1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved.
5 * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved.
6 * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
10 *
11 * a) Redistributions of source code must retain the above copyright notice,
12 *    this list of conditions and the following disclaimer.
13 *
14 * b) Redistributions in binary form must reproduce the above copyright
15 *    notice, this list of conditions and the following disclaimer in
16 *    the documentation and/or other materials provided with the distribution.
17 *
18 * c) Neither the name of Cisco Systems, Inc. nor the names of its
19 *    contributors may be used to endorse or promote products derived
20 *    from this software without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
24 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
26 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
27 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
28 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
29 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
30 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
31 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
32 * THE POSSIBILITY OF SUCH DAMAGE.
33 */
34
35#include <sys/cdefs.h>
36__FBSDID("$FreeBSD$");
37
38#define _IP_VHL
39#include <netinet/sctp_os.h>
40#include <netinet/sctp_pcb.h>
41#ifdef INET6
42#endif
43#include <netinet/sctp_var.h>
44#include <netinet/sctp_sysctl.h>
45#include <netinet/sctp_timer.h>
46#include <netinet/sctputil.h>
47#include <netinet/sctp_output.h>
48#include <netinet/sctp_header.h>
49#include <netinet/sctp_indata.h>
50#include <netinet/sctp_asconf.h>
51#include <netinet/sctp_input.h>
52#include <netinet/sctp.h>
53#include <netinet/sctp_uio.h>
54#if defined(INET) || defined(INET6)
55#include <netinet/udp.h>
56#endif
57
58void
59sctp_audit_retranmission_queue(struct sctp_association *asoc)
60{
61	struct sctp_tmit_chunk *chk;
62
63	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit invoked on send queue cnt:%d onqueue:%d\n",
64	    asoc->sent_queue_retran_cnt,
65	    asoc->sent_queue_cnt);
66	asoc->sent_queue_retran_cnt = 0;
67	asoc->sent_queue_cnt = 0;
68	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
69		if (chk->sent == SCTP_DATAGRAM_RESEND) {
70			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
71		}
72		asoc->sent_queue_cnt++;
73	}
74	TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) {
75		if (chk->sent == SCTP_DATAGRAM_RESEND) {
76			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
77		}
78	}
79	TAILQ_FOREACH(chk, &asoc->asconf_send_queue, sctp_next) {
80		if (chk->sent == SCTP_DATAGRAM_RESEND) {
81			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
82		}
83	}
84	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit completes retran:%d onqueue:%d\n",
85	    asoc->sent_queue_retran_cnt,
86	    asoc->sent_queue_cnt);
87}
88
89static int
90sctp_threshold_management(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
91    struct sctp_nets *net, uint16_t threshold)
92{
93	if (net) {
94		net->error_count++;
95		SCTPDBG(SCTP_DEBUG_TIMER4, "Error count for %p now %d thresh:%d\n",
96		    (void *)net, net->error_count,
97		    net->failure_threshold);
98		if (net->error_count > net->failure_threshold) {
99			/* We had a threshold failure */
100			if (net->dest_state & SCTP_ADDR_REACHABLE) {
101				net->dest_state &= ~SCTP_ADDR_REACHABLE;
102				net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY;
103				net->dest_state &= ~SCTP_ADDR_PF;
104				sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN,
105				    stcb, 0,
106				    (void *)net, SCTP_SO_NOT_LOCKED);
107			}
108		} else if ((net->pf_threshold < net->failure_threshold) &&
109		    (net->error_count > net->pf_threshold)) {
110			if (!(net->dest_state & SCTP_ADDR_PF)) {
111				net->dest_state |= SCTP_ADDR_PF;
112				net->last_active = sctp_get_tick_count();
113				sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
114				sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT,
115				    inp, stcb, net,
116				    SCTP_FROM_SCTP_TIMER + SCTP_LOC_1);
117				sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net);
118			}
119		}
120	}
121	if (stcb == NULL)
122		return (0);
123
124	if (net) {
125		if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) {
126			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
127				sctp_misc_ints(SCTP_THRESHOLD_INCR,
128				    stcb->asoc.overall_error_count,
129				    (stcb->asoc.overall_error_count + 1),
130				    SCTP_FROM_SCTP_TIMER,
131				    __LINE__);
132			}
133			stcb->asoc.overall_error_count++;
134		}
135	} else {
136		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
137			sctp_misc_ints(SCTP_THRESHOLD_INCR,
138			    stcb->asoc.overall_error_count,
139			    (stcb->asoc.overall_error_count + 1),
140			    SCTP_FROM_SCTP_TIMER,
141			    __LINE__);
142		}
143		stcb->asoc.overall_error_count++;
144	}
145	SCTPDBG(SCTP_DEBUG_TIMER4, "Overall error count for %p now %d thresh:%u state:%x\n",
146	    (void *)&stcb->asoc, stcb->asoc.overall_error_count,
147	    (uint32_t)threshold,
148	    ((net == NULL) ? (uint32_t)0 : (uint32_t)net->dest_state));
149	/*
150	 * We specifically do not do >= to give the assoc one more change
151	 * before we fail it.
152	 */
153	if (stcb->asoc.overall_error_count > threshold) {
154		/* Abort notification sends a ULP notify */
155		struct mbuf *op_err;
156
157		op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
158		    "Association error counter exceeded");
159		inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_2;
160		sctp_abort_an_association(inp, stcb, op_err, SCTP_SO_NOT_LOCKED);
161		return (1);
162	}
163	return (0);
164}
165
166/*
167 * sctp_find_alternate_net() returns a non-NULL pointer as long as there
168 * exists nets, which are not being deleted.
169 */
170struct sctp_nets *
171sctp_find_alternate_net(struct sctp_tcb *stcb,
172    struct sctp_nets *net,
173    int mode)
174{
175	/* Find and return an alternate network if possible */
176	struct sctp_nets *alt, *mnet, *min_errors_net = NULL, *max_cwnd_net = NULL;
177	bool looped;
178
179	/* JRS 5/14/07 - Initialize min_errors to an impossible value. */
180	int min_errors = -1;
181	uint32_t max_cwnd = 0;
182
183	if (stcb->asoc.numnets == 1) {
184		/* No selection can be made. */
185		return (TAILQ_FIRST(&stcb->asoc.nets));
186	}
187	/*
188	 * JRS 5/14/07 - If mode is set to 2, use the CMT PF find alternate
189	 * net algorithm. This algorithm chooses the active destination (not
190	 * in PF state) with the largest cwnd value. If all destinations are
191	 * in PF state, unreachable, or unconfirmed, choose the desination
192	 * that is in PF state with the lowest error count. In case of a
193	 * tie, choose the destination that was most recently active.
194	 */
195	if (mode == 2) {
196		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
197			/*
198			 * JRS 5/14/07 - If the destination is unreachable
199			 * or unconfirmed, skip it.
200			 */
201			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
202			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
203				continue;
204			}
205			/*
206			 * JRS 5/14/07 -  If the destination is reachable
207			 * but in PF state, compare the error count of the
208			 * destination to the minimum error count seen thus
209			 * far. Store the destination with the lower error
210			 * count.  If the error counts are equal, store the
211			 * destination that was most recently active.
212			 */
213			if (mnet->dest_state & SCTP_ADDR_PF) {
214				/*
215				 * JRS 5/14/07 - If the destination under
216				 * consideration is the current destination,
217				 * work as if the error count is one higher.
218				 * The actual error count will not be
219				 * incremented until later in the t3
220				 * handler.
221				 */
222				if (mnet == net) {
223					if (min_errors == -1) {
224						min_errors = mnet->error_count + 1;
225						min_errors_net = mnet;
226					} else if (mnet->error_count + 1 < min_errors) {
227						min_errors = mnet->error_count + 1;
228						min_errors_net = mnet;
229					} else if (mnet->error_count + 1 == min_errors
230					    && mnet->last_active > min_errors_net->last_active) {
231						min_errors_net = mnet;
232						min_errors = mnet->error_count + 1;
233					}
234					continue;
235				} else {
236					if (min_errors == -1) {
237						min_errors = mnet->error_count;
238						min_errors_net = mnet;
239					} else if (mnet->error_count < min_errors) {
240						min_errors = mnet->error_count;
241						min_errors_net = mnet;
242					} else if (mnet->error_count == min_errors
243					    && mnet->last_active > min_errors_net->last_active) {
244						min_errors_net = mnet;
245						min_errors = mnet->error_count;
246					}
247					continue;
248				}
249			}
250			/*
251			 * JRS 5/14/07 - If the destination is reachable and
252			 * not in PF state, compare the cwnd of the
253			 * destination to the highest cwnd seen thus far.
254			 * Store the destination with the higher cwnd value.
255			 * If the cwnd values are equal, randomly choose one
256			 * of the two destinations.
257			 */
258			if (max_cwnd < mnet->cwnd) {
259				max_cwnd_net = mnet;
260				max_cwnd = mnet->cwnd;
261			} else if (max_cwnd == mnet->cwnd) {
262				uint32_t rndval;
263				uint8_t this_random;
264
265				if (stcb->asoc.hb_random_idx > 3) {
266					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
267					memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values));
268					this_random = stcb->asoc.hb_random_values[0];
269					stcb->asoc.hb_random_idx++;
270					stcb->asoc.hb_ect_randombit = 0;
271				} else {
272					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
273					stcb->asoc.hb_random_idx++;
274					stcb->asoc.hb_ect_randombit = 0;
275				}
276				if (this_random % 2 == 1) {
277					max_cwnd_net = mnet;
278					max_cwnd = mnet->cwnd;	/* Useless? */
279				}
280			}
281		}
282		if (max_cwnd_net == NULL) {
283			if (min_errors_net == NULL) {
284				return (net);
285			}
286			return (min_errors_net);
287		} else {
288			return (max_cwnd_net);
289		}
290	}			/* JRS 5/14/07 - If mode is set to 1, use the
291				 * CMT policy for choosing an alternate net. */
292	else if (mode == 1) {
293		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
294			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
295			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
296				/*
297				 * will skip ones that are not-reachable or
298				 * unconfirmed
299				 */
300				continue;
301			}
302			if (max_cwnd < mnet->cwnd) {
303				max_cwnd_net = mnet;
304				max_cwnd = mnet->cwnd;
305			} else if (max_cwnd == mnet->cwnd) {
306				uint32_t rndval;
307				uint8_t this_random;
308
309				if (stcb->asoc.hb_random_idx > 3) {
310					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
311					memcpy(stcb->asoc.hb_random_values, &rndval,
312					    sizeof(stcb->asoc.hb_random_values));
313					this_random = stcb->asoc.hb_random_values[0];
314					stcb->asoc.hb_random_idx = 0;
315					stcb->asoc.hb_ect_randombit = 0;
316				} else {
317					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
318					stcb->asoc.hb_random_idx++;
319					stcb->asoc.hb_ect_randombit = 0;
320				}
321				if (this_random % 2) {
322					max_cwnd_net = mnet;
323					max_cwnd = mnet->cwnd;
324				}
325			}
326		}
327		if (max_cwnd_net) {
328			return (max_cwnd_net);
329		}
330	}
331	/* Look for an alternate net, which is active. */
332	if ((net != NULL) && ((net->dest_state & SCTP_ADDR_BEING_DELETED) == 0)) {
333		alt = TAILQ_NEXT(net, sctp_next);;
334	} else {
335		alt = TAILQ_FIRST(&stcb->asoc.nets);
336	}
337	looped = false;
338	for (;;) {
339		if (alt == NULL) {
340			if (!looped) {
341				alt = TAILQ_FIRST(&stcb->asoc.nets);
342				looped = true;
343			}
344			/* Definitely out of candidates. */
345			if (alt == NULL) {
346				break;
347			}
348		}
349		if (alt->ro.ro_nh == NULL) {
350			if (alt->ro._s_addr) {
351				sctp_free_ifa(alt->ro._s_addr);
352				alt->ro._s_addr = NULL;
353			}
354			alt->src_addr_selected = 0;
355		}
356		if (((alt->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE) &&
357		    (alt->ro.ro_nh != NULL) &&
358		    (!(alt->dest_state & SCTP_ADDR_UNCONFIRMED)) &&
359		    (alt != net)) {
360			/* Found an alternate net, which is reachable. */
361			break;
362		}
363		alt = TAILQ_NEXT(alt, sctp_next);
364	}
365
366	if (alt == NULL) {
367		/*
368		 * In case no active alternate net has been found, look for
369		 * an alternate net, which is confirmed.
370		 */
371		if ((net != NULL) && ((net->dest_state & SCTP_ADDR_BEING_DELETED) == 0)) {
372			alt = TAILQ_NEXT(net, sctp_next);;
373		} else {
374			alt = TAILQ_FIRST(&stcb->asoc.nets);
375		}
376		looped = false;
377		for (;;) {
378			if (alt == NULL) {
379				if (!looped) {
380					alt = TAILQ_FIRST(&stcb->asoc.nets);
381					looped = true;
382				}
383				/* Definitely out of candidates. */
384				if (alt == NULL) {
385					break;
386				}
387			}
388			if ((!(alt->dest_state & SCTP_ADDR_UNCONFIRMED)) &&
389			    (alt != net)) {
390				/*
391				 * Found an alternate net, which is
392				 * confirmed.
393				 */
394				break;
395			}
396			alt = TAILQ_NEXT(alt, sctp_next);
397		}
398	}
399	if (alt == NULL) {
400		/*
401		 * In case no confirmed alternate net has been found, just
402		 * return net, if it is not being deleted. In the other case
403		 * just return the first net.
404		 */
405		if ((net != NULL) && ((net->dest_state & SCTP_ADDR_BEING_DELETED) == 0)) {
406			alt = net;
407		}
408		if (alt == NULL) {
409			alt = TAILQ_FIRST(&stcb->asoc.nets);
410		}
411	}
412	return (alt);
413}
414
415static void
416sctp_backoff_on_timeout(struct sctp_tcb *stcb,
417    struct sctp_nets *net,
418    int win_probe,
419    int num_marked, int num_abandoned)
420{
421	if (net->RTO == 0) {
422		if (net->RTO_measured) {
423			net->RTO = stcb->asoc.minrto;
424		} else {
425			net->RTO = stcb->asoc.initial_rto;
426		}
427	}
428	net->RTO <<= 1;
429	if (net->RTO > stcb->asoc.maxrto) {
430		net->RTO = stcb->asoc.maxrto;
431	}
432	if ((win_probe == 0) && (num_marked || num_abandoned)) {
433		/* We don't apply penalty to window probe scenarios */
434		/* JRS - Use the congestion control given in the CC module */
435		stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout(stcb, net);
436	}
437}
438
439#ifndef INVARIANTS
440static void
441sctp_recover_sent_list(struct sctp_tcb *stcb)
442{
443	struct sctp_tmit_chunk *chk, *nchk;
444	struct sctp_association *asoc;
445
446	asoc = &stcb->asoc;
447	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
448		if (SCTP_TSN_GE(asoc->last_acked_seq, chk->rec.data.tsn)) {
449			SCTP_PRINTF("Found chk:%p tsn:%x <= last_acked_seq:%x\n",
450			    (void *)chk, chk->rec.data.tsn, asoc->last_acked_seq);
451			if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
452				if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) {
453					asoc->strmout[chk->rec.data.sid].chunks_on_queues--;
454				}
455			}
456			if ((asoc->strmout[chk->rec.data.sid].chunks_on_queues == 0) &&
457			    (asoc->strmout[chk->rec.data.sid].state == SCTP_STREAM_RESET_PENDING) &&
458			    TAILQ_EMPTY(&asoc->strmout[chk->rec.data.sid].outqueue)) {
459				asoc->trigger_reset = 1;
460			}
461			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
462			if (PR_SCTP_ENABLED(chk->flags)) {
463				if (asoc->pr_sctp_cnt != 0)
464					asoc->pr_sctp_cnt--;
465			}
466			if (chk->data) {
467				/* sa_ignore NO_NULL_CHK */
468				sctp_free_bufspace(stcb, asoc, chk, 1);
469				sctp_m_freem(chk->data);
470				chk->data = NULL;
471				if (asoc->prsctp_supported && PR_SCTP_BUF_ENABLED(chk->flags)) {
472					asoc->sent_queue_cnt_removeable--;
473				}
474			}
475			asoc->sent_queue_cnt--;
476			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
477		}
478	}
479	SCTP_PRINTF("after recover order is as follows\n");
480	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
481		SCTP_PRINTF("chk:%p TSN:%x\n", (void *)chk, chk->rec.data.tsn);
482	}
483}
484#endif
485
486static int
487sctp_mark_all_for_resend(struct sctp_tcb *stcb,
488    struct sctp_nets *net,
489    struct sctp_nets *alt,
490    int window_probe,
491    int *num_marked,
492    int *num_abandoned)
493{
494
495	/*
496	 * Mark all chunks (well not all) that were sent to *net for
497	 * retransmission. Move them to alt for there destination as well...
498	 * We only mark chunks that have been outstanding long enough to
499	 * have received feed-back.
500	 */
501	struct sctp_tmit_chunk *chk, *nchk;
502	struct sctp_nets *lnets;
503	struct timeval now, min_wait, tv;
504	int cur_rto;
505	int cnt_abandoned;
506	int audit_tf, num_mk, fir;
507	unsigned int cnt_mk;
508	uint32_t orig_flight, orig_tf;
509	uint32_t tsnlast, tsnfirst;
510	int recovery_cnt = 0;
511
512	/* none in flight now */
513	audit_tf = 0;
514	fir = 0;
515	/*
516	 * figure out how long a data chunk must be pending before we can
517	 * mark it ..
518	 */
519	(void)SCTP_GETTIME_TIMEVAL(&now);
520	/* get cur rto in micro-seconds */
521	cur_rto = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv;
522	cur_rto *= 1000;
523	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
524		sctp_log_fr(cur_rto,
525		    stcb->asoc.peers_rwnd,
526		    window_probe,
527		    SCTP_FR_T3_MARK_TIME);
528		sctp_log_fr(net->flight_size, 0, 0, SCTP_FR_CWND_REPORT);
529		sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT);
530	}
531	tv.tv_sec = cur_rto / 1000000;
532	tv.tv_usec = cur_rto % 1000000;
533	min_wait = now;
534	timevalsub(&min_wait, &tv);
535	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
536		/*
537		 * if we hit here, we don't have enough seconds on the clock
538		 * to account for the RTO. We just let the lower seconds be
539		 * the bounds and don't worry about it. This may mean we
540		 * will mark a lot more than we should.
541		 */
542		min_wait.tv_sec = min_wait.tv_usec = 0;
543	}
544	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
545		sctp_log_fr(cur_rto, now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME);
546		sctp_log_fr(0, min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME);
547	}
548	/*
549	 * Our rwnd will be incorrect here since we are not adding back the
550	 * cnt * mbuf but we will fix that down below.
551	 */
552	orig_flight = net->flight_size;
553	orig_tf = stcb->asoc.total_flight;
554
555	net->fast_retran_ip = 0;
556	/* Now on to each chunk */
557	cnt_abandoned = 0;
558	num_mk = cnt_mk = 0;
559	tsnfirst = tsnlast = 0;
560#ifndef INVARIANTS
561start_again:
562#endif
563	TAILQ_FOREACH_SAFE(chk, &stcb->asoc.sent_queue, sctp_next, nchk) {
564		if (SCTP_TSN_GE(stcb->asoc.last_acked_seq, chk->rec.data.tsn)) {
565			/* Strange case our list got out of order? */
566			SCTP_PRINTF("Our list is out of order? last_acked:%x chk:%x\n",
567			    (unsigned int)stcb->asoc.last_acked_seq, (unsigned int)chk->rec.data.tsn);
568			recovery_cnt++;
569#ifdef INVARIANTS
570			panic("last acked >= chk on sent-Q");
571#else
572			SCTP_PRINTF("Recover attempts a restart cnt:%d\n", recovery_cnt);
573			sctp_recover_sent_list(stcb);
574			if (recovery_cnt < 10) {
575				goto start_again;
576			} else {
577				SCTP_PRINTF("Recovery fails %d times??\n", recovery_cnt);
578			}
579#endif
580		}
581		if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) {
582			/*
583			 * found one to mark: If it is less than
584			 * DATAGRAM_ACKED it MUST not be a skipped or marked
585			 * TSN but instead one that is either already set
586			 * for retransmission OR one that needs
587			 * retransmission.
588			 */
589
590			/* validate its been outstanding long enough */
591			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
592				sctp_log_fr(chk->rec.data.tsn,
593				    chk->sent_rcv_time.tv_sec,
594				    chk->sent_rcv_time.tv_usec,
595				    SCTP_FR_T3_MARK_TIME);
596			}
597			if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) {
598				/*
599				 * we have reached a chunk that was sent
600				 * some seconds past our min.. forget it we
601				 * will find no more to send.
602				 */
603				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
604					sctp_log_fr(0,
605					    chk->sent_rcv_time.tv_sec,
606					    chk->sent_rcv_time.tv_usec,
607					    SCTP_FR_T3_STOPPED);
608				}
609				continue;
610			} else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) &&
611			    (window_probe == 0)) {
612				/*
613				 * we must look at the micro seconds to
614				 * know.
615				 */
616				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
617					/*
618					 * ok it was sent after our boundary
619					 * time.
620					 */
621					continue;
622				}
623			}
624			if (stcb->asoc.prsctp_supported && PR_SCTP_TTL_ENABLED(chk->flags)) {
625				/* Is it expired? */
626				if (timevalcmp(&now, &chk->rec.data.timetodrop, >)) {
627					/* Yes so drop it */
628					if (chk->data) {
629						(void)sctp_release_pr_sctp_chunk(stcb,
630						    chk,
631						    1,
632						    SCTP_SO_NOT_LOCKED);
633						cnt_abandoned++;
634					}
635					continue;
636				}
637			}
638			if (stcb->asoc.prsctp_supported && PR_SCTP_RTX_ENABLED(chk->flags)) {
639				/* Has it been retransmitted tv_sec times? */
640				if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) {
641					if (chk->data) {
642						(void)sctp_release_pr_sctp_chunk(stcb,
643						    chk,
644						    1,
645						    SCTP_SO_NOT_LOCKED);
646						cnt_abandoned++;
647					}
648					continue;
649				}
650			}
651			if (chk->sent < SCTP_DATAGRAM_RESEND) {
652				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
653				num_mk++;
654				if (fir == 0) {
655					fir = 1;
656					tsnfirst = chk->rec.data.tsn;
657				}
658				tsnlast = chk->rec.data.tsn;
659				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
660					sctp_log_fr(chk->rec.data.tsn, chk->snd_count,
661					    0, SCTP_FR_T3_MARKED);
662				}
663
664				if (chk->rec.data.chunk_was_revoked) {
665					/* deflate the cwnd */
666					chk->whoTo->cwnd -= chk->book_size;
667					chk->rec.data.chunk_was_revoked = 0;
668				}
669				net->marked_retrans++;
670				stcb->asoc.marked_retrans++;
671				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
672					sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO,
673					    chk->whoTo->flight_size,
674					    chk->book_size,
675					    (uint32_t)(uintptr_t)chk->whoTo,
676					    chk->rec.data.tsn);
677				}
678				sctp_flight_size_decrease(chk);
679				sctp_total_flight_decrease(stcb, chk);
680				stcb->asoc.peers_rwnd += chk->send_size;
681				stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh);
682			}
683			chk->sent = SCTP_DATAGRAM_RESEND;
684			chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
685			SCTP_STAT_INCR(sctps_markedretrans);
686
687			/* reset the TSN for striking and other FR stuff */
688			chk->rec.data.doing_fast_retransmit = 0;
689			/* Clear any time so NO RTT is being done */
690
691			if (chk->do_rtt) {
692				if (chk->whoTo->rto_needed == 0) {
693					chk->whoTo->rto_needed = 1;
694				}
695			}
696			chk->do_rtt = 0;
697			if (alt != net) {
698				sctp_free_remote_addr(chk->whoTo);
699				chk->no_fr_allowed = 1;
700				chk->whoTo = alt;
701				atomic_add_int(&alt->ref_count, 1);
702			} else {
703				chk->no_fr_allowed = 0;
704				if (TAILQ_EMPTY(&stcb->asoc.send_queue)) {
705					chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq;
706				} else {
707					chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.tsn;
708				}
709			}
710			/*
711			 * CMT: Do not allow FRs on retransmitted TSNs.
712			 */
713			if (stcb->asoc.sctp_cmt_on_off > 0) {
714				chk->no_fr_allowed = 1;
715			}
716#ifdef THIS_SHOULD_NOT_BE_DONE
717		} else if (chk->sent == SCTP_DATAGRAM_ACKED) {
718			/* remember highest acked one */
719			could_be_sent = chk;
720#endif
721		}
722		if (chk->sent == SCTP_DATAGRAM_RESEND) {
723			cnt_mk++;
724		}
725	}
726	if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) {
727		/* we did not subtract the same things? */
728		audit_tf = 1;
729	}
730
731	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
732		sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT);
733	}
734#ifdef SCTP_DEBUG
735	if (num_mk) {
736		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
737		    tsnlast);
738		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%u\n",
739		    num_mk,
740		    stcb->asoc.peers_rwnd);
741	}
742#endif
743	*num_marked = num_mk;
744	*num_abandoned = cnt_abandoned;
745	/*
746	 * Now check for a ECN Echo that may be stranded And include the
747	 * cnt_mk'd to have all resends in the control queue.
748	 */
749	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
750		if (chk->sent == SCTP_DATAGRAM_RESEND) {
751			cnt_mk++;
752		}
753		if ((chk->whoTo == net) &&
754		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
755			sctp_free_remote_addr(chk->whoTo);
756			chk->whoTo = alt;
757			if (chk->sent != SCTP_DATAGRAM_RESEND) {
758				chk->sent = SCTP_DATAGRAM_RESEND;
759				chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
760				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
761				cnt_mk++;
762			}
763			atomic_add_int(&alt->ref_count, 1);
764		}
765	}
766#ifdef THIS_SHOULD_NOT_BE_DONE
767	if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) {
768		/* fix it so we retransmit the highest acked anyway */
769		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
770		cnt_mk++;
771		could_be_sent->sent = SCTP_DATAGRAM_RESEND;
772	}
773#endif
774	if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) {
775#ifdef INVARIANTS
776		SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n",
777		    cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk);
778#endif
779#ifndef SCTP_AUDITING_ENABLED
780		stcb->asoc.sent_queue_retran_cnt = cnt_mk;
781#endif
782	}
783	if (audit_tf) {
784		SCTPDBG(SCTP_DEBUG_TIMER4,
785		    "Audit total flight due to negative value net:%p\n",
786		    (void *)net);
787		stcb->asoc.total_flight = 0;
788		stcb->asoc.total_flight_count = 0;
789		/* Clear all networks flight size */
790		TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) {
791			lnets->flight_size = 0;
792			SCTPDBG(SCTP_DEBUG_TIMER4,
793			    "Net:%p c-f cwnd:%d ssthresh:%d\n",
794			    (void *)lnets, lnets->cwnd, lnets->ssthresh);
795		}
796		TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
797			if (chk->sent < SCTP_DATAGRAM_RESEND) {
798				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
799					sctp_misc_ints(SCTP_FLIGHT_LOG_UP,
800					    chk->whoTo->flight_size,
801					    chk->book_size,
802					    (uint32_t)(uintptr_t)chk->whoTo,
803					    chk->rec.data.tsn);
804				}
805
806				sctp_flight_size_increase(chk);
807				sctp_total_flight_increase(stcb, chk);
808			}
809		}
810	}
811	/* We return 1 if we only have a window probe outstanding */
812	return (0);
813}
814
815int
816sctp_t3rxt_timer(struct sctp_inpcb *inp,
817    struct sctp_tcb *stcb,
818    struct sctp_nets *net)
819{
820	struct sctp_nets *alt;
821	int win_probe, num_mk, num_abandoned;
822
823	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
824		sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT);
825	}
826	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
827		struct sctp_nets *lnet;
828
829		TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
830			if (net == lnet) {
831				sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3);
832			} else {
833				sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3);
834			}
835		}
836	}
837	/* Find an alternate and mark those for retransmission */
838	if ((stcb->asoc.peers_rwnd == 0) &&
839	    (stcb->asoc.total_flight < net->mtu)) {
840		SCTP_STAT_INCR(sctps_timowindowprobe);
841		win_probe = 1;
842	} else {
843		win_probe = 0;
844	}
845
846	if (win_probe == 0) {
847		/* We don't do normal threshold management on window probes */
848		if (sctp_threshold_management(inp, stcb, net,
849		    stcb->asoc.max_send_times)) {
850			/* Association was destroyed */
851			return (1);
852		} else {
853			if (net != stcb->asoc.primary_destination) {
854				/* send a immediate HB if our RTO is stale */
855				struct timeval now;
856				unsigned int ms_goneby;
857
858				(void)SCTP_GETTIME_TIMEVAL(&now);
859				if (net->last_sent_time.tv_sec) {
860					ms_goneby = (now.tv_sec - net->last_sent_time.tv_sec) * 1000;
861				} else {
862					ms_goneby = 0;
863				}
864				if ((net->dest_state & SCTP_ADDR_PF) == 0) {
865					if ((ms_goneby > net->RTO) || (net->RTO == 0)) {
866						/*
867						 * no recent feed back in an
868						 * RTO or more, request a
869						 * RTT update
870						 */
871						sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
872					}
873				}
874			}
875		}
876	} else {
877		/*
878		 * For a window probe we don't penalize the net's but only
879		 * the association. This may fail it if SACKs are not coming
880		 * back. If sack's are coming with rwnd locked at 0, we will
881		 * continue to hold things waiting for rwnd to raise
882		 */
883		if (sctp_threshold_management(inp, stcb, NULL,
884		    stcb->asoc.max_send_times)) {
885			/* Association was destroyed */
886			return (1);
887		}
888	}
889	if (stcb->asoc.sctp_cmt_on_off > 0) {
890		if (net->pf_threshold < net->failure_threshold) {
891			alt = sctp_find_alternate_net(stcb, net, 2);
892		} else {
893			/*
894			 * CMT: Using RTX_SSTHRESH policy for CMT. If CMT is
895			 * being used, then pick dest with largest ssthresh
896			 * for any retransmission.
897			 */
898			alt = sctp_find_alternate_net(stcb, net, 1);
899			/*
900			 * CUCv2: If a different dest is picked for the
901			 * retransmission, then new (rtx-)pseudo_cumack
902			 * needs to be tracked for orig dest. Let CUCv2
903			 * track new (rtx-) pseudo-cumack always.
904			 */
905			net->find_pseudo_cumack = 1;
906			net->find_rtx_pseudo_cumack = 1;
907		}
908	} else {
909		alt = sctp_find_alternate_net(stcb, net, 0);
910	}
911
912	num_mk = 0;
913	num_abandoned = 0;
914	(void)sctp_mark_all_for_resend(stcb, net, alt, win_probe,
915	    &num_mk, &num_abandoned);
916	/* FR Loss recovery just ended with the T3. */
917	stcb->asoc.fast_retran_loss_recovery = 0;
918
919	/* CMT FR loss recovery ended with the T3 */
920	net->fast_retran_loss_recovery = 0;
921	if ((stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) &&
922	    (net->flight_size == 0)) {
923		(*stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) (stcb, net);
924	}
925
926	/*
927	 * setup the sat loss recovery that prevents satellite cwnd advance.
928	 */
929	stcb->asoc.sat_t3_loss_recovery = 1;
930	stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq;
931
932	/* Backoff the timer and cwnd */
933	sctp_backoff_on_timeout(stcb, net, win_probe, num_mk, num_abandoned);
934	if ((!(net->dest_state & SCTP_ADDR_REACHABLE)) ||
935	    (net->dest_state & SCTP_ADDR_PF)) {
936		/* Move all pending over too */
937		sctp_move_chunks_from_net(stcb, net);
938
939		/*
940		 * Get the address that failed, to force a new src address
941		 * selecton and a route allocation.
942		 */
943		if (net->ro._s_addr) {
944			sctp_free_ifa(net->ro._s_addr);
945			net->ro._s_addr = NULL;
946		}
947		net->src_addr_selected = 0;
948
949		/* Force a route allocation too */
950		RO_NHFREE(&net->ro);
951
952		/* Was it our primary? */
953		if ((stcb->asoc.primary_destination == net) && (alt != net)) {
954			/*
955			 * Yes, note it as such and find an alternate note:
956			 * this means HB code must use this to resent the
957			 * primary if it goes active AND if someone does a
958			 * change-primary then this flag must be cleared
959			 * from any net structures.
960			 */
961			if (stcb->asoc.alternate) {
962				sctp_free_remote_addr(stcb->asoc.alternate);
963			}
964			stcb->asoc.alternate = alt;
965			atomic_add_int(&stcb->asoc.alternate->ref_count, 1);
966		}
967	}
968	/*
969	 * Special case for cookie-echo'ed case, we don't do output but must
970	 * await the COOKIE-ACK before retransmission
971	 */
972	if (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED) {
973		/*
974		 * Here we just reset the timer and start again since we
975		 * have not established the asoc
976		 */
977		sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net);
978		return (0);
979	}
980	if (stcb->asoc.prsctp_supported) {
981		struct sctp_tmit_chunk *lchk;
982
983		lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc);
984		/* C3. See if we need to send a Fwd-TSN */
985		if (SCTP_TSN_GT(stcb->asoc.advanced_peer_ack_point, stcb->asoc.last_acked_seq)) {
986			send_forward_tsn(stcb, &stcb->asoc);
987			for (; lchk != NULL; lchk = TAILQ_NEXT(lchk, sctp_next)) {
988				if (lchk->whoTo != NULL) {
989					break;
990				}
991			}
992			if (lchk != NULL) {
993				/* Assure a timer is up */
994				sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo);
995			}
996		}
997	}
998	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
999		sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX);
1000	}
1001	return (0);
1002}
1003
1004int
1005sctp_t1init_timer(struct sctp_inpcb *inp,
1006    struct sctp_tcb *stcb,
1007    struct sctp_nets *net)
1008{
1009	/* bump the thresholds */
1010	if (stcb->asoc.delayed_connection) {
1011		/*
1012		 * special hook for delayed connection. The library did NOT
1013		 * complete the rest of its sends.
1014		 */
1015		stcb->asoc.delayed_connection = 0;
1016		sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1017		return (0);
1018	}
1019	if (SCTP_GET_STATE(stcb) != SCTP_STATE_COOKIE_WAIT) {
1020		return (0);
1021	}
1022	if (sctp_threshold_management(inp, stcb, net,
1023	    stcb->asoc.max_init_times)) {
1024		/* Association was destroyed */
1025		return (1);
1026	}
1027	stcb->asoc.dropped_special_cnt = 0;
1028	sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0, 0);
1029	if (stcb->asoc.initial_init_rto_max < net->RTO) {
1030		net->RTO = stcb->asoc.initial_init_rto_max;
1031	}
1032	if (stcb->asoc.numnets > 1) {
1033		/* If we have more than one addr use it */
1034		struct sctp_nets *alt;
1035
1036		alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0);
1037		if (alt != stcb->asoc.primary_destination) {
1038			sctp_move_chunks_from_net(stcb, stcb->asoc.primary_destination);
1039			stcb->asoc.primary_destination = alt;
1040		}
1041	}
1042	/* Send out a new init */
1043	sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1044	return (0);
1045}
1046
1047/*
1048 * For cookie and asconf we actually need to find and mark for resend, then
1049 * increment the resend counter (after all the threshold management stuff of
1050 * course).
1051 */
1052int
1053sctp_cookie_timer(struct sctp_inpcb *inp,
1054    struct sctp_tcb *stcb,
1055    struct sctp_nets *net SCTP_UNUSED)
1056{
1057	struct sctp_nets *alt;
1058	struct sctp_tmit_chunk *cookie;
1059
1060	/* first before all else we must find the cookie */
1061	TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) {
1062		if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) {
1063			break;
1064		}
1065	}
1066	if (cookie == NULL) {
1067		if (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED) {
1068			/* FOOBAR! */
1069			struct mbuf *op_err;
1070
1071			op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
1072			    "Cookie timer expired, but no cookie");
1073			inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_3;
1074			sctp_abort_an_association(inp, stcb, op_err, SCTP_SO_NOT_LOCKED);
1075		} else {
1076#ifdef INVARIANTS
1077			panic("Cookie timer expires in wrong state?");
1078#else
1079			SCTP_PRINTF("Strange in state %d not cookie-echoed yet c-e timer expires?\n", SCTP_GET_STATE(stcb));
1080			return (0);
1081#endif
1082		}
1083		return (0);
1084	}
1085	/* Ok we found the cookie, threshold management next */
1086	if (sctp_threshold_management(inp, stcb, cookie->whoTo,
1087	    stcb->asoc.max_init_times)) {
1088		/* Assoc is over */
1089		return (1);
1090	}
1091	/*
1092	 * Cleared threshold management, now lets backoff the address and
1093	 * select an alternate
1094	 */
1095	stcb->asoc.dropped_special_cnt = 0;
1096	sctp_backoff_on_timeout(stcb, cookie->whoTo, 1, 0, 0);
1097	alt = sctp_find_alternate_net(stcb, cookie->whoTo, 0);
1098	if (alt != cookie->whoTo) {
1099		sctp_free_remote_addr(cookie->whoTo);
1100		cookie->whoTo = alt;
1101		atomic_add_int(&alt->ref_count, 1);
1102	}
1103	/* Now mark the retran info */
1104	if (cookie->sent != SCTP_DATAGRAM_RESEND) {
1105		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1106	}
1107	cookie->sent = SCTP_DATAGRAM_RESEND;
1108	cookie->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1109	/*
1110	 * Now call the output routine to kick out the cookie again, Note we
1111	 * don't mark any chunks for retran so that FR will need to kick in
1112	 * to move these (or a send timer).
1113	 */
1114	return (0);
1115}
1116
1117int
1118sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
1119{
1120	struct sctp_nets *alt, *net;
1121	struct sctp_tmit_chunk *strrst = NULL, *chk = NULL;
1122
1123	if (stcb->asoc.stream_reset_outstanding == 0) {
1124		return (0);
1125	}
1126	/* find the existing STRRESET, we use the seq number we sent out on */
1127	(void)sctp_find_stream_reset(stcb, stcb->asoc.str_reset_seq_out, &strrst);
1128	if (strrst == NULL) {
1129		return (0);
1130	}
1131	net = strrst->whoTo;
1132	/* do threshold management */
1133	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1134		/* Assoc is over */
1135		return (1);
1136	}
1137	/*
1138	 * Cleared threshold management, now lets backoff the address and
1139	 * select an alternate
1140	 */
1141	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1142	alt = sctp_find_alternate_net(stcb, net, 0);
1143	strrst->whoTo = alt;
1144	atomic_add_int(&alt->ref_count, 1);
1145
1146	/* See if a ECN Echo is also stranded */
1147	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1148		if ((chk->whoTo == net) &&
1149		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1150			sctp_free_remote_addr(chk->whoTo);
1151			if (chk->sent != SCTP_DATAGRAM_RESEND) {
1152				chk->sent = SCTP_DATAGRAM_RESEND;
1153				chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1154				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1155			}
1156			chk->whoTo = alt;
1157			atomic_add_int(&alt->ref_count, 1);
1158		}
1159	}
1160	if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1161		/*
1162		 * If the address went un-reachable, we need to move to
1163		 * alternates for ALL chk's in queue
1164		 */
1165		sctp_move_chunks_from_net(stcb, net);
1166	}
1167	sctp_free_remote_addr(net);
1168
1169	/* mark the retran info */
1170	if (strrst->sent != SCTP_DATAGRAM_RESEND)
1171		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1172	strrst->sent = SCTP_DATAGRAM_RESEND;
1173	strrst->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1174
1175	/* restart the timer */
1176	sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, alt);
1177	return (0);
1178}
1179
1180int
1181sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1182    struct sctp_nets *net)
1183{
1184	struct sctp_nets *alt;
1185	struct sctp_tmit_chunk *asconf, *chk;
1186
1187	/* is this a first send, or a retransmission? */
1188	if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) {
1189		/* compose a new ASCONF chunk and send it */
1190		sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED);
1191	} else {
1192		/*
1193		 * Retransmission of the existing ASCONF is needed
1194		 */
1195
1196		/* find the existing ASCONF */
1197		asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue);
1198		if (asconf == NULL) {
1199			return (0);
1200		}
1201		net = asconf->whoTo;
1202		/* do threshold management */
1203		if (sctp_threshold_management(inp, stcb, net,
1204		    stcb->asoc.max_send_times)) {
1205			/* Assoc is over */
1206			return (1);
1207		}
1208		if (asconf->snd_count > stcb->asoc.max_send_times) {
1209			/*
1210			 * Something is rotten: our peer is not responding
1211			 * to ASCONFs but apparently is to other chunks.
1212			 * i.e. it is not properly handling the chunk type
1213			 * upper bits. Mark this peer as ASCONF incapable
1214			 * and cleanup.
1215			 */
1216			SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n");
1217			sctp_asconf_cleanup(stcb);
1218			return (0);
1219		}
1220		/*
1221		 * cleared threshold management, so now backoff the net and
1222		 * select an alternate
1223		 */
1224		sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1225		alt = sctp_find_alternate_net(stcb, net, 0);
1226		if (asconf->whoTo != alt) {
1227			asconf->whoTo = alt;
1228			atomic_add_int(&alt->ref_count, 1);
1229		}
1230
1231		/* See if an ECN Echo is also stranded */
1232		TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1233			if ((chk->whoTo == net) &&
1234			    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1235				sctp_free_remote_addr(chk->whoTo);
1236				chk->whoTo = alt;
1237				if (chk->sent != SCTP_DATAGRAM_RESEND) {
1238					chk->sent = SCTP_DATAGRAM_RESEND;
1239					chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1240					sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1241				}
1242				atomic_add_int(&alt->ref_count, 1);
1243			}
1244		}
1245		TAILQ_FOREACH(chk, &stcb->asoc.asconf_send_queue, sctp_next) {
1246			if (chk->whoTo != alt) {
1247				sctp_free_remote_addr(chk->whoTo);
1248				chk->whoTo = alt;
1249				atomic_add_int(&alt->ref_count, 1);
1250			}
1251			if (asconf->sent != SCTP_DATAGRAM_RESEND && chk->sent != SCTP_DATAGRAM_UNSENT)
1252				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1253			chk->sent = SCTP_DATAGRAM_RESEND;
1254			chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1255		}
1256		if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1257			/*
1258			 * If the address went un-reachable, we need to move
1259			 * to the alternate for ALL chunks in queue
1260			 */
1261			sctp_move_chunks_from_net(stcb, net);
1262		}
1263		sctp_free_remote_addr(net);
1264
1265		/* mark the retran info */
1266		if (asconf->sent != SCTP_DATAGRAM_RESEND)
1267			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1268		asconf->sent = SCTP_DATAGRAM_RESEND;
1269		asconf->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1270
1271		/* send another ASCONF if any and we can do */
1272		sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED);
1273	}
1274	return (0);
1275}
1276
1277/* Mobility adaptation */
1278void
1279sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
1280{
1281	if (stcb->asoc.deleted_primary == NULL) {
1282		SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n");
1283		sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1284		return;
1285	}
1286	SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary ");
1287	SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa);
1288	sctp_free_remote_addr(stcb->asoc.deleted_primary);
1289	stcb->asoc.deleted_primary = NULL;
1290	sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1291	return;
1292}
1293
1294/*
1295 * For the shutdown and shutdown-ack, we do not keep one around on the
1296 * control queue. This means we must generate a new one and call the general
1297 * chunk output routine, AFTER having done threshold management.
1298 * It is assumed that net is non-NULL.
1299 */
1300int
1301sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1302    struct sctp_nets *net)
1303{
1304	struct sctp_nets *alt;
1305
1306	/* first threshold management */
1307	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1308		/* Assoc is over */
1309		return (1);
1310	}
1311	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1312	/* second select an alternative */
1313	alt = sctp_find_alternate_net(stcb, net, 0);
1314
1315	/* third generate a shutdown into the queue for out net */
1316	sctp_send_shutdown(stcb, alt);
1317
1318	/* fourth restart timer */
1319	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt);
1320	return (0);
1321}
1322
1323int
1324sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1325    struct sctp_nets *net)
1326{
1327	struct sctp_nets *alt;
1328
1329	/* first threshold management */
1330	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1331		/* Assoc is over */
1332		return (1);
1333	}
1334	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1335	/* second select an alternative */
1336	alt = sctp_find_alternate_net(stcb, net, 0);
1337
1338	/* third generate a shutdown into the queue for out net */
1339	sctp_send_shutdown_ack(stcb, alt);
1340
1341	/* fourth restart timer */
1342	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt);
1343	return (0);
1344}
1345
1346static void
1347sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp,
1348    struct sctp_tcb *stcb)
1349{
1350	struct sctp_stream_queue_pending *sp;
1351	unsigned int i, chks_in_queue = 0;
1352	int being_filled = 0;
1353
1354	/*
1355	 * This function is ONLY called when the send/sent queues are empty.
1356	 */
1357	if ((stcb == NULL) || (inp == NULL))
1358		return;
1359
1360	if (stcb->asoc.sent_queue_retran_cnt) {
1361		SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n",
1362		    stcb->asoc.sent_queue_retran_cnt);
1363		stcb->asoc.sent_queue_retran_cnt = 0;
1364	}
1365	if (stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1366		/* No stream scheduler information, initialize scheduler */
1367		stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc, 0);
1368		if (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1369			/* yep, we lost a stream or two */
1370			SCTP_PRINTF("Found additional streams NOT managed by scheduler, corrected\n");
1371		} else {
1372			/* no streams lost */
1373			stcb->asoc.total_output_queue_size = 0;
1374		}
1375	}
1376	/* Check to see if some data queued, if so report it */
1377	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
1378		if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) {
1379			TAILQ_FOREACH(sp, &stcb->asoc.strmout[i].outqueue, next) {
1380				if (sp->msg_is_complete)
1381					being_filled++;
1382				chks_in_queue++;
1383			}
1384		}
1385	}
1386	if (chks_in_queue != stcb->asoc.stream_queue_cnt) {
1387		SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n",
1388		    stcb->asoc.stream_queue_cnt, chks_in_queue);
1389	}
1390	if (chks_in_queue) {
1391		/* call the output queue function */
1392		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1393		if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1394		    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1395			/*
1396			 * Probably should go in and make it go back through
1397			 * and add fragments allowed
1398			 */
1399			if (being_filled == 0) {
1400				SCTP_PRINTF("Still nothing moved %d chunks are stuck\n",
1401				    chks_in_queue);
1402			}
1403		}
1404	} else {
1405		SCTP_PRINTF("Found no chunks on any queue tot:%lu\n",
1406		    (u_long)stcb->asoc.total_output_queue_size);
1407		stcb->asoc.total_output_queue_size = 0;
1408	}
1409}
1410
1411int
1412sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1413    struct sctp_nets *net)
1414{
1415	uint8_t net_was_pf;
1416
1417	if (net->dest_state & SCTP_ADDR_PF) {
1418		net_was_pf = 1;
1419	} else {
1420		net_was_pf = 0;
1421	}
1422	if (net->hb_responded == 0) {
1423		if (net->ro._s_addr) {
1424			/*
1425			 * Invalidate the src address if we did not get a
1426			 * response last time.
1427			 */
1428			sctp_free_ifa(net->ro._s_addr);
1429			net->ro._s_addr = NULL;
1430			net->src_addr_selected = 0;
1431		}
1432		sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1433		if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1434			/* Assoc is over */
1435			return (1);
1436		}
1437	}
1438	/* Zero PBA, if it needs it */
1439	if (net->partial_bytes_acked) {
1440		net->partial_bytes_acked = 0;
1441	}
1442	if ((stcb->asoc.total_output_queue_size > 0) &&
1443	    (TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1444	    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1445		sctp_audit_stream_queues_for_size(inp, stcb);
1446	}
1447	if (!(net->dest_state & SCTP_ADDR_NOHB) &&
1448	    !((net_was_pf == 0) && (net->dest_state & SCTP_ADDR_PF))) {
1449		/*
1450		 * when move to PF during threshold mangement, a HB has been
1451		 * queued in that routine
1452		 */
1453		uint32_t ms_gone_by;
1454
1455		if ((net->last_sent_time.tv_sec > 0) ||
1456		    (net->last_sent_time.tv_usec > 0)) {
1457			struct timeval diff;
1458
1459			SCTP_GETTIME_TIMEVAL(&diff);
1460			timevalsub(&diff, &net->last_sent_time);
1461			ms_gone_by = (uint32_t)(diff.tv_sec * 1000) +
1462			    (uint32_t)(diff.tv_usec / 1000);
1463		} else {
1464			ms_gone_by = 0xffffffff;
1465		}
1466		if ((ms_gone_by >= net->heart_beat_delay) ||
1467		    (net->dest_state & SCTP_ADDR_PF)) {
1468			sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
1469		}
1470	}
1471	return (0);
1472}
1473
1474void
1475sctp_pathmtu_timer(struct sctp_inpcb *inp,
1476    struct sctp_tcb *stcb,
1477    struct sctp_nets *net)
1478{
1479	uint32_t next_mtu, mtu;
1480
1481	next_mtu = sctp_get_next_mtu(net->mtu);
1482
1483	if ((next_mtu > net->mtu) && (net->port == 0)) {
1484		if ((net->src_addr_selected == 0) ||
1485		    (net->ro._s_addr == NULL) ||
1486		    (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1487			if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1488				sctp_free_ifa(net->ro._s_addr);
1489				net->ro._s_addr = NULL;
1490				net->src_addr_selected = 0;
1491			} else if (net->ro._s_addr == NULL) {
1492#if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1493				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1494					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1495
1496					/* KAME hack: embed scopeid */
1497					(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
1498				}
1499#endif
1500
1501				net->ro._s_addr = sctp_source_address_selection(inp,
1502				    stcb,
1503				    (sctp_route_t *)&net->ro,
1504				    net, 0, stcb->asoc.vrf_id);
1505#if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1506				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1507					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1508
1509					(void)sa6_recoverscope(sin6);
1510				}
1511#endif				/* INET6 */
1512			}
1513			if (net->ro._s_addr)
1514				net->src_addr_selected = 1;
1515		}
1516		if (net->ro._s_addr) {
1517			mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_nh);
1518#if defined(INET) || defined(INET6)
1519			if (net->port) {
1520				mtu -= sizeof(struct udphdr);
1521			}
1522#endif
1523			if (mtu > next_mtu) {
1524				net->mtu = next_mtu;
1525			} else {
1526				net->mtu = mtu;
1527			}
1528		}
1529	}
1530	/* restart the timer */
1531	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
1532}
1533
1534void
1535sctp_autoclose_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
1536{
1537	struct timeval tn, *tim_touse;
1538	struct sctp_association *asoc;
1539	uint32_t ticks_gone_by;
1540
1541	(void)SCTP_GETTIME_TIMEVAL(&tn);
1542	if (stcb->asoc.sctp_autoclose_ticks > 0 &&
1543	    sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) {
1544		/* Auto close is on */
1545		asoc = &stcb->asoc;
1546		/* pick the time to use */
1547		if (asoc->time_last_rcvd.tv_sec >
1548		    asoc->time_last_sent.tv_sec) {
1549			tim_touse = &asoc->time_last_rcvd;
1550		} else {
1551			tim_touse = &asoc->time_last_sent;
1552		}
1553		/* Now has long enough transpired to autoclose? */
1554		ticks_gone_by = sctp_secs_to_ticks((uint32_t)(tn.tv_sec - tim_touse->tv_sec));
1555		if (ticks_gone_by >= asoc->sctp_autoclose_ticks) {
1556			/*
1557			 * autoclose time has hit, call the output routine,
1558			 * which should do nothing just to be SURE we don't
1559			 * have hanging data. We can then safely check the
1560			 * queues and know that we are clear to send
1561			 * shutdown
1562			 */
1563			sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED);
1564			/* Are we clean? */
1565			if (TAILQ_EMPTY(&asoc->send_queue) &&
1566			    TAILQ_EMPTY(&asoc->sent_queue)) {
1567				/*
1568				 * there is nothing queued to send, so I'm
1569				 * done...
1570				 */
1571				if (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_SENT) {
1572					/* only send SHUTDOWN 1st time thru */
1573					struct sctp_nets *net;
1574
1575					if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) ||
1576					    (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
1577						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
1578					}
1579					SCTP_SET_STATE(stcb, SCTP_STATE_SHUTDOWN_SENT);
1580					sctp_stop_timers_for_shutdown(stcb);
1581					if (stcb->asoc.alternate) {
1582						net = stcb->asoc.alternate;
1583					} else {
1584						net = stcb->asoc.primary_destination;
1585					}
1586					sctp_send_shutdown(stcb, net);
1587					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
1588					    stcb->sctp_ep, stcb, net);
1589					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
1590					    stcb->sctp_ep, stcb, NULL);
1591				}
1592			}
1593		} else {
1594			/*
1595			 * No auto close at this time, reset t-o to check
1596			 * later
1597			 */
1598			uint32_t tmp;
1599
1600			/* fool the timer startup to use the time left */
1601			tmp = asoc->sctp_autoclose_ticks;
1602			asoc->sctp_autoclose_ticks -= ticks_gone_by;
1603			sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb, NULL);
1604			/* restore the real tick value */
1605			asoc->sctp_autoclose_ticks = tmp;
1606		}
1607	}
1608}
1609