this is a pmap bug.  on mac68k, pmap_collect() does:

		pmap_remove(pmap, VM_MIN_ADDRESS, VM_MAX_ADDRESS);

this will remove wired mappings even though pmap_collect()
is not allowed to remove wired mappings.

I fixed one instance of this problem a while back on x86, and I thought
this was reported to the maintainers of affected platforms, but maybe
I'm mistaken about that.  I guess I should go open a PR for each platform
which still has this bug.  note that even on x86 there appears to be
another bug which triggers this same panic.

-Chuck


On Sun, Feb 11, 2001 at 04:39:45PM -0800, Alex Zepeda wrote:
> I noticed that a few build proceses hung when I attempted to kill them via 
> Ctrl-C.  Eventually a ps -xa hung, and well, the processes that hung were 
> listed in ps, but kill said they didn't exist.   A reboot command panic'd 
> the kernel.  Here's the trace:
> 
> panic: uvm_fault_unwire_locked: unwiring non-wired memory
> Stopped in pid 145 (ntpd) at _kdbprinttrap+0x5a:	unlk	a6
> ...
> kdbprinttrap
> panic
> uvm_fault_unwire_locked
> uvm_unmap_remove
> uvm_unmap
> uvm_deallocate
> exit1
> syssys_exit
> syscall(1)
> trap0
> 
> Hmm... maybe it's time to get a disk so I can at least keep a kernel dump.
> 
> - alex