Re: [PATCH 0/2] pinctrl: meson: use one uniform 'function' name

[Kevin Hilman]

Yixun Lan <yixun.lan@xxxxxxxxxxx> writes:

> Hi Jerome:
>
> On 01/10/2018 03:28 PM, Jerome Brunet wrote:
>> On Wed, 2018-01-10 at 10:12 +0800, Yixun Lan wrote:
>>>
>>> On 01/08/18 16:52, Jerome Brunet wrote:
>>>> On Mon, 2018-01-08 at 15:33 +0800, Yixun Lan wrote:
>>>>> These two patches are general improvement for meson pinctrl driver.
>>>>> It make the two pinctrl trees (ee/ao) to share one uniform 'function' name for
>>>>> one hardware block even its pin groups live inside two differet hardware domains,
>>>>> which for example EE vs AO domain here.
>>>>>
>>>>> This idea is motivated by Martin's question at [1]
>>>>>
>>>>> [1]
>>>>>  http://lkml.kernel.org/r/CAFBinCCuQ-NK747+GHDkhZty_UMMgzCYOYFcNTrRDJgU8OM=Gw@xxxxxxxxxxxxxx
>>>>>
>>>>>
>>>>> Yixun Lan (2):
>>>>>   pinctrl: meson: introduce a macro to have name/groups seperated
>>>>>   pinctrl: meson-axg: correct the pin expansion of UART_AO_B
>>>>>
>>>>>  drivers/pinctrl/meson/pinctrl-meson-axg.c | 4 ++--
>>>>>  drivers/pinctrl/meson/pinctrl-meson.h     | 8 +++++---
>>>>>  2 files changed, 7 insertions(+), 5 deletions(-)
>>>>
>>>> Hi Yixun,
>>>>
>>>> Honestly, I don't like the idea. I think it adds an unnecessary complexity.
>>>> I don't see the point of FUNCTION_EX(uart_ao_b, _z) when you could simply write 
>>>> FUNCTION(uart_ao_b_z) ... especially when there is just a couple of function per
>>>> SoC available on different domains.
>>>>
>>>> A pinctrl driver can already be challenging to understand at first, let's keep
>>>> it simple and avoid adding more macros.
>>>>
>>>
>>> Hi Jeromeï
>>>   In my opinion, the idea of keeping one uniform 'function' in DT (thus
>>> introducing another macro) is worth considering. It would make the DT
>>> part much clean.
>> 
>> Ok this is your opinion. I don't share it. Keeping function names tidy is good,
>> I don't think we need another macro to do so.
>> 
>>>   And yes, it's a trade-off here, either we 1) do more in code to make
>>> DT clean or 2) do nothing in the code level to make DT live with it.
>> 
>> I don't see how adding a macro doing just string concatenation is going to make
>> anything more clean. It does not prevent one to write FUNCTION_EX(uart_ao_b,
>> _gpioz), resulting in uart_ao_b_gpioz, which is what is apparently considered
>> 'not clean'
>> 
> for the benefits of introducing macro 'FUNCTION_EX', it will end with
>  .name = "uart_ao_b", -> same for both EE, AO domain, and it will match
> the DT part (although still different for '.groups')
>
>
>> BTW, there no cleanness issue here, the name is just out of the 'usual scheme'
>> but there is no problem with. If you want to change this, and
>> s/uart_ao_b_gpioz/uart_ao_b_z/, now is the time to change it. 
>> 
> I'd rather *NOT* to push a pinctrl patch for just changing
> 'uart_ao_b_gpioz' to 'uart_ao_b_z' (it's a cosmetic change, and still
> end with two different name - 'uart_ao_b_gpioz/z' & 'uart_ao_b' in DT)

FWIW, I agree with Jerome.

Rather than this patch adding a(nother) hard-to-understand macro, please
submit a pinctrl rename/cleanup to s/uart_ao_b_gpioz/uart_ao_b_z/.
Since there are not any users of the _gpioz name, now is the time to do
it.

We're already using the _x _y _z suffixes all over the place, and IMO,
adding this new macro would make that even more confusing that it
already is.

Kevin