1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
45
46#include <linux/slab.h>
47#include <linux/types.h>
48#include <linux/random.h>
49#include <net/sctp/sctp.h>
50#include <net/sctp/sm.h>
51
52
53
54
55static struct sctp_transport *sctp_transport_init(struct net *net,
56 struct sctp_transport *peer,
57 const union sctp_addr *addr,
58 gfp_t gfp)
59{
60
61 peer->af_specific = sctp_get_af_specific(addr->sa.sa_family);
62 memcpy(&peer->ipaddr, addr, peer->af_specific->sockaddr_len);
63 memset(&peer->saddr, 0, sizeof(union sctp_addr));
64
65 peer->sack_generation = 0;
66
67
68
69
70
71
72
73 peer->rto = msecs_to_jiffies(net->sctp.rto_initial);
74
75 peer->last_time_heard = 0;
76 peer->last_time_ecne_reduced = jiffies;
77
78 peer->param_flags = SPP_HB_DISABLE |
79 SPP_PMTUD_ENABLE |
80 SPP_SACKDELAY_ENABLE;
81
82
83 peer->pathmaxrxt = net->sctp.max_retrans_path;
84 peer->pf_retrans = net->sctp.pf_retrans;
85
86 INIT_LIST_HEAD(&peer->transmitted);
87 INIT_LIST_HEAD(&peer->send_ready);
88 INIT_LIST_HEAD(&peer->transports);
89
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);
95
96
97 get_random_bytes(&peer->hb_nonce, sizeof(peer->hb_nonce));
98
99 refcount_set(&peer->refcnt, 1);
100
101 return peer;
102}
103
104
105struct sctp_transport *sctp_transport_new(struct net *net,
106 const union sctp_addr *addr,
107 gfp_t gfp)
108{
109 struct sctp_transport *transport;
110
111 transport = kzalloc(sizeof(*transport), gfp);
112 if (!transport)
113 goto fail;
114
115 if (!sctp_transport_init(net, transport, addr, gfp))
116 goto fail_init;
117
118 SCTP_DBG_OBJCNT_INC(transport);
119
120 return transport;
121
122fail_init:
123 kfree(transport);
124
125fail:
126 return NULL;
127}
128
129
130
131
132void sctp_transport_free(struct sctp_transport *transport)
133{
134
135 if (del_timer(&transport->hb_timer))
136 sctp_transport_put(transport);
137
138
139
140
141
142
143 if (del_timer(&transport->T3_rtx_timer))
144 sctp_transport_put(transport);
145
146 if (del_timer(&transport->reconf_timer))
147 sctp_transport_put(transport);
148
149
150 if (del_timer(&transport->proto_unreach_timer))
151 sctp_association_put(transport->asoc);
152
153 sctp_transport_put(transport);
154}
155
156static void sctp_transport_destroy_rcu(struct rcu_head *head)
157{
158 struct sctp_transport *transport;
159
160 transport = container_of(head, struct sctp_transport, rcu);
161
162 dst_release(transport->dst);
163 kfree(transport);
164 SCTP_DBG_OBJCNT_DEC(transport);
165}
166
167
168
169
170static void sctp_transport_destroy(struct sctp_transport *transport)
171{
172 if (unlikely(refcount_read(&transport->refcnt))) {
173 WARN(1, "Attempt to destroy undead transport %p!\n", transport);
174 return;
175 }
176
177 sctp_packet_free(&transport->packet);
178
179 if (transport->asoc)
180 sctp_association_put(transport->asoc);
181
182 call_rcu(&transport->rcu, sctp_transport_destroy_rcu);
183}
184
185
186
187
188void sctp_transport_reset_t3_rtx(struct sctp_transport *transport)
189{
190
191
192
193
194
195
196
197
198 if (!timer_pending(&transport->T3_rtx_timer))
199 if (!mod_timer(&transport->T3_rtx_timer,
200 jiffies + transport->rto))
201 sctp_transport_hold(transport);
202}
203
204void sctp_transport_reset_hb_timer(struct sctp_transport *transport)
205{
206 unsigned long expires;
207
208
209 expires = jiffies + sctp_transport_timeout(transport);
210 if ((time_before(transport->hb_timer.expires, expires) ||
211 !timer_pending(&transport->hb_timer)) &&
212 !mod_timer(&transport->hb_timer,
213 expires + prandom_u32_max(transport->rto)))
214 sctp_transport_hold(transport);
215}
216
217void sctp_transport_reset_reconf_timer(struct sctp_transport *transport)
218{
219 if (!timer_pending(&transport->reconf_timer))
220 if (!mod_timer(&transport->reconf_timer,
221 jiffies + transport->rto))
222 sctp_transport_hold(transport);
223}
224
225
226
227
228
229void sctp_transport_set_owner(struct sctp_transport *transport,
230 struct sctp_association *asoc)
231{
232 transport->asoc = asoc;
233 sctp_association_hold(asoc);
234}
235
236
237void sctp_transport_pmtu(struct sctp_transport *transport, struct sock *sk)
238{
239
240 if (!transport->dst || transport->dst->obsolete) {
241 sctp_transport_dst_release(transport);
242 transport->af_specific->get_dst(transport, &transport->saddr,
243 &transport->fl, sk);
244 }
245
246 if (transport->param_flags & SPP_PMTUD_DISABLE) {
247 struct sctp_association *asoc = transport->asoc;
248
249 if (!transport->pathmtu && asoc && asoc->pathmtu)
250 transport->pathmtu = asoc->pathmtu;
251 if (transport->pathmtu)
252 return;
253 }
254
255 if (transport->dst)
256 transport->pathmtu = sctp_dst_mtu(transport->dst);
257 else
258 transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
259}
260
261bool sctp_transport_update_pmtu(struct sctp_transport *t, u32 pmtu)
262{
263 struct dst_entry *dst = sctp_transport_dst_check(t);
264 struct sock *sk = t->asoc->base.sk;
265 bool change = true;
266
267 if (unlikely(pmtu < SCTP_DEFAULT_MINSEGMENT)) {
268 pr_warn_ratelimited("%s: Reported pmtu %d too low, using default minimum of %d\n",
269 __func__, pmtu, SCTP_DEFAULT_MINSEGMENT);
270
271 pmtu = SCTP_DEFAULT_MINSEGMENT;
272 }
273 pmtu = SCTP_TRUNC4(pmtu);
274
275 if (dst) {
276 struct sctp_pf *pf = sctp_get_pf_specific(dst->ops->family);
277 union sctp_addr addr;
278
279 pf->af->from_sk(&addr, sk);
280 pf->to_sk_daddr(&t->ipaddr, sk);
281 dst->ops->update_pmtu(dst, sk, NULL, pmtu);
282 pf->to_sk_daddr(&addr, sk);
283
284 dst = sctp_transport_dst_check(t);
285 }
286
287 if (!dst) {
288 t->af_specific->get_dst(t, &t->saddr, &t->fl, sk);
289 dst = t->dst;
290 }
291
292 if (dst) {
293
294 pmtu = sctp_dst_mtu(dst);
295 change = t->pathmtu != pmtu;
296 }
297 t->pathmtu = pmtu;
298
299 return change;
300}
301
302
303
304
305void sctp_transport_route(struct sctp_transport *transport,
306 union sctp_addr *saddr, struct sctp_sock *opt)
307{
308 struct sctp_association *asoc = transport->asoc;
309 struct sctp_af *af = transport->af_specific;
310
311 sctp_transport_dst_release(transport);
312 af->get_dst(transport, saddr, &transport->fl, sctp_opt2sk(opt));
313
314 if (saddr)
315 memcpy(&transport->saddr, saddr, sizeof(union sctp_addr));
316 else
317 af->get_saddr(opt, transport, &transport->fl);
318
319 sctp_transport_pmtu(transport, sctp_opt2sk(opt));
320
321
322
323
324 if (transport->dst && asoc &&
325 (!asoc->peer.primary_path || transport == asoc->peer.active_path))
326 opt->pf->to_sk_saddr(&transport->saddr, asoc->base.sk);
327}
328
329
330int sctp_transport_hold(struct sctp_transport *transport)
331{
332 return refcount_inc_not_zero(&transport->refcnt);
333}
334
335
336
337
338void sctp_transport_put(struct sctp_transport *transport)
339{
340 if (refcount_dec_and_test(&transport->refcnt))
341 sctp_transport_destroy(transport);
342}
343
344
345void sctp_transport_update_rto(struct sctp_transport *tp, __u32 rtt)
346{
347 if (unlikely(!tp->rto_pending))
348
349 pr_debug("%s: rto_pending not set on transport %p!\n", __func__, tp);
350
351 if (tp->rttvar || tp->srtt) {
352 struct net *net = sock_net(tp->asoc->base.sk);
353
354
355
356
357
358
359
360
361
362
363
364 tp->rttvar = tp->rttvar - (tp->rttvar >> net->sctp.rto_beta)
365 + (((__u32)abs((__s64)tp->srtt - (__s64)rtt)) >> net->sctp.rto_beta);
366 tp->srtt = tp->srtt - (tp->srtt >> net->sctp.rto_alpha)
367 + (rtt >> net->sctp.rto_alpha);
368 } else {
369
370
371
372 tp->srtt = rtt;
373 tp->rttvar = rtt >> 1;
374 }
375
376
377
378
379 if (tp->rttvar == 0)
380 tp->rttvar = SCTP_CLOCK_GRANULARITY;
381
382
383 tp->rto = tp->srtt + (tp->rttvar << 2);
384
385
386
387
388 if (tp->rto < tp->asoc->rto_min)
389 tp->rto = tp->asoc->rto_min;
390
391
392
393
394 if (tp->rto > tp->asoc->rto_max)
395 tp->rto = tp->asoc->rto_max;
396
397 sctp_max_rto(tp->asoc, tp);
398 tp->rtt = rtt;
399
400
401
402
403 tp->rto_pending = 0;
404
405 pr_debug("%s: transport:%p, rtt:%d, srtt:%d rttvar:%d, rto:%ld\n",
406 __func__, tp, rtt, tp->srtt, tp->rttvar, tp->rto);
407}
408
409
410
411
412void sctp_transport_raise_cwnd(struct sctp_transport *transport,
413 __u32 sack_ctsn, __u32 bytes_acked)
414{
415 struct sctp_association *asoc = transport->asoc;
416 __u32 cwnd, ssthresh, flight_size, pba, pmtu;
417
418 cwnd = transport->cwnd;
419 flight_size = transport->flight_size;
420
421
422 if (asoc->fast_recovery &&
423 TSN_lte(asoc->fast_recovery_exit, sack_ctsn))
424 asoc->fast_recovery = 0;
425
426 ssthresh = transport->ssthresh;
427 pba = transport->partial_bytes_acked;
428 pmtu = transport->asoc->pathmtu;
429
430 if (cwnd <= ssthresh) {
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445 if (asoc->fast_recovery)
446 return;
447
448
449
450
451
452
453 if (flight_size < cwnd)
454 return;
455
456 if (bytes_acked > pmtu)
457 cwnd += pmtu;
458 else
459 cwnd += bytes_acked;
460
461 pr_debug("%s: slow start: transport:%p, bytes_acked:%d, "
462 "cwnd:%d, ssthresh:%d, flight_size:%d, pba:%d\n",
463 __func__, transport, bytes_acked, cwnd, ssthresh,
464 flight_size, pba);
465 } else {
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488 pba += bytes_acked;
489 if (pba > cwnd && flight_size < cwnd)
490 pba = cwnd;
491 if (pba >= cwnd && flight_size >= cwnd) {
492 pba = pba - cwnd;
493 cwnd += pmtu;
494 }
495
496 pr_debug("%s: congestion avoidance: transport:%p, "
497 "bytes_acked:%d, cwnd:%d, ssthresh:%d, "
498 "flight_size:%d, pba:%d\n", __func__,
499 transport, bytes_acked, cwnd, ssthresh,
500 flight_size, pba);
501 }
502
503 transport->cwnd = cwnd;
504 transport->partial_bytes_acked = pba;
505}
506
507
508
509
510void sctp_transport_lower_cwnd(struct sctp_transport *transport,
511 enum sctp_lower_cwnd reason)
512{
513 struct sctp_association *asoc = transport->asoc;
514
515 switch (reason) {
516 case SCTP_LOWER_CWND_T3_RTX:
517
518
519
520
521
522
523
524 transport->ssthresh = max(transport->cwnd/2,
525 4*asoc->pathmtu);
526 transport->cwnd = asoc->pathmtu;
527
528
529 asoc->fast_recovery = 0;
530 break;
531
532 case SCTP_LOWER_CWND_FAST_RTX:
533
534
535
536
537
538
539
540
541
542
543
544
545 if (asoc->fast_recovery)
546 return;
547
548
549 asoc->fast_recovery = 1;
550 asoc->fast_recovery_exit = asoc->next_tsn - 1;
551
552 transport->ssthresh = max(transport->cwnd/2,
553 4*asoc->pathmtu);
554 transport->cwnd = transport->ssthresh;
555 break;
556
557 case SCTP_LOWER_CWND_ECNE:
558
559
560
561
562
563
564
565
566
567
568
569
570 if (time_after(jiffies, transport->last_time_ecne_reduced +
571 transport->rtt)) {
572 transport->ssthresh = max(transport->cwnd/2,
573 4*asoc->pathmtu);
574 transport->cwnd = transport->ssthresh;
575 transport->last_time_ecne_reduced = jiffies;
576 }
577 break;
578
579 case SCTP_LOWER_CWND_INACTIVE:
580
581
582
583
584
585
586
587
588 transport->cwnd = max(transport->cwnd/2,
589 4*asoc->pathmtu);
590
591 transport->ssthresh = transport->cwnd;
592 break;
593 }
594
595 transport->partial_bytes_acked = 0;
596
597 pr_debug("%s: transport:%p, reason:%d, cwnd:%d, ssthresh:%d\n",
598 __func__, transport, reason, transport->cwnd,
599 transport->ssthresh);
600}
601
602
603
604
605
606
607
608
609
610
611
612void sctp_transport_burst_limited(struct sctp_transport *t)
613{
614 struct sctp_association *asoc = t->asoc;
615 u32 old_cwnd = t->cwnd;
616 u32 max_burst_bytes;
617
618 if (t->burst_limited || asoc->max_burst == 0)
619 return;
620
621 max_burst_bytes = t->flight_size + (asoc->max_burst * asoc->pathmtu);
622 if (max_burst_bytes < old_cwnd) {
623 t->cwnd = max_burst_bytes;
624 t->burst_limited = old_cwnd;
625 }
626}
627
628
629
630
631void sctp_transport_burst_reset(struct sctp_transport *t)
632{
633 if (t->burst_limited) {
634 t->cwnd = t->burst_limited;
635 t->burst_limited = 0;
636 }
637}
638
639
640unsigned long sctp_transport_timeout(struct sctp_transport *trans)
641{
642
643 unsigned long timeout = trans->rto >> 1;
644
645 if (trans->state != SCTP_UNCONFIRMED &&
646 trans->state != SCTP_PF)
647 timeout += trans->hbinterval;
648
649 return max_t(unsigned long, timeout, HZ / 5);
650}
651
652
653void sctp_transport_reset(struct sctp_transport *t)
654{
655 struct sctp_association *asoc = t->asoc;
656
657
658
659
660
661
662 t->cwnd = min(4*asoc->pathmtu, max_t(__u32, 2*asoc->pathmtu, 4380));
663 t->burst_limited = 0;
664 t->ssthresh = asoc->peer.i.a_rwnd;
665 t->rto = asoc->rto_initial;
666 sctp_max_rto(asoc, t);
667 t->rtt = 0;
668 t->srtt = 0;
669 t->rttvar = 0;
670
671
672 t->partial_bytes_acked = 0;
673 t->flight_size = 0;
674 t->error_count = 0;
675 t->rto_pending = 0;
676 t->hb_sent = 0;
677
678
679 t->cacc.changeover_active = 0;
680 t->cacc.cycling_changeover = 0;
681 t->cacc.next_tsn_at_change = 0;
682 t->cacc.cacc_saw_newack = 0;
683}
684
685
686void sctp_transport_immediate_rtx(struct sctp_transport *t)
687{
688
689 if (del_timer(&t->T3_rtx_timer))
690 sctp_transport_put(t);
691
692 sctp_retransmit(&t->asoc->outqueue, t, SCTP_RTXR_T3_RTX);
693 if (!timer_pending(&t->T3_rtx_timer)) {
694 if (!mod_timer(&t->T3_rtx_timer, jiffies + t->rto))
695 sctp_transport_hold(t);
696 }
697}
698
699
700void sctp_transport_dst_release(struct sctp_transport *t)
701{
702 dst_release(t->dst);
703 t->dst = NULL;
704 t->dst_pending_confirm = 0;
705}
706
707
708void sctp_transport_dst_confirm(struct sctp_transport *t)
709{
710 t->dst_pending_confirm = 1;
711}
712