Administrator Guide

Handling of Switch-Initiated Trac
When the control processor (CP) initiates a control packet, the following processing occurs:
• TCP/UDP port number is extracted from the sockaddr structure in the in_selectsrc call which is called as part of the connect system
call or in the ip_output function. If the destination TCP/UDP port number belongs to a congured management application, then
sin_port of destination sockaddr structure is set to Management EIS ID 2 so that route lookup can be done in the management EIS
routing table.
• To ensure that protocol separation is done only for switch initiated trac where the application acts as client, only the destination
TCP/UDP port is compared and not the source TCP/UDP port. The source TCP/UDP port becomes a known port number when the
box acts as server.
• TFTP is an exception to the preceding logic.
• For TFTP, data transfer is initiated on port 69, but the data transfer ports are chosen independently by the sender and receiver during
initialization of the connection. The ports are chosen at random according to the parameters of the networking stack, typically from the
range of temporary ports.
• If route lookup in EIS routing table succeeds, the application-specic packet count is incremented. This counter is viewed using the
show management application pkt-cntr command. This counter is cleared using clear management application
pkt-cntr command.
• If the route lookup in the EIS routing table fails or if management port is down, then packets are dropped. The application-specic
count of the dropped packets is incremented and is viewed using the show management application pkt-drop-cntr
command. This counter is cleared using clear management application pkt-drop-cntr command.
• Packets whose destination TCP/UDP port does not match a congured management application, take the regular route lookup ow in
the IP stack.
• In the ARP layer, for all ARP packets received through the management interface, a double route lookup is done, one in the default
routing table and another in the management EIS routing table. This is because in the ARP layer, we do not have TCP/UDP port
information to decide the table in which the route lookup should be done.
• The show arp command is enhanced to show the routing table type for the ARP entry.
• For the clear arp-cache command, upon receiving the ARP delete request, the route corresponding to the destination IP is
identied. The ARP entries learned in the management EIS routing table are also cleared.
• Therefore, a separate control over clearing the ARP entries learned via routes in the EIS table is not present. If the ARP entry for a
destination is cleared in the default routing table, then if an ARP entry for the destination exists in the EIS table, that entry is also
cleared.
• Because fallback support is removed, if the management port is down or the route lookup in EIS table fails packets are dropped.
Therefore, switch-initiated trac sessions that used to work previously via fallback may not work now.
Handling of Switch-Destined Trac
• The switch processes all trac received on the management port destined to the management port IP address or the front-end port
destined to the front-end IP address.
• If the source TCP/UDP port number matches a congured EIS or non-EIS management application and the source IP address is a
management Port IP address, then the EIS route lookup is done for the response trac and hence is sent out of the management port.
In this case, the source IP address is a management port IP address only if the trac was originally destined to the management port
IP.
• ICMP-based applications like ping and traceroute are exceptions to the preceding logic since we do not have TCP/UDP port number.
So if source IP address of the packet matches the management port IP address EIS route lookup is done.
• Management application packet counter is incremented if EIS route lookup succeeds and packet is sent out of the management port.
386
Internet Group Management Protocol (IGMP)