Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!cs.utexas.edu!uunet!convex!rigel!harper
From: harper@rigel.uucp (David Harper)
Newsgroups: comp.sys.ibm.pc
Subject: Re: IBM-AT BIOS ROMS
Message-ID: <1817@convex.UUCP>
Date: 20 Sep 89 17:45:24 GMT
References: <1989Sep19.021545.8110@NCoast.ORG>
Sender: usenet@convex.UUCP
Reply-To: harper@rigel.UUCP (David Harper)
Organization: Convex Computer Corporation, Richardson, Tx.
Lines: 52

In article <1989Sep19.021545.8110@NCoast.ORG> dchou@NCoast.ORG (David Chou) writes:
>I have recently tried to update a hard disk in an IBM-AT with a
>Miniscribe 3085 and have found that the geometry for this disk is not
>in the BIOS tables (only 14 are in the tables).  I prefer not to use
>a TSR such as Disk Manager.  Fortunately (or unfortunately) I have a
>copy of the IBM AT Tech Reference Manual and have located the disk
>geometry tables.  It is therefore possible for me to copy my BIOS ROMS
>into 27128 EPROMS and alter one of the existing disk types.  Unfortunately,
>I suspect that the BIOS does a checksum on the BIOS ROMS, but cannot
>find where it is being done or where the checksum is being performed.
>Is there anyone out there who has any information on how this is being
>done?  I have already found the code where the system performs checksums
>on auxiliary ROMS, but this does not seem applicable for the BIOS ROMS.
>Incidentally, the system is one of the original 6MHz ATs with 128K
>stacked DRAMS.
>
>Thanks in advance.
>
>David Chou     ncoast!dchou@hal.cwru.edu

I have been faced with similar problems in the past.  On one BIOS which came  
out of an American Research Corp PC/AT I was able to track down the routine
which does the checksum and modify it so that it would never fail.  I think
this involved changing a 'jump equal' instruction to a 'jump always' instruction.
Later, I wanted to perform the same trick on a Phoenix BIOS but never was able
to locate the checksum routine.  Also, I was unable to find the checksum itself;
I had throught that I might simply be able to change that to the new value.  
Eventually, I sold the drive I was trying to integrate so the problem went away.  
About two days after that ( 8-( ) I thought of a way that should work, but I 
must stress that it has not been tried yet.  Briefly, the steps to be performed 
are as follows:
     1) Determine the checksum of each EPROM.  This can generally be
	done by simply reading the chip in an EPROM programmer.  Usually
	they return the checksum read in when the operation completes.
     2) Find the table entry that you want to use and modify it for the
	correct parameters.  I can't remember the table format off the
	top of my head (I'm at work and my copy of the AT Tech Ref manual
	is at home) but the book clearly shows what goes where.
     3) Pick one or more table entries that you think you will never use and
	adjust their values so that the net result is the same checksum that
	you originally started with.  This must be done with each EPROM
	individually.
I don't know if all EPROM checksums work the same but the one that I defeated
worked by adding up the individual byte values from each successive location
into a 16 bit value, starting over at zero again on overflow.  If this is the
case with your EPROMs the above should work.  As I say, I never had a chance 
to try this approach and I would be interested to hear if it works.


Dave Harper    -     Convex Computer Corp.            E-mail address:
3000 Waterview Pky.  Richardson, TX 75081             harper@convex.COM
(214) 497-4525 (W)   (214) 727-4206 (H)