On Mon, 6 Mar 2000, Frank van der Linden wrote:

> Date: Mon, 6 Mar 2000 16:40:20 +0100
> From: Frank van der Linden <frank@wins.uva.nl>
> To: George Sollish <gsollish@mail.gisco.net>
> Cc: port-i386@netbsd.org
> Subject: Re: bootselector error 3
> 
> On Mon, Mar 06, 2000 at 10:19:30AM -0500, George Sollish wrote:
> > I am building a multiboot i586-166 box with a new WD 13.6G drive.  I
> > loaded 1.4.1 into the top 6G first, with a 50M partition for DR-DOS on the
> > bottom loaded next.  At this point the 1.4.1 bootselector could boot
> > either OS from the hard drive.  I then loaded FreeBSD 3.2 into the first
> > of two 3.5G partitions.  Now the 1.4.1 bootselector returns error 3 on
> > default. Before the default (set to NetBSD) takes over I can F1 to DR-DOS,
> > but any other F-key returns error 3.
> > 
> > I can boot either BSD from floppy.  I did notice that the BIOS geometry
> > was different on the FreeBSD side, and reset it with their fdisk -i.
> > Afterwards I reinstalled the NetBSD bootselector and power-cycled just to
> > be sure.
> 
> What seems to have happened here is that the FreeBSD install, presumably
> because it used a different geometry somehow, changed the partition
> boundaries. What could have happened is that there is now a mismatch
> between the 2 ways that the partition boundaries are stored in the MBR:
> by their CHS and logical sector numbers.
> 
> "Error 3" means that the bootselector loaded the first sector of the
> partition, but did not find a valid bootsector signature.
> 
> You can study the output of "fdisk wd0" on NetBSD, and see if the
> CHS values match the logical sector values. If not, you'll have to
> bring them back in sync (I assume that you know what the correct
> values are..)
> 
> I don't know why FreeBSD had different ideas about the geometry; I
> have installed 4 OSs (amongst which were FreeBSD and NetBSD) on
> one 12.7G disk, without problems.
> 
> - Frank

After chasing my tail for several hours I reloaded 1.4.1 and our
bootselector now works again on all three installed OS's.  FreeBSD still
complains about the NetBSD slice ending outside their idea of its
endpoint, but this is the extreme upper end of the drive and neither OS
sees all of it anyway if its really a 13.6G (12971M is the hardware size
both OS's return).  Unless you see problems ahead, I'll continue with it
as it stands.

Thanks.

George E Sollish	Chief Engineer  	 Auto Gear Equipment
			Host	 		 Classic-FM's Listening Room
			Project Manager		 The Payne Lake Project