blob: e745c118f2396e19bdd35387ed27551aeef4e252 [file] [log] [blame]
Vlad Yasevich60c778b2008-01-11 09:57:09 -05001/* SCTP kernel implementation
Linus Torvalds1da177e2005-04-16 15:20:36 -07002 * 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 *
Vlad Yasevich60c778b2008-01-11 09:57:09 -05008 * This file is part of the SCTP kernel implementation
Linus Torvalds1da177e2005-04-16 15:20:36 -07009 *
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 *
Vlad Yasevich60c778b2008-01-11 09:57:09 -050014 * This SCTP implementation is free software;
Linus Torvalds1da177e2005-04-16 15:20:36 -070015 * 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 *
Vlad Yasevich60c778b2008-01-11 09:57:09 -050020 * This SCTP implementation is distributed in the hope that it
Linus Torvalds1da177e2005-04-16 15:20:36 -070021 * 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, write to
28 * the Free Software Foundation, 59 Temple Place - Suite 330,
29 * Boston, MA 02111-1307, USA.
30 *
31 * Please send any bug reports or fixes you make to the
32 * email address(es):
33 * lksctp developers <lksctp-developers@lists.sourceforge.net>
34 *
35 * Or submit a bug report through the following website:
36 * http://www.sf.net/projects/lksctp
37 *
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Karl Knutson <karl@athena.chicago.il.us>
41 * Jon Grimm <jgrimm@us.ibm.com>
42 * Xingang Guo <xingang.guo@intel.com>
43 * Hui Huang <hui.huang@nokia.com>
44 * Sridhar Samudrala <sri@us.ibm.com>
45 * Ardelle Fan <ardelle.fan@intel.com>
46 *
47 * Any bugs reported given to us we will try to fix... any fixes shared will
48 * be incorporated into the next SCTP release.
49 */
50
51#include <linux/types.h>
Sridhar Samudralaad8fec12006-07-21 14:48:50 -070052#include <linux/random.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070053#include <net/sctp/sctp.h>
54#include <net/sctp/sm.h>
55
56/* 1st Level Abstractions. */
57
58/* Initialize a new transport from provided memory. */
59static struct sctp_transport *sctp_transport_init(struct sctp_transport *peer,
60 const union sctp_addr *addr,
Al Virodd0fc662005-10-07 07:46:04 +010061 gfp_t gfp)
Linus Torvalds1da177e2005-04-16 15:20:36 -070062{
63 /* Copy in the address. */
64 peer->ipaddr = *addr;
65 peer->af_specific = sctp_get_af_specific(addr->sa.sa_family);
66 peer->asoc = NULL;
67
68 peer->dst = NULL;
69 memset(&peer->saddr, 0, sizeof(union sctp_addr));
70
71 /* From 6.3.1 RTO Calculation:
72 *
73 * C1) Until an RTT measurement has been made for a packet sent to the
74 * given destination transport address, set RTO to the protocol
75 * parameter 'RTO.Initial'.
76 */
Vlad Yasevichb6157d82007-10-24 15:59:16 -040077 peer->last_rto = peer->rto = msecs_to_jiffies(sctp_rto_initial);
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 peer->rtt = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -070079 peer->rttvar = 0;
80 peer->srtt = 0;
81 peer->rto_pending = 0;
Vlad Yasevicha6465232008-06-04 12:38:43 -070082 peer->fast_recovery = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -070083
84 peer->last_time_heard = jiffies;
85 peer->last_time_used = jiffies;
86 peer->last_time_ecne_reduced = jiffies;
87
Frank Filz3f7a87d2005-06-20 13:14:57 -070088 peer->init_sent_count = 0;
89
Frank Filz52ccb8e2005-12-22 11:36:46 -080090 peer->param_flags = SPP_HB_DISABLE |
91 SPP_PMTUD_ENABLE |
92 SPP_SACKDELAY_ENABLE;
93 peer->hbinterval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -070094
95 /* Initialize the default path max_retrans. */
Frank Filz52ccb8e2005-12-22 11:36:46 -080096 peer->pathmaxrxt = sctp_max_retrans_path;
Linus Torvalds1da177e2005-04-16 15:20:36 -070097 peer->error_count = 0;
98
99 INIT_LIST_HEAD(&peer->transmitted);
100 INIT_LIST_HEAD(&peer->send_ready);
101 INIT_LIST_HEAD(&peer->transports);
102
Florian Westphal6d0ccba2008-07-18 23:04:39 -0700103 peer->T3_rtx_timer.expires = 0;
104 peer->hb_timer.expires = 0;
105
Pavel Emelyanovb24b8a22008-01-23 21:20:07 -0800106 setup_timer(&peer->T3_rtx_timer, sctp_generate_t3_rtx_event,
107 (unsigned long)peer);
108 setup_timer(&peer->hb_timer, sctp_generate_heartbeat_event,
109 (unsigned long)peer);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110
Sridhar Samudralaad8fec12006-07-21 14:48:50 -0700111 /* Initialize the 64-bit random nonce sent with heartbeat. */
112 get_random_bytes(&peer->hb_nonce, sizeof(peer->hb_nonce));
113
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114 atomic_set(&peer->refcnt, 1);
115 peer->dead = 0;
116
117 peer->malloced = 0;
118
119 /* Initialize the state information for SFR-CACC */
120 peer->cacc.changeover_active = 0;
121 peer->cacc.cycling_changeover = 0;
122 peer->cacc.next_tsn_at_change = 0;
123 peer->cacc.cacc_saw_newack = 0;
124
125 return peer;
126}
127
128/* Allocate and initialize a new transport. */
Alexey Dobriyan3182cd82005-07-11 20:57:47 -0700129struct sctp_transport *sctp_transport_new(const union sctp_addr *addr,
Al Virodd0fc662005-10-07 07:46:04 +0100130 gfp_t gfp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131{
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900132 struct sctp_transport *transport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900134 transport = t_new(struct sctp_transport, gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135 if (!transport)
136 goto fail;
137
138 if (!sctp_transport_init(transport, addr, gfp))
139 goto fail_init;
140
141 transport->malloced = 1;
142 SCTP_DBG_OBJCNT_INC(transport);
143
144 return transport;
145
146fail_init:
147 kfree(transport);
148
149fail:
150 return NULL;
151}
152
153/* This transport is no longer needed. Free up if possible, or
154 * delay until it last reference count.
155 */
156void sctp_transport_free(struct sctp_transport *transport)
157{
158 transport->dead = 1;
159
160 /* Try to delete the heartbeat timer. */
161 if (del_timer(&transport->hb_timer))
162 sctp_transport_put(transport);
163
164 /* Delete the T3_rtx timer if it's active.
165 * There is no point in not doing this now and letting
166 * structure hang around in memory since we know
167 * the tranport is going away.
168 */
169 if (timer_pending(&transport->T3_rtx_timer) &&
170 del_timer(&transport->T3_rtx_timer))
171 sctp_transport_put(transport);
172
173
174 sctp_transport_put(transport);
175}
176
177/* Destroy the transport data structure.
178 * Assumes there are no more users of this structure.
179 */
180static void sctp_transport_destroy(struct sctp_transport *transport)
181{
182 SCTP_ASSERT(transport->dead, "Transport is not dead", return);
183
184 if (transport->asoc)
185 sctp_association_put(transport->asoc);
186
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900187 sctp_packet_free(&transport->packet);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188
189 dst_release(transport->dst);
190 kfree(transport);
191 SCTP_DBG_OBJCNT_DEC(transport);
192}
193
194/* Start T3_rtx timer if it is not already running and update the heartbeat
195 * timer. This routine is called every time a DATA chunk is sent.
196 */
Vlad Yasevich62aeaff2008-06-04 12:39:11 -0700197void sctp_transport_reset_timers(struct sctp_transport *transport, int force)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198{
199 /* RFC 2960 6.3.2 Retransmission Timer Rules
200 *
201 * R1) Every time a DATA chunk is sent to any address(including a
202 * retransmission), if the T3-rtx timer of that address is not running
203 * start it running so that it will expire after the RTO of that
204 * address.
205 */
206
Vlad Yasevich62aeaff2008-06-04 12:39:11 -0700207 if (force || !timer_pending(&transport->T3_rtx_timer))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 if (!mod_timer(&transport->T3_rtx_timer,
209 jiffies + transport->rto))
210 sctp_transport_hold(transport);
211
212 /* When a data chunk is sent, reset the heartbeat interval. */
213 if (!mod_timer(&transport->hb_timer,
214 sctp_transport_timeout(transport)))
215 sctp_transport_hold(transport);
216}
217
218/* This transport has been assigned to an association.
219 * Initialize fields from the association or from the sock itself.
220 * Register the reference count in the association.
221 */
222void sctp_transport_set_owner(struct sctp_transport *transport,
223 struct sctp_association *asoc)
224{
225 transport->asoc = asoc;
226 sctp_association_hold(asoc);
227}
228
229/* Initialize the pmtu of a transport. */
230void sctp_transport_pmtu(struct sctp_transport *transport)
231{
232 struct dst_entry *dst;
233
234 dst = transport->af_specific->get_dst(NULL, &transport->ipaddr, NULL);
235
236 if (dst) {
Frank Filz52ccb8e2005-12-22 11:36:46 -0800237 transport->pathmtu = dst_mtu(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 dst_release(dst);
239 } else
Frank Filz52ccb8e2005-12-22 11:36:46 -0800240 transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241}
242
Vlad Yasevichc910b472007-06-07 13:47:03 -0400243/* this is a complete rip-off from __sk_dst_check
244 * the cookie is always 0 since this is how it's used in the
245 * pmtu code
246 */
247static struct dst_entry *sctp_transport_dst_check(struct sctp_transport *t)
248{
249 struct dst_entry *dst = t->dst;
250
251 if (dst && dst->obsolete && dst->ops->check(dst, 0) == NULL) {
252 dst_release(t->dst);
253 t->dst = NULL;
254 return NULL;
255 }
256
257 return dst;
258}
259
260void sctp_transport_update_pmtu(struct sctp_transport *t, u32 pmtu)
261{
262 struct dst_entry *dst;
263
264 if (unlikely(pmtu < SCTP_DEFAULT_MINSEGMENT)) {
265 printk(KERN_WARNING "%s: Reported pmtu %d too low, "
266 "using default minimum of %d\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -0800267 __func__, pmtu,
Vlad Yasevichc910b472007-06-07 13:47:03 -0400268 SCTP_DEFAULT_MINSEGMENT);
269 /* Use default minimum segment size and disable
270 * pmtu discovery on this transport.
271 */
272 t->pathmtu = SCTP_DEFAULT_MINSEGMENT;
Vlad Yasevichc910b472007-06-07 13:47:03 -0400273 } else {
274 t->pathmtu = pmtu;
275 }
276
277 dst = sctp_transport_dst_check(t);
278 if (dst)
279 dst->ops->update_pmtu(dst, pmtu);
280}
281
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282/* Caches the dst entry and source address for a transport's destination
283 * address.
284 */
285void sctp_transport_route(struct sctp_transport *transport,
286 union sctp_addr *saddr, struct sctp_sock *opt)
287{
288 struct sctp_association *asoc = transport->asoc;
289 struct sctp_af *af = transport->af_specific;
290 union sctp_addr *daddr = &transport->ipaddr;
291 struct dst_entry *dst;
292
293 dst = af->get_dst(asoc, daddr, saddr);
294
295 if (saddr)
296 memcpy(&transport->saddr, saddr, sizeof(union sctp_addr));
297 else
YOSHIFUJI Hideakie5117102008-05-29 19:55:05 +0900298 af->get_saddr(opt, asoc, dst, daddr, &transport->saddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299
300 transport->dst = dst;
Frank Filz52ccb8e2005-12-22 11:36:46 -0800301 if ((transport->param_flags & SPP_PMTUD_DISABLE) && transport->pathmtu) {
302 return;
303 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 if (dst) {
Frank Filz52ccb8e2005-12-22 11:36:46 -0800305 transport->pathmtu = dst_mtu(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306
307 /* Initialize sk->sk_rcv_saddr, if the transport is the
308 * association's active path for getsockname().
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900309 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 if (asoc && (transport == asoc->peer.active_path))
Neil Hormanbf031ff2005-12-02 20:32:29 -0800311 opt->pf->af->to_sk_saddr(&transport->saddr,
312 asoc->base.sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313 } else
Frank Filz52ccb8e2005-12-22 11:36:46 -0800314 transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315}
316
317/* Hold a reference to a transport. */
318void sctp_transport_hold(struct sctp_transport *transport)
319{
320 atomic_inc(&transport->refcnt);
321}
322
323/* Release a reference to a transport and clean up
324 * if there are no more references.
325 */
326void sctp_transport_put(struct sctp_transport *transport)
327{
328 if (atomic_dec_and_test(&transport->refcnt))
329 sctp_transport_destroy(transport);
330}
331
332/* Update transport's RTO based on the newly calculated RTT. */
333void sctp_transport_update_rto(struct sctp_transport *tp, __u32 rtt)
334{
335 /* Check for valid transport. */
336 SCTP_ASSERT(tp, "NULL transport", return);
337
338 /* We should not be doing any RTO updates unless rto_pending is set. */
339 SCTP_ASSERT(tp->rto_pending, "rto_pending not set", return);
340
341 if (tp->rttvar || tp->srtt) {
342 /* 6.3.1 C3) When a new RTT measurement R' is made, set
343 * RTTVAR <- (1 - RTO.Beta) * RTTVAR + RTO.Beta * |SRTT - R'|
344 * SRTT <- (1 - RTO.Alpha) * SRTT + RTO.Alpha * R'
345 */
346
347 /* Note: The above algorithm has been rewritten to
348 * express rto_beta and rto_alpha as inverse powers
349 * of two.
350 * For example, assuming the default value of RTO.Alpha of
351 * 1/8, rto_alpha would be expressed as 3.
352 */
353 tp->rttvar = tp->rttvar - (tp->rttvar >> sctp_rto_beta)
354 + ((abs(tp->srtt - rtt)) >> sctp_rto_beta);
355 tp->srtt = tp->srtt - (tp->srtt >> sctp_rto_alpha)
356 + (rtt >> sctp_rto_alpha);
357 } else {
358 /* 6.3.1 C2) When the first RTT measurement R is made, set
359 * SRTT <- R, RTTVAR <- R/2.
360 */
361 tp->srtt = rtt;
362 tp->rttvar = rtt >> 1;
363 }
364
365 /* 6.3.1 G1) Whenever RTTVAR is computed, if RTTVAR = 0, then
366 * adjust RTTVAR <- G, where G is the CLOCK GRANULARITY.
367 */
368 if (tp->rttvar == 0)
369 tp->rttvar = SCTP_CLOCK_GRANULARITY;
370
371 /* 6.3.1 C3) After the computation, update RTO <- SRTT + 4 * RTTVAR. */
372 tp->rto = tp->srtt + (tp->rttvar << 2);
373
374 /* 6.3.1 C6) Whenever RTO is computed, if it is less than RTO.Min
375 * seconds then it is rounded up to RTO.Min seconds.
376 */
377 if (tp->rto < tp->asoc->rto_min)
378 tp->rto = tp->asoc->rto_min;
379
380 /* 6.3.1 C7) A maximum value may be placed on RTO provided it is
381 * at least RTO.max seconds.
382 */
383 if (tp->rto > tp->asoc->rto_max)
384 tp->rto = tp->asoc->rto_max;
385
386 tp->rtt = rtt;
Vlad Yasevichb6157d82007-10-24 15:59:16 -0400387 tp->last_rto = tp->rto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388
389 /* Reset rto_pending so that a new RTT measurement is started when a
390 * new data chunk is sent.
391 */
392 tp->rto_pending = 0;
393
394 SCTP_DEBUG_PRINTK("%s: transport: %p, rtt: %d, srtt: %d "
Harvey Harrison0dc47872008-03-05 20:47:47 -0800395 "rttvar: %d, rto: %ld\n", __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 tp, rtt, tp->srtt, tp->rttvar, tp->rto);
397}
398
399/* This routine updates the transport's cwnd and partial_bytes_acked
400 * parameters based on the bytes acked in the received SACK.
401 */
402void sctp_transport_raise_cwnd(struct sctp_transport *transport,
403 __u32 sack_ctsn, __u32 bytes_acked)
404{
405 __u32 cwnd, ssthresh, flight_size, pba, pmtu;
406
407 cwnd = transport->cwnd;
408 flight_size = transport->flight_size;
409
Vlad Yasevicha6465232008-06-04 12:38:43 -0700410 /* See if we need to exit Fast Recovery first */
411 if (transport->fast_recovery &&
412 TSN_lte(transport->fast_recovery_exit, sack_ctsn))
413 transport->fast_recovery = 0;
414
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 /* The appropriate cwnd increase algorithm is performed if, and only
Vlad Yasevicha6465232008-06-04 12:38:43 -0700416 * if the cumulative TSN whould advanced and the congestion window is
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 * being fully utilized.
418 */
Vlad Yasevicha6465232008-06-04 12:38:43 -0700419 if (TSN_lte(sack_ctsn, transport->asoc->ctsn_ack_point) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 (flight_size < cwnd))
421 return;
422
423 ssthresh = transport->ssthresh;
424 pba = transport->partial_bytes_acked;
Frank Filz52ccb8e2005-12-22 11:36:46 -0800425 pmtu = transport->asoc->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426
427 if (cwnd <= ssthresh) {
Vlad Yasevicha6465232008-06-04 12:38:43 -0700428 /* RFC 4960 7.2.1
429 * o When cwnd is less than or equal to ssthresh, an SCTP
430 * endpoint MUST use the slow-start algorithm to increase
431 * cwnd only if the current congestion window is being fully
432 * utilized, an incoming SACK advances the Cumulative TSN
433 * Ack Point, and the data sender is not in Fast Recovery.
434 * Only when these three conditions are met can the cwnd be
435 * increased; otherwise, the cwnd MUST not be increased.
436 * If these conditions are met, then cwnd MUST be increased
437 * by, at most, the lesser of 1) the total size of the
438 * previously outstanding DATA chunk(s) acknowledged, and
439 * 2) the destination's path MTU. This upper bound protects
440 * against the ACK-Splitting attack outlined in [SAVAGE99].
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 */
Vlad Yasevicha6465232008-06-04 12:38:43 -0700442 if (transport->fast_recovery)
443 return;
444
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 if (bytes_acked > pmtu)
446 cwnd += pmtu;
447 else
448 cwnd += bytes_acked;
449 SCTP_DEBUG_PRINTK("%s: SLOW START: transport: %p, "
450 "bytes_acked: %d, cwnd: %d, ssthresh: %d, "
451 "flight_size: %d, pba: %d\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -0800452 __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 transport, bytes_acked, cwnd,
454 ssthresh, flight_size, pba);
455 } else {
456 /* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh,
457 * upon each SACK arrival that advances the Cumulative TSN Ack
458 * Point, increase partial_bytes_acked by the total number of
459 * bytes of all new chunks acknowledged in that SACK including
460 * chunks acknowledged by the new Cumulative TSN Ack and by
461 * Gap Ack Blocks.
462 *
463 * When partial_bytes_acked is equal to or greater than cwnd
464 * and before the arrival of the SACK the sender had cwnd or
465 * more bytes of data outstanding (i.e., before arrival of the
466 * SACK, flightsize was greater than or equal to cwnd),
467 * increase cwnd by MTU, and reset partial_bytes_acked to
468 * (partial_bytes_acked - cwnd).
469 */
470 pba += bytes_acked;
471 if (pba >= cwnd) {
472 cwnd += pmtu;
473 pba = ((cwnd < pba) ? (pba - cwnd) : 0);
474 }
475 SCTP_DEBUG_PRINTK("%s: CONGESTION AVOIDANCE: "
476 "transport: %p, bytes_acked: %d, cwnd: %d, "
477 "ssthresh: %d, flight_size: %d, pba: %d\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -0800478 __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 transport, bytes_acked, cwnd,
480 ssthresh, flight_size, pba);
481 }
482
483 transport->cwnd = cwnd;
484 transport->partial_bytes_acked = pba;
485}
486
487/* This routine is used to lower the transport's cwnd when congestion is
488 * detected.
489 */
490void sctp_transport_lower_cwnd(struct sctp_transport *transport,
491 sctp_lower_cwnd_t reason)
492{
493 switch (reason) {
494 case SCTP_LOWER_CWND_T3_RTX:
495 /* RFC 2960 Section 7.2.3, sctpimpguide
496 * When the T3-rtx timer expires on an address, SCTP should
497 * perform slow start by:
498 * ssthresh = max(cwnd/2, 4*MTU)
499 * cwnd = 1*MTU
500 * partial_bytes_acked = 0
501 */
502 transport->ssthresh = max(transport->cwnd/2,
Frank Filz52ccb8e2005-12-22 11:36:46 -0800503 4*transport->asoc->pathmtu);
504 transport->cwnd = transport->asoc->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505 break;
506
507 case SCTP_LOWER_CWND_FAST_RTX:
508 /* RFC 2960 7.2.4 Adjust the ssthresh and cwnd of the
509 * destination address(es) to which the missing DATA chunks
510 * were last sent, according to the formula described in
511 * Section 7.2.3.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900512 *
513 * RFC 2960 7.2.3, sctpimpguide Upon detection of packet
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 * losses from SACK (see Section 7.2.4), An endpoint
515 * should do the following:
516 * ssthresh = max(cwnd/2, 4*MTU)
517 * cwnd = ssthresh
518 * partial_bytes_acked = 0
519 */
Vlad Yasevicha6465232008-06-04 12:38:43 -0700520 if (transport->fast_recovery)
521 return;
522
523 /* Mark Fast recovery */
524 transport->fast_recovery = 1;
525 transport->fast_recovery_exit = transport->asoc->next_tsn - 1;
526
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 transport->ssthresh = max(transport->cwnd/2,
Frank Filz52ccb8e2005-12-22 11:36:46 -0800528 4*transport->asoc->pathmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 transport->cwnd = transport->ssthresh;
530 break;
531
532 case SCTP_LOWER_CWND_ECNE:
533 /* RFC 2481 Section 6.1.2.
534 * If the sender receives an ECN-Echo ACK packet
535 * then the sender knows that congestion was encountered in the
536 * network on the path from the sender to the receiver. The
537 * indication of congestion should be treated just as a
538 * congestion loss in non-ECN Capable TCP. That is, the TCP
539 * source halves the congestion window "cwnd" and reduces the
540 * slow start threshold "ssthresh".
541 * A critical condition is that TCP does not react to
542 * congestion indications more than once every window of
543 * data (or more loosely more than once every round-trip time).
544 */
545 if ((jiffies - transport->last_time_ecne_reduced) >
546 transport->rtt) {
547 transport->ssthresh = max(transport->cwnd/2,
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900548 4*transport->asoc->pathmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 transport->cwnd = transport->ssthresh;
550 transport->last_time_ecne_reduced = jiffies;
551 }
552 break;
553
554 case SCTP_LOWER_CWND_INACTIVE:
555 /* RFC 2960 Section 7.2.1, sctpimpguide
556 * When the endpoint does not transmit data on a given
557 * transport address, the cwnd of the transport address
558 * should be adjusted to max(cwnd/2, 4*MTU) per RTO.
559 * NOTE: Although the draft recommends that this check needs
560 * to be done every RTO interval, we do it every hearbeat
561 * interval.
562 */
563 if ((jiffies - transport->last_time_used) > transport->rto)
564 transport->cwnd = max(transport->cwnd/2,
Frank Filz52ccb8e2005-12-22 11:36:46 -0800565 4*transport->asoc->pathmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 break;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -0700567 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568
569 transport->partial_bytes_acked = 0;
570 SCTP_DEBUG_PRINTK("%s: transport: %p reason: %d cwnd: "
Harvey Harrison0dc47872008-03-05 20:47:47 -0800571 "%d ssthresh: %d\n", __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 transport, reason,
573 transport->cwnd, transport->ssthresh);
574}
575
576/* What is the next timeout value for this transport? */
577unsigned long sctp_transport_timeout(struct sctp_transport *t)
578{
579 unsigned long timeout;
Sridhar Samudralaad8fec12006-07-21 14:48:50 -0700580 timeout = t->rto + sctp_jitter(t->rto);
581 if (t->state != SCTP_UNCONFIRMED)
582 timeout += t->hbinterval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 timeout += jiffies;
584 return timeout;
585}
Vlad Yasevich749bf922007-03-19 17:02:30 -0700586
587/* Reset transport variables to their initial values */
588void sctp_transport_reset(struct sctp_transport *t)
589{
590 struct sctp_association *asoc = t->asoc;
591
592 /* RFC 2960 (bis), Section 5.2.4
593 * All the congestion control parameters (e.g., cwnd, ssthresh)
594 * related to this peer MUST be reset to their initial values
595 * (see Section 6.2.1)
596 */
597 t->cwnd = min(4*asoc->pathmtu, max_t(__u32, 2*asoc->pathmtu, 4380));
Vlad Yasevich289f4242007-03-22 12:26:25 -0700598 t->ssthresh = asoc->peer.i.a_rwnd;
Vlad Yasevichb6157d82007-10-24 15:59:16 -0400599 t->last_rto = t->rto = asoc->rto_initial;
Vlad Yasevich749bf922007-03-19 17:02:30 -0700600 t->rtt = 0;
601 t->srtt = 0;
602 t->rttvar = 0;
603
604 /* Reset these additional varibles so that we have a clean
605 * slate.
606 */
607 t->partial_bytes_acked = 0;
608 t->flight_size = 0;
609 t->error_count = 0;
610 t->rto_pending = 0;
Vlad Yasevicha6465232008-06-04 12:38:43 -0700611 t->fast_recovery = 0;
Vlad Yasevich749bf922007-03-19 17:02:30 -0700612
613 /* Initialize the state information for SFR-CACC */
614 t->cacc.changeover_active = 0;
615 t->cacc.cycling_changeover = 0;
616 t->cacc.next_tsn_at_change = 0;
617 t->cacc.cacc_saw_newack = 0;
618}