750   race condition: PRP NodesTable vs HSR-PRP RedBox

This tissue has following status: green

Created: 16 Jul 2012

Links:

Page: 20

Clause: 4.2.7.4.1

Paragraph:

Category: Issue for edition 2 of this part

Issue: The sending rules for a DANP with NodesTable allow the DANP to send a frame without trailer, if the destination is registered as a SAN in its NodesTable.

When a HSR RedBox in HSR-PRP mode receives such a frame on its interlink port, it will create an entry in its ProxyNodeTable and generate a sequence number on behalf of the DANP (5.4.2), usually starting with zero. This number is then registered in the duplicate detection of HSR devices. If the DANP now sends a frame to a HSR target, it will use its own sequence counter, which might happen to be zero.

The result: a legitimate frame is dropped inside the HSR ring, even though we have full redundancy.

Usually a frame targeted at a PRP SAN will not be visible by the HSR-PRP RedBox because of the MAC learning of the switches inside the PRP LAN. But you only need to reboot or replace the switch for this to happen.

Proposal: I see two possible solutions:

A) PRP nodes must always append a trailer, even when communicating with a SAN.

B) Remove the ProxyNodeTable from the HSR-PRP RedBox. The HSR-PRP RedBox must not forward frames without PRP trailer from the interlink to the HSR ring. SANs inside a PRP network can still hear HSR nodes but are no longer able to communicate with them.

 

Privacy | Contact | Disclaimer