From owner-freebsd-net@FreeBSD.ORG Thu Oct 4 21:38:20 2007 Return-Path: Delivered-To: freebsd-net@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 8DCFE16A421 for ; Thu, 4 Oct 2007 21:38:20 +0000 (UTC) (envelope-from wawa@yandex-team.ru) Received: from relanium.yandex.ru (relanium.yandex.ru [213.180.193.88]) by mx1.freebsd.org (Postfix) with ESMTP id 0DD0913C46E for ; Thu, 4 Oct 2007 21:38:19 +0000 (UTC) (envelope-from wawa@yandex-team.ru) Received: from [87.250.227.227] (v3-227-227.yandex.net [87.250.227.227]) by relanium.yandex.ru (8.14.1/8.14.1) with ESMTP id l94Lc7wc008088 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Fri, 5 Oct 2007 01:38:16 +0400 (MSD) (envelope-from wawa@yandex-team.ru) Message-ID: <47055D3A.2020604@yandex-team.ru> Date: Fri, 05 Oct 2007 01:38:02 +0400 From: Vladimir Ivanov Organization: Yandex LLC User-Agent: Thunderbird 1.5.0.13 (X11/20070824) MIME-Version: 1.0 To: rmkml , "freebsd-net@freebsd.org" References: <47055447.3020206@yandex-team.ru> In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Antivirus: Dr.Web (R) for Mail Servers on relanium.yandex.ru host X-Antivirus-Code: 100000 Cc: Subject: Re: SMPable version of EM driver X-BeenThere: freebsd-net@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Networking and TCP/IP with FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 04 Oct 2007 21:38:20 -0000 rmkml wrote: > thx Vladimir, > I have litle tested your em driver, > quick comparaison : > kernel: sys=10% intr=25% user=15% > kernel+polling: sys=7% intr=5% user=15% > kernel+yandex_em: sys=30% intr=1% user=15% > same trafic (tcpreplay ~230Mbps) > fbsd62release_amd64 on intel core 2 duo E6600+4Go > It is possible reduce sys cpu ? I'm not sure what that digits mean :-). There is one more (see RX_KTHREADS_NUM) than usual thread started by the driver. That's why the results isn't easy to compare. Much of L/A calculation techniques depends of "number of running threads". > Regards > Rmkml > > > On Fri, 5 Oct 2007, Vladimir Ivanov wrote: > >> Date: Fri, 05 Oct 2007 00:59:51 +0400 >> From: Vladimir Ivanov >> To: rmkml , "freebsd-net@freebsd.org" >> >> Subject: Re: SMPable version of EM driver >> >> rmkml wrote: >>> Hi Vladimir, >>> very thank for your work on intel em driver ! >>> just commented line 772 and added 773 : >>> 771: /* Send a copy of the frame to the BPF listener */ >>> 772: /* ETHER_BPF_MTAP(ifp, m_head); */ >>> 773: BPF_MTAP(ifp, m_head); >>> what is ETHER_BPF_MTAP() ? >> >> look at RELENG_6. >> They've moved vlan promisc hack from driver level into ethersubr. >> Briefly: we've to disable hardware vlan tagging if we want to tcpdump >> or bridge trunked port. >> >>> Best Regards >>> Rmkml >> >> Regards, >> Vladimir >> >>