Logic table derived from PurgeChange.logic

Variables

1. isEgress
   "MPS is acting as a gateway, so next hop is not necessarily same as shortcut destination. Shortcut next hop is an IP address on the MPS/NHS. [per cache entry]"
   • notEgress
   • egress
2. MPOA-NH
   "Whether CIP has been associated with the reply, i.e.shortcut destination is an MPC. [per cache entry]"
   • CIP
   • noCIP
3. dest-type
   "where the next hop is with respect to the IP interfaces of the NHS/MPS.nonAuth implies a non-authoritative reply has been issued from the NHRP cache. so is necessarily remote too. (egress implies remote)? [no, NBMA-ID change is example where dest can be local/subnet] [per cache entry]"
   • local
   • subnet
   • remote
   • nonAuth
4. Change
   "The type of indication of the change.
   + Route change [Dest/Mask]: oldNHIP -> newNHIP
   + MPC change [Elan+MAC]: oldMPC ATM -> new MPC ATM
   + ARP change [IP] oldMAC->newMAC (subnet implied by IP)
   + Registration (NHRP) [IP] oldNBMA->newNBMA
   + Port failure [IP] ifOper=down, llc_read fail? (Should we handle this, or have it translate to route changes?)
   + Neighbour change [MPC-ATM] (An existing MPC changes device type, to MPS)
   + Egress device sent a purge for some reason [SrcIP,cie.DestIP,cie.mask]:dataATM->null
   (For NHRP SrcIP==NHCIP, for MPOA SrcIP==MPSIP(of router i/f on elan)
   + other device sent a purge for some reason [SrcIP,cie.DestIP,cie.mask]:dataATM->null
   (is different case, but how is it distinguished?)
   + Configuration/Policy change (e.g. security) [???]"
   • route
   • MPC
   • ARP
   • Register
   • PortFail
   • Neighbour
   • RecvPurge

Outcomes

1. reality
   "Indicates whether this case can actually happen due to dependencies between the variables."
   • possible
   • impossible
2. SeeRouteChange
   "A poll or active indication from routing subsystem that the existing IP/Mask -> next hop is no longer valid
   

   SeeRouteChange(dest,mask,was_via)}

   "
   • SeeRouteChange
   • none
3. Context
   "How the change is identified to the MPS/NHS
   + RouteChange from routing subsystem (poll or message) also I/F fail?
   + RequestPlus i.e. during resolution request processing, le_arp, arp etc, and also asynchronous le_arp invalidation
   + Registration (upon receipt of NHRP registration pkt)
   + Purge (upon receipt of a purge pkt)"
   • RouteChange
   • RequestPlus
   • Register
   • Purge
4. pg-path
   "Purges destinations that were routed through existing next hop implying reply history match of next hop IP and old route next hop AND the masked compare using the old route mask gives equality of reply's destination and the old route dest AND there is no other route using the same next hop with a more specific mask. (Refer to day book 16/4/97 for sample config)
   

   pg-path(RouteDestIP, oldNH){
     For all History in ReplyHistoryTable
     if ( (History.NHIP == oldNH) && 
        (History.mask & RouteDestIP) == (History.mask & History.targetIP)) {
             lookup(History.dest)=>[new.NHIP, new.PrefixLen] ;
             if ( (new.NHIP = oldNH) && 
                  (new.PrefixLen > History.PrefixLen)  {
                  do_nothing();
             } else {
                  send_purge ( History.{target, prefixLen, ...})
             }
     } else {
       do_nothing();
     }
   }

   "
   • purgePath
   • none
5. pg-MAC
   "Purges all MPOA reply history records with shortcut set up set to this MAC on this ELAN.
   RH.MAC == MAC && RH.ifIndex == ifIndex(ELAN)
   

   pg-MAC(MAC,ELAN) {
   	// alternate implementation using assumption that MAC unique in subnet
   	for all ARP records in ELAN
   		if arp.MAC == MAC
   			pg-IP(arp.IP)   // problem with Multicast IPs?
   }

   "
   • purgeMAC
   • none
6. pg-IP
   "Purge all replies with Next Hop IP equal original IP, or IP with changed ATM.
   

   pg-IP (IP) {
   	for all in RH table
   		if RH.nextHopIp == IP	
   			purge RH
   }

   "
   • purgeNHIP
   • none
7. pg-MskegressIP
   "Purge all replies for which MPS/NHS is egress, with egress IP address equal to the original route's IP or the IP with the changed ATM address masked with the the mask of the subnet on which the MPS has an interface. (IP = egressMPS.IP's interface)
   (RH.nextHopIp & mask(IP.ifIndex)) == (IP & mask(IP.ifIndex)))
   

   pg-MskegressIP(EIP) {
   	lookup(IpAddrTable, EIP) [using masked compare ipAdEntNetMask] => Lmask
   	for all RH in ReplyHistoryTable
   		if ((RH.nextHopIp & Lmask == (EIP & mask(RH.ifIndex))) && RH.isEgress)
   			send_purge(RH)
   
   }

   "
   • pg-EgressIPSub
   • none
8. pg-NBMA
   "Purges based on match of final destination protocol address, and an ATM address that was previously used in the reply.
   

   pg-NBMA (DestIP, Prefix, NBMA) {// also have sourceIP of either NHC or MPS
       for all RH in ReplyHistoryTable
           if (DestIP == RH.DestIP)&& (NBMA==RH.NBMA)&&(Prefix>RH.prefix)
                purge(RH)
   }
   // should it be ::-
   //if (DestIP/prefix == RH.DestIP/prefix)&&(NBMA==RH.NBMA)&&(Prefix>RH.prefix)
   

   "
   • pg-NBMA
   • none

Rules

1. "Notification of route changes"
   Conditions
   • reality != impossible
   • Change = route
   
   Outcomes
   • SeeRouteChange = SeeRouteChange
   
   
2. "Notification of route changes (default). Don't know what happens with a port fail, as may or may not get detection by routing system."
   Conditions
   • reality != impossible
   • Change != PortFail
   • Change != route
   
   Outcomes
   • SeeRouteChange = none
   
   
3. "For destinations local to NHS/MPS, it must be the egress"
   Conditions
   • isEgress = notEgress
   • dest-type = local
   
   Outcomes
   • reality = impossible
   
   
4. "MPOA subnets require imposition. [could this change if subnet type changes?]"
   Conditions
   • Change = MPC
   • MPOA-NH = noCIP
   
   Outcomes
   • reality = impossible
   
   
5. "MPOA subnets require imposition. [could this change if subnet type changes?]"
   Conditions
   • Change = Neighbour
   • MPOA-NH = noCIP
   
   Outcomes
   • reality = impossible
   
   
6. "MPOA subnets require IP-ARP. [could this change if subnet type changes?]"
   Conditions
   • Change = ARP
   • MPOA-NH = noCIP
   
   Outcomes
   • reality = impossible
   
   
7. "NHRP subnets require registration [could this change if subnet type changes?]"
   Conditions
   • Change = Register
   • MPOA-NH != noCIP
   
   Outcomes
   • reality = impossible
   
   
8. "Local knowledge is authoritative"
   Conditions
   • isEgress = egress
   • dest-type = nonAuth
   
   Outcomes
   • reality = impossible
   
   
9. "MAC changes can 'always' be converted to use NH-IP. pg-IP(invArp(MAC,ELAN))"
   Conditions
   • pg-MAC = purgeMAC
   
   Outcomes
   • pg-IP = purgeNHIP
   
   
10. "Egress point changed.
    pg-path(dest,was_via) -- note that lost mask!"
    Conditions
    • SeeRouteChange = SeeRouteChange
    • reality != impossible
    
    Outcomes
    • pg-path = purgePath
    
    
11. "Egress shuts down [Note: need to match control address of E-MPC/E-NHC to reply history, else cache entries purged on different(new) egress point]
    

    pg-IP (cie.DestIP)
    

    "
    Conditions
    • Change = RecvPurge
    • reality != impossible
    
    Outcomes
    • pg-IP = purgeNHIP
    
    
12. "LANE bridging change
    

    	pg-MAC(ELAN,MAC)	// have old and new MPC-Data ATM addresses too?
    

    "
    Conditions
    • Change = MPC
    • MPOA-NH = CIP
    • reality != impossible
    
    Outcomes
    • pg-MAC = purgeMAC
    
    
13. "LANE bridging change, ARP change (unlikely?)
    

    	pg-MAC(ELAN(IP),oldMAC)
    

    "
    Conditions
    • Change = ARP
    • MPOA-NH = CIP
    • reality != impossible
    
    Outcomes
    • pg-MAC = purgeMAC
    
    
14. "Device ATM address changes (unlikely?)
    

    	NHRP_Registration_change(IP, oldNBMA, newNBMA) {
    		pg-IP(IP)
    	}
    

    "
    Conditions
    • Change = Register
    • MPOA-NH = noCIP
    • reality != impossible
    
    Outcomes
    • pg-IP = purgeNHIP
    
    
15. "lose local ATM I/F. In NHRP no requirement to purge, but MPOA equivalent of MPC losing keep-alive. Addresses purged are all those with IP on subnet. [Not sure if this should be sent or not, Keep-alive mechanism may be enough! [problem that can't send any purges anyway?] [What happens with NBMA differences, where NHS/MPS is egressing?]
    

    port_fail(IP) {
    	pg-MskegressIP(IP)
    }
    

    "
    Conditions
    • reality != impossible
    • Change = PortFail
    • isEgress = notEgress
    • MPOA-NH = CIP
    
    Outcomes
    • pg-MskegressIP = pg-EgressIPSub
    
    
16. "Lose other I/F
    

    port_fail(IP) {
    	pg-MskegressIP(IP)
    }
    

    "
    Conditions
    • reality != impossible
    • Change = PortFail
    • isEgress = egress
    
    Outcomes
    • pg-MskegressIP = pg-EgressIPSub
    
    
17. "NHC/MPC purge.
    

    purge_packet(CIE_IP, CIE_Prefix, CIE_NBMA) {
    		pg-NBMA(CIE_IP, CIE_Prefix, CIE_NBMA)
    }
    

    "
    Conditions
    • reality != impossible
    • Change = RecvPurge
    
    Outcomes
    • pg-NBMA = pg-NBMA
    
    
18. "Context for changes"
    Conditions
    • Change = route
    
    Outcomes
    • Context = RouteChange
    
    
19. "Context for changes"
    Conditions
    • Change = MPC
    
    Outcomes
    • Context = RequestPlus
    
    
20. "Context for changes"
    Conditions
    • Change = ARP
    
    Outcomes
    • Context = RequestPlus
    
    
21. "Context for changes"
    Conditions
    • Change = Register
    
    Outcomes
    • Context = Register
    
    
22. "Context for changes (sililar to route change at least)"
    Conditions
    • Change = PortFail
    
    Outcomes
    • Context = RouteChange
    
    
23. "Context for changes"
    Conditions
    • Change = Neighbour
    
    Outcomes
    • Context = RequestPlus
    
    
24. "Context for changes"
    Conditions
    • Change = RecvPurge
    
    Outcomes
    • Context = Purge
    
    

Variables Summary Table

Outcomes Summary Table

Logic Table