Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!usc!samsung!uunet!math.fu-berlin.de!unido!aega84!tmcsys!lothar
From: lothar@tmcsys.UUCP (L. Hirschbiegel)
Newsgroups: comp.unix.sysv386
Subject: Re: Does WINDOWS run under VP/ix?
Message-ID: <388@tmcsys.UUCP>
Date: 1 Jun 91 23:30:15 GMT
References: <1991May17.010926.338@gorgon.uucp> <1991May17.105152.1870@metro.ucc.su.OZ.AU> <314@harem.clydeunix.com>
Reply-To: lothar@tmcsys.UUCP (L. Hirschbiegel)
Organization: Private Site
Lines: 28

In article <314@harem.clydeunix.com> wes@harem.clydeunix.com (Barnacle Wes) writes:
+In article <1991May17.105152.1870@metro.ucc.su.OZ.AU>, glenn@suphys.physics.su.OZ.AU (Glenn Geers) writes:
+> [ ... ]  What I'm saying is that you could run a lot of different
+> protected mode operating systems if you had a suitable supervisor program
+> to manage segment swapping - I'd hate to wait for a 4Gb swap! - and memory
+> fragmentation and compaction would be a pain!
+
+I think you'd have to re-write the VM handler code in the unix kernel
+pretty extensively to do this; all of the REAL operating systems :-)
+running in such a system are going to want control of the Global
+Descriptor Table, and I think the GDT would differ greatly from system
+to system.  The 386 allows you multiple virtual 8086s, but not multiple
+virtual 386s.
+
+	Wes Peters

I'm not really sure about this, but didn't Jolitz et al. describe something
like this for their BSD/386 port? They have mentioned the possiblity of some
kind of "supervisor kernel" above the real unix, handling interrupts, traps,
exceptions etc. out of regular internal unix kernel context.
This would force the kernel to handle itself as some kind of [system] process
and surely increase the number of context switches some orders of magnitude.
Does anybody know whether they finally realised something in this direction?

-- 
------------------------------------------------------------------------
L. Hirschbiegel, AEG - A84, Frankfurt (Germany) / email: unido!aega84!lh