[1/3] dt-bindings: thermal: imx8mm-thermal: Add support for i.MX8MP

Add thermal binding doc for Freescale's i.MX8MP Thermal Monitoring Unit.

Signed-off-by: Anson Huang <Anson.Huang@nxp.com>
---
 Documentation/devicetree/bindings/thermal/imx8mm-thermal.txt | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

On 08/03/2020 16:27, Anson Huang wrote:
> i.MX8MP shares same TMU with i.MX8MM, the only difference is i.MX8MP
> has two thermal sensors while i.MX8MM ONLY has one, add multiple sensors
> support for i.MX8MM TMU driver.
> 
> Signed-off-by: Anson Huang <Anson.Huang@nxp.com>
> ---
>  drivers/thermal/imx8mm_thermal.c | 108 +++++++++++++++++++++++++++++++++------
>  1 file changed, 93 insertions(+), 15 deletions(-)
> 
> diff --git a/drivers/thermal/imx8mm_thermal.c b/drivers/thermal/imx8mm_thermal.c
> index d597ceb..8a87ed0 100644
> --- a/drivers/thermal/imx8mm_thermal.c
> +++ b/drivers/thermal/imx8mm_thermal.c
> @@ -10,34 +10,75 @@
>  #include <linux/io.h>
>  #include <linux/module.h>
>  #include <linux/of.h>
> -#include <linux/of_address.h>
> +#include <linux/of_device.h>
>  #include <linux/platform_device.h>
>  #include <linux/thermal.h>
>  
>  #include "thermal_core.h"
>  
>  #define TER			0x0	/* TMU enable */
> +#define TPS			0x4
>  #define TRITSR			0x20	/* TMU immediate temp */
>  
>  #define TER_EN			BIT(31)
>  #define TRITSR_VAL_MASK		0xff
>  
> -#define TEMP_LOW_LIMIT		10
> +#define PROBE_SEL_ALL		GENMASK(31, 30)
>  
> -struct imx8mm_tmu {
> +#define PROBE0_STATUS_OFFSET	30
> +#define PROBE0_VAL_OFFSET	16
> +#define SIGN_BIT		BIT(7)
> +#define TEMP_VAL_MASK		GENMASK(6, 0)
> +
> +#define VER1_TEMP_LOW_LIMIT	10
> +#define VER2_TEMP_LOW_LIMIT	-40
> +#define VER2_TEMP_HIGH_LIMIT	125
> +
> +#define TMU_VER1		0x1
> +#define TMU_VER2		0x2
> +
> +struct thermal_soc_data {
> +	u32 num_sensors;
> +	u32 version;
> +};
> +
> +struct tmu_sensor {
> +	struct imx8mm_tmu *priv;
> +	u32 hw_id;
>  	struct thermal_zone_device *tzd;
> +};
> +
> +struct imx8mm_tmu {
>  	void __iomem *base;
>  	struct clk *clk;
> +	const struct thermal_soc_data *socdata;
> +	struct tmu_sensor sensors[0];
>  };
>  
>  static int tmu_get_temp(void *data, int *temp)
>  {
> -	struct imx8mm_tmu *tmu = data;
> +	struct tmu_sensor *sensor = data;
> +	struct imx8mm_tmu *tmu = sensor->priv;
> +	bool ready;
>  	u32 val;
>  
> -	val = readl_relaxed(tmu->base + TRITSR) & TRITSR_VAL_MASK;
> -	if (val < TEMP_LOW_LIMIT)
> -		return -EAGAIN;
> +	if (tmu->socdata->version == TMU_VER1) {

Don't do this here, implement a callback to read the temp, store it in
the socdata and call it directly from here.

So you end up with something simple like:

	*temp = tmu->socdata->get_temp(...);

> +		val = readl_relaxed(tmu->base + TRITSR) & TRITSR_VAL_MASK;
> +		if (val < VER1_TEMP_LOW_LIMIT)
> +			return -EAGAIN;> +	} else {
> +		val = readl_relaxed(tmu->base + TRITSR);
> +		ready = val & (1 << (sensor->hw_id + PROBE0_STATUS_OFFSET));

	test_bit()?

> +		val = (val >> (sensor->hw_id * PROBE0_VAL_OFFSET))
> +		      & TRITSR_VAL_MASK;
> +		if (val & SIGN_BIT) /* negative */
> +			val = (~(val & TEMP_VAL_MASK) + 1);

Please have a look at the different bitops available to simplify this
decoding.

> +		*temp = val;
> +		if (!ready || *temp < VER2_TEMP_LOW_LIMIT ||
> +		    *temp > VER2_TEMP_HIGH_LIMIT)
> +			return -EAGAIN;
> +	}
>  
>  	*temp = val * 1000;
>  
> @@ -50,14 +91,21 @@ static struct thermal_zone_of_device_ops tmu_tz_ops = {
>  
>  static int imx8mm_tmu_probe(struct platform_device *pdev)
>  {
> +	const struct thermal_soc_data *data;
>  	struct imx8mm_tmu *tmu;
>  	u32 val;
>  	int ret;
> +	int i;
> +
> +	data = of_device_get_match_data(&pdev->dev);
>  
> -	tmu = devm_kzalloc(&pdev->dev, sizeof(struct imx8mm_tmu), GFP_KERNEL);
> +	tmu = devm_kzalloc(&pdev->dev, struct_size(tmu, sensors,
> +			   data->num_sensors), GFP_KERNEL);
>  	if (!tmu)
>  		return -ENOMEM;
>  
> +	tmu->socdata = data;
> +
>  	tmu->base = devm_platform_ioremap_resource(pdev, 0);
>  	if (IS_ERR(tmu->base))
>  		return PTR_ERR(tmu->base);
> @@ -77,16 +125,35 @@ static int imx8mm_tmu_probe(struct platform_device *pdev)
>  		return ret;
>  	}
>  
> -	tmu->tzd = devm_thermal_zone_of_sensor_register(&pdev->dev, 0,
> -							tmu, &tmu_tz_ops);
> -	if (IS_ERR(tmu->tzd)) {
> -		dev_err(&pdev->dev,
> -			"failed to register thermal zone sensor: %d\n", ret);
> -		return PTR_ERR(tmu->tzd);
> +	/* disable the monitor during initialization */
> +	val = readl_relaxed(tmu->base + TER);
> +	val &= ~TER_EN;
> +	writel_relaxed(val, tmu->base + TER);

Could you wrap those calls inside a small helper function with a self
described name?

> +
> +	for (i = 0; i < data->num_sensors; i++) {
> +		tmu->sensors[i].priv = tmu;
> +		tmu->sensors[i].tzd =
> +			devm_thermal_zone_of_sensor_register(&pdev->dev, i,
> +							     &tmu->sensors[i],
> +							     &tmu_tz_ops);
> +		if (IS_ERR(tmu->sensors[i].tzd)) {
> +			dev_err(&pdev->dev,
> +				"failed to register thermal zone sensor[%d]: %d\n",
> +				i, ret);
> +			return PTR_ERR(tmu->sensors[i].tzd);
> +		}
> +		tmu->sensors[i].hw_id = i;

May be you can store the offset directly, so it is not computed every
time the temperature is read?

>  	}
>  
>  	platform_set_drvdata(pdev, tmu);
>  
> +	/* enable all the probes for V2 TMU */
> +	if (tmu->socdata->version == TMU_VER2) {
> +		val = readl_relaxed(tmu->base + TPS);
> +		val |= PROBE_SEL_ALL;
> +		writel_relaxed(val, tmu->base + TPS);
> +	}

Same comment as before about putting these in a helper

>  	/* enable the monitor */
>  	val = readl_relaxed(tmu->base + TER);
>  	val |= TER_EN;
> @@ -111,8 +178,19 @@ static int imx8mm_tmu_remove(struct platform_device *pdev)
>  	return 0;
>  }
>  
> +static struct thermal_soc_data imx8mm_tmu_data = {
> +	.num_sensors = 1,
> +	.version = TMU_VER1,
> +};
> +
> +static struct thermal_soc_data imx8mp_tmu_data = {
> +	.num_sensors = 2,
> +	.version = TMU_VER2,
> +};
> +
>  static const struct of_device_id imx8mm_tmu_table[] = {
> -	{ .compatible = "fsl,imx8mm-tmu", },
> +	{ .compatible = "fsl,imx8mm-tmu", .data = &imx8mm_tmu_data, },
> +	{ .compatible = "fsl,imx8mp-tmu", .data = &imx8mp_tmu_data, },
>  	{ },
>  };
>  
>

Hi, Daniel

> Subject: Re: [PATCH 2/3] thermal: imx8mm: Add i.MX8MP support
> 
> On 08/03/2020 16:27, Anson Huang wrote:
> > i.MX8MP shares same TMU with i.MX8MM, the only difference is i.MX8MP
> > has two thermal sensors while i.MX8MM ONLY has one, add multiple
> > sensors support for i.MX8MM TMU driver.
> >
> > Signed-off-by: Anson Huang <Anson.Huang@nxp.com>
> > ---
> >  drivers/thermal/imx8mm_thermal.c | 108
> > +++++++++++++++++++++++++++++++++------
> >  1 file changed, 93 insertions(+), 15 deletions(-)
> >
> > diff --git a/drivers/thermal/imx8mm_thermal.c
> > b/drivers/thermal/imx8mm_thermal.c
> > index d597ceb..8a87ed0 100644
> > --- a/drivers/thermal/imx8mm_thermal.c
> > +++ b/drivers/thermal/imx8mm_thermal.c
> > @@ -10,34 +10,75 @@
> >  #include <linux/io.h>
> >  #include <linux/module.h>
> >  #include <linux/of.h>
> > -#include <linux/of_address.h>
> > +#include <linux/of_device.h>
> >  #include <linux/platform_device.h>
> >  #include <linux/thermal.h>
> >
> >  #include "thermal_core.h"
> >
> >  #define TER			0x0	/* TMU enable */
> > +#define TPS			0x4
> >  #define TRITSR			0x20	/* TMU immediate temp */
> >
> >  #define TER_EN			BIT(31)
> >  #define TRITSR_VAL_MASK		0xff
> >
> > -#define TEMP_LOW_LIMIT		10
> > +#define PROBE_SEL_ALL		GENMASK(31, 30)
> >
> > -struct imx8mm_tmu {
> > +#define PROBE0_STATUS_OFFSET	30
> > +#define PROBE0_VAL_OFFSET	16
> > +#define SIGN_BIT		BIT(7)
> > +#define TEMP_VAL_MASK		GENMASK(6, 0)
> > +
> > +#define VER1_TEMP_LOW_LIMIT	10
> > +#define VER2_TEMP_LOW_LIMIT	-40
> > +#define VER2_TEMP_HIGH_LIMIT	125
> > +
> > +#define TMU_VER1		0x1
> > +#define TMU_VER2		0x2
> > +
> > +struct thermal_soc_data {
> > +	u32 num_sensors;
> > +	u32 version;
> > +};
> > +
> > +struct tmu_sensor {
> > +	struct imx8mm_tmu *priv;
> > +	u32 hw_id;
> >  	struct thermal_zone_device *tzd;
> > +};
> > +
> > +struct imx8mm_tmu {
> >  	void __iomem *base;
> >  	struct clk *clk;
> > +	const struct thermal_soc_data *socdata;
> > +	struct tmu_sensor sensors[0];
> >  };
> >
> >  static int tmu_get_temp(void *data, int *temp)  {
> > -	struct imx8mm_tmu *tmu = data;
> > +	struct tmu_sensor *sensor = data;
> > +	struct imx8mm_tmu *tmu = sensor->priv;
> > +	bool ready;
> >  	u32 val;
> >
> > -	val = readl_relaxed(tmu->base + TRITSR) & TRITSR_VAL_MASK;
> > -	if (val < TEMP_LOW_LIMIT)
> > -		return -EAGAIN;
> > +	if (tmu->socdata->version == TMU_VER1) {
> 
> Don't do this here, implement a callback to read the temp, store it in the
> socdata and call it directly from here.
> 
> So you end up with something simple like:
> 
> 	*temp = tmu->socdata->get_temp(...);
> 

Make sense, do it in V2.


> > +		val = readl_relaxed(tmu->base + TRITSR) &
> TRITSR_VAL_MASK;
> > +		if (val < VER1_TEMP_LOW_LIMIT)
> > +			return -EAGAIN;> +	} else {
> > +		val = readl_relaxed(tmu->base + TRITSR);
> > +		ready = val & (1 << (sensor->hw_id +
> PROBE0_STATUS_OFFSET));
> 
> 	test_bit()?

OK.

> 
> > +		val = (val >> (sensor->hw_id * PROBE0_VAL_OFFSET))
> > +		      & TRITSR_VAL_MASK;
> > +		if (val & SIGN_BIT) /* negative */
> > +			val = (~(val & TEMP_VAL_MASK) + 1);
> 
> Please have a look at the different bitops available to simplify this decoding.

I can ONLY find the FIELD_GET for getting the temperature value field, for the positive
and negative value check, I can't find any API for it.


> 
> > +		*temp = val;
> > +		if (!ready || *temp < VER2_TEMP_LOW_LIMIT ||
> > +		    *temp > VER2_TEMP_HIGH_LIMIT)
> > +			return -EAGAIN;
> > +	}
> >
> >  	*temp = val * 1000;
> >
> > @@ -50,14 +91,21 @@ static struct thermal_zone_of_device_ops
> > tmu_tz_ops = {
> >
> >  static int imx8mm_tmu_probe(struct platform_device *pdev)  {
> > +	const struct thermal_soc_data *data;
> >  	struct imx8mm_tmu *tmu;
> >  	u32 val;
> >  	int ret;
> > +	int i;
> > +
> > +	data = of_device_get_match_data(&pdev->dev);
> >
> > -	tmu = devm_kzalloc(&pdev->dev, sizeof(struct imx8mm_tmu),
> GFP_KERNEL);
> > +	tmu = devm_kzalloc(&pdev->dev, struct_size(tmu, sensors,
> > +			   data->num_sensors), GFP_KERNEL);
> >  	if (!tmu)
> >  		return -ENOMEM;
> >
> > +	tmu->socdata = data;
> > +
> >  	tmu->base = devm_platform_ioremap_resource(pdev, 0);
> >  	if (IS_ERR(tmu->base))
> >  		return PTR_ERR(tmu->base);
> > @@ -77,16 +125,35 @@ static int imx8mm_tmu_probe(struct
> platform_device *pdev)
> >  		return ret;
> >  	}
> >
> > -	tmu->tzd = devm_thermal_zone_of_sensor_register(&pdev->dev, 0,
> > -							tmu, &tmu_tz_ops);
> > -	if (IS_ERR(tmu->tzd)) {
> > -		dev_err(&pdev->dev,
> > -			"failed to register thermal zone sensor: %d\n", ret);
> > -		return PTR_ERR(tmu->tzd);
> > +	/* disable the monitor during initialization */
> > +	val = readl_relaxed(tmu->base + TER);
> > +	val &= ~TER_EN;
> > +	writel_relaxed(val, tmu->base + TER);
> 
> Could you wrap those calls inside a small helper function with a self
> described name?

OK.

> 
> > +
> > +	for (i = 0; i < data->num_sensors; i++) {
> > +		tmu->sensors[i].priv = tmu;
> > +		tmu->sensors[i].tzd =
> > +			devm_thermal_zone_of_sensor_register(&pdev->dev,
> i,
> > +							     &tmu->sensors[i],
> > +							     &tmu_tz_ops);
> > +		if (IS_ERR(tmu->sensors[i].tzd)) {
> > +			dev_err(&pdev->dev,
> > +				"failed to register thermal zone
> sensor[%d]: %d\n",
> > +				i, ret);
> > +			return PTR_ERR(tmu->sensors[i].tzd);
> > +		}
> > +		tmu->sensors[i].hw_id = i;
> 
> May be you can store the offset directly, so it is not computed every time the
> temperature is read?

There are 2 place need to identify the sensor ID, the ready bit and the temperature
value field, so I think the hw_id is necessary, and it also make the logic easy to read.


> 
> >  	}
> >
> >  	platform_set_drvdata(pdev, tmu);
> >
> > +	/* enable all the probes for V2 TMU */
> > +	if (tmu->socdata->version == TMU_VER2) {
> > +		val = readl_relaxed(tmu->base + TPS);
> > +		val |= PROBE_SEL_ALL;
> > +		writel_relaxed(val, tmu->base + TPS);
> > +	}
> 
> Same comment as before about putting these in a helper

OK

Please help review V2 patch.

Thanks,
Anson