +               highest_perf = cppc_perf.nominal_perf;
> +
> +       WRITE_ONCE(cpudata->highest_perf, highest_perf);
>
>         WRITE_ONCE(cpudata->nominal_perf, cppc_perf.nominal_perf);
>         WRITE_ONCE(cpudata->lowest_nonlinear_perf,
> --
> 2.34.1
>