Re: [PATCH] ARM: dts: at91: add serial MFD sub-node for usart

[Codrin.Ciubotariu]

On 02.11.2020 14:29, Lee Jones wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
> 
> On Mon, 02 Nov 2020, Codrin.Ciubotariu@xxxxxxxxxxxxx wrote:
> 
>> On 02.11.2020 11:01, Lee Jones wrote:
>>> On Fri, 30 Oct 2020, Nicolas Ferre wrote:
>>>
>>>> On 30/10/2020 at 12:07, Codrin Ciubotariu wrote:
>>>>> The "atmel,at91sam9260-usart" driver is a MFD driver, so it needs sub-nodes
>>>>> to match the registered platform device. For this reason, we add a serial
>>>>> subnode to all the "atmel,at91sam9260-usart" serial compatible nods. This
>>>>> will also remove the boot warning:
>>>>> "atmel_usart_serial: Failed to locate of_node [id: -2]"
>>>>
>>>> I don't remember this warning was raised previously even if the MFD driver
>>>> was added a while ago (Sept. 2018).
>>>>
>>>> I would say it's due to 466a62d7642f ("mfd: core: Make a best effort attempt
>>>> to match devices with the correct of_nodes") which was added on mid August
>>>> and corrected with 22380b65dc70 ("mfd: mfd-core: Ensure disabled devices are
>>>> ignored without error") but maybe not covering our case.
>>>>
>>>> So, well, I don't know what's the best option to this change. Moreover, I
>>>> would say that all other USART related properties go into the child not if
>>>> there is a need for one.
>>>>
>>>> Lee, I suspect that we're not the only ones experiencing this ugly warning
>>>> during the boot log: can you point us out how to deal with it for our
>>>> existing atmel_serial.c users?
>>>
>>> You should not be instantiating drivers through Device Tree which are
>>> not described there.  If the correct representation of the H/W already
>>> exists in Device Tree i.e. no SPI and UART IP really exists, use the
>>> MFD core API to register them utilising the platform API instead.
>>>
>>> This should do it:
>>>
>>> diff --git a/drivers/mfd/at91-usart.c b/drivers/mfd/at91-usart.c
>>> index 6a8351a4588e2..939bd2332a4f6 100644
>>> --- a/drivers/mfd/at91-usart.c
>>> +++ b/drivers/mfd/at91-usart.c
>>> @@ -17,12 +17,10 @@
>>>
>>>    static const struct mfd_cell at91_usart_spi_subdev = {
>>>           .name = "at91_usart_spi",
>>> -       .of_compatible = "microchip,at91sam9g45-usart-spi",
>>>    };
>>>
>>>    static const struct mfd_cell at91_usart_serial_subdev = {
>>>           .name = "atmel_usart_serial",
>>> -       .of_compatible = "atmel,at91rm9200-usart-serial",
>>>    };
>>>
>>>    static int at91_usart_mode_probe(struct platform_device *pdev)
>>
>> [snip]
>>
>> Hi Lee, thank you for looking through our usart driver and for sharing
>> your thoughts. Removing the usage of compatible string means that for
>> similar serial/SPI IPs we would need to create new platform drivers.
> 
> Why would you need to do that?

In the case we will have to support another similar IP, but with a 
different set of features. Not a new platform driver from scratch, but 
at least a new struct platform_driver for each variant.

> 
>> This is not ideal, but it's a solution. What I proposed is more
>> flexible, but, as you pointed out, I am not sure it correctly describes
>> the HW, because the decision of whether to use this IP as a serial or a
>> SPI is a configurable one.
>>
>> Thanks and best regards,
>> Codrin
> 
> --
> Lee Jones [李琼斯]
> Senior Technical Lead - Developer Services
> Linaro.org │ Open source software for Arm SoCs
> Follow Linaro: Facebook | Twitter | Blog
>