Greetings,

It would hang on anything that hit the partition table.
Even booting a 3.5" bootable dos floppy, as soon
as the partition table was read, it would hang the machine.
CTRL-ALT-DEL works, as well as numlock responds.
I let it sit for about 10 minutes to see if maybe something
would timeout.  As soon as the HDD LED would blink,
the machine would just freeze.  For NetBSD, it wouldn't
even get as far as the NetBSD boot loader text.  Right
at the Award "flower box"  hardware list after all posts...
the HDD light would blink looking for the active partition 
and "freezerama"

The SCSI settings- although I tried every combination
imaginable- from enable/disable int13 extensions,
to >8 GB for DOS support enable/disable..  nothing fixed 
this freeze.  I tried tons of other settings too, such as
disable sync, initiate wide negot. and several others.
For all testing, I took everything off the bus and tried with 
just this singular drive too.  Nothing.

In fact, the only way I could get the machine back
after each test  (since -nothing- bootable; floppy or 
cdrom or anything could get past this)  was to low-level 
the drive in the controller's own format routine in the BIOS itself.

I am aware of the MBR bug as well as the bootloader
bug.  I tried Megabytes, Cylinders and Sectors for 
partitioning, as well as 1.4.1 and Current (04/16 snapshot)
images to prep the drive- same results.  Anything done with 
a NetBSD disklabel would make any drive prepped
with this BIOS a "hang on boot" drive.  I forgot to
mention I tried two additional drives this morning
right before assuming it wasn't geometry any more- an 
8GB and an 18GB.  All IBMS UW, and exact same model 
drives that are working perfectly in two other 1.4.X 
systems with 2940UW's.

Simply flashing the controller off  the dreaded 2.20 bios 
backwards to a 1.X SCSI BIOS fixed it.  I actually tried to
get it to break with the 1.X BIOS- by changing the int13, >8
GIG, and trying wacked BIOS Geometries.  (1023 cyl, 555
cyl, and several other geometries that had no business working).
They all worked and booted perfectly.  No matter what
tomfoolery I could dream up on the drive with the 1.X 
SCSI BIOS, it prepped and booted perfectly.

Adaptec has this 2.20 bios available on their site- as well
as the 1.3X as well.  It's quite easy to reproduce.  =)

The strangest thing is the 2.20 Adaptec SCSI bios has no 
issues with fdisk/format, nor with format/boot of NT, Win2K 
or several other OS installs I tried.

So, it's definately a real hair-pulling "gotcha" for anyone
out there that might be playing with a 2940UW + 2.20
BIOS.  =)

Cheers,
Steve

----- Original Message ----- 
From: "Frank van der Linden" <frank@wins.uva.nl>
To: "Steve Paul" <stevep@mccue.com>
Cc: <port-i386@netbsd.org>
Sent: Tuesday, May 09, 2000 1:26 AM
Subject: Re: SOLVED- 4.5GB SCSI Disk/geometry woes on 1.4.2


> On Mon, May 08, 2000 at 03:37:52PM -0700, Steve Paul wrote:
> > The solution- the 2940UW BIOS revision.  2940UW's
> > that have the 2.20 revision BIOS will not work.  This revision 
> > works fine for everything- NT, Windows, DOS.. You can FDISK
> > and FORMAT and boot fine- but upon running NetBSD
> > the machine will hang on boot. My other two working NetBSD
> > boxes had pre 2.00 BIOS's.  I'd still be pulling my hair out
> > if I didn't notice the BIOS numbers for that 3-4 seconds 
> > during the Adaptec <CTRL-A> screen.
> 
> At what point during the boot did it hang? There was a problem with
> the bootloader that was recently fixed. Also, did you have an option
> called "BIOS int13 extensions" or "support for > 8G drives" enabled?
> 
> - Frank
>