[c-nsp] Negotiation problem with Catalyst 2950 and Cisco 2821

[Michael Loftis]

--On December 3, 2005 12:51:26 PM -0700 John Neiberger 
<jneiberger at gmail.com> wrote:

> We have the opposite philosophy. We set everything to auto. We have
> *far* more problems with hard-set switch settings. It's been my
> experience--and I have documentation at work to back this up--that
> hard setting your devices to 100/Full is the worst possible choice.

I have to agreee, especially in large networks, or networks where you don't 
control a lot of the connected devices.  Autoneg is well established and 
works very well nowadays.  The old lock it in attitude comes from devices 
that didn't or don't support autoneg.  The problems are outlined by Mr. 
Neiberger, there is no standard when 'hard wiring.'  When I code PHY 
drivers I attempt autoneg, then if that fails, if the PHY supports 
detection, then run at HDX at the detected (10meg or 100meg) rate, and if 
all else fails, and we do have a link I'll lock it to 10meg/HDX.  This is 
of course barring any specific settings from user level configuration.

Others do it differently.  I've seen devices that leave autoneg on when 
locked into 10M/HDX but advertise 100M/FDX during the autonegotiation 
process and other odd behaviors.  In my experience autoneg has become the 
safest bet, as I said, become, it's not always been like that.

>
> The short story is that there is no standard method of behavior when a
> NIC or switch port is hard set. Some devices still participate in Nway
> autonegotiation even when hard set, while others completely disable
> Nway. You run into problem when you connect devices that behave
> differently. Some NICs that still participate in Nway will see that
> there is no autonegiating link partner, so they'll try to be "helpful"
> and fall back to half duplex under the assumption that they are not
> connected to a device capable of full duplex.
>
> I have successfully got Cisco to change their documentation to reflect
> this behavior, but I don't have the link to that documentation here at
> home. I'd have to go look for it.
>
> John
>
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>



--
"Genius might be described as a supreme capacity for getting its possessors
into trouble of all kinds."
-- Samuel Butler