[no subject]

[Claus Stovgaard <claus.stovgaard@xxxxxxxxx>]

Hi All.

Will like to discuss the current state of using spidev from devicetree.

I am working with some custom hardware having a Xilinx ZynqMP as main
device with some added programmable logic accessible via SPI. As the
SPI communication depend on what is loaded in the programmable device,
and when it is loaded, it is handled solely by user space, because it
is the user space application there decided when to load the FPGA and
with what.

So currently I am using a small patch just adding a simple compatible
for spidev, on our 4.19 kernel.
{ .compatible = "linux,spidev" },
And assign it in the devicetree for the spislave of the programmable
logic so it can be accessed using spidev.

I would prefer not to have patches for something like this, and I
think this is trivial and should be a feature from Linux. Many in my
situation might just reuse one of the "compatible" lines already
present om spidev.c for their devicetree even though it does not fit
the hardware. I don't just think it is a very nice solution.

Then I was thinking about creating a patch with a compatible string
like "unknown,custom-hardware" or similar. Something which can
indicate that there is a custom spislave, and create an interface for
accessing via the kernel, but don't have a better option when loading
devicetree.

What is yours response to the idea of creating a custom-hardware
binding for spidev, intended to be used for programmable hardware
unknown at the devicetree time.

Or is there a better way for creating spidev interfaces in this situation.

Regards
Claus