[Lenny400] [Letux-kernel] L4Re/Fiasco.OC on the Letux 400

[Paul Boddie]

On Tuesday 17. April 2018 21.34.32 H. Nikolaus Schaller wrote:
>

[Testing the tentative 4.x kernel on the Letux 400]

> same for me... I just keep rebasing it on latest upstream
> release-candidates. And doing a test compilation.
> 
> What I didn't find time for was to open the L400 and solder a connector
> for access to the serial console. This seems to be the first step before
> testing anything. Even if the device boots the very old kernel from flash.

In fact, I've done all the L4 porting work without a serial console, but this 
builds on all the effort getting things to work on the Ben NanoNote, since I 
have had quite a bit more experience of getting the Ben to do what I want. 
Also, the bootloader on the Ben initialises the LCD, so I was able to gain 
experience under less adverse circumstances, although I think I understand the 
LCD peripheral fairly well and could probably write a quick routine to 
initialise it. Writing colours and bit patterns to the display was a simple 
but sufficient way of seeing how far the CPU was getting.

So, I just had to be fairly confident about the bootstrap code and the kernel 
working just as it does on the Ben, and instead of using the display for 
debugging, the Caps Lock and Num Lock LEDs were used for simple signalling 
purposes. Doing this for the Linux kernel would probably be slightly more 
difficult but not impossible. Indeed, I may attempt such tricks with Linux on 
the Ben since recent kernels do not work with it, either.

[...]

> As you indicate it is important to know that your kernel patches can not
> be completely wrong :) So it is another stimulus to finally try to boot.

Indeed. Although the structure of the above patches is rather different to any 
Linux kernel patches, the knowledge encoded by them both is the same, at least 
with regard to the hardware interfacing.

[...]

> Oh, that would be an interesting task. What I don't know is if the current
> QtMoko has any hard coded focus on armhf. So switching to mipsel
> (soft-float) may be easy or may not...

I will have to investigate. QtMoko should be portable, as indeed should Qt be, 
but I was also thinking about Fiasco.OC plus runtime environment plus QtMoko 
on things like the GTA04, Pyra, PocketBeagle, and other ARM-based devices. But 
the feasibility or desirability of this is speculation at this point.

> > Anyway, I hope this is of interest to someone out there!
> 
> Yes, definitively.

Good to know! :-)

Paul