No trunk information on H3C switches

[Morten Brekkevold]

While testing NAV with an H3C switch, we found that the trunk status of ports was not found, but the native VLAN of all ports was.

NAV gets its VLAN information from Q-BRIDGE-MIB::dot1qVlanCurrentTable, which should reflect the current VLAN configuration of the device. The H3C switch in question does not populate this table at all. It does, however, populate the Q-BRIDGE-MIB::dot1qVlanStaticTable, which reflects the statically configured VLANs of the device. We interpret the MIB to say that the former table should be a superset of the latter, but the H3C does not adhere to this.

I think a solution to this is that NAV should fall back to the Q-BRIDGE-MIB::dot1qVlanStaticTable if the Q-BRIDGE-MIB::dot1qVlanCurrentTable is unpopulated.

[Morten Brekkevold (mbrekkevold)]

I've implemented the necessary changes, but I am none the wiser. I can now get information about egress and untagged ports on the H3C switch, but still nothing to indicate which ports are trunk ports.

In fact, the result tells me that all configured VLANs are transmitted *untagged* on port 8, yet the NOC tells me port 1 is configured for trunking.

The changes made should not change NAV's behavior on any other known device, though, so they can likely be merged for 3.9 and left until we figure this one out (a lot of code cleanup went into this patch as well, which is a plus).

[Morten Brekkevold (mbrekkevold)]

In combination with the parent and grandparent changesets, this should fix the problem: http://metanav.uninett.no/hg/default/rev/4dc5f4691bd9

BUT: The H3C switch still appears to report no trunk ports, but multiple untagged VLANs on a port not configured as trunk.

Until we figure out what's going on with the test switch, this can't really be considered fixed. The mentioned changes do clean the affected code a great deal, and the old behavior was kept, so we'll leave this in for now.

[Morten Brekkevold (mbrekkevold)]

H3C confirmed and fixed a bug with this exact issue in their switch software. Apparently the bitstrings in Q-BRIDGE-MIB::dot1qVlanStaticTable columns were reversed.

Updating the software on our H3C switch reveals a slight typo in the above fix, which has been fixed here: http://metanav.uninett.no/hg/series/3.9.x/rev/5a04a8dee552

With the updated H3C software (5.20 Feature 2210L04) and these patches, the issue seems to be resolved.