[Letux-kernel] Status LX16

[H. Nikolaus Schaller]

> Am 22.12.2024 um 19:50 schrieb Andreas Kemnade <andreas at kemnade.info>:
> 
> Am Sun, 22 Dec 2024 19:33:36 +0100
> schrieb "H. Nikolaus Schaller" <hns at goldelico.com>:
> 
>>> Am 22.12.2024 um 19:25 schrieb Andreas Kemnade <andreas at kemnade.info>:
>>> 
>>> Am Sun, 22 Dec 2024 18:55:28 +0100
>>> schrieb "H. Nikolaus Schaller" <hns at goldelico.com>:
>>> 
>>>> One other aspect which could be better than it is, is the management of the device tree for variants. We still have a single lx16.dtb
>>>> which must be compiled into the kernel binary. Even if we had multiple .dtb in /boot. I still have no idea how to manage that
>>>> better than we did so far with #if in the DTS file.  
>>> 
>>> If I remember correctly, the u-boot can load a dtb from file, the one
>>> in the kernel is only a fallback.  
>> 
>> Well, when starting the X1600 project we investigated that but it did not work.
>> Something was fundamentally different for MIPS from ARM.
>> 
>> I do not exactly remember what it was.
>> 
> I investigated u-boot and found everything was there... when I thought
> about patching u-boot.
> 
>> Maybe that the other ingenic devices (CI20 and others) are all expected to
>> be configured for builtin-dtb. And then the LX16 would fall out of the scheme.
>> Or that the arch/mips/boot/ingenic/Makefile can't do that or something...
>> 
> the builtin-dtb (even if configured) is a fallback. If something comes
> from the bootloader, it is used instead of the builtin-dtb. At least
> that was what I found from code research. I did not remember whether I
> practically tested that.

So we should investigate that later. It seems to be the right way to do,
but AFAIR there was some practical reason why we did not do it right from
the beginning.

May be something as simple as that it is not possible to tell the U-Boot
binary which *.dtb to load. If that must be hard-coded into U-Boot all the
flexibility is gone and a .dtb built into the kernel is equally "flexible" :)

Or even simpler, that the vendor 5.15 kernel was not prepared to handle
that. So the result would be that one has to switch u-boot and kernel between.
But that would be no longer a limitation. I think it is at least 6 months ago
thatcI booted a vendor kernel, if not 12.

BR,
Nikolaus