Re: [rtc-linux] [PATCH 2/6] mfd: max77620: add core driver for MAX77620/MAX20024

[Krzysztof Kozlowski]

W dniu 08.01.2016 o 18:16, Laxman Dewangan pisze:
> Hi Krzysztof,
> Thanks for review.
> I will fix most of your comment on my next patch.
> 
> Answering to some of comment/query.
> 
> On Friday 08 January 2016 07:05 AM, Krzysztof Kozlowski wrote:
>> ()2016-01-07 23:38 GMT+09:00 Laxman Dewangan <ldewangan@xxxxxxxxxx>:
>> +                               dev_err(dev,
>> +                                   "FPS enable-input %u is not
>> supported\n",
>> +                                       pval);
>> Indentation of arguments does not seem equal here or maybe this is
>> just my email client. Have you run this through checkpatch? And
>> sparse? And coccicheck (that one definitely not because kbuild is
>> complaining)?
> I ran checkpatch before I sent.

Anyway please be sure that indentation is consistent.

> 
>> +               chip->rmap[i] = devm_regmap_init_i2c(chip->clients[i],
>> +               (const struct regmap_config
>> *)&max77620_regmap_config[i]);
>> Indentation looks weird here (or again this is my email client...).
>> The cast is even weirder?!? Why casting?
> There is some parameter difference for MAX77620 and MAX20024. I have
> only one structure for it and changing tun time so I have not define
> this structure as constant.
> Now API needs const type structure and hence casting it.

I don't quite get... usually there is no need of casting pointer to a
writable memory when function accepts pointer to const.

> 
> However, I have  define different structure for MAX77620 and MAX20024
> which are const type and hence no need to explicitly casting here. This
> will be in my next patch.

You mean v2? Okay, let's wait for that...

> 
> +static inline int max77620_reg_update(struct device *dev, int sid,
> +               unsigned int reg, unsigned int mask, unsigned int val)
> +{
> +       struct max77620_chip *chip = dev_get_drvdata(dev);
> +
> +       return regmap_update_bits(chip->rmap[sid], reg, mask, val);
> +}
> 
>> I think all these shouldn't be static inlines in header. Although some
>> of them are one-liners but rest are not. Let the compiler decide what
>> to do with these wrappers.
> 
> If I dont make inline from header then this will complain as unused
> static function on related C compilation if it is not used on C. This
> header included from all sub module driver and they are not using all
> these APIs.
> 
> To avoid compilation warning,  I need to use inline here.

Because this shouldn't be defined in header at the first place. Instead
define it in main MFD driver with EXPORT_SYMBOL() and put in headers
just declaration.

Best regards,
Krzysztof