xref: /freebsd/sys/netinet/sctp_timer.c (revision 1e413cf93298b5b97441a21d9a50fdcd0ee9945e)
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 #include <netinet6/sctp6_var.h>
41 #endif
42 #include <netinet/sctp_var.h>
43 #include <netinet/sctp_sysctl.h>
44 #include <netinet/sctp_timer.h>
45 #include <netinet/sctputil.h>
46 #include <netinet/sctp_output.h>
47 #include <netinet/sctp_header.h>
48 #include <netinet/sctp_indata.h>
49 #include <netinet/sctp_asconf.h>
50 #include <netinet/sctp_input.h>
51 #include <netinet/sctp.h>
52 #include <netinet/sctp_uio.h>
53 
54 
55 
56 void
57 sctp_early_fr_timer(struct sctp_inpcb *inp,
58     struct sctp_tcb *stcb,
59     struct sctp_nets *net)
60 {
61 	struct sctp_tmit_chunk *chk, *tp2;
62 	struct timeval now, min_wait, tv;
63 	unsigned int cur_rtt, cnt = 0, cnt_resend = 0;
64 
65 	/* an early FR is occuring. */
66 	(void)SCTP_GETTIME_TIMEVAL(&now);
67 	/* get cur rto in micro-seconds */
68 	if (net->lastsa == 0) {
69 		/* Hmm no rtt estimate yet? */
70 		cur_rtt = stcb->asoc.initial_rto >> 2;
71 	} else {
72 
73 		cur_rtt = ((net->lastsa >> 2) + net->lastsv) >> 1;
74 	}
75 	if (cur_rtt < sctp_early_fr_msec) {
76 		cur_rtt = sctp_early_fr_msec;
77 	}
78 	cur_rtt *= 1000;
79 	tv.tv_sec = cur_rtt / 1000000;
80 	tv.tv_usec = cur_rtt % 1000000;
81 	min_wait = now;
82 	timevalsub(&min_wait, &tv);
83 	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
84 		/*
85 		 * if we hit here, we don't have enough seconds on the clock
86 		 * to account for the RTO. We just let the lower seconds be
87 		 * the bounds and don't worry about it. This may mean we
88 		 * will mark a lot more than we should.
89 		 */
90 		min_wait.tv_sec = min_wait.tv_usec = 0;
91 	}
92 	chk = TAILQ_LAST(&stcb->asoc.sent_queue, sctpchunk_listhead);
93 	for (; chk != NULL; chk = tp2) {
94 		tp2 = TAILQ_PREV(chk, sctpchunk_listhead, sctp_next);
95 		if (chk->whoTo != net) {
96 			continue;
97 		}
98 		if (chk->sent == SCTP_DATAGRAM_RESEND)
99 			cnt_resend++;
100 		else if ((chk->sent > SCTP_DATAGRAM_UNSENT) &&
101 		    (chk->sent < SCTP_DATAGRAM_RESEND)) {
102 			/* pending, may need retran */
103 			if (chk->sent_rcv_time.tv_sec > min_wait.tv_sec) {
104 				/*
105 				 * we have reached a chunk that was sent
106 				 * some seconds past our min.. forget it we
107 				 * will find no more to send.
108 				 */
109 				continue;
110 			} else if (chk->sent_rcv_time.tv_sec == min_wait.tv_sec) {
111 				/*
112 				 * we must look at the micro seconds to
113 				 * know.
114 				 */
115 				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
116 					/*
117 					 * ok it was sent after our boundary
118 					 * time.
119 					 */
120 					continue;
121 				}
122 			}
123 			if (sctp_logging_level & SCTP_EARLYFR_LOGGING_ENABLE) {
124 				sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count,
125 				    4, SCTP_FR_MARKED_EARLY);
126 			}
127 			SCTP_STAT_INCR(sctps_earlyfrmrkretrans);
128 			chk->sent = SCTP_DATAGRAM_RESEND;
129 			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
130 			/* double book size since we are doing an early FR */
131 			chk->book_size_scale++;
132 			cnt += chk->send_size;
133 			if ((cnt + net->flight_size) > net->cwnd) {
134 				/* Mark all we could possibly resend */
135 				break;
136 			}
137 		}
138 	}
139 	if (cnt) {
140 		/*
141 		 * JRS - Use the congestion control given in the congestion
142 		 * control module
143 		 */
144 		stcb->asoc.cc_functions.sctp_cwnd_update_after_fr_timer(inp, stcb, net);
145 	} else if (cnt_resend) {
146 		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_EARLY_FR_TMR, SCTP_SO_NOT_LOCKED);
147 	}
148 	/* Restart it? */
149 	if (net->flight_size < net->cwnd) {
150 		SCTP_STAT_INCR(sctps_earlyfrstrtmr);
151 		sctp_timer_start(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net);
152 	}
153 }
154 
155 void
156 sctp_audit_retranmission_queue(struct sctp_association *asoc)
157 {
158 	struct sctp_tmit_chunk *chk;
159 
160 	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit invoked on send queue cnt:%d onqueue:%d\n",
161 	    asoc->sent_queue_retran_cnt,
162 	    asoc->sent_queue_cnt);
163 	asoc->sent_queue_retran_cnt = 0;
164 	asoc->sent_queue_cnt = 0;
165 	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
166 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
167 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
168 		}
169 		asoc->sent_queue_cnt++;
170 	}
171 	TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) {
172 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
173 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
174 		}
175 	}
176 	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit completes retran:%d onqueue:%d\n",
177 	    asoc->sent_queue_retran_cnt,
178 	    asoc->sent_queue_cnt);
179 }
180 
181 int
182 sctp_threshold_management(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
183     struct sctp_nets *net, uint16_t threshold)
184 {
185 	if (net) {
186 		net->error_count++;
187 		SCTPDBG(SCTP_DEBUG_TIMER4, "Error count for %p now %d thresh:%d\n",
188 		    net, net->error_count,
189 		    net->failure_threshold);
190 		if (net->error_count > net->failure_threshold) {
191 			/* We had a threshold failure */
192 			if (net->dest_state & SCTP_ADDR_REACHABLE) {
193 				net->dest_state &= ~SCTP_ADDR_REACHABLE;
194 				net->dest_state |= SCTP_ADDR_NOT_REACHABLE;
195 				net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY;
196 				if (net == stcb->asoc.primary_destination) {
197 					net->dest_state |= SCTP_ADDR_WAS_PRIMARY;
198 				}
199 				/*
200 				 * JRS 5/14/07 - If a destination is
201 				 * unreachable, the PF bit is turned off.
202 				 * This allows an unambiguous use of the PF
203 				 * bit for destinations that are reachable
204 				 * but potentially failed. If the
205 				 * destination is set to the unreachable
206 				 * state, also set the destination to the PF
207 				 * state.
208 				 */
209 				/*
210 				 * Add debug message here if destination is
211 				 * not in PF state.
212 				 */
213 				/* Stop any running T3 timers here? */
214 				if (sctp_cmt_on_off && sctp_cmt_pf) {
215 					net->dest_state &= ~SCTP_ADDR_PF;
216 					SCTPDBG(SCTP_DEBUG_TIMER4, "Destination %p moved from PF to unreachable.\n",
217 					    net);
218 				}
219 				sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN,
220 				    stcb,
221 				    SCTP_FAILED_THRESHOLD,
222 				    (void *)net, SCTP_SO_NOT_LOCKED);
223 			}
224 		}
225 		/*********HOLD THIS COMMENT FOR PATCH OF ALTERNATE
226 		 *********ROUTING CODE
227 		 */
228 		/*********HOLD THIS COMMENT FOR END OF PATCH OF ALTERNATE
229 		 *********ROUTING CODE
230 		 */
231 	}
232 	if (stcb == NULL)
233 		return (0);
234 
235 	if (net) {
236 		if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) {
237 			if (sctp_logging_level & SCTP_THRESHOLD_LOGGING) {
238 				sctp_misc_ints(SCTP_THRESHOLD_INCR,
239 				    stcb->asoc.overall_error_count,
240 				    (stcb->asoc.overall_error_count + 1),
241 				    SCTP_FROM_SCTP_TIMER,
242 				    __LINE__);
243 			}
244 			stcb->asoc.overall_error_count++;
245 		}
246 	} else {
247 		if (sctp_logging_level & SCTP_THRESHOLD_LOGGING) {
248 			sctp_misc_ints(SCTP_THRESHOLD_INCR,
249 			    stcb->asoc.overall_error_count,
250 			    (stcb->asoc.overall_error_count + 1),
251 			    SCTP_FROM_SCTP_TIMER,
252 			    __LINE__);
253 		}
254 		stcb->asoc.overall_error_count++;
255 	}
256 	SCTPDBG(SCTP_DEBUG_TIMER4, "Overall error count for %p now %d thresh:%u state:%x\n",
257 	    &stcb->asoc, stcb->asoc.overall_error_count,
258 	    (uint32_t) threshold,
259 	    ((net == NULL) ? (uint32_t) 0 : (uint32_t) net->dest_state));
260 	/*
261 	 * We specifically do not do >= to give the assoc one more change
262 	 * before we fail it.
263 	 */
264 	if (stcb->asoc.overall_error_count > threshold) {
265 		/* Abort notification sends a ULP notify */
266 		struct mbuf *oper;
267 
268 		oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
269 		    0, M_DONTWAIT, 1, MT_DATA);
270 		if (oper) {
271 			struct sctp_paramhdr *ph;
272 			uint32_t *ippp;
273 
274 			SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
275 			    sizeof(uint32_t);
276 			ph = mtod(oper, struct sctp_paramhdr *);
277 			ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
278 			ph->param_length = htons(SCTP_BUF_LEN(oper));
279 			ippp = (uint32_t *) (ph + 1);
280 			*ippp = htonl(SCTP_FROM_SCTP_TIMER + SCTP_LOC_1);
281 		}
282 		inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_1;
283 		printf("Aborting association threshold:%d overall error count:%d\n",
284 		    threshold,
285 		    stcb->asoc.overall_error_count);
286 		sctp_abort_an_association(inp, stcb, SCTP_FAILED_THRESHOLD, oper, SCTP_SO_NOT_LOCKED);
287 		return (1);
288 	}
289 	return (0);
290 }
291 
292 struct sctp_nets *
293 sctp_find_alternate_net(struct sctp_tcb *stcb,
294     struct sctp_nets *net,
295     int mode)
296 {
297 	/* Find and return an alternate network if possible */
298 	struct sctp_nets *alt, *mnet, *min_errors_net = NULL, *max_cwnd_net = NULL;
299 	int once;
300 
301 	/* JRS 5/14/07 - Initialize min_errors to an impossible value. */
302 	int min_errors = -1;
303 	uint32_t max_cwnd = 0;
304 
305 	if (stcb->asoc.numnets == 1) {
306 		/* No others but net */
307 		return (TAILQ_FIRST(&stcb->asoc.nets));
308 	}
309 	/*
310 	 * JRS 5/14/07 - If mode is set to 2, use the CMT PF find alternate
311 	 * net algorithm. This algorithm chooses the active destination (not
312 	 * in PF state) with the largest cwnd value. If all destinations are
313 	 * in PF state, unreachable, or unconfirmed, choose the desination
314 	 * that is in PF state with the lowest error count. In case of a
315 	 * tie, choose the destination that was most recently active.
316 	 */
317 	if (mode == 2) {
318 		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
319 			/*
320 			 * JRS 5/14/07 - If the destination is unreachable
321 			 * or unconfirmed, skip it.
322 			 */
323 			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
324 			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
325 				continue;
326 			}
327 			/*
328 			 * JRS 5/14/07 -  If the destination is reachable
329 			 * but in PF state, compare the error count of the
330 			 * destination to the minimum error count seen thus
331 			 * far. Store the destination with the lower error
332 			 * count.  If the error counts are equal, store the
333 			 * destination that was most recently active.
334 			 */
335 			if (mnet->dest_state & SCTP_ADDR_PF) {
336 				/*
337 				 * JRS 5/14/07 - If the destination under
338 				 * consideration is the current destination,
339 				 * work as if the error count is one higher.
340 				 * The actual error count will not be
341 				 * incremented until later in the t3
342 				 * handler.
343 				 */
344 				if (mnet == net) {
345 					if (min_errors == -1) {
346 						min_errors = mnet->error_count + 1;
347 						min_errors_net = mnet;
348 					} else if (mnet->error_count + 1 < min_errors) {
349 						min_errors = mnet->error_count + 1;
350 						min_errors_net = mnet;
351 					} else if (mnet->error_count + 1 == min_errors
352 					    && mnet->last_active > min_errors_net->last_active) {
353 						min_errors_net = mnet;
354 						min_errors = mnet->error_count + 1;
355 					}
356 					continue;
357 				} else {
358 					if (min_errors == -1) {
359 						min_errors = mnet->error_count;
360 						min_errors_net = mnet;
361 					} else if (mnet->error_count < min_errors) {
362 						min_errors = mnet->error_count;
363 						min_errors_net = mnet;
364 					} else if (mnet->error_count == min_errors
365 					    && mnet->last_active > min_errors_net->last_active) {
366 						min_errors_net = mnet;
367 						min_errors = mnet->error_count;
368 					}
369 					continue;
370 				}
371 			}
372 			/*
373 			 * JRS 5/14/07 - If the destination is reachable and
374 			 * not in PF state, compare the cwnd of the
375 			 * destination to the highest cwnd seen thus far.
376 			 * Store the destination with the higher cwnd value.
377 			 * If the cwnd values are equal, randomly choose one
378 			 * of the two destinations.
379 			 */
380 			if (max_cwnd < mnet->cwnd) {
381 				max_cwnd_net = mnet;
382 				max_cwnd = mnet->cwnd;
383 			} else if (max_cwnd == mnet->cwnd) {
384 				uint32_t rndval;
385 				uint8_t this_random;
386 
387 				if (stcb->asoc.hb_random_idx > 3) {
388 					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
389 					memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values));
390 					this_random = stcb->asoc.hb_random_values[0];
391 					stcb->asoc.hb_random_idx++;
392 					stcb->asoc.hb_ect_randombit = 0;
393 				} else {
394 					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
395 					stcb->asoc.hb_random_idx++;
396 					stcb->asoc.hb_ect_randombit = 0;
397 				}
398 				if (this_random % 2 == 1) {
399 					max_cwnd_net = mnet;
400 					max_cwnd = mnet->cwnd;
401 					//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_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_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_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_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_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_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_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_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_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_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_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_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_logging_level & SCTP_FR_LOGGING_ENABLE) {
933 		sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT);
934 	}
935 	if (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_cmt_on_off && 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_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_cmt_on_off && 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_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;
1318 
1319 	/* is this a first send, or a retransmission? */
1320 	if (stcb->asoc.asconf_sent == 0) {
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 		TAILQ_FOREACH(asconf, &stcb->asoc.control_send_queue,
1330 		    sctp_next) {
1331 			if (asconf->rec.chunk_id.id == SCTP_ASCONF) {
1332 				break;
1333 			}
1334 		}
1335 		if (asconf == NULL) {
1336 			return (0);
1337 		}
1338 		/* do threshold management */
1339 		if (sctp_threshold_management(inp, stcb, asconf->whoTo,
1340 		    stcb->asoc.max_send_times)) {
1341 			/* Assoc is over */
1342 			return (1);
1343 		}
1344 		if (asconf->snd_count > stcb->asoc.max_send_times) {
1345 			/*
1346 			 * Something is rotten: our peer is not responding
1347 			 * to ASCONFs but apparently is to other chunks.
1348 			 * i.e. it is not properly handling the chunk type
1349 			 * upper bits. Mark this peer as ASCONF incapable
1350 			 * and cleanup.
1351 			 */
1352 			SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n");
1353 			sctp_asconf_cleanup(stcb, net);
1354 			return (0);
1355 		}
1356 		/*
1357 		 * cleared threshold management, so now backoff the net and
1358 		 * select an alternate
1359 		 */
1360 		sctp_backoff_on_timeout(stcb, asconf->whoTo, 1, 0);
1361 		alt = sctp_find_alternate_net(stcb, asconf->whoTo, 0);
1362 		sctp_free_remote_addr(asconf->whoTo);
1363 		asconf->whoTo = alt;
1364 		atomic_add_int(&alt->ref_count, 1);
1365 
1366 		/* See if an ECN Echo is also stranded */
1367 		TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1368 			if ((chk->whoTo == net) &&
1369 			    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1370 				sctp_free_remote_addr(chk->whoTo);
1371 				chk->whoTo = alt;
1372 				if (chk->sent != SCTP_DATAGRAM_RESEND) {
1373 					chk->sent = SCTP_DATAGRAM_RESEND;
1374 					sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1375 				}
1376 				atomic_add_int(&alt->ref_count, 1);
1377 			}
1378 		}
1379 		if (net->dest_state & SCTP_ADDR_NOT_REACHABLE) {
1380 			/*
1381 			 * If the address went un-reachable, we need to move
1382 			 * to the alternate for ALL chunks in queue
1383 			 */
1384 			sctp_move_all_chunks_to_alt(stcb, net, alt);
1385 		}
1386 		/* mark the retran info */
1387 		if (asconf->sent != SCTP_DATAGRAM_RESEND)
1388 			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1389 		asconf->sent = SCTP_DATAGRAM_RESEND;
1390 	}
1391 	return (0);
1392 }
1393 
1394 /* Mobility adaptation */
1395 void
1396 sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1397     struct sctp_nets *net)
1398 {
1399 	if (stcb->asoc.deleted_primary == NULL) {
1400 		SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n");
1401 		sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1402 		return;
1403 	}
1404 	SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary ");
1405 	SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa);
1406 	sctp_free_remote_addr(stcb->asoc.deleted_primary);
1407 	stcb->asoc.deleted_primary = NULL;
1408 	sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1409 	return;
1410 }
1411 
1412 /*
1413  * For the shutdown and shutdown-ack, we do not keep one around on the
1414  * control queue. This means we must generate a new one and call the general
1415  * chunk output routine, AFTER having done threshold management.
1416  */
1417 int
1418 sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1419     struct sctp_nets *net)
1420 {
1421 	struct sctp_nets *alt;
1422 
1423 	/* first threshold managment */
1424 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1425 		/* Assoc is over */
1426 		return (1);
1427 	}
1428 	/* second select an alternative */
1429 	alt = sctp_find_alternate_net(stcb, net, 0);
1430 
1431 	/* third generate a shutdown into the queue for out net */
1432 	if (alt) {
1433 		sctp_send_shutdown(stcb, alt);
1434 	} else {
1435 		/*
1436 		 * if alt is NULL, there is no dest to send to??
1437 		 */
1438 		return (0);
1439 	}
1440 	/* fourth restart timer */
1441 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt);
1442 	return (0);
1443 }
1444 
1445 int
1446 sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1447     struct sctp_nets *net)
1448 {
1449 	struct sctp_nets *alt;
1450 
1451 	/* first threshold managment */
1452 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1453 		/* Assoc is over */
1454 		return (1);
1455 	}
1456 	/* second select an alternative */
1457 	alt = sctp_find_alternate_net(stcb, net, 0);
1458 
1459 	/* third generate a shutdown into the queue for out net */
1460 	sctp_send_shutdown_ack(stcb, alt);
1461 
1462 	/* fourth restart timer */
1463 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt);
1464 	return (0);
1465 }
1466 
1467 static void
1468 sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp,
1469     struct sctp_tcb *stcb)
1470 {
1471 	struct sctp_stream_out *outs;
1472 	struct sctp_stream_queue_pending *sp;
1473 	unsigned int chks_in_queue = 0;
1474 	int being_filled = 0;
1475 
1476 	/*
1477 	 * This function is ONLY called when the send/sent queues are empty.
1478 	 */
1479 	if ((stcb == NULL) || (inp == NULL))
1480 		return;
1481 
1482 	if (stcb->asoc.sent_queue_retran_cnt) {
1483 		SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n",
1484 		    stcb->asoc.sent_queue_retran_cnt);
1485 		stcb->asoc.sent_queue_retran_cnt = 0;
1486 	}
1487 	SCTP_TCB_SEND_LOCK(stcb);
1488 	if (TAILQ_EMPTY(&stcb->asoc.out_wheel)) {
1489 		int i, cnt = 0;
1490 
1491 		/* Check to see if a spoke fell off the wheel */
1492 		for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
1493 			if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) {
1494 				sctp_insert_on_wheel(stcb, &stcb->asoc, &stcb->asoc.strmout[i], 1);
1495 				cnt++;
1496 			}
1497 		}
1498 		if (cnt) {
1499 			/* yep, we lost a spoke or two */
1500 			SCTP_PRINTF("Found an additional %d streams NOT on outwheel, corrected\n", cnt);
1501 		} else {
1502 			/* no spokes lost, */
1503 			stcb->asoc.total_output_queue_size = 0;
1504 		}
1505 		SCTP_TCB_SEND_UNLOCK(stcb);
1506 		return;
1507 	}
1508 	SCTP_TCB_SEND_UNLOCK(stcb);
1509 	/* Check to see if some data queued, if so report it */
1510 	TAILQ_FOREACH(outs, &stcb->asoc.out_wheel, next_spoke) {
1511 		if (!TAILQ_EMPTY(&outs->outqueue)) {
1512 			TAILQ_FOREACH(sp, &outs->outqueue, next) {
1513 				if (sp->msg_is_complete)
1514 					being_filled++;
1515 				chks_in_queue++;
1516 			}
1517 		}
1518 	}
1519 	if (chks_in_queue != stcb->asoc.stream_queue_cnt) {
1520 		SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n",
1521 		    stcb->asoc.stream_queue_cnt, chks_in_queue);
1522 	}
1523 	if (chks_in_queue) {
1524 		/* call the output queue function */
1525 		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1526 		if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1527 		    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1528 			/*
1529 			 * Probably should go in and make it go back through
1530 			 * and add fragments allowed
1531 			 */
1532 			if (being_filled == 0) {
1533 				SCTP_PRINTF("Still nothing moved %d chunks are stuck\n",
1534 				    chks_in_queue);
1535 			}
1536 		}
1537 	} else {
1538 		SCTP_PRINTF("Found no chunks on any queue tot:%lu\n",
1539 		    (u_long)stcb->asoc.total_output_queue_size);
1540 		stcb->asoc.total_output_queue_size = 0;
1541 	}
1542 }
1543 
1544 int
1545 sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1546     struct sctp_nets *net, int cnt_of_unconf)
1547 {
1548 	int ret;
1549 
1550 	if (net) {
1551 		if (net->hb_responded == 0) {
1552 			if (net->ro._s_addr) {
1553 				/*
1554 				 * Invalidate the src address if we did not
1555 				 * get a response last time.
1556 				 */
1557 				sctp_free_ifa(net->ro._s_addr);
1558 				net->ro._s_addr = NULL;
1559 				net->src_addr_selected = 0;
1560 			}
1561 			sctp_backoff_on_timeout(stcb, net, 1, 0);
1562 		}
1563 		/* Zero PBA, if it needs it */
1564 		if (net->partial_bytes_acked) {
1565 			net->partial_bytes_acked = 0;
1566 		}
1567 	}
1568 	if ((stcb->asoc.total_output_queue_size > 0) &&
1569 	    (TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1570 	    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1571 		sctp_audit_stream_queues_for_size(inp, stcb);
1572 	}
1573 	/* Send a new HB, this will do threshold managment, pick a new dest */
1574 	if (cnt_of_unconf == 0) {
1575 		if (sctp_send_hb(stcb, 0, NULL) < 0) {
1576 			return (1);
1577 		}
1578 	} else {
1579 		/*
1580 		 * this will send out extra hb's up to maxburst if there are
1581 		 * any unconfirmed addresses.
1582 		 */
1583 		uint32_t cnt_sent = 0;
1584 
1585 		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1586 			if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) &&
1587 			    (net->dest_state & SCTP_ADDR_REACHABLE)) {
1588 				cnt_sent++;
1589 				if (net->hb_responded == 0) {
1590 					/* Did we respond last time? */
1591 					if (net->ro._s_addr) {
1592 						sctp_free_ifa(net->ro._s_addr);
1593 						net->ro._s_addr = NULL;
1594 						net->src_addr_selected = 0;
1595 					}
1596 				}
1597 				ret = sctp_send_hb(stcb, 1, net);
1598 				if (ret < 0)
1599 					return 1;
1600 				else if (ret == 0) {
1601 					break;
1602 				}
1603 				if (cnt_sent >= sctp_hb_maxburst)
1604 					break;
1605 			}
1606 		}
1607 	}
1608 	return (0);
1609 }
1610 
1611 int
1612 sctp_is_hb_timer_running(struct sctp_tcb *stcb)
1613 {
1614 	if (SCTP_OS_TIMER_PENDING(&stcb->asoc.hb_timer.timer)) {
1615 		/* its running */
1616 		return (1);
1617 	} else {
1618 		/* nope */
1619 		return (0);
1620 	}
1621 }
1622 
1623 int
1624 sctp_is_sack_timer_running(struct sctp_tcb *stcb)
1625 {
1626 	if (SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) {
1627 		/* its running */
1628 		return (1);
1629 	} else {
1630 		/* nope */
1631 		return (0);
1632 	}
1633 }
1634 
1635 #define SCTP_NUMBER_OF_MTU_SIZES 18
1636 static uint32_t mtu_sizes[] = {
1637 	68,
1638 	296,
1639 	508,
1640 	512,
1641 	544,
1642 	576,
1643 	1006,
1644 	1492,
1645 	1500,
1646 	1536,
1647 	2002,
1648 	2048,
1649 	4352,
1650 	4464,
1651 	8166,
1652 	17914,
1653 	32000,
1654 	65535
1655 };
1656 
1657 
1658 static uint32_t
1659 sctp_getnext_mtu(struct sctp_inpcb *inp, uint32_t cur_mtu)
1660 {
1661 	/* select another MTU that is just bigger than this one */
1662 	int i;
1663 
1664 	for (i = 0; i < SCTP_NUMBER_OF_MTU_SIZES; i++) {
1665 		if (cur_mtu < mtu_sizes[i]) {
1666 			/* no max_mtu is bigger than this one */
1667 			return (mtu_sizes[i]);
1668 		}
1669 	}
1670 	/* here return the highest allowable */
1671 	return (cur_mtu);
1672 }
1673 
1674 
1675 void
1676 sctp_pathmtu_timer(struct sctp_inpcb *inp,
1677     struct sctp_tcb *stcb,
1678     struct sctp_nets *net)
1679 {
1680 	uint32_t next_mtu;
1681 
1682 	/* restart the timer in any case */
1683 	next_mtu = sctp_getnext_mtu(inp, net->mtu);
1684 	if (next_mtu <= net->mtu) {
1685 		/* nothing to do */
1686 		return;
1687 	} {
1688 		uint32_t mtu;
1689 
1690 		if ((net->src_addr_selected == 0) ||
1691 		    (net->ro._s_addr == NULL) ||
1692 		    (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1693 			if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1694 				sctp_free_ifa(net->ro._s_addr);
1695 				net->ro._s_addr = NULL;
1696 				net->src_addr_selected = 0;
1697 			} else if (net->ro._s_addr == NULL) {
1698 				net->ro._s_addr = sctp_source_address_selection(inp,
1699 				    stcb,
1700 				    (sctp_route_t *) & net->ro,
1701 				    net, 0, stcb->asoc.vrf_id);
1702 			}
1703 			if (net->ro._s_addr)
1704 				net->src_addr_selected = 1;
1705 		}
1706 		if (net->ro._s_addr) {
1707 			mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_rt);
1708 			if (mtu > next_mtu) {
1709 				net->mtu = next_mtu;
1710 			}
1711 		}
1712 	}
1713 	/* restart the timer */
1714 	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
1715 }
1716 
1717 void
1718 sctp_autoclose_timer(struct sctp_inpcb *inp,
1719     struct sctp_tcb *stcb,
1720     struct sctp_nets *net)
1721 {
1722 	struct timeval tn, *tim_touse;
1723 	struct sctp_association *asoc;
1724 	int ticks_gone_by;
1725 
1726 	(void)SCTP_GETTIME_TIMEVAL(&tn);
1727 	if (stcb->asoc.sctp_autoclose_ticks &&
1728 	    sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) {
1729 		/* Auto close is on */
1730 		asoc = &stcb->asoc;
1731 		/* pick the time to use */
1732 		if (asoc->time_last_rcvd.tv_sec >
1733 		    asoc->time_last_sent.tv_sec) {
1734 			tim_touse = &asoc->time_last_rcvd;
1735 		} else {
1736 			tim_touse = &asoc->time_last_sent;
1737 		}
1738 		/* Now has long enough transpired to autoclose? */
1739 		ticks_gone_by = SEC_TO_TICKS(tn.tv_sec - tim_touse->tv_sec);
1740 		if ((ticks_gone_by > 0) &&
1741 		    (ticks_gone_by >= (int)asoc->sctp_autoclose_ticks)) {
1742 			/*
1743 			 * autoclose time has hit, call the output routine,
1744 			 * which should do nothing just to be SURE we don't
1745 			 * have hanging data. We can then safely check the
1746 			 * queues and know that we are clear to send
1747 			 * shutdown
1748 			 */
1749 			sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED);
1750 			/* Are we clean? */
1751 			if (TAILQ_EMPTY(&asoc->send_queue) &&
1752 			    TAILQ_EMPTY(&asoc->sent_queue)) {
1753 				/*
1754 				 * there is nothing queued to send, so I'm
1755 				 * done...
1756 				 */
1757 				if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) {
1758 					/* only send SHUTDOWN 1st time thru */
1759 					sctp_send_shutdown(stcb, stcb->asoc.primary_destination);
1760 					if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
1761 					    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
1762 						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
1763 					}
1764 					SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
1765 					SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
1766 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
1767 					    stcb->sctp_ep, stcb,
1768 					    asoc->primary_destination);
1769 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
1770 					    stcb->sctp_ep, stcb,
1771 					    asoc->primary_destination);
1772 				}
1773 			}
1774 		} else {
1775 			/*
1776 			 * No auto close at this time, reset t-o to check
1777 			 * later
1778 			 */
1779 			int tmp;
1780 
1781 			/* fool the timer startup to use the time left */
1782 			tmp = asoc->sctp_autoclose_ticks;
1783 			asoc->sctp_autoclose_ticks -= ticks_gone_by;
1784 			sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb,
1785 			    net);
1786 			/* restore the real tick value */
1787 			asoc->sctp_autoclose_ticks = tmp;
1788 		}
1789 	}
1790 }
1791 
1792 void
1793 sctp_iterator_timer(struct sctp_iterator *it)
1794 {
1795 	int iteration_count = 0;
1796 	int inp_skip = 0;
1797 
1798 	/*
1799 	 * only one iterator can run at a time. This is the only way we can
1800 	 * cleanly pull ep's from underneath all the running interators when
1801 	 * a ep is freed.
1802 	 */
1803 	SCTP_ITERATOR_LOCK();
1804 	if (it->inp == NULL) {
1805 		/* iterator is complete */
1806 done_with_iterator:
1807 		SCTP_ITERATOR_UNLOCK();
1808 		SCTP_INP_INFO_WLOCK();
1809 		TAILQ_REMOVE(&sctppcbinfo.iteratorhead, it, sctp_nxt_itr);
1810 		/* stopping the callout is not needed, in theory */
1811 		SCTP_INP_INFO_WUNLOCK();
1812 		(void)SCTP_OS_TIMER_STOP(&it->tmr.timer);
1813 		if (it->function_atend != NULL) {
1814 			(*it->function_atend) (it->pointer, it->val);
1815 		}
1816 		SCTP_FREE(it, SCTP_M_ITER);
1817 		return;
1818 	}
1819 select_a_new_ep:
1820 	SCTP_INP_WLOCK(it->inp);
1821 	while (((it->pcb_flags) &&
1822 	    ((it->inp->sctp_flags & it->pcb_flags) != it->pcb_flags)) ||
1823 	    ((it->pcb_features) &&
1824 	    ((it->inp->sctp_features & it->pcb_features) != it->pcb_features))) {
1825 		/* endpoint flags or features don't match, so keep looking */
1826 		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
1827 			SCTP_INP_WUNLOCK(it->inp);
1828 			goto done_with_iterator;
1829 		}
1830 		SCTP_INP_WUNLOCK(it->inp);
1831 		it->inp = LIST_NEXT(it->inp, sctp_list);
1832 		if (it->inp == NULL) {
1833 			goto done_with_iterator;
1834 		}
1835 		SCTP_INP_WLOCK(it->inp);
1836 	}
1837 	if ((it->inp->inp_starting_point_for_iterator != NULL) &&
1838 	    (it->inp->inp_starting_point_for_iterator != it)) {
1839 		SCTP_PRINTF("Iterator collision, waiting for one at %p\n",
1840 		    it->inp);
1841 		SCTP_INP_WUNLOCK(it->inp);
1842 		goto start_timer_return;
1843 	}
1844 	/* mark the current iterator on the endpoint */
1845 	it->inp->inp_starting_point_for_iterator = it;
1846 	SCTP_INP_WUNLOCK(it->inp);
1847 	SCTP_INP_RLOCK(it->inp);
1848 	/* now go through each assoc which is in the desired state */
1849 	if (it->done_current_ep == 0) {
1850 		if (it->function_inp != NULL)
1851 			inp_skip = (*it->function_inp) (it->inp, it->pointer, it->val);
1852 		it->done_current_ep = 1;
1853 	}
1854 	if (it->stcb == NULL) {
1855 		/* run the per instance function */
1856 		it->stcb = LIST_FIRST(&it->inp->sctp_asoc_list);
1857 	}
1858 	SCTP_INP_RUNLOCK(it->inp);
1859 	if ((inp_skip) || it->stcb == NULL) {
1860 		if (it->function_inp_end != NULL) {
1861 			inp_skip = (*it->function_inp_end) (it->inp,
1862 			    it->pointer,
1863 			    it->val);
1864 		}
1865 		goto no_stcb;
1866 	}
1867 	if ((it->stcb) &&
1868 	    (it->stcb->asoc.stcb_starting_point_for_iterator == it)) {
1869 		it->stcb->asoc.stcb_starting_point_for_iterator = NULL;
1870 	}
1871 	while (it->stcb) {
1872 		SCTP_TCB_LOCK(it->stcb);
1873 		if (it->asoc_state && ((it->stcb->asoc.state & it->asoc_state) != it->asoc_state)) {
1874 			/* not in the right state... keep looking */
1875 			SCTP_TCB_UNLOCK(it->stcb);
1876 			goto next_assoc;
1877 		}
1878 		/* mark the current iterator on the assoc */
1879 		it->stcb->asoc.stcb_starting_point_for_iterator = it;
1880 		/* see if we have limited out the iterator loop */
1881 		iteration_count++;
1882 		if (iteration_count > SCTP_ITERATOR_MAX_AT_ONCE) {
1883 	start_timer_return:
1884 			/* set a timer to continue this later */
1885 			if (it->stcb)
1886 				SCTP_TCB_UNLOCK(it->stcb);
1887 			sctp_timer_start(SCTP_TIMER_TYPE_ITERATOR,
1888 			    (struct sctp_inpcb *)it, NULL, NULL);
1889 			SCTP_ITERATOR_UNLOCK();
1890 			return;
1891 		}
1892 		/* run function on this one */
1893 		(*it->function_assoc) (it->inp, it->stcb, it->pointer, it->val);
1894 
1895 		/*
1896 		 * we lie here, it really needs to have its own type but
1897 		 * first I must verify that this won't effect things :-0
1898 		 */
1899 		if (it->no_chunk_output == 0)
1900 			sctp_chunk_output(it->inp, it->stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1901 
1902 		SCTP_TCB_UNLOCK(it->stcb);
1903 next_assoc:
1904 		it->stcb = LIST_NEXT(it->stcb, sctp_tcblist);
1905 		if (it->stcb == NULL) {
1906 			if (it->function_inp_end != NULL) {
1907 				inp_skip = (*it->function_inp_end) (it->inp,
1908 				    it->pointer,
1909 				    it->val);
1910 			}
1911 		}
1912 	}
1913 no_stcb:
1914 	/* done with all assocs on this endpoint, move on to next endpoint */
1915 	it->done_current_ep = 0;
1916 	SCTP_INP_WLOCK(it->inp);
1917 	it->inp->inp_starting_point_for_iterator = NULL;
1918 	SCTP_INP_WUNLOCK(it->inp);
1919 	if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
1920 		it->inp = NULL;
1921 	} else {
1922 		SCTP_INP_INFO_RLOCK();
1923 		it->inp = LIST_NEXT(it->inp, sctp_list);
1924 		SCTP_INP_INFO_RUNLOCK();
1925 	}
1926 	if (it->inp == NULL) {
1927 		goto done_with_iterator;
1928 	}
1929 	goto select_a_new_ep;
1930 }
1931