RS> In that case it looks pretty certain that for some reason OS2 is
RS> having a problem with that size drive, particularly in the light of a
RS> specific mention of something similar about >2GB drives in the APAR.

PE> I've had a thought.

You want to be very careful of that stuff, can get dangerous.

PE> My 2.5 gig drive is 4969 cylinders, 16 heads and 63 sectors.
PE> The normal thing to do is scale down the number of cylinders
PE> by 4 and increase the number of heads to 64.

Nope, thats only viable BELOW 528MB or 504MB depending on which
MB you are using. Because IDE has a limit of 16 heads. Tho that
is what the LARGE entry in many bios like the Award does, but
in that case the bios itself translates that into the original
numbers for the commands down the cable. AND thats just faking
it up for *DOS*, OS2 doesnt have a problem with cylinders >1024.

PE> That won't work for me, because 4969/4 > 1024.
PE> So maybe I cannot have bios remapping?

As far as normal OS2 is concerned, there is no need for any of that.
Those original CHS figures with 4969 cylinders should work fine.

PE> Another thing is that according to the Linux docs,
PE> LBA is a method of addressing the disk using a
PE> block number INSTEAD of using the physical sectors.

Well, while that is sort of try, there isnt quite
the absolute distinction that you are implying there.

PE> So what is wrong with enabling LBA on all my drives?

Nothing, it should work fine and the big question
is why OS2 freaks on those 2.5GB drives install wise.

PE> How can that EVER cause any harm?

The only potential for harm arises when you have the drive configured
partition wise with LBA turned off, and THEN turn it on. The original
config had a sequence of sectors, essentially thru the sector count
in the CHS setting of 63, then the head increments, then the cylinder.
Now you wont necessarily always get the SAME sector sequence when you
enable LBA and so the formatting at the partition and directory structure
level can be very seriously fucked.

PE> That ISN'T remapping of physical address, that is just ALLOWING
PE> a smart application to use an ADDITIONAL form of addressing?

No. Think about the definition of a partition for example. With CHS,
LBA disabled, the space the partition occupys is essentially defined
as a set of tracks between the lowest cylinder number and the highest,
with all the heads and all the sectors per track in that group.

Now if you enable LBA and say the sectors per track CHANGES, you will
then find that things come comprehensively unstuck on just which complete
collection of sectors are actually within a particular partition etc.

RS> Corse that assumes you really did try it properly just have a
RS> SINGLE 2.5GB hard drive installed, no other drives at all, you
RS> zeroed out the bytes in the first physical sector. You disabled LBA
RS> in the motherboard bios and then tried installing OS2 in a partition
RS> below the 1024 cylinder limit and THAT much simpler config failed.

PE> I never did this.  I may have come close to trying that
PE> when I was first having problems, I cannot remember.

In the situation where there is some real doubt about whether there
is just some significant wart with OS2 and a >2GB drive, you really
do need to go right back to that VERY minimal config, and zero out
all the data in the first physical sector, and see if it will install
from scratch from OS2 booted off floppys.

And you need to do that zeroing of the first physical
sector TWICE, both with a test with LBA enabled, and for
the test with LBA disabled, using those 4969/16/63 numbers.

If you do have a LARGE drive type too, it would be worth trying that
as well, but from memory the AMI Winbios does have that alternative.

AND this test is with just one 2.5GB installed, eliminating that
question mark about physical drives 3 & 4 with that bios as well.
@EOT:

---