[Letux-kernel] [Tinkerphones] CI20 - was Re: New LetuxOS Kernels and some tricks and thoughts

[Paul Boddie]

On Tuesday 11. June 2019 17.38.20 H. Nikolaus Schaller wrote:
> Hi Paul,
> also adding the letux kernel mailing list because discussion becomes quite
> kernel focussed...

OK, maybe my message will bounce, though, because I am not subscribed to that 
list.

[...]

> I have just booted sucessfully the letux-5.2-rc4 kernel. Log is attached.
> It seems to ignore NAND completely.

Yes, support for the specific kind of NAND on the CI20 was dropped in the 
mainline kernel because it is the "wrong" type, or something. There was 
argument about it being unreliable and no-one being able to know when it 
fails, or something like that, so the kernel developers just pushed it over 
the side. Obviously, this came as a surprise to anyone thinking they could 
upgrade their kernel and expect everything to keep working.

Anyway, I tried booting the 5.1.8 kernel (from upstream) attached to the 
serial cable and it did actually boot in the end. I don't get any logging to 
UART0 despite changing the device tree file and doing "make uImage" but maybe 
that isn't sufficient. Also, I couldn't figure out how to make systemd start a 
terminal over UART0 (/dev/tty0), which was easily done using /etc/inittab 
before some people decided they knew better about how things should be set up.

So, I ended up investigating using the UART4 connector which is unreliable 
mechanically for those of us using female header connectors.

> The only problem seems to be "spidev at 0 enforce active low on chipselect
> handle" which indicates a missing DT property.
> 
> What I wonder is where the LEDs (LED0 .. LED3) near the Ethernet controller
> chip DM9000C are connected to... I could not locate them in the
> schematics...

Aren't they the MMC indicators?

> It mentions only the RED/BLUE D5 and two wires LED1 and LED2 going from
> DM9000C to the RJ45 socket which has the two Ethernet status LEDs.
> 
> It is likely that they go to some SD8..SD15 of the DM9000C but it seems that
> the schematics do not exactly match my (pink CI20 V2a) board. But I also
> could not locate an LED or GPIO controller driver for the DM9000C... So
> there is no easy way to get them blinking (heartbeat, cpu, mmc activity).

So it seems that the LEDs are a general problem given that they do not 
indicate MMC activity and made me think that the board hadn't booted. Since I 
don't have the board connected with a serial cable and Ethernet at the same 
time, I don't know whether the networking actually works, but ifconfig does 
seem to show eth0 set up, albeit without an address.

Maybe I will try and let the board boot on the network again and just wait for 
it to appear or not appear.

Paul