On Tue, Feb 19, 2019 at 6:54 AM Rob Herring robh@kernel.org wrote:
I believe using eDP connector binding wouldn't help much in my case and it won't improve accuracy of hardware description while adding unnecessary code duplication (edp-connector will be pretty much simple-panel).
Since currently there're no standalone connector drivers, implementing one requires significant refactoring of the code that I'm not familiar.
I'm not talking about drivers. I'm talking about bindings. Those are not necessarily 1-1. There's no reason the simple panel driver can't have an 'edp-connector' entry.
These aren't independent things. Bindings are useless without drivers.
Also how are you going to address mainlined platforms that have eDP and use panel bindings? E.g. rk3399 supports eDP and uses panel binding - see rk3399-gru-kevin.dts as example.
Also, since you have EDID, you should be using that for timing data IMO, and the binding needs to have enough information to support that. It may if DP-aux comes from the bridge chip or it may not if you need to describe that connection. Again, this is independent from what Linux chooses to do. If Linux chooses to have its own timing information that's its choice. Another OS may choose to use EDID.
I see nothing wrong here - anx6345 driver reads EDID (over AUX) and uses timing from it, if reading EDID failed it fallback to timings defined in panel driver.
Rob