18:21 <hexdump0815> digetx: i tested 5.15.0 with the drm fix from a few days ago on my 2gb 1366x768 nyan and it was working well - yesterday i tested it on the other nyan with 4gb 1920x1080 and ran into another problem i already saw before

18:22 <hexdump0815> the problem is that on boot sometimes the console works and sometimes it does not - xorg in the end always works in both cases

18:22 <hexdump0815> i managed to collect debug info based on your script for both cases

18:23 <hexdump0815> working console case: https://pastebin.com/raw/qqUCTzGF

18:23 <hexdump0815> non working console case: https://pastebin.com/raw/rSsqSEHR

18:23 <hexdump0815> both are from the same machine - the chance is much higher to end up without a working console but from time to time it sometimes works

18:27 <hexdump0815> from the dmesg i noticed two things: there is an i2c timeout (i guess while getting info from the panel?) in the nonworking case and later the real full hd mode of the panel is getting pruned in the non working case so that it ends up at the non native 1366x768 which maybe results in the black screen

18:27 <hexdump0815> i guess xorg is doing its own mode probing later and succeeds at the proper full hd mode in both cases

18:28 <hexdump0815> any idea what might be the problem here? the hardware is fully ok i would say as the legacy kernel with linux and chromeos are both work perfectly find on this device

18:30 <hexdump0815> correction to above: the chance is much higher to end up with a non working console than with a working one

18:30 <hexdump0815> oh - no correction needed it was correct abvoe already, just worded differently :)

19:41 <digetx> failing I2C indeed should be the problem, there are too many reasons why it may fail, I can't fix it without having device on my hands

19:43 <digetx> there is no proper device-tree for the 4gb model with 1080 panel, perhaps a simple workaround could be to add and use device-tree that has the 1080 panel and not rely on DDC

19:47 <digetx> it's wrong to use incorrect panel anyways because userspace will see two modes supported by the 1080 panel and the 768 mode simply can't work

23:33 <hexdump0815> digetx: do you have an example dts maybe how to supply the panel timing via dtb? do i have to disable or switch anything to make it work this way?

23:34 <hexdump0815> i might also have a look at the different chromeos dts files - there are multiple revisions of them at least - maybe this gives a hint on where the i2c problem might come from