Re: [PATCH v5 05/10] power: supply: rt5033_charger: Add RT5033 charger device driver

[Jakob Hauser <jahau@rocketmail.com>]

Hi Christophe, Hi all,

On 14.05.23 16:31, Christophe JAILLET wrote:
> Le 14/05/2023 à 14:31, Jakob Hauser a écrit :

...

>> +static int rt5033_charger_probe(struct platform_device *pdev)
>> +{
>> +    struct rt5033_charger *charger;
>> +    struct power_supply_config psy_cfg = {};
>> +    int ret;
>> +
>> +    charger = devm_kzalloc(&pdev->dev, sizeof(*charger), GFP_KERNEL);
>> +    if (!charger)
>> +        return -ENOMEM;
>> +
>> +    platform_set_drvdata(pdev, charger);
>> +    charger->dev = &pdev->dev;
>> +    charger->regmap = dev_get_regmap(pdev->dev.parent, NULL);
>> +
>> +    psy_cfg.of_node = pdev->dev.of_node;
>> +    psy_cfg.drv_data = charger;
>> +
>> +    charger->psy = devm_power_supply_register(&pdev->dev,
>> +                          &rt5033_charger_desc,
>> +                          &psy_cfg);
>> +    if (IS_ERR(charger->psy))
>> +        return dev_err_probe(&pdev->dev, PTR_ERR(charger->psy),
>> +                     "Failed to register power supply\n");
>> +
>> +    charger->chg = rt5033_charger_dt_init(charger);
>> +    if (IS_ERR_OR_NULL(charger->chg))
> 
> Hi,
> 
> Nit: charger->chg can't be NULL.
> 
>> +        return -ENODEV;
> 
> Why bother returning specific error code in rt5033_charger_dt_init() if 
> they are eaten here.
> 
> return PTR_ERR(charger->chg)?
> 

Thanks for the heads-up.

...

Writing towards the list:

The way it is done in the current patchset is taken from the original 
patchset of March 2015 [2]. I kept the original as far as possible.

By now I'm not happy with the way of initializing "struct 
rt5033_charger_data". I realized this in the course of the review. As I 
didn't want to disturb the review with this, I had planned a small 
clean-up patch after this review is finished.

The cause of the complicated handling of "struct rt5033_charger_data" 
lies inside of the "struct rt5033_charger". There the "struct 
rt5033_charger_data" is initialized as pointer *chg.

The clean-up would be:

  - Inside of "struct rt5033_charger" change the
    "struct rt5033_charger_data" to non-pointer "chg". It is then
    initialized right away.

       struct rt5033_charger_data      chg;

  - Change function rt5033_charger_dt_init() from type
    "struct rt5033_charger_data" to type "int".

       static int rt5033_charger_dt_init(struct rt5033_charger *charger)

  - In the probe function, call the function rt5033_charger_dt_init() in
    the same way like e.g. the following rt5033_charger_reg_init():

       ret = rt5033_charger_dt_init(charger);
               if (ret)
                       return ret;

  - Within function rt5033_charger_dt_init() and all other functions
    using the charger data, get the address of the already-initialized
    struct &charger->chg.

       struct rt5033_charger_data *chg = &charger->chg;

This would also solve the issue reported by Christophe because the 
errors inside function rt5033_charger_dt_init() would be passed to the 
probe function by the "ret =" and being returned there with "return ret".

I'm not sure how to handle this now. I would prefer to get the review of 
this patchset finished and send a clean-up patch afterwards.

[2] 
https://lore.kernel.org/lkml/1425864191-4121-1-git-send-email-beomho.seo@samsung.com/T/#u

Kind regards,
Jakob