input.c 22 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748
  1. /*
  2. * net/dccp/input.c
  3. *
  4. * An implementation of the DCCP protocol
  5. * Arnaldo Carvalho de Melo <acme@conectiva.com.br>
  6. *
  7. * This program is free software; you can redistribute it and/or
  8. * modify it under the terms of the GNU General Public License
  9. * as published by the Free Software Foundation; either version
  10. * 2 of the License, or (at your option) any later version.
  11. */
  12. #include <linux/dccp.h>
  13. #include <linux/skbuff.h>
  14. #include <linux/slab.h>
  15. #include <net/sock.h>
  16. #include "ackvec.h"
  17. #include "ccid.h"
  18. #include "dccp.h"
  19. /* rate-limit for syncs in reply to sequence-invalid packets; RFC 4340, 7.5.4 */
  20. int sysctl_dccp_sync_ratelimit __read_mostly = HZ / 8;
  21. static void dccp_enqueue_skb(struct sock *sk, struct sk_buff *skb)
  22. {
  23. __skb_pull(skb, dccp_hdr(skb)->dccph_doff * 4);
  24. __skb_queue_tail(&sk->sk_receive_queue, skb);
  25. skb_set_owner_r(skb, sk);
  26. sk->sk_data_ready(sk, 0);
  27. }
  28. static void dccp_fin(struct sock *sk, struct sk_buff *skb)
  29. {
  30. /*
  31. * On receiving Close/CloseReq, both RD/WR shutdown are performed.
  32. * RFC 4340, 8.3 says that we MAY send further Data/DataAcks after
  33. * receiving the closing segment, but there is no guarantee that such
  34. * data will be processed at all.
  35. */
  36. sk->sk_shutdown = SHUTDOWN_MASK;
  37. sock_set_flag(sk, SOCK_DONE);
  38. dccp_enqueue_skb(sk, skb);
  39. }
  40. static int dccp_rcv_close(struct sock *sk, struct sk_buff *skb)
  41. {
  42. int queued = 0;
  43. switch (sk->sk_state) {
  44. /*
  45. * We ignore Close when received in one of the following states:
  46. * - CLOSED (may be a late or duplicate packet)
  47. * - PASSIVE_CLOSEREQ (the peer has sent a CloseReq earlier)
  48. * - RESPOND (already handled by dccp_check_req)
  49. */
  50. case DCCP_CLOSING:
  51. /*
  52. * Simultaneous-close: receiving a Close after sending one. This
  53. * can happen if both client and server perform active-close and
  54. * will result in an endless ping-pong of crossing and retrans-
  55. * mitted Close packets, which only terminates when one of the
  56. * nodes times out (min. 64 seconds). Quicker convergence can be
  57. * achieved when one of the nodes acts as tie-breaker.
  58. * This is ok as both ends are done with data transfer and each
  59. * end is just waiting for the other to acknowledge termination.
  60. */
  61. if (dccp_sk(sk)->dccps_role != DCCP_ROLE_CLIENT)
  62. break;
  63. /* fall through */
  64. case DCCP_REQUESTING:
  65. case DCCP_ACTIVE_CLOSEREQ:
  66. dccp_send_reset(sk, DCCP_RESET_CODE_CLOSED);
  67. dccp_done(sk);
  68. break;
  69. case DCCP_OPEN:
  70. case DCCP_PARTOPEN:
  71. /* Give waiting application a chance to read pending data */
  72. queued = 1;
  73. dccp_fin(sk, skb);
  74. dccp_set_state(sk, DCCP_PASSIVE_CLOSE);
  75. /* fall through */
  76. case DCCP_PASSIVE_CLOSE:
  77. /*
  78. * Retransmitted Close: we have already enqueued the first one.
  79. */
  80. sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_HUP);
  81. }
  82. return queued;
  83. }
  84. static int dccp_rcv_closereq(struct sock *sk, struct sk_buff *skb)
  85. {
  86. int queued = 0;
  87. /*
  88. * Step 7: Check for unexpected packet types
  89. * If (S.is_server and P.type == CloseReq)
  90. * Send Sync packet acknowledging P.seqno
  91. * Drop packet and return
  92. */
  93. if (dccp_sk(sk)->dccps_role != DCCP_ROLE_CLIENT) {
  94. dccp_send_sync(sk, DCCP_SKB_CB(skb)->dccpd_seq, DCCP_PKT_SYNC);
  95. return queued;
  96. }
  97. /* Step 13: process relevant Client states < CLOSEREQ */
  98. switch (sk->sk_state) {
  99. case DCCP_REQUESTING:
  100. dccp_send_close(sk, 0);
  101. dccp_set_state(sk, DCCP_CLOSING);
  102. break;
  103. case DCCP_OPEN:
  104. case DCCP_PARTOPEN:
  105. /* Give waiting application a chance to read pending data */
  106. queued = 1;
  107. dccp_fin(sk, skb);
  108. dccp_set_state(sk, DCCP_PASSIVE_CLOSEREQ);
  109. /* fall through */
  110. case DCCP_PASSIVE_CLOSEREQ:
  111. sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_HUP);
  112. }
  113. return queued;
  114. }
  115. static u16 dccp_reset_code_convert(const u8 code)
  116. {
  117. const u16 error_code[] = {
  118. [DCCP_RESET_CODE_CLOSED] = 0, /* normal termination */
  119. [DCCP_RESET_CODE_UNSPECIFIED] = 0, /* nothing known */
  120. [DCCP_RESET_CODE_ABORTED] = ECONNRESET,
  121. [DCCP_RESET_CODE_NO_CONNECTION] = ECONNREFUSED,
  122. [DCCP_RESET_CODE_CONNECTION_REFUSED] = ECONNREFUSED,
  123. [DCCP_RESET_CODE_TOO_BUSY] = EUSERS,
  124. [DCCP_RESET_CODE_AGGRESSION_PENALTY] = EDQUOT,
  125. [DCCP_RESET_CODE_PACKET_ERROR] = ENOMSG,
  126. [DCCP_RESET_CODE_BAD_INIT_COOKIE] = EBADR,
  127. [DCCP_RESET_CODE_BAD_SERVICE_CODE] = EBADRQC,
  128. [DCCP_RESET_CODE_OPTION_ERROR] = EILSEQ,
  129. [DCCP_RESET_CODE_MANDATORY_ERROR] = EOPNOTSUPP,
  130. };
  131. return code >= DCCP_MAX_RESET_CODES ? 0 : error_code[code];
  132. }
  133. static void dccp_rcv_reset(struct sock *sk, struct sk_buff *skb)
  134. {
  135. u16 err = dccp_reset_code_convert(dccp_hdr_reset(skb)->dccph_reset_code);
  136. sk->sk_err = err;
  137. /* Queue the equivalent of TCP fin so that dccp_recvmsg exits the loop */
  138. dccp_fin(sk, skb);
  139. if (err && !sock_flag(sk, SOCK_DEAD))
  140. sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
  141. dccp_time_wait(sk, DCCP_TIME_WAIT, 0);
  142. }
  143. static void dccp_event_ack_recv(struct sock *sk, struct sk_buff *skb)
  144. {
  145. struct dccp_sock *dp = dccp_sk(sk);
  146. if (dp->dccps_hc_rx_ackvec != NULL)
  147. dccp_ackvec_check_rcv_ackno(dp->dccps_hc_rx_ackvec, sk,
  148. DCCP_SKB_CB(skb)->dccpd_ack_seq);
  149. }
  150. static void dccp_deliver_input_to_ccids(struct sock *sk, struct sk_buff *skb)
  151. {
  152. const struct dccp_sock *dp = dccp_sk(sk);
  153. /* Don't deliver to RX CCID when node has shut down read end. */
  154. if (!(sk->sk_shutdown & RCV_SHUTDOWN))
  155. ccid_hc_rx_packet_recv(dp->dccps_hc_rx_ccid, sk, skb);
  156. /*
  157. * Until the TX queue has been drained, we can not honour SHUT_WR, since
  158. * we need received feedback as input to adjust congestion control.
  159. */
  160. if (sk->sk_write_queue.qlen > 0 || !(sk->sk_shutdown & SEND_SHUTDOWN))
  161. ccid_hc_tx_packet_recv(dp->dccps_hc_tx_ccid, sk, skb);
  162. }
  163. static int dccp_check_seqno(struct sock *sk, struct sk_buff *skb)
  164. {
  165. const struct dccp_hdr *dh = dccp_hdr(skb);
  166. struct dccp_sock *dp = dccp_sk(sk);
  167. u64 lswl, lawl, seqno = DCCP_SKB_CB(skb)->dccpd_seq,
  168. ackno = DCCP_SKB_CB(skb)->dccpd_ack_seq;
  169. /*
  170. * Step 5: Prepare sequence numbers for Sync
  171. * If P.type == Sync or P.type == SyncAck,
  172. * If S.AWL <= P.ackno <= S.AWH and P.seqno >= S.SWL,
  173. * / * P is valid, so update sequence number variables
  174. * accordingly. After this update, P will pass the tests
  175. * in Step 6. A SyncAck is generated if necessary in
  176. * Step 15 * /
  177. * Update S.GSR, S.SWL, S.SWH
  178. * Otherwise,
  179. * Drop packet and return
  180. */
  181. if (dh->dccph_type == DCCP_PKT_SYNC ||
  182. dh->dccph_type == DCCP_PKT_SYNCACK) {
  183. if (between48(ackno, dp->dccps_awl, dp->dccps_awh) &&
  184. dccp_delta_seqno(dp->dccps_swl, seqno) >= 0)
  185. dccp_update_gsr(sk, seqno);
  186. else
  187. return -1;
  188. }
  189. /*
  190. * Step 6: Check sequence numbers
  191. * Let LSWL = S.SWL and LAWL = S.AWL
  192. * If P.type == CloseReq or P.type == Close or P.type == Reset,
  193. * LSWL := S.GSR + 1, LAWL := S.GAR
  194. * If LSWL <= P.seqno <= S.SWH
  195. * and (P.ackno does not exist or LAWL <= P.ackno <= S.AWH),
  196. * Update S.GSR, S.SWL, S.SWH
  197. * If P.type != Sync,
  198. * Update S.GAR
  199. */
  200. lswl = dp->dccps_swl;
  201. lawl = dp->dccps_awl;
  202. if (dh->dccph_type == DCCP_PKT_CLOSEREQ ||
  203. dh->dccph_type == DCCP_PKT_CLOSE ||
  204. dh->dccph_type == DCCP_PKT_RESET) {
  205. lswl = ADD48(dp->dccps_gsr, 1);
  206. lawl = dp->dccps_gar;
  207. }
  208. if (between48(seqno, lswl, dp->dccps_swh) &&
  209. (ackno == DCCP_PKT_WITHOUT_ACK_SEQ ||
  210. between48(ackno, lawl, dp->dccps_awh))) {
  211. dccp_update_gsr(sk, seqno);
  212. if (dh->dccph_type != DCCP_PKT_SYNC &&
  213. ackno != DCCP_PKT_WITHOUT_ACK_SEQ &&
  214. after48(ackno, dp->dccps_gar))
  215. dp->dccps_gar = ackno;
  216. } else {
  217. unsigned long now = jiffies;
  218. /*
  219. * Step 6: Check sequence numbers
  220. * Otherwise,
  221. * If P.type == Reset,
  222. * Send Sync packet acknowledging S.GSR
  223. * Otherwise,
  224. * Send Sync packet acknowledging P.seqno
  225. * Drop packet and return
  226. *
  227. * These Syncs are rate-limited as per RFC 4340, 7.5.4:
  228. * at most 1 / (dccp_sync_rate_limit * HZ) Syncs per second.
  229. */
  230. if (time_before(now, (dp->dccps_rate_last +
  231. sysctl_dccp_sync_ratelimit)))
  232. return 0;
  233. DCCP_WARN("Step 6 failed for %s packet, "
  234. "(LSWL(%llu) <= P.seqno(%llu) <= S.SWH(%llu)) and "
  235. "(P.ackno %s or LAWL(%llu) <= P.ackno(%llu) <= S.AWH(%llu), "
  236. "sending SYNC...\n", dccp_packet_name(dh->dccph_type),
  237. (unsigned long long) lswl, (unsigned long long) seqno,
  238. (unsigned long long) dp->dccps_swh,
  239. (ackno == DCCP_PKT_WITHOUT_ACK_SEQ) ? "doesn't exist"
  240. : "exists",
  241. (unsigned long long) lawl, (unsigned long long) ackno,
  242. (unsigned long long) dp->dccps_awh);
  243. dp->dccps_rate_last = now;
  244. if (dh->dccph_type == DCCP_PKT_RESET)
  245. seqno = dp->dccps_gsr;
  246. dccp_send_sync(sk, seqno, DCCP_PKT_SYNC);
  247. return -1;
  248. }
  249. return 0;
  250. }
  251. static int __dccp_rcv_established(struct sock *sk, struct sk_buff *skb,
  252. const struct dccp_hdr *dh, const unsigned len)
  253. {
  254. struct dccp_sock *dp = dccp_sk(sk);
  255. switch (dccp_hdr(skb)->dccph_type) {
  256. case DCCP_PKT_DATAACK:
  257. case DCCP_PKT_DATA:
  258. /*
  259. * FIXME: schedule DATA_DROPPED (RFC 4340, 11.7.2) if and when
  260. * - sk_shutdown == RCV_SHUTDOWN, use Code 1, "Not Listening"
  261. * - sk_receive_queue is full, use Code 2, "Receive Buffer"
  262. */
  263. dccp_enqueue_skb(sk, skb);
  264. return 0;
  265. case DCCP_PKT_ACK:
  266. goto discard;
  267. case DCCP_PKT_RESET:
  268. /*
  269. * Step 9: Process Reset
  270. * If P.type == Reset,
  271. * Tear down connection
  272. * S.state := TIMEWAIT
  273. * Set TIMEWAIT timer
  274. * Drop packet and return
  275. */
  276. dccp_rcv_reset(sk, skb);
  277. return 0;
  278. case DCCP_PKT_CLOSEREQ:
  279. if (dccp_rcv_closereq(sk, skb))
  280. return 0;
  281. goto discard;
  282. case DCCP_PKT_CLOSE:
  283. if (dccp_rcv_close(sk, skb))
  284. return 0;
  285. goto discard;
  286. case DCCP_PKT_REQUEST:
  287. /* Step 7
  288. * or (S.is_server and P.type == Response)
  289. * or (S.is_client and P.type == Request)
  290. * or (S.state >= OPEN and P.type == Request
  291. * and P.seqno >= S.OSR)
  292. * or (S.state >= OPEN and P.type == Response
  293. * and P.seqno >= S.OSR)
  294. * or (S.state == RESPOND and P.type == Data),
  295. * Send Sync packet acknowledging P.seqno
  296. * Drop packet and return
  297. */
  298. if (dp->dccps_role != DCCP_ROLE_LISTEN)
  299. goto send_sync;
  300. goto check_seq;
  301. case DCCP_PKT_RESPONSE:
  302. if (dp->dccps_role != DCCP_ROLE_CLIENT)
  303. goto send_sync;
  304. check_seq:
  305. if (dccp_delta_seqno(dp->dccps_osr,
  306. DCCP_SKB_CB(skb)->dccpd_seq) >= 0) {
  307. send_sync:
  308. dccp_send_sync(sk, DCCP_SKB_CB(skb)->dccpd_seq,
  309. DCCP_PKT_SYNC);
  310. }
  311. break;
  312. case DCCP_PKT_SYNC:
  313. dccp_send_sync(sk, DCCP_SKB_CB(skb)->dccpd_seq,
  314. DCCP_PKT_SYNCACK);
  315. /*
  316. * From RFC 4340, sec. 5.7
  317. *
  318. * As with DCCP-Ack packets, DCCP-Sync and DCCP-SyncAck packets
  319. * MAY have non-zero-length application data areas, whose
  320. * contents receivers MUST ignore.
  321. */
  322. goto discard;
  323. }
  324. DCCP_INC_STATS_BH(DCCP_MIB_INERRS);
  325. discard:
  326. __kfree_skb(skb);
  327. return 0;
  328. }
  329. int dccp_rcv_established(struct sock *sk, struct sk_buff *skb,
  330. const struct dccp_hdr *dh, const unsigned len)
  331. {
  332. struct dccp_sock *dp = dccp_sk(sk);
  333. if (dccp_check_seqno(sk, skb))
  334. goto discard;
  335. if (dccp_parse_options(sk, NULL, skb))
  336. return 1;
  337. if (DCCP_SKB_CB(skb)->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ)
  338. dccp_event_ack_recv(sk, skb);
  339. if (dp->dccps_hc_rx_ackvec != NULL &&
  340. dccp_ackvec_add(dp->dccps_hc_rx_ackvec, sk,
  341. DCCP_SKB_CB(skb)->dccpd_seq,
  342. DCCP_ACKVEC_STATE_RECEIVED))
  343. goto discard;
  344. dccp_deliver_input_to_ccids(sk, skb);
  345. return __dccp_rcv_established(sk, skb, dh, len);
  346. discard:
  347. __kfree_skb(skb);
  348. return 0;
  349. }
  350. EXPORT_SYMBOL_GPL(dccp_rcv_established);
  351. static int dccp_rcv_request_sent_state_process(struct sock *sk,
  352. struct sk_buff *skb,
  353. const struct dccp_hdr *dh,
  354. const unsigned len)
  355. {
  356. /*
  357. * Step 4: Prepare sequence numbers in REQUEST
  358. * If S.state == REQUEST,
  359. * If (P.type == Response or P.type == Reset)
  360. * and S.AWL <= P.ackno <= S.AWH,
  361. * / * Set sequence number variables corresponding to the
  362. * other endpoint, so P will pass the tests in Step 6 * /
  363. * Set S.GSR, S.ISR, S.SWL, S.SWH
  364. * / * Response processing continues in Step 10; Reset
  365. * processing continues in Step 9 * /
  366. */
  367. if (dh->dccph_type == DCCP_PKT_RESPONSE) {
  368. const struct inet_connection_sock *icsk = inet_csk(sk);
  369. struct dccp_sock *dp = dccp_sk(sk);
  370. long tstamp = dccp_timestamp();
  371. if (!between48(DCCP_SKB_CB(skb)->dccpd_ack_seq,
  372. dp->dccps_awl, dp->dccps_awh)) {
  373. dccp_pr_debug("invalid ackno: S.AWL=%llu, "
  374. "P.ackno=%llu, S.AWH=%llu\n",
  375. (unsigned long long)dp->dccps_awl,
  376. (unsigned long long)DCCP_SKB_CB(skb)->dccpd_ack_seq,
  377. (unsigned long long)dp->dccps_awh);
  378. goto out_invalid_packet;
  379. }
  380. /*
  381. * If option processing (Step 8) failed, return 1 here so that
  382. * dccp_v4_do_rcv() sends a Reset. The Reset code depends on
  383. * the option type and is set in dccp_parse_options().
  384. */
  385. if (dccp_parse_options(sk, NULL, skb))
  386. return 1;
  387. /* Obtain usec RTT sample from SYN exchange (used by TFRC). */
  388. if (likely(dp->dccps_options_received.dccpor_timestamp_echo))
  389. dp->dccps_syn_rtt = dccp_sample_rtt(sk, 10 * (tstamp -
  390. dp->dccps_options_received.dccpor_timestamp_echo));
  391. /* Stop the REQUEST timer */
  392. inet_csk_clear_xmit_timer(sk, ICSK_TIME_RETRANS);
  393. WARN_ON(sk->sk_send_head == NULL);
  394. kfree_skb(sk->sk_send_head);
  395. sk->sk_send_head = NULL;
  396. /*
  397. * Set ISR, GSR from packet. ISS was set in dccp_v{4,6}_connect
  398. * and GSS in dccp_transmit_skb(). Setting AWL/AWH and SWL/SWH
  399. * is done as part of activating the feature values below, since
  400. * these settings depend on the local/remote Sequence Window
  401. * features, which were undefined or not confirmed until now.
  402. */
  403. dp->dccps_gsr = dp->dccps_isr = DCCP_SKB_CB(skb)->dccpd_seq;
  404. dccp_sync_mss(sk, icsk->icsk_pmtu_cookie);
  405. /*
  406. * Step 10: Process REQUEST state (second part)
  407. * If S.state == REQUEST,
  408. * / * If we get here, P is a valid Response from the
  409. * server (see Step 4), and we should move to
  410. * PARTOPEN state. PARTOPEN means send an Ack,
  411. * don't send Data packets, retransmit Acks
  412. * periodically, and always include any Init Cookie
  413. * from the Response * /
  414. * S.state := PARTOPEN
  415. * Set PARTOPEN timer
  416. * Continue with S.state == PARTOPEN
  417. * / * Step 12 will send the Ack completing the
  418. * three-way handshake * /
  419. */
  420. dccp_set_state(sk, DCCP_PARTOPEN);
  421. /*
  422. * If feature negotiation was successful, activate features now;
  423. * an activation failure means that this host could not activate
  424. * one ore more features (e.g. insufficient memory), which would
  425. * leave at least one feature in an undefined state.
  426. */
  427. if (dccp_feat_activate_values(sk, &dp->dccps_featneg))
  428. goto unable_to_proceed;
  429. /* Make sure socket is routed, for correct metrics. */
  430. icsk->icsk_af_ops->rebuild_header(sk);
  431. if (!sock_flag(sk, SOCK_DEAD)) {
  432. sk->sk_state_change(sk);
  433. sk_wake_async(sk, SOCK_WAKE_IO, POLL_OUT);
  434. }
  435. if (sk->sk_write_pending || icsk->icsk_ack.pingpong ||
  436. icsk->icsk_accept_queue.rskq_defer_accept) {
  437. /* Save one ACK. Data will be ready after
  438. * several ticks, if write_pending is set.
  439. *
  440. * It may be deleted, but with this feature tcpdumps
  441. * look so _wonderfully_ clever, that I was not able
  442. * to stand against the temptation 8) --ANK
  443. */
  444. /*
  445. * OK, in DCCP we can as well do a similar trick, its
  446. * even in the draft, but there is no need for us to
  447. * schedule an ack here, as dccp_sendmsg does this for
  448. * us, also stated in the draft. -acme
  449. */
  450. __kfree_skb(skb);
  451. return 0;
  452. }
  453. dccp_send_ack(sk);
  454. return -1;
  455. }
  456. out_invalid_packet:
  457. /* dccp_v4_do_rcv will send a reset */
  458. DCCP_SKB_CB(skb)->dccpd_reset_code = DCCP_RESET_CODE_PACKET_ERROR;
  459. return 1;
  460. unable_to_proceed:
  461. DCCP_SKB_CB(skb)->dccpd_reset_code = DCCP_RESET_CODE_ABORTED;
  462. /*
  463. * We mark this socket as no longer usable, so that the loop in
  464. * dccp_sendmsg() terminates and the application gets notified.
  465. */
  466. dccp_set_state(sk, DCCP_CLOSED);
  467. sk->sk_err = ECOMM;
  468. return 1;
  469. }
  470. static int dccp_rcv_respond_partopen_state_process(struct sock *sk,
  471. struct sk_buff *skb,
  472. const struct dccp_hdr *dh,
  473. const unsigned len)
  474. {
  475. struct dccp_sock *dp = dccp_sk(sk);
  476. u32 sample = dp->dccps_options_received.dccpor_timestamp_echo;
  477. int queued = 0;
  478. switch (dh->dccph_type) {
  479. case DCCP_PKT_RESET:
  480. inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK);
  481. break;
  482. case DCCP_PKT_DATA:
  483. if (sk->sk_state == DCCP_RESPOND)
  484. break;
  485. case DCCP_PKT_DATAACK:
  486. case DCCP_PKT_ACK:
  487. /*
  488. * FIXME: we should be reseting the PARTOPEN (DELACK) timer
  489. * here but only if we haven't used the DELACK timer for
  490. * something else, like sending a delayed ack for a TIMESTAMP
  491. * echo, etc, for now were not clearing it, sending an extra
  492. * ACK when there is nothing else to do in DELACK is not a big
  493. * deal after all.
  494. */
  495. /* Stop the PARTOPEN timer */
  496. if (sk->sk_state == DCCP_PARTOPEN)
  497. inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK);
  498. /* Obtain usec RTT sample from SYN exchange (used by TFRC). */
  499. if (likely(sample)) {
  500. long delta = dccp_timestamp() - sample;
  501. dp->dccps_syn_rtt = dccp_sample_rtt(sk, 10 * delta);
  502. }
  503. dp->dccps_osr = DCCP_SKB_CB(skb)->dccpd_seq;
  504. dccp_set_state(sk, DCCP_OPEN);
  505. if (dh->dccph_type == DCCP_PKT_DATAACK ||
  506. dh->dccph_type == DCCP_PKT_DATA) {
  507. __dccp_rcv_established(sk, skb, dh, len);
  508. queued = 1; /* packet was queued
  509. (by __dccp_rcv_established) */
  510. }
  511. break;
  512. }
  513. return queued;
  514. }
  515. int dccp_rcv_state_process(struct sock *sk, struct sk_buff *skb,
  516. struct dccp_hdr *dh, unsigned len)
  517. {
  518. struct dccp_sock *dp = dccp_sk(sk);
  519. struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
  520. const int old_state = sk->sk_state;
  521. int queued = 0;
  522. /*
  523. * Step 3: Process LISTEN state
  524. *
  525. * If S.state == LISTEN,
  526. * If P.type == Request or P contains a valid Init Cookie option,
  527. * (* Must scan the packet's options to check for Init
  528. * Cookies. Only Init Cookies are processed here,
  529. * however; other options are processed in Step 8. This
  530. * scan need only be performed if the endpoint uses Init
  531. * Cookies *)
  532. * (* Generate a new socket and switch to that socket *)
  533. * Set S := new socket for this port pair
  534. * S.state = RESPOND
  535. * Choose S.ISS (initial seqno) or set from Init Cookies
  536. * Initialize S.GAR := S.ISS
  537. * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init
  538. * Cookies Continue with S.state == RESPOND
  539. * (* A Response packet will be generated in Step 11 *)
  540. * Otherwise,
  541. * Generate Reset(No Connection) unless P.type == Reset
  542. * Drop packet and return
  543. */
  544. if (sk->sk_state == DCCP_LISTEN) {
  545. if (dh->dccph_type == DCCP_PKT_REQUEST) {
  546. if (inet_csk(sk)->icsk_af_ops->conn_request(sk,
  547. skb) < 0)
  548. return 1;
  549. goto discard;
  550. }
  551. if (dh->dccph_type == DCCP_PKT_RESET)
  552. goto discard;
  553. /* Caller (dccp_v4_do_rcv) will send Reset */
  554. dcb->dccpd_reset_code = DCCP_RESET_CODE_NO_CONNECTION;
  555. return 1;
  556. }
  557. if (sk->sk_state != DCCP_REQUESTING && sk->sk_state != DCCP_RESPOND) {
  558. if (dccp_check_seqno(sk, skb))
  559. goto discard;
  560. /*
  561. * Step 8: Process options and mark acknowledgeable
  562. */
  563. if (dccp_parse_options(sk, NULL, skb))
  564. return 1;
  565. if (dcb->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ)
  566. dccp_event_ack_recv(sk, skb);
  567. if (dp->dccps_hc_rx_ackvec != NULL &&
  568. dccp_ackvec_add(dp->dccps_hc_rx_ackvec, sk,
  569. DCCP_SKB_CB(skb)->dccpd_seq,
  570. DCCP_ACKVEC_STATE_RECEIVED))
  571. goto discard;
  572. dccp_deliver_input_to_ccids(sk, skb);
  573. }
  574. /*
  575. * Step 9: Process Reset
  576. * If P.type == Reset,
  577. * Tear down connection
  578. * S.state := TIMEWAIT
  579. * Set TIMEWAIT timer
  580. * Drop packet and return
  581. */
  582. if (dh->dccph_type == DCCP_PKT_RESET) {
  583. dccp_rcv_reset(sk, skb);
  584. return 0;
  585. /*
  586. * Step 7: Check for unexpected packet types
  587. * If (S.is_server and P.type == Response)
  588. * or (S.is_client and P.type == Request)
  589. * or (S.state == RESPOND and P.type == Data),
  590. * Send Sync packet acknowledging P.seqno
  591. * Drop packet and return
  592. */
  593. } else if ((dp->dccps_role != DCCP_ROLE_CLIENT &&
  594. dh->dccph_type == DCCP_PKT_RESPONSE) ||
  595. (dp->dccps_role == DCCP_ROLE_CLIENT &&
  596. dh->dccph_type == DCCP_PKT_REQUEST) ||
  597. (sk->sk_state == DCCP_RESPOND &&
  598. dh->dccph_type == DCCP_PKT_DATA)) {
  599. dccp_send_sync(sk, dcb->dccpd_seq, DCCP_PKT_SYNC);
  600. goto discard;
  601. } else if (dh->dccph_type == DCCP_PKT_CLOSEREQ) {
  602. if (dccp_rcv_closereq(sk, skb))
  603. return 0;
  604. goto discard;
  605. } else if (dh->dccph_type == DCCP_PKT_CLOSE) {
  606. if (dccp_rcv_close(sk, skb))
  607. return 0;
  608. goto discard;
  609. }
  610. switch (sk->sk_state) {
  611. case DCCP_CLOSED:
  612. dcb->dccpd_reset_code = DCCP_RESET_CODE_NO_CONNECTION;
  613. return 1;
  614. case DCCP_REQUESTING:
  615. queued = dccp_rcv_request_sent_state_process(sk, skb, dh, len);
  616. if (queued >= 0)
  617. return queued;
  618. __kfree_skb(skb);
  619. return 0;
  620. case DCCP_RESPOND:
  621. case DCCP_PARTOPEN:
  622. queued = dccp_rcv_respond_partopen_state_process(sk, skb,
  623. dh, len);
  624. break;
  625. }
  626. if (dh->dccph_type == DCCP_PKT_ACK ||
  627. dh->dccph_type == DCCP_PKT_DATAACK) {
  628. switch (old_state) {
  629. case DCCP_PARTOPEN:
  630. sk->sk_state_change(sk);
  631. sk_wake_async(sk, SOCK_WAKE_IO, POLL_OUT);
  632. break;
  633. }
  634. } else if (unlikely(dh->dccph_type == DCCP_PKT_SYNC)) {
  635. dccp_send_sync(sk, dcb->dccpd_seq, DCCP_PKT_SYNCACK);
  636. goto discard;
  637. }
  638. if (!queued) {
  639. discard:
  640. __kfree_skb(skb);
  641. }
  642. return 0;
  643. }
  644. EXPORT_SYMBOL_GPL(dccp_rcv_state_process);
  645. /**
  646. * dccp_sample_rtt - Validate and finalise computation of RTT sample
  647. * @delta: number of microseconds between packet and acknowledgment
  648. * The routine is kept generic to work in different contexts. It should be
  649. * called immediately when the ACK used for the RTT sample arrives.
  650. */
  651. u32 dccp_sample_rtt(struct sock *sk, long delta)
  652. {
  653. /* dccpor_elapsed_time is either zeroed out or set and > 0 */
  654. delta -= dccp_sk(sk)->dccps_options_received.dccpor_elapsed_time * 10;
  655. if (unlikely(delta <= 0)) {
  656. DCCP_WARN("unusable RTT sample %ld, using min\n", delta);
  657. return DCCP_SANE_RTT_MIN;
  658. }
  659. if (unlikely(delta > DCCP_SANE_RTT_MAX)) {
  660. DCCP_WARN("RTT sample %ld too large, using max\n", delta);
  661. return DCCP_SANE_RTT_MAX;
  662. }
  663. return delta;
  664. }