Re: [RFC PATCH 2/2] thermal: add a virtual sensor to aggregate temperatures

[Alexandre Bailon <abailon@baylibre.com>]

Hi Daniel,

On 20/08/2021 14:52, Daniel wrote:
> On 19/08/2021 14:32, Alexandre Bailon wrote:
>> This adds a virtual thermal sensor that reads temperature from
>> hardware sensor and return an aggregated temperature.
>> Currently, this only return the max temperature.
>>
>> Signed-off-by: Alexandre Bailon <abailon@baylibre.com>
>> ---
>>   drivers/thermal/Kconfig              |   8 ++
>>   drivers/thermal/Makefile             |   1 +
>>   drivers/thermal/thermal_aggregator.c | 134 +++++++++++++++++++++++++++
>>   3 files changed, 143 insertions(+)
>>   create mode 100644 drivers/thermal/thermal_aggregator.c
>>
>> diff --git a/drivers/thermal/Kconfig b/drivers/thermal/Kconfig
>> index 7a4ba50ba97d0..f9c152cfb95bc 100644
>> --- a/drivers/thermal/Kconfig
>> +++ b/drivers/thermal/Kconfig
>> @@ -228,6 +228,14 @@ config THERMAL_MMIO
>>   	  register or shared memory, is a potential candidate to work with this
>>   	  driver.
>>   
>> +config THERMAL_AGGREGATOR
> We discussed the virtual sensor doing aggregation but may be we can give
> it another purpose in the future like returning a constant temp or
> low/high pass filter.
>
> It may make sense to use a more generic name like virtual sensor for
> example.
Indeed, this would make more sense.
>
>> +	tristate "Generic thermal aggregator driver"
>> +	depends on TERMAL_OF || COMPILE_TEST
> s/TERMAL_OF/THERMAL_OF/
>
>> +	help
>> +	  This option enables the generic thermal sensor aggregator.
>> +	  This driver creates a thermal sensor that reads the hardware sensors
>> +	  and aggregate the temperature.
>> +
>>   config HISI_THERMAL
>>   	tristate "Hisilicon thermal driver"
>>   	depends on ARCH_HISI || COMPILE_TEST
>> diff --git a/drivers/thermal/Makefile b/drivers/thermal/Makefile
>> index 9729a2b089919..5b20ef15aebbe 100644
>> --- a/drivers/thermal/Makefile
>> +++ b/drivers/thermal/Makefile
>> @@ -60,3 +60,4 @@ obj-$(CONFIG_UNIPHIER_THERMAL)	+= uniphier_thermal.o
>>   obj-$(CONFIG_AMLOGIC_THERMAL)     += amlogic_thermal.o
>>   obj-$(CONFIG_SPRD_THERMAL)	+= sprd_thermal.o
>>   obj-$(CONFIG_KHADAS_MCU_FAN_THERMAL)	+= khadas_mcu_fan.o
>> +obj-$(CONFIG_THERMAL_AGGREGATOR) += thermal_aggregator.o
>> diff --git a/drivers/thermal/thermal_aggregator.c b/drivers/thermal/thermal_aggregator.c
>> new file mode 100644
>> index 0000000000000..76f871dbfee9e
>> --- /dev/null
>> +++ b/drivers/thermal/thermal_aggregator.c
>> @@ -0,0 +1,134 @@
>> +#include <linux/err.h>
>> +#include <linux/export.h>
>> +#include <linux/module.h>
>> +#include <linux/of_device.h>
>> +#include <linux/of_platform.h>
>> +#include <linux/slab.h>
>> +#include <linux/thermal.h>
>> +#include <linux/types.h>
>> +#include <linux/string.h>
>> +
>> +const char *aggr_types[] = {
>> +	"min",
>> +	"max",
>> +	"avg",
>> +};
>> +
>> +struct thermal_aggr;
>> +
>> +typedef int (*aggregate_fn)(struct thermal_aggr *aggr);
>> +
>> +struct thermal_aggr_sensor {
>> +	struct thermal_sensor *sensor;
>> +
>> +	struct list_head node;
>> +};
>> +
>> +struct thermal_aggr {
>> +	struct list_head sensors;
>> +	aggregate_fn *aggregate;
>> +	//struct thermal_zone_device *tz;
>> +};
>> +
>> +static int thermal_aggr_read_temp(void *data, int *temperature)
>> +{
>> +	struct thermal_aggr *aggr = data;
>> +	struct thermal_aggr_sensor *sensor;
>> +	int max_temp = 0;
>> +	int temp;
>> +
> What happens if a hardware sensor module is unloaded ?
Hum, I don't know how to deal with it.
Maybe adding refcounting to sensors to prevent module unloading ?
>
>> +	list_for_each_entry(sensor, &aggr->sensors, node) {
>> +		if (!sensor->sensor) {
>> +			return -ENODEV;
>> +		}
>
>
>
>> +		sensor->sensor->ops->get_temp(sensor->sensor->sensor_data, &temp);
>> +		max_temp = max(max_temp, temp);
>> +	}
>> +
>> +	*temperature = max_temp;
>> +
>> +	return 0;
>> +}
>> +
>> +static const struct thermal_zone_of_device_ops thermal_aggr_max_ops = {
>> +	.get_temp = thermal_aggr_read_temp,
> 	.get_temp = thermal_virtual_sensor_get_temp ?

Actually, I though I could create a thermol_zone_of_device_ops for each 
type of operation
(min, max, etc) we would support and would just to register the right 
ops at probe time.

>> +};
>> +
>> +static int thermal_aggr_probe(struct platform_device *pdev)
>> +{
>> +	struct thermal_aggr *aggr;
>> +	struct device *dev = &pdev->dev;
>> +	struct of_phandle_args args;
>> +	int count;
>> +	int ret;
>> +	int i;
>> +
>> +	count = of_count_phandle_with_args(dev->of_node,
>> +					   "thermal-sensors",
>> +					   "#thermal-sensor-cells");
>> +	if (count <= 0)
>> +		return -EINVAL;
>> +
>> +	aggr = kzalloc(sizeof(*aggr), GFP_KERNEL);
> 	devm_kzalloc
>
>> +	INIT_LIST_HEAD(&aggr->sensors);
>> +
>> +	for (i = 0; i < count; i++) {
>> +		struct thermal_sensor *sensor;
>> +		struct thermal_aggr_sensor *aggr_sensor;
>> +		int id;
>> +
>> +		ret = of_parse_phandle_with_args(dev->of_node,
>> +						 "thermal-sensors",
>> +						 "#thermal-sensor-cells",
>> +						 i,
>> +						 &args);
>> +		if (ret) {
>> +			return ret;
>> +		}
>> +
>> +		id = args.args_count ? args.args[0] : 0;
>> +		sensor = thermal_of_get_sensor(args.np, id);
>> +		if (sensor == NULL) {
>> +			return -EPROBE_DEFER;
>> +		}
>> +
>> +		aggr_sensor = kzalloc(sizeof(*aggr_sensor), GFP_KERNEL);
> 		devm_kzalloc
>
>> +		aggr_sensor->np = args.np;
> Why the 'np' and id are stored, they won't be needed anymore, no ?
Right. At some point, I had issues with the sensors that was not available.
I though it was an probe orderings issue and I tried to get the sensor 
later from the callback.
It was an issue with the sensor module itself, and not with the ordering 
but I forgot to remove
np and id that not useful anymore.
>
>> +		aggr_sensor->id = id;
>> +		aggr_sensor->sensor = sensor;
>> +		list_add(&aggr_sensor->node, &aggr->sensors);
>> +	}
>> +
>> +	/*tzdev = */devm_thermal_zone_of_sensor_register(dev, 0, aggr, &thermal_aggr_max_ops);
>> +
>> +	return 0;
>> +}
>> +
>> +static int thermal_aggr_remove(struct platform_device *pdev)
>> +{
>> +	return 0;
>> +}
>> +
>> +static const struct of_device_id thermal_aggr_of_match[] = {
>> +	{
>> +		.compatible = "generic,thermal-aggregator",  "^virtual,.*":
> As stated in the documentation
> Documentation/devicetree/bindings/vendor-prefixes.yaml
>
>    "^virtual,.*":
>      description: Used for virtual device without specific vendor.
>
> I suggest something like:
>
> 	.compatible = "virtual,thermal-sensor",
OK, makes sense to me.

Thanks you for the review.
Alexandre

>
>
>> +	},
>> +	{
>> +	},
>> +};
>> +MODULE_DEVICE_TABLE(of, thermal_aggr_of_match);
>> +
>> +static struct platform_driver thermal_aggr = {
>> +	.probe = thermal_aggr_probe,
>> +	.remove = thermal_aggr_remove,
>> +	.driver = {
>> +		.name = "thermal-aggregator",
>> +		.of_match_table = thermal_aggr_of_match,
>> +	},
>> +};
>> +
>> +module_platform_driver(thermal_aggr);
>> +MODULE_AUTHOR("Alexandre Bailon <abailon@baylibre.com>");
>> +MODULE_DESCRIPTION("Thermal sensor aggregator");
>> +MODULE_LICENSE("GPL v2");
>>
>