blob: d3ae493d234aaebe0e60c97e17479659e4a4c181 [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 Dobriyan3182cd82005-07-11 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. */
214void sctp_transport_pmtu(struct sctp_transport *transport)
215{
216 struct dst_entry *dst;
217
218 dst = transport->af_specific->get_dst(NULL, &transport->ipaddr, NULL);
219
220 if (dst) {
Frank Filz52ccb8e2005-12-22 11:36:46 -0800221 transport->pathmtu = dst_mtu(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 dst_release(dst);
223 } else
Frank Filz52ccb8e2005-12-22 11:36:46 -0800224 transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225}
226
Vlad Yasevichc910b472007-06-07 13:47:03 -0400227/* this is a complete rip-off from __sk_dst_check
228 * the cookie is always 0 since this is how it's used in the
229 * pmtu code
230 */
231static struct dst_entry *sctp_transport_dst_check(struct sctp_transport *t)
232{
233 struct dst_entry *dst = t->dst;
234
235 if (dst && dst->obsolete && dst->ops->check(dst, 0) == NULL) {
236 dst_release(t->dst);
237 t->dst = NULL;
238 return NULL;
239 }
240
241 return dst;
242}
243
244void sctp_transport_update_pmtu(struct sctp_transport *t, u32 pmtu)
245{
246 struct dst_entry *dst;
247
248 if (unlikely(pmtu < SCTP_DEFAULT_MINSEGMENT)) {
Joe Perches145ce502010-08-24 13:21:08 +0000249 pr_warn("%s: Reported pmtu %d too low, using default minimum of %d\n",
250 __func__, pmtu,
251 SCTP_DEFAULT_MINSEGMENT);
Vlad Yasevichc910b472007-06-07 13:47:03 -0400252 /* Use default minimum segment size and disable
253 * pmtu discovery on this transport.
254 */
255 t->pathmtu = SCTP_DEFAULT_MINSEGMENT;
Vlad Yasevichc910b472007-06-07 13:47:03 -0400256 } else {
257 t->pathmtu = pmtu;
258 }
259
260 dst = sctp_transport_dst_check(t);
261 if (dst)
262 dst->ops->update_pmtu(dst, pmtu);
263}
264
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265/* Caches the dst entry and source address for a transport's destination
266 * address.
267 */
268void sctp_transport_route(struct sctp_transport *transport,
269 union sctp_addr *saddr, struct sctp_sock *opt)
270{
271 struct sctp_association *asoc = transport->asoc;
272 struct sctp_af *af = transport->af_specific;
273 union sctp_addr *daddr = &transport->ipaddr;
274 struct dst_entry *dst;
275
276 dst = af->get_dst(asoc, daddr, saddr);
277
278 if (saddr)
279 memcpy(&transport->saddr, saddr, sizeof(union sctp_addr));
280 else
YOSHIFUJI Hideakie5117102008-05-29 19:55:05 +0900281 af->get_saddr(opt, asoc, dst, daddr, &transport->saddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282
283 transport->dst = dst;
Frank Filz52ccb8e2005-12-22 11:36:46 -0800284 if ((transport->param_flags & SPP_PMTUD_DISABLE) && transport->pathmtu) {
285 return;
286 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 if (dst) {
Frank Filz52ccb8e2005-12-22 11:36:46 -0800288 transport->pathmtu = dst_mtu(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289
290 /* Initialize sk->sk_rcv_saddr, if the transport is the
291 * association's active path for getsockname().
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900292 */
Vlad Yasevicha78102e2009-11-11 11:54:37 +0000293 if (asoc && (!asoc->peer.primary_path ||
294 (transport == asoc->peer.active_path)))
Neil Hormanbf031ff2005-12-02 20:32:29 -0800295 opt->pf->af->to_sk_saddr(&transport->saddr,
296 asoc->base.sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297 } else
Frank Filz52ccb8e2005-12-22 11:36:46 -0800298 transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299}
300
301/* Hold a reference to a transport. */
302void sctp_transport_hold(struct sctp_transport *transport)
303{
304 atomic_inc(&transport->refcnt);
305}
306
307/* Release a reference to a transport and clean up
308 * if there are no more references.
309 */
310void sctp_transport_put(struct sctp_transport *transport)
311{
312 if (atomic_dec_and_test(&transport->refcnt))
313 sctp_transport_destroy(transport);
314}
315
316/* Update transport's RTO based on the newly calculated RTT. */
317void sctp_transport_update_rto(struct sctp_transport *tp, __u32 rtt)
318{
319 /* Check for valid transport. */
320 SCTP_ASSERT(tp, "NULL transport", return);
321
322 /* We should not be doing any RTO updates unless rto_pending is set. */
323 SCTP_ASSERT(tp->rto_pending, "rto_pending not set", return);
324
325 if (tp->rttvar || tp->srtt) {
326 /* 6.3.1 C3) When a new RTT measurement R' is made, set
327 * RTTVAR <- (1 - RTO.Beta) * RTTVAR + RTO.Beta * |SRTT - R'|
328 * SRTT <- (1 - RTO.Alpha) * SRTT + RTO.Alpha * R'
329 */
330
331 /* Note: The above algorithm has been rewritten to
332 * express rto_beta and rto_alpha as inverse powers
333 * of two.
334 * For example, assuming the default value of RTO.Alpha of
335 * 1/8, rto_alpha would be expressed as 3.
336 */
337 tp->rttvar = tp->rttvar - (tp->rttvar >> sctp_rto_beta)
338 + ((abs(tp->srtt - rtt)) >> sctp_rto_beta);
339 tp->srtt = tp->srtt - (tp->srtt >> sctp_rto_alpha)
340 + (rtt >> sctp_rto_alpha);
341 } else {
342 /* 6.3.1 C2) When the first RTT measurement R is made, set
343 * SRTT <- R, RTTVAR <- R/2.
344 */
345 tp->srtt = rtt;
346 tp->rttvar = rtt >> 1;
347 }
348
349 /* 6.3.1 G1) Whenever RTTVAR is computed, if RTTVAR = 0, then
350 * adjust RTTVAR <- G, where G is the CLOCK GRANULARITY.
351 */
352 if (tp->rttvar == 0)
353 tp->rttvar = SCTP_CLOCK_GRANULARITY;
354
355 /* 6.3.1 C3) After the computation, update RTO <- SRTT + 4 * RTTVAR. */
356 tp->rto = tp->srtt + (tp->rttvar << 2);
357
358 /* 6.3.1 C6) Whenever RTO is computed, if it is less than RTO.Min
359 * seconds then it is rounded up to RTO.Min seconds.
360 */
361 if (tp->rto < tp->asoc->rto_min)
362 tp->rto = tp->asoc->rto_min;
363
364 /* 6.3.1 C7) A maximum value may be placed on RTO provided it is
365 * at least RTO.max seconds.
366 */
367 if (tp->rto > tp->asoc->rto_max)
368 tp->rto = tp->asoc->rto_max;
369
370 tp->rtt = rtt;
371
372 /* Reset rto_pending so that a new RTT measurement is started when a
373 * new data chunk is sent.
374 */
375 tp->rto_pending = 0;
376
377 SCTP_DEBUG_PRINTK("%s: transport: %p, rtt: %d, srtt: %d "
Harvey Harrison0dc47872008-03-05 20:47:47 -0800378 "rttvar: %d, rto: %ld\n", __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 tp, rtt, tp->srtt, tp->rttvar, tp->rto);
380}
381
382/* This routine updates the transport's cwnd and partial_bytes_acked
383 * parameters based on the bytes acked in the received SACK.
384 */
385void sctp_transport_raise_cwnd(struct sctp_transport *transport,
386 __u32 sack_ctsn, __u32 bytes_acked)
387{
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400388 struct sctp_association *asoc = transport->asoc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 __u32 cwnd, ssthresh, flight_size, pba, pmtu;
390
391 cwnd = transport->cwnd;
392 flight_size = transport->flight_size;
393
Vlad Yasevicha6465232008-06-04 12:38:43 -0700394 /* See if we need to exit Fast Recovery first */
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400395 if (asoc->fast_recovery &&
396 TSN_lte(asoc->fast_recovery_exit, sack_ctsn))
397 asoc->fast_recovery = 0;
Vlad Yasevicha6465232008-06-04 12:38:43 -0700398
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 /* The appropriate cwnd increase algorithm is performed if, and only
Vlad Yasevicha6465232008-06-04 12:38:43 -0700400 * if the cumulative TSN whould advanced and the congestion window is
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 * being fully utilized.
402 */
Vlad Yasevicha6465232008-06-04 12:38:43 -0700403 if (TSN_lte(sack_ctsn, transport->asoc->ctsn_ack_point) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 (flight_size < cwnd))
405 return;
406
407 ssthresh = transport->ssthresh;
408 pba = transport->partial_bytes_acked;
Frank Filz52ccb8e2005-12-22 11:36:46 -0800409 pmtu = transport->asoc->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410
411 if (cwnd <= ssthresh) {
Vlad Yasevicha6465232008-06-04 12:38:43 -0700412 /* RFC 4960 7.2.1
413 * o When cwnd is less than or equal to ssthresh, an SCTP
414 * endpoint MUST use the slow-start algorithm to increase
415 * cwnd only if the current congestion window is being fully
416 * utilized, an incoming SACK advances the Cumulative TSN
417 * Ack Point, and the data sender is not in Fast Recovery.
418 * Only when these three conditions are met can the cwnd be
419 * increased; otherwise, the cwnd MUST not be increased.
420 * If these conditions are met, then cwnd MUST be increased
421 * by, at most, the lesser of 1) the total size of the
422 * previously outstanding DATA chunk(s) acknowledged, and
423 * 2) the destination's path MTU. This upper bound protects
424 * against the ACK-Splitting attack outlined in [SAVAGE99].
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 */
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400426 if (asoc->fast_recovery)
Vlad Yasevicha6465232008-06-04 12:38:43 -0700427 return;
428
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 if (bytes_acked > pmtu)
430 cwnd += pmtu;
431 else
432 cwnd += bytes_acked;
433 SCTP_DEBUG_PRINTK("%s: SLOW START: transport: %p, "
434 "bytes_acked: %d, cwnd: %d, ssthresh: %d, "
435 "flight_size: %d, pba: %d\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -0800436 __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 transport, bytes_acked, cwnd,
438 ssthresh, flight_size, pba);
439 } else {
440 /* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh,
441 * upon each SACK arrival that advances the Cumulative TSN Ack
442 * Point, increase partial_bytes_acked by the total number of
443 * bytes of all new chunks acknowledged in that SACK including
444 * chunks acknowledged by the new Cumulative TSN Ack and by
445 * Gap Ack Blocks.
446 *
447 * When partial_bytes_acked is equal to or greater than cwnd
448 * and before the arrival of the SACK the sender had cwnd or
449 * more bytes of data outstanding (i.e., before arrival of the
450 * SACK, flightsize was greater than or equal to cwnd),
451 * increase cwnd by MTU, and reset partial_bytes_acked to
452 * (partial_bytes_acked - cwnd).
453 */
454 pba += bytes_acked;
455 if (pba >= cwnd) {
456 cwnd += pmtu;
457 pba = ((cwnd < pba) ? (pba - cwnd) : 0);
458 }
459 SCTP_DEBUG_PRINTK("%s: CONGESTION AVOIDANCE: "
460 "transport: %p, bytes_acked: %d, cwnd: %d, "
461 "ssthresh: %d, flight_size: %d, pba: %d\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -0800462 __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 transport, bytes_acked, cwnd,
464 ssthresh, flight_size, pba);
465 }
466
467 transport->cwnd = cwnd;
468 transport->partial_bytes_acked = pba;
469}
470
471/* This routine is used to lower the transport's cwnd when congestion is
472 * detected.
473 */
474void sctp_transport_lower_cwnd(struct sctp_transport *transport,
475 sctp_lower_cwnd_t reason)
476{
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400477 struct sctp_association *asoc = transport->asoc;
478
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 switch (reason) {
480 case SCTP_LOWER_CWND_T3_RTX:
481 /* RFC 2960 Section 7.2.3, sctpimpguide
482 * When the T3-rtx timer expires on an address, SCTP should
483 * perform slow start by:
484 * ssthresh = max(cwnd/2, 4*MTU)
485 * cwnd = 1*MTU
486 * partial_bytes_acked = 0
487 */
488 transport->ssthresh = max(transport->cwnd/2,
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400489 4*asoc->pathmtu);
490 transport->cwnd = asoc->pathmtu;
Vlad Yasevich33ce8282009-09-04 18:20:58 -0400491
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400492 /* T3-rtx also clears fast recovery */
493 asoc->fast_recovery = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 break;
495
496 case SCTP_LOWER_CWND_FAST_RTX:
497 /* RFC 2960 7.2.4 Adjust the ssthresh and cwnd of the
498 * destination address(es) to which the missing DATA chunks
499 * were last sent, according to the formula described in
500 * Section 7.2.3.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900501 *
502 * RFC 2960 7.2.3, sctpimpguide Upon detection of packet
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 * losses from SACK (see Section 7.2.4), An endpoint
504 * should do the following:
505 * ssthresh = max(cwnd/2, 4*MTU)
506 * cwnd = ssthresh
507 * partial_bytes_acked = 0
508 */
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400509 if (asoc->fast_recovery)
Vlad Yasevicha6465232008-06-04 12:38:43 -0700510 return;
511
512 /* Mark Fast recovery */
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400513 asoc->fast_recovery = 1;
514 asoc->fast_recovery_exit = asoc->next_tsn - 1;
Vlad Yasevicha6465232008-06-04 12:38:43 -0700515
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 transport->ssthresh = max(transport->cwnd/2,
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400517 4*asoc->pathmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 transport->cwnd = transport->ssthresh;
519 break;
520
521 case SCTP_LOWER_CWND_ECNE:
522 /* RFC 2481 Section 6.1.2.
523 * If the sender receives an ECN-Echo ACK packet
524 * then the sender knows that congestion was encountered in the
525 * network on the path from the sender to the receiver. The
526 * indication of congestion should be treated just as a
527 * congestion loss in non-ECN Capable TCP. That is, the TCP
528 * source halves the congestion window "cwnd" and reduces the
529 * slow start threshold "ssthresh".
530 * A critical condition is that TCP does not react to
531 * congestion indications more than once every window of
532 * data (or more loosely more than once every round-trip time).
533 */
Wei Yongjunf61f6f82009-03-02 09:46:13 +0000534 if (time_after(jiffies, transport->last_time_ecne_reduced +
535 transport->rtt)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 transport->ssthresh = max(transport->cwnd/2,
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400537 4*asoc->pathmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 transport->cwnd = transport->ssthresh;
539 transport->last_time_ecne_reduced = jiffies;
540 }
541 break;
542
543 case SCTP_LOWER_CWND_INACTIVE:
544 /* RFC 2960 Section 7.2.1, sctpimpguide
545 * When the endpoint does not transmit data on a given
546 * transport address, the cwnd of the transport address
547 * should be adjusted to max(cwnd/2, 4*MTU) per RTO.
548 * NOTE: Although the draft recommends that this check needs
549 * to be done every RTO interval, we do it every hearbeat
550 * interval.
551 */
Vlad Yasevich245cba72009-11-23 15:53:58 -0500552 transport->cwnd = max(transport->cwnd/2,
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400553 4*asoc->pathmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 break;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -0700555 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556
557 transport->partial_bytes_acked = 0;
558 SCTP_DEBUG_PRINTK("%s: transport: %p reason: %d cwnd: "
Harvey Harrison0dc47872008-03-05 20:47:47 -0800559 "%d ssthresh: %d\n", __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 transport, reason,
561 transport->cwnd, transport->ssthresh);
562}
563
Vlad Yasevich46d5a802009-11-23 15:54:00 -0500564/* Apply Max.Burst limit to the congestion window:
565 * sctpimpguide-05 2.14.2
566 * D) When the time comes for the sender to
567 * transmit new DATA chunks, the protocol parameter Max.Burst MUST
568 * first be applied to limit how many new DATA chunks may be sent.
569 * The limit is applied by adjusting cwnd as follows:
570 * if ((flightsize+ Max.Burst * MTU) < cwnd)
571 * cwnd = flightsize + Max.Burst * MTU
572 */
573
574void sctp_transport_burst_limited(struct sctp_transport *t)
575{
576 struct sctp_association *asoc = t->asoc;
577 u32 old_cwnd = t->cwnd;
578 u32 max_burst_bytes;
579
580 if (t->burst_limited)
581 return;
582
583 max_burst_bytes = t->flight_size + (asoc->max_burst * asoc->pathmtu);
584 if (max_burst_bytes < old_cwnd) {
585 t->cwnd = max_burst_bytes;
586 t->burst_limited = old_cwnd;
587 }
588}
589
590/* Restore the old cwnd congestion window, after the burst had it's
591 * desired effect.
592 */
593void sctp_transport_burst_reset(struct sctp_transport *t)
594{
595 if (t->burst_limited) {
596 t->cwnd = t->burst_limited;
597 t->burst_limited = 0;
598 }
599}
600
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601/* What is the next timeout value for this transport? */
602unsigned long sctp_transport_timeout(struct sctp_transport *t)
603{
604 unsigned long timeout;
Sridhar Samudralaad8fec12006-07-21 14:48:50 -0700605 timeout = t->rto + sctp_jitter(t->rto);
606 if (t->state != SCTP_UNCONFIRMED)
607 timeout += t->hbinterval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 timeout += jiffies;
609 return timeout;
610}
Vlad Yasevich749bf922007-03-19 17:02:30 -0700611
612/* Reset transport variables to their initial values */
613void sctp_transport_reset(struct sctp_transport *t)
614{
615 struct sctp_association *asoc = t->asoc;
616
617 /* RFC 2960 (bis), Section 5.2.4
618 * All the congestion control parameters (e.g., cwnd, ssthresh)
619 * related to this peer MUST be reset to their initial values
620 * (see Section 6.2.1)
621 */
622 t->cwnd = min(4*asoc->pathmtu, max_t(__u32, 2*asoc->pathmtu, 4380));
Vlad Yasevich46d5a802009-11-23 15:54:00 -0500623 t->burst_limited = 0;
Vlad Yasevich289f4242007-03-22 12:26:25 -0700624 t->ssthresh = asoc->peer.i.a_rwnd;
Andrei Pelinescu-Onciul5fdd4ba2009-11-29 00:14:02 -0800625 t->rto = asoc->rto_initial;
Vlad Yasevich749bf922007-03-19 17:02:30 -0700626 t->rtt = 0;
627 t->srtt = 0;
628 t->rttvar = 0;
629
630 /* Reset these additional varibles so that we have a clean
631 * slate.
632 */
633 t->partial_bytes_acked = 0;
634 t->flight_size = 0;
635 t->error_count = 0;
636 t->rto_pending = 0;
Vlad Yasevichfaee47c2009-02-13 08:33:43 +0000637 t->hb_sent = 0;
Vlad Yasevich749bf922007-03-19 17:02:30 -0700638
639 /* Initialize the state information for SFR-CACC */
640 t->cacc.changeover_active = 0;
641 t->cacc.cycling_changeover = 0;
642 t->cacc.next_tsn_at_change = 0;
643 t->cacc.cacc_saw_newack = 0;
644}