Path: utzoo!dciem!nrcaer!scs!helios!uunet!dalcs!lane
From: lane@dalcs.UUCP (John Wright/Dr. Pat Lane)
Newsgroups: comp.sys.ibm.pc
Subject: Bug(?!) in Western Digital controller.
Message-ID: <3015@dalcs.UUCP>
Date: 4 Oct 88 04:46:43 GMT
Article-I.D.: dalcs.3015
Organization: Math, Stats & CS, Dalhousie University, Halifax, NS, Canada
Lines: 52

I think I have detected a bug or a fault in a Western Digital controller
and I'd like to know if anyone else has seen anything like this and if it's
a known bug in the controller (Western Digital claimed not to know of it).

I have an XT clone with a V20 CPU, Phoenix ROM BIOS ver 2.51, ST225 hard
disk and a Western Digital WD1002S-WX2 controller.  A stamp on the board
has all sorts of code numbers on it, including a "Rev X5" and the format
program comes up as "WX2 Format Revision 1.0S (c) WDC 1986";  I'm told this
is a "Super Bios" controller.

Several hard disk testing programs I have do an "ECC READ/WRITE" test which
consists of writing a test pattern to a disk sector, doing a "long" read 
(INT 13 function 0Ah) to read in the test pattern and its 4 ECC bytes.  It
then repeatedly alters a few bits in the buffer, does a long write (INT 13
function 0Bh) of the altered test pattern with the same ECC bytes and does 
a normal read to see if the test pattern comes back corrected.  Finally it
does a more drastic alteration of the test pattern, does a long write, and
a normal read, this time expecting a CRC error to be generated.

On this machine, the test works anywhere on side 0 but fails anywhere on
the othe sides.  Upon *extensive* experimentation, I have discovered that
the controller seems to be mishandling the long write function.  It works
fine when head 0 is selected, but, and I kid you not, it ignores the head
number (in register DH) and writes on side 0 regardless of which head was
selected!  This clearly explains the ECC test behaviour.  Long reads and
regular writes seem to not be affected.

I called Western Digital and (after the typical 1/2 hour wait on hold)
talked to a (typically competant sounding) engineer who said he'd never
heard of that being a known bug in any WD controller (and said he'd been
there long enough that he would have heard).  He wanted to know if I had
had any specific problems with applications.  I can't say that I have but
I was having problems with the machine hanging on disk accesses.  Doing
a low-level re-format seems to have cured that but I wonder now if that
was caused by this problem.  However, I can't think of any normal program
that would do long writes...?

Questions:
1. Has anyone else had problems like this?
2. Can anyone tell me if this is a known or likely bug in the controller ROM
   BIOS or if is likely to be a fault in my specific unit.
3. Can anyone tell me how this problem with long writes might affect the
   system in normal operations?  I mean I can live with it failing ECC tests
   but if its going to start hanging because sectors are winding up on the
   wrong disk side, I want a new controller!

Thanks for any hints on this one.
-- 
John Wright      /////////////////     Phone:  902-424-3805  or  902-424-6527
Post: c/o Dr Pat Lane, Biology Dept, Dalhousie U, Halifax N.S., CANADA B3H-4H8 
Cdn/Bitnet: lane@cs.dal.cdn    Arpa: lane%dalcs.uucp@uunet.uu.net
Uucp: lane@dalcs.uucp or {uunet,watmath,utai,garfield}!dalcs!lane