🚀 go-pugleaf

In article <4h7cfg$84q@pellew.ntu.edu.au>,
	radams@banda.ntu.edu.au (Rob Adams) writes:
>I am trying to install Linux on my 386, I have a Future Domain TMC950 card
>with its BIOS chip removed, I have been able to get the drive recognised 
>by using the tmc8xx=0xCA000,5 commandline.

Good luck with it, but my advise is to exchange this with any other
available SCSI card.

The Linux Seagate/FutureDomain 8xx/9xx driver barely works. It has
some serious problems, some of which I was - while I was using it -
able to kludge around by hard-coding the geometries, but basically
it is not good enough for any "production" use. To be honest, I don't
know why it is still part of the Linux kernel. It hasn't been touched
in quite some time and still needs a *lot* of work.

After some frustration with it, even with kludges (and more kludges)
applied, I gave it up and swapped to an old Adaptec 1452B, which has
worked 100% ever since.


>My problem is in the setting up of partitions, the drive is a Seagate
>ST31200N (1.2Gig) SCSI drive. It has C2700/H9/S84 according to Seagate
>AND the Kernal startup, but fdisk wants to argue that I cant have more
>than 64 (or is it 63) sectors...

Yep, you have geometry problems, same as I did.

If you have no other choice and you need to kludge around like I did
for a while (and it did actually work, and was able to get lilo to
work and therefore make the drive bootable), then edit seagate.c
in /usr/src/linux/drivers/scsi, being careful to examine the function
where the drive geometry is determined. There is a problem in the way
the driver interrogates the card, so what I did to work around it
was to test the unit/lun and hard code in the geometries of the drive.
This pretty much fixed all of the fdisk/lilo related problems.

You'll see from the comments in that function (I think it was
seagate_st0x_biosparam()) which indicate that further work is
required to get BIOS parameter translation working correctly.

Other aspects of the driver are also not particularly good. I kept
on having problems with "shadow" drives appearing during boot which
did not exist, and strange errors returning from drives that were
there which caused them not to be correctly identified and mounted.
The straw that broke the camel's back was the inability to get the
tape driver working on the card (yet strangely enough, it worked
just fine when I used an 8-bit TMC855, but not at all on an the 950).

-- 
David Nugent: davidn@blaze.net.au, davidn@unique.blaze.net.au
Unix/Linux/DOS/Win/Os2 Microsystem Technical Support [FOR HIRE!]
http://www.blaze.net.au/~davidn