transport.c 21 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712
  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_transport_put(transport);
  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. !timer_pending(&transport->hb_timer)) &&
  178. !mod_timer(&transport->hb_timer,
  179. expires + prandom_u32_max(transport->rto)))
  180. sctp_transport_hold(transport);
  181. }
  182. void sctp_transport_reset_reconf_timer(struct sctp_transport *transport)
  183. {
  184. if (!timer_pending(&transport->reconf_timer))
  185. if (!mod_timer(&transport->reconf_timer,
  186. jiffies + transport->rto))
  187. sctp_transport_hold(transport);
  188. }
  189. /* This transport has been assigned to an association.
  190. * Initialize fields from the association or from the sock itself.
  191. * Register the reference count in the association.
  192. */
  193. void sctp_transport_set_owner(struct sctp_transport *transport,
  194. struct sctp_association *asoc)
  195. {
  196. transport->asoc = asoc;
  197. sctp_association_hold(asoc);
  198. }
  199. /* Initialize the pmtu of a transport. */
  200. void sctp_transport_pmtu(struct sctp_transport *transport, struct sock *sk)
  201. {
  202. /* If we don't have a fresh route, look one up */
  203. if (!transport->dst || transport->dst->obsolete) {
  204. sctp_transport_dst_release(transport);
  205. transport->af_specific->get_dst(transport, &transport->saddr,
  206. &transport->fl, sk);
  207. }
  208. if (transport->dst) {
  209. transport->pathmtu = SCTP_TRUNC4(dst_mtu(transport->dst));
  210. } else
  211. transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
  212. }
  213. bool sctp_transport_update_pmtu(struct sctp_transport *t, u32 pmtu)
  214. {
  215. struct dst_entry *dst = sctp_transport_dst_check(t);
  216. struct sock *sk = t->asoc->base.sk;
  217. bool change = true;
  218. if (unlikely(pmtu < SCTP_DEFAULT_MINSEGMENT)) {
  219. pr_warn_ratelimited("%s: Reported pmtu %d too low, using default minimum of %d\n",
  220. __func__, pmtu, SCTP_DEFAULT_MINSEGMENT);
  221. /* Use default minimum segment instead */
  222. pmtu = SCTP_DEFAULT_MINSEGMENT;
  223. }
  224. pmtu = SCTP_TRUNC4(pmtu);
  225. if (dst) {
  226. struct sctp_pf *pf = sctp_get_pf_specific(dst->ops->family);
  227. union sctp_addr addr;
  228. pf->af->from_sk(&addr, sk);
  229. pf->to_sk_daddr(&t->ipaddr, sk);
  230. dst->ops->update_pmtu(dst, sk, NULL, pmtu, true);
  231. pf->to_sk_daddr(&addr, sk);
  232. dst = sctp_transport_dst_check(t);
  233. }
  234. if (!dst) {
  235. t->af_specific->get_dst(t, &t->saddr, &t->fl, sk);
  236. dst = t->dst;
  237. }
  238. if (dst) {
  239. /* Re-fetch, as under layers may have a higher minimum size */
  240. pmtu = SCTP_TRUNC4(dst_mtu(dst));
  241. change = t->pathmtu != pmtu;
  242. }
  243. t->pathmtu = pmtu;
  244. return change;
  245. }
  246. /* Caches the dst entry and source address for a transport's destination
  247. * address.
  248. */
  249. void sctp_transport_route(struct sctp_transport *transport,
  250. union sctp_addr *saddr, struct sctp_sock *opt)
  251. {
  252. struct sctp_association *asoc = transport->asoc;
  253. struct sctp_af *af = transport->af_specific;
  254. af->get_dst(transport, saddr, &transport->fl, sctp_opt2sk(opt));
  255. if (saddr)
  256. memcpy(&transport->saddr, saddr, sizeof(union sctp_addr));
  257. else
  258. af->get_saddr(opt, transport, &transport->fl);
  259. if ((transport->param_flags & SPP_PMTUD_DISABLE) && transport->pathmtu) {
  260. return;
  261. }
  262. if (transport->dst) {
  263. transport->pathmtu = SCTP_TRUNC4(dst_mtu(transport->dst));
  264. /* Initialize sk->sk_rcv_saddr, if the transport is the
  265. * association's active path for getsockname().
  266. */
  267. if (asoc && (!asoc->peer.primary_path ||
  268. (transport == asoc->peer.active_path)))
  269. opt->pf->to_sk_saddr(&transport->saddr,
  270. asoc->base.sk);
  271. } else
  272. transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
  273. }
  274. /* Hold a reference to a transport. */
  275. int sctp_transport_hold(struct sctp_transport *transport)
  276. {
  277. return refcount_inc_not_zero(&transport->refcnt);
  278. }
  279. /* Release a reference to a transport and clean up
  280. * if there are no more references.
  281. */
  282. void sctp_transport_put(struct sctp_transport *transport)
  283. {
  284. if (refcount_dec_and_test(&transport->refcnt))
  285. sctp_transport_destroy(transport);
  286. }
  287. /* Update transport's RTO based on the newly calculated RTT. */
  288. void sctp_transport_update_rto(struct sctp_transport *tp, __u32 rtt)
  289. {
  290. if (unlikely(!tp->rto_pending))
  291. /* We should not be doing any RTO updates unless rto_pending is set. */
  292. pr_debug("%s: rto_pending not set on transport %p!\n", __func__, tp);
  293. if (tp->rttvar || tp->srtt) {
  294. struct net *net = sock_net(tp->asoc->base.sk);
  295. /* 6.3.1 C3) When a new RTT measurement R' is made, set
  296. * RTTVAR <- (1 - RTO.Beta) * RTTVAR + RTO.Beta * |SRTT - R'|
  297. * SRTT <- (1 - RTO.Alpha) * SRTT + RTO.Alpha * R'
  298. */
  299. /* Note: The above algorithm has been rewritten to
  300. * express rto_beta and rto_alpha as inverse powers
  301. * of two.
  302. * For example, assuming the default value of RTO.Alpha of
  303. * 1/8, rto_alpha would be expressed as 3.
  304. */
  305. tp->rttvar = tp->rttvar - (tp->rttvar >> net->sctp.rto_beta)
  306. + (((__u32)abs((__s64)tp->srtt - (__s64)rtt)) >> net->sctp.rto_beta);
  307. tp->srtt = tp->srtt - (tp->srtt >> net->sctp.rto_alpha)
  308. + (rtt >> net->sctp.rto_alpha);
  309. } else {
  310. /* 6.3.1 C2) When the first RTT measurement R is made, set
  311. * SRTT <- R, RTTVAR <- R/2.
  312. */
  313. tp->srtt = rtt;
  314. tp->rttvar = rtt >> 1;
  315. }
  316. /* 6.3.1 G1) Whenever RTTVAR is computed, if RTTVAR = 0, then
  317. * adjust RTTVAR <- G, where G is the CLOCK GRANULARITY.
  318. */
  319. if (tp->rttvar == 0)
  320. tp->rttvar = SCTP_CLOCK_GRANULARITY;
  321. /* 6.3.1 C3) After the computation, update RTO <- SRTT + 4 * RTTVAR. */
  322. tp->rto = tp->srtt + (tp->rttvar << 2);
  323. /* 6.3.1 C6) Whenever RTO is computed, if it is less than RTO.Min
  324. * seconds then it is rounded up to RTO.Min seconds.
  325. */
  326. if (tp->rto < tp->asoc->rto_min)
  327. tp->rto = tp->asoc->rto_min;
  328. /* 6.3.1 C7) A maximum value may be placed on RTO provided it is
  329. * at least RTO.max seconds.
  330. */
  331. if (tp->rto > tp->asoc->rto_max)
  332. tp->rto = tp->asoc->rto_max;
  333. sctp_max_rto(tp->asoc, tp);
  334. tp->rtt = rtt;
  335. /* Reset rto_pending so that a new RTT measurement is started when a
  336. * new data chunk is sent.
  337. */
  338. tp->rto_pending = 0;
  339. pr_debug("%s: transport:%p, rtt:%d, srtt:%d rttvar:%d, rto:%ld\n",
  340. __func__, tp, rtt, tp->srtt, tp->rttvar, tp->rto);
  341. }
  342. /* This routine updates the transport's cwnd and partial_bytes_acked
  343. * parameters based on the bytes acked in the received SACK.
  344. */
  345. void sctp_transport_raise_cwnd(struct sctp_transport *transport,
  346. __u32 sack_ctsn, __u32 bytes_acked)
  347. {
  348. struct sctp_association *asoc = transport->asoc;
  349. __u32 cwnd, ssthresh, flight_size, pba, pmtu;
  350. cwnd = transport->cwnd;
  351. flight_size = transport->flight_size;
  352. /* See if we need to exit Fast Recovery first */
  353. if (asoc->fast_recovery &&
  354. TSN_lte(asoc->fast_recovery_exit, sack_ctsn))
  355. asoc->fast_recovery = 0;
  356. ssthresh = transport->ssthresh;
  357. pba = transport->partial_bytes_acked;
  358. pmtu = transport->asoc->pathmtu;
  359. if (cwnd <= ssthresh) {
  360. /* RFC 4960 7.2.1
  361. * o When cwnd is less than or equal to ssthresh, an SCTP
  362. * endpoint MUST use the slow-start algorithm to increase
  363. * cwnd only if the current congestion window is being fully
  364. * utilized, an incoming SACK advances the Cumulative TSN
  365. * Ack Point, and the data sender is not in Fast Recovery.
  366. * Only when these three conditions are met can the cwnd be
  367. * increased; otherwise, the cwnd MUST not be increased.
  368. * If these conditions are met, then cwnd MUST be increased
  369. * by, at most, the lesser of 1) the total size of the
  370. * previously outstanding DATA chunk(s) acknowledged, and
  371. * 2) the destination's path MTU. This upper bound protects
  372. * against the ACK-Splitting attack outlined in [SAVAGE99].
  373. */
  374. if (asoc->fast_recovery)
  375. return;
  376. /* The appropriate cwnd increase algorithm is performed
  377. * if, and only if the congestion window is being fully
  378. * utilized. Note that RFC4960 Errata 3.22 removed the
  379. * other condition on ctsn moving.
  380. */
  381. if (flight_size < cwnd)
  382. return;
  383. if (bytes_acked > pmtu)
  384. cwnd += pmtu;
  385. else
  386. cwnd += bytes_acked;
  387. pr_debug("%s: slow start: transport:%p, bytes_acked:%d, "
  388. "cwnd:%d, ssthresh:%d, flight_size:%d, pba:%d\n",
  389. __func__, transport, bytes_acked, cwnd, ssthresh,
  390. flight_size, pba);
  391. } else {
  392. /* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh,
  393. * upon each SACK arrival, increase partial_bytes_acked
  394. * by the total number of bytes of all new chunks
  395. * acknowledged in that SACK including chunks
  396. * acknowledged by the new Cumulative TSN Ack and by Gap
  397. * Ack Blocks. (updated by RFC4960 Errata 3.22)
  398. *
  399. * When partial_bytes_acked is greater than cwnd and
  400. * before the arrival of the SACK the sender had less
  401. * bytes of data outstanding than cwnd (i.e., before
  402. * arrival of the SACK, flightsize was less than cwnd),
  403. * reset partial_bytes_acked to cwnd. (RFC 4960 Errata
  404. * 3.26)
  405. *
  406. * When partial_bytes_acked is equal to or greater than
  407. * cwnd and before the arrival of the SACK the sender
  408. * had cwnd or more bytes of data outstanding (i.e.,
  409. * before arrival of the SACK, flightsize was greater
  410. * than or equal to cwnd), partial_bytes_acked is reset
  411. * to (partial_bytes_acked - cwnd). Next, cwnd is
  412. * increased by MTU. (RFC 4960 Errata 3.12)
  413. */
  414. pba += bytes_acked;
  415. if (pba > cwnd && flight_size < cwnd)
  416. pba = cwnd;
  417. if (pba >= cwnd && flight_size >= cwnd) {
  418. pba = pba - cwnd;
  419. cwnd += pmtu;
  420. }
  421. pr_debug("%s: congestion avoidance: transport:%p, "
  422. "bytes_acked:%d, cwnd:%d, ssthresh:%d, "
  423. "flight_size:%d, pba:%d\n", __func__,
  424. transport, bytes_acked, cwnd, ssthresh,
  425. flight_size, pba);
  426. }
  427. transport->cwnd = cwnd;
  428. transport->partial_bytes_acked = pba;
  429. }
  430. /* This routine is used to lower the transport's cwnd when congestion is
  431. * detected.
  432. */
  433. void sctp_transport_lower_cwnd(struct sctp_transport *transport,
  434. enum sctp_lower_cwnd reason)
  435. {
  436. struct sctp_association *asoc = transport->asoc;
  437. switch (reason) {
  438. case SCTP_LOWER_CWND_T3_RTX:
  439. /* RFC 2960 Section 7.2.3, sctpimpguide
  440. * When the T3-rtx timer expires on an address, SCTP should
  441. * perform slow start by:
  442. * ssthresh = max(cwnd/2, 4*MTU)
  443. * cwnd = 1*MTU
  444. * partial_bytes_acked = 0
  445. */
  446. transport->ssthresh = max(transport->cwnd/2,
  447. 4*asoc->pathmtu);
  448. transport->cwnd = asoc->pathmtu;
  449. /* T3-rtx also clears fast recovery */
  450. asoc->fast_recovery = 0;
  451. break;
  452. case SCTP_LOWER_CWND_FAST_RTX:
  453. /* RFC 2960 7.2.4 Adjust the ssthresh and cwnd of the
  454. * destination address(es) to which the missing DATA chunks
  455. * were last sent, according to the formula described in
  456. * Section 7.2.3.
  457. *
  458. * RFC 2960 7.2.3, sctpimpguide Upon detection of packet
  459. * losses from SACK (see Section 7.2.4), An endpoint
  460. * should do the following:
  461. * ssthresh = max(cwnd/2, 4*MTU)
  462. * cwnd = ssthresh
  463. * partial_bytes_acked = 0
  464. */
  465. if (asoc->fast_recovery)
  466. return;
  467. /* Mark Fast recovery */
  468. asoc->fast_recovery = 1;
  469. asoc->fast_recovery_exit = asoc->next_tsn - 1;
  470. transport->ssthresh = max(transport->cwnd/2,
  471. 4*asoc->pathmtu);
  472. transport->cwnd = transport->ssthresh;
  473. break;
  474. case SCTP_LOWER_CWND_ECNE:
  475. /* RFC 2481 Section 6.1.2.
  476. * If the sender receives an ECN-Echo ACK packet
  477. * then the sender knows that congestion was encountered in the
  478. * network on the path from the sender to the receiver. The
  479. * indication of congestion should be treated just as a
  480. * congestion loss in non-ECN Capable TCP. That is, the TCP
  481. * source halves the congestion window "cwnd" and reduces the
  482. * slow start threshold "ssthresh".
  483. * A critical condition is that TCP does not react to
  484. * congestion indications more than once every window of
  485. * data (or more loosely more than once every round-trip time).
  486. */
  487. if (time_after(jiffies, transport->last_time_ecne_reduced +
  488. transport->rtt)) {
  489. transport->ssthresh = max(transport->cwnd/2,
  490. 4*asoc->pathmtu);
  491. transport->cwnd = transport->ssthresh;
  492. transport->last_time_ecne_reduced = jiffies;
  493. }
  494. break;
  495. case SCTP_LOWER_CWND_INACTIVE:
  496. /* RFC 2960 Section 7.2.1, sctpimpguide
  497. * When the endpoint does not transmit data on a given
  498. * transport address, the cwnd of the transport address
  499. * should be adjusted to max(cwnd/2, 4*MTU) per RTO.
  500. * NOTE: Although the draft recommends that this check needs
  501. * to be done every RTO interval, we do it every hearbeat
  502. * interval.
  503. */
  504. transport->cwnd = max(transport->cwnd/2,
  505. 4*asoc->pathmtu);
  506. /* RFC 4960 Errata 3.27.2: also adjust sshthresh */
  507. transport->ssthresh = transport->cwnd;
  508. break;
  509. }
  510. transport->partial_bytes_acked = 0;
  511. pr_debug("%s: transport:%p, reason:%d, cwnd:%d, ssthresh:%d\n",
  512. __func__, transport, reason, transport->cwnd,
  513. transport->ssthresh);
  514. }
  515. /* Apply Max.Burst limit to the congestion window:
  516. * sctpimpguide-05 2.14.2
  517. * D) When the time comes for the sender to
  518. * transmit new DATA chunks, the protocol parameter Max.Burst MUST
  519. * first be applied to limit how many new DATA chunks may be sent.
  520. * The limit is applied by adjusting cwnd as follows:
  521. * if ((flightsize+ Max.Burst * MTU) < cwnd)
  522. * cwnd = flightsize + Max.Burst * MTU
  523. */
  524. void sctp_transport_burst_limited(struct sctp_transport *t)
  525. {
  526. struct sctp_association *asoc = t->asoc;
  527. u32 old_cwnd = t->cwnd;
  528. u32 max_burst_bytes;
  529. if (t->burst_limited || asoc->max_burst == 0)
  530. return;
  531. max_burst_bytes = t->flight_size + (asoc->max_burst * asoc->pathmtu);
  532. if (max_burst_bytes < old_cwnd) {
  533. t->cwnd = max_burst_bytes;
  534. t->burst_limited = old_cwnd;
  535. }
  536. }
  537. /* Restore the old cwnd congestion window, after the burst had it's
  538. * desired effect.
  539. */
  540. void sctp_transport_burst_reset(struct sctp_transport *t)
  541. {
  542. if (t->burst_limited) {
  543. t->cwnd = t->burst_limited;
  544. t->burst_limited = 0;
  545. }
  546. }
  547. /* What is the next timeout value for this transport? */
  548. unsigned long sctp_transport_timeout(struct sctp_transport *trans)
  549. {
  550. /* RTO + timer slack +/- 50% of RTO */
  551. unsigned long timeout = trans->rto >> 1;
  552. if (trans->state != SCTP_UNCONFIRMED &&
  553. trans->state != SCTP_PF)
  554. timeout += trans->hbinterval;
  555. return max_t(unsigned long, timeout, HZ / 5);
  556. }
  557. /* Reset transport variables to their initial values */
  558. void sctp_transport_reset(struct sctp_transport *t)
  559. {
  560. struct sctp_association *asoc = t->asoc;
  561. /* RFC 2960 (bis), Section 5.2.4
  562. * All the congestion control parameters (e.g., cwnd, ssthresh)
  563. * related to this peer MUST be reset to their initial values
  564. * (see Section 6.2.1)
  565. */
  566. t->cwnd = min(4*asoc->pathmtu, max_t(__u32, 2*asoc->pathmtu, 4380));
  567. t->burst_limited = 0;
  568. t->ssthresh = asoc->peer.i.a_rwnd;
  569. t->rto = asoc->rto_initial;
  570. sctp_max_rto(asoc, t);
  571. t->rtt = 0;
  572. t->srtt = 0;
  573. t->rttvar = 0;
  574. /* Reset these additional variables so that we have a clean slate. */
  575. t->partial_bytes_acked = 0;
  576. t->flight_size = 0;
  577. t->error_count = 0;
  578. t->rto_pending = 0;
  579. t->hb_sent = 0;
  580. /* Initialize the state information for SFR-CACC */
  581. t->cacc.changeover_active = 0;
  582. t->cacc.cycling_changeover = 0;
  583. t->cacc.next_tsn_at_change = 0;
  584. t->cacc.cacc_saw_newack = 0;
  585. }
  586. /* Schedule retransmission on the given transport */
  587. void sctp_transport_immediate_rtx(struct sctp_transport *t)
  588. {
  589. /* Stop pending T3_rtx_timer */
  590. if (del_timer(&t->T3_rtx_timer))
  591. sctp_transport_put(t);
  592. sctp_retransmit(&t->asoc->outqueue, t, SCTP_RTXR_T3_RTX);
  593. if (!timer_pending(&t->T3_rtx_timer)) {
  594. if (!mod_timer(&t->T3_rtx_timer, jiffies + t->rto))
  595. sctp_transport_hold(t);
  596. }
  597. }
  598. /* Drop dst */
  599. void sctp_transport_dst_release(struct sctp_transport *t)
  600. {
  601. dst_release(t->dst);
  602. t->dst = NULL;
  603. t->dst_pending_confirm = 0;
  604. }
  605. /* Schedule neighbour confirm */
  606. void sctp_transport_dst_confirm(struct sctp_transport *t)
  607. {
  608. t->dst_pending_confirm = 1;
  609. }