Michael K. Sanders wrote:

> I seem to have gotten it fixed, thanks to Michael Hitch.
> 
> Michael wrote:
> 
>   There was a bug in procfs that was fixed a while back that might 
>   be causing the problem.  If I remember correctly, there was an
>   uninitialized variable.  It may have been in miscfs/procfs/procfs_mem.c,
>   and I think the variable was kva.  The current file initializes it
>   to VM_MIN_KERNEL_ADDRESS just prior to the vm_map_find() call.  If
>   you're building your own kernel, you might try fixing that and see
>   if it helps.
> 
> I changed this and all seems well so far. The panic I was getting was
> from pmap_enter_ptpage, which I remember from my attempts at running a
> system with 4MB RAM. Might this be related to the random crashes
> people have experienced on machines without a lot of RAM?
> 
> So to all those having random crash problems (arrrgh!), try this and
> see if it helps. Fixing an uninitialized variable certainly can't _hurt_.

Could this cause random crashes even on machines with lots of RAM?

Is it possible that the crashes would show up as complete lockups with
no panic or other error messages?  I think someone somewhere said that
with the 1.0 kernel, you would get lockups instead of error messages because
of some other bug (in the debugger?).

I have had random lockups on my A3000+16MB/64MB every now and then.
Sometimes the system crashes in less than 24 hours, sometimes it runs
a long time without problems.  Currently the uptime is >12 days, and
the system has been used by 5-6 users simultaneously every day.

I have been thinking that the crashes could occur due to an uninitialized
variable.  One sign of that is (I think) that whenever I shutdown to
single user mode and then return to multi user without turning off the
computer first, the machine tends to crash pretty soon.

Any comments, Gurus?

  -jm