Re: [RFC v1 3/3] thermal/drivers/mediatek/lvts_thermal: add mt7988 support

[Frank Wunderlich <frank-w@public-files.de>]

Am 13. September 2023 10:16:51 MESZ schrieb AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>:
Hi angelo,

thanks for first look

>Il 11/09/23 20:33, Frank Wunderlich ha scritto:
>> From: Frank Wunderlich <frank-w@public-files.de>
>> 
>> Add Support for mediatek fologic 880/MT7988.
>> 
>> Signed-off-by: Frank Wunderlich <frank-w@public-files.de>
>> ---
>>   drivers/thermal/mediatek/lvts_thermal.c | 73 +++++++++++++++++++++++++
>>   1 file changed, 73 insertions(+)
>> 
>> diff --git a/drivers/thermal/mediatek/lvts_thermal.c b/drivers/thermal/mediatek/lvts_thermal.c
>> index c1004b4da3b6..48b257a3c80e 100644
>> --- a/drivers/thermal/mediatek/lvts_thermal.c
>> +++ b/drivers/thermal/mediatek/lvts_thermal.c
>> @@ -82,6 +82,8 @@
>>   #define LVTS_GOLDEN_TEMP_DEFAULT	50
>>   #define LVTS_COEFF_A_MT8195			-250460
>>   #define LVTS_COEFF_B_MT8195			250460
>> +#define LVTS_COEFF_A_MT7988			-204650
>> +#define LVTS_COEFF_B_MT7988			204650
>>     #define LVTS_MSR_IMMEDIATE_MODE		0
>>   #define LVTS_MSR_FILTERED_MODE		1
>> @@ -1272,6 +1274,67 @@ static int lvts_remove(struct platform_device *pdev)
>>   	return 0;
>>   }
>>   +/*
>> + * LVTS MT7988
>> + */
>> +#define LVTS_HW_SHUTDOWN_MT7988	117000
>
>Are you sure that this chip's Tj is >117°C ?!
>
>Looks a bit high... if it is exactly 117°C, I would suggest cutting earlier,
>either at 110 (safe side) or 115: after all, this is a life-saver feature and
>the chip is actually never meant to *constantly* work at 110°C (as it would
>degrade fast and say goodbye earlier than "planned").

I took values from SDK

https://git01.mediatek.com/plugins/gitiles/openwrt/feeds/mtk-openwrt-feeds/+/refs/heads/master/target/linux/mediatek/files-5.4/drivers/thermal/mediatek/soc_temp_lvts.c#1483

>> +//enum mt7988_lvts_domain { MT7988_AP_DOMAIN, MT7988_NUM_DOMAIN };
>> +
>> +enum mt7988_lvts_sensor_enum {
>> +	MT7988_TS3_0,
>> +	MT7988_TS3_1,
>> +	MT7988_TS3_2,
>> +	MT7988_TS3_3,
>> +	MT7988_TS4_0,
>> +	MT7988_TS4_1,
>> +	MT7988_TS4_2,
>> +	MT7988_TS4_3,
>> +	MT7988_NUM_TS
>> +};

>This enumeration should be definitions in bindings (mediatek,lvts-thermal.h).
>
>Besides, the LVTS is about internal temperatures, so those TS3_x and 4_x can
>be renamed like what was done for MT8192 and MT8195: this is because you will
>never see TS3_2 being CPU2 on a board and CPU4 on another, being those - again -
>internal to the SoC, hence unchangeable.

Right these sensors are internally only and i took naming from sdk to avoid confusion. And i have not more information about these internal sensors (special meaning),but their values are packed together to get the resulting (average) temperature.

>Another reason is that you'll anyway have to refer to those sensors in the
>devicetree to configure thermal trips and such, so... :-)

In device tree it will look like this:

https://github.com/frank-w/BPI-Router-Linux/blob/6.5-lvts/arch/arm64/boot/dts/mediatek/mt7988a.dtsi#L771

Daniel has also defined thermal trips there,but these are untested atm. I only verified temperature itself i get from sysfs as far as i can (start at ~40°C and reaching ~70 while running).

>Regards,
>Angelo

regards Frank