NEC D-2363 disk geometry for Rimfire 3200 controller
Paul Martin
unido!athen!ecrcvax!paul at uunet.uu.net
Tue Aug 22 01:14:10 AEST 1989
Can anyone out there in net land give some advice in calculating the
geometry of a NEC D-2363 900MB disk running on the following combination
of hardware and software :
Rimfire 3200 controller,
Sun 3/280S, OS 4.0.1
Our problem is that the NEC D-2363 disk is supposed to deliver a disk
capacity of approx 900MB. We have a similar disk using a Sun Xylogics 7053
SMD controller on a second Sun server and this indeed does provide us with
the correct disk capacity. The capacity difference stems from the fact
that the recommended geometry of the *SAME* disk is substantially
different for the two controllers, particularly in the number of heads.
For the Xylogics controller, /etc/format.dat uses 27 heads, 964 cylinders
and 67 sec/cyl giving a total disk capacity of 1743876 sectors or 851 MB
of real data space.
For the Rimfire rfutil suggests the use of 19 heads, 1022 cylinders and 67
sec/cyl giving a total disk capacity of 1301006 sectors or *ONLY* 635MB.
I tried changing the geometry of the disk as seen by the Rimfire
controller to make it match that of the Xylogics controller. I was able to
format, label and newfs all the disk partitions without any problems. The
UNIX commands newfs, df, mount, umount and dkinfo all worked - everything
except for fsck which complained about a corrupt super block and
instructed the use of the "-b" option to use an alternate super block.
This made no difference every superblock seemed to be corrupted in the
same way. However I was still able to mount/umount all the disk partitions
?? Perhaps some guru can explain that one ?
Well I have now reverted back to the rfutil recommended geometry as this
is the only labelling that I could find that would allow fsck to work. I
really did not want to leave a system that would always hang on reboot -
even if it meant losing more than 200 MB of disk space.
Obviously there is something wrong with the geometry given in rfutil - and
I would rather not experiment on our network master machine. Can anyone
else throw some light on my confusion.
Thanks in advance for any assistance,
================================================================================
Paul Martin
European Computer-Industry Research Centre ECRC, Arabellastr. 17/II
US: paul%ecrcvax.uucp at pyramid.pyramid.com 8000 Muenchen 81, West Germany
..!pyramid!ecrcvax!paul Tel. (089) 92699 124
Europe: paul at ecrcvax.uucp
..!unido!ecrcvax!paul
================================================================================
More information about the Comp.sys.sun
mailing list