Re: Trouble booting on hx2490 (WAS: Re: status of port: is this project dead?)

On Sunday 07 January 2007 23:33, Richard Purdie wrote:
> On Sun, 2007-01-07 at 23:21 +0000, Tvrtko A. Ursulin wrote:
> > On Sunday 07 January 2007 14:12, Tvrtko A. Ursulin wrote:
> > This one looks cleaner, penguin is gone replaced with a nice progress bar
> > and it seems to know more about the device. All I get is a one or more
> > horizontal colour bands in the middle of the screen.
> >
> > But I figured out how to capture some output (see below). I don't
> > understand from did it invent initrd because it isn't specified in
> > default.txt?
>
> Would it be possible that the LCD parameters are wrong but the kernel
> boots? Try adding a GPIO call early in the kernel boot process to turn
> on the green mail LED (GPIO 106)?

It has a green LED? :) Yes I see now.. I've never read the manual so I only
had blue and orange. ;)

> A serial lead will help a lot!

One would have hoped... I got the cable today but I'm afraid I did not make
any real progress with it.

I'll throw in a bunch of facts from my testing today. It's best I can do at
the moment since I failed to establish a pattern:

First boot I tried with the serial connection was partially successful. It got
to the point of panicking since my rootfs didn't have /dev/mmcblk* nodes.
Console on tty0 worked, ttyS0 was silent.

That was pretty encouraging. So I fixed the rootfs and retried. Now it mounted
root but started getting error 2's when reading from mmc2 (bad crc if I am
not mistaking). So I remembered seeing that rootdelay parameter and increased
it to 4.

Third boot and everything was peachy. After couple seconds of silence I got an
OE splash screen. Unfortunately it got to ~20% and got stuck there. Nothing
on the serial so don't know what happened.

At this point I was pretty optimistic, I thought - it's simple now. Oh how I
was wrong.

Anyway to cut the long story short - never again it got as far as userspace. A
couple of times I got kernel printing out on tty0 and then it was mostly
failing with that mmc0 error 2 problem.

Vast majority of attempts produced only colour bands which I mentioned before.
Last ten or so attempts were completely "coulour banded" and that where I
decided to call it a day.

Never ever I got anything from the serial line. I suspected my USB-RS232
dongle but when damn ActiveSync was automatically started I saw handshake
lines being pulled. Something which never happened when booting the kernel
(and yes I have turned on hardware handshaking in some attempts).

I have try to cycle different kernels (yours, mine, last angstrom image). Mine
worked a couple of times, yours once (hey, that logo is pretty big!),
angstrom never. But I think that's a red herring.

So I guess turning on that LED early in boot could be next. Then moving it
farther and farther in init sequence until I see at which point it doesn't
lit up. But that sounds pretty damn painful - I think I'll want to reconfirm
that the serial link is healthy and plan for that is to find a terminal
emulator for ppc and see if it will talk with minicom.

To get there I need to study bitbake a bit because the last time I naively
modified some sources it had just fetched a fresh copy and overwrote my
changes.

Tvrtko
Received on Tue Jan 09 2007 - 19:03:50 EST