Re: [PATCH V4 2/2] cpufreq: scmi: Register for limit change notifications

From: Vincent Guittot
Date: Tue May 28 2024 - 05:08:58 EST


Hi Sibi,

On Thu, 28 Mar 2024 at 08:42, Sibi Sankar <quic_sibis@xxxxxxxxxxx> wrote:
>
> Register for limit change notifications if supported and use the throttled
> frequency from the notification to apply HW pressure.
>
> Signed-off-by: Sibi Sankar <quic_sibis@xxxxxxxxxxx>
> ---
>
> v4:
> * Use a interim variable to show the khz calc. [Lukasz]
> * Use driver_data to pass on the handle and scmi_dev instead of using
> global variables. Dropped Lukasz's Rb due to adding these minor
> changes.
>
> drivers/cpufreq/scmi-cpufreq.c | 44 ++++++++++++++++++++++++++++++++++
> 1 file changed, 44 insertions(+)
>
> diff --git a/drivers/cpufreq/scmi-cpufreq.c b/drivers/cpufreq/scmi-cpufreq.c
> index 3b4f6bfb2f4c..d946b7a08258 100644
> --- a/drivers/cpufreq/scmi-cpufreq.c
> +++ b/drivers/cpufreq/scmi-cpufreq.c
> @@ -21,11 +21,18 @@
> #include <linux/types.h>
> #include <linux/units.h>
>
> +struct scmi_cpufreq_driver_data {
> + struct scmi_device *sdev;
> + const struct scmi_handle *handle;
> +};
> +
> struct scmi_data {
> int domain_id;
> int nr_opp;
> struct device *cpu_dev;
> + struct cpufreq_policy *policy;
> cpumask_var_t opp_shared_cpus;
> + struct notifier_block limit_notify_nb;
> };
>
> static struct scmi_protocol_handle *ph;
> @@ -174,6 +181,22 @@ static struct freq_attr *scmi_cpufreq_hw_attr[] = {
> NULL,
> };
>
> +static int scmi_limit_notify_cb(struct notifier_block *nb, unsigned long event, void *data)
> +{
> + struct scmi_data *priv = container_of(nb, struct scmi_data, limit_notify_nb);
> + struct scmi_perf_limits_report *limit_notify = data;
> + struct cpufreq_policy *policy = priv->policy;
> + unsigned int limit_freq_khz;
> +
> + limit_freq_khz = limit_notify->range_max_freq / HZ_PER_KHZ;
> +
> + policy->max = clamp(limit_freq_khz, policy->cpuinfo.min_freq, policy->cpuinfo.max_freq);
> +
> + cpufreq_update_pressure(policy);

I noticed your patch while looking for other things in the archive but
I don't think this is the right way to do it.

cpufreq_update_pressure() aims to set to the scheduler the aggregation
of all cappings set to cpufreq through the pm_qos and
freq_qos_add_request(). Calling this function directly in scmi
notification callback will overwrite the pm_qos aggregation. And at
the opposite, any update of a pm_qos constraint will overwrite scmi
notification. Instead you should better set a pm_qos constraint like
others

> +
> + return NOTIFY_OK;
> +}
> +
> static int scmi_cpufreq_init(struct cpufreq_policy *policy)
> {
> int ret, nr_opp, domain;
> @@ -181,6 +204,7 @@ static int scmi_cpufreq_init(struct cpufreq_policy *policy)
> struct device *cpu_dev;
> struct scmi_data *priv;
> struct cpufreq_frequency_table *freq_table;
> + struct scmi_cpufreq_driver_data *data = cpufreq_get_driver_data();
>
> cpu_dev = get_cpu_device(policy->cpu);
> if (!cpu_dev) {
> @@ -294,6 +318,17 @@ static int scmi_cpufreq_init(struct cpufreq_policy *policy)
> }
> }
>
> + priv->limit_notify_nb.notifier_call = scmi_limit_notify_cb;
> + ret = data->handle->notify_ops->devm_event_notifier_register(data->sdev, SCMI_PROTOCOL_PERF,
> + SCMI_EVENT_PERFORMANCE_LIMITS_CHANGED,
> + &domain,
> + &priv->limit_notify_nb);
> + if (ret)
> + dev_warn(cpu_dev,
> + "failed to register for limits change notifier for domain %d\n", domain);
> +
> + priv->policy = policy;
> +
> return 0;
>
> out_free_opp:
> @@ -366,12 +401,21 @@ static int scmi_cpufreq_probe(struct scmi_device *sdev)
> int ret;
> struct device *dev = &sdev->dev;
> const struct scmi_handle *handle;
> + struct scmi_cpufreq_driver_data *data;
>
> handle = sdev->handle;
>
> if (!handle)
> return -ENODEV;
>
> + data = devm_kzalloc(dev, sizeof(*data), GFP_KERNEL);
> + if (!data)
> + return -ENOMEM;
> +
> + data->sdev = sdev;
> + data->handle = handle;
> + scmi_cpufreq_driver.driver_data = data;
> +
> perf_ops = handle->devm_protocol_get(sdev, SCMI_PROTOCOL_PERF, &ph);
> if (IS_ERR(perf_ops))
> return PTR_ERR(perf_ops);
> --
> 2.34.1
>
>