> >         - You said:
> >       >                 tar -tvf /zip1/netbsd-usr.tar
> >       >                 ==> panic: buffer larger than expected
> >       It's no clear to me if you mean the MAXBSIZE message here. Because
> >       it looks like both the source & destination of the copy were ffs-type
> >       filesystems.
> 
> You're both right. I forgot the word "MAXBSIZE" ! The correct message was
> 	panic: buffer larger than MAXBSIZE expected
> It's also correct: the /zip1 contained a FFS-type filesystem (source). 
> But the destination exactly was the terminal (only listing tar-file - 
> t-option). May be this messages comes, because I changed the media just
> before the above command from a MSDOS-type fs to ffs.
But then the panic should have happened at umount-time, not after you
mounted an ffs filesystem. I'm puzzled.....
> 
[ ... ]
> 
> >    - All those panics look like somehow data transfer through SCSI
> >      ends up at places that are outside the planned DMA area... I don't
> >      really get this :-( And why does this not happen on Markus' Falcon,
> >      is this hardware????? I think about a way to verify this.
> 
> Tja, my question ! To help to clear my hardware I can say that my actual
> Falcon has a NEW motherboard which is very newer as my last one of last year;
> this is from about first quarter of 1993. But the disk is a very old MFM-disk
> from about 1988 from CDC-WREN V or something (588 MB raw disk space).
> However the ZIP (IOMEGA) is very new (from Oct 1995)
It might also be a combination af things. That's what makes this kind of things
so hard to fix.
> >    - The fact that sometimes your Falcon must be turned off for a while
> >      gives me the idea that the kernel 'forgets' to initialize something.
> 
> Haeh' can you explain this a little ??? What happens, what is the difference
> when Falcon is switched of a _longer_ time or a _shorter_ time ?
>From a hardware point of view switching it of longer than 30 seconds should
be equal to eternity. Usually > 10 seconds should be enough, but with enough
capacitors, some things might hold on a bit longer - not RAM though.

I took the liberty of uploading a slightly modified BOOTX kernel to your
ftp area. It's called netbsd.gz (I wanted to give it a more descriptive
name, but this was denied). I tweaked some timings a little and I checked
the DMA pointer a bit harder. Just to see if it is really pointing were
it was meant to. It is a polling kernel. Let's see....

Leo.