tor-spec.txt 40 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934
  1. $Id$
  2. Tor Protocol Specification
  3. Roger Dingledine
  4. Nick Mathewson
  5. Note: This is an attempt to specify Tor as currently implemented. Future
  6. versions of Tor will implement improved protocols, and compatibility is not
  7. guaranteed.
  8. This is not a design document; most design criteria are not examined. For
  9. more information on why Tor acts as it does, see tor-design.pdf.
  10. TODO: (very soon)
  11. - REASON_CONNECTFAILED should include an IP.
  12. - Copy prose from tor-design to make everything more readable.
  13. 0. Notation:
  14. PK -- a public key.
  15. SK -- a private key
  16. K -- a key for a symmetric cypher
  17. a|b -- concatenation of 'a' and 'b'.
  18. [A0 B1 C2] -- a three-byte sequence, containing the bytes with
  19. hexadecimal values A0, B1, and C2, in that order.
  20. All numeric values are encoded in network (big-endian) order.
  21. Unless otherwise specified, all symmetric ciphers are AES in counter
  22. mode, with an IV of all 0 bytes. Asymmetric ciphers are either RSA
  23. with 1024-bit keys and exponents of 65537, or DH where the generator
  24. is 2 and the modulus is the safe prime from rfc2409, section 6.2,
  25. whose hex representation is:
  26. "FFFFFFFFFFFFFFFFC90FDAA22168C234C4C6628B80DC1CD129024E08"
  27. "8A67CC74020BBEA63B139B22514A08798E3404DDEF9519B3CD3A431B"
  28. "302B0A6DF25F14374FE1356D6D51C245E485B576625E7EC6F44C42E9"
  29. "A637ED6B0BFF5CB6F406B7EDEE386BFB5A899FA5AE9F24117C4B1FE6"
  30. "49286651ECE65381FFFFFFFFFFFFFFFF"
  31. All "hashes" are 20-byte SHA1 cryptographic digests.
  32. When we refer to "the hash of a public key", we mean the SHA1 hash of the
  33. DER encoding of an ASN.1 RSA public key (as specified in PKCS.1).
  34. 1. System overview
  35. Onion Routing is a distributed overlay network designed to anonymize
  36. low-latency TCP-based applications such as web browsing, secure shell,
  37. and instant messaging. Clients choose a path through the network and
  38. build a ``circuit'', in which each node (or ``onion router'' or ``OR'')
  39. in the path knows its predecessor and successor, but no other nodes in
  40. the circuit. Traffic flowing down the circuit is sent in fixed-size
  41. ``cells'', which are unwrapped by a symmetric key at each node (like
  42. the layers of an onion) and relayed downstream.
  43. 2. Connections
  44. There are two ways to connect to an onion router (OR). The first is
  45. as an onion proxy (OP), which allows the OP to authenticate the OR
  46. without authenticating itself. The second is as another OR, which
  47. allows mutual authentication.
  48. Tor uses TLS for link encryption. All implementations MUST support
  49. the TLS ciphersuite "TLS_EDH_RSA_WITH_DES_192_CBC3_SHA", and SHOULD
  50. support "TLS_DHE_RSA_WITH_AES_128_CBC_SHA" if it is available.
  51. Implementations MAY support other ciphersuites, but MUST NOT
  52. support any suite without ephemeral keys, symmetric keys of at
  53. least 128 bits, and digests of at least 160 bits.
  54. An OP or OR always sends a two-certificate chain, consisting of a
  55. certificate using a short-term connection key and a second, self-
  56. signed certificate containing the OR's identity key. The commonName of the
  57. first certificate is the OR's nickname, and the commonName of the second
  58. certificate is the OR's nickname, followed by a space and the string
  59. "<identity>".
  60. All parties receiving certificates must confirm that the identity key is
  61. as expected. (When initiating a connection, the expected identity key is
  62. the one given in the directory; when creating a connection because of an
  63. EXTEND cell, the expected identity key is the one given in the cell.) If
  64. the key is not as expected, the party must close the connection.
  65. All parties SHOULD reject connections to or from ORs that have malformed
  66. or missing certificates. ORs MAY accept or reject connections from OPs
  67. with malformed or missing certificates.
  68. Once a TLS connection is established, the two sides send cells
  69. (specified below) to one another. Cells are sent serially. All
  70. cells are 512 bytes long. Cells may be sent embedded in TLS
  71. records of any size or divided across TLS records, but the framing
  72. of TLS records MUST NOT leak information about the type or contents
  73. of the cells.
  74. TLS connections are not permanent. An OP or an OR may close a
  75. connection to an OR if there are no circuits running over the
  76. connection, and an amount of time (KeepalivePeriod, defaults to 5
  77. minutes) has passed.
  78. (As an exception, directory servers may try to stay connected to all of
  79. the ORs.)
  80. 3. Cell Packet format
  81. The basic unit of communication for onion routers and onion
  82. proxies is a fixed-width "cell". Each cell contains the following
  83. fields:
  84. CircID [2 bytes]
  85. Command [1 byte]
  86. Payload (padded with 0 bytes) [509 bytes]
  87. [Total size: 512 bytes]
  88. The CircID field determines which circuit, if any, the cell is
  89. associated with.
  90. The 'Command' field holds one of the following values:
  91. 0 -- PADDING (Padding) (See Sec 6.2)
  92. 1 -- CREATE (Create a circuit) (See Sec 4)
  93. 2 -- CREATED (Acknowledge create) (See Sec 4)
  94. 3 -- RELAY (End-to-end data) (See Sec 5)
  95. 4 -- DESTROY (Stop using a circuit) (See Sec 4)
  96. 5 -- CREATE_FAST (Create a circuit, no PK) (See sec 4)
  97. 6 -- CREATED_FAST (Circtuit created, no PK) (See Sec 4)
  98. The interpretation of 'Payload' depends on the type of the cell.
  99. PADDING: Payload is unused.
  100. CREATE: Payload contains the handshake challenge.
  101. CREATED: Payload contains the handshake response.
  102. RELAY: Payload contains the relay header and relay body.
  103. DESTROY: Payload is unused.
  104. Upon receiving any other value for the command field, an OR must
  105. drop the cell.
  106. The payload is padded with 0 bytes.
  107. PADDING cells are currently used to implement connection keepalive.
  108. If there is no other traffic, ORs and OPs send one another a PADDING
  109. cell every few minutes.
  110. CREATE, CREATED, and DESTROY cells are used to manage circuits;
  111. see section 4 below.
  112. RELAY cells are used to send commands and data along a circuit; see
  113. section 5 below.
  114. 4. Circuit management
  115. 4.1. CREATE and CREATED cells
  116. Users set up circuits incrementally, one hop at a time. To create a
  117. new circuit, OPs send a CREATE cell to the first node, with the
  118. first half of the DH handshake; that node responds with a CREATED
  119. cell with the second half of the DH handshake plus the first 20 bytes
  120. of derivative key data (see section 4.2). To extend a circuit past
  121. the first hop, the OP sends an EXTEND relay cell (see section 5)
  122. which instructs the last node in the circuit to send a CREATE cell
  123. to extend the circuit.
  124. The payload for a CREATE cell is an 'onion skin', which consists
  125. of the first step of the DH handshake data (also known as g^x).
  126. The data is encrypted to Bob's PK as follows: Suppose Bob's PK
  127. modulus is L octets long. If the data to be encrypted is shorter
  128. than L-42, then it is encrypted directly (with OAEP padding: see
  129. ftp://ftp.rsasecurity.com/pub/pkcs/pkcs-1/pkcs-1v2-1.pdf). If the
  130. data is at least as long as L-42, then a randomly generated 16-byte
  131. symmetric key is prepended to the data, after which the first L-16-42
  132. bytes of the data are encrypted with Bob's PK; and the rest of the
  133. data is encrypted with the symmetric key.
  134. So in this case, the onion skin on the wire looks like:
  135. RSA-encrypted:
  136. OAEP padding [42 bytes]
  137. Symmetric key [16 bytes]
  138. First part of g^x [70 bytes]
  139. Symmetrically encrypted:
  140. Second part of g^x [58 bytes]
  141. The relay payload for an EXTEND relay cell consists of:
  142. Address [4 bytes]
  143. Port [2 bytes]
  144. Onion skin [186 bytes]
  145. Public key hash [20 bytes]
  146. The port and address field denote the IPV4 address and port of the next
  147. onion router in the circuit; the public key hash is the SHA1 hash of the
  148. PKCS#1 ASN1 encoding of the next onion router's identity (signing) key.
  149. The payload for a CREATED cell, or the relay payload for an
  150. EXTENDED cell, contains:
  151. DH data (g^y) [128 bytes]
  152. Derivative key data (KH) [20 bytes] <see 4.2 below>
  153. The CircID for a CREATE cell is an arbitrarily chosen 2-byte integer,
  154. selected by the node (OP or OR) that sends the CREATE cell. To prevent
  155. CircID collisions, when one OR sends a CREATE cell to another, it chooses
  156. from only one half of the possible values based on the ORs' public
  157. identity keys: if the sending OR has a lower key, it chooses a CircID with
  158. an MSB of 0; otherwise, it chooses a CircID with an MSB of 1.
  159. Public keys are compared numerically by modulus.
  160. (Older versions of Tor compared OR nicknames, and did it in a broken and
  161. unreliable way. To support versions of Tor earlier than 0.0.9pre6,
  162. implementations should notice when the other side of a connection is
  163. sending CREATE cells with the "wrong" MSG, and switch accordingly.)
  164. 4.1.1. CREATE_FAST/CREATED_FAST cells
  165. When initializing the first hop of a circuit, the OP has already
  166. established the OR's identity and negotiated a secret key using TLS.
  167. Because of this, it is not always necessary for the OP to perform the
  168. public key operations to create a circuit. In this case, the
  169. OP SHOULD send a CREATE_FAST cell instead of a CREATE cell for the first
  170. hop only. The OR responds with a CREATED_FAST cell, and the circuit is
  171. created.
  172. A CREATE_FAST cell contains:
  173. Key material (X) [20 bytes]
  174. A CREATED_FAST cell contains:
  175. Key material (Y) [20 bytes]
  176. Derivative key data [20 bytes]
  177. [Versions of Tor before 0.1.0.6-rc did not support these cell types;
  178. clients should not send CREATE_FAST cells to older Tor servers.]
  179. 4.2. Setting circuit keys
  180. Once the handshake between the OP and an OR is completed, both servers can
  181. now calculate g^xy with ordinary DH. Before computing g^xy, both client
  182. and server MUST verify that the received g^x/g^y value is not degenerate;
  183. that is, it must be strictly greater than 1 and strictly less than p-1
  184. where p is the DH modulus. Implementations MUST NOT complete a handshake
  185. with degenerate keys. Implementions MAY discard other "weak" g^x values.
  186. (Discarding degenerate keys is critical for security; if bad keys are not
  187. discarded, an attacker can substitute the server's CREATED cell's g^y with
  188. 0 or 1, thus creating a known g^xy and impersonating the server.)
  189. (The mainline Tor implementation discards all g^x values that are less
  190. than 2^24, that are greater than p-2^24, or that have more than 1024-16
  191. identical bits. This constitutes a negligible portion of the keyspace;
  192. the chances of stumbling on such a key at random are astronomically
  193. small. Nevertheless, implementors may wish to make their implementations
  194. discard such keys.)
  195. From the base key material g^xy, they compute derivative key material as
  196. follows. First, the server represents g^xy as a big-endian unsigned
  197. integer. Next, the server computes 100 bytes of key data as K = SHA1(g^xy
  198. | [00]) | SHA1(g^xy | [01]) | ... SHA1(g^xy | [04]) where "00" is a single
  199. octet whose value is zero, [01] is a single octet whose value is one, etc.
  200. The first 20 bytes of K form KH, bytes 21-40 form the forward digest Df,
  201. 41-60 form the backward digest Db, 61-76 form Kf, and 77-92 form Kb.
  202. KH is used in the handshake response to demonstrate knowledge of the
  203. computed shared key. Df is used to seed the integrity-checking hash
  204. for the stream of data going from the OP to the OR, and Db seeds the
  205. integrity-checking hash for the data stream from the OR to the OP. Kf
  206. is used to encrypt the stream of data going from the OP to the OR, and
  207. Kb is used to encrypt the stream of data going from the OR to the OP.
  208. The fast-setup case uses the same formula, except that X|Y is used
  209. in place of g^xy in determining K. That is,
  210. K = SHA1(X|Y | [00]) | SHA1(X|Y | [01]) | ... SHA1(X|Y| | [04])
  211. The values KH, Kf, Kb, Df, and Db are established and used as before.
  212. 4.3. Creating circuits
  213. When creating a circuit through the network, the circuit creator
  214. (OP) performs the following steps:
  215. 1. Choose an onion router as an exit node (R_N), such that the onion
  216. router's exit policy does not exclude all pending streams
  217. that need a circuit.
  218. 2. Choose a chain of (N-1) onion routers
  219. (R_1...R_N-1) to constitute the path, such that no router
  220. appears in the path twice.
  221. 3. If not already connected to the first router in the chain,
  222. open a new connection to that router.
  223. 4. Choose a circID not already in use on the connection with the
  224. first router in the chain; send a CREATE cell along the
  225. connection, to be received by the first onion router.
  226. 5. Wait until a CREATED cell is received; finish the handshake
  227. and extract the forward key Kf_1 and the backward key Kb_1.
  228. 6. For each subsequent onion router R (R_2 through R_N), extend
  229. the circuit to R.
  230. To extend the circuit by a single onion router R_M, the OP performs
  231. these steps:
  232. 1. Create an onion skin, encrypted to R_M's public key.
  233. 2. Send the onion skin in a relay EXTEND cell along
  234. the circuit (see section 5).
  235. 3. When a relay EXTENDED cell is received, verify KH, and
  236. calculate the shared keys. The circuit is now extended.
  237. When an onion router receives an EXTEND relay cell, it sends a CREATE
  238. cell to the next onion router, with the enclosed onion skin as its
  239. payload. The initiating onion router chooses some circID not yet
  240. used on the connection between the two onion routers. (But see
  241. section 4.1. above, concerning choosing circIDs based on
  242. lexicographic order of nicknames.)
  243. As an extension (called router twins), if the desired next onion
  244. router R in the circuit is down, and some other onion router R'
  245. has the same public keys as R, then it's ok to extend to R' rather than R.
  246. When an onion router receives a CREATE cell, if it already has a
  247. circuit on the given connection with the given circID, it drops the
  248. cell. Otherwise, after receiving the CREATE cell, it completes the
  249. DH handshake, and replies with a CREATED cell. Upon receiving a
  250. CREATED cell, an onion router packs it payload into an EXTENDED relay
  251. cell (see section 5), and sends that cell up the circuit. Upon
  252. receiving the EXTENDED relay cell, the OP can retrieve g^y.
  253. (As an optimization, OR implementations may delay processing onions
  254. until a break in traffic allows time to do so without harming
  255. network latency too greatly.)
  256. 4.4. Tearing down circuits
  257. Circuits are torn down when an unrecoverable error occurs along
  258. the circuit, or when all streams on a circuit are closed and the
  259. circuit's intended lifetime is over. Circuits may be torn down
  260. either completely or hop-by-hop.
  261. To tear down a circuit completely, an OR or OP sends a DESTROY
  262. cell to the adjacent nodes on that circuit, using the appropriate
  263. direction's circID.
  264. Upon receiving an outgoing DESTROY cell, an OR frees resources
  265. associated with the corresponding circuit. If it's not the end of
  266. the circuit, it sends a DESTROY cell for that circuit to the next OR
  267. in the circuit. If the node is the end of the circuit, then it tears
  268. down any associated edge connections (see section 5.1).
  269. After a DESTROY cell has been processed, an OR ignores all data or
  270. destroy cells for the corresponding circuit.
  271. (The rest of this section is not currently used; on errors, circuits
  272. are destroyed, not truncated.)
  273. To tear down part of a circuit, the OP may send a RELAY_TRUNCATE cell
  274. signaling a given OR (Stream ID zero). That OR sends a DESTROY
  275. cell to the next node in the circuit, and replies to the OP with a
  276. RELAY_TRUNCATED cell.
  277. When an unrecoverable error occurs along one connection in a
  278. circuit, the nodes on either side of the connection should, if they
  279. are able, act as follows: the node closer to the OP should send a
  280. RELAY_TRUNCATED cell towards the OP; the node farther from the OP
  281. should send a DESTROY cell down the circuit.
  282. 4.5. Routing relay cells
  283. When an OR receives a RELAY cell, it checks the cell's circID and
  284. determines whether it has a corresponding circuit along that
  285. connection. If not, the OR drops the RELAY cell.
  286. Otherwise, if the OR is not at the OP edge of the circuit (that is,
  287. either an 'exit node' or a non-edge node), it de/encrypts the payload
  288. with AES/CTR, as follows:
  289. 'Forward' relay cell (same direction as CREATE):
  290. Use Kf as key; decrypt.
  291. 'Back' relay cell (opposite direction from CREATE):
  292. Use Kb as key; encrypt.
  293. The OR then decides whether it recognizes the relay cell, by
  294. inspecting the payload as described in section 5.1 below. If the OR
  295. recognizes the cell, it processes the contents of the relay cell.
  296. Otherwise, it passes the decrypted relay cell along the circuit if
  297. the circuit continues. If the OR at the end of the circuit
  298. encounters an unrecognized relay cell, an error has occurred: the OR
  299. sends a DESTROY cell to tear down the circuit.
  300. When a relay cell arrives at an OP, the OP decrypts the payload
  301. with AES/CTR as follows:
  302. OP receives data cell:
  303. For I=N...1,
  304. Decrypt with Kb_I. If the payload is recognized (see
  305. section 5.1), then stop and process the payload.
  306. For more information, see section 5 below.
  307. 5. Application connections and stream management
  308. 5.1. Relay cells
  309. Within a circuit, the OP and the exit node use the contents of
  310. RELAY packets to tunnel end-to-end commands and TCP connections
  311. ("Streams") across circuits. End-to-end commands can be initiated
  312. by either edge; streams are initiated by the OP.
  313. The payload of each unencrypted RELAY cell consists of:
  314. Relay command [1 byte]
  315. 'Recognized' [2 bytes]
  316. StreamID [2 bytes]
  317. Digest [4 bytes]
  318. Length [2 bytes]
  319. Data [498 bytes]
  320. The relay commands are:
  321. 1 -- RELAY_BEGIN
  322. 2 -- RELAY_DATA
  323. 3 -- RELAY_END
  324. 4 -- RELAY_CONNECTED
  325. 5 -- RELAY_SENDME
  326. 6 -- RELAY_EXTEND
  327. 7 -- RELAY_EXTENDED
  328. 8 -- RELAY_TRUNCATE
  329. 9 -- RELAY_TRUNCATED
  330. 10 -- RELAY_DROP
  331. 11 -- RELAY_RESOLVE
  332. 12 -- RELAY_RESOLVED
  333. The 'Recognized' field in any unencrypted relay payload is always
  334. set to zero; the 'digest' field is computed as the first four bytes
  335. of the running SHA-1 digest of all the bytes that have travelled
  336. over this circuit, seeded from Df or Db respectively (obtained in
  337. section 4.2 above), and including this RELAY cell's entire payload
  338. (taken with the digest field set to zero).
  339. When the 'recognized' field of a RELAY cell is zero, and the digest
  340. is correct, the cell is considered "recognized" for the purposes of
  341. decryption (see section 4.5 above).
  342. All RELAY cells pertaining to the same tunneled stream have the
  343. same stream ID. StreamIDs are chosen randomly by the OP. RELAY
  344. cells that affect the entire circuit rather than a particular
  345. stream use a StreamID of zero.
  346. The 'Length' field of a relay cell contains the number of bytes in
  347. the relay payload which contain real payload data. The remainder of
  348. the payload is padded with NUL bytes.
  349. 5.2. Opening streams and transferring data
  350. To open a new anonymized TCP connection, the OP chooses an open
  351. circuit to an exit that may be able to connect to the destination
  352. address, selects an arbitrary StreamID not yet used on that circuit,
  353. and constructs a RELAY_BEGIN cell with a payload encoding the address
  354. and port of the destination host. The payload format is:
  355. ADDRESS | ':' | PORT | [00]
  356. where ADDRESS can be a DNS hostname, or an IPv4 address in
  357. dotted-quad format, or an IPv6 address surrounded by square brackets;
  358. and where PORT is encoded in decimal.
  359. [What is the [00] for? -NM]
  360. [It's so the payload is easy to parse out with string funcs -RD]
  361. Upon receiving this cell, the exit node resolves the address as
  362. necessary, and opens a new TCP connection to the target port. If the
  363. address cannot be resolved, or a connection can't be established, the
  364. exit node replies with a RELAY_END cell. (See 5.4 below.)
  365. Otherwise, the exit node replies with a RELAY_CONNECTED cell, whose
  366. payload is in one of the following formats:
  367. The IPv4 address to which the connection was made [4 octets]
  368. A number of seconds (TTL) for which the address may be cached [4 octets]
  369. or
  370. Four zero-valued octets [4 octets]
  371. An address type (6) [1 octet]
  372. The IPv6 address to which the connection was made [16 octets]
  373. A number of seconds (TTL) for which the address may be cached [4 octets]
  374. [XXXX Versions of Tor before 0.1.1.6 ignore and do not generate the TTL
  375. field. No version of Tor currently generates the IPv6 format.]
  376. The OP waits for a RELAY_CONNECTED cell before sending any data.
  377. Once a connection has been established, the OP and exit node
  378. package stream data in RELAY_DATA cells, and upon receiving such
  379. cells, echo their contents to the corresponding TCP stream.
  380. RELAY_DATA cells sent to unrecognized streams are dropped.
  381. Relay RELAY_DROP cells are long-range dummies; upon receiving such
  382. a cell, the OR or OP must drop it.
  383. 5.3. Closing streams
  384. When an anonymized TCP connection is closed, or an edge node
  385. encounters error on any stream, it sends a 'RELAY_END' cell along the
  386. circuit (if possible) and closes the TCP connection immediately. If
  387. an edge node receives a 'RELAY_END' cell for any stream, it closes
  388. the TCP connection completely, and sends nothing more along the
  389. circuit for that stream.
  390. The payload of a RELAY_END cell begins with a single 'reason' byte to
  391. describe why the stream is closing, plus optional data (depending on
  392. the reason.) The values are:
  393. 1 -- REASON_MISC (catch-all for unlisted reasons)
  394. 2 -- REASON_RESOLVEFAILED (couldn't look up hostname)
  395. 3 -- REASON_CONNECTREFUSED (remote host refused connection) [*]
  396. 4 -- REASON_EXITPOLICY (OR refuses to connect to host or port)
  397. 5 -- REASON_DESTROY (Circuit is being destroyed)
  398. 6 -- REASON_DONE (Anonymized TCP connection was closed)
  399. 7 -- REASON_TIMEOUT (Connection timed out, or OR timed out
  400. while connecting)
  401. 8 -- (unallocated) [**]
  402. 9 -- REASON_HIBERNATING (OR is temporarily hibernating)
  403. 10 -- REASON_INTERNAL (Internal error at the OR)
  404. 11 -- REASON_RESOURCELIMIT (OR has no resources to fulfill request)
  405. 12 -- REASON_CONNRESET (Connection was unexpectedly reset)
  406. 13 -- REASON_TORPROTOCOL (Sent when closing connection because of
  407. Tor protocol violations.)
  408. (With REASON_EXITPOLICY, the 4-byte IPv4 address or 16-byte IPv6 address
  409. forms the optional data; no other reason currently has extra data.
  410. As of 0.1.1.6, the body also contains a 4-byte TTL.)
  411. OPs and ORs MUST accept reasons not on the above list, since future
  412. versions of Tor may provide more fine-grained reasons.
  413. [*] Older versions of Tor also send this reason when connections are
  414. reset.
  415. [**] Due to a bug in versions of Tor through 0095, error reason 8 must
  416. remain allocated until that version is obsolete.
  417. --- [The rest of this section describes unimplemented functionality.]
  418. Because TCP connections can be half-open, we follow an equivalent
  419. to TCP's FIN/FIN-ACK/ACK protocol to close streams.
  420. An exit connection can have a TCP stream in one of three states:
  421. 'OPEN', 'DONE_PACKAGING', and 'DONE_DELIVERING'. For the purposes
  422. of modeling transitions, we treat 'CLOSED' as a fourth state,
  423. although connections in this state are not, in fact, tracked by the
  424. onion router.
  425. A stream begins in the 'OPEN' state. Upon receiving a 'FIN' from
  426. the corresponding TCP connection, the edge node sends a 'RELAY_FIN'
  427. cell along the circuit and changes its state to 'DONE_PACKAGING'.
  428. Upon receiving a 'RELAY_FIN' cell, an edge node sends a 'FIN' to
  429. the corresponding TCP connection (e.g., by calling
  430. shutdown(SHUT_WR)) and changing its state to 'DONE_DELIVERING'.
  431. When a stream in already in 'DONE_DELIVERING' receives a 'FIN', it
  432. also sends a 'RELAY_FIN' along the circuit, and changes its state
  433. to 'CLOSED'. When a stream already in 'DONE_PACKAGING' receives a
  434. 'RELAY_FIN' cell, it sends a 'FIN' and changes its state to
  435. 'CLOSED'.
  436. If an edge node encounters an error on any stream, it sends a
  437. 'RELAY_END' cell (if possible) and closes the stream immediately.
  438. 5.4. Remote hostname lookup
  439. To find the address associated with a hostname, the OP sends a
  440. RELAY_RESOLVE cell containing the hostname to be resolved. (For a reverse
  441. lookup, the OP sends a RELAY_RESOLVE cell containing an in-addr.arpa
  442. address.) The OR replies with a RELAY_RESOLVED cell containing a status
  443. byte, and any number of answers. Each answer is of the form:
  444. Type (1 octet)
  445. Length (1 octet)
  446. Value (variable-width)
  447. TTL (4 octets)
  448. "Length" is the length of the Value field.
  449. "Type" is one of:
  450. 0x00 -- Hostname
  451. 0x04 -- IPv4 address
  452. 0x06 -- IPv6 address
  453. 0xF0 -- Error, transient
  454. 0xF1 -- Error, nontransient
  455. If any answer has a type of 'Error', then no other answer may be given.
  456. The RELAY_RESOLVE cell must use a nonzero, distinct streamID; the
  457. corresponding RELAY_RESOLVED cell must use the same streamID. No stream
  458. is actually created by the OR when resolving the name.
  459. 6. Flow control
  460. 6.1. Link throttling
  461. Each node should do appropriate bandwidth throttling to keep its
  462. user happy.
  463. Communicants rely on TCP's default flow control to push back when they
  464. stop reading.
  465. 6.2. Link padding
  466. Currently nodes are not required to do any sort of link padding or
  467. dummy traffic. Because strong attacks exist even with link padding,
  468. and because link padding greatly increases the bandwidth requirements
  469. for running a node, we plan to leave out link padding until this
  470. tradeoff is better understood.
  471. 6.3. Circuit-level flow control
  472. To control a circuit's bandwidth usage, each OR keeps track of
  473. two 'windows', consisting of how many RELAY_DATA cells it is
  474. allowed to package for transmission, and how many RELAY_DATA cells
  475. it is willing to deliver to streams outside the network.
  476. Each 'window' value is initially set to 1000 data cells
  477. in each direction (cells that are not data cells do not affect
  478. the window). When an OR is willing to deliver more cells, it sends a
  479. RELAY_SENDME cell towards the OP, with Stream ID zero. When an OR
  480. receives a RELAY_SENDME cell with stream ID zero, it increments its
  481. packaging window.
  482. Each of these cells increments the corresponding window by 100.
  483. The OP behaves identically, except that it must track a packaging
  484. window and a delivery window for every OR in the circuit.
  485. An OR or OP sends cells to increment its delivery window when the
  486. corresponding window value falls under some threshold (900).
  487. If a packaging window reaches 0, the OR or OP stops reading from
  488. TCP connections for all streams on the corresponding circuit, and
  489. sends no more RELAY_DATA cells until receiving a RELAY_SENDME cell.
  490. [this stuff is badly worded; copy in the tor-design section -RD]
  491. 6.4. Stream-level flow control
  492. Edge nodes use RELAY_SENDME cells to implement end-to-end flow
  493. control for individual connections across circuits. Similarly to
  494. circuit-level flow control, edge nodes begin with a window of cells
  495. (500) per stream, and increment the window by a fixed value (50)
  496. upon receiving a RELAY_SENDME cell. Edge nodes initiate RELAY_SENDME
  497. cells when both a) the window is <= 450, and b) there are less than
  498. ten cell payloads remaining to be flushed at that edge.
  499. 7. Directories and routers
  500. 7.1. Extensible information format
  501. Router descriptors and directories both obey the following lightweight
  502. extensible information format.
  503. The highest level object is a Document, which consists of one or more Items.
  504. Every Item begins with a KeywordLine, followed by one or more Objects. A
  505. KeywordLine begins with a Keyword, optionally followed by a space and more
  506. non-newline characters, and ends with a newline. A Keyword is a sequence of
  507. one or more characters in the set [A-Za-z0-9-]. An Object is a block of
  508. encoded data in pseudo-Open-PGP-style armor. (cf. RFC 2440)
  509. More formally:
  510. Document ::= (Item | NL)+
  511. Item ::= KeywordLine Object*
  512. KeywordLine ::= Keyword NL | Keyword SP ArgumentsChar+ NL
  513. Keyword = KeywordChar+
  514. KeywordChar ::= 'A' ... 'Z' | 'a' ... 'z' | '0' ... '9' | '-'
  515. ArgumentChar ::= any printing ASCII character except NL.
  516. Object ::= BeginLine Base-64-encoded-data EndLine
  517. BeginLine ::= "-----BEGIN " Keyword "-----" NL
  518. EndLine ::= "-----END " Keyword "-----" NL
  519. The BeginLine and EndLine of an Object must use the same keyword.
  520. When interpreting a Document, software MUST reject any document containing a
  521. KeywordLine that starts with a keyword it doesn't recognize.
  522. The "opt" keyword is reserved for non-critical future extensions. All
  523. implementations MUST ignore any item of the form "opt keyword ....." when
  524. they would not recognize "keyword ....."; and MUST treat "opt keyword ....."
  525. as synonymous with "keyword ......" when keyword is recognized.
  526. 7.2. Router descriptor format.
  527. Every router descriptor MUST start with a "router" Item; MUST end with a
  528. "router-signature" Item and an extra NL; and MUST contain exactly one
  529. instance of each of the following Items: "published" "onion-key" "link-key"
  530. "signing-key" "bandwidth". Additionally, a router descriptor MAY contain any
  531. number of "accept", "reject", "fingerprint", "uptime", and "opt" Items.
  532. Other than "router" and "router-signature", the items may appear in any
  533. order.
  534. The items' formats are as follows:
  535. "router" nickname address (ORPort SocksPort DirPort)?
  536. Indicates the beginning of a router descriptor. "address" must be an
  537. IPv4 address in dotted-quad format. The Port values will soon be
  538. deprecated; using them here is equivalent to using them in a "ports"
  539. item.
  540. "ports" ORPort SocksPort DirPort
  541. Indicates the TCP ports at which this OR exposes functionality.
  542. ORPort is a port at which this OR accepts TLS connections for the main
  543. OR protocol; SocksPort is the port at which this OR accepts SOCKS
  544. connections; and DirPort is the port at which this OR accepts
  545. directory-related HTTP connections. If any port is not supported, the
  546. value 0 is given instead of a port number.
  547. "bandwidth" bandwidth-avg bandwidth-burst bandwidth-observed
  548. Estimated bandwidth for this router, in bytes per second. The
  549. "average" bandwidth is the volume per second that the OR is willing
  550. to sustain over long periods; the "burst" bandwidth is the volume
  551. that the OR is willing to sustain in very short intervals. The
  552. "observed" value is an estimate of the capacity this server can
  553. handle. The server remembers the max bandwidth sustained output
  554. over any ten second period in the past day, and another sustained
  555. input. The "observed" value is the lesser of these two numbers.
  556. "platform" string
  557. A human-readable string describing the system on which this OR is
  558. running. This MAY include the operating system, and SHOULD include
  559. the name and version of the software implementing the Tor protocol.
  560. "published" YYYY-MM-DD HH:MM:SS
  561. The time, in GMT, when this descriptor was generated.
  562. "fingerprint"
  563. A fingerprint (20 byte SHA1 hash of asn1 encoded public key, encoded
  564. in hex, with spaces after every 4 characters) for this router's
  565. identity key.
  566. [We didn't start parsing this line until Tor 0.1.0.6-rc; it should
  567. be marked with "opt" until earlier versions of Tor are obsolete.]
  568. "hibernating" 0|1
  569. If the value is 1, then the Tor server was hibernating when the
  570. descriptor was published, and shouldn't be used to build circuits.
  571. [We didn't start parsing this line until Tor 0.1.0.6-rc; it should
  572. be marked with "opt" until earlier versions of Tor are obsolete.]
  573. "uptime"
  574. The number of seconds that this OR process has been running.
  575. "onion-key" NL a public key in PEM format
  576. This key is used to encrypt EXTEND cells for this OR. The key MUST
  577. be accepted for at least XXXX hours after any new key is published in
  578. a subsequent descriptor.
  579. "signing-key" NL a public key in PEM format
  580. The OR's long-term identity key.
  581. "accept" exitpattern
  582. "reject" exitpattern
  583. These lines, in order, describe the rules that an OR follows when
  584. deciding whether to allow a new stream to a given address. The
  585. 'exitpattern' syntax is described below.
  586. "router-signature" NL Signature NL
  587. The "SIGNATURE" object contains a signature of the PKCS1-padded SHA1
  588. hash of the entire router descriptor, taken from the beginning of the
  589. "router" line, through the newline after the "router-signature" line.
  590. The router descriptor is invalid unless the signature is performed
  591. with the router's identity key.
  592. "contact" info NL
  593. Describes a way to contact the server's administrator, preferably
  594. including an email address and a PGP key fingerprint.
  595. "family" names NL
  596. 'Names' is a space-separated list of server nicknames. If two ORs
  597. list one another in their "family" entries, then OPs should treat
  598. them as a single OR for the purpose of path selection.
  599. For example, if node A's descriptor contains "family B", and node B's
  600. descriptor contains "family A", then node A and node B should never
  601. be used on the same circuit.
  602. "read-history" YYYY-MM-DD HH:MM:SS (NSEC s) NUM,NUM,NUM,NUM,NUM... NL
  603. "write-history" YYYY-MM-DD HH:MM:SS (NSEC s) NUM,NUM,NUM,NUM,NUM... NL
  604. Declare how much bandwidth the OR has used recently. Usage is divided
  605. into intervals of NSEC seconds. The YYYY-MM-DD HH:MM:SS field defines
  606. the end of the most recent interval. The numbers are the number of
  607. bytes used in the most recent intervals, ordered from oldest to newest.
  608. [We didn't start parsing these lines until Tor 0.1.0.6-rc; they should
  609. be marked with "opt" until earlier versions of Tor are obsolete.]
  610. nickname ::= between 1 and 19 alphanumeric characters, case-insensitive.
  611. exitpattern ::= addrspec ":" portspec
  612. portspec ::= "*" | port | port "-" port
  613. port ::= an integer between 1 and 65535, inclusive.
  614. addrspec ::= "*" | ip4spec | ip6spec
  615. ipv4spec ::= ip4 | ip4 "/" num_ip4_bits | ip4 "/" ip4mask
  616. ip4 ::= an IPv4 address in dotted-quad format
  617. ip4mask ::= an IPv4 mask in dotted-quad format
  618. num_ip4_bits ::= an integer between 0 and 32
  619. ip6spec ::= ip6 | ip6 "/" num_ip6_bits
  620. ip6 ::= an IPv6 address, surrounded by square brackets.
  621. num_ip6_bits ::= an integer between 0 and 128
  622. Ports are required; if they are not included in the router
  623. line, they must appear in the "ports" lines.
  624. 7.3. Directory format
  625. A Directory begins with a "signed-directory" item, followed by one each of
  626. the following, in any order: "recommended-software", "published",
  627. "router-status", "dir-signing-key". It may include any number of "opt"
  628. items. After these items, a directory includes any number of router
  629. descriptors, and a single "directory-signature" item.
  630. "signed-directory"
  631. Indicates the start of a directory.
  632. "published" YYYY-MM-DD HH:MM:SS
  633. The time at which this directory was generated and signed, in GMT.
  634. "dir-signing-key"
  635. The key used to sign this directory; see "signing-key" for format.
  636. "recommended-software" comma-separated-version-list
  637. A list of which versions of which implementations are currently
  638. believed to be secure and compatible with the network.
  639. "running-routers" space-separated-list
  640. A description of which routers are currently believed to be up or
  641. down. Every entry consists of an optional "!", followed by either an
  642. OR's nickname, or "$" followed by a hexadecimal encoding of the hash
  643. of an OR's identity key. If the "!" is included, the router is
  644. believed not to be running; otherwise, it is believed to be running.
  645. If a router's nickname is given, exactly one router of that nickname
  646. will appear in the directory, and that router is "approved" by the
  647. directory server. If a hashed identity key is given, that OR is not
  648. "approved". [XXXX The 'running-routers' line is only provided for
  649. backward compatibility. New code should parse 'router-status'
  650. instead.]
  651. "router-status" space-separated-list
  652. A description of which routers are currently believed to be up or
  653. down, and which are verified or unverified. Contains one entry for
  654. every router that the directory server knows. Each entry is of the
  655. format:
  656. !name=$digest [Verified router, currently not live.]
  657. name=$digest [Verified router, currently live.]
  658. !$digest [Unverified router, currently not live.]
  659. or $digest [Unverified router, currently live.]
  660. (where 'name' is the router's nickname and 'digest' is a hexadecimal
  661. encoding of the hash of the routers' identity key).
  662. When parsing this line, clients should only mark a router as
  663. 'verified' if its nickname AND digest match the one provided.
  664. "directory-signature" nickname-of-dirserver NL Signature
  665. The signature is computed by computing the SHA-1 hash of the
  666. directory, from the characters "signed-directory", through the newline
  667. after "directory-signature". This digest is then padded with PKCS.1,
  668. and signed with the directory server's signing key.
  669. If software encounters an unrecognized keyword in a single router descriptor,
  670. it MUST reject only that router descriptor, and continue using the
  671. others. Because this mechanism is used to add 'critical' extensions to
  672. future versions of the router descriptor format, implementation should treat
  673. it as a normal occurrence and not, for example, report it to the user as an
  674. error. [Versions of Tor prior to 0.1.1 did this.]
  675. If software encounters an unrecognized keyword in the directory header,
  676. it SHOULD reject the entire directory.
  677. 7.4. Network-status descriptor
  678. A "network-status" (a.k.a "running-routers") document is a truncated
  679. directory that contains only the current status of a list of nodes, not
  680. their actual descriptors. It contains exactly one of each of the following
  681. entries.
  682. "network-status"
  683. Must appear first.
  684. "published" YYYY-MM-DD HH:MM:SS
  685. (see 7.3 above)
  686. "router-status" list
  687. (see 7.3 above)
  688. "directory-signature" NL signature
  689. (see 7.3 above)
  690. 7.5. Behavior of a directory server
  691. lists nodes that are connected currently
  692. speaks HTTP on a socket, spits out directory on request
  693. Directory servers listen on a certain port (the DirPort), and speak a
  694. limited version of HTTP 1.0. Clients send either GET or POST commands.
  695. The basic interactions are:
  696. "%s %s HTTP/1.0\r\nContent-Length: %lu\r\nHost: %s\r\n\r\n",
  697. command, url, content-length, host.
  698. Get "/tor/" to fetch a full directory.
  699. Get "/tor/dir.z" to fetch a compressed full directory.
  700. Get "/tor/running-routers" to fetch a network-status descriptor.
  701. Post "/tor/" to post a server descriptor, with the body of the
  702. request containing the descriptor.
  703. "host" is used to specify the address:port of the dirserver, so
  704. the request can survive going through HTTP proxies.
  705. A.1. Differences between spec and implementation
  706. - The current specification requires all ORs to have IPv4 addresses, but
  707. allows servers to exit and resolve to IPv6 addresses, and to declare IPv6
  708. addresses in their exit policies. The current codebase has no IPv6
  709. support at all.