> Out of curiousity, does anyone know why the physical disk geometry is
> not being used here?  I've noted BSD has this geometry from way back.
> I do understand that not all blocks are available due to MSCP and the
> RQDX3 label info etc. but I had hoped that unlike modern SCSI disks where
> I wonder how much the smart ffs fights the smart drive, at least MSCP
> is almost as dumb as the SMD stuff that ffs is designed for and might
> actually do better if the geometry matched.

17*8*1024 = 139264, available: 138672, missing: 592

Assuming that partitions end on cylinder boundaries, using the 
physical disk geometry would leave some sectors unused:

17*8*1019 = 138584, unused: 88

So specifying the physical paramters for sectors/tracks/cylinders
might result in a better filesystem layout. The tradeoff will be
losing 88 blocks ("nn blocks unallocated in last cylinder ...");

> From the Ultrix 4.2A /etc/disktab file:
> 
>  rd53|RD53|DEC RD53 Winchester:\
> 	:ty=winchester:ns#17:nt#8:nc#1024:\
> 	:pa#32768:ba#8192:fa#1024:\
> 	:pb#50160:bb#4096:fb#1024:\
> 	:pc#138672:bc#8192:fc#1024:\    
            ^^^^^^
                  same total size as above

I would not be surprised if the only way to address a given sector
using MSCP is by giving the logical block-number. Anything else is
done by the controller.

I really don't know, but to me it looks like somebody factored the
number of available sectors and used one of the possible triples as
the new disk-geometry ...

> | sectors/track: 18
> | tracks/cylinder: 8
> | cylinders: 963

Ciao,
	bertram