transport.c 21 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702
  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. timer_setup(&peer->T3_rtx_timer, sctp_generate_t3_rtx_event, 0);
  80. timer_setup(&peer->hb_timer, sctp_generate_heartbeat_event, 0);
  81. timer_setup(&peer->reconf_timer, sctp_generate_reconf_event, 0);
  82. timer_setup(&peer->proto_unreach_timer,
  83. sctp_generate_proto_unreach_event, 0);
  84. /* Initialize the 64-bit random nonce sent with heartbeat. */
  85. get_random_bytes(&peer->hb_nonce, sizeof(peer->hb_nonce));
  86. refcount_set(&peer->refcnt, 1);
  87. return peer;
  88. }
  89. /* Allocate and initialize a new transport. */
  90. struct sctp_transport *sctp_transport_new(struct net *net,
  91. const union sctp_addr *addr,
  92. gfp_t gfp)
  93. {
  94. struct sctp_transport *transport;
  95. transport = kzalloc(sizeof(*transport), gfp);
  96. if (!transport)
  97. goto fail;
  98. if (!sctp_transport_init(net, transport, addr, gfp))
  99. goto fail_init;
  100. SCTP_DBG_OBJCNT_INC(transport);
  101. return transport;
  102. fail_init:
  103. kfree(transport);
  104. fail:
  105. return NULL;
  106. }
  107. /* This transport is no longer needed. Free up if possible, or
  108. * delay until it last reference count.
  109. */
  110. void sctp_transport_free(struct sctp_transport *transport)
  111. {
  112. /* Try to delete the heartbeat timer. */
  113. if (del_timer(&transport->hb_timer))
  114. sctp_transport_put(transport);
  115. /* Delete the T3_rtx timer if it's active.
  116. * There is no point in not doing this now and letting
  117. * structure hang around in memory since we know
  118. * the tranport is going away.
  119. */
  120. if (del_timer(&transport->T3_rtx_timer))
  121. sctp_transport_put(transport);
  122. if (del_timer(&transport->reconf_timer))
  123. sctp_transport_put(transport);
  124. /* Delete the ICMP proto unreachable timer if it's active. */
  125. if (del_timer(&transport->proto_unreach_timer))
  126. sctp_association_put(transport->asoc);
  127. sctp_transport_put(transport);
  128. }
  129. static void sctp_transport_destroy_rcu(struct rcu_head *head)
  130. {
  131. struct sctp_transport *transport;
  132. transport = container_of(head, struct sctp_transport, rcu);
  133. dst_release(transport->dst);
  134. kfree(transport);
  135. SCTP_DBG_OBJCNT_DEC(transport);
  136. }
  137. /* Destroy the transport data structure.
  138. * Assumes there are no more users of this structure.
  139. */
  140. static void sctp_transport_destroy(struct sctp_transport *transport)
  141. {
  142. if (unlikely(refcount_read(&transport->refcnt))) {
  143. WARN(1, "Attempt to destroy undead transport %p!\n", transport);
  144. return;
  145. }
  146. sctp_packet_free(&transport->packet);
  147. if (transport->asoc)
  148. sctp_association_put(transport->asoc);
  149. call_rcu(&transport->rcu, sctp_transport_destroy_rcu);
  150. }
  151. /* Start T3_rtx timer if it is not already running and update the heartbeat
  152. * timer. This routine is called every time a DATA chunk is sent.
  153. */
  154. void sctp_transport_reset_t3_rtx(struct sctp_transport *transport)
  155. {
  156. /* RFC 2960 6.3.2 Retransmission Timer Rules
  157. *
  158. * R1) Every time a DATA chunk is sent to any address(including a
  159. * retransmission), if the T3-rtx timer of that address is not running
  160. * start it running so that it will expire after the RTO of that
  161. * address.
  162. */
  163. if (!timer_pending(&transport->T3_rtx_timer))
  164. if (!mod_timer(&transport->T3_rtx_timer,
  165. jiffies + transport->rto))
  166. sctp_transport_hold(transport);
  167. }
  168. void sctp_transport_reset_hb_timer(struct sctp_transport *transport)
  169. {
  170. unsigned long expires;
  171. /* When a data chunk is sent, reset the heartbeat interval. */
  172. expires = jiffies + sctp_transport_timeout(transport);
  173. if (time_before(transport->hb_timer.expires, expires) &&
  174. !mod_timer(&transport->hb_timer,
  175. expires + prandom_u32_max(transport->rto)))
  176. sctp_transport_hold(transport);
  177. }
  178. void sctp_transport_reset_reconf_timer(struct sctp_transport *transport)
  179. {
  180. if (!timer_pending(&transport->reconf_timer))
  181. if (!mod_timer(&transport->reconf_timer,
  182. jiffies + transport->rto))
  183. sctp_transport_hold(transport);
  184. }
  185. /* This transport has been assigned to an association.
  186. * Initialize fields from the association or from the sock itself.
  187. * Register the reference count in the association.
  188. */
  189. void sctp_transport_set_owner(struct sctp_transport *transport,
  190. struct sctp_association *asoc)
  191. {
  192. transport->asoc = asoc;
  193. sctp_association_hold(asoc);
  194. }
  195. /* Initialize the pmtu of a transport. */
  196. void sctp_transport_pmtu(struct sctp_transport *transport, struct sock *sk)
  197. {
  198. /* If we don't have a fresh route, look one up */
  199. if (!transport->dst || transport->dst->obsolete) {
  200. sctp_transport_dst_release(transport);
  201. transport->af_specific->get_dst(transport, &transport->saddr,
  202. &transport->fl, sk);
  203. }
  204. if (transport->param_flags & SPP_PMTUD_DISABLE) {
  205. struct sctp_association *asoc = transport->asoc;
  206. if (!transport->pathmtu && asoc && asoc->pathmtu)
  207. transport->pathmtu = asoc->pathmtu;
  208. if (transport->pathmtu)
  209. return;
  210. }
  211. if (transport->dst)
  212. transport->pathmtu = sctp_dst_mtu(transport->dst);
  213. else
  214. transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
  215. }
  216. bool sctp_transport_update_pmtu(struct sctp_transport *t, u32 pmtu)
  217. {
  218. struct dst_entry *dst = sctp_transport_dst_check(t);
  219. bool change = true;
  220. if (unlikely(pmtu < SCTP_DEFAULT_MINSEGMENT)) {
  221. pr_warn_ratelimited("%s: Reported pmtu %d too low, using default minimum of %d\n",
  222. __func__, pmtu, SCTP_DEFAULT_MINSEGMENT);
  223. /* Use default minimum segment instead */
  224. pmtu = SCTP_DEFAULT_MINSEGMENT;
  225. }
  226. pmtu = SCTP_TRUNC4(pmtu);
  227. if (dst) {
  228. dst->ops->update_pmtu(dst, t->asoc->base.sk, NULL, pmtu);
  229. dst = sctp_transport_dst_check(t);
  230. }
  231. if (!dst) {
  232. t->af_specific->get_dst(t, &t->saddr, &t->fl, t->asoc->base.sk);
  233. dst = t->dst;
  234. }
  235. if (dst) {
  236. /* Re-fetch, as under layers may have a higher minimum size */
  237. pmtu = SCTP_TRUNC4(dst_mtu(dst));
  238. change = t->pathmtu != pmtu;
  239. }
  240. t->pathmtu = pmtu;
  241. return change;
  242. }
  243. /* Caches the dst entry and source address for a transport's destination
  244. * address.
  245. */
  246. void sctp_transport_route(struct sctp_transport *transport,
  247. union sctp_addr *saddr, struct sctp_sock *opt)
  248. {
  249. struct sctp_association *asoc = transport->asoc;
  250. struct sctp_af *af = transport->af_specific;
  251. sctp_transport_dst_release(transport);
  252. af->get_dst(transport, saddr, &transport->fl, sctp_opt2sk(opt));
  253. if (saddr)
  254. memcpy(&transport->saddr, saddr, sizeof(union sctp_addr));
  255. else
  256. af->get_saddr(opt, transport, &transport->fl);
  257. sctp_transport_pmtu(transport, sctp_opt2sk(opt));
  258. /* Initialize sk->sk_rcv_saddr, if the transport is the
  259. * association's active path for getsockname().
  260. */
  261. if (transport->dst && asoc &&
  262. (!asoc->peer.primary_path || transport == asoc->peer.active_path))
  263. opt->pf->to_sk_saddr(&transport->saddr, asoc->base.sk);
  264. }
  265. /* Hold a reference to a transport. */
  266. int sctp_transport_hold(struct sctp_transport *transport)
  267. {
  268. return refcount_inc_not_zero(&transport->refcnt);
  269. }
  270. /* Release a reference to a transport and clean up
  271. * if there are no more references.
  272. */
  273. void sctp_transport_put(struct sctp_transport *transport)
  274. {
  275. if (refcount_dec_and_test(&transport->refcnt))
  276. sctp_transport_destroy(transport);
  277. }
  278. /* Update transport's RTO based on the newly calculated RTT. */
  279. void sctp_transport_update_rto(struct sctp_transport *tp, __u32 rtt)
  280. {
  281. if (unlikely(!tp->rto_pending))
  282. /* We should not be doing any RTO updates unless rto_pending is set. */
  283. pr_debug("%s: rto_pending not set on transport %p!\n", __func__, tp);
  284. if (tp->rttvar || tp->srtt) {
  285. struct net *net = sock_net(tp->asoc->base.sk);
  286. /* 6.3.1 C3) When a new RTT measurement R' is made, set
  287. * RTTVAR <- (1 - RTO.Beta) * RTTVAR + RTO.Beta * |SRTT - R'|
  288. * SRTT <- (1 - RTO.Alpha) * SRTT + RTO.Alpha * R'
  289. */
  290. /* Note: The above algorithm has been rewritten to
  291. * express rto_beta and rto_alpha as inverse powers
  292. * of two.
  293. * For example, assuming the default value of RTO.Alpha of
  294. * 1/8, rto_alpha would be expressed as 3.
  295. */
  296. tp->rttvar = tp->rttvar - (tp->rttvar >> net->sctp.rto_beta)
  297. + (((__u32)abs((__s64)tp->srtt - (__s64)rtt)) >> net->sctp.rto_beta);
  298. tp->srtt = tp->srtt - (tp->srtt >> net->sctp.rto_alpha)
  299. + (rtt >> net->sctp.rto_alpha);
  300. } else {
  301. /* 6.3.1 C2) When the first RTT measurement R is made, set
  302. * SRTT <- R, RTTVAR <- R/2.
  303. */
  304. tp->srtt = rtt;
  305. tp->rttvar = rtt >> 1;
  306. }
  307. /* 6.3.1 G1) Whenever RTTVAR is computed, if RTTVAR = 0, then
  308. * adjust RTTVAR <- G, where G is the CLOCK GRANULARITY.
  309. */
  310. if (tp->rttvar == 0)
  311. tp->rttvar = SCTP_CLOCK_GRANULARITY;
  312. /* 6.3.1 C3) After the computation, update RTO <- SRTT + 4 * RTTVAR. */
  313. tp->rto = tp->srtt + (tp->rttvar << 2);
  314. /* 6.3.1 C6) Whenever RTO is computed, if it is less than RTO.Min
  315. * seconds then it is rounded up to RTO.Min seconds.
  316. */
  317. if (tp->rto < tp->asoc->rto_min)
  318. tp->rto = tp->asoc->rto_min;
  319. /* 6.3.1 C7) A maximum value may be placed on RTO provided it is
  320. * at least RTO.max seconds.
  321. */
  322. if (tp->rto > tp->asoc->rto_max)
  323. tp->rto = tp->asoc->rto_max;
  324. sctp_max_rto(tp->asoc, tp);
  325. tp->rtt = rtt;
  326. /* Reset rto_pending so that a new RTT measurement is started when a
  327. * new data chunk is sent.
  328. */
  329. tp->rto_pending = 0;
  330. pr_debug("%s: transport:%p, rtt:%d, srtt:%d rttvar:%d, rto:%ld\n",
  331. __func__, tp, rtt, tp->srtt, tp->rttvar, tp->rto);
  332. }
  333. /* This routine updates the transport's cwnd and partial_bytes_acked
  334. * parameters based on the bytes acked in the received SACK.
  335. */
  336. void sctp_transport_raise_cwnd(struct sctp_transport *transport,
  337. __u32 sack_ctsn, __u32 bytes_acked)
  338. {
  339. struct sctp_association *asoc = transport->asoc;
  340. __u32 cwnd, ssthresh, flight_size, pba, pmtu;
  341. cwnd = transport->cwnd;
  342. flight_size = transport->flight_size;
  343. /* See if we need to exit Fast Recovery first */
  344. if (asoc->fast_recovery &&
  345. TSN_lte(asoc->fast_recovery_exit, sack_ctsn))
  346. asoc->fast_recovery = 0;
  347. ssthresh = transport->ssthresh;
  348. pba = transport->partial_bytes_acked;
  349. pmtu = transport->asoc->pathmtu;
  350. if (cwnd <= ssthresh) {
  351. /* RFC 4960 7.2.1
  352. * o When cwnd is less than or equal to ssthresh, an SCTP
  353. * endpoint MUST use the slow-start algorithm to increase
  354. * cwnd only if the current congestion window is being fully
  355. * utilized, an incoming SACK advances the Cumulative TSN
  356. * Ack Point, and the data sender is not in Fast Recovery.
  357. * Only when these three conditions are met can the cwnd be
  358. * increased; otherwise, the cwnd MUST not be increased.
  359. * If these conditions are met, then cwnd MUST be increased
  360. * by, at most, the lesser of 1) the total size of the
  361. * previously outstanding DATA chunk(s) acknowledged, and
  362. * 2) the destination's path MTU. This upper bound protects
  363. * against the ACK-Splitting attack outlined in [SAVAGE99].
  364. */
  365. if (asoc->fast_recovery)
  366. return;
  367. /* The appropriate cwnd increase algorithm is performed
  368. * if, and only if the congestion window is being fully
  369. * utilized. Note that RFC4960 Errata 3.22 removed the
  370. * other condition on ctsn moving.
  371. */
  372. if (flight_size < cwnd)
  373. return;
  374. if (bytes_acked > pmtu)
  375. cwnd += pmtu;
  376. else
  377. cwnd += bytes_acked;
  378. pr_debug("%s: slow start: transport:%p, bytes_acked:%d, "
  379. "cwnd:%d, ssthresh:%d, flight_size:%d, pba:%d\n",
  380. __func__, transport, bytes_acked, cwnd, ssthresh,
  381. flight_size, pba);
  382. } else {
  383. /* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh,
  384. * upon each SACK arrival, increase partial_bytes_acked
  385. * by the total number of bytes of all new chunks
  386. * acknowledged in that SACK including chunks
  387. * acknowledged by the new Cumulative TSN Ack and by Gap
  388. * Ack Blocks. (updated by RFC4960 Errata 3.22)
  389. *
  390. * When partial_bytes_acked is greater than cwnd and
  391. * before the arrival of the SACK the sender had less
  392. * bytes of data outstanding than cwnd (i.e., before
  393. * arrival of the SACK, flightsize was less than cwnd),
  394. * reset partial_bytes_acked to cwnd. (RFC 4960 Errata
  395. * 3.26)
  396. *
  397. * When partial_bytes_acked is equal to or greater than
  398. * cwnd and before the arrival of the SACK the sender
  399. * had cwnd or more bytes of data outstanding (i.e.,
  400. * before arrival of the SACK, flightsize was greater
  401. * than or equal to cwnd), partial_bytes_acked is reset
  402. * to (partial_bytes_acked - cwnd). Next, cwnd is
  403. * increased by MTU. (RFC 4960 Errata 3.12)
  404. */
  405. pba += bytes_acked;
  406. if (pba > cwnd && flight_size < cwnd)
  407. pba = cwnd;
  408. if (pba >= cwnd && flight_size >= cwnd) {
  409. pba = pba - cwnd;
  410. cwnd += pmtu;
  411. }
  412. pr_debug("%s: congestion avoidance: transport:%p, "
  413. "bytes_acked:%d, cwnd:%d, ssthresh:%d, "
  414. "flight_size:%d, pba:%d\n", __func__,
  415. transport, bytes_acked, cwnd, ssthresh,
  416. flight_size, pba);
  417. }
  418. transport->cwnd = cwnd;
  419. transport->partial_bytes_acked = pba;
  420. }
  421. /* This routine is used to lower the transport's cwnd when congestion is
  422. * detected.
  423. */
  424. void sctp_transport_lower_cwnd(struct sctp_transport *transport,
  425. enum sctp_lower_cwnd reason)
  426. {
  427. struct sctp_association *asoc = transport->asoc;
  428. switch (reason) {
  429. case SCTP_LOWER_CWND_T3_RTX:
  430. /* RFC 2960 Section 7.2.3, sctpimpguide
  431. * When the T3-rtx timer expires on an address, SCTP should
  432. * perform slow start by:
  433. * ssthresh = max(cwnd/2, 4*MTU)
  434. * cwnd = 1*MTU
  435. * partial_bytes_acked = 0
  436. */
  437. transport->ssthresh = max(transport->cwnd/2,
  438. 4*asoc->pathmtu);
  439. transport->cwnd = asoc->pathmtu;
  440. /* T3-rtx also clears fast recovery */
  441. asoc->fast_recovery = 0;
  442. break;
  443. case SCTP_LOWER_CWND_FAST_RTX:
  444. /* RFC 2960 7.2.4 Adjust the ssthresh and cwnd of the
  445. * destination address(es) to which the missing DATA chunks
  446. * were last sent, according to the formula described in
  447. * Section 7.2.3.
  448. *
  449. * RFC 2960 7.2.3, sctpimpguide Upon detection of packet
  450. * losses from SACK (see Section 7.2.4), An endpoint
  451. * should do the following:
  452. * ssthresh = max(cwnd/2, 4*MTU)
  453. * cwnd = ssthresh
  454. * partial_bytes_acked = 0
  455. */
  456. if (asoc->fast_recovery)
  457. return;
  458. /* Mark Fast recovery */
  459. asoc->fast_recovery = 1;
  460. asoc->fast_recovery_exit = asoc->next_tsn - 1;
  461. transport->ssthresh = max(transport->cwnd/2,
  462. 4*asoc->pathmtu);
  463. transport->cwnd = transport->ssthresh;
  464. break;
  465. case SCTP_LOWER_CWND_ECNE:
  466. /* RFC 2481 Section 6.1.2.
  467. * If the sender receives an ECN-Echo ACK packet
  468. * then the sender knows that congestion was encountered in the
  469. * network on the path from the sender to the receiver. The
  470. * indication of congestion should be treated just as a
  471. * congestion loss in non-ECN Capable TCP. That is, the TCP
  472. * source halves the congestion window "cwnd" and reduces the
  473. * slow start threshold "ssthresh".
  474. * A critical condition is that TCP does not react to
  475. * congestion indications more than once every window of
  476. * data (or more loosely more than once every round-trip time).
  477. */
  478. if (time_after(jiffies, transport->last_time_ecne_reduced +
  479. transport->rtt)) {
  480. transport->ssthresh = max(transport->cwnd/2,
  481. 4*asoc->pathmtu);
  482. transport->cwnd = transport->ssthresh;
  483. transport->last_time_ecne_reduced = jiffies;
  484. }
  485. break;
  486. case SCTP_LOWER_CWND_INACTIVE:
  487. /* RFC 2960 Section 7.2.1, sctpimpguide
  488. * When the endpoint does not transmit data on a given
  489. * transport address, the cwnd of the transport address
  490. * should be adjusted to max(cwnd/2, 4*MTU) per RTO.
  491. * NOTE: Although the draft recommends that this check needs
  492. * to be done every RTO interval, we do it every hearbeat
  493. * interval.
  494. */
  495. transport->cwnd = max(transport->cwnd/2,
  496. 4*asoc->pathmtu);
  497. /* RFC 4960 Errata 3.27.2: also adjust sshthresh */
  498. transport->ssthresh = transport->cwnd;
  499. break;
  500. }
  501. transport->partial_bytes_acked = 0;
  502. pr_debug("%s: transport:%p, reason:%d, cwnd:%d, ssthresh:%d\n",
  503. __func__, transport, reason, transport->cwnd,
  504. transport->ssthresh);
  505. }
  506. /* Apply Max.Burst limit to the congestion window:
  507. * sctpimpguide-05 2.14.2
  508. * D) When the time comes for the sender to
  509. * transmit new DATA chunks, the protocol parameter Max.Burst MUST
  510. * first be applied to limit how many new DATA chunks may be sent.
  511. * The limit is applied by adjusting cwnd as follows:
  512. * if ((flightsize+ Max.Burst * MTU) < cwnd)
  513. * cwnd = flightsize + Max.Burst * MTU
  514. */
  515. void sctp_transport_burst_limited(struct sctp_transport *t)
  516. {
  517. struct sctp_association *asoc = t->asoc;
  518. u32 old_cwnd = t->cwnd;
  519. u32 max_burst_bytes;
  520. if (t->burst_limited || asoc->max_burst == 0)
  521. return;
  522. max_burst_bytes = t->flight_size + (asoc->max_burst * asoc->pathmtu);
  523. if (max_burst_bytes < old_cwnd) {
  524. t->cwnd = max_burst_bytes;
  525. t->burst_limited = old_cwnd;
  526. }
  527. }
  528. /* Restore the old cwnd congestion window, after the burst had it's
  529. * desired effect.
  530. */
  531. void sctp_transport_burst_reset(struct sctp_transport *t)
  532. {
  533. if (t->burst_limited) {
  534. t->cwnd = t->burst_limited;
  535. t->burst_limited = 0;
  536. }
  537. }
  538. /* What is the next timeout value for this transport? */
  539. unsigned long sctp_transport_timeout(struct sctp_transport *trans)
  540. {
  541. /* RTO + timer slack +/- 50% of RTO */
  542. unsigned long timeout = trans->rto >> 1;
  543. if (trans->state != SCTP_UNCONFIRMED &&
  544. trans->state != SCTP_PF)
  545. timeout += trans->hbinterval;
  546. return max_t(unsigned long, timeout, HZ / 5);
  547. }
  548. /* Reset transport variables to their initial values */
  549. void sctp_transport_reset(struct sctp_transport *t)
  550. {
  551. struct sctp_association *asoc = t->asoc;
  552. /* RFC 2960 (bis), Section 5.2.4
  553. * All the congestion control parameters (e.g., cwnd, ssthresh)
  554. * related to this peer MUST be reset to their initial values
  555. * (see Section 6.2.1)
  556. */
  557. t->cwnd = min(4*asoc->pathmtu, max_t(__u32, 2*asoc->pathmtu, 4380));
  558. t->burst_limited = 0;
  559. t->ssthresh = asoc->peer.i.a_rwnd;
  560. t->rto = asoc->rto_initial;
  561. sctp_max_rto(asoc, t);
  562. t->rtt = 0;
  563. t->srtt = 0;
  564. t->rttvar = 0;
  565. /* Reset these additional variables so that we have a clean slate. */
  566. t->partial_bytes_acked = 0;
  567. t->flight_size = 0;
  568. t->error_count = 0;
  569. t->rto_pending = 0;
  570. t->hb_sent = 0;
  571. /* Initialize the state information for SFR-CACC */
  572. t->cacc.changeover_active = 0;
  573. t->cacc.cycling_changeover = 0;
  574. t->cacc.next_tsn_at_change = 0;
  575. t->cacc.cacc_saw_newack = 0;
  576. }
  577. /* Schedule retransmission on the given transport */
  578. void sctp_transport_immediate_rtx(struct sctp_transport *t)
  579. {
  580. /* Stop pending T3_rtx_timer */
  581. if (del_timer(&t->T3_rtx_timer))
  582. sctp_transport_put(t);
  583. sctp_retransmit(&t->asoc->outqueue, t, SCTP_RTXR_T3_RTX);
  584. if (!timer_pending(&t->T3_rtx_timer)) {
  585. if (!mod_timer(&t->T3_rtx_timer, jiffies + t->rto))
  586. sctp_transport_hold(t);
  587. }
  588. }
  589. /* Drop dst */
  590. void sctp_transport_dst_release(struct sctp_transport *t)
  591. {
  592. dst_release(t->dst);
  593. t->dst = NULL;
  594. t->dst_pending_confirm = 0;
  595. }
  596. /* Schedule neighbour confirm */
  597. void sctp_transport_dst_confirm(struct sctp_transport *t)
  598. {
  599. t->dst_pending_confirm = 1;
  600. }