[PATCH RESEND v3 0/9] Add thermal sensor driver support for Mediatek MT8196

[PATCH RESEND v3 0/9] Add thermal sensor driver support for Mediatek MT8196

[Laura Nao]

This patch series extends the MediaTek LVTS thermal driver to support the
MT8196 SoC.

MT8196 requires a different implementation of the lvts_temp_to_raw()
function.

To support this, the series introduces:

- A new struct lvts_platform_ops to allow platform-specific
  conversion logic between raw sensor values and temperature
- A variant of the lvts_temp_to_raw() implementation
- Platform data and controller definitions for MT8196

Link to v2: https://lore.kernel.org/all/20250730152128.311109-1-laura.nao@collabora.com/

Changes in v3:
- Make ops in struct lvts_data a pointer to const struct lvts_platform_ops
- Changed mediatek,mt8188-efuse const entry in eFuse binding to an enum
with mediatek,mt8196-efuse and mediatek,mt8188-efuse, dropped mediatek,mt8196-efuse
const entry

Laura Nao (9):
  dt-bindings: thermal: mediatek: Add LVTS thermal controller support
    for MT8196
  thermal/drivers/mediatek/lvts: Make number of calibration offsets
    configurable
  thermal/drivers/mediatek/lvts: Guard against zero temp_factor in
    lvts_raw_to_temp
  thermal: mediatek: lvts: Add platform ops to support alternative
    conversion logic
  thermal/drivers/mediatek/lvts: Add lvts_temp_to_raw variant
  thermal/drivers/mediatek/lvts: Add support for ATP mode
  thermal/drivers/mediatek/lvts: Support MSR offset for 16-bit
    calibration data
  thermal/drivers/mediatek/lvts_thermal: Add MT8196 support
  dt-bindings: nvmem: mediatek: efuse: Add support for MT8196

 .../bindings/nvmem/mediatek,efuse.yaml        |   4 +-
 .../thermal/mediatek,lvts-thermal.yaml        |   2 +
 drivers/thermal/mediatek/lvts_thermal.c       | 305 ++++++++++++++++--
 .../thermal/mediatek,lvts-thermal.h           |  26 ++
 4 files changed, 314 insertions(+), 23 deletions(-)

-- 
2.39.5

[PATCH RESEND v3 1/9] dt-bindings: thermal: mediatek: Add LVTS thermal controller support for MT8196

[Laura Nao]

Add LVTS thermal controller binding for MediaTek MT8196.

Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Signed-off-by: Laura Nao <laura.nao@collabora.com>
---
 .../thermal/mediatek,lvts-thermal.yaml        |  2 ++
 .../thermal/mediatek,lvts-thermal.h           | 26 +++++++++++++++++++
 2 files changed, 28 insertions(+)

diff --git a/Documentation/devicetree/bindings/thermal/mediatek,lvts-thermal.yaml b/Documentation/devicetree/bindings/thermal/mediatek,lvts-thermal.yaml
index 0259cd3ce9c5..beccdabe110b 100644
--- a/Documentation/devicetree/bindings/thermal/mediatek,lvts-thermal.yaml
+++ b/Documentation/devicetree/bindings/thermal/mediatek,lvts-thermal.yaml
@@ -26,6 +26,8 @@ properties:
       - mediatek,mt8192-lvts-mcu
       - mediatek,mt8195-lvts-ap
       - mediatek,mt8195-lvts-mcu
+      - mediatek,mt8196-lvts-ap
+      - mediatek,mt8196-lvts-mcu
 
   reg:
     maxItems: 1
diff --git a/include/dt-bindings/thermal/mediatek,lvts-thermal.h b/include/dt-bindings/thermal/mediatek,lvts-thermal.h
index ddc7302a510a..0ec8ad184d47 100644
--- a/include/dt-bindings/thermal/mediatek,lvts-thermal.h
+++ b/include/dt-bindings/thermal/mediatek,lvts-thermal.h
@@ -80,4 +80,30 @@
 #define MT8192_AP_MD1   15
 #define MT8192_AP_MD2   16
 
+#define MT8196_MCU_MEDIUM_CPU6_0        0
+#define MT8196_MCU_MEDIUM_CPU6_1        1
+#define MT8196_MCU_DSU2                 2
+#define MT8196_MCU_DSU3                 3
+#define MT8196_MCU_LITTLE_CPU3          4
+#define MT8196_MCU_LITTLE_CPU0          5
+#define MT8196_MCU_LITTLE_CPU1          6
+#define MT8196_MCU_LITTLE_CPU2          7
+#define MT8196_MCU_MEDIUM_CPU4_0        8
+#define MT8196_MCU_MEDIUM_CPU4_1        9
+#define MT8196_MCU_MEDIUM_CPU5_0        10
+#define MT8196_MCU_MEDIUM_CPU5_1        11
+#define MT8196_MCU_DSU0                 12
+#define MT8196_MCU_DSU1                 13
+#define MT8196_MCU_BIG_CPU7_0           14
+#define MT8196_MCU_BIG_CPU7_1           15
+
+#define MT8196_AP_TOP0                  0
+#define MT8196_AP_TOP1                  1
+#define MT8196_AP_TOP2                  2
+#define MT8196_AP_TOP3                  3
+#define MT8196_AP_BOT0                  4
+#define MT8196_AP_BOT1                  5
+#define MT8196_AP_BOT2                  6
+#define MT8196_AP_BOT3                  7
+
 #endif /* __MEDIATEK_LVTS_DT_H */
-- 
2.39.5

[PATCH RESEND v3 2/9] thermal/drivers/mediatek/lvts: Make number of calibration offsets configurable

[Laura Nao]

MT8196/MT6991 use 2-byte eFuse calibration data, whereas other SoCs
supported by the driver rely on 3 bytes. Make the number of calibration
bytes per sensor configurable, enabling support for SoCs with varying
calibration formats.

Reviewed-by: Fei Shao <fshao@chromium.org>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Signed-off-by: Laura Nao <laura.nao@collabora.com>
---
 drivers/thermal/mediatek/lvts_thermal.c | 32 +++++++++++++++++--------
 1 file changed, 22 insertions(+), 10 deletions(-)

diff --git a/drivers/thermal/mediatek/lvts_thermal.c b/drivers/thermal/mediatek/lvts_thermal.c
index ab55b20cda47..1c54d0b75b1a 100644
--- a/drivers/thermal/mediatek/lvts_thermal.c
+++ b/drivers/thermal/mediatek/lvts_thermal.c
@@ -96,12 +96,14 @@
 
 #define LVTS_MINIMUM_THRESHOLD		20000
 
+#define LVTS_MAX_CAL_OFFSETS		3
+
 static int golden_temp = LVTS_GOLDEN_TEMP_DEFAULT;
 static int golden_temp_offset;
 
 struct lvts_sensor_data {
 	int dt_id;
-	u8 cal_offsets[3];
+	u8 cal_offsets[LVTS_MAX_CAL_OFFSETS];
 };
 
 struct lvts_ctrl_data {
@@ -127,6 +129,7 @@ struct lvts_data {
 	const struct lvts_ctrl_data *lvts_ctrl;
 	const u32 *conn_cmd;
 	const u32 *init_cmd;
+	int num_cal_offsets;
 	int num_lvts_ctrl;
 	int num_conn_cmd;
 	int num_init_cmd;
@@ -711,7 +714,7 @@ static int lvts_calibration_init(struct device *dev, struct lvts_ctrl *lvts_ctrl
 					u8 *efuse_calibration,
 					size_t calib_len)
 {
-	int i;
+	int i, j;
 	u32 gt;
 
 	/* A zero value for gt means that device has invalid efuse data */
@@ -720,17 +723,18 @@ static int lvts_calibration_init(struct device *dev, struct lvts_ctrl *lvts_ctrl
 	lvts_for_each_valid_sensor(i, lvts_ctrl_data) {
 		const struct lvts_sensor_data *sensor =
 					&lvts_ctrl_data->lvts_sensor[i];
+		u32 calib = 0;
 
-		if (sensor->cal_offsets[0] >= calib_len ||
-		    sensor->cal_offsets[1] >= calib_len ||
-		    sensor->cal_offsets[2] >= calib_len)
-			return -EINVAL;
+		for (j = 0; j < lvts_ctrl->lvts_data->num_cal_offsets; j++) {
+			u8 offset = sensor->cal_offsets[j];
+
+			if (offset >= calib_len)
+				return -EINVAL;
+			calib |= efuse_calibration[offset] << (8 * j);
+		}
 
 		if (gt) {
-			lvts_ctrl->calibration[i] =
-				(efuse_calibration[sensor->cal_offsets[0]] << 0) +
-				(efuse_calibration[sensor->cal_offsets[1]] << 8) +
-				(efuse_calibration[sensor->cal_offsets[2]] << 16);
+			lvts_ctrl->calibration[i] = calib;
 		} else if (lvts_ctrl->lvts_data->def_calibration) {
 			lvts_ctrl->calibration[i] = lvts_ctrl->lvts_data->def_calibration;
 		} else {
@@ -1763,6 +1767,7 @@ static const struct lvts_data mt7988_lvts_ap_data = {
 	.temp_factor	= LVTS_COEFF_A_MT7988,
 	.temp_offset	= LVTS_COEFF_B_MT7988,
 	.gt_calib_bit_offset = 24,
+	.num_cal_offsets = 3,
 };
 
 static const struct lvts_data mt8186_lvts_data = {
@@ -1776,6 +1781,7 @@ static const struct lvts_data mt8186_lvts_data = {
 	.temp_offset	= LVTS_COEFF_B_MT7988,
 	.gt_calib_bit_offset = 24,
 	.def_calibration = 19000,
+	.num_cal_offsets = 3,
 };
 
 static const struct lvts_data mt8188_lvts_mcu_data = {
@@ -1789,6 +1795,7 @@ static const struct lvts_data mt8188_lvts_mcu_data = {
 	.temp_offset	= LVTS_COEFF_B_MT8195,
 	.gt_calib_bit_offset = 20,
 	.def_calibration = 35000,
+	.num_cal_offsets = 3,
 };
 
 static const struct lvts_data mt8188_lvts_ap_data = {
@@ -1802,6 +1809,7 @@ static const struct lvts_data mt8188_lvts_ap_data = {
 	.temp_offset	= LVTS_COEFF_B_MT8195,
 	.gt_calib_bit_offset = 20,
 	.def_calibration = 35000,
+	.num_cal_offsets = 3,
 };
 
 static const struct lvts_data mt8192_lvts_mcu_data = {
@@ -1815,6 +1823,7 @@ static const struct lvts_data mt8192_lvts_mcu_data = {
 	.temp_offset	= LVTS_COEFF_B_MT8195,
 	.gt_calib_bit_offset = 24,
 	.def_calibration = 35000,
+	.num_cal_offsets = 3,
 };
 
 static const struct lvts_data mt8192_lvts_ap_data = {
@@ -1828,6 +1837,7 @@ static const struct lvts_data mt8192_lvts_ap_data = {
 	.temp_offset	= LVTS_COEFF_B_MT8195,
 	.gt_calib_bit_offset = 24,
 	.def_calibration = 35000,
+	.num_cal_offsets = 3,
 };
 
 static const struct lvts_data mt8195_lvts_mcu_data = {
@@ -1841,6 +1851,7 @@ static const struct lvts_data mt8195_lvts_mcu_data = {
 	.temp_offset	= LVTS_COEFF_B_MT8195,
 	.gt_calib_bit_offset = 24,
 	.def_calibration = 35000,
+	.num_cal_offsets = 3,
 };
 
 static const struct lvts_data mt8195_lvts_ap_data = {
@@ -1854,6 +1865,7 @@ static const struct lvts_data mt8195_lvts_ap_data = {
 	.temp_offset	= LVTS_COEFF_B_MT8195,
 	.gt_calib_bit_offset = 24,
 	.def_calibration = 35000,
+	.num_cal_offsets = 3,
 };
 
 static const struct of_device_id lvts_of_match[] = {
-- 
2.39.5

[PATCH RESEND v3 3/9] thermal/drivers/mediatek/lvts: Guard against zero temp_factor in lvts_raw_to_temp

[Laura Nao]

Add a guard against zero temp_factor in lvts_raw_to_temp() to prevent
division by zero and ensure safe conversion.

Fixes: 6725a29321e4 ("thermal/drivers/mediatek/lvts_thermal: Make coeff configurable")
Reviewed-by: Chen-Yu Tsai <wenst@chromium.org>
Reviewed-by: Fei Shao <fshao@chromium.org>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Signed-off-by: Laura Nao <laura.nao@collabora.com>
---
 drivers/thermal/mediatek/lvts_thermal.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)

diff --git a/drivers/thermal/mediatek/lvts_thermal.c b/drivers/thermal/mediatek/lvts_thermal.c
index 1c54d0b75b1a..4ef549386add 100644
--- a/drivers/thermal/mediatek/lvts_thermal.c
+++ b/drivers/thermal/mediatek/lvts_thermal.c
@@ -284,11 +284,14 @@ static int lvts_raw_to_temp(u32 raw_temp, int temp_factor)
 
 static u32 lvts_temp_to_raw(int temperature, int temp_factor)
 {
-	u32 raw_temp = ((s64)(golden_temp_offset - temperature)) << 14;
+	u32 raw_temp;
 
-	raw_temp = div_s64(raw_temp, -temp_factor);
+	if (temp_factor == 0)
+		return temperature;
 
-	return raw_temp;
+	raw_temp = ((s64)(golden_temp_offset - temperature)) << 14;
+
+	return div_s64(raw_temp, -temp_factor);
 }
 
 static int lvts_get_temp(struct thermal_zone_device *tz, int *temp)
@@ -1346,6 +1349,9 @@ static int lvts_probe(struct platform_device *pdev)
 	if (irq < 0)
 		return irq;
 
+	if (!lvts_data->temp_factor)
+		dev_warn(dev, "temp_factor should never be zero; check platform data.\n");
+
 	golden_temp_offset = lvts_data->temp_offset;
 
 	ret = lvts_domain_init(dev, lvts_td, lvts_data);
-- 
2.39.5

Re: [PATCH RESEND v3 3/9] thermal/drivers/mediatek/lvts: Guard against zero temp_factor in lvts_raw_to_temp

[Daniel Lezcano]

On 10/16/25 16:21, Laura Nao wrote:
> Add a guard against zero temp_factor in lvts_raw_to_temp() to prevent
> division by zero and ensure safe conversion.

Is the temp_factor something else than a ro data statically initialized 
by the lvts_data structure ?

It is pointless to handle the case where the temp_factor is zero. If we 
read the temperature the kernel crashes immediately (which means it was 
not tested).

The temp_factor is an internal value of the driver. If the temp_factor 
is zero, the driver is buggy and should be fixed.


> Fixes: 6725a29321e4 ("thermal/drivers/mediatek/lvts_thermal: Make coeff configurable")
> Reviewed-by: Chen-Yu Tsai <wenst@chromium.org>
> Reviewed-by: Fei Shao <fshao@chromium.org>
> Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
> Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
> Signed-off-by: Laura Nao <laura.nao@collabora.com>
> ---
>   drivers/thermal/mediatek/lvts_thermal.c | 12 +++++++++---
>   1 file changed, 9 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/thermal/mediatek/lvts_thermal.c b/drivers/thermal/mediatek/lvts_thermal.c
> index 1c54d0b75b1a..4ef549386add 100644
> --- a/drivers/thermal/mediatek/lvts_thermal.c
> +++ b/drivers/thermal/mediatek/lvts_thermal.c
> @@ -284,11 +284,14 @@ static int lvts_raw_to_temp(u32 raw_temp, int temp_factor)
>   
>   static u32 lvts_temp_to_raw(int temperature, int temp_factor)
>   {
> -	u32 raw_temp = ((s64)(golden_temp_offset - temperature)) << 14;
> +	u32 raw_temp;
>   
> -	raw_temp = div_s64(raw_temp, -temp_factor);
> +	if (temp_factor == 0)
> +		return temperature;
>   
> -	return raw_temp;
> +	raw_temp = ((s64)(golden_temp_offset - temperature)) << 14;
> +
> +	return div_s64(raw_temp, -temp_factor);
>   }
>   
>   static int lvts_get_temp(struct thermal_zone_device *tz, int *temp)
> @@ -1346,6 +1349,9 @@ static int lvts_probe(struct platform_device *pdev)
>   	if (irq < 0)
>   		return irq;
>   
> +	if (!lvts_data->temp_factor)
> +		dev_warn(dev, "temp_factor should never be zero; check platform data.\n");
> +
>   	golden_temp_offset = lvts_data->temp_offset;
>   
>   	ret = lvts_domain_init(dev, lvts_td, lvts_data);
If





-- 
<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

Re: [PATCH RESEND v3 3/9] thermal/drivers/mediatek/lvts: Guard against zero temp_factor in lvts_raw_to_temp

[Laura Nao]

Hi Daniel,

On 11/10/25 13:42, Daniel Lezcano wrote:
> On 10/16/25 16:21, Laura Nao wrote:
>> Add a guard against zero temp_factor in lvts_raw_to_temp() to prevent
>> division by zero and ensure safe conversion. 
>
> Is the temp_factor something else than a ro data statically initialized by the lvts_data structure ?
>
> It is pointless to handle the case where the temp_factor is zero. If we read the temperature the kernel crashes immediately (which means it was not tested).
>
> The temp_factor is an internal value of the driver. If the temp_factor is zero, the driver is buggy and should be fixed.
>
>

That’s right - if temp_factor is zero, that indicates broken platform 
data. I propose failing the probe instead of adding a runtime guard, 
since this condition should never happen in a valid configuration. This 
way, we make the developer aware of the issue early and avoid a kernel 
crash at runtime.

I'll send out a v4 to drop the "if (temp_factor == 0)" checks and 
replace the warning with an error:

if (!lvts_data->temp_factor)
	return dev_err_probe(dev, -EINVAL, "temp_factor should never be zero; check platform data.\n");

Best,

Laura

>> Fixes: 6725a29321e4 ("thermal/drivers/mediatek/lvts_thermal: Make coeff configurable")
>> Reviewed-by: Chen-Yu Tsai <wenst@chromium.org>
>> Reviewed-by: Fei Shao <fshao@chromium.org>
>> Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
>> Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
>> Signed-off-by: Laura Nao <laura.nao@collabora.com>
>> ---
>>   drivers/thermal/mediatek/lvts_thermal.c | 12 +++++++++---
>>   1 file changed, 9 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/thermal/mediatek/lvts_thermal.c b/drivers/thermal/mediatek/lvts_thermal.c
>> index 1c54d0b75b1a..4ef549386add 100644
>> --- a/drivers/thermal/mediatek/lvts_thermal.c
>> +++ b/drivers/thermal/mediatek/lvts_thermal.c
>> @@ -284,11 +284,14 @@ static int lvts_raw_to_temp(u32 raw_temp, int temp_factor)
>>     static u32 lvts_temp_to_raw(int temperature, int temp_factor)
>>   {
>> -    u32 raw_temp = ((s64)(golden_temp_offset - temperature)) << 14;
>> +    u32 raw_temp;
>>   -    raw_temp = div_s64(raw_temp, -temp_factor);
>> +    if (temp_factor == 0)
>> +        return temperature;
>>   -    return raw_temp;
>> +    raw_temp = ((s64)(golden_temp_offset - temperature)) << 14;
>> +
>> +    return div_s64(raw_temp, -temp_factor);
>>   }
>>     static int lvts_get_temp(struct thermal_zone_device *tz, int *temp)
>> @@ -1346,6 +1349,9 @@ static int lvts_probe(struct platform_device *pdev)
>>       if (irq < 0)
>>           return irq;
>>   +    if (!lvts_data->temp_factor)
>> +        dev_warn(dev, "temp_factor should never be zero; check platform data.\n");
>> +
>>       golden_temp_offset = lvts_data->temp_offset;
>>         ret = lvts_domain_init(dev, lvts_td, lvts_data); 
> If
>
>
>
>
>

[PATCH RESEND v3 4/9] thermal: mediatek: lvts: Add platform ops to support alternative conversion logic

[Laura Nao]

Introduce lvts_platform_ops struct to support SoC-specific versions of
lvts_raw_to_temp() and lvts_temp_to_raw() conversion functions.

This is in preparation for supporting SoCs like MT8196/MT6991, which
require a different lvts_temp_to_raw() implementation.

Reviewed-by: Fei Shao <fshao@chromium.org>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Signed-off-by: Laura Nao <laura.nao@collabora.com>
---
 drivers/thermal/mediatek/lvts_thermal.c | 27 ++++++++++++++++++++++---
 1 file changed, 24 insertions(+), 3 deletions(-)

diff --git a/drivers/thermal/mediatek/lvts_thermal.c b/drivers/thermal/mediatek/lvts_thermal.c
index 4ef549386add..df1c0f059ad0 100644
--- a/drivers/thermal/mediatek/lvts_thermal.c
+++ b/drivers/thermal/mediatek/lvts_thermal.c
@@ -125,8 +125,14 @@ struct lvts_ctrl_data {
 			continue; \
 		else
 
+struct lvts_platform_ops {
+	int (*lvts_raw_to_temp)(u32 raw_temp, int temp_factor);
+	u32 (*lvts_temp_to_raw)(int temperature, int temp_factor);
+};
+
 struct lvts_data {
 	const struct lvts_ctrl_data *lvts_ctrl;
+	const struct lvts_platform_ops *ops;
 	const u32 *conn_cmd;
 	const u32 *init_cmd;
 	int num_cal_offsets;
@@ -300,6 +306,7 @@ static int lvts_get_temp(struct thermal_zone_device *tz, int *temp)
 	struct lvts_ctrl *lvts_ctrl = container_of(lvts_sensor, struct lvts_ctrl,
 						   sensors[lvts_sensor->id]);
 	const struct lvts_data *lvts_data = lvts_ctrl->lvts_data;
+	const struct lvts_platform_ops *ops = lvts_data->ops;
 	void __iomem *msr = lvts_sensor->msr;
 	u32 value;
 	int rc;
@@ -332,7 +339,7 @@ static int lvts_get_temp(struct thermal_zone_device *tz, int *temp)
 	if (rc)
 		return -EAGAIN;
 
-	*temp = lvts_raw_to_temp(value & 0xFFFF, lvts_data->temp_factor);
+	*temp = ops->lvts_raw_to_temp(value & 0xFFFF, lvts_data->temp_factor);
 
 	return 0;
 }
@@ -400,10 +407,11 @@ static int lvts_set_trips(struct thermal_zone_device *tz, int low, int high)
 	struct lvts_ctrl *lvts_ctrl = container_of(lvts_sensor, struct lvts_ctrl,
 						   sensors[lvts_sensor->id]);
 	const struct lvts_data *lvts_data = lvts_ctrl->lvts_data;
+	const struct lvts_platform_ops *ops = lvts_data->ops;
 	void __iomem *base = lvts_sensor->base;
-	u32 raw_low = lvts_temp_to_raw(low != -INT_MAX ? low : LVTS_MINIMUM_THRESHOLD,
+	u32 raw_low = ops->lvts_temp_to_raw(low != -INT_MAX ? low : LVTS_MINIMUM_THRESHOLD,
 				       lvts_data->temp_factor);
-	u32 raw_high = lvts_temp_to_raw(high, lvts_data->temp_factor);
+	u32 raw_high = ops->lvts_temp_to_raw(high, lvts_data->temp_factor);
 	bool should_update_thresh;
 
 	lvts_sensor->low_thresh = low;
@@ -1763,6 +1771,11 @@ static const struct lvts_ctrl_data mt8195_lvts_ap_data_ctrl[] = {
 	}
 };
 
+static const struct lvts_platform_ops lvts_platform_ops_v1 = {
+	.lvts_raw_to_temp = lvts_raw_to_temp,
+	.lvts_temp_to_raw = lvts_temp_to_raw,
+};
+
 static const struct lvts_data mt7988_lvts_ap_data = {
 	.lvts_ctrl	= mt7988_lvts_ap_data_ctrl,
 	.conn_cmd	= mt7988_conn_cmds,
@@ -1774,6 +1787,7 @@ static const struct lvts_data mt7988_lvts_ap_data = {
 	.temp_offset	= LVTS_COEFF_B_MT7988,
 	.gt_calib_bit_offset = 24,
 	.num_cal_offsets = 3,
+	.ops = &lvts_platform_ops_v1,
 };
 
 static const struct lvts_data mt8186_lvts_data = {
@@ -1788,6 +1802,7 @@ static const struct lvts_data mt8186_lvts_data = {
 	.gt_calib_bit_offset = 24,
 	.def_calibration = 19000,
 	.num_cal_offsets = 3,
+	.ops = &lvts_platform_ops_v1,
 };
 
 static const struct lvts_data mt8188_lvts_mcu_data = {
@@ -1802,6 +1817,7 @@ static const struct lvts_data mt8188_lvts_mcu_data = {
 	.gt_calib_bit_offset = 20,
 	.def_calibration = 35000,
 	.num_cal_offsets = 3,
+	.ops = &lvts_platform_ops_v1,
 };
 
 static const struct lvts_data mt8188_lvts_ap_data = {
@@ -1816,6 +1832,7 @@ static const struct lvts_data mt8188_lvts_ap_data = {
 	.gt_calib_bit_offset = 20,
 	.def_calibration = 35000,
 	.num_cal_offsets = 3,
+	.ops = &lvts_platform_ops_v1,
 };
 
 static const struct lvts_data mt8192_lvts_mcu_data = {
@@ -1830,6 +1847,7 @@ static const struct lvts_data mt8192_lvts_mcu_data = {
 	.gt_calib_bit_offset = 24,
 	.def_calibration = 35000,
 	.num_cal_offsets = 3,
+	.ops = &lvts_platform_ops_v1,
 };
 
 static const struct lvts_data mt8192_lvts_ap_data = {
@@ -1844,6 +1862,7 @@ static const struct lvts_data mt8192_lvts_ap_data = {
 	.gt_calib_bit_offset = 24,
 	.def_calibration = 35000,
 	.num_cal_offsets = 3,
+	.ops = &lvts_platform_ops_v1,
 };
 
 static const struct lvts_data mt8195_lvts_mcu_data = {
@@ -1858,6 +1877,7 @@ static const struct lvts_data mt8195_lvts_mcu_data = {
 	.gt_calib_bit_offset = 24,
 	.def_calibration = 35000,
 	.num_cal_offsets = 3,
+	.ops = &lvts_platform_ops_v1,
 };
 
 static const struct lvts_data mt8195_lvts_ap_data = {
@@ -1872,6 +1892,7 @@ static const struct lvts_data mt8195_lvts_ap_data = {
 	.gt_calib_bit_offset = 24,
 	.def_calibration = 35000,
 	.num_cal_offsets = 3,
+	.ops = &lvts_platform_ops_v1,
 };
 
 static const struct of_device_id lvts_of_match[] = {
-- 
2.39.5

Re: [PATCH RESEND v3 4/9] thermal: mediatek: lvts: Add platform ops to support alternative conversion logic

[Daniel Lezcano]

On 10/16/25 16:21, Laura Nao wrote:
> Introduce lvts_platform_ops struct to support SoC-specific versions of
> lvts_raw_to_temp() and lvts_temp_to_raw() conversion functions.
> 
> This is in preparation for supporting SoCs like MT8196/MT6991, which
> require a different lvts_temp_to_raw() implementation.
> 
> Reviewed-by: Fei Shao <fshao@chromium.org>
> Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
> Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
> Signed-off-by: Laura Nao <laura.nao@collabora.com>
> ---
>   drivers/thermal/mediatek/lvts_thermal.c | 27 ++++++++++++++++++++++---
>   1 file changed, 24 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/thermal/mediatek/lvts_thermal.c b/drivers/thermal/mediatek/lvts_thermal.c
> index 4ef549386add..df1c0f059ad0 100644
> --- a/drivers/thermal/mediatek/lvts_thermal.c
> +++ b/drivers/thermal/mediatek/lvts_thermal.c
> @@ -125,8 +125,14 @@ struct lvts_ctrl_data {
>   			continue; \
>   		else
>   
> +struct lvts_platform_ops {
> +	int (*lvts_raw_to_temp)(u32 raw_temp, int temp_factor);
> +	u32 (*lvts_temp_to_raw)(int temperature, int temp_factor);
> +};
> +
>   struct lvts_data {
>   	const struct lvts_ctrl_data *lvts_ctrl;
> +	const struct lvts_platform_ops *ops;
>   	const u32 *conn_cmd;
>   	const u32 *init_cmd;
>   	int num_cal_offsets;
> @@ -300,6 +306,7 @@ static int lvts_get_temp(struct thermal_zone_device *tz, int *temp)
>   	struct lvts_ctrl *lvts_ctrl = container_of(lvts_sensor, struct lvts_ctrl,
>   						   sensors[lvts_sensor->id]);
>   	const struct lvts_data *lvts_data = lvts_ctrl->lvts_data;
> +	const struct lvts_platform_ops *ops = lvts_data->ops;
>   	void __iomem *msr = lvts_sensor->msr;
>   	u32 value;
>   	int rc;
> @@ -332,7 +339,7 @@ static int lvts_get_temp(struct thermal_zone_device *tz, int *temp)
>   	if (rc)
>   		return -EAGAIN;
>   
> -	*temp = lvts_raw_to_temp(value & 0xFFFF, lvts_data->temp_factor);
> +	*temp = ops->lvts_raw_to_temp(value & 0xFFFF, lvts_data->temp_factor);

Don't do this in each functions. It does not help for the readability.

May be something like:

int lvts_raw_to_temp(u32 raw_temp, const struct lvts_ctrl_data)
{
	return data->ops->lvts_temp_to_raw(raw_temp, data->temp_factor);
}

or

int lvts_raw_to_temp(u32 raw_temp, const struct lvts_ctrl_data)
{
	int temperature;

	if (data->ops->lvts_temp_to_raw)
		return data->ops->lvts_temp_to_raw(raw_temp, data->temp_factor);

	temperature = ((s64)(raw_temp & 0xFFFF) * temp_factor) >> 14;
         temperature += golden_temp_offset;

         return temperature;
}

... and get rid of all the lvts_platform_ops_v1

(btw _v1 is confusing, it suggests there multiple versions of the same SoC)

[ ... ]

-- 
<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

Re: [PATCH RESEND v3 4/9] thermal: mediatek: lvts: Add platform ops to support alternative conversion logic

[Laura Nao]

Hi Daniel,

On 11/10/25 14:06, Daniel Lezcano wrote:
> On 10/16/25 16:21, Laura Nao wrote:
>> Introduce lvts_platform_ops struct to support SoC-specific versions of
>> lvts_raw_to_temp() and lvts_temp_to_raw() conversion functions.
>>
>> This is in preparation for supporting SoCs like MT8196/MT6991, which
>> require a different lvts_temp_to_raw() implementation.
>>
>> Reviewed-by: Fei Shao <fshao@chromium.org>
>> Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
>> Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
>> Signed-off-by: Laura Nao <laura.nao@collabora.com>
>> ---
>>   drivers/thermal/mediatek/lvts_thermal.c | 27 ++++++++++++++++++++++---
>>   1 file changed, 24 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/thermal/mediatek/lvts_thermal.c b/drivers/thermal/mediatek/lvts_thermal.c
>> index 4ef549386add..df1c0f059ad0 100644
>> --- a/drivers/thermal/mediatek/lvts_thermal.c
>> +++ b/drivers/thermal/mediatek/lvts_thermal.c
>> @@ -125,8 +125,14 @@ struct lvts_ctrl_data {
>>               continue; \
>>           else
>>   +struct lvts_platform_ops {
>> +    int (*lvts_raw_to_temp)(u32 raw_temp, int temp_factor);
>> +    u32 (*lvts_temp_to_raw)(int temperature, int temp_factor);
>> +};
>> +
>>   struct lvts_data {
>>       const struct lvts_ctrl_data *lvts_ctrl;
>> +    const struct lvts_platform_ops *ops;
>>       const u32 *conn_cmd;
>>       const u32 *init_cmd;
>>       int num_cal_offsets;
>> @@ -300,6 +306,7 @@ static int lvts_get_temp(struct thermal_zone_device *tz, int *temp)
>>       struct lvts_ctrl *lvts_ctrl = container_of(lvts_sensor, struct lvts_ctrl,
>>                              sensors[lvts_sensor->id]);
>>       const struct lvts_data *lvts_data = lvts_ctrl->lvts_data;
>> +    const struct lvts_platform_ops *ops = lvts_data->ops;
>>       void __iomem *msr = lvts_sensor->msr;
>>       u32 value;
>>       int rc;
>> @@ -332,7 +339,7 @@ static int lvts_get_temp(struct thermal_zone_device *tz, int *temp)
>>       if (rc)
>>           return -EAGAIN;
>>   -    *temp = lvts_raw_to_temp(value & 0xFFFF, lvts_data->temp_factor);
>> +    *temp = ops->lvts_raw_to_temp(value & 0xFFFF, lvts_data->temp_factor); 
>
> Don't do this in each functions. It does not help for the readability.
>
> May be something like:
>
> int lvts_raw_to_temp(u32 raw_temp, const struct lvts_ctrl_data)
> {
>     return data->ops->lvts_temp_to_raw(raw_temp, data->temp_factor);
> }
>
> or
>
> int lvts_raw_to_temp(u32 raw_temp, const struct lvts_ctrl_data)
> {
>     int temperature;
>
>     if (data->ops->lvts_temp_to_raw)
>         return data->ops->lvts_temp_to_raw(raw_temp, data->temp_factor);
>
>     temperature = ((s64)(raw_temp & 0xFFFF) * temp_factor) >> 14;
>         temperature += golden_temp_offset;
>
>         return temperature;
> }
>
> ... and get rid of all the lvts_platform_ops_v1
>
> (btw _v1 is confusing, it suggests there multiple versions of the same SoC)
>

Right, the first option looks more efficient. Since temp_offset is 
already part of lvts_data, the function would look like:

int lvts_raw_to_temp(u32 raw_temp, const struct lvts_data *lvts_data)
{
	return lvts_data->ops->lvts_raw_to_temp(raw_temp, lvts_data->temp_factor);
}

and the same pattern applies for temp_to_raw().

This change will require renaming the existing 
lvts_raw_to_temp()/lvts_temp_to_raw()/lvts_temp_to_raw_v2() 
implementations. I agree the current _v1 and _v2 suffixes aren’t very 
descriptive. Since lvts_temp_to_raw_v2() version is only used by MT8196,
it could be renamed to lvts_temp_to_raw_mt8196(), with the corresponding 
platform ops defined as lvts_platform_ops_mt8196. The base 
lvts_raw_to_temp()/lvts_temp_to_raw() are shared across the other SoCs,
so they could be renamed after the first supported platform (mt7988) and 
reused for all SoCs that share the same logic.

I'll send out a v4 with the proposed changes.

Thanks,

Laura

> [ ... ]
>

[PATCH RESEND v3 5/9] thermal/drivers/mediatek/lvts: Add lvts_temp_to_raw variant

[Laura Nao]

MT8196/MT6991 require a different version of lvts_temp_to_raw(),
specifically the multiplicative inverse of the existing implementation.
Introduce a variant of the function with inverted calculation logic to
match this requirement.

This ensures accurate raw value generation for temperature
thresholds, avoiding spurious thermal interrupts or unintended hardware
resets on MT8196/MT6991.

Reviewed-by: Chen-Yu Tsai <wenst@chromium.org>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Reviewed-by: Fei Shao <fshao@chromium.org>
Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Signed-off-by: Laura Nao <laura.nao@collabora.com>
---
 drivers/thermal/mediatek/lvts_thermal.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/drivers/thermal/mediatek/lvts_thermal.c b/drivers/thermal/mediatek/lvts_thermal.c
index df1c0f059ad0..31796a5b8858 100644
--- a/drivers/thermal/mediatek/lvts_thermal.c
+++ b/drivers/thermal/mediatek/lvts_thermal.c
@@ -300,6 +300,18 @@ static u32 lvts_temp_to_raw(int temperature, int temp_factor)
 	return div_s64(raw_temp, -temp_factor);
 }
 
+static u32 lvts_temp_to_raw_v2(int temperature, int temp_factor)
+{
+	u32 raw_temp;
+
+	if (temp_factor == 0)
+		return temperature;
+
+	raw_temp = temperature - golden_temp_offset;
+
+	return div_s64((s64)temp_factor << 14, raw_temp);
+}
+
 static int lvts_get_temp(struct thermal_zone_device *tz, int *temp)
 {
 	struct lvts_sensor *lvts_sensor = thermal_zone_device_priv(tz);
-- 
2.39.5

Re: [PATCH RESEND v3 5/9] thermal/drivers/mediatek/lvts: Add lvts_temp_to_raw variant

[Daniel Lezcano]

On 10/16/25 16:21, Laura Nao wrote:
> MT8196/MT6991 require a different version of lvts_temp_to_raw(),
> specifically the multiplicative inverse of the existing implementation.
> Introduce a variant of the function with inverted calculation logic to
> match this requirement.
> 
> This ensures accurate raw value generation for temperature
> thresholds, avoiding spurious thermal interrupts or unintended hardware
> resets on MT8196/MT6991.
> 
> Reviewed-by: Chen-Yu Tsai <wenst@chromium.org>
> Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
> Reviewed-by: Fei Shao <fshao@chromium.org>
> Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
> Signed-off-by: Laura Nao <laura.nao@collabora.com>
> ---
>   drivers/thermal/mediatek/lvts_thermal.c | 12 ++++++++++++
>   1 file changed, 12 insertions(+)
> 
> diff --git a/drivers/thermal/mediatek/lvts_thermal.c b/drivers/thermal/mediatek/lvts_thermal.c
> index df1c0f059ad0..31796a5b8858 100644
> --- a/drivers/thermal/mediatek/lvts_thermal.c
> +++ b/drivers/thermal/mediatek/lvts_thermal.c
> @@ -300,6 +300,18 @@ static u32 lvts_temp_to_raw(int temperature, int temp_factor)
>   	return div_s64(raw_temp, -temp_factor);
>   }
>   
> +static u32 lvts_temp_to_raw_v2(int temperature, int temp_factor)

Same comment as the previous patch ... _v2 ...




-- 
<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

[PATCH RESEND v3 6/9] thermal/drivers/mediatek/lvts: Add support for ATP mode

[Laura Nao]

MT8196/MT6991 uses ATP (Abnormal Temperature Prevention) mode to detect
abnormal temperature conditions, which involves reading temperature data
from a dedicated set of registers separate from the ones used for
immediate and filtered modes.

Add support for ATP mode and its relative registers to ensure accurate
temperature readings and proper thermal management on MT8196/MT6991
devices.

While at it, convert mode defines to enum.

Reviewed-by: Chen-Yu Tsai <wenst@chromium.org>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Reviewed-by: Fei Shao <fshao@chromium.org>
Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Signed-off-by: Laura Nao <laura.nao@collabora.com>
---
 drivers/thermal/mediatek/lvts_thermal.c | 44 +++++++++++++++++++++----
 1 file changed, 37 insertions(+), 7 deletions(-)

diff --git a/drivers/thermal/mediatek/lvts_thermal.c b/drivers/thermal/mediatek/lvts_thermal.c
index 31796a5b8858..574e45400214 100644
--- a/drivers/thermal/mediatek/lvts_thermal.c
+++ b/drivers/thermal/mediatek/lvts_thermal.c
@@ -44,6 +44,10 @@
 #define LVTS_EDATA01(__base)	(__base + 0x0058)
 #define LVTS_EDATA02(__base)	(__base + 0x005C)
 #define LVTS_EDATA03(__base)	(__base + 0x0060)
+#define LVTS_ATP0(__base)		(__base + 0x0070)
+#define LVTS_ATP1(__base)		(__base + 0x0074)
+#define LVTS_ATP2(__base)		(__base + 0x0078)
+#define LVTS_ATP3(__base)		(__base + 0x007C)
 #define LVTS_MSR0(__base)		(__base + 0x0090)
 #define LVTS_MSR1(__base)		(__base + 0x0094)
 #define LVTS_MSR2(__base)		(__base + 0x0098)
@@ -88,9 +92,6 @@
 #define LVTS_COEFF_A_MT7988			-204650
 #define LVTS_COEFF_B_MT7988			204650
 
-#define LVTS_MSR_IMMEDIATE_MODE		0
-#define LVTS_MSR_FILTERED_MODE		1
-
 #define LVTS_MSR_READ_TIMEOUT_US	400
 #define LVTS_MSR_READ_WAIT_US		(LVTS_MSR_READ_TIMEOUT_US / 2)
 
@@ -101,6 +102,12 @@
 static int golden_temp = LVTS_GOLDEN_TEMP_DEFAULT;
 static int golden_temp_offset;
 
+enum lvts_msr_mode {
+	LVTS_MSR_IMMEDIATE_MODE,
+	LVTS_MSR_FILTERED_MODE,
+	LVTS_MSR_ATP_MODE,
+};
+
 struct lvts_sensor_data {
 	int dt_id;
 	u8 cal_offsets[LVTS_MAX_CAL_OFFSETS];
@@ -110,7 +117,7 @@ struct lvts_ctrl_data {
 	struct lvts_sensor_data lvts_sensor[LVTS_SENSOR_MAX];
 	u8 valid_sensor_mask;
 	int offset;
-	int mode;
+	enum lvts_msr_mode mode;
 };
 
 #define VALID_SENSOR_MAP(s0, s1, s2, s3) \
@@ -211,6 +218,10 @@ static const struct debugfs_reg32 lvts_regs[] = {
 	LVTS_DEBUG_FS_REGS(LVTS_EDATA01),
 	LVTS_DEBUG_FS_REGS(LVTS_EDATA02),
 	LVTS_DEBUG_FS_REGS(LVTS_EDATA03),
+	LVTS_DEBUG_FS_REGS(LVTS_ATP0),
+	LVTS_DEBUG_FS_REGS(LVTS_ATP1),
+	LVTS_DEBUG_FS_REGS(LVTS_ATP2),
+	LVTS_DEBUG_FS_REGS(LVTS_ATP3),
 	LVTS_DEBUG_FS_REGS(LVTS_MSR0),
 	LVTS_DEBUG_FS_REGS(LVTS_MSR1),
 	LVTS_DEBUG_FS_REGS(LVTS_MSR2),
@@ -625,6 +636,13 @@ static int lvts_sensor_init(struct device *dev, struct lvts_ctrl *lvts_ctrl,
 		LVTS_IMMD3(lvts_ctrl->base)
 	};
 
+	void __iomem *atp_regs[] = {
+		LVTS_ATP0(lvts_ctrl->base),
+		LVTS_ATP1(lvts_ctrl->base),
+		LVTS_ATP2(lvts_ctrl->base),
+		LVTS_ATP3(lvts_ctrl->base)
+	};
+
 	int i;
 
 	lvts_for_each_valid_sensor(i, lvts_ctrl_data) {
@@ -660,8 +678,20 @@ static int lvts_sensor_init(struct device *dev, struct lvts_ctrl *lvts_ctrl,
 		/*
 		 * Each sensor has its own register address to read from.
 		 */
-		lvts_sensor[i].msr = lvts_ctrl_data->mode == LVTS_MSR_IMMEDIATE_MODE ?
-			imm_regs[i] : msr_regs[i];
+		switch (lvts_ctrl_data->mode) {
+		case LVTS_MSR_IMMEDIATE_MODE:
+			lvts_sensor[i].msr = imm_regs[i];
+			break;
+		case LVTS_MSR_FILTERED_MODE:
+			lvts_sensor[i].msr = msr_regs[i];
+			break;
+		case LVTS_MSR_ATP_MODE:
+			lvts_sensor[i].msr = atp_regs[i];
+			break;
+		default:
+			lvts_sensor[i].msr = imm_regs[i];
+			break;
+		}
 
 		lvts_sensor[i].low_thresh = INT_MIN;
 		lvts_sensor[i].high_thresh = INT_MIN;
@@ -911,7 +941,7 @@ static void lvts_ctrl_monitor_enable(struct device *dev, struct lvts_ctrl *lvts_
 	u32 sensor_map = 0;
 	int i;
 
-	if (lvts_ctrl->mode != LVTS_MSR_FILTERED_MODE)
+	if (lvts_ctrl->mode == LVTS_MSR_IMMEDIATE_MODE)
 		return;
 
 	if (enable) {
-- 
2.39.5

[PATCH RESEND v3 7/9] thermal/drivers/mediatek/lvts: Support MSR offset for 16-bit calibration data

[Laura Nao]

On MT8196/MT6991, per-sensor calibration data read from eFuses is
16-bit. When the LVTS controller operates in 16-bit mode, a fixed offset
must be added to MSR values during post-processing to obtain correct
temperature readings. Introduce a new msr_offset field in lvts_data,
program the respective register and apply the offset to the calibration
data read from eFuses.

Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Signed-off-by: Laura Nao <laura.nao@collabora.com>
---
 drivers/thermal/mediatek/lvts_thermal.c | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)

diff --git a/drivers/thermal/mediatek/lvts_thermal.c b/drivers/thermal/mediatek/lvts_thermal.c
index 574e45400214..8ca5760cbb76 100644
--- a/drivers/thermal/mediatek/lvts_thermal.c
+++ b/drivers/thermal/mediatek/lvts_thermal.c
@@ -44,6 +44,7 @@
 #define LVTS_EDATA01(__base)	(__base + 0x0058)
 #define LVTS_EDATA02(__base)	(__base + 0x005C)
 #define LVTS_EDATA03(__base)	(__base + 0x0060)
+#define LVTS_MSROFT(__base)		(__base + 0x006C)
 #define LVTS_ATP0(__base)		(__base + 0x0070)
 #define LVTS_ATP1(__base)		(__base + 0x0074)
 #define LVTS_ATP2(__base)		(__base + 0x0078)
@@ -150,6 +151,7 @@ struct lvts_data {
 	int temp_offset;
 	int gt_calib_bit_offset;
 	unsigned int def_calibration;
+	u16 msr_offset;
 };
 
 struct lvts_sensor {
@@ -218,6 +220,7 @@ static const struct debugfs_reg32 lvts_regs[] = {
 	LVTS_DEBUG_FS_REGS(LVTS_EDATA01),
 	LVTS_DEBUG_FS_REGS(LVTS_EDATA02),
 	LVTS_DEBUG_FS_REGS(LVTS_EDATA03),
+	LVTS_DEBUG_FS_REGS(LVTS_MSROFT),
 	LVTS_DEBUG_FS_REGS(LVTS_ATP0),
 	LVTS_DEBUG_FS_REGS(LVTS_ATP1),
 	LVTS_DEBUG_FS_REGS(LVTS_ATP2),
@@ -788,6 +791,8 @@ static int lvts_calibration_init(struct device *dev, struct lvts_ctrl *lvts_ctrl
 
 		if (gt) {
 			lvts_ctrl->calibration[i] = calib;
+			if (lvts_ctrl->lvts_data->msr_offset)
+				lvts_ctrl->calibration[i] += lvts_ctrl->lvts_data->msr_offset;
 		} else if (lvts_ctrl->lvts_data->def_calibration) {
 			lvts_ctrl->calibration[i] = lvts_ctrl->lvts_data->def_calibration;
 		} else {
@@ -1095,6 +1100,17 @@ static int lvts_ctrl_calibrate(struct device *dev, struct lvts_ctrl *lvts_ctrl)
 	for (i = 0; i < LVTS_SENSOR_MAX; i++)
 		writel(lvts_ctrl->calibration[i], lvts_edata[i]);
 
+	/* LVTS_MSROFT : Constant offset applied to MSR values
+	 * for post-processing
+	 *
+	 * Bits:
+	 *
+	 * 20-0 : Constant data added to MSR values
+	 */
+	if (lvts_ctrl->lvts_data->msr_offset)
+		writel(lvts_ctrl->lvts_data->msr_offset,
+		       LVTS_MSROFT(lvts_ctrl->base));
+
 	return 0;
 }
 
-- 
2.39.5

[PATCH RESEND v3 8/9] thermal/drivers/mediatek/lvts_thermal: Add MT8196 support

[Laura Nao]

Add LVTS driver support for MT8196.

Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Signed-off-by: Laura Nao <laura.nao@collabora.com>
---
 drivers/thermal/mediatek/lvts_thermal.c | 164 ++++++++++++++++++++++++
 1 file changed, 164 insertions(+)

diff --git a/drivers/thermal/mediatek/lvts_thermal.c b/drivers/thermal/mediatek/lvts_thermal.c
index 8ca5760cbb76..33f68697d1e3 100644
--- a/drivers/thermal/mediatek/lvts_thermal.c
+++ b/drivers/thermal/mediatek/lvts_thermal.c
@@ -92,6 +92,10 @@
 #define LVTS_COEFF_B_MT8195			250460
 #define LVTS_COEFF_A_MT7988			-204650
 #define LVTS_COEFF_B_MT7988			204650
+#define LVTS_COEFF_A_MT8196			391460
+#define LVTS_COEFF_B_MT8196			-391460
+
+#define LVTS_MSR_OFFSET_MT8196		-984
 
 #define LVTS_MSR_READ_TIMEOUT_US	400
 #define LVTS_MSR_READ_WAIT_US		(LVTS_MSR_READ_TIMEOUT_US / 2)
@@ -761,6 +765,39 @@ static int lvts_sensor_init(struct device *dev, struct lvts_ctrl *lvts_ctrl,
  * <-----ap--tc#3-----> <-----sensor#7-----> <-----sensor#8----->
  *  0x40 | 0x41 | 0x42 | 0x43 | 0x44 | 0x45 | 0x46 | 0x47 | 0x48
  *
+ * MT8196 :
+ * Stream index map for MCU Domain mt8196 :
+ *
+ * <-sensor#1--> <-sensor#0--> <-sensor#3--> <-sensor#2-->
+ *  0x04 | 0x05 | 0x06 | 0x07 | 0x08 | 0x09 | 0x0A | 0x0B
+ *
+ * <-sensor#5--> <-sensor#4--> <-sensor#7--> <-sensor#6-->
+ *  0x0C | 0x0D | 0x0E | 0x0F | 0x10 | 0x11 | 0x12 | 0x13
+ *
+ * <-sensor#9--> <-sensor#8--> <-sensor#11-> <-sensor#10->
+ *  0x14 | 0x15 | 0x16 | 0x17 | 0x18 | 0x19 | 0x1A | 0X1B
+ *
+ * <-sensor#13-> <-sensor#12-> <-sensor#15-> <-sensor#14->
+ *  0x1C | 0x1D | 0x1E | 0x1F | 0x20 | 0x21 | 0x22 | 0x23
+ *
+ * Stream index map for APU Domain mt8196 :
+ *
+ * <-sensor#1--> <-sensor#0--> <-sensor#3--> <-sensor#2-->
+ *  0x24 | 0x25 | 0x26 | 0x27 | 0x28 | 0x29 | 0x2A | 0x2B
+ *
+ * Stream index map for GPU Domain mt8196 :
+ *
+ * <-sensor#1--> <-sensor#0-->
+ *  0x2C | 0x2D | 0x2E | 0x2F
+ *
+ * Stream index map for AP Domain mt8196 :
+ *
+ * <-sensor#1--> <-sensor#0--> <-sensor#3--> <-sensor#2-->
+ *  0x30 | 0x31 | 0x32 | 0x33 | 0x34 | 0x35 | 0x36 | 0x37
+ *
+ * <-sensor#5--> <-sensor#4--> <-sensor#6--> <-sensor#7-->
+ *  0x38 | 0x39 | 0x3A | 0x3B | 0x3C | 0x3D | 0x3E | 0x3F
+ *
  * Note: In some cases, values don't strictly follow a little endian ordering.
  * The data description gives byte offsets constituting each calibration value
  * for each sensor.
@@ -1829,11 +1866,112 @@ static const struct lvts_ctrl_data mt8195_lvts_ap_data_ctrl[] = {
 	}
 };
 
+static const struct lvts_ctrl_data mt8196_lvts_mcu_data_ctrl[] = {
+	{
+		.lvts_sensor = {
+			{ .dt_id = MT8196_MCU_MEDIUM_CPU6_0,
+			  .cal_offsets = { 0x06, 0x07 } },
+			{ .dt_id = MT8196_MCU_MEDIUM_CPU6_1,
+			  .cal_offsets = { 0x04, 0x05 } },
+			{ .dt_id = MT8196_MCU_DSU2,
+			  .cal_offsets = { 0x0A, 0x0B } },
+			{ .dt_id = MT8196_MCU_DSU3,
+			  .cal_offsets = { 0x08, 0x09 } }
+		},
+		VALID_SENSOR_MAP(1, 1, 1, 1),
+		.offset = 0x0,
+		.mode = LVTS_MSR_ATP_MODE,
+	},
+	{
+		.lvts_sensor = {
+			{ .dt_id = MT8196_MCU_LITTLE_CPU3,
+			  .cal_offsets = { 0x0E, 0x0F } },
+			{ .dt_id = MT8196_MCU_LITTLE_CPU0,
+			  .cal_offsets = { 0x0C, 0x0D } },
+			{ .dt_id = MT8196_MCU_LITTLE_CPU1,
+			  .cal_offsets = { 0x12, 0x13 } },
+			{ .dt_id = MT8196_MCU_LITTLE_CPU2,
+			  .cal_offsets = { 0x10, 0x11 } }
+		},
+		VALID_SENSOR_MAP(1, 1, 1, 1),
+		.offset = 0x100,
+		.mode = LVTS_MSR_ATP_MODE,
+	},
+	{
+		.lvts_sensor = {
+			{ .dt_id = MT8196_MCU_MEDIUM_CPU4_0,
+			  .cal_offsets = { 0x16, 0x17 } },
+			{ .dt_id = MT8196_MCU_MEDIUM_CPU4_1,
+			  .cal_offsets = { 0x14, 0x15 } },
+			{ .dt_id = MT8196_MCU_MEDIUM_CPU5_0,
+			  .cal_offsets = { 0x1A, 0x1B } },
+			{ .dt_id = MT8196_MCU_MEDIUM_CPU5_1,
+			  .cal_offsets = { 0x18, 0x19 } }
+		},
+		VALID_SENSOR_MAP(1, 1, 1, 1),
+		.offset = 0x200,
+		.mode = LVTS_MSR_ATP_MODE,
+	},
+	{
+		.lvts_sensor = {
+			{ .dt_id = MT8196_MCU_DSU0,
+			  .cal_offsets = { 0x1E, 0x1F } },
+			{ .dt_id = MT8196_MCU_DSU1,
+			  .cal_offsets = { 0x1C, 0x1D } },
+			{ .dt_id = MT8196_MCU_BIG_CPU7_0,
+			  .cal_offsets = { 0x22, 0x23 } },
+			{ .dt_id = MT8196_MCU_BIG_CPU7_1,
+			  .cal_offsets = { 0x20, 0x21 } }
+		},
+		VALID_SENSOR_MAP(1, 1, 1, 1),
+		.offset = 0x300,
+		.mode = LVTS_MSR_ATP_MODE,
+	}
+};
+
+static const struct lvts_ctrl_data mt8196_lvts_ap_data_ctrl[] = {
+	{
+		.lvts_sensor = {
+			{ .dt_id = MT8196_AP_TOP0,
+			  .cal_offsets = { 0x32, 0x33 } },
+			{ .dt_id = MT8196_AP_TOP1,
+			  .cal_offsets = { 0x30, 0x31 } },
+			{ .dt_id = MT8196_AP_TOP2,
+			  .cal_offsets = { 0x36, 0x37 } },
+			{ .dt_id = MT8196_AP_TOP3,
+			  .cal_offsets = { 0x34, 0x35 } }
+		},
+		VALID_SENSOR_MAP(1, 1, 1, 1),
+		.offset = 0x0,
+		.mode = LVTS_MSR_ATP_MODE,
+	},
+	{
+		.lvts_sensor = {
+			{ .dt_id = MT8196_AP_BOT0,
+			  .cal_offsets = { 0x3A, 0x3B } },
+			{ .dt_id = MT8196_AP_BOT1,
+			  .cal_offsets = { 0x38, 0x39 } },
+			{ .dt_id = MT8196_AP_BOT2,
+			  .cal_offsets = { 0x3E, 0x3F } },
+			{ .dt_id = MT8196_AP_BOT3,
+			  .cal_offsets = { 0x3C, 0x3D } }
+		},
+		VALID_SENSOR_MAP(1, 1, 1, 1),
+		.offset = 0x100,
+		.mode = LVTS_MSR_ATP_MODE,
+	}
+};
+
 static const struct lvts_platform_ops lvts_platform_ops_v1 = {
 	.lvts_raw_to_temp = lvts_raw_to_temp,
 	.lvts_temp_to_raw = lvts_temp_to_raw,
 };
 
+static const struct lvts_platform_ops lvts_platform_ops_v2 = {
+	.lvts_raw_to_temp = lvts_raw_to_temp,
+	.lvts_temp_to_raw = lvts_temp_to_raw_v2,
+};
+
 static const struct lvts_data mt7988_lvts_ap_data = {
 	.lvts_ctrl	= mt7988_lvts_ap_data_ctrl,
 	.conn_cmd	= mt7988_conn_cmds,
@@ -1953,6 +2091,30 @@ static const struct lvts_data mt8195_lvts_ap_data = {
 	.ops = &lvts_platform_ops_v1,
 };
 
+static const struct lvts_data mt8196_lvts_mcu_data = {
+	.lvts_ctrl	= mt8196_lvts_mcu_data_ctrl,
+	.num_lvts_ctrl	= ARRAY_SIZE(mt8196_lvts_mcu_data_ctrl),
+	.temp_factor	= LVTS_COEFF_A_MT8196,
+	.temp_offset	= LVTS_COEFF_B_MT8196,
+	.gt_calib_bit_offset = 0,
+	.def_calibration = 14437,
+	.num_cal_offsets = 2,
+	.msr_offset = LVTS_MSR_OFFSET_MT8196,
+	.ops = &lvts_platform_ops_v2,
+};
+
+static const struct lvts_data mt8196_lvts_ap_data = {
+	.lvts_ctrl	= mt8196_lvts_ap_data_ctrl,
+	.num_lvts_ctrl	= ARRAY_SIZE(mt8196_lvts_ap_data_ctrl),
+	.temp_factor	= LVTS_COEFF_A_MT8196,
+	.temp_offset	= LVTS_COEFF_B_MT8196,
+	.gt_calib_bit_offset = 0,
+	.def_calibration = 14437,
+	.num_cal_offsets = 2,
+	.msr_offset = LVTS_MSR_OFFSET_MT8196,
+	.ops = &lvts_platform_ops_v2,
+};
+
 static const struct of_device_id lvts_of_match[] = {
 	{ .compatible = "mediatek,mt7988-lvts-ap", .data = &mt7988_lvts_ap_data },
 	{ .compatible = "mediatek,mt8186-lvts", .data = &mt8186_lvts_data },
@@ -1962,6 +2124,8 @@ static const struct of_device_id lvts_of_match[] = {
 	{ .compatible = "mediatek,mt8192-lvts-ap", .data = &mt8192_lvts_ap_data },
 	{ .compatible = "mediatek,mt8195-lvts-mcu", .data = &mt8195_lvts_mcu_data },
 	{ .compatible = "mediatek,mt8195-lvts-ap", .data = &mt8195_lvts_ap_data },
+	{ .compatible = "mediatek,mt8196-lvts-mcu", .data = &mt8196_lvts_mcu_data },
+	{ .compatible = "mediatek,mt8196-lvts-ap", .data = &mt8196_lvts_ap_data },
 	{},
 };
 MODULE_DEVICE_TABLE(of, lvts_of_match);
-- 
2.39.5

[PATCH RESEND v3 9/9] dt-bindings: nvmem: mediatek: efuse: Add support for MT8196

[Laura Nao]

The MT8196 eFuse layout is compatible with MT8186 and shares the same
decoding scheme for the gpu-speedbin cell.

Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Acked-by: Rob Herring (Arm) <robh@kernel.org>
Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Signed-off-by: Laura Nao <laura.nao@collabora.com>
---
 Documentation/devicetree/bindings/nvmem/mediatek,efuse.yaml | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/Documentation/devicetree/bindings/nvmem/mediatek,efuse.yaml b/Documentation/devicetree/bindings/nvmem/mediatek,efuse.yaml
index 4dc0d42df3e6..c90b026e40bd 100644
--- a/Documentation/devicetree/bindings/nvmem/mediatek,efuse.yaml
+++ b/Documentation/devicetree/bindings/nvmem/mediatek,efuse.yaml
@@ -25,7 +25,9 @@ properties:
   compatible:
     oneOf:
       - items:
-          - const: mediatek,mt8188-efuse
+          - enum:
+              - mediatek,mt8196-efuse
+              - mediatek,mt8188-efuse
           - const: mediatek,mt8186-efuse
       - const: mediatek,mt8186-efuse
 
-- 
2.39.5

Aw: [PATCH RESEND v3 0/9] Add thermal sensor driver support for Mediatek MT8196

[frank-w]

> Gesendet: Donnerstag, 16. Oktober 2025 um 16:21
> Von: "Laura Nao" <laura.nao@collabora.com>
> Betreff: [PATCH RESEND v3 0/9] Add thermal sensor driver support for Mediatek MT8196
>
> This patch series extends the MediaTek LVTS thermal driver to support the
> MT8196 SoC.
> 
> MT8196 requires a different implementation of the lvts_temp_to_raw()
> function.
> 
> To support this, the series introduces:
> 
> - A new struct lvts_platform_ops to allow platform-specific
>   conversion logic between raw sensor values and temperature
> - A variant of the lvts_temp_to_raw() implementation
> - Platform data and controller definitions for MT8196

Hi

tested the series against mt7988 and mt7987 (support not yet upstreamed, so added the necessary fields for it to my 6.17 version) 
and it does not break these boards.

Tested-by: Frank Wunderlich <frank-w@public-files.de>

regards Frank

Re: [PATCH RESEND v3 0/9] Add thermal sensor driver support for Mediatek MT8196

[AngeloGioacchino Del Regno]

Il 16/10/25 16:21, Laura Nao ha scritto:
> This patch series extends the MediaTek LVTS thermal driver to support the
> MT8196 SoC.
> 
> MT8196 requires a different implementation of the lvts_temp_to_raw()
> function.
> 
> To support this, the series introduces:
> 
> - A new struct lvts_platform_ops to allow platform-specific
>    conversion logic between raw sensor values and temperature
> - A variant of the lvts_temp_to_raw() implementation
> - Platform data and controller definitions for MT8196
> 
> Link to v2: https://lore.kernel.org/all/20250730152128.311109-1-laura.nao@collabora.com/
> 
> Changes in v3:
> - Make ops in struct lvts_data a pointer to const struct lvts_platform_ops
> - Changed mediatek,mt8188-efuse const entry in eFuse binding to an enum
> with mediatek,mt8196-efuse and mediatek,mt8188-efuse, dropped mediatek,mt8196-efuse
> const entry
> 

I think this series is ready to be picked.

Please don't let this slip through the cracks.

Regards,
Angelo

> Laura Nao (9):
>    dt-bindings: thermal: mediatek: Add LVTS thermal controller support
>      for MT8196
>    thermal/drivers/mediatek/lvts: Make number of calibration offsets
>      configurable
>    thermal/drivers/mediatek/lvts: Guard against zero temp_factor in
>      lvts_raw_to_temp
>    thermal: mediatek: lvts: Add platform ops to support alternative
>      conversion logic
>    thermal/drivers/mediatek/lvts: Add lvts_temp_to_raw variant
>    thermal/drivers/mediatek/lvts: Add support for ATP mode
>    thermal/drivers/mediatek/lvts: Support MSR offset for 16-bit
>      calibration data
>    thermal/drivers/mediatek/lvts_thermal: Add MT8196 support
>    dt-bindings: nvmem: mediatek: efuse: Add support for MT8196
> 
>   .../bindings/nvmem/mediatek,efuse.yaml        |   4 +-
>   .../thermal/mediatek,lvts-thermal.yaml        |   2 +
>   drivers/thermal/mediatek/lvts_thermal.c       | 305 ++++++++++++++++--
>   .../thermal/mediatek,lvts-thermal.h           |  26 ++
>   4 files changed, 314 insertions(+), 23 deletions(-)
>