Subject: Re: Installing use current snapshot RAMDISK kernel
To: Sageev George <sageev@bmsrs.usc.edu>
From: Gregory McGarry <g.mcgarry@ieee.org>
List: port-hp300
Date: 07/12/2002 16:16:18
Sageev George wrote:

> First, let me say that by following your instructions
> to replace the netbsd install kernel with a generic
> kernel on exported filesystem, I was able to get the
> -current snapshot (4/15/2002) up and running (both on a 
> local drive, and a diskless system).  

Excellent!  The snapshot you are installing is actually
older than the 1.6_BETAx releases available at

ftp://releng.netbsd.org/pub/NetBSD-daily/

You might prefer the beta release since there are many
problems that have been fixed.
 
> Because I felt that the network booting was straight 
> in my mind, _and_ in response to your request, I did 
> try an install using the sysinst in the RAMDISK kernel.

Thanks for testing it out.

> Although the netbsd.ramdisk (from the 4/15/2002 snapshot) 
> did load over the network, there was an error at the end, 
> when the installation program usually starts.  I had seen 
> this behavior before, when I tried using "dd" to dump the 
> miniroot filesystem to my swap partition.  As happened here,
> even though the kernel loaded, I got a lot of "illegal 
> instruction" errors, either in the upgrade program, or 
> just running binaries that were located in the filesystem.
> 
> Am I doing something wrong here?  Console output is 
> below.

No, you're not doing anything wrong.  This is a known problem
with the snapshot that I didn't clarify in the README file
available with the snapshot.  I have now updated that file.

The problem is that the install kernels (such as the one in
the ramdisk kernel) didn't include support for floating point
interface on 68040 cpus.  It has been fixed in the 1.6_BETA
snapshots.

I apologise for this.  I hope it didn't cause you to waste
time on it.  Also, thanks for persisting with testing the
installation media.  Unfortunately the 1.6 release has
had very little testing, so your persistence is valuable.
 
	-- Gregory McGarry <g.mcgarry@ieee.org>