sctp_timer.c revision 243157
1/*-
2 * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved.
3 * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved.
4 * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions are met:
8 *
9 * a) Redistributions of source code must retain the above copyright notice,
10 *    this list of conditions and the following disclaimer.
11 *
12 * b) Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in
14 *    the documentation and/or other materials provided with the distribution.
15 *
16 * c) Neither the name of Cisco Systems, Inc. nor the names of its
17 *    contributors may be used to endorse or promote products derived
18 *    from this software without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
22 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
24 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
30 * THE POSSIBILITY OF SUCH DAMAGE.
31 */
32
33#include <sys/cdefs.h>
34__FBSDID("$FreeBSD: head/sys/netinet/sctp_timer.c 243157 2012-11-16 19:39:10Z tuexen $");
35
36#define _IP_VHL
37#include <netinet/sctp_os.h>
38#include <netinet/sctp_pcb.h>
39#ifdef INET6
40#endif
41#include <netinet/sctp_var.h>
42#include <netinet/sctp_sysctl.h>
43#include <netinet/sctp_timer.h>
44#include <netinet/sctputil.h>
45#include <netinet/sctp_output.h>
46#include <netinet/sctp_header.h>
47#include <netinet/sctp_indata.h>
48#include <netinet/sctp_asconf.h>
49#include <netinet/sctp_input.h>
50#include <netinet/sctp.h>
51#include <netinet/sctp_uio.h>
52#include <netinet/udp.h>
53
54
55void
56sctp_audit_retranmission_queue(struct sctp_association *asoc)
57{
58	struct sctp_tmit_chunk *chk;
59
60	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit invoked on send queue cnt:%d onqueue:%d\n",
61	    asoc->sent_queue_retran_cnt,
62	    asoc->sent_queue_cnt);
63	asoc->sent_queue_retran_cnt = 0;
64	asoc->sent_queue_cnt = 0;
65	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
66		if (chk->sent == SCTP_DATAGRAM_RESEND) {
67			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
68		}
69		asoc->sent_queue_cnt++;
70	}
71	TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) {
72		if (chk->sent == SCTP_DATAGRAM_RESEND) {
73			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
74		}
75	}
76	TAILQ_FOREACH(chk, &asoc->asconf_send_queue, sctp_next) {
77		if (chk->sent == SCTP_DATAGRAM_RESEND) {
78			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
79		}
80	}
81	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit completes retran:%d onqueue:%d\n",
82	    asoc->sent_queue_retran_cnt,
83	    asoc->sent_queue_cnt);
84}
85
86int
87sctp_threshold_management(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
88    struct sctp_nets *net, uint16_t threshold)
89{
90	if (net) {
91		net->error_count++;
92		SCTPDBG(SCTP_DEBUG_TIMER4, "Error count for %p now %d thresh:%d\n",
93		    (void *)net, net->error_count,
94		    net->failure_threshold);
95		if (net->error_count > net->failure_threshold) {
96			/* We had a threshold failure */
97			if (net->dest_state & SCTP_ADDR_REACHABLE) {
98				net->dest_state &= ~SCTP_ADDR_REACHABLE;
99				net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY;
100				net->dest_state &= ~SCTP_ADDR_PF;
101				sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN,
102				    stcb, 0,
103				    (void *)net, SCTP_SO_NOT_LOCKED);
104			}
105		} else if ((net->pf_threshold < net->failure_threshold) &&
106		    (net->error_count > net->pf_threshold)) {
107			if (!(net->dest_state & SCTP_ADDR_PF)) {
108				net->dest_state |= SCTP_ADDR_PF;
109				net->last_active = sctp_get_tick_count();
110				sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
111				sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_TIMER + SCTP_LOC_3);
112				sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net);
113			}
114		}
115	}
116	if (stcb == NULL)
117		return (0);
118
119	if (net) {
120		if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) {
121			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
122				sctp_misc_ints(SCTP_THRESHOLD_INCR,
123				    stcb->asoc.overall_error_count,
124				    (stcb->asoc.overall_error_count + 1),
125				    SCTP_FROM_SCTP_TIMER,
126				    __LINE__);
127			}
128			stcb->asoc.overall_error_count++;
129		}
130	} else {
131		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
132			sctp_misc_ints(SCTP_THRESHOLD_INCR,
133			    stcb->asoc.overall_error_count,
134			    (stcb->asoc.overall_error_count + 1),
135			    SCTP_FROM_SCTP_TIMER,
136			    __LINE__);
137		}
138		stcb->asoc.overall_error_count++;
139	}
140	SCTPDBG(SCTP_DEBUG_TIMER4, "Overall error count for %p now %d thresh:%u state:%x\n",
141	    (void *)&stcb->asoc, stcb->asoc.overall_error_count,
142	    (uint32_t) threshold,
143	    ((net == NULL) ? (uint32_t) 0 : (uint32_t) net->dest_state));
144	/*
145	 * We specifically do not do >= to give the assoc one more change
146	 * before we fail it.
147	 */
148	if (stcb->asoc.overall_error_count > threshold) {
149		/* Abort notification sends a ULP notify */
150		struct mbuf *oper;
151
152		oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
153		    0, M_DONTWAIT, 1, MT_DATA);
154		if (oper) {
155			struct sctp_paramhdr *ph;
156			uint32_t *ippp;
157
158			SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
159			    sizeof(uint32_t);
160			ph = mtod(oper, struct sctp_paramhdr *);
161			ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
162			ph->param_length = htons(SCTP_BUF_LEN(oper));
163			ippp = (uint32_t *) (ph + 1);
164			*ippp = htonl(SCTP_FROM_SCTP_TIMER + SCTP_LOC_1);
165		}
166		inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_1;
167		sctp_abort_an_association(inp, stcb, oper, SCTP_SO_NOT_LOCKED);
168		return (1);
169	}
170	return (0);
171}
172
173/*
174 * sctp_find_alternate_net() returns a non-NULL pointer as long
175 * the argument net is non-NULL.
176 */
177struct sctp_nets *
178sctp_find_alternate_net(struct sctp_tcb *stcb,
179    struct sctp_nets *net,
180    int mode)
181{
182	/* Find and return an alternate network if possible */
183	struct sctp_nets *alt, *mnet, *min_errors_net = NULL, *max_cwnd_net = NULL;
184	int once;
185
186	/* JRS 5/14/07 - Initialize min_errors to an impossible value. */
187	int min_errors = -1;
188	uint32_t max_cwnd = 0;
189
190	if (stcb->asoc.numnets == 1) {
191		/* No others but net */
192		return (TAILQ_FIRST(&stcb->asoc.nets));
193	}
194	/*
195	 * JRS 5/14/07 - If mode is set to 2, use the CMT PF find alternate
196	 * net algorithm. This algorithm chooses the active destination (not
197	 * in PF state) with the largest cwnd value. If all destinations are
198	 * in PF state, unreachable, or unconfirmed, choose the desination
199	 * that is in PF state with the lowest error count. In case of a
200	 * tie, choose the destination that was most recently active.
201	 */
202	if (mode == 2) {
203		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
204			/*
205			 * JRS 5/14/07 - If the destination is unreachable
206			 * or unconfirmed, skip it.
207			 */
208			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
209			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
210				continue;
211			}
212			/*
213			 * JRS 5/14/07 -  If the destination is reachable
214			 * but in PF state, compare the error count of the
215			 * destination to the minimum error count seen thus
216			 * far. Store the destination with the lower error
217			 * count.  If the error counts are equal, store the
218			 * destination that was most recently active.
219			 */
220			if (mnet->dest_state & SCTP_ADDR_PF) {
221				/*
222				 * JRS 5/14/07 - If the destination under
223				 * consideration is the current destination,
224				 * work as if the error count is one higher.
225				 * The actual error count will not be
226				 * incremented until later in the t3
227				 * handler.
228				 */
229				if (mnet == net) {
230					if (min_errors == -1) {
231						min_errors = mnet->error_count + 1;
232						min_errors_net = mnet;
233					} else if (mnet->error_count + 1 < min_errors) {
234						min_errors = mnet->error_count + 1;
235						min_errors_net = mnet;
236					} else if (mnet->error_count + 1 == min_errors
237					    && mnet->last_active > min_errors_net->last_active) {
238						min_errors_net = mnet;
239						min_errors = mnet->error_count + 1;
240					}
241					continue;
242				} else {
243					if (min_errors == -1) {
244						min_errors = mnet->error_count;
245						min_errors_net = mnet;
246					} else if (mnet->error_count < min_errors) {
247						min_errors = mnet->error_count;
248						min_errors_net = mnet;
249					} else if (mnet->error_count == min_errors
250					    && mnet->last_active > min_errors_net->last_active) {
251						min_errors_net = mnet;
252						min_errors = mnet->error_count;
253					}
254					continue;
255				}
256			}
257			/*
258			 * JRS 5/14/07 - If the destination is reachable and
259			 * not in PF state, compare the cwnd of the
260			 * destination to the highest cwnd seen thus far.
261			 * Store the destination with the higher cwnd value.
262			 * If the cwnd values are equal, randomly choose one
263			 * of the two destinations.
264			 */
265			if (max_cwnd < mnet->cwnd) {
266				max_cwnd_net = mnet;
267				max_cwnd = mnet->cwnd;
268			} else if (max_cwnd == mnet->cwnd) {
269				uint32_t rndval;
270				uint8_t this_random;
271
272				if (stcb->asoc.hb_random_idx > 3) {
273					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
274					memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values));
275					this_random = stcb->asoc.hb_random_values[0];
276					stcb->asoc.hb_random_idx++;
277					stcb->asoc.hb_ect_randombit = 0;
278				} else {
279					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
280					stcb->asoc.hb_random_idx++;
281					stcb->asoc.hb_ect_randombit = 0;
282				}
283				if (this_random % 2 == 1) {
284					max_cwnd_net = mnet;
285					max_cwnd = mnet->cwnd;	/* Useless? */
286				}
287			}
288		}
289		if (max_cwnd_net == NULL) {
290			if (min_errors_net == NULL) {
291				return (net);
292			}
293			return (min_errors_net);
294		} else {
295			return (max_cwnd_net);
296		}
297	}
298	/*
299	 * JRS 5/14/07 - If mode is set to 1, use the CMT policy for
300	 * choosing an alternate net.
301	 */
302	else if (mode == 1) {
303		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
304			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
305			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
306				/*
307				 * will skip ones that are not-reachable or
308				 * unconfirmed
309				 */
310				continue;
311			}
312			if (max_cwnd < mnet->cwnd) {
313				max_cwnd_net = mnet;
314				max_cwnd = mnet->cwnd;
315			} else if (max_cwnd == mnet->cwnd) {
316				uint32_t rndval;
317				uint8_t this_random;
318
319				if (stcb->asoc.hb_random_idx > 3) {
320					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
321					memcpy(stcb->asoc.hb_random_values, &rndval,
322					    sizeof(stcb->asoc.hb_random_values));
323					this_random = stcb->asoc.hb_random_values[0];
324					stcb->asoc.hb_random_idx = 0;
325					stcb->asoc.hb_ect_randombit = 0;
326				} else {
327					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
328					stcb->asoc.hb_random_idx++;
329					stcb->asoc.hb_ect_randombit = 0;
330				}
331				if (this_random % 2) {
332					max_cwnd_net = mnet;
333					max_cwnd = mnet->cwnd;
334				}
335			}
336		}
337		if (max_cwnd_net) {
338			return (max_cwnd_net);
339		}
340	}
341	mnet = net;
342	once = 0;
343
344	if (mnet == NULL) {
345		mnet = TAILQ_FIRST(&stcb->asoc.nets);
346		if (mnet == NULL) {
347			return (NULL);
348		}
349	}
350	do {
351		alt = TAILQ_NEXT(mnet, sctp_next);
352		if (alt == NULL) {
353			once++;
354			if (once > 1) {
355				break;
356			}
357			alt = TAILQ_FIRST(&stcb->asoc.nets);
358			if (alt == NULL) {
359				return (NULL);
360			}
361		}
362		if (alt->ro.ro_rt == NULL) {
363			if (alt->ro._s_addr) {
364				sctp_free_ifa(alt->ro._s_addr);
365				alt->ro._s_addr = NULL;
366			}
367			alt->src_addr_selected = 0;
368		}
369		/* sa_ignore NO_NULL_CHK */
370		if (((alt->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE) &&
371		    (alt->ro.ro_rt != NULL) &&
372		    (!(alt->dest_state & SCTP_ADDR_UNCONFIRMED))) {
373			/* Found a reachable address */
374			break;
375		}
376		mnet = alt;
377	} while (alt != NULL);
378
379	if (alt == NULL) {
380		/* Case where NO insv network exists (dormant state) */
381		/* we rotate destinations */
382		once = 0;
383		mnet = net;
384		do {
385			if (mnet == NULL) {
386				return (TAILQ_FIRST(&stcb->asoc.nets));
387			}
388			alt = TAILQ_NEXT(mnet, sctp_next);
389			if (alt == NULL) {
390				once++;
391				if (once > 1) {
392					break;
393				}
394				alt = TAILQ_FIRST(&stcb->asoc.nets);
395			}
396			/* sa_ignore NO_NULL_CHK */
397			if ((!(alt->dest_state & SCTP_ADDR_UNCONFIRMED)) &&
398			    (alt != net)) {
399				/* Found an alternate address */
400				break;
401			}
402			mnet = alt;
403		} while (alt != NULL);
404	}
405	if (alt == NULL) {
406		return (net);
407	}
408	return (alt);
409}
410
411static void
412sctp_backoff_on_timeout(struct sctp_tcb *stcb,
413    struct sctp_nets *net,
414    int win_probe,
415    int num_marked, int num_abandoned)
416{
417	if (net->RTO == 0) {
418		net->RTO = stcb->asoc.minrto;
419	}
420	net->RTO <<= 1;
421	if (net->RTO > stcb->asoc.maxrto) {
422		net->RTO = stcb->asoc.maxrto;
423	}
424	if ((win_probe == 0) && (num_marked || num_abandoned)) {
425		/* We don't apply penalty to window probe scenarios */
426		/* JRS - Use the congestion control given in the CC module */
427		stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout(stcb, net);
428	}
429}
430
431#ifndef INVARIANTS
432static void
433sctp_recover_sent_list(struct sctp_tcb *stcb)
434{
435	struct sctp_tmit_chunk *chk, *nchk;
436	struct sctp_association *asoc;
437
438	asoc = &stcb->asoc;
439	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
440		if (SCTP_TSN_GE(asoc->last_acked_seq, chk->rec.data.TSN_seq)) {
441			SCTP_PRINTF("Found chk:%p tsn:%x <= last_acked_seq:%x\n",
442			    (void *)chk, chk->rec.data.TSN_seq, asoc->last_acked_seq);
443			if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
444				if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
445					asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
446				}
447			}
448			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
449			if (chk->pr_sctp_on) {
450				if (asoc->pr_sctp_cnt != 0)
451					asoc->pr_sctp_cnt--;
452			}
453			if (chk->data) {
454				/* sa_ignore NO_NULL_CHK */
455				sctp_free_bufspace(stcb, asoc, chk, 1);
456				sctp_m_freem(chk->data);
457				chk->data = NULL;
458				if (asoc->peer_supports_prsctp && PR_SCTP_BUF_ENABLED(chk->flags)) {
459					asoc->sent_queue_cnt_removeable--;
460				}
461			}
462			asoc->sent_queue_cnt--;
463			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
464		}
465	}
466	SCTP_PRINTF("after recover order is as follows\n");
467	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
468		SCTP_PRINTF("chk:%p TSN:%x\n", (void *)chk, chk->rec.data.TSN_seq);
469	}
470}
471
472#endif
473
474static int
475sctp_mark_all_for_resend(struct sctp_tcb *stcb,
476    struct sctp_nets *net,
477    struct sctp_nets *alt,
478    int window_probe,
479    int *num_marked,
480    int *num_abandoned)
481{
482
483	/*
484	 * Mark all chunks (well not all) that were sent to *net for
485	 * retransmission. Move them to alt for there destination as well...
486	 * We only mark chunks that have been outstanding long enough to
487	 * have received feed-back.
488	 */
489	struct sctp_tmit_chunk *chk, *nchk;
490	struct sctp_nets *lnets;
491	struct timeval now, min_wait, tv;
492	int cur_rto;
493	int cnt_abandoned;
494	int audit_tf, num_mk, fir;
495	unsigned int cnt_mk;
496	uint32_t orig_flight, orig_tf;
497	uint32_t tsnlast, tsnfirst;
498	int recovery_cnt = 0;
499
500
501	/* none in flight now */
502	audit_tf = 0;
503	fir = 0;
504	/*
505	 * figure out how long a data chunk must be pending before we can
506	 * mark it ..
507	 */
508	(void)SCTP_GETTIME_TIMEVAL(&now);
509	/* get cur rto in micro-seconds */
510	cur_rto = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv;
511	cur_rto *= 1000;
512	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
513		sctp_log_fr(cur_rto,
514		    stcb->asoc.peers_rwnd,
515		    window_probe,
516		    SCTP_FR_T3_MARK_TIME);
517		sctp_log_fr(net->flight_size, 0, 0, SCTP_FR_CWND_REPORT);
518		sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT);
519	}
520	tv.tv_sec = cur_rto / 1000000;
521	tv.tv_usec = cur_rto % 1000000;
522	min_wait = now;
523	timevalsub(&min_wait, &tv);
524	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
525		/*
526		 * if we hit here, we don't have enough seconds on the clock
527		 * to account for the RTO. We just let the lower seconds be
528		 * the bounds and don't worry about it. This may mean we
529		 * will mark a lot more than we should.
530		 */
531		min_wait.tv_sec = min_wait.tv_usec = 0;
532	}
533	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
534		sctp_log_fr(cur_rto, now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME);
535		sctp_log_fr(0, min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME);
536	}
537	/*
538	 * Our rwnd will be incorrect here since we are not adding back the
539	 * cnt * mbuf but we will fix that down below.
540	 */
541	orig_flight = net->flight_size;
542	orig_tf = stcb->asoc.total_flight;
543
544	net->fast_retran_ip = 0;
545	/* Now on to each chunk */
546	cnt_abandoned = 0;
547	num_mk = cnt_mk = 0;
548	tsnfirst = tsnlast = 0;
549#ifndef INVARIANTS
550start_again:
551#endif
552	TAILQ_FOREACH_SAFE(chk, &stcb->asoc.sent_queue, sctp_next, nchk) {
553		if (SCTP_TSN_GE(stcb->asoc.last_acked_seq, chk->rec.data.TSN_seq)) {
554			/* Strange case our list got out of order? */
555			SCTP_PRINTF("Our list is out of order? last_acked:%x chk:%x",
556			    (unsigned int)stcb->asoc.last_acked_seq, (unsigned int)chk->rec.data.TSN_seq);
557			recovery_cnt++;
558#ifdef INVARIANTS
559			panic("last acked >= chk on sent-Q");
560#else
561			SCTP_PRINTF("Recover attempts a restart cnt:%d\n", recovery_cnt);
562			sctp_recover_sent_list(stcb);
563			if (recovery_cnt < 10) {
564				goto start_again;
565			} else {
566				SCTP_PRINTF("Recovery fails %d times??\n", recovery_cnt);
567			}
568#endif
569		}
570		if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) {
571			/*
572			 * found one to mark: If it is less than
573			 * DATAGRAM_ACKED it MUST not be a skipped or marked
574			 * TSN but instead one that is either already set
575			 * for retransmission OR one that needs
576			 * retransmission.
577			 */
578
579			/* validate its been outstanding long enough */
580			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
581				sctp_log_fr(chk->rec.data.TSN_seq,
582				    chk->sent_rcv_time.tv_sec,
583				    chk->sent_rcv_time.tv_usec,
584				    SCTP_FR_T3_MARK_TIME);
585			}
586			if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) {
587				/*
588				 * we have reached a chunk that was sent
589				 * some seconds past our min.. forget it we
590				 * will find no more to send.
591				 */
592				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
593					sctp_log_fr(0,
594					    chk->sent_rcv_time.tv_sec,
595					    chk->sent_rcv_time.tv_usec,
596					    SCTP_FR_T3_STOPPED);
597				}
598				continue;
599			} else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) &&
600			    (window_probe == 0)) {
601				/*
602				 * we must look at the micro seconds to
603				 * know.
604				 */
605				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
606					/*
607					 * ok it was sent after our boundary
608					 * time.
609					 */
610					continue;
611				}
612			}
613			if (stcb->asoc.peer_supports_prsctp && PR_SCTP_TTL_ENABLED(chk->flags)) {
614				/* Is it expired? */
615				if (timevalcmp(&now, &chk->rec.data.timetodrop, >)) {
616					/* Yes so drop it */
617					if (chk->data) {
618						(void)sctp_release_pr_sctp_chunk(stcb,
619						    chk,
620						    1,
621						    SCTP_SO_NOT_LOCKED);
622						cnt_abandoned++;
623					}
624					continue;
625				}
626			}
627			if (stcb->asoc.peer_supports_prsctp && PR_SCTP_RTX_ENABLED(chk->flags)) {
628				/* Has it been retransmitted tv_sec times? */
629				if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) {
630					if (chk->data) {
631						(void)sctp_release_pr_sctp_chunk(stcb,
632						    chk,
633						    1,
634						    SCTP_SO_NOT_LOCKED);
635						cnt_abandoned++;
636					}
637					continue;
638				}
639			}
640			if (chk->sent < SCTP_DATAGRAM_RESEND) {
641				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
642				num_mk++;
643				if (fir == 0) {
644					fir = 1;
645					tsnfirst = chk->rec.data.TSN_seq;
646				}
647				tsnlast = chk->rec.data.TSN_seq;
648				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
649					sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count,
650					    0, SCTP_FR_T3_MARKED);
651				}
652				if (chk->rec.data.chunk_was_revoked) {
653					/* deflate the cwnd */
654					chk->whoTo->cwnd -= chk->book_size;
655					chk->rec.data.chunk_was_revoked = 0;
656				}
657				net->marked_retrans++;
658				stcb->asoc.marked_retrans++;
659				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
660					sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO,
661					    chk->whoTo->flight_size,
662					    chk->book_size,
663					    (uintptr_t) chk->whoTo,
664					    chk->rec.data.TSN_seq);
665				}
666				sctp_flight_size_decrease(chk);
667				sctp_total_flight_decrease(stcb, chk);
668				stcb->asoc.peers_rwnd += chk->send_size;
669				stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh);
670			}
671			chk->sent = SCTP_DATAGRAM_RESEND;
672			SCTP_STAT_INCR(sctps_markedretrans);
673
674			/* reset the TSN for striking and other FR stuff */
675			chk->rec.data.doing_fast_retransmit = 0;
676			/* Clear any time so NO RTT is being done */
677
678			if (chk->do_rtt) {
679				if (chk->whoTo->rto_needed == 0) {
680					chk->whoTo->rto_needed = 1;
681				}
682			}
683			chk->do_rtt = 0;
684			if (alt != net) {
685				sctp_free_remote_addr(chk->whoTo);
686				chk->no_fr_allowed = 1;
687				chk->whoTo = alt;
688				atomic_add_int(&alt->ref_count, 1);
689			} else {
690				chk->no_fr_allowed = 0;
691				if (TAILQ_EMPTY(&stcb->asoc.send_queue)) {
692					chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq;
693				} else {
694					chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.TSN_seq;
695				}
696			}
697			/*
698			 * CMT: Do not allow FRs on retransmitted TSNs.
699			 */
700			if (stcb->asoc.sctp_cmt_on_off > 0) {
701				chk->no_fr_allowed = 1;
702			}
703#ifdef THIS_SHOULD_NOT_BE_DONE
704		} else if (chk->sent == SCTP_DATAGRAM_ACKED) {
705			/* remember highest acked one */
706			could_be_sent = chk;
707#endif
708		}
709		if (chk->sent == SCTP_DATAGRAM_RESEND) {
710			cnt_mk++;
711		}
712	}
713	if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) {
714		/* we did not subtract the same things? */
715		audit_tf = 1;
716	}
717	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
718		sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT);
719	}
720#ifdef SCTP_DEBUG
721	if (num_mk) {
722		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
723		    tsnlast);
724		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%ld\n",
725		    num_mk, (u_long)stcb->asoc.peers_rwnd);
726		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
727		    tsnlast);
728		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%d\n",
729		    num_mk,
730		    (int)stcb->asoc.peers_rwnd);
731	}
732#endif
733	*num_marked = num_mk;
734	*num_abandoned = cnt_abandoned;
735	/*
736	 * Now check for a ECN Echo that may be stranded And include the
737	 * cnt_mk'd to have all resends in the control queue.
738	 */
739	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
740		if (chk->sent == SCTP_DATAGRAM_RESEND) {
741			cnt_mk++;
742		}
743		if ((chk->whoTo == net) &&
744		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
745			sctp_free_remote_addr(chk->whoTo);
746			chk->whoTo = alt;
747			if (chk->sent != SCTP_DATAGRAM_RESEND) {
748				chk->sent = SCTP_DATAGRAM_RESEND;
749				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
750				cnt_mk++;
751			}
752			atomic_add_int(&alt->ref_count, 1);
753		}
754	}
755#ifdef THIS_SHOULD_NOT_BE_DONE
756	if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) {
757		/* fix it so we retransmit the highest acked anyway */
758		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
759		cnt_mk++;
760		could_be_sent->sent = SCTP_DATAGRAM_RESEND;
761	}
762#endif
763	if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) {
764#ifdef INVARIANTS
765		SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n",
766		    cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk);
767#endif
768#ifndef SCTP_AUDITING_ENABLED
769		stcb->asoc.sent_queue_retran_cnt = cnt_mk;
770#endif
771	}
772	if (audit_tf) {
773		SCTPDBG(SCTP_DEBUG_TIMER4,
774		    "Audit total flight due to negative value net:%p\n",
775		    (void *)net);
776		stcb->asoc.total_flight = 0;
777		stcb->asoc.total_flight_count = 0;
778		/* Clear all networks flight size */
779		TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) {
780			lnets->flight_size = 0;
781			SCTPDBG(SCTP_DEBUG_TIMER4,
782			    "Net:%p c-f cwnd:%d ssthresh:%d\n",
783			    (void *)lnets, lnets->cwnd, lnets->ssthresh);
784		}
785		TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
786			if (chk->sent < SCTP_DATAGRAM_RESEND) {
787				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
788					sctp_misc_ints(SCTP_FLIGHT_LOG_UP,
789					    chk->whoTo->flight_size,
790					    chk->book_size,
791					    (uintptr_t) chk->whoTo,
792					    chk->rec.data.TSN_seq);
793				}
794				sctp_flight_size_increase(chk);
795				sctp_total_flight_increase(stcb, chk);
796			}
797		}
798	}
799	/* We return 1 if we only have a window probe outstanding */
800	return (0);
801}
802
803
804int
805sctp_t3rxt_timer(struct sctp_inpcb *inp,
806    struct sctp_tcb *stcb,
807    struct sctp_nets *net)
808{
809	struct sctp_nets *alt;
810	int win_probe, num_mk, num_abandoned;
811
812	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
813		sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT);
814	}
815	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
816		struct sctp_nets *lnet;
817
818		TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
819			if (net == lnet) {
820				sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3);
821			} else {
822				sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3);
823			}
824		}
825	}
826	/* Find an alternate and mark those for retransmission */
827	if ((stcb->asoc.peers_rwnd == 0) &&
828	    (stcb->asoc.total_flight < net->mtu)) {
829		SCTP_STAT_INCR(sctps_timowindowprobe);
830		win_probe = 1;
831	} else {
832		win_probe = 0;
833	}
834
835	if (win_probe == 0) {
836		/* We don't do normal threshold management on window probes */
837		if (sctp_threshold_management(inp, stcb, net,
838		    stcb->asoc.max_send_times)) {
839			/* Association was destroyed */
840			return (1);
841		} else {
842			if (net != stcb->asoc.primary_destination) {
843				/* send a immediate HB if our RTO is stale */
844				struct timeval now;
845				unsigned int ms_goneby;
846
847				(void)SCTP_GETTIME_TIMEVAL(&now);
848				if (net->last_sent_time.tv_sec) {
849					ms_goneby = (now.tv_sec - net->last_sent_time.tv_sec) * 1000;
850				} else {
851					ms_goneby = 0;
852				}
853				if ((net->dest_state & SCTP_ADDR_PF) == 0) {
854					if ((ms_goneby > net->RTO) || (net->RTO == 0)) {
855						/*
856						 * no recent feed back in an
857						 * RTO or more, request a
858						 * RTT update
859						 */
860						sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
861					}
862				}
863			}
864		}
865	} else {
866		/*
867		 * For a window probe we don't penalize the net's but only
868		 * the association. This may fail it if SACKs are not coming
869		 * back. If sack's are coming with rwnd locked at 0, we will
870		 * continue to hold things waiting for rwnd to raise
871		 */
872		if (sctp_threshold_management(inp, stcb, NULL,
873		    stcb->asoc.max_send_times)) {
874			/* Association was destroyed */
875			return (1);
876		}
877	}
878	if (stcb->asoc.sctp_cmt_on_off > 0) {
879		if (net->pf_threshold < net->failure_threshold) {
880			alt = sctp_find_alternate_net(stcb, net, 2);
881		} else {
882			/*
883			 * CMT: Using RTX_SSTHRESH policy for CMT. If CMT is
884			 * being used, then pick dest with largest ssthresh
885			 * for any retransmission.
886			 */
887			alt = sctp_find_alternate_net(stcb, net, 1);
888			/*
889			 * CUCv2: If a different dest is picked for the
890			 * retransmission, then new (rtx-)pseudo_cumack
891			 * needs to be tracked for orig dest. Let CUCv2
892			 * track new (rtx-) pseudo-cumack always.
893			 */
894			net->find_pseudo_cumack = 1;
895			net->find_rtx_pseudo_cumack = 1;
896		}
897	} else {
898		alt = sctp_find_alternate_net(stcb, net, 0);
899	}
900
901	num_mk = 0;
902	num_abandoned = 0;
903	(void)sctp_mark_all_for_resend(stcb, net, alt, win_probe,
904	    &num_mk, &num_abandoned);
905	/* FR Loss recovery just ended with the T3. */
906	stcb->asoc.fast_retran_loss_recovery = 0;
907
908	/* CMT FR loss recovery ended with the T3 */
909	net->fast_retran_loss_recovery = 0;
910	if ((stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) &&
911	    (net->flight_size == 0)) {
912		(*stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) (stcb, net);
913	}
914	/*
915	 * setup the sat loss recovery that prevents satellite cwnd advance.
916	 */
917	stcb->asoc.sat_t3_loss_recovery = 1;
918	stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq;
919
920	/* Backoff the timer and cwnd */
921	sctp_backoff_on_timeout(stcb, net, win_probe, num_mk, num_abandoned);
922	if ((!(net->dest_state & SCTP_ADDR_REACHABLE)) ||
923	    (net->dest_state & SCTP_ADDR_PF)) {
924		/* Move all pending over too */
925		sctp_move_chunks_from_net(stcb, net);
926
927		/*
928		 * Get the address that failed, to force a new src address
929		 * selecton and a route allocation.
930		 */
931		if (net->ro._s_addr) {
932			sctp_free_ifa(net->ro._s_addr);
933			net->ro._s_addr = NULL;
934		}
935		net->src_addr_selected = 0;
936
937		/* Force a route allocation too */
938		if (net->ro.ro_rt) {
939			RTFREE(net->ro.ro_rt);
940			net->ro.ro_rt = NULL;
941		}
942		/* Was it our primary? */
943		if ((stcb->asoc.primary_destination == net) && (alt != net)) {
944			/*
945			 * Yes, note it as such and find an alternate note:
946			 * this means HB code must use this to resent the
947			 * primary if it goes active AND if someone does a
948			 * change-primary then this flag must be cleared
949			 * from any net structures.
950			 */
951			if (stcb->asoc.alternate) {
952				sctp_free_remote_addr(stcb->asoc.alternate);
953			}
954			stcb->asoc.alternate = alt;
955			atomic_add_int(&stcb->asoc.alternate->ref_count, 1);
956		}
957	}
958	/*
959	 * Special case for cookie-echo'ed case, we don't do output but must
960	 * await the COOKIE-ACK before retransmission
961	 */
962	if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
963		/*
964		 * Here we just reset the timer and start again since we
965		 * have not established the asoc
966		 */
967		sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net);
968		return (0);
969	}
970	if (stcb->asoc.peer_supports_prsctp) {
971		struct sctp_tmit_chunk *lchk;
972
973		lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc);
974		/* C3. See if we need to send a Fwd-TSN */
975		if (SCTP_TSN_GT(stcb->asoc.advanced_peer_ack_point, stcb->asoc.last_acked_seq)) {
976			send_forward_tsn(stcb, &stcb->asoc);
977			if (lchk) {
978				/* Assure a timer is up */
979				sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo);
980			}
981		}
982	}
983	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
984		sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX);
985	}
986	return (0);
987}
988
989int
990sctp_t1init_timer(struct sctp_inpcb *inp,
991    struct sctp_tcb *stcb,
992    struct sctp_nets *net)
993{
994	/* bump the thresholds */
995	if (stcb->asoc.delayed_connection) {
996		/*
997		 * special hook for delayed connection. The library did NOT
998		 * complete the rest of its sends.
999		 */
1000		stcb->asoc.delayed_connection = 0;
1001		sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1002		return (0);
1003	}
1004	if (SCTP_GET_STATE((&stcb->asoc)) != SCTP_STATE_COOKIE_WAIT) {
1005		return (0);
1006	}
1007	if (sctp_threshold_management(inp, stcb, net,
1008	    stcb->asoc.max_init_times)) {
1009		/* Association was destroyed */
1010		return (1);
1011	}
1012	stcb->asoc.dropped_special_cnt = 0;
1013	sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0, 0);
1014	if (stcb->asoc.initial_init_rto_max < net->RTO) {
1015		net->RTO = stcb->asoc.initial_init_rto_max;
1016	}
1017	if (stcb->asoc.numnets > 1) {
1018		/* If we have more than one addr use it */
1019		struct sctp_nets *alt;
1020
1021		alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0);
1022		if (alt != stcb->asoc.primary_destination) {
1023			sctp_move_chunks_from_net(stcb, stcb->asoc.primary_destination);
1024			stcb->asoc.primary_destination = alt;
1025		}
1026	}
1027	/* Send out a new init */
1028	sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1029	return (0);
1030}
1031
1032/*
1033 * For cookie and asconf we actually need to find and mark for resend, then
1034 * increment the resend counter (after all the threshold management stuff of
1035 * course).
1036 */
1037int
1038sctp_cookie_timer(struct sctp_inpcb *inp,
1039    struct sctp_tcb *stcb,
1040    struct sctp_nets *net SCTP_UNUSED)
1041{
1042	struct sctp_nets *alt;
1043	struct sctp_tmit_chunk *cookie;
1044
1045	/* first before all else we must find the cookie */
1046	TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) {
1047		if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) {
1048			break;
1049		}
1050	}
1051	if (cookie == NULL) {
1052		if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
1053			/* FOOBAR! */
1054			struct mbuf *oper;
1055
1056			oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
1057			    0, M_DONTWAIT, 1, MT_DATA);
1058			if (oper) {
1059				struct sctp_paramhdr *ph;
1060				uint32_t *ippp;
1061
1062				SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
1063				    sizeof(uint32_t);
1064				ph = mtod(oper, struct sctp_paramhdr *);
1065				ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1066				ph->param_length = htons(SCTP_BUF_LEN(oper));
1067				ippp = (uint32_t *) (ph + 1);
1068				*ippp = htonl(SCTP_FROM_SCTP_TIMER + SCTP_LOC_3);
1069			}
1070			inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_4;
1071			sctp_abort_an_association(inp, stcb, oper, SCTP_SO_NOT_LOCKED);
1072		} else {
1073#ifdef INVARIANTS
1074			panic("Cookie timer expires in wrong state?");
1075#else
1076			SCTP_PRINTF("Strange in state %d not cookie-echoed yet c-e timer expires?\n", SCTP_GET_STATE(&stcb->asoc));
1077			return (0);
1078#endif
1079		}
1080		return (0);
1081	}
1082	/* Ok we found the cookie, threshold management next */
1083	if (sctp_threshold_management(inp, stcb, cookie->whoTo,
1084	    stcb->asoc.max_init_times)) {
1085		/* Assoc is over */
1086		return (1);
1087	}
1088	/*
1089	 * cleared theshold management now lets backoff the address & select
1090	 * an alternate
1091	 */
1092	stcb->asoc.dropped_special_cnt = 0;
1093	sctp_backoff_on_timeout(stcb, cookie->whoTo, 1, 0, 0);
1094	alt = sctp_find_alternate_net(stcb, cookie->whoTo, 0);
1095	if (alt != cookie->whoTo) {
1096		sctp_free_remote_addr(cookie->whoTo);
1097		cookie->whoTo = alt;
1098		atomic_add_int(&alt->ref_count, 1);
1099	}
1100	/* Now mark the retran info */
1101	if (cookie->sent != SCTP_DATAGRAM_RESEND) {
1102		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1103	}
1104	cookie->sent = SCTP_DATAGRAM_RESEND;
1105	/*
1106	 * Now call the output routine to kick out the cookie again, Note we
1107	 * don't mark any chunks for retran so that FR will need to kick in
1108	 * to move these (or a send timer).
1109	 */
1110	return (0);
1111}
1112
1113int
1114sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1115    struct sctp_nets *net)
1116{
1117	struct sctp_nets *alt;
1118	struct sctp_tmit_chunk *strrst = NULL, *chk = NULL;
1119
1120	if (stcb->asoc.stream_reset_outstanding == 0) {
1121		return (0);
1122	}
1123	/* find the existing STRRESET, we use the seq number we sent out on */
1124	(void)sctp_find_stream_reset(stcb, stcb->asoc.str_reset_seq_out, &strrst);
1125	if (strrst == NULL) {
1126		return (0);
1127	}
1128	/* do threshold management */
1129	if (sctp_threshold_management(inp, stcb, strrst->whoTo,
1130	    stcb->asoc.max_send_times)) {
1131		/* Assoc is over */
1132		return (1);
1133	}
1134	/*
1135	 * cleared theshold management now lets backoff the address & select
1136	 * an alternate
1137	 */
1138	sctp_backoff_on_timeout(stcb, strrst->whoTo, 1, 0, 0);
1139	alt = sctp_find_alternate_net(stcb, strrst->whoTo, 0);
1140	sctp_free_remote_addr(strrst->whoTo);
1141	strrst->whoTo = alt;
1142	atomic_add_int(&alt->ref_count, 1);
1143
1144	/* See if a ECN Echo is also stranded */
1145	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1146		if ((chk->whoTo == net) &&
1147		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1148			sctp_free_remote_addr(chk->whoTo);
1149			if (chk->sent != SCTP_DATAGRAM_RESEND) {
1150				chk->sent = SCTP_DATAGRAM_RESEND;
1151				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1152			}
1153			chk->whoTo = alt;
1154			atomic_add_int(&alt->ref_count, 1);
1155		}
1156	}
1157	if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1158		/*
1159		 * If the address went un-reachable, we need to move to
1160		 * alternates for ALL chk's in queue
1161		 */
1162		sctp_move_chunks_from_net(stcb, net);
1163	}
1164	/* mark the retran info */
1165	if (strrst->sent != SCTP_DATAGRAM_RESEND)
1166		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1167	strrst->sent = SCTP_DATAGRAM_RESEND;
1168
1169	/* restart the timer */
1170	sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, strrst->whoTo);
1171	return (0);
1172}
1173
1174int
1175sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1176    struct sctp_nets *net)
1177{
1178	struct sctp_nets *alt;
1179	struct sctp_tmit_chunk *asconf, *chk;
1180
1181	/* is this a first send, or a retransmission? */
1182	if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) {
1183		/* compose a new ASCONF chunk and send it */
1184		sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED);
1185	} else {
1186		/*
1187		 * Retransmission of the existing ASCONF is needed
1188		 */
1189
1190		/* find the existing ASCONF */
1191		asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue);
1192		if (asconf == NULL) {
1193			return (0);
1194		}
1195		/* do threshold management */
1196		if (sctp_threshold_management(inp, stcb, asconf->whoTo,
1197		    stcb->asoc.max_send_times)) {
1198			/* Assoc is over */
1199			return (1);
1200		}
1201		if (asconf->snd_count > stcb->asoc.max_send_times) {
1202			/*
1203			 * Something is rotten: our peer is not responding
1204			 * to ASCONFs but apparently is to other chunks.
1205			 * i.e. it is not properly handling the chunk type
1206			 * upper bits. Mark this peer as ASCONF incapable
1207			 * and cleanup.
1208			 */
1209			SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n");
1210			sctp_asconf_cleanup(stcb, net);
1211			return (0);
1212		}
1213		/*
1214		 * cleared threshold management, so now backoff the net and
1215		 * select an alternate
1216		 */
1217		sctp_backoff_on_timeout(stcb, asconf->whoTo, 1, 0, 0);
1218		alt = sctp_find_alternate_net(stcb, asconf->whoTo, 0);
1219		if (asconf->whoTo != alt) {
1220			sctp_free_remote_addr(asconf->whoTo);
1221			asconf->whoTo = alt;
1222			atomic_add_int(&alt->ref_count, 1);
1223		}
1224		/* See if an ECN Echo is also stranded */
1225		TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1226			if ((chk->whoTo == net) &&
1227			    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1228				sctp_free_remote_addr(chk->whoTo);
1229				chk->whoTo = alt;
1230				if (chk->sent != SCTP_DATAGRAM_RESEND) {
1231					chk->sent = SCTP_DATAGRAM_RESEND;
1232					sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1233				}
1234				atomic_add_int(&alt->ref_count, 1);
1235			}
1236		}
1237		TAILQ_FOREACH(chk, &stcb->asoc.asconf_send_queue, sctp_next) {
1238			if (chk->whoTo != alt) {
1239				sctp_free_remote_addr(chk->whoTo);
1240				chk->whoTo = alt;
1241				atomic_add_int(&alt->ref_count, 1);
1242			}
1243			if (asconf->sent != SCTP_DATAGRAM_RESEND && chk->sent != SCTP_DATAGRAM_UNSENT)
1244				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1245			chk->sent = SCTP_DATAGRAM_RESEND;
1246		}
1247		if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1248			/*
1249			 * If the address went un-reachable, we need to move
1250			 * to the alternate for ALL chunks in queue
1251			 */
1252			sctp_move_chunks_from_net(stcb, net);
1253		}
1254		/* mark the retran info */
1255		if (asconf->sent != SCTP_DATAGRAM_RESEND)
1256			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1257		asconf->sent = SCTP_DATAGRAM_RESEND;
1258
1259		/* send another ASCONF if any and we can do */
1260		sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED);
1261	}
1262	return (0);
1263}
1264
1265/* Mobility adaptation */
1266void
1267sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1268    struct sctp_nets *net SCTP_UNUSED)
1269{
1270	if (stcb->asoc.deleted_primary == NULL) {
1271		SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n");
1272		sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1273		return;
1274	}
1275	SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary ");
1276	SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa);
1277	sctp_free_remote_addr(stcb->asoc.deleted_primary);
1278	stcb->asoc.deleted_primary = NULL;
1279	sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1280	return;
1281}
1282
1283/*
1284 * For the shutdown and shutdown-ack, we do not keep one around on the
1285 * control queue. This means we must generate a new one and call the general
1286 * chunk output routine, AFTER having done threshold management.
1287 * It is assumed that net is non-NULL.
1288 */
1289int
1290sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1291    struct sctp_nets *net)
1292{
1293	struct sctp_nets *alt;
1294
1295	/* first threshold managment */
1296	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1297		/* Assoc is over */
1298		return (1);
1299	}
1300	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1301	/* second select an alternative */
1302	alt = sctp_find_alternate_net(stcb, net, 0);
1303
1304	/* third generate a shutdown into the queue for out net */
1305	sctp_send_shutdown(stcb, alt);
1306
1307	/* fourth restart timer */
1308	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt);
1309	return (0);
1310}
1311
1312int
1313sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1314    struct sctp_nets *net)
1315{
1316	struct sctp_nets *alt;
1317
1318	/* first threshold managment */
1319	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1320		/* Assoc is over */
1321		return (1);
1322	}
1323	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1324	/* second select an alternative */
1325	alt = sctp_find_alternate_net(stcb, net, 0);
1326
1327	/* third generate a shutdown into the queue for out net */
1328	sctp_send_shutdown_ack(stcb, alt);
1329
1330	/* fourth restart timer */
1331	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt);
1332	return (0);
1333}
1334
1335static void
1336sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp,
1337    struct sctp_tcb *stcb)
1338{
1339	struct sctp_stream_queue_pending *sp;
1340	unsigned int i, chks_in_queue = 0;
1341	int being_filled = 0;
1342
1343	/*
1344	 * This function is ONLY called when the send/sent queues are empty.
1345	 */
1346	if ((stcb == NULL) || (inp == NULL))
1347		return;
1348
1349	if (stcb->asoc.sent_queue_retran_cnt) {
1350		SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n",
1351		    stcb->asoc.sent_queue_retran_cnt);
1352		stcb->asoc.sent_queue_retran_cnt = 0;
1353	}
1354	if (stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1355		/* No stream scheduler information, initialize scheduler */
1356		stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc, 0);
1357		if (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1358			/* yep, we lost a stream or two */
1359			SCTP_PRINTF("Found additional streams NOT managed by scheduler, corrected\n");
1360		} else {
1361			/* no streams lost */
1362			stcb->asoc.total_output_queue_size = 0;
1363		}
1364	}
1365	/* Check to see if some data queued, if so report it */
1366	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
1367		if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) {
1368			TAILQ_FOREACH(sp, &stcb->asoc.strmout[i].outqueue, next) {
1369				if (sp->msg_is_complete)
1370					being_filled++;
1371				chks_in_queue++;
1372			}
1373		}
1374	}
1375	if (chks_in_queue != stcb->asoc.stream_queue_cnt) {
1376		SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n",
1377		    stcb->asoc.stream_queue_cnt, chks_in_queue);
1378	}
1379	if (chks_in_queue) {
1380		/* call the output queue function */
1381		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1382		if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1383		    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1384			/*
1385			 * Probably should go in and make it go back through
1386			 * and add fragments allowed
1387			 */
1388			if (being_filled == 0) {
1389				SCTP_PRINTF("Still nothing moved %d chunks are stuck\n",
1390				    chks_in_queue);
1391			}
1392		}
1393	} else {
1394		SCTP_PRINTF("Found no chunks on any queue tot:%lu\n",
1395		    (u_long)stcb->asoc.total_output_queue_size);
1396		stcb->asoc.total_output_queue_size = 0;
1397	}
1398}
1399
1400int
1401sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1402    struct sctp_nets *net)
1403{
1404	uint8_t net_was_pf;
1405
1406	if (net->dest_state & SCTP_ADDR_PF) {
1407		net_was_pf = 1;
1408	} else {
1409		net_was_pf = 0;
1410	}
1411	if (net->hb_responded == 0) {
1412		if (net->ro._s_addr) {
1413			/*
1414			 * Invalidate the src address if we did not get a
1415			 * response last time.
1416			 */
1417			sctp_free_ifa(net->ro._s_addr);
1418			net->ro._s_addr = NULL;
1419			net->src_addr_selected = 0;
1420		}
1421		sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1422		if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1423			/* Assoc is over */
1424			return (1);
1425		}
1426	}
1427	/* Zero PBA, if it needs it */
1428	if (net->partial_bytes_acked) {
1429		net->partial_bytes_acked = 0;
1430	}
1431	if ((stcb->asoc.total_output_queue_size > 0) &&
1432	    (TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1433	    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1434		sctp_audit_stream_queues_for_size(inp, stcb);
1435	}
1436	if (!(net->dest_state & SCTP_ADDR_NOHB) &&
1437	    !((net_was_pf == 0) && (net->dest_state & SCTP_ADDR_PF))) {
1438		/*
1439		 * when move to PF during threshold mangement, a HB has been
1440		 * queued in that routine
1441		 */
1442		uint32_t ms_gone_by;
1443
1444		if ((net->last_sent_time.tv_sec > 0) ||
1445		    (net->last_sent_time.tv_usec > 0)) {
1446			struct timeval diff;
1447
1448			SCTP_GETTIME_TIMEVAL(&diff);
1449			timevalsub(&diff, &net->last_sent_time);
1450			ms_gone_by = (uint32_t) (diff.tv_sec * 1000) +
1451			    (uint32_t) (diff.tv_usec / 1000);
1452		} else {
1453			ms_gone_by = 0xffffffff;
1454		}
1455		if ((ms_gone_by >= net->heart_beat_delay) ||
1456		    (net->dest_state & SCTP_ADDR_PF)) {
1457			sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
1458		}
1459	}
1460	return (0);
1461}
1462
1463void
1464sctp_pathmtu_timer(struct sctp_inpcb *inp,
1465    struct sctp_tcb *stcb,
1466    struct sctp_nets *net)
1467{
1468	uint32_t next_mtu, mtu;
1469
1470	next_mtu = sctp_get_next_mtu(net->mtu);
1471
1472	if ((next_mtu > net->mtu) && (net->port == 0)) {
1473		if ((net->src_addr_selected == 0) ||
1474		    (net->ro._s_addr == NULL) ||
1475		    (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1476			if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1477				sctp_free_ifa(net->ro._s_addr);
1478				net->ro._s_addr = NULL;
1479				net->src_addr_selected = 0;
1480			} else if (net->ro._s_addr == NULL) {
1481#if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1482				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1483					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1484
1485					/* KAME hack: embed scopeid */
1486					(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
1487				}
1488#endif
1489
1490				net->ro._s_addr = sctp_source_address_selection(inp,
1491				    stcb,
1492				    (sctp_route_t *) & net->ro,
1493				    net, 0, stcb->asoc.vrf_id);
1494#if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1495				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1496					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1497
1498					(void)sa6_recoverscope(sin6);
1499				}
1500#endif				/* INET6 */
1501			}
1502			if (net->ro._s_addr)
1503				net->src_addr_selected = 1;
1504		}
1505		if (net->ro._s_addr) {
1506			mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_rt);
1507			if (net->port) {
1508				mtu -= sizeof(struct udphdr);
1509			}
1510			if (mtu > next_mtu) {
1511				net->mtu = next_mtu;
1512			}
1513		}
1514	}
1515	/* restart the timer */
1516	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
1517}
1518
1519void
1520sctp_autoclose_timer(struct sctp_inpcb *inp,
1521    struct sctp_tcb *stcb,
1522    struct sctp_nets *net)
1523{
1524	struct timeval tn, *tim_touse;
1525	struct sctp_association *asoc;
1526	int ticks_gone_by;
1527
1528	(void)SCTP_GETTIME_TIMEVAL(&tn);
1529	if (stcb->asoc.sctp_autoclose_ticks &&
1530	    sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) {
1531		/* Auto close is on */
1532		asoc = &stcb->asoc;
1533		/* pick the time to use */
1534		if (asoc->time_last_rcvd.tv_sec >
1535		    asoc->time_last_sent.tv_sec) {
1536			tim_touse = &asoc->time_last_rcvd;
1537		} else {
1538			tim_touse = &asoc->time_last_sent;
1539		}
1540		/* Now has long enough transpired to autoclose? */
1541		ticks_gone_by = SEC_TO_TICKS(tn.tv_sec - tim_touse->tv_sec);
1542		if ((ticks_gone_by > 0) &&
1543		    (ticks_gone_by >= (int)asoc->sctp_autoclose_ticks)) {
1544			/*
1545			 * autoclose time has hit, call the output routine,
1546			 * which should do nothing just to be SURE we don't
1547			 * have hanging data. We can then safely check the
1548			 * queues and know that we are clear to send
1549			 * shutdown
1550			 */
1551			sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED);
1552			/* Are we clean? */
1553			if (TAILQ_EMPTY(&asoc->send_queue) &&
1554			    TAILQ_EMPTY(&asoc->sent_queue)) {
1555				/*
1556				 * there is nothing queued to send, so I'm
1557				 * done...
1558				 */
1559				if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) {
1560					/* only send SHUTDOWN 1st time thru */
1561					struct sctp_nets *netp;
1562
1563					if (stcb->asoc.alternate) {
1564						netp = stcb->asoc.alternate;
1565					} else {
1566						netp = stcb->asoc.primary_destination;
1567					}
1568					sctp_send_shutdown(stcb, netp);
1569					if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
1570					    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
1571						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
1572					}
1573					SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
1574					SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
1575					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
1576					    stcb->sctp_ep, stcb,
1577					    netp);
1578					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
1579					    stcb->sctp_ep, stcb,
1580					    netp);
1581				}
1582			}
1583		} else {
1584			/*
1585			 * No auto close at this time, reset t-o to check
1586			 * later
1587			 */
1588			int tmp;
1589
1590			/* fool the timer startup to use the time left */
1591			tmp = asoc->sctp_autoclose_ticks;
1592			asoc->sctp_autoclose_ticks -= ticks_gone_by;
1593			sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb,
1594			    net);
1595			/* restore the real tick value */
1596			asoc->sctp_autoclose_ticks = tmp;
1597		}
1598	}
1599}
1600