Re: [PATCH 2/3] clk: twl: add clock driver for TWL6032

[Krzysztof Kozlowski]

On 28/08/2023 18:24, Andreas Kemnade wrote:
> Hi,
> 
> On Thu, 24 Aug 2023 09:04:28 +0200
> Krzysztof Kozlowski <krzysztof.kozlowski@xxxxxxxxxx> wrote:
> 
>>>>> +
>>>>> +MODULE_DESCRIPTION("Clock driver for TWL Series Devices");
>>>>> +MODULE_ALIAS("platform:twl-clk");    
>>>>
>>>> This alias is unnecessary?
>>>>  
>>> The question is whether this driver should have a separate dt
>>> node (and if a separate node, then one per clock as the clk-palmas
>>> driver) or not. See Rob's review of the binding document.
>>> So we have basically #clock-cells = <1>; in the twl parent
>>> and a call to mfd_add_device() there in the former case and I guess
>>> that alias is needed then.
>>>   
>>
>> You should not need the alias in any of these cases. platform alias for
>> platform driver means you have incomplete tables and use alias instead
>> of tables. Preference is to use tables.
> 
> Is there a general consensus that MODULE_ALIAS("platform:.*") should be
> exorcised? Of course for this new driver I will avoid it now anyways.

Whether "general" I don't know, but I was removing it quite a lot in the
past. I think I removed all at some point, now I guess we have them back :/.

MODULE_ALIAS is not the correct way to solve module matching problem. ID
table with the correct way. Alias is just a workaround which now works,
but later might stop (e.g. ID table will come with additional features).

Best regards,
Krzysztof