From owner-freebsd-hackers Wed Dec 24 06:50:06 1997 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.7/8.8.7) id GAA11565 for hackers-outgoing; Wed, 24 Dec 1997 06:50:06 -0800 (PST) (envelope-from owner-freebsd-hackers@FreeBSD.ORG) Received: from mail.netwalk.com (root@mail.netwalk.com [205.156.197.5]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id GAA11501 for ; Wed, 24 Dec 1997 06:49:56 -0800 (PST) (envelope-from randy@tcs5-28.netwalk.net) Received: from tcs5-28.netwalk.net (randy@tcs5-28.netwalk.net [206.175.76.28]) by mail.netwalk.com (8.8.5/8.6.9) with ESMTP id JAA13357; Wed, 24 Dec 1997 09:49:54 -0500 Received: (from randy@localhost) by tcs5-28.netwalk.net (8.8.5/8.8.5) id JAA05852; Wed, 24 Dec 1997 09:49:29 -0500 Message-ID: <19971224094929.26472@netwalk.com> Date: Wed, 24 Dec 1997 09:49:29 -0500 From: dh2@netwalk.com To: Kapil Chowksey Cc: hackers@FreeBSD.ORG Subject: Re: (?) lnc0: Transmit underflow error -- Resetting Reply-To: dh2@netwalk.com References: <19971223230921.57532@netwalk.com> <199712241526.KAA17065@tarang.hss.hns.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Mailer: Mutt 0.85 In-Reply-To: <199712241526.KAA17065@tarang.hss.hns.com>; from Kapil Chowksey on Wed, Dec 24, 1997 at 10:26:26AM -0500 Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk Kapil Chowksey: > On Tue, 23 December, dh2@netwalk.com wrote: > > I'm trying to do transfers of a large amount of data between my FreeBSD > > 3.0-971208-SNAP box with a NE2100 Lance card and a Linux 2.0.32 box with an > > NE2000. > > > > My effective transfer rate is measely 1.6kbytes per second over a 10Mbps > > ethernet with FTP, and I'm getting loads of these in the FreeBSD error log: > > Ditto here. I have a PCI based NE2100 card (lnc1) on an HP Vectra > pentium 200Mhz class PC which gives equally pathetic performance (both > under 2.2.5-RELEASE and 3.0-971208-SNAP) while ftp'ing to any other > host (Solaris, HPUX, EtherExpress Linux's, ed0 based FreeBSDs). > > It seems that packet sizes greater than 738 get delayed by 1-2 seconds > in the driver itself ! That's interesting. I found last night that if I cranked the lnc0 interface's MTU down to 512, I got slightly better performance, but going even further down to 256 made a much bigger difference. For example, FTPing a 540Meg file between the machines, at MTU=512, it was still going to take > 100min to ftp it over a 10-base ethernet. At MTU=256 however, it took only 15 min, just a few minutes longer than I'd expect at an optimized config. It appears that segment size is related as you said. Though indications here are that there's a significant difference with numbers below 738 -- between 256 and 512 for sure. > I am not getting any messages from lnc driver however. Is there a > maintainer for lnc driver currently ? I'd also be interested. Would be happy to try suggestions and patches here. Thanks, Randall Hopper