> 
> pefo@enea.se (Per Fogelstrom)  wrote:
>  > Suggestion: Put an printf in the interrupt handler and print out the
>  > cause register. If you get there tell me what you got.
> 
> err, I hope this doesn't sound stupid but the interrupt handler is
> in locore.S from what I can see and I'm not familiar enough with this
> to figure out how to output from locore... Sorry, I'm a 68k person
> myself... 
> 
> If you have a code fragment you'd like me to try.........
> 
No, not stupid at all. I wasn't specific enough. I suspect it is an
non expected interrupt that kills the system. Have seen it many times
before :-). This is a code fragment from 'pmax/trap.c':

/*
 * Handle an interrupt.
 * Called from MachKernIntr() or MachUserIntr()
 * Note: curproc might be NULL.
 */
interrupt(statusReg, causeReg, pc)
        unsigned statusReg;     /* status register at time of the exception */
        unsigned causeReg;      /* cause register at time of exception */
        unsigned pc;            /* program counter where to continue */
{
        register unsigned mask;
        struct clockframe cf;

#ifdef DEBUG
        trp->status = statusReg;
        trp->cause = causeReg;
        trp->vadr = 0;
        trp->pc = pc;
        trp->ra = 0;
        trp->code = 0;
        if (++trp == &trapdebug[TRAPSIZE])
                trp = trapdebug;
#endif
    --->
    |   cnt.v_intr++;
    |   mask = causeReg & statusReg;    /* pending interrupts & enable mask */
    |   if (pmax_hardware_intr)
    |           splx((*pmax_hardware_intr)(mask, pc, statusReg, causeReg));
    |
    |
    |  Now here it's beginning to be interesting. Values of interest is
       'statusReg' and 'causeReg'. Having a printf here could be anoying
       when the clock is started but to me it looks that you don't get
       that far. Just put in:

	printf("Interrupt cause = %x, status %x\n", causeReg, statusReg);

	and see what happens.

Per