Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 23 Apr 1998 09:19:30 -0700 (PDT)
From:      patl@phoenix.volant.org
To:        Jonathan Lemon <jlemon@americantv.com>
Cc:        freebsd-hackers@FreeBSD.ORG
Subject:   VM architecture (was Re: Protected mode instructions which reduce to noop.)
Message-ID:  <ML-3.3.893348370.3628.patl@asimov>
In-Reply-To: <19980423103837.32719@right.PCS>

next in thread | previous in thread | raw e-mail | index | archive | help
> What VM type architecture were you referring to?

>From the context, I assumed that he was reffering to IBM's
mainframe system.  (Originally VM/370 IIRC.)  It is a smallish
OK kernel that simple provides multiple virtual machines that
run other OSes.  (Or other copies of itself.  Back in the late
70s the recursion depth was around five or six before something
failed.)  The sub-OSes had access to real or virtual disks,
virtual card readers, card punches, and printers, terminals, etc.

The basic mechanism is to run the sub-OSes without kernel privilege
and to trap on Priveleged Instruction exceptions.  The instruction,
registers, and memory would then be examined and the requested
operation translated as necessary and performed.  Or the exception
would be passed back up to the client OS, as appropriate.

There was also a mechanism, using a privileged diagnostic instruction,
for a VM-aware OS to directly request the services of the VM kernel.
CMS was the user interface to VM.  It was an OS that actually required
VM underneath since it didn't attempt to perform any of the functions
that VM could provide for it.  The standard mode of operation had
each on-line user in a separate virtual machine running their own
copy of CMS.  There was also usually one copy of DOS (no, not much
like PC-DOS), OS1, or OS2 (no, not OS/2) running batch jobs in it's
own virtual machine.  OS1 and OS2 could be configured to know that
they were running under VM and delegate some of their operations for
better system-wide efficiency.  (One of the standard techniques was
to tell OS1 that it had a full complement of physical memory and that
it shouldn't page.  This was a big win even when it was the only
thing running on a VM system because VM's paging algorythms were
better than OS1's.)



-Pat

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?ML-3.3.893348370.3628.patl>