Message-ID: <c5459b90-d157-4393-a5e5-dca76c2decf1@roeck-us.net>
Date: Thu, 15 Jan 2026 22:11:30 -0800
From: Guenter Roeck <linux@...ck-us.net>
To: Mayank Mahajan <mayankmahajan.x@....com>, corbet@....net,
 robh@...nel.org, krzk+dt@...nel.org, conor+dt@...nel.org,
 linux-hwmon@...r.kernel.org, devicetree@...r.kernel.org,
 linux-doc@...r.kernel.org, linux-kernel@...r.kernel.org
Cc: priyanka.jain@....com, vikash.bansal@....com
Subject: Re: [PATCH v3 2/3] hwmon: (tmp108) Add support for P3T1035 and
 P3T2030

On 1/15/26 03:14, Mayank Mahajan wrote:
> Add support for the P3T1035 & P3T2030 temperature sensor. While mostly
> compatible with the TMP108, P3T1035 uses an 8-bit configuration register
> instead of the 16-bit layout used by TMP108. Updated driver to handle
> this difference during configuration read/write.
> 
> Signed-off-by: Mayank Mahajan <mayankmahajan.x@....com>
> ---
> V1 -> V2:
> - Disabled hysteresis in is_visible function for P3T1035.
> - Added tables for conversion rate similar to the LM75 driver.
> - Implemented different bus access depending on the chip being used.
>    - Removed regmap for 8 bits; now we are using one regmap as before.
>    - Added read and write functions for i2c and i3c for use with regmap.
>    - Mapped the 8-bit configuration register to a 16 bit value for P3T1035.
> V2 -> V3:
> - Remove changes not relevant to adding a new device in the driver.
> - Address warnings due to incorrect usage of casting operations.
> - Remove the usage of P3T2030 as it's functionally identical to P3T1035.
> 
>   drivers/hwmon/Kconfig  |   2 +-
>   drivers/hwmon/tmp108.c | 227 +++++++++++++++++++++++++++++++++--------
>   2 files changed, 186 insertions(+), 43 deletions(-)
> 
> diff --git a/drivers/hwmon/Kconfig b/drivers/hwmon/Kconfig
> index 157678b821fc..31969bddc812 100644
> --- a/drivers/hwmon/Kconfig
> +++ b/drivers/hwmon/Kconfig
> @@ -2398,7 +2398,7 @@ config SENSORS_TMP108
>   	select REGMAP_I3C if I3C
>   	help
>   	  If you say yes here you get support for Texas Instruments TMP108
> -	  sensor chips and NXP P3T1085.
> +	  sensor chips, NXP temperature sensors P3T1035, P3T1085 and P3T2030.
> 
>   	  This driver can also be built as a module. If so, the module
>   	  will be called tmp108.
> diff --git a/drivers/hwmon/tmp108.c b/drivers/hwmon/tmp108.c
> index 60a237cbedbc..38a2203c3bd9 100644
> --- a/drivers/hwmon/tmp108.c
> +++ b/drivers/hwmon/tmp108.c
> @@ -17,9 +17,16 @@
>   #include <linux/regmap.h>
>   #include <linux/regulator/consumer.h>
>   #include <linux/slab.h>
> +#include <linux/util_macros.h>
> 
>   #define	DRIVER_NAME "tmp108"
> 
> +enum tmp108_hw_id {
> +	P3T1035_ID,		/* For sensors p3t1035 and p3t2030 */
> +	P3T1085_ID,
> +	TMP108_ID,
> +};
> +
>   #define	TMP108_REG_TEMP		0x00
>   #define	TMP108_REG_CONF		0x01
>   #define	TMP108_REG_TLOW		0x02
> @@ -61,6 +68,7 @@
>   #define TMP108_CONVRATE_1HZ		TMP108_CONF_CR0		/* Default */
>   #define TMP108_CONVRATE_4HZ		TMP108_CONF_CR1
>   #define TMP108_CONVRATE_16HZ		(TMP108_CONF_CR0|TMP108_CONF_CR1)
> +#define TMP108_CONVRATE_SHIFT		13
> 
>   #define TMP108_CONF_HYSTERESIS_MASK	(TMP108_CONF_HYS0|TMP108_CONF_HYS1)
>   #define TMP108_HYSTERESIS_0C		0x0000
> @@ -71,11 +79,21 @@
>   #define TMP108_CONVERSION_TIME_MS	30	/* in milli-seconds */
> 
>   struct tmp108 {
> -	struct regmap *regmap;
> -	u16 orig_config;
> -	unsigned long ready_time;
> +	struct regmap		*regmap;
> +	u16			orig_config;
> +	unsigned long		ready_time;
> +	enum tmp108_hw_id	hw_id;
> +	bool			config_reg_16bits;
> +	u8			reg_buf[1];
> +	u8			val_buf[3];
> +	unsigned int		sample_times[4];
>   };
> 
> +static const u16 tmp108_sample_set_masks[] = { 3 << TMP108_CONVRATE_SHIFT,
> +					       2 << TMP108_CONVRATE_SHIFT,
> +					       1 << TMP108_CONVRATE_SHIFT,
> +					       0 << TMP108_CONVRATE_SHIFT };
> +

Unnecessary. See below.

>   /* convert 12-bit TMP108 register value to milliCelsius */
>   static inline int tmp108_temp_reg_to_mC(s16 val)
>   {
> @@ -94,6 +112,8 @@ static int tmp108_read(struct device *dev, enum hwmon_sensor_types type,
>   	struct tmp108 *tmp108 = dev_get_drvdata(dev);
>   	unsigned int regval;
>   	int err, hyst;
> +	u16 conv_rate;
> +	u8 index;
> 
>   	if (type == hwmon_chip) {
>   		if (attr == hwmon_chip_update_interval) {
> @@ -101,21 +121,10 @@ static int tmp108_read(struct device *dev, enum hwmon_sensor_types type,
>   					  &regval);
>   			if (err < 0)
>   				return err;
> -			switch (regval & TMP108_CONF_CONVRATE_MASK) {
> -			case TMP108_CONVRATE_0P25HZ:
> -			default:
> -				*temp = 4000;
> -				break;
> -			case TMP108_CONVRATE_1HZ:
> -				*temp = 1000;
> -				break;
> -			case TMP108_CONVRATE_4HZ:
> -				*temp = 250;
> -				break;
> -			case TMP108_CONVRATE_16HZ:
> -				*temp = 63;
> -				break;
> -			}
> +			conv_rate = regval & TMP108_CONF_CONVRATE_MASK;
> +			index = find_closest_descending(conv_rate, tmp108_sample_set_masks,
> +							(int)ARRAY_SIZE(tmp108_sample_set_masks));
> +			*temp = tmp108->sample_times[index];

(regval & TMP108_CONF_CONVRATE_MASK) >> TMP108_CONVRATE_SHIFT, or alternatively
FIELD_GET(TMP108_CONF_CONVRATE_MASK, regval), yields 0..3. With a sample_times
array of { 4000, 1000, 250, 63 } or { 4000, 1000, 250, 125 }, the code above
could simply be
			*temp = tmp108->sample_times[FIELD_GET(TMP108_CONF_CONVRATE_MASK, regval)];
which would both be easier to understand and much simpler.

>   			return 0;
>   		}
>   		return -EOPNOTSUPP;
> @@ -192,22 +201,17 @@ static int tmp108_write(struct device *dev, enum hwmon_sensor_types type,
>   {
>   	struct tmp108 *tmp108 = dev_get_drvdata(dev);
>   	u32 regval, mask;
> +	u8 index;
>   	int err;
> 
>   	if (type == hwmon_chip) {
>   		if (attr == hwmon_chip_update_interval) {
> -			if (temp < 156)
> -				mask = TMP108_CONVRATE_16HZ;
> -			else if (temp < 625)
> -				mask = TMP108_CONVRATE_4HZ;
> -			else if (temp < 2500)
> -				mask = TMP108_CONVRATE_1HZ;
> -			else
> -				mask = TMP108_CONVRATE_0P25HZ;
> +			index = find_closest(temp, tmp108->sample_times,
> +					     (int)ARRAY_SIZE(tmp108->sample_times));

I don't see why the type cast would be needed. Other users of find_closest()
don't need it either.

>   			return regmap_update_bits(tmp108->regmap,
>   						  TMP108_REG_CONF,
>   						  TMP108_CONF_CONVRATE_MASK,
> -						  mask);
> +						  tmp108_sample_set_masks[index]);

Use GENMASK().

>   		}
>   		return -EOPNOTSUPP;
>   	}
> @@ -251,6 +255,8 @@ static int tmp108_write(struct device *dev, enum hwmon_sensor_types type,
>   static umode_t tmp108_is_visible(const void *data, enum hwmon_sensor_types type,
>   				 u32 attr, int channel)
>   {
> +	const struct tmp108 *tmp108 = data;
> +
>   	if (type == hwmon_chip && attr == hwmon_chip_update_interval)
>   		return 0644;
> 
> @@ -264,8 +270,11 @@ static umode_t tmp108_is_visible(const void *data, enum hwmon_sensor_types type,
>   		return 0444;
>   	case hwmon_temp_min:
>   	case hwmon_temp_max:
> +		return 0644;
>   	case hwmon_temp_min_hyst:
>   	case hwmon_temp_max_hyst:
> +		if (tmp108->hw_id == P3T1035_ID)
> +			return 0;
>   		return 0644;
>   	default:
>   		return 0;
> @@ -311,6 +320,105 @@ static bool tmp108_is_volatile_reg(struct device *dev, unsigned int reg)
>   	return reg == TMP108_REG_TEMP || reg == TMP108_REG_CONF;
>   }
> 
> +static int tmp108_i2c_reg_read(void *context, unsigned int reg, unsigned int *val)
> +{
> +	struct i2c_client *client = context;
> +	struct tmp108 *tmp108 = i2c_get_clientdata(client);
> +	int ret;
> +
> +	if (reg == TMP108_REG_CONF && !tmp108->config_reg_16bits)
> +		ret = i2c_smbus_read_byte_data(client, TMP108_REG_CONF);
> +	else
> +		ret = i2c_smbus_read_word_swapped(client, reg);
> +
> +	if (ret < 0)
> +		return ret;
> +
> +	if (reg == TMP108_REG_CONF && !tmp108->config_reg_16bits)
> +		*val = ret << 8;
> +	else
> +		*val = ret;

This evaluates reg and tmp108->config_reg_16bits twice. Try

	if (reg == TMP108_REG_CONF && !tmp108->config_reg_16bits) {
		ret = i2c_smbus_read_byte_data(client, TMP108_REG_CONF);
		if (ret < 0)
			return ret;
		*val = ret << 8;
		return 0;
	}
	ret = i2c_smbus_read_word_swapped(client, reg);
	if (ret < 0)
		return ret;
	*val = ret;
	return 0;

instead to reduce runtime overhead.

> +
> +	return 0;
> +}
> +
> +static int tmp108_i2c_reg_write(void *context, unsigned int reg, unsigned int val)
> +{
> +	struct i2c_client *client = context;
> +	struct tmp108 *tmp108 = i2c_get_clientdata(client);
> +
> +	if (reg == TMP108_REG_CONF && !tmp108->config_reg_16bits)
> +		return i2c_smbus_write_byte_data(client, reg, val >> 8);
> +	return i2c_smbus_write_word_swapped(client, reg, val);
> +}
> +
> +static const struct regmap_bus tmp108_i2c_regmap_bus = {
> +	.reg_read = tmp108_i2c_reg_read,
> +	.reg_write = tmp108_i2c_reg_write,
> +};
> +
> +static int tmp108_i3c_reg_read(void *context, unsigned int reg, unsigned int *val)
> +{
> +	struct i3c_device *i3cdev = context;
> +	struct tmp108 *tmp108 = i3cdev_get_drvdata(i3cdev);
> +	struct i3c_xfer xfers[] = {
> +		{
> +			.rnw = false,
> +			.len = 1,
> +			.data.out = tmp108->reg_buf,
> +		},
> +		{
> +			.rnw = true,
> +			.len = 2,
> +			.data.in = tmp108->val_buf,

What is the point of having reg_buf and val_buf allocated instead
of just using local variables/arrays ?

> +		},
> +	};
> +	int ret;
> +
> +	tmp108->reg_buf[0] = reg;
> +
> +	if (reg == TMP108_REG_CONF && !tmp108->config_reg_16bits)
> +		xfers[1].len--;
> +
> +	ret = i3c_device_do_xfers(i3cdev, xfers, 2, I3C_SDR);
> +	if (ret < 0)
> +		return ret;
> +
> +	*val = tmp108->val_buf[0] << 8;
> +	if (!(reg == TMP108_REG_CONF && !tmp108->config_reg_16bits))

Please refrain from using double negations.
	if (reg != TMP108_REG_CONF || tmp108->config_reg_16bits)
is much easier to understand.

> +		*val |= tmp108->val_buf[1];
> +
> +	return 0;
> +}
> +
> +static int tmp108_i3c_reg_write(void *context, unsigned int reg, unsigned int val)
> +{
> +	struct i3c_device *i3cdev = context;
> +	struct tmp108 *tmp108 = i3cdev_get_drvdata(i3cdev);
> +	struct i3c_xfer xfers[] = {
> +		{
> +			.rnw = false,
> +			.len = 3,
> +			.data.out = tmp108->val_buf,
> +		},
> +	};
> +
> +	tmp108->val_buf[0] = reg;
> +	tmp108->val_buf[1] = (val >> 8) & 0xff;
> +
> +	if (reg == TMP108_REG_CONF && !tmp108->config_reg_16bits)
> +		xfers[0].len--;
> +	else
> +		tmp108->val_buf[2] = val & 0xff;
> +
> +	return i3c_device_do_xfers(i3cdev, xfers, 1, I3C_SDR);
> +}
> +
> +static const struct regmap_bus tmp108_i3c_regmap_bus = {
> +	.reg_read = tmp108_i3c_reg_read,
> +	.reg_write = tmp108_i3c_reg_write,
> +};
> +
>   static const struct regmap_config tmp108_regmap_config = {
>   	.reg_bits = 8,
>   	.val_bits = 16,
> @@ -323,7 +431,8 @@ static const struct regmap_config tmp108_regmap_config = {
>   	.use_single_write = true,
>   };
> 
> -static int tmp108_common_probe(struct device *dev, struct regmap *regmap, char *name)
> +static int tmp108_common_probe(struct device *dev, struct regmap *regmap, char *name,
> +			       enum tmp108_hw_id hw_id)
>   {
>   	struct device *hwmon_dev;
>   	struct tmp108 *tmp108;
> @@ -340,6 +449,14 @@ static int tmp108_common_probe(struct device *dev, struct regmap *regmap, char *
> 
>   	dev_set_drvdata(dev, tmp108);
>   	tmp108->regmap = regmap;
> +	tmp108->hw_id = hw_id;
> +	tmp108->config_reg_16bits = (hw_id == P3T1035_ID) ? false : true;
> +	if (hw_id == P3T1035_ID)
> +		memcpy(tmp108->sample_times, (unsigned int[]){ 125, 250, 1000, 4000 },
> +		       sizeof(tmp108->sample_times));
> +	else
> +		memcpy(tmp108->sample_times, (unsigned int[]){ 63, 250, 1000, 4000 },
> +		       sizeof(tmp108->sample_times));

You'd think that the repeated 0-day complaints have an effect.
Just make tmp108->sample_times a pointer and create two ushort arrays where the values
match the index values.

	struct tmp108 {
		ushort *sample_times;
	};

	ushort p3t_1035_sample_times[] = {4000, 1000, 250, 125};
	ushort tmp108_sample_times[] = {4000, 1000, 250, 125};

	if (hw_id == P3T1035_ID)
		tmp108->sample_times = p3t_1035_sample_times;
	else
		tmp108->sample_times = tmp108_sample_times;

Something like
	tmp108->sample_times = (ushort []) {4000, 1000, 250, 125};
might work as well, but I did not test it.
		
The memcpy is really unnecessary here.

> 
>   	err = regmap_read(tmp108->regmap, TMP108_REG_CONF, &config);
>   	if (err < 0) {
> @@ -351,7 +468,6 @@ static int tmp108_common_probe(struct device *dev, struct regmap *regmap, char *
>   	/* Only continuous mode is supported. */
>   	config &= ~TMP108_CONF_MODE_MASK;
>   	config |= TMP108_MODE_CONTINUOUS;
> -
>   	/* Only comparator mode is supported. */
>   	config &= ~TMP108_CONF_TM;
> 
> @@ -384,17 +500,33 @@ static int tmp108_probe(struct i2c_client *client)
>   {
>   	struct device *dev = &client->dev;
>   	struct regmap *regmap;
> +	enum tmp108_hw_id hw_id;
> +	const void *of_data;
> 
>   	if (!i2c_check_functionality(client->adapter,
> -				     I2C_FUNC_SMBUS_WORD_DATA))
> +				     I2C_FUNC_SMBUS_BYTE_DATA | I2C_FUNC_SMBUS_WORD_DATA))
>   		return dev_err_probe(dev, -ENODEV,
>   				     "adapter doesn't support SMBus word transactions\n");
> 
> -	regmap = devm_regmap_init_i2c(client, &tmp108_regmap_config);
> +	regmap = devm_regmap_init(dev, &tmp108_i2c_regmap_bus, client, &tmp108_regmap_config);
>   	if (IS_ERR(regmap))
>   		return dev_err_probe(dev, PTR_ERR(regmap), "regmap init failed");
> 
> -	return tmp108_common_probe(dev, regmap, client->name);
> +	/* Prefer OF match data (DT-first systems) */
> +	of_data = device_get_match_data(&client->dev);
> +	if (of_data) {
> +		hw_id = (unsigned long)of_data;
> +	} else {
> +		/* Fall back to legacy I2C ID table */
> +		const struct i2c_device_id *id = i2c_client_get_device_id(client);
> +
> +		if (!id) {
> +			return dev_err_probe(dev, -ENODEV, "No matching device ID for i2c device\n");
> +		}
> +		hw_id = (unsigned long)id->driver_data;
> +	}

That complexity is unnecessary. Just use i2c_get_match_data().

> +
> +	return tmp108_common_probe(dev, regmap, client->name, hw_id);
>   }
> 
>   static int tmp108_suspend(struct device *dev)
> @@ -420,16 +552,18 @@ static int tmp108_resume(struct device *dev)
>   static DEFINE_SIMPLE_DEV_PM_OPS(tmp108_dev_pm_ops, tmp108_suspend, tmp108_resume);
> 
>   static const struct i2c_device_id tmp108_i2c_ids[] = {
> -	{ "p3t1085" },
> -	{ "tmp108" },
> -	{ }
> +	{ "p3t1035", P3T1035_ID },
> +	{ "p3t1085", P3T1085_ID },
> +	{ "tmp108", TMP108_ID },
> +	{ /* sentinel */ },
>   };
>   MODULE_DEVICE_TABLE(i2c, tmp108_i2c_ids);
> 
>   static const struct of_device_id tmp108_of_ids[] = {
> -	{ .compatible = "nxp,p3t1085", },
> -	{ .compatible = "ti,tmp108", },
> -	{}
> +	{ .compatible = "nxp,p3t1035", .data = (void *)(uintptr_t)P3T1035_ID },
> +	{ .compatible = "nxp,p3t1085", .data = (void *)(uintptr_t)P3T1085_ID },
> +	{ .compatible = "ti,tmp108", .data = (void *)(uintptr_t)TMP108_ID },
> +	{ /* sentinel */ },
>   };
>   MODULE_DEVICE_TABLE(of, tmp108_of_ids);
> 
> @@ -444,8 +578,9 @@ static struct i2c_driver tmp108_driver = {
>   };
> 
>   static const struct i3c_device_id p3t1085_i3c_ids[] = {
> -	I3C_DEVICE(0x011b, 0x1529, NULL),
> -	{}
> +	I3C_DEVICE(0x011B, 0x1529, (void *)P3T1085_ID),
> +	I3C_DEVICE(0x011B, 0x152B, (void *)P3T1035_ID),
> +	{ /* sentinel */ },

I know that some people like that comment, and I accept it for new drivers.
I do _not_ accept it being changed in existing drivers.

>   };
>   MODULE_DEVICE_TABLE(i3c, p3t1085_i3c_ids);
> 
> @@ -453,13 +588,21 @@ static int p3t1085_i3c_probe(struct i3c_device *i3cdev)
>   {
>   	struct device *dev = i3cdev_to_dev(i3cdev);
>   	struct regmap *regmap;
> +	const struct i3c_device_id *id;
> +	enum tmp108_hw_id hw_id;
> 
> -	regmap = devm_regmap_init_i3c(i3cdev, &tmp108_regmap_config);
> +	regmap = devm_regmap_init(dev, &tmp108_i3c_regmap_bus, i3cdev, &tmp108_regmap_config);
>   	if (IS_ERR(regmap))
>   		return dev_err_probe(dev, PTR_ERR(regmap),
>   				     "Failed to register i3c regmap\n");
> 
> -	return tmp108_common_probe(dev, regmap, "p3t1085_i3c");
> +	id = i3c_device_match_id(i3cdev, p3t1085_i3c_ids);
> +	if (!id) {
> +		return dev_err_probe(dev, -ENODEV, "No matching device ID for i3c device\n");
> +	}

Unnecessary error check since the id already matches or the function would not
have been called.

> +	hw_id = (enum tmp108_hw_id)(uintptr_t)id->data;
> +
> +	return tmp108_common_probe(dev, regmap, "p3t1085_i3c", hw_id);
>   }
> 
>   static struct i3c_driver p3t1085_driver = {
> --
> 2.34.1
> 

Powered by blists - more mailing lists