On Mon, Oct 11, 1999 at 10:27:30AM +0200, Hauke Fath wrote:
> At 15:24 10.10.99 +0000, jiho wrote:
> >I want to change the BIOS geometry because it got fouled up somehow.  The
> >installation got it right the first time, at 255 heads.  Now it's 15 heads.
> >Yes, I ran the DOS fdisk, but I've run that without incident on this drive,
> >since installing NetBSD and its boot menu.  (That's the BIOS geometry, what
> >would DOS do to it?)
> >
> >I tried to install Linux.  The Linux fdisk complained about a "corrupt"
> >partition table.  The only choices were to wipe the whole thing out, or 
> >abort.
> >  Obviously I aborted.  Lovely.

The rule of thumb on sysinst is: don't try to outsmart it. It will get
the BIOS geometry right, except in some very rare cases. Don't mess
with it the values, it'll lead to bad results. If the BIOS geometry
values sysinst sees seem bad to you, first check the BIOS setup
if they are reasonable in there. Only modify them in sysinst as
a last resort, if they seem really screwed up. This should only
be needed in about 1% of all cases.

The problem in the above (Jim's) case is a bug that was present in 1.4.1,
where partitions could end up with a off by one error (off by one sector
for the end of the partition, to be exact)..


> an AHA2940UW. Sysinst offered the "BIOS geometry" and the "geometry" that 
> the drive announced during boot. The former didn't get me anywhere, while 
> the latter worked; but I found that sysinst had messed up the MBR partition 
> table by inserting the cylinder numbers from the "drive geometry". Luckily 
> I had scribbled down the original values (I had set up a small DOS 
> partition and set up the MBR partition tabl ewith pfdisk), so I was able to 
> manually fix the cylinder values.

See "don't mess with the values" above. Why didn't the BIOS geometry
"get you anywhere"?

- Frank