> > Re. fatal page faults during panic ... I can't help but wonder of 
> > something is trying to operate on curproc while it's NULL.  Note that 
> > disk_unbusy() is typically called in an interrupt context.
>
> [ ... ]
>
> Thud.  I wonder if there are any non-panic situations when mi_switch() can
> be called when curproc is NULL?

mi_switch() cannot legally be called if curproc is NULL.  if curproc
is NULL, then sleep() or tsleep() has been called while in an
interrupt context, and that's a big no-no.

whatever code is doing this is buggy.


cgd