On Tue, Apr 17, 2007 at 03:29:52PM +0200, Anders Lindgren wrote:

> On Tue, 17 Apr 2007, Andrew Doran wrote:
> 
> >Hi,
> >
> >With Nick's help I've figured this one out. There are two problems:
> >
> >The first is an issue with the UP scheduler locking scheme. That's fixed by
> >ditching it completely and always using the MP scheme. I have checked that
> >in on the yamt-idlelwp branch. I think that should be enough to make arm
> >work when not under load.
> 
>   Dare I ask what happens under load? :-)

Then you run into the second problem: if a RW lock or mutex is released and
it has waiters, the system can panic with a failed lock assertion. I'll see
about getting that checked in to the branch later today.

Cheers,
Andrew