Re: [PATCH RFC 2/3] clk: cpufreq helper for voltage scaling

[Mike Turquette <mturquette@linaro.org>]

Quoting Nishanth Menon (2013-07-31 07:36:57)
> On 07/07/2013 08:44 PM, Mike Turquette wrote:
> > This patch builds on "clk: notifier handler for dynamic voltage scaling"
> > by adding a CPUfreq-specific helper function for registering a clock
> > rate-change notifier to scale regulator voltage as a function of clock
> > rate.
> >
> > In particular this patch creates an instance of struct
> > cpufreq_frequency_table and also calculate the voltage scaling latency,
> > both of which are returned to the caller for use in the CPUfreq driver.
> >
> > Signed-off-by: Mike Turquette <mturquette@linaro.org>
> > ---
> >   drivers/clk/clk-voltage-notifier.c | 71 ++++++++++++++++++++++++++++++++++++++
> >   include/linux/clk.h                |  6 ++++
> >   2 files changed, 77 insertions(+)
> 
> As stage 1 of the transition, I dont have any complaints, but for SoC 
> level, at least on OMAP, it is not just regulator and clock, we have 
> additional tweaks to do - like ABB, AVS etc.
> 
> What would you suggest as the strategy to introduce SoC specific 
> handlers for voltage notifiers that can co-exist with this?

I would suggest to use the clock rate-change notifier handlers. Every
version of these "helper" functions that I have posted are simply that:
helper functions.

There is nothing stopping you today from using the rate-change notifiers
that were merged back in v3.4 and scaling multiple regulators in a
specific order.

This patch is solving the simple case of one clock, one regulator. You
may construct your own notifier handler as needed and I have said this
many times in the past.

I would also point out that if a DT binding existed to properly express
operating points as tuples of clock/clock-rate and support for multiple
regulator/voltage pairs then I would have implemented that in this
patch. As it stands the OPP library today doesn't go into that
complexity nor does the operating-points DT binding. This patch chooses
the path of least resistance.

> 
> >
> > diff --git a/drivers/clk/clk-voltage-notifier.c b/drivers/clk/clk-voltage-notifier.c
> > index cb6b85f..8e1707f 100644
> > --- a/drivers/clk/clk-voltage-notifier.c
> > +++ b/drivers/clk/clk-voltage-notifier.c
> > @@ -10,6 +10,7 @@
> >    */
> >
> >   #include <linux/clk.h>
> > +#include <linux/cpufreq.h>
> 
> for example - we should not depend on cpufreq in a generic clock based 
> dvfs approach. we would like to allow devfreq also to transition to this.

See patch #1 in this series.

Patch #2 builds on patch #1 to handle the very common case of CPUfreq.
It is purely a helper and the goal is to reduce the duplicated
boilerplate that would go into every CPUfreq driver if this patch did
not exist.

I would also like to see devfreq drivers use this mechanism and they can
do exactly that with the stuff made available in patch #1.

There is clearly no dependency on CPUfreq for you to scale voltage from
a clock rate-change notifier.

Regards,
Mike

> 
> >   #include <linux/device.h>
> >   #include <linux/module.h>
> >   #include <linux/notifier.h>
> > @@ -133,3 +134,73 @@ void of_clk_volt_notifier_unregister(struct notifier_block *nb)
> >       kfree(vsd);
> >   }
> >   EXPORT_SYMBOL_GPL(of_clk_volt_notifier_unregister);
> > +
> > +void of_clk_cpufreq_notifier_unregister(struct notifier_block *nb,
> > +             struct cpufreq_frequency_table *freq_table)
> > +{
> > +     struct volt_scale_data *data = to_volt_scale_data(nb);
> > +     struct device *dev = data->dev;
> > +
> > +     opp_free_cpufreq_table(dev, &freq_table);
> > +     of_clk_volt_notifier_unregister(nb);
> > +}
> > +EXPORT_SYMBOL_GPL(of_clk_cpufreq_notifier_unregister);
> > +
> > +/**
> > + * of_clk_cpufreq_notifer_register - register clock notifier to scale voltage
> > + * @dev: device that owns this scaling operation
> > + * @np: pointer to device's DT node
> > + * @clk: clock whose rate-change notifier triggers voltage scaling
> > + * @supply: regulator id string
> > + * @freq_table: pointer to uninitialized CPUfreq frequency scaling table
> > + * @voltage_latency: worst case voltage scaling latency
> > + */
> > +struct notifier_block *of_clk_cpufreq_notifier_register(struct device *dev,
> > +             struct device_node *np, struct clk *clk, const char *supply,
> > +             struct cpufreq_frequency_table **table,
> > +             unsigned int *voltage_latency)
> > +{
> > +     struct notifier_block *nb;
> > +     struct volt_scale_data *vsd;
> > +     struct opp *opp;
> > +     unsigned long min, max;
> > +     int ret, i;
> > +
> > +     nb = of_clk_volt_notifier_register(dev, np, clk, supply,
> > +                     voltage_latency);
> > +     if (IS_ERR(nb))
> > +             return nb;
> > +
> > +     vsd = to_volt_scale_data(nb);
> > +
> > +     ret = opp_init_cpufreq_table(dev, table);
> > +     if (ret) {
> > +             pr_err("failed to init cpufreq table: %d\n", ret);
> > +             goto err_unregister;
> > +     }
> > +
> > +     /* find the min and max voltages and compute worst-case latency */
> > +     for (i = 0; table[0][i].frequency != CPUFREQ_TABLE_END; i++)
> > +             ;
> > +
> > +     rcu_read_lock();
> > +     opp = opp_find_freq_exact(dev,
> > +                     table[0][0].frequency * 1000, true);
> > +     min = opp_get_voltage(opp);
> > +     opp = opp_find_freq_exact(dev,
> > +                     table[0][i-1].frequency * 1000, true);
> > +     max = opp_get_voltage(opp);
> > +     rcu_read_unlock();
> > +
> > +     *voltage_latency = regulator_set_voltage_time(vsd->reg, min, max);
> > +     if (*voltage_latency < 0)
> > +             pr_warn("%s: failed to calculate voltage latency, %d\n",
> > +                             __func__, ret);
> > +
> > +     return nb;
> > +
> > +err_unregister:
> > +     of_clk_volt_notifier_unregister(nb);
> > +     return ERR_PTR(ret);
> > +}
> > +EXPORT_SYMBOL_GPL(of_clk_cpufreq_notifier_register);
> > diff --git a/include/linux/clk.h b/include/linux/clk.h
> > index 85ea520..3104883 100644
> > --- a/include/linux/clk.h
> > +++ b/include/linux/clk.h
> > @@ -86,6 +86,12 @@ struct notifier_block *of_clk_volt_notifier_register(struct device *dev,
> >               struct device_node *np, struct clk *clk, const char *supply,
> >               int *voltage_latency);
> >   void of_clk_volt_notifier_unregister(struct notifier_block *nb);
> > +struct notifier_block *of_clk_cpufreq_notifier_register(struct device *dev,
> > +             struct device_node *np, struct clk *clk, const char *supply,
> > +             struct cpufreq_frequency_table **table,
> > +             unsigned int *voltage_latency);
> > +void of_clk_cpufreq_notifier_unregister(struct notifier_block *nb,
> > +             struct cpufreq_frequency_table *freq_table);
> >
> >   #endif
> >
> >
> 
> 
> -- 
> Regards,
> Nishanth Menon