Subject: Gateway 2000 disk problems and 1.0
To: None <current-users@NetBSD.ORG>
From: John C. Hayward <johnh@david.wheaton.edu>
List: current-users
Date: 01/28/1995 22:06:29
Dear NetBSDers,
   I have a lab of 11 486's running NetBSD-1.0.  We had been running 
NetBSD-current as of early September and over the Christmas updated to
1.0 including the new file system. 
   Since updating I have noticed occasionally
disk errors on some of the stations.  These machines came from the factory
with energy conservation measures including disk power down timer.  Many
times when there would be a seek error indicating it took too long to 
complete when doing disk activity after the disk had spun down.  On some
of the machine these seemed to be followed by complaints about a particular
block having IO errors.
   Reading the whole Unix partition resulted in no errors.  Repartitioning
to DOS and doing a surface scan produced no errors.
   I have disabled the power saving feature related to the disk and it seems
that the disk error messages have disappeared as well.
   However first one then another and now a third machine has become "stuck"
in that attempting to change the active partition reports success but on 
reboot the old active partition is still active.  One of this is "stuck"
with the Unix partition and the other two are stuck on DOS.
   I have written a DOS program which does an int 13 to read and write
sectors in attempting to figure out the problem.  On one of the "stuck"
machines the only differences noted from a machine which is not stuck are
that: 
1) on return from a write sector where sector is 1 or 2 on track 0
  AL is 50 on the stuck machine while on a working machine 01 is in AL.
  AH is 0 and carry clear which would seem to indicate a successful write
2) when writing sector 1 or 2 to a different contents that what it previously
  had then reading that sector back the original contents are still there
  indicating that the write really failed.

It seems that sector 1 and 2 on track 0 are now "readonly" for some reason.

Another item which may be related is that I have changed the IO memory 
address on the SMC Ethernet card from cc00 to ec00 on at least 2 of the
machines which are experiencing this problem.

My attempts to contact Gateway at this point have been fruitless (to talk
to tech support you have to call in when there is an open line - they do
not have a phone queue - no reply to e-mail yet either).

Since the I386 people on current have considerable experience I thought
someone might know:

1) Has anyone else had problems with NetBSD and power saving disk option?

2) I could not see any mention of a "lock partition table" in the BIOS
   but one could imagine to protect against viruses which attack the partition
   table someone coming up with a way to lock the partition table.
   Any one know if this is possible?

3) Since I moved the IO address of the Ethernet card so It's last address
   is EFFF is probing of that address space (or just beyond it) the cause
   of problems.

4) I have documentation for a Zenith BIOS and there is an function which
   allows formating of a track.  It requires the sector information be
   stored in memory before doing the format.  The controller I have is
   a western digital built into the mother board.  Will this work with
   this controller and what is the exact format of the sector information?
   I would like to try an format track 0 and see if that resolves my 
   problem.

Any help and pointers would be appreciated.
johnh...