Re: [RFC PATCH v2 08/10] rtc: bd70528: Initial support for ROHM bd70528 RTC

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

On 1/27/19 11:48 PM, Matti Vaittinen wrote:
> Thanks again Guenter,
> 
> On Sat, Jan 26, 2019 at 08:30:24AM -0800, Guenter Roeck wrote:
>> On 1/25/19 3:05 AM, Matti Vaittinen wrote:
>>> +/*
>>> + * We read regs RTC_SEC => RTC_YEAR
>>> + * this struct is ordered according to chip registers.
>>> + * Keep it u8 only to avoid padding issues.
>>> + */
>>> +struct bd70528_rtc_day {
>>> +	u8 sec;
>>> +	u8 min;
>>> +	u8 hour;
>>> +};
>>> +
>>> +struct bd70528_rtc_data {
>>> +	struct bd70528_rtc_day time;
>>> +	u8 week;
>>> +	u8 day;
>>> +	u8 month;
>>> +	u8 year;
>>> +};
>>> +
>>> +struct bd70528_rtc_wake {
>>> +	struct bd70528_rtc_day time;
>>> +	u8 ctrl;
>>> +};
>>> +
>>> +struct bd70528_rtc_alm {
>>> +	struct bd70528_rtc_data data;
>>> +	u8 alm_mask;
>>> +	u8 alm_repeat;
>>> +};
>>
>> At least some of the above are directly associated with chip registers.
>> I don't think this will work for all architectures without explicit packed
>> attribute.
> 
> Allright. I was thinking of that but thought that most of the
> architectures using this PMIC would handle alignments fine if I used
> only u8 members. I did consider using __attribute__((packed)) - but I'm
> not sure if we hit into troubles with that too. I guess some people
> would like to compile kernel with other compiler(s) but gcc - although
> I'm not sure if this should be taken into account. I'll try doing some
> study on this - unless someone replies to this and just tells how this
> should be done. (I am pretty sure I can find the answer from mail
> archives though). I'll try adding some packing hint for compiler at v3.
> 

Use __packed ?

>>> +		if ((!enable) == (!(*old_state & BD70528_WAKE_STATE_BIT)))
>>> +			return 0;
>>
>> I think
>> 		if (enable == !!(*old_state & BD70528_WAKE_STATE_BIT))
>> would be much better readable. Even if not, there are way too many ()
>> in the above conditional.
> 
> Allright. I'll fix this
>   
>>> +	if (alm.alm_mask & BD70528_MASK_ALM_EN)
>>> +		a->enabled = 0;
>>> +	else
>>> +		a->enabled = 1;
>>> +
>> Without conditional:
>> 	a->enabled = !(alm.alm_mask & BD70528_MASK_ALM_EN);
>>
> 
> Right. Much nicer, thanks! I'll change this.
> 
>>> +static int bd70528_set_time(struct device *dev, struct rtc_time *t)
>>> +{
>>> +	int ret, old_states;
>>> +	struct bd70528_rtc_data rtc_data;
>>> +	struct bd70528_rtc *r = dev_get_drvdata(dev);
>>> +	struct bd70528 *bd70528 = r->mfd;
>>> +
>>> +	ret = bd70528_disable_rtc_based_timers(r, &old_states);
>>> +
>>
>> AFAICS the disable/enable functions are only called once. Since they
>> also apply set / clear a mutex, I find that a bit confusing. I think
>> it would be better to fold the code into this function. If anything,
>> I could imagine something like
>>
>> 	mutex_lock();
>> 	ret = bd70528_set_time_locked();
>> 	mutex_unlock()
>>
>> to simplify error handling.
> 
> Yep. Makes sense. I'll tidy this.
> 
>>> +	ret = regmap_bulk_read(bd70528->chip.regmap,
>>> +			       BD70528_REG_RTC_START, &rtc_data,
>>> +			       sizeof(rtc_data));
>>> +
>>> +	tm2rtc(t, &rtc_data);
>>> +
>>> +	ret = regmap_bulk_write(bd70528->chip.regmap,
>>> +				BD70528_REG_RTC_START, &rtc_data,
>>> +				sizeof(rtc_data));
>>> +
>>> +	ret = bd70528_re_enable_rtc_based_timers(r, old_states);
>>> +
>>
>> Kind of off that all the error returns are ignored here.
> 
> And I'll fix this too.
> 
> Br,
> 	Matti Vaittinen
>