OK, it happened again.  My info from the first time is appended and here's
the stack trace this time.  Again hand copied and I'm afraid I rebooted
before I could double-check all the numbers this time (so the hex in the
first dump may be more accurate).

I note that the new dump is missing the aarpresolve/ddp routines so that
would seem to let netatalk off the hook as a cause.  _mi_switch is common,
but called by a different route.  Kind of looks like the mc... routines
really are the common thread.

FWIW the machine hung over the weekend, but I couldn't use the debugger.  I
broke into it, but couldn't type anything.

_Debugger
_panic
_trap
faultstkadj(4aac00,548b80,80,4aac00,527e00) + 0
_mcstart(4aac28,1,4aac00,7efd8c,c9374) + 5e
_mc_tint(4aac00) + aa
_mcintr(4aac00,7efe18,34b4,7efda4,0) + 58
_psc_lev3_intr(7efda4) + 36
_lev3intr(?)
_mi_switch(0,2004,527e00,0,7eff80) + 14
_sys_select
_syscall
_trap0

>I've had the machine die twice in the last week with a VM fault, followed by
>mc0: receive FIFO overflow, so I guess I should pay attention.  The kernel
>this
>happened with is at ftp://macbsd.jpl.nasa.gov/pub/users/hotz/netbsd.hotz4
>together with the applied patches and the config file if anyone is that
>ambitious.
>
>The partial stack trace (copied by hand) is:
>_Debugger
>_panic
>_trap
>faultstkadj(4aac00,549100,33def884,4aac00,7f2c9c) + 0
>_mcstart(4aac28,33def801,4aac00,7f2b94,c9374) + 5e
>_mc_tint(4aac00) + aa
>_mcintr(4aac00,7f2c18,34b4,7f2bac,2404) + 58
>_psc_lev3_intr(7f2bac) + 36
>_lev3intr(?)
>_aarpresolve(4aac28,544a80,7f2c9c,7f2c62) + 14
>_ether_output(4aac28,544a80,7f2c9c,0,532080) + 22a
>_ddp_route
>_ddp_input
>_atintr
>_netintr
>_trap
>rei(?)
>_mi_switch
>_tslup
>_sys_select
>_syscall
>_trap0

Signature failed Preliminary Design Review.
Feasibility of a new signature is currently being evaluated.
h.b.hotz@jpl.nasa.gov, or hbhotz@oxy.edu