[Letux-kernel] [PATCH 00/20] A bunch of JZ4730 fixups for letux-kernel

[Paul Boddie]

On Saturday, 26 December 2020 14:54:53 CET H.  Nikolaus Schaller wrote:
> Oh,
> nice. Yes, I'll try it even if it does not fix anything.

It probably won't, but it might still be useful in seeing how the timers 
behave as a result. I didn't figure out how to change the width of the timer 
counter, and the documentation wasn't readily helpful.

(Who thought it was a good idea to write documentation in YAML?!)

> Simplifying code is also important, like I have done for the I2C driver.

I'll look at that a bit more. A message will be sent about it in a while.

> Maybe a suggestion: according to upstream rules, commit messages should be
> imperative (i.e. tell what the patch is doing and not what we have done).
> I.e. "Simplify JZ4730 TCU driver modifications"
> 
> And it is usually helpful to prefix by the subsystem abbreviation, i.e.
> 
> "clk: ingenic: Simplify JZ4730 TCU driver modifications"
> 
> But it is not highest priority, especially as we will finally squeeze
> all our fixes into a single patch...

Indeed. I'm sure I have seen people make a strong case for avoiding the 
imperative in other projects precisely because the reader can be left with the 
understanding that the fix (or whatever) isn't made and that the message is 
actually instructing the reader to continue the work, which is what the 
imperative is for, after all.

> In the meantime I have added CONFIG_DRM_PANEL_SIMPLE and now we get a
> /dev/fb0.
> 
> But display remains black. There are three potential reasons:
> a) backlight is not enabled

It may be possible to control the backlight via a special file.

> b) LCD timings are no longer defined in the DT but the panel-simple driver
> c) panel is not enabled by hardware setup (LCD_EN/TP204) which is an output
> of the P89LPC915 So this may be an I2C issue...
> 
> It may even be a mix of all of them.

Maybe. I had forgotten the involvement of panel-simple, which I am sure was 
discussed a while ago. Looking at my notes, there was a patch that lets the 
panel-simple driver read panel details from the device tree, and this requires 
the compatible attribute to be given as "panel-dpi". This code is now 
upstream, apparently.

I now see that there is code in the driver to read timings for non-DPI panels 
(panel_simple_parse_panel_timing_node). However, I don't see a way of 
specifying a generic "panel-simple" and have the driver activated on that 
basis. None of the Chunghwa panels in the file are comparable to the one used 
by the Mipsbook.

The panel_dpi_probe function reads the timings plus width and height from the 
device tree. In contrast, panel_simple_parse_panel_timing_node only reads the 
timings. Both functions supposedly calculate the appropriate bus flags. I have 
no idea at the moment how the appropriate bus format is obtained. For this 
panel, we need to end up with MEDIA_BUS_FMT_RGB565_1X16, I think.

Paul