From owner-freebsd-net@FreeBSD.ORG Mon Sep 29 21:08:44 2008 Return-Path: Delivered-To: net@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 82B471065687; Mon, 29 Sep 2008 21:08:44 +0000 (UTC) (envelope-from luigi@onelab2.iet.unipi.it) Received: from onelab2.iet.unipi.it (onelab2.iet.unipi.it [131.114.9.129]) by mx1.freebsd.org (Postfix) with ESMTP id 3B59B8FC1F; Mon, 29 Sep 2008 21:08:44 +0000 (UTC) (envelope-from luigi@onelab2.iet.unipi.it) Received: by onelab2.iet.unipi.it (Postfix, from userid 275) id F2476730A8; Mon, 29 Sep 2008 23:12:18 +0200 (CEST) Date: Mon, 29 Sep 2008 23:12:18 +0200 From: Luigi Rizzo To: Robert Watson Message-ID: <20080929211218.GC69054@onelab2.iet.unipi.it> References: <20080929043134.GD54819@cdnetworks.co.kr> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.4.2.3i Cc: pyunyh@gmail.com, stable@freebsd.org, net@freebsd.org Subject: Re: 7.1-PRERELEASE : bad network performance (nfe0) 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: Mon, 29 Sep 2008 21:08:44 -0000 On Mon, Sep 29, 2008 at 09:10:29PM +0100, Robert Watson wrote: > > On Mon, 29 Sep 2008, Arno J. Klaassen wrote: > > >However, the "request/respones" tests are awfull for my notebook (test > >repeated on the notebook for the sake of conviction) : > > Is it possible to rerun these tests with a 7.0 kernel of the same general > configuration? That would help us determine if it's a regression between > 7.0 and 7.1, or perhaps a more general issue between 6.x and 7.x. I > wouldn't reject a hardware, driver, or general stack issue at this point as > things are still fairly unclear. If it's definitely between 7.0 and 7.1 > that the problem arises, trying a series of kernels spaced at, say, one > month intervals in that period would be quite helpful in narrowing down the > source. two things: + the 'nfe' driver is not in 6.x so i wonder how these numbers were derived in 6.x -- perhaps using a backport, or using the nve driver instead ? In any case we cannot easily compare 6.x and 7.x behaviour with nfe. + with the nfe driver and the MCP67 chipset i have found a tendency of the card to stall at high data rates and with some system load (e.g. massive scp while some X applications is spinning). This was completely repeatable and caused the network card to become deaf (it could transmit though) and it required an ifconfig down/up to come back to life, watchdogs and timeouts did not fix it. Additionally i have found some bug in the polling implementation which may or may not relate to more generic interrupt handling. This was described in a thread in april. A patch to address the stalls is at http://info.iet.unipi.it/~luigi/FreeBSD/nfe-20080426.1044.diff (i have been running this on my home pc since late april) A related patch changes slightly the implementation of polling: http://info.iet.unipi.it/~luigi/FreeBSD/nfe-20080427.diff At the time when i raised the problem on the mailing list apparently others were not seeing the problem so i did not pursue the integration in the system - but if this is a significant problem in 7.1R then it is worth a try. cheers luigi