Re: [PATCH 1/6] thermal: of: Add support for hardware-tracked trip points

[Mikko Perttunen]

Inline.

On 01/07/14 00:08, Stephen Warren wrote:
> On 06/27/2014 02:11 AM, Mikko Perttunen wrote:
> > This adds support for hardware-tracked trip points to the device tree
> > thermal sensor framework.
> >
> > The framework supports an arbitrary number of trip points. Whenever
> > the current temperature is updated, the trip points immediately
> > below and above the current temperature are found. A sensor driver
> > callback `set_trips' is then called with the temperatures.
> > If there is no trip point above or below the current temperature,
> > the passed trip temperature will be LONG_MAX or LONG_MIN respectively.
> > In this callback, the driver should program the hardware such that
> > it is notified when either of these trip points are triggered.
> > When a trip point is triggered, the driver should call
> > `thermal_zone_device_update' for the respective thermal zone. This
> > will cause the trip points to be updated again.
> >
> > If the `set_trips' callback is not implemented (is NULL), the framework
> > behaves as before.
>
> Is there no "core thermal" code? I would have expected this new feature
> to be implemented in "core" code rather than of/dt "support" code.
> Perhaps there would also be some additions to the of/dt code, but I'd
> still expect the bulk of the feature to be complete independant of
> of/dt. Systems still using board files or ACPI or ... would surely
> benefit from this too?

The thermal core only supports a fixed number of trip points for each 
driver and the core informs the driver of any changes to those, so 
drivers using the core framework can already have hardware trip points, 
but just a fixed number of them.

The way of-thermal works, is it reads all the trip points from the 
device tree, registers a new thermal_zone_device with that number of 
trip points and then handles the trip points completely independently. 
Of course, if we're just polling, this is fine, since the thermal core
also knows about those trip points and will trigger cooling when polling 
the each zone. However, the driver doesn't, so it cannot setup any 
interrupts to call thermal_zone_device_update.

> > diff --git a/drivers/thermal/of-thermal.c b/drivers/thermal/of-thermal.c
>
> > +static int of_thermal_set_trips(struct thermal_zone_device *tz, long temp)
>
> s/tz/tzd/ or s/tz/tzdev/ ? Since "tz" says "thermal zone" to me, but
> it's actually a "thermal zone device".

I followed the existing convention; "tz" is the name used most often by 
both the core and the of-thermal framework.

> > +	struct __thermal_zone *data = tz->devdata;
>
> s/data/tz/ ? "data" is a rather generic term, and "tz" seems like a good
> abbreviation for a __thermal_zone.

Same, though here both "data" and "tz" seem to be used..

> > +	for (i = 0; i < data->ntrips; ++i) {
> > +		struct __thermal_trip *trip = data->trips + i;
> > +		long trip_low = trip->temperature - trip->hysteresis;
> > +
> > +		if (trip_low < temp && trip_low > low)
> > +			low = trip_low;
> > +
> > +		if (trip->temperature > temp && trip->temperature < high)
> > +			high = trip->temperature;
> > +	}
>
> That seems to always apply hysteresis to the low end of a trip object.
> Don't you need to apply the hysteresis to either the low or high end of
> the range, depending on whether the temperature is currently below/above
> the range, and hence which direction the edge will be crossed?

I believe applying only to the low end is correct. Say that we have a 
trip point at 40C and hysteresis of 2C. When we exceed 40C cooling will 
start immediately, but it will only be stopped when we cool down to 38C. 
At that point there is again a 2C gap between the current temperature 
and the trip point. It would seem that this is the interpretation used 
by our downstream kernel and also some people on the Internet (however 
trustworthy they may be..)

If you don't feel this is right, please elaborate.

> Similar comments elsewhere. Perhaps the patch is consistent with some
> existing confusing naming style though?
>
> > +static int of_thermal_update_trips(struct thermal_zone_device *tz)
> > +{
> > +	long temp;
> > +	int err;
> > +
> > +	err = of_thermal_get_temp(tz, &temp);
> > +	if (err)
> > +		return err;
> > +
> > +	err = of_thermal_set_trips(tz, temp);
>
> Doesn't this patch modify of_thermal_get_temp() to call
> of_thermal_set_trips() itself?

You're right. I suppose this function is unneeded.

> > @@ -384,7 +467,8 @@ thermal_zone_of_add_sensor(struct device_node *zone,
> >   struct thermal_zone_device *
> >   thermal_zone_of_sensor_register(struct device *dev, int sensor_id,
> >   				void *data, int (*get_temp)(void *, long *),
> > -				int (*get_trend)(void *, long *))
> > +				int (*get_trend)(void *, long *),
> > +				int (*set_trips)(void *, long, long))
>
> Passing in a struct containing fields for all the ops seem better than
> forever extending this function with more and more function pointers.

Very true.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/