From owner-freebsd-stable@FreeBSD.ORG Fri Jan 25 15:56:50 2013 Return-Path: Delivered-To: freebsd-stable@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:1900:2254:206a::19:1]) by hub.freebsd.org (Postfix) with ESMTP id ECF32496 for ; Fri, 25 Jan 2013 15:56:50 +0000 (UTC) (envelope-from danny@cs.huji.ac.il) Received: from kabab.cs.huji.ac.il (kabab.cs.huji.ac.il [132.65.16.84]) by mx1.freebsd.org (Postfix) with ESMTP id A5E0577A for ; Fri, 25 Jan 2013 15:56:50 +0000 (UTC) Received: from pampa.cs.huji.ac.il ([132.65.80.32]) by kabab.cs.huji.ac.il with esmtp id 1Tyle4-000DG8-NZ; Fri, 25 Jan 2013 17:56:49 +0200 X-Mailer: exmh version 2.7.2 01/07/2005 with nmh-1.3 To: Damien Fleuriot Subject: Re: bge bad performance In-reply-to: <727F9491-A88F-4A4E-BD4F-61235F31500E@my.gd> References: <727F9491-A88F-4A4E-BD4F-61235F31500E@my.gd> Comments: In-reply-to Damien Fleuriot message dated "Fri, 25 Jan 2013 16:02:53 +0100." Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: quoted-printable Date: Fri, 25 Jan 2013 17:56:48 +0200 From: Daniel Braniss Message-ID: Cc: "freebsd-stable@freebsd.org" X-BeenThere: freebsd-stable@freebsd.org X-Mailman-Version: 2.1.14 Precedence: list List-Id: Production branch of FreeBSD source code List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 25 Jan 2013 15:56:51 -0000 > > On 25 Jan 2013, at 11:45, Daniel Braniss wrot= e: >=20 > > Hi, > > It seems that I have more issues with the bge, > > Broadcom NetXtreme Gigabit Ethernet, ASIC rev. 0x5720000 > >=20 > > ifconfig says: > > bge2: flags=3D8843 metric 0 m= tu 15=3D00 > > options=3Dc019b > O,LINKSTATE> > > ether xx... > > inet6 xxx prefixlen 64 scopeid 0x3 > > inet xxx... netmask 0xfffff000 broadcast yyy > > nd6 options=3D21 > > media: Ethernet autoselect (1000baseT ) > > status: active > > > > iperf reports > > =5B 3=5D 0.0-10.0 sec 230 MBytes 193 Mbits/sec > >=20 >=20 > Post your iperf command line and explain the test environment, as in di= rect connection vs switched network link ? >=20 I run iperf -s on the same host for a long time, so not to influence the results. the network is switched, and I try out several other host to be able to a- compare b- detect if the network is too busy. c- to somehow 'normalize' what promped me to check iperf in this case was that tar via NFS started = to=20 give read timeouts. so I run perf -c: with=20 dev.bge.2.forced_collapse: 0 or 1 dev.bge.2.msi: 1 pe-04> iperf -c minbari ------------------------------------------------------------ Client connecting to minbari, TCP port 5001 TCP window size: 257 KByte (default) ------------------------------------------------------------ =5B 3=5D local 132.65.16.35 port 12835 connected with 132.65.16.212 port= 5001 =5B ID=5D Interval Transfer Bandwidth =5B 3=5D 0.0-10.0 sec 238 MBytes 200 Mbits/sec setting: dev.bge.2.forced_collapse:1 dev.bge.2.msi:1 pe-04> iperf -c minbari ------------------------------------------------------------ Client connecting to minbari, TCP port 5001 TCP window size: 257 KByte (default) ------------------------------------------------------------ =5B 3=5D local 132.65.16.35 port 23207 connected with 132.65.16.212 port= 5001 =5B ID=5D Interval Transfer Bandwidth =5B 3=5D 0.0-10.0 sec 506 MBytes 425 Mbits/sec the same command from a similar host, but with bce: Client connecting to minbari, TCP port 5001 TCP window size: 257 KByte (default) ------------------------------------------------------------ =5B 3=5D local 132.65.80.2 port 30664 connected with 132.65.16.212 port = 5001 =5B ID=5D Interval Transfer Bandwidth =5B 3=5D 0.0-10.0 sec 976 MBytes 818 Mbits/sec In case it's not obvious, all hosts are server class, the network shows n= o=20 errors, neither the hosts. cheers danny