> 
> I've been seeing the following panic, fairly regularly:
> 
> 	panic: unwire: page not in pmap
> 	
> the kernel debugger then starts up.  The panic seems to be occuring under
> moderate load (compiler jobs, etc).
> 
There are three types of panics that happens sometimes, this
is one of them. It is very seldom, and I've not gone any deeper
into the kernel to see where it is. But I do have some ideas what
it can be :-) This panic is probably because of either a mapping
change without locking the function first, or a forgotten flush 
of the TLB. The other two panics are:
	KSP invalid trap
This can happen sometimes when a swapped-out process get swapped
in. I'm not sure why this can happen, in cpu_swapin() i actually
frob all pages in the kernel stack to be sure they are in physical
memory before trying to use them. The last panic is:
	uba zero uentry
This is the most annoying error but it only happens when more than
one disk is used on the same MSCP controller. There may be permutations
of this panic, but it is probably because of a circular list that gets
null terminated somewhere.

> I'm not at all familiar with the kernel debugger -- is there something I 
> can do in the way of a core dump at this point (under the assumption that 
> someone might want to take a look at it to see why the pmap code is failing)?  
> 
The kernel debugger isn't useful for much yet, lots of functions
isn't implemented. I used it to get an easy process list when
hunting errors earlier, but the panics that are left is somewhat
harder to find. Someday, I (or someone else that does it before)
will fix kernel crash dumps, and then it will be easier to find
why this type of panic happens.

-- Ragge