Subject: Re: [gwr@mc.com: Re: kern/1991: device number defined inconsistently on sparc/Ultrix/OSF1 Alpha]
To: Taras Ivanenko <ivanenko@ctpa03.mit.edu>
From: None <Chris_G_Demetriou@NIAGARA.NECTAR.CS.CMU.EDU>
List: netbsd-bugs
Date: 02/05/1996 23:42:09
> My impression is that OSF1 converts the numbers into its own format
> when doing some operations. I have no idea why it is doing that but
> the effect on NetBSD was disastrous. I can not change much in OSF1
> kernel, it is probably DEC's proprietary code but I can change the way
> NetBSD sees the devices. For the reference, I have 

This impression is correct.  It's one of the most annoying things i've
run into with OSF/1, and it seems to be ... one hell of a bug (in
OSF/1).

Actually, as far as i know, this applies to "Digital UNIX" as well as
earlier versions of "DEC OSF/1", and i use the names interchangeably.


At some point, the device node format was changed in OSF/1 so that it
was 12/20 (from 8/8).  At that time, some compatibility code was put
in the kernel to convert from "old format" to "new format" when old
device nodes were read in to core, and back when they're written back
to disk.

The lose is that the converted ("Fixed!") device node information is
what's exported by the NFS server.

I had two solutions to this:
	(1) not use OSF/1 as my NFS server for my root partitions.
	(2) hack OSF/1 so that it didn't do the annoying translation.
oh yeah, my other solution was to get disk booting working on the
machines i was using...  8-)

I can probably provide a binary patch to OSF/1 (i.e. tell you what
bits to change in one of your kernel build's .o's) to disable the
conversion.  as long as you have OK device nodes on your OSF/1 system
(which you can get by just running MAKEDEV), it's safe to use.


chris