Interestingly enough I was skewered by Charles many times over one
weekend about this very issue.  Maybe I was the only one that was
paying attention.  I had fixed things much in the way that you (and I
at the time) suggested.  There is a better way.  First, I confess that
specifying the cyl/head/sec for wd drives was my doing.  It seemed to
make sense at the time, give me info for sysinst and make the wd look
like a scsi.  But I now understand that that both disks should just
show the capacity.  That said, then what is wrong is sysinst and not
the kernel.  The driver should not recompute.  What is wrong about
sysinst (and I believe this is only for ide) is that it calculates a
disk size by multiplying up the bios CHS sizes.  It should instead use
"disk->geom[4]" which is the disk capacity.  Sysinst then can fudge
from this whatever number of cylinders it likes.  You can use HS from
the disk or HS from bios or ...


Manuel Bouyer <bouyer@antioche.lip6.fr> writes:

> On Sep 30, Phil Nelson wrote
> > 
> > Well, I just got another computer to play with and found something
> > interesting:
> > 
> > System 1:
> >   wd0 at wdc0 drive 0: <IBM-DHEA-38451>
> >   wd0: using 16-sector 16-bit pio transfers, lba mode
> >   wd0 8063MB, 16383 cyl, 16 head, 63 sec, 512 bytes/sect x 16514064 sectors
> > 
> > System 2:
> >   wd0 at wdc0 drive 0: <IBM-DTTA-371010>
> >   wd0: using 16-sector 16-bit pio transfers, lba mode
> >   wd0 9641MB, 16383 cyl, 16 head, 63 sec, 512 bytes/sect x 19746720 sectors
> > 
> > This is the first time I have seen an IDE disk have a return where
> > cyl*head*sec != sectors.  I presume this is something sysinst may
> > need to deal with.  (And possibly other places also.)
> > 
> 
> "Read the ATA-4 draft" :). For >8Gb drives, the drive should not
> report > 16383*16*63 for BIOS compatibility.
> I guess the driver should just recompute the number of cylinders in this
> case ...
> 
> --
> Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
> --