2025-01-31T12:54:04+0000 Miod Vallat <miod@online.fr>:
> Thanks to off-list debugging with Miguel Landaeta, I have been able to
> improve these changes a lot.
> 
> The previous iteration of this diff would not attach on his system
> because the keyboard controller is found before the GPIO controller.
> This is now fixed by postponing GPIO interrupt registration a bit.
> 
> While I was working with interrupts, I have also reworked the attachment
> logic.
> It is now as follows: if your pckbc@acpi node has GPIO interrupts, or
> regular interrupts not configured in legacy, ISA-compatible (i.e.
> edge-triggered, rising front), then the legacy pckbc@isa has no chance
> to behave properly, and we attach; otherwise we let it claim the
> keyboard unless you change the device flags to 1 to force attachment.
> 
> This should remove the need for a list of "known to need acpi
> attachment" devices, and in fact, I have removed it now.
> 
> Please test.

Tested on my Thinkpad E14 Gen 6 (AMD) and found no regressions. Thanks!