The problem was as follows:
- there were two nodes with the same MAC address, sending traffic
  at high frequency
- the above triggers a lot of learning requests to rtud, as a result
- rtu HASH table representation in SHMEM is constantly udpated
- the aging process runs in background (separate process) and updates
  the time the entry was last used.

The update process could not the time of last access in the HASH
table in SHMEM because the learing process was constantly occupying
the SHMEM. The age out process was never aging out entry (even if this
one is within mutex) since the access time was never updated.

Thus, now both, update of last access time and age-out process are
both within mutex that allows the learning and aging-out processes
to both access SHMEM and HW.