2.6 kernel

[Alan Cox]

On Tue, Dec 16, 2003 at 01:28:09PM -0500, Gene C. wrote:
> To me it seemed to matter since it was not using the LBA values but the 
> "actual" (real) values of the device.

Real ? which real 8)

With old disks doing C/H/S its all fairly easy, because there is a real
geometry. Once you get drives too big for CHS (about 528Mb if I remember rightly)
then you switch to LBA, LBA28, LBA48. In those modes its more like scsi, what
you are really talking about is block offset from zero. The entire CHS concept
is so much irrelevance wiht modern drives that its not too big an issue how you
chop it up. You also have to allow for the fun corner cases

	-	Drive strapped to report below real size to bios
		(and possibly partitioned otherwise)
	-	Partition table and other offsets in cylinders

Thats where it gets really nasty. If the partition table laid down by windows 
says x sectors per track and we use a different geometry then all track based
calculations between Linux and Windows will disagree. End result-> not pretty.