xref: /freebsd/sys/netinet/sctp_timer.c (revision d6af161a344d22e8cf5b33f84941afab682ba816)
1 /*-
2  * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved.
3  *
4  * Redistribution and use in source and binary forms, with or without
5  * modification, are permitted provided that the following conditions are met:
6  *
7  * a) Redistributions of source code must retain the above copyright notice,
8  *   this list of conditions and the following disclaimer.
9  *
10  * b) Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in
12  *   the documentation and/or other materials provided with the distribution.
13  *
14  * c) Neither the name of Cisco Systems, Inc. nor the names of its
15  *    contributors may be used to endorse or promote products derived
16  *    from this software without specific prior written permission.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
20  * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
22  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
23  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
24  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
25  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
26  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
27  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
28  * THE POSSIBILITY OF SUCH DAMAGE.
29  */
30 
31 /* $KAME: sctp_timer.c,v 1.29 2005/03/06 16:04:18 itojun Exp $	 */
32 
33 #include <sys/cdefs.h>
34 __FBSDID("$FreeBSD$");
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 
53 
54 
55 void
56 sctp_early_fr_timer(struct sctp_inpcb *inp,
57     struct sctp_tcb *stcb,
58     struct sctp_nets *net)
59 {
60 	struct sctp_tmit_chunk *chk, *tp2;
61 	struct timeval now, min_wait, tv;
62 	unsigned int cur_rtt, cnt = 0, cnt_resend = 0;
63 
64 	/* an early FR is occuring. */
65 	(void)SCTP_GETTIME_TIMEVAL(&now);
66 	/* get cur rto in micro-seconds */
67 	if (net->lastsa == 0) {
68 		/* Hmm no rtt estimate yet? */
69 		cur_rtt = stcb->asoc.initial_rto >> 2;
70 	} else {
71 
72 		cur_rtt = ((net->lastsa >> 2) + net->lastsv) >> 1;
73 	}
74 	if (cur_rtt < SCTP_BASE_SYSCTL(sctp_early_fr_msec)) {
75 		cur_rtt = SCTP_BASE_SYSCTL(sctp_early_fr_msec);
76 	}
77 	cur_rtt *= 1000;
78 	tv.tv_sec = cur_rtt / 1000000;
79 	tv.tv_usec = cur_rtt % 1000000;
80 	min_wait = now;
81 	timevalsub(&min_wait, &tv);
82 	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
83 		/*
84 		 * if we hit here, we don't have enough seconds on the clock
85 		 * to account for the RTO. We just let the lower seconds be
86 		 * the bounds and don't worry about it. This may mean we
87 		 * will mark a lot more than we should.
88 		 */
89 		min_wait.tv_sec = min_wait.tv_usec = 0;
90 	}
91 	chk = TAILQ_LAST(&stcb->asoc.sent_queue, sctpchunk_listhead);
92 	for (; chk != NULL; chk = tp2) {
93 		tp2 = TAILQ_PREV(chk, sctpchunk_listhead, sctp_next);
94 		if (chk->whoTo != net) {
95 			continue;
96 		}
97 		if (chk->sent == SCTP_DATAGRAM_RESEND)
98 			cnt_resend++;
99 		else if ((chk->sent > SCTP_DATAGRAM_UNSENT) &&
100 		    (chk->sent < SCTP_DATAGRAM_RESEND)) {
101 			/* pending, may need retran */
102 			if (chk->sent_rcv_time.tv_sec > min_wait.tv_sec) {
103 				/*
104 				 * we have reached a chunk that was sent
105 				 * some seconds past our min.. forget it we
106 				 * will find no more to send.
107 				 */
108 				continue;
109 			} else if (chk->sent_rcv_time.tv_sec == min_wait.tv_sec) {
110 				/*
111 				 * we must look at the micro seconds to
112 				 * know.
113 				 */
114 				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
115 					/*
116 					 * ok it was sent after our boundary
117 					 * time.
118 					 */
119 					continue;
120 				}
121 			}
122 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_EARLYFR_LOGGING_ENABLE) {
123 				sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count,
124 				    4, SCTP_FR_MARKED_EARLY);
125 			}
126 			SCTP_STAT_INCR(sctps_earlyfrmrkretrans);
127 			chk->sent = SCTP_DATAGRAM_RESEND;
128 			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
129 			/* double book size since we are doing an early FR */
130 			chk->book_size_scale++;
131 			cnt += chk->send_size;
132 			if ((cnt + net->flight_size) > net->cwnd) {
133 				/* Mark all we could possibly resend */
134 				break;
135 			}
136 		}
137 	}
138 	if (cnt) {
139 		/*
140 		 * JRS - Use the congestion control given in the congestion
141 		 * control module
142 		 */
143 		stcb->asoc.cc_functions.sctp_cwnd_update_after_fr_timer(inp, stcb, net);
144 	} else if (cnt_resend) {
145 		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_EARLY_FR_TMR, SCTP_SO_NOT_LOCKED);
146 	}
147 	/* Restart it? */
148 	if (net->flight_size < net->cwnd) {
149 		SCTP_STAT_INCR(sctps_earlyfrstrtmr);
150 		sctp_timer_start(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net);
151 	}
152 }
153 
154 void
155 sctp_audit_retranmission_queue(struct sctp_association *asoc)
156 {
157 	struct sctp_tmit_chunk *chk;
158 
159 	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit invoked on send queue cnt:%d onqueue:%d\n",
160 	    asoc->sent_queue_retran_cnt,
161 	    asoc->sent_queue_cnt);
162 	asoc->sent_queue_retran_cnt = 0;
163 	asoc->sent_queue_cnt = 0;
164 	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
165 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
166 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
167 		}
168 		asoc->sent_queue_cnt++;
169 	}
170 	TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) {
171 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
172 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
173 		}
174 	}
175 	TAILQ_FOREACH(chk, &asoc->asconf_send_queue, sctp_next) {
176 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
177 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
178 		}
179 	}
180 	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit completes retran:%d onqueue:%d\n",
181 	    asoc->sent_queue_retran_cnt,
182 	    asoc->sent_queue_cnt);
183 }
184 
185 int
186 sctp_threshold_management(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
187     struct sctp_nets *net, uint16_t threshold)
188 {
189 	if (net) {
190 		net->error_count++;
191 		SCTPDBG(SCTP_DEBUG_TIMER4, "Error count for %p now %d thresh:%d\n",
192 		    net, net->error_count,
193 		    net->failure_threshold);
194 		if (net->error_count > net->failure_threshold) {
195 			/* We had a threshold failure */
196 			if (net->dest_state & SCTP_ADDR_REACHABLE) {
197 				net->dest_state &= ~SCTP_ADDR_REACHABLE;
198 				net->dest_state |= SCTP_ADDR_NOT_REACHABLE;
199 				net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY;
200 				if (net == stcb->asoc.primary_destination) {
201 					net->dest_state |= SCTP_ADDR_WAS_PRIMARY;
202 				}
203 				/*
204 				 * JRS 5/14/07 - If a destination is
205 				 * unreachable, the PF bit is turned off.
206 				 * This allows an unambiguous use of the PF
207 				 * bit for destinations that are reachable
208 				 * but potentially failed. If the
209 				 * destination is set to the unreachable
210 				 * state, also set the destination to the PF
211 				 * state.
212 				 */
213 				/*
214 				 * Add debug message here if destination is
215 				 * not in PF state.
216 				 */
217 				/* Stop any running T3 timers here? */
218 				if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && SCTP_BASE_SYSCTL(sctp_cmt_pf)) {
219 					net->dest_state &= ~SCTP_ADDR_PF;
220 					SCTPDBG(SCTP_DEBUG_TIMER4, "Destination %p moved from PF to unreachable.\n",
221 					    net);
222 				}
223 				sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN,
224 				    stcb,
225 				    SCTP_FAILED_THRESHOLD,
226 				    (void *)net, SCTP_SO_NOT_LOCKED);
227 			}
228 		}
229 		/*********HOLD THIS COMMENT FOR PATCH OF ALTERNATE
230 		 *********ROUTING CODE
231 		 */
232 		/*********HOLD THIS COMMENT FOR END OF PATCH OF ALTERNATE
233 		 *********ROUTING CODE
234 		 */
235 	}
236 	if (stcb == NULL)
237 		return (0);
238 
239 	if (net) {
240 		if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) {
241 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
242 				sctp_misc_ints(SCTP_THRESHOLD_INCR,
243 				    stcb->asoc.overall_error_count,
244 				    (stcb->asoc.overall_error_count + 1),
245 				    SCTP_FROM_SCTP_TIMER,
246 				    __LINE__);
247 			}
248 			stcb->asoc.overall_error_count++;
249 		}
250 	} else {
251 		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
252 			sctp_misc_ints(SCTP_THRESHOLD_INCR,
253 			    stcb->asoc.overall_error_count,
254 			    (stcb->asoc.overall_error_count + 1),
255 			    SCTP_FROM_SCTP_TIMER,
256 			    __LINE__);
257 		}
258 		stcb->asoc.overall_error_count++;
259 	}
260 	SCTPDBG(SCTP_DEBUG_TIMER4, "Overall error count for %p now %d thresh:%u state:%x\n",
261 	    &stcb->asoc, stcb->asoc.overall_error_count,
262 	    (uint32_t) threshold,
263 	    ((net == NULL) ? (uint32_t) 0 : (uint32_t) net->dest_state));
264 	/*
265 	 * We specifically do not do >= to give the assoc one more change
266 	 * before we fail it.
267 	 */
268 	if (stcb->asoc.overall_error_count > threshold) {
269 		/* Abort notification sends a ULP notify */
270 		struct mbuf *oper;
271 
272 		oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
273 		    0, M_DONTWAIT, 1, MT_DATA);
274 		if (oper) {
275 			struct sctp_paramhdr *ph;
276 			uint32_t *ippp;
277 
278 			SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
279 			    sizeof(uint32_t);
280 			ph = mtod(oper, struct sctp_paramhdr *);
281 			ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
282 			ph->param_length = htons(SCTP_BUF_LEN(oper));
283 			ippp = (uint32_t *) (ph + 1);
284 			*ippp = htonl(SCTP_FROM_SCTP_TIMER + SCTP_LOC_1);
285 		}
286 		inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_1;
287 		sctp_abort_an_association(inp, stcb, SCTP_FAILED_THRESHOLD, oper, SCTP_SO_NOT_LOCKED);
288 		return (1);
289 	}
290 	return (0);
291 }
292 
293 struct sctp_nets *
294 sctp_find_alternate_net(struct sctp_tcb *stcb,
295     struct sctp_nets *net,
296     int mode)
297 {
298 	/* Find and return an alternate network if possible */
299 	struct sctp_nets *alt, *mnet, *min_errors_net = NULL, *max_cwnd_net = NULL;
300 	int once;
301 
302 	/* JRS 5/14/07 - Initialize min_errors to an impossible value. */
303 	int min_errors = -1;
304 	uint32_t max_cwnd = 0;
305 
306 	if (stcb->asoc.numnets == 1) {
307 		/* No others but net */
308 		return (TAILQ_FIRST(&stcb->asoc.nets));
309 	}
310 	/*
311 	 * JRS 5/14/07 - If mode is set to 2, use the CMT PF find alternate
312 	 * net algorithm. This algorithm chooses the active destination (not
313 	 * in PF state) with the largest cwnd value. If all destinations are
314 	 * in PF state, unreachable, or unconfirmed, choose the desination
315 	 * that is in PF state with the lowest error count. In case of a
316 	 * tie, choose the destination that was most recently active.
317 	 */
318 	if (mode == 2) {
319 		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
320 			/*
321 			 * JRS 5/14/07 - If the destination is unreachable
322 			 * or unconfirmed, skip it.
323 			 */
324 			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
325 			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
326 				continue;
327 			}
328 			/*
329 			 * JRS 5/14/07 -  If the destination is reachable
330 			 * but in PF state, compare the error count of the
331 			 * destination to the minimum error count seen thus
332 			 * far. Store the destination with the lower error
333 			 * count.  If the error counts are equal, store the
334 			 * destination that was most recently active.
335 			 */
336 			if (mnet->dest_state & SCTP_ADDR_PF) {
337 				/*
338 				 * JRS 5/14/07 - If the destination under
339 				 * consideration is the current destination,
340 				 * work as if the error count is one higher.
341 				 * The actual error count will not be
342 				 * incremented until later in the t3
343 				 * handler.
344 				 */
345 				if (mnet == net) {
346 					if (min_errors == -1) {
347 						min_errors = mnet->error_count + 1;
348 						min_errors_net = mnet;
349 					} else if (mnet->error_count + 1 < min_errors) {
350 						min_errors = mnet->error_count + 1;
351 						min_errors_net = mnet;
352 					} else if (mnet->error_count + 1 == min_errors
353 					    && mnet->last_active > min_errors_net->last_active) {
354 						min_errors_net = mnet;
355 						min_errors = mnet->error_count + 1;
356 					}
357 					continue;
358 				} else {
359 					if (min_errors == -1) {
360 						min_errors = mnet->error_count;
361 						min_errors_net = mnet;
362 					} else if (mnet->error_count < min_errors) {
363 						min_errors = mnet->error_count;
364 						min_errors_net = mnet;
365 					} else if (mnet->error_count == min_errors
366 					    && mnet->last_active > min_errors_net->last_active) {
367 						min_errors_net = mnet;
368 						min_errors = mnet->error_count;
369 					}
370 					continue;
371 				}
372 			}
373 			/*
374 			 * JRS 5/14/07 - If the destination is reachable and
375 			 * not in PF state, compare the cwnd of the
376 			 * destination to the highest cwnd seen thus far.
377 			 * Store the destination with the higher cwnd value.
378 			 * If the cwnd values are equal, randomly choose one
379 			 * of the two destinations.
380 			 */
381 			if (max_cwnd < mnet->cwnd) {
382 				max_cwnd_net = mnet;
383 				max_cwnd = mnet->cwnd;
384 			} else if (max_cwnd == mnet->cwnd) {
385 				uint32_t rndval;
386 				uint8_t this_random;
387 
388 				if (stcb->asoc.hb_random_idx > 3) {
389 					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
390 					memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values));
391 					this_random = stcb->asoc.hb_random_values[0];
392 					stcb->asoc.hb_random_idx++;
393 					stcb->asoc.hb_ect_randombit = 0;
394 				} else {
395 					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
396 					stcb->asoc.hb_random_idx++;
397 					stcb->asoc.hb_ect_randombit = 0;
398 				}
399 				if (this_random % 2 == 1) {
400 					max_cwnd_net = mnet;
401 					max_cwnd = mnet->cwnd;	/* Useless? */
402 				}
403 			}
404 		}
405 		/*
406 		 * JRS 5/14/07 - After all destination have been considered
407 		 * as alternates, check to see if there was some active
408 		 * destination (not in PF state).  If not, check to see if
409 		 * there was some PF destination with the minimum number of
410 		 * errors.  If not, return the original destination.  If
411 		 * there is a min_errors_net, remove the PF flag from that
412 		 * destination, set the cwnd to one or two MTUs, and return
413 		 * the destination as an alt. If there was some active
414 		 * destination with a highest cwnd, return the destination
415 		 * as an alt.
416 		 */
417 		if (max_cwnd_net == NULL) {
418 			if (min_errors_net == NULL) {
419 				return (net);
420 			}
421 			min_errors_net->dest_state &= ~SCTP_ADDR_PF;
422 			min_errors_net->cwnd = min_errors_net->mtu * SCTP_BASE_SYSCTL(sctp_cmt_pf);
423 			if (SCTP_OS_TIMER_PENDING(&min_errors_net->rxt_timer.timer)) {
424 				sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep,
425 				    stcb, min_errors_net,
426 				    SCTP_FROM_SCTP_TIMER + SCTP_LOC_2);
427 			}
428 			SCTPDBG(SCTP_DEBUG_TIMER4, "Destination %p moved from PF to active with %d errors.\n",
429 			    min_errors_net, min_errors_net->error_count);
430 			return (min_errors_net);
431 		} else {
432 			return (max_cwnd_net);
433 		}
434 	}
435 	/*
436 	 * JRS 5/14/07 - If mode is set to 1, use the CMT policy for
437 	 * choosing an alternate net.
438 	 */
439 	else if (mode == 1) {
440 		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
441 			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
442 			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)
443 			    ) {
444 				/*
445 				 * will skip ones that are not-reachable or
446 				 * unconfirmed
447 				 */
448 				continue;
449 			}
450 			if (max_cwnd < mnet->cwnd) {
451 				max_cwnd_net = mnet;
452 				max_cwnd = mnet->cwnd;
453 			} else if (max_cwnd == mnet->cwnd) {
454 				uint32_t rndval;
455 				uint8_t this_random;
456 
457 				if (stcb->asoc.hb_random_idx > 3) {
458 					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
459 					memcpy(stcb->asoc.hb_random_values, &rndval,
460 					    sizeof(stcb->asoc.hb_random_values));
461 					this_random = stcb->asoc.hb_random_values[0];
462 					stcb->asoc.hb_random_idx = 0;
463 					stcb->asoc.hb_ect_randombit = 0;
464 				} else {
465 					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
466 					stcb->asoc.hb_random_idx++;
467 					stcb->asoc.hb_ect_randombit = 0;
468 				}
469 				if (this_random % 2) {
470 					max_cwnd_net = mnet;
471 					max_cwnd = mnet->cwnd;
472 				}
473 			}
474 		}
475 		if (max_cwnd_net) {
476 			return (max_cwnd_net);
477 		}
478 	}
479 	mnet = net;
480 	once = 0;
481 
482 	if (mnet == NULL) {
483 		mnet = TAILQ_FIRST(&stcb->asoc.nets);
484 	}
485 	do {
486 		alt = TAILQ_NEXT(mnet, sctp_next);
487 		if (alt == NULL) {
488 			once++;
489 			if (once > 1) {
490 				break;
491 			}
492 			alt = TAILQ_FIRST(&stcb->asoc.nets);
493 		}
494 		if (alt->ro.ro_rt == NULL) {
495 			if (alt->ro._s_addr) {
496 				sctp_free_ifa(alt->ro._s_addr);
497 				alt->ro._s_addr = NULL;
498 			}
499 			alt->src_addr_selected = 0;
500 		}
501 		if (
502 		    ((alt->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE) &&
503 		    (alt->ro.ro_rt != NULL) &&
504 		/* sa_ignore NO_NULL_CHK */
505 		    (!(alt->dest_state & SCTP_ADDR_UNCONFIRMED))
506 		    ) {
507 			/* Found a reachable address */
508 			break;
509 		}
510 		mnet = alt;
511 	} while (alt != NULL);
512 
513 	if (alt == NULL) {
514 		/* Case where NO insv network exists (dormant state) */
515 		/* we rotate destinations */
516 		once = 0;
517 		mnet = net;
518 		do {
519 			alt = TAILQ_NEXT(mnet, sctp_next);
520 			if (alt == NULL) {
521 				once++;
522 				if (once > 1) {
523 					break;
524 				}
525 				alt = TAILQ_FIRST(&stcb->asoc.nets);
526 			}
527 			/* sa_ignore NO_NULL_CHK */
528 			if ((!(alt->dest_state & SCTP_ADDR_UNCONFIRMED)) &&
529 			    (alt != net)) {
530 				/* Found an alternate address */
531 				break;
532 			}
533 			mnet = alt;
534 		} while (alt != NULL);
535 	}
536 	if (alt == NULL) {
537 		return (net);
538 	}
539 	return (alt);
540 }
541 
542 
543 
544 static void
545 sctp_backoff_on_timeout(struct sctp_tcb *stcb,
546     struct sctp_nets *net,
547     int win_probe,
548     int num_marked)
549 {
550 	if (net->RTO == 0) {
551 		net->RTO = stcb->asoc.minrto;
552 	}
553 	net->RTO <<= 1;
554 	if (net->RTO > stcb->asoc.maxrto) {
555 		net->RTO = stcb->asoc.maxrto;
556 	}
557 	if ((win_probe == 0) && num_marked) {
558 		/* We don't apply penalty to window probe scenarios */
559 		/* JRS - Use the congestion control given in the CC module */
560 		stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout(stcb, net);
561 	}
562 }
563 
564 static int
565 sctp_mark_all_for_resend(struct sctp_tcb *stcb,
566     struct sctp_nets *net,
567     struct sctp_nets *alt,
568     int window_probe,
569     int *num_marked)
570 {
571 
572 	/*
573 	 * Mark all chunks (well not all) that were sent to *net for
574 	 * retransmission. Move them to alt for there destination as well...
575 	 * We only mark chunks that have been outstanding long enough to
576 	 * have received feed-back.
577 	 */
578 	struct sctp_tmit_chunk *chk, *tp2, *could_be_sent = NULL;
579 	struct sctp_nets *lnets;
580 	struct timeval now, min_wait, tv;
581 	int cur_rtt;
582 	int audit_tf, num_mk, fir;
583 	unsigned int cnt_mk;
584 	uint32_t orig_flight, orig_tf;
585 	uint32_t tsnlast, tsnfirst;
586 
587 
588 	/* none in flight now */
589 	audit_tf = 0;
590 	fir = 0;
591 	/*
592 	 * figure out how long a data chunk must be pending before we can
593 	 * mark it ..
594 	 */
595 	(void)SCTP_GETTIME_TIMEVAL(&now);
596 	/* get cur rto in micro-seconds */
597 	cur_rtt = (((net->lastsa >> 2) + net->lastsv) >> 1);
598 	cur_rtt *= 1000;
599 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
600 		sctp_log_fr(cur_rtt,
601 		    stcb->asoc.peers_rwnd,
602 		    window_probe,
603 		    SCTP_FR_T3_MARK_TIME);
604 		sctp_log_fr(net->flight_size,
605 		    SCTP_OS_TIMER_PENDING(&net->fr_timer.timer),
606 		    SCTP_OS_TIMER_ACTIVE(&net->fr_timer.timer),
607 		    SCTP_FR_CWND_REPORT);
608 		sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT);
609 	}
610 	tv.tv_sec = cur_rtt / 1000000;
611 	tv.tv_usec = cur_rtt % 1000000;
612 	min_wait = now;
613 	timevalsub(&min_wait, &tv);
614 	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
615 		/*
616 		 * if we hit here, we don't have enough seconds on the clock
617 		 * to account for the RTO. We just let the lower seconds be
618 		 * the bounds and don't worry about it. This may mean we
619 		 * will mark a lot more than we should.
620 		 */
621 		min_wait.tv_sec = min_wait.tv_usec = 0;
622 	}
623 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
624 		sctp_log_fr(cur_rtt, now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME);
625 		sctp_log_fr(0, min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME);
626 	}
627 	/*
628 	 * Our rwnd will be incorrect here since we are not adding back the
629 	 * cnt * mbuf but we will fix that down below.
630 	 */
631 	orig_flight = net->flight_size;
632 	orig_tf = stcb->asoc.total_flight;
633 
634 	net->fast_retran_ip = 0;
635 	/* Now on to each chunk */
636 	num_mk = cnt_mk = 0;
637 	tsnfirst = tsnlast = 0;
638 	chk = TAILQ_FIRST(&stcb->asoc.sent_queue);
639 	for (; chk != NULL; chk = tp2) {
640 		tp2 = TAILQ_NEXT(chk, sctp_next);
641 		if ((compare_with_wrap(stcb->asoc.last_acked_seq,
642 		    chk->rec.data.TSN_seq,
643 		    MAX_TSN)) ||
644 		    (stcb->asoc.last_acked_seq == chk->rec.data.TSN_seq)) {
645 			/* Strange case our list got out of order? */
646 			SCTP_PRINTF("Our list is out of order?\n");
647 			panic("Out of order list");
648 		}
649 		if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) {
650 			/*
651 			 * found one to mark: If it is less than
652 			 * DATAGRAM_ACKED it MUST not be a skipped or marked
653 			 * TSN but instead one that is either already set
654 			 * for retransmission OR one that needs
655 			 * retransmission.
656 			 */
657 
658 			/* validate its been outstanding long enough */
659 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
660 				sctp_log_fr(chk->rec.data.TSN_seq,
661 				    chk->sent_rcv_time.tv_sec,
662 				    chk->sent_rcv_time.tv_usec,
663 				    SCTP_FR_T3_MARK_TIME);
664 			}
665 			if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) {
666 				/*
667 				 * we have reached a chunk that was sent
668 				 * some seconds past our min.. forget it we
669 				 * will find no more to send.
670 				 */
671 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
672 					sctp_log_fr(0,
673 					    chk->sent_rcv_time.tv_sec,
674 					    chk->sent_rcv_time.tv_usec,
675 					    SCTP_FR_T3_STOPPED);
676 				}
677 				continue;
678 			} else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) &&
679 			    (window_probe == 0)) {
680 				/*
681 				 * we must look at the micro seconds to
682 				 * know.
683 				 */
684 				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
685 					/*
686 					 * ok it was sent after our boundary
687 					 * time.
688 					 */
689 					if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
690 						sctp_log_fr(0,
691 						    chk->sent_rcv_time.tv_sec,
692 						    chk->sent_rcv_time.tv_usec,
693 						    SCTP_FR_T3_STOPPED);
694 					}
695 					continue;
696 				}
697 			}
698 			if (PR_SCTP_TTL_ENABLED(chk->flags)) {
699 				/* Is it expired? */
700 				if ((now.tv_sec > chk->rec.data.timetodrop.tv_sec) ||
701 				    ((chk->rec.data.timetodrop.tv_sec == now.tv_sec) &&
702 				    (now.tv_usec > chk->rec.data.timetodrop.tv_usec))) {
703 					/* Yes so drop it */
704 					if (chk->data) {
705 						(void)sctp_release_pr_sctp_chunk(stcb,
706 						    chk,
707 						    (SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_SENT),
708 						    &stcb->asoc.sent_queue, SCTP_SO_NOT_LOCKED);
709 					}
710 				}
711 				continue;
712 			}
713 			if (PR_SCTP_RTX_ENABLED(chk->flags)) {
714 				/* Has it been retransmitted tv_sec times? */
715 				if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) {
716 					if (chk->data) {
717 						(void)sctp_release_pr_sctp_chunk(stcb,
718 						    chk,
719 						    (SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_SENT),
720 						    &stcb->asoc.sent_queue, SCTP_SO_NOT_LOCKED);
721 					}
722 				}
723 				continue;
724 			}
725 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
726 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
727 				num_mk++;
728 				if (fir == 0) {
729 					fir = 1;
730 					tsnfirst = chk->rec.data.TSN_seq;
731 				}
732 				tsnlast = chk->rec.data.TSN_seq;
733 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
734 					sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count,
735 					    0, SCTP_FR_T3_MARKED);
736 				}
737 				if (chk->rec.data.chunk_was_revoked) {
738 					/* deflate the cwnd */
739 					chk->whoTo->cwnd -= chk->book_size;
740 					chk->rec.data.chunk_was_revoked = 0;
741 				}
742 				net->marked_retrans++;
743 				stcb->asoc.marked_retrans++;
744 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
745 					sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO,
746 					    chk->whoTo->flight_size,
747 					    chk->book_size,
748 					    (uintptr_t) chk->whoTo,
749 					    chk->rec.data.TSN_seq);
750 				}
751 				sctp_flight_size_decrease(chk);
752 				sctp_total_flight_decrease(stcb, chk);
753 				stcb->asoc.peers_rwnd += chk->send_size;
754 				stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh);
755 			}
756 			chk->sent = SCTP_DATAGRAM_RESEND;
757 			SCTP_STAT_INCR(sctps_markedretrans);
758 
759 			/* reset the TSN for striking and other FR stuff */
760 			chk->rec.data.doing_fast_retransmit = 0;
761 			/* Clear any time so NO RTT is being done */
762 			chk->do_rtt = 0;
763 			if (alt != net) {
764 				sctp_free_remote_addr(chk->whoTo);
765 				chk->no_fr_allowed = 1;
766 				chk->whoTo = alt;
767 				atomic_add_int(&alt->ref_count, 1);
768 			} else {
769 				chk->no_fr_allowed = 0;
770 				if (TAILQ_EMPTY(&stcb->asoc.send_queue)) {
771 					chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq;
772 				} else {
773 					chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.TSN_seq;
774 				}
775 			}
776 			/*
777 			 * CMT: Do not allow FRs on retransmitted TSNs.
778 			 */
779 			if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 1) {
780 				chk->no_fr_allowed = 1;
781 			}
782 		} else if (chk->sent == SCTP_DATAGRAM_ACKED) {
783 			/* remember highest acked one */
784 			could_be_sent = chk;
785 		}
786 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
787 			cnt_mk++;
788 		}
789 	}
790 	if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) {
791 		/* we did not subtract the same things? */
792 		audit_tf = 1;
793 	}
794 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
795 		sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT);
796 	}
797 #ifdef SCTP_DEBUG
798 	if (num_mk) {
799 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
800 		    tsnlast);
801 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%ld\n",
802 		    num_mk, (u_long)stcb->asoc.peers_rwnd);
803 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
804 		    tsnlast);
805 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%d\n",
806 		    num_mk,
807 		    (int)stcb->asoc.peers_rwnd);
808 	}
809 #endif
810 	*num_marked = num_mk;
811 	if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) {
812 		/* fix it so we retransmit the highest acked anyway */
813 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
814 		cnt_mk++;
815 		could_be_sent->sent = SCTP_DATAGRAM_RESEND;
816 	}
817 	if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) {
818 #ifdef INVARIANTS
819 		SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n",
820 		    cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk);
821 #endif
822 #ifndef SCTP_AUDITING_ENABLED
823 		stcb->asoc.sent_queue_retran_cnt = cnt_mk;
824 #endif
825 	}
826 	/* Now check for a ECN Echo that may be stranded */
827 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
828 		if ((chk->whoTo == net) &&
829 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
830 			sctp_free_remote_addr(chk->whoTo);
831 			chk->whoTo = alt;
832 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
833 				chk->sent = SCTP_DATAGRAM_RESEND;
834 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
835 			}
836 			atomic_add_int(&alt->ref_count, 1);
837 		}
838 	}
839 	if (audit_tf) {
840 		SCTPDBG(SCTP_DEBUG_TIMER4,
841 		    "Audit total flight due to negative value net:%p\n",
842 		    net);
843 		stcb->asoc.total_flight = 0;
844 		stcb->asoc.total_flight_count = 0;
845 		/* Clear all networks flight size */
846 		TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) {
847 			lnets->flight_size = 0;
848 			SCTPDBG(SCTP_DEBUG_TIMER4,
849 			    "Net:%p c-f cwnd:%d ssthresh:%d\n",
850 			    lnets, lnets->cwnd, lnets->ssthresh);
851 		}
852 		TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
853 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
854 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
855 					sctp_misc_ints(SCTP_FLIGHT_LOG_UP,
856 					    chk->whoTo->flight_size,
857 					    chk->book_size,
858 					    (uintptr_t) chk->whoTo,
859 					    chk->rec.data.TSN_seq);
860 				}
861 				sctp_flight_size_increase(chk);
862 				sctp_total_flight_increase(stcb, chk);
863 			}
864 		}
865 	}
866 	/*
867 	 * Setup the ecn nonce re-sync point. We do this since
868 	 * retranmissions are NOT setup for ECN. This means that do to
869 	 * Karn's rule, we don't know the total of the peers ecn bits.
870 	 */
871 	chk = TAILQ_FIRST(&stcb->asoc.send_queue);
872 	if (chk == NULL) {
873 		stcb->asoc.nonce_resync_tsn = stcb->asoc.sending_seq;
874 	} else {
875 		stcb->asoc.nonce_resync_tsn = chk->rec.data.TSN_seq;
876 	}
877 	stcb->asoc.nonce_wait_for_ecne = 0;
878 	stcb->asoc.nonce_sum_check = 0;
879 	/* We return 1 if we only have a window probe outstanding */
880 	return (0);
881 }
882 
883 static void
884 sctp_move_all_chunks_to_alt(struct sctp_tcb *stcb,
885     struct sctp_nets *net,
886     struct sctp_nets *alt)
887 {
888 	struct sctp_association *asoc;
889 	struct sctp_stream_out *outs;
890 	struct sctp_tmit_chunk *chk;
891 	struct sctp_stream_queue_pending *sp;
892 
893 	if (net == alt)
894 		/* nothing to do */
895 		return;
896 
897 	asoc = &stcb->asoc;
898 
899 	/*
900 	 * now through all the streams checking for chunks sent to our bad
901 	 * network.
902 	 */
903 	TAILQ_FOREACH(outs, &asoc->out_wheel, next_spoke) {
904 		/* now clean up any chunks here */
905 		TAILQ_FOREACH(sp, &outs->outqueue, next) {
906 			if (sp->net == net) {
907 				sctp_free_remote_addr(sp->net);
908 				sp->net = alt;
909 				atomic_add_int(&alt->ref_count, 1);
910 			}
911 		}
912 	}
913 	/* Now check the pending queue */
914 	TAILQ_FOREACH(chk, &asoc->send_queue, sctp_next) {
915 		if (chk->whoTo == net) {
916 			sctp_free_remote_addr(chk->whoTo);
917 			chk->whoTo = alt;
918 			atomic_add_int(&alt->ref_count, 1);
919 		}
920 	}
921 
922 }
923 
924 int
925 sctp_t3rxt_timer(struct sctp_inpcb *inp,
926     struct sctp_tcb *stcb,
927     struct sctp_nets *net)
928 {
929 	struct sctp_nets *alt;
930 	int win_probe, num_mk;
931 
932 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
933 		sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT);
934 	}
935 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
936 		struct sctp_nets *lnet;
937 
938 		TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
939 			if (net == lnet) {
940 				sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3);
941 			} else {
942 				sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3);
943 			}
944 		}
945 	}
946 	/* Find an alternate and mark those for retransmission */
947 	if ((stcb->asoc.peers_rwnd == 0) &&
948 	    (stcb->asoc.total_flight < net->mtu)) {
949 		SCTP_STAT_INCR(sctps_timowindowprobe);
950 		win_probe = 1;
951 	} else {
952 		win_probe = 0;
953 	}
954 
955 	/*
956 	 * JRS 5/14/07 - If CMT PF is on and the destination if not already
957 	 * in PF state, set the destination to PF state and store the
958 	 * current time as the time that the destination was last active. In
959 	 * addition, find an alternate destination with PF-based
960 	 * find_alt_net().
961 	 */
962 	if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && SCTP_BASE_SYSCTL(sctp_cmt_pf)) {
963 		if ((net->dest_state & SCTP_ADDR_PF) != SCTP_ADDR_PF) {
964 			net->dest_state |= SCTP_ADDR_PF;
965 			net->last_active = sctp_get_tick_count();
966 			SCTPDBG(SCTP_DEBUG_TIMER4, "Destination %p moved from active to PF.\n",
967 			    net);
968 		}
969 		alt = sctp_find_alternate_net(stcb, net, 2);
970 	} else if (SCTP_BASE_SYSCTL(sctp_cmt_on_off)) {
971 		/*
972 		 * CMT: Using RTX_SSTHRESH policy for CMT. If CMT is being
973 		 * used, then pick dest with largest ssthresh for any
974 		 * retransmission.
975 		 */
976 		alt = net;
977 		alt = sctp_find_alternate_net(stcb, alt, 1);
978 		/*
979 		 * CUCv2: If a different dest is picked for the
980 		 * retransmission, then new (rtx-)pseudo_cumack needs to be
981 		 * tracked for orig dest. Let CUCv2 track new (rtx-)
982 		 * pseudo-cumack always.
983 		 */
984 		net->find_pseudo_cumack = 1;
985 		net->find_rtx_pseudo_cumack = 1;
986 	} else {		/* CMT is OFF */
987 		alt = sctp_find_alternate_net(stcb, net, 0);
988 	}
989 
990 	(void)sctp_mark_all_for_resend(stcb, net, alt, win_probe, &num_mk);
991 	/* FR Loss recovery just ended with the T3. */
992 	stcb->asoc.fast_retran_loss_recovery = 0;
993 
994 	/* CMT FR loss recovery ended with the T3 */
995 	net->fast_retran_loss_recovery = 0;
996 
997 	/*
998 	 * setup the sat loss recovery that prevents satellite cwnd advance.
999 	 */
1000 	stcb->asoc.sat_t3_loss_recovery = 1;
1001 	stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq;
1002 
1003 	/* Backoff the timer and cwnd */
1004 	sctp_backoff_on_timeout(stcb, net, win_probe, num_mk);
1005 	if (win_probe == 0) {
1006 		/* We don't do normal threshold management on window probes */
1007 		if (sctp_threshold_management(inp, stcb, net,
1008 		    stcb->asoc.max_send_times)) {
1009 			/* Association was destroyed */
1010 			return (1);
1011 		} else {
1012 			if (net != stcb->asoc.primary_destination) {
1013 				/* send a immediate HB if our RTO is stale */
1014 				struct timeval now;
1015 				unsigned int ms_goneby;
1016 
1017 				(void)SCTP_GETTIME_TIMEVAL(&now);
1018 				if (net->last_sent_time.tv_sec) {
1019 					ms_goneby = (now.tv_sec - net->last_sent_time.tv_sec) * 1000;
1020 				} else {
1021 					ms_goneby = 0;
1022 				}
1023 				if ((ms_goneby > net->RTO) || (net->RTO == 0)) {
1024 					/*
1025 					 * no recent feed back in an RTO or
1026 					 * more, request a RTT update
1027 					 */
1028 					if (sctp_send_hb(stcb, 1, net) < 0)
1029 						return 1;
1030 				}
1031 			}
1032 		}
1033 	} else {
1034 		/*
1035 		 * For a window probe we don't penalize the net's but only
1036 		 * the association. This may fail it if SACKs are not coming
1037 		 * back. If sack's are coming with rwnd locked at 0, we will
1038 		 * continue to hold things waiting for rwnd to raise
1039 		 */
1040 		if (sctp_threshold_management(inp, stcb, NULL,
1041 		    stcb->asoc.max_send_times)) {
1042 			/* Association was destroyed */
1043 			return (1);
1044 		}
1045 	}
1046 	if (net->dest_state & SCTP_ADDR_NOT_REACHABLE) {
1047 		/* Move all pending over too */
1048 		sctp_move_all_chunks_to_alt(stcb, net, alt);
1049 
1050 		/*
1051 		 * Get the address that failed, to force a new src address
1052 		 * selecton and a route allocation.
1053 		 */
1054 		if (net->ro._s_addr) {
1055 			sctp_free_ifa(net->ro._s_addr);
1056 			net->ro._s_addr = NULL;
1057 		}
1058 		net->src_addr_selected = 0;
1059 
1060 		/* Force a route allocation too */
1061 		if (net->ro.ro_rt) {
1062 			RTFREE(net->ro.ro_rt);
1063 			net->ro.ro_rt = NULL;
1064 		}
1065 		/* Was it our primary? */
1066 		if ((stcb->asoc.primary_destination == net) && (alt != net)) {
1067 			/*
1068 			 * Yes, note it as such and find an alternate note:
1069 			 * this means HB code must use this to resent the
1070 			 * primary if it goes active AND if someone does a
1071 			 * change-primary then this flag must be cleared
1072 			 * from any net structures.
1073 			 */
1074 			if (sctp_set_primary_addr(stcb,
1075 			    (struct sockaddr *)NULL,
1076 			    alt) == 0) {
1077 				net->dest_state |= SCTP_ADDR_WAS_PRIMARY;
1078 			}
1079 		}
1080 	} else if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && SCTP_BASE_SYSCTL(sctp_cmt_pf) && (net->dest_state & SCTP_ADDR_PF) == SCTP_ADDR_PF) {
1081 		/*
1082 		 * JRS 5/14/07 - If the destination hasn't failed completely
1083 		 * but is in PF state, a PF-heartbeat needs to be sent
1084 		 * manually.
1085 		 */
1086 		if (sctp_send_hb(stcb, 1, net) < 0)
1087 			return 1;
1088 	}
1089 	/*
1090 	 * Special case for cookie-echo'ed case, we don't do output but must
1091 	 * await the COOKIE-ACK before retransmission
1092 	 */
1093 	if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
1094 		/*
1095 		 * Here we just reset the timer and start again since we
1096 		 * have not established the asoc
1097 		 */
1098 		sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net);
1099 		return (0);
1100 	}
1101 	if (stcb->asoc.peer_supports_prsctp) {
1102 		struct sctp_tmit_chunk *lchk;
1103 
1104 		lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc);
1105 		/* C3. See if we need to send a Fwd-TSN */
1106 		if (compare_with_wrap(stcb->asoc.advanced_peer_ack_point,
1107 		    stcb->asoc.last_acked_seq, MAX_TSN)) {
1108 			/*
1109 			 * ISSUE with ECN, see FWD-TSN processing for notes
1110 			 * on issues that will occur when the ECN NONCE
1111 			 * stuff is put into SCTP for cross checking.
1112 			 */
1113 			send_forward_tsn(stcb, &stcb->asoc);
1114 			if (lchk) {
1115 				/* Assure a timer is up */
1116 				sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo);
1117 			}
1118 		}
1119 	}
1120 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
1121 		sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX);
1122 	}
1123 	return (0);
1124 }
1125 
1126 int
1127 sctp_t1init_timer(struct sctp_inpcb *inp,
1128     struct sctp_tcb *stcb,
1129     struct sctp_nets *net)
1130 {
1131 	/* bump the thresholds */
1132 	if (stcb->asoc.delayed_connection) {
1133 		/*
1134 		 * special hook for delayed connection. The library did NOT
1135 		 * complete the rest of its sends.
1136 		 */
1137 		stcb->asoc.delayed_connection = 0;
1138 		sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1139 		return (0);
1140 	}
1141 	if (SCTP_GET_STATE((&stcb->asoc)) != SCTP_STATE_COOKIE_WAIT) {
1142 		return (0);
1143 	}
1144 	if (sctp_threshold_management(inp, stcb, net,
1145 	    stcb->asoc.max_init_times)) {
1146 		/* Association was destroyed */
1147 		return (1);
1148 	}
1149 	stcb->asoc.dropped_special_cnt = 0;
1150 	sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0);
1151 	if (stcb->asoc.initial_init_rto_max < net->RTO) {
1152 		net->RTO = stcb->asoc.initial_init_rto_max;
1153 	}
1154 	if (stcb->asoc.numnets > 1) {
1155 		/* If we have more than one addr use it */
1156 		struct sctp_nets *alt;
1157 
1158 		alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0);
1159 		if ((alt != NULL) && (alt != stcb->asoc.primary_destination)) {
1160 			sctp_move_all_chunks_to_alt(stcb, stcb->asoc.primary_destination, alt);
1161 			stcb->asoc.primary_destination = alt;
1162 		}
1163 	}
1164 	/* Send out a new init */
1165 	sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1166 	return (0);
1167 }
1168 
1169 /*
1170  * For cookie and asconf we actually need to find and mark for resend, then
1171  * increment the resend counter (after all the threshold management stuff of
1172  * course).
1173  */
1174 int
1175 sctp_cookie_timer(struct sctp_inpcb *inp,
1176     struct sctp_tcb *stcb,
1177     struct sctp_nets *net)
1178 {
1179 	struct sctp_nets *alt;
1180 	struct sctp_tmit_chunk *cookie;
1181 
1182 	/* first before all else we must find the cookie */
1183 	TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) {
1184 		if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) {
1185 			break;
1186 		}
1187 	}
1188 	if (cookie == NULL) {
1189 		if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
1190 			/* FOOBAR! */
1191 			struct mbuf *oper;
1192 
1193 			oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
1194 			    0, M_DONTWAIT, 1, MT_DATA);
1195 			if (oper) {
1196 				struct sctp_paramhdr *ph;
1197 				uint32_t *ippp;
1198 
1199 				SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
1200 				    sizeof(uint32_t);
1201 				ph = mtod(oper, struct sctp_paramhdr *);
1202 				ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1203 				ph->param_length = htons(SCTP_BUF_LEN(oper));
1204 				ippp = (uint32_t *) (ph + 1);
1205 				*ippp = htonl(SCTP_FROM_SCTP_TIMER + SCTP_LOC_3);
1206 			}
1207 			inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_4;
1208 			sctp_abort_an_association(inp, stcb, SCTP_INTERNAL_ERROR,
1209 			    oper, SCTP_SO_NOT_LOCKED);
1210 		} else {
1211 #ifdef INVARIANTS
1212 			panic("Cookie timer expires in wrong state?");
1213 #else
1214 			SCTP_PRINTF("Strange in state %d not cookie-echoed yet c-e timer expires?\n", SCTP_GET_STATE(&stcb->asoc));
1215 			return (0);
1216 #endif
1217 		}
1218 		return (0);
1219 	}
1220 	/* Ok we found the cookie, threshold management next */
1221 	if (sctp_threshold_management(inp, stcb, cookie->whoTo,
1222 	    stcb->asoc.max_init_times)) {
1223 		/* Assoc is over */
1224 		return (1);
1225 	}
1226 	/*
1227 	 * cleared theshold management now lets backoff the address & select
1228 	 * an alternate
1229 	 */
1230 	stcb->asoc.dropped_special_cnt = 0;
1231 	sctp_backoff_on_timeout(stcb, cookie->whoTo, 1, 0);
1232 	alt = sctp_find_alternate_net(stcb, cookie->whoTo, 0);
1233 	if (alt != cookie->whoTo) {
1234 		sctp_free_remote_addr(cookie->whoTo);
1235 		cookie->whoTo = alt;
1236 		atomic_add_int(&alt->ref_count, 1);
1237 	}
1238 	/* Now mark the retran info */
1239 	if (cookie->sent != SCTP_DATAGRAM_RESEND) {
1240 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1241 	}
1242 	cookie->sent = SCTP_DATAGRAM_RESEND;
1243 	/*
1244 	 * Now call the output routine to kick out the cookie again, Note we
1245 	 * don't mark any chunks for retran so that FR will need to kick in
1246 	 * to move these (or a send timer).
1247 	 */
1248 	return (0);
1249 }
1250 
1251 int
1252 sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1253     struct sctp_nets *net)
1254 {
1255 	struct sctp_nets *alt;
1256 	struct sctp_tmit_chunk *strrst = NULL, *chk = NULL;
1257 
1258 	if (stcb->asoc.stream_reset_outstanding == 0) {
1259 		return (0);
1260 	}
1261 	/* find the existing STRRESET, we use the seq number we sent out on */
1262 	(void)sctp_find_stream_reset(stcb, stcb->asoc.str_reset_seq_out, &strrst);
1263 	if (strrst == NULL) {
1264 		return (0);
1265 	}
1266 	/* do threshold management */
1267 	if (sctp_threshold_management(inp, stcb, strrst->whoTo,
1268 	    stcb->asoc.max_send_times)) {
1269 		/* Assoc is over */
1270 		return (1);
1271 	}
1272 	/*
1273 	 * cleared theshold management now lets backoff the address & select
1274 	 * an alternate
1275 	 */
1276 	sctp_backoff_on_timeout(stcb, strrst->whoTo, 1, 0);
1277 	alt = sctp_find_alternate_net(stcb, strrst->whoTo, 0);
1278 	sctp_free_remote_addr(strrst->whoTo);
1279 	strrst->whoTo = alt;
1280 	atomic_add_int(&alt->ref_count, 1);
1281 
1282 	/* See if a ECN Echo is also stranded */
1283 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1284 		if ((chk->whoTo == net) &&
1285 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1286 			sctp_free_remote_addr(chk->whoTo);
1287 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
1288 				chk->sent = SCTP_DATAGRAM_RESEND;
1289 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1290 			}
1291 			chk->whoTo = alt;
1292 			atomic_add_int(&alt->ref_count, 1);
1293 		}
1294 	}
1295 	if (net->dest_state & SCTP_ADDR_NOT_REACHABLE) {
1296 		/*
1297 		 * If the address went un-reachable, we need to move to
1298 		 * alternates for ALL chk's in queue
1299 		 */
1300 		sctp_move_all_chunks_to_alt(stcb, net, alt);
1301 	}
1302 	/* mark the retran info */
1303 	if (strrst->sent != SCTP_DATAGRAM_RESEND)
1304 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1305 	strrst->sent = SCTP_DATAGRAM_RESEND;
1306 
1307 	/* restart the timer */
1308 	sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, strrst->whoTo);
1309 	return (0);
1310 }
1311 
1312 int
1313 sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1314     struct sctp_nets *net)
1315 {
1316 	struct sctp_nets *alt;
1317 	struct sctp_tmit_chunk *asconf, *chk, *nchk;
1318 
1319 	/* is this a first send, or a retransmission? */
1320 	if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) {
1321 		/* compose a new ASCONF chunk and send it */
1322 		sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED);
1323 	} else {
1324 		/*
1325 		 * Retransmission of the existing ASCONF is needed
1326 		 */
1327 
1328 		/* find the existing ASCONF */
1329 		asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue);
1330 		if (asconf == NULL) {
1331 			return (0);
1332 		}
1333 		/* do threshold management */
1334 		if (sctp_threshold_management(inp, stcb, asconf->whoTo,
1335 		    stcb->asoc.max_send_times)) {
1336 			/* Assoc is over */
1337 			return (1);
1338 		}
1339 		if (asconf->snd_count > stcb->asoc.max_send_times) {
1340 			/*
1341 			 * Something is rotten: our peer is not responding
1342 			 * to ASCONFs but apparently is to other chunks.
1343 			 * i.e. it is not properly handling the chunk type
1344 			 * upper bits. Mark this peer as ASCONF incapable
1345 			 * and cleanup.
1346 			 */
1347 			SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n");
1348 			sctp_asconf_cleanup(stcb, net);
1349 			return (0);
1350 		}
1351 		/*
1352 		 * cleared threshold management, so now backoff the net and
1353 		 * select an alternate
1354 		 */
1355 		sctp_backoff_on_timeout(stcb, asconf->whoTo, 1, 0);
1356 		alt = sctp_find_alternate_net(stcb, asconf->whoTo, 0);
1357 		if (asconf->whoTo != alt) {
1358 			sctp_free_remote_addr(asconf->whoTo);
1359 			asconf->whoTo = alt;
1360 			atomic_add_int(&alt->ref_count, 1);
1361 		}
1362 		/* See if an ECN Echo is also stranded */
1363 		TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1364 			if ((chk->whoTo == net) &&
1365 			    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1366 				sctp_free_remote_addr(chk->whoTo);
1367 				chk->whoTo = alt;
1368 				if (chk->sent != SCTP_DATAGRAM_RESEND) {
1369 					chk->sent = SCTP_DATAGRAM_RESEND;
1370 					sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1371 				}
1372 				atomic_add_int(&alt->ref_count, 1);
1373 			}
1374 		}
1375 		for (chk = asconf; chk; chk = nchk) {
1376 			nchk = TAILQ_NEXT(chk, sctp_next);
1377 			if (chk->whoTo != alt) {
1378 				sctp_free_remote_addr(chk->whoTo);
1379 				chk->whoTo = alt;
1380 				atomic_add_int(&alt->ref_count, 1);
1381 			}
1382 			if (asconf->sent != SCTP_DATAGRAM_RESEND && chk->sent != SCTP_DATAGRAM_UNSENT)
1383 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1384 			chk->sent = SCTP_DATAGRAM_RESEND;
1385 		}
1386 		if (net->dest_state & SCTP_ADDR_NOT_REACHABLE) {
1387 			/*
1388 			 * If the address went un-reachable, we need to move
1389 			 * to the alternate for ALL chunks in queue
1390 			 */
1391 			sctp_move_all_chunks_to_alt(stcb, net, alt);
1392 			net = alt;
1393 		}
1394 		/* mark the retran info */
1395 		if (asconf->sent != SCTP_DATAGRAM_RESEND)
1396 			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1397 		asconf->sent = SCTP_DATAGRAM_RESEND;
1398 
1399 		/* send another ASCONF if any and we can do */
1400 		sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED);
1401 	}
1402 	return (0);
1403 }
1404 
1405 /* Mobility adaptation */
1406 void
1407 sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1408     struct sctp_nets *net)
1409 {
1410 	if (stcb->asoc.deleted_primary == NULL) {
1411 		SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n");
1412 		sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1413 		return;
1414 	}
1415 	SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary ");
1416 	SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa);
1417 	sctp_free_remote_addr(stcb->asoc.deleted_primary);
1418 	stcb->asoc.deleted_primary = NULL;
1419 	sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1420 	return;
1421 }
1422 
1423 /*
1424  * For the shutdown and shutdown-ack, we do not keep one around on the
1425  * control queue. This means we must generate a new one and call the general
1426  * chunk output routine, AFTER having done threshold management.
1427  */
1428 int
1429 sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1430     struct sctp_nets *net)
1431 {
1432 	struct sctp_nets *alt;
1433 
1434 	/* first threshold managment */
1435 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1436 		/* Assoc is over */
1437 		return (1);
1438 	}
1439 	/* second select an alternative */
1440 	alt = sctp_find_alternate_net(stcb, net, 0);
1441 
1442 	/* third generate a shutdown into the queue for out net */
1443 	if (alt) {
1444 		sctp_send_shutdown(stcb, alt);
1445 	} else {
1446 		/*
1447 		 * if alt is NULL, there is no dest to send to??
1448 		 */
1449 		return (0);
1450 	}
1451 	/* fourth restart timer */
1452 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt);
1453 	return (0);
1454 }
1455 
1456 int
1457 sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1458     struct sctp_nets *net)
1459 {
1460 	struct sctp_nets *alt;
1461 
1462 	/* first threshold managment */
1463 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1464 		/* Assoc is over */
1465 		return (1);
1466 	}
1467 	/* second select an alternative */
1468 	alt = sctp_find_alternate_net(stcb, net, 0);
1469 
1470 	/* third generate a shutdown into the queue for out net */
1471 	sctp_send_shutdown_ack(stcb, alt);
1472 
1473 	/* fourth restart timer */
1474 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt);
1475 	return (0);
1476 }
1477 
1478 static void
1479 sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp,
1480     struct sctp_tcb *stcb)
1481 {
1482 	struct sctp_stream_out *outs;
1483 	struct sctp_stream_queue_pending *sp;
1484 	unsigned int chks_in_queue = 0;
1485 	int being_filled = 0;
1486 
1487 	/*
1488 	 * This function is ONLY called when the send/sent queues are empty.
1489 	 */
1490 	if ((stcb == NULL) || (inp == NULL))
1491 		return;
1492 
1493 	if (stcb->asoc.sent_queue_retran_cnt) {
1494 		SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n",
1495 		    stcb->asoc.sent_queue_retran_cnt);
1496 		stcb->asoc.sent_queue_retran_cnt = 0;
1497 	}
1498 	SCTP_TCB_SEND_LOCK(stcb);
1499 	if (TAILQ_EMPTY(&stcb->asoc.out_wheel)) {
1500 		int i, cnt = 0;
1501 
1502 		/* Check to see if a spoke fell off the wheel */
1503 		for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
1504 			if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) {
1505 				sctp_insert_on_wheel(stcb, &stcb->asoc, &stcb->asoc.strmout[i], 1);
1506 				cnt++;
1507 			}
1508 		}
1509 		if (cnt) {
1510 			/* yep, we lost a spoke or two */
1511 			SCTP_PRINTF("Found an additional %d streams NOT on outwheel, corrected\n", cnt);
1512 		} else {
1513 			/* no spokes lost, */
1514 			stcb->asoc.total_output_queue_size = 0;
1515 		}
1516 		SCTP_TCB_SEND_UNLOCK(stcb);
1517 		return;
1518 	}
1519 	SCTP_TCB_SEND_UNLOCK(stcb);
1520 	/* Check to see if some data queued, if so report it */
1521 	TAILQ_FOREACH(outs, &stcb->asoc.out_wheel, next_spoke) {
1522 		if (!TAILQ_EMPTY(&outs->outqueue)) {
1523 			TAILQ_FOREACH(sp, &outs->outqueue, next) {
1524 				if (sp->msg_is_complete)
1525 					being_filled++;
1526 				chks_in_queue++;
1527 			}
1528 		}
1529 	}
1530 	if (chks_in_queue != stcb->asoc.stream_queue_cnt) {
1531 		SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n",
1532 		    stcb->asoc.stream_queue_cnt, chks_in_queue);
1533 	}
1534 	if (chks_in_queue) {
1535 		/* call the output queue function */
1536 		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1537 		if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1538 		    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1539 			/*
1540 			 * Probably should go in and make it go back through
1541 			 * and add fragments allowed
1542 			 */
1543 			if (being_filled == 0) {
1544 				SCTP_PRINTF("Still nothing moved %d chunks are stuck\n",
1545 				    chks_in_queue);
1546 			}
1547 		}
1548 	} else {
1549 		SCTP_PRINTF("Found no chunks on any queue tot:%lu\n",
1550 		    (u_long)stcb->asoc.total_output_queue_size);
1551 		stcb->asoc.total_output_queue_size = 0;
1552 	}
1553 }
1554 
1555 int
1556 sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1557     struct sctp_nets *net, int cnt_of_unconf)
1558 {
1559 	int ret;
1560 
1561 	if (net) {
1562 		if (net->hb_responded == 0) {
1563 			if (net->ro._s_addr) {
1564 				/*
1565 				 * Invalidate the src address if we did not
1566 				 * get a response last time.
1567 				 */
1568 				sctp_free_ifa(net->ro._s_addr);
1569 				net->ro._s_addr = NULL;
1570 				net->src_addr_selected = 0;
1571 			}
1572 			sctp_backoff_on_timeout(stcb, net, 1, 0);
1573 		}
1574 		/* Zero PBA, if it needs it */
1575 		if (net->partial_bytes_acked) {
1576 			net->partial_bytes_acked = 0;
1577 		}
1578 	}
1579 	if ((stcb->asoc.total_output_queue_size > 0) &&
1580 	    (TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1581 	    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1582 		sctp_audit_stream_queues_for_size(inp, stcb);
1583 	}
1584 	/* Send a new HB, this will do threshold managment, pick a new dest */
1585 	if (cnt_of_unconf == 0) {
1586 		if (sctp_send_hb(stcb, 0, NULL) < 0) {
1587 			return (1);
1588 		}
1589 	} else {
1590 		/*
1591 		 * this will send out extra hb's up to maxburst if there are
1592 		 * any unconfirmed addresses.
1593 		 */
1594 		uint32_t cnt_sent = 0;
1595 
1596 		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1597 			if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) &&
1598 			    (net->dest_state & SCTP_ADDR_REACHABLE)) {
1599 				cnt_sent++;
1600 				if (net->hb_responded == 0) {
1601 					/* Did we respond last time? */
1602 					if (net->ro._s_addr) {
1603 						sctp_free_ifa(net->ro._s_addr);
1604 						net->ro._s_addr = NULL;
1605 						net->src_addr_selected = 0;
1606 					}
1607 				}
1608 				ret = sctp_send_hb(stcb, 1, net);
1609 				if (ret < 0)
1610 					return 1;
1611 				else if (ret == 0) {
1612 					break;
1613 				}
1614 				if (cnt_sent >= SCTP_BASE_SYSCTL(sctp_hb_maxburst))
1615 					break;
1616 			}
1617 		}
1618 	}
1619 	return (0);
1620 }
1621 
1622 int
1623 sctp_is_hb_timer_running(struct sctp_tcb *stcb)
1624 {
1625 	if (SCTP_OS_TIMER_PENDING(&stcb->asoc.hb_timer.timer)) {
1626 		/* its running */
1627 		return (1);
1628 	} else {
1629 		/* nope */
1630 		return (0);
1631 	}
1632 }
1633 
1634 int
1635 sctp_is_sack_timer_running(struct sctp_tcb *stcb)
1636 {
1637 	if (SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) {
1638 		/* its running */
1639 		return (1);
1640 	} else {
1641 		/* nope */
1642 		return (0);
1643 	}
1644 }
1645 
1646 #define SCTP_NUMBER_OF_MTU_SIZES 18
1647 static uint32_t mtu_sizes[] = {
1648 	68,
1649 	296,
1650 	508,
1651 	512,
1652 	544,
1653 	576,
1654 	1006,
1655 	1492,
1656 	1500,
1657 	1536,
1658 	2002,
1659 	2048,
1660 	4352,
1661 	4464,
1662 	8166,
1663 	17914,
1664 	32000,
1665 	65535
1666 };
1667 
1668 
1669 static uint32_t
1670 sctp_getnext_mtu(struct sctp_inpcb *inp, uint32_t cur_mtu)
1671 {
1672 	/* select another MTU that is just bigger than this one */
1673 	int i;
1674 
1675 	for (i = 0; i < SCTP_NUMBER_OF_MTU_SIZES; i++) {
1676 		if (cur_mtu < mtu_sizes[i]) {
1677 			/* no max_mtu is bigger than this one */
1678 			return (mtu_sizes[i]);
1679 		}
1680 	}
1681 	/* here return the highest allowable */
1682 	return (cur_mtu);
1683 }
1684 
1685 
1686 void
1687 sctp_pathmtu_timer(struct sctp_inpcb *inp,
1688     struct sctp_tcb *stcb,
1689     struct sctp_nets *net)
1690 {
1691 	uint32_t next_mtu, mtu;
1692 
1693 	next_mtu = sctp_getnext_mtu(inp, net->mtu);
1694 
1695 	if ((next_mtu > net->mtu) && (net->port == 0)) {
1696 		if ((net->src_addr_selected == 0) ||
1697 		    (net->ro._s_addr == NULL) ||
1698 		    (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1699 			if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1700 				sctp_free_ifa(net->ro._s_addr);
1701 				net->ro._s_addr = NULL;
1702 				net->src_addr_selected = 0;
1703 			} else if (net->ro._s_addr == NULL) {
1704 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1705 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1706 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1707 
1708 					/* KAME hack: embed scopeid */
1709 					(void)sa6_embedscope(sin6, MODULE_GLOBAL(MOD_INET6, ip6_use_defzone));
1710 				}
1711 #endif
1712 
1713 				net->ro._s_addr = sctp_source_address_selection(inp,
1714 				    stcb,
1715 				    (sctp_route_t *) & net->ro,
1716 				    net, 0, stcb->asoc.vrf_id);
1717 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1718 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1719 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1720 
1721 					(void)sa6_recoverscope(sin6);
1722 				}
1723 #endif				/* INET6 */
1724 			}
1725 			if (net->ro._s_addr)
1726 				net->src_addr_selected = 1;
1727 		}
1728 		if (net->ro._s_addr) {
1729 			mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_rt);
1730 			if (mtu > next_mtu) {
1731 				net->mtu = next_mtu;
1732 			}
1733 		}
1734 	}
1735 	/* restart the timer */
1736 	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
1737 }
1738 
1739 void
1740 sctp_autoclose_timer(struct sctp_inpcb *inp,
1741     struct sctp_tcb *stcb,
1742     struct sctp_nets *net)
1743 {
1744 	struct timeval tn, *tim_touse;
1745 	struct sctp_association *asoc;
1746 	int ticks_gone_by;
1747 
1748 	(void)SCTP_GETTIME_TIMEVAL(&tn);
1749 	if (stcb->asoc.sctp_autoclose_ticks &&
1750 	    sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) {
1751 		/* Auto close is on */
1752 		asoc = &stcb->asoc;
1753 		/* pick the time to use */
1754 		if (asoc->time_last_rcvd.tv_sec >
1755 		    asoc->time_last_sent.tv_sec) {
1756 			tim_touse = &asoc->time_last_rcvd;
1757 		} else {
1758 			tim_touse = &asoc->time_last_sent;
1759 		}
1760 		/* Now has long enough transpired to autoclose? */
1761 		ticks_gone_by = SEC_TO_TICKS(tn.tv_sec - tim_touse->tv_sec);
1762 		if ((ticks_gone_by > 0) &&
1763 		    (ticks_gone_by >= (int)asoc->sctp_autoclose_ticks)) {
1764 			/*
1765 			 * autoclose time has hit, call the output routine,
1766 			 * which should do nothing just to be SURE we don't
1767 			 * have hanging data. We can then safely check the
1768 			 * queues and know that we are clear to send
1769 			 * shutdown
1770 			 */
1771 			sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED);
1772 			/* Are we clean? */
1773 			if (TAILQ_EMPTY(&asoc->send_queue) &&
1774 			    TAILQ_EMPTY(&asoc->sent_queue)) {
1775 				/*
1776 				 * there is nothing queued to send, so I'm
1777 				 * done...
1778 				 */
1779 				if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) {
1780 					/* only send SHUTDOWN 1st time thru */
1781 					sctp_send_shutdown(stcb, stcb->asoc.primary_destination);
1782 					if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
1783 					    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
1784 						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
1785 					}
1786 					SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
1787 					SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
1788 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
1789 					    stcb->sctp_ep, stcb,
1790 					    asoc->primary_destination);
1791 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
1792 					    stcb->sctp_ep, stcb,
1793 					    asoc->primary_destination);
1794 				}
1795 			}
1796 		} else {
1797 			/*
1798 			 * No auto close at this time, reset t-o to check
1799 			 * later
1800 			 */
1801 			int tmp;
1802 
1803 			/* fool the timer startup to use the time left */
1804 			tmp = asoc->sctp_autoclose_ticks;
1805 			asoc->sctp_autoclose_ticks -= ticks_gone_by;
1806 			sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb,
1807 			    net);
1808 			/* restore the real tick value */
1809 			asoc->sctp_autoclose_ticks = tmp;
1810 		}
1811 	}
1812 }
1813 
1814 void
1815 sctp_iterator_timer(struct sctp_iterator *it)
1816 {
1817 	int iteration_count = 0;
1818 	int inp_skip = 0;
1819 
1820 	/*
1821 	 * only one iterator can run at a time. This is the only way we can
1822 	 * cleanly pull ep's from underneath all the running interators when
1823 	 * a ep is freed.
1824 	 */
1825 	SCTP_ITERATOR_LOCK();
1826 	if (it->inp == NULL) {
1827 		/* iterator is complete */
1828 done_with_iterator:
1829 		SCTP_ITERATOR_UNLOCK();
1830 		SCTP_INP_INFO_WLOCK();
1831 		TAILQ_REMOVE(&SCTP_BASE_INFO(iteratorhead), it, sctp_nxt_itr);
1832 		/* stopping the callout is not needed, in theory */
1833 		SCTP_INP_INFO_WUNLOCK();
1834 		(void)SCTP_OS_TIMER_STOP(&it->tmr.timer);
1835 		if (it->function_atend != NULL) {
1836 			(*it->function_atend) (it->pointer, it->val);
1837 		}
1838 		SCTP_FREE(it, SCTP_M_ITER);
1839 		return;
1840 	}
1841 select_a_new_ep:
1842 	SCTP_INP_WLOCK(it->inp);
1843 	while (((it->pcb_flags) &&
1844 	    ((it->inp->sctp_flags & it->pcb_flags) != it->pcb_flags)) ||
1845 	    ((it->pcb_features) &&
1846 	    ((it->inp->sctp_features & it->pcb_features) != it->pcb_features))) {
1847 		/* endpoint flags or features don't match, so keep looking */
1848 		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
1849 			SCTP_INP_WUNLOCK(it->inp);
1850 			goto done_with_iterator;
1851 		}
1852 		SCTP_INP_WUNLOCK(it->inp);
1853 		it->inp = LIST_NEXT(it->inp, sctp_list);
1854 		if (it->inp == NULL) {
1855 			goto done_with_iterator;
1856 		}
1857 		SCTP_INP_WLOCK(it->inp);
1858 	}
1859 	if ((it->inp->inp_starting_point_for_iterator != NULL) &&
1860 	    (it->inp->inp_starting_point_for_iterator != it)) {
1861 		SCTP_PRINTF("Iterator collision, waiting for one at %p\n",
1862 		    it->inp);
1863 		SCTP_INP_WUNLOCK(it->inp);
1864 		goto start_timer_return;
1865 	}
1866 	/* mark the current iterator on the endpoint */
1867 	it->inp->inp_starting_point_for_iterator = it;
1868 	SCTP_INP_WUNLOCK(it->inp);
1869 	SCTP_INP_RLOCK(it->inp);
1870 	/* now go through each assoc which is in the desired state */
1871 	if (it->done_current_ep == 0) {
1872 		if (it->function_inp != NULL)
1873 			inp_skip = (*it->function_inp) (it->inp, it->pointer, it->val);
1874 		it->done_current_ep = 1;
1875 	}
1876 	if (it->stcb == NULL) {
1877 		/* run the per instance function */
1878 		it->stcb = LIST_FIRST(&it->inp->sctp_asoc_list);
1879 	}
1880 	SCTP_INP_RUNLOCK(it->inp);
1881 	if ((inp_skip) || it->stcb == NULL) {
1882 		if (it->function_inp_end != NULL) {
1883 			inp_skip = (*it->function_inp_end) (it->inp,
1884 			    it->pointer,
1885 			    it->val);
1886 		}
1887 		goto no_stcb;
1888 	}
1889 	if ((it->stcb) &&
1890 	    (it->stcb->asoc.stcb_starting_point_for_iterator == it)) {
1891 		it->stcb->asoc.stcb_starting_point_for_iterator = NULL;
1892 	}
1893 	while (it->stcb) {
1894 		SCTP_TCB_LOCK(it->stcb);
1895 		if (it->asoc_state && ((it->stcb->asoc.state & it->asoc_state) != it->asoc_state)) {
1896 			/* not in the right state... keep looking */
1897 			SCTP_TCB_UNLOCK(it->stcb);
1898 			goto next_assoc;
1899 		}
1900 		/* mark the current iterator on the assoc */
1901 		it->stcb->asoc.stcb_starting_point_for_iterator = it;
1902 		/* see if we have limited out the iterator loop */
1903 		iteration_count++;
1904 		if (iteration_count > SCTP_ITERATOR_MAX_AT_ONCE) {
1905 	start_timer_return:
1906 			/* set a timer to continue this later */
1907 			if (it->stcb)
1908 				SCTP_TCB_UNLOCK(it->stcb);
1909 			sctp_timer_start(SCTP_TIMER_TYPE_ITERATOR,
1910 			    (struct sctp_inpcb *)it, NULL, NULL);
1911 			SCTP_ITERATOR_UNLOCK();
1912 			return;
1913 		}
1914 		/* run function on this one */
1915 		(*it->function_assoc) (it->inp, it->stcb, it->pointer, it->val);
1916 
1917 		/*
1918 		 * we lie here, it really needs to have its own type but
1919 		 * first I must verify that this won't effect things :-0
1920 		 */
1921 		if (it->no_chunk_output == 0)
1922 			sctp_chunk_output(it->inp, it->stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1923 
1924 		SCTP_TCB_UNLOCK(it->stcb);
1925 next_assoc:
1926 		it->stcb = LIST_NEXT(it->stcb, sctp_tcblist);
1927 		if (it->stcb == NULL) {
1928 			if (it->function_inp_end != NULL) {
1929 				inp_skip = (*it->function_inp_end) (it->inp,
1930 				    it->pointer,
1931 				    it->val);
1932 			}
1933 		}
1934 	}
1935 no_stcb:
1936 	/* done with all assocs on this endpoint, move on to next endpoint */
1937 	it->done_current_ep = 0;
1938 	SCTP_INP_WLOCK(it->inp);
1939 	it->inp->inp_starting_point_for_iterator = NULL;
1940 	SCTP_INP_WUNLOCK(it->inp);
1941 	if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
1942 		it->inp = NULL;
1943 	} else {
1944 		SCTP_INP_INFO_RLOCK();
1945 		it->inp = LIST_NEXT(it->inp, sctp_list);
1946 		SCTP_INP_INFO_RUNLOCK();
1947 	}
1948 	if (it->inp == NULL) {
1949 		goto done_with_iterator;
1950 	}
1951 	goto select_a_new_ep;
1952 }
1953