2.6.24-rc7 update status

[Werner Almesberger]

Andy Green wrote:
> Maybe JinKyoungKwon's issue posted here yesterday is the same, he didn't
> mention if it is GTA-01.  He says he can boot by hand from U-Boot,
> something to try.

That's how I'm usually booting. The kernel is loaded, uncompresses,
then ... silence :-(

If a system doesn't make it past the splash screen, but comes up when
booted manually, this might be some u-boot configuration problem. Or
an old u-boot, with bug 887.

> If the motion sensors are still giving trouble, when I am done with
> Glamo MMC I am wondering if it, along with HDQ and possible vibrator
> problems Nod mentioned, can be helped by implementing the FIQ stuff at
> this point.

The acceleration sensor problem currently seems to be that it
permanently asserts its interrupt. I already fished a few bugs out of
the interrupt logic, and the schematics look fine, but there seems to
be more. The bitbanging seems to be okay, though.

One problem with the acceleration sensors is that they want to get
serviced at at least 100Hz. This means that combined worst-case
latency of interrupt handling, workqueue scheduling, and getting the
communication done must be all below 10ms, which may be a little tough.

Sometimes I wish we had some fast and/or flexible MCU sit somewhere,
tapping into all our serial links ...

We can of course play some digital filtering tricks to cope with
losses. Some good interrupt-driven FSM would help as well, of course.

Anyway, that's optimization. First the interrupts have to work
properly at all.

- Werner