Re: [PATCH 1/2] thermal: devfreq_cooling: Use PM QoS to set frequency limits

[Matthias Kaehlcke]

On Tue, Jan 14, 2020 at 04:08:38PM +0000, Leonard Crestez wrote:
> On 10.01.2020 19:49, Matthias Kaehlcke wrote:
> > Now that devfreq supports limiting the frequency range of a device
> > through PM QoS make use of it instead of disabling OPPs that should
> > not be used.
> > 
> > Signed-off-by: Matthias Kaehlcke <mka@xxxxxxxxxxxx>
> > ---
> 
> It is not obvious but this changes behavior when min max requests 
> conflict (min > max): with PM QoS a MIN_FREQUENCY request takes 
> precedence but if higher OPPs are disabled then this will override 
> MIN_FREQUENCY.

Thanks for pointing this out.

> There are very few users of this functionality so I don't think there 
> are any systems that depend on this behaving one way or the other but 
> perhaps it should be mentioned in commit message?

Sounds good, I'll add a note.

> As far as I can tell the only user of devfreq_cooling in upstream is 
> drivers/gpu/drm/panfrost?

Indeed, apparently GPUs are the primary devfreq device used for cooling,
and unfortunately most GPU drivers are not in upstream.

> >   drivers/thermal/devfreq_cooling.c | 66 ++++++++++---------------------
> >   1 file changed, 20 insertions(+), 46 deletions(-)
> > 
> > diff --git a/drivers/thermal/devfreq_cooling.c b/drivers/thermal/devfreq_cooling.c
> > index ef59256887ff..3a63603afcf2 100644
> > --- a/drivers/thermal/devfreq_cooling.c
> > +++ b/drivers/thermal/devfreq_cooling.c
> > @@ -24,11 +24,13 @@
> >   #include <linux/idr.h>
> >   #include <linux/slab.h>
> >   #include <linux/pm_opp.h>
> > +#include <linux/pm_qos.h>
> >   #include <linux/thermal.h>
> >   
> >   #include <trace/events/thermal.h>
> >   
> > -#define SCALE_ERROR_MITIGATION 100
> > +#define HZ_PER_KHZ		1000
> > +#define SCALE_ERROR_MITIGATION	100
> >   
> >   static DEFINE_IDA(devfreq_ida);
> >   
> > @@ -65,49 +67,9 @@ struct devfreq_cooling_device {
> >   	struct devfreq_cooling_power *power_ops;
> >   	u32 res_util;
> >   	int capped_state;
> > +	struct dev_pm_qos_request req_max_freq;
> >   };
> >   
> > -/**
> > - * partition_enable_opps() - disable all opps above a given state
> > - * @dfc:	Pointer to devfreq we are operating on
> > - * @cdev_state:	cooling device state we're setting
> > - *
> > - * Go through the OPPs of the device, enabling all OPPs until
> > - * @cdev_state and disabling those frequencies above it.
> > - */
> > -static int partition_enable_opps(struct devfreq_cooling_device *dfc,
> > -				 unsigned long cdev_state)
> > -{
> > -	int i;
> > -	struct device *dev = dfc->devfreq->dev.parent;
> > -
> > -	for (i = 0; i < dfc->freq_table_size; i++) {
> > -		struct dev_pm_opp *opp;
> > -		int ret = 0;
> > -		unsigned int freq = dfc->freq_table[i];
> > -		bool want_enable = i >= cdev_state ? true : false;
> > -
> > -		opp = dev_pm_opp_find_freq_exact(dev, freq, !want_enable);
> > -
> > -		if (PTR_ERR(opp) == -ERANGE)
> > -			continue;
> > -		else if (IS_ERR(opp))
> > -			return PTR_ERR(opp);
> > -
> > -		dev_pm_opp_put(opp);
> > -
> > -		if (want_enable)
> > -			ret = dev_pm_opp_enable(dev, freq);
> > -		else
> > -			ret = dev_pm_opp_disable(dev, freq);
> > -
> > -		if (ret)
> > -			return ret;
> > -	}
> > -
> > -	return 0;
> > -}
> > -
> >   static int devfreq_cooling_get_max_state(struct thermal_cooling_device *cdev,
> >   					 unsigned long *state)
> >   {
> > @@ -134,7 +96,7 @@ static int devfreq_cooling_set_cur_state(struct thermal_cooling_device *cdev,
> >   	struct devfreq_cooling_device *dfc = cdev->devdata;
> >   	struct devfreq *df = dfc->devfreq;
> >   	struct device *dev = df->dev.parent;
> > -	int ret;
> > +	unsigned long freq;
> >   
> >   	if (state == dfc->cooling_state)
> >   		return 0;
> > @@ -144,9 +106,10 @@ static int devfreq_cooling_set_cur_state(struct thermal_cooling_device *cdev,
> >   	if (state >= dfc->freq_table_size)
> >   		return -EINVAL;
> >   
> > -	ret = partition_enable_opps(dfc, state);
> > -	if (ret)
> > -		return ret;
> > +	freq = dfc->freq_table[state];
> > +
> > +	dev_pm_qos_update_request(&dfc->req_max_freq,
> > +				  DIV_ROUND_UP(freq, HZ_PER_KHZ));
> >   
> >   	dfc->cooling_state = state;
> >   
> > @@ -529,6 +492,12 @@ of_devfreq_cooling_register_power(struct device_node *np, struct devfreq *df,
> >   	if (err)
> >   		goto free_dfc;
> >   
> > +	err = dev_pm_qos_add_request(df->dev.parent, &dfc->req_max_freq,
> > +				     DEV_PM_QOS_MAX_FREQUENCY,
> > +				     PM_QOS_MAX_FREQUENCY_DEFAULT_VALUE);
> > +	if (err < 0)
> > +		goto remove_qos_req;
> > +
> >   	err = ida_simple_get(&devfreq_ida, 0, 0, GFP_KERNEL);
> >   	if (err < 0)
> >   		goto free_tables;
> > @@ -552,6 +521,10 @@ of_devfreq_cooling_register_power(struct device_node *np, struct devfreq *df,
> >   
> >   release_ida:
> >   	ida_simple_remove(&devfreq_ida, dfc->id);
> > +
> > +remove_qos_req:
> > +	dev_pm_qos_remove_request(&dfc->req_max_freq); > +
> 
> A quirk of the dev_pm_qos API is that dev_pm_qos_remove_request prints a 
> WARN splat if !dev_pm_qos_request_active and this can true on 
> dev_pm_qos_add_request error.
> 
> I dealt with this by checking dev_pm_qos_request_active explicitly but 
> perhaps dev_pm_qos API could be changed? In general "free/release" 
> functions shouldn't complain if there's nothing to do.

I think we should be good here if we jump to 'free_tables' if
_add_request() fails, as requested by Chanwoo. Then _remove_request() is
only called when _add_request() was successful.

> >   free_tables:
> >   	kfree(dfc->power_table);
> >   	kfree(dfc->freq_table);
> > @@ -600,6 +573,7 @@ void devfreq_cooling_unregister(struct thermal_cooling_device *cdev)
> >   
> >   	thermal_cooling_device_unregister(dfc->cdev);
> >   	ida_simple_remove(&devfreq_ida, dfc->id);
> > +	dev_pm_qos_remove_request(&dfc->req_max_freq);
> >   	kfree(dfc->power_table);
> >   	kfree(dfc->freq_table);