> If the instruction is different in the kernel file, it is most likely
memory I would guess, although > > there are a few other possibilities
as well (such as compiler error or kernel memory being overwritten).

Just noticed (after rebooting from another db> Trap) that just before
disk syncing on reboot it says:
Kloader_alloc_memory: can't allocate memory

Just wondering if this could be the reason?  I'm currently running
netbsd on the z50 with just the 16mb built in memory.  Could that be a
problem, should I get a 32mb chip for it asap?

Cheers
steve
---------------------------------------------------------------------
Steve Dobson                                  Tel:   +44 01904 433953
Experimental Officer                          Fax:   +44 01904 433902
Department of Archaeology                     Email: sd20@york.ac.uk
The King's Manor
University of York
York, YO1 7EP, UK
---------------------------------------------------------------------

-----Original Message-----
From: Matthew Orgass [mailto:darkstar@city-net.com] 
Sent: 16 August 2004 20:47
To: S Dobson
Cc: port-hpcmips@NetBSD.org
Subject: Re: debug


On 2004-08-16 sd20@york.ac.uk wrote:

> Ok, my z50 crashed again today, I was connected to the network, but 
> just looking at my local files with 'cd' and 'ls' - It didn't put me 
> into debug it just froze after spitting out this message:
>
> Trap: reserved instruction in kernel mode status=0x120782, cause=0x28,

> epc=0x8028f448, Vaddr=0xc3311f8 pid=7 cmd=ioflush usp=0x0 
> ksp=0x80001450 Stopped in pid7 (ioflush) at 0x8028f448: spec12 
> zero,zero,zero

  If I am reading the disassembly code right, this is movz
zero,zero,zero which is a Mips IV instruction.  It's encoding contains
only a single one bit, so it might be a hardware problem.  If you have
built cross tools (or installed the compiler natively) you can do:
/usr/mtools/bin/mipsel--netbsd-objdump -D netbsd | grep "^8028f448" to
see what your kernel says (mtools = your too directory; for native
objdump just use objdump).  This should be done on the kernel you are
actually using, as it might be a disk error, and if that shows the same
instruction (objdump might also consider it to be "ffs zero, zero" or
not disassemble it; the code should be 0xA)  then check, if possible,
the same kernel that has not been copied to that disk.  If the
instruction is different in the kernel file, it is most likely memory I
would guess, although there are a few other possibilities as well (such
as compiler error or kernel memory being overwritten).

> 1> win CE often complains about my 1gb IBM microdrive  - sometimes I
> need to take it out and put it back in for it to be readable.  Don't 
> know if that's normal but could netbsd also have these problems 
> causing kernel errors?

  A disk error might be able cause this, but I have found CE to often be
upset at disks for no apparant reason.

> 2> Don't know if this was right, but after installing netbsd onto the 
> 2> CF
> card using boot.exe in win CE the boot loader would complain saying 
> that it can't find \storage card\netbsd-GENERIC.  So I copied (and 
> renamed) netbsd-GENERIC to root (/netbsd) and altered /etc/rc.d/sysdb 
> to contain the line: kvm_mkdb "/netbsd" - don't know if that's ok or 
> not but it seemed to stop it from complaining ;-)

  I think what you did is the right thing to do.

Matthew Orgass
darkstar@city-net.com