Re: [PATCH v6 3/5] thermal: renesas: rzg3e: Add thermal driver for the Renesas RZ/G3E SoC

[Daniel Lezcano <daniel.lezcano@linaro.org>]

On 31/07/2025 19:19, John Madieu wrote:
> Hi Daniel,
> 
> Thanks for your review.
> 
>> -----Original Message-----
>> From: Daniel Lezcano <daniel.lezcano@linaro.org>
>> Sent: Wednesday, July 16, 2025 11:11 PM
>> To: John Madieu <john.madieu.xa@bp.renesas.com>
>> Subject: Re: [PATCH v6 3/5] thermal: renesas: rzg3e: Add thermal driver
>> for the Renesas RZ/G3E SoC
>>
>> On Thu, May 22, 2025 at 08:22:46PM +0200, John Madieu wrote:
>>> The RZ/G3E SoC integrates a Temperature Sensor Unit (TSU) block
>>> designed to monitor the chip's junction temperature. This sensor is
>>> connected to channel 1 of the APB port clock/reset and provides
>> temperature measurements.
>>>
>>> It also requires calibration values stored in the system controller
>>> registers for accurate temperature measurement. Add a driver for the
>> Renesas RZ/G3E TSU.
>>>
>>> Signed-off-by: John Madieu <john.madieu.xa@bp.renesas.com>
>>> ---

[ ... ]

>>> +static int rzg3e_thermal_get_temp(struct thermal_zone_device *zone,
>>> +int *temp) {
>>> +	struct rzg3e_thermal_priv *priv = thermal_zone_device_priv(zone);
>>> +	u32 val;
>>> +	int ret;
>>> +
>>> +	if (priv->mode == THERMAL_DEVICE_DISABLED)
>>> +		return -EBUSY;

[ ... ]

>>> +	reinit_completion(&priv->conv_complete);
>>> +
>>> +	/* Enable ADC interrupt */
>>> +	writel(TSU_SIER_ADIE, priv->base + TSU_SIER);
>>
>> Why enable irq here ?
>>
> 
> I did it this way because, in 'set_trips' callback, the
> driver does trigger conversion to check whether the current
> temperature is part of the window or not, and triggers the
> comparison interrupt accordingly. Because of that, I did not
> want the conversion-complete interrupt to also be triggered.
> 
> That's the reason why I enable conversion-complete interrupt
> in 'get_temp', to make sure its interrupt is being triggered
> only when the thermal core calls it.
> 
> Should I do it another way ?

I don't ATM, the approach is very unusual so I'm still trying to figure 
out what is this completion approach and readl_poll_timeout_atomic. At 
the first glance I would say it is wrong.


>>> +	/* Verify no ongoing conversion */
>>> +	ret = readl_poll_timeout_atomic(priv->base + TSU_SSR, val,
>>> +					!(val & TSU_SSR_CONV_RUNNING),
>>> +					TSU_POLL_DELAY_US, TSU_TIMEOUT_US);
>>> +	if (ret) {
>>> +		dev_err(priv->dev, "ADC conversion timed out\n");
>>> +		return ret;
>>> +	}
>>> +
>>> +	/* Start conversion */
>>> +	writel(TSU_STRGR_ADST, priv->base + TSU_STRGR);
>>> +
>>> +	if (!wait_for_completion_timeout(&priv->conv_complete,
>>> +					 msecs_to_jiffies(100))) {
>>> +		dev_err(priv->dev, "ADC conversion completion timeout\n");
>>> +		return -ETIMEDOUT;
>>> +	}
>>
>> Can you explain what is happening here ?
>>
> 
> I might not get what you are asking, but since I compute the
> temperature in the hard IRQ handler, I just wait for it to complete
> and notify the completion so I can grab the processed value to notify
> the thermal core.
> 
> Please let me know if this does not answer your question.

Can you describe how the sensor works ? And perhaps if you have a 
pointer to some documentation ?
  [ ... ]

-- 
<http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs

Follow Linaro:  <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog