blob: 47f82bd794d915188bad037463c2aa14175a55ef [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
Jeff Kirsher4b2f13a2013-12-06 06:28:48 -080027 * along with GNU CC; see the file COPYING. If not, see
28 * <http://www.gnu.org/licenses/>.
Linus Torvalds1da177e2005-04-16 15:20:36 -070029 *
30 * Please send any bug reports or fixes you make to the
31 * email address(es):
Daniel Borkmann91705c62013-07-23 14:51:47 +020032 * lksctp developers <linux-sctp@vger.kernel.org>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070034 * 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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070042 */
43
Joe Perches145ce502010-08-24 13:21:08 +000044#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
45
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090046#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070047#include <linux/types.h>
Sridhar Samudralaad8fec12006-07-21 14:48:50 -070048#include <linux/random.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070049#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. */
Eric W. Biederman89bf3452012-08-07 07:26:14 +000055static struct sctp_transport *sctp_transport_init(struct net *net,
56 struct sctp_transport *peer,
Linus Torvalds1da177e2005-04-16 15:20:36 -070057 const union sctp_addr *addr,
Al Virodd0fc662005-10-07 07:46:04 +010058 gfp_t gfp)
Linus Torvalds1da177e2005-04-16 15:20:36 -070059{
60 /* Copy in the address. */
61 peer->ipaddr = *addr;
62 peer->af_specific = sctp_get_af_specific(addr->sa.sa_family);
Linus Torvalds1da177e2005-04-16 15:20:36 -070063 memset(&peer->saddr, 0, sizeof(union sctp_addr));
64
Neil Horman42448542012-06-30 03:04:26 +000065 peer->sack_generation = 0;
66
Linus Torvalds1da177e2005-04-16 15:20:36 -070067 /* 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 */
Eric W. Biedermane1fc3b12012-08-07 07:29:57 +000073 peer->rto = msecs_to_jiffies(net->sctp.rto_initial);
Linus Torvalds1da177e2005-04-16 15:20:36 -070074
Thomas Gleixner8b0e1952016-12-25 12:30:41 +010075 peer->last_time_heard = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -070076 peer->last_time_ecne_reduced = jiffies;
77
Frank Filz52ccb8e2005-12-22 11:36:46 -080078 peer->param_flags = SPP_HB_DISABLE |
79 SPP_PMTUD_ENABLE |
80 SPP_SACKDELAY_ENABLE;
Linus Torvalds1da177e2005-04-16 15:20:36 -070081
82 /* Initialize the default path max_retrans. */
Eric W. Biedermane1fc3b12012-08-07 07:29:57 +000083 peer->pathmaxrxt = net->sctp.max_retrans_path;
84 peer->pf_retrans = net->sctp.pf_retrans;
Linus Torvalds1da177e2005-04-16 15:20:36 -070085
86 INIT_LIST_HEAD(&peer->transmitted);
87 INIT_LIST_HEAD(&peer->send_ready);
88 INIT_LIST_HEAD(&peer->transports);
89
Kees Cook9c3b5752017-10-24 01:45:31 -070090 timer_setup(&peer->T3_rtx_timer, sctp_generate_t3_rtx_event, 0);
91 timer_setup(&peer->hb_timer, sctp_generate_heartbeat_event, 0);
92 timer_setup(&peer->reconf_timer, sctp_generate_reconf_event, 0);
93 timer_setup(&peer->proto_unreach_timer,
94 sctp_generate_proto_unreach_event, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -070095
Sridhar Samudralaad8fec12006-07-21 14:48:50 -070096 /* Initialize the 64-bit random nonce sent with heartbeat. */
97 get_random_bytes(&peer->hb_nonce, sizeof(peer->hb_nonce));
98
Reshetova, Elenaa4b2b582017-07-04 15:53:27 +030099 refcount_set(&peer->refcnt, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100
101 return peer;
102}
103
104/* Allocate and initialize a new transport. */
Eric W. Biederman89bf3452012-08-07 07:26:14 +0000105struct sctp_transport *sctp_transport_new(struct net *net,
106 const union sctp_addr *addr,
Al Virodd0fc662005-10-07 07:46:04 +0100107 gfp_t gfp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108{
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900109 struct sctp_transport *transport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110
Daniel Borkmann939cfa72013-06-17 11:40:04 +0200111 transport = kzalloc(sizeof(*transport), gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112 if (!transport)
113 goto fail;
114
Eric W. Biederman89bf3452012-08-07 07:26:14 +0000115 if (!sctp_transport_init(net, transport, addr, gfp))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116 goto fail_init;
117
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118 SCTP_DBG_OBJCNT_INC(transport);
119
120 return transport;
121
122fail_init:
123 kfree(transport);
124
125fail:
126 return NULL;
127}
128
129/* This transport is no longer needed. Free up if possible, or
130 * delay until it last reference count.
131 */
132void sctp_transport_free(struct sctp_transport *transport)
133{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 /* Try to delete the heartbeat timer. */
135 if (del_timer(&transport->hb_timer))
136 sctp_transport_put(transport);
137
138 /* Delete the T3_rtx timer if it's active.
139 * There is no point in not doing this now and letting
140 * structure hang around in memory since we know
141 * the tranport is going away.
142 */
Ying Xue25cc4ae2013-02-03 20:32:57 +0000143 if (del_timer(&transport->T3_rtx_timer))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144 sctp_transport_put(transport);
145
Xin Long7b9438d2017-01-18 00:44:43 +0800146 if (del_timer(&transport->reconf_timer))
147 sctp_transport_put(transport);
148
Wei Yongjun55fa0cf2010-05-09 16:56:07 +0000149 /* Delete the ICMP proto unreachable timer if it's active. */
Ying Xue25cc4ae2013-02-03 20:32:57 +0000150 if (del_timer(&transport->proto_unreach_timer))
Wei Yongjun55fa0cf2010-05-09 16:56:07 +0000151 sctp_association_put(transport->asoc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152
153 sctp_transport_put(transport);
154}
155
Thomas Graf45122ca262012-12-06 09:25:05 +0000156static void sctp_transport_destroy_rcu(struct rcu_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157{
Thomas Graf45122ca262012-12-06 09:25:05 +0000158 struct sctp_transport *transport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159
Thomas Graf45122ca262012-12-06 09:25:05 +0000160 transport = container_of(head, struct sctp_transport, rcu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161
162 dst_release(transport->dst);
163 kfree(transport);
164 SCTP_DBG_OBJCNT_DEC(transport);
165}
166
Thomas Graf45122ca262012-12-06 09:25:05 +0000167/* Destroy the transport data structure.
168 * Assumes there are no more users of this structure.
169 */
170static void sctp_transport_destroy(struct sctp_transport *transport)
171{
Reshetova, Elenaa4b2b582017-07-04 15:53:27 +0300172 if (unlikely(refcount_read(&transport->refcnt))) {
Daniel Borkmannbb333812013-06-28 19:49:40 +0200173 WARN(1, "Attempt to destroy undead transport %p!\n", transport);
174 return;
175 }
Thomas Graf45122ca262012-12-06 09:25:05 +0000176
Daniel Borkmann8c986532013-02-01 04:37:43 +0000177 sctp_packet_free(&transport->packet);
178
179 if (transport->asoc)
180 sctp_association_put(transport->asoc);
Daniel Borkmann771085d2013-08-09 16:25:21 +0200181
182 call_rcu(&transport->rcu, sctp_transport_destroy_rcu);
Thomas Graf45122ca262012-12-06 09:25:05 +0000183}
184
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185/* Start T3_rtx timer if it is not already running and update the heartbeat
186 * timer. This routine is called every time a DATA chunk is sent.
187 */
Marcelo Ricardo Leitnerba6f5e32016-04-06 15:15:19 -0300188void sctp_transport_reset_t3_rtx(struct sctp_transport *transport)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189{
190 /* RFC 2960 6.3.2 Retransmission Timer Rules
191 *
192 * R1) Every time a DATA chunk is sent to any address(including a
193 * retransmission), if the T3-rtx timer of that address is not running
194 * start it running so that it will expire after the RTO of that
195 * address.
196 */
197
Vlad Yasevichd9efc222010-04-30 22:41:09 -0400198 if (!timer_pending(&transport->T3_rtx_timer))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 if (!mod_timer(&transport->T3_rtx_timer,
200 jiffies + transport->rto))
201 sctp_transport_hold(transport);
Marcelo Ricardo Leitnerba6f5e32016-04-06 15:15:19 -0300202}
203
204void sctp_transport_reset_hb_timer(struct sctp_transport *transport)
205{
206 unsigned long expires;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207
208 /* When a data chunk is sent, reset the heartbeat interval. */
Marcelo Ricardo Leitnerba6f5e32016-04-06 15:15:19 -0300209 expires = jiffies + sctp_transport_timeout(transport);
210 if (time_before(transport->hb_timer.expires, expires) &&
211 !mod_timer(&transport->hb_timer,
212 expires + prandom_u32_max(transport->rto)))
213 sctp_transport_hold(transport);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214}
215
Xin Long7b9438d2017-01-18 00:44:43 +0800216void sctp_transport_reset_reconf_timer(struct sctp_transport *transport)
217{
218 if (!timer_pending(&transport->reconf_timer))
219 if (!mod_timer(&transport->reconf_timer,
220 jiffies + transport->rto))
221 sctp_transport_hold(transport);
222}
223
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224/* This transport has been assigned to an association.
225 * Initialize fields from the association or from the sock itself.
226 * Register the reference count in the association.
227 */
228void sctp_transport_set_owner(struct sctp_transport *transport,
229 struct sctp_association *asoc)
230{
231 transport->asoc = asoc;
232 sctp_association_hold(asoc);
233}
234
235/* Initialize the pmtu of a transport. */
Vlad Yasevich9914ae32011-04-26 21:51:31 +0000236void sctp_transport_pmtu(struct sctp_transport *transport, struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237{
Vlad Yasevichda0420b2011-04-26 21:54:17 +0000238 /* If we don't have a fresh route, look one up */
David S. Millerf5b0a872012-07-19 12:31:33 -0700239 if (!transport->dst || transport->dst->obsolete) {
Julian Anastasovc86a7732017-02-06 23:14:13 +0200240 sctp_transport_dst_release(transport);
Vlad Yasevichda0420b2011-04-26 21:54:17 +0000241 transport->af_specific->get_dst(transport, &transport->saddr,
David S. Miller8663c932011-05-06 16:32:47 -0700242 &transport->fl, sk);
Vlad Yasevichda0420b2011-04-26 21:54:17 +0000243 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244
Vlad Yasevichda0420b2011-04-26 21:54:17 +0000245 if (transport->dst) {
Marcelo Ricardo Leitnere2f036a2016-09-21 08:45:55 -0300246 transport->pathmtu = SCTP_TRUNC4(dst_mtu(transport->dst));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 } else
Frank Filz52ccb8e2005-12-22 11:36:46 -0800248 transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249}
250
Marcelo Ricardo Leitnerb6c57342018-01-05 11:17:18 -0200251bool sctp_transport_update_pmtu(struct sctp_transport *t, u32 pmtu)
Vlad Yasevichc910b472007-06-07 13:47:03 -0400252{
Xin Long3ebfdf02017-04-04 13:39:55 +0800253 struct dst_entry *dst = sctp_transport_dst_check(t);
Marcelo Ricardo Leitnerb6c57342018-01-05 11:17:18 -0200254 bool change = true;
Vlad Yasevichc910b472007-06-07 13:47:03 -0400255
256 if (unlikely(pmtu < SCTP_DEFAULT_MINSEGMENT)) {
Marcelo Ricardo Leitnerb6c57342018-01-05 11:17:18 -0200257 pr_warn_ratelimited("%s: Reported pmtu %d too low, using default minimum of %d\n",
258 __func__, pmtu, SCTP_DEFAULT_MINSEGMENT);
259 /* Use default minimum segment instead */
260 pmtu = SCTP_DEFAULT_MINSEGMENT;
Vlad Yasevichc910b472007-06-07 13:47:03 -0400261 }
Marcelo Ricardo Leitnerb6c57342018-01-05 11:17:18 -0200262 pmtu = SCTP_TRUNC4(pmtu);
Vlad Yasevichc910b472007-06-07 13:47:03 -0400263
David S. Miller02f3d4c2012-07-16 03:57:14 -0700264 if (dst) {
Xin Long3ebfdf02017-04-04 13:39:55 +0800265 dst->ops->update_pmtu(dst, t->asoc->base.sk, NULL, pmtu);
David S. Miller02f3d4c2012-07-16 03:57:14 -0700266 dst = sctp_transport_dst_check(t);
David S. Miller02f3d4c2012-07-16 03:57:14 -0700267 }
Xin Long3ebfdf02017-04-04 13:39:55 +0800268
Marcelo Ricardo Leitnerb6c57342018-01-05 11:17:18 -0200269 if (!dst) {
Xin Long3ebfdf02017-04-04 13:39:55 +0800270 t->af_specific->get_dst(t, &t->saddr, &t->fl, t->asoc->base.sk);
Marcelo Ricardo Leitnerb6c57342018-01-05 11:17:18 -0200271 dst = t->dst;
272 }
273
274 if (dst) {
275 /* Re-fetch, as under layers may have a higher minimum size */
276 pmtu = SCTP_TRUNC4(dst_mtu(dst));
277 change = t->pathmtu != pmtu;
278 }
279 t->pathmtu = pmtu;
280
281 return change;
Vlad Yasevichc910b472007-06-07 13:47:03 -0400282}
283
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284/* Caches the dst entry and source address for a transport's destination
285 * address.
286 */
287void sctp_transport_route(struct sctp_transport *transport,
288 union sctp_addr *saddr, struct sctp_sock *opt)
289{
290 struct sctp_association *asoc = transport->asoc;
291 struct sctp_af *af = transport->af_specific;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292
David S. Miller8663c932011-05-06 16:32:47 -0700293 af->get_dst(transport, saddr, &transport->fl, sctp_opt2sk(opt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294
295 if (saddr)
296 memcpy(&transport->saddr, saddr, sizeof(union sctp_addr));
297 else
David S. Miller8663c932011-05-06 16:32:47 -0700298 af->get_saddr(opt, transport, &transport->fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299
Frank Filz52ccb8e2005-12-22 11:36:46 -0800300 if ((transport->param_flags & SPP_PMTUD_DISABLE) && transport->pathmtu) {
301 return;
302 }
Vlad Yasevichda0420b2011-04-26 21:54:17 +0000303 if (transport->dst) {
Marcelo Ricardo Leitnere2f036a2016-09-21 08:45:55 -0300304 transport->pathmtu = SCTP_TRUNC4(dst_mtu(transport->dst));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305
306 /* Initialize sk->sk_rcv_saddr, if the transport is the
307 * association's active path for getsockname().
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900308 */
Vlad Yasevicha78102e2009-11-11 11:54:37 +0000309 if (asoc && (!asoc->peer.primary_path ||
310 (transport == asoc->peer.active_path)))
Jason Gunthorpe299ee122014-07-30 12:40:53 -0600311 opt->pf->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. */
Xin Long1eed6772016-01-22 01:49:07 +0800318int sctp_transport_hold(struct sctp_transport *transport)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319{
Reshetova, Elenaa4b2b582017-07-04 15:53:27 +0300320 return refcount_inc_not_zero(&transport->refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321}
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{
Reshetova, Elenaa4b2b582017-07-04 15:53:27 +0300328 if (refcount_dec_and_test(&transport->refcnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329 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{
Daniel Borkmannbb333812013-06-28 19:49:40 +0200335 if (unlikely(!tp->rto_pending))
336 /* We should not be doing any RTO updates unless rto_pending is set. */
337 pr_debug("%s: rto_pending not set on transport %p!\n", __func__, tp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338
339 if (tp->rttvar || tp->srtt) {
Eric W. Biedermane1fc3b12012-08-07 07:29:57 +0000340 struct net *net = sock_net(tp->asoc->base.sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 /* 6.3.1 C3) When a new RTT measurement R' is made, set
342 * RTTVAR <- (1 - RTO.Beta) * RTTVAR + RTO.Beta * |SRTT - R'|
343 * SRTT <- (1 - RTO.Alpha) * SRTT + RTO.Alpha * R'
344 */
345
346 /* Note: The above algorithm has been rewritten to
347 * express rto_beta and rto_alpha as inverse powers
348 * of two.
349 * For example, assuming the default value of RTO.Alpha of
350 * 1/8, rto_alpha would be expressed as 3.
351 */
Eric W. Biedermane1fc3b12012-08-07 07:29:57 +0000352 tp->rttvar = tp->rttvar - (tp->rttvar >> net->sctp.rto_beta)
Andrew Morton79211c82015-11-09 14:58:13 -0800353 + (((__u32)abs((__s64)tp->srtt - (__s64)rtt)) >> net->sctp.rto_beta);
Eric W. Biedermane1fc3b12012-08-07 07:29:57 +0000354 tp->srtt = tp->srtt - (tp->srtt >> net->sctp.rto_alpha)
355 + (rtt >> net->sctp.rto_alpha);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 } else {
357 /* 6.3.1 C2) When the first RTT measurement R is made, set
358 * SRTT <- R, RTTVAR <- R/2.
359 */
360 tp->srtt = rtt;
361 tp->rttvar = rtt >> 1;
362 }
363
364 /* 6.3.1 G1) Whenever RTTVAR is computed, if RTTVAR = 0, then
365 * adjust RTTVAR <- G, where G is the CLOCK GRANULARITY.
366 */
367 if (tp->rttvar == 0)
368 tp->rttvar = SCTP_CLOCK_GRANULARITY;
369
370 /* 6.3.1 C3) After the computation, update RTO <- SRTT + 4 * RTTVAR. */
371 tp->rto = tp->srtt + (tp->rttvar << 2);
372
373 /* 6.3.1 C6) Whenever RTO is computed, if it is less than RTO.Min
374 * seconds then it is rounded up to RTO.Min seconds.
375 */
376 if (tp->rto < tp->asoc->rto_min)
377 tp->rto = tp->asoc->rto_min;
378
379 /* 6.3.1 C7) A maximum value may be placed on RTO provided it is
380 * at least RTO.max seconds.
381 */
382 if (tp->rto > tp->asoc->rto_max)
383 tp->rto = tp->asoc->rto_max;
384
Michele Baldessari196d6752012-12-01 04:49:42 +0000385 sctp_max_rto(tp->asoc, tp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 tp->rtt = rtt;
387
388 /* Reset rto_pending so that a new RTT measurement is started when a
389 * new data chunk is sent.
390 */
391 tp->rto_pending = 0;
392
Daniel Borkmannbb333812013-06-28 19:49:40 +0200393 pr_debug("%s: transport:%p, rtt:%d, srtt:%d rttvar:%d, rto:%ld\n",
394 __func__, tp, rtt, tp->srtt, tp->rttvar, tp->rto);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395}
396
397/* This routine updates the transport's cwnd and partial_bytes_acked
398 * parameters based on the bytes acked in the received SACK.
399 */
400void sctp_transport_raise_cwnd(struct sctp_transport *transport,
401 __u32 sack_ctsn, __u32 bytes_acked)
402{
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400403 struct sctp_association *asoc = transport->asoc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 __u32 cwnd, ssthresh, flight_size, pba, pmtu;
405
406 cwnd = transport->cwnd;
407 flight_size = transport->flight_size;
408
Vlad Yasevicha6465232008-06-04 12:38:43 -0700409 /* See if we need to exit Fast Recovery first */
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400410 if (asoc->fast_recovery &&
411 TSN_lte(asoc->fast_recovery_exit, sack_ctsn))
412 asoc->fast_recovery = 0;
Vlad Yasevicha6465232008-06-04 12:38:43 -0700413
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 ssthresh = transport->ssthresh;
415 pba = transport->partial_bytes_acked;
Frank Filz52ccb8e2005-12-22 11:36:46 -0800416 pmtu = transport->asoc->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417
418 if (cwnd <= ssthresh) {
Vlad Yasevicha6465232008-06-04 12:38:43 -0700419 /* RFC 4960 7.2.1
420 * o When cwnd is less than or equal to ssthresh, an SCTP
421 * endpoint MUST use the slow-start algorithm to increase
422 * cwnd only if the current congestion window is being fully
423 * utilized, an incoming SACK advances the Cumulative TSN
424 * Ack Point, and the data sender is not in Fast Recovery.
425 * Only when these three conditions are met can the cwnd be
426 * increased; otherwise, the cwnd MUST not be increased.
427 * If these conditions are met, then cwnd MUST be increased
428 * by, at most, the lesser of 1) the total size of the
429 * previously outstanding DATA chunk(s) acknowledged, and
430 * 2) the destination's path MTU. This upper bound protects
431 * against the ACK-Splitting attack outlined in [SAVAGE99].
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 */
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400433 if (asoc->fast_recovery)
Vlad Yasevicha6465232008-06-04 12:38:43 -0700434 return;
435
Marcelo Ricardo Leitner4ccbd0b2017-06-23 19:59:35 -0300436 /* The appropriate cwnd increase algorithm is performed
437 * if, and only if the congestion window is being fully
438 * utilized. Note that RFC4960 Errata 3.22 removed the
439 * other condition on ctsn moving.
440 */
441 if (flight_size < cwnd)
442 return;
443
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 if (bytes_acked > pmtu)
445 cwnd += pmtu;
446 else
447 cwnd += bytes_acked;
Daniel Borkmannbb333812013-06-28 19:49:40 +0200448
449 pr_debug("%s: slow start: transport:%p, bytes_acked:%d, "
450 "cwnd:%d, ssthresh:%d, flight_size:%d, pba:%d\n",
451 __func__, transport, bytes_acked, cwnd, ssthresh,
452 flight_size, pba);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 } else {
454 /* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh,
Marcelo Ricardo Leitnere56f777a2017-06-23 19:59:34 -0300455 * upon each SACK arrival, increase partial_bytes_acked
456 * by the total number of bytes of all new chunks
457 * acknowledged in that SACK including chunks
458 * acknowledged by the new Cumulative TSN Ack and by Gap
459 * Ack Blocks. (updated by RFC4960 Errata 3.22)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 *
Marcelo Ricardo Leitner4ccbd0b2017-06-23 19:59:35 -0300461 * When partial_bytes_acked is greater than cwnd and
462 * before the arrival of the SACK the sender had less
463 * bytes of data outstanding than cwnd (i.e., before
464 * arrival of the SACK, flightsize was less than cwnd),
465 * reset partial_bytes_acked to cwnd. (RFC 4960 Errata
466 * 3.26)
467 *
Marcelo Ricardo Leitnerd0b53f42017-06-23 19:59:33 -0300468 * When partial_bytes_acked is equal to or greater than
469 * cwnd and before the arrival of the SACK the sender
470 * had cwnd or more bytes of data outstanding (i.e.,
471 * before arrival of the SACK, flightsize was greater
472 * than or equal to cwnd), partial_bytes_acked is reset
473 * to (partial_bytes_acked - cwnd). Next, cwnd is
474 * increased by MTU. (RFC 4960 Errata 3.12)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475 */
476 pba += bytes_acked;
Marcelo Ricardo Leitner4ccbd0b2017-06-23 19:59:35 -0300477 if (pba > cwnd && flight_size < cwnd)
478 pba = cwnd;
479 if (pba >= cwnd && flight_size >= cwnd) {
Marcelo Ricardo Leitnerd0b53f42017-06-23 19:59:33 -0300480 pba = pba - cwnd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 cwnd += pmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 }
Daniel Borkmannbb333812013-06-28 19:49:40 +0200483
484 pr_debug("%s: congestion avoidance: transport:%p, "
485 "bytes_acked:%d, cwnd:%d, ssthresh:%d, "
486 "flight_size:%d, pba:%d\n", __func__,
487 transport, bytes_acked, cwnd, ssthresh,
488 flight_size, pba);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 }
490
491 transport->cwnd = cwnd;
492 transport->partial_bytes_acked = pba;
493}
494
495/* This routine is used to lower the transport's cwnd when congestion is
496 * detected.
497 */
498void sctp_transport_lower_cwnd(struct sctp_transport *transport,
Xin Long233e7932017-08-05 19:59:51 +0800499 enum sctp_lower_cwnd reason)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500{
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400501 struct sctp_association *asoc = transport->asoc;
502
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 switch (reason) {
504 case SCTP_LOWER_CWND_T3_RTX:
505 /* RFC 2960 Section 7.2.3, sctpimpguide
506 * When the T3-rtx timer expires on an address, SCTP should
507 * perform slow start by:
508 * ssthresh = max(cwnd/2, 4*MTU)
509 * cwnd = 1*MTU
510 * partial_bytes_acked = 0
511 */
512 transport->ssthresh = max(transport->cwnd/2,
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400513 4*asoc->pathmtu);
514 transport->cwnd = asoc->pathmtu;
Vlad Yasevich33ce8282009-09-04 18:20:58 -0400515
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400516 /* T3-rtx also clears fast recovery */
517 asoc->fast_recovery = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 break;
519
520 case SCTP_LOWER_CWND_FAST_RTX:
521 /* RFC 2960 7.2.4 Adjust the ssthresh and cwnd of the
522 * destination address(es) to which the missing DATA chunks
523 * were last sent, according to the formula described in
524 * Section 7.2.3.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900525 *
526 * RFC 2960 7.2.3, sctpimpguide Upon detection of packet
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 * losses from SACK (see Section 7.2.4), An endpoint
528 * should do the following:
529 * ssthresh = max(cwnd/2, 4*MTU)
530 * cwnd = ssthresh
531 * partial_bytes_acked = 0
532 */
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400533 if (asoc->fast_recovery)
Vlad Yasevicha6465232008-06-04 12:38:43 -0700534 return;
535
536 /* Mark Fast recovery */
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400537 asoc->fast_recovery = 1;
538 asoc->fast_recovery_exit = asoc->next_tsn - 1;
Vlad Yasevicha6465232008-06-04 12:38:43 -0700539
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 transport->ssthresh = max(transport->cwnd/2,
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400541 4*asoc->pathmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 transport->cwnd = transport->ssthresh;
543 break;
544
545 case SCTP_LOWER_CWND_ECNE:
546 /* RFC 2481 Section 6.1.2.
547 * If the sender receives an ECN-Echo ACK packet
548 * then the sender knows that congestion was encountered in the
549 * network on the path from the sender to the receiver. The
550 * indication of congestion should be treated just as a
551 * congestion loss in non-ECN Capable TCP. That is, the TCP
552 * source halves the congestion window "cwnd" and reduces the
553 * slow start threshold "ssthresh".
554 * A critical condition is that TCP does not react to
555 * congestion indications more than once every window of
556 * data (or more loosely more than once every round-trip time).
557 */
Wei Yongjunf61f6f82009-03-02 09:46:13 +0000558 if (time_after(jiffies, transport->last_time_ecne_reduced +
559 transport->rtt)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 transport->ssthresh = max(transport->cwnd/2,
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400561 4*asoc->pathmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 transport->cwnd = transport->ssthresh;
563 transport->last_time_ecne_reduced = jiffies;
564 }
565 break;
566
567 case SCTP_LOWER_CWND_INACTIVE:
568 /* RFC 2960 Section 7.2.1, sctpimpguide
569 * When the endpoint does not transmit data on a given
570 * transport address, the cwnd of the transport address
571 * should be adjusted to max(cwnd/2, 4*MTU) per RTO.
572 * NOTE: Although the draft recommends that this check needs
573 * to be done every RTO interval, we do it every hearbeat
574 * interval.
575 */
Vlad Yasevich245cba72009-11-23 15:53:58 -0500576 transport->cwnd = max(transport->cwnd/2,
Vlad Yasevichcf9b4812010-04-30 22:41:10 -0400577 4*asoc->pathmtu);
Marcelo Ricardo Leitnera02d0362017-06-23 19:59:36 -0300578 /* RFC 4960 Errata 3.27.2: also adjust sshthresh */
579 transport->ssthresh = transport->cwnd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 break;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -0700581 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582
583 transport->partial_bytes_acked = 0;
Daniel Borkmannbb333812013-06-28 19:49:40 +0200584
585 pr_debug("%s: transport:%p, reason:%d, cwnd:%d, ssthresh:%d\n",
586 __func__, transport, reason, transport->cwnd,
587 transport->ssthresh);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588}
589
Vlad Yasevich46d5a802009-11-23 15:54:00 -0500590/* Apply Max.Burst limit to the congestion window:
591 * sctpimpguide-05 2.14.2
592 * D) When the time comes for the sender to
593 * transmit new DATA chunks, the protocol parameter Max.Burst MUST
594 * first be applied to limit how many new DATA chunks may be sent.
595 * The limit is applied by adjusting cwnd as follows:
596 * if ((flightsize+ Max.Burst * MTU) < cwnd)
597 * cwnd = flightsize + Max.Burst * MTU
598 */
599
600void sctp_transport_burst_limited(struct sctp_transport *t)
601{
602 struct sctp_association *asoc = t->asoc;
603 u32 old_cwnd = t->cwnd;
604 u32 max_burst_bytes;
605
wangweidong78ac8142013-12-04 17:32:39 +0800606 if (t->burst_limited || asoc->max_burst == 0)
Vlad Yasevich46d5a802009-11-23 15:54:00 -0500607 return;
608
609 max_burst_bytes = t->flight_size + (asoc->max_burst * asoc->pathmtu);
610 if (max_burst_bytes < old_cwnd) {
611 t->cwnd = max_burst_bytes;
612 t->burst_limited = old_cwnd;
613 }
614}
615
616/* Restore the old cwnd congestion window, after the burst had it's
617 * desired effect.
618 */
619void sctp_transport_burst_reset(struct sctp_transport *t)
620{
621 if (t->burst_limited) {
622 t->cwnd = t->burst_limited;
623 t->burst_limited = 0;
624 }
625}
626
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627/* What is the next timeout value for this transport? */
Daniel Borkmann8f610592014-06-30 13:52:08 +0200628unsigned long sctp_transport_timeout(struct sctp_transport *trans)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629{
Daniel Borkmann8f610592014-06-30 13:52:08 +0200630 /* RTO + timer slack +/- 50% of RTO */
Marcelo Ricardo Leitnerba6f5e32016-04-06 15:15:19 -0300631 unsigned long timeout = trans->rto >> 1;
Daniel Borkmann8f610592014-06-30 13:52:08 +0200632
633 if (trans->state != SCTP_UNCONFIRMED &&
634 trans->state != SCTP_PF)
635 timeout += trans->hbinterval;
636
Marcelo Ricardo Leitnerba6f5e32016-04-06 15:15:19 -0300637 return timeout;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638}
Vlad Yasevich749bf922007-03-19 17:02:30 -0700639
640/* Reset transport variables to their initial values */
641void sctp_transport_reset(struct sctp_transport *t)
642{
643 struct sctp_association *asoc = t->asoc;
644
645 /* RFC 2960 (bis), Section 5.2.4
646 * All the congestion control parameters (e.g., cwnd, ssthresh)
647 * related to this peer MUST be reset to their initial values
648 * (see Section 6.2.1)
649 */
650 t->cwnd = min(4*asoc->pathmtu, max_t(__u32, 2*asoc->pathmtu, 4380));
Vlad Yasevich46d5a802009-11-23 15:54:00 -0500651 t->burst_limited = 0;
Vlad Yasevich289f4242007-03-22 12:26:25 -0700652 t->ssthresh = asoc->peer.i.a_rwnd;
Andrei Pelinescu-Onciul5fdd4ba2009-11-29 00:14:02 -0800653 t->rto = asoc->rto_initial;
Michele Baldessari196d6752012-12-01 04:49:42 +0000654 sctp_max_rto(asoc, t);
Vlad Yasevich749bf922007-03-19 17:02:30 -0700655 t->rtt = 0;
656 t->srtt = 0;
657 t->rttvar = 0;
658
Masahiro Yamadab564d622017-02-27 14:29:06 -0800659 /* Reset these additional variables so that we have a clean slate. */
Vlad Yasevich749bf922007-03-19 17:02:30 -0700660 t->partial_bytes_acked = 0;
661 t->flight_size = 0;
662 t->error_count = 0;
663 t->rto_pending = 0;
Vlad Yasevichfaee47c2009-02-13 08:33:43 +0000664 t->hb_sent = 0;
Vlad Yasevich749bf922007-03-19 17:02:30 -0700665
666 /* Initialize the state information for SFR-CACC */
667 t->cacc.changeover_active = 0;
668 t->cacc.cycling_changeover = 0;
669 t->cacc.next_tsn_at_change = 0;
670 t->cacc.cacc_saw_newack = 0;
671}
Michio Hondaddc4bbe2011-06-17 11:03:23 +0900672
673/* Schedule retransmission on the given transport */
674void sctp_transport_immediate_rtx(struct sctp_transport *t)
675{
676 /* Stop pending T3_rtx_timer */
Ying Xue25cc4ae2013-02-03 20:32:57 +0000677 if (del_timer(&t->T3_rtx_timer))
Michio Hondaddc4bbe2011-06-17 11:03:23 +0900678 sctp_transport_put(t);
Ying Xue25cc4ae2013-02-03 20:32:57 +0000679
Michio Hondaddc4bbe2011-06-17 11:03:23 +0900680 sctp_retransmit(&t->asoc->outqueue, t, SCTP_RTXR_T3_RTX);
681 if (!timer_pending(&t->T3_rtx_timer)) {
682 if (!mod_timer(&t->T3_rtx_timer, jiffies + t->rto))
683 sctp_transport_hold(t);
684 }
Michio Hondaddc4bbe2011-06-17 11:03:23 +0900685}
Julian Anastasovc86a7732017-02-06 23:14:13 +0200686
687/* Drop dst */
688void sctp_transport_dst_release(struct sctp_transport *t)
689{
690 dst_release(t->dst);
691 t->dst = NULL;
692 t->dst_pending_confirm = 0;
693}
694
695/* Schedule neighbour confirm */
696void sctp_transport_dst_confirm(struct sctp_transport *t)
697{
698 t->dst_pending_confirm = 1;
699}