tor-spec.txt 37 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833
  1. $Id$
  2. Tor Protocol Specification
  3. Roger Dingledine
  4. Nick Mathewson
  5. Note: This document aims to specify Tor as implemented in 0.1.2.x
  6. and earlier. Future versions of Tor may implement improved protocols, and
  7. compatibility is not guaranteed.
  8. This specification is not a design document; most design criteria
  9. are not examined. For more information on why Tor acts as it does,
  10. see tor-design.pdf.
  11. 0. Preliminaries
  12. 0.1. Notation and encoding
  13. PK -- a public key.
  14. SK -- a private key.
  15. K -- a key for a symmetric cypher.
  16. a|b -- concatenation of 'a' and 'b'.
  17. [A0 B1 C2] -- a three-byte sequence, containing the bytes with
  18. hexadecimal values A0, B1, and C2, in that order.
  19. All numeric values are encoded in network (big-endian) order.
  20. H(m) -- a cryptographic hash of m.
  21. 0.2. Security parameters
  22. Tor uses a stream cipher, a public-key cipher, the Diffie-Hellman
  23. protocol, and a hash function.
  24. KEY_LEN -- the length of the stream cipher's key, in bytes.
  25. PK_ENC_LEN -- the length of a public-key encrypted message, in bytes.
  26. PK_PAD_LEN -- the number of bytes added in padding for public-key
  27. encryption, in bytes. (The largest number of bytes that can be encrypted
  28. in a single public-key operation is therefore PK_ENC_LEN-PK_PAD_LEN.)
  29. DH_LEN -- the number of bytes used to represent a member of the
  30. Diffie-Hellman group.
  31. DH_SEC_LEN -- the number of bytes used in a Diffie-Hellman private key (x).
  32. HASH_LEN -- the length of the hash function's output, in bytes.
  33. PAYLOAD_LEN -- The longest allowable cell payload, in bytes. (509)
  34. CELL_LEN -- The length of a Tor cell, in bytes.
  35. 0.3. Ciphers
  36. For a stream cipher, we use 128-bit AES in counter mode, with an IV of all
  37. 0 bytes.
  38. For a public-key cipher, we use RSA with 1024-bit keys and a fixed
  39. exponent of 65537. We use OAEP-MGF1 padding, with SHA-1 as its digest
  40. function. We leave optional the "Label" parameter unset. (For OAEP
  41. padding, see ftp://ftp.rsasecurity.com/pub/pkcs/pkcs-1/pkcs-1v2-1.pdf)
  42. [Nick, what does "we leave optional the Label parameter unset" mean? -RD]
  43. For Diffie-Hellman, we use a generator (g) of 2. For the modulus (p), we
  44. use the 1024-bit safe prime from rfc2409 section 6.2 whose hex
  45. representation is:
  46. "FFFFFFFFFFFFFFFFC90FDAA22168C234C4C6628B80DC1CD129024E08"
  47. "8A67CC74020BBEA63B139B22514A08798E3404DDEF9519B3CD3A431B"
  48. "302B0A6DF25F14374FE1356D6D51C245E485B576625E7EC6F44C42E9"
  49. "A637ED6B0BFF5CB6F406B7EDEE386BFB5A899FA5AE9F24117C4B1FE6"
  50. "49286651ECE65381FFFFFFFFFFFFFFFF"
  51. As an optimization, implementations SHOULD choose DH private keys (x) of
  52. 320 bits. Implementations that do this MUST never use any DH key more
  53. than once.
  54. [May other implementations reuse their DH keys?? -RD]
  55. [Probably not. Conceivably, you could get away with changing DH keys once
  56. per second, but there are too many oddball attacks for me to be
  57. comfortable that this is safe. -NM]
  58. For a hash function, we use SHA-1.
  59. KEY_LEN=16.
  60. DH_LEN=128; DH_SEC_LEN=40.
  61. PK_ENC_LEN=128; PK_PAD_LEN=42.
  62. HASH_LEN=20.
  63. When we refer to "the hash of a public key", we mean the SHA-1 hash of the
  64. DER encoding of an ASN.1 RSA public key (as specified in PKCS.1).
  65. All "random" values should be generated with a cryptographically strong
  66. random number generator, unless otherwise noted.
  67. The "hybrid encryption" of a byte sequence M with a public key PK is
  68. computed as follows:
  69. 1. If M is less than PK_ENC_LEN-PK_PAD_LEN, pad and encrypt M with PK.
  70. 2. Otherwise, generate a KEY_LEN byte random key K.
  71. Let M1 = the first PK_ENC_LEN-PK_PAD_LEN-KEY_LEN bytes of M,
  72. and let M2 = the rest of M.
  73. Pad and encrypt K|M1 with PK. Encrypt M2 with our stream cipher,
  74. using the key K. Concatenate these encrypted values.
  75. [XXX Note that this "hybrid encryption" approach does not prevent
  76. an attacker from adding or removing bytes to the end of M. It also
  77. allows attackers to modify the bytes not covered by the OAEP --
  78. see Goldberg's PET2006 paper for details. We will add a MAC to this
  79. scheme one day. -RD]
  80. 0.4. Other parameter values
  81. CELL_LEN=512
  82. 1. System overview
  83. Tor is a distributed overlay network designed to anonymize
  84. low-latency TCP-based applications such as web browsing, secure shell,
  85. and instant messaging. Clients choose a path through the network and
  86. build a ``circuit'', in which each node (or ``onion router'' or ``OR'')
  87. in the path knows its predecessor and successor, but no other nodes in
  88. the circuit. Traffic flowing down the circuit is sent in fixed-size
  89. ``cells'', which are unwrapped by a symmetric key at each node (like
  90. the layers of an onion) and relayed downstream.
  91. 1.1. Keys and names
  92. Every Tor server has multiple public/private keypairs:
  93. - A long-term signing-only "Identity key" used to sign documents and
  94. certificates, and used to establish server identity.
  95. - A medium-term "Onion key" used to decrypt onion skins when accepting
  96. circuit extend attempts. (See 5.1.) Old keys MUST be accepted for at
  97. least one week after they are no longer advertised. Because of this,
  98. servers MUST retain old keys for a while after they're rotated.
  99. - A short-term "Connection key" used to negotiate TLS connections.
  100. Tor implementations MAY rotate this key as often as they like, and
  101. SHOULD rotate this key at least once a day.
  102. Tor servers are also identified by "nicknames"; these are specified in
  103. dir-spec.txt.
  104. 2. Connections
  105. Tor uses TLS/SSLv3 for link authentication and encryption. All
  106. implementations MUST support the SSLv3 ciphersuite
  107. "SSL_DHE_RSA_WITH_3DES_EDE_CBC_SHA", and SHOULD support the TLS
  108. ciphersuite "TLS_DHE_RSA_WITH_AES_128_CBC_SHA" if it is available.
  109. Implementations MAY support other TLS ciphersuites, but MUST NOT
  110. support any suite that lacks ephemeral keys, or whose symmetric keys are
  111. less then KEY_LEN bits, or whose digests are less than HASH_LEN bits.
  112. Implementations SHOULD NOT allow other SSLv3 ciphersuites.
  113. Even though the connection protocol is identical, we will think of the
  114. initiator as either an onion router (OR) if it is willing to relay
  115. traffic for other Tor users, or an onion proxy (OP) if it only handles
  116. local requests. Onion proxies SHOULD NOT provide long-term-trackable
  117. identifiers in their handshakes.
  118. During the TLS handshake, the connection initiator always sends a
  119. two-certificate chain, consisting of an X.509 certificate using a
  120. short-term connection public key and a second, self- signed X.509
  121. certificate containing its identity key. The commonName of the first
  122. certificate is the OR's nickname, and the commonName of the second
  123. certificate is the OR's nickname, followed by a space and the string
  124. "<identity>". The other party sends a similar certificate chain.
  125. Implementations running Protocol 1 and earlier use an
  126. organizationName of "Tor" or "TOR". Future implementations (which
  127. support the version negotiation protocol in section 4.1) MUST NOT
  128. have either of these values for their organizationName.
  129. All parties receiving certificates must confirm that the identity key is
  130. as expected. (When initiating a connection, the expected identity key is
  131. the one given in the directory; when creating a connection because of an
  132. EXTEND cell, the expected identity key is the one given in the cell.) If
  133. the key is not as expected, the party must close the connection.
  134. When connecting to an OR, all parties SHOULD reject the connection if that
  135. OR has a malformed or missing certificate. When accepting an incoming
  136. connection, an OR SHOULD NOT reject incoming connections from parties with
  137. malformed or missing certificates. (However, an OR should not believe
  138. that an incoming connection is from another OR unless the certificates
  139. are present and well-formed.)
  140. [Before version 0.1.2.8-rc, ORs rejected incoming connections from ORs and
  141. OPs alike if their certificates were missing or malformed.]
  142. Once a TLS connection is established, the two sides send cells
  143. (specified below) to one another. Cells are sent serially. All
  144. cells are CELL_LEN bytes long. Cells may be sent embedded in TLS
  145. records of any size or divided across TLS records, but the framing
  146. of TLS records MUST NOT leak information about the type or contents
  147. of the cells.
  148. TLS connections are not permanent. Either side may close a connection
  149. if there are no circuits running over it and an amount of time
  150. (KeepalivePeriod, defaults to 5 minutes) has passed.
  151. (As an exception, directory servers may try to stay connected to all of
  152. the ORs -- though this will be phased out for the Tor 0.1.2.x release.)
  153. 3. Cell Packet format
  154. The basic unit of communication for onion routers and onion
  155. proxies is a fixed-width "cell".
  156. On a version 1 connection, each cell contains the following
  157. fields:
  158. CircID [2 bytes]
  159. Command [1 byte]
  160. Payload (padded with 0 bytes) [PAYLOAD_LEN bytes]
  161. The CircID field determines which circuit, if any, the cell is
  162. associated with.
  163. The 'Command' field holds one of the following values:
  164. 0 -- PADDING (Padding) (See Sec 7.2)
  165. 1 -- CREATE (Create a circuit) (See Sec 5.1)
  166. 2 -- CREATED (Acknowledge create) (See Sec 5.1)
  167. 3 -- RELAY (End-to-end data) (See Sec 5.5 and 6)
  168. 4 -- DESTROY (Stop using a circuit) (See Sec 5.4)
  169. 5 -- CREATE_FAST (Create a circuit, no PK) (See Sec 5.1)
  170. 6 -- CREATED_FAST (Circuit created, no PK) (See Sec 5.1)
  171. The interpretation of 'Payload' depends on the type of the cell.
  172. PADDING: Payload is unused.
  173. CREATE: Payload contains the handshake challenge.
  174. CREATED: Payload contains the handshake response.
  175. RELAY: Payload contains the relay header and relay body.
  176. DESTROY: Payload contains a reason for closing the circuit.
  177. (see 5.4)
  178. Upon receiving any other value for the command field, an OR must
  179. drop the cell. [XXXX Versions prior to 0.1.0.?? logged a warning
  180. when dropping the cell; this is bad behavior. -NM]
  181. The payload is padded with 0 bytes.
  182. PADDING cells are currently used to implement connection keepalive.
  183. If there is no other traffic, ORs and OPs send one another a PADDING
  184. cell every few minutes.
  185. CREATE, CREATED, and DESTROY cells are used to manage circuits;
  186. see section 4 below.
  187. RELAY cells are used to send commands and data along a circuit; see
  188. section 5 below.
  189. 4. [This section deliberately left blank.]
  190. 5. Circuit management
  191. 5.1. CREATE and CREATED cells
  192. Users set up circuits incrementally, one hop at a time. To create a
  193. new circuit, OPs send a CREATE cell to the first node, with the
  194. first half of the DH handshake; that node responds with a CREATED
  195. cell with the second half of the DH handshake plus the first 20 bytes
  196. of derivative key data (see section 5.2). To extend a circuit past
  197. the first hop, the OP sends an EXTEND relay cell (see section 5)
  198. which instructs the last node in the circuit to send a CREATE cell
  199. to extend the circuit.
  200. The payload for a CREATE cell is an 'onion skin', which consists
  201. of the first step of the DH handshake data (also known as g^x).
  202. This value is hybrid-encrypted (see 0.3) to Bob's onion key, giving
  203. an onion-skin of:
  204. PK-encrypted:
  205. Padding padding [PK_PAD_LEN bytes]
  206. Symmetric key [KEY_LEN bytes]
  207. First part of g^x [PK_ENC_LEN-PK_PAD_LEN-KEY_LEN bytes]
  208. Symmetrically encrypted:
  209. Second part of g^x [DH_LEN-(PK_ENC_LEN-PK_PAD_LEN-KEY_LEN)
  210. bytes]
  211. The relay payload for an EXTEND relay cell consists of:
  212. Address [4 bytes]
  213. Port [2 bytes]
  214. Onion skin [DH_LEN+KEY_LEN+PK_PAD_LEN bytes]
  215. Identity fingerprint [HASH_LEN bytes]
  216. The port and address field denote the IPV4 address and port of the next
  217. onion router in the circuit; the public key hash is the hash of the PKCS#1
  218. ASN1 encoding of the next onion router's identity (signing) key. (See 0.3
  219. above.) (Including this hash allows the extending OR verify that it is
  220. indeed connected to the correct target OR, and prevents certain
  221. man-in-the-middle attacks.)
  222. The payload for a CREATED cell, or the relay payload for an
  223. EXTENDED cell, contains:
  224. DH data (g^y) [DH_LEN bytes]
  225. Derivative key data (KH) [HASH_LEN bytes] <see 5.2 below>
  226. The CircID for a CREATE cell is an arbitrarily chosen 2-byte integer,
  227. selected by the node (OP or OR) that sends the CREATE cell. To prevent
  228. CircID collisions, when one node sends a CREATE cell to another, it chooses
  229. from only one half of the possible values based on the ORs' public
  230. identity keys: if the sending node has a lower key, it chooses a CircID with
  231. an MSB of 0; otherwise, it chooses a CircID with an MSB of 1.
  232. (An OP with no public key MAY choose any CircID it wishes, since an OP
  233. never needs to process a CREATE cell.)
  234. Public keys are compared numerically by modulus.
  235. As usual with DH, x and y MUST be generated randomly.
  236. [
  237. To implement backward-compatible version negotiation, parties MUST
  238. drop CREATE cells with all-[00] onion-skins.
  239. ]
  240. 5.1.1. CREATE_FAST/CREATED_FAST cells
  241. When initializing the first hop of a circuit, the OP has already
  242. established the OR's identity and negotiated a secret key using TLS.
  243. Because of this, it is not always necessary for the OP to perform the
  244. public key operations to create a circuit. In this case, the
  245. OP MAY send a CREATE_FAST cell instead of a CREATE cell for the first
  246. hop only. The OR responds with a CREATED_FAST cell, and the circuit is
  247. created.
  248. A CREATE_FAST cell contains:
  249. Key material (X) [HASH_LEN bytes]
  250. A CREATED_FAST cell contains:
  251. Key material (Y) [HASH_LEN bytes]
  252. Derivative key data [HASH_LEN bytes] (See 5.2 below)
  253. The values of X and Y must be generated randomly.
  254. [Versions of Tor before 0.1.0.6-rc did not support these cell types;
  255. clients should not send CREATE_FAST cells to older Tor servers.]
  256. If an OR sees a circuit created with CREATE_FAST, the OR is sure to be the
  257. first hop of a circuit. ORs SHOULD reject attempts to create streams with
  258. RELAY_BEGIN exiting the circuit at the first hop: letting Tor be used as a
  259. single hop proxy makes exit nodes a more attractive target for compromise.
  260. 5.2. Setting circuit keys
  261. Once the handshake between the OP and an OR is completed, both can
  262. now calculate g^xy with ordinary DH. Before computing g^xy, both client
  263. and server MUST verify that the received g^x or g^y value is not degenerate;
  264. that is, it must be strictly greater than 1 and strictly less than p-1
  265. where p is the DH modulus. Implementations MUST NOT complete a handshake
  266. with degenerate keys. Implementations MUST NOT discard other "weak"
  267. g^x values.
  268. (Discarding degenerate keys is critical for security; if bad keys
  269. are not discarded, an attacker can substitute the server's CREATED
  270. cell's g^y with 0 or 1, thus creating a known g^xy and impersonating
  271. the server. Discarding other keys may allow attacks to learn bits of
  272. the private key.)
  273. (The mainline Tor implementation, in the 0.1.1.x-alpha series, discarded
  274. all g^x values less than 2^24, greater than p-2^24, or having more than
  275. 1024-16 identical bits. This served no useful purpose, and we stopped.)
  276. If CREATE or EXTEND is used to extend a circuit, the client and server
  277. base their key material on K0=g^xy, represented as a big-endian unsigned
  278. integer.
  279. If CREATE_FAST is used, the client and server base their key material on
  280. K0=X|Y.
  281. From the base key material K0, they compute KEY_LEN*2+HASH_LEN*3 bytes of
  282. derivative key data as
  283. K = H(K0 | [00]) | H(K0 | [01]) | H(K0 | [02]) | ...
  284. The first HASH_LEN bytes of K form KH; the next HASH_LEN form the forward
  285. digest Df; the next HASH_LEN 41-60 form the backward digest Db; the next
  286. KEY_LEN 61-76 form Kf, and the final KEY_LEN form Kb. Excess bytes from K
  287. are discarded.
  288. KH is used in the handshake response to demonstrate knowledge of the
  289. computed shared key. Df is used to seed the integrity-checking hash
  290. for the stream of data going from the OP to the OR, and Db seeds the
  291. integrity-checking hash for the data stream from the OR to the OP. Kf
  292. is used to encrypt the stream of data going from the OP to the OR, and
  293. Kb is used to encrypt the stream of data going from the OR to the OP.
  294. 5.3. Creating circuits
  295. When creating a circuit through the network, the circuit creator
  296. (OP) performs the following steps:
  297. 1. Choose an onion router as an exit node (R_N), such that the onion
  298. router's exit policy includes at least one pending stream that
  299. needs a circuit (if there are any).
  300. 2. Choose a chain of (N-1) onion routers
  301. (R_1...R_N-1) to constitute the path, such that no router
  302. appears in the path twice.
  303. 3. If not already connected to the first router in the chain,
  304. open a new connection to that router.
  305. 4. Choose a circID not already in use on the connection with the
  306. first router in the chain; send a CREATE cell along the
  307. connection, to be received by the first onion router.
  308. 5. Wait until a CREATED cell is received; finish the handshake
  309. and extract the forward key Kf_1 and the backward key Kb_1.
  310. 6. For each subsequent onion router R (R_2 through R_N), extend
  311. the circuit to R.
  312. To extend the circuit by a single onion router R_M, the OP performs
  313. these steps:
  314. 1. Create an onion skin, encrypted to R_M's public onion key.
  315. 2. Send the onion skin in a relay EXTEND cell along
  316. the circuit (see section 5).
  317. 3. When a relay EXTENDED cell is received, verify KH, and
  318. calculate the shared keys. The circuit is now extended.
  319. When an onion router receives an EXTEND relay cell, it sends a CREATE
  320. cell to the next onion router, with the enclosed onion skin as its
  321. payload. The initiating onion router chooses some circID not yet
  322. used on the connection between the two onion routers. (But see
  323. section 5.1. above, concerning choosing circIDs based on
  324. lexicographic order of nicknames.)
  325. When an onion router receives a CREATE cell, if it already has a
  326. circuit on the given connection with the given circID, it drops the
  327. cell. Otherwise, after receiving the CREATE cell, it completes the
  328. DH handshake, and replies with a CREATED cell. Upon receiving a
  329. CREATED cell, an onion router packs it payload into an EXTENDED relay
  330. cell (see section 5), and sends that cell up the circuit. Upon
  331. receiving the EXTENDED relay cell, the OP can retrieve g^y.
  332. (As an optimization, OR implementations may delay processing onions
  333. until a break in traffic allows time to do so without harming
  334. network latency too greatly.)
  335. 5.4. Tearing down circuits
  336. Circuits are torn down when an unrecoverable error occurs along
  337. the circuit, or when all streams on a circuit are closed and the
  338. circuit's intended lifetime is over. Circuits may be torn down
  339. either completely or hop-by-hop.
  340. To tear down a circuit completely, an OR or OP sends a DESTROY
  341. cell to the adjacent nodes on that circuit, using the appropriate
  342. direction's circID.
  343. Upon receiving an outgoing DESTROY cell, an OR frees resources
  344. associated with the corresponding circuit. If it's not the end of
  345. the circuit, it sends a DESTROY cell for that circuit to the next OR
  346. in the circuit. If the node is the end of the circuit, then it tears
  347. down any associated edge connections (see section 6.1).
  348. After a DESTROY cell has been processed, an OR ignores all data or
  349. destroy cells for the corresponding circuit.
  350. To tear down part of a circuit, the OP may send a RELAY_TRUNCATE cell
  351. signaling a given OR (Stream ID zero). That OR sends a DESTROY
  352. cell to the next node in the circuit, and replies to the OP with a
  353. RELAY_TRUNCATED cell.
  354. When an unrecoverable error occurs along one connection in a
  355. circuit, the nodes on either side of the connection should, if they
  356. are able, act as follows: the node closer to the OP should send a
  357. RELAY_TRUNCATED cell towards the OP; the node farther from the OP
  358. should send a DESTROY cell down the circuit.
  359. The payload of a RELAY_TRUNCATED or DESTROY cell contains a single octet,
  360. describing why the circuit is being closed or truncated. When sending a
  361. TRUNCATED or DESTROY cell because of another TRUNCATED or DESTROY cell,
  362. the error code should be propagated. The origin of a circuit always sets
  363. this error code to 0, to avoid leaking its version.
  364. The error codes are:
  365. 0 -- NONE (No reason given.)
  366. 1 -- PROTOCOL (Tor protocol violation.)
  367. 2 -- INTERNAL (Internal error.)
  368. 3 -- REQUESTED (A client sent a TRUNCATE command.)
  369. 4 -- HIBERNATING (Not currently operating; trying to save bandwidth.)
  370. 5 -- RESOURCELIMIT (Out of memory, sockets, or circuit IDs.)
  371. 6 -- CONNECTFAILED (Unable to reach server.)
  372. 7 -- OR_IDENTITY (Connected to server, but its OR identity was not
  373. as expected.)
  374. 8 -- OR_CONN_CLOSED (The OR connection that was carrying this circuit
  375. died.)
  376. 9 -- FINISHED (The circuit has expired for being dirty or old.)
  377. 10 -- TIMEOUT (Circuit construction took too long)
  378. 11 -- DESTROYED (The circuit was destroyed w/o client TRUNCATE)
  379. 12 -- NOSUCHSERVICE (Request for unknown hidden service)
  380. [Versions of Tor prior to 0.1.0.11 didn't send reasons; implementations
  381. MUST accept empty TRUNCATED and DESTROY cells.]
  382. 5.5. Routing relay cells
  383. When an OR receives a RELAY cell, it checks the cell's circID and
  384. determines whether it has a corresponding circuit along that
  385. connection. If not, the OR drops the RELAY cell.
  386. Otherwise, if the OR is not at the OP edge of the circuit (that is,
  387. either an 'exit node' or a non-edge node), it de/encrypts the payload
  388. with the stream cipher, as follows:
  389. 'Forward' relay cell (same direction as CREATE):
  390. Use Kf as key; decrypt.
  391. 'Back' relay cell (opposite direction from CREATE):
  392. Use Kb as key; encrypt.
  393. Note that in counter mode, decrypt and encrypt are the same operation.
  394. The OR then decides whether it recognizes the relay cell, by
  395. inspecting the payload as described in section 6.1 below. If the OR
  396. recognizes the cell, it processes the contents of the relay cell.
  397. Otherwise, it passes the decrypted relay cell along the circuit if
  398. the circuit continues. If the OR at the end of the circuit
  399. encounters an unrecognized relay cell, an error has occurred: the OR
  400. sends a DESTROY cell to tear down the circuit.
  401. When a relay cell arrives at an OP, the OP decrypts the payload
  402. with the stream cipher as follows:
  403. OP receives data cell:
  404. For I=N...1,
  405. Decrypt with Kb_I. If the payload is recognized (see
  406. section 6..1), then stop and process the payload.
  407. For more information, see section 6 below.
  408. 6. Application connections and stream management
  409. 6.1. Relay cells
  410. Within a circuit, the OP and the exit node use the contents of
  411. RELAY packets to tunnel end-to-end commands and TCP connections
  412. ("Streams") across circuits. End-to-end commands can be initiated
  413. by either edge; streams are initiated by the OP.
  414. The payload of each unencrypted RELAY cell consists of:
  415. Relay command [1 byte]
  416. 'Recognized' [2 bytes]
  417. StreamID [2 bytes]
  418. Digest [4 bytes]
  419. Length [2 bytes]
  420. Data [CELL_LEN-14 bytes]
  421. The relay commands are:
  422. 1 -- RELAY_BEGIN [forward]
  423. 2 -- RELAY_DATA [forward or backward]
  424. 3 -- RELAY_END [forward or backward]
  425. 4 -- RELAY_CONNECTED [backward]
  426. 5 -- RELAY_SENDME [forward or backward] [sometimes control]
  427. 6 -- RELAY_EXTEND [forward] [control]
  428. 7 -- RELAY_EXTENDED [backward] [control]
  429. 8 -- RELAY_TRUNCATE [forward] [control]
  430. 9 -- RELAY_TRUNCATED [backward] [control]
  431. 10 -- RELAY_DROP [forward or backward] [control]
  432. 11 -- RELAY_RESOLVE [forward]
  433. 12 -- RELAY_RESOLVED [backward]
  434. 13 -- RELAY_BEGIN_DIR [forward]
  435. 32..40 -- Used for hidden services; see rend-spec.txt.
  436. Commands labelled as "forward" must only be sent by the originator
  437. of the circuit. Commands labelled as "backward" must only be sent by
  438. other nodes in the circuit back to the originator. Commands marked
  439. as either can be sent either by the originator or other nodes.
  440. The 'recognized' field in any unencrypted relay payload is always set
  441. to zero; the 'digest' field is computed as the first four bytes of
  442. the running digest of all the bytes that have been destined for
  443. this hop of the circuit or originated from this hop of the circuit,
  444. seeded from Df or Db respectively (obtained in section 5.2 above),
  445. and including this RELAY cell's entire payload (taken with the digest
  446. field set to zero).
  447. When the 'recognized' field of a RELAY cell is zero, and the digest
  448. is correct, the cell is considered "recognized" for the purposes of
  449. decryption (see section 5.5 above).
  450. (The digest does not include any bytes from relay cells that do
  451. not start or end at this hop of the circuit. That is, it does not
  452. include forwarded data. Therefore if 'recognized' is zero but the
  453. digest does not match, the running digest at that node should
  454. not be updated, and the cell should be forwarded on.)
  455. All RELAY cells pertaining to the same tunneled stream have the
  456. same stream ID. StreamIDs are chosen arbitrarily by the OP. RELAY
  457. cells that affect the entire circuit rather than a particular
  458. stream use a StreamID of zero -- they are marked in the table above
  459. as "[control]" style cells. (Sendme cells are marked as "sometimes
  460. control" because they can take include a StreamID or not depending
  461. on their purpose -- see Section 7.)
  462. The 'Length' field of a relay cell contains the number of bytes in
  463. the relay payload which contain real payload data. The remainder of
  464. the payload is padded with NUL bytes.
  465. If the RELAY cell is recognized but the relay command is not
  466. understood, the cell must be dropped and ignored. Its contents
  467. still count with respect to the digests, though. [Before
  468. 0.1.1.10, Tor closed circuits when it received an unknown relay
  469. command. Perhaps this will be more forward-compatible. -RD]
  470. 6.2. Opening streams and transferring data
  471. To open a new anonymized TCP connection, the OP chooses an open
  472. circuit to an exit that may be able to connect to the destination
  473. address, selects an arbitrary StreamID not yet used on that circuit,
  474. and constructs a RELAY_BEGIN cell with a payload encoding the address
  475. and port of the destination host. The payload format is:
  476. ADDRESS | ':' | PORT | [00]
  477. where ADDRESS can be a DNS hostname, or an IPv4 address in
  478. dotted-quad format, or an IPv6 address surrounded by square brackets;
  479. and where PORT is a decimal integer between 1 and 65535, inclusive.
  480. [What is the [00] for? -NM]
  481. [It's so the payload is easy to parse out with string funcs -RD]
  482. Upon receiving this cell, the exit node resolves the address as
  483. necessary, and opens a new TCP connection to the target port. If the
  484. address cannot be resolved, or a connection can't be established, the
  485. exit node replies with a RELAY_END cell. (See 6.4 below.)
  486. Otherwise, the exit node replies with a RELAY_CONNECTED cell, whose
  487. payload is in one of the following formats:
  488. The IPv4 address to which the connection was made [4 octets]
  489. A number of seconds (TTL) for which the address may be cached [4 octets]
  490. or
  491. Four zero-valued octets [4 octets]
  492. An address type (6) [1 octet]
  493. The IPv6 address to which the connection was made [16 octets]
  494. A number of seconds (TTL) for which the address may be cached [4 octets]
  495. [XXXX Versions of Tor before 0.1.1.6 ignore and do not generate the TTL
  496. field. No version of Tor currently generates the IPv6 format.
  497. Tor servers before 0.1.2.0 set the TTL field to a fixed value. Later
  498. versions set the TTL to the last value seen from a DNS server, and expire
  499. their own cached entries after a fixed interval. This prevents certain
  500. attacks.]
  501. The OP waits for a RELAY_CONNECTED cell before sending any data.
  502. Once a connection has been established, the OP and exit node
  503. package stream data in RELAY_DATA cells, and upon receiving such
  504. cells, echo their contents to the corresponding TCP stream.
  505. RELAY_DATA cells sent to unrecognized streams are dropped.
  506. Relay RELAY_DROP cells are long-range dummies; upon receiving such
  507. a cell, the OR or OP must drop it.
  508. 6.2.1. Opening a directory stream
  509. If a Tor server is a directory server, it should respond to a
  510. RELAY_BEGIN_DIR cell as if it had received a BEGIN cell requesting a
  511. connection to its directory port. RELAY_BEGIN_DIR cells ignore exit
  512. policy, since the stream is local to the Tor process.
  513. If the Tor server is not running a directory service, it should respond
  514. with a REASON_NOTDIRECTORY RELAY_END cell.
  515. Clients MUST generate an all-zero payload for RELAY_BEGIN_DIR cells,
  516. and servers MUST ignore the payload.
  517. [RELAY_BEGIN_DIR was not supported before Tor 0.1.2.2-alpha; clients
  518. SHOULD NOT send it to routers running earlier versions of Tor.]
  519. 6.3. Closing streams
  520. When an anonymized TCP connection is closed, or an edge node
  521. encounters error on any stream, it sends a 'RELAY_END' cell along the
  522. circuit (if possible) and closes the TCP connection immediately. If
  523. an edge node receives a 'RELAY_END' cell for any stream, it closes
  524. the TCP connection completely, and sends nothing more along the
  525. circuit for that stream.
  526. The payload of a RELAY_END cell begins with a single 'reason' byte to
  527. describe why the stream is closing, plus optional data (depending on
  528. the reason.) The values are:
  529. 1 -- REASON_MISC (catch-all for unlisted reasons)
  530. 2 -- REASON_RESOLVEFAILED (couldn't look up hostname)
  531. 3 -- REASON_CONNECTREFUSED (remote host refused connection) [*]
  532. 4 -- REASON_EXITPOLICY (OR refuses to connect to host or port)
  533. 5 -- REASON_DESTROY (Circuit is being destroyed)
  534. 6 -- REASON_DONE (Anonymized TCP connection was closed)
  535. 7 -- REASON_TIMEOUT (Connection timed out, or OR timed out
  536. while connecting)
  537. 8 -- (unallocated) [**]
  538. 9 -- REASON_HIBERNATING (OR is temporarily hibernating)
  539. 10 -- REASON_INTERNAL (Internal error at the OR)
  540. 11 -- REASON_RESOURCELIMIT (OR has no resources to fulfill request)
  541. 12 -- REASON_CONNRESET (Connection was unexpectedly reset)
  542. 13 -- REASON_TORPROTOCOL (Sent when closing connection because of
  543. Tor protocol violations.)
  544. 14 -- REASON_NOTDIRECTORY (Client sent RELAY_BEGIN_DIR to a
  545. non-directory server.)
  546. (With REASON_EXITPOLICY, the 4-byte IPv4 address or 16-byte IPv6 address
  547. forms the optional data; no other reason currently has extra data.
  548. As of 0.1.1.6, the body also contains a 4-byte TTL.)
  549. OPs and ORs MUST accept reasons not on the above list, since future
  550. versions of Tor may provide more fine-grained reasons.
  551. [*] Older versions of Tor also send this reason when connections are
  552. reset.
  553. [**] Due to a bug in versions of Tor through 0095, error reason 8 must
  554. remain allocated until that version is obsolete.
  555. --- [The rest of this section describes unimplemented functionality.]
  556. Because TCP connections can be half-open, we follow an equivalent
  557. to TCP's FIN/FIN-ACK/ACK protocol to close streams.
  558. An exit connection can have a TCP stream in one of three states:
  559. 'OPEN', 'DONE_PACKAGING', and 'DONE_DELIVERING'. For the purposes
  560. of modeling transitions, we treat 'CLOSED' as a fourth state,
  561. although connections in this state are not, in fact, tracked by the
  562. onion router.
  563. A stream begins in the 'OPEN' state. Upon receiving a 'FIN' from
  564. the corresponding TCP connection, the edge node sends a 'RELAY_FIN'
  565. cell along the circuit and changes its state to 'DONE_PACKAGING'.
  566. Upon receiving a 'RELAY_FIN' cell, an edge node sends a 'FIN' to
  567. the corresponding TCP connection (e.g., by calling
  568. shutdown(SHUT_WR)) and changing its state to 'DONE_DELIVERING'.
  569. When a stream in already in 'DONE_DELIVERING' receives a 'FIN', it
  570. also sends a 'RELAY_FIN' along the circuit, and changes its state
  571. to 'CLOSED'. When a stream already in 'DONE_PACKAGING' receives a
  572. 'RELAY_FIN' cell, it sends a 'FIN' and changes its state to
  573. 'CLOSED'.
  574. If an edge node encounters an error on any stream, it sends a
  575. 'RELAY_END' cell (if possible) and closes the stream immediately.
  576. 6.4. Remote hostname lookup
  577. To find the address associated with a hostname, the OP sends a
  578. RELAY_RESOLVE cell containing the hostname to be resolved. (For a reverse
  579. lookup, the OP sends a RELAY_RESOLVE cell containing an in-addr.arpa
  580. address.) The OR replies with a RELAY_RESOLVED cell containing a status
  581. byte, and any number of answers. Each answer is of the form:
  582. Type (1 octet)
  583. Length (1 octet)
  584. Value (variable-width)
  585. TTL (4 octets)
  586. "Length" is the length of the Value field.
  587. "Type" is one of:
  588. 0x00 -- Hostname
  589. 0x04 -- IPv4 address
  590. 0x06 -- IPv6 address
  591. 0xF0 -- Error, transient
  592. 0xF1 -- Error, nontransient
  593. If any answer has a type of 'Error', then no other answer may be given.
  594. The RELAY_RESOLVE cell must use a nonzero, distinct streamID; the
  595. corresponding RELAY_RESOLVED cell must use the same streamID. No stream
  596. is actually created by the OR when resolving the name.
  597. 7. Flow control
  598. 7.1. Link throttling
  599. Each node should do appropriate bandwidth throttling to keep its
  600. user happy.
  601. Communicants rely on TCP's default flow control to push back when they
  602. stop reading.
  603. 7.2. Link padding
  604. Link padding can be created by sending PADDING cells along the
  605. connection; relay cells of type "DROP" can be used for long-range
  606. padding.
  607. Currently nodes are not required to do any sort of link padding or
  608. dummy traffic. Because strong attacks exist even with link padding,
  609. and because link padding greatly increases the bandwidth requirements
  610. for running a node, we plan to leave out link padding until this
  611. tradeoff is better understood.
  612. 7.3. Circuit-level flow control
  613. To control a circuit's bandwidth usage, each OR keeps track of two
  614. 'windows', consisting of how many RELAY_DATA cells it is allowed to
  615. originate (package for transmission), and how many RELAY_DATA cells
  616. it is willing to consume (receive for local streams). These limits
  617. do not apply to cells that the OR receives from one host and relays
  618. to another.
  619. Each 'window' value is initially set to 1000 data cells
  620. in each direction (cells that are not data cells do not affect
  621. the window). When an OR is willing to deliver more cells, it sends a
  622. RELAY_SENDME cell towards the OP, with Stream ID zero. When an OR
  623. receives a RELAY_SENDME cell with stream ID zero, it increments its
  624. packaging window.
  625. Each of these cells increments the corresponding window by 100.
  626. The OP behaves identically, except that it must track a packaging
  627. window and a delivery window for every OR in the circuit.
  628. An OR or OP sends cells to increment its delivery window when the
  629. corresponding window value falls under some threshold (900).
  630. If a packaging window reaches 0, the OR or OP stops reading from
  631. TCP connections for all streams on the corresponding circuit, and
  632. sends no more RELAY_DATA cells until receiving a RELAY_SENDME cell.
  633. [this stuff is badly worded; copy in the tor-design section -RD]
  634. 7.4. Stream-level flow control
  635. Edge nodes use RELAY_SENDME cells to implement end-to-end flow
  636. control for individual connections across circuits. Similarly to
  637. circuit-level flow control, edge nodes begin with a window of cells
  638. (500) per stream, and increment the window by a fixed value (50)
  639. upon receiving a RELAY_SENDME cell. Edge nodes initiate RELAY_SENDME
  640. cells when both a) the window is <= 450, and b) there are less than
  641. ten cell payloads remaining to be flushed at that edge.
  642. A.1. Differences between spec and implementation
  643. - The current specification requires all ORs to have IPv4 addresses, but
  644. allows servers to exit and resolve to IPv6 addresses, and to declare IPv6
  645. addresses in their exit policies. The current codebase has no IPv6
  646. support at all.