Re: [PATCH] regulator: rt5033: Use linear ranges to map all voltage selection

[ChiYuan Huang]

Axel Lin <axel.lin@xxxxxxxxxx> 於 2021年7月1日 週四 下午12:41寫道：
>
>
>
> cy_huang <u0084500@xxxxxxxxx> 於 2021年7月1日 週四 上午11:54寫道：
>>
>> From: ChiYuan Huang <cy_huang@xxxxxxxxxxx>
>>
>> Instead of linear mapping, Use linear range to map all voltage selection.
>>
>> Signed-off-by: ChiYuan Huang <cy_huang@xxxxxxxxxxx>
>> ---
>> Even though commit 6549c46af855 ("regulator: rt5033: Fix n_voltages settings for BUCK and LDO")
>> can fix the linear mapping to the correct min/max voltage
>> But there're still non-step ranges for the reserved value.
>>
>> To use the linear range can fix it for mapping all voltage selection.
>> ---
>>  drivers/regulator/rt5033-regulator.c | 23 +++++++++++++++--------
>>  1 file changed, 15 insertions(+), 8 deletions(-)
>>
>> diff --git a/drivers/regulator/rt5033-regulator.c b/drivers/regulator/rt5033-regulator.c
>> index 0e73116..2ff607c 100644
>> --- a/drivers/regulator/rt5033-regulator.c
>> +++ b/drivers/regulator/rt5033-regulator.c
>> @@ -13,6 +13,16 @@
>>  #include <linux/mfd/rt5033-private.h>
>>  #include <linux/regulator/of_regulator.h>
>>
>> +static const struct linear_range rt5033_buck_ranges[] = {
>> +       REGULATOR_LINEAR_RANGE(1000000, 0, 20, 100000),
>> +       REGULATOR_LINEAR_RANGE(3000000, 21, 31, 0),
>> +};
>> +
>> +static const struct linear_range rt5033_ldo_ranges[] = {
>> +       REGULATOR_LINEAR_RANGE(1200000, 0, 18, 100000),
>> +       REGULATOR_LINEAR_RANGE(3000000, 19, 31, 0),
>> +};
>> +
>>  static const struct regulator_ops rt5033_safe_ldo_ops = {
>>         .is_enabled             = regulator_is_enabled_regmap,
>>         .enable                 = regulator_enable_regmap,
>> @@ -24,8 +34,7 @@ static const struct regulator_ops rt5033_buck_ops = {
>>         .is_enabled             = regulator_is_enabled_regmap,
>>         .enable                 = regulator_enable_regmap,
>>         .disable                = regulator_disable_regmap,
>> -       .list_voltage           = regulator_list_voltage_linear,
>> -       .map_voltage            = regulator_map_voltage_linear,
>> +       .list_voltage           = regulator_list_voltage_linear_range,
>>         .get_voltage_sel        = regulator_get_voltage_sel_regmap,
>>         .set_voltage_sel        = regulator_set_voltage_sel_regmap,
>>  };
>> @@ -39,9 +48,8 @@ static const struct regulator_desc rt5033_supported_regulators[] = {
>>                 .ops            = &rt5033_buck_ops,
>>                 .type           = REGULATOR_VOLTAGE,
>>                 .owner          = THIS_MODULE,
>> -               .n_voltages     = RT5033_REGULATOR_BUCK_VOLTAGE_STEP_NUM,
>> -               .min_uV         = RT5033_REGULATOR_BUCK_VOLTAGE_MIN,
>> -               .uV_step        = RT5033_REGULATOR_BUCK_VOLTAGE_STEP,
>> +               .linear_ranges  = rt5033_buck_ranges,
>> +               .n_linear_ranges = ARRAY_SIZE(rt5033_buck_ranges),
>>
>
> If you want to use linear range here, you need to change RT5033_REGULATOR_BUCK_VOLTAGE_STEP_NUM back to 32
> rather than delete the .n_voltages setting.

Sorry, I really forget the N_VOLTAGES.
>
> I'm fifty-fifty about the change because I don't see any benefit with converting to linear range (even though in theory it's correct).
> The voltage of all entries in the second linear range is the *same* as the latest selector of the first linear range.
> When the regulator core to choose the best selector, it will always select the latest selector of the first linear range if it meets the requested range anyway.
> (Because the entries in the second linear range are not *better*, it's just the same.)
>
> If the initial version is this driver is using linear range then it's fine.
> But given the initial version is using linear so when I fix the n_voltages setting I decide to not change it to linear range.
> This makes it easier to fix older versions if necessary.
> (I'm not sure if linear range is available in some old kernel versions, the initial version of this driver was committed in 2014).
>
>
>From the regulator register in probe, it will get the current voltage
from the IC.
If the vout sel is not is over N_VOLTAGES, it will return the error number.

But as I think it's the side effect to change the vout step num.
To use the linear range is just to guarantee all vout sel range are included.

That's my initial thoughts.
> Regards,
> Axel
>