install problems for 386/ix on PS/2

A. Lester Buck buck at siswat.UUCP
Tue Mar 27 15:18:22 AEST 1990

I have had some problems installing 386/ix on a PS/2 Model 80 with 70 MB
ESDI disk.

IBM documentation lists this disk as having 36 sectors/track, 7 heads, and
583 cylinders.  After a complete low level format and defect scan, using the
advanced diagnostics on the reference diskette (control-A mode), the disk is
reported to have 36 sectors/track, 7 heads, and 569 cylinders with zero
defects found.  (I don't know what the extra 14 cylinders are reserved for.)

The 386/ix boot disk reports the disk configuration as 32 sectors/track,
64 heads, and 70 cylinders.  The installation disk is able to change ONLY
the cylinders parameter.  Clearly some type of mapping is occurring inside
386/ix.  Since the actual number of sectors (36*7*583 = 146916) is less
than the remapped number (32*64*70 = 143360), things should work out if
the mapping is consistent.  But the complete bad sector scan in 386/ix
always bombs out at the end of the disk, reporting the disk as unusable.

Apparently, the disk is treated as really having only 32 sectors/track,
and the actual number of sectors (7*583 = 4081) is less than the remapped
number of sectors (64*70 = 4480).  I successfully installed by backing
down the number of cylinders from 70 to 63 (64*63 = 4032 < 4081).

What is the point of this remapping scheme?  The net effect is to discard
(36-32)/36 = 11% of usable disk space.  Why can't 386/ix read the disk
geometry from the ESDI interface, like any other ESDI interface?  And if
386/ix is going to remap the disk geometry, how about doing the calculation

A. Lester Buck     buck at  ...!texbell!moray!siswat!buck

More information about the Comp.unix.i386 mailing list