Error message displayed on deleting vlan member "Failed to get port by bridge port ID"

[allas-nvidia]

Description:

When we remove vlan member from vlan we can see error in log:

ERR swss#orchagent: :- update: Failed to get port by bridge port ID ...

It seems there is a race which can cause the scenario, when we receive notifications about FDB FLUSH on port which was already removed.

removeBridgePort (portsorch.cpp) called as a result of "remove vlan member" command, will call flushFDBEntries and at the end of this function ( removeBridgePort ) will set port.m_bridge_port_id = SAI_NULL_OBJECT_ID;
As a result of flushFDBEntries sai will notify FdbOrch with attr oid which will be passed to the function FdbOrch::update( fdborch.cpp) as bridge_port_id.
However, the function getPortByBridgePortId will return false - as m_bridge_port_id was set to SAI_NULL_OBJECT_ID at the end of the function removeBridgePort and the error "Failed to get port by bridge port ID" will appear in log.

Steps to reproduce the issue:

1. sudo config vlan add 3
2. sudo config vlan member add 3 Ethernet8
3. sudo config vlan member del 3 Ethernet8

Describe the results you received:

Error in log: "Failed to get port by bridge port ID"

Describe the results you expected:

No error in log

Output of show version:

(paste your output here)

Additional information you deem important (e.g. issue happens only occasionally):

[prsunny]

@madhanmellanox , can you please take a look?

[madhanmellanox]

@allas-nvidia in bool PortsOrch::removeBridgePort(Port &port), we call
//Flush the FDB entires corresponding to the port
gFdbOrch->flushFDBEntries(port.m_bridge_port_id, SAI_NULL_OBJECT_ID);

we don't m_bridge_port_id to SAI_NULL_OBJECT_ID, we set vlan_id to SAI_NULL_OBJECT_ID.

[prsunny]

In this flow, flushFDBEntries is called from two places:

1. removeVlanMember
2. removeBridgePort

So I'm assuming flush event is generated in both cases and for the first one, FdbOrch::update correctly process the event but for the second flush event, it throws the Failed to get port by bridge port error. Is this correct understanding?

When you repro the issue, can you please confirm if the STATEDB entry is cleaned?

redis-cli -n 6 keys *FDB*

[allas-nvidia]

Yes, STATEDB entry is cleaned

[liat-grozovik]

fix was integrated to both master and 202012. issue considered to be closed