On 7/8/26 08:30, Krzysztof Kozlowski wrote:
> On Tue, Jul 07, 2026 at 09:22:27PM +0200, Daniel Lezcano wrote:
>> +static const struct hwmon_ops t14s_ec_hwmon_ops = {
>> +	.is_visible = t14s_ec_hwmon_is_visible,
>> +	.read = t14s_ec_hwmon_read,
>> +	.read_string = t14s_ec_hwmon_read_string,
>> +};
>> +
>> +static const struct hwmon_channel_info *t14s_ec_hwmon_info[] = {
>> +	HWMON_CHANNEL_INFO(chip, HWMON_C_REGISTER_TZ),
>> +	HWMON_CHANNEL_INFO(temp,
>> +			   HWMON_T_INPUT | HWMON_T_LABEL,
>> +			   HWMON_T_INPUT | HWMON_T_LABEL,
>> +			   HWMON_T_INPUT | HWMON_T_LABEL,
>> +			   HWMON_T_INPUT | HWMON_T_LABEL,
>> +			   HWMON_T_INPUT | HWMON_T_LABEL,
>> +			   HWMON_T_INPUT | HWMON_T_LABEL),
>> +	NULL
>> +};
>> +
>> +static const struct hwmon_chip_info t14s_ec_chip_info = {
>> +	.ops = &t14s_ec_hwmon_ops,
>> +	.info = t14s_ec_hwmon_info,
>> +};
>> +
>> +static int t14s_ec_hwmon_probe(struct t14s_ec *ec)
>> +{
>> +	struct device *dev;
>> +	struct t14s_ec_hwmon_sys_thermx sys_thermx[] = {
>> +		{ .label = "soc",	.reg = T14S_EC_SYS_THERM0 },
>> +		{ .label = "keyboard",	.reg = T14S_EC_SYS_THERM1 },
>> +		{ .label = "base",	.reg = T14S_EC_SYS_THERM2 },
>> +		{ .label = "charging",	.reg = T14S_EC_SYS_THERM3 },
>> +		{ .label = "qtm",	.reg = T14S_EC_SYS_THERM6 },
>> +		{ .label = "ssd",	.reg = T14S_EC_SYS_THERM7 },
>> +	};
>> +
>> +	ec->ec_hwmon.sys_thermx = devm_kmemdup_array(ec->dev, sys_thermx,
>> +						     ARRAY_SIZE(sys_thermx),
>> +						     sizeof(sys_thermx[0]), GFP_KERNEL);
>> +	if (!ec->ec_hwmon.sys_thermx)
>> +		return -ENOMEM;
>> +
>> +	dev = devm_hwmon_device_register_with_info(ec->dev, "t14s_ec", ec,
>> +						   &t14s_ec_chip_info, NULL);
> 
> Last time I commented this looks like actual ABI break. You did not respond to
> that, so I assume my finding was right, thus:
> 1. binding is not correct (although driver should be fixed, not binding)
> 2. the code still breaks users

Sorry I misunderstood it was a question expecting an answer but an 
affirmation.

By ABI break do you mean:

  * new driver with old DT will fail ?
  * new DT with old driver will break ?

* If no thermal-cells is specified, new code will fail with -EINVAL
   -> is that one a problem ?

* If thermal-cells is specified, but no thermal zone, new code does not fail

* If thermal-cells is specified, old code does not fail

* If thermal-cells is not specified, but a thermal zone is specified, 
schema is not respected

Can you clarify ?

Thanks