Speed and Duplex Mismatch

[vontech615]

I have a Netvanta 1544.  It is throwing the following message...

LLDP LLDP: Speed Duplex mis-match on giga-swx 0/12. Speed/Duplex:

"1G/full-duplex" Neighbor Speed/Duplex: "100M/full-duplex"

It is throwing this message on all of the ports.  I think I know the issue but not sure how to solve it.  We have NEC DT700 IP Phones.  On the phones we have a Gigabit adapter to makes the LAN and PC ports gigabit capable. So PC and Phone are on 1 physical port on the switch.  However, the phone itself is only Fast Ethernet. Please see links to pics so you can an idea.   I believe through LLDP the switch thinks there is a mismatch somewhere because the phone is only 100M.  The errors are driving me nuts.

The adapter attached to the phone...

http://i.imgur.com/etcSHTn.jpg

This is the adapter LED's.  The red light is due to it being only 100m from the phone to the adapter.  The Cable plugged into the LAN port goes to my Adtran.

http://i.imgur.com/wtDN96H.jpg

[Anonymous]

- Thanks for posting your question on the forum!

I would agree with your idea that, via LLDP, the phone is reporting itself as having a speed of 100Mbps while the switch reports that it is running at speed of 1Gbps. The easiest way to get rid of the error messages is to disable LLDP on the switchport(s) that are reporting the error. It will not affect any functionality on your network, but only disable LLDP messages from being shared. The command can be issued from the switchport config level: "no lldp send-and-receive"

If the message is only popping up while you are in the CLI of the switch, you can issue the "no events" command and that will stop the events from being displayed during your session.

More information regarding LLDP in AOS can be found in this guide:

Please do not hesitate to let us know if you have any questions.

Thanks,

Noor

[jayh]

In your particular case the adapter is likely passing LLDP through rather than actively participating in LLDP and re-writing the appropriate speeds.  Do you still get the chatter if you plug the phone directly into the port? 

There's an LLDP bug that has been around for quite a while.  We've seen it with Polycom phones as well.  I suspect it is on the Adtran side but could be some subtle non-compliance with RFCs on the other side as well.  FWIW, we have customers with H/P PoE switches and the same Polycom phones that don't have the problem. 

It seems to be cosmetic only.  If the speeds were truly different, communication would fail.  If duplex were truly mismatched you would see runts and CRC errors as well as miserable throughput. 

"No events" will allow you to use the CLI without the spew.  Disabling LLDP is a workaround but if you're using LLDP for voice VLAN assignment you're stuck leaving it on.

Some dumb switches and other devices will pass LLDP and CDP packets through which can cause confusing results.  Your speed adapters may fall into that category.  Disabling LLDP on any port connected to an unmanaged switch, bridge, or wireless AP would be good practice.

Previous threads re this problem:

[vontech615]

Here is a link to a log output of the show interfaces command.

https://skydrive.live.com/?cid=a552efbe03fccb30&id=A552EFBE03FCCB30%212233&action=Share

I'm not totally sure what to look for but I do notice that there's no CRC and/or Runt's going on on these interfaces.

[jayh · Accepted solution]

There is in fact no speed or duplex mismatch.  The warnings about it are in error, generated by a problem with a protocol called LLDP (link-layer discovery protocol).

LLDP is a standards-based derivative of proprietary CDP (Cisco Discovery Protocol).

It is supposed to communicate only between two directly-connected interfaces on a physical link.  Each interface identifies a set of parameters about its host such as make/model/software, trunking and VLANs (including voice VLAN) and speed and duplex.

The 1544 switch compares its settings with those of the other end as reported by LLDP.  If it detects a problem with the link such as the speed and/or duplex not matching, different native VLANs, etc., then it throws a warning on the console.

What is happening here is that LLDP *shows* a speed mismatch when in fact there is none.  This pops up as an error, and causes a lot of noise on the console.  If in fact one side of a link were 100M and the other 1G  then no traffic would pass, so the warning is obviously bogus.

We have seen similar erroneous LLDP speed mismatch reports between Adtran switches and Polycom phones, see the discussions referenced in other links.

In this particular case, I suspect the speed adapter boxes at the phones are passing the LLDP from the phone (at 100M) to the switch (at 1G) which causes the errors.  If so, the adapter boxes aren't handling LLDP correctly.  They should either send LLDP based on their own directly-connected interfaces to each end, or drop the LLDP frames completely.  As LLDP is supposed to only address directly-connected interfaces, the device shouldn't pass it through as received on one interface out another.

Short answer:  You don't have a speed mismatch.  The error messages can be ignored or hidden with "no events".

[Anonymous]

As an update, in R10.11.0 AOS, a switchport command was added that allows an AOS switch the ability to ignore these LLDP speed-duplex mismatch messages.  If you add the command "no lldp receive 802.3-info mac-phy-config" to any applicable switchports, the switch will no longer log every LLDP update indicating a mismatch as an event in the event-history.