slow softintrs

[Michael <macallan%netbsd.org@localhost>]

Hello,

I keep getting KASSERT() panics from soft interrupt handlers on SMP  
macppc, usually it's about curcpu or curthread not being what some  
mutex expects. Since I don't see this on sparc64 for instance I  
suspect the slow softintr code. To confirm or contradict this:
- is anyone else seeing those panics on any SMP powerpc machine?
- which other archs use slow softintrs? Do we support SMP on any of  
them?
Finally - is anyone working on fast softintr support for powerpc?

have fun
Michael