[Letux-kernel] Raspi3B+ support

[H. Nikolaus Schaller]

Hi,
I am still fighting with upstream kernel...

The issue that upstream kernel has a different
naming scheme for the .dtb (bcm2837 vs. bcm2710) is
now fixed by simply making a copy to the "old style"
name.

But I still have no console output with the upstream
DTBs.

And there seems to be a dependency from .config.

If I enable CONFIG_MMC_BCM2837 and CONFIG_DMA_BCM2837
(which are only existing upstream), there is a blinking
pattern on the green LED. It blinks twice.

According to

https://elinux.org/R-Pi_Troubleshooting#Green_LED_blinks_in_a_specific_pattern

this should say that the SD card is unreadable...

Well, it is readable since I just have to copy a proper
kernel/device tree and then it works again. So bootloader
can read the SD.

What I suspect is something else: the kernel ends in
some kernel panic which makes the boot-loader to restart
and something left the MMC interface in a bad state so
that the boot loader can't read the SD card on second attempt
unless I do a power cycle.

So in summary:
* bcm2710-rpi-3-b-plus.dtb	built from github tree boots with letux-kernel (CONFIG_MMC_BCM2835=y and CONFIG_DMA_BCM2837=y)
* bcm2837-rpi-3-b-plus.4.16.dtb	provided by Mark renamed to bcm2710-rpi-3-b-plus.dtb does not boot with letux-kernel
* bcm2837-rpi-3-b-plus.dtb	renamed to bcm2710-rpi-3-b-plus.dtb does not boot with letux-kernel

Looks as if I should really focus on analysing the differences
of the github and kernel.org DTB file used for bcm2710
resp. bcm2837. And play around with bcm2837.dts

It starts with that the serial0/1 aliases are defined completely
different.

Having a working console with the bcm2837.dtb will make things easier...

BR,
Nikolaus