Subject: Re: CATS -current panics during boot
To: Dr. Stephen Borrill <sborrill@xemplar.co.uk>
From: Richard Earnshaw <rearnsha@arm.com>
List: port-arm32
Date: 10/27/1998 15:15:22
> On 25th Oct, Todd Whitesel wrote:
> > root file system type: ffs
> > panic: Prefetch abort in non-USR mode (frame=0xf35a8d44)
> >
> > Stopped in init at _Debugger+0x10: ldmdb r11, {r11, r13, r15}
> > db>
> >
> > if I type "trace" I get
> >
> > db>trace
> > _Debugger(_Debugger+0x10)
> > _panic(_panic+0x14)
> > _prefetch_abort_handler(_prefetch_abort_handler+0x10)
> > _cnopen(_cnopen+0x10)
> > _spec_open(_spec_open+0x10)
> > _vn_open(_vn_open+0x10)
> > _sys_open(_sys_open+0x10)
> > _syscall(_syscall+0x10)
> > db>
>
> I get a very similar output with a fresh -current (and Richard
> Earnshaws wdc patch). The wdc patch made the machine recognise the
> discs, but then I got a panic. However, I've got _resethandler instead
> of _prefetch_abort_handler in the above list.
>
> Also, I've got:
>
> panic: Branch to never-never land (zero)..... were (sic) dead
> prior to the above.
Hmm, I saw a "Branch to never-never..." while I was testing the above
patch; but since at the time other things weren't working (it wasn't
taking interrupts at that time), I assumed the two were related. Perhaps
this is a race condition somewhere. I haven't rebooted it more than once
since I managed to get things working.
It looks to me as though cdevsw[major(dev)].d_open in dev/cons.c is a NULL
(or nearly NULL) pointer.
Richard.