idnits 2.17.1 draft-ietf-tram-turnbis-00.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- == There are 4 instances of lines with private range IPv4 addresses in the document. If these are generic example addresses, they should be changed to use any of the ranges defined in RFC 6890 (or successor): 192.0.2.x, 198.51.100.x or 203.0.113.x. -- The document has examples using IPv4 documentation addresses according to RFC6890, but does not use any IPv6 documentation addresses. Maybe there should be IPv6 examples, too? Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (August 27, 2014) is 3530 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Missing Reference: '0x4001' is mentioned on line 657, but not defined ** Obsolete normative reference: RFC 5389 (Obsoleted by RFC 8489) -- Obsolete informational reference (is this intentional?): RFC 5245 (Obsoleted by RFC 8445, RFC 8839) -- Obsolete informational reference (is this intentional?): RFC 6156 (Obsoleted by RFC 8656) == Outdated reference: A later version (-12) exists of draft-ietf-tram-turn-server-discovery-00 -- Obsolete informational reference (is this intentional?): RFC 5766 (Obsoleted by RFC 8656) Summary: 1 error (**), 0 flaws (~~), 4 warnings (==), 5 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 TRAM WG T. Reddy, Ed. 3 Internet-Draft Cisco Systems, Inc. 4 Intended status: Standards Track A. Johnston, Ed. 5 Expires: February 28, 2015 Avaya 6 P. Matthews 7 Alcatel-Lucent 8 J. Rosenberg 9 jdrosen.net 10 August 27, 2014 12 Traversal Using Relays around NAT (TURN): Relay Extensions to Session 13 Traversal Utilities for NAT (STUN) 14 draft-ietf-tram-turnbis-00 16 Abstract 18 If a host is located behind a NAT, then in certain situations it can 19 be impossible for that host to communicate directly with other hosts 20 (peers). In these situations, it is necessary for the host to use 21 the services of an intermediate node that acts as a communication 22 relay. This specification defines a protocol, called TURN (Traversal 23 Using Relays around NAT), that allows the host to control the 24 operation of the relay and to exchange packets with its peers using 25 the relay. TURN differs from some other relay control protocols in 26 that it allows a client to communicate with multiple peers using a 27 single relay address. 29 The TURN protocol was designed to be used as part of the ICE 30 (Interactive Connectivity Establishment) approach to NAT traversal, 31 though it also can be used without ICE. 33 Status of This Memo 35 This Internet-Draft is submitted in full conformance with the 36 provisions of BCP 78 and BCP 79. 38 Internet-Drafts are working documents of the Internet Engineering 39 Task Force (IETF). Note that other groups may also distribute 40 working documents as Internet-Drafts. The list of current Internet- 41 Drafts is at http://datatracker.ietf.org/drafts/current/. 43 Internet-Drafts are draft documents valid for a maximum of six months 44 and may be updated, replaced, or obsoleted by other documents at any 45 time. It is inappropriate to use Internet-Drafts as reference 46 material or to cite them other than as "work in progress." 48 This Internet-Draft will expire on February 28, 2015. 50 Copyright Notice 52 Copyright (c) 2014 IETF Trust and the persons identified as the 53 document authors. All rights reserved. 55 This document is subject to BCP 78 and the IETF Trust's Legal 56 Provisions Relating to IETF Documents 57 (http://trustee.ietf.org/license-info) in effect on the date of 58 publication of this document. Please review these documents 59 carefully, as they describe your rights and restrictions with respect 60 to this document. Code Components extracted from this document must 61 include Simplified BSD License text as described in Section 4.e of 62 the Trust Legal Provisions and are provided without warranty as 63 described in the Simplified BSD License. 65 Table of Contents 67 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 4 68 2. Overview of Operation . . . . . . . . . . . . . . . . . . . . 5 69 2.1. Transports . . . . . . . . . . . . . . . . . . . . . . . 8 70 2.2. Allocations . . . . . . . . . . . . . . . . . . . . . . . 9 71 2.3. Permissions . . . . . . . . . . . . . . . . . . . . . . . 11 72 2.4. Send Mechanism . . . . . . . . . . . . . . . . . . . . . 12 73 2.5. Channels . . . . . . . . . . . . . . . . . . . . . . . . 14 74 2.6. Unprivileged TURN Servers . . . . . . . . . . . . . . . . 16 75 2.7. Avoiding IP Fragmentation . . . . . . . . . . . . . . . . 16 76 2.8. RTP Support . . . . . . . . . . . . . . . . . . . . . . . 18 77 2.9. Discovery of Servers . . . . . . . . . . . . . . . . . . 18 78 3. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 18 79 4. General Behavior . . . . . . . . . . . . . . . . . . . . . . 20 80 5. Allocations . . . . . . . . . . . . . . . . . . . . . . . . . 22 81 6. Creating an Allocation . . . . . . . . . . . . . . . . . . . 23 82 6.1. Sending an Allocate Request . . . . . . . . . . . . . . . 23 83 6.2. Receiving an Allocate Request . . . . . . . . . . . . . . 25 84 6.3. Receiving an Allocate Success Response . . . . . . . . . 29 85 6.4. Receiving an Allocate Error Response . . . . . . . . . . 30 86 7. Refreshing an Allocation . . . . . . . . . . . . . . . . . . 32 87 7.1. Sending a Refresh Request . . . . . . . . . . . . . . . . 32 88 7.2. Receiving a Refresh Request . . . . . . . . . . . . . . . 32 89 7.3. Receiving a Refresh Response . . . . . . . . . . . . . . 33 90 8. Permissions . . . . . . . . . . . . . . . . . . . . . . . . . 33 91 9. CreatePermission . . . . . . . . . . . . . . . . . . . . . . 34 92 9.1. Forming a CreatePermission Request . . . . . . . . . . . 35 93 9.2. Receiving a CreatePermission Request . . . . . . . . . . 35 94 9.3. Receiving a CreatePermission Response . . . . . . . . . . 36 95 10. Send and Data Methods . . . . . . . . . . . . . . . . . . . . 36 96 10.1. Forming a Send Indication . . . . . . . . . . . . . . . 36 97 10.2. Receiving a Send Indication . . . . . . . . . . . . . . 36 98 10.3. Receiving a UDP Datagram . . . . . . . . . . . . . . . . 37 99 10.4. Receiving a Data Indication . . . . . . . . . . . . . . 38 100 11. Channels . . . . . . . . . . . . . . . . . . . . . . . . . . 38 101 11.1. Sending a ChannelBind Request . . . . . . . . . . . . . 40 102 11.2. Receiving a ChannelBind Request . . . . . . . . . . . . 40 103 11.3. Receiving a ChannelBind Response . . . . . . . . . . . . 41 104 11.4. The ChannelData Message . . . . . . . . . . . . . . . . 42 105 11.5. Sending a ChannelData Message . . . . . . . . . . . . . 42 106 11.6. Receiving a ChannelData Message . . . . . . . . . . . . 43 107 11.7. Relaying Data from the Peer . . . . . . . . . . . . . . 44 108 12. IP Header Fields . . . . . . . . . . . . . . . . . . . . . . 44 109 13. New STUN Methods . . . . . . . . . . . . . . . . . . . . . . 45 110 14. New STUN Attributes . . . . . . . . . . . . . . . . . . . . . 46 111 14.1. CHANNEL-NUMBER . . . . . . . . . . . . . . . . . . . . . 46 112 14.2. LIFETIME . . . . . . . . . . . . . . . . . . . . . . . . 47 113 14.3. XOR-PEER-ADDRESS . . . . . . . . . . . . . . . . . . . . 47 114 14.4. DATA . . . . . . . . . . . . . . . . . . . . . . . . . . 47 115 14.5. XOR-RELAYED-ADDRESS . . . . . . . . . . . . . . . . . . 47 116 14.6. EVEN-PORT . . . . . . . . . . . . . . . . . . . . . . . 47 117 14.7. REQUESTED-TRANSPORT . . . . . . . . . . . . . . . . . . 48 118 14.8. DONT-FRAGMENT . . . . . . . . . . . . . . . . . . . . . 48 119 14.9. RESERVATION-TOKEN . . . . . . . . . . . . . . . . . . . 48 120 15. New STUN Error Response Codes . . . . . . . . . . . . . . . . 49 121 16. Detailed Example . . . . . . . . . . . . . . . . . . . . . . 49 122 17. Security Considerations . . . . . . . . . . . . . . . . . . . 56 123 17.1. Outsider Attacks . . . . . . . . . . . . . . . . . . . . 56 124 17.1.1. Obtaining Unauthorized Allocations . . . . . . . . . 56 125 17.1.2. Offline Dictionary Attacks . . . . . . . . . . . . . 57 126 17.1.3. Faked Refreshes and Permissions . . . . . . . . . . 57 127 17.1.4. Fake Data . . . . . . . . . . . . . . . . . . . . . 57 128 17.1.5. Impersonating a Server . . . . . . . . . . . . . . . 58 129 17.1.6. Eavesdropping Traffic . . . . . . . . . . . . . . . 58 130 17.1.7. TURN Loop Attack . . . . . . . . . . . . . . . . . . 59 131 17.2. Firewall Considerations . . . . . . . . . . . . . . . . 60 132 17.2.1. Faked Permissions . . . . . . . . . . . . . . . . . 60 133 17.2.2. Blacklisted IP Addresses . . . . . . . . . . . . . . 61 134 17.2.3. Running Servers on Well-Known Ports . . . . . . . . 61 135 17.3. Insider Attacks . . . . . . . . . . . . . . . . . . . . 61 136 17.3.1. DoS against TURN Server . . . . . . . . . . . . . . 61 137 17.3.2. Anonymous Relaying of Malicious Traffic . . . . . . 62 138 17.3.3. Manipulating Other Allocations . . . . . . . . . . . 62 139 17.4. Other Considerations . . . . . . . . . . . . . . . . . . 62 140 18. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 62 141 19. IAB Considerations . . . . . . . . . . . . . . . . . . . . . 63 142 20. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 64 143 21. References . . . . . . . . . . . . . . . . . . . . . . . . . 64 144 21.1. Normative References . . . . . . . . . . . . . . . . . . 65 145 21.2. Informative References . . . . . . . . . . . . . . . . . 65 147 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 67 149 1. Introduction 151 A host behind a NAT may wish to exchange packets with other hosts, 152 some of which may also be behind NATs. To do this, the hosts 153 involved can use "hole punching" techniques (see [RFC5128]) in an 154 attempt discover a direct communication path; that is, a 155 communication path that goes from one host to another through 156 intervening NATs and routers, but does not traverse any relays. 158 As described in [RFC5128] and [RFC4787], hole punching techniques 159 will fail if both hosts are behind NATs that are not well behaved. 160 For example, if both hosts are behind NATs that have a mapping 161 behavior of "address-dependent mapping" or "address- and port- 162 dependent mapping", then hole punching techniques generally fail. 164 When a direct communication path cannot be found, it is necessary to 165 use the services of an intermediate host that acts as a relay for the 166 packets. This relay typically sits in the public Internet and relays 167 packets between two hosts that both sit behind NATs. 169 This specification defines a protocol, called TURN, that allows a 170 host behind a NAT (called the TURN client) to request that another 171 host (called the TURN server) act as a relay. The client can arrange 172 for the server to relay packets to and from certain other hosts 173 (called peers) and can control aspects of how the relaying is done. 174 The client does this by obtaining an IP address and port on the 175 server, called the relayed transport address. When a peer sends a 176 packet to the relayed transport address, the server relays the packet 177 to the client. When the client sends a data packet to the server, 178 the server relays it to the appropriate peer using the relayed 179 transport address as the source. 181 A client using TURN must have some way to communicate the relayed 182 transport address to its peers, and to learn each peer's IP address 183 and port (more precisely, each peer's server-reflexive transport 184 address, see Section 2). How this is done is out of the scope of the 185 TURN protocol. One way this might be done is for the client and 186 peers to exchange email messages. Another way is for the client and 187 its peers to use a special-purpose "introduction" or "rendezvous" 188 protocol (see [RFC5128] for more details). 190 If TURN is used with ICE [RFC5245], then the relayed transport 191 address and the IP addresses and ports of the peers are included in 192 the ICE candidate information that the rendezvous protocol must 193 carry. For example, if TURN and ICE are used as part of a multimedia 194 solution using SIP [RFC3261], then SIP serves the role of the 195 rendezvous protocol, carrying the ICE candidate information inside 196 the body of SIP messages. If TURN and ICE are used with some other 197 rendezvous protocol, then [I-D.rosenberg-mmusic-ice-nonsip] provides 198 guidance on the services the rendezvous protocol must perform. 200 Though the use of a TURN server to enable communication between two 201 hosts behind NATs is very likely to work, it comes at a high cost to 202 the provider of the TURN server, since the server typically needs a 203 high-bandwidth connection to the Internet . As a consequence, it is 204 best to use a TURN server only when a direct communication path 205 cannot be found. When the client and a peer use ICE to determine the 206 communication path, ICE will use hole punching techniques to search 207 for a direct path first and only use a TURN server when a direct path 208 cannot be found. 210 TURN was originally invented to support multimedia sessions signaled 211 using SIP. Since SIP supports forking, TURN supports multiple peers 212 per relayed transport address; a feature not supported by other 213 approaches (e.g., SOCKS [RFC1928]). However, care has been taken to 214 make sure that TURN is suitable for other types of applications. 216 TURN was designed as one piece in the larger ICE approach to NAT 217 traversal. Implementors of TURN are urged to investigate ICE and 218 seriously consider using it for their application. However, it is 219 possible to use TURN without ICE. 221 TURN is an extension to the STUN (Session Traversal Utilities for 222 NAT) protocol [RFC5389]. Most, though not all, TURN messages are 223 STUN-formatted messages. A reader of this document should be 224 familiar with STUN. 226 2. Overview of Operation 228 This section gives an overview of the operation of TURN. It is non- 229 normative. 231 In a typical configuration, a TURN client is connected to a private 232 network [RFC1918] and through one or more NATs to the public 233 Internet. On the public Internet is a TURN server. Elsewhere in the 234 Internet are one or more peers with which the TURN client wishes to 235 communicate. These peers may or may not be behind one or more NATs. 236 The client uses the server as a relay to send packets to these peers 237 and to receive packets from these peers. 239 Peer A 240 Server-Reflexive +---------+ 241 Transport Address | | 242 192.0.2.150:32102 | | 243 | /| | 244 TURN | / ^| Peer A | 245 Client's Server | / || | 246 Host Transport Transport | // || | 247 Address Address | // |+---------+ 248 10.1.1.2:49721 192.0.2.15:3478 |+-+ // Peer A 249 | | ||N| / Host Transport 250 | +-+ | ||A|/ Address 251 | | | | v|T| 192.168.100.2:49582 252 | | | | /+-+ 253 +---------+| | | |+---------+ / +---------+ 254 | || |N| || | // | | 255 | TURN |v | | v| TURN |/ | | 256 | Client |----|A|----------| Server |------------------| Peer B | 257 | | | |^ | |^ ^| | 258 | | |T|| | || || | 259 +---------+ | || +---------+| |+---------+ 260 | || | | 261 | || | | 262 +-+| | | 263 | | | 264 | | | 265 Client's | Peer B 266 Server-Reflexive Relayed Transport 267 Transport Address Transport Address Address 268 192.0.2.1:7000 192.0.2.15:50000 192.0.2.210:49191 270 Figure 1 272 Figure 1 shows a typical deployment. In this figure, the TURN client 273 and the TURN server are separated by a NAT, with the client on the 274 private side and the server on the public side of the NAT. This NAT 275 is assumed to be a "bad" NAT; for example, it might have a mapping 276 property of "address-and-port-dependent mapping" (see [RFC4787]). 278 The client talks to the server from a (IP address, port) combination 279 called the client's HOST TRANSPORT ADDRESS. (The combination of an 280 IP address and port is called a TRANSPORT ADDRESS.) 282 The client sends TURN messages from its host transport address to a 283 transport address on the TURN server that is known as the TURN SERVER 284 TRANSPORT ADDRESS. The client learns the TURN server transport 285 address through some unspecified means (e.g., configuration), and 286 this address is typically used by many clients simultaneously. 288 Since the client is behind a NAT, the server sees packets from the 289 client as coming from a transport address on the NAT itself. This 290 address is known as the client's SERVER-REFLEXIVE transport address; 291 packets sent by the server to the client's server-reflexive transport 292 address will be forwarded by the NAT to the client's host transport 293 address. 295 The client uses TURN commands to create and manipulate an ALLOCATION 296 on the server. An allocation is a data structure on the server. 297 This data structure contains, amongst other things, the RELAYED 298 TRANSPORT ADDRESS for the allocation. The relayed transport address 299 is the transport address on the server that peers can use to have the 300 server relay data to the client. An allocation is uniquely 301 identified by its relayed transport address. 303 Once an allocation is created, the client can send application data 304 to the server along with an indication of to which peer the data is 305 to be sent, and the server will relay this data to the appropriate 306 peer. The client sends the application data to the server inside a 307 TURN message; at the server, the data is extracted from the TURN 308 message and sent to the peer in a UDP datagram. In the reverse 309 direction, a peer can send application data in a UDP datagram to the 310 relayed transport address for the allocation; the server will then 311 encapsulate this data inside a TURN message and send it to the client 312 along with an indication of which peer sent the data. Since the TURN 313 message always contains an indication of which peer the client is 314 communicating with, the client can use a single allocation to 315 communicate with multiple peers. 317 When the peer is behind a NAT, then the client must identify the peer 318 using its server-reflexive transport address rather than its host 319 transport address. For example, to send application data to Peer A 320 in the example above, the client must specify 192.0.2.150:32102 (Peer 321 A's server-reflexive transport address) rather than 322 192.168.100.2:49582 (Peer A's host transport address). 324 Each allocation on the server belongs to a single client and has 325 exactly one relayed transport address that is used only by that 326 allocation. Thus, when a packet arrives at a relayed transport 327 address on the server, the server knows for which client the data is 328 intended. 330 The client may have multiple allocations on a server at the same 331 time. 333 2.1. Transports 335 TURN, as defined in this specification, always uses UDP between the 336 server and the peer. However, this specification allows the use of 337 any one of UDP, TCP, Transport Layer Security (TLS) over TCP or 338 Datagram Transport Layer Security (DTLS) over UDP to carry the TURN 339 messages between the client and the server. 341 +----------------------------+---------------------+ 342 | TURN client to TURN server | TURN server to peer | 343 +----------------------------+---------------------+ 344 | UDP | UDP | 345 | TCP | UDP | 346 | TLS-over-TCP | UDP | 347 | DTLS-over-UDP | UDP | 348 +----------------------------+---------------------+ 350 If TCP or TLS-over-TCP is used between the client and the server, 351 then the server will convert between these transports and UDP 352 transport when relaying data to/from the peer. 354 Since this version of TURN only supports UDP between the server and 355 the peer, it is expected that most clients will prefer to use UDP 356 between the client and the server as well. That being the case, some 357 readers may wonder: Why also support TCP and TLS-over-TCP? 359 TURN supports TCP transport between the client and the server because 360 some firewalls are configured to block UDP entirely. These firewalls 361 block UDP but not TCP, in part because TCP has properties that make 362 the intention of the nodes being protected by the firewall more 363 obvious to the firewall. For example, TCP has a three-way handshake 364 that makes in clearer that the protected node really wishes to have 365 that particular connection established, while for UDP the best the 366 firewall can do is guess which flows are desired by using filtering 367 rules. Also, TCP has explicit connection teardown; while for UDP, 368 the firewall has to use timers to guess when the flow is finished. 370 TURN supports TLS-over-TCP transport and DTLS-over-UDP transport 371 between the client and the server because (D)TLS provides additional 372 security properties not provided by TURN's default digest 373 authentication; properties that some clients may wish to take 374 advantage of. In particular, (D)TLS provides a way for the client to 375 ascertain that it is talking to the correct server, and provides for 376 confidentiality of TURN control messages. TURN does not require 377 (D)TLS because the overhead of using (D)TLS is higher than that of 378 digest authentication; for example, using (D)TLS likely means that 379 most application data will be doubly encrypted (once by (D)TLS and 380 once to ensure it is still encrypted in the UDP datagram). 382 There is an extension to TURN for TCP transport between the server 383 and the peers [RFC6062]. For this reason, allocations that use UDP 384 between the server and the peers are known as UDP allocations, while 385 allocations that use TCP between the server and the peers are known 386 as TCP allocations. This specification describes only UDP 387 allocations. 389 Editor's Note: Should we merge RFC 6062 text on TCP transport into 390 this document? 392 TURN as defined in this specification, only supports IPv4. All IP 393 addresses in this specification must be IPv4 addresses. TURN usage 394 for IPv6 and for relaying between IPv4 and IPv6 is defined in 395 [RFC6156]. 397 Editor's Note: Should we merge RFC 6156 text on IPv6 into this 398 document? 400 In some applications for TURN, the client may send and receive 401 packets other than TURN packets on the host transport address it uses 402 to communicate with the server. This can happen, for example, when 403 using TURN with ICE. In these cases, the client can distinguish TURN 404 packets from other packets by examining the source address of the 405 arriving packet: those arriving from the TURN server will be TURN 406 packets. 408 2.2. Allocations 410 To create an allocation on the server, the client uses an Allocate 411 transaction. The client sends an Allocate request to the server, and 412 the server replies with an Allocate success response containing the 413 allocated relayed transport address. The client can include 414 attributes in the Allocate request that describe the type of 415 allocation it desires (e.g., the lifetime of the allocation). Since 416 relaying data has security implications, the server requires that the 417 client authenticate itself, typically using STUN's long-term 418 credential mechanism, to show that it is authorized to use the 419 server. 421 Once a relayed transport address is allocated, a client must keep the 422 allocation alive. To do this, the client periodically sends a 423 Refresh request to the server. TURN deliberately uses a different 424 method (Refresh rather than Allocate) for refreshes to ensure that 425 the client is informed if the allocation vanishes for some reason. 427 The frequency of the Refresh transaction is determined by the 428 lifetime of the allocation. The default lifetime of an allocation is 429 10 minutes -- this value was chosen to be long enough so that 430 refreshing is not typically a burden on the client, while expiring 431 allocations where the client has unexpectedly quit in a timely 432 manner. However, the client can request a longer lifetime in the 433 Allocate request and may modify its request in a Refresh request, and 434 the server always indicates the actual lifetime in the response. The 435 client must issue a new Refresh transaction within "lifetime" seconds 436 of the previous Allocate or Refresh transaction. Once a client no 437 longer wishes to use an allocation, it should delete the allocation 438 using a Refresh request with a requested lifetime of 0. 440 Both the server and client keep track of a value known as the 441 5-TUPLE. At the client, the 5-tuple consists of the client's host 442 transport address, the server transport address, and the transport 443 protocol used by the client to communicate with the server. At the 444 server, the 5-tuple value is the same except that the client's host 445 transport address is replaced by the client's server-reflexive 446 address, since that is the client's address as seen by the server. 448 Both the client and the server remember the 5-tuple used in the 449 Allocate request. Subsequent messages between the client and the 450 server use the same 5-tuple. In this way, the client and server know 451 which allocation is being referred to. If the client wishes to 452 allocate a second relayed transport address, it must create a second 453 allocation using a different 5-tuple (e.g., by using a different 454 client host address or port). 456 NOTE: While the terminology used in this document refers to 457 5-tuples, the TURN server can store whatever identifier it likes 458 that yields identical results. Specifically, an implementation 459 may use a file-descriptor in place of a 5-tuple to represent a TCP 460 connection. 462 TURN TURN Peer Peer 463 client server A B 464 |-- Allocate request --------------->| | | 465 | | | | 466 |<--------------- Allocate failure --| | | 467 | (401 Unauthorized) | | | 468 | | | | 469 |-- Allocate request --------------->| | | 470 | | | | 471 |<---------- Allocate success resp --| | | 472 | (192.0.2.15:50000) | | | 473 // // // // 474 | | | | 475 |-- Refresh request ---------------->| | | 476 | | | | 477 |<----------- Refresh success resp --| | | 478 | | | | 480 Figure 2 482 In Figure 2, the client sends an Allocate request to the server 483 without credentials. Since the server requires that all requests be 484 authenticated using STUN's long-term credential mechanism, the server 485 rejects the request with a 401 (Unauthorized) error code. The client 486 then tries again, this time including credentials (not shown). This 487 time, the server accepts the Allocate request and returns an Allocate 488 success response containing (amongst other things) the relayed 489 transport address assigned to the allocation. Sometime later, the 490 client decides to refresh the allocation and thus sends a Refresh 491 request to the server. The refresh is accepted and the server 492 replies with a Refresh success response. 494 2.3. Permissions 496 To ease concerns amongst enterprise IT administrators that TURN could 497 be used to bypass corporate firewall security, TURN includes the 498 notion of permissions. TURN permissions mimic the address-restricted 499 filtering mechanism of NATs that comply with [RFC4787]. 501 An allocation can have zero or more permissions. Each permission 502 consists of an IP address and a lifetime. When the server receives a 503 UDP datagram on the allocation's relayed transport address, it first 504 checks the list of permissions. If the source IP address of the 505 datagram matches a permission, the application data is relayed to the 506 client, otherwise the UDP datagram is silently discarded. 508 A permission expires after 5 minutes if it is not refreshed, and 509 there is no way to explicitly delete a permission. This behavior was 510 selected to match the behavior of a NAT that complies with [RFC4787]. 512 The client can install or refresh a permission using either a 513 CreatePermission request or a ChannelBind request. Using the 514 CreatePermission request, multiple permissions can be installed or 515 refreshed with a single request -- this is important for applications 516 that use ICE. For security reasons, permissions can only be 517 installed or refreshed by transactions that can be authenticated; 518 thus, Send indications and ChannelData messages (which are used to 519 send data to peers) do not install or refresh any permissions. 521 Note that permissions are within the context of an allocation, so 522 adding or expiring a permission in one allocation does not affect 523 other allocations. 525 2.4. Send Mechanism 527 There are two mechanisms for the client and peers to exchange 528 application data using the TURN server. The first mechanism uses the 529 Send and Data methods, the second way uses channels. Common to both 530 ways is the ability of the client to communicate with multiple peers 531 using a single allocated relayed transport address; thus, both ways 532 include a means for the client to indicate to the server which peer 533 should receive the data, and for the server to indicate to the client 534 which peer sent the data. 536 The Send mechanism uses Send and Data indications. Send indications 537 are used to send application data from the client to the server, 538 while Data indications are used to send application data from the 539 server to the client. 541 When using the Send mechanism, the client sends a Send indication to 542 the TURN server containing (a) an XOR-PEER-ADDRESS attribute 543 specifying the (server-reflexive) transport address of the peer and 544 (b) a DATA attribute holding the application data. When the TURN 545 server receives the Send indication, it extracts the application data 546 from the DATA attribute and sends it in a UDP datagram to the peer, 547 using the allocated relay address as the source address. Note that 548 there is no need to specify the relayed transport address, since it 549 is implied by the 5-tuple used for the Send indication. 551 In the reverse direction, UDP datagrams arriving at the relayed 552 transport address on the TURN server are converted into Data 553 indications and sent to the client, with the server-reflexive 554 transport address of the peer included in an XOR-PEER-ADDRESS 555 attribute and the data itself in a DATA attribute. Since the relayed 556 transport address uniquely identified the allocation, the server 557 knows which client should receive the data. 559 Send and Data indications cannot be authenticated, since the long- 560 term credential mechanism of STUN does not support authenticating 561 indications. This is not as big an issue as it might first appear, 562 since the client-to-server leg is only half of the total path to the 563 peer. Applications that want proper security should encrypt the data 564 sent between the client and a peer. 566 Because Send indications are not authenticated, it is possible for an 567 attacker to send bogus Send indications to the server, which will 568 then relay these to a peer. To partly mitigate this attack, TURN 569 requires that the client install a permission towards a peer before 570 sending data to it using a Send indication. 572 TURN TURN Peer Peer 573 client server A B 574 | | | | 575 |-- CreatePermission req (Peer A) -->| | | 576 |<-- CreatePermission success resp --| | | 577 | | | | 578 |--- Send ind (Peer A)-------------->| | | 579 | |=== data ===>| | 580 | | | | 581 | |<== data ====| | 582 |<-------------- Data ind (Peer A) --| | | 583 | | | | 584 | | | | 585 |--- Send ind (Peer B)-------------->| | | 586 | | dropped | | 587 | | | | 588 | |<== data ==================| 589 | dropped | | | 590 | | | | 592 Figure 3 594 In Figure 3, the client has already created an allocation and now 595 wishes to send data to its peers. The client first creates a 596 permission by sending the server a CreatePermission request 597 specifying Peer A's (server-reflexive) IP address in the XOR-PEER- 598 ADDRESS attribute; if this was not done, the server would not relay 599 data between the client and the server. The client then sends data 600 to Peer A using a Send indication; at the server, the application 601 data is extracted and forwarded in a UDP datagram to Peer A, using 602 the relayed transport address as the source transport address. When 603 a UDP datagram from Peer A is received at the relayed transport 604 address, the contents are placed into a Data indication and forwarded 605 to the client. Later, the client attempts to exchange data with Peer 606 B; however, no permission has been installed for Peer B, so the Send 607 indication from the client and the UDP datagram from the peer are 608 both dropped by the server. 610 2.5. Channels 612 For some applications (e.g., Voice over IP), the 36 bytes of overhead 613 that a Send indication or Data indication adds to the application 614 data can substantially increase the bandwidth required between the 615 client and the server. To remedy this, TURN offers a second way for 616 the client and server to associate data with a specific peer. 618 This second way uses an alternate packet format known as the 619 ChannelData message. The ChannelData message does not use the STUN 620 header used by other TURN messages, but instead has a 4-byte header 621 that includes a number known as a channel number. Each channel 622 number in use is bound to a specific peer and thus serves as a 623 shorthand for the peer's host transport address. 625 To bind a channel to a peer, the client sends a ChannelBind request 626 to the server, and includes an unbound channel number and the 627 transport address of the peer. Once the channel is bound, the client 628 can use a ChannelData message to send the server data destined for 629 the peer. Similarly, the server can relay data from that peer 630 towards the client using a ChannelData message. 632 Channel bindings last for 10 minutes unless refreshed -- this 633 lifetime was chosen to be longer than the permission lifetime. 634 Channel bindings are refreshed by sending another ChannelBind request 635 rebinding the channel to the peer. Like permissions (but unlike 636 allocations), there is no way to explicitly delete a channel binding; 637 the client must simply wait for it to time out. 639 TURN TURN Peer Peer 640 client server A B 641 | | | | 642 |-- ChannelBind req ---------------->| | | 643 | (Peer A to 0x4001) | | | 644 | | | | 645 |<---------- ChannelBind succ resp --| | | 646 | | | | 647 |-- [0x4001] data ------------------>| | | 648 | |=== data ===>| | 649 | | | | 650 | |<== data ====| | 651 |<------------------ [0x4001] data --| | | 652 | | | | 653 |--- Send ind (Peer A)-------------->| | | 654 | |=== data ===>| | 655 | | | | 656 | |<== data ====| | 657 |<------------------ [0x4001] data --| | | 658 | | | | 660 Figure 4 662 Figure 4 shows the channel mechanism in use. The client has already 663 created an allocation and now wishes to bind a channel to Peer A. To 664 do this, the client sends a ChannelBind request to the server, 665 specifying the transport address of Peer A and a channel number 666 (0x4001). After that, the client can send application data 667 encapsulated inside ChannelData messages to Peer A: this is shown as 668 "[0x4001] data" where 0x4001 is the channel number. When the 669 ChannelData message arrives at the server, the server transfers the 670 data to a UDP datagram and sends it to Peer A (which is the peer 671 bound to channel number 0x4001). 673 In the reverse direction, when Peer A sends a UDP datagram to the 674 relayed transport address, this UDP datagram arrives at the server on 675 the relayed transport address assigned to the allocation. Since the 676 UDP datagram was received from Peer A, which has a channel number 677 assigned to it, the server encapsulates the data into a ChannelData 678 message when sending the data to the client. 680 Once a channel has been bound, the client is free to intermix 681 ChannelData messages and Send indications. In the figure, the client 682 later decides to use a Send indication rather than a ChannelData 683 message to send additional data to Peer A. The client might decide 684 to do this, for example, so it can use the DONT-FRAGMENT attribute 685 (see the next section). However, once a channel is bound, the server 686 will always use a ChannelData message, as shown in the call flow. 688 Note that ChannelData messages can only be used for peers to which 689 the client has bound a channel. In the example above, Peer A has 690 been bound to a channel, but Peer B has not, so application data to 691 and from Peer B would use the Send mechanism. 693 2.6. Unprivileged TURN Servers 695 This version of TURN is designed so that the server can be 696 implemented as an application that runs in user space under commonly 697 available operating systems without requiring special privileges. 698 This design decision was made to make it easy to deploy a TURN 699 server: for example, to allow a TURN server to be integrated into a 700 peer-to-peer application so that one peer can offer NAT traversal 701 services to another peer. 703 This design decision has the following implications for data relayed 704 by a TURN server: 706 o The value of the Diffserv field may not be preserved across the 707 server; 709 o The Time to Live (TTL) field may be reset, rather than 710 decremented, across the server; 712 o The Explicit Congestion Notification (ECN) field may be reset by 713 the server; 715 o ICMP messages are not relayed by the server; 717 o There is no end-to-end fragmentation, since the packet is re- 718 assembled at the server. 720 Future work may specify alternate TURN semantics that address these 721 limitations. 723 2.7. Avoiding IP Fragmentation 725 For reasons described in [Frag-Harmful], applications, especially 726 those sending large volumes of data, should try hard to avoid having 727 their packets fragmented. Applications using TCP can more or less 728 ignore this issue because fragmentation avoidance is now a standard 729 part of TCP, but applications using UDP (and thus any application 730 using this version of TURN) must handle fragmentation avoidance 731 themselves. 733 The application running on the client and the peer can take one of 734 two approaches to avoid IP fragmentation. 736 The first approach is to avoid sending large amounts of application 737 data in the TURN messages/UDP datagrams exchanged between the client 738 and the peer. This is the approach taken by most VoIP (Voice-over- 739 IP) applications. In this approach, the application exploits the 740 fact that the IP specification [RFC0791] specifies that IP packets up 741 to 576 bytes should never need to be fragmented. 743 The exact amount of application data that can be included while 744 avoiding fragmentation depends on the details of the TURN session 745 between the client and the server: whether UDP, TCP, or (D)TLS 746 transport is used, whether ChannelData messages or Send/Data 747 indications are used, and whether any additional attributes (such as 748 the DONT-FRAGMENT attribute) are included. Another factor, which is 749 hard to determine, is whether the MTU is reduced somewhere along the 750 path for other reasons, such as the use of IP-in-IP tunneling. 752 As a guideline, sending a maximum of 500 bytes of application data in 753 a single TURN message (by the client on the client-to-server leg) or 754 a UDP datagram (by the peer on the peer-to-server leg) will generally 755 avoid IP fragmentation. To further reduce the chance of 756 fragmentation, it is recommended that the client use ChannelData 757 messages when transferring significant volumes of data, since the 758 overhead of the ChannelData message is less than Send and Data 759 indications. 761 The second approach the client and peer can take to avoid 762 fragmentation is to use a path MTU discovery algorithm to determine 763 the maximum amount of application data that can be sent without 764 fragmentation. 766 Unfortunately, because servers implementing this version of TURN do 767 not relay ICMP messages, the classic path MTU discovery algorithm 768 defined in [RFC1191] is not able to discover the MTU of the 769 transmission path between the client and the peer. (Even if they did 770 relay ICMP messages, the algorithm would not always work since ICMP 771 messages are often filtered out by combined NAT/firewall devices). 773 So the client and server need to use a path MTU discovery algorithm 774 that does not require ICMP messages. The Packetized Path MTU 775 Discovery algorithm defined in [RFC4821] is one such algorithm. 777 The details of how to use the algorithm of [RFC4821] with TURN are 778 still under investigation. However, as a step towards this goal, 779 this version of TURN supports a DONT-FRAGMENT attribute. When the 780 client includes this attribute in a Send indication, this tells the 781 server to set the DF bit in the resulting UDP datagram that it sends 782 to the peer. Since some servers may be unable to set the DF bit, the 783 client should also include this attribute in the Allocate request -- 784 any server that does not support the DONT-FRAGMENT attribute will 785 indicate this by rejecting the Allocate request. 787 2.8. RTP Support 789 One of the envisioned uses of TURN is as a relay for clients and 790 peers wishing to exchange real-time data (e.g., voice or video) using 791 RTP. To facilitate the use of TURN for this purpose, TURN includes 792 some special support for older versions of RTP. 794 Old versions of RTP [RFC3550] required that the RTP stream be on an 795 even port number and the associated RTP Control Protocol (RTCP) 796 stream, if present, be on the next highest port. To allow clients to 797 work with peers that still require this, TURN allows the client to 798 request that the server allocate a relayed transport address with an 799 even port number, and to optionally request the server reserve the 800 next-highest port number for a subsequent allocation. 802 2.9. Discovery of Servers 804 Methods of TURN server discovery, including using anycast, are 805 described in [I-D.ietf-tram-turn-server-discovery]. 807 3. Terminology 809 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 810 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 811 document are to be interpreted as described in RFC 2119 [RFC2119]. 813 Readers are expected to be familiar with [RFC5389] and the terms 814 defined there. 816 The following terms are used in this document: 818 TURN: The protocol spoken between a TURN client and a TURN server. 819 It is an extension to the STUN protocol [RFC5389]. The protocol 820 allows a client to allocate and use a relayed transport address. 822 TURN client: A STUN client that implements this specification. 824 TURN server: A STUN server that implements this specification. It 825 relays data between a TURN client and its peer(s). 827 Peer: A host with which the TURN client wishes to communicate. The 828 TURN server relays traffic between the TURN client and its 829 peer(s). The peer does not interact with the TURN server using 830 the protocol defined in this document; rather, the peer receives 831 data sent by the TURN server and the peer sends data towards the 832 TURN server. 834 Transport Address: The combination of an IP address and a port. 836 Host Transport Address: A transport address on a client or a peer. 838 Server-Reflexive Transport Address: A transport address on the 839 "public side" of a NAT. This address is allocated by the NAT to 840 correspond to a specific host transport address. 842 Relayed Transport Address: A transport address on the TURN server 843 that is used for relaying packets between the client and a peer. 844 A peer sends to this address on the TURN server, and the packet is 845 then relayed to the client. 847 TURN Server Transport Address: A transport address on the TURN 848 server that is used for sending TURN messages to the server. This 849 is the transport address that the client uses to communicate with 850 the server. 852 Peer Transport Address: The transport address of the peer as seen by 853 the server. When the peer is behind a NAT, this is the peer's 854 server-reflexive transport address. 856 Allocation: The relayed transport address granted to a client 857 through an Allocate request, along with related state, such as 858 permissions and expiration timers. 860 5-tuple: The combination (client IP address and port, server IP 861 address and port, and transport protocol (currently one of UDP, 862 TCP, or (D)TLS)) used to communicate between the client and the 863 server. The 5-tuple uniquely identifies this communication 864 stream. The 5-tuple also uniquely identifies the Allocation on 865 the server. 867 Channel: A channel number and associated peer transport address. 868 Once a channel number is bound to a peer's transport address, the 869 client and server can use the more bandwidth-efficient ChannelData 870 message to exchange data. 872 Permission: The IP address and transport protocol (but not the port) 873 of a peer that is permitted to send traffic to the TURN server and 874 have that traffic relayed to the TURN client. The TURN server 875 will only forward traffic to its client from peers that match an 876 existing permission. 878 Realm: A string used to describe the server or a context within the 879 server. The realm tells the client which username and password 880 combination to use to authenticate requests. 882 Nonce: A string chosen at random by the server and included in the 883 message-digest. To prevent reply attacks, the server should 884 change the nonce regularly. 886 4. General Behavior 888 This section contains general TURN processing rules that apply to all 889 TURN messages. 891 TURN is an extension to STUN. All TURN messages, with the exception 892 of the ChannelData message, are STUN-formatted messages. All the 893 base processing rules described in [RFC5389] apply to STUN-formatted 894 messages. This means that all the message-forming and message- 895 processing descriptions in this document are implicitly prefixed with 896 the rules of [RFC5389]. 898 [RFC5389] specifies an authentication mechanism called the long-term 899 credential mechanism. TURN servers and clients MUST implement this 900 mechanism. The server MUST demand that all requests from the client 901 be authenticated using this mechanism, or that a equally strong or 902 stronger mechanism for client authentication is used. 904 Note that the long-term credential mechanism applies only to requests 905 and cannot be used to authenticate indications; thus, indications in 906 TURN are never authenticated. If the server requires requests to be 907 authenticated, then the server's administrator MUST choose a realm 908 value that will uniquely identify the username and password 909 combination that the client must use, even if the client uses 910 multiple servers under different administrations. The server's 911 administrator MAY choose to allocate a unique username to each 912 client, or MAY choose to allocate the same username to more than one 913 client (for example, to all clients from the same department or 914 company). For each allocation, the server SHOULD generate a new 915 random nonce when the allocation is first attempted following the 916 randomness recommendations in [RFC4086] and SHOULD expire the nonce 917 at least once every hour during the lifetime of the allocation. 919 All requests after the initial Allocate must use the same username as 920 that used to create the allocation, to prevent attackers from 921 hijacking the client's allocation. Specifically, if the server 922 requires the use of the long-term credential mechanism, and if a non- 923 Allocate request passes authentication under this mechanism, and if 924 the 5-tuple identifies an existing allocation, but the request does 925 not use the same username as used to create the allocation, then the 926 request MUST be rejected with a 441 (Wrong Credentials) error. 928 When a TURN message arrives at the server from the client, the server 929 uses the 5-tuple in the message to identify the associated 930 allocation. For all TURN messages (including ChannelData) EXCEPT an 931 Allocate request, if the 5-tuple does not identify an existing 932 allocation, then the message MUST either be rejected with a 437 933 Allocation Mismatch error (if it is a request) or silently ignored 934 (if it is an indication or a ChannelData message). A client 935 receiving a 437 error response to a request other than Allocate MUST 936 assume the allocation no longer exists. 938 [RFC5389] defines a number of attributes, including the SOFTWARE and 939 FINGERPRINT attributes. The client SHOULD include the SOFTWARE 940 attribute in all Allocate and Refresh requests and MAY include it in 941 any other requests or indications. The server SHOULD include the 942 SOFTWARE attribute in all Allocate and Refresh responses (either 943 success or failure) and MAY include it in other responses or 944 indications. The client and the server MAY include the FINGERPRINT 945 attribute in any STUN-formatted messages defined in this document. 947 TURN does not use the backwards-compatibility mechanism described in 948 [RFC5389]. 950 TURN, as defined in this specification, only supports IPv4. The 951 client's IP address, the server's IP address, and all IP addresses 952 appearing in a relayed transport address MUST be IPv4 addresses. 954 By default, TURN runs on the same ports as STUN: 3478 for TURN over 955 UDP and TCP, and 5349 for TURN over (D)TLS. However, TURN has its 956 own set of Service Record (SRV) names: "turn" for UDP and TCP, and 957 "turns" for (D)TLS. Either the SRV procedures or the ALTERNATE- 958 SERVER procedures, both described in Section 6, can be used to run 959 TURN on a different port. 961 To ensure interoperability, a TURN server MUST support the use of UDP 962 transport between the client and the server, and SHOULD support the 963 use of TCP and (D)TLS transport. 965 When UDP transport is used between the client and the server, the 966 client will retransmit a request if it does not receive a response 967 within a certain timeout period. Because of this, the server may 968 receive two (or more) requests with the same 5-tuple and same 969 transaction id. STUN requires that the server recognize this case 970 and treat the request as idempotent (see [RFC5389]). Some 971 implementations may choose to meet this requirement by remembering 972 all received requests and the corresponding responses for 40 seconds. 974 Other implementations may choose to reprocess the request and arrange 975 that such reprocessing returns essentially the same response. To aid 976 implementors who choose the latter approach (the so-called "stateless 977 stack approach"), this specification includes some implementation 978 notes on how this might be done. Implementations are free to choose 979 either approach or choose some other approach that gives the same 980 results. 982 When TCP transport is used between the client and the server, it is 983 possible that a bit error will cause a length field in a TURN packet 984 to become corrupted, causing the receiver to lose synchronization 985 with the incoming stream of TURN messages. A client or server that 986 detects a long sequence of invalid TURN messages over TCP transport 987 SHOULD close the corresponding TCP connection to help the other end 988 detect this situation more rapidly. 990 To mitigate either intentional or unintentional denial-of-service 991 attacks against the server by clients with valid usernames and 992 passwords, it is RECOMMENDED that the server impose limits on both 993 the number of allocations active at one time for a given username and 994 on the amount of bandwidth those allocations can use. The server 995 should reject new allocations that would exceed the limit on the 996 allowed number of allocations active at one time with a 486 997 (Allocation Quota Exceeded) (see Section 6.2), and should discard 998 application data traffic that exceeds the bandwidth quota. 1000 5. Allocations 1002 All TURN operations revolve around allocations, and all TURN messages 1003 are associated with an allocation. An allocation conceptually 1004 consists of the following state data: 1006 o the relayed transport address; 1008 o the 5-tuple: (client's IP address, client's port, server IP 1009 address, server port, transport protocol); 1011 o the authentication information; 1013 o the time-to-expiry; 1015 o a list of permissions; 1017 o a list of channel to peer bindings. 1019 The relayed transport address is the transport address allocated by 1020 the server for communicating with peers, while the 5-tuple describes 1021 the communication path between the client and the server. On the 1022 client, the 5-tuple uses the client's host transport address; on the 1023 server, the 5-tuple uses the client's server-reflexive transport 1024 address. 1026 Both the relayed transport address and the 5-tuple MUST be unique 1027 across all allocations, so either one can be used to uniquely 1028 identify the allocation. 1030 The authentication information (e.g., username, password, realm, and 1031 nonce) is used to both verify subsequent requests and to compute the 1032 message integrity of responses. The username, realm, and nonce 1033 values are initially those used in the authenticated Allocate request 1034 that creates the allocation, though the server can change the nonce 1035 value during the lifetime of the allocation using a 438 (Stale Nonce) 1036 reply. Note that, rather than storing the password explicitly, for 1037 security reasons, it may be desirable for the server to store the key 1038 value, which is an MD5 hash over the username, realm, and password 1039 (see [RFC5389]). 1041 Editor's Note: Remove MD5 based on the changes in STUN bis draft. 1043 The time-to-expiry is the time in seconds left until the allocation 1044 expires. Each Allocate or Refresh transaction sets this timer, which 1045 then ticks down towards 0. By default, each Allocate or Refresh 1046 transaction resets this timer to the default lifetime value of 600 1047 seconds (10 minutes), but the client can request a different value in 1048 the Allocate and Refresh request. Allocations can only be refreshed 1049 using the Refresh request; sending data to a peer does not refresh an 1050 allocation. When an allocation expires, the state data associated 1051 with the allocation can be freed. 1053 The list of permissions is described in Section 8 and the list of 1054 channels is described in Section 11. 1056 6. Creating an Allocation 1058 An allocation on the server is created using an Allocate transaction. 1060 6.1. Sending an Allocate Request 1062 The client forms an Allocate request as follows. 1064 The client first picks a host transport address. It is RECOMMENDED 1065 that the client pick a currently unused transport address, typically 1066 by allowing the underlying OS to pick a currently unused port for a 1067 new socket. 1069 The client then picks a transport protocol to use between the client 1070 and the server. The transport protocol MUST be one of UDP, TCP, TLS- 1071 over-TCP or DTLS-over-UDP. Since this specification only allows UDP 1072 between the server and the peers, it is RECOMMENDED that the client 1073 pick UDP unless it has a reason to use a different transport. One 1074 reason to pick a different transport would be that the client 1075 believes, either through configuration or by experiment, that it is 1076 unable to contact any TURN server using UDP. See Section 2.1 for 1077 more discussion. 1079 The client also picks a server transport address, which SHOULD be 1080 done as follows. The client receives (perhaps through configuration) 1081 a domain name for a TURN server. The client then uses the DNS 1082 procedures described in [RFC5389], but using an SRV service name of 1083 "turn" (or "turns" for TURN over (D)TLS) instead of "stun" (or 1084 "stuns"). For example, to find servers in the example.com domain, 1085 the client performs a lookup for '_turn._udp.example.com', 1086 '_turn._tcp.example.com', and '_turns._tcp.example.com' if the client 1087 wants to communicate with the server using UDP, TCP, TLS-over-TCP, or 1088 DTLS-over-UDP, respectively. 1090 The client MUST include a REQUESTED-TRANSPORT attribute in the 1091 request. This attribute specifies the transport protocol between the 1092 server and the peers (note that this is NOT the transport protocol 1093 that appears in the 5-tuple). In this specification, the REQUESTED- 1094 TRANSPORT type is always UDP. This attribute is included to allow 1095 future extensions to specify other protocols. 1097 If the client wishes the server to initialize the time-to-expiry 1098 field of the allocation to some value other than the default 1099 lifetime, then it MAY include a LIFETIME attribute specifying its 1100 desired value. This is just a request, and the server may elect to 1101 use a different value. Note that the server will ignore requests to 1102 initialize the field to less than the default value. 1104 If the client wishes to later use the DONT-FRAGMENT attribute in one 1105 or more Send indications on this allocation, then the client SHOULD 1106 include the DONT-FRAGMENT attribute in the Allocate request. This 1107 allows the client to test whether this attribute is supported by the 1108 server. 1110 If the client requires the port number of the relayed transport 1111 address be even, the client includes the EVEN-PORT attribute. If 1112 this attribute is not included, then the port can be even or odd. By 1113 setting the R bit in the EVEN-PORT attribute to 1, the client can 1114 request that the server reserve the next highest port number (on the 1115 same IP address) for a subsequent allocation. If the R bit is 0, no 1116 such request is made. 1118 The client MAY also include a RESERVATION-TOKEN attribute in the 1119 request to ask the server to use a previously reserved port for the 1120 allocation. If the RESERVATION-TOKEN attribute is included, then the 1121 client MUST omit the EVEN-PORT attribute. 1123 Once constructed, the client sends the Allocate request on the 1124 5-tuple. 1126 6.2. Receiving an Allocate Request 1128 When the server receives an Allocate request, it performs the 1129 following checks: 1131 1. The server MUST require that the request be authenticated. This 1132 authentication MUST be done using the long-term credential 1133 mechanism of [RFC5389] unless the client and server agree to use 1134 another mechanism through some procedure outside the scope of 1135 this document. 1137 2. The server checks if the 5-tuple is currently in use by an 1138 existing allocation. If yes, the server rejects the request with 1139 a 437 (Allocation Mismatch) error. 1141 3. The server checks if the request contains a REQUESTED-TRANSPORT 1142 attribute. If the REQUESTED-TRANSPORT attribute is not included 1143 or is malformed, the server rejects the request with a 400 (Bad 1144 Request) error. Otherwise, if the attribute is included but 1145 specifies a protocol other that UDP, the server rejects the 1146 request with a 442 (Unsupported Transport Protocol) error. 1148 4. The request may contain a DONT-FRAGMENT attribute. If it does, 1149 but the server does not support sending UDP datagrams with the DF 1150 bit set to 1 (see Section 12), then the server treats the DONT- 1151 FRAGMENT attribute in the Allocate request as an unknown 1152 comprehension-required attribute. 1154 5. The server checks if the request contains a RESERVATION-TOKEN 1155 attribute. If yes, and the request also contains an EVEN-PORT 1156 attribute, then the server rejects the request with a 400 (Bad 1157 Request) error. Otherwise, it checks to see if the token is 1158 valid (i.e., the token is in range and has not expired and the 1159 corresponding relayed transport address is still available). If 1160 the token is not valid for some reason, the server rejects the 1161 request with a 508 (Insufficient Capacity) error. 1163 6. The server checks if the request contains an EVEN-PORT attribute. 1164 If yes, then the server checks that it can satisfy the request 1165 (i.e., can allocate a relayed transport address as described 1166 below). If the server cannot satisfy the request, then the 1167 server rejects the request with a 508 (Insufficient Capacity) 1168 error. 1170 7. At any point, the server MAY choose to reject the request with a 1171 486 (Allocation Quota Reached) error if it feels the client is 1172 trying to exceed some locally defined allocation quota. The 1173 server is free to define this allocation quota any way it wishes, 1174 but SHOULD define it based on the username used to authenticate 1175 the request, and not on the client's transport address. 1177 8. Also at any point, the server MAY choose to reject the request 1178 with a 300 (Try Alternate) error if it wishes to redirect the 1179 client to a different server. The use of this error code and 1180 attribute follow the specification in [RFC5389]. 1182 If all the checks pass, the server creates the allocation. The 1183 5-tuple is set to the 5-tuple from the Allocate request, while the 1184 list of permissions and the list of channels are initially empty. 1186 The server chooses a relayed transport address for the allocation as 1187 follows: 1189 o If the request contains a RESERVATION-TOKEN, the server uses the 1190 previously reserved transport address corresponding to the 1191 included token (if it is still available). Note that the 1192 reservation is a server-wide reservation and is not specific to a 1193 particular allocation, since the Allocate request containing the 1194 RESERVATION-TOKEN uses a different 5-tuple than the Allocate 1195 request that made the reservation. The 5-tuple for the Allocate 1196 request containing the RESERVATION-TOKEN attribute can be any 1197 allowed 5-tuple; it can use a different client IP address and 1198 port, a different transport protocol, and even different server IP 1199 address and port (provided, of course, that the server IP address 1200 and port are ones on which the server is listening for TURN 1201 requests). 1203 o If the request contains an EVEN-PORT attribute with the R bit set 1204 to 0, then the server allocates a relayed transport address with 1205 an even port number. 1207 o If the request contains an EVEN-PORT attribute with the R bit set 1208 to 1, then the server looks for a pair of port numbers N and N+1 1209 on the same IP address, where N is even. Port N is used in the 1210 current allocation, while the relayed transport address with port 1211 N+1 is assigned a token and reserved for a future allocation. The 1212 server MUST hold this reservation for at least 30 seconds, and MAY 1213 choose to hold longer (e.g., until the allocation with port N 1214 expires). The server then includes the token in a RESERVATION- 1215 TOKEN attribute in the success response. 1217 o Otherwise, the server allocates any available relayed transport 1218 address. 1220 In all cases, the server SHOULD only allocate ports from the range 1221 49152 - 65535 (the Dynamic and/or Private Port range [Port-Numbers]), 1222 unless the TURN server application knows, through some means not 1223 specified here, that other applications running on the same host as 1224 the TURN server application will not be impacted by allocating ports 1225 outside this range. This condition can often be satisfied by running 1226 the TURN server application on a dedicated machine and/or by 1227 arranging that any other applications on the machine allocate ports 1228 before the TURN server application starts. In any case, the TURN 1229 server SHOULD NOT allocate ports in the range 0 - 1023 (the Well- 1230 Known Port range) to discourage clients from using TURN to run 1231 standard services. 1233 NOTE: The use of randomized port assignments to avoid certain 1234 types of attacks is described in [RFC6056]. It is RECOMMENDED 1235 that a TURN server implement a randomized port assignment 1236 algorithm from [RFC6056]. This is especially applicable to 1237 servers that choose to pre-allocate a number of ports from the 1238 underlying OS and then later assign them to allocations; for 1239 example, a server may choose this technique to implement the EVEN- 1240 PORT attribute. 1242 Editor's Note: Should we recommend a specific algorithm from RFC 1243 6056? 1245 The server determines the initial value of the time-to-expiry field 1246 as follows. If the request contains a LIFETIME attribute, then the 1247 server computes the minimum of the client's proposed lifetime and the 1248 server's maximum allowed lifetime. If this computed value is greater 1249 than the default lifetime, then the server uses the computed lifetime 1250 as the initial value of the time-to-expiry field. Otherwise, the 1251 server uses the default lifetime. It is RECOMMENDED that the server 1252 use a maximum allowed lifetime value of no more than 3600 seconds (1 1253 hour). Servers that implement allocation quotas or charge users for 1254 allocations in some way may wish to use a smaller maximum allowed 1255 lifetime (perhaps as small as the default lifetime) to more quickly 1256 remove orphaned allocations (that is, allocations where the 1257 corresponding client has crashed or terminated or the client 1258 connection has been lost for some reason). Also, note that the time- 1259 to-expiry is recomputed with each successful Refresh request, and 1260 thus the value computed here applies only until the first refresh. 1262 Once the allocation is created, the server replies with a success 1263 response. The success response contains: 1265 o An XOR-RELAYED-ADDRESS attribute containing the relayed transport 1266 address. 1268 o A LIFETIME attribute containing the current value of the time-to- 1269 expiry timer. 1271 o A RESERVATION-TOKEN attribute (if a second relayed transport 1272 address was reserved). 1274 o An XOR-MAPPED-ADDRESS attribute containing the client's IP address 1275 and port (from the 5-tuple). 1277 NOTE: The XOR-MAPPED-ADDRESS attribute is included in the response 1278 as a convenience to the client. TURN itself does not make use of 1279 this value, but clients running ICE can often need this value and 1280 can thus avoid having to do an extra Binding transaction with some 1281 STUN server to learn it. 1283 The response (either success or error) is sent back to the client on 1284 the 5-tuple. 1286 NOTE: When the Allocate request is sent over UDP, section 7.3.1 of 1287 [RFC5389] requires that the server handle the possible 1288 retransmissions of the request so that retransmissions do not 1289 cause multiple allocations to be created. Implementations may 1290 achieve this using the so-called "stateless stack approach" as 1291 follows. To detect retransmissions when the original request was 1292 successful in creating an allocation, the server can store the 1293 transaction id that created the request with the allocation data 1294 and compare it with incoming Allocate requests on the same 1295 5-tuple. Once such a request is detected, the server can stop 1296 parsing the request and immediately generate a success response. 1297 When building this response, the value of the LIFETIME attribute 1298 can be taken from the time-to-expiry field in the allocate state 1299 data, even though this value may differ slightly from the LIFETIME 1300 value originally returned. In addition, the server may need to 1301 store an indication of any reservation token returned in the 1302 original response, so that this may be returned in any 1303 retransmitted responses. 1305 For the case where the original request was unsuccessful in 1306 creating an allocation, the server may choose to do nothing 1307 special. Note, however, that there is a rare case where the 1308 server rejects the original request but accepts the retransmitted 1309 request (because conditions have changed in the brief intervening 1310 time period). If the client receives the first failure response, 1311 it will ignore the second (success) response and believe that an 1312 allocation was not created. An allocation created in this matter 1313 will eventually timeout, since the client will not refresh it. 1314 Furthermore, if the client later retries with the same 5-tuple but 1315 different transaction id, it will receive a 437 (Allocation 1316 Mismatch), which will cause it to retry with a different 5-tuple. 1317 The server may use a smaller maximum lifetime value to minimize 1318 the lifetime of allocations "orphaned" in this manner. 1320 6.3. Receiving an Allocate Success Response 1322 If the client receives an Allocate success response, then it MUST 1323 check that the mapped address and the relayed transport address are 1324 in an address family that the client understands and is prepared to 1325 handle. This specification only covers the case where these two 1326 addresses are IPv4 addresses. If these two addresses are not in an 1327 address family which the client is prepared to handle, then the 1328 client MUST delete the allocation (Section 7) and MUST NOT attempt to 1329 create another allocation on that server until it believes the 1330 mismatch has been fixed. 1332 The IETF is currently considering mechanisms for transitioning 1333 between IPv4 and IPv6 that could result in a client originating an 1334 Allocate request over IPv6, but the request would arrive at the 1335 server over IPv4, or vice versa. 1337 Editor's Note: This text on IPv6 should be updated. 1339 Otherwise, the client creates its own copy of the allocation data 1340 structure to track what is happening on the server. In particular, 1341 the client needs to remember the actual lifetime received back from 1342 the server, rather than the value sent to the server in the request. 1343 The client must also remember the 5-tuple used for the request and 1344 the username and password it used to authenticate the request to 1345 ensure that it reuses them for subsequent messages. The client also 1346 needs to track the channels and permissions it establishes on the 1347 server. 1349 The client will probably wish to send the relayed transport address 1350 to peers (using some method not specified here) so the peers can 1351 communicate with it. The client may also wish to use the server- 1352 reflexive address it receives in the XOR-MAPPED-ADDRESS attribute in 1353 its ICE processing. 1355 6.4. Receiving an Allocate Error Response 1357 If the client receives an Allocate error response, then the 1358 processing depends on the actual error code returned: 1360 o (Request timed out): There is either a problem with the server, or 1361 a problem reaching the server with the chosen transport. The 1362 client considers the current transaction as having failed but MAY 1363 choose to retry the Allocate request using a different transport 1364 (e.g., TCP instead of UDP). 1366 o 300 (Try Alternate): The server would like the client to use the 1367 server specified in the ALTERNATE-SERVER attribute instead. The 1368 client considers the current transaction as having failed, but 1369 SHOULD try the Allocate request with the alternate server before 1370 trying any other servers (e.g., other servers discovered using the 1371 SRV procedures). When trying the Allocate request with the 1372 alternate server, the client follows the ALTERNATE-SERVER 1373 procedures specified in [RFC5389]. 1375 o 400 (Bad Request): The server believes the client's request is 1376 malformed for some reason. The client considers the current 1377 transaction as having failed. The client MAY notify the user or 1378 operator and SHOULD NOT retry the request with this server until 1379 it believes the problem has been fixed. 1381 o 401 (Unauthorized): If the client has followed the procedures of 1382 the long-term credential mechanism and still gets this error, then 1383 the server is not accepting the client's credentials. In this 1384 case, the client considers the current transaction as having 1385 failed and SHOULD notify the user or operator. The client SHOULD 1386 NOT send any further requests to this server until it believes the 1387 problem has been fixed. 1389 o 403 (Forbidden): The request is valid, but the server is refusing 1390 to perform it, likely due to administrative restrictions. The 1391 client considers the current transaction as having failed. The 1392 client MAY notify the user or operator and SHOULD NOT retry the 1393 same request with this server until it believes the problem has 1394 been fixed. 1396 o 420 (Unknown Attribute): If the client included a DONT-FRAGMENT 1397 attribute in the request and the server rejected the request with 1398 a 420 error code and listed the DONT-FRAGMENT attribute in the 1399 UNKNOWN-ATTRIBUTES attribute in the error response, then the 1400 client now knows that the server does not support the DONT- 1401 FRAGMENT attribute. The client considers the current transaction 1402 as having failed but MAY choose to retry the Allocate request 1403 without the DONT-FRAGMENT attribute. 1405 o 437 (Allocation Mismatch): This indicates that the client has 1406 picked a 5-tuple that the server sees as already in use. One way 1407 this could happen is if an intervening NAT assigned a mapped 1408 transport address that was used by another client that recently 1409 crashed. The client considers the current transaction as having 1410 failed. The client SHOULD pick another client transport address 1411 and retry the Allocate request (using a different transaction id). 1412 The client SHOULD try three different client transport addresses 1413 before giving up on this server. Once the client gives up on the 1414 server, it SHOULD NOT try to create another allocation on the 1415 server for 2 minutes. 1417 o 438 (Stale Nonce): See the procedures for the long-term credential 1418 mechanism [RFC5389]. 1420 o 441 (Wrong Credentials): The client should not receive this error 1421 in response to a Allocate request. The client MAY notify the user 1422 or operator and SHOULD NOT retry the same request with this server 1423 until it believes the problem has been fixed. 1425 o 442 (Unsupported Transport Address): The client should not receive 1426 this error in response to a request for a UDP allocation. The 1427 client MAY notify the user or operator and SHOULD NOT reattempt 1428 the request with this server until it believes the problem has 1429 been fixed. 1431 o 486 (Allocation Quota Reached): The server is currently unable to 1432 create any more allocations with this username. The client 1433 considers the current transaction as having failed. The client 1434 SHOULD wait at least 1 minute before trying to create any more 1435 allocations on the server. 1437 o 508 (Insufficient Capacity): The server has no more relayed 1438 transport addresses available, or has none with the requested 1439 properties, or the one that was reserved is no longer available. 1440 The client considers the current operation as having failed. If 1441 the client is using either the EVEN-PORT or the RESERVATION-TOKEN 1442 attribute, then the client MAY choose to remove or modify this 1443 attribute and try again immediately. Otherwise, the client SHOULD 1444 wait at least 1 minute before trying to create any more 1445 allocations on this server. 1447 An unknown error response MUST be handled as described in [RFC5389]. 1449 7. Refreshing an Allocation 1451 A Refresh transaction can be used to either (a) refresh an existing 1452 allocation and update its time-to-expiry or (b) delete an existing 1453 allocation. 1455 If a client wishes to continue using an allocation, then the client 1456 MUST refresh it before it expires. It is suggested that the client 1457 refresh the allocation roughly 1 minute before it expires. If a 1458 client no longer wishes to use an allocation, then it SHOULD 1459 explicitly delete the allocation. A client MAY refresh an allocation 1460 at any time for other reasons. 1462 7.1. Sending a Refresh Request 1464 If the client wishes to immediately delete an existing allocation, it 1465 includes a LIFETIME attribute with a value of 0. All other forms of 1466 the request refresh the allocation. 1468 The Refresh transaction updates the time-to-expiry timer of an 1469 allocation. If the client wishes the server to set the time-to- 1470 expiry timer to something other than the default lifetime, it 1471 includes a LIFETIME attribute with the requested value. The server 1472 then computes a new time-to-expiry value in the same way as it does 1473 for an Allocate transaction, with the exception that a requested 1474 lifetime of 0 causes the server to immediately delete the allocation. 1476 7.2. Receiving a Refresh Request 1478 When the server receives a Refresh request, it processes as per 1479 Section 4 plus the specific rules mentioned here. 1481 The server computes a value called the "desired lifetime" as follows: 1482 if the request contains a LIFETIME attribute and the attribute value 1483 is 0, then the "desired lifetime" is 0. Otherwise, if the request 1484 contains a LIFETIME attribute, then the server computes the minimum 1485 of the client's requested lifetime and the server's maximum allowed 1486 lifetime. If this computed value is greater than the default 1487 lifetime, then the "desired lifetime" is the computed value. 1488 Otherwise, the "desired lifetime" is the default lifetime. 1490 Subsequent processing depends on the "desired lifetime" value: 1492 o If the "desired lifetime" is 0, then the request succeeds and the 1493 allocation is deleted. 1495 o If the "desired lifetime" is non-zero, then the request succeeds 1496 and the allocation's time-to-expiry is set to the "desired 1497 lifetime". 1499 If the request succeeds, then the server sends a success response 1500 containing: 1502 o A LIFETIME attribute containing the current value of the time-to- 1503 expiry timer. 1505 NOTE: A server need not do anything special to implement 1506 idempotency of Refresh requests over UDP using the "stateless 1507 stack approach". Retransmitted Refresh requests with a non-zero 1508 "desired lifetime" will simply refresh the allocation. A 1509 retransmitted Refresh request with a zero "desired lifetime" will 1510 cause a 437 (Allocation Mismatch) response if the allocation has 1511 already been deleted, but the client will treat this as equivalent 1512 to a success response (see below). 1514 7.3. Receiving a Refresh Response 1516 If the client receives a success response to its Refresh request with 1517 a non-zero lifetime, it updates its copy of the allocation data 1518 structure with the time-to-expiry value contained in the response. 1520 If the client receives a 437 (Allocation Mismatch) error response to 1521 a request to delete the allocation, then the allocation no longer 1522 exists and it should consider its request as having effectively 1523 succeeded. 1525 8. Permissions 1527 For each allocation, the server keeps a list of zero or more 1528 permissions. Each permission consists of an IP address and an 1529 associated time-to-expiry. While a permission exists, all peers 1530 using the IP address in the permission are allowed to send data to 1531 the client. The time-to-expiry is the number of seconds until the 1532 permission expires. Within the context of an allocation, a 1533 permission is uniquely identified by its associated IP address. 1535 By sending either CreatePermission requests or ChannelBind requests, 1536 the client can cause the server to install or refresh a permission 1537 for a given IP address. This causes one of two things to happen: 1539 o If no permission for that IP address exists, then a permission is 1540 created with the given IP address and a time-to-expiry equal to 1541 Permission Lifetime. 1543 o If a permission for that IP address already exists, then the time- 1544 to-expiry for that permission is reset to Permission Lifetime. 1546 The Permission Lifetime MUST be 300 seconds (= 5 minutes). 1548 Each permission's time-to-expiry decreases down once per second until 1549 it reaches 0; at which point, the permission expires and is deleted. 1551 CreatePermission and ChannelBind requests may be freely intermixed on 1552 a permission. A given permission may be initially installed and/or 1553 refreshed with a CreatePermission request, and then later refreshed 1554 with a ChannelBind request, or vice versa. 1556 When a UDP datagram arrives at the relayed transport address for the 1557 allocation, the server extracts the source IP address from the IP 1558 header. The server then compares this address with the IP address 1559 associated with each permission in the list of permissions for the 1560 allocation. If no match is found, relaying is not permitted, and the 1561 server silently discards the UDP datagram. If an exact match is 1562 found, then the permission check is considered to have succeeded and 1563 the server continues to process the UDP datagram as specified 1564 elsewhere (Section 10.3). Note that only addresses are compared and 1565 port numbers are not considered. 1567 The permissions for one allocation are totally unrelated to the 1568 permissions for a different allocation. If an allocation expires, 1569 all its permissions expire with it. 1571 NOTE: Though TURN permissions expire after 5 minutes, many NATs 1572 deployed at the time of publication expire their UDP bindings 1573 considerably faster. Thus, an application using TURN will 1574 probably wish to send some sort of keep-alive traffic at a much 1575 faster rate. Applications using ICE should follow the keep-alive 1576 guidelines of ICE [RFC5245], and applications not using ICE are 1577 advised to do something similar. 1579 9. CreatePermission 1581 TURN supports two ways for the client to install or refresh 1582 permissions on the server. This section describes one way: the 1583 CreatePermission request. 1585 A CreatePermission request may be used in conjunction with either the 1586 Send mechanism in Section 10 or the Channel mechanism in Section 11. 1588 9.1. Forming a CreatePermission Request 1590 The client who wishes to install or refresh one or more permissions 1591 can send a CreatePermission request to the server. 1593 When forming a CreatePermission request, the client MUST include at 1594 least one XOR-PEER-ADDRESS attribute, and MAY include more than one 1595 such attribute. The IP address portion of each XOR-PEER-ADDRESS 1596 attribute contains the IP address for which a permission should be 1597 installed or refreshed. The port portion of each XOR-PEER-ADDRESS 1598 attribute will be ignored and can be any arbitrary value. The 1599 various XOR-PEER-ADDRESS attributes can appear in any order. 1601 9.2. Receiving a CreatePermission Request 1603 When the server receives the CreatePermission request, it processes 1604 as per Section 4 plus the specific rules mentioned here. 1606 The message is checked for validity. The CreatePermission request 1607 MUST contain at least one XOR-PEER-ADDRESS attribute and MAY contain 1608 multiple such attributes. If no such attribute exists, or if any of 1609 these attributes are invalid, then a 400 (Bad Request) error is 1610 returned. If the request is valid, but the server is unable to 1611 satisfy the request due to some capacity limit or similar, then a 508 1612 (Insufficient Capacity) error is returned. 1614 The server MAY impose restrictions on the IP address allowed in the 1615 XOR-PEER-ADDRESS attribute -- if a value is not allowed, the server 1616 rejects the request with a 403 (Forbidden) error. 1618 If the message is valid and the server is capable of carrying out the 1619 request, then the server installs or refreshes a permission for the 1620 IP address contained in each XOR-PEER-ADDRESS attribute as described 1621 in Section 8. The port portion of each attribute is ignored and may 1622 be any arbitrary value. 1624 The server then responds with a CreatePermission success response. 1625 There are no mandatory attributes in the success response. 1627 NOTE: A server need not do anything special to implement 1628 idempotency of CreatePermission requests over UDP using the 1629 "stateless stack approach". Retransmitted CreatePermission 1630 requests will simply refresh the permissions. 1632 9.3. Receiving a CreatePermission Response 1634 If the client receives a valid CreatePermission success response, 1635 then the client updates its data structures to indicate that the 1636 permissions have been installed or refreshed. 1638 10. Send and Data Methods 1640 TURN supports two mechanisms for sending and receiving data from 1641 peers. This section describes the use of the Send and Data 1642 mechanisms, while Section 11 describes the use of the Channel 1643 mechanism. 1645 10.1. Forming a Send Indication 1647 The client can use a Send indication to pass data to the server for 1648 relaying to a peer. A client may use a Send indication even if a 1649 channel is bound to that peer. However, the client MUST ensure that 1650 there is a permission installed for the IP address of the peer to 1651 which the Send indication is being sent; this prevents a third party 1652 from using a TURN server to send data to arbitrary destinations. 1654 When forming a Send indication, the client MUST include an XOR-PEER- 1655 ADDRESS attribute and a DATA attribute. The XOR-PEER-ADDRESS 1656 attribute contains the transport address of the peer to which the 1657 data is to be sent, and the DATA attribute contains the actual 1658 application data to be sent to the peer. 1660 The client MAY include a DONT-FRAGMENT attribute in the Send 1661 indication if it wishes the server to set the DF bit on the UDP 1662 datagram sent to the peer. 1664 10.2. Receiving a Send Indication 1666 When the server receives a Send indication, it processes as per 1667 Section 4 plus the specific rules mentioned here. 1669 The message is first checked for validity. The Send indication MUST 1670 contain both an XOR-PEER-ADDRESS attribute and a DATA attribute. If 1671 one of these attributes is missing or invalid, then the message is 1672 discarded. Note that the DATA attribute is allowed to contain zero 1673 bytes of data. 1675 The Send indication may also contain the DONT-FRAGMENT attribute. If 1676 the server is unable to set the DF bit on outgoing UDP datagrams when 1677 this attribute is present, then the server acts as if the DONT- 1678 FRAGMENT attribute is an unknown comprehension-required attribute 1679 (and thus the Send indication is discarded). 1681 The server also checks that there is a permission installed for the 1682 IP address contained in the XOR-PEER-ADDRESS attribute. If no such 1683 permission exists, the message is discarded. Note that a Send 1684 indication never causes the server to refresh the permission. 1686 The server MAY impose restrictions on the IP address and port values 1687 allowed in the XOR-PEER-ADDRESS attribute -- if a value is not 1688 allowed, the server silently discards the Send indication. 1690 If everything is OK, then the server forms a UDP datagram as follows: 1692 o the source transport address is the relayed transport address of 1693 the allocation, where the allocation is determined by the 5-tuple 1694 on which the Send indication arrived; 1696 o the destination transport address is taken from the XOR-PEER- 1697 ADDRESS attribute; 1699 o the data following the UDP header is the contents of the value 1700 field of the DATA attribute. 1702 The handling of the DONT-FRAGMENT attribute (if present), is 1703 described in Section 12. 1705 The resulting UDP datagram is then sent to the peer. 1707 10.3. Receiving a UDP Datagram 1709 When the server receives a UDP datagram at a currently allocated 1710 relayed transport address, the server looks up the allocation 1711 associated with the relayed transport address. The server then 1712 checks to see whether the set of permissions for the allocation allow 1713 the relaying of the UDP datagram as described in Section 8. 1715 If relaying is permitted, then the server checks if there is a 1716 channel bound to the peer that sent the UDP datagram (see 1717 Section 11). If a channel is bound, then processing proceeds as 1718 described in Section 11.7. 1720 If relaying is permitted but no channel is bound to the peer, then 1721 the server forms and sends a Data indication. The Data indication 1722 MUST contain both an XOR-PEER-ADDRESS and a DATA attribute. The DATA 1723 attribute is set to the value of the 'data octets' field from the 1724 datagram, and the XOR-PEER-ADDRESS attribute is set to the source 1725 transport address of the received UDP datagram. The Data indication 1726 is then sent on the 5-tuple associated with the allocation. 1728 10.4. Receiving a Data Indication 1730 When the client receives a Data indication, it checks that the Data 1731 indication contains both an XOR-PEER-ADDRESS and a DATA attribute, 1732 and discards the indication if it does not. The client SHOULD also 1733 check that the XOR-PEER-ADDRESS attribute value contains an IP 1734 address with which the client believes there is an active permission, 1735 and discard the Data indication otherwise. Note that the DATA 1736 attribute is allowed to contain zero bytes of data. 1738 NOTE: The latter check protects the client against an attacker who 1739 somehow manages to trick the server into installing permissions 1740 not desired by the client. 1742 If the Data indication passes the above checks, the client delivers 1743 the data octets inside the DATA attribute to the application, along 1744 with an indication that they were received from the peer whose 1745 transport address is given by the XOR-PEER-ADDRESS attribute. 1747 11. Channels 1749 Channels provide a way for the client and server to send application 1750 data using ChannelData messages, which have less overhead than Send 1751 and Data indications. 1753 The ChannelData message (see Section 11.4) starts with a two-byte 1754 field that carries the channel number. The values of this field are 1755 allocated as follows: 1757 0x0000 through 0x3FFF: These values can never be used for channel 1758 numbers. 1760 0x4000 through 0x7FFF: These values are the allowed channel 1761 numbers (16,384 possible values). 1763 0x8000 through 0xFFFF: These values are reserved for future use. 1765 Because of this division, ChannelData messages can be distinguished 1766 from STUN-formatted messages (e.g., Allocate request, Send 1767 indication, etc.) by examining the first two bits of the message: 1769 0b00: STUN-formatted message (since the first two bits of a STUN- 1770 formatted message are always zero). 1772 0b01: ChannelData message (since the channel number is the first 1773 field in the ChannelData message and channel numbers fall in the 1774 range 0x4000 - 0x7FFF). 1776 0b10: Reserved 1778 0b11: Reserved 1780 The reserved values may be used in the future to extend the range of 1781 channel numbers. Thus, an implementation MUST NOT assume that a TURN 1782 message always starts with a 0 bit. 1784 Channel bindings are always initiated by the client. The client can 1785 bind a channel to a peer at any time during the lifetime of the 1786 allocation. The client may bind a channel to a peer before 1787 exchanging data with it, or after exchanging data with it (using Send 1788 and Data indications) for some time, or may choose never to bind a 1789 channel to it. The client can also bind channels to some peers while 1790 not binding channels to other peers. 1792 Channel bindings are specific to an allocation, so that the use of a 1793 channel number or peer transport address in a channel binding in one 1794 allocation has no impact on their use in a different allocation. If 1795 an allocation expires, all its channel bindings expire with it. 1797 A channel binding consists of: 1799 o a channel number; 1801 o a transport address (of the peer); and 1803 o A time-to-expiry timer. 1805 Within the context of an allocation, a channel binding is uniquely 1806 identified either by the channel number or by the peer's transport 1807 address. Thus, the same channel cannot be bound to two different 1808 transport addresses, nor can the same transport address be bound to 1809 two different channels. 1811 A channel binding lasts for 10 minutes unless refreshed. Refreshing 1812 the binding (by the server receiving a ChannelBind request rebinding 1813 the channel to the same peer) resets the time-to-expiry timer back to 1814 10 minutes. 1816 When the channel binding expires, the channel becomes unbound. Once 1817 unbound, the channel number can be bound to a different transport 1818 address, and the transport address can be bound to a different 1819 channel number. To prevent race conditions, the client MUST wait 5 1820 minutes after the channel binding expires before attempting to bind 1821 the channel number to a different transport address or the transport 1822 address to a different channel number. 1824 When binding a channel to a peer, the client SHOULD be prepared to 1825 receive ChannelData messages on the channel from the server as soon 1826 as it has sent the ChannelBind request. Over UDP, it is possible for 1827 the client to receive ChannelData messages from the server before it 1828 receives a ChannelBind success response. 1830 In the other direction, the client MAY elect to send ChannelData 1831 messages before receiving the ChannelBind success response. Doing 1832 so, however, runs the risk of having the ChannelData messages dropped 1833 by the server if the ChannelBind request does not succeed for some 1834 reason (e.g., packet lost if the request is sent over UDP, or the 1835 server being unable to fulfill the request). A client that wishes to 1836 be safe should either queue the data or use Send indications until 1837 the channel binding is confirmed. 1839 11.1. Sending a ChannelBind Request 1841 A channel binding is created or refreshed using a ChannelBind 1842 transaction. A ChannelBind transaction also creates or refreshes a 1843 permission towards the peer (see Section 8). 1845 To initiate the ChannelBind transaction, the client forms a 1846 ChannelBind request. The channel to be bound is specified in a 1847 CHANNEL-NUMBER attribute, and the peer's transport address is 1848 specified in an XOR-PEER-ADDRESS attribute. Section 11.2 describes 1849 the restrictions on these attributes. 1851 Rebinding a channel to the same transport address that it is already 1852 bound to provides a way to refresh a channel binding and the 1853 corresponding permission without sending data to the peer. Note 1854 however, that permissions need to be refreshed more frequently than 1855 channels. 1857 11.2. Receiving a ChannelBind Request 1859 When the server receives a ChannelBind request, it processes as per 1860 Section 4 plus the specific rules mentioned here. 1862 The server checks the following: 1864 o The request contains both a CHANNEL-NUMBER and an XOR-PEER-ADDRESS 1865 attribute; 1867 o The channel number is in the range 0x4000 through 0x7FFE 1868 (inclusive); 1870 o The channel number is not currently bound to a different transport 1871 address (same transport address is OK); 1873 o The transport address is not currently bound to a different 1874 channel number. 1876 If any of these tests fail, the server replies with a 400 (Bad 1877 Request) error. 1879 The server MAY impose restrictions on the IP address and port values 1880 allowed in the XOR-PEER-ADDRESS attribute -- if a value is not 1881 allowed, the server rejects the request with a 403 (Forbidden) error. 1883 If the request is valid, but the server is unable to fulfill the 1884 request due to some capacity limit or similar, the server replies 1885 with a 508 (Insufficient Capacity) error. 1887 Otherwise, the server replies with a ChannelBind success response. 1888 There are no required attributes in a successful ChannelBind 1889 response. 1891 If the server can satisfy the request, then the server creates or 1892 refreshes the channel binding using the channel number in the 1893 CHANNEL-NUMBER attribute and the transport address in the XOR-PEER- 1894 ADDRESS attribute. The server also installs or refreshes a 1895 permission for the IP address in the XOR-PEER-ADDRESS attribute as 1896 described in Section 8. 1898 NOTE: A server need not do anything special to implement 1899 idempotency of ChannelBind requests over UDP using the "stateless 1900 stack approach". Retransmitted ChannelBind requests will simply 1901 refresh the channel binding and the corresponding permission. 1902 Furthermore, the client must wait 5 minutes before binding a 1903 previously bound channel number or peer address to a different 1904 channel, eliminating the possibility that the transaction would 1905 initially fail but succeed on a retransmission. 1907 11.3. Receiving a ChannelBind Response 1909 When the client receives a ChannelBind success response, it updates 1910 its data structures to record that the channel binding is now active. 1911 It also updates its data structures to record that the corresponding 1912 permission has been installed or refreshed. 1914 If the client receives a ChannelBind failure response that indicates 1915 that the channel information is out-of-sync between the client and 1916 the server (e.g., an unexpected 400 "Bad Request" response), then it 1917 is RECOMMENDED that the client immediately delete the allocation and 1918 start afresh with a new allocation. 1920 11.4. The ChannelData Message 1922 The ChannelData message is used to carry application data between the 1923 client and the server. It has the following format: 1925 0 1 2 3 1926 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 1927 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 1928 | Channel Number | Length | 1929 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 1930 | | 1931 / Application Data / 1932 / / 1933 | | 1934 | +-------------------------------+ 1935 | | 1936 +-------------------------------+ 1938 The Channel Number field specifies the number of the channel on which 1939 the data is traveling, and thus the address of the peer that is 1940 sending or is to receive the data. 1942 The Length field specifies the length in bytes of the application 1943 data field (i.e., it does not include the size of the ChannelData 1944 header). Note that 0 is a valid length. 1946 The Application Data field carries the data the client is trying to 1947 send to the peer, or that the peer is sending to the client. 1949 11.5. Sending a ChannelData Message 1951 Once a client has bound a channel to a peer, then when the client has 1952 data to send to that peer it may use either a ChannelData message or 1953 a Send indication; that is, the client is not obligated to use the 1954 channel when it exists and may freely intermix the two message types 1955 when sending data to the peer. The server, on the other hand, MUST 1956 use the ChannelData message if a channel has been bound to the peer. 1958 The fields of the ChannelData message are filled in as described in 1959 Section 11.4. 1961 Over TCP and TLS-over-TCP, the ChannelData message MUST be padded to 1962 a multiple of four bytes in order to ensure the alignment of 1963 subsequent messages. The padding is not reflected in the length 1964 field of the ChannelData message, so the actual size of a ChannelData 1965 message (including padding) is (4 + Length) rounded up to the nearest 1966 multiple of 4. Over UDP, the padding is not required but MAY be 1967 included. 1969 The ChannelData message is then sent on the 5-tuple associated with 1970 the allocation. 1972 11.6. Receiving a ChannelData Message 1974 The receiver of the ChannelData message uses the first two bits to 1975 distinguish it from STUN-formatted messages, as described above. If 1976 the message uses a value in the reserved range (0x8000 through 1977 0xFFFF), then the message is silently discarded. 1979 If the ChannelData message is received in a UDP datagram, and if the 1980 UDP datagram is too short to contain the claimed length of the 1981 ChannelData message (i.e., the UDP header length field value is less 1982 than the ChannelData header length field value + 4 + 8), then the 1983 message is silently discarded. 1985 If the ChannelData message is received over TCP or over TLS-over-TCP, 1986 then the actual length of the ChannelData message is as described in 1987 Section 11.5. 1989 If the ChannelData message is received on a channel that is not bound 1990 to any peer, then the message is silently discarded. 1992 On the client, it is RECOMMENDED that the client discard the 1993 ChannelData message if the client believes there is no active 1994 permission towards the peer. On the server, the receipt of a 1995 ChannelData message MUST NOT refresh either the channel binding or 1996 the permission towards the peer. 1998 On the server, if no errors are detected, the server relays the 1999 application data to the peer by forming a UDP datagram as follows: 2001 o the source transport address is the relayed transport address of 2002 the allocation, where the allocation is determined by the 5-tuple 2003 on which the ChannelData message arrived; 2005 o the destination transport address is the transport address to 2006 which the channel is bound; 2008 o the data following the UDP header is the contents of the data 2009 field of the ChannelData message. 2011 The resulting UDP datagram is then sent to the peer. Note that if 2012 the Length field in the ChannelData message is 0, then there will be 2013 no data in the UDP datagram, but the UDP datagram is still formed and 2014 sent. 2016 11.7. Relaying Data from the Peer 2018 When the server receives a UDP datagram on the relayed transport 2019 address associated with an allocation, the server processes it as 2020 described in Section 10.3. If that section indicates that a 2021 ChannelData message should be sent (because there is a channel bound 2022 to the peer that sent to the UDP datagram), then the server forms and 2023 sends a ChannelData message as described in Section 11.5. 2025 12. IP Header Fields 2027 This section describes how the server sets various fields in the IP 2028 header when relaying between the client and the peer or vice versa. 2029 The descriptions in this section apply: (a) when the server sends a 2030 UDP datagram to the peer, or (b) when the server sends a Data 2031 indication or ChannelData message to the client over UDP transport. 2032 The descriptions in this section do not apply to TURN messages sent 2033 over TCP or TLS transport from the server to the client. 2035 The descriptions below have two parts: a preferred behavior and an 2036 alternate behavior. The server SHOULD implement the preferred 2037 behavior, but if that is not possible for a particular field, then it 2038 SHOULD implement the alternative behavior. 2040 Time to Live (TTL) field 2042 Preferred Behavior: If the incoming value is 0, then the drop the 2043 incoming packet. Otherwise, set the outgoing Time to Live/Hop 2044 Count to one less than the incoming value. 2046 Alternate Behavior: Set the outgoing value to the default for 2047 outgoing packets. 2049 Differentiated Services Code Point (DSCP) field [RFC2474] 2051 Preferred Behavior: Set the outgoing value to the incoming value, 2052 unless the server includes a differentiated services classifier 2053 and marker [RFC2474]. 2055 Alternate Behavior: Set the outgoing value to a fixed value, which 2056 by default is Best Effort unless configured otherwise. 2058 In both cases, if the server is immediately adjacent to a 2059 differentiated services classifier and marker, then DSCP MAY be 2060 set to any arbitrary value in the direction towards the 2061 classifier. 2063 Explicit Congestion Notification (ECN) field [RFC3168] 2065 Preferred Behavior: Set the outgoing value to the incoming value, 2066 UNLESS the server is doing Active Queue Management, the incoming 2067 ECN field is ECT(1) (=0b01) or ECT(0) (=0b10), and the server 2068 wishes to indicate that congestion has been experienced, in which 2069 case set the outgoing value to CE (=0b11). 2071 Alternate Behavior: Set the outgoing value to Not-ECT (=0b00). 2073 IPv4 Fragmentation fields 2075 Preferred Behavior: When the server sends a packet to a peer in 2076 response to a Send indication containing the DONT-FRAGMENT 2077 attribute, then set the DF bit in the outgoing IP header to 1. In 2078 all other cases when sending an outgoing packet containing 2079 application data (e.g., Data indication, ChannelData message, or 2080 DONT-FRAGMENT attribute not included in the Send indication), copy 2081 the DF bit from the DF bit of the incoming packet that contained 2082 the application data. 2084 Set the other fragmentation fields (Identification, More 2085 Fragments, Fragment Offset) as appropriate for a packet 2086 originating from the server. 2088 Alternate Behavior: As described in the Preferred Behavior, except 2089 always assume the incoming DF bit is 0. 2091 In both the Preferred and Alternate Behaviors, the resulting 2092 packet may be too large for the outgoing link. If this is the 2093 case, then the normal fragmentation rules apply [RFC1122]. 2095 IPv4 Options 2097 Preferred Behavior: The outgoing packet is sent without any IPv4 2098 options. 2100 Alternate Behavior: Same as preferred. 2102 13. New STUN Methods 2104 This section lists the codepoints for the new STUN methods defined in 2105 this specification. See elsewhere in this document for the semantics 2106 of these new methods. 2108 0x003 : Allocate (only request/response semantics defined) 2109 0x004 : Refresh (only request/response semantics defined) 2110 0x006 : Send (only indication semantics defined) 2111 0x007 : Data (only indication semantics defined) 2112 0x008 : CreatePermission (only request/response semantics defined 2113 0x009 : ChannelBind (only request/response semantics defined) 2115 14. New STUN Attributes 2117 This STUN extension defines the following new attributes: 2119 0x000C: CHANNEL-NUMBER 2120 0x000D: LIFETIME 2121 0x0010: Reserved (was BANDWIDTH) 2122 0x0012: XOR-PEER-ADDRESS 2123 0x0013: DATA 2124 0x0016: XOR-RELAYED-ADDRESS 2125 0x0018: EVEN-PORT 2126 0x0019: REQUESTED-TRANSPORT 2127 0x001A: DONT-FRAGMENT 2128 0x0021: Reserved (was TIMER-VAL) 2129 0x0022: RESERVATION-TOKEN 2131 Some of these attributes have lengths that are not multiples of 4. 2132 By the rules of STUN, any attribute whose length is not a multiple of 2133 4 bytes MUST be immediately followed by 1 to 3 padding bytes to 2134 ensure the next attribute (if any) would start on a 4-byte boundary 2135 (see [RFC5389]). 2137 14.1. CHANNEL-NUMBER 2139 The CHANNEL-NUMBER attribute contains the number of the channel. The 2140 value portion of this attribute is 4 bytes long and consists of a 2141 16-bit unsigned integer, followed by a two-octet RFFU (Reserved For 2142 Future Use) field, which MUST be set to 0 on transmission and MUST be 2143 ignored on reception. 2145 0 1 2 3 2146 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 2147 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 2148 | Channel Number | RFFU = 0 | 2149 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 2151 14.2. LIFETIME 2153 The LIFETIME attribute represents the duration for which the server 2154 will maintain an allocation in the absence of a refresh. The value 2155 portion of this attribute is 4-bytes long and consists of a 32-bit 2156 unsigned integral value representing the number of seconds remaining 2157 until expiration. 2159 14.3. XOR-PEER-ADDRESS 2161 The XOR-PEER-ADDRESS specifies the address and port of the peer as 2162 seen from the TURN server. (For example, the peer's server-reflexive 2163 transport address if the peer is behind a NAT.) It is encoded in the 2164 same way as XOR-MAPPED-ADDRESS [RFC5389]. 2166 14.4. DATA 2168 The DATA attribute is present in all Send and Data indications. The 2169 value portion of this attribute is variable length and consists of 2170 the application data (that is, the data that would immediately follow 2171 the UDP header if the data was been sent directly between the client 2172 and the peer). If the length of this attribute is not a multiple of 2173 4, then padding must be added after this attribute. 2175 14.5. XOR-RELAYED-ADDRESS 2177 The XOR-RELAYED-ADDRESS is present in Allocate responses. It 2178 specifies the address and port that the server allocated to the 2179 client. It is encoded in the same way as XOR-MAPPED-ADDRESS 2180 [RFC5389]. 2182 14.6. EVEN-PORT 2184 This attribute allows the client to request that the port in the 2185 relayed transport address be even, and (optionally) that the server 2186 reserve the next-higher port number. The value portion of this 2187 attribute is 1 byte long. Its format is: 2189 0 2190 0 1 2 3 4 5 6 7 2191 +-+-+-+-+-+-+-+-+ 2192 |R| RFFU | 2193 +-+-+-+-+-+-+-+-+ 2195 The value contains a single 1-bit flag: 2197 R: If 1, the server is requested to reserve the next-higher port 2198 number (on the same IP address) for a subsequent allocation. If 2199 0, no such reservation is requested. 2201 The other 7 bits of the attribute's value must be set to zero on 2202 transmission and ignored on reception. 2204 Since the length of this attribute is not a multiple of 4, padding 2205 must immediately follow this attribute. 2207 14.7. REQUESTED-TRANSPORT 2209 This attribute is used by the client to request a specific transport 2210 protocol for the allocated transport address. The value of this 2211 attribute is 4 bytes with the following format: 2213 0 1 2 3 2214 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 2215 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 2216 | Protocol | RFFU | 2217 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 2219 The Protocol field specifies the desired protocol. The codepoints 2220 used in this field are taken from those allowed in the Protocol field 2221 in the IPv4 header and the NextHeader field in the IPv6 header 2222 [Protocol-Numbers]. This specification only allows the use of 2223 codepoint 17 (User Datagram Protocol). 2225 The RFFU field MUST be set to zero on transmission and MUST be 2226 ignored on reception. It is reserved for future uses. 2228 14.8. DONT-FRAGMENT 2230 This attribute is used by the client to request that the server set 2231 the DF (Don't Fragment) bit in the IP header when relaying the 2232 application data onward to the peer. This attribute has no value 2233 part and thus the attribute length field is 0. 2235 14.9. RESERVATION-TOKEN 2237 The RESERVATION-TOKEN attribute contains a token that uniquely 2238 identifies a relayed transport address being held in reserve by the 2239 server. The server includes this attribute in a success response to 2240 tell the client about the token, and the client includes this 2241 attribute in a subsequent Allocate request to request the server use 2242 that relayed transport address for the allocation. 2244 The attribute value is 8 bytes and contains the token value. 2246 15. New STUN Error Response Codes 2248 This document defines the following new error response codes: 2250 403 (Forbidden): The request was valid but cannot be performed due 2251 to administrative or similar restrictions. 2253 437 (Allocation Mismatch): A request was received by the server that 2254 requires an allocation to be in place, but no allocation exists, 2255 or a request was received that requires no allocation, but an 2256 allocation exists. 2258 441 (Wrong Credentials): The credentials in the (non-Allocate) 2259 request do not match those used to create the allocation. 2261 442 (Unsupported Transport Protocol): The Allocate request asked the 2262 server to use a transport protocol between the server and the peer 2263 that the server does not support. NOTE: This does NOT refer to 2264 the transport protocol used in the 5-tuple. 2266 486 (Allocation Quota Reached): No more allocations using this 2267 username can be created at the present time. 2269 508 (Insufficient Capacity): The server is unable to carry out the 2270 request due to some capacity limit being reached. In an Allocate 2271 response, this could be due to the server having no more relayed 2272 transport addresses available at that time, having none with the 2273 requested properties, or the one that corresponds to the specified 2274 reservation token is not available. 2276 16. Detailed Example 2278 This section gives an example of the use of TURN, showing in detail 2279 the contents of the messages exchanged. The example uses the network 2280 diagram shown in the Overview (Figure 1). 2282 For each message, the attributes included in the message and their 2283 values are shown. For convenience, values are shown in a human- 2284 readable format rather than showing the actual octets; for example, 2285 "XOR-RELAYED-ADDRESS=192.0.2.15:9000" shows that the XOR-RELAYED- 2286 ADDRESS attribute is included with an address of 192.0.2.15 and a 2287 port of 9000, here the address and port are shown before the xor-ing 2288 is done. For attributes with string-like values (e.g., 2289 SOFTWARE="Example client, version 1.03" and 2290 NONCE="adl7W7PeDU4hKE72jdaQvbAMcr6h39sm"), the value of the attribute 2291 is shown in quotes for readability, but these quotes do not appear in 2292 the actual value. 2294 TURN TURN Peer Peer 2295 client server A B 2296 | | | | 2297 |--- Allocate request -------------->| | | 2298 | Transaction-Id=0xA56250D3F17ABE679422DE85 | | 2299 | SOFTWARE="Example client, version 1.03" | | 2300 | LIFETIME=3600 (1 hour) | | | 2301 | REQUESTED-TRANSPORT=17 (UDP) | | | 2302 | DONT-FRAGMENT | | | 2303 | | | | 2304 |<-- Allocate error response --------| | | 2305 | Transaction-Id=0xA56250D3F17ABE679422DE85 | | 2306 | SOFTWARE="Example server, version 1.17" | | 2307 | ERROR-CODE=401 (Unauthorized) | | | 2308 | REALM="example.com" | | | 2309 | NONCE="adl7W7PeDU4hKE72jdaQvbAMcr6h39sm" | | 2310 | | | | 2311 |--- Allocate request -------------->| | | 2312 | Transaction-Id=0xC271E932AD7446A32C234492 | | 2313 | SOFTWARE="Example client 1.03" | | | 2314 | LIFETIME=3600 (1 hour) | | | 2315 | REQUESTED-TRANSPORT=17 (UDP) | | | 2316 | DONT-FRAGMENT | | | 2317 | USERNAME="George" | | | 2318 | REALM="example.com" | | | 2319 | NONCE="adl7W7PeDU4hKE72jdaQvbAMcr6h39sm" | | 2320 | MESSAGE-INTEGRITY=... | | | 2321 | | | | 2322 |<-- Allocate success response ------| | | 2323 | Transaction-Id=0xC271E932AD7446A32C234492 | | 2324 | SOFTWARE="Example server, version 1.17" | | 2325 | LIFETIME=1200 (20 minutes) | | | 2326 | XOR-RELAYED-ADDRESS=192.0.2.15:50000 | | 2327 | XOR-MAPPED-ADDRESS=192.0.2.1:7000 | | 2328 | MESSAGE-INTEGRITY=... | | | 2330 The client begins by selecting a host transport address to use for 2331 the TURN session; in this example, the client has selected 2332 10.1.1.2:49721 as shown in Figure 1. The client then sends an 2333 Allocate request to the server at the server transport address. The 2334 client randomly selects a 96-bit transaction id of 2335 0xA56250D3F17ABE679422DE85 for this transaction; this is encoded in 2336 the transaction id field in the fixed header. The client includes a 2337 SOFTWARE attribute that gives information about the client's 2338 software; here the value is "Example client, version 1.03" to 2339 indicate that this is version 1.03 of something called the Example 2340 client. The client includes the LIFETIME attribute because it wishes 2341 the allocation to have a longer lifetime than the default of 10 2342 minutes; the value of this attribute is 3600 seconds, which 2343 corresponds to 1 hour. The client must always include a REQUESTED- 2344 TRANSPORT attribute in an Allocate request and the only value allowed 2345 by this specification is 17, which indicates UDP transport between 2346 the server and the peers. The client also includes the DONT-FRAGMENT 2347 attribute because it wishes to use the DONT-FRAGMENT attribute later 2348 in Send indications; this attribute consists of only an attribute 2349 header, there is no value part. We assume the client has not 2350 recently interacted with the server, thus the client does not include 2351 USERNAME, REALM, NONCE, or MESSAGE-INTEGRITY attribute. Finally, 2352 note that the order of attributes in a message is arbitrary (except 2353 for the MESSAGE-INTEGRITY and FINGERPRINT attributes) and the client 2354 could have used a different order. 2356 Servers require any request to be authenticated. Thus, when the 2357 server receives the initial Allocate request, it rejects the request 2358 because the request does not contain the authentication attributes. 2359 Following the procedures of the long-term credential mechanism of 2360 STUN [RFC5389], the server includes an ERROR-CODE attribute with a 2361 value of 401 (Unauthorized), a REALM attribute that specifies the 2362 authentication realm used by the server (in this case, the server's 2363 domain "example.com"), and a nonce value in a NONCE attribute. The 2364 server also includes a SOFTWARE attribute that gives information 2365 about the server's software. 2367 The client, upon receipt of the 401 error, re-attempts the Allocate 2368 request, this time including the authentication attributes. The 2369 client selects a new transaction id, and then populates the new 2370 Allocate request with the same attributes as before. The client 2371 includes a USERNAME attribute and uses the realm value received from 2372 the server to help it determine which value to use; here the client 2373 is configured to use the username "George" for the realm 2374 "example.com". The client also includes the REALM and NONCE 2375 attributes, which are just copied from the 401 error response. 2376 Finally, the client includes a MESSAGE-INTEGRITY attribute as the 2377 last attribute in the message, whose value is a Hashed Message 2378 Authentication Code - Secure Hash Algorithm 1 (HMAC-SHA1) hash over 2379 the contents of the message (shown as just "..." above); this HMAC- 2380 SHA1 computation includes a password value. Thus, an attacker cannot 2381 compute the message integrity value without somehow knowing the 2382 secret password. 2384 The server, upon receipt of the authenticated Allocate request, 2385 checks that everything is OK, then creates an allocation. The server 2386 replies with an Allocate success response. The server includes a 2387 LIFETIME attribute giving the lifetime of the allocation; here, the 2388 server has reduced the client's requested 1-hour lifetime to just 20 2389 minutes, because this particular server doesn't allow lifetimes 2390 longer than 20 minutes. The server includes an XOR-RELAYED-ADDRESS 2391 attribute whose value is the relayed transport address of the 2392 allocation. The server includes an XOR-MAPPED-ADDRESS attribute 2393 whose value is the server-reflexive address of the client; this value 2394 is not used otherwise in TURN but is returned as a convenience to the 2395 client. The server includes a MESSAGE-INTEGRITY attribute to 2396 authenticate the response and to ensure its integrity; note that the 2397 response does not contain the USERNAME, REALM, and NONCE attributes. 2398 The server also includes a SOFTWARE attribute. 2400 TURN TURN Peer Peer 2401 client server A B 2402 |--- CreatePermission request ------>| | | 2403 | Transaction-Id=0xE5913A8F460956CA277D3319 | | 2404 | XOR-PEER-ADDRESS=192.0.2.150:0 | | | 2405 | USERNAME="George" | | | 2406 | REALM="example.com" | | | 2407 | NONCE="adl7W7PeDU4hKE72jdaQvbAMcr6h39sm" | | 2408 | MESSAGE-INTEGRITY=... | | | 2409 | | | | 2410 |<-- CreatePermission success resp.--| | | 2411 | Transaction-Id=0xE5913A8F460956CA277D3319 | | 2412 | MESSAGE-INTEGRITY=... | | | 2414 The client then creates a permission towards Peer A in preparation 2415 for sending it some application data. This is done through a 2416 CreatePermission request. The XOR-PEER-ADDRESS attribute contains 2417 the IP address for which a permission is established (the IP address 2418 of peer A); note that the port number in the attribute is ignored 2419 when used in a CreatePermission request, and here it has been set to 2420 0; also, note how the client uses Peer A's server-reflexive IP 2421 address and not its (private) host address. The client uses the same 2422 username, realm, and nonce values as in the previous request on the 2423 allocation. Though it is allowed to do so, the client has chosen not 2424 to include a SOFTWARE attribute in this request. 2426 The server receives the CreatePermission request, creates the 2427 corresponding permission, and then replies with a CreatePermission 2428 success response. Like the client, the server chooses not to include 2429 the SOFTWARE attribute in its reply. Again, note how success 2430 responses contain a MESSAGE-INTEGRITY attribute (assuming the server 2431 uses the long-term credential mechanism), but no USERNAME, REALM, and 2432 NONCE attributes. 2434 TURN TURN Peer Peer 2435 client server A B 2436 |--- Send indication --------------->| | | 2437 | Transaction-Id=0x1278E9ACA2711637EF7D3328 | | 2438 | XOR-PEER-ADDRESS=192.0.2.150:32102 | | 2439 | DONT-FRAGMENT | | | 2440 | DATA=... | | | 2441 | |-- UDP dgm ->| | 2442 | | data=... | | 2443 | | | | 2444 | |<- UDP dgm --| | 2445 | | data=... | | 2446 |<-- Data indication ----------------| | | 2447 | Transaction-Id=0x8231AE8F9242DA9FF287FEFF | | 2448 | XOR-PEER-ADDRESS=192.0.2.150:32102 | | 2449 | DATA=... | | | 2451 The client now sends application data to Peer A using a Send 2452 indication. Peer A's server-reflexive transport address is specified 2453 in the XOR-PEER-ADDRESS attribute, and the application data (shown 2454 here as just "...") is specified in the DATA attribute. The client 2455 is doing a form of path MTU discovery at the application layer and 2456 thus specifies (by including the DONT-FRAGMENT attribute) that the 2457 server should set the DF bit in the UDP datagram to send to the peer. 2458 Indications cannot be authenticated using the long-term credential 2459 mechanism of STUN, so no MESSAGE-INTEGRITY attribute is included in 2460 the message. An application wishing to ensure that its data is not 2461 altered or forged must integrity-protect its data at the application 2462 level. 2464 Upon receipt of the Send indication, the server extracts the 2465 application data and sends it in a UDP datagram to Peer A, with the 2466 relayed transport address as the source transport address of the 2467 datagram, and with the DF bit set as requested. Note that, had the 2468 client not previously established a permission for Peer A's server- 2469 reflexive IP address, then the server would have silently discarded 2470 the Send indication instead. 2472 Peer A then replies with its own UDP datagram containing application 2473 data. The datagram is sent to the relayed transport address on the 2474 server. When this arrives, the server creates a Data indication 2475 containing the source of the UDP datagram in the XOR-PEER-ADDRESS 2476 attribute, and the data from the UDP datagram in the DATA attribute. 2477 The resulting Data indication is then sent to the client. 2479 TURN TURN Peer Peer 2480 client server A B 2481 |--- ChannelBind request ----------->| | | 2482 | Transaction-Id=0x6490D3BC175AFF3D84513212 | | 2483 | CHANNEL-NUMBER=0x4000 | | | 2484 | XOR-PEER-ADDRESS=192.0.2.210:49191 | | 2485 | USERNAME="George" | | | 2486 | REALM="example.com" | | | 2487 | NONCE="adl7W7PeDU4hKE72jdaQvbAMcr6h39sm" | | 2488 | MESSAGE-INTEGRITY=... | | | 2489 | | | | 2490 |<-- ChannelBind success response ---| | | 2491 | Transaction-Id=0x6490D3BC175AFF3D84513212 | | 2492 | MESSAGE-INTEGRITY=... | | | 2494 The client now binds a channel to Peer B, specifying a free channel 2495 number (0x4000) in the CHANNEL-NUMBER attribute, and Peer B's 2496 transport address in the XOR-PEER-ADDRESS attribute. As before, the 2497 client re-uses the username, realm, and nonce from its last request 2498 in the message. 2500 Upon receipt of the request, the server binds the channel number to 2501 the peer, installs a permission for Peer B's IP address, and then 2502 replies with ChannelBind success response. 2504 TURN TURN Peer Peer 2505 client server A B 2506 |--- ChannelData ------------------->| | | 2507 | Channel-number=0x4000 |--- UDP datagram --------->| 2508 | Data=... | Data=... | 2509 | | | | 2510 | |<-- UDP datagram ----------| 2511 | | Data=... | | 2512 |<-- ChannelData --------------------| | | 2513 | Channel-number=0x4000 | | | 2514 | Data=... | | | 2516 The client now sends a ChannelData message to the server with data 2517 destined for Peer B. The ChannelData message is not a STUN message, 2518 and thus has no transaction id. Instead, it has only three fields: a 2519 channel number, data, and data length; here the channel number field 2520 is 0x4000 (the channel the client just bound to Peer B). When the 2521 server receives the ChannelData message, it checks that the channel 2522 is currently bound (which it is) and then sends the data onward to 2523 Peer B in a UDP datagram, using the relayed transport address as the 2524 source transport address and 192.0.2.210:49191 (the value of the XOR- 2525 PEER-ADDRESS attribute in the ChannelBind request) as the destination 2526 transport address. 2528 Later, Peer B sends a UDP datagram back to the relayed transport 2529 address. This causes the server to send a ChannelData message to the 2530 client containing the data from the UDP datagram. The server knows 2531 to which client to send the ChannelData message because of the 2532 relayed transport address at which the UDP datagram arrived, and 2533 knows to use channel 0x4000 because this is the channel bound to 2534 192.0.2.210:49191. Note that if there had not been any channel 2535 number bound to that address, the server would have used a Data 2536 indication instead. 2538 TURN TURN Peer Peer 2539 client server A B 2540 |--- Refresh request --------------->| | | 2541 | Transaction-Id=0x0864B3C27ADE9354B4312414 | | 2542 | SOFTWARE="Example client 1.03" | | | 2543 | USERNAME="George" | | | 2544 | REALM="example.com" | | | 2545 | NONCE="adl7W7PeDU4hKE72jdaQvbAMcr6h39sm" | | 2546 | MESSAGE-INTEGRITY=... | | | 2547 | | | | 2548 |<-- Refresh error response ---------| | | 2549 | Transaction-Id=0x0864B3C27ADE9354B4312414 | | 2550 | SOFTWARE="Example server, version 1.17" | | 2551 | ERROR-CODE=438 (Stale Nonce) | | | 2552 | REALM="example.com" | | | 2553 | NONCE="npSw1Xw239bBwGYhjNWgz2yH47sxB2j" | | 2554 | | | | 2555 |--- Refresh request --------------->| | | 2556 | Transaction-Id=0x427BD3E625A85FC731DC4191 | | 2557 | SOFTWARE="Example client 1.03" | | | 2558 | USERNAME="George" | | | 2559 | REALM="example.com" | | | 2560 | NONCE="npSw1Xw239bBwGYhjNWgz2yH47sxB2j" | | 2561 | MESSAGE-INTEGRITY=... | | | 2562 | | | | 2563 |<-- Refresh success response -------| | | 2564 | Transaction-Id=0x427BD3E625A85FC731DC4191 | | 2565 | SOFTWARE="Example server, version 1.17" | | 2566 | LIFETIME=600 (10 minutes) | | | 2568 Sometime before the 20 minute lifetime is up, the client refreshes 2569 the allocation. This is done using a Refresh request. As before, 2570 the client includes the latest username, realm, and nonce values in 2571 the request. The client also includes the SOFTWARE attribute, 2572 following the recommended practice of always including this attribute 2573 in Allocate and Refresh messages. When the server receives the 2574 Refresh request, it notices that the nonce value has expired, and so 2575 replies with 438 (Stale Nonce) error given a new nonce value. The 2576 client then reattempts the request, this time with the new nonce 2577 value. This second attempt is accepted, and the server replies with 2578 a success response. Note that the client did not include a LIFETIME 2579 attribute in the request, so the server refreshes the allocation for 2580 the default lifetime of 10 minutes (as can be seen by the LIFETIME 2581 attribute in the success response). 2583 17. Security Considerations 2585 This section considers attacks that are possible in a TURN 2586 deployment, and discusses how they are mitigated by mechanisms in the 2587 protocol or recommended practices in the implementation. 2589 Most of the attacks on TURN are mitigated by the server requiring 2590 requests be authenticated. Thus, this specification requires the use 2591 of authentication. The mandatory-to-implement mechanism is the long- 2592 term credential mechanism of STUN. Other authentication mechanisms 2593 of equal or stronger security properties may be used. However, it is 2594 important to ensure that they can be invoked in an inter-operable 2595 way. 2597 17.1. Outsider Attacks 2599 Outsider attacks are ones where the attacker has no credentials in 2600 the system, and is attempting to disrupt the service seen by the 2601 client or the server. 2603 17.1.1. Obtaining Unauthorized Allocations 2605 An attacker might wish to obtain allocations on a TURN server for any 2606 number of nefarious purposes. A TURN server provides a mechanism for 2607 sending and receiving packets while cloaking the actual IP address of 2608 the client. This makes TURN servers an attractive target for 2609 attackers who wish to use it to mask their true identity. 2611 An attacker might also wish to simply utilize the services of a TURN 2612 server without paying for them. Since TURN services require 2613 resources from the provider, it is anticipated that their usage will 2614 come with a cost. 2616 These attacks are prevented using the long-term credential mechanism, 2617 which allows the TURN server to determine the identity of the 2618 requestor and whether the requestor is allowed to obtain the 2619 allocation. 2621 17.1.2. Offline Dictionary Attacks 2623 The long-term credential mechanism used by TURN is subject to offline 2624 dictionary attacks. An attacker that is capable of eavesdropping on 2625 a message exchange between a client and server can determine the 2626 password by trying a number of candidate passwords and seeing if one 2627 of them is correct. This attack works when the passwords are low 2628 entropy, such as a word from the dictionary. This attack can be 2629 mitigated by using strong passwords with large entropy. In 2630 situations where even stronger mitigation is required, (D)TLS 2631 transport between the client and the server can be used. 2633 17.1.3. Faked Refreshes and Permissions 2635 An attacker might wish to attack an active allocation by sending it a 2636 Refresh request with an immediate expiration, in order to delete it 2637 and disrupt service to the client. This is prevented by 2638 authentication of refreshes. Similarly, an attacker wishing to send 2639 CreatePermission requests to create permissions to undesirable 2640 destinations is prevented from doing so through authentication. The 2641 motivations for such an attack are described in Section 17.2. 2643 17.1.4. Fake Data 2645 An attacker might wish to send data to the client or the peer, as if 2646 they came from the peer or client, respectively. To do that, the 2647 attacker can send the client a faked Data Indication or ChannelData 2648 message, or send the TURN server a faked Send Indication or 2649 ChannelData message. 2651 Since indications and ChannelData messages are not authenticated, 2652 this attack is not prevented by TURN. However, this attack is 2653 generally present in IP-based communications and is not substantially 2654 worsened by TURN. Consider a normal, non-TURN IP session between 2655 hosts A and B. An attacker can send packets to B as if they came 2656 from A by sending packets towards A with a spoofed IP address of B. 2657 This attack requires the attacker to know the IP addresses of A and 2658 B. With TURN, an attacker wishing to send packets towards a client 2659 using a Data indication needs to know its IP address (and port), the 2660 IP address and port of the TURN server, and the IP address and port 2661 of the peer (for inclusion in the XOR-PEER-ADDRESS attribute). To 2662 send a fake ChannelData message to a client, an attacker needs to 2663 know the IP address and port of the client, the IP address and port 2664 of the TURN server, and the channel number. This particular 2665 combination is mildly more guessable than in the non-TURN case. 2667 These attacks are more properly mitigated by application-layer 2668 authentication techniques. In the case of real-time traffic, usage 2669 of SRTP [RFC3711] prevents these attacks. 2671 In some situations, the TURN server may be situated in the network 2672 such that it is able to send to hosts to which the client cannot 2673 directly send. This can happen, for example, if the server is 2674 located behind a firewall that allows packets from outside the 2675 firewall to be delivered to the server, but not to other hosts behind 2676 the firewall. In these situations, an attacker could send the server 2677 a Send indication with an XOR-PEER-ADDRESS attribute containing the 2678 transport address of one of the other hosts behind the firewall. If 2679 the server was to allow relaying of traffic to arbitrary peers, then 2680 this would provide a way for the attacker to attack arbitrary hosts 2681 behind the firewall. 2683 To mitigate this attack, TURN requires that the client establish a 2684 permission to a host before sending it data. Thus, an attacker can 2685 only attack hosts with which the client is already communicating, 2686 unless the attacker is able to create authenticated requests. 2687 Furthermore, the server administrator may configure the server to 2688 restrict the range of IP addresses and ports to which it will relay 2689 data. To provide even greater security, the server administrator can 2690 require that the client use (D)TLS for all communication between the 2691 client and the server. 2693 17.1.5. Impersonating a Server 2695 When a client learns a relayed address from a TURN server, it uses 2696 that relayed address in application protocols to receive traffic. 2697 Therefore, an attacker wishing to intercept or redirect that traffic 2698 might try to impersonate a TURN server and provide the client with a 2699 faked relayed address. 2701 This attack is prevented through the long-term credential mechanism, 2702 which provides message integrity for responses in addition to 2703 verifying that they came from the server. Furthermore, an attacker 2704 cannot replay old server responses as the transaction id in the STUN 2705 header prevents this. Replay attacks are further thwarted through 2706 frequent changes to the nonce value. 2708 17.1.6. Eavesdropping Traffic 2710 TURN concerns itself primarily with authentication and message 2711 integrity. Confidentiality is only a secondary concern, as TURN 2712 control messages do not include information that is particularly 2713 sensitive. The primary protocol content of the messages is the IP 2714 address of the peer. If it is important to prevent an eavesdropper 2715 on a TURN connection from learning this, TURN can be run over (D)TLS. 2717 Confidentiality for the application data relayed by TURN is best 2718 provided by the application protocol itself, since running TURN over 2719 (D)TLS does not protect application data between the server and the 2720 peer. If confidentiality of application data is important, then the 2721 application should encrypt or otherwise protect its data. For 2722 example, for real-time media, confidentiality can be provided by 2723 using SRTP. 2725 17.1.7. TURN Loop Attack 2727 An attacker might attempt to cause data packets to loop indefinitely 2728 between two TURN servers. The attack goes as follows. First, the 2729 attacker sends an Allocate request to server A, using the source 2730 address of server B. Server A will send its response to server B, 2731 and for the attack to succeed, the attacker must have the ability to 2732 either view or guess the contents of this response, so that the 2733 attacker can learn the allocated relayed transport address. The 2734 attacker then sends an Allocate request to server B, using the source 2735 address of server A. Again, the attacker must be able to view or 2736 guess the contents of the response, so it can send learn the 2737 allocated relayed transport address. Using the same spoofed source 2738 address technique, the attacker then binds a channel number on server 2739 A to the relayed transport address on server B, and similarly binds 2740 the same channel number on server B to the relayed transport address 2741 on server A. Finally, the attacker sends a ChannelData message to 2742 server A. 2744 The result is a data packet that loops from the relayed transport 2745 address on server A to the relayed transport address on server B, 2746 then from server B's transport address to server A's transport 2747 address, and then around the loop again. 2749 This attack is mitigated as follows. By requiring all requests to be 2750 authenticated and/or by randomizing the port number allocated for the 2751 relayed transport address, the server forces the attacker to either 2752 intercept or view responses sent to a third party (in this case, the 2753 other server) so that the attacker can authenticate the requests and 2754 learn the relayed transport address. Without one of these two 2755 measures, an attacker can guess the contents of the responses without 2756 needing to see them, which makes the attack much easier to perform. 2757 Furthermore, by requiring authenticated requests, the server forces 2758 the attacker to have credentials acceptable to the server, which 2759 turns this from an outsider attack into an insider attack and allows 2760 the attack to be traced back to the client initiating it. 2762 The attack can be further mitigated by imposing a per-username limit 2763 on the bandwidth used to relay data by allocations owned by that 2764 username, to limit the impact of this attack on other allocations. 2765 More mitigation can be achieved by decrementing the TTL when relaying 2766 data packets (if the underlying OS allows this). 2768 17.2. Firewall Considerations 2770 A key security consideration of TURN is that TURN should not weaken 2771 the protections afforded by firewalls deployed between a client and a 2772 TURN server. It is anticipated that TURN servers will often be 2773 present on the public Internet, and clients may often be inside 2774 enterprise networks with corporate firewalls. If TURN servers 2775 provide a 'backdoor' for reaching into the enterprise, TURN will be 2776 blocked by these firewalls. 2778 TURN servers therefore emulate the behavior of NAT devices that 2779 implement address-dependent filtering [RFC4787], a property common in 2780 many firewalls as well. When a NAT or firewall implements this 2781 behavior, packets from an outside IP address are only allowed to be 2782 sent to an internal IP address and port if the internal IP address 2783 and port had recently sent a packet to that outside IP address. TURN 2784 servers introduce the concept of permissions, which provide exactly 2785 this same behavior on the TURN server. An attacker cannot send a 2786 packet to a TURN server and expect it to be relayed towards the 2787 client, unless the client has tried to contact the attacker first. 2789 It is important to note that some firewalls have policies that are 2790 even more restrictive than address-dependent filtering. Firewalls 2791 can also be configured with address- and port-dependent filtering, or 2792 can be configured to disallow inbound traffic entirely. In these 2793 cases, if a client is allowed to connect the TURN server, 2794 communications to the client will be less restrictive than what the 2795 firewall would normally allow. 2797 17.2.1. Faked Permissions 2799 In firewalls and NAT devices, permissions are granted implicitly 2800 through the traversal of a packet from the inside of the network 2801 towards the outside peer. Thus, a permission cannot, by definition, 2802 be created by any entity except one inside the firewall or NAT. With 2803 TURN, this restriction no longer holds. Since the TURN server sits 2804 outside the firewall, at attacker outside the firewall can now send a 2805 message to the TURN server and try to create a permission for itself. 2807 This attack is prevented because all messages that create permissions 2808 (i.e., ChannelBind and CreatePermission) are authenticated. 2810 17.2.2. Blacklisted IP Addresses 2812 Many firewalls can be configured with blacklists that prevent a 2813 client behind the firewall from sending packets to, or receiving 2814 packets from, ranges of blacklisted IP addresses. This is 2815 accomplished by inspecting the source and destination addresses of 2816 packets entering and exiting the firewall, respectively. 2818 This feature is also present in TURN, since TURN servers are allowed 2819 to arbitrarily restrict the range of addresses of peers that they 2820 will relay to. 2822 17.2.3. Running Servers on Well-Known Ports 2824 A malicious client behind a firewall might try to connect to a TURN 2825 server and obtain an allocation which it then uses to run a server. 2826 For example, a client might try to run a DNS server or FTP server. 2828 This is not possible in TURN. A TURN server will never accept 2829 traffic from a peer for which the client has not installed a 2830 permission. Thus, peers cannot just connect to the allocated port in 2831 order to obtain the service. 2833 17.3. Insider Attacks 2835 In insider attacks, a client has legitimate credentials but defies 2836 the trust relationship that goes with those credentials. These 2837 attacks cannot be prevented by cryptographic means but need to be 2838 considered in the design of the protocol. 2840 17.3.1. DoS against TURN Server 2842 A client wishing to disrupt service to other clients might obtain an 2843 allocation and then flood it with traffic, in an attempt to swamp the 2844 server and prevent it from servicing other legitimate clients. This 2845 is mitigated by the recommendation that the server limit the amount 2846 of bandwidth it will relay for a given username. This won't prevent 2847 a client from sending a large amount of traffic, but it allows the 2848 server to immediately discard traffic in excess. 2850 Since each allocation uses a port number on the IP address of the 2851 TURN server, the number of allocations on a server is finite. An 2852 attacker might attempt to consume all of them by requesting a large 2853 number of allocations. This is prevented by the recommendation that 2854 the server impose a limit of the number of allocations active at a 2855 time for a given username. 2857 17.3.2. Anonymous Relaying of Malicious Traffic 2859 TURN servers provide a degree of anonymization. A client can send 2860 data to peers without revealing its own IP address. TURN servers may 2861 therefore become attractive vehicles for attackers to launch attacks 2862 against targets without fear of detection. Indeed, it is possible 2863 for a client to chain together multiple TURN servers, such that any 2864 number of relays can be used before a target receives a packet. 2866 Administrators who are worried about this attack can maintain logs 2867 that capture the actual source IP and port of the client, and perhaps 2868 even every permission that client installs. This will allow for 2869 forensic tracing to determine the original source, should it be 2870 discovered that an attack is being relayed through a TURN server. 2872 17.3.3. Manipulating Other Allocations 2874 An attacker might attempt to disrupt service to other users of the 2875 TURN server by sending Refresh requests or CreatePermission requests 2876 that (through source address spoofing) appear to be coming from 2877 another user of the TURN server. TURN prevents this by requiring 2878 that the credentials used in CreatePermission, Refresh, and 2879 ChannelBind messages match those used to create the initial 2880 allocation. Thus, the fake requests from the attacker will be 2881 rejected. 2883 17.4. Other Considerations 2885 Any relay addresses learned through an Allocate request will not 2886 operate properly with IPsec Authentication Header (AH) [RFC4302] in 2887 transport or tunnel mode. However, tunnel-mode IPsec Encapsulating 2888 Security Payload (ESP) [RFC4303] should still operate. 2890 18. IANA Considerations 2892 Since TURN is an extension to STUN [RFC5389], the methods, 2893 attributes, and error codes defined in this specification are new 2894 methods, attributes, and error codes for STUN. IANA has added these 2895 new protocol elements to the IANA registry of STUN protocol elements. 2897 The codepoints for the new STUN methods defined in this specification 2898 are listed in Section 13. 2900 The codepoints for the new STUN attributes defined in this 2901 specification are listed in Section 14. 2903 The codepoints for the new STUN error codes defined in this 2904 specification are listed in Section 15. 2906 IANA has allocated the SRV service name of "turn" for TURN over UDP 2907 or TCP, and the service name of "turns" for TURN over (D)TLS. 2909 IANA has created a registry for TURN channel numbers, initially 2910 populated as follows: 2912 o 0x0000 through 0x3FFF: Reserved and not available for use, since 2913 they conflict with the STUN header. 2915 o 0x4000 through 0x7FFF: A TURN implementation is free to use 2916 channel numbers in this range. 2918 o 0x8000 through 0xFFFF: Unassigned. 2920 Any change to this registry must be made through an IETF Standards 2921 Action. 2923 19. IAB Considerations 2925 The IAB has studied the problem of "Unilateral Self Address Fixing" 2926 (UNSAF), which is the general process by which a client attempts to 2927 determine its address in another realm on the other side of a NAT 2928 through a collaborative protocol-reflection mechanism [RFC3424]. The 2929 TURN extension is an example of a protocol that performs this type of 2930 function. The IAB has mandated that any protocols developed for this 2931 purpose document a specific set of considerations. These 2932 considerations and the responses for TURN are documented in this 2933 section. 2935 Consideration 1: Precise definition of a specific, limited-scope 2936 problem that is to be solved with the UNSAF proposal. A short-term 2937 fix should not be generalized to solve other problems. Such 2938 generalizations lead to the prolonged dependence on and usage of the 2939 supposed short-term fix -- meaning that it is no longer accurate to 2940 call it "short-term". 2942 Response: TURN is a protocol for communication between a relay (= 2943 TURN server) and its client. The protocol allows a client that is 2944 behind a NAT to obtain and use a public IP address on the relay. As 2945 a convenience to the client, TURN also allows the client to determine 2946 its server-reflexive transport address. 2948 Consideration 2: Description of an exit strategy/transition plan. 2949 The better short-term fixes are the ones that will naturally see less 2950 and less use as the appropriate technology is deployed. 2952 Response: TURN will no longer be needed once there are no longer any 2953 NATs. Unfortunately, as of the date of publication of this document, 2954 it no longer seems very likely that NATs will go away any time soon. 2955 However, the need for TURN will also decrease as the number of NATs 2956 with the mapping property of Endpoint-Independent Mapping [RFC4787] 2957 increases. 2959 Consideration 3: Discussion of specific issues that may render 2960 systems more "brittle". For example, approaches that involve using 2961 data at multiple network layers create more dependencies, increase 2962 debugging challenges, and make it harder to transition. 2964 Response: TURN is "brittle" in that it requires the NAT bindings 2965 between the client and the server to be maintained unchanged for the 2966 lifetime of the allocation. This is typically done using keep- 2967 alives. If this is not done, then the client will lose its 2968 allocation and can no longer exchange data with its peers. 2970 Consideration 4: Identify requirements for longer-term, sound 2971 technical solutions; contribute to the process of finding the right 2972 longer-term solution. 2974 Response: The need for TURN will be reduced once NATs implement the 2975 recommendations for NAT UDP behavior documented in [RFC4787]. 2976 Applications are also strongly urged to use ICE [RFC5245] to 2977 communicate with peers; though ICE uses TURN, it does so only as a 2978 last resort, and uses it in a controlled manner. 2980 Consideration 5: Discussion of the impact of the noted practical 2981 issues with existing deployed NATs and experience reports. 2983 Response: Some NATs deployed today exhibit a mapping behavior other 2984 than Endpoint-Independent mapping. These NATs are difficult to work 2985 with, as they make it difficult or impossible for protocols like ICE 2986 to use server-reflexive transport addresses on those NATs. A client 2987 behind such a NAT is often forced to use a relay protocol like TURN 2988 because "UDP hole punching" techniques [RFC5128] do not work. 2990 20. Acknowledgements 2992 Most of the text in this note comes from the original TURN 2993 specification, [RFC5766]. The authors would like to thank Rohan Mahy 2994 co-author of orginal TURN specification and everyone who had 2995 contributed to that document. 2997 21. References 2998 21.1. Normative References 3000 [RFC5389] Rosenberg, J., Mahy, R., Matthews, P., and D. Wing, 3001 "Session Traversal Utilities for NAT (STUN)", RFC 5389, 3002 October 2008. 3004 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 3005 Requirement Levels", BCP 14, RFC 2119, March 1997. 3007 [RFC2474] Nichols, K., Blake, S., Baker, F., and D. Black, 3008 "Definition of the Differentiated Services Field (DS 3009 Field) in the IPv4 and IPv6 Headers", RFC 2474, December 3010 1998. 3012 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 3013 of Explicit Congestion Notification (ECN) to IP", RFC 3014 3168, September 2001. 3016 [RFC1122] Braden, R., "Requirements for Internet Hosts - 3017 Communication Layers", STD 3, RFC 1122, October 1989. 3019 21.2. Informative References 3021 [RFC1191] Mogul, J. and S. Deering, "Path MTU discovery", RFC 1191, 3022 November 1990. 3024 [RFC0791] Postel, J., "Internet Protocol", STD 5, RFC 791, September 3025 1981. 3027 [RFC1918] Rekhter, Y., Moskowitz, R., Karrenberg, D., Groot, G., and 3028 E. Lear, "Address Allocation for Private Internets", BCP 3029 5, RFC 1918, February 1996. 3031 [RFC3424] Daigle, L. and IAB, "IAB Considerations for UNilateral 3032 Self-Address Fixing (UNSAF) Across Network Address 3033 Translation", RFC 3424, November 2002. 3035 [RFC4787] Audet, F. and C. Jennings, "Network Address Translation 3036 (NAT) Behavioral Requirements for Unicast UDP", BCP 127, 3037 RFC 4787, January 2007. 3039 [RFC5245] Rosenberg, J., "Interactive Connectivity Establishment 3040 (ICE): A Protocol for Network Address Translator (NAT) 3041 Traversal for Offer/Answer Protocols", RFC 5245, April 3042 2010. 3044 [RFC6062] Perreault, S. and J. Rosenberg, "Traversal Using Relays 3045 around NAT (TURN) Extensions for TCP Allocations", RFC 3046 6062, November 2010. 3048 [RFC6156] Camarillo, G., Novo, O., and S. Perreault, "Traversal 3049 Using Relays around NAT (TURN) Extension for IPv6", RFC 3050 6156, April 2011. 3052 [RFC6056] Larsen, M. and F. Gont, "Recommendations for Transport- 3053 Protocol Port Randomization", BCP 156, RFC 6056, January 3054 2011. 3056 [RFC5128] Srisuresh, P., Ford, B., and D. Kegel, "State of Peer-to- 3057 Peer (P2P) Communication across Network Address 3058 Translators (NATs)", RFC 5128, March 2008. 3060 [RFC1928] Leech, M., Ganis, M., Lee, Y., Kuris, R., Koblas, D., and 3061 L. Jones, "SOCKS Protocol Version 5", RFC 1928, March 3062 1996. 3064 [RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V. 3065 Jacobson, "RTP: A Transport Protocol for Real-Time 3066 Applications", STD 64, RFC 3550, July 2003. 3068 [RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and K. 3069 Norrman, "The Secure Real-time Transport Protocol (SRTP)", 3070 RFC 3711, March 2004. 3072 [RFC4302] Kent, S., "IP Authentication Header", RFC 4302, December 3073 2005. 3075 [RFC4303] Kent, S., "IP Encapsulating Security Payload (ESP)", RFC 3076 4303, December 2005. 3078 [RFC4821] Mathis, M. and J. Heffner, "Packetization Layer Path MTU 3079 Discovery", RFC 4821, March 2007. 3081 [RFC3261] Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, 3082 A., Peterson, J., Sparks, R., Handley, M., and E. 3083 Schooler, "SIP: Session Initiation Protocol", RFC 3261, 3084 June 2002. 3086 [I-D.rosenberg-mmusic-ice-nonsip] 3087 Rosenberg, J., "Guidelines for Usage of Interactive 3088 Connectivity Establishment (ICE) by non Session Initiation 3089 Protocol (SIP) Protocols", draft-rosenberg-mmusic-ice- 3090 nonsip-01 (work in progress), July 2008. 3092 [I-D.ietf-tram-turn-server-discovery] 3093 Patil, P., Reddy, T., and D. Wing, "TURN Server Auto 3094 Discovery", draft-ietf-tram-turn-server-discovery-00 (work 3095 in progress), July 2014. 3097 [RFC4086] Eastlake, D., Schiller, J., and S. Crocker, "Randomness 3098 Requirements for Security", BCP 106, RFC 4086, June 2005. 3100 [RFC5766] Mahy, R., Matthews, P., and J. Rosenberg, "Traversal Using 3101 Relays around NAT (TURN): Relay Extensions to Session 3102 Traversal Utilities for NAT (STUN)", RFC 5766, April 2010. 3104 [Port-Numbers] 3105 "IANA Port Numbers Registry", 2005, 3106 . 3108 [Frag-Harmful] 3109 "Fragmentation Considered Harmful", . 3112 [Protocol-Numbers] 3113 "IANA Protocol Numbers Registry", 2005, 3114 . 3116 Authors' Addresses 3118 Tirumaleswar Reddy (editor) 3119 Cisco Systems, Inc. 3120 Cessna Business Park, Varthur Hobl 3121 Sarjapur Marathalli Outer Ring Road 3122 Bangalore, Karnataka 560103 3123 India 3125 Email: tireddy@cisco.com 3127 Alan Johnston (editor) 3128 Avaya 3129 St. Louis, MO 3130 USA 3132 Email: alan.b.johnston@gmail.com 3133 Philip Matthews 3134 Alcatel-Lucent 3135 600 March Road 3136 Ottawa, Ontario 3137 Canada 3139 Email: philip_matthews@magma.ca 3141 Jonathan Rosenberg 3142 jdrosen.net 3143 Edison, NJ 3144 USA 3146 Email: jdrosen@jdrosen.net 3147 URI: http://www.jdrosen.net