From owner-freebsd-hackers  Thu Sep  4 07:07:33 1997
Return-Path: <owner-freebsd-hackers>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.7/8.8.7) id HAA00134
          for hackers-outgoing; Thu, 4 Sep 1997 07:07:33 -0700 (PDT)
Received: from word.smith.net.au (word.smith.net.au [202.0.75.3])
          by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id HAA00127
          for <freebsd-hackers@FreeBSD.ORG>; Thu, 4 Sep 1997 07:07:27 -0700 (PDT)
Received: from word.smith.net.au (localhost.smith.net.au [127.0.0.1])
	by word.smith.net.au (8.8.7/8.8.5) with ESMTP id XAA01388;
	Thu, 4 Sep 1997 23:36:03 +0930 (CST)
Message-Id: <199709041406.XAA01388@word.smith.net.au>
X-Mailer: exmh version 2.0zeta 7/24/97
To: Brian Campbell <brianc@pobox.com>
cc: freebsd-hackers@FreeBSD.ORG
Subject: Re: 2.2-stable swap usage? 
In-reply-to: Your message of "Thu, 04 Sep 1997 00:08:12 -0400."
             <19970904000812.60761@pobox.com> 
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Thu, 04 Sep 1997 23:36:02 +0930
From: Mike Smith <mike@smith.net.au>
Sender: owner-freebsd-hackers@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

> Mike Smith said:
> > The simple answer is that once swap is allocated to a process, it is 
> > never freed.  You have, in the case above, 24M worth of text which at 
> > some stage has been swapped out, and thus has had swap allocated to it. 
> > It doesn't mean you have 24M worth of swap currently "in use".
> 
> John S. Dyson wrote:
> > Once the pages in MFS or any other process are paged out, those pages will be
> > persistant in swap until the process exits (or the memory is explicitly
> > deallocated by the process.)
> 
> Ok.  So, contrary to what Mike Smith says, pages that belonged to
> a process that has since exited will no longer be marked in-use by
> swap?

I didn't say that; I said once it's been given to a process, it's never 
taken back.  If it's backing memory that the process has freed, it will 
be returned.  Once the process exits, naturally swap will be freed; the 
system would starve without this.

> If they are still marked "in-use", is there a program other than
> pstat that gives a more accurate picture of how many [active] pages
> are in swap?

That'd be horrifically difficult to determine at any given point in 
time.  You could look at the figures given by 'systat -vmstat' and in 
particular the amount of "active" memory, but how you go about deciding 
what is in use by processes as opposed to by the buffer cache is not 
clear.

> > MFS doesn't deallocate any of it's memory usage.
> 
> So, if 90% of MFS is consumed by files which are later unlinked
> (and not in use by any process), is swap thereafter limited to 10%
> of its original size?

That sounds reasonable.

mike