Debian Codegraph @

rfc6886.html

raw
1 <pre>Independent Submission S. Cheshire
2 Request for Comments: 6886 M. Krochmal
3 Category: Informational Apple Inc.
4 ISSN: 2070-1721 April 2013
5
6 <span class="h1">NAT Port Mapping Protocol (NAT-PMP)</span>
7
8 Abstract
9
10 This document describes a protocol for automating the process of
11 creating Network Address Translation (NAT) port mappings. Included
12 in the protocol is a method for retrieving the external IPv4 address
13 of a NAT gateway, thus allowing a client to make its external IPv4
14 address and port known to peers that may wish to communicate with it.
15 From 2005 onwards, this protocol was implemented in Apple products
16 including Mac OS X, Bonjour for Windows, and AirPort wireless base
17 stations. In 2013, NAT Port Mapping Protocol (NAT-PMP) was
18 superseded by the IETF Standards Track RFC "Port Control Protocol
19 (PCP)", which builds on NAT-PMP and uses a compatible packet format,
20 but adds a number of significant enhancements.
21
22 Status of This Memo
23
24 This document is not an Internet Standards Track specification; it is
25 published for informational purposes.
26
27 This is a contribution to the RFC Series, independently of any other
28 RFC stream. The RFC Editor has chosen to publish this document at
29 its discretion and makes no statement about its value for
30 implementation or deployment. Documents approved for publication by
31 the RFC Editor are not a candidate for any level of Internet
32 Standard; see <a href="./rfc5741#section-2">Section&nbsp;2 of RFC 5741</a>.
33
34 Information about the current status of this document, any errata,
35 and how to provide feedback on it may be obtained at
36 <a href="http://www.rfc-editor.org/info/rfc6886">http://www.rfc-editor.org/info/rfc6886</a>.
37
38 Copyright Notice
39
40 Copyright (c) 2013 IETF Trust and the persons identified as the
41 document authors. All rights reserved.
42
43 This document is subject to <a href="https://www.rfc-editor.org/bcp/bcp78">BCP 78</a> and the IETF Trust's Legal
44 Provisions Relating to IETF Documents
45 (<a href="http://trustee.ietf.org/license-info">http://trustee.ietf.org/license-info</a>) in effect on the date of
46 publication of this document. Please review these documents
47 carefully, as they describe your rights and restrictions with respect
48 to this document.
49
50
51
52 <span class="grey">Cheshire &amp; Krochmal Informational [Page 1]</span></pre>
53 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-2" ></span>
54 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
55
56
57 Table of Contents
58
59 <a href="#section-1">1</a>. Introduction ....................................................<a href="#page-3">3</a>
60 <a href="#section-1.1">1.1</a>. Transition to Port Control Protocol ........................<a href="#page-4">4</a>
61 <a href="#section-2">2</a>. Conventions and Terminology Used in This Document ...............<a href="#page-5">5</a>
62 <a href="#section-3">3</a>. Protocol and Packet Format ......................................<a href="#page-5">5</a>
63 <a href="#section-3.1">3.1</a>. Requests and Responses .....................................<a href="#page-6">6</a>
64 <a href="#section-3.2">3.2</a>. Determining the External Address ...........................<a href="#page-7">7</a>
65 <a href="#section-3.3">3.3</a>. Requesting a Mapping ......................................<a href="#page-10">10</a>
66 <a href="#section-3.4">3.4</a>. Destroying a Mapping ......................................<a href="#page-13">13</a>
67 <a href="#section-3.5">3.5</a>. Result Codes ..............................................<a href="#page-14">14</a>
68 <a href="#section-3.6">3.6</a>. Seconds Since Start of Epoch ..............................<a href="#page-16">16</a>
69 <a href="#section-3.7">3.7</a>. Recreating Mappings on NAT Gateway Reboot .................<a href="#page-16">16</a>
70 <a href="#section-3.8">3.8</a>. NAT Gateways with NAT Function Disabled ...................<a href="#page-18">18</a>
71 <a href="#section-3.9">3.9</a>. All Mappings Are Bidirectional ............................<a href="#page-19">19</a>
72 <a href="#section-4">4</a>. UNSAF Considerations ...........................................<a href="#page-20">20</a>
73 <a href="#section-4.1">4.1</a>. Scope .....................................................<a href="#page-20">20</a>
74 <a href="#section-4.2">4.2</a>. Transition Plan ...........................................<a href="#page-20">20</a>
75 <a href="#section-4.3">4.3</a>. Failure Cases .............................................<a href="#page-21">21</a>
76 <a href="#section-4.4">4.4</a>. Long-Term Solution ........................................<a href="#page-23">23</a>
77 <a href="#section-4.5">4.5</a>. Existing Deployed NATs ....................................<a href="#page-23">23</a>
78 <a href="#section-5">5</a>. Security Considerations ........................................<a href="#page-23">23</a>
79 <a href="#section-6">6</a>. IANA Considerations ............................................<a href="#page-24">24</a>
80 <a href="#section-7">7</a>. Acknowledgments ................................................<a href="#page-24">24</a>
81 <a href="#section-8">8</a>. Deployment History .............................................<a href="#page-25">25</a>
82 <a href="#section-9">9</a>. Noteworthy Features of NAT Port Mapping Protocol and PCP .......<a href="#page-26">26</a>
83 <a href="#section-9.1">9.1</a>. Simplicity ................................................<a href="#page-27">27</a>
84 <a href="#section-9.2">9.2</a>. Focused Scope .............................................<a href="#page-27">27</a>
85 <a href="#section-9.3">9.3</a>. Efficiency ................................................<a href="#page-27">27</a>
86 <a href="#section-9.4">9.4</a>. Atomic Allocation Operations ..............................<a href="#page-29">29</a>
87 <a href="#section-9.5">9.5</a>. Garbage Collection ........................................<a href="#page-29">29</a>
88 <a href="#section-9.6">9.6</a>. State Change Announcements ................................<a href="#page-30">30</a>
89 <a href="#section-9.7">9.7</a>. Soft State Recovery .......................................<a href="#page-31">31</a>
90 <a href="#section-9.8">9.8</a>. On-Path NAT Discovery .....................................<a href="#page-31">31</a>
91 <a href="#section-10">10</a>. References ....................................................<a href="#page-32">32</a>
92 <a href="#section-10.1">10.1</a>. Normative References .....................................<a href="#page-32">32</a>
93 <a href="#section-10.2">10.2</a>. Informative References ...................................<a href="#page-32">32</a>
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108 <span class="grey">Cheshire &amp; Krochmal Informational [Page 2]</span></pre>
109 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-3" ></span>
110 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
111
112
113 <span class="h2"><a class="selflink" id="section-1" href="#section-1">1</a>. Introduction</span>
114
115 Network Address Translation (NAT) is a method of sharing one public
116 Internet address with a number of devices. This document is focused
117 on devices that are formally classified as "NAPTs" (Network
118 Address/Port Translators) [<a href="./rfc2663" title="&quot;IP Network Address Translator (NAT) Terminology and Considerations&quot;">RFC2663</a>]. A full description of NAT is
119 beyond the scope of this document. The following brief overview will
120 cover the aspects relevant to this port mapping protocol. For more
121 information on NAT, see "Traditional IP Network Address Translator
122 (Traditional NAT)" [<a href="./rfc3022" title="&quot;Traditional IP Network Address Translator (Traditional NAT)&quot;">RFC3022</a>].
123
124 NATs have one or more external IP addresses. A private network is
125 set up behind the NAT. Client devices on that private network behind
126 the NAT are assigned private addresses, and those client devices use
127 the private address of the NAT device as their default gateway.
128
129 When a packet from any device behind the NAT is sent to an address on
130 the public Internet, the packet first passes through the NAT box.
131 The NAT box looks at the source port and address. In some cases, a
132 NAT will also keep track of the destination port and address. The
133 NAT then creates a mapping from the internal address and internal
134 port to an external address and external port if a mapping does not
135 already exist.
136
137 The NAT box replaces the internal address and port in the packet with
138 the external entries from the mapping and sends the packet on to the
139 next gateway.
140
141 When a packet from any address on the Internet is received on the
142 NAT's external side, the NAT will look up the destination address and
143 port (external address and port) in the list of mappings. If an
144 entry is found, it will contain the internal address and port to
145 which the packet should be sent. The NAT gateway will then rewrite
146 the destination address and port with those from the mapping. The
147 packet will then be forwarded to the new destination addresses. If
148 the packet did not match any mapping, the packet will most likely be
149 dropped. Various NATs implement different strategies to handle this.
150 The important thing to note is that if there is no mapping, the NAT
151 does not know to which internal address the packet should be sent.
152
153 Mappings are usually created automatically as a result of observing
154 outbound packets. There are a few exceptions. Some NATs may allow
155 manually created permanent mappings that map an external port to a
156 specific internal IP address and port. Such a mapping allows
157 incoming connections to the device with that internal address. Some
158 NATs also implement a default mapping where any inbound packet that
159
160
161
162
163
164 <span class="grey">Cheshire &amp; Krochmal Informational [Page 3]</span></pre>
165 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-4" ></span>
166 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
167
168
169 does not match any other more specific mapping will be forwarded to a
170 specified internal address. Both types of mappings are usually set
171 up manually through some configuration tool. Such manual
172 configuration of port mappings is unreasonably onerous for most
173 residential NAT users.
174
175 Without these manually created inbound port mappings, clients behind
176 the NAT would be unable to receive inbound connections, which
177 represents a loss of connectivity when compared to the original
178 Internet architecture [<a href="#ref-ETEAISD" title="&quot;End-to-end arguments in system design&quot;">ETEAISD</a>]. For those who view this loss of
179 connectivity as a bad thing, NAT-PMP allows clients to operate more
180 like a host directly connected to the unrestricted public Internet,
181 with an unrestricted public IPv4 address. NAT-PMP allows client
182 hosts to communicate with the NAT gateway to request the creation of
183 inbound mappings on demand. Having created a NAT mapping to allow
184 inbound connections, the client can then record its external IPv4
185 address and external port in a public registry (e.g., the worldwide
186 Domain Name System) or otherwise make it accessible to peers that
187 wish to communicate with it.
188
189 <span class="h3"><a class="selflink" id="section-1.1" href="#section-1.1">1.1</a>. Transition to Port Control Protocol</span>
190
191 NAT-PMP enjoyed almost a decade of useful service, and operational
192 experience with NAT-PMP informed the design of its IETF Standards
193 Track successor, Port Control Protocol (PCP) [<a href="./rfc6887" title="&quot;Port Control Protocol (PCP)&quot;">RFC6887</a>]. PCP builds
194 on NAT-PMP, using the same UDP ports 5350 and 5351, and a compatible
195 packet format. PCP also adds significant enhancements, including
196 IPv6 support, management of outbound mappings, management of firewall
197 rules, full compatibility with large-scale NATs with a pool of
198 external addresses, error lifetimes, and an extension mechanism to
199 enable future enhancements.
200
201 Because of the significant enhancements in PCP, all existing NAT-PMP
202 implementations are encouraged to migrate to PCP. The version number
203 in the packet header is 0 for NAT-PMP and 2 for PCP, so the packets
204 are easily distinguished. (Version number 1 was used by a vendor
205 that shipped products that use a protocol that is incompatible with
206 the IETF Standard. PCP implementations MUST NOT use version
207 number 1.)
208
209 For NAT-PMP servers, adding PCP support is simple. When packets are
210 received, if the version number is 2, then the packet is interpreted
211 as a PCP request, the request is handled, and replies and updates
212 pertaining to that mapping are sent using PCP format. If the version
213 number is 0, then the existing code handles the request exactly as it
214 already does, and replies and updates pertaining to that mapping are
215
216
217
218
219
220 <span class="grey">Cheshire &amp; Krochmal Informational [Page 4]</span></pre>
221 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-5" ></span>
222 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
223
224
225 sent using NAT-PMP format. If the version number is 1 or any other
226 unsupported version, then result code 1 (Unsupported Version) is
227 returned.
228
229 NAT-PMP clients should add PCP support, and should default to sending
230 requests using PCP format, which will cause clients to prefer the
231 newer format when possible. If a PCP request is sent to an old
232 NAT-PMP server that doesn't understand the new PCP format, then it
233 will return result code 1 (Unsupported Version), and the client
234 should then immediately retry the same request using NAT-PMP format.
235 The presence of the Unsupported Version reply allows fast fail-over
236 to NAT-PMP format, without waiting for timeouts, retransmissions, or
237 other arbitrary delays. It is recommended that clients always try
238 PCP first for every new request, retransmission, and renewal, and
239 only try NAT-PMP in response to an "Unsupported Version" error.
240 Clients SHOULD NOT record that a given server only speaks NAT-PMP and
241 subsequently default to NAT-PMP for that server, since NAT firmware
242 gets updated, and even a NAT gateway that speaks only NAT-PMP today,
243 may be updated to speak PCP tomorrow.
244
245 <span class="h2"><a class="selflink" id="section-2" href="#section-2">2</a>. Conventions and Terminology Used in This Document</span>
246
247 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
248 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
249 document are to be interpreted as described in "Key words for use in
250 RFCs to Indicate Requirement Levels" [<a href="./rfc2119" title="&quot;Key words for use in RFCs to Indicate Requirement Levels&quot;">RFC2119</a>].
251
252 <span class="h2"><a class="selflink" id="section-3" href="#section-3">3</a>. Protocol and Packet Format</span>
253
254 The NAT Port Mapping Protocol runs over UDP. Every packet starts
255 with an 8-bit version followed by an 8-bit operation code.
256
257 All numeric quantities in NAT-PMP larger than a single byte (e.g.,
258 error values, Seconds Since Start of Epoch, and mapping lifetime) are
259 transmitted in the traditional IETF network byte order (i.e., most
260 significant byte first).
261
262 Non-numeric quantities in NAT-PMP larger than a single byte (e.g.,
263 the NAT gateway's external IP address) are transmitted in the natural
264 byte order, with no byte swapping.
265
266 This document specifies version 0 of the protocol. Any NAT-PMP
267 gateway implementing this version of the protocol, receiving a
268 request with a version number other than 0, MUST return result code 1
269 (Unsupported Version), indicating the highest version number it does
270 support (i.e., 0) in the version field of the response.
271
272
273
274
275
276 <span class="grey">Cheshire &amp; Krochmal Informational [Page 5]</span></pre>
277 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-6" ></span>
278 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
279
280
281 Opcodes between 0 and 127 are client requests. Opcodes from 128 to
282 255 are corresponding server responses. Responses always contain a
283 16-bit result code in network byte order. A result code of zero
284 indicates success. Responses also contain a 32-bit unsigned integer
285 corresponding to the number of seconds since the NAT gateway was
286 rebooted or since its port mapping state was otherwise reset.
287
288 This protocol SHOULD only be used when the client determines that its
289 primary IPv4 address is in one of the private IPv4 address ranges
290 defined in "Address Allocation for Private Internets" [<a href="./rfc1918" title="&quot;Address Allocation for Private Internets&quot;">RFC1918</a>].
291 This includes the address ranges 10/8, 172.16/12, and 192.168/16.
292
293 Clients always send their NAT-PMP requests to their default gateway,
294 as learned via DHCP [<a href="./rfc2131" title="&quot;Dynamic Host Configuration Protocol&quot;">RFC2131</a>], or similar means. This protocol is
295 designed for small home networks, with a single logical link (subnet)
296 where the client's default gateway is also the NAT for that network.
297 For more complicated networks where the NAT is some device other than
298 the client's default gateway, this protocol is not appropriate.
299
300 <span class="h3"><a class="selflink" id="section-3.1" href="#section-3.1">3.1</a>. Requests and Responses</span>
301
302 NAT gateways are often low-cost devices, with limited memory and CPU
303 speed. For this reason, to avoid making excessive demands on the NAT
304 gateway, clients SHOULD NOT issue multiple concurrent requests. If a
305 client needs to perform multiple requests (e.g., on boot, wake from
306 sleep, network connection, etc.), it SHOULD queue them and issue them
307 serially, one at a time. Once the NAT gateway responds to one
308 request the client machine may issue the next. In the case of a fast
309 NAT gateway, the client may be able to complete requests at a rate of
310 hundreds per second. In the case of a slow NAT gateway that takes
311 perhaps half a second to respond to a NAT-PMP request, the client
312 SHOULD respect this and allow the NAT gateway to operate at the pace
313 it can manage, and not overload it by issuing requests faster than
314 the rate it's answering them.
315
316 To determine the external IPv4 address, or to request a port mapping,
317 a NAT-PMP client sends its request packet to port 5351 of its
318 configured gateway address, and waits 250 ms for a response. If no
319 NAT-PMP response is received from the gateway after 250 ms, the
320 client retransmits its request and waits 500 ms. The client SHOULD
321 repeat this process with the interval between attempts doubling each
322 time. If, after sending its ninth attempt (and then waiting for 64
323 seconds), the client has still received no response, then it SHOULD
324 conclude that this gateway does not support NAT Port Mapping Protocol
325 and MAY log an error message indicating this fact. In addition, if
326 the NAT-PMP client receives an "ICMP Port Unreachable" message from
327
328
329
330
331
332 <span class="grey">Cheshire &amp; Krochmal Informational [Page 6]</span></pre>
333 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-7" ></span>
334 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
335
336
337 the gateway for port 5351, then it can skip any remaining
338 retransmissions and conclude immediately that the gateway does not
339 support NAT-PMP.
340
341 As a performance optimization the client MAY record this information
342 and use it to suppress further attempts to use NAT-PMP, but the
343 client should not retain this information for too long. In
344 particular, any event that may indicate a potential change of gateway
345 or a change in gateway configuration (hardware link change
346 indication, change of gateway MAC address, acquisition of new DHCP
347 lease, receipt of NAT-PMP announcement packet from gateway, etc.)
348 should cause the client to discard its previous information regarding
349 the gateway's lack of NAT-PMP support, and send its next NAT-PMP
350 request packet normally.
351
352 When deleting a port mapping, the client uses the same initial 250 ms
353 timeout, doubling on each successive interval, except that clients
354 may choose not to try the full nine times before giving up. This is
355 because mapping deletion requests are in some sense advisory. They
356 are useful for efficiency, but not required for correctness; it is
357 always possible for client software to crash, or for power to fail,
358 or for a client device to be physically unplugged from the network
359 before it gets a chance to send its mapping deletion request(s), so
360 NAT gateways already need to cope with this case. Because of this,
361 it may be acceptable for a client to retry only once or twice before
362 giving up on deleting its port mapping(s), but a client SHOULD always
363 send at least one deletion request whenever possible, to reduce the
364 amount of stale state that accumulates on NAT gateways.
365
366 A client need not continue trying to delete a port mapping after the
367 time when that mapping would naturally have expired anyway.
368
369 <span class="h3"><a class="selflink" id="section-3.2" href="#section-3.2">3.2</a>. Determining the External Address</span>
370
371 To determine the external address, the client behind the NAT sends
372 the following UDP payload to port 5351 of the configured gateway
373 address:
374
375 0 1
376 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5
377 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
378 | Vers = 0 | OP = 0 |
379 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
380
381
382
383
384
385
386
387
388 <span class="grey">Cheshire &amp; Krochmal Informational [Page 7]</span></pre>
389 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-8" ></span>
390 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
391
392
393 A compatible NAT gateway MUST generate a response with the following
394 format:
395
396 0 1 2 3
397 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
398 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
399 | Vers = 0 | OP = 128 + 0 | Result Code (net byte order) |
400 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
401 | Seconds Since Start of Epoch (in network byte order) |
402 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
403 | External IPv4 Address (a.b.c.d) |
404 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
405
406 This response indicates that the NAT gateway implements this version
407 of the protocol, and returns the external IPv4 address of the NAT
408 gateway. If the result code is non-zero, the value of the External
409 IPv4 Address field is undefined (MUST be set to zero on transmission,
410 and MUST be ignored on reception).
411
412 The NAT gateway MUST fill in the Seconds Since Start of Epoch field
413 with the time elapsed since its port mapping table was initialized on
414 startup, or reset for any other reason (see <a href="#section-3.6">Section 3.6</a>, "Seconds
415 Since Start of Epoch").
416
417 Upon receiving a response packet, the client MUST check the source IP
418 address, and silently discard the packet if the address is not the
419 address of the gateway to which the request was sent.
420
421 <span class="h4"><a class="selflink" id="section-3.2.1" href="#section-3.2.1">3.2.1</a>. Announcing Address Changes</span>
422
423 Upon boot, acquisition of an external IPv4 address, subsequent change
424 of the external IPv4 address, reboot, or any other event that may
425 indicate possible loss or change of NAT mapping state, the NAT
426 gateway MUST send a gratuitous response to the link-local multicast
427 address 224.0.0.1, port 5350, with the packet format above, to notify
428 clients of the external IPv4 address and Seconds Since Start of
429 Epoch.
430
431 To accommodate packet loss, the NAT gateway SHOULD multicast 10
432 address notifications. The interval between the first two
433 notifications SHOULD be 250 ms, and the interval between each
434 subsequent notification SHOULD double. The Seconds Since Start of
435 Epoch field in each transmission MUST be updated appropriately to
436 reflect the passage of time, so as not to trigger unnecessary
437 additional mapping renewals (see <a href="#section-3.7">Section 3.7</a>, "Recreating Mappings on
438 NAT Gateway Reboot").
439
440
441
442
443
444 <span class="grey">Cheshire &amp; Krochmal Informational [Page 8]</span></pre>
445 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-9" ></span>
446 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
447
448
449 Upon receiving a gratuitous address announcement packet, the client
450 MUST check the source IP address, and silently discard the packet if
451 the address is not the address of the client's current configured
452 gateway. This is to guard against inadvertent misconfigurations
453 where there may be more than one NAT gateway active on the network.
454
455 If the source IP address is correct, then the Seconds Since Start of
456 Epoch field is checked as described in <a href="#section-3.6">Section 3.6</a>, and if the value
457 is outside the expected plausible range, indicating that a NAT
458 gateway state loss has occurred, then the NAT-PMP client promptly
459 recreates all its active port mapping leases, as described in <a href="#section-3.7">Section</a>
460 <a href="#section-3.7">3.7</a>, "Recreating Mappings on NAT Gateway Reboot".
461
462 IMPLEMENTATION NOTE: Earlier implementations of NAT-PMP used UDP port
463 5351 as the destination both for client requests (address and port
464 mapping) and for address announcements. NAT-PMP servers would listen
465 on UDP 5351 for client requests, and NAT-PMP clients would listen on
466 UDP 5351 for server announcements. However, implementers encountered
467 difficulties when a single device is acting in both roles, for
468 example, a home computer with Internet Sharing enabled. This
469 computer is acting in the role of NAT-PMP server to its DHCP clients,
470 yet, at the same time, it has to act in the role of NAT-PMP client in
471 order to determine whether it is, itself, behind another NAT gateway.
472 While in principle it might be possible on some operating systems for
473 two processes to coordinate sharing of a single UDP port, on many
474 platforms this is difficult or even impossible, so, for pragmatic
475 engineering reasons, it is convenient to have clients listen on UDP
476 5350 and servers listen on UDP 5351.
477
478 IMPLEMENTATION NOTE: A given host may have more than one independent
479 NAT-PMP client running at the same time, and address announcements
480 need to be available to all of them. Clients should therefore set
481 the SO_REUSEPORT option or equivalent in order to allow other
482 processes to also listen on port 5350. Additionally, implementers
483 have encountered issues when one or more processes on the same device
484 listen to port 5350 on *all* addresses. Clients should therefore
485 bind specifically to 224.0.0.1:5350, not to 0.0.0.0:5350.
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500 <span class="grey">Cheshire &amp; Krochmal Informational [Page 9]</span></pre>
501 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-10" ></span>
502 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
503
504
505 <span class="h3"><a class="selflink" id="section-3.3" href="#section-3.3">3.3</a>. Requesting a Mapping</span>
506
507 To create a mapping, the client sends a UDP packet to port 5351 of
508 the gateway's internal IP address with the following format:
509
510 0 1 2 3
511 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
512 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
513 | Vers = 0 | OP = x | Reserved |
514 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
515 | Internal Port | Suggested External Port |
516 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
517 | Requested Port Mapping Lifetime in Seconds |
518 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
519
520 Opcodes supported:
521 1 - Map UDP
522 2 - Map TCP
523
524 The Reserved field MUST be set to zero on transmission and MUST be
525 ignored on reception.
526
527 The Ports and Lifetime are transmitted in the traditional network
528 byte order (i.e., most significant byte first).
529
530 The Internal Port is set to the local port on which the client is
531 listening.
532
533 If the client would prefer to have a high-numbered "anonymous"
534 external port assigned, then it should set the Suggested External
535 Port to zero, which indicates to the gateway that it should allocate
536 a high-numbered port of its choosing. If the client would prefer
537 instead to have the mapped external port be the same as its local
538 internal port if possible (e.g., a web server listening on port 80
539 that would ideally like to have external port 80), then it should set
540 the Suggested External Port to the desired value. However, the
541 gateway is not obliged to assign the port suggested, and may choose
542 not to, either for policy reasons (e.g., port 80 is reserved and
543 clients may not request it) or because that port has already been
544 assigned to some other client. Because of this, some product
545 developers have questioned the value of having the Suggested External
546 Port field at all. The reason is for failure recovery. Most low-
547 cost home NAT gateways do not record temporary port mappings in
548 persistent storage, so if the gateway crashes or is rebooted, all the
549 mappings are lost. A renewal packet is formatted identically to an
550 initial mapping request packet, except that for renewals the client
551 sets the Suggested External Port field to the port the gateway
552 actually assigned, rather than the port the client originally wanted.
553
554
555
556 <span class="grey">Cheshire &amp; Krochmal Informational [Page 10]</span></pre>
557 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-11" ></span>
558 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
559
560
561 When a freshly rebooted NAT gateway receives a renewal packet from a
562 client, it appears to the gateway just like an ordinary initial
563 request for a port mapping, except that in this case the Suggested
564 External Port is likely to be one that the NAT gateway *is* willing
565 to allocate (it allocated it to this client right before the reboot,
566 so it should presumably be willing to allocate it again). This
567 improves the stability of external ports across NAT gateway restarts.
568
569 The RECOMMENDED Port Mapping Lifetime is 7200 seconds (two hours).
570
571 After sending the port mapping request, the client then waits for the
572 NAT gateway to respond. If after 250 ms the client hasn't received a
573 response from the gateway, the client SHOULD reissue its request as
574 described above in <a href="#section-3.1">Section 3.1</a>, "Requests and Responses".
575
576 The NAT gateway responds with the following packet format:
577
578 0 1 2 3
579 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
580 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
581 | Vers = 0 | OP = 128 + x | Result Code |
582 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
583 | Seconds Since Start of Epoch |
584 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
585 | Internal Port | Mapped External Port |
586 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
587 | Port Mapping Lifetime in Seconds |
588 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
589
590 The epoch time, ports, and lifetime are transmitted in the
591 traditional network byte order (i.e., most significant byte first).
592
593 The 'x' in the OP field MUST match what the client requested. Some
594 NAT gateways are incapable of creating a UDP port mapping without
595 also creating a corresponding TCP port mapping, and vice versa, and
596 these gateways MUST NOT implement NAT Port Mapping Protocol until
597 this deficiency is fixed. A NAT gateway that implements this
598 protocol MUST be able to create TCP-only and UDP-only port mappings.
599 If a NAT gateway silently creates a pair of mappings for a client
600 that only requested one mapping, then it may expose that client to
601 receiving inbound UDP packets or inbound TCP connection requests that
602 it did not ask for and does not want.
603
604 While a NAT gateway MUST NOT automatically create mappings for TCP
605 when the client requests UDP, and vice versa, the NAT gateway MUST
606 reserve the companion port so the same client can choose to map it in
607 the future. For example, if a client requests to map TCP port 80,
608
609
610
611
612 <span class="grey">Cheshire &amp; Krochmal Informational [Page 11]</span></pre>
613 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-12" ></span>
614 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
615
616
617 as long as the client maintains the lease for that TCP port mapping,
618 another client with a different internal IP address MUST NOT be able
619 to successfully acquire the mapping for UDP port 80.
620
621 The client normally requests the external port matching the internal
622 port. If that external port is not available, the NAT gateway MUST
623 return an available external port if possible, or return an error
624 code if no external ports are available.
625
626 The source address of the packet MUST be used for the internal
627 address in the mapping. This protocol is not intended to facilitate
628 one device behind a NAT creating mappings for other devices. If
629 there are legacy devices that require inbound mappings, permanent
630 mappings can be created manually by the user through an
631 administrative interface, just as they are today.
632
633 If a mapping already exists for a given internal address and port
634 (whether that mapping was created explicitly using NAT-PMP,
635 implicitly as a result of an outgoing TCP SYN packet, or manually by
636 a human administrator) and that client requests another mapping for
637 the same internal port (possibly requesting a different external
638 port), then the mapping request should succeed, returning the
639 already-assigned external port. This is necessary to handle the case
640 where a client requests a mapping with suggested external port X, and
641 is granted a mapping with actual external port Y, but the
642 acknowledgment packet gets lost. When the client retransmits its
643 mapping request, it should get back the same positive acknowledgment
644 as was sent (and lost) the first time.
645
646 The NAT gateway MUST NOT accept mapping requests destined to the NAT
647 gateway's external IP address or received on its external network
648 interface. Only packets received on the internal interface(s) with a
649 destination address matching the internal address(es) of the NAT
650 gateway should be allowed.
651
652 The NAT gateway MUST fill in the Seconds Since Start of Epoch field
653 with the time elapsed since its port mapping table was initialized on
654 startup or reset for any other reason (see <a href="#section-3.6">Section 3.6</a>, "Seconds
655 Since Start of Epoch").
656
657 The Port Mapping Lifetime is an unsigned integer in seconds. The NAT
658 gateway MAY reduce the lifetime from what the client requested. The
659 NAT gateway SHOULD NOT offer a lease lifetime greater than that
660 requested by the client.
661
662 Upon receiving the response packet, the client MUST check the source
663 IP address, and silently discard the packet if the address is not the
664 address of the gateway to which the request was sent.
665
666
667
668 <span class="grey">Cheshire &amp; Krochmal Informational [Page 12]</span></pre>
669 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-13" ></span>
670 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
671
672
673 The client SHOULD begin trying to renew the mapping halfway to expiry
674 time, like DHCP. The renewal packet should look exactly the same as
675 a request packet, except that the client SHOULD set the Suggested
676 External Port to what the NAT gateway previously mapped, not what the
677 client originally suggested. As described above, this enables the
678 gateway to automatically recover its mapping state after a crash or
679 reboot.
680
681 <span class="h3"><a class="selflink" id="section-3.4" href="#section-3.4">3.4</a>. Destroying a Mapping</span>
682
683 A mapping may be destroyed in a variety of ways. If a client fails
684 to renew a mapping, then at the time its lifetime expires, the
685 mapping MUST be automatically deleted. In the common case where the
686 gateway device is a combined DHCP server and NAT gateway, when a
687 client's DHCP address lease expires, the gateway device MAY
688 automatically delete any mappings belonging to that client.
689 Otherwise, a new client being assigned the same IP address could
690 receive unexpected inbound UDP packets or inbound TCP connection
691 requests that it did not ask for and does not want.
692
693 A client MAY also send an explicit packet to request deletion of a
694 mapping that is no longer needed. A client requests explicit
695 deletion of a mapping by sending a message to the NAT gateway
696 requesting the mapping, with the Requested Lifetime in Seconds set to
697 zero. The Suggested External Port MUST be set to zero by the client
698 on sending, and MUST be ignored by the gateway on reception.
699
700 When a mapping is destroyed successfully as a result of the client
701 explicitly requesting the deletion, the NAT gateway MUST send a
702 response packet that is formatted as defined in <a href="#section-3.3">Section 3.3</a>,
703 "Requesting a Mapping". The response MUST contain a result code of
704 0, the internal port as indicated in the deletion request, an
705 external port of 0, and a lifetime of 0. The NAT gateway MUST
706 respond to a request to destroy a mapping that does not exist as if
707 the request were successful. This is because of the case where the
708 acknowledgment is lost, and the client retransmits its request to
709 delete the mapping. In this case, the second request to delete the
710 mapping MUST return the same response packet as the first request.
711
712 If the deletion request was unsuccessful, the response MUST contain a
713 non-zero result code and the requested mapping; the lifetime is
714 undefined (MUST be set to zero on transmission, and MUST be ignored
715 on reception). If the client attempts to delete a port mapping that
716 was manually assigned by some kind of configuration tool, the NAT
717 gateway MUST respond with a "Not Authorized" error, result code 2.
718
719
720
721
722
723
724 <span class="grey">Cheshire &amp; Krochmal Informational [Page 13]</span></pre>
725 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-14" ></span>
726 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
727
728
729 When a mapping is destroyed as a result of its lifetime expiring or
730 for any other reason, if the NAT gateway's internal state indicates
731 that there are still active TCP connections traversing that now-
732 defunct mapping, then the NAT gateway SHOULD send appropriately
733 constructed TCP RST (reset) packets both to the local client and to
734 the remote peer on the Internet to terminate that TCP connection.
735
736 A client can request the explicit deletion of all its UDP or TCP
737 mappings by sending the same deletion request to the NAT gateway with
738 the external port, internal port, and lifetime set to zero. A client
739 MAY choose to do this when it first acquires a new IP address in
740 order to protect itself from port mappings that were performed by a
741 previous owner of the IP address. After receiving such a deletion
742 request, the gateway MUST delete all its UDP or TCP port mappings
743 (depending on the opcode). The gateway responds to such a deletion
744 request with a response as described above, with the internal port
745 set to zero. If the gateway is unable to delete a port mapping, for
746 example, because the mapping was manually configured by the
747 administrator, the gateway MUST still delete as many port mappings as
748 possible, but respond with a non-zero result code. The exact result
749 code to return depends on the cause of the failure. If the gateway
750 is able to successfully delete all port mappings as requested, it
751 MUST respond with a result code of zero.
752
753 <span class="h3"><a class="selflink" id="section-3.5" href="#section-3.5">3.5</a>. Result Codes</span>
754
755 Currently defined result codes:
756
757 0 - Success
758 1 - Unsupported Version
759 2 - Not Authorized/Refused
760 (e.g., box supports mapping, but user has turned feature off)
761 3 - Network Failure
762 (e.g., NAT box itself has not obtained a DHCP lease)
763 4 - Out of resources
764 (NAT box cannot create any more mappings at this time)
765 5 - Unsupported opcode
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780 <span class="grey">Cheshire &amp; Krochmal Informational [Page 14]</span></pre>
781 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-15" ></span>
782 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
783
784
785 If the version in the request is not zero, then the NAT-PMP server
786 MUST return the following "Unsupported Version" error response to the
787 client:
788
789 0 1 2 3
790 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
791 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
792 | Vers = 0 | OP = 0 | Result Code = 1 |
793 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
794 | Seconds Since Start of Epoch (in network byte order) |
795 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
796
797 If the opcode in the request is 128 or greater, then this is not a
798 request; it's a response, and the NAT-PMP server MUST silently ignore
799 it. Otherwise, if the opcode in the request is less than 128, but is
800 not a supported opcode (currently 0, 1, or 2), then the entire
801 request MUST be returned to the sender, with the top bit of the
802 opcode set (to indicate that this is a response) and the result code
803 set to 5 (Unsupported opcode).
804
805 For version 0 and a supported opcode (0, 1, or 2), if the operation
806 fails for some reason (Not Authorized, Network Failure, or Out of
807 resources), then a valid response MUST be sent to the client, with
808 the top bit of the opcode set (to indicate that this is a response)
809 and the result code set appropriately. Other fields in the response
810 MUST be set appropriately. Specifically:
811
812 o Seconds Since Start of Epoch MUST be set correctly
813
814 o The External IPv4 Address should be set to the correct address, or
815 to 0.0.0.0, as appropriate.
816
817 o The Internal Port MUST be set to the client's requested Internal
818 Port. This is particularly important, because the client needs
819 this information to identify which request suffered the failure.
820
821 o The Mapped External Port and Port Mapping Lifetime MUST be set
822 appropriately -- i.e., zero if no successful port mapping was
823 created.
824
825 Should future NAT-PMP opcodes be defined, their error responses MUST
826 similarly be specified to include sufficient information to identify
827 which request suffered the failure. By design, NAT-PMP messages do
828 not contain any transaction identifiers. All NAT-PMP messages are
829 idempotent and self-describing; therefore, the specifications of
830 future NAT-PMP messages need to include enough information for their
831 responses to be self-describing.
832
833
834
835
836 <span class="grey">Cheshire &amp; Krochmal Informational [Page 15]</span></pre>
837 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-16" ></span>
838 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
839
840
841 Clients MUST be able to properly handle result codes not defined in
842 this document. Undefined results codes MUST be treated as fatal
843 errors of the request.
844
845 <span class="h3"><a class="selflink" id="section-3.6" href="#section-3.6">3.6</a>. Seconds Since Start of Epoch</span>
846
847 Every packet sent by the NAT gateway includes a Seconds Since Start
848 of Epoch (SSSoE) field. If the NAT gateway resets or loses the state
849 of its port mapping table, due to reboot, power failure, or any other
850 reason, it MUST reset its epoch time and begin counting SSSoE from
851 zero again. Whenever a client receives any packet from the NAT
852 gateway, either unsolicited or in response to a client request, the
853 client computes its own conservative estimate of the expected SSSoE
854 value by taking the SSSoE value in the last packet it received from
855 the gateway and adding 7/8 (87.5%) of the time elapsed according to
856 the client's local clock since that packet was received. If the
857 SSSoE in the newly received packet is less than the client's
858 conservative estimate by more than 2 seconds, then the client
859 concludes that the NAT gateway has undergone a reboot or other loss
860 of port mapping state, and the client MUST immediately renew all its
861 active port mapping leases as described in <a href="#section-3.7">Section 3.7</a>, "Recreating
862 Mappings on NAT Gateway Reboot".
863
864 <span class="h3"><a class="selflink" id="section-3.7" href="#section-3.7">3.7</a>. Recreating Mappings on NAT Gateway Reboot</span>
865
866 The NAT gateway MAY store mappings in persistent storage so that,
867 when it is powered off or rebooted, it remembers the port mapping
868 state of the network.
869
870 However, maintaining this state is not essential for correct
871 operation. When the NAT gateway powers on or clears its port mapping
872 state as the result of a configuration change, it MUST reset the
873 epoch time and re-announce its IPv4 address as described in <a href="#section-3.2.1">Section</a>
874 <a href="#section-3.2.1">3.2.1</a>, "Announcing Address Changes". Reception of this packet where
875 time has apparently gone backwards serves as a hint to clients on the
876 network that they SHOULD immediately send renewal packets (to
877 immediately recreate their mappings) instead of waiting until the
878 originally scheduled time for those renewals. Clients who miss
879 receiving those gateway announcement packets for any reason will
880 still renew their mappings at the originally scheduled time and cause
881 their mappings to be recreated; it will just take a little longer for
882 these clients.
883
884
885
886
887
888
889
890
891
892 <span class="grey">Cheshire &amp; Krochmal Informational [Page 16]</span></pre>
893 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-17" ></span>
894 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
895
896
897 A mapping renewal packet is formatted identically to an original
898 mapping request; from the point of view of the client, it is a
899 renewal of an existing mapping, but from the point of view of the
900 freshly rebooted NAT gateway, it appears as a new mapping request.
901
902 This self-healing property of the protocol is very important.
903
904 The remarkable reliability of the Internet as a whole derives in
905 large part from the fact that important state is held in the
906 endpoints, not in the network itself [<a href="#ref-ETEAISD" title="&quot;End-to-end arguments in system design&quot;">ETEAISD</a>]. Power-cycling an
907 Ethernet switch results only in a brief interruption in the flow of
908 packets; established TCP connections through that switch are not
909 broken, merely delayed for a few seconds. Indeed, a failing Ethernet
910 switch can even be replaced with a new one, and as long as the cables
911 are transferred over reasonably quickly, after the upgrade all the
912 TCP connections that were previously going through the old switch
913 will be unbroken and now going through the new one. The same is true
914 of IP routers, wireless base stations, etc. The one exception is NAT
915 gateways. Because the port mapping state is required for the NAT
916 gateway to know where to forward inbound packets, loss of that state
917 breaks connectivity through the NAT gateway. By allowing clients to
918 detect when this loss of NAT gateway state has occurred, and recreate
919 it on demand, we turn hard state in the network into soft state, and
920 allow it to be recovered automatically when needed.
921
922 Without this automatic recreation of soft state in the NAT gateway,
923 reliable long-term networking would not be achieved. As mentioned
924 above, the reliability of the Internet does not come from trying to
925 build a perfect network in which errors never happen, but from
926 accepting that in any sufficiently large system there will always be
927 some component somewhere that's failing, and designing mechanisms
928 that can handle those failures and recover. To illustrate this point
929 with an example, consider the following scenario: Imagine a network
930 security camera that has a web interface and accepts incoming
931 connections from web browser clients. Imagine this network security
932 camera uses NAT-PMP or a similar protocol to set up an inbound port
933 mapping in the NAT gateway so that it can receive incoming
934 connections from clients on the other side of the NAT gateway. Now,
935 this camera may well operate for weeks, months, or even years.
936 During that time, it's possible that the NAT gateway could experience
937 a power failure or be rebooted. The user could upgrade the NAT
938 gateway's firmware, or even replace the entire NAT gateway device
939 with a newer model. The general point is that if the camera operates
940 for a long enough period of time, some kind of disruption to the NAT
941 gateway becomes inevitable. The question is not whether the NAT
942 gateway will lose its port mappings, but when, and how often. If the
943 network camera and devices like it on the network can detect when the
944 NAT gateway has lost its port mappings, and recreate them
945
946
947
948 <span class="grey">Cheshire &amp; Krochmal Informational [Page 17]</span></pre>
949 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-18" ></span>
950 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
951
952
953 automatically, then these disruptions are self-correcting and largely
954 invisible to the end user. If, on the other hand, the disruptions
955 are not self-correcting, and after a NAT gateway reboot the user has
956 to manually reset or reboot all the other devices on the network too,
957 then these disruptions are *very* visible to the end user. This
958 aspect of the design is part of what makes the difference between a
959 protocol that keeps on working indefinitely over a time scale of
960 months or years, and a protocol that works in brief testing, but in
961 the real world is continually failing and requiring manual
962 intervention to get it going again.
963
964 When a client renews its port mappings as the result of receiving a
965 packet where the Seconds Since Start of Epoch (SSSoE) field indicates
966 that a reboot or similar loss of state has occurred, the client MUST
967 first delay by a random amount of time selected with uniform random
968 distribution in the range 0 to 5 seconds, and then send its first
969 port mapping request. After that request is acknowledged by the
970 gateway, the client may then send its second request, and so on, as
971 rapidly as the gateway allows. The requests SHOULD be issued
972 serially, one at a time; the client SHOULD NOT issue multiple
973 concurrent requests.
974
975 The discussion in this section focuses on recreating inbound port
976 mappings after loss of NAT gateway state, because that is the more
977 serious problem. Losing port mappings for outgoing connections
978 destroys those currently active connections, but does not prevent
979 clients from establishing new outgoing connections. In contrast,
980 losing inbound port mappings not only destroys all existing inbound
981 connections, but also prevents the reception of any new inbound
982 connections until the port mapping is recreated. Accordingly, we
983 consider recovery of inbound port mappings more important. However,
984 clients that want outgoing connections to survive a NAT gateway
985 reboot can also achieve that using NAT-PMP, in the common case of a
986 residential NAT gateway with a single, relatively stable, external IP
987 address. After initiating an outbound TCP connection (which will
988 cause the NAT gateway to establish an implicit port mapping), the
989 client should send the NAT gateway a port mapping request for the
990 source port of its TCP connection, which will cause the NAT gateway
991 to send a response giving the external port it allocated for that
992 mapping. The client can then store this information, and use it
993 later to recreate the mapping if it determines that the NAT gateway
994 has lost its mapping state.
995
996 <span class="h3"><a class="selflink" id="section-3.8" href="#section-3.8">3.8</a>. NAT Gateways with NAT Function Disabled</span>
997
998 Note that only devices that are *currently* acting in the role of NAT
999 gateway should participate in NAT-PMP protocol exchanges with
1000 clients. A network device that is capable of NAT (and NAT-PMP) but
1001
1002
1003
1004 <span class="grey">Cheshire &amp; Krochmal Informational [Page 18]</span></pre>
1005 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-19" ></span>
1006 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
1007
1008
1009 is currently configured not to perform that function (e.g., it is
1010 acting as a traditional IP router, forwarding packets without
1011 modifying them) MUST NOT respond to NAT-PMP requests from clients nor
1012 send spontaneous NAT-PMP address-change announcements.
1013
1014 In particular, a network device not currently acting in the role of
1015 NAT gateway should not even respond to NAT-PMP requests by returning
1016 an error code such as 2, "Not Authorized/Refused", since to do so is
1017 misleading to clients -- it suggests that NAT port mapping is
1018 necessary on this network for the client to successfully receive
1019 inbound connections, but is not available because the administrator
1020 has chosen to disable that functionality.
1021
1022 Clients should also be careful to avoid making unfounded assumptions,
1023 such as the assumption that if the client has an IPv4 address in one
1024 of the private IPv4 address ranges [<a href="./rfc1918" title="&quot;Address Allocation for Private Internets&quot;">RFC1918</a>], then that means NAT
1025 necessarily must be in use. Net 10/8 has enough addresses to build a
1026 private network with millions of hosts and thousands of
1027 interconnected subnets, all without any use of NAT. Many
1028 organizations have built such private networks that benefit from
1029 using standard TCP/IP technology, but by choice do not connect to the
1030 public Internet. The purpose of NAT-PMP is to mitigate some of the
1031 damage caused by NAT. It would be an ironic and unwanted side effect
1032 of this protocol if it were to lead well-meaning but misguided
1033 developers to create products that refuse to work on a private
1034 network *unless* they can find a NAT gateway to talk to.
1035 Consequently, a client finding that NAT-PMP is not available on its
1036 network should not give up, but should proceed on the assumption that
1037 the network may be a traditional routed IP network, with no address
1038 translation being used. This assumption may not always be true, but
1039 it is better than the alternative of falsely assuming the worst and
1040 not even trying to use normal (non-NAT) IP networking.
1041
1042 If a network device not currently acting in the role of NAT gateway
1043 receives UDP packets addressed to port 5351, it SHOULD respond
1044 immediately with an "ICMP Port Unreachable" message to tell the
1045 client that it needn't continue with timeouts and retransmissions,
1046 and it should assume that NAT-PMP is not available and not needed on
1047 this network. Typically, this behavior can be achieved merely by not
1048 having an open socket listening on UDP port 5351.
1049
1050 <span class="h3"><a class="selflink" id="section-3.9" href="#section-3.9">3.9</a>. All Mappings Are Bidirectional</span>
1051
1052 All NAT mappings, whether created implicitly by an outbound packet,
1053 created explicitly using NAT-PMP, or configured statically, are
1054 bidirectional. This means that when an outbound packet from a
1055 particular internal address and port is translated to an external
1056
1057
1058
1059
1060 <span class="grey">Cheshire &amp; Krochmal Informational [Page 19]</span></pre>
1061 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-20" ></span>
1062 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
1063
1064
1065 address and port, replies addressed to that external address and port
1066 need to be translated back to the corresponding internal address and
1067 port.
1068
1069 The converse is also true. When an inbound packet is received that
1070 is addressed to an external address and port that matches an existing
1071 mapping (implicit, explicit, or static), it is translated to the
1072 corresponding internal address and port and forwarded. Outbound
1073 replies from that internal address and port need to be translated to
1074 the correct external address and port so that they are correctly
1075 recognized by the remote peer.
1076
1077 In particular, if an outbound UDP reply that matches an existing
1078 explicit or static mapping is instead treated like a "new" outbound
1079 UDP packet, and a new dynamic mapping is created (with a different
1080 external address and port), then at the time that packet arrives at
1081 the remote peer it will not be recognized as a valid reply. For TCP
1082 this bug is quickly spotted because all TCP implementations will
1083 ignore replies with the wrong apparent source address and port. For
1084 UDP this bug can more easily go unnoticed because some UDP clients
1085 neglect to check the source address and port of replies; thus, they
1086 will appear to work some of the time with NAT gateways that put the
1087 wrong source address and port in outbound packets. All NAT gateways
1088 MUST ensure that mappings, however created, are bidirectional.
1089
1090 <span class="h2"><a class="selflink" id="section-4" href="#section-4">4</a>. UNSAF Considerations</span>
1091
1092 The document "IAB Considerations for UNilateral Self-Address Fixing
1093 (UNSAF) Across Network Address Translation (NAT)" [<a href="./rfc3424" title="&quot;IAB Considerations for UNilateral Self-Address Fixing (UNSAF) Across Network Address Translation&quot;">RFC3424</a>] covers a
1094 number of issues when working with NATs. It outlines some
1095 requirements for any document that attempts to work around problems
1096 associated with NATs. This section addresses those requirements.
1097
1098 <span class="h3"><a class="selflink" id="section-4.1" href="#section-4.1">4.1</a>. Scope</span>
1099
1100 This protocol addresses the needs of TCP and UDP transport peers that
1101 are separated from the public Internet by exactly one IPv4 NAT. Such
1102 peers must have access to some form of directory server for
1103 registering the public IPv4 address and port at which they can be
1104 reached.
1105
1106 <span class="h3"><a class="selflink" id="section-4.2" href="#section-4.2">4.2</a>. Transition Plan</span>
1107
1108 Any client making use of this protocol SHOULD implement IPv6 support.
1109 If a client supports IPv6 and is running on a device with a global
1110 IPv6 address, that IPv6 address SHOULD be preferred to the IPv4
1111 external address learned via this NAT mapping protocol. In case
1112 other clients do not have IPv6 connectivity, both the IPv4 and IPv6
1113
1114
1115
1116 <span class="grey">Cheshire &amp; Krochmal Informational [Page 20]</span></pre>
1117 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-21" ></span>
1118 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
1119
1120
1121 addresses SHOULD be registered with whatever form of directory server
1122 is used. Preference SHOULD be given to IPv6 addresses when
1123 available. By implementing support for IPv6 and using this protocol
1124 for IPv4, vendors can ship products today that will work under both
1125 scenarios. As IPv6 becomes more widely deployed, clients of this
1126 protocol following these recommendations will transparently make use
1127 of IPv6.
1128
1129 <span class="h3"><a class="selflink" id="section-4.3" href="#section-4.3">4.3</a>. Failure Cases</span>
1130
1131 Aside from NATs that do not implement this protocol, there are a
1132 number of situations where this protocol may not work.
1133
1134 <span class="h4"><a class="selflink" id="section-4.3.1" href="#section-4.3.1">4.3.1</a>. NAT behind NAT</span>
1135
1136 Some people's primary IPv4 address, assigned by their ISP, may itself
1137 be a NAT address. In addition, some people may have an external IPv4
1138 address, but may then double NAT themselves, perhaps by choice or
1139 perhaps by accident. Although it might be possible in principle for
1140 one NAT gateway to recursively request a mapping from the next one,
1141 this document does not advocate that and does not try to prescribe
1142 how it would be done.
1143
1144 It would be a lot of work to implement nested NAT port mapping
1145 correctly, and there are a number of reasons why the end result might
1146 not be as useful as we might hope. Consider the case of an ISP that
1147 offers each of its customers only a single NAT address. This ISP
1148 could instead have chosen to provide each customer with a single
1149 public IPv4 address, or, if the ISP insists on running NAT, it could
1150 have chosen to allow each customer a reasonable number of addresses,
1151 enough for each customer device to have its own NAT address directly
1152 from the ISP. If, instead, this ISP chooses to allow each customer
1153 just one and only one NAT address, forcing said customer to run
1154 nested NAT in order to use more than one device, it seems unlikely
1155 that such an ISP would be so obliging as to provide a NAT service
1156 that supports NAT-PMP. Supposing that such an ISP did wish to offer
1157 its customers NAT service with NAT-PMP so as to give them the ability
1158 to receive inbound connections, this ISP could easily choose to allow
1159 each client to request a reasonable number of DHCP addresses from
1160 that gateway. Remember that Net 10/8 [<a href="./rfc1918" title="&quot;Address Allocation for Private Internets&quot;">RFC1918</a>] allows for over 16
1161 million addresses, so NAT addresses are not in any way in short
1162 supply. A single NAT gateway with 16 million available addresses is
1163 likely to run out of packet forwarding capacity before it runs out of
1164 internal addresses to hand out. In this way, the ISP could offer
1165 single-level NAT with NAT-PMP, obviating the need to support nested
1166 NAT-PMP. In addition, an ISP that is motivated to provide their
1167 customers with unhindered access to the Internet by allowing incoming
1168 as well as outgoing connections has better ways to offer this
1169
1170
1171
1172 <span class="grey">Cheshire &amp; Krochmal Informational [Page 21]</span></pre>
1173 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-22" ></span>
1174 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
1175
1176
1177 service. Such an ISP could offer its customers real public IPv4
1178 addresses instead of NAT addresses, or could choose to offer its
1179 customers full IPv6 connectivity, where no mapping or translation is
1180 required at all.
1181
1182 Note: In the nine years since NAT-PMP was designed, the pool of
1183 available IPv4 addresses has been exhausted, and many ISPs now offer
1184 translated IPv4 addresses out of necessity. Such ISPs have indicated
1185 a willingness to offer PCP service to their customers.
1186
1187 <span class="h4"><a class="selflink" id="section-4.3.2" href="#section-4.3.2">4.3.2</a>. NATs with Multiple External IPv4 Addresses</span>
1188
1189 If a NAT maps internal addresses to multiple external addresses, then
1190 it SHOULD pick one of those external addresses as the one it will
1191 support for inbound connections, and for the purposes of this
1192 protocol it SHOULD act as if that address were its only address.
1193
1194 <span class="h4"><a class="selflink" id="section-4.3.3" href="#section-4.3.3">4.3.3</a>. NATs and Routed Private Networks</span>
1195
1196 In some cases, a large network may be subnetted. Some sites may
1197 install a NAT gateway and subnet the private network. Such
1198 subnetting breaks this protocol because the router address is not
1199 necessarily the address of the device performing NAT.
1200
1201 Addressing this problem is not a high priority. Any site with the
1202 resources to set up such a configuration should have the resources to
1203 add manual mappings or attain a range of globally unique addresses.
1204
1205 Not all NATs will support this protocol. In the case where a client
1206 is run behind a NAT that does not support this protocol, the software
1207 relying on the functionality of this protocol may be unusable.
1208
1209 <span class="h4"><a class="selflink" id="section-4.3.4" href="#section-4.3.4">4.3.4</a>. Communication between Hosts behind the Same NAT</span>
1210
1211 NAT gateways supporting NAT-PMP should also implement "hairpin
1212 translation". Hairpin translation means supporting communication
1213 between two local clients being served by the same NAT gateway.
1214
1215 Suppose device A is listening on internal address and port
1216 10.0.0.2:80 for incoming connections. Using NAT-PMP, device A has
1217 obtained a mapping to external address and port x.x.x.x:80, and has
1218 recorded this external address and port in a public directory of some
1219 kind. For example, it could have created a DNS SRV record containing
1220 this information, and recorded it, using DNS Dynamic Update
1221 [<a href="./rfc3007" title="&quot;Secure Domain Name System (DNS) Dynamic Update&quot;">RFC3007</a>], in a publicly accessible DNS server. Suppose then that
1222 device B, behind the same NAT gateway as device A, but unknowing or
1223 uncaring of this fact, retrieves device A's DNS SRV record and
1224 attempts to open a TCP connection to x.x.x.x:80. The outgoing
1225
1226
1227
1228 <span class="grey">Cheshire &amp; Krochmal Informational [Page 22]</span></pre>
1229 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-23" ></span>
1230 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
1231
1232
1233 packets addressed to this public Internet address will be sent to the
1234 NAT gateway for translation and forwarding. Having translated the
1235 source address and port number on the outgoing packet, the NAT
1236 gateway needs to be smart enough to recognize that the destination
1237 address is in fact itself, and then feed this packet back into its
1238 packet reception engine, to perform the destination port mapping
1239 lookup to translate and forward this packet to device A at address
1240 and port 10.0.0.2:80.
1241
1242 <span class="h4"><a class="selflink" id="section-4.3.5" href="#section-4.3.5">4.3.5</a>. Non-UDP/TCP Transport Traffic</span>
1243
1244 Any communication over transport protocols other than TCP and UDP
1245 will not be served by this protocol. Examples are Generic Routing
1246 Encapsulation (GRE), Authentication Header (AH), and Encapsulating
1247 Security Payload (ESP).
1248
1249 <span class="h3"><a class="selflink" id="section-4.4" href="#section-4.4">4.4</a>. Long-Term Solution</span>
1250
1251 As IPv6 is deployed, clients of this protocol supporting IPv6 will be
1252 able to bypass this protocol and the NAT when communicating with
1253 other IPv6 devices. In order to ensure this transition, any client
1254 implementing this protocol SHOULD also implement IPv6 and use this
1255 solution only when IPv6 is not available to both peers.
1256
1257 <span class="h3"><a class="selflink" id="section-4.5" href="#section-4.5">4.5</a>. Existing Deployed NATs</span>
1258
1259 Existing deployed NATs will not support this protocol. This protocol
1260 will only work with NATs that are upgraded to support it.
1261
1262 <span class="h2"><a class="selflink" id="section-5" href="#section-5">5</a>. Security Considerations</span>
1263
1264 As discussed in <a href="#section-3.2">Section 3.2</a>, "Determining the External Address", only
1265 a client on the internal side of the NAT may create port mappings,
1266 and it may do so only on its own behalf. By using IP address
1267 spoofing, it's possible for one client to delete the port mappings of
1268 another client. It's also possible for one client to create port
1269 mappings on behalf of another client. In cases where this is a
1270 concern, it can be dealt with using IPsec [<a href="./rfc4301" title="&quot;Security Architecture for the Internet Protocol&quot;">RFC4301</a>].
1271
1272 The multicast announcements described in <a href="#section-3.2.1">Section 3.2.1</a>, "Announcing
1273 Address Changes", could be spoofed, facilitating a denial-of-service
1274 attack. This makes NAT-PMP unsuitable for use on LANs with large
1275 numbers of hosts where one or more of the hosts may be untrustworthy.
1276
1277 Another concern is that rogue software running on a local host could
1278 create port mappings for unsuspecting hosts, thereby rendering them
1279 vulnerable to external attack. However, it's not clear how realistic
1280 this threat model is, since rogue software on a local host could
1281
1282
1283
1284 <span class="grey">Cheshire &amp; Krochmal Informational [Page 23]</span></pre>
1285 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-24" ></span>
1286 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
1287
1288
1289 attack such unsuspecting hosts directly itself, without resorting to
1290 such a convoluted indirect technique. This concern is also a little
1291 misguided because it is based on the assumption that a NAT gateway
1292 and a firewall are the same thing, which they are not.
1293
1294 Some people view the property of NATs blocking inbound connections as
1295 a security benefit that is undermined by this protocol. The authors
1296 of this document have a different point of view. In the days before
1297 NAT became prevalent, all hosts had unique public IP addresses, and
1298 had unhindered ability to communicate with any other host on the
1299 Internet (a configuration that is still surprisingly common). Using
1300 NAT breaks this unhindered connectivity, relegating hosts to second-
1301 class status, unable to receive inbound connections. This protocol
1302 goes some way to partially reverse that damage. The purpose of a NAT
1303 gateway should be to allow several hosts to share a single address,
1304 not to simultaneously impede those host's ability to communicate
1305 freely. Security is most properly provided by end-to-end
1306 cryptographic security, and/or by explicit firewall functionality, as
1307 appropriate. Blocking of certain connections should occur only as a
1308 result of explicit and intentional firewall policy, not as an
1309 accidental side effect of some other technology.
1310
1311 However, since many users do have an expectation that their NAT
1312 gateways can function as a kind of firewall, any NAT gateway
1313 implementing this protocol SHOULD have an administrative mechanism to
1314 disable it, thereby restoring the pre-NAT-PMP behavior.
1315
1316 <span class="h2"><a class="selflink" id="section-6" href="#section-6">6</a>. IANA Considerations</span>
1317
1318 UDP ports 5350 and 5351 have been assigned for use by NAT-PMP, and
1319 subsequently by its successor, Port Control Protocol [<a href="./rfc6887" title="&quot;Port Control Protocol (PCP)&quot;">RFC6887</a>].
1320
1321 No further IANA services are required by this document.
1322
1323 <span class="h2"><a class="selflink" id="section-7" href="#section-7">7</a>. Acknowledgments</span>
1324
1325 The concepts described in this document have been explored,
1326 developed, and implemented with help from Mark Baugher, Bob Bradley,
1327 Josh Graessley, Rory McGuire, Rob Newberry, Roger Pantos, John
1328 Saxton, Kiren Sekar, Jessica Vazquez, and James Woodyatt.
1329
1330 Special credit goes to Mike Bell, the Apple Vice President who
1331 recognized the need for a clean, elegant, reliable Port Mapping
1332 Protocol, and made the decision early on that Apple's AirPort base
1333 stations would support NAT-PMP.
1334
1335
1336
1337
1338
1339
1340 <span class="grey">Cheshire &amp; Krochmal Informational [Page 24]</span></pre>
1341 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-25" ></span>
1342 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
1343
1344
1345 <span class="h2"><a class="selflink" id="section-8" href="#section-8">8</a>. Deployment History</span>
1346
1347 In August 2004, NAT-PMP client software first became available to the
1348 public through Apple's Darwin Open Source code. In April 2005,
1349 NAT-PMP implementations began shipping to end users with the launch
1350 of Mac OS X 10.4 Tiger and Bonjour for Windows 1.0, and in June 2005
1351 the protocol was first publicly documented in the original draft
1352 version of this document.
1353
1354 The NAT-PMP client in Mac OS X 10.4 Tiger and Bonjour for Windows
1355 exists as part of the mDNSResponder/mdnsd system service. When a
1356 client advertises a service using Wide Area Bonjour [<a href="./rfc6763" title="&quot;DNS-Based Service Discovery&quot;">RFC6763</a>], and
1357 the machine is behind a NAT-PMP-capable NAT gateway, and the machine
1358 is so configured, the mDNSResponder system service automatically uses
1359 NAT-PMP to set up an inbound port mapping, and then records the
1360 external IPv4 address and port in the global DNS. Existing client
1361 software using the Bonjour programming APIs [<a href="#ref-Bonjour" title="&quot;Bonjour&quot;">Bonjour</a>] got this new
1362 NAT traversal functionality automatically. The logic behind this
1363 decision was that if client software publishes its information into
1364 the global DNS via Wide Area Bonjour service advertising, then it's
1365 reasonable to infer an expectation that this information should
1366 actually be usable by the peers retrieving it. Generally speaking,
1367 recording a private IPv4 address like 10.0.0.2 in the public DNS is
1368 likely to be pointless because that address is not reachable from
1369 clients on the other side of the NAT gateway. In the case of a home
1370 user with a single computer directly connected to their Cable or DSL
1371 modem, with a single global IPv4 address and no NAT gateway (a common
1372 configuration at that time), publishing the machine's global IPv4
1373 address into the global DNS is useful, because that IPv4 address is
1374 globally reachable. In contrast, a home user using a NAT gateway to
1375 share a single global IPv4 address between several computers loses
1376 this ability to receive inbound connections. This breaks many peer-
1377 to-peer collaborative applications, like the multi-user text editor
1378 SubEthaEdit [<a href="#ref-SEE">SEE</a>]. For many users, moving from one computer with a
1379 global IPv4 address, to two computers using NAT to share a single
1380 global IPv4 address, loss of inbound reachability was an unwanted
1381 side effect of using NAT for address sharing. Automatically creating
1382 the necessary inbound port mappings helped remedy this unwanted side
1383 effect of NAT.
1384
1385 The server side of the NAT-PMP protocol is implemented in Apple's
1386 AirPort Extreme, AirPort Express, and Time Capsule wireless base
1387 stations, and in the Internet Sharing feature of Mac OS X 10.4 and
1388 later. Some third-party NAT vendors, such as Peplink, also offer
1389 NAT-PMP in their products.
1390
1391
1392
1393
1394
1395
1396 <span class="grey">Cheshire &amp; Krochmal Informational [Page 25]</span></pre>
1397 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-26" ></span>
1398 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
1399
1400
1401 In Mac OS X 10.4 Tiger, the NAT-PMP client was invoked automatically
1402 as a side effect of clients requesting Wide Area Bonjour service
1403 registrations. Using NAT-PMP without an associated Wide Area Bonjour
1404 service registration required use of a third-party client library.
1405
1406 In October 2007, Mac OS X 10.5 Leopard added the "DNSServiceNATPort-
1407 MappingCreate" API, which made NAT-PMP client functionality directly
1408 available, so software could use it with other directory and
1409 rendezvous mechanisms in addition to Wide Area Bonjour DNS Updates.
1410
1411 In 2013, NAT-PMP was superseded by the IETF Standards Track Port
1412 Control Protocol [<a href="./rfc6887" title="&quot;Port Control Protocol (PCP)&quot;">RFC6887</a>]. PCP builds on NAT-PMP and uses a
1413 compatible packet format, and adds a number of significant
1414 enhancements, including IPv6 support, management of outbound
1415 mappings, management of firewall rules, full compatibility with
1416 large-scale NATs with a pool of external addresses, error lifetimes,
1417 and an extension mechanism to enable future enhancements.
1418
1419 <span class="h2"><a class="selflink" id="section-9" href="#section-9">9</a>. Noteworthy Features of NAT Port Mapping Protocol and PCP</span>
1420
1421 Some readers have asked how NAT-PMP and PCP compare to other similar
1422 solutions, particularly the UPnP Forum's Internet Gateway Device
1423 (IGD) Device Control Protocol [<a href="#ref-IGD" title="&quot;Internet Gateway Device (IGD) Standardized Device Control Protocol V 1.0&quot;">IGD</a>].
1424
1425 The answer is that although the Universal Plug and Play (UPnP) IGD
1426 protocol is often used as a way for client devices to create port
1427 mappings programmatically, it's not ideal for that task. Whereas
1428 NAT-PMP was explicitly designed to be used primarily by software
1429 entities managing their own port mappings, UPnP IGD is more tailored
1430 towards being used by humans configuring all the settings of their
1431 gateway using some GUI tool. This difference in emphasis leads to
1432 protocol differences. For example, while it is reasonable and
1433 sensible to require software entities to renew their mappings
1434 periodically to prove that they are still there (like a device
1435 renewing its DHCP address lease), it would be unreasonable to require
1436 the same thing of a human user. When a human user configures their
1437 gateway, they expect it to stay configured that way until they decide
1438 to change it. If they configure a port mapping, they expect it to
1439 stay configured until they decide to delete it.
1440
1441 Because of this focus on being a general administration protocol for
1442 all aspects of home gateway configuration, UPnP IGD is a large and
1443 complicated collection of protocols (360 pages of specification
1444 spread over 13 separate documents, not counting supporting protocol
1445 specifications like Simple Service Discovery Protocol (SSDP) and
1446 Extensible Markup Language (XML)). While it may be a fine way for
1447
1448
1449
1450
1451
1452 <span class="grey">Cheshire &amp; Krochmal Informational [Page 26]</span></pre>
1453 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-27" ></span>
1454 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
1455
1456
1457 human users to configure their home gateways, it is not especially
1458 suited to the task of programmatically creating dynamic port
1459 mappings.
1460
1461 The requirements for a good port mapping protocol, requirements that
1462 are met by NAT-PMP, are outlined below.
1463
1464 <span class="h3"><a class="selflink" id="section-9.1" href="#section-9.1">9.1</a>. Simplicity</span>
1465
1466 Many home gateways, and many of the devices that connect to them, are
1467 small, low-cost devices, with limited RAM, flash memory, and CPU
1468 resources. Protocols they use should be considerate of this,
1469 supporting a small number of simple operations that can be
1470 implemented easily with a small amount of code. A quick comparison,
1471 based on page count of the respective documents alone, suggests that
1472 NAT-PMP is at least ten times simpler than UPnP IGD.
1473
1474 <span class="h3"><a class="selflink" id="section-9.2" href="#section-9.2">9.2</a>. Focused Scope</span>
1475
1476 The more things a protocol can do, the more chance there is that
1477 something it does could be exploited for malicious purposes. NAT-PMP
1478 is tightly focused on the specific task of creating port mappings.
1479 Were the protocol to be misused in some way, this helps limit the
1480 scope of what mischief could be performed using the protocol.
1481
1482 Because UPnP IGD allows control over all home gateway configuration
1483 settings, the potential for mischief is far greater. For example, a
1484 UPnP IGD home gateway allows messages that tell it to change the DNS
1485 server addresses that it sends to clients in its DHCP packets. Using
1486 this mechanism, a single item of malicious web content (e.g., a rogue
1487 Flash banner advert on a web page) can make a persistent change to
1488 the home gateway's configuration without the user's knowledge, such
1489 that all future DNS requests by all local clients will be sent to a
1490 rogue DNS server. This allows criminals to perform a variety of
1491 mischief, such as hijacking connections to bank web sites and
1492 redirecting them to the criminals' web servers instead [<a href="#ref-VU347812">VU347812</a>].
1493
1494 <span class="h3"><a class="selflink" id="section-9.3" href="#section-9.3">9.3</a>. Efficiency</span>
1495
1496 In addition to low-cost home gateways, many of the clients will also
1497 be similarly constrained low-cost devices with limited RAM resources.
1498
1499 When implementing a NAT-PMP client on a constrained device, it's
1500 beneficial to have well-defined bounds on RAM requirements that are
1501 fixed and known in advance. For example, when requesting the
1502 gateway's external IPv4 address, a NAT-PMP client on Ethernet knows
1503
1504
1505
1506
1507
1508 <span class="grey">Cheshire &amp; Krochmal Informational [Page 27]</span></pre>
1509 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-28" ></span>
1510 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
1511
1512
1513 that to receive the reply it will require 14 bytes for the Ethernet
1514 header, 20 bytes for the IPv4 header, 8 bytes for the UDP header, and
1515 12 bytes for the NAT-PMP payload, making a total of 54 bytes.
1516
1517 In contrast, UPnP IGD uses an XML reply of unbounded size. It is not
1518 uncommon for a UPnP IGD device to return an XML document 4000 to 8000
1519 bytes in size to communicate its 4-byte external IPv4 address, and
1520 the protocol specification places no upper bound on how large the XML
1521 response may be, so there's nothing to stop the reply being even
1522 larger. This means that developers of UPnP client devices can only
1523 guess at how much memory they may need to receive the XML reply.
1524 Operational experience suggests that 10,000 bytes is usually enough
1525 for most UPnP IGD home gateways today, but that's no guarantee that
1526 some future UPnP IGD home gateway might not return a perfectly legal
1527 XML reply much larger than that.
1528
1529 In addition, because the XML reply is too large to fit in a single
1530 UDP packet, UPnP IGD has to use a TCP connection, thereby adding the
1531 overhead of TCP connection setup and teardown.
1532
1533 The process of discovering a UPnP IGD home gateway's external IPv4
1534 address consists of:
1535
1536 o SSDP transaction to discover the TCP port to use, and the "URL" of
1537 the XML document to fetch from the gateway. Following the SSDP
1538 specification, this is 3 multicast requests, eliciting 9 unicast
1539 responses.
1540
1541 o HTTP "GET" request to get the device description. Typically, 16
1542 packets: 3 for TCP connection setup, 9 packets of data exchange,
1543 and a 4-packet FIN-ACK-FIN-ACK sequence to close the connection.
1544
1545 o HTTP "POST" to request the external IPv4 address. Typically, 14
1546 packets: 3 for TCP connection setup, 7 packets of data exchange,
1547 and a 4-packet FIN-ACK-FIN-ACK sequence to close the connection.
1548
1549 To retrieve the external IPv4 address NAT-PMP takes a 2-packet UDP
1550 exchange (44-byte request, 54-byte response); the same thing using
1551 UPnP IGD takes 42 packets and thousands of bytes.
1552
1553 Similarly, UPnP IGD's HTTP "POST" request for a port mapping is
1554 typically a 14-packet exchange, compared with NAT-PMP's 2-packet UDP
1555 exchange.
1556
1557
1558
1559
1560
1561
1562
1563
1564 <span class="grey">Cheshire &amp; Krochmal Informational [Page 28]</span></pre>
1565 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-29" ></span>
1566 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
1567
1568
1569 <span class="h3"><a class="selflink" id="section-9.4" href="#section-9.4">9.4</a>. Atomic Allocation Operations</span>
1570
1571 Some of the useful properties of NAT-PMP were inspired by DHCP, a
1572 reliable and successful protocol. For example, DHCP allows a client
1573 to request a desired IP address, but if that address is already in
1574 use the DHCP server will instead assign some other available address.
1575
1576 Correspondingly, NAT-PMP allows a client to request a desired
1577 external port, and if that external port is already in use by some
1578 other client, the NAT-PMP server will instead assign some other
1579 available external port.
1580
1581 UPnP IGD does not do this. If a UPnP IGD client requests an external
1582 port that has already been allocated, then one of two things happens.
1583
1584 Some UPnP IGD home gateways just silently overwrite the old mapping
1585 with the new one, causing the previous client to lose connectivity.
1586 If the previous client renews its port mapping, then it in turn
1587 overwrites the new mapping, and the two clients fight over the same
1588 external port indefinitely, neither achieving reliable connectivity.
1589
1590 Other IGD home gateways return a "Conflict" error if the port is
1591 already in use, which does at least tell the client what happened,
1592 but doesn't tell the client what to do. Instead of the NAT gateway
1593 (which does know which ports are available) assigning one to the
1594 client, the NAT gateway makes the client (which doesn't know) keep
1595 guessing until it gets lucky. This problem remains mild as long as
1596 not many clients are using UPnP IGD, but gets progressively worse as
1597 the number of clients on the network requesting port mappings goes
1598 up. In addition, UPnP IGD works particularly badly in conjunction
1599 with the emerging policy of allocating pre-assigned port ranges to
1600 each client. If a client is assigned TCP port range 63488-64511, and
1601 the UPnP IGD client requests TCP port 80, trying successive
1602 incrementing ports until it succeeds, then the UPnP IGD client will
1603 have to issue 63,409 requests before it succeeds.
1604
1605 <span class="h3"><a class="selflink" id="section-9.5" href="#section-9.5">9.5</a>. Garbage Collection</span>
1606
1607 In any system that operates for a long period of time (as a home
1608 gateway should), it is important that garbage data does not
1609 accumulate indefinitely until the system runs out of memory and
1610 fails.
1611
1612 Similar to how DHCP leases an IP address to a client for a finite
1613 length of time, NAT-PMP leases an external port to a client for a
1614 finite length of time. The NAT-PMP client must renew the port
1615 mapping before it expires, or, like an unrenewed DHCP address, it
1616 will be reclaimed. If a laptop computer is abruptly disconnected
1617
1618
1619
1620 <span class="grey">Cheshire &amp; Krochmal Informational [Page 29]</span></pre>
1621 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-30" ></span>
1622 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
1623
1624
1625 from the network without the opportunity to delete its port mappings,
1626 the NAT gateway will reclaim those mappings when they are not
1627 renewed.
1628
1629 In principle, UPnP IGD should allow clients to specify a lifetime on
1630 port mappings. However, a Google search for "UPnP NewLeaseDuration"
1631 shows that in practice pretty much every client uses
1632 "&lt;NewLeaseDuration&gt;0&lt;/NewLeaseDuration&gt;" to request an infinite
1633 lease, and the protocol has no way for the NAT gateway to decline
1634 that infinite lease request and require the client to renew it at
1635 reasonable intervals. Furthermore, anecdotal evidence is that if the
1636 client requests a lease other than zero, there are IGD home gateways
1637 that will ignore the request, fail in other ways, or even crash
1638 completely. As a client implementer then, you would be well advised
1639 not to attempt to request a lease other than zero, unless you want to
1640 suffer the support costs and bad publicity of lots of people
1641 complaining that your device brought down their entire network.
1642
1643 Because none of the early UPnP IGD clients requested port mapping
1644 leases, many UPnP IGD home gateway vendors never tested that
1645 functionality, and got away with shipping home gateways where that
1646 functionality was buggy or nonexistent. Because there are so many
1647 buggy UPnP IGD home gateways already deployed, client writers wisely
1648 stick to the well-trodden path of only requesting infinite leases.
1649 Because there are now few (if any) clients attempting to request non-
1650 zero leases, home gateway vendors have little incentive to expend
1651 resources implementing a feature no one uses.
1652
1653 This unfortunate consequence of the way UPnP IGD was developed and
1654 deployed means that in practice it has no usable port mapping lease
1655 facility today, and therefore when run for a long period of time UPnP
1656 IGD home gateways have no good way to avoid accumulating an unbounded
1657 number of stale port mappings.
1658
1659 <span class="h3"><a class="selflink" id="section-9.6" href="#section-9.6">9.6</a>. State Change Announcements</span>
1660
1661 When using DHCP on the external interface, as is the norm for home
1662 gateways, there is no guarantee that a UPnP IGD home gateway's
1663 external IPv4 address will remain unchanged. Indeed, some ISPs
1664 change their customer's IPv4 address every 24 hours (possibly in an
1665 effort to make it harder for their customers to "run a server" at
1666 home). What this means is that if the home gateway's external IPv4
1667 address changes, it needs to inform its clients, so that they can
1668 make any necessary updates to global directory information (e.g.,
1669 performing a Dynamic DNS update to update their address record).
1670
1671 When a NAT-PMP gateway's external IPv4 address changes, it broadcasts
1672 announcement packets to inform clients of this. UPnP IGD does not.
1673
1674
1675
1676 <span class="grey">Cheshire &amp; Krochmal Informational [Page 30]</span></pre>
1677 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-31" ></span>
1678 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
1679
1680
1681 <span class="h3"><a class="selflink" id="section-9.7" href="#section-9.7">9.7</a>. Soft State Recovery</span>
1682
1683 When run for a long enough period of time, any network will have
1684 devices that fail, get rebooted, suffer power outages, or lose state
1685 for other reasons. A home gateway that runs for long enough is
1686 likely to suffer some such incident eventually. After losing state,
1687 it has no record of the port mappings it created, and clients suffer
1688 a consequent loss of connectivity.
1689
1690 To handle this case, NAT-PMP has the "Seconds Since Start of Epoch"
1691 mechanism. After a reboot or other loss of state, a NAT-PMP gateway
1692 broadcasts announcement packets giving its external IPv4 address,
1693 with the Seconds Since Start of Epoch field reset to begin counting
1694 from zero again. When a NAT-PMP client observes packets from its
1695 NAT-PMP gateway where the gateway's notion of time has apparently
1696 gone backwards compared to the client's, the client knows the gateway
1697 has probably lost state, and immediately recreates its mappings to
1698 restore connectivity.
1699
1700 UPnP IGD has no equivalent mechanism.
1701
1702 <span class="h3"><a class="selflink" id="section-9.8" href="#section-9.8">9.8</a>. On-Path NAT Discovery</span>
1703
1704 For any given host, it is only useful to request NAT port mappings in
1705 the NAT gateway through which that host's packets are flowing. A NAT
1706 port mapping is a request for packets to be translated in a certain
1707 way; the NAT gateway can only perform that translation if it's
1708 actually forwarding inbound and outbound packets for that host.
1709
1710 This is why NAT-PMP sends its requests to the host's default router,
1711 since this is the device that is forwarding (and possibly
1712 translating) inbound and outbound packets for that host. (In a
1713 larger network with multiple hops between a host and its NAT gateway,
1714 some other mechanism would need to be used to discover the correct
1715 on-path NAT for a host; this is possible, but outside the scope of
1716 this document.)
1717
1718 In contrast, UPnP IGD does not limit itself to using only on-path
1719 NATs. UPnP IGD uses a multicast SSDP query, and uses any device it
1720 finds on the local network claiming UPnP IGD capability, regardless
1721 of whether any inbound or outbound traffic is actually flowing
1722 through that device. Over the past few years this led to many bug
1723 reports being sent to Apple with the general form: "Port Mapping
1724 doesn't work on my Mac and that's a bug because everything else on my
1725 network says UPnP IGD is working fine." Upon investigation it always
1726 turned out that: (i) these people had NAT gateways that either didn't
1727 support port mapping requests, or had that capability disabled, and
1728 (ii) for some reason they also had some other old NAT device still
1729
1730
1731
1732 <span class="grey">Cheshire &amp; Krochmal Informational [Page 31]</span></pre>
1733 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-32" ></span>
1734 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
1735
1736
1737 connected to their network, and those other NAT devices were
1738 advertising UPnP IGD capability, even though they were not the active
1739 NAT gateway for the network. This led to UPnP IGD clients falsely
1740 reporting that they were "working fine", and only the Mac correctly
1741 reporting that it was unable to make any useful port mappings. In
1742 many cases the people reporting this "bug" had devices like game
1743 consoles on their home network that for many years had been reporting
1744 that UPnP IGD was "working fine", yet during those years they had
1745 never once successfully received any inbound network packet or
1746 connection. The irony is that, for these people who were reporting
1747 bugs to Apple, UPnP IGD "working fine" had been indistinguishable
1748 from UPnP IGD doing nothing useful at all. It was only when Back to
1749 My Mac [<a href="./rfc6281" title="&quot;Understanding Apple's Back to My Mac (BTMM) Service&quot;">RFC6281</a>] started reporting that it was unable to make any
1750 functional port mappings that these people discovered they'd never
1751 had any working port mappings on their NAT gateway.
1752
1753 <span class="h2"><a class="selflink" id="section-10" href="#section-10">10</a>. References</span>
1754
1755 <span class="h3"><a class="selflink" id="section-10.1" href="#section-10.1">10.1</a>. Normative References</span>
1756
1757 [<a id="ref-RFC1918">RFC1918</a>] Rekhter, Y., Moskowitz, B., Karrenberg, D., de Groot, G.,
1758 and E. Lear, "Address Allocation for Private Internets",
1759 <a href="https://www.rfc-editor.org/bcp/bcp5">BCP 5</a>, <a href="./rfc1918">RFC 1918</a>, February 1996.
1760
1761 [<a id="ref-RFC2119">RFC2119</a>] Bradner, S., "Key words for use in RFCs to Indicate
1762 Requirement Levels", <a href="https://www.rfc-editor.org/bcp/bcp14">BCP 14</a>, <a href="./rfc2119">RFC 2119</a>, March 1997.
1763
1764 <span class="h3"><a class="selflink" id="section-10.2" href="#section-10.2">10.2</a>. Informative References</span>
1765
1766 [<a id="ref-Bonjour">Bonjour</a>] Apple "Bonjour" &lt;<a href="http://developer.apple.com/bonjour/">http://developer.apple.com/bonjour/</a>&gt;.
1767
1768 [<a id="ref-ETEAISD">ETEAISD</a>] J. Saltzer, D. Reed and D. Clark: "End-to-end arguments in
1769 system design", ACM Trans. Comp. Sys., 2(4):277-88,
1770 November 1984.
1771
1772 [<a id="ref-IGD">IGD</a>] UPnP Standards "Internet Gateway Device (IGD) Standardized
1773 Device Control Protocol V 1.0", November 2001,
1774 &lt;<a href="http://www.upnp.org/standardizeddcps/igd.asp">http://www.upnp.org/standardizeddcps/igd.asp</a>&gt;.
1775
1776 [<a id="ref-RFC2131">RFC2131</a>] Droms, R., "Dynamic Host Configuration Protocol", <a href="./rfc2131">RFC</a>
1777 <a href="./rfc2131">2131</a>, March 1997.
1778
1779 [<a id="ref-RFC2663">RFC2663</a>] Srisuresh, P. and M. Holdrege, "IP Network Address
1780 Translator (NAT) Terminology and Considerations", <a href="./rfc2663">RFC</a>
1781 <a href="./rfc2663">2663</a>, August 1999.
1782
1783 [<a id="ref-RFC3007">RFC3007</a>] Wellington, B., "Secure Domain Name System (DNS) Dynamic
1784 Update", <a href="./rfc3007">RFC 3007</a>, November 2000.
1785
1786
1787
1788 <span class="grey">Cheshire &amp; Krochmal Informational [Page 32]</span></pre>
1789 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-33" ></span>
1790 <span class="grey"><a href="./rfc6886">RFC 6886</a> NAT-PMP April 2013</span>
1791
1792
1793 [<a id="ref-RFC3022">RFC3022</a>] Srisuresh, P. and K. Egevang, "Traditional IP Network
1794 Address Translator (Traditional NAT)", <a href="./rfc3022">RFC 3022</a>, January
1795 2001.
1796
1797 [<a id="ref-RFC3424">RFC3424</a>] Daigle, L., Ed., and IAB, "IAB Considerations for
1798 UNilateral Self-Address Fixing (UNSAF) Across Network
1799 Address Translation", <a href="./rfc3424">RFC 3424</a>, November 2002.
1800
1801 [<a id="ref-RFC4301">RFC4301</a>] Kent, S. and K. Seo, "Security Architecture for the
1802 Internet Protocol", <a href="./rfc4301">RFC 4301</a>, December 2005.
1803
1804 [<a id="ref-RFC6281">RFC6281</a>] Cheshire, S., Zhu, Z., Wakikawa, R., and L. Zhang,
1805 "Understanding Apple's Back to My Mac (BTMM) Service", <a href="./rfc6281">RFC</a>
1806 <a href="./rfc6281">6281</a>, June 2011.
1807
1808 [<a id="ref-RFC6763">RFC6763</a>] Cheshire, S. and M. Krochmal, "DNS-Based Service
1809 Discovery", <a href="./rfc6763">RFC 6763</a>, February 2013.
1810
1811 [<a id="ref-RFC6887">RFC6887</a>] Wing, D., Ed., Cheshire, S., Boucadair, M., Penno, R., and
1812 P. Selkirk, "Port Control Protocol (PCP)", <a href="./rfc6887">RFC 6887</a>, April
1813 2013.
1814
1815 [<a id="ref-SEE">SEE</a>] SubEthaEdit, &lt;<a href="http://www.codingmonkeys.de/subethaedit/">http://www.codingmonkeys.de/subethaedit/</a>&gt;.
1816
1817 [<a id="ref-VU347812">VU347812</a>] United States Computer Emergency Readiness Team
1818 Vulnerability Note VU#347812,
1819 &lt;<a href="http://www.kb.cert.org/vuls/id/347812">http://www.kb.cert.org/vuls/id/347812</a>&gt;.
1820
1821 Authors' Addresses
1822
1823 Stuart Cheshire
1824 Apple Inc.
1825 1 Infinite Loop
1826 Cupertino, CA 95014
1827 USA
1828
1829 EMail: [email protected]
1830
1831 Marc Krochmal
1832 Apple Inc.
1833 1 Infinite Loop
1834 Cupertino, CA 95014
1835 USA
1836
1837 EMail: [email protected]
1838
1839
1840
1841
1842
1843
1844 Cheshire &amp; Krochmal Informational [Page 33]
1845 </pre>