Subject: Re: reworked idea for reducing BIOS headache.
To: None <jnemeth@cue.bc.ca, phil@steelhead.cs.wwu.edu, toddpw@best.com>
From: Ross Harvey <ross@teraflop.com>
List: tech-install
Date: 10/03/1998 15:05:50
> From: John Nemeth <jnemeth@cue.bc.ca>
> On Sep 29,  9:06pm, Phil Nelson wrote:
> } 
> } >Okay, so set the first partition in the MBR to cover the whole disk minus
> } >the first track, type it to NetBSD, and put a disklabel in that partition.
> } 
> } This is what sysinst does if the whole disk is selected.  The problem is
> } knowing that is the first track in relation to the BIOS geometry for
> } that disk.  This is the real problem that needs to be addressed.  
> } 
> } I may generate a sysinst that will skip only the first sector and try
> } that road again.  That would eliminate the need to worry with getting
>
>      Why stop there?  Don't skip any sectors and start NetBSD in the
> first sector on the disk.  If NetBSD is going to use the whole disk
> then you don't need an MS-DOS style partition table.  This has always
> worked fine for me.

But it hasn't always worked fine for everyone else. I think that some BIOS
versions need to recognize something and will not simply execute the first
sector.

And BTW, we've said this before, several times. :-)

> } big enough to cover /dev/?d?a, it would work.  (Assuming it is true
> } that BIOSes don't really complain at booting something from the first
> } track.)
>
>      The BIOS doesn't care.  It simply loads and executes the code in
> the first sector.  It's up to that code to figure what to do next, and
> if it complains about booting something in the first track, you can
> simply replace it with generic code.

Most BIOS programs don't care, right. Sadly, some of them seem to. They don't
all `simply load and execute code in the first sector'.

I usually install NetBSD this way, by just giving it the whole drive, and
it usually works, sure. But I've seen a sector-0-boot drive that works with
one BIOS get the (IIRC) `no operating system' message when the same drive
is installed on another PC. Note here that geometry issues can't explain
this failure, as sector 0 maps to sector 0 in all geometries. These were
both Pentium-II system boards with bios programs from the same era, but
from different companies. I wouldn't be suprised if this was only one
particular BIOS vendor, and perhaps we could just quirk it.

This hasn't just been observed by NetBSD'ers either. FreeBSD used to have
a handbook section discussing what they called `dangerously dedicated'
drives. I think it discussed this general issue with similar assertions.

  --Ross Harvey