Re: [PATCH v1 6/7] soc: starfive: Add dphy pmu support

[Changhuang Liang]

On 2023/4/12 5:15, Conor Dooley wrote:
> On Mon, Apr 10, 2023 at 11:47:42PM -0700, Changhuang Liang wrote:
[...]
>> +++ b/MAINTAINERS
>> @@ -19944,6 +19944,7 @@ F:	include/dt-bindings/reset/starfive?jh71*.h
>>  
>>  STARFIVE JH71XX PMU CONTROLLER DRIVER
>>  M:	Walker Chen <walker.chen@xxxxxxxxxxxxxxxx>
>> +M:	Changhuang Liang <changhuang.liang@xxxxxxxxxxxxxxxx>
> 
> Unmentioned in the commit message, plus I would like an R-b or an Ack
> from Walker.
> 

OK, I will make a discuss with Walker.

>>  S:	Supported
>>  F:	Documentation/devicetree/bindings/power/starfive*
>>  F:	drivers/soc/starfive/jh71xx_pmu.c
>> diff --git a/drivers/soc/starfive/jh71xx_pmu.c b/drivers/soc/starfive/jh71xx_pmu.c
>> index 990db6735c48..d4092ca4dccf 100644
[...]
>> @@ -94,6 +97,8 @@ static int jh71xx_pmu_get_state(struct jh71xx_pmu_dev *pmd, u32 mask, bool *is_o
>>  
>>  	if (pmu->match_data->pmu_type == JH71XX_PMU_GENERAL)
>>  		offset = JH71XX_PMU_CURR_POWER_MODE;
>> +	else if (pmu->match_data->pmu_type == JH71XX_PMU_DPHY)
> 
> There are only two options for this "enum", so why `else if`?
> 

OK, will change to else.

>> +		offset = JH71XX_PMU_DPHY_SWITCH;
>>  
>>  	regmap_read(pmu->base, offset, &val);
>>  
>> @@ -170,6 +175,23 @@ static int jh71xx_pmu_general_set_state(struct jh71xx_pmu_dev *pmd, u32 mask, bo
>>  	return 0;
>>  }
>>  
[...]
>>  static int jh71xx_pmu_set_state(struct jh71xx_pmu_dev *pmd, u32 mask, bool on)
>>  {
>>  	struct jh71xx_pmu *pmu = pmd->pmu;
>> @@ -191,6 +213,8 @@ static int jh71xx_pmu_set_state(struct jh71xx_pmu_dev *pmd, u32 mask, bool on)
>>  
>>  	if (pmu->match_data->pmu_type == JH71XX_PMU_GENERAL)
>>  		ret = jh71xx_pmu_general_set_state(pmd, mask, on);
>> +	else if (pmu->match_data->pmu_type == JH71XX_PMU_DPHY)
>> +		ret = jh71xx_pmu_dphy_set_state(pmd, mask, on);
> 
> Perhaps I am verging on over-complication, but I dislike this carry on.
> Is this the only time we'll see a power domain provider coming out of
> a syscon, or are there likely to be more?
> Either way, I think having an ops struct w/ both parse_dt() and the
> set_state() implementations would be neater than what you have here.
> 
> Very much open to dissenting opinions there though. Emil? Walker?
> 
> Cheers,
> Conor.
> 

"else if" will change to "else"
As far as I know, there are only two types power domain on the JH7110 SoC.
One is the original, another one is coming out of a syscon.

>>  
>>  	return ret;
>>  }
[...]
>> 2.25.1
>>