01:08 <smaeul> LordKalma1: the driver you linked does not use SPI at all. so you can stick the drm_display_mode in drivers/gpu/drm/panel/panel-simple.c and call it a day. this would sidestep the whole file/module authorship question.

01:09 <smaeul> I'm assuming it's only a "SPI" driver because of copy/paste. and I bet the panel doesn't use an ST7701s either, that was just a convenient file to hijack

04:10 <jernej> smaeul: ST7701s really uses SPI, if it's connected to RGB bus

04:11 <jernej> but yeah, worth to check if it is really there

07:36 <LordKalma> smaeul, the panel in question does use the ST7701S (notice the S) and it does use SPI. There are _very_ similar looking drivers in there honestly

07:37 <LordKalma> In the meanwhile, the found out that the driver works fine on 5.8, but looks weird, and there are other problems on higher kernels. Seems to be a clocks issue, haha so weird

07:37 <LordKalma> no wonder why the manufacturer used 5.8 and not an LTS release

08:51 <LordKalma> problem reproduced on a 5.10 kernel

08:52 <LordKalma> wonder what happened between 5.8 and 5.10 to break graphics for this device

08:53 <bauen1> LordKalma: perhaps git-bisect will help you

08:54 <LordKalma> haha newton's method on big? amazing

08:55 <LordKalma> I currently don't have the compilation infrastructure setup, only other community member, but we'll get to the bottom of this, we hope

08:58 <bauen1> LordKalma: git bisect is basically binary search over the commits, so at worst it will take log2(n) where n = amount of commits, so still a lot but doable

08:59 <LordKalma> yeah, reading the docs, interesting

08:59 <LordKalma> 09:54 <LordKalma> haha newton's method on git*? amazing

08:59 <LordKalma> is what I meant

12:43 <smaeul> LordKalma: if the panel uses ST7701S, I would expect to see some long series of SPI transactions inside st7701s_prepare

12:43 <LordKalma> it uses that IC inside

12:43 <smaeul> for example, https://github.com/smaeul/linux/commit/451ecd29714d04bc83dae1a8789632809fa2c2da is a driver for some generic 480x480 panel using ST7701S

12:43 <LordKalma> dunno how the stuff happens

12:43 <smaeul> but https://github.com/ruilvo/linux-stable/blob/v5.15-feature-st7701s/drivers/gpu/drm/panel/panel-sitronix-st7701s.c doesn't perform any SPI transactions

12:45 <LordKalma> that's a fair point

12:45 <LordKalma> other drivers seem to look exactly like that

12:45 <LordKalma> but you're right that it doesn't seem to do anything special with the SPI bus

12:47 <smaeul> so presumably the chip is configured with all of the right parameters already (in NVM). At that point Linux doesn't care what the controller is, it just has to turn the panel power on, and panel-simple is designed for that case

12:59 <DuClare> Or maybe it isn't configured right and that's why it looks wrong :P

13:12 <LordKalma> we discovered it looks good on 5.8

13:12 <LordKalma> but not above

13:14 <LordKalma> and it's not just the display. I quote "waterfall is updating better than 10x slower than it should. on the console i see lots of errors from the qt app about the rx buffer being full"

13:14 <LordKalma> sounds like clocks gone wrong

13:15 <LordKalma> do you know if there is anything noteworthy about that between 5.8 and 5.10?