On Tue, 2002-07-23 at 08:55, anton wilson wrote:

> Strangely enough I'm getting processes exiting with preemption count of 1 
> when I use my patch.
> What causes such a problem?

It probably means you are off-by-1 somewhere in your scheduler locking
that is keeping the preempt_count at 1 instead of at 0 in normal
situations.

This also means you are not preempting.

A good check to see if your patch is right would be to diff, for
example, kernel/sched.c in 2.4-ac against yours... apply my patch to a
recent 2.4-ac and see the differences.

I have also attached a preempt-kernel patch for O(1) scheduler on 2.4.18
- if this applies cleanly then everything should be fine.  The patch is
not entirely up-to-date but may be of use.

	Robert Love