xref: /linux/net/sctp/transport.c (revision 74ce1896c6c65b2f8cccbf59162d542988835835)
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 
44 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
45 
46 #include <linux/slab.h>
47 #include <linux/types.h>
48 #include <linux/random.h>
49 #include <net/sctp/sctp.h>
50 #include <net/sctp/sm.h>
51 
52 /* 1st Level Abstractions.  */
53 
54 /* Initialize a new transport from provided memory.  */
55 static struct sctp_transport *sctp_transport_init(struct net *net,
56 						  struct sctp_transport *peer,
57 						  const union sctp_addr *addr,
58 						  gfp_t gfp)
59 {
60 	/* Copy in the address.  */
61 	peer->ipaddr = *addr;
62 	peer->af_specific = sctp_get_af_specific(addr->sa.sa_family);
63 	memset(&peer->saddr, 0, sizeof(union sctp_addr));
64 
65 	peer->sack_generation = 0;
66 
67 	/* From 6.3.1 RTO Calculation:
68 	 *
69 	 * C1) Until an RTT measurement has been made for a packet sent to the
70 	 * given destination transport address, set RTO to the protocol
71 	 * parameter 'RTO.Initial'.
72 	 */
73 	peer->rto = msecs_to_jiffies(net->sctp.rto_initial);
74 
75 	peer->last_time_heard = 0;
76 	peer->last_time_ecne_reduced = jiffies;
77 
78 	peer->param_flags = SPP_HB_DISABLE |
79 			    SPP_PMTUD_ENABLE |
80 			    SPP_SACKDELAY_ENABLE;
81 
82 	/* Initialize the default path max_retrans.  */
83 	peer->pathmaxrxt  = net->sctp.max_retrans_path;
84 	peer->pf_retrans  = net->sctp.pf_retrans;
85 
86 	INIT_LIST_HEAD(&peer->transmitted);
87 	INIT_LIST_HEAD(&peer->send_ready);
88 	INIT_LIST_HEAD(&peer->transports);
89 
90 	setup_timer(&peer->T3_rtx_timer, sctp_generate_t3_rtx_event,
91 		    (unsigned long)peer);
92 	setup_timer(&peer->hb_timer, sctp_generate_heartbeat_event,
93 		    (unsigned long)peer);
94 	setup_timer(&peer->reconf_timer, sctp_generate_reconf_event,
95 		    (unsigned long)peer);
96 	setup_timer(&peer->proto_unreach_timer,
97 		    sctp_generate_proto_unreach_event, (unsigned long)peer);
98 
99 	/* Initialize the 64-bit random nonce sent with heartbeat. */
100 	get_random_bytes(&peer->hb_nonce, sizeof(peer->hb_nonce));
101 
102 	refcount_set(&peer->refcnt, 1);
103 
104 	return peer;
105 }
106 
107 /* Allocate and initialize a new transport.  */
108 struct sctp_transport *sctp_transport_new(struct net *net,
109 					  const union sctp_addr *addr,
110 					  gfp_t gfp)
111 {
112 	struct sctp_transport *transport;
113 
114 	transport = kzalloc(sizeof(*transport), gfp);
115 	if (!transport)
116 		goto fail;
117 
118 	if (!sctp_transport_init(net, transport, addr, gfp))
119 		goto fail_init;
120 
121 	SCTP_DBG_OBJCNT_INC(transport);
122 
123 	return transport;
124 
125 fail_init:
126 	kfree(transport);
127 
128 fail:
129 	return NULL;
130 }
131 
132 /* This transport is no longer needed.  Free up if possible, or
133  * delay until it last reference count.
134  */
135 void sctp_transport_free(struct sctp_transport *transport)
136 {
137 	/* Try to delete the heartbeat timer.  */
138 	if (del_timer(&transport->hb_timer))
139 		sctp_transport_put(transport);
140 
141 	/* Delete the T3_rtx timer if it's active.
142 	 * There is no point in not doing this now and letting
143 	 * structure hang around in memory since we know
144 	 * the tranport is going away.
145 	 */
146 	if (del_timer(&transport->T3_rtx_timer))
147 		sctp_transport_put(transport);
148 
149 	if (del_timer(&transport->reconf_timer))
150 		sctp_transport_put(transport);
151 
152 	/* Delete the ICMP proto unreachable timer if it's active. */
153 	if (del_timer(&transport->proto_unreach_timer))
154 		sctp_association_put(transport->asoc);
155 
156 	sctp_transport_put(transport);
157 }
158 
159 static void sctp_transport_destroy_rcu(struct rcu_head *head)
160 {
161 	struct sctp_transport *transport;
162 
163 	transport = container_of(head, struct sctp_transport, rcu);
164 
165 	dst_release(transport->dst);
166 	kfree(transport);
167 	SCTP_DBG_OBJCNT_DEC(transport);
168 }
169 
170 /* Destroy the transport data structure.
171  * Assumes there are no more users of this structure.
172  */
173 static void sctp_transport_destroy(struct sctp_transport *transport)
174 {
175 	if (unlikely(refcount_read(&transport->refcnt))) {
176 		WARN(1, "Attempt to destroy undead transport %p!\n", transport);
177 		return;
178 	}
179 
180 	sctp_packet_free(&transport->packet);
181 
182 	if (transport->asoc)
183 		sctp_association_put(transport->asoc);
184 
185 	call_rcu(&transport->rcu, sctp_transport_destroy_rcu);
186 }
187 
188 /* Start T3_rtx timer if it is not already running and update the heartbeat
189  * timer.  This routine is called every time a DATA chunk is sent.
190  */
191 void sctp_transport_reset_t3_rtx(struct sctp_transport *transport)
192 {
193 	/* RFC 2960 6.3.2 Retransmission Timer Rules
194 	 *
195 	 * R1) Every time a DATA chunk is sent to any address(including a
196 	 * retransmission), if the T3-rtx timer of that address is not running
197 	 * start it running so that it will expire after the RTO of that
198 	 * address.
199 	 */
200 
201 	if (!timer_pending(&transport->T3_rtx_timer))
202 		if (!mod_timer(&transport->T3_rtx_timer,
203 			       jiffies + transport->rto))
204 			sctp_transport_hold(transport);
205 }
206 
207 void sctp_transport_reset_hb_timer(struct sctp_transport *transport)
208 {
209 	unsigned long expires;
210 
211 	/* When a data chunk is sent, reset the heartbeat interval.  */
212 	expires = jiffies + sctp_transport_timeout(transport);
213 	if (time_before(transport->hb_timer.expires, expires) &&
214 	    !mod_timer(&transport->hb_timer,
215 		       expires + prandom_u32_max(transport->rto)))
216 		sctp_transport_hold(transport);
217 }
218 
219 void sctp_transport_reset_reconf_timer(struct sctp_transport *transport)
220 {
221 	if (!timer_pending(&transport->reconf_timer))
222 		if (!mod_timer(&transport->reconf_timer,
223 			       jiffies + transport->rto))
224 			sctp_transport_hold(transport);
225 }
226 
227 /* This transport has been assigned to an association.
228  * Initialize fields from the association or from the sock itself.
229  * Register the reference count in the association.
230  */
231 void sctp_transport_set_owner(struct sctp_transport *transport,
232 			      struct sctp_association *asoc)
233 {
234 	transport->asoc = asoc;
235 	sctp_association_hold(asoc);
236 }
237 
238 /* Initialize the pmtu of a transport. */
239 void sctp_transport_pmtu(struct sctp_transport *transport, struct sock *sk)
240 {
241 	/* If we don't have a fresh route, look one up */
242 	if (!transport->dst || transport->dst->obsolete) {
243 		sctp_transport_dst_release(transport);
244 		transport->af_specific->get_dst(transport, &transport->saddr,
245 						&transport->fl, sk);
246 	}
247 
248 	if (transport->dst) {
249 		transport->pathmtu = SCTP_TRUNC4(dst_mtu(transport->dst));
250 	} else
251 		transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
252 }
253 
254 void sctp_transport_update_pmtu(struct sctp_transport *t, u32 pmtu)
255 {
256 	struct dst_entry *dst = sctp_transport_dst_check(t);
257 
258 	if (unlikely(pmtu < SCTP_DEFAULT_MINSEGMENT)) {
259 		pr_warn("%s: Reported pmtu %d too low, using default minimum of %d\n",
260 			__func__, pmtu, SCTP_DEFAULT_MINSEGMENT);
261 		/* Use default minimum segment size and disable
262 		 * pmtu discovery on this transport.
263 		 */
264 		t->pathmtu = SCTP_DEFAULT_MINSEGMENT;
265 	} else {
266 		t->pathmtu = pmtu;
267 	}
268 
269 	if (dst) {
270 		dst->ops->update_pmtu(dst, t->asoc->base.sk, NULL, pmtu);
271 		dst = sctp_transport_dst_check(t);
272 	}
273 
274 	if (!dst)
275 		t->af_specific->get_dst(t, &t->saddr, &t->fl, t->asoc->base.sk);
276 }
277 
278 /* Caches the dst entry and source address for a transport's destination
279  * address.
280  */
281 void sctp_transport_route(struct sctp_transport *transport,
282 			  union sctp_addr *saddr, struct sctp_sock *opt)
283 {
284 	struct sctp_association *asoc = transport->asoc;
285 	struct sctp_af *af = transport->af_specific;
286 
287 	af->get_dst(transport, saddr, &transport->fl, sctp_opt2sk(opt));
288 
289 	if (saddr)
290 		memcpy(&transport->saddr, saddr, sizeof(union sctp_addr));
291 	else
292 		af->get_saddr(opt, transport, &transport->fl);
293 
294 	if ((transport->param_flags & SPP_PMTUD_DISABLE) && transport->pathmtu) {
295 		return;
296 	}
297 	if (transport->dst) {
298 		transport->pathmtu = SCTP_TRUNC4(dst_mtu(transport->dst));
299 
300 		/* Initialize sk->sk_rcv_saddr, if the transport is the
301 		 * association's active path for getsockname().
302 		 */
303 		if (asoc && (!asoc->peer.primary_path ||
304 				(transport == asoc->peer.active_path)))
305 			opt->pf->to_sk_saddr(&transport->saddr,
306 					     asoc->base.sk);
307 	} else
308 		transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
309 }
310 
311 /* Hold a reference to a transport.  */
312 int sctp_transport_hold(struct sctp_transport *transport)
313 {
314 	return refcount_inc_not_zero(&transport->refcnt);
315 }
316 
317 /* Release a reference to a transport and clean up
318  * if there are no more references.
319  */
320 void sctp_transport_put(struct sctp_transport *transport)
321 {
322 	if (refcount_dec_and_test(&transport->refcnt))
323 		sctp_transport_destroy(transport);
324 }
325 
326 /* Update transport's RTO based on the newly calculated RTT. */
327 void sctp_transport_update_rto(struct sctp_transport *tp, __u32 rtt)
328 {
329 	if (unlikely(!tp->rto_pending))
330 		/* We should not be doing any RTO updates unless rto_pending is set.  */
331 		pr_debug("%s: rto_pending not set on transport %p!\n", __func__, tp);
332 
333 	if (tp->rttvar || tp->srtt) {
334 		struct net *net = sock_net(tp->asoc->base.sk);
335 		/* 6.3.1 C3) When a new RTT measurement R' is made, set
336 		 * RTTVAR <- (1 - RTO.Beta) * RTTVAR + RTO.Beta * |SRTT - R'|
337 		 * SRTT <- (1 - RTO.Alpha) * SRTT + RTO.Alpha * R'
338 		 */
339 
340 		/* Note:  The above algorithm has been rewritten to
341 		 * express rto_beta and rto_alpha as inverse powers
342 		 * of two.
343 		 * For example, assuming the default value of RTO.Alpha of
344 		 * 1/8, rto_alpha would be expressed as 3.
345 		 */
346 		tp->rttvar = tp->rttvar - (tp->rttvar >> net->sctp.rto_beta)
347 			+ (((__u32)abs((__s64)tp->srtt - (__s64)rtt)) >> net->sctp.rto_beta);
348 		tp->srtt = tp->srtt - (tp->srtt >> net->sctp.rto_alpha)
349 			+ (rtt >> net->sctp.rto_alpha);
350 	} else {
351 		/* 6.3.1 C2) When the first RTT measurement R is made, set
352 		 * SRTT <- R, RTTVAR <- R/2.
353 		 */
354 		tp->srtt = rtt;
355 		tp->rttvar = rtt >> 1;
356 	}
357 
358 	/* 6.3.1 G1) Whenever RTTVAR is computed, if RTTVAR = 0, then
359 	 * adjust RTTVAR <- G, where G is the CLOCK GRANULARITY.
360 	 */
361 	if (tp->rttvar == 0)
362 		tp->rttvar = SCTP_CLOCK_GRANULARITY;
363 
364 	/* 6.3.1 C3) After the computation, update RTO <- SRTT + 4 * RTTVAR. */
365 	tp->rto = tp->srtt + (tp->rttvar << 2);
366 
367 	/* 6.3.1 C6) Whenever RTO is computed, if it is less than RTO.Min
368 	 * seconds then it is rounded up to RTO.Min seconds.
369 	 */
370 	if (tp->rto < tp->asoc->rto_min)
371 		tp->rto = tp->asoc->rto_min;
372 
373 	/* 6.3.1 C7) A maximum value may be placed on RTO provided it is
374 	 * at least RTO.max seconds.
375 	 */
376 	if (tp->rto > tp->asoc->rto_max)
377 		tp->rto = tp->asoc->rto_max;
378 
379 	sctp_max_rto(tp->asoc, tp);
380 	tp->rtt = rtt;
381 
382 	/* Reset rto_pending so that a new RTT measurement is started when a
383 	 * new data chunk is sent.
384 	 */
385 	tp->rto_pending = 0;
386 
387 	pr_debug("%s: transport:%p, rtt:%d, srtt:%d rttvar:%d, rto:%ld\n",
388 		 __func__, tp, rtt, tp->srtt, tp->rttvar, tp->rto);
389 }
390 
391 /* This routine updates the transport's cwnd and partial_bytes_acked
392  * parameters based on the bytes acked in the received SACK.
393  */
394 void sctp_transport_raise_cwnd(struct sctp_transport *transport,
395 			       __u32 sack_ctsn, __u32 bytes_acked)
396 {
397 	struct sctp_association *asoc = transport->asoc;
398 	__u32 cwnd, ssthresh, flight_size, pba, pmtu;
399 
400 	cwnd = transport->cwnd;
401 	flight_size = transport->flight_size;
402 
403 	/* See if we need to exit Fast Recovery first */
404 	if (asoc->fast_recovery &&
405 	    TSN_lte(asoc->fast_recovery_exit, sack_ctsn))
406 		asoc->fast_recovery = 0;
407 
408 	ssthresh = transport->ssthresh;
409 	pba = transport->partial_bytes_acked;
410 	pmtu = transport->asoc->pathmtu;
411 
412 	if (cwnd <= ssthresh) {
413 		/* RFC 4960 7.2.1
414 		 * o  When cwnd is less than or equal to ssthresh, an SCTP
415 		 *    endpoint MUST use the slow-start algorithm to increase
416 		 *    cwnd only if the current congestion window is being fully
417 		 *    utilized, an incoming SACK advances the Cumulative TSN
418 		 *    Ack Point, and the data sender is not in Fast Recovery.
419 		 *    Only when these three conditions are met can the cwnd be
420 		 *    increased; otherwise, the cwnd MUST not be increased.
421 		 *    If these conditions are met, then cwnd MUST be increased
422 		 *    by, at most, the lesser of 1) the total size of the
423 		 *    previously outstanding DATA chunk(s) acknowledged, and
424 		 *    2) the destination's path MTU.  This upper bound protects
425 		 *    against the ACK-Splitting attack outlined in [SAVAGE99].
426 		 */
427 		if (asoc->fast_recovery)
428 			return;
429 
430 		/* The appropriate cwnd increase algorithm is performed
431 		 * if, and only if the congestion window is being fully
432 		 * utilized.  Note that RFC4960 Errata 3.22 removed the
433 		 * other condition on ctsn moving.
434 		 */
435 		if (flight_size < cwnd)
436 			return;
437 
438 		if (bytes_acked > pmtu)
439 			cwnd += pmtu;
440 		else
441 			cwnd += bytes_acked;
442 
443 		pr_debug("%s: slow start: transport:%p, bytes_acked:%d, "
444 			 "cwnd:%d, ssthresh:%d, flight_size:%d, pba:%d\n",
445 			 __func__, transport, bytes_acked, cwnd, ssthresh,
446 			 flight_size, pba);
447 	} else {
448 		/* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh,
449 		 * upon each SACK arrival, increase partial_bytes_acked
450 		 * by the total number of bytes of all new chunks
451 		 * acknowledged in that SACK including chunks
452 		 * acknowledged by the new Cumulative TSN Ack and by Gap
453 		 * Ack Blocks. (updated by RFC4960 Errata 3.22)
454 		 *
455 		 * When partial_bytes_acked is greater than cwnd and
456 		 * before the arrival of the SACK the sender had less
457 		 * bytes of data outstanding than cwnd (i.e., before
458 		 * arrival of the SACK, flightsize was less than cwnd),
459 		 * reset partial_bytes_acked to cwnd. (RFC 4960 Errata
460 		 * 3.26)
461 		 *
462 		 * When partial_bytes_acked is equal to or greater than
463 		 * cwnd and before the arrival of the SACK the sender
464 		 * had cwnd or more bytes of data outstanding (i.e.,
465 		 * before arrival of the SACK, flightsize was greater
466 		 * than or equal to cwnd), partial_bytes_acked is reset
467 		 * to (partial_bytes_acked - cwnd). Next, cwnd is
468 		 * increased by MTU. (RFC 4960 Errata 3.12)
469 		 */
470 		pba += bytes_acked;
471 		if (pba > cwnd && flight_size < cwnd)
472 			pba = cwnd;
473 		if (pba >= cwnd && flight_size >= cwnd) {
474 			pba = pba - cwnd;
475 			cwnd += pmtu;
476 		}
477 
478 		pr_debug("%s: congestion avoidance: transport:%p, "
479 			 "bytes_acked:%d, cwnd:%d, ssthresh:%d, "
480 			 "flight_size:%d, pba:%d\n", __func__,
481 			 transport, bytes_acked, cwnd, ssthresh,
482 			 flight_size, pba);
483 	}
484 
485 	transport->cwnd = cwnd;
486 	transport->partial_bytes_acked = pba;
487 }
488 
489 /* This routine is used to lower the transport's cwnd when congestion is
490  * detected.
491  */
492 void sctp_transport_lower_cwnd(struct sctp_transport *transport,
493 			       enum sctp_lower_cwnd reason)
494 {
495 	struct sctp_association *asoc = transport->asoc;
496 
497 	switch (reason) {
498 	case SCTP_LOWER_CWND_T3_RTX:
499 		/* RFC 2960 Section 7.2.3, sctpimpguide
500 		 * When the T3-rtx timer expires on an address, SCTP should
501 		 * perform slow start by:
502 		 *      ssthresh = max(cwnd/2, 4*MTU)
503 		 *      cwnd = 1*MTU
504 		 *      partial_bytes_acked = 0
505 		 */
506 		transport->ssthresh = max(transport->cwnd/2,
507 					  4*asoc->pathmtu);
508 		transport->cwnd = asoc->pathmtu;
509 
510 		/* T3-rtx also clears fast recovery */
511 		asoc->fast_recovery = 0;
512 		break;
513 
514 	case SCTP_LOWER_CWND_FAST_RTX:
515 		/* RFC 2960 7.2.4 Adjust the ssthresh and cwnd of the
516 		 * destination address(es) to which the missing DATA chunks
517 		 * were last sent, according to the formula described in
518 		 * Section 7.2.3.
519 		 *
520 		 * RFC 2960 7.2.3, sctpimpguide Upon detection of packet
521 		 * losses from SACK (see Section 7.2.4), An endpoint
522 		 * should do the following:
523 		 *      ssthresh = max(cwnd/2, 4*MTU)
524 		 *      cwnd = ssthresh
525 		 *      partial_bytes_acked = 0
526 		 */
527 		if (asoc->fast_recovery)
528 			return;
529 
530 		/* Mark Fast recovery */
531 		asoc->fast_recovery = 1;
532 		asoc->fast_recovery_exit = asoc->next_tsn - 1;
533 
534 		transport->ssthresh = max(transport->cwnd/2,
535 					  4*asoc->pathmtu);
536 		transport->cwnd = transport->ssthresh;
537 		break;
538 
539 	case SCTP_LOWER_CWND_ECNE:
540 		/* RFC 2481 Section 6.1.2.
541 		 * If the sender receives an ECN-Echo ACK packet
542 		 * then the sender knows that congestion was encountered in the
543 		 * network on the path from the sender to the receiver. The
544 		 * indication of congestion should be treated just as a
545 		 * congestion loss in non-ECN Capable TCP. That is, the TCP
546 		 * source halves the congestion window "cwnd" and reduces the
547 		 * slow start threshold "ssthresh".
548 		 * A critical condition is that TCP does not react to
549 		 * congestion indications more than once every window of
550 		 * data (or more loosely more than once every round-trip time).
551 		 */
552 		if (time_after(jiffies, transport->last_time_ecne_reduced +
553 					transport->rtt)) {
554 			transport->ssthresh = max(transport->cwnd/2,
555 						  4*asoc->pathmtu);
556 			transport->cwnd = transport->ssthresh;
557 			transport->last_time_ecne_reduced = jiffies;
558 		}
559 		break;
560 
561 	case SCTP_LOWER_CWND_INACTIVE:
562 		/* RFC 2960 Section 7.2.1, sctpimpguide
563 		 * When the endpoint does not transmit data on a given
564 		 * transport address, the cwnd of the transport address
565 		 * should be adjusted to max(cwnd/2, 4*MTU) per RTO.
566 		 * NOTE: Although the draft recommends that this check needs
567 		 * to be done every RTO interval, we do it every hearbeat
568 		 * interval.
569 		 */
570 		transport->cwnd = max(transport->cwnd/2,
571 					 4*asoc->pathmtu);
572 		/* RFC 4960 Errata 3.27.2: also adjust sshthresh */
573 		transport->ssthresh = transport->cwnd;
574 		break;
575 	}
576 
577 	transport->partial_bytes_acked = 0;
578 
579 	pr_debug("%s: transport:%p, reason:%d, cwnd:%d, ssthresh:%d\n",
580 		 __func__, transport, reason, transport->cwnd,
581 		 transport->ssthresh);
582 }
583 
584 /* Apply Max.Burst limit to the congestion window:
585  * sctpimpguide-05 2.14.2
586  * D) When the time comes for the sender to
587  * transmit new DATA chunks, the protocol parameter Max.Burst MUST
588  * first be applied to limit how many new DATA chunks may be sent.
589  * The limit is applied by adjusting cwnd as follows:
590  * 	if ((flightsize+ Max.Burst * MTU) < cwnd)
591  * 		cwnd = flightsize + Max.Burst * MTU
592  */
593 
594 void sctp_transport_burst_limited(struct sctp_transport *t)
595 {
596 	struct sctp_association *asoc = t->asoc;
597 	u32 old_cwnd = t->cwnd;
598 	u32 max_burst_bytes;
599 
600 	if (t->burst_limited || asoc->max_burst == 0)
601 		return;
602 
603 	max_burst_bytes = t->flight_size + (asoc->max_burst * asoc->pathmtu);
604 	if (max_burst_bytes < old_cwnd) {
605 		t->cwnd = max_burst_bytes;
606 		t->burst_limited = old_cwnd;
607 	}
608 }
609 
610 /* Restore the old cwnd congestion window, after the burst had it's
611  * desired effect.
612  */
613 void sctp_transport_burst_reset(struct sctp_transport *t)
614 {
615 	if (t->burst_limited) {
616 		t->cwnd = t->burst_limited;
617 		t->burst_limited = 0;
618 	}
619 }
620 
621 /* What is the next timeout value for this transport? */
622 unsigned long sctp_transport_timeout(struct sctp_transport *trans)
623 {
624 	/* RTO + timer slack +/- 50% of RTO */
625 	unsigned long timeout = trans->rto >> 1;
626 
627 	if (trans->state != SCTP_UNCONFIRMED &&
628 	    trans->state != SCTP_PF)
629 		timeout += trans->hbinterval;
630 
631 	return timeout;
632 }
633 
634 /* Reset transport variables to their initial values */
635 void sctp_transport_reset(struct sctp_transport *t)
636 {
637 	struct sctp_association *asoc = t->asoc;
638 
639 	/* RFC 2960 (bis), Section 5.2.4
640 	 * All the congestion control parameters (e.g., cwnd, ssthresh)
641 	 * related to this peer MUST be reset to their initial values
642 	 * (see Section 6.2.1)
643 	 */
644 	t->cwnd = min(4*asoc->pathmtu, max_t(__u32, 2*asoc->pathmtu, 4380));
645 	t->burst_limited = 0;
646 	t->ssthresh = asoc->peer.i.a_rwnd;
647 	t->rto = asoc->rto_initial;
648 	sctp_max_rto(asoc, t);
649 	t->rtt = 0;
650 	t->srtt = 0;
651 	t->rttvar = 0;
652 
653 	/* Reset these additional variables so that we have a clean slate. */
654 	t->partial_bytes_acked = 0;
655 	t->flight_size = 0;
656 	t->error_count = 0;
657 	t->rto_pending = 0;
658 	t->hb_sent = 0;
659 
660 	/* Initialize the state information for SFR-CACC */
661 	t->cacc.changeover_active = 0;
662 	t->cacc.cycling_changeover = 0;
663 	t->cacc.next_tsn_at_change = 0;
664 	t->cacc.cacc_saw_newack = 0;
665 }
666 
667 /* Schedule retransmission on the given transport */
668 void sctp_transport_immediate_rtx(struct sctp_transport *t)
669 {
670 	/* Stop pending T3_rtx_timer */
671 	if (del_timer(&t->T3_rtx_timer))
672 		sctp_transport_put(t);
673 
674 	sctp_retransmit(&t->asoc->outqueue, t, SCTP_RTXR_T3_RTX);
675 	if (!timer_pending(&t->T3_rtx_timer)) {
676 		if (!mod_timer(&t->T3_rtx_timer, jiffies + t->rto))
677 			sctp_transport_hold(t);
678 	}
679 }
680 
681 /* Drop dst */
682 void sctp_transport_dst_release(struct sctp_transport *t)
683 {
684 	dst_release(t->dst);
685 	t->dst = NULL;
686 	t->dst_pending_confirm = 0;
687 }
688 
689 /* Schedule neighbour confirm */
690 void sctp_transport_dst_confirm(struct sctp_transport *t)
691 {
692 	t->dst_pending_confirm = 1;
693 }
694