From owner-freebsd-stable Tue Apr 29 16:07:02 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.5/8.8.5) id QAA16481 for stable-outgoing; Tue, 29 Apr 1997 16:07:02 -0700 (PDT) Received: from dyson.iquest.net (dyson.iquest.net [198.70.144.127]) by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id QAA16469 for ; Tue, 29 Apr 1997 16:06:52 -0700 (PDT) Received: (from root@localhost) by dyson.iquest.net (8.8.4/8.6.9) id SAA07175; Tue, 29 Apr 1997 18:05:13 -0500 (EST) From: "John S. Dyson" Message-Id: <199704292305.SAA07175@dyson.iquest.net> Subject: Re: Swap leaks (was Re: Memory usage on NFS server) In-Reply-To: <199704292201.IAA11983@gsms01.alcatel.com.au> from Peter Jeremy at "Apr 30, 97 08:01:22 am" To: peter.jeremy@alcatel.com.au (Peter Jeremy) Date: Tue, 29 Apr 1997 18:05:03 -0500 (EST) Cc: freebsd-stable@freefall.FreeBSD.ORG X-Mailer: ELM [version 2.4ME+ PL31 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-stable@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk > On Tue, 29 Apr 1997 10:03:57 +0200 (SAT), John Hay > wrote: > > Inn then typically show a usage of > >~70M according to top, but as soon as I kill and restart it the swap > >usage go down to ~5M. > > Some of the BSD/386-derived Unices have a memory leak in the VM > subsystem which matches the above behaviour: When a process fork()d > whilst the system was paging, swap space allocated to the child was > not released until the _parent_ exit()d. This primarily bit long- > running daemons that fork() regularly: cron, inetd, innd and sendmail. > > I thought FreeBSD 2.1 and 2.2 included fixes for this VM problem. > Can someone please confirm this. (In which case John's problem is > something else). > The swap-leak problem is fixed as completely as resonable. Once swap space is allocated for a given object in a process, it stays around until the object is destroyed. Swap space is not recovered when pages are paged back in. That is a performance design decision. John