Debian Codegraph @

rfc2075.html

raw
1 <pre>Network Working Group C. Partridge
2 Request for Comments: 2075 BBN
3 Category: Experimental January 1997
4
5
6 <span class="h1">IP Echo Host Service</span>
7
8 Status of this Memo
9
10 This memo defines an Experimental Protocol for the Internet
11 community. This memo does not specify an Internet standard of any
12 kind. Discussion and suggestions for improvement are requested.
13 Distribution of this memo is unlimited.
14
15 Abstract
16
17 This memo describes how to implement an IP echo host. IP echo hosts
18 send back IP datagrams after exchanging the source and destination IP
19 addresses. The effect is that datagrams sent to the echo host are
20 sent back to the source, as if they originated at the echo host.
21
22 Introduction
23
24 An IP echo host returns IP datagrams to their original source host,
25 with the IP source and destination addresses reversed, so that the
26 returning datagram appears to be coming from the echo host to the
27 original source. IP echo hosts are tremendously useful for debugging
28 applications and protocols. They allow researchers to create looped
29 back conversations across the Internet, exposing their traffic to all
30 the vagaries of Internet behavior (congestion, cross traffic,
31 variable round-trip times and the like) without having to distribute
32 prototype software to a large number of test machines.
33
34 IP echo hosts were heavily used on the Internet in the late 1970s and
35 early 1980s to debug various Internet transport and application
36 protocols. But, for reasons unclear, at the current date there are
37 no echo hosts on the Internet and few people are even aware of the
38 concept. The goal of this memo is to document the concept in the
39 hopes it will be revived.
40
41 Implementation Details
42
43 While the basic idea of a echo host is simple, there are a few
44 implementation details that require attention. This section
45 describes those implementation details. The presentation works from
46 the simplest to most difficult issues.
47
48
49
50
51
52 <span class="grey">Partridge Experimental [Page 1]</span></pre>
53 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-2" ></span>
54 <span class="grey"><a href="./rfc2075">RFC 2075</a> IP Echo Host Service January 1997</span>
55
56
57 The most straightforward situation is when an echo host receives an
58 IP datagram with no options and whose protocol field has a value
59 other than 1 (ICMP). In this case, the echo host modifies the header
60 by exchanging the source and destination addresses, decrements the
61 TTL by one and updates the IP header checksum. The host then
62 transmits the updated IP datagram back to the original source of the
63 datagram.
64
65 NOTE: If the TTL is zero or less after decrementing, the datagram
66 MUST not be echoed. In general, an echo host is required to do
67 all the various sanity checks that a router or host would do to an
68 IP datagram before accepting the datagram for echoing (see STD 3,
69 <a href="./rfc1122">RFC 1122</a>, and <a href="./rfc1812">RFC 1812</a>).
70
71 The TTL MUST be decremented for security reasons noted below.
72 Observe, however, that the effect is that hosts using an echo path
73 through an echo host SHOULD set their TTL to twice the normal
74 value to be sure of achieving connectivity over the echo path.
75
76 If an arriving IP datagram has options, the echo host's
77 responsibilities are more complex. In general, the IP source and
78 destination are always exchanged and TTL and checksum updated, but in
79 certain situations, other special actions may have to take place.
80
81 If the datagram contains an incomplete source route option (i.e. the
82 echo host is not the final destination), the datagram MUST be
83 discarded. If the datagram contains a complete source route option,
84 the source route option MUST be reversed, and the datagram (with
85 source and destination IP addresses exchanged and updated TTL) MUST
86 be sent back along the reverse source route.
87
88 More generally, the goal with any option is to update the option such
89 that when the echoed packet is received at the original source, the
90 option fields will contain data which makes sense for a datagram
91 originating at the echo host.
92
93 There is one option for which it is unclear what the correct action.
94 The timestamp option is sometimes used for round-trip time
95 estimation. If the option is reset at the echo host, then a history
96 of roughly half of the trip delay will be lost. But if the option is
97 not reset, then the timestamp option will appear inconsistent with
98 the source and destination addresses of the datagram. To try to
99 balance these two issues, the following rules are suggested:
100
101 1. If the first entry in the timestamp option contains the IP
102 address of the source host, the entry SHOULD be rewritten to
103 contain the IP address of the echo host, and the timestamp option
104 pointer SHOULD be truncated so that this timestamp is the only one
105
106
107
108 <span class="grey">Partridge Experimental [Page 2]</span></pre>
109 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-3" ></span>
110 <span class="grey"><a href="./rfc2075">RFC 2075</a> IP Echo Host Service January 1997</span>
111
112
113 in the list. (This rewrite makes the option appear consistent
114 with the new source and destination IP addresses, and retains the
115 source timestamp, while losing information about the path to the
116 echo host).
117
118 2. If the first entry in the timestamp option does not contain the
119 IP address of the source host, the entry SHOULD be echoed back
120 unchanged. The echo host SHOULD NOT appear in the timestamp
121 option. (This approach retains the entire history of the path,
122 though observe that on a symmetric route, it means every router
123 may appear twice in the path).
124
125 Finally, if the IP datagram contains an ICMP packet (i.e. the IP
126 protocol field value is 1), the datagram SHOULD be discarded. The
127 reason for this rule is that the most likely reason for receiving an
128 ICMP datagram is that an echoed datagram has encountered a problem at
129 some router in the path and the router has sent back an ICMP
130 datagram. Echoing the ICMP datagram back to the router may confuse
131 the router and thus SHOULD be avoided. (This rule simply follows the
132 Internet maxim of being conservative in what we send).
133
134 However, in some cases the ICMP datagram will have useful information
135 for the source host which it would be desirable to echo. A
136 sophisticated echo host MAY choose to echo ICMP datagrams according
137 to the following rules:
138
139 1. Any ICMP datagram in which the destination address in the
140 encapsulated IP header (the header within the ICMP datagram)
141 matches the source address of the ICMP datagram MAY be safely
142 echoed.
143
144 2. ICMP Source Quench and ICMP Destination Unreachable with a code
145 of 4 (fragmentation needed and DF set) MAY be sent to the
146 *destination* of the encapsulated IP datagram if the source IP
147 address of the encapsulated IP datagram is that of the echo host.
148 When the ICMP message is sent on, it SHOULD be rewritten as an
149 ICMP message from the echo host to the source.
150
151 3. All other ICMP messages MUST be discarded.
152
153 These rules were chosen to try to ensure that end-to-end ICMP
154 messages are passed through, as are messages from routers which are
155 fairly safe and useful (or necessary) to the end system, but that
156 potentially dangerous messages such as Redirects are suppressed.
157 (The ICMP Destination Unreachable with code 4 is required for MTU
158 discovery under <a href="./rfc1191">RFC-1191</a>).
159
160
161
162
163
164 <span class="grey">Partridge Experimental [Page 3]</span></pre>
165 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-4" ></span>
166 <span class="grey"><a href="./rfc2075">RFC 2075</a> IP Echo Host Service January 1997</span>
167
168
169 Security Considerations
170
171 Echo hosts pose a number of security concerns related to address
172 spoofing.
173
174 First, echo hosts provide obvious ways to extend attacks that make
175 use of address spoofing. A malevolent host can write an third
176 party's IP address as the source address of a datagram sent to an
177 echo host and thus cause the echo host to send a datagram to the
178 third party. In general, this trick does not create a new security
179 hole (the malevolent host could just as well have sent the datagram
180 with a forged source address straight to the third party host). But
181 there are some new twists to the problem.
182
183 One exception is if the echo host is a host inside a firewall that
184 accepts datagrams from hosts outside the firewall. In that case, a
185 malevolent host outside the firewall may be able to use the echo host
186 to make its packets appear to originate from inside the firewall
187 (from the echo host). In general, a good firewall will catch these
188 cases (the source address of the datagrams sent to the echo host will
189 be for a host inside the firewall and testing for interior source
190 addresses on datagrams arriving at an exterior interface is a
191 standard firewall filter) but since the primary purpose of echo hosts
192 is for wide scale Internet testing, there seems no reason to invite
193 danger. So we recommend that echo hosts SHOULD NOT be placed inside
194 firewalls.
195
196 Second, address spoofing can be used to cause flooding of the
197 network. In this case, a malevolent host sends a datagram to an echo
198 host with the source address of another echo host. This trick will
199 cause datagrams to circulate between the two echo hosts. The
200 requirement that the echo host decrement the TTL by one ensures that
201 each datagram will eventually die, but a sufficiently malevolent host
202 sending a large number of datagrams with high TTLs to an echo host
203 can cause considerable disruption. There are a number of possible
204 ways to repair this problem (such as requiring sources to
205 authenticate themselves before sending datagrams to be echoed). A
206 simple protection is simply to limit the number of packets echoed
207 back to any one source per second. For instance, one might limit a
208 source to a packet rate equal to 10% of the interface bandwidth (for
209 a 10 Mb/s Ethernet this would be about 75 maximum sized packets per
210 second).
211
212 One variation of this attack is to generate e-mail addressed to the
213 echo host (e.g., [email protected]). This e-mail will loop over the
214 network a number of times until the SMTP server determines the
215 message has too many Received-From: lines.
216
217
218
219
220 <span class="grey">Partridge Experimental [Page 4]</span></pre>
221 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-5" ></span>
222 <span class="grey"><a href="./rfc2075">RFC 2075</a> IP Echo Host Service January 1997</span>
223
224
225 A third variation of the flooding trick is to place a multicast or
226 broadcast address as the source of the IP datagram sent to an echo
227 server. Since this results in an illegal arriving IP datagram, the
228 echo server MUST discard the datagram. (This warning serves as a
229 reminder that echo servers MUST do the standard checks for an illegal
230 datagram before echoing).
231
232 Implementation Note
233
234 Echo hosts are often implemented as virtual interfaces on an existing
235 host or router. One can think of the echo host's IP address as a
236 second IP address for the host, with the semantics that all datagrams
237 sent to that address get echoed. Observe that when an echo host is
238 supported as a module within a larger host implementation, an easy
239 implementation mistake to make is to accidentally put the non-echo
240 address of a host into an echoed packet. For a variety of reasons
241 (including security and correct operation of echo paths) implementors
242 MUST ensure this NEVER happens.
243
244 Acknowledgements
245
246 This memo was stimulated by a conversation with Jon Crowcroft in
247 which we both lamented the demise of some beloved IP echo hosts
248 (e.g., goonhilly-echo.arpa). It has been considerably improved by
249 comments from various members of the End2End-Interest mailing list,
250 including Bob Braden, Mark Handley, Christian Huitema, Dave Mills,
251 Tim Salo, Vern Schryver, Lansing Sloan, and Rich Stevens.
252
253 The author is emphatically not the inventor of echo hosts. Enquiries
254 to the usual suspects suggest that echo hosts were created by persons
255 unknown (probably at BBN) very early in the development of IP. I'd
256 like to thank those persons who created echo hosts and apologize for
257 any errors in describing their invention.
258
259 Author's Address
260
261 Craig Partridge
262 BBN Corporation
263 10 Moulton St
264 Cambridge MA 02138
265
266 EMail: [email protected]
267
268
269
270
271
272
273
274
275
276 Partridge Experimental [Page 5]
277 </pre>