Robert Black <r.black@ic.ac.uk> writes:

> 
> I take it that it is safe to assume that interrupt code which is interrupting
> the kernel never, under any circumstances, reads or writes userland memory or
> does anything else which could potentially cause a fault.

That's not really true.  There are two cases where user-level memory
is accessed from an interrupt handler:

1) to modify a profiling counter
2) to modify some I/O buffer

In both cases, the buffer is wired ahead of time so that a fault
shouldn't occur.  (In the case of profiling counters, however, the
memory may have been unmapped, so there's a special mechanism for
handling faults.)