On Thu, Jul 03, 2025 at 10:34:23AM +0200, Hans de Goede wrote:
Hi Thomas,
On 3-Jul-25 8:47 AM, Thomas Zimmermann wrote:
quoted
Hi
Am 02.07.25 um 22:43 schrieb Krzysztof Kozlowski:
quoted
On 30/06/2025 10:40, Hans de Goede wrote:
quoted
quoted
No one asks to drop them from the driver. I only want specific front
compatible which will list and constrain the properties. It is not
contradictory to your statements, U-boot support, driver support. I
really do not see ANY argument why this cannot follow standard DT rules.
So what you are saying is that you want something like:
framebuffer0: framebuffer@1d385000 {
compatible = "qcom.simple-framebuffer-sm8650-mdss", "simple-framebuffer";
}
and that the binding for qcom.simple-framebuffer-sm8650-mdss
can then list interconnects ?
IMO yes (after adjusting above to coding style), but as mentioned in
other response you can just get an ack or opinion from Rob or Conor.
But does that work with *any* device that requires interconnects? The next such simple-framebuffer device should work out of the box *without* the kernel knowing anything about it. That's one of the key features of the simple-framebuffer. If we have to maintainer per-device feature sets, it breaks that assumption.
The driver code for this can still be generic and since the driver
will bind to the fallback plain "simple-framebuffer" compatible
this should also work for new platforms.
The e.g. "qcom.simple-framebuffer-sm8650-mdss" compatible would
purely be something in the dt-bindings to document which simplefb
implementations will have interconnects and which ones will not.
The driver does not necessarily need to check these more
precise compatibles, it can still just check for the generic
presence of interconnects.
This ship has kind of sailed though. This binding has been used by
plenty of firmwares and bootloaders over the years, and has been
deployed on plenty of devices already.
Good luck fixing it in all of them, and then updating every device.
Maxime