Subject: Re: 1.3.3 i386 install crash and burn
To: Ken Harrenstien <klh@us.oracle.com>
From: Manuel Bouyer <bouyer@antioche.lip6.fr>
List: netbsd-bugs
Date: 03/03/1999 19:49:41
On Mar 3, Ken Harrenstien wrote
> Shouldn't it be possible to get it from the drive itself?  This is
> SCSI, after all.  The install did give me a hard time about the geometry
> and I had to change the head/sector factors to get the #-sector count
> under 64.  Then later it asked me whether to use real or bogus
> geometries, with no clue as to what the implications of either might
> be; the INSTALL document doesn't mention this question at all.
> So yes, it's possible the geometry was "incorrect" in some sense.

Well, no. For the MBR (which holds the boot program, so it has to be
understood by the BIOS), whe have to use a geometry compatible with the
bios's limitations, and most goemotries reported by SCSI disks don't work
(the number of cylinders it too big, this is also why most IDE drives reports
geometries with 16 tracks, 63 sectors). The BIOS of the SCSI board will do
the translation, but usually there's no other way to find which geometry they
uses for the BIOS (the newest Asus BIOS prints the geometry they use. Old
one don't and I don't think adaptec do it either) than ask the bios itself.
NetBSD can't ask the bios yet, but there are some work being done for this.


--
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
--