From owner-freebsd-stable Wed Dec 26 11:43:41 2001 Delivered-To: freebsd-stable@freebsd.org Received: from mail2.uniserve.com (mail2.uniserve.com [204.244.156.10]) by hub.freebsd.org (Postfix) with ESMTP id 6A18A37B405 for ; Wed, 26 Dec 2001 11:43:34 -0800 (PST) Received: from mail2.uniserve.com ([204.244.156.10]) by mail2.uniserve.com with esmtp (Exim 3.13 #1) id 16JJxZ-000BAW-00; Wed, 26 Dec 2001 11:43:29 -0800 Date: Wed, 26 Dec 2001 11:43:29 -0800 (PST) From: Tom X-Sender: tom@athena.uniserve.ca To: Allen Landsidel Cc: sthaug@nethelp.no, freebsd-stable@FreeBSD.ORG Subject: Re: 4.5 PRERELEASE - Call for testing In-Reply-To: <5.1.0.14.0.20011226141409.00b02048@rfnj.org> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG On Wed, 26 Dec 2001, Allen Landsidel wrote: > At 08:04 PM 12/26/2001 +0100, sthaug@nethelp.no wrote: > > > Tom's reply stated that if you turned auto-negotiation off by forcing a > > > speed/duplex setting on either end that it would cause problems. This > > > simply isn't true because you can't turn auto-negotating off, and > > forcing a > > > speed/duplex setting doesn't do this.. what it does do is forces the other > > > end to negotiate this setting, if it is available. The negotiation still > > > occurs. > > > >It's certainly possible that we're using the wrong terminology. The > >observation still stands (supported by *lots* of examples in practice): > >If you manually set duplex at one end, and leave it unconfigured (ie. > >auto-negotiation) at the other end, you will often get a duplex mismatch. > > > >Please don't try to claim that this doesn't happen. It does - even with > >new equipment from well known manufacturers. > > I'll claim I've never seen it so long as I was using auto-negotiating > devices.. nics from linksys, 3com, intel and others.. switches and hubs > from the same three. > > You (and Tom) have still both failed to address the question posed : What > about "dumb" devices such as unmanaged switches. My 16port > auto-negotiating 10/100 switch has no way *at all* of setting the duplex > *or* speed of the ports.. are you stating that such devices (and there are > tons of them) won't work if you force the NIC to a certain speed/duplex > setting? Exactly. If you disable auto-negotation on one end, auto-negotation won't work. If one end only supports auto-negotation, you should use auto-negotation on both ends. If one end can't support auto-negotation, the auto-negotation side will auto-detect the speed and use half-duplex, which is the default for devices incapable of auto-negotiation. You can't use full-duplex at all if one end only supports auto-negotation and the other end does not (or auto-negotiate is disabled). Also, by setting a manual setting, you disable autonegotation. I know that you dispute this, but if check the ifconfig manpage, the source, and the ethernet standards, you will see that auto-negotiation and manual settings are mutually exclusive. The same thing goes for switches. If you you have a managed switch, and you set a manual setting auto-negotation is disabled. So back to my original point: it is not a good idea to just enable auto-negotation on one end of the link. I would recommend people use auto everwhere. Even the cheap devices support it properly now. Auto-negotiation is going to become more critical as more options are added to the ethernet standard (ie. master/slave, flow control). Tom To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message