new kernel ... / meaning of stable-tracking

Werner Almesberger werner at
Mon Sep 29 20:15:15 CEST 2008

Andy Green wrote:
> But, this is not settable by i2c but only set by variant.  And on our
> variant it doesn't do that but just has it waking the device on press.

Ooooh. Now I see it "Access: R". Wow, that's what I call wicked.
I always had that one filed under "things that don't work for some
mysterious reason".

By the way, yes, the 50606 has a similar feature, only that it's not
optional there. Since our kernel normally just clears it and defers
further cleanup to user space, it's actually a little weaker than what
we have on GTA02 if the kernel is still alive. But if the kernel is
stuck to the point where it can't answer PMU interrupts anymore, then
that will get the GTA01 to shut down, where the GTA02 wouldn't.

There is another mechanism to tell the PMU that the kernel is alive
and healthy, the KEEPACT heartbeat. Unfortunately, the corresponding
signal isn't connected to the CPU in GTA02, although this could be
done by putting 0R on the unpopulated R1546.

I don't quite remember why we never implemented this.

- Werner

More information about the openmoko-kernel mailing list