From owner-freebsd-current  Tue Aug 27 10:44:00 1996
Return-Path: owner-current
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.5/8.7.3) id KAA17213
          for current-outgoing; Tue, 27 Aug 1996 10:44:00 -0700 (PDT)
Received: from dyson.iquest.net (dyson.iquest.net [198.70.144.127])
          by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id KAA17189
          for <current@FreeBSD.org>; Tue, 27 Aug 1996 10:43:50 -0700 (PDT)
Received: (from root@localhost) by dyson.iquest.net (8.7.5/8.6.9) id MAA00794; Tue, 27 Aug 1996 12:34:40 -0500 (EST)
From: "John S. Dyson" <toor@dyson.iquest.net>
Message-Id: <199608271734.MAA00794@dyson.iquest.net>
Subject: Re: FreeBSD malloc.c, -lmalloc, and squid.
To: terry@lambert.org (Terry Lambert)
Date: Tue, 27 Aug 1996 12:34:40 -0500 (EST)
Cc: stesin@gu.kiev.ua, angio@aros.net, squid-users@nlanr.net,
        current@FreeBSD.org
In-Reply-To: <199608271653.JAA24972@phaeton.artisoft.com> from "Terry Lambert" at Aug 27, 96 09:53:42 am
Reply-To: dyson@FreeBSD.org
X-Mailer: ELM [version 2.4 PL24 ME8]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-current@FreeBSD.org
X-Loop: FreeBSD.org
Precedence: bulk

> 
> Secondly, the sbrk()-based system recovery of memory assumes too
> much allocation locality: the recovery of memory (via sbrk() back
> to the system) can only occur at the end of the allocation region.
> 
Actually, we have an madvise(2) in current that supports "freeing" pages, but
keeping the VM space.  That doesn't do EXACTLY what you say above, but
it can help.  I have been running with an augmented version of
phkmalloc with that change for a long time now.

>
> I would like to see an extention to the mmap() semantics for sparse
> reclaimation on a per-page basis:  When a full page is no longer on
> use, it would be returned individually to the system.
> 
You can do an munmap that does not correspond directly to an mmap.  Also,
as I said above, we have an madvise(2) call (arg:MADV_FREE) that releases
the page, without giving up the vm space.  Specifically, the page is
marked clean, and might/might-not be demand-zeroed the next time that you
use it.


John