>I've asked a similar question some time ago wrt. IWM access and even got
>a reply from you, remember? And I have exactly the same problem right
>now. No, not exactly: the IWM driver stubs don't hurt anything else
>unless I run them so the rest of the kernel remains stable.
>
>The only alternative that comes to mind is an LKM. But LKM support is
>sparse and hasn't grown much since Terry Lambert left the project....

It might also be possible to map the parts of device memory you want into
a process that runs your test code. Kind of like how dt maps in screen
memory into itself to blow data directly to the screen. You'll have to
make some modifications to the kernel but I think this might provide
a nice solution.

The only problem with this is that it isn't 100% foolproof. For instance,
some of my code used to produce a problem that made the SCSI driver think
that the SCSI bus was requesting a phase change. Who'd have thought of
video code causing SCSI interrupts?! You start producing spurrious interrupts
that don't get cleared and you start having some serious problems.

Good luck!

_______________________________________________________________________
 Michael Zucca - mrz5149@rit.cs.rit.edu - http://www.rit.edu/~mrz5149/
 "I will choose a path that's clear. I will choose Freewill. "
  --Rush, Freewill
_______________________________________________________________________