The switch does not learn MAC address from LLDP multicasts

[vladakoci]

While performing troubleshooting for

https://supportforums.cisco.com/message/3874519   

I realized the 3750 switch does not learn MAC address from LLDP multicasts generated every 30 seconds by end device connected to the port .

It surprised me, not sure if this is expected behaviour or not. It is not dependent on if port security is set or not.

For the issue above I had a Polycom IP phone connected to the port. It generates LLDP multicast every 30 seconds, but does not do anything else.

The switch sees the phone as LLDP neighbour, but does not learn its MAC address.

I reproduced this scenario later on with Cisco switches only, 'works' the same way.

Does anyone know is this expected or a bugy behaviour ?

Thank you,

Vlad

[hostettle]

Hi,

I would not be surprised by that.

See 802.1Q-2011 section 7.5: "Frames that carry control information to determine the active topology and current extent of each VLAN, i.e., spanning tree and MVRP PDUs, and frames from other link constrained protocols, such as EAPOL and LLDP, are not forwarded".

"Are not forwarded" means not relayed from a port to another port, through the MAC 802.3 switching function. So, the learning process is not invoked.

Best regards,

Michel

[vladakoci]

Thank you Michel,

I had my own internal explanation to that the same like you provided, means that most likely if the switch does not forward the frame and processes it just within the switch it does not learn the MAC address. But I have not found any explicit document that would say LLDP multicast are not used for learning process.

I compared those to LOOP protocol (  used by keepalive ) frames. These frames are also not ( normally ) forwarded, but the switch learns MAC address from them. But maybe  this comparison is not good, as in case of loops LOOP frames would be forwarded, so there is probably no reason why they should not be used for learning.

I wonder if there is an explicit list of all protocols that are not used for learning process. that would be very helpful.

Vlad

[hostettle · Accepted solution]

Hi Vlad,

> But I have not found any explicit document that would say LLDP multicast

> are not used for learning process.

802.1Q-2011, § 7.5 is explicit on this subject.

> I compared those to LOOP protocol (  used by keepalive ) frames.

> These frames are also not ( normally ) forwarded

I think you are refering to a specific function of the link OAM sub familly included in the Slow protocols familly. This loop function inverts SA and DA in the MAC frame received, and returns the frame to the same port. So, there is no need of forwarding between ports!

If it is a loop through another port, it is not a link OAM protocol but a network OAM protocol wih other characteristics.

> I wonder if there is an explicit list of all protocols that are not used

> for learning process. that would be very helpful.

There is not a exhaustiv list, it depends on the implementation.

For example, for EVPLAN in a WAN bridge, you have systematically PAUSE ad LLDP discard: see G.8011.2 (01/2009) table 8-2.

Generally, this question concerns (1) the SA addresses when the DA addresses are with OUI = 00-80-C2, (2) the configuration of the filtering database.

Best regards,

Michel

[vladakoci]

Thank you Michel.

I should have been more precise and say" But I have not found any publicly available explicit document" but I clearly understand it is mentioned  in IEEE document that I do not have an access to.

[vladakoci]

I've performed several other tests and think the answer above might not be Cisco-wise correct.

I tested on WS-C3560CG-8PC-S and this switch learns the MAC address from the LLDP multicast.

I tested also on this stack

-- ----- ----- ---------- ----------

* 1 54 WS-C3750X-48P 12.2(55)SE6 C3750E-UNIVERSALK9-M

2 54 WS-C3750X-48P 12.2(55)SE6 C3750E-UNIVERSALK9-M

3 52 WS-C3750-48P 12.2(55)SE6 C3750-IPBASEK9-M

4 52 WS-C3750-48P 12.2(55)SE6 C3750-IPBASEK9-M

5 52 WS-C3750-48P 12.2(55)SE6 C3750-IPBASEK9-M

6 52 WS-C3750-48P 12.2(55)SE6 C3750-IPBASEK9-M

and can see ports on 3750X learn MAC address from LLDP multicast, but on WS-C3750-48P not. This looks like an inconsistency or a Cisco bug.

Does anyone have an authoritative answer if a Cisco switch is supposed to learn a MAC address from a LLDP multicast ?

Thank you,

Vlad

[hostettle]

Hi Vlad,

Unfortunally, you don't bring any consistant information, traces.

LLDP has a reserved DA address = 01-80-C2-00-00-0E.

All the frames with a reserved address have not there SA address learned. They are not relayed.

It is same as STP frames.

If you see these SA addresses, it is not due to the learning process.

Best regards,

Michel Hostettler

[vladakoci]

Cisco confirms it is a bug that MAC address is learnt from LLDP multicasts on platforms such as C3560, C3750X

http://tools.cisco.com/Support/BugToolKit/search/getBugDetails.do?method=fetchBugDetails&bugId=CSCsm95391

[mhnedirli]

can you look to the "sh lldp neighbors detail" i have 3560 sw and it shows mac address as chassis id