From owner-freebsd-current@FreeBSD.ORG Mon Jan 19 03:24:24 2004 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 B35C516A4DB for ; Mon, 19 Jan 2004 03:24:24 -0800 (PST) Received: from mail.drunkencomputing.de (ratz.drunkencomputing.de [195.244.235.248]) by mx1.FreeBSD.org (Postfix) with ESMTP id 8D08A43D2F for ; Mon, 19 Jan 2004 03:24:20 -0800 (PST) (envelope-from hscholz@raisdorf.net) Received: from localhost (localhost [127.0.0.1]) by mail.drunkencomputing.de (Postfix) with ESMTP id 55B998B115; Mon, 19 Jan 2004 12:24:11 +0100 (CET) Received: from mail.drunkencomputing.de ([127.0.0.1])port 10024) with ESMTP id 56141-10; Mon, 19 Jan 2004 12:24:10 +0100 (CET) Received: from goanna.raisdorf.net (pD958CA44.dip.t-dialin.net [217.88.202.68]) by mail.drunkencomputing.de (Postfix) with ESMTP id EB6888B10F; Mon, 19 Jan 2004 12:24:06 +0100 (CET) Date: Mon, 19 Jan 2004 12:24:17 +0100 From: Hendrik Scholz To: Andreas Moeller Message-Id: <20040119122417.69b2fa94@goanna.raisdorf.net> In-Reply-To: <4009ADE7.2060101@gmx.net> References: <4009ADE7.2060101@gmx.net> X-Mailer: Sylpheed version 0.9.8claws (GTK+ 1.2.10; i386-portbld-freebsd5.2) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at drunkencomputing.de cc: current@FreeBSD.org Subject: Re: fxp(4) TCP checksum offloading X-BeenThere: freebsd-current@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list Reply-To: hscholz@raisdorf.net List-Id: Discussions about the use of FreeBSD-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 19 Jan 2004 11:24:24 -0000 Hi! On Sat, 17 Jan 2004 22:49:27 +0100 Andreas Moeller wrote: > I got an Intel PRO/100 S NIC (Intel 82550 chip) which supports TCP > checksum offloading. I am also able to load the microcode necessary in > > order to activate that feature: > > fxp0: Microcode loaded, int_delay: 1000 usec bundle_max: 6 That ain't TCP checksum offloading, that is IRQ mitigation. When using IRQ mitigation the FIFO buffer in the network card is used to 'delay' interrupts. Up to 6 packets will be buffered for up to 1000usec before raising an interrupt. In case more than one packet arrives during the 1000usec period only one interrupt is triggered instead of multiple thus reducing the time spend for context switching and handling the interrupts. IRQ mitigation is enabled using the link0 feature as described in the fxp manpage. fxp1: port 0x8440-0x847f mem 0xfc320000-0xfc33ffff,0xfc362000-0xfc362fff irq 23 at device 5.0 on pci4 [0.08] root@cave:~ > ifconfig fxp1 link0 [0.07] root@cave:~ > ifconfig fxp1 fxp1: flags=9802 mtu 1500 options=3 ether 00:02:b3:b0:60:8c media: Ethernet autoselect (none) status: no carrier [0.07] root@cave:~ > ifconfig fxp1 -link0 [0.06] root@cave:~ > ifconfig fxp1 fxp1: flags=8802 mtu 1500 options=3 ether 00:02:b3:b0:60:8c media: Ethernet autoselect (none) status: no carrier [0.06] root@cave:~ > Additionally the options shows that RX and TX checksums are enabled by the driver. > Unfortunately, when enabled, network connections (at least those based > > on TCP) start timing out while ping remains working flawlessly. The > system in question is running latest -CURRENT source as of today. I cannot reproduce this problem since my box is running 1000baseSX. As of 5.2-REL you have to enable the sshd and in case security is set to 'extreme' sshd won't be fired up. Do you have sshd_enable="YES" in /etc/rc.conf? > Any ideas? (Is there a way to unload the microcode again besides > rebooting the system?) $ ifconfig fxp0 -link0 Hendrik -- Hendrik Scholz - - http://raisdorf.net/ cell phone: 404-606-5324 (US) 0160-1570-272 (DE)