[c-nsp] TestUnusedPortLoopback failing on storm-control enabled shut port?

[Peter Rathlev]

Before starting the heavy support apparatus:

C6k/Sup720 PFC3B running SXI1 AIS. After some "cleaning up", where our
defaults for storm-control were applied to some administratively shut
ports, we're seeing messages like these:

003716: Aug 22 22:58:17.225 CEST: %HA_EM-6-LOG: Mandatory.go_unusedportlpbk.tcl: GOLD EEM TCL policy for TestUnusedPortLoopback
003717: Aug 22 22:58:16.544 CEST: %CONST_DIAG-SP-3-HM_PORT_TEST_FAIL: Module 6 TestUnusedPortLoopback Port(s)[1] failed. System operation continues.

The result is the same when we run the test manually with "diagnostic
start module 6 test 34 port 1":

003724: Aug 23 10:57:52.173 CEST: %DIAG-SP-6-TEST_RUNNING: Module 6: Running TestUnusedPortLoopback{ID=34} ...
003725: Aug 23 10:57:57.643 CEST: %DIAG-SP-3-TEST_FAIL: Module 6: TestUnusedPortLoopback{ID=34} has failed. Error code = 0x2B (DIAG_CHECK_ETHER_PAK_ERROR)

Sometimes the tests aren't failing, but most of the time they are.

It seems that when we remove the (recently applied) storm-control
configuration, the tests can complete successfully. The interface
configuration of this interfaces when failing is this:

Current configuration : 135 bytes
!
interface GigabitEthernet6/1
 no ip address
 shutdown
 storm-control broadcast level 50.00
 storm-control multicast level 50.00
end

It's port 1 on the supervisor itself. We haven't tested removing
storm-control thoroughly yet, but it seems to be the thing. We normally
aren't using the supervisor ports, except for temporary connections in
special situations.

We began seeing this on one of 32 similar core switches a few days ago.
Now it has started appearing on another, though the log messages are
further apart. The two supervisors are the same age and hardware
revisions (Sup:5.2, PFC:2.3, MSFC3:2.5).

Is TAC the only way?

-- 
Peter