BGP: resolve lifecycle-node check via getContainingNode(), document the FSM gating

Move BgpRouter::isLifecycleNode() out of line and look up the node with
getContainingNode() instead of getParentModule(), and document why the FSM
only re-arms reconnection on lifecycle-capable nodes: enabling auto-reconnect
unconditionally re-opens a listening socket on an already-bound port and
crashes plain (non-lifecycle) BGP scenarios.

Author: avarga

src/inet/routing/bgpv4/BgpRouter.cc

@@ -78,6 +78,17 @@ void BgpRouter::recordStatistics()
     bgpModule->recordScalar("UpdateMsgRcv", statTab[5]);
 }
 
+// A node is "lifecycle-capable" if it contains a NodeStatus ("status") submodule,
+// i.e. it can be shut down, restarted or crashed at runtime. Only on such nodes does
+// the FSM re-arm a connection after a loss (see BgpFsm.cc): a restarted node must
+// reconnect, whereas a plain BGP router keeps its historical behavior of not
+// auto-reconnecting. Enabling auto-reconnect unconditionally would also re-open a
+// listening socket on an already-bound port, crashing non-lifecycle scenarios.
+bool BgpRouter::isLifecycleNode() const
+{
+    return getContainingNode(bgpModule)->getSubmodule("status") != nullptr;
+}
+
 void BgpRouter::closeSessions(bool abort)
 {
     for (auto& elem : _BGPSessions) {

src/inet/routing/bgpv4/BgpRouter.h

@@ -120,7 +120,7 @@ class INET_API BgpRouter : public TcpSocket::BufferingCallback
     cMessage *getCancelEvent(cMessage *msg) { return bgpModule->cancelEvent(msg); }
     IIpv4RoutingTable *getIPRoutingTable() { return rt; }
     std::vector<BgpRoutingTableEntry *> getBGPRoutingTable() { return bgpRoutingTable; }
-    bool isLifecycleNode() const { return bgpModule->getParentModule()->getSubmodule("status") != nullptr; }
+    bool isLifecycleNode() const;
 
     /**
      * \brief active listenSocket for a given session (used by fsm)