Re: [PATCH v2 1/2] spi: stm32_qspi: Add transfer_one_message() spi callback

[Patrice CHOTARD <patrice.chotard@xxxxxxxxxxx>]

On 8/10/22 15:40, Mark Brown wrote:
> On Wed, Aug 10, 2022 at 03:31:59PM +0200, Patrice CHOTARD wrote:
>> On 8/10/22 15:23, Mark Brown wrote:
>>> On Wed, Aug 10, 2022 at 03:15:08PM +0200, Patrice CHOTARD wrote:
> 
>>> Yes.  Though I'm not clear if the bindings actually want to enforce it
>>> there, it's a device level property not a controller level one so it
>>> might not be something where controller support gets validated.
> 
>> Ah yes, i see, parallel-memories should not be used in our qspi controller node.
>> So i can't reuse parallel-memories for my purpose.
> 
>> So i need to add a new proprietary property at controller level as done in the v1 ?
> 
> Can't the controller figure this out by looking at the properties of the
> connected devices?  You'd need to just return an error if we ever
> triggered transfer_one_message() on a device that can't support the
> operation.

It should be a solution.

I just noticed another point, property parallel-memories is an array of uint64 which represent device's size.
In case a FPGA is connected to the qspi 8 line bus, parallel-memories property will be set with what ?
simply random value to make dtbs_check happy ?

IMHO, adding a new proprietary property would be cleaner.