From owner-freebsd-current@FreeBSD.ORG Sat Jul 5 06:25:08 2003 Return-Path: Delivered-To: freebsd-current@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 9238C37B401; Sat, 5 Jul 2003 06:25:08 -0700 (PDT) Received: from freebie.xs4all.nl (freebie.xs4all.nl [213.84.32.253]) by mx1.FreeBSD.org (Postfix) with ESMTP id 396F543FEC; Sat, 5 Jul 2003 06:25:07 -0700 (PDT) (envelope-from wkb@freebie.xs4all.nl) Received: from freebie.xs4all.nl (localhost [127.0.0.1]) by freebie.xs4all.nl (8.12.9/8.12.9) with ESMTP id h65DP6YK011353; Sat, 5 Jul 2003 15:25:06 +0200 (CEST) (envelope-from wkb@freebie.xs4all.nl) Received: (from wkb@localhost) by freebie.xs4all.nl (8.12.9/8.12.9/Submit) id h65DP5oJ011352; Sat, 5 Jul 2003 15:25:05 +0200 (CEST) Date: Sat, 5 Jul 2003 15:25:05 +0200 From: Wilko Bulte To: Maxime Henrion Message-ID: <20030705132505.GA11334@freebie.xs4all.nl> References: <20030704001335.GM42121@elvis.mu.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20030704001335.GM42121@elvis.mu.org> User-Agent: Mutt/1.4i X-OS: FreeBSD 4.8-STABLE X-PGP: finger wilko@freebsd.org cc: current@freebsd.org Subject: Re: Call for testers: patch to convert dc(4) to busdma X-BeenThere: freebsd-current@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list List-Id: Discussions about the use of FreeBSD-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sat, 05 Jul 2003 13:25:08 -0000 Could it be there is something fishy with mbuf handling? I ran netperf from my DS10 Alpha box towards a 4-stable netperf-receiver (called freebie) As in: ds10#ls netperf tcp_range_script udp_rr_script netserver tcp_rr_script udp_stream_script snapshot_script tcp_stream_script ds10#./tcp_range_script freebie ------------------------------------ Testing with the following command line: /usr/local/netperf/netperf -l 60 -H freebie -t TCP_STREAM -i 10,2 -I 99,3 -- -m 1 -s 32768 -S 32768 TCP STREAM TEST to freebie : +/-1.5% @ 99% conf. : histogram After some time one gets: Out of mbuf address space! Consider increasing NMBCLUSTERS All mbufs or mbuf clusters exhausted, please see tuning(7). Out of mbuf address space! Consider increasing NMBCLUSTERS All mbufs or mbuf clusters exhausted, please see tuning(7). Initially things were like: ds10#netstat -m mbuf usage: GEN cache: 0/0 (in use/in pool) CPU #0 cache: 2195/2240 (in use/in pool) Total: 2195/2240 (in use/in pool) Mbuf cache high watermark: 512 Maximum possible: 18176 Allocated mbuf types: 2195 mbufs allocated to data 12% of mbuf map consumed mbuf cluster usage: GEN cache: 0/0 (in use/in pool) CPU #0 cache: 2189/2192 (in use/in pool) Total: 2189/2192 (in use/in pool) Cluster cache high watermark: 128 Maximum possible: 9088 24% of cluster map consumed 4944 KBytes of wired memory reserved (99% in use) 0 requests for memory denied 0 requests for memory delayed 0 calls to protocol drain routines and after the machine ran out of buffers it displayed: ds10#netstat -m mbuf usage: GEN cache: 0/0 (in use/in pool) CPU #0 cache: 9061/9088 (in use/in pool) Total: 9061/9088 (in use/in pool) Mbuf cache high watermark: 512 Maximum possible: 18176 Allocated mbuf types: 9061 mbufs allocated to data 50% of mbuf map consumed mbuf cluster usage: GEN cache: 0/0 (in use/in pool) CPU #0 cache: 9058/9072 (in use/in pool) Total: 9058/9072 (in use/in pool) Cluster cache high watermark: 128 Maximum possible: 9088 99% of cluster map consumed 20416 KBytes of wired memory reserved (99% in use) 0 requests for memory denied 0 requests for memory delayed 0 calls to protocol drain routines ds10# During the test run there was a steady increase in the % for the cluster map. Wilko -- | / o / /_ _ wilko@FreeBSD.org |/|/ / / /( (_) Bulte