(October 11, 1995)
Daniel Z. Tabor Jr.
New Jersey Institute of Technology
Control and Error Messaging:
Physical-based networks can report errors via hardware components and signaling.
Software-based networks (the Internet) cannot and needs a means of error control.
To allow gateways to send error or control messages to other gateways or hosts, the
Internet Control Message Protocol (ICMP) was added.
Internet Control Message Protocol (ICMP):
ICMP is:
A required part of the IP protocol which must always be included.
Provides communication between IP software on two different machines (not just source and destination
machines).
Not restricted to gateways. It provides a single mechanism that is used for ALL control and information
messages.
Reports errors but does not correct them. The source node must take action to correct problems.
The source node may not be able to handle all the problems and must trust that remote operators can fix problems
of that type:
Gateways routing to incorrect locations
Gateways using corrupted routing tables.
ICMP is not considered a high-level protocol.
Since ICMP messages are encapsulated within IP datagrams, they:
Receive no additional reliability or priority.
May be lost or discarded.
Contribute to congestion.
ICMP error messages are NOT generated from datagrams carrying ICMP messages and in error.
ICMP Message Format:
Each has it’s own format, but all begin with the same three fields:
TYPE (8-bits):
Message type.
CODE (8-bits):
Further information on the message type.
CHECKSUM (16-bits):
Same checksum algorithm as used in the IP protocol, but only covers the ICMP datagram.
If ICMP message is reporting errors, it includes the entire header and first 64-bits of the IP datagram that
caused the problem.
Higher-layer protocols in TCP/IP are designed to encode crucial information in the first 64-bits of a
datagram/message for this reason.
Message Types:
Echo Request/Reply
Echo Request/Reply:
Serve as a debugging tool for networks, where echo requests are sent and echo replies are received from
destinations who are reachable and responding.
Optionally, sample data packets (56K) are sent and received by programs which implement the echo
request and reply ICMP messages (PING).
Echo Requests and Replies ensure that the major pieces of the transport system work, including:
IP software on the source host.
Intermediate gateways which route over the sending path.
The destination host is running, as are it’s IP and ICMP protocol modules.
Return path gateways work and their routing tables are correct.
TYPE (8-bits):
Request = 8
Reply = 0
SEQUENCE NUMBER (16-bits):
Used by the sender to match replies to requests sent.
Message Types:
Unreachable Destinations
Unreachable Destinations:
Used when a gateway cannot deliver an IP datagram (sent back to original source).
CODE (8-bits):
Integer value that further describes the problem which occurred.
Network Unreachable errors usually imply Routing Failures.
Host Unreachable errors usually imply Delivery Failures.
Failures may be caused by:
The destination being out-of-service.
A nonexistent destination address.
Gateways that do not have a route to the destination.
Although gateways send unreachable messages because of non-existent routes or delivery problems, not
all such errors can be detected.
Code Value
Meaning
0 1 2 3 5 6 7 12
Network Unreachable
Host Unreachable
Protocol Unreachable
Port Unreachable
Source Route Failed
Destination Network Unknown
Destination Host Unknown
Host Unreachable for Type of Service
Message Types:
Source Quench
Source-Quenching:
Used in the case of congested networks or for flow-control.
Congestion is a network issue (global to the whole network).
Flow-Control is an end-to-end issue and/or a point-to-point issue.
Source-Quench messages can be used with different methods of congestion control:
Monitoring of congestion conditions.
Source-Quench message for each datagram dropped.
Source-Quench sent when queues begin to fill (prior to congestion occurring).
Message Types:
Time Exceeded
Time Exceeded:
Because of routing cycles (in which several gateways constantly loop packets to each other):
The TTL field (hop count) will reach zero and the datagram is dropped.
Or they are dropped if a time-out occurs while waiting for fragments of the datagram to arrive.
A message is sent in either case.
CODE (8-bits):
0 = TTL count exceeded.
1 = Fragment reassembly time exceeded
The timer is started at destination after the first fragment is received.
Message Types:
Parameter Problem
Parameter Problem:
Sent when incorrect datagram header information is received or when other parameter problems occur.
Used in severe problems.
POINTER (8-bits):
Identifies the octet in the datagram that caused the problem.
Message Types:
Redirect Message
Redirect Message:
Sent to hosts requesting that they change their routes since the one they choose was non-optimal.
Gateways are assumed to know correct routes. Hosts begin with minimal routing information and update
their tables with new routes sent to them by default gateways.
Redirect:
Is a simple tool which permits hosts to boot knowing only one gateway IP address.
This does not solve the route propagation problem in a general way.
GATEWAY IP ADDRESS (32-bits):
Address of the gateway that the host should use to reach the destination mentioned in the datagram header
(INTERNET HEADER field in the next sequence of bits)
Gateways only send ICMP-Redirect messages to hosts and not to other gateways (Gateway to Gateway Protocol is
used for that).
Message Types:
Timestamp Message
Timestamp Request and Reply:
Simple technique for synchronizing clocks by requesting a machine’s current value for the time of day.
Timestamp values are given in milliseconds since midnight Universal Time.
TYPE (8-bits):
Request = 13
Reply = 14
IDENTIFIER and SEQ. NUMBER (16-bits each):
Both are used by the source to associate replies with requests sent.
Timestamp fields used to compute the time delay estimates:
ORIGINAL (32-bits):
Filled by the sender just before sending the timestamp request.
RECEIVE (32-bits):
Filled by the destination exactly when the request is received.
TRANSIT (32-bits):
Filled when the reply message is sent by the destination.
Delay estimates are not consistent over the Internet using Timestamp-ICMP messages.
Statistical analysis is needed for truly accurate measurement.
Message Types:
Address Mask Message
Address Mask Request and Reply:
Used to determine the subnet mask used at the destination.
Must know the subnet mask to participate in subnet addressing at either end.
TYPE (8-bits):
Request = 17
Reply = 18
ADDRESS MASK (32-bits):
Used to contain the 32-bit subnet mask used at the destination.
Debugging and Measurement Tools: PING
PING - utilizes the ICMP echo request and reply messages to test the accessibility and functionality of IP
equipment.
When equipment is pinged, a response is expected.
The round-trip time transmission times are displayed in milliseconds.
Some implementations give PING access to other IP datagram header features including:
Record Route
Timestamp
Loose and Strict Source Routing.
PING uses these features to gather statistics about network operations and test the integrity and
composition of a path to a remote node.
PING only tests as far up (in the conceptual layers) as the IP layer in the remote host.
PING can be used on a loopback address to test local TCP/IP stacks and network hardware equipment.
Debugging and Measurement Tools: Echo
Echo is a service normally provided in the application layer that will return characters received from a
remote TCP or UDP connection.
It provides an excellent confidence test that the destination is responding.
Since the Echo application resides (conceptually) above the IP layer, it can check the transport layer (whereas
PING cannot).
Debugging and Measurement Tools: Netstat
Netstat command provides information about the network interface and a means of examining various
aspects of the protocol stack’s functionality.
Netstat options:
netstat - i
Network drivers and interface cards are operational
Number of good or erred frames that have been encountered.
netstat - r
Option for displaying the router table contents.
netstat - a
Option allows the active server processes to be viewed, showing the server program waiting for incoming
connections, connections established, and IP address and port of the remote node.
