blob: b026ba0c69922e09eb5c150298f650ea7bb3d1c4 [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
Joe Perches145ce502010-08-24 13:21:08 +000051#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
52
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090053#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070054#include <linux/types.h>
Sridhar Samudralaad8fec12006-07-21 14:48:50 -070055#include <linux/random.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070056#include <net/sctp/sctp.h>
57#include <net/sctp/sm.h>
58
59/* 1st Level Abstractions. */
60
61/* Initialize a new transport from provided memory. */
62static struct sctp_transport *sctp_transport_init(struct sctp_transport *peer,
63 const union sctp_addr *addr,
Al Virodd0fc662005-10-07 07:46:04 +010064 gfp_t gfp)
Linus Torvalds1da177e2005-04-16 15:20:36 -070065{
66 /* Copy in the address. */
67 peer->ipaddr = *addr;
68 peer->af_specific = sctp_get_af_specific(addr->sa.sa_family);
Linus Torvalds1da177e2005-04-16 15:20:36 -070069 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 */
Andrei Pelinescu-Onciul5fdd4ba2009-11-29 00:14:02 -080077 peer->rto = msecs_to_jiffies(sctp_rto_initial);
Linus Torvalds1da177e2005-04-16 15:20:36 -070078
79 peer->last_time_heard = jiffies;
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 peer->last_time_ecne_reduced = jiffies;
81
Frank Filz52ccb8e2005-12-22 11:36:46 -080082 peer->param_flags = SPP_HB_DISABLE |
83 SPP_PMTUD_ENABLE |
84 SPP_SACKDELAY_ENABLE;
Linus Torvalds1da177e2005-04-16 15:20:36 -070085
86 /* Initialize the default path max_retrans. */
Frank Filz52ccb8e2005-12-22 11:36:46 -080087 peer->pathmaxrxt = sctp_max_retrans_path;
Linus Torvalds1da177e2005-04-16 15:20:36 -070088
89 INIT_LIST_HEAD(&peer->transmitted);
90 INIT_LIST_HEAD(&peer->send_ready);
91 INIT_LIST_HEAD(&peer->transports);
92
Pavel Emelyanovb24b8a22008-01-23 21:20:07 -080093 setup_timer(&peer->T3_rtx_timer, sctp_generate_t3_rtx_event,
94 (unsigned long)peer);
95 setup_timer(&peer->hb_timer, sctp_generate_heartbeat_event,
96 (unsigned long)peer);
Vlad Yasevich50b5d6a2010-05-06 00:56:07 -070097 setup_timer(&peer->proto_unreach_timer,
98 sctp_generate_proto_unreach_event, (unsigned long)peer);
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
Sridhar Samudralaad8fec12006-07-21 14:48:50 -0700100 /* Initialize the 64-bit random nonce sent with heartbeat. */
101 get_random_bytes(&peer->hb_nonce, sizeof(peer->hb_nonce));
102
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103 atomic_set(&peer->refcnt, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104
105 return peer;
106}
107
108/* Allocate and initialize a new transport. */
Alexey Dobriyan3182cd842005-07-11 20:57:47 -0700109struct sctp_transport *sctp_transport_new(const union sctp_addr *addr,
Al Virodd0fc662005-10-07 07:46:04 +0100110 gfp_t gfp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111{
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900112 struct sctp_transport *transport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900114 transport = t_new(struct sctp_transport, gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115 if (!transport)
116 goto fail;
117
118 if (!sctp_transport_init(transport, addr, gfp))
119 goto fail_init;
120
121 transport->malloced = 1;
122 SCTP_DBG_OBJCNT_INC(transport);
123
124 return transport;
125
126fail_init:
127 kfree(transport);
128
129fail:
130 return NULL;
131}
132
133/* This transport is no longer needed. Free up if possible, or
134 * delay until it last reference count.
135 */
136void sctp_transport_free(struct sctp_transport *transport)
137{
138 transport->dead = 1;
139
140 /* Try to delete the heartbeat timer. */
141 if (del_timer(&transport->hb_timer))
142 sctp_transport_put(transport);
143
144 /* Delete the T3_rtx timer if it's active.
145 * There is no point in not doing this now and letting
146 * structure hang around in memory since we know
147 * the tranport is going away.
148 */
149 if (timer_pending(&transport->T3_rtx_timer) &&
150 del_timer(&transport->T3_rtx_timer))
151 sctp_transport_put(transport);
152
Wei Yongjun55fa0cf2010-05-09 16:56:07 +0000153 /* Delete the ICMP proto unreachable timer if it's active. */
154 if (timer_pending(&transport->proto_unreach_timer) &&
155 del_timer(&transport->proto_unreach_timer))
156 sctp_association_put(transport->asoc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157
158 sctp_transport_put(transport);
159}
160
161/* Destroy the transport data structure.
162 * Assumes there are no more users of this structure.
163 */
164static void sctp_transport_destroy(struct sctp_transport *transport)
165{
166 SCTP_ASSERT(transport->dead, "Transport is not dead", return);
167
168 if (transport->asoc)
169 sctp_association_put(transport->asoc);
170
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900171 sctp_packet_free(&transport->packet);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172
173 dst_release(transport->dst);
174 kfree(transport);
175 SCTP_DBG_OBJCNT_DEC(transport);
176}
177
178/* Start T3_rtx timer if it is not already running and update the heartbeat
179 * timer. This routine is called every time a DATA chunk is sent.
180 */
Vlad Yasevichd9efc222010-04-30 22:41:09 -0400181void sctp_transport_reset_timers(struct sctp_transport *transport)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182{
183 /* RFC 2960 6.3.2 Retransmission Timer Rules
184 *
185 * R1) Every time a DATA chunk is sent to any address(including a
186 * retransmission), if the T3-rtx timer of that address is not running
187 * start it running so that it will expire after the RTO of that
188 * address.
189 */
190
Vlad Yasevichd9efc222010-04-30 22:41:09 -0400191 if (!timer_pending(&transport->T3_rtx_timer))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 if (!mod_timer(&transport->T3_rtx_timer,
193 jiffies + transport->rto))
194 sctp_transport_hold(transport);
195
196 /* When a data chunk is sent, reset the heartbeat interval. */
197 if (!mod_timer(&transport->hb_timer,
198 sctp_transport_timeout(transport)))
199 sctp_transport_hold(transport);
200}
201
202/* This transport has been assigned to an association.
203 * Initialize fields from the association or from the sock itself.
204 * Register the reference count in the association.
205 */
206void sctp_transport_set_owner(struct sctp_transport *transport,
207 struct sctp_association *asoc)
208{
209 transport->asoc = asoc;
210 sctp_association_hold(asoc);
211}
212
213/* Initialize the pmtu of a transport. */
Vlad Yasevich9914ae32011-04-26 21:51:31 +0000214void sctp_transport_pmtu(struct sctp_transport *transport, struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215{
Vlad Yasevichda0420b2011-04-26 21:54:17 +0000216 /* If we don't have a fresh route, look one up */
217 if (!transport->dst || transport->dst->obsolete > 1) {
218 dst_release(transport->dst);
219 transport->af_specific->get_dst(transport, &transport->saddr,
David S. Miller8663c932011-05-06 16:32:47 -0700220 &transport->fl, sk);
Vlad Yasevichda0420b2011-04-26 21:54:17 +0000221 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222
Vlad Yasevichda0420b2011-04-26 21:54:17 +0000223 if (transport->dst) {
224 transport->pathmtu = dst_mtu(transport->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 } else
Frank Filz52ccb8e2005-12-22 11:36:46 -0800226 transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227}
228
Vlad Yasevichc910b472007-06-07 13:47:03 -0400229void sctp_transport_update_pmtu(struct sctp_transport *t, u32 pmtu)
230{
231 struct dst_entry *dst;
232
233 if (unlikely(pmtu < SCTP_DEFAULT_MINSEGMENT)) {
Joe Perches145ce502010-08-24 13:21:08 +0000234 pr_warn("%s: Reported pmtu %d too low, using default minimum of %d\n",
235 __func__, pmtu,
236 SCTP_DEFAULT_MINSEGMENT);
Vlad Yasevichc910b472007-06-07 13:47:03 -0400237 /* Use default minimum segment size and disable
238 * pmtu discovery on this transport.
239 */
240 t->pathmtu = SCTP_DEFAULT_MINSEGMENT;
Vlad Yasevichc910b472007-06-07 13:47:03 -0400241 } else {
242 t->pathmtu = pmtu;
243 }
244
245 dst = sctp_transport_dst_check(t);
246 if (dst)
247 dst->ops->update_pmtu(dst, pmtu);
248}
249
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250/* Caches the dst entry and source address for a transport's destination
251 * address.
252 */
253void sctp_transport_route(struct sctp_transport *transport,
254 union sctp_addr *saddr, struct sctp_sock *opt)
255{
256 struct sctp_association *asoc = transport->asoc;
257 struct sctp_af *af = transport->af_specific;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258
David S. Miller8663c932011-05-06 16:32:47 -0700259 af->get_dst(transport, saddr, &transport->fl, sctp_opt2sk(opt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260
261 if (saddr)
262 memcpy(&transport->saddr, saddr, sizeof(union sctp_addr));
263 else
David S. Miller8663c932011-05-06 16:32:47 -0700264 af->get_saddr(opt, transport, &transport->fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265
Frank Filz52ccb8e2005-12-22 11:36:46 -0800266 if ((transport->param_flags & SPP_PMTUD_DISABLE) && transport->pathmtu) {
267 return;
268 }
Vlad Yasevichda0420b2011-04-26 21:54:17 +0000269 if (transport->dst) {
270 transport->pathmtu = dst_mtu(transport->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271
272 /* Initialize sk->sk_rcv_saddr, if the transport is the
273 * association's active path for getsockname().
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900274 */
Vlad Yasevicha78102e2009-11-11 11:54:37 +0000275 if (asoc && (!asoc->peer.primary_path ||
276 (transport == asoc->peer.active_path)))
Neil Hormanbf031ff2005-12-02 20:32:29 -0800277 opt->pf->af->to_sk_saddr(&transport->saddr,
278 asoc->base.sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 } else
Frank Filz52ccb8e2005-12-22 11:36:46 -0800280 transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281}
282
283/* Hold a reference to a transport. */
284void sctp_transport_hold(struct sctp_transport *transport)
285{
286 atomic_inc(&transport->refcnt);
287}
288
289/* Release a reference to a transport and clean up
290 * if there are no more references.
291 */
292void sctp_transport_put(struct sctp_transport *transport)
293{
294 if (atomic_dec_and_test(&transport->refcnt))
295 sctp_transport_destroy(transport);
296}
297
298/* Update transport's RTO based on the newly calculated RTT. */
299void sctp_transport_update_rto(struct sctp_transport *tp, __u32 rtt)
300{
301 /* Check for valid transport. */
302 SCTP_ASSERT(tp, "NULL transport", return);
303
304 /* We should not be doing any RTO updates unless rto_pending is set. */
305 SCTP_ASSERT(tp->rto_pending, "rto_pending not set", return);
306
307 if (tp->rttvar || tp->srtt) {
308 /* 6.3.1 C3) When a new RTT measurement R' is made, set
309 * RTTVAR <- (1 - RTO.Beta) * RTTVAR + RTO.Beta * |SRTT - R'|
310 * SRTT <- (1 - RTO.Alpha) * SRTT + RTO.Alpha * R'
311 */
312
313 /* Note: The above algorithm has been rewritten to
314 * express rto_beta and rto_alpha as inverse powers
315 * of two.
316 * For example, assuming the default value of RTO.Alpha of
317 * 1/8, rto_alpha would be expressed as 3.
318 */
319 tp->rttvar = tp->rttvar - (tp->rttvar >> sctp_rto_beta)
320 + ((abs(tp->srtt - rtt)) >> sctp_rto_beta);
321 tp->srtt = tp->srtt - (tp->srtt >> sctp_rto_alpha)
322 + (rtt >> sctp_rto_alpha);
323 } else {
324 /* 6.3.1 C2) When the first RTT measurement R is made, set
325 * SRTT <- R, RTTVAR <- R/2.
326 */
327 tp->srtt = rtt;
328 tp->rttvar = rtt >> 1;
329 }
330
331 /* 6.3.1 G1) Whenever RTTVAR is computed, if RTTVAR = 0, then
332 * adjust RTTVAR <- G, where G is the CLOCK GRANULARITY.
333 */
334 if (tp->rttvar == 0)
335 tp->rttvar = SCTP_CLOCK_GRANULARITY;
336
337 /* 6.3.1 C3) After the computation, update RTO <- SRTT + 4 * RTTVAR. */
338 tp->rto = tp->srtt + (tp->rttvar << 2);
339
340 /* 6.3.1 C6) Whenever RTO is computed, if it is less than RTO.Min
341 * seconds then it is rounded up to RTO.Min seconds.
342 */
343 if (tp->rto < tp->asoc->rto_min)
344 tp->rto = tp->asoc->rto_min;
345
346 /* 6.3.1 C7) A maximum value may be placed on RTO provided it is
347 * at least RTO.max seconds.
348 */
349 if (tp->rto > tp->asoc->rto_max)
350 tp->rto = tp->asoc->rto_max;
351
352 tp->rtt = rtt;
353
354 /* Reset rto_pending so that a new RTT measurement is started when a
355 * new data chunk is sent.
356 */
357 tp->rto_pending = 0;
358
359 SCTP_DEBUG_PRINTK("%s: transport: %p, rtt: %d, srtt: %d "
Harvey Harrison0dc47872008-03-05 20:47:47 -0800360 "rttvar: %d, rto: %ld\n", __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361 tp, rtt, tp->srtt, tp->rttvar, tp->rto);
362}
363
364/* This routine updates the transport's cwnd and partial_bytes_acked
365 * parameters based on the bytes acked in the received SACK.
366 */
367void sctp_transport_raise_cwnd(struct sctp_transport *transport,
368 __u32 sack_ctsn, __u32 bytes_acked)
369{
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400370 struct sctp_association *asoc = transport->asoc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 __u32 cwnd, ssthresh, flight_size, pba, pmtu;
372
373 cwnd = transport->cwnd;
374 flight_size = transport->flight_size;
375
Vlad Yasevicha6465232008-06-04 12:38:43 -0700376 /* See if we need to exit Fast Recovery first */
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400377 if (asoc->fast_recovery &&
378 TSN_lte(asoc->fast_recovery_exit, sack_ctsn))
379 asoc->fast_recovery = 0;
Vlad Yasevicha6465232008-06-04 12:38:43 -0700380
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 /* The appropriate cwnd increase algorithm is performed if, and only
Vlad Yasevicha6465232008-06-04 12:38:43 -0700382 * if the cumulative TSN whould advanced and the congestion window is
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 * being fully utilized.
384 */
Vlad Yasevicha6465232008-06-04 12:38:43 -0700385 if (TSN_lte(sack_ctsn, transport->asoc->ctsn_ack_point) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 (flight_size < cwnd))
387 return;
388
389 ssthresh = transport->ssthresh;
390 pba = transport->partial_bytes_acked;
Frank Filz52ccb8e2005-12-22 11:36:46 -0800391 pmtu = transport->asoc->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392
393 if (cwnd <= ssthresh) {
Vlad Yasevicha6465232008-06-04 12:38:43 -0700394 /* RFC 4960 7.2.1
395 * o When cwnd is less than or equal to ssthresh, an SCTP
396 * endpoint MUST use the slow-start algorithm to increase
397 * cwnd only if the current congestion window is being fully
398 * utilized, an incoming SACK advances the Cumulative TSN
399 * Ack Point, and the data sender is not in Fast Recovery.
400 * Only when these three conditions are met can the cwnd be
401 * increased; otherwise, the cwnd MUST not be increased.
402 * If these conditions are met, then cwnd MUST be increased
403 * by, at most, the lesser of 1) the total size of the
404 * previously outstanding DATA chunk(s) acknowledged, and
405 * 2) the destination's path MTU. This upper bound protects
406 * against the ACK-Splitting attack outlined in [SAVAGE99].
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 */
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400408 if (asoc->fast_recovery)
Vlad Yasevicha6465232008-06-04 12:38:43 -0700409 return;
410
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 if (bytes_acked > pmtu)
412 cwnd += pmtu;
413 else
414 cwnd += bytes_acked;
415 SCTP_DEBUG_PRINTK("%s: SLOW START: transport: %p, "
416 "bytes_acked: %d, cwnd: %d, ssthresh: %d, "
417 "flight_size: %d, pba: %d\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -0800418 __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 transport, bytes_acked, cwnd,
420 ssthresh, flight_size, pba);
421 } else {
422 /* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh,
423 * upon each SACK arrival that advances the Cumulative TSN Ack
424 * Point, increase partial_bytes_acked by the total number of
425 * bytes of all new chunks acknowledged in that SACK including
426 * chunks acknowledged by the new Cumulative TSN Ack and by
427 * Gap Ack Blocks.
428 *
429 * When partial_bytes_acked is equal to or greater than cwnd
430 * and before the arrival of the SACK the sender had cwnd or
431 * more bytes of data outstanding (i.e., before arrival of the
432 * SACK, flightsize was greater than or equal to cwnd),
433 * increase cwnd by MTU, and reset partial_bytes_acked to
434 * (partial_bytes_acked - cwnd).
435 */
436 pba += bytes_acked;
437 if (pba >= cwnd) {
438 cwnd += pmtu;
439 pba = ((cwnd < pba) ? (pba - cwnd) : 0);
440 }
441 SCTP_DEBUG_PRINTK("%s: CONGESTION AVOIDANCE: "
442 "transport: %p, bytes_acked: %d, cwnd: %d, "
443 "ssthresh: %d, flight_size: %d, pba: %d\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -0800444 __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 transport, bytes_acked, cwnd,
446 ssthresh, flight_size, pba);
447 }
448
449 transport->cwnd = cwnd;
450 transport->partial_bytes_acked = pba;
451}
452
453/* This routine is used to lower the transport's cwnd when congestion is
454 * detected.
455 */
456void sctp_transport_lower_cwnd(struct sctp_transport *transport,
457 sctp_lower_cwnd_t reason)
458{
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400459 struct sctp_association *asoc = transport->asoc;
460
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 switch (reason) {
462 case SCTP_LOWER_CWND_T3_RTX:
463 /* RFC 2960 Section 7.2.3, sctpimpguide
464 * When the T3-rtx timer expires on an address, SCTP should
465 * perform slow start by:
466 * ssthresh = max(cwnd/2, 4*MTU)
467 * cwnd = 1*MTU
468 * partial_bytes_acked = 0
469 */
470 transport->ssthresh = max(transport->cwnd/2,
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400471 4*asoc->pathmtu);
472 transport->cwnd = asoc->pathmtu;
Vlad Yasevich33ce8282009-09-04 18:20:58 -0400473
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400474 /* T3-rtx also clears fast recovery */
475 asoc->fast_recovery = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 break;
477
478 case SCTP_LOWER_CWND_FAST_RTX:
479 /* RFC 2960 7.2.4 Adjust the ssthresh and cwnd of the
480 * destination address(es) to which the missing DATA chunks
481 * were last sent, according to the formula described in
482 * Section 7.2.3.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900483 *
484 * RFC 2960 7.2.3, sctpimpguide Upon detection of packet
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 * losses from SACK (see Section 7.2.4), An endpoint
486 * should do the following:
487 * ssthresh = max(cwnd/2, 4*MTU)
488 * cwnd = ssthresh
489 * partial_bytes_acked = 0
490 */
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400491 if (asoc->fast_recovery)
Vlad Yasevicha6465232008-06-04 12:38:43 -0700492 return;
493
494 /* Mark Fast recovery */
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400495 asoc->fast_recovery = 1;
496 asoc->fast_recovery_exit = asoc->next_tsn - 1;
Vlad Yasevicha6465232008-06-04 12:38:43 -0700497
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 transport->ssthresh = max(transport->cwnd/2,
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400499 4*asoc->pathmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500 transport->cwnd = transport->ssthresh;
501 break;
502
503 case SCTP_LOWER_CWND_ECNE:
504 /* RFC 2481 Section 6.1.2.
505 * If the sender receives an ECN-Echo ACK packet
506 * then the sender knows that congestion was encountered in the
507 * network on the path from the sender to the receiver. The
508 * indication of congestion should be treated just as a
509 * congestion loss in non-ECN Capable TCP. That is, the TCP
510 * source halves the congestion window "cwnd" and reduces the
511 * slow start threshold "ssthresh".
512 * A critical condition is that TCP does not react to
513 * congestion indications more than once every window of
514 * data (or more loosely more than once every round-trip time).
515 */
Wei Yongjunf61f6f82009-03-02 09:46:13 +0000516 if (time_after(jiffies, transport->last_time_ecne_reduced +
517 transport->rtt)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 transport->ssthresh = max(transport->cwnd/2,
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400519 4*asoc->pathmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 transport->cwnd = transport->ssthresh;
521 transport->last_time_ecne_reduced = jiffies;
522 }
523 break;
524
525 case SCTP_LOWER_CWND_INACTIVE:
526 /* RFC 2960 Section 7.2.1, sctpimpguide
527 * When the endpoint does not transmit data on a given
528 * transport address, the cwnd of the transport address
529 * should be adjusted to max(cwnd/2, 4*MTU) per RTO.
530 * NOTE: Although the draft recommends that this check needs
531 * to be done every RTO interval, we do it every hearbeat
532 * interval.
533 */
Vlad Yasevich245cba72009-11-23 15:53:58 -0500534 transport->cwnd = max(transport->cwnd/2,
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400535 4*asoc->pathmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 break;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -0700537 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538
539 transport->partial_bytes_acked = 0;
540 SCTP_DEBUG_PRINTK("%s: transport: %p reason: %d cwnd: "
Harvey Harrison0dc47872008-03-05 20:47:47 -0800541 "%d ssthresh: %d\n", __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 transport, reason,
543 transport->cwnd, transport->ssthresh);
544}
545
Vlad Yasevich46d5a802009-11-23 15:54:00 -0500546/* Apply Max.Burst limit to the congestion window:
547 * sctpimpguide-05 2.14.2
548 * D) When the time comes for the sender to
549 * transmit new DATA chunks, the protocol parameter Max.Burst MUST
550 * first be applied to limit how many new DATA chunks may be sent.
551 * The limit is applied by adjusting cwnd as follows:
552 * if ((flightsize+ Max.Burst * MTU) < cwnd)
553 * cwnd = flightsize + Max.Burst * MTU
554 */
555
556void sctp_transport_burst_limited(struct sctp_transport *t)
557{
558 struct sctp_association *asoc = t->asoc;
559 u32 old_cwnd = t->cwnd;
560 u32 max_burst_bytes;
561
562 if (t->burst_limited)
563 return;
564
565 max_burst_bytes = t->flight_size + (asoc->max_burst * asoc->pathmtu);
566 if (max_burst_bytes < old_cwnd) {
567 t->cwnd = max_burst_bytes;
568 t->burst_limited = old_cwnd;
569 }
570}
571
572/* Restore the old cwnd congestion window, after the burst had it's
573 * desired effect.
574 */
575void sctp_transport_burst_reset(struct sctp_transport *t)
576{
577 if (t->burst_limited) {
578 t->cwnd = t->burst_limited;
579 t->burst_limited = 0;
580 }
581}
582
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583/* What is the next timeout value for this transport? */
584unsigned long sctp_transport_timeout(struct sctp_transport *t)
585{
586 unsigned long timeout;
Sridhar Samudralaad8fec12006-07-21 14:48:50 -0700587 timeout = t->rto + sctp_jitter(t->rto);
588 if (t->state != SCTP_UNCONFIRMED)
589 timeout += t->hbinterval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 timeout += jiffies;
591 return timeout;
592}
Vlad Yasevich749bf922007-03-19 17:02:30 -0700593
594/* Reset transport variables to their initial values */
595void sctp_transport_reset(struct sctp_transport *t)
596{
597 struct sctp_association *asoc = t->asoc;
598
599 /* RFC 2960 (bis), Section 5.2.4
600 * All the congestion control parameters (e.g., cwnd, ssthresh)
601 * related to this peer MUST be reset to their initial values
602 * (see Section 6.2.1)
603 */
604 t->cwnd = min(4*asoc->pathmtu, max_t(__u32, 2*asoc->pathmtu, 4380));
Vlad Yasevich46d5a802009-11-23 15:54:00 -0500605 t->burst_limited = 0;
Vlad Yasevich289f4242007-03-22 12:26:25 -0700606 t->ssthresh = asoc->peer.i.a_rwnd;
Andrei Pelinescu-Onciul5fdd4ba2009-11-29 00:14:02 -0800607 t->rto = asoc->rto_initial;
Vlad Yasevich749bf922007-03-19 17:02:30 -0700608 t->rtt = 0;
609 t->srtt = 0;
610 t->rttvar = 0;
611
612 /* Reset these additional varibles so that we have a clean
613 * slate.
614 */
615 t->partial_bytes_acked = 0;
616 t->flight_size = 0;
617 t->error_count = 0;
618 t->rto_pending = 0;
Vlad Yasevichfaee47c2009-02-13 08:33:43 +0000619 t->hb_sent = 0;
Vlad Yasevich749bf922007-03-19 17:02:30 -0700620
621 /* Initialize the state information for SFR-CACC */
622 t->cacc.changeover_active = 0;
623 t->cacc.cycling_changeover = 0;
624 t->cacc.next_tsn_at_change = 0;
625 t->cacc.cacc_saw_newack = 0;
626}
Michio Hondaddc4bbe2011-06-17 11:03:23 +0900627
628/* Schedule retransmission on the given transport */
629void sctp_transport_immediate_rtx(struct sctp_transport *t)
630{
631 /* Stop pending T3_rtx_timer */
632 if (timer_pending(&t->T3_rtx_timer)) {
633 (void)del_timer(&t->T3_rtx_timer);
634 sctp_transport_put(t);
635 }
636 sctp_retransmit(&t->asoc->outqueue, t, SCTP_RTXR_T3_RTX);
637 if (!timer_pending(&t->T3_rtx_timer)) {
638 if (!mod_timer(&t->T3_rtx_timer, jiffies + t->rto))
639 sctp_transport_hold(t);
640 }
641 return;
642}