[Letux-kernel] jz4780 read out LCDC setup for HDMI

[Paul Boddie]

On Saturday, 18 July 2020 20:03:35 CEST H. Nikolaus Schaller wrote:
> 
> Here is a hdmi-5.8-rc5.txt created on the latest kernel but by the same
> script.

> If you do a diff to the working hdmi-3.0.8.txt you may check for significant
> and insignificant differences. Most worrying to me is that the LCDC has some
> quite different parameters. Using exactly the same panel it should be
> almost the same...

Yes, there are some curious differences.

The LCD configuration register has negative hsync polarity in the 5.8-rc5 
kernel register dump, presumably because I didn't set the appropriate mode 
flags to assert positive hsync and negative vsync. Looking for guidance, I 
found the drivers/gpu/drm/bridge/nwl-dsi.c file that uses a mode_fixup 
operation to introduce these settings. And there was also a way of setting the 
bus flags for pixel clock polarity. So, thanks to our friends at NXP and 
Purism for the examples!

I'm not going to claim that the other register differences are not important, 
but there can be some legitimate variation. One thing that indicates that the 
peripheral is active in the 5.8-rc5 configuration is the presence of a value 
in the JZ_REG_LCD_IID register. Here, it is using one of the values that I 
have set in the driver for the JZ_REG_LCD_FID0 and JZ_REG_LCD_FID1 registers, 
indicating that a frame successfully caused an interrupt condition. I don't 
think the 3.0.8 kernel uses interrupts in the same way, and that influences 
the control register and various other registers.

Testing some changes to the mode and bus flags doesn't give a picture, and 
although the modetest utility claims that the "CRTC" uses positive hsync and 
vsync, a message I sent to various driver maintainers a while ago also 
indicated this. Only inspection of the registers might indicate success or 
otherwise, here.

Anyway, I just pushed these most recent changes.

Paul