Re: [PATCH v3 4/5] hwmon: tmp108: Add support for I3C device

[Guenter Roeck <linux@roeck-us.net>]

On 11/11/24 15:20, Frank Li wrote:
> On Mon, Nov 11, 2024 at 10:37:04AM -0800, Guenter Roeck wrote:
>> On 11/11/24 10:10, Guenter Roeck wrote:
>>> On 11/11/24 10:04, Guenter Roeck wrote:
>>> [ ... ]
>>>>> +static int p3t1085_i3c_probe(struct i3c_device *i3cdev)
>>>>> +{
>>>>> +    struct device *dev = i3cdev_to_dev(i3cdev);
>>>>> +    struct regmap *regmap;
>>>>> +
>>>>> +    regmap = devm_regmap_init_i3c(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");
>>>>> +}
>>>>> +
>>>>> +static struct i3c_driver p3t1085_driver = {
>>>>> +    .driver = {
>>>>> +        .name = "p3t1085_i3c",
>>>>> +    },
>>>>> +    .probe = p3t1085_i3c_probe,
>>>>> +    .id_table = p3t1085_i3c_ids,
>>>>> +};
>>>>> +module_i3c_driver(p3t1085_driver);
>>>>> +#endif
>>>>
>>>> While looking at i3c code, I found module_i3c_i2c_driver(). Can we use
>>>> that function to register both i2c and i3c in one call ?
>>>>
>>> Answering my own question: No, because devm_regmap_init_i3c()
>>> does not provide a dummy function if i3C is not enabled.
>>>
>>
>> I do have another concern, though: What happens if the i2c part of the driver
>> registers and the i3c part fails to register ? module_i3c_i2c_driver() handles
>> that situation by unregistering the i2c driver, but I don't really know
>> what happens if a single module registers two drivers and one of them fails.
> 
> After use module_i3c_i2c_driver(), and remove #ifdef I3C, and disable I3C
> in config, build passed.
> 
> It possible cause by
> 
> static inline int i3c_i2c_driver_register(struct i3c_driver *i3cdrv,
> 					  struct i2c_driver *i2cdrv)
> {
> 	int ret;
> 
> 	ret = i2c_add_driver(i2cdrv);
> 	if (ret || !IS_ENABLED(CONFIG_I3C))
> 		return ret;
> 
> ^^^ !IS_ENABLED(CONFIG_I3C) is true, so linker skip below part. So no
> ref to i3cdrv, so linker remove all related codes.
> 

Yes, but I don't think we can rely on the compiler removing the call to
devm_regmap_init_i3c().

Thanks,
Guenter