> > > > 
> > > > Feb 15 17:00:07 kaptah /netbsd: com0: 2 silo overflows, 0
> 
> OK, this is likely a MP issue then, something blocking interrupts
> for too long (maybe an issue with the kernel big lock ?)
> I suspect a UP kernel won't have this issue

Would there be any merit in getting a kernel stack trace at the
point at which the silo overflow is detected?
It is likely to be where interrupts are enabled after they have
been disabled for a long time.

	David

-- 
David Laight: david@l8s.co.uk