On 11-06-21, 09:51, Pierre-Louis Bossart wrote:
Thanks Vinod for your feedback,
If you look at the *existing* code, we don't handle any "resources" with the platform devices, we use the platform_device_info.data to pass the link information. It's a void pointer. We do not touch the resource field in the platform_device_into at all.
Yes that is true I dont disagree on that part. My ask here is to make it better, it can be followed up after this but I would at least like to agree on the direction.
[...]
That's it. We did not change anything else, all the other fields are identical. We are only changing the TYPE of device and the interfaces for probe/remove but using the same information and the same device hierarchy.
The move in itself is okay but I dont think that should be the end goal.
What we suggested in this patch is only an iso-functionality change. I believe from Greg's and your feedback that there is no objection on that small step.
This is not the end-goal indeed. The second step would be to remove the intel_init.c file. I fully agree with you Vinod that this can be moved into the SOF driver, and we could do this in a follow-up step. We can also improve the partition between 'context' used by the child driver and information passed by the parent on SHIM registers and bases.
I think we'd need to agree on the details of the second step, Bard and I can work on a proposal, but I don't see a disconnect on the direction to simplify the interface. That's the right thing to do.
That is agreeable to me. Looking forward to updates to clean these bits up
Thanks