transport.c 21 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693
  1. /* SCTP kernel implementation
  2. * Copyright (c) 1999-2000 Cisco, Inc.
  3. * Copyright (c) 1999-2001 Motorola, Inc.
  4. * Copyright (c) 2001-2003 International Business Machines Corp.
  5. * Copyright (c) 2001 Intel Corp.
  6. * Copyright (c) 2001 La Monte H.P. Yarroll
  7. *
  8. * This file is part of the SCTP kernel implementation
  9. *
  10. * This module provides the abstraction for an SCTP tranport representing
  11. * a remote transport address. For local transport addresses, we just use
  12. * union sctp_addr.
  13. *
  14. * This SCTP implementation is free software;
  15. * you can redistribute it and/or modify it under the terms of
  16. * the GNU General Public License as published by
  17. * the Free Software Foundation; either version 2, or (at your option)
  18. * any later version.
  19. *
  20. * This SCTP implementation is distributed in the hope that it
  21. * will be useful, but WITHOUT ANY WARRANTY; without even the implied
  22. * ************************
  23. * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
  24. * See the GNU General Public License for more details.
  25. *
  26. * You should have received a copy of the GNU General Public License
  27. * along with GNU CC; see the file COPYING. If not, see
  28. * <http://www.gnu.org/licenses/>.
  29. *
  30. * Please send any bug reports or fixes you make to the
  31. * email address(es):
  32. * lksctp developers <linux-sctp@vger.kernel.org>
  33. *
  34. * Written or modified by:
  35. * La Monte H.P. Yarroll <piggy@acm.org>
  36. * Karl Knutson <karl@athena.chicago.il.us>
  37. * Jon Grimm <jgrimm@us.ibm.com>
  38. * Xingang Guo <xingang.guo@intel.com>
  39. * Hui Huang <hui.huang@nokia.com>
  40. * Sridhar Samudrala <sri@us.ibm.com>
  41. * Ardelle Fan <ardelle.fan@intel.com>
  42. */
  43. #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  44. #include <linux/slab.h>
  45. #include <linux/types.h>
  46. #include <linux/random.h>
  47. #include <net/sctp/sctp.h>
  48. #include <net/sctp/sm.h>
  49. /* 1st Level Abstractions. */
  50. /* Initialize a new transport from provided memory. */
  51. static struct sctp_transport *sctp_transport_init(struct net *net,
  52. struct sctp_transport *peer,
  53. const union sctp_addr *addr,
  54. gfp_t gfp)
  55. {
  56. /* Copy in the address. */
  57. peer->ipaddr = *addr;
  58. peer->af_specific = sctp_get_af_specific(addr->sa.sa_family);
  59. memset(&peer->saddr, 0, sizeof(union sctp_addr));
  60. peer->sack_generation = 0;
  61. /* From 6.3.1 RTO Calculation:
  62. *
  63. * C1) Until an RTT measurement has been made for a packet sent to the
  64. * given destination transport address, set RTO to the protocol
  65. * parameter 'RTO.Initial'.
  66. */
  67. peer->rto = msecs_to_jiffies(net->sctp.rto_initial);
  68. peer->last_time_heard = 0;
  69. peer->last_time_ecne_reduced = jiffies;
  70. peer->param_flags = SPP_HB_DISABLE |
  71. SPP_PMTUD_ENABLE |
  72. SPP_SACKDELAY_ENABLE;
  73. /* Initialize the default path max_retrans. */
  74. peer->pathmaxrxt = net->sctp.max_retrans_path;
  75. peer->pf_retrans = net->sctp.pf_retrans;
  76. INIT_LIST_HEAD(&peer->transmitted);
  77. INIT_LIST_HEAD(&peer->send_ready);
  78. INIT_LIST_HEAD(&peer->transports);
  79. setup_timer(&peer->T3_rtx_timer, sctp_generate_t3_rtx_event,
  80. (unsigned long)peer);
  81. setup_timer(&peer->hb_timer, sctp_generate_heartbeat_event,
  82. (unsigned long)peer);
  83. setup_timer(&peer->reconf_timer, sctp_generate_reconf_event,
  84. (unsigned long)peer);
  85. setup_timer(&peer->proto_unreach_timer,
  86. sctp_generate_proto_unreach_event, (unsigned long)peer);
  87. /* Initialize the 64-bit random nonce sent with heartbeat. */
  88. get_random_bytes(&peer->hb_nonce, sizeof(peer->hb_nonce));
  89. refcount_set(&peer->refcnt, 1);
  90. return peer;
  91. }
  92. /* Allocate and initialize a new transport. */
  93. struct sctp_transport *sctp_transport_new(struct net *net,
  94. const union sctp_addr *addr,
  95. gfp_t gfp)
  96. {
  97. struct sctp_transport *transport;
  98. transport = kzalloc(sizeof(*transport), gfp);
  99. if (!transport)
  100. goto fail;
  101. if (!sctp_transport_init(net, transport, addr, gfp))
  102. goto fail_init;
  103. SCTP_DBG_OBJCNT_INC(transport);
  104. return transport;
  105. fail_init:
  106. kfree(transport);
  107. fail:
  108. return NULL;
  109. }
  110. /* This transport is no longer needed. Free up if possible, or
  111. * delay until it last reference count.
  112. */
  113. void sctp_transport_free(struct sctp_transport *transport)
  114. {
  115. /* Try to delete the heartbeat timer. */
  116. if (del_timer(&transport->hb_timer))
  117. sctp_transport_put(transport);
  118. /* Delete the T3_rtx timer if it's active.
  119. * There is no point in not doing this now and letting
  120. * structure hang around in memory since we know
  121. * the tranport is going away.
  122. */
  123. if (del_timer(&transport->T3_rtx_timer))
  124. sctp_transport_put(transport);
  125. if (del_timer(&transport->reconf_timer))
  126. sctp_transport_put(transport);
  127. /* Delete the ICMP proto unreachable timer if it's active. */
  128. if (del_timer(&transport->proto_unreach_timer))
  129. sctp_association_put(transport->asoc);
  130. sctp_transport_put(transport);
  131. }
  132. static void sctp_transport_destroy_rcu(struct rcu_head *head)
  133. {
  134. struct sctp_transport *transport;
  135. transport = container_of(head, struct sctp_transport, rcu);
  136. dst_release(transport->dst);
  137. kfree(transport);
  138. SCTP_DBG_OBJCNT_DEC(transport);
  139. }
  140. /* Destroy the transport data structure.
  141. * Assumes there are no more users of this structure.
  142. */
  143. static void sctp_transport_destroy(struct sctp_transport *transport)
  144. {
  145. if (unlikely(refcount_read(&transport->refcnt))) {
  146. WARN(1, "Attempt to destroy undead transport %p!\n", transport);
  147. return;
  148. }
  149. sctp_packet_free(&transport->packet);
  150. if (transport->asoc)
  151. sctp_association_put(transport->asoc);
  152. call_rcu(&transport->rcu, sctp_transport_destroy_rcu);
  153. }
  154. /* Start T3_rtx timer if it is not already running and update the heartbeat
  155. * timer. This routine is called every time a DATA chunk is sent.
  156. */
  157. void sctp_transport_reset_t3_rtx(struct sctp_transport *transport)
  158. {
  159. /* RFC 2960 6.3.2 Retransmission Timer Rules
  160. *
  161. * R1) Every time a DATA chunk is sent to any address(including a
  162. * retransmission), if the T3-rtx timer of that address is not running
  163. * start it running so that it will expire after the RTO of that
  164. * address.
  165. */
  166. if (!timer_pending(&transport->T3_rtx_timer))
  167. if (!mod_timer(&transport->T3_rtx_timer,
  168. jiffies + transport->rto))
  169. sctp_transport_hold(transport);
  170. }
  171. void sctp_transport_reset_hb_timer(struct sctp_transport *transport)
  172. {
  173. unsigned long expires;
  174. /* When a data chunk is sent, reset the heartbeat interval. */
  175. expires = jiffies + sctp_transport_timeout(transport);
  176. if (time_before(transport->hb_timer.expires, expires) &&
  177. !mod_timer(&transport->hb_timer,
  178. expires + prandom_u32_max(transport->rto)))
  179. sctp_transport_hold(transport);
  180. }
  181. void sctp_transport_reset_reconf_timer(struct sctp_transport *transport)
  182. {
  183. if (!timer_pending(&transport->reconf_timer))
  184. if (!mod_timer(&transport->reconf_timer,
  185. jiffies + transport->rto))
  186. sctp_transport_hold(transport);
  187. }
  188. /* This transport has been assigned to an association.
  189. * Initialize fields from the association or from the sock itself.
  190. * Register the reference count in the association.
  191. */
  192. void sctp_transport_set_owner(struct sctp_transport *transport,
  193. struct sctp_association *asoc)
  194. {
  195. transport->asoc = asoc;
  196. sctp_association_hold(asoc);
  197. }
  198. /* Initialize the pmtu of a transport. */
  199. void sctp_transport_pmtu(struct sctp_transport *transport, struct sock *sk)
  200. {
  201. /* If we don't have a fresh route, look one up */
  202. if (!transport->dst || transport->dst->obsolete) {
  203. sctp_transport_dst_release(transport);
  204. transport->af_specific->get_dst(transport, &transport->saddr,
  205. &transport->fl, sk);
  206. }
  207. if (transport->dst) {
  208. transport->pathmtu = SCTP_TRUNC4(dst_mtu(transport->dst));
  209. } else
  210. transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
  211. }
  212. void sctp_transport_update_pmtu(struct sctp_transport *t, u32 pmtu)
  213. {
  214. struct dst_entry *dst = sctp_transport_dst_check(t);
  215. if (unlikely(pmtu < SCTP_DEFAULT_MINSEGMENT)) {
  216. pr_warn("%s: Reported pmtu %d too low, using default minimum of %d\n",
  217. __func__, pmtu, SCTP_DEFAULT_MINSEGMENT);
  218. /* Use default minimum segment size and disable
  219. * pmtu discovery on this transport.
  220. */
  221. t->pathmtu = SCTP_DEFAULT_MINSEGMENT;
  222. } else {
  223. t->pathmtu = pmtu;
  224. }
  225. if (dst) {
  226. dst->ops->update_pmtu(dst, t->asoc->base.sk, NULL, pmtu);
  227. dst = sctp_transport_dst_check(t);
  228. }
  229. if (!dst)
  230. t->af_specific->get_dst(t, &t->saddr, &t->fl, t->asoc->base.sk);
  231. }
  232. /* Caches the dst entry and source address for a transport's destination
  233. * address.
  234. */
  235. void sctp_transport_route(struct sctp_transport *transport,
  236. union sctp_addr *saddr, struct sctp_sock *opt)
  237. {
  238. struct sctp_association *asoc = transport->asoc;
  239. struct sctp_af *af = transport->af_specific;
  240. af->get_dst(transport, saddr, &transport->fl, sctp_opt2sk(opt));
  241. if (saddr)
  242. memcpy(&transport->saddr, saddr, sizeof(union sctp_addr));
  243. else
  244. af->get_saddr(opt, transport, &transport->fl);
  245. if ((transport->param_flags & SPP_PMTUD_DISABLE) && transport->pathmtu) {
  246. return;
  247. }
  248. if (transport->dst) {
  249. transport->pathmtu = SCTP_TRUNC4(dst_mtu(transport->dst));
  250. /* Initialize sk->sk_rcv_saddr, if the transport is the
  251. * association's active path for getsockname().
  252. */
  253. if (asoc && (!asoc->peer.primary_path ||
  254. (transport == asoc->peer.active_path)))
  255. opt->pf->to_sk_saddr(&transport->saddr,
  256. asoc->base.sk);
  257. } else
  258. transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
  259. }
  260. /* Hold a reference to a transport. */
  261. int sctp_transport_hold(struct sctp_transport *transport)
  262. {
  263. return refcount_inc_not_zero(&transport->refcnt);
  264. }
  265. /* Release a reference to a transport and clean up
  266. * if there are no more references.
  267. */
  268. void sctp_transport_put(struct sctp_transport *transport)
  269. {
  270. if (refcount_dec_and_test(&transport->refcnt))
  271. sctp_transport_destroy(transport);
  272. }
  273. /* Update transport's RTO based on the newly calculated RTT. */
  274. void sctp_transport_update_rto(struct sctp_transport *tp, __u32 rtt)
  275. {
  276. if (unlikely(!tp->rto_pending))
  277. /* We should not be doing any RTO updates unless rto_pending is set. */
  278. pr_debug("%s: rto_pending not set on transport %p!\n", __func__, tp);
  279. if (tp->rttvar || tp->srtt) {
  280. struct net *net = sock_net(tp->asoc->base.sk);
  281. /* 6.3.1 C3) When a new RTT measurement R' is made, set
  282. * RTTVAR <- (1 - RTO.Beta) * RTTVAR + RTO.Beta * |SRTT - R'|
  283. * SRTT <- (1 - RTO.Alpha) * SRTT + RTO.Alpha * R'
  284. */
  285. /* Note: The above algorithm has been rewritten to
  286. * express rto_beta and rto_alpha as inverse powers
  287. * of two.
  288. * For example, assuming the default value of RTO.Alpha of
  289. * 1/8, rto_alpha would be expressed as 3.
  290. */
  291. tp->rttvar = tp->rttvar - (tp->rttvar >> net->sctp.rto_beta)
  292. + (((__u32)abs((__s64)tp->srtt - (__s64)rtt)) >> net->sctp.rto_beta);
  293. tp->srtt = tp->srtt - (tp->srtt >> net->sctp.rto_alpha)
  294. + (rtt >> net->sctp.rto_alpha);
  295. } else {
  296. /* 6.3.1 C2) When the first RTT measurement R is made, set
  297. * SRTT <- R, RTTVAR <- R/2.
  298. */
  299. tp->srtt = rtt;
  300. tp->rttvar = rtt >> 1;
  301. }
  302. /* 6.3.1 G1) Whenever RTTVAR is computed, if RTTVAR = 0, then
  303. * adjust RTTVAR <- G, where G is the CLOCK GRANULARITY.
  304. */
  305. if (tp->rttvar == 0)
  306. tp->rttvar = SCTP_CLOCK_GRANULARITY;
  307. /* 6.3.1 C3) After the computation, update RTO <- SRTT + 4 * RTTVAR. */
  308. tp->rto = tp->srtt + (tp->rttvar << 2);
  309. /* 6.3.1 C6) Whenever RTO is computed, if it is less than RTO.Min
  310. * seconds then it is rounded up to RTO.Min seconds.
  311. */
  312. if (tp->rto < tp->asoc->rto_min)
  313. tp->rto = tp->asoc->rto_min;
  314. /* 6.3.1 C7) A maximum value may be placed on RTO provided it is
  315. * at least RTO.max seconds.
  316. */
  317. if (tp->rto > tp->asoc->rto_max)
  318. tp->rto = tp->asoc->rto_max;
  319. sctp_max_rto(tp->asoc, tp);
  320. tp->rtt = rtt;
  321. /* Reset rto_pending so that a new RTT measurement is started when a
  322. * new data chunk is sent.
  323. */
  324. tp->rto_pending = 0;
  325. pr_debug("%s: transport:%p, rtt:%d, srtt:%d rttvar:%d, rto:%ld\n",
  326. __func__, tp, rtt, tp->srtt, tp->rttvar, tp->rto);
  327. }
  328. /* This routine updates the transport's cwnd and partial_bytes_acked
  329. * parameters based on the bytes acked in the received SACK.
  330. */
  331. void sctp_transport_raise_cwnd(struct sctp_transport *transport,
  332. __u32 sack_ctsn, __u32 bytes_acked)
  333. {
  334. struct sctp_association *asoc = transport->asoc;
  335. __u32 cwnd, ssthresh, flight_size, pba, pmtu;
  336. cwnd = transport->cwnd;
  337. flight_size = transport->flight_size;
  338. /* See if we need to exit Fast Recovery first */
  339. if (asoc->fast_recovery &&
  340. TSN_lte(asoc->fast_recovery_exit, sack_ctsn))
  341. asoc->fast_recovery = 0;
  342. ssthresh = transport->ssthresh;
  343. pba = transport->partial_bytes_acked;
  344. pmtu = transport->asoc->pathmtu;
  345. if (cwnd <= ssthresh) {
  346. /* RFC 4960 7.2.1
  347. * o When cwnd is less than or equal to ssthresh, an SCTP
  348. * endpoint MUST use the slow-start algorithm to increase
  349. * cwnd only if the current congestion window is being fully
  350. * utilized, an incoming SACK advances the Cumulative TSN
  351. * Ack Point, and the data sender is not in Fast Recovery.
  352. * Only when these three conditions are met can the cwnd be
  353. * increased; otherwise, the cwnd MUST not be increased.
  354. * If these conditions are met, then cwnd MUST be increased
  355. * by, at most, the lesser of 1) the total size of the
  356. * previously outstanding DATA chunk(s) acknowledged, and
  357. * 2) the destination's path MTU. This upper bound protects
  358. * against the ACK-Splitting attack outlined in [SAVAGE99].
  359. */
  360. if (asoc->fast_recovery)
  361. return;
  362. /* The appropriate cwnd increase algorithm is performed
  363. * if, and only if the congestion window is being fully
  364. * utilized. Note that RFC4960 Errata 3.22 removed the
  365. * other condition on ctsn moving.
  366. */
  367. if (flight_size < cwnd)
  368. return;
  369. if (bytes_acked > pmtu)
  370. cwnd += pmtu;
  371. else
  372. cwnd += bytes_acked;
  373. pr_debug("%s: slow start: transport:%p, bytes_acked:%d, "
  374. "cwnd:%d, ssthresh:%d, flight_size:%d, pba:%d\n",
  375. __func__, transport, bytes_acked, cwnd, ssthresh,
  376. flight_size, pba);
  377. } else {
  378. /* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh,
  379. * upon each SACK arrival, increase partial_bytes_acked
  380. * by the total number of bytes of all new chunks
  381. * acknowledged in that SACK including chunks
  382. * acknowledged by the new Cumulative TSN Ack and by Gap
  383. * Ack Blocks. (updated by RFC4960 Errata 3.22)
  384. *
  385. * When partial_bytes_acked is greater than cwnd and
  386. * before the arrival of the SACK the sender had less
  387. * bytes of data outstanding than cwnd (i.e., before
  388. * arrival of the SACK, flightsize was less than cwnd),
  389. * reset partial_bytes_acked to cwnd. (RFC 4960 Errata
  390. * 3.26)
  391. *
  392. * When partial_bytes_acked is equal to or greater than
  393. * cwnd and before the arrival of the SACK the sender
  394. * had cwnd or more bytes of data outstanding (i.e.,
  395. * before arrival of the SACK, flightsize was greater
  396. * than or equal to cwnd), partial_bytes_acked is reset
  397. * to (partial_bytes_acked - cwnd). Next, cwnd is
  398. * increased by MTU. (RFC 4960 Errata 3.12)
  399. */
  400. pba += bytes_acked;
  401. if (pba > cwnd && flight_size < cwnd)
  402. pba = cwnd;
  403. if (pba >= cwnd && flight_size >= cwnd) {
  404. pba = pba - cwnd;
  405. cwnd += pmtu;
  406. }
  407. pr_debug("%s: congestion avoidance: transport:%p, "
  408. "bytes_acked:%d, cwnd:%d, ssthresh:%d, "
  409. "flight_size:%d, pba:%d\n", __func__,
  410. transport, bytes_acked, cwnd, ssthresh,
  411. flight_size, pba);
  412. }
  413. transport->cwnd = cwnd;
  414. transport->partial_bytes_acked = pba;
  415. }
  416. /* This routine is used to lower the transport's cwnd when congestion is
  417. * detected.
  418. */
  419. void sctp_transport_lower_cwnd(struct sctp_transport *transport,
  420. sctp_lower_cwnd_t reason)
  421. {
  422. struct sctp_association *asoc = transport->asoc;
  423. switch (reason) {
  424. case SCTP_LOWER_CWND_T3_RTX:
  425. /* RFC 2960 Section 7.2.3, sctpimpguide
  426. * When the T3-rtx timer expires on an address, SCTP should
  427. * perform slow start by:
  428. * ssthresh = max(cwnd/2, 4*MTU)
  429. * cwnd = 1*MTU
  430. * partial_bytes_acked = 0
  431. */
  432. transport->ssthresh = max(transport->cwnd/2,
  433. 4*asoc->pathmtu);
  434. transport->cwnd = asoc->pathmtu;
  435. /* T3-rtx also clears fast recovery */
  436. asoc->fast_recovery = 0;
  437. break;
  438. case SCTP_LOWER_CWND_FAST_RTX:
  439. /* RFC 2960 7.2.4 Adjust the ssthresh and cwnd of the
  440. * destination address(es) to which the missing DATA chunks
  441. * were last sent, according to the formula described in
  442. * Section 7.2.3.
  443. *
  444. * RFC 2960 7.2.3, sctpimpguide Upon detection of packet
  445. * losses from SACK (see Section 7.2.4), An endpoint
  446. * should do the following:
  447. * ssthresh = max(cwnd/2, 4*MTU)
  448. * cwnd = ssthresh
  449. * partial_bytes_acked = 0
  450. */
  451. if (asoc->fast_recovery)
  452. return;
  453. /* Mark Fast recovery */
  454. asoc->fast_recovery = 1;
  455. asoc->fast_recovery_exit = asoc->next_tsn - 1;
  456. transport->ssthresh = max(transport->cwnd/2,
  457. 4*asoc->pathmtu);
  458. transport->cwnd = transport->ssthresh;
  459. break;
  460. case SCTP_LOWER_CWND_ECNE:
  461. /* RFC 2481 Section 6.1.2.
  462. * If the sender receives an ECN-Echo ACK packet
  463. * then the sender knows that congestion was encountered in the
  464. * network on the path from the sender to the receiver. The
  465. * indication of congestion should be treated just as a
  466. * congestion loss in non-ECN Capable TCP. That is, the TCP
  467. * source halves the congestion window "cwnd" and reduces the
  468. * slow start threshold "ssthresh".
  469. * A critical condition is that TCP does not react to
  470. * congestion indications more than once every window of
  471. * data (or more loosely more than once every round-trip time).
  472. */
  473. if (time_after(jiffies, transport->last_time_ecne_reduced +
  474. transport->rtt)) {
  475. transport->ssthresh = max(transport->cwnd/2,
  476. 4*asoc->pathmtu);
  477. transport->cwnd = transport->ssthresh;
  478. transport->last_time_ecne_reduced = jiffies;
  479. }
  480. break;
  481. case SCTP_LOWER_CWND_INACTIVE:
  482. /* RFC 2960 Section 7.2.1, sctpimpguide
  483. * When the endpoint does not transmit data on a given
  484. * transport address, the cwnd of the transport address
  485. * should be adjusted to max(cwnd/2, 4*MTU) per RTO.
  486. * NOTE: Although the draft recommends that this check needs
  487. * to be done every RTO interval, we do it every hearbeat
  488. * interval.
  489. */
  490. transport->cwnd = max(transport->cwnd/2,
  491. 4*asoc->pathmtu);
  492. /* RFC 4960 Errata 3.27.2: also adjust sshthresh */
  493. transport->ssthresh = transport->cwnd;
  494. break;
  495. }
  496. transport->partial_bytes_acked = 0;
  497. pr_debug("%s: transport:%p, reason:%d, cwnd:%d, ssthresh:%d\n",
  498. __func__, transport, reason, transport->cwnd,
  499. transport->ssthresh);
  500. }
  501. /* Apply Max.Burst limit to the congestion window:
  502. * sctpimpguide-05 2.14.2
  503. * D) When the time comes for the sender to
  504. * transmit new DATA chunks, the protocol parameter Max.Burst MUST
  505. * first be applied to limit how many new DATA chunks may be sent.
  506. * The limit is applied by adjusting cwnd as follows:
  507. * if ((flightsize+ Max.Burst * MTU) < cwnd)
  508. * cwnd = flightsize + Max.Burst * MTU
  509. */
  510. void sctp_transport_burst_limited(struct sctp_transport *t)
  511. {
  512. struct sctp_association *asoc = t->asoc;
  513. u32 old_cwnd = t->cwnd;
  514. u32 max_burst_bytes;
  515. if (t->burst_limited || asoc->max_burst == 0)
  516. return;
  517. max_burst_bytes = t->flight_size + (asoc->max_burst * asoc->pathmtu);
  518. if (max_burst_bytes < old_cwnd) {
  519. t->cwnd = max_burst_bytes;
  520. t->burst_limited = old_cwnd;
  521. }
  522. }
  523. /* Restore the old cwnd congestion window, after the burst had it's
  524. * desired effect.
  525. */
  526. void sctp_transport_burst_reset(struct sctp_transport *t)
  527. {
  528. if (t->burst_limited) {
  529. t->cwnd = t->burst_limited;
  530. t->burst_limited = 0;
  531. }
  532. }
  533. /* What is the next timeout value for this transport? */
  534. unsigned long sctp_transport_timeout(struct sctp_transport *trans)
  535. {
  536. /* RTO + timer slack +/- 50% of RTO */
  537. unsigned long timeout = trans->rto >> 1;
  538. if (trans->state != SCTP_UNCONFIRMED &&
  539. trans->state != SCTP_PF)
  540. timeout += trans->hbinterval;
  541. return timeout;
  542. }
  543. /* Reset transport variables to their initial values */
  544. void sctp_transport_reset(struct sctp_transport *t)
  545. {
  546. struct sctp_association *asoc = t->asoc;
  547. /* RFC 2960 (bis), Section 5.2.4
  548. * All the congestion control parameters (e.g., cwnd, ssthresh)
  549. * related to this peer MUST be reset to their initial values
  550. * (see Section 6.2.1)
  551. */
  552. t->cwnd = min(4*asoc->pathmtu, max_t(__u32, 2*asoc->pathmtu, 4380));
  553. t->burst_limited = 0;
  554. t->ssthresh = asoc->peer.i.a_rwnd;
  555. t->rto = asoc->rto_initial;
  556. sctp_max_rto(asoc, t);
  557. t->rtt = 0;
  558. t->srtt = 0;
  559. t->rttvar = 0;
  560. /* Reset these additional variables so that we have a clean slate. */
  561. t->partial_bytes_acked = 0;
  562. t->flight_size = 0;
  563. t->error_count = 0;
  564. t->rto_pending = 0;
  565. t->hb_sent = 0;
  566. /* Initialize the state information for SFR-CACC */
  567. t->cacc.changeover_active = 0;
  568. t->cacc.cycling_changeover = 0;
  569. t->cacc.next_tsn_at_change = 0;
  570. t->cacc.cacc_saw_newack = 0;
  571. }
  572. /* Schedule retransmission on the given transport */
  573. void sctp_transport_immediate_rtx(struct sctp_transport *t)
  574. {
  575. /* Stop pending T3_rtx_timer */
  576. if (del_timer(&t->T3_rtx_timer))
  577. sctp_transport_put(t);
  578. sctp_retransmit(&t->asoc->outqueue, t, SCTP_RTXR_T3_RTX);
  579. if (!timer_pending(&t->T3_rtx_timer)) {
  580. if (!mod_timer(&t->T3_rtx_timer, jiffies + t->rto))
  581. sctp_transport_hold(t);
  582. }
  583. }
  584. /* Drop dst */
  585. void sctp_transport_dst_release(struct sctp_transport *t)
  586. {
  587. dst_release(t->dst);
  588. t->dst = NULL;
  589. t->dst_pending_confirm = 0;
  590. }
  591. /* Schedule neighbour confirm */
  592. void sctp_transport_dst_confirm(struct sctp_transport *t)
  593. {
  594. t->dst_pending_confirm = 1;
  595. }