xref: /freebsd/sys/netinet/sctp_timer.c (revision 83416c885d74830565cf956f2a27c4c122949c8b)
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 #ifndef INVARIANTS
565 static void
566 sctp_recover_sent_list(struct sctp_tcb *stcb)
567 {
568 	struct sctp_tmit_chunk *chk, *tp2;
569 	struct sctp_association *asoc;
570 
571 	asoc = &stcb->asoc;
572 	chk = TAILQ_FIRST(&stcb->asoc.sent_queue);
573 	for (; chk != NULL; chk = tp2) {
574 		tp2 = TAILQ_NEXT(chk, sctp_next);
575 		if ((compare_with_wrap(stcb->asoc.last_acked_seq,
576 		    chk->rec.data.TSN_seq,
577 		    MAX_TSN)) ||
578 		    (stcb->asoc.last_acked_seq == chk->rec.data.TSN_seq)) {
579 
580 			SCTP_PRINTF("Found chk:%p tsn:%x <= last_acked_seq:%x\n",
581 			    chk, chk->rec.data.TSN_seq, stcb->asoc.last_acked_seq);
582 			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
583 			if (chk->pr_sctp_on) {
584 				if (asoc->pr_sctp_cnt != 0)
585 					asoc->pr_sctp_cnt--;
586 			}
587 			if (chk->data) {
588 				/* sa_ignore NO_NULL_CHK */
589 				sctp_free_bufspace(stcb, asoc, chk, 1);
590 				sctp_m_freem(chk->data);
591 				if (PR_SCTP_BUF_ENABLED(chk->flags)) {
592 					asoc->sent_queue_cnt_removeable--;
593 				}
594 			}
595 			chk->data = NULL;
596 			asoc->sent_queue_cnt--;
597 			sctp_free_a_chunk(stcb, chk);
598 		}
599 	}
600 	SCTP_PRINTF("after recover order is as follows\n");
601 	chk = TAILQ_FIRST(&stcb->asoc.sent_queue);
602 	for (; chk != NULL; chk = tp2) {
603 		tp2 = TAILQ_NEXT(chk, sctp_next);
604 		SCTP_PRINTF("chk:%p TSN:%x\n", chk, chk->rec.data.TSN_seq);
605 	}
606 }
607 
608 #endif
609 
610 static int
611 sctp_mark_all_for_resend(struct sctp_tcb *stcb,
612     struct sctp_nets *net,
613     struct sctp_nets *alt,
614     int window_probe,
615     int *num_marked)
616 {
617 
618 	/*
619 	 * Mark all chunks (well not all) that were sent to *net for
620 	 * retransmission. Move them to alt for there destination as well...
621 	 * We only mark chunks that have been outstanding long enough to
622 	 * have received feed-back.
623 	 */
624 	struct sctp_tmit_chunk *chk, *tp2, *could_be_sent = NULL;
625 	struct sctp_nets *lnets;
626 	struct timeval now, min_wait, tv;
627 	int cur_rtt;
628 	int audit_tf, num_mk, fir;
629 	unsigned int cnt_mk;
630 	uint32_t orig_flight, orig_tf;
631 	uint32_t tsnlast, tsnfirst;
632 	int recovery_cnt = 0;
633 
634 
635 	/* none in flight now */
636 	audit_tf = 0;
637 	fir = 0;
638 	/*
639 	 * figure out how long a data chunk must be pending before we can
640 	 * mark it ..
641 	 */
642 	(void)SCTP_GETTIME_TIMEVAL(&now);
643 	/* get cur rto in micro-seconds */
644 	cur_rtt = (((net->lastsa >> 2) + net->lastsv) >> 1);
645 	cur_rtt *= 1000;
646 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
647 		sctp_log_fr(cur_rtt,
648 		    stcb->asoc.peers_rwnd,
649 		    window_probe,
650 		    SCTP_FR_T3_MARK_TIME);
651 		sctp_log_fr(net->flight_size,
652 		    SCTP_OS_TIMER_PENDING(&net->fr_timer.timer),
653 		    SCTP_OS_TIMER_ACTIVE(&net->fr_timer.timer),
654 		    SCTP_FR_CWND_REPORT);
655 		sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT);
656 	}
657 	tv.tv_sec = cur_rtt / 1000000;
658 	tv.tv_usec = cur_rtt % 1000000;
659 	min_wait = now;
660 	timevalsub(&min_wait, &tv);
661 	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
662 		/*
663 		 * if we hit here, we don't have enough seconds on the clock
664 		 * to account for the RTO. We just let the lower seconds be
665 		 * the bounds and don't worry about it. This may mean we
666 		 * will mark a lot more than we should.
667 		 */
668 		min_wait.tv_sec = min_wait.tv_usec = 0;
669 	}
670 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
671 		sctp_log_fr(cur_rtt, now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME);
672 		sctp_log_fr(0, min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME);
673 	}
674 	/*
675 	 * Our rwnd will be incorrect here since we are not adding back the
676 	 * cnt * mbuf but we will fix that down below.
677 	 */
678 	orig_flight = net->flight_size;
679 	orig_tf = stcb->asoc.total_flight;
680 
681 	net->fast_retran_ip = 0;
682 	/* Now on to each chunk */
683 	num_mk = cnt_mk = 0;
684 	tsnfirst = tsnlast = 0;
685 #ifndef INVARIANTS
686 start_again:
687 #endif
688 	chk = TAILQ_FIRST(&stcb->asoc.sent_queue);
689 	for (; chk != NULL; chk = tp2) {
690 		tp2 = TAILQ_NEXT(chk, sctp_next);
691 		if ((compare_with_wrap(stcb->asoc.last_acked_seq,
692 		    chk->rec.data.TSN_seq,
693 		    MAX_TSN)) ||
694 		    (stcb->asoc.last_acked_seq == chk->rec.data.TSN_seq)) {
695 			/* Strange case our list got out of order? */
696 			SCTP_PRINTF("Our list is out of order? last_acked:%x chk:%x",
697 			    (unsigned int)stcb->asoc.last_acked_seq, (unsigned int)chk->rec.data.TSN_seq);
698 			recovery_cnt++;
699 #ifdef INVARIANTS
700 			panic("last acked >= chk on sent-Q");
701 #else
702 			SCTP_PRINTF("Recover attempts a restart cnt:%d\n", recovery_cnt);
703 			sctp_recover_sent_list(stcb);
704 			if (recovery_cnt < 10) {
705 				goto start_again;
706 			} else {
707 				SCTP_PRINTF("Recovery fails %d times??\n", recovery_cnt);
708 			}
709 #endif
710 		}
711 		if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) {
712 			/*
713 			 * found one to mark: If it is less than
714 			 * DATAGRAM_ACKED it MUST not be a skipped or marked
715 			 * TSN but instead one that is either already set
716 			 * for retransmission OR one that needs
717 			 * retransmission.
718 			 */
719 
720 			/* validate its been outstanding long enough */
721 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
722 				sctp_log_fr(chk->rec.data.TSN_seq,
723 				    chk->sent_rcv_time.tv_sec,
724 				    chk->sent_rcv_time.tv_usec,
725 				    SCTP_FR_T3_MARK_TIME);
726 			}
727 			if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) {
728 				/*
729 				 * we have reached a chunk that was sent
730 				 * some seconds past our min.. forget it we
731 				 * will find no more to send.
732 				 */
733 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
734 					sctp_log_fr(0,
735 					    chk->sent_rcv_time.tv_sec,
736 					    chk->sent_rcv_time.tv_usec,
737 					    SCTP_FR_T3_STOPPED);
738 				}
739 				continue;
740 			} else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) &&
741 			    (window_probe == 0)) {
742 				/*
743 				 * we must look at the micro seconds to
744 				 * know.
745 				 */
746 				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
747 					/*
748 					 * ok it was sent after our boundary
749 					 * time.
750 					 */
751 					if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
752 						sctp_log_fr(0,
753 						    chk->sent_rcv_time.tv_sec,
754 						    chk->sent_rcv_time.tv_usec,
755 						    SCTP_FR_T3_STOPPED);
756 					}
757 					continue;
758 				}
759 			}
760 			if (PR_SCTP_TTL_ENABLED(chk->flags)) {
761 				/* Is it expired? */
762 				if ((now.tv_sec > chk->rec.data.timetodrop.tv_sec) ||
763 				    ((chk->rec.data.timetodrop.tv_sec == now.tv_sec) &&
764 				    (now.tv_usec > chk->rec.data.timetodrop.tv_usec))) {
765 					/* Yes so drop it */
766 					if (chk->data) {
767 						(void)sctp_release_pr_sctp_chunk(stcb,
768 						    chk,
769 						    (SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_SENT),
770 						    &stcb->asoc.sent_queue, SCTP_SO_NOT_LOCKED);
771 					}
772 				}
773 				continue;
774 			}
775 			if (PR_SCTP_RTX_ENABLED(chk->flags)) {
776 				/* Has it been retransmitted tv_sec times? */
777 				if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) {
778 					if (chk->data) {
779 						(void)sctp_release_pr_sctp_chunk(stcb,
780 						    chk,
781 						    (SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_SENT),
782 						    &stcb->asoc.sent_queue, SCTP_SO_NOT_LOCKED);
783 					}
784 				}
785 				continue;
786 			}
787 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
788 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
789 				num_mk++;
790 				if (fir == 0) {
791 					fir = 1;
792 					tsnfirst = chk->rec.data.TSN_seq;
793 				}
794 				tsnlast = chk->rec.data.TSN_seq;
795 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
796 					sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count,
797 					    0, SCTP_FR_T3_MARKED);
798 				}
799 				if (chk->rec.data.chunk_was_revoked) {
800 					/* deflate the cwnd */
801 					chk->whoTo->cwnd -= chk->book_size;
802 					chk->rec.data.chunk_was_revoked = 0;
803 				}
804 				net->marked_retrans++;
805 				stcb->asoc.marked_retrans++;
806 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
807 					sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO,
808 					    chk->whoTo->flight_size,
809 					    chk->book_size,
810 					    (uintptr_t) chk->whoTo,
811 					    chk->rec.data.TSN_seq);
812 				}
813 				sctp_flight_size_decrease(chk);
814 				sctp_total_flight_decrease(stcb, chk);
815 				stcb->asoc.peers_rwnd += chk->send_size;
816 				stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh);
817 			}
818 			chk->sent = SCTP_DATAGRAM_RESEND;
819 			SCTP_STAT_INCR(sctps_markedretrans);
820 
821 			/* reset the TSN for striking and other FR stuff */
822 			chk->rec.data.doing_fast_retransmit = 0;
823 			/* Clear any time so NO RTT is being done */
824 			chk->do_rtt = 0;
825 			if (alt != net) {
826 				sctp_free_remote_addr(chk->whoTo);
827 				chk->no_fr_allowed = 1;
828 				chk->whoTo = alt;
829 				atomic_add_int(&alt->ref_count, 1);
830 			} else {
831 				chk->no_fr_allowed = 0;
832 				if (TAILQ_EMPTY(&stcb->asoc.send_queue)) {
833 					chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq;
834 				} else {
835 					chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.TSN_seq;
836 				}
837 			}
838 			/*
839 			 * CMT: Do not allow FRs on retransmitted TSNs.
840 			 */
841 			if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 1) {
842 				chk->no_fr_allowed = 1;
843 			}
844 		} else if (chk->sent == SCTP_DATAGRAM_ACKED) {
845 			/* remember highest acked one */
846 			could_be_sent = chk;
847 		}
848 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
849 			cnt_mk++;
850 		}
851 	}
852 	if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) {
853 		/* we did not subtract the same things? */
854 		audit_tf = 1;
855 	}
856 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
857 		sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT);
858 	}
859 #ifdef SCTP_DEBUG
860 	if (num_mk) {
861 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
862 		    tsnlast);
863 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%ld\n",
864 		    num_mk, (u_long)stcb->asoc.peers_rwnd);
865 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
866 		    tsnlast);
867 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%d\n",
868 		    num_mk,
869 		    (int)stcb->asoc.peers_rwnd);
870 	}
871 #endif
872 	*num_marked = num_mk;
873 	if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) {
874 		/* fix it so we retransmit the highest acked anyway */
875 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
876 		cnt_mk++;
877 		could_be_sent->sent = SCTP_DATAGRAM_RESEND;
878 	}
879 	if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) {
880 #ifdef INVARIANTS
881 		SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n",
882 		    cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk);
883 #endif
884 #ifndef SCTP_AUDITING_ENABLED
885 		stcb->asoc.sent_queue_retran_cnt = cnt_mk;
886 #endif
887 	}
888 	/* Now check for a ECN Echo that may be stranded */
889 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
890 		if ((chk->whoTo == net) &&
891 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
892 			sctp_free_remote_addr(chk->whoTo);
893 			chk->whoTo = alt;
894 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
895 				chk->sent = SCTP_DATAGRAM_RESEND;
896 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
897 			}
898 			atomic_add_int(&alt->ref_count, 1);
899 		}
900 	}
901 	if (audit_tf) {
902 		SCTPDBG(SCTP_DEBUG_TIMER4,
903 		    "Audit total flight due to negative value net:%p\n",
904 		    net);
905 		stcb->asoc.total_flight = 0;
906 		stcb->asoc.total_flight_count = 0;
907 		/* Clear all networks flight size */
908 		TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) {
909 			lnets->flight_size = 0;
910 			SCTPDBG(SCTP_DEBUG_TIMER4,
911 			    "Net:%p c-f cwnd:%d ssthresh:%d\n",
912 			    lnets, lnets->cwnd, lnets->ssthresh);
913 		}
914 		TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
915 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
916 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
917 					sctp_misc_ints(SCTP_FLIGHT_LOG_UP,
918 					    chk->whoTo->flight_size,
919 					    chk->book_size,
920 					    (uintptr_t) chk->whoTo,
921 					    chk->rec.data.TSN_seq);
922 				}
923 				sctp_flight_size_increase(chk);
924 				sctp_total_flight_increase(stcb, chk);
925 			}
926 		}
927 	}
928 	/*
929 	 * Setup the ecn nonce re-sync point. We do this since
930 	 * retranmissions are NOT setup for ECN. This means that do to
931 	 * Karn's rule, we don't know the total of the peers ecn bits.
932 	 */
933 	chk = TAILQ_FIRST(&stcb->asoc.send_queue);
934 	if (chk == NULL) {
935 		stcb->asoc.nonce_resync_tsn = stcb->asoc.sending_seq;
936 	} else {
937 		stcb->asoc.nonce_resync_tsn = chk->rec.data.TSN_seq;
938 	}
939 	stcb->asoc.nonce_wait_for_ecne = 0;
940 	stcb->asoc.nonce_sum_check = 0;
941 	/* We return 1 if we only have a window probe outstanding */
942 	return (0);
943 }
944 
945 static void
946 sctp_move_all_chunks_to_alt(struct sctp_tcb *stcb,
947     struct sctp_nets *net,
948     struct sctp_nets *alt)
949 {
950 	struct sctp_association *asoc;
951 	struct sctp_stream_out *outs;
952 	struct sctp_tmit_chunk *chk;
953 	struct sctp_stream_queue_pending *sp;
954 
955 	if (net == alt)
956 		/* nothing to do */
957 		return;
958 
959 	asoc = &stcb->asoc;
960 
961 	/*
962 	 * now through all the streams checking for chunks sent to our bad
963 	 * network.
964 	 */
965 	TAILQ_FOREACH(outs, &asoc->out_wheel, next_spoke) {
966 		/* now clean up any chunks here */
967 		TAILQ_FOREACH(sp, &outs->outqueue, next) {
968 			if (sp->net == net) {
969 				sctp_free_remote_addr(sp->net);
970 				sp->net = alt;
971 				atomic_add_int(&alt->ref_count, 1);
972 			}
973 		}
974 	}
975 	/* Now check the pending queue */
976 	TAILQ_FOREACH(chk, &asoc->send_queue, sctp_next) {
977 		if (chk->whoTo == net) {
978 			sctp_free_remote_addr(chk->whoTo);
979 			chk->whoTo = alt;
980 			atomic_add_int(&alt->ref_count, 1);
981 		}
982 	}
983 
984 }
985 
986 int
987 sctp_t3rxt_timer(struct sctp_inpcb *inp,
988     struct sctp_tcb *stcb,
989     struct sctp_nets *net)
990 {
991 	struct sctp_nets *alt;
992 	int win_probe, num_mk;
993 
994 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
995 		sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT);
996 	}
997 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
998 		struct sctp_nets *lnet;
999 
1000 		TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
1001 			if (net == lnet) {
1002 				sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3);
1003 			} else {
1004 				sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3);
1005 			}
1006 		}
1007 	}
1008 	/* Find an alternate and mark those for retransmission */
1009 	if ((stcb->asoc.peers_rwnd == 0) &&
1010 	    (stcb->asoc.total_flight < net->mtu)) {
1011 		SCTP_STAT_INCR(sctps_timowindowprobe);
1012 		win_probe = 1;
1013 	} else {
1014 		win_probe = 0;
1015 	}
1016 
1017 	/*
1018 	 * JRS 5/14/07 - If CMT PF is on and the destination if not already
1019 	 * in PF state, set the destination to PF state and store the
1020 	 * current time as the time that the destination was last active. In
1021 	 * addition, find an alternate destination with PF-based
1022 	 * find_alt_net().
1023 	 */
1024 	if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && SCTP_BASE_SYSCTL(sctp_cmt_pf)) {
1025 		if ((net->dest_state & SCTP_ADDR_PF) != SCTP_ADDR_PF) {
1026 			net->dest_state |= SCTP_ADDR_PF;
1027 			net->last_active = sctp_get_tick_count();
1028 			SCTPDBG(SCTP_DEBUG_TIMER4, "Destination %p moved from active to PF.\n",
1029 			    net);
1030 		}
1031 		alt = sctp_find_alternate_net(stcb, net, 2);
1032 	} else if (SCTP_BASE_SYSCTL(sctp_cmt_on_off)) {
1033 		/*
1034 		 * CMT: Using RTX_SSTHRESH policy for CMT. If CMT is being
1035 		 * used, then pick dest with largest ssthresh for any
1036 		 * retransmission.
1037 		 */
1038 		alt = net;
1039 		alt = sctp_find_alternate_net(stcb, alt, 1);
1040 		/*
1041 		 * CUCv2: If a different dest is picked for the
1042 		 * retransmission, then new (rtx-)pseudo_cumack needs to be
1043 		 * tracked for orig dest. Let CUCv2 track new (rtx-)
1044 		 * pseudo-cumack always.
1045 		 */
1046 		net->find_pseudo_cumack = 1;
1047 		net->find_rtx_pseudo_cumack = 1;
1048 	} else {		/* CMT is OFF */
1049 		alt = sctp_find_alternate_net(stcb, net, 0);
1050 	}
1051 
1052 	(void)sctp_mark_all_for_resend(stcb, net, alt, win_probe, &num_mk);
1053 	/* FR Loss recovery just ended with the T3. */
1054 	stcb->asoc.fast_retran_loss_recovery = 0;
1055 
1056 	/* CMT FR loss recovery ended with the T3 */
1057 	net->fast_retran_loss_recovery = 0;
1058 
1059 	/*
1060 	 * setup the sat loss recovery that prevents satellite cwnd advance.
1061 	 */
1062 	stcb->asoc.sat_t3_loss_recovery = 1;
1063 	stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq;
1064 
1065 	/* Backoff the timer and cwnd */
1066 	sctp_backoff_on_timeout(stcb, net, win_probe, num_mk);
1067 	if (win_probe == 0) {
1068 		/* We don't do normal threshold management on window probes */
1069 		if (sctp_threshold_management(inp, stcb, net,
1070 		    stcb->asoc.max_send_times)) {
1071 			/* Association was destroyed */
1072 			return (1);
1073 		} else {
1074 			if (net != stcb->asoc.primary_destination) {
1075 				/* send a immediate HB if our RTO is stale */
1076 				struct timeval now;
1077 				unsigned int ms_goneby;
1078 
1079 				(void)SCTP_GETTIME_TIMEVAL(&now);
1080 				if (net->last_sent_time.tv_sec) {
1081 					ms_goneby = (now.tv_sec - net->last_sent_time.tv_sec) * 1000;
1082 				} else {
1083 					ms_goneby = 0;
1084 				}
1085 				if ((ms_goneby > net->RTO) || (net->RTO == 0)) {
1086 					/*
1087 					 * no recent feed back in an RTO or
1088 					 * more, request a RTT update
1089 					 */
1090 					if (sctp_send_hb(stcb, 1, net) < 0)
1091 						return 1;
1092 				}
1093 			}
1094 		}
1095 	} else {
1096 		/*
1097 		 * For a window probe we don't penalize the net's but only
1098 		 * the association. This may fail it if SACKs are not coming
1099 		 * back. If sack's are coming with rwnd locked at 0, we will
1100 		 * continue to hold things waiting for rwnd to raise
1101 		 */
1102 		if (sctp_threshold_management(inp, stcb, NULL,
1103 		    stcb->asoc.max_send_times)) {
1104 			/* Association was destroyed */
1105 			return (1);
1106 		}
1107 	}
1108 	if (net->dest_state & SCTP_ADDR_NOT_REACHABLE) {
1109 		/* Move all pending over too */
1110 		sctp_move_all_chunks_to_alt(stcb, net, alt);
1111 
1112 		/*
1113 		 * Get the address that failed, to force a new src address
1114 		 * selecton and a route allocation.
1115 		 */
1116 		if (net->ro._s_addr) {
1117 			sctp_free_ifa(net->ro._s_addr);
1118 			net->ro._s_addr = NULL;
1119 		}
1120 		net->src_addr_selected = 0;
1121 
1122 		/* Force a route allocation too */
1123 		if (net->ro.ro_rt) {
1124 			RTFREE(net->ro.ro_rt);
1125 			net->ro.ro_rt = NULL;
1126 		}
1127 		/* Was it our primary? */
1128 		if ((stcb->asoc.primary_destination == net) && (alt != net)) {
1129 			/*
1130 			 * Yes, note it as such and find an alternate note:
1131 			 * this means HB code must use this to resent the
1132 			 * primary if it goes active AND if someone does a
1133 			 * change-primary then this flag must be cleared
1134 			 * from any net structures.
1135 			 */
1136 			if (sctp_set_primary_addr(stcb,
1137 			    (struct sockaddr *)NULL,
1138 			    alt) == 0) {
1139 				net->dest_state |= SCTP_ADDR_WAS_PRIMARY;
1140 			}
1141 		}
1142 	} else if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && SCTP_BASE_SYSCTL(sctp_cmt_pf) && (net->dest_state & SCTP_ADDR_PF) == SCTP_ADDR_PF) {
1143 		/*
1144 		 * JRS 5/14/07 - If the destination hasn't failed completely
1145 		 * but is in PF state, a PF-heartbeat needs to be sent
1146 		 * manually.
1147 		 */
1148 		if (sctp_send_hb(stcb, 1, net) < 0)
1149 			return 1;
1150 	}
1151 	/*
1152 	 * Special case for cookie-echo'ed case, we don't do output but must
1153 	 * await the COOKIE-ACK before retransmission
1154 	 */
1155 	if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
1156 		/*
1157 		 * Here we just reset the timer and start again since we
1158 		 * have not established the asoc
1159 		 */
1160 		sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net);
1161 		return (0);
1162 	}
1163 	if (stcb->asoc.peer_supports_prsctp) {
1164 		struct sctp_tmit_chunk *lchk;
1165 
1166 		lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc);
1167 		/* C3. See if we need to send a Fwd-TSN */
1168 		if (compare_with_wrap(stcb->asoc.advanced_peer_ack_point,
1169 		    stcb->asoc.last_acked_seq, MAX_TSN)) {
1170 			/*
1171 			 * ISSUE with ECN, see FWD-TSN processing for notes
1172 			 * on issues that will occur when the ECN NONCE
1173 			 * stuff is put into SCTP for cross checking.
1174 			 */
1175 			send_forward_tsn(stcb, &stcb->asoc);
1176 			if (lchk) {
1177 				/* Assure a timer is up */
1178 				sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo);
1179 			}
1180 		}
1181 	}
1182 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
1183 		sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX);
1184 	}
1185 	return (0);
1186 }
1187 
1188 int
1189 sctp_t1init_timer(struct sctp_inpcb *inp,
1190     struct sctp_tcb *stcb,
1191     struct sctp_nets *net)
1192 {
1193 	/* bump the thresholds */
1194 	if (stcb->asoc.delayed_connection) {
1195 		/*
1196 		 * special hook for delayed connection. The library did NOT
1197 		 * complete the rest of its sends.
1198 		 */
1199 		stcb->asoc.delayed_connection = 0;
1200 		sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1201 		return (0);
1202 	}
1203 	if (SCTP_GET_STATE((&stcb->asoc)) != SCTP_STATE_COOKIE_WAIT) {
1204 		return (0);
1205 	}
1206 	if (sctp_threshold_management(inp, stcb, net,
1207 	    stcb->asoc.max_init_times)) {
1208 		/* Association was destroyed */
1209 		return (1);
1210 	}
1211 	stcb->asoc.dropped_special_cnt = 0;
1212 	sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0);
1213 	if (stcb->asoc.initial_init_rto_max < net->RTO) {
1214 		net->RTO = stcb->asoc.initial_init_rto_max;
1215 	}
1216 	if (stcb->asoc.numnets > 1) {
1217 		/* If we have more than one addr use it */
1218 		struct sctp_nets *alt;
1219 
1220 		alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0);
1221 		if ((alt != NULL) && (alt != stcb->asoc.primary_destination)) {
1222 			sctp_move_all_chunks_to_alt(stcb, stcb->asoc.primary_destination, alt);
1223 			stcb->asoc.primary_destination = alt;
1224 		}
1225 	}
1226 	/* Send out a new init */
1227 	sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1228 	return (0);
1229 }
1230 
1231 /*
1232  * For cookie and asconf we actually need to find and mark for resend, then
1233  * increment the resend counter (after all the threshold management stuff of
1234  * course).
1235  */
1236 int
1237 sctp_cookie_timer(struct sctp_inpcb *inp,
1238     struct sctp_tcb *stcb,
1239     struct sctp_nets *net)
1240 {
1241 	struct sctp_nets *alt;
1242 	struct sctp_tmit_chunk *cookie;
1243 
1244 	/* first before all else we must find the cookie */
1245 	TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) {
1246 		if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) {
1247 			break;
1248 		}
1249 	}
1250 	if (cookie == NULL) {
1251 		if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
1252 			/* FOOBAR! */
1253 			struct mbuf *oper;
1254 
1255 			oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
1256 			    0, M_DONTWAIT, 1, MT_DATA);
1257 			if (oper) {
1258 				struct sctp_paramhdr *ph;
1259 				uint32_t *ippp;
1260 
1261 				SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
1262 				    sizeof(uint32_t);
1263 				ph = mtod(oper, struct sctp_paramhdr *);
1264 				ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1265 				ph->param_length = htons(SCTP_BUF_LEN(oper));
1266 				ippp = (uint32_t *) (ph + 1);
1267 				*ippp = htonl(SCTP_FROM_SCTP_TIMER + SCTP_LOC_3);
1268 			}
1269 			inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_4;
1270 			sctp_abort_an_association(inp, stcb, SCTP_INTERNAL_ERROR,
1271 			    oper, SCTP_SO_NOT_LOCKED);
1272 		} else {
1273 #ifdef INVARIANTS
1274 			panic("Cookie timer expires in wrong state?");
1275 #else
1276 			SCTP_PRINTF("Strange in state %d not cookie-echoed yet c-e timer expires?\n", SCTP_GET_STATE(&stcb->asoc));
1277 			return (0);
1278 #endif
1279 		}
1280 		return (0);
1281 	}
1282 	/* Ok we found the cookie, threshold management next */
1283 	if (sctp_threshold_management(inp, stcb, cookie->whoTo,
1284 	    stcb->asoc.max_init_times)) {
1285 		/* Assoc is over */
1286 		return (1);
1287 	}
1288 	/*
1289 	 * cleared theshold management now lets backoff the address & select
1290 	 * an alternate
1291 	 */
1292 	stcb->asoc.dropped_special_cnt = 0;
1293 	sctp_backoff_on_timeout(stcb, cookie->whoTo, 1, 0);
1294 	alt = sctp_find_alternate_net(stcb, cookie->whoTo, 0);
1295 	if (alt != cookie->whoTo) {
1296 		sctp_free_remote_addr(cookie->whoTo);
1297 		cookie->whoTo = alt;
1298 		atomic_add_int(&alt->ref_count, 1);
1299 	}
1300 	/* Now mark the retran info */
1301 	if (cookie->sent != SCTP_DATAGRAM_RESEND) {
1302 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1303 	}
1304 	cookie->sent = SCTP_DATAGRAM_RESEND;
1305 	/*
1306 	 * Now call the output routine to kick out the cookie again, Note we
1307 	 * don't mark any chunks for retran so that FR will need to kick in
1308 	 * to move these (or a send timer).
1309 	 */
1310 	return (0);
1311 }
1312 
1313 int
1314 sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1315     struct sctp_nets *net)
1316 {
1317 	struct sctp_nets *alt;
1318 	struct sctp_tmit_chunk *strrst = NULL, *chk = NULL;
1319 
1320 	if (stcb->asoc.stream_reset_outstanding == 0) {
1321 		return (0);
1322 	}
1323 	/* find the existing STRRESET, we use the seq number we sent out on */
1324 	(void)sctp_find_stream_reset(stcb, stcb->asoc.str_reset_seq_out, &strrst);
1325 	if (strrst == NULL) {
1326 		return (0);
1327 	}
1328 	/* do threshold management */
1329 	if (sctp_threshold_management(inp, stcb, strrst->whoTo,
1330 	    stcb->asoc.max_send_times)) {
1331 		/* Assoc is over */
1332 		return (1);
1333 	}
1334 	/*
1335 	 * cleared theshold management now lets backoff the address & select
1336 	 * an alternate
1337 	 */
1338 	sctp_backoff_on_timeout(stcb, strrst->whoTo, 1, 0);
1339 	alt = sctp_find_alternate_net(stcb, strrst->whoTo, 0);
1340 	sctp_free_remote_addr(strrst->whoTo);
1341 	strrst->whoTo = alt;
1342 	atomic_add_int(&alt->ref_count, 1);
1343 
1344 	/* See if a ECN Echo is also stranded */
1345 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1346 		if ((chk->whoTo == net) &&
1347 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1348 			sctp_free_remote_addr(chk->whoTo);
1349 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
1350 				chk->sent = SCTP_DATAGRAM_RESEND;
1351 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1352 			}
1353 			chk->whoTo = alt;
1354 			atomic_add_int(&alt->ref_count, 1);
1355 		}
1356 	}
1357 	if (net->dest_state & SCTP_ADDR_NOT_REACHABLE) {
1358 		/*
1359 		 * If the address went un-reachable, we need to move to
1360 		 * alternates for ALL chk's in queue
1361 		 */
1362 		sctp_move_all_chunks_to_alt(stcb, net, alt);
1363 	}
1364 	/* mark the retran info */
1365 	if (strrst->sent != SCTP_DATAGRAM_RESEND)
1366 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1367 	strrst->sent = SCTP_DATAGRAM_RESEND;
1368 
1369 	/* restart the timer */
1370 	sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, strrst->whoTo);
1371 	return (0);
1372 }
1373 
1374 int
1375 sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1376     struct sctp_nets *net)
1377 {
1378 	struct sctp_nets *alt;
1379 	struct sctp_tmit_chunk *asconf, *chk, *nchk;
1380 
1381 	/* is this a first send, or a retransmission? */
1382 	if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) {
1383 		/* compose a new ASCONF chunk and send it */
1384 		sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED);
1385 	} else {
1386 		/*
1387 		 * Retransmission of the existing ASCONF is needed
1388 		 */
1389 
1390 		/* find the existing ASCONF */
1391 		asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue);
1392 		if (asconf == NULL) {
1393 			return (0);
1394 		}
1395 		/* do threshold management */
1396 		if (sctp_threshold_management(inp, stcb, asconf->whoTo,
1397 		    stcb->asoc.max_send_times)) {
1398 			/* Assoc is over */
1399 			return (1);
1400 		}
1401 		if (asconf->snd_count > stcb->asoc.max_send_times) {
1402 			/*
1403 			 * Something is rotten: our peer is not responding
1404 			 * to ASCONFs but apparently is to other chunks.
1405 			 * i.e. it is not properly handling the chunk type
1406 			 * upper bits. Mark this peer as ASCONF incapable
1407 			 * and cleanup.
1408 			 */
1409 			SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n");
1410 			sctp_asconf_cleanup(stcb, net);
1411 			return (0);
1412 		}
1413 		/*
1414 		 * cleared threshold management, so now backoff the net and
1415 		 * select an alternate
1416 		 */
1417 		sctp_backoff_on_timeout(stcb, asconf->whoTo, 1, 0);
1418 		alt = sctp_find_alternate_net(stcb, asconf->whoTo, 0);
1419 		if (asconf->whoTo != alt) {
1420 			sctp_free_remote_addr(asconf->whoTo);
1421 			asconf->whoTo = alt;
1422 			atomic_add_int(&alt->ref_count, 1);
1423 		}
1424 		/* See if an ECN Echo is also stranded */
1425 		TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1426 			if ((chk->whoTo == net) &&
1427 			    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1428 				sctp_free_remote_addr(chk->whoTo);
1429 				chk->whoTo = alt;
1430 				if (chk->sent != SCTP_DATAGRAM_RESEND) {
1431 					chk->sent = SCTP_DATAGRAM_RESEND;
1432 					sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1433 				}
1434 				atomic_add_int(&alt->ref_count, 1);
1435 			}
1436 		}
1437 		for (chk = asconf; chk; chk = nchk) {
1438 			nchk = TAILQ_NEXT(chk, sctp_next);
1439 			if (chk->whoTo != alt) {
1440 				sctp_free_remote_addr(chk->whoTo);
1441 				chk->whoTo = alt;
1442 				atomic_add_int(&alt->ref_count, 1);
1443 			}
1444 			if (asconf->sent != SCTP_DATAGRAM_RESEND && chk->sent != SCTP_DATAGRAM_UNSENT)
1445 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1446 			chk->sent = SCTP_DATAGRAM_RESEND;
1447 		}
1448 		if (net->dest_state & SCTP_ADDR_NOT_REACHABLE) {
1449 			/*
1450 			 * If the address went un-reachable, we need to move
1451 			 * to the alternate for ALL chunks in queue
1452 			 */
1453 			sctp_move_all_chunks_to_alt(stcb, net, alt);
1454 			net = alt;
1455 		}
1456 		/* mark the retran info */
1457 		if (asconf->sent != SCTP_DATAGRAM_RESEND)
1458 			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1459 		asconf->sent = SCTP_DATAGRAM_RESEND;
1460 
1461 		/* send another ASCONF if any and we can do */
1462 		sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED);
1463 	}
1464 	return (0);
1465 }
1466 
1467 /* Mobility adaptation */
1468 void
1469 sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1470     struct sctp_nets *net)
1471 {
1472 	if (stcb->asoc.deleted_primary == NULL) {
1473 		SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n");
1474 		sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1475 		return;
1476 	}
1477 	SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary ");
1478 	SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa);
1479 	sctp_free_remote_addr(stcb->asoc.deleted_primary);
1480 	stcb->asoc.deleted_primary = NULL;
1481 	sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1482 	return;
1483 }
1484 
1485 /*
1486  * For the shutdown and shutdown-ack, we do not keep one around on the
1487  * control queue. This means we must generate a new one and call the general
1488  * chunk output routine, AFTER having done threshold management.
1489  */
1490 int
1491 sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1492     struct sctp_nets *net)
1493 {
1494 	struct sctp_nets *alt;
1495 
1496 	/* first threshold managment */
1497 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1498 		/* Assoc is over */
1499 		return (1);
1500 	}
1501 	/* second select an alternative */
1502 	alt = sctp_find_alternate_net(stcb, net, 0);
1503 
1504 	/* third generate a shutdown into the queue for out net */
1505 	if (alt) {
1506 		sctp_send_shutdown(stcb, alt);
1507 	} else {
1508 		/*
1509 		 * if alt is NULL, there is no dest to send to??
1510 		 */
1511 		return (0);
1512 	}
1513 	/* fourth restart timer */
1514 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt);
1515 	return (0);
1516 }
1517 
1518 int
1519 sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1520     struct sctp_nets *net)
1521 {
1522 	struct sctp_nets *alt;
1523 
1524 	/* first threshold managment */
1525 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1526 		/* Assoc is over */
1527 		return (1);
1528 	}
1529 	/* second select an alternative */
1530 	alt = sctp_find_alternate_net(stcb, net, 0);
1531 
1532 	/* third generate a shutdown into the queue for out net */
1533 	sctp_send_shutdown_ack(stcb, alt);
1534 
1535 	/* fourth restart timer */
1536 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt);
1537 	return (0);
1538 }
1539 
1540 static void
1541 sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp,
1542     struct sctp_tcb *stcb)
1543 {
1544 	struct sctp_stream_out *outs;
1545 	struct sctp_stream_queue_pending *sp;
1546 	unsigned int chks_in_queue = 0;
1547 	int being_filled = 0;
1548 
1549 	/*
1550 	 * This function is ONLY called when the send/sent queues are empty.
1551 	 */
1552 	if ((stcb == NULL) || (inp == NULL))
1553 		return;
1554 
1555 	if (stcb->asoc.sent_queue_retran_cnt) {
1556 		SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n",
1557 		    stcb->asoc.sent_queue_retran_cnt);
1558 		stcb->asoc.sent_queue_retran_cnt = 0;
1559 	}
1560 	SCTP_TCB_SEND_LOCK(stcb);
1561 	if (TAILQ_EMPTY(&stcb->asoc.out_wheel)) {
1562 		int i, cnt = 0;
1563 
1564 		/* Check to see if a spoke fell off the wheel */
1565 		for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
1566 			if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) {
1567 				sctp_insert_on_wheel(stcb, &stcb->asoc, &stcb->asoc.strmout[i], 1);
1568 				cnt++;
1569 			}
1570 		}
1571 		if (cnt) {
1572 			/* yep, we lost a spoke or two */
1573 			SCTP_PRINTF("Found an additional %d streams NOT on outwheel, corrected\n", cnt);
1574 		} else {
1575 			/* no spokes lost, */
1576 			stcb->asoc.total_output_queue_size = 0;
1577 		}
1578 		SCTP_TCB_SEND_UNLOCK(stcb);
1579 		return;
1580 	}
1581 	SCTP_TCB_SEND_UNLOCK(stcb);
1582 	/* Check to see if some data queued, if so report it */
1583 	TAILQ_FOREACH(outs, &stcb->asoc.out_wheel, next_spoke) {
1584 		if (!TAILQ_EMPTY(&outs->outqueue)) {
1585 			TAILQ_FOREACH(sp, &outs->outqueue, next) {
1586 				if (sp->msg_is_complete)
1587 					being_filled++;
1588 				chks_in_queue++;
1589 			}
1590 		}
1591 	}
1592 	if (chks_in_queue != stcb->asoc.stream_queue_cnt) {
1593 		SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n",
1594 		    stcb->asoc.stream_queue_cnt, chks_in_queue);
1595 	}
1596 	if (chks_in_queue) {
1597 		/* call the output queue function */
1598 		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1599 		if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1600 		    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1601 			/*
1602 			 * Probably should go in and make it go back through
1603 			 * and add fragments allowed
1604 			 */
1605 			if (being_filled == 0) {
1606 				SCTP_PRINTF("Still nothing moved %d chunks are stuck\n",
1607 				    chks_in_queue);
1608 			}
1609 		}
1610 	} else {
1611 		SCTP_PRINTF("Found no chunks on any queue tot:%lu\n",
1612 		    (u_long)stcb->asoc.total_output_queue_size);
1613 		stcb->asoc.total_output_queue_size = 0;
1614 	}
1615 }
1616 
1617 int
1618 sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1619     struct sctp_nets *net, int cnt_of_unconf)
1620 {
1621 	int ret;
1622 
1623 	if (net) {
1624 		if (net->hb_responded == 0) {
1625 			if (net->ro._s_addr) {
1626 				/*
1627 				 * Invalidate the src address if we did not
1628 				 * get a response last time.
1629 				 */
1630 				sctp_free_ifa(net->ro._s_addr);
1631 				net->ro._s_addr = NULL;
1632 				net->src_addr_selected = 0;
1633 			}
1634 			sctp_backoff_on_timeout(stcb, net, 1, 0);
1635 		}
1636 		/* Zero PBA, if it needs it */
1637 		if (net->partial_bytes_acked) {
1638 			net->partial_bytes_acked = 0;
1639 		}
1640 	}
1641 	if ((stcb->asoc.total_output_queue_size > 0) &&
1642 	    (TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1643 	    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1644 		sctp_audit_stream_queues_for_size(inp, stcb);
1645 	}
1646 	/* Send a new HB, this will do threshold managment, pick a new dest */
1647 	if (cnt_of_unconf == 0) {
1648 		if (sctp_send_hb(stcb, 0, NULL) < 0) {
1649 			return (1);
1650 		}
1651 	} else {
1652 		/*
1653 		 * this will send out extra hb's up to maxburst if there are
1654 		 * any unconfirmed addresses.
1655 		 */
1656 		uint32_t cnt_sent = 0;
1657 
1658 		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1659 			if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) &&
1660 			    (net->dest_state & SCTP_ADDR_REACHABLE)) {
1661 				cnt_sent++;
1662 				if (net->hb_responded == 0) {
1663 					/* Did we respond last time? */
1664 					if (net->ro._s_addr) {
1665 						sctp_free_ifa(net->ro._s_addr);
1666 						net->ro._s_addr = NULL;
1667 						net->src_addr_selected = 0;
1668 					}
1669 				}
1670 				ret = sctp_send_hb(stcb, 1, net);
1671 				if (ret < 0)
1672 					return 1;
1673 				else if (ret == 0) {
1674 					break;
1675 				}
1676 				if (cnt_sent >= SCTP_BASE_SYSCTL(sctp_hb_maxburst))
1677 					break;
1678 			}
1679 		}
1680 	}
1681 	return (0);
1682 }
1683 
1684 int
1685 sctp_is_hb_timer_running(struct sctp_tcb *stcb)
1686 {
1687 	if (SCTP_OS_TIMER_PENDING(&stcb->asoc.hb_timer.timer)) {
1688 		/* its running */
1689 		return (1);
1690 	} else {
1691 		/* nope */
1692 		return (0);
1693 	}
1694 }
1695 
1696 int
1697 sctp_is_sack_timer_running(struct sctp_tcb *stcb)
1698 {
1699 	if (SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) {
1700 		/* its running */
1701 		return (1);
1702 	} else {
1703 		/* nope */
1704 		return (0);
1705 	}
1706 }
1707 
1708 #define SCTP_NUMBER_OF_MTU_SIZES 18
1709 static uint32_t mtu_sizes[] = {
1710 	68,
1711 	296,
1712 	508,
1713 	512,
1714 	544,
1715 	576,
1716 	1006,
1717 	1492,
1718 	1500,
1719 	1536,
1720 	2002,
1721 	2048,
1722 	4352,
1723 	4464,
1724 	8166,
1725 	17914,
1726 	32000,
1727 	65535
1728 };
1729 
1730 
1731 static uint32_t
1732 sctp_getnext_mtu(struct sctp_inpcb *inp, uint32_t cur_mtu)
1733 {
1734 	/* select another MTU that is just bigger than this one */
1735 	int i;
1736 
1737 	for (i = 0; i < SCTP_NUMBER_OF_MTU_SIZES; i++) {
1738 		if (cur_mtu < mtu_sizes[i]) {
1739 			/* no max_mtu is bigger than this one */
1740 			return (mtu_sizes[i]);
1741 		}
1742 	}
1743 	/* here return the highest allowable */
1744 	return (cur_mtu);
1745 }
1746 
1747 
1748 void
1749 sctp_pathmtu_timer(struct sctp_inpcb *inp,
1750     struct sctp_tcb *stcb,
1751     struct sctp_nets *net)
1752 {
1753 	uint32_t next_mtu, mtu;
1754 
1755 	next_mtu = sctp_getnext_mtu(inp, net->mtu);
1756 
1757 	if ((next_mtu > net->mtu) && (net->port == 0)) {
1758 		if ((net->src_addr_selected == 0) ||
1759 		    (net->ro._s_addr == NULL) ||
1760 		    (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1761 			if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1762 				sctp_free_ifa(net->ro._s_addr);
1763 				net->ro._s_addr = NULL;
1764 				net->src_addr_selected = 0;
1765 			} else if (net->ro._s_addr == NULL) {
1766 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1767 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1768 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1769 
1770 					/* KAME hack: embed scopeid */
1771 					(void)sa6_embedscope(sin6, MODULE_GLOBAL(MOD_INET6, ip6_use_defzone));
1772 				}
1773 #endif
1774 
1775 				net->ro._s_addr = sctp_source_address_selection(inp,
1776 				    stcb,
1777 				    (sctp_route_t *) & net->ro,
1778 				    net, 0, stcb->asoc.vrf_id);
1779 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1780 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1781 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1782 
1783 					(void)sa6_recoverscope(sin6);
1784 				}
1785 #endif				/* INET6 */
1786 			}
1787 			if (net->ro._s_addr)
1788 				net->src_addr_selected = 1;
1789 		}
1790 		if (net->ro._s_addr) {
1791 			mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_rt);
1792 			if (mtu > next_mtu) {
1793 				net->mtu = next_mtu;
1794 			}
1795 		}
1796 	}
1797 	/* restart the timer */
1798 	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
1799 }
1800 
1801 void
1802 sctp_autoclose_timer(struct sctp_inpcb *inp,
1803     struct sctp_tcb *stcb,
1804     struct sctp_nets *net)
1805 {
1806 	struct timeval tn, *tim_touse;
1807 	struct sctp_association *asoc;
1808 	int ticks_gone_by;
1809 
1810 	(void)SCTP_GETTIME_TIMEVAL(&tn);
1811 	if (stcb->asoc.sctp_autoclose_ticks &&
1812 	    sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) {
1813 		/* Auto close is on */
1814 		asoc = &stcb->asoc;
1815 		/* pick the time to use */
1816 		if (asoc->time_last_rcvd.tv_sec >
1817 		    asoc->time_last_sent.tv_sec) {
1818 			tim_touse = &asoc->time_last_rcvd;
1819 		} else {
1820 			tim_touse = &asoc->time_last_sent;
1821 		}
1822 		/* Now has long enough transpired to autoclose? */
1823 		ticks_gone_by = SEC_TO_TICKS(tn.tv_sec - tim_touse->tv_sec);
1824 		if ((ticks_gone_by > 0) &&
1825 		    (ticks_gone_by >= (int)asoc->sctp_autoclose_ticks)) {
1826 			/*
1827 			 * autoclose time has hit, call the output routine,
1828 			 * which should do nothing just to be SURE we don't
1829 			 * have hanging data. We can then safely check the
1830 			 * queues and know that we are clear to send
1831 			 * shutdown
1832 			 */
1833 			sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED);
1834 			/* Are we clean? */
1835 			if (TAILQ_EMPTY(&asoc->send_queue) &&
1836 			    TAILQ_EMPTY(&asoc->sent_queue)) {
1837 				/*
1838 				 * there is nothing queued to send, so I'm
1839 				 * done...
1840 				 */
1841 				if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) {
1842 					/* only send SHUTDOWN 1st time thru */
1843 					sctp_send_shutdown(stcb, stcb->asoc.primary_destination);
1844 					if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
1845 					    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
1846 						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
1847 					}
1848 					SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
1849 					SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
1850 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
1851 					    stcb->sctp_ep, stcb,
1852 					    asoc->primary_destination);
1853 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
1854 					    stcb->sctp_ep, stcb,
1855 					    asoc->primary_destination);
1856 				}
1857 			}
1858 		} else {
1859 			/*
1860 			 * No auto close at this time, reset t-o to check
1861 			 * later
1862 			 */
1863 			int tmp;
1864 
1865 			/* fool the timer startup to use the time left */
1866 			tmp = asoc->sctp_autoclose_ticks;
1867 			asoc->sctp_autoclose_ticks -= ticks_gone_by;
1868 			sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb,
1869 			    net);
1870 			/* restore the real tick value */
1871 			asoc->sctp_autoclose_ticks = tmp;
1872 		}
1873 	}
1874 }
1875 
1876 void
1877 sctp_iterator_timer(struct sctp_iterator *it)
1878 {
1879 	int iteration_count = 0;
1880 	int inp_skip = 0;
1881 
1882 	/*
1883 	 * only one iterator can run at a time. This is the only way we can
1884 	 * cleanly pull ep's from underneath all the running interators when
1885 	 * a ep is freed.
1886 	 */
1887 	SCTP_ITERATOR_LOCK();
1888 	if (it->inp == NULL) {
1889 		/* iterator is complete */
1890 done_with_iterator:
1891 		SCTP_ITERATOR_UNLOCK();
1892 		SCTP_INP_INFO_WLOCK();
1893 		TAILQ_REMOVE(&SCTP_BASE_INFO(iteratorhead), it, sctp_nxt_itr);
1894 		/* stopping the callout is not needed, in theory */
1895 		SCTP_INP_INFO_WUNLOCK();
1896 		(void)SCTP_OS_TIMER_STOP(&it->tmr.timer);
1897 		if (it->function_atend != NULL) {
1898 			(*it->function_atend) (it->pointer, it->val);
1899 		}
1900 		SCTP_FREE(it, SCTP_M_ITER);
1901 		return;
1902 	}
1903 select_a_new_ep:
1904 	SCTP_INP_WLOCK(it->inp);
1905 	while (((it->pcb_flags) &&
1906 	    ((it->inp->sctp_flags & it->pcb_flags) != it->pcb_flags)) ||
1907 	    ((it->pcb_features) &&
1908 	    ((it->inp->sctp_features & it->pcb_features) != it->pcb_features))) {
1909 		/* endpoint flags or features don't match, so keep looking */
1910 		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
1911 			SCTP_INP_WUNLOCK(it->inp);
1912 			goto done_with_iterator;
1913 		}
1914 		SCTP_INP_WUNLOCK(it->inp);
1915 		it->inp = LIST_NEXT(it->inp, sctp_list);
1916 		if (it->inp == NULL) {
1917 			goto done_with_iterator;
1918 		}
1919 		SCTP_INP_WLOCK(it->inp);
1920 	}
1921 	if ((it->inp->inp_starting_point_for_iterator != NULL) &&
1922 	    (it->inp->inp_starting_point_for_iterator != it)) {
1923 		SCTP_PRINTF("Iterator collision, waiting for one at %p\n",
1924 		    it->inp);
1925 		SCTP_INP_WUNLOCK(it->inp);
1926 		goto start_timer_return;
1927 	}
1928 	/* mark the current iterator on the endpoint */
1929 	it->inp->inp_starting_point_for_iterator = it;
1930 	SCTP_INP_WUNLOCK(it->inp);
1931 	SCTP_INP_RLOCK(it->inp);
1932 	/* now go through each assoc which is in the desired state */
1933 	if (it->done_current_ep == 0) {
1934 		if (it->function_inp != NULL)
1935 			inp_skip = (*it->function_inp) (it->inp, it->pointer, it->val);
1936 		it->done_current_ep = 1;
1937 	}
1938 	if (it->stcb == NULL) {
1939 		/* run the per instance function */
1940 		it->stcb = LIST_FIRST(&it->inp->sctp_asoc_list);
1941 	}
1942 	SCTP_INP_RUNLOCK(it->inp);
1943 	if ((inp_skip) || it->stcb == NULL) {
1944 		if (it->function_inp_end != NULL) {
1945 			inp_skip = (*it->function_inp_end) (it->inp,
1946 			    it->pointer,
1947 			    it->val);
1948 		}
1949 		goto no_stcb;
1950 	}
1951 	if ((it->stcb) &&
1952 	    (it->stcb->asoc.stcb_starting_point_for_iterator == it)) {
1953 		it->stcb->asoc.stcb_starting_point_for_iterator = NULL;
1954 	}
1955 	while (it->stcb) {
1956 		SCTP_TCB_LOCK(it->stcb);
1957 		if (it->asoc_state && ((it->stcb->asoc.state & it->asoc_state) != it->asoc_state)) {
1958 			/* not in the right state... keep looking */
1959 			SCTP_TCB_UNLOCK(it->stcb);
1960 			goto next_assoc;
1961 		}
1962 		/* mark the current iterator on the assoc */
1963 		it->stcb->asoc.stcb_starting_point_for_iterator = it;
1964 		/* see if we have limited out the iterator loop */
1965 		iteration_count++;
1966 		if (iteration_count > SCTP_ITERATOR_MAX_AT_ONCE) {
1967 	start_timer_return:
1968 			/* set a timer to continue this later */
1969 			if (it->stcb)
1970 				SCTP_TCB_UNLOCK(it->stcb);
1971 			sctp_timer_start(SCTP_TIMER_TYPE_ITERATOR,
1972 			    (struct sctp_inpcb *)it, NULL, NULL);
1973 			SCTP_ITERATOR_UNLOCK();
1974 			return;
1975 		}
1976 		/* run function on this one */
1977 		(*it->function_assoc) (it->inp, it->stcb, it->pointer, it->val);
1978 
1979 		/*
1980 		 * we lie here, it really needs to have its own type but
1981 		 * first I must verify that this won't effect things :-0
1982 		 */
1983 		if (it->no_chunk_output == 0)
1984 			sctp_chunk_output(it->inp, it->stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1985 
1986 		SCTP_TCB_UNLOCK(it->stcb);
1987 next_assoc:
1988 		it->stcb = LIST_NEXT(it->stcb, sctp_tcblist);
1989 		if (it->stcb == NULL) {
1990 			if (it->function_inp_end != NULL) {
1991 				inp_skip = (*it->function_inp_end) (it->inp,
1992 				    it->pointer,
1993 				    it->val);
1994 			}
1995 		}
1996 	}
1997 no_stcb:
1998 	/* done with all assocs on this endpoint, move on to next endpoint */
1999 	it->done_current_ep = 0;
2000 	SCTP_INP_WLOCK(it->inp);
2001 	it->inp->inp_starting_point_for_iterator = NULL;
2002 	SCTP_INP_WUNLOCK(it->inp);
2003 	if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
2004 		it->inp = NULL;
2005 	} else {
2006 		SCTP_INP_INFO_RLOCK();
2007 		it->inp = LIST_NEXT(it->inp, sctp_list);
2008 		SCTP_INP_INFO_RUNLOCK();
2009 	}
2010 	if (it->inp == NULL) {
2011 		goto done_with_iterator;
2012 	}
2013 	goto select_a_new_ep;
2014 }
2015