3 Tor Protocol Specification
8 Note: This document aims to specify Tor as implemented in 0.2.1.x. Future
9 versions of Tor may implement improved protocols, and compatibility is not
10 guaranteed. Compatibility notes are given for versions 0.1.1.15-rc and
11 later; earlier versions are not compatible with the Tor network as of this
14 This specification is not a design document; most design criteria
15 are not examined. For more information on why Tor acts as it does,
20 0.1. Notation and encoding
24 K -- a key for a symmetric cypher.
26 a|b -- concatenation of 'a' and 'b'.
28 [A0 B1 C2] -- a three-byte sequence, containing the bytes with
29 hexadecimal values A0, B1, and C2, in that order.
31 All numeric values are encoded in network (big-endian) order.
33 H(m) -- a cryptographic hash of m.
35 0.2. Security parameters
37 Tor uses a stream cipher, a public-key cipher, the Diffie-Hellman
38 protocol, and a hash function.
40 KEY_LEN -- the length of the stream cipher's key, in bytes.
42 PK_ENC_LEN -- the length of a public-key encrypted message, in bytes.
43 PK_PAD_LEN -- the number of bytes added in padding for public-key
44 encryption, in bytes. (The largest number of bytes that can be encrypted
45 in a single public-key operation is therefore PK_ENC_LEN-PK_PAD_LEN.)
47 DH_LEN -- the number of bytes used to represent a member of the
49 DH_SEC_LEN -- the number of bytes used in a Diffie-Hellman private key (x).
51 HASH_LEN -- the length of the hash function's output, in bytes.
53 PAYLOAD_LEN -- The longest allowable cell payload, in bytes. (509)
55 CELL_LEN -- The length of a Tor cell, in bytes.
59 For a stream cipher, we use 128-bit AES in counter mode, with an IV of all
62 For a public-key cipher, we use RSA with 1024-bit keys and a fixed
63 exponent of 65537. We use OAEP-MGF1 padding, with SHA-1 as its digest
64 function. We leave optional the "Label" parameter unset. (For OAEP
65 padding, see ftp://ftp.rsasecurity.com/pub/pkcs/pkcs-1/pkcs-1v2-1.pdf)
67 [Nick, what does "we leave optional the Label parameter unset" mean? -RD]
69 For Diffie-Hellman, we use a generator (g) of 2. For the modulus (p), we
70 use the 1024-bit safe prime from rfc2409 section 6.2 whose hex
73 "FFFFFFFFFFFFFFFFC90FDAA22168C234C4C6628B80DC1CD129024E08"
74 "8A67CC74020BBEA63B139B22514A08798E3404DDEF9519B3CD3A431B"
75 "302B0A6DF25F14374FE1356D6D51C245E485B576625E7EC6F44C42E9"
76 "A637ED6B0BFF5CB6F406B7EDEE386BFB5A899FA5AE9F24117C4B1FE6"
77 "49286651ECE65381FFFFFFFFFFFFFFFF"
79 As an optimization, implementations SHOULD choose DH private keys (x) of
80 320 bits. Implementations that do this MUST never use any DH key more
82 [May other implementations reuse their DH keys?? -RD]
83 [Probably not. Conceivably, you could get away with changing DH keys once
84 per second, but there are too many oddball attacks for me to be
85 comfortable that this is safe. -NM]
87 For a hash function, we use SHA-1.
90 DH_LEN=128; DH_SEC_LEN=40.
91 PK_ENC_LEN=128; PK_PAD_LEN=42.
94 When we refer to "the hash of a public key", we mean the SHA-1 hash of the
95 DER encoding of an ASN.1 RSA public key (as specified in PKCS.1).
97 All "random" values should be generated with a cryptographically strong
98 random number generator, unless otherwise noted.
100 The "hybrid encryption" of a byte sequence M with a public key PK is
102 1. If M is less than PK_ENC_LEN-PK_PAD_LEN, pad and encrypt M with PK.
103 2. Otherwise, generate a KEY_LEN byte random key K.
104 Let M1 = the first PK_ENC_LEN-PK_PAD_LEN-KEY_LEN bytes of M,
105 and let M2 = the rest of M.
106 Pad and encrypt K|M1 with PK. Encrypt M2 with our stream cipher,
107 using the key K. Concatenate these encrypted values.
108 [XXX Note that this "hybrid encryption" approach does not prevent
109 an attacker from adding or removing bytes to the end of M. It also
110 allows attackers to modify the bytes not covered by the OAEP --
111 see Goldberg's PET2006 paper for details. We will add a MAC to this
114 0.4. Other parameter values
120 Tor is a distributed overlay network designed to anonymize
121 low-latency TCP-based applications such as web browsing, secure shell,
122 and instant messaging. Clients choose a path through the network and
123 build a ``circuit'', in which each node (or ``onion router'' or ``OR'')
124 in the path knows its predecessor and successor, but no other nodes in
125 the circuit. Traffic flowing down the circuit is sent in fixed-size
126 ``cells'', which are unwrapped by a symmetric key at each node (like
127 the layers of an onion) and relayed downstream.
131 Every Tor server has multiple public/private keypairs:
133 - A long-term signing-only "Identity key" used to sign documents and
134 certificates, and used to establish server identity.
135 - A medium-term "Onion key" used to decrypt onion skins when accepting
136 circuit extend attempts. (See 5.1.) Old keys MUST be accepted for at
137 least one week after they are no longer advertised. Because of this,
138 servers MUST retain old keys for a while after they're rotated.
139 - A short-term "Connection key" used to negotiate TLS connections.
140 Tor implementations MAY rotate this key as often as they like, and
141 SHOULD rotate this key at least once a day.
143 Tor servers are also identified by "nicknames"; these are specified in
148 Connections between two Tor servers, or between a client and a server,
149 use TLS/SSLv3 for link authentication and encryption. All
150 implementations MUST support the SSLv3 ciphersuite
151 "SSL_DHE_RSA_WITH_3DES_EDE_CBC_SHA", and SHOULD support the TLS
152 ciphersuite "TLS_DHE_RSA_WITH_AES_128_CBC_SHA" if it is available.
154 There are three acceptable ways to perform a TLS handshake when
155 connecting to a Tor server: "certificates up-front", "renegotiation", and
156 "backwards-compatible renegotiation". ("Backwards-compatible
157 renegotiation" is, as the name implies, compatible with both other
160 Before Tor 0.2.0.21, only "certificates up-front" was supported. In Tor
161 0.2.0.21 or later, "backwards-compatible renegotiation" is used.
163 In "certificates up-front", the connection initiator always sends a
164 two-certificate chain, consisting of an X.509 certificate using a
165 short-term connection public key and a second, self- signed X.509
166 certificate containing its identity key. The other party sends a similar
167 certificate chain. The initiator's ClientHello MUST NOT include any
168 ciphersuites other than:
169 TLS_DHE_RSA_WITH_AES_256_CBC_SHA
170 TLS_DHE_RSA_WITH_AES_128_CBC_SHA
171 SSL_DHE_RSA_WITH_3DES_EDE_CBC_SHA
172 SSL_DHE_DSS_WITH_3DES_EDE_CBC_SHA
174 In "renegotiation", the connection initiator sends no certificates, and
175 the responder sends a single connection certificate. Once the TLS
176 handshake is complete, the initiator renegotiates the handshake, with each
177 parties sending a two-certificate chain as in "certificates up-front".
178 The initiator's ClientHello MUST include at least once ciphersuite not in
179 the list above. The responder SHOULD NOT select any ciphersuite besides
180 those in the list above.
181 [The above "should not" is because some of the ciphers that
182 clients list may be fake.]
184 In "backwards-compatible renegotiation", the connection initiator's
185 ClientHello MUST include at least one ciphersuite other than those listed
186 above. The connection responder examines the initiator's ciphersuite list
187 to see whether it includes any ciphers other than those included in the
188 list above. If extra ciphers are included, the responder proceeds as in
189 "renegotiation": it sends a single certificate and does not request
190 client certificates. Otherwise (in the case that no extra ciphersuites
191 are included in the ClientHello) the responder proceeds as in
192 "certificates up-front": it requests client certificates, and sends a
193 two-certificate chain. In either case, once the responder has sent its
194 certificate or certificates, the initiator counts them. If two
195 certificates have been sent, it proceeds as in "certificates up-front";
196 otherwise, it proceeds as in "renegotiation".
198 All new implementations of the Tor server protocol MUST support
199 "backwards-compatible renegotiation"; clients SHOULD do this too. If
200 this is not possible, new client implementations MUST support both
201 "renegotiation" and "certificates up-front" and use the router's
202 published link protocols list (see dir-spec.txt on the "protocols" entry)
203 to decide which to use.
205 In all of the above handshake variants, certificates sent in the clear
206 SHOULD NOT include any strings to identify the host as a Tor server. In
207 the "renegotation" and "backwards-compatible renegotiation", the
208 initiator SHOULD chose a list of ciphersuites and TLS extensions chosen
209 to mimic one used by a popular web browser.
211 Responders MUST NOT select any TLS ciphersuite that lacks ephemeral keys,
212 or whose symmetric keys are less then KEY_LEN bits, or whose digests are
213 less than HASH_LEN bits. Responders SHOULD NOT select any SSLv3
214 ciphersuite other than those listed above.
216 Even though the connection protocol is identical, we will think of the
217 initiator as either an onion router (OR) if it is willing to relay
218 traffic for other Tor users, or an onion proxy (OP) if it only handles
219 local requests. Onion proxies SHOULD NOT provide long-term-trackable
220 identifiers in their handshakes.
222 In all handshake variants, once all certificates are exchanged, all
223 parties receiving certificates must confirm that the identity key is as
224 expected. (When initiating a connection, the expected identity key is
225 the one given in the directory; when creating a connection because of an
226 EXTEND cell, the expected identity key is the one given in the cell.) If
227 the key is not as expected, the party must close the connection.
229 When connecting to an OR, all parties SHOULD reject the connection if that
230 OR has a malformed or missing certificate. When accepting an incoming
231 connection, an OR SHOULD NOT reject incoming connections from parties with
232 malformed or missing certificates. (However, an OR should not believe
233 that an incoming connection is from another OR unless the certificates
234 are present and well-formed.)
236 [Before version 0.1.2.8-rc, ORs rejected incoming connections from ORs and
237 OPs alike if their certificates were missing or malformed.]
239 Once a TLS connection is established, the two sides send cells
240 (specified below) to one another. Cells are sent serially. All
241 cells are CELL_LEN bytes long. Cells may be sent embedded in TLS
242 records of any size or divided across TLS records, but the framing
243 of TLS records MUST NOT leak information about the type or contents
246 TLS connections are not permanent. Either side MAY close a connection
247 if there are no circuits running over it and an amount of time
248 (KeepalivePeriod, defaults to 5 minutes) has passed since the last time
249 any traffic was transmitted over the TLS connection. Clients SHOULD
250 also hold a TLS connection with no circuits open, if it is likely that a
251 circuit will be built soon using that connection.
253 (As an exception, directory servers may try to stay connected to all of
254 the ORs -- though this will be phased out for the Tor 0.1.2.x release.)
256 3. Cell Packet format
258 The basic unit of communication for onion routers and onion
259 proxies is a fixed-width "cell".
261 On a version 1 connection, each cell contains the following
266 Payload (padded with 0 bytes) [PAYLOAD_LEN bytes]
268 On a version 2 connection, all cells are as in version 1 connections,
269 except for the initial VERSIONS cell, whose format is:
271 Circuit [2 octets; set to 0]
272 Command [1 octet; set to 7 for VERSIONS]
273 Length [2 octets; big-endian integer]
274 Payload [Length bytes]
276 The CircID field determines which circuit, if any, the cell is
279 The 'Command' field holds one of the following values:
280 0 -- PADDING (Padding) (See Sec 7.2)
281 1 -- CREATE (Create a circuit) (See Sec 5.1)
282 2 -- CREATED (Acknowledge create) (See Sec 5.1)
283 3 -- RELAY (End-to-end data) (See Sec 5.5 and 6)
284 4 -- DESTROY (Stop using a circuit) (See Sec 5.4)
285 5 -- CREATE_FAST (Create a circuit, no PK) (See Sec 5.1)
286 6 -- CREATED_FAST (Circuit created, no PK) (See Sec 5.1)
287 7 -- VERSIONS (Negotiate proto version) (See Sec 4)
288 8 -- NETINFO (Time and address info) (See Sec 4)
290 The interpretation of 'Payload' depends on the type of the cell.
291 PADDING: Payload is unused.
292 CREATE: Payload contains the handshake challenge.
293 CREATED: Payload contains the handshake response.
294 RELAY: Payload contains the relay header and relay body.
295 DESTROY: Payload contains a reason for closing the circuit.
297 Upon receiving any other value for the command field, an OR must
298 drop the cell. Since more cell types may be added in the future, ORs
299 should generally not warn when encountering unrecognized commands.
301 The payload is padded with 0 bytes.
303 PADDING cells are currently used to implement connection keepalive.
304 If there is no other traffic, ORs and OPs send one another a PADDING
305 cell every few minutes.
307 CREATE, CREATED, and DESTROY cells are used to manage circuits;
310 RELAY cells are used to send commands and data along a circuit; see
313 VERSIONS and NETINFO cells are used to set up connections. See section 4
316 4. Negotiating and initializing connections
318 4.1. Negotiating versions with VERSIONS cells
320 There are multiple instances of the Tor link connection protocol. Any
321 connection negotiated using the "certificates up front" handshake (see
322 section 2 above) is "version 1". In any connection where both parties
323 have behaved as in the "renegotiation" handshake, the link protocol
324 version is 2 or higher.
326 To determine the version, in any connection where the "renegotiation"
327 handshake was used (that is, where the server sent only one certificate
328 at first and where the client did not send any certificates until
329 renegotiation), both parties MUST send a VERSIONS cell immediately after
330 the renegotiation is finished, before any other cells are sent. Parties
331 MUST NOT send any other cells on a connection until they have received a
334 The payload in a VERSIONS cell is a series of big-endian two-byte
335 integers. Both parties MUST select as the link protocol version the
336 highest number contained both in the VERSIONS cell they sent and in the
337 versions cell they received. If they have no such version in common,
338 they cannot communicate and MUST close the connection.
340 Since the version 1 link protocol does not use the "renegotiation"
341 handshake, implementations MUST NOT list version 1 in their VERSIONS
346 If version 2 or higher is negotiated, each party sends the other a
347 NETINFO cell. The cell's payload is:
350 Other OR's address [variable]
351 Number of addresses [1 byte]
352 This OR's addresses [variable]
354 The address format is a type/length/value sequence as given in section
355 6.4 below. The timestamp is a big-endian unsigned integer number of
356 seconds since the unix epoch.
358 Implementations MAY use the timestamp value to help decide if their
359 clocks are skewed. Initiators MAY use "other OR's address" to help
360 learn which address their connections are originating from, if they do
361 not know it. Initiators SHOULD use "this OR's address" to make sure
362 that they have connected to another OR at its canonical address.
364 [As of 0.2.0.23-rc, implementations use none of the above values.]
367 5. Circuit management
369 5.1. CREATE and CREATED cells
371 Users set up circuits incrementally, one hop at a time. To create a
372 new circuit, OPs send a CREATE cell to the first node, with the
373 first half of the DH handshake; that node responds with a CREATED
374 cell with the second half of the DH handshake plus the first 20 bytes
375 of derivative key data (see section 5.2). To extend a circuit past
376 the first hop, the OP sends an EXTEND relay cell (see section 5)
377 which instructs the last node in the circuit to send a CREATE cell
378 to extend the circuit.
380 The payload for a CREATE cell is an 'onion skin', which consists
381 of the first step of the DH handshake data (also known as g^x).
382 This value is hybrid-encrypted (see 0.3) to Bob's onion key, giving
385 Padding padding [PK_PAD_LEN bytes]
386 Symmetric key [KEY_LEN bytes]
387 First part of g^x [PK_ENC_LEN-PK_PAD_LEN-KEY_LEN bytes]
388 Symmetrically encrypted:
389 Second part of g^x [DH_LEN-(PK_ENC_LEN-PK_PAD_LEN-KEY_LEN)
392 The relay payload for an EXTEND relay cell consists of:
395 Onion skin [DH_LEN+KEY_LEN+PK_PAD_LEN bytes]
396 Identity fingerprint [HASH_LEN bytes]
398 The port and address field denote the IPV4 address and port of the next
399 onion router in the circuit; the public key hash is the hash of the PKCS#1
400 ASN1 encoding of the next onion router's identity (signing) key. (See 0.3
401 above.) Including this hash allows the extending OR verify that it is
402 indeed connected to the correct target OR, and prevents certain
403 man-in-the-middle attacks.
405 The payload for a CREATED cell, or the relay payload for an
406 EXTENDED cell, contains:
407 DH data (g^y) [DH_LEN bytes]
408 Derivative key data (KH) [HASH_LEN bytes] <see 5.2 below>
410 The CircID for a CREATE cell is an arbitrarily chosen 2-byte integer,
411 selected by the node (OP or OR) that sends the CREATE cell. To prevent
412 CircID collisions, when one node sends a CREATE cell to another, it chooses
413 from only one half of the possible values based on the ORs' public
414 identity keys: if the sending node has a lower key, it chooses a CircID with
415 an MSB of 0; otherwise, it chooses a CircID with an MSB of 1.
417 (An OP with no public key MAY choose any CircID it wishes, since an OP
418 never needs to process a CREATE cell.)
420 Public keys are compared numerically by modulus.
422 As usual with DH, x and y MUST be generated randomly.
424 5.1.1. CREATE_FAST/CREATED_FAST cells
426 When initializing the first hop of a circuit, the OP has already
427 established the OR's identity and negotiated a secret key using TLS.
428 Because of this, it is not always necessary for the OP to perform the
429 public key operations to create a circuit. In this case, the
430 OP MAY send a CREATE_FAST cell instead of a CREATE cell for the first
431 hop only. The OR responds with a CREATED_FAST cell, and the circuit is
434 A CREATE_FAST cell contains:
436 Key material (X) [HASH_LEN bytes]
438 A CREATED_FAST cell contains:
440 Key material (Y) [HASH_LEN bytes]
441 Derivative key data [HASH_LEN bytes] (See 5.2 below)
443 The values of X and Y must be generated randomly.
445 If an OR sees a circuit created with CREATE_FAST, the OR is sure to be the
446 first hop of a circuit. ORs SHOULD reject attempts to create streams with
447 RELAY_BEGIN exiting the circuit at the first hop: letting Tor be used as a
448 single hop proxy makes exit nodes a more attractive target for compromise.
450 5.2. Setting circuit keys
452 Once the handshake between the OP and an OR is completed, both can
453 now calculate g^xy with ordinary DH. Before computing g^xy, both client
454 and server MUST verify that the received g^x or g^y value is not degenerate;
455 that is, it must be strictly greater than 1 and strictly less than p-1
456 where p is the DH modulus. Implementations MUST NOT complete a handshake
457 with degenerate keys. Implementations MUST NOT discard other "weak"
460 (Discarding degenerate keys is critical for security; if bad keys
461 are not discarded, an attacker can substitute the server's CREATED
462 cell's g^y with 0 or 1, thus creating a known g^xy and impersonating
463 the server. Discarding other keys may allow attacks to learn bits of
466 If CREATE or EXTEND is used to extend a circuit, the client and server
467 base their key material on K0=g^xy, represented as a big-endian unsigned
470 If CREATE_FAST is used, the client and server base their key material on
473 From the base key material K0, they compute KEY_LEN*2+HASH_LEN*3 bytes of
474 derivative key data as
475 K = H(K0 | [00]) | H(K0 | [01]) | H(K0 | [02]) | ...
477 The first HASH_LEN bytes of K form KH; the next HASH_LEN form the forward
478 digest Df; the next HASH_LEN 41-60 form the backward digest Db; the next
479 KEY_LEN 61-76 form Kf, and the final KEY_LEN form Kb. Excess bytes from K
482 KH is used in the handshake response to demonstrate knowledge of the
483 computed shared key. Df is used to seed the integrity-checking hash
484 for the stream of data going from the OP to the OR, and Db seeds the
485 integrity-checking hash for the data stream from the OR to the OP. Kf
486 is used to encrypt the stream of data going from the OP to the OR, and
487 Kb is used to encrypt the stream of data going from the OR to the OP.
489 5.3. Creating circuits
491 When creating a circuit through the network, the circuit creator
492 (OP) performs the following steps:
494 1. Choose an onion router as an exit node (R_N), such that the onion
495 router's exit policy includes at least one pending stream that
496 needs a circuit (if there are any).
498 2. Choose a chain of (N-1) onion routers
499 (R_1...R_N-1) to constitute the path, such that no router
500 appears in the path twice.
502 3. If not already connected to the first router in the chain,
503 open a new connection to that router.
505 4. Choose a circID not already in use on the connection with the
506 first router in the chain; send a CREATE cell along the
507 connection, to be received by the first onion router.
509 5. Wait until a CREATED cell is received; finish the handshake
510 and extract the forward key Kf_1 and the backward key Kb_1.
512 6. For each subsequent onion router R (R_2 through R_N), extend
515 To extend the circuit by a single onion router R_M, the OP performs
518 1. Create an onion skin, encrypted to R_M's public onion key.
520 2. Send the onion skin in a relay EXTEND cell along
521 the circuit (see section 5).
523 3. When a relay EXTENDED cell is received, verify KH, and
524 calculate the shared keys. The circuit is now extended.
526 When an onion router receives an EXTEND relay cell, it sends a CREATE
527 cell to the next onion router, with the enclosed onion skin as its
528 payload. As special cases, if the extend cell includes a digest of
529 all zeroes, or asks to extend back to the relay that sent the extend
530 cell, the circuit will fail and be torn down. The initiating onion
531 router chooses some circID not yet used on the connection between the
532 two onion routers. (But see section 5.1. above, concerning choosing
533 circIDs based on lexicographic order of nicknames.)
535 When an onion router receives a CREATE cell, if it already has a
536 circuit on the given connection with the given circID, it drops the
537 cell. Otherwise, after receiving the CREATE cell, it completes the
538 DH handshake, and replies with a CREATED cell. Upon receiving a
539 CREATED cell, an onion router packs it payload into an EXTENDED relay
540 cell (see section 5), and sends that cell up the circuit. Upon
541 receiving the EXTENDED relay cell, the OP can retrieve g^y.
543 (As an optimization, OR implementations may delay processing onions
544 until a break in traffic allows time to do so without harming
545 network latency too greatly.)
547 5.3.1. Canonical connections
549 It is possible for an attacker to launch a man-in-the-middle attack
550 against a connection by telling OR Alice to extend to OR Bob at some
551 address X controlled by the attacker. The attacker cannot read the
552 encrypted traffic, but the attacker is now in a position to count all
553 bytes sent between Alice and Bob (assuming Alice was not already
556 To prevent this, when an OR we gets an extend request, it SHOULD use an
557 existing OR connection if the ID matches, and ANY of the following
559 - The IP matches the requested IP.
560 - The OR knows that the IP of the connection it's using is canonical
561 because it was listed in the NETINFO cell.
562 - The OR knows that the IP of the connection it's using is canonical
563 because it was listed in the server descriptor.
565 [This is not implemented in Tor 0.2.0.23-rc.]
567 5.4. Tearing down circuits
569 Circuits are torn down when an unrecoverable error occurs along
570 the circuit, or when all streams on a circuit are closed and the
571 circuit's intended lifetime is over. Circuits may be torn down
572 either completely or hop-by-hop.
574 To tear down a circuit completely, an OR or OP sends a DESTROY
575 cell to the adjacent nodes on that circuit, using the appropriate
578 Upon receiving an outgoing DESTROY cell, an OR frees resources
579 associated with the corresponding circuit. If it's not the end of
580 the circuit, it sends a DESTROY cell for that circuit to the next OR
581 in the circuit. If the node is the end of the circuit, then it tears
582 down any associated edge connections (see section 6.1).
584 After a DESTROY cell has been processed, an OR ignores all data or
585 destroy cells for the corresponding circuit.
587 To tear down part of a circuit, the OP may send a RELAY_TRUNCATE cell
588 signaling a given OR (Stream ID zero). That OR sends a DESTROY
589 cell to the next node in the circuit, and replies to the OP with a
590 RELAY_TRUNCATED cell.
592 When an unrecoverable error occurs along one connection in a
593 circuit, the nodes on either side of the connection should, if they
594 are able, act as follows: the node closer to the OP should send a
595 RELAY_TRUNCATED cell towards the OP; the node farther from the OP
596 should send a DESTROY cell down the circuit.
598 The payload of a RELAY_TRUNCATED or DESTROY cell contains a single octet,
599 describing why the circuit is being closed or truncated. When sending a
600 TRUNCATED or DESTROY cell because of another TRUNCATED or DESTROY cell,
601 the error code should be propagated. The origin of a circuit always sets
602 this error code to 0, to avoid leaking its version.
605 0 -- NONE (No reason given.)
606 1 -- PROTOCOL (Tor protocol violation.)
607 2 -- INTERNAL (Internal error.)
608 3 -- REQUESTED (A client sent a TRUNCATE command.)
609 4 -- HIBERNATING (Not currently operating; trying to save bandwidth.)
610 5 -- RESOURCELIMIT (Out of memory, sockets, or circuit IDs.)
611 6 -- CONNECTFAILED (Unable to reach server.)
612 7 -- OR_IDENTITY (Connected to server, but its OR identity was not
614 8 -- OR_CONN_CLOSED (The OR connection that was carrying this circuit
616 9 -- FINISHED (The circuit has expired for being dirty or old.)
617 10 -- TIMEOUT (Circuit construction took too long)
618 11 -- DESTROYED (The circuit was destroyed w/o client TRUNCATE)
619 12 -- NOSUCHSERVICE (Request for unknown hidden service)
621 5.5. Routing relay cells
623 When an OR receives a RELAY cell, it checks the cell's circID and
624 determines whether it has a corresponding circuit along that
625 connection. If not, the OR drops the RELAY cell.
627 Otherwise, if the OR is not at the OP edge of the circuit (that is,
628 either an 'exit node' or a non-edge node), it de/encrypts the payload
629 with the stream cipher, as follows:
630 'Forward' relay cell (same direction as CREATE):
631 Use Kf as key; decrypt.
632 'Back' relay cell (opposite direction from CREATE):
633 Use Kb as key; encrypt.
634 Note that in counter mode, decrypt and encrypt are the same operation.
636 The OR then decides whether it recognizes the relay cell, by
637 inspecting the payload as described in section 6.1 below. If the OR
638 recognizes the cell, it processes the contents of the relay cell.
639 Otherwise, it passes the decrypted relay cell along the circuit if
640 the circuit continues. If the OR at the end of the circuit
641 encounters an unrecognized relay cell, an error has occurred: the OR
642 sends a DESTROY cell to tear down the circuit.
644 When a relay cell arrives at an OP, the OP decrypts the payload
645 with the stream cipher as follows:
646 OP receives data cell:
648 Decrypt with Kb_I. If the payload is recognized (see
649 section 6..1), then stop and process the payload.
651 For more information, see section 6 below.
653 6. Application connections and stream management
657 Within a circuit, the OP and the exit node use the contents of
658 RELAY packets to tunnel end-to-end commands and TCP connections
659 ("Streams") across circuits. End-to-end commands can be initiated
660 by either edge; streams are initiated by the OP.
662 The payload of each unencrypted RELAY cell consists of:
663 Relay command [1 byte]
664 'Recognized' [2 bytes]
668 Data [CELL_LEN-14 bytes]
670 The relay commands are:
671 1 -- RELAY_BEGIN [forward]
672 2 -- RELAY_DATA [forward or backward]
673 3 -- RELAY_END [forward or backward]
674 4 -- RELAY_CONNECTED [backward]
675 5 -- RELAY_SENDME [forward or backward] [sometimes control]
676 6 -- RELAY_EXTEND [forward] [control]
677 7 -- RELAY_EXTENDED [backward] [control]
678 8 -- RELAY_TRUNCATE [forward] [control]
679 9 -- RELAY_TRUNCATED [backward] [control]
680 10 -- RELAY_DROP [forward or backward] [control]
681 11 -- RELAY_RESOLVE [forward]
682 12 -- RELAY_RESOLVED [backward]
683 13 -- RELAY_BEGIN_DIR [forward]
685 32..40 -- Used for hidden services; see rend-spec.txt.
687 Commands labelled as "forward" must only be sent by the originator
688 of the circuit. Commands labelled as "backward" must only be sent by
689 other nodes in the circuit back to the originator. Commands marked
690 as either can be sent either by the originator or other nodes.
692 The 'recognized' field in any unencrypted relay payload is always set
693 to zero; the 'digest' field is computed as the first four bytes of
694 the running digest of all the bytes that have been destined for
695 this hop of the circuit or originated from this hop of the circuit,
696 seeded from Df or Db respectively (obtained in section 5.2 above),
697 and including this RELAY cell's entire payload (taken with the digest
700 When the 'recognized' field of a RELAY cell is zero, and the digest
701 is correct, the cell is considered "recognized" for the purposes of
702 decryption (see section 5.5 above).
704 (The digest does not include any bytes from relay cells that do
705 not start or end at this hop of the circuit. That is, it does not
706 include forwarded data. Therefore if 'recognized' is zero but the
707 digest does not match, the running digest at that node should
708 not be updated, and the cell should be forwarded on.)
710 All RELAY cells pertaining to the same tunneled stream have the
711 same stream ID. StreamIDs are chosen arbitrarily by the OP. RELAY
712 cells that affect the entire circuit rather than a particular
713 stream use a StreamID of zero -- they are marked in the table above
714 as "[control]" style cells. (Sendme cells are marked as "sometimes
715 control" because they can take include a StreamID or not depending
716 on their purpose -- see Section 7.)
718 The 'Length' field of a relay cell contains the number of bytes in
719 the relay payload which contain real payload data. The remainder of
720 the payload is padded with NUL bytes.
722 If the RELAY cell is recognized but the relay command is not
723 understood, the cell must be dropped and ignored. Its contents
724 still count with respect to the digests, though.
726 6.2. Opening streams and transferring data
728 To open a new anonymized TCP connection, the OP chooses an open
729 circuit to an exit that may be able to connect to the destination
730 address, selects an arbitrary StreamID not yet used on that circuit,
731 and constructs a RELAY_BEGIN cell with a payload encoding the address
732 and port of the destination host. The payload format is:
734 ADDRESS | ':' | PORT | [00]
736 where ADDRESS can be a DNS hostname, or an IPv4 address in
737 dotted-quad format, or an IPv6 address surrounded by square brackets;
738 and where PORT is a decimal integer between 1 and 65535, inclusive.
740 [What is the [00] for? -NM]
741 [It's so the payload is easy to parse out with string funcs -RD]
743 Upon receiving this cell, the exit node resolves the address as
744 necessary, and opens a new TCP connection to the target port. If the
745 address cannot be resolved, or a connection can't be established, the
746 exit node replies with a RELAY_END cell. (See 6.4 below.)
747 Otherwise, the exit node replies with a RELAY_CONNECTED cell, whose
748 payload is in one of the following formats:
749 The IPv4 address to which the connection was made [4 octets]
750 A number of seconds (TTL) for which the address may be cached [4 octets]
752 Four zero-valued octets [4 octets]
753 An address type (6) [1 octet]
754 The IPv6 address to which the connection was made [16 octets]
755 A number of seconds (TTL) for which the address may be cached [4 octets]
756 [XXXX No version of Tor currently generates the IPv6 format.]
758 [Tor servers before 0.1.2.0 set the TTL field to a fixed value. Later
759 versions set the TTL to the last value seen from a DNS server, and expire
760 their own cached entries after a fixed interval. This prevents certain
763 The OP waits for a RELAY_CONNECTED cell before sending any data.
764 Once a connection has been established, the OP and exit node
765 package stream data in RELAY_DATA cells, and upon receiving such
766 cells, echo their contents to the corresponding TCP stream.
767 RELAY_DATA cells sent to unrecognized streams are dropped.
769 Relay RELAY_DROP cells are long-range dummies; upon receiving such
770 a cell, the OR or OP must drop it.
772 6.2.1. Opening a directory stream
774 If a Tor server is a directory server, it should respond to a
775 RELAY_BEGIN_DIR cell as if it had received a BEGIN cell requesting a
776 connection to its directory port. RELAY_BEGIN_DIR cells ignore exit
777 policy, since the stream is local to the Tor process.
779 If the Tor server is not running a directory service, it should respond
780 with a REASON_NOTDIRECTORY RELAY_END cell.
782 Clients MUST generate an all-zero payload for RELAY_BEGIN_DIR cells,
783 and servers MUST ignore the payload.
785 [RELAY_BEGIN_DIR was not supported before Tor 0.1.2.2-alpha; clients
786 SHOULD NOT send it to routers running earlier versions of Tor.]
790 When an anonymized TCP connection is closed, or an edge node
791 encounters error on any stream, it sends a 'RELAY_END' cell along the
792 circuit (if possible) and closes the TCP connection immediately. If
793 an edge node receives a 'RELAY_END' cell for any stream, it closes
794 the TCP connection completely, and sends nothing more along the
795 circuit for that stream.
797 The payload of a RELAY_END cell begins with a single 'reason' byte to
798 describe why the stream is closing, plus optional data (depending on
799 the reason.) The values are:
801 1 -- REASON_MISC (catch-all for unlisted reasons)
802 2 -- REASON_RESOLVEFAILED (couldn't look up hostname)
803 3 -- REASON_CONNECTREFUSED (remote host refused connection) [*]
804 4 -- REASON_EXITPOLICY (OR refuses to connect to host or port)
805 5 -- REASON_DESTROY (Circuit is being destroyed)
806 6 -- REASON_DONE (Anonymized TCP connection was closed)
807 7 -- REASON_TIMEOUT (Connection timed out, or OR timed out
809 8 -- (unallocated) [**]
810 9 -- REASON_HIBERNATING (OR is temporarily hibernating)
811 10 -- REASON_INTERNAL (Internal error at the OR)
812 11 -- REASON_RESOURCELIMIT (OR has no resources to fulfill request)
813 12 -- REASON_CONNRESET (Connection was unexpectedly reset)
814 13 -- REASON_TORPROTOCOL (Sent when closing connection because of
815 Tor protocol violations.)
816 14 -- REASON_NOTDIRECTORY (Client sent RELAY_BEGIN_DIR to a
817 non-directory server.)
819 (With REASON_EXITPOLICY, the 4-byte IPv4 address or 16-byte IPv6 address
820 forms the optional data, along with a 4-byte TTL; no other reason
821 currently has extra data.)
823 OPs and ORs MUST accept reasons not on the above list, since future
824 versions of Tor may provide more fine-grained reasons.
826 [*] Older versions of Tor also send this reason when connections are
828 [**] Due to a bug in versions of Tor through 0095, error reason 8 must
829 remain allocated until that version is obsolete.
831 --- [The rest of this section describes unimplemented functionality.]
833 Because TCP connections can be half-open, we follow an equivalent
834 to TCP's FIN/FIN-ACK/ACK protocol to close streams.
836 An exit connection can have a TCP stream in one of three states:
837 'OPEN', 'DONE_PACKAGING', and 'DONE_DELIVERING'. For the purposes
838 of modeling transitions, we treat 'CLOSED' as a fourth state,
839 although connections in this state are not, in fact, tracked by the
842 A stream begins in the 'OPEN' state. Upon receiving a 'FIN' from
843 the corresponding TCP connection, the edge node sends a 'RELAY_FIN'
844 cell along the circuit and changes its state to 'DONE_PACKAGING'.
845 Upon receiving a 'RELAY_FIN' cell, an edge node sends a 'FIN' to
846 the corresponding TCP connection (e.g., by calling
847 shutdown(SHUT_WR)) and changing its state to 'DONE_DELIVERING'.
849 When a stream in already in 'DONE_DELIVERING' receives a 'FIN', it
850 also sends a 'RELAY_FIN' along the circuit, and changes its state
851 to 'CLOSED'. When a stream already in 'DONE_PACKAGING' receives a
852 'RELAY_FIN' cell, it sends a 'FIN' and changes its state to
855 If an edge node encounters an error on any stream, it sends a
856 'RELAY_END' cell (if possible) and closes the stream immediately.
858 6.4. Remote hostname lookup
860 To find the address associated with a hostname, the OP sends a
861 RELAY_RESOLVE cell containing the hostname to be resolved with a nul
862 terminating byte. (For a reverse lookup, the OP sends a RELAY_RESOLVE
863 cell containing an in-addr.arpa address.) The OR replies with a
864 RELAY_RESOLVED cell containing a status byte, and any number of
865 answers. Each answer is of the form:
868 Value (variable-width)
870 "Length" is the length of the Value field.
875 0xF0 -- Error, transient
876 0xF1 -- Error, nontransient
878 If any answer has a type of 'Error', then no other answer may be given.
880 The RELAY_RESOLVE cell must use a nonzero, distinct streamID; the
881 corresponding RELAY_RESOLVED cell must use the same streamID. No stream
882 is actually created by the OR when resolving the name.
888 Each node should do appropriate bandwidth throttling to keep its
891 Communicants rely on TCP's default flow control to push back when they
896 Link padding can be created by sending PADDING cells along the
897 connection; relay cells of type "DROP" can be used for long-range
900 Currently nodes are not required to do any sort of link padding or
901 dummy traffic. Because strong attacks exist even with link padding,
902 and because link padding greatly increases the bandwidth requirements
903 for running a node, we plan to leave out link padding until this
904 tradeoff is better understood.
906 7.3. Circuit-level flow control
908 To control a circuit's bandwidth usage, each OR keeps track of two
909 'windows', consisting of how many RELAY_DATA cells it is allowed to
910 originate (package for transmission), and how many RELAY_DATA cells
911 it is willing to consume (receive for local streams). These limits
912 do not apply to cells that the OR receives from one host and relays
915 Each 'window' value is initially set to 1000 data cells
916 in each direction (cells that are not data cells do not affect
917 the window). When an OR is willing to deliver more cells, it sends a
918 RELAY_SENDME cell towards the OP, with Stream ID zero. When an OR
919 receives a RELAY_SENDME cell with stream ID zero, it increments its
922 Each of these cells increments the corresponding window by 100.
924 The OP behaves identically, except that it must track a packaging
925 window and a delivery window for every OR in the circuit.
927 An OR or OP sends cells to increment its delivery window when the
928 corresponding window value falls under some threshold (900).
930 If a packaging window reaches 0, the OR or OP stops reading from
931 TCP connections for all streams on the corresponding circuit, and
932 sends no more RELAY_DATA cells until receiving a RELAY_SENDME cell.
933 [this stuff is badly worded; copy in the tor-design section -RD]
935 7.4. Stream-level flow control
937 Edge nodes use RELAY_SENDME cells to implement end-to-end flow
938 control for individual connections across circuits. Similarly to
939 circuit-level flow control, edge nodes begin with a window of cells
940 (500) per stream, and increment the window by a fixed value (50)
941 upon receiving a RELAY_SENDME cell. Edge nodes initiate RELAY_SENDME
942 cells when both a) the window is <= 450, and b) there are less than
943 ten cell payloads remaining to be flushed at that edge.
946 A.1. Differences between spec and implementation
948 - The current specification requires all ORs to have IPv4 addresses, but
949 allows servers to exit and resolve to IPv6 addresses, and to declare IPv6
950 addresses in their exit policies. The current codebase has no IPv6