Handling Fragmented And Unfragmented Traffic - Cisco Catalyst 3750 Software Configuration Manual

Chapter 25 Configuring Network Security with ACLs

Handling Fragmented and Unfragmented Traffic

IP packets can be fragmented as they cross the network. When this happens, only the fragment
containing the beginning of the packet contains the Layer 4 information, such as TCP or UDP port
numbers, ICMP type and code, and so on. All other fragments are missing this information.
Some ACEs do not check Layer 4 information and therefore can be applied to all packet fragments. ACEs
that do test Layer 4 information cannot be applied in the standard manner to most of the fragments in a
fragmented IP packet. When the fragment contains no Layer 4 information and the ACE tests some Layer
4 information, the matching rules are modified:
•
•
Consider access list 102, configured with these commands, applied to three fragmented packets:
Switch(config)# access-list 102 permit tcp any host 10.1.1.1 eq smtp
Switch(config)# access-list 102 deny tcp any host 10.1.1.2 eq telnet
Switch(config)# access-list 102 permit tcp any host 10.1.1.2
Switch(config)# access-list 102 deny tcp any any
Note In the first and second ACEs in the examples, the eq keyword after the destination address means to test
for the TCP-destination-port well-known numbers equaling Simple Mail Transfer Protocol (SMTP) and
Telnet, respectively.
•
•
•
78-15870-01
Permit ACEs that check the Layer 3 information in the fragment (including protocol type, such as
TCP, UDP, and so on) are considered to match the fragment regardless of what the missing Layer 4
information might have been.
Deny ACEs that check Layer 4 information never match a fragment unless the fragment contains
Layer 4 information.
Packet A is a TCP packet from host 10.2.2.2., port 65000, going to host 10.1.1.1 on the SMTP port.
If this packet is fragmented, the first fragment matches the first ACE (a permit) as if it were a
complete packet because all Layer 4 information is present. The remaining fragments also match the
first ACE, even though they do not contain the SMTP port information, because the first ACE only
checks Layer 3 information when applied to fragments. The information in this example is that the
packet is TCP and that the destination is 10.1.1.1.
Packet B is from host 10.2.2.2, port 65001, going to host 10.1.1.2 on the Telnet port. If this packet
is fragmented, the first fragment matches the second ACE (a deny) because all Layer 3 and Layer 4
information is present. The remaining fragments in the packet do not match the second ACE because
they are missing Layer 4 information. Instead, they match the third ACE (a permit).
Because the first fragment was denied, host 10.1.1.2 cannot reassemble a complete packet, so packet
B is effectively denied. However, the later fragments that are permitted will consume bandwidth on
the network and resources of host 10.1.1.2 as it tries to reassemble the packet.
Fragmented packet C is from host 10.2.2.2, port 65001, going to host 10.1.1.3, port ftp. If this packet
is fragmented, the first fragment matches the fourth ACE (a deny). All other fragments also match
the fourth ACE because that ACE does not check any Layer 4 information and because Layer 3
information in all fragments shows that they are being sent to host 10.1.1.3, and the earlier permit
ACEs were checking different hosts.
Catalyst 3750 Metro Switch Software Configuration Guide
Understanding ACLs
25-5