]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - net/sctp/transport.c
Merge tag 'trace-v4.15-rc4-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rosted...
[mirror_ubuntu-bionic-kernel.git] / net / sctp / transport.c
CommitLineData
60c778b2 1/* SCTP kernel implementation
1da177e4
LT
2 * Copyright (c) 1999-2000 Cisco, Inc.
3 * Copyright (c) 1999-2001 Motorola, Inc.
4 * Copyright (c) 2001-2003 International Business Machines Corp.
5 * Copyright (c) 2001 Intel Corp.
6 * Copyright (c) 2001 La Monte H.P. Yarroll
7 *
60c778b2 8 * This file is part of the SCTP kernel implementation
1da177e4
LT
9 *
10 * This module provides the abstraction for an SCTP tranport representing
11 * a remote transport address. For local transport addresses, we just use
12 * union sctp_addr.
13 *
60c778b2 14 * This SCTP implementation is free software;
1da177e4
LT
15 * you can redistribute it and/or modify it under the terms of
16 * the GNU General Public License as published by
17 * the Free Software Foundation; either version 2, or (at your option)
18 * any later version.
19 *
60c778b2 20 * This SCTP implementation is distributed in the hope that it
1da177e4
LT
21 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
22 * ************************
23 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
24 * See the GNU General Public License for more details.
25 *
26 * You should have received a copy of the GNU General Public License
4b2f13a2
JK
27 * along with GNU CC; see the file COPYING. If not, see
28 * <http://www.gnu.org/licenses/>.
1da177e4
LT
29 *
30 * Please send any bug reports or fixes you make to the
31 * email address(es):
91705c61 32 * lksctp developers <linux-sctp@vger.kernel.org>
1da177e4 33 *
1da177e4
LT
34 * Written or modified by:
35 * La Monte H.P. Yarroll <piggy@acm.org>
36 * Karl Knutson <karl@athena.chicago.il.us>
37 * Jon Grimm <jgrimm@us.ibm.com>
38 * Xingang Guo <xingang.guo@intel.com>
39 * Hui Huang <hui.huang@nokia.com>
40 * Sridhar Samudrala <sri@us.ibm.com>
41 * Ardelle Fan <ardelle.fan@intel.com>
1da177e4
LT
42 */
43
145ce502
JP
44#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
45
5a0e3ad6 46#include <linux/slab.h>
1da177e4 47#include <linux/types.h>
ad8fec17 48#include <linux/random.h>
1da177e4
LT
49#include <net/sctp/sctp.h>
50#include <net/sctp/sm.h>
51
52/* 1st Level Abstractions. */
53
54/* Initialize a new transport from provided memory. */
89bf3450
EB
55static struct sctp_transport *sctp_transport_init(struct net *net,
56 struct sctp_transport *peer,
1da177e4 57 const union sctp_addr *addr,
dd0fc66f 58 gfp_t gfp)
1da177e4
LT
59{
60 /* Copy in the address. */
61 peer->ipaddr = *addr;
62 peer->af_specific = sctp_get_af_specific(addr->sa.sa_family);
1da177e4
LT
63 memset(&peer->saddr, 0, sizeof(union sctp_addr));
64
4244854d
NH
65 peer->sack_generation = 0;
66
1da177e4
LT
67 /* From 6.3.1 RTO Calculation:
68 *
69 * C1) Until an RTT measurement has been made for a packet sent to the
70 * given destination transport address, set RTO to the protocol
71 * parameter 'RTO.Initial'.
72 */
e1fc3b14 73 peer->rto = msecs_to_jiffies(net->sctp.rto_initial);
1da177e4 74
8b0e1953 75 peer->last_time_heard = 0;
1da177e4
LT
76 peer->last_time_ecne_reduced = jiffies;
77
52ccb8e9
FF
78 peer->param_flags = SPP_HB_DISABLE |
79 SPP_PMTUD_ENABLE |
80 SPP_SACKDELAY_ENABLE;
1da177e4
LT
81
82 /* Initialize the default path max_retrans. */
e1fc3b14
EB
83 peer->pathmaxrxt = net->sctp.max_retrans_path;
84 peer->pf_retrans = net->sctp.pf_retrans;
1da177e4
LT
85
86 INIT_LIST_HEAD(&peer->transmitted);
87 INIT_LIST_HEAD(&peer->send_ready);
88 INIT_LIST_HEAD(&peer->transports);
89
9c3b5751
KC
90 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);
1da177e4 95
ad8fec17
SS
96 /* Initialize the 64-bit random nonce sent with heartbeat. */
97 get_random_bytes(&peer->hb_nonce, sizeof(peer->hb_nonce));
98
a4b2b58e 99 refcount_set(&peer->refcnt, 1);
1da177e4
LT
100
101 return peer;
102}
103
104/* Allocate and initialize a new transport. */
89bf3450
EB
105struct sctp_transport *sctp_transport_new(struct net *net,
106 const union sctp_addr *addr,
dd0fc66f 107 gfp_t gfp)
1da177e4 108{
d808ad9a 109 struct sctp_transport *transport;
1da177e4 110
939cfa75 111 transport = kzalloc(sizeof(*transport), gfp);
1da177e4
LT
112 if (!transport)
113 goto fail;
114
89bf3450 115 if (!sctp_transport_init(net, transport, addr, gfp))
1da177e4
LT
116 goto fail_init;
117
1da177e4
LT
118 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{
1da177e4
LT
134 /* 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 */
25cc4ae9 143 if (del_timer(&transport->T3_rtx_timer))
1da177e4
LT
144 sctp_transport_put(transport);
145
7b9438de
XL
146 if (del_timer(&transport->reconf_timer))
147 sctp_transport_put(transport);
148
55fa0cfd 149 /* Delete the ICMP proto unreachable timer if it's active. */
25cc4ae9 150 if (del_timer(&transport->proto_unreach_timer))
55fa0cfd 151 sctp_association_put(transport->asoc);
1da177e4
LT
152
153 sctp_transport_put(transport);
154}
155
45122ca2 156static void sctp_transport_destroy_rcu(struct rcu_head *head)
1da177e4 157{
45122ca2 158 struct sctp_transport *transport;
1da177e4 159
45122ca2 160 transport = container_of(head, struct sctp_transport, rcu);
1da177e4
LT
161
162 dst_release(transport->dst);
163 kfree(transport);
164 SCTP_DBG_OBJCNT_DEC(transport);
165}
166
45122ca2
TG
167/* 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{
a4b2b58e 172 if (unlikely(refcount_read(&transport->refcnt))) {
bb33381d
DB
173 WARN(1, "Attempt to destroy undead transport %p!\n", transport);
174 return;
175 }
45122ca2 176
8c98653f
DB
177 sctp_packet_free(&transport->packet);
178
179 if (transport->asoc)
180 sctp_association_put(transport->asoc);
771085d6
DB
181
182 call_rcu(&transport->rcu, sctp_transport_destroy_rcu);
45122ca2
TG
183}
184
1da177e4
LT
185/* 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 */
ba6f5e33 188void sctp_transport_reset_t3_rtx(struct sctp_transport *transport)
1da177e4
LT
189{
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
d9efc223 198 if (!timer_pending(&transport->T3_rtx_timer))
1da177e4
LT
199 if (!mod_timer(&transport->T3_rtx_timer,
200 jiffies + transport->rto))
201 sctp_transport_hold(transport);
ba6f5e33
MRL
202}
203
204void sctp_transport_reset_hb_timer(struct sctp_transport *transport)
205{
206 unsigned long expires;
1da177e4
LT
207
208 /* When a data chunk is sent, reset the heartbeat interval. */
ba6f5e33
MRL
209 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);
1da177e4
LT
214}
215
7b9438de
XL
216void 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
1da177e4
LT
224/* 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. */
9914ae3c 236void sctp_transport_pmtu(struct sctp_transport *transport, struct sock *sk)
1da177e4 237{
da0420be 238 /* If we don't have a fresh route, look one up */
f5b0a874 239 if (!transport->dst || transport->dst->obsolete) {
c86a773c 240 sctp_transport_dst_release(transport);
da0420be 241 transport->af_specific->get_dst(transport, &transport->saddr,
8663c938 242 &transport->fl, sk);
da0420be 243 }
1da177e4 244
da0420be 245 if (transport->dst) {
e2f036a9 246 transport->pathmtu = SCTP_TRUNC4(dst_mtu(transport->dst));
1da177e4 247 } else
52ccb8e9 248 transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
1da177e4
LT
249}
250
b6c5734d 251bool sctp_transport_update_pmtu(struct sctp_transport *t, u32 pmtu)
c910b47e 252{
3ebfdf08 253 struct dst_entry *dst = sctp_transport_dst_check(t);
b6c5734d 254 bool change = true;
c910b47e
VY
255
256 if (unlikely(pmtu < SCTP_DEFAULT_MINSEGMENT)) {
b6c5734d
MRL
257 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;
c910b47e 261 }
b6c5734d 262 pmtu = SCTP_TRUNC4(pmtu);
c910b47e 263
02f3d4ce 264 if (dst) {
3ebfdf08 265 dst->ops->update_pmtu(dst, t->asoc->base.sk, NULL, pmtu);
02f3d4ce 266 dst = sctp_transport_dst_check(t);
02f3d4ce 267 }
3ebfdf08 268
b6c5734d 269 if (!dst) {
3ebfdf08 270 t->af_specific->get_dst(t, &t->saddr, &t->fl, t->asoc->base.sk);
b6c5734d
MRL
271 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;
c910b47e
VY
282}
283
1da177e4
LT
284/* 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;
1da177e4 292
8663c938 293 af->get_dst(transport, saddr, &transport->fl, sctp_opt2sk(opt));
1da177e4
LT
294
295 if (saddr)
296 memcpy(&transport->saddr, saddr, sizeof(union sctp_addr));
297 else
8663c938 298 af->get_saddr(opt, transport, &transport->fl);
1da177e4 299
52ccb8e9
FF
300 if ((transport->param_flags & SPP_PMTUD_DISABLE) && transport->pathmtu) {
301 return;
302 }
da0420be 303 if (transport->dst) {
e2f036a9 304 transport->pathmtu = SCTP_TRUNC4(dst_mtu(transport->dst));
1da177e4
LT
305
306 /* Initialize sk->sk_rcv_saddr, if the transport is the
307 * association's active path for getsockname().
d808ad9a 308 */
a78102e7
VY
309 if (asoc && (!asoc->peer.primary_path ||
310 (transport == asoc->peer.active_path)))
299ee123
JG
311 opt->pf->to_sk_saddr(&transport->saddr,
312 asoc->base.sk);
1da177e4 313 } else
52ccb8e9 314 transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
1da177e4
LT
315}
316
317/* Hold a reference to a transport. */
1eed6779 318int sctp_transport_hold(struct sctp_transport *transport)
1da177e4 319{
a4b2b58e 320 return refcount_inc_not_zero(&transport->refcnt);
1da177e4
LT
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{
a4b2b58e 328 if (refcount_dec_and_test(&transport->refcnt))
1da177e4
LT
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{
bb33381d
DB
335 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);
1da177e4
LT
338
339 if (tp->rttvar || tp->srtt) {
e1fc3b14 340 struct net *net = sock_net(tp->asoc->base.sk);
1da177e4
LT
341 /* 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 */
e1fc3b14 352 tp->rttvar = tp->rttvar - (tp->rttvar >> net->sctp.rto_beta)
79211c8e 353 + (((__u32)abs((__s64)tp->srtt - (__s64)rtt)) >> net->sctp.rto_beta);
e1fc3b14
EB
354 tp->srtt = tp->srtt - (tp->srtt >> net->sctp.rto_alpha)
355 + (rtt >> net->sctp.rto_alpha);
1da177e4
LT
356 } 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
196d6759 385 sctp_max_rto(tp->asoc, tp);
1da177e4
LT
386 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
bb33381d
DB
393 pr_debug("%s: transport:%p, rtt:%d, srtt:%d rttvar:%d, rto:%ld\n",
394 __func__, tp, rtt, tp->srtt, tp->rttvar, tp->rto);
1da177e4
LT
395}
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{
cf9b4812 403 struct sctp_association *asoc = transport->asoc;
1da177e4
LT
404 __u32 cwnd, ssthresh, flight_size, pba, pmtu;
405
406 cwnd = transport->cwnd;
407 flight_size = transport->flight_size;
408
a6465234 409 /* See if we need to exit Fast Recovery first */
cf9b4812
VY
410 if (asoc->fast_recovery &&
411 TSN_lte(asoc->fast_recovery_exit, sack_ctsn))
412 asoc->fast_recovery = 0;
a6465234 413
1da177e4
LT
414 ssthresh = transport->ssthresh;
415 pba = transport->partial_bytes_acked;
52ccb8e9 416 pmtu = transport->asoc->pathmtu;
1da177e4
LT
417
418 if (cwnd <= ssthresh) {
a6465234
VY
419 /* 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].
1da177e4 432 */
cf9b4812 433 if (asoc->fast_recovery)
a6465234
VY
434 return;
435
4ccbd0b0
MRL
436 /* 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
1da177e4
LT
444 if (bytes_acked > pmtu)
445 cwnd += pmtu;
446 else
447 cwnd += bytes_acked;
bb33381d
DB
448
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);
1da177e4
LT
453 } else {
454 /* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh,
e56f777a
MRL
455 * 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)
1da177e4 460 *
4ccbd0b0
MRL
461 * 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 *
d0b53f40
MRL
468 * 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)
1da177e4
LT
475 */
476 pba += bytes_acked;
4ccbd0b0
MRL
477 if (pba > cwnd && flight_size < cwnd)
478 pba = cwnd;
479 if (pba >= cwnd && flight_size >= cwnd) {
d0b53f40 480 pba = pba - cwnd;
1da177e4 481 cwnd += pmtu;
1da177e4 482 }
bb33381d
DB
483
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);
1da177e4
LT
489 }
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,
233e7936 499 enum sctp_lower_cwnd reason)
1da177e4 500{
cf9b4812
VY
501 struct sctp_association *asoc = transport->asoc;
502
1da177e4
LT
503 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,
cf9b4812
VY
513 4*asoc->pathmtu);
514 transport->cwnd = asoc->pathmtu;
33ce8281 515
cf9b4812
VY
516 /* T3-rtx also clears fast recovery */
517 asoc->fast_recovery = 0;
1da177e4
LT
518 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.
d808ad9a
YH
525 *
526 * RFC 2960 7.2.3, sctpimpguide Upon detection of packet
1da177e4
LT
527 * 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 */
cf9b4812 533 if (asoc->fast_recovery)
a6465234
VY
534 return;
535
536 /* Mark Fast recovery */
cf9b4812
VY
537 asoc->fast_recovery = 1;
538 asoc->fast_recovery_exit = asoc->next_tsn - 1;
a6465234 539
1da177e4 540 transport->ssthresh = max(transport->cwnd/2,
cf9b4812 541 4*asoc->pathmtu);
1da177e4
LT
542 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 */
f61f6f82
WY
558 if (time_after(jiffies, transport->last_time_ecne_reduced +
559 transport->rtt)) {
1da177e4 560 transport->ssthresh = max(transport->cwnd/2,
cf9b4812 561 4*asoc->pathmtu);
1da177e4
LT
562 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 */
245cba7e 576 transport->cwnd = max(transport->cwnd/2,
cf9b4812 577 4*asoc->pathmtu);
a02d036c
MRL
578 /* RFC 4960 Errata 3.27.2: also adjust sshthresh */
579 transport->ssthresh = transport->cwnd;
1da177e4 580 break;
3ff50b79 581 }
1da177e4
LT
582
583 transport->partial_bytes_acked = 0;
bb33381d
DB
584
585 pr_debug("%s: transport:%p, reason:%d, cwnd:%d, ssthresh:%d\n",
586 __func__, transport, reason, transport->cwnd,
587 transport->ssthresh);
1da177e4
LT
588}
589
46d5a808
VY
590/* 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
78ac814f 606 if (t->burst_limited || asoc->max_burst == 0)
46d5a808
VY
607 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
1da177e4 627/* What is the next timeout value for this transport? */
8f61059a 628unsigned long sctp_transport_timeout(struct sctp_transport *trans)
1da177e4 629{
8f61059a 630 /* RTO + timer slack +/- 50% of RTO */
ba6f5e33 631 unsigned long timeout = trans->rto >> 1;
8f61059a
DB
632
633 if (trans->state != SCTP_UNCONFIRMED &&
634 trans->state != SCTP_PF)
635 timeout += trans->hbinterval;
636
ba6f5e33 637 return timeout;
1da177e4 638}
749bf921
VY
639
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));
46d5a808 651 t->burst_limited = 0;
289f4249 652 t->ssthresh = asoc->peer.i.a_rwnd;
5fdd4bae 653 t->rto = asoc->rto_initial;
196d6759 654 sctp_max_rto(asoc, t);
749bf921
VY
655 t->rtt = 0;
656 t->srtt = 0;
657 t->rttvar = 0;
658
b564d62e 659 /* Reset these additional variables so that we have a clean slate. */
749bf921
VY
660 t->partial_bytes_acked = 0;
661 t->flight_size = 0;
662 t->error_count = 0;
663 t->rto_pending = 0;
faee47cd 664 t->hb_sent = 0;
749bf921
VY
665
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}
ddc4bbee
MH
672
673/* Schedule retransmission on the given transport */
674void sctp_transport_immediate_rtx(struct sctp_transport *t)
675{
676 /* Stop pending T3_rtx_timer */
25cc4ae9 677 if (del_timer(&t->T3_rtx_timer))
ddc4bbee 678 sctp_transport_put(t);
25cc4ae9 679
ddc4bbee
MH
680 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 }
ddc4bbee 685}
c86a773c
JA
686
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}