Date:   Mon, 27 Mar 2023 18:37:32 +0200
From:   "Rafael J. Wysocki" <rafael@...nel.org>
To:     Ricardo Neri <ricardo.neri-calderon@...ux.intel.com>
Cc:     "Peter Zijlstra (Intel)" <peterz@...radead.org>,
        Juri Lelli <juri.lelli@...hat.com>,
        Vincent Guittot <vincent.guittot@...aro.org>,
        Ricardo Neri <ricardo.neri@...el.com>,
        "Ravi V. Shankar" <ravi.v.shankar@...el.com>,
        Ben Segall <bsegall@...gle.com>,
        Daniel Bristot de Oliveira <bristot@...hat.com>,
        Dietmar Eggemann <dietmar.eggemann@....com>,
        Len Brown <len.brown@...el.com>, Mel Gorman <mgorman@...e.de>,
        "Rafael J. Wysocki" <rafael.j.wysocki@...el.com>,
        Srinivas Pandruvada <srinivas.pandruvada@...ux.intel.com>,
        Steven Rostedt <rostedt@...dmis.org>,
        Tim Chen <tim.c.chen@...ux.intel.com>,
        Valentin Schneider <vschneid@...hat.com>,
        Lukasz Luba <lukasz.luba@....com>,
        Ionela Voinescu <ionela.voinescu@....com>, x86@...nel.org,
        "Joel Fernandes (Google)" <joel@...lfernandes.org>,
        linux-kernel@...r.kernel.org, linux-pm@...r.kernel.org,
        "Tim C . Chen" <tim.c.chen@...el.com>
Subject: Re: [PATCH v3 13/24] thermal: intel: hfi: Store per-CPU IPCC scores

On Tue, Feb 7, 2023 at 6:02 AM Ricardo Neri
<ricardo.neri-calderon@...ux.intel.com> wrote:
>
> The scheduler reads the IPCC scores when balancing load. These reads can
> be quite frequent. Hardware can also update the HFI table frequently.
> Concurrent access may cause a lot of lock contention. It gets worse as the
> number of CPUs increases.
>
> Instead, create separate per-CPU IPCC scores that the scheduler can read
> without the HFI table lock.
>
> Cc: Ben Segall <bsegall@...gle.com>
> Cc: Daniel Bristot de Oliveira <bristot@...hat.com>
> Cc: Dietmar Eggemann <dietmar.eggemann@....com>
> Cc: Ionela Voinescu <ionela.voinescu@....com>
> Cc: Joel Fernandes (Google) <joel@...lfernandes.org>
> Cc: Len Brown <len.brown@...el.com>
> Cc: Lukasz Luba <lukasz.luba@....com>
> Cc: Mel Gorman <mgorman@...e.de>
> Cc: Rafael J. Wysocki <rafael.j.wysocki@...el.com>
> Cc: Srinivas Pandruvada <srinivas.pandruvada@...ux.intel.com>
> Cc: Steven Rostedt <rostedt@...dmis.org>
> Cc: Tim C. Chen <tim.c.chen@...el.com>
> Cc: Valentin Schneider <vschneid@...hat.com>
> Cc: x86@...nel.org
> Cc: linux-pm@...r.kernel.org
> Cc: linux-kernel@...r.kernel.org
> Suggested-by: Peter Zijlstra (Intel) <peterz@...radead.org>
> Signed-off-by: Ricardo Neri <ricardo.neri-calderon@...ux.intel.com>
> ---
> Changes since v2:
>  * Only create these per-CPU variables when Intel Thread Director is
>    supported.
>
> Changes since v1:
>  * Added this patch.
> ---
>  drivers/thermal/intel/intel_hfi.c | 46 +++++++++++++++++++++++++++++++
>  1 file changed, 46 insertions(+)
>
> diff --git a/drivers/thermal/intel/intel_hfi.c b/drivers/thermal/intel/intel_hfi.c
> index 2527ae3836c7..b06021828892 100644
> --- a/drivers/thermal/intel/intel_hfi.c
> +++ b/drivers/thermal/intel/intel_hfi.c
> @@ -29,6 +29,7 @@
>  #include <linux/kernel.h>
>  #include <linux/math.h>
>  #include <linux/mutex.h>
> +#include <linux/percpu.h>
>  #include <linux/percpu-defs.h>
>  #include <linux/printk.h>
>  #include <linux/processor.h>
> @@ -170,6 +171,43 @@ static struct workqueue_struct *hfi_updates_wq;
>  #define HFI_UPDATE_INTERVAL            HZ
>  #define HFI_MAX_THERM_NOTIFY_COUNT     16
>
> +#ifdef CONFIG_IPC_CLASSES

It would be good to provide a (concise) description of this variable.

> +static int __percpu *hfi_ipcc_scores;
> +
> +static int alloc_hfi_ipcc_scores(void)
> +{
> +       if (!cpu_feature_enabled(X86_FEATURE_ITD))
> +               return 0;
> +
> +       hfi_ipcc_scores = __alloc_percpu(sizeof(*hfi_ipcc_scores) *
> +                                        hfi_features.nr_classes,
> +                                        sizeof(*hfi_ipcc_scores));
> +
> +       return !hfi_ipcc_scores;

I would do

if (!hfi_ipcc_scores)
        return -ENOMEM;

return 0;

Or make the function return bool.

> +}
> +
> +static void set_hfi_ipcc_score(void *caps, int cpu)
> +{
> +       int i, *hfi_class;
> +
> +       if (!cpu_feature_enabled(X86_FEATURE_ITD))
> +               return;
> +
> +       hfi_class = per_cpu_ptr(hfi_ipcc_scores, cpu);
> +
> +       for (i = 0;  i < hfi_features.nr_classes; i++) {
> +               struct hfi_cpu_data *class_caps;
> +
> +               class_caps = caps + i * hfi_features.class_stride;
> +               WRITE_ONCE(hfi_class[i], class_caps->perf_cap);

As it stands, it is unclear why WRITE_ONCE() is needed here.

> +       }
> +}
> +
> +#else
> +static int alloc_hfi_ipcc_scores(void) { return 0; }
> +static void set_hfi_ipcc_score(void *caps, int cpu) { }
> +#endif /* CONFIG_IPC_CLASSES */
> +
>  static void get_hfi_caps(struct hfi_instance *hfi_instance,
>                          struct thermal_genl_cpu_caps *cpu_caps)
>  {
> @@ -192,6 +230,8 @@ static void get_hfi_caps(struct hfi_instance *hfi_instance,
>                 cpu_caps[i].efficiency = caps->ee_cap << 2;
>
>                 ++i;
> +
> +               set_hfi_ipcc_score(caps, cpu);
>         }
>         raw_spin_unlock_irq(&hfi_instance->table_lock);
>  }
> @@ -580,8 +620,14 @@ void __init intel_hfi_init(void)
>         if (!hfi_updates_wq)
>                 goto err_nomem;
>
> +       if (alloc_hfi_ipcc_scores())
> +               goto err_ipcc;
> +
>         return;
>
> +err_ipcc:
> +       destroy_workqueue(hfi_updates_wq);
> +
>  err_nomem:
>         for (j = 0; j < i; ++j) {
>                 hfi_instance = &hfi_instances[j];
> --

Powered by blists - more mailing lists