> On Wed, 21 May 1997 16:20:30 -0400 
>  Scott Bartram <scottb@orionsoft.com> wrote:
> 
>  > I've got a current (5-17) kernel running on a 1+ that I'm trying to use
>  > as an Appletalk server. I consistently receive esp-type panics when the
>  > system gets any load. Is the 1+ a hopeless cause or what?
> 
>  > data fault: pc=f803c670 addr=f85e7f24 ser=80<INVAL>
>  > panic: kernel fault
>  > syncing disks... 4 4 done
> 
> ...this doesn't really look like an "esp-type" panic at all.  In fact, given
> that the disks sync'd up, and the crash dump succeeded, I'd say it's
> something else entirely...

Very well could be. As I said earlier, my main indication was that every
time I actually saw it panic it was preceeded by a flurry of "esp"
diagnostic messages.

> Could you add:
> 
> makeoptions 	DEBUG="-g"
> 
> to your kernel config file, config, make clean, and make, and then
> boot the resulting kernel?  That will also produce a netbsd.gdb that
> has a full debugging symbol table, useful for examining the port-mortem.
> 
> Once it crashes again, remember the "pc" value (that's the program counter),
> and do:
> 
> 	gdb -k netbsd.gdb /var/crash/netbsd.N.core
> 	list *<whatever the pc value was>
> 	bt

Will do that.

> ...that'll provide at least some better info... also, what sort of
> activity is going on when the crash occurs?

A combination of network traffic (NFS/IP and (mostly) Appletalk) combined
with something like a 'find . -print ...' that generates heavy disk IO.

FYI, the system has now been up 25 hours serving 5-10 Appletalk users but
I haven't stressed it with anything else.

scott