Subject: Re: segmentation faults with new loadbsd
To: Waldi Ravens <waldi@moacs.indiv.nl.net>
From: Leo Weppelman <leo@ahwau.ahold.nl>
List: port-atari
Date: 05/05/1995 11:57:34
Hi Waldi,

> 
> 10/16 is the exact amount of physically available memory. I wasn't even
> aware that loadbsd supports a -T option, are there any other options
> available?
Yep, try 'loadbsd -h'

> BTW when I tried to start loadbsd under MagiC, it reported `Unknown FPU
> type'. Allthough MagiC does fully replace TOS, it's hard to believe it
> also replaces the hardware. :-))
> 
> I think the difference is that TOS is lazy and sets the _FPU cookie
> to 0x20000 (type unknown), while MagiC sets 0x60000 (68882). Loadbsd
> probably expects to find the lazy TOS value.
I do indeed check the cookie jar, only I got the FPU values wrong (one
bit shifted right :-( ), I will correct this. It's just lucky it works
for TOS.
>  
> > Did you install the distribution sets? 
> 
> No, this won't be possible until I can mount an msdos, minix or ext2
> filesystem under NetBSD. Somehow the archives must be accessible by
> gzip&tar under NetBSD, and I'm definitely not going to use floppies,
> that would take ages (well, at least the best part of my life). :-)
> 
The msdosfs-code is maintained by mycroft. I read your (next) mail
about the msdosfs bootsector values and I think that it is best
to work it out a bit and forward it to mycroft. For the time being
however, I think our best bet is to get the 'NHEADS' and 'SectPerTrack'
in line with the checks. It won't matter for Atari and it can easily
be done with a little program. At least you are able to continue
the installation :-) .

> I did play a bit with ufs on floppy:
> 
> 	# newfs -N /dev/rfd0a
> 	... 720 sectors in 80 cylinders of 1 tracks, 9 sectors
> 	...
> 	# newfs /dev/rfd0a
> No problem, just the error message on the disk label.
> 	# fsck /dev/rfd0a
> 	BAD SUPERBLOCK: MAGIC NUMBER WRONG
> Amazing!
The problem will be that the floppy-device your using is 360Kb. I
wonder why newfs doesn't complain, this is obviously wrong!
Floppy devices are coded as:
	a  - 360Kb
   b  - 720Kb
   c  - 1.44Mb
I have already prepared a new MAKEDEV script on my machine to update
the one on lamp. It has the above explanation about floppy formats
and it _also_ created fdxc! You can not yet use the 1.44Mb floppy's
however. I fixed a bug in the floppy driver yesterday evening :-(
I think I fixed it but I cannot try it because my drive can't handle
it.
The new MAKEDEV also creates an rd0 entry :-)
> 
> Well, after an hour or so of running newfs and fsck, one of them
> changed its mind about the magic number. Fsck reported and repaired
> missing BLK(S) IN BITMAP (whatever that may be). Mount only needed
> one attempt, so there I sat with a new ufs with the enormous amount
> of 223 1K-blocks. :-)
> 

> > It makes me very curious about what's wrong with Benni's configuration.
> > Well just wait and see what results he gets.
> 
> I'm getting more and more curious about your configuration, as I happen
> to encounter the same (or similar) problems Benni had (or still has).
I was referring to his memory config. The problems you have with MSDOS
fs depend probably on the version of AHDI who made the fs'. On my
filesystems, the unused values are zero.
> 
> I also tried to mount a CD-ROM (mount_cd9660). After a promising start
> some error message appears about trying to access blocks past the end
> of the device or something like that (sorry, forgot to write it down).
I am using the generic CD-ROM code. I have never tried it as I do not have
a CD-ROM. Maybe I should take one from work home to test it.

Leo.