From owner-freebsd-net@FreeBSD.ORG Wed May 26 13:34:40 2004 Return-Path: Delivered-To: freebsd-net@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id A680316A4CE; Wed, 26 May 2004 13:34:40 -0700 (PDT) Received: from freefall.freebsd.org (freefall.freebsd.org [216.136.204.21]) by mx1.FreeBSD.org (Postfix) with ESMTP id 9E8C643D1D; Wed, 26 May 2004 13:34:40 -0700 (PDT) (envelope-from bmilekic@FreeBSD.org) Received: from freefall.freebsd.org (bmilekic@localhost [127.0.0.1]) i4QKYVRS018372; Wed, 26 May 2004 13:34:31 -0700 (PDT) (envelope-from bmilekic@freefall.freebsd.org) Received: (from bmilekic@localhost) by freefall.freebsd.org (8.12.11/8.12.11/Submit) id i4QKYVTR018371; Wed, 26 May 2004 13:34:31 -0700 (PDT) (envelope-from bmilekic) Date: Wed, 26 May 2004 13:34:31 -0700 From: Bosko Milekic To: freebsd-net@freebsd.org Message-ID: <20040526203431.GA17806@freefall.freebsd.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.4.1i cc: freebsd-hackers@freebsd.org Subject: Network buffer allocations: mbuma, PLEASE TEST X-BeenThere: freebsd-net@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list List-Id: Networking and TCP/IP with FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 26 May 2004 20:34:40 -0000 Hi, If you're running -CURRENT, please test this: http://people.freebsd.org/~bmilekic/code/mbuma2.diff It is several extensions to UMA and mbuf & cluster allocation built on top of it. Once you apply the patch from src/, you need to rebuild and reinstall src/usr.bin/netstat, src/usr.bin/systat, and then a new kernel. When you're configuring your new kernel, you should remove the NMBCLUSTERS compile-time option, it's no longer needed. Clusters will still be capped off according to maxusers (which is auto-tuned itself). Alternately, if you want theoretically unlimited number of clusters, you can tune the boot-time kern.ipc.nmbclusters tunable to zero. Unless final issues arise I'm going to commit this tomorrow morning; it's been tested already quite a bit, and performance considered. A paper is available and was presented at BSDCan 2004; in case you missed it: http://www.unixdaemons.com/~bmilekic/netbuf_bmilekic.pdf It has been looked at for quite some time now. Additional code cleanups will need to occur following commit, maybe. Future work is also possible, see the paper if you're interested in taking some of it on. Oh, and keep me in the CC; I have no idea if I'm subscribed to these lists anymore. You should also follow up to this thread on -net and not on -hackers (trim -hackers from CC in the future). Thanks and happy hacking! Regards, -- Bosko Milekic bmilekic@technokratis.com bmilekic@FreeBSD.org