SCSI-disk formatting

Per Lindgren pl at mizar.docs.uu.se
Wed Aug 15 19:15:23 AEST 1990


I have a (hopefully) simple question about the formatting of SCSI disks.
My departement purchased two MiniScribe 9380S SCSI disks this summer. Now
I have some strange behaviour with the format program. Ignorant as I am
about the internals of SCSI, I assumed the format process was the same as
for a disk hooked to a Xylogics 753 controller. Apparently it isn't.  To
start with, there are no defect lists on the disk (or on paper) that
format can access. When I asked the one we purchased the disks from, he
informed me of the fact that SCSI takes care of defective sectors
"internally", my Sun should never be bothered with that. So, sometimes
(but not always) when I ask the disk to give me the "original" defect list
it just gives me an error (offending/unknown command), and *sometimes* it
seems to produce a defect list:

0x89898989      17  0x04e8833b
   1       2     12   10926      -1       0
   2     182      2    8780      -1       0
    ...
    ...
  17    1045      7    9930      -1       0

When during format the program tries to write the defect list it fails
with a command error. As far as I can see, all failures occur when format
is trying to write to the last 2 cylinders on the disk (the alternate
cylinders).  I have tried to vary the number of cylindes (i.e. use fewer),
but the same error occurs on the last two cylinders.  Note that it is not
a media error, it is an illegal command error. 

Some questions:

- When formatting a SCSI disk, is it really necessary for the format
  program to know how many heads etc. that the disk has? Isn't SCSI
  addressed "sequentially"?

- Is the format program on my Sun (3/60, SunOS 4.0.3) "stupid" enough not
  to treat SCSI disks in the right way?

In the end I managed to get the disk thru the format program (with the
errors mentioned above) and to create a file system on it. But can I trust
the disk?  Pleas answer with email. If anything interesting crops up I'll
summarise it for the net.

  --------------------------------------------------------------------------
  |Per Lindgren			  |	pl at mizar.docs.uu.se		   |
  |Dept. of Computer Systems      |	pl at kuling.docs.uu.se		   |
  |Uppsala University, SWEDEN     |		   			   |
  --------------------------------------------------------------------------



More information about the Comp.sys.sun mailing list