xref: /freebsd/sys/netinet/sctp_timer.c (revision 145992504973bd16cf3518af9ba5ce185fefa82a)
1 /*-
2  * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved.
3  * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved.
4  * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions are met:
8  *
9  * a) Redistributions of source code must retain the above copyright notice,
10  *    this list of conditions and the following disclaimer.
11  *
12  * b) Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in
14  *    the documentation and/or other materials provided with the distribution.
15  *
16  * c) Neither the name of Cisco Systems, Inc. nor the names of its
17  *    contributors may be used to endorse or promote products derived
18  *    from this software without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
22  * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
24  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
30  * THE POSSIBILITY OF SUCH DAMAGE.
31  */
32 
33 #include <sys/cdefs.h>
34 __FBSDID("$FreeBSD$");
35 
36 #define _IP_VHL
37 #include <netinet/sctp_os.h>
38 #include <netinet/sctp_pcb.h>
39 #ifdef INET6
40 #endif
41 #include <netinet/sctp_var.h>
42 #include <netinet/sctp_sysctl.h>
43 #include <netinet/sctp_timer.h>
44 #include <netinet/sctputil.h>
45 #include <netinet/sctp_output.h>
46 #include <netinet/sctp_header.h>
47 #include <netinet/sctp_indata.h>
48 #include <netinet/sctp_asconf.h>
49 #include <netinet/sctp_input.h>
50 #include <netinet/sctp.h>
51 #include <netinet/sctp_uio.h>
52 #include <netinet/udp.h>
53 
54 
55 void
56 sctp_audit_retranmission_queue(struct sctp_association *asoc)
57 {
58 	struct sctp_tmit_chunk *chk;
59 
60 	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit invoked on send queue cnt:%d onqueue:%d\n",
61 	    asoc->sent_queue_retran_cnt,
62 	    asoc->sent_queue_cnt);
63 	asoc->sent_queue_retran_cnt = 0;
64 	asoc->sent_queue_cnt = 0;
65 	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
66 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
67 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
68 		}
69 		asoc->sent_queue_cnt++;
70 	}
71 	TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) {
72 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
73 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
74 		}
75 	}
76 	TAILQ_FOREACH(chk, &asoc->asconf_send_queue, sctp_next) {
77 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
78 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
79 		}
80 	}
81 	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit completes retran:%d onqueue:%d\n",
82 	    asoc->sent_queue_retran_cnt,
83 	    asoc->sent_queue_cnt);
84 }
85 
86 int
87 sctp_threshold_management(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
88     struct sctp_nets *net, uint16_t threshold)
89 {
90 	if (net) {
91 		net->error_count++;
92 		SCTPDBG(SCTP_DEBUG_TIMER4, "Error count for %p now %d thresh:%d\n",
93 		    (void *)net, net->error_count,
94 		    net->failure_threshold);
95 		if (net->error_count > net->failure_threshold) {
96 			/* We had a threshold failure */
97 			if (net->dest_state & SCTP_ADDR_REACHABLE) {
98 				net->dest_state &= ~SCTP_ADDR_REACHABLE;
99 				net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY;
100 				net->dest_state &= ~SCTP_ADDR_PF;
101 				sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN,
102 				    stcb, 0,
103 				    (void *)net, SCTP_SO_NOT_LOCKED);
104 			}
105 		} else if ((net->pf_threshold < net->failure_threshold) &&
106 		    (net->error_count > net->pf_threshold)) {
107 			if (!(net->dest_state & SCTP_ADDR_PF)) {
108 				net->dest_state |= SCTP_ADDR_PF;
109 				net->last_active = sctp_get_tick_count();
110 				sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
111 				sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_TIMER + SCTP_LOC_3);
112 				sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net);
113 			}
114 		}
115 	}
116 	if (stcb == NULL)
117 		return (0);
118 
119 	if (net) {
120 		if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) {
121 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
122 				sctp_misc_ints(SCTP_THRESHOLD_INCR,
123 				    stcb->asoc.overall_error_count,
124 				    (stcb->asoc.overall_error_count + 1),
125 				    SCTP_FROM_SCTP_TIMER,
126 				    __LINE__);
127 			}
128 			stcb->asoc.overall_error_count++;
129 		}
130 	} else {
131 		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
132 			sctp_misc_ints(SCTP_THRESHOLD_INCR,
133 			    stcb->asoc.overall_error_count,
134 			    (stcb->asoc.overall_error_count + 1),
135 			    SCTP_FROM_SCTP_TIMER,
136 			    __LINE__);
137 		}
138 		stcb->asoc.overall_error_count++;
139 	}
140 	SCTPDBG(SCTP_DEBUG_TIMER4, "Overall error count for %p now %d thresh:%u state:%x\n",
141 	    (void *)&stcb->asoc, stcb->asoc.overall_error_count,
142 	    (uint32_t) threshold,
143 	    ((net == NULL) ? (uint32_t) 0 : (uint32_t) net->dest_state));
144 	/*
145 	 * We specifically do not do >= to give the assoc one more change
146 	 * before we fail it.
147 	 */
148 	if (stcb->asoc.overall_error_count > threshold) {
149 		/* Abort notification sends a ULP notify */
150 		struct mbuf *oper;
151 
152 		oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
153 		    0, M_DONTWAIT, 1, MT_DATA);
154 		if (oper) {
155 			struct sctp_paramhdr *ph;
156 			uint32_t *ippp;
157 
158 			SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
159 			    sizeof(uint32_t);
160 			ph = mtod(oper, struct sctp_paramhdr *);
161 			ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
162 			ph->param_length = htons(SCTP_BUF_LEN(oper));
163 			ippp = (uint32_t *) (ph + 1);
164 			*ippp = htonl(SCTP_FROM_SCTP_TIMER + SCTP_LOC_1);
165 		}
166 		inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_1;
167 		sctp_abort_an_association(inp, stcb, oper, SCTP_SO_NOT_LOCKED);
168 		return (1);
169 	}
170 	return (0);
171 }
172 
173 /*
174  * sctp_find_alternate_net() returns a non-NULL pointer as long
175  * the argument net is non-NULL.
176  */
177 struct sctp_nets *
178 sctp_find_alternate_net(struct sctp_tcb *stcb,
179     struct sctp_nets *net,
180     int mode)
181 {
182 	/* Find and return an alternate network if possible */
183 	struct sctp_nets *alt, *mnet, *min_errors_net = NULL, *max_cwnd_net = NULL;
184 	int once;
185 
186 	/* JRS 5/14/07 - Initialize min_errors to an impossible value. */
187 	int min_errors = -1;
188 	uint32_t max_cwnd = 0;
189 
190 	if (stcb->asoc.numnets == 1) {
191 		/* No others but net */
192 		return (TAILQ_FIRST(&stcb->asoc.nets));
193 	}
194 	/*
195 	 * JRS 5/14/07 - If mode is set to 2, use the CMT PF find alternate
196 	 * net algorithm. This algorithm chooses the active destination (not
197 	 * in PF state) with the largest cwnd value. If all destinations are
198 	 * in PF state, unreachable, or unconfirmed, choose the desination
199 	 * that is in PF state with the lowest error count. In case of a
200 	 * tie, choose the destination that was most recently active.
201 	 */
202 	if (mode == 2) {
203 		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
204 			/*
205 			 * JRS 5/14/07 - If the destination is unreachable
206 			 * or unconfirmed, skip it.
207 			 */
208 			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
209 			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
210 				continue;
211 			}
212 			/*
213 			 * JRS 5/14/07 -  If the destination is reachable
214 			 * but in PF state, compare the error count of the
215 			 * destination to the minimum error count seen thus
216 			 * far. Store the destination with the lower error
217 			 * count.  If the error counts are equal, store the
218 			 * destination that was most recently active.
219 			 */
220 			if (mnet->dest_state & SCTP_ADDR_PF) {
221 				/*
222 				 * JRS 5/14/07 - If the destination under
223 				 * consideration is the current destination,
224 				 * work as if the error count is one higher.
225 				 * The actual error count will not be
226 				 * incremented until later in the t3
227 				 * handler.
228 				 */
229 				if (mnet == net) {
230 					if (min_errors == -1) {
231 						min_errors = mnet->error_count + 1;
232 						min_errors_net = mnet;
233 					} else if (mnet->error_count + 1 < min_errors) {
234 						min_errors = mnet->error_count + 1;
235 						min_errors_net = mnet;
236 					} else if (mnet->error_count + 1 == min_errors
237 					    && mnet->last_active > min_errors_net->last_active) {
238 						min_errors_net = mnet;
239 						min_errors = mnet->error_count + 1;
240 					}
241 					continue;
242 				} else {
243 					if (min_errors == -1) {
244 						min_errors = mnet->error_count;
245 						min_errors_net = mnet;
246 					} else if (mnet->error_count < min_errors) {
247 						min_errors = mnet->error_count;
248 						min_errors_net = mnet;
249 					} else if (mnet->error_count == min_errors
250 					    && mnet->last_active > min_errors_net->last_active) {
251 						min_errors_net = mnet;
252 						min_errors = mnet->error_count;
253 					}
254 					continue;
255 				}
256 			}
257 			/*
258 			 * JRS 5/14/07 - If the destination is reachable and
259 			 * not in PF state, compare the cwnd of the
260 			 * destination to the highest cwnd seen thus far.
261 			 * Store the destination with the higher cwnd value.
262 			 * If the cwnd values are equal, randomly choose one
263 			 * of the two destinations.
264 			 */
265 			if (max_cwnd < mnet->cwnd) {
266 				max_cwnd_net = mnet;
267 				max_cwnd = mnet->cwnd;
268 			} else if (max_cwnd == mnet->cwnd) {
269 				uint32_t rndval;
270 				uint8_t this_random;
271 
272 				if (stcb->asoc.hb_random_idx > 3) {
273 					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
274 					memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values));
275 					this_random = stcb->asoc.hb_random_values[0];
276 					stcb->asoc.hb_random_idx++;
277 					stcb->asoc.hb_ect_randombit = 0;
278 				} else {
279 					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
280 					stcb->asoc.hb_random_idx++;
281 					stcb->asoc.hb_ect_randombit = 0;
282 				}
283 				if (this_random % 2 == 1) {
284 					max_cwnd_net = mnet;
285 					max_cwnd = mnet->cwnd;	/* Useless? */
286 				}
287 			}
288 		}
289 		if (max_cwnd_net == NULL) {
290 			if (min_errors_net == NULL) {
291 				return (net);
292 			}
293 			return (min_errors_net);
294 		} else {
295 			return (max_cwnd_net);
296 		}
297 	}
298 	/*
299 	 * JRS 5/14/07 - If mode is set to 1, use the CMT policy for
300 	 * choosing an alternate net.
301 	 */
302 	else if (mode == 1) {
303 		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
304 			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
305 			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
306 				/*
307 				 * will skip ones that are not-reachable or
308 				 * unconfirmed
309 				 */
310 				continue;
311 			}
312 			if (max_cwnd < mnet->cwnd) {
313 				max_cwnd_net = mnet;
314 				max_cwnd = mnet->cwnd;
315 			} else if (max_cwnd == mnet->cwnd) {
316 				uint32_t rndval;
317 				uint8_t this_random;
318 
319 				if (stcb->asoc.hb_random_idx > 3) {
320 					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
321 					memcpy(stcb->asoc.hb_random_values, &rndval,
322 					    sizeof(stcb->asoc.hb_random_values));
323 					this_random = stcb->asoc.hb_random_values[0];
324 					stcb->asoc.hb_random_idx = 0;
325 					stcb->asoc.hb_ect_randombit = 0;
326 				} else {
327 					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
328 					stcb->asoc.hb_random_idx++;
329 					stcb->asoc.hb_ect_randombit = 0;
330 				}
331 				if (this_random % 2) {
332 					max_cwnd_net = mnet;
333 					max_cwnd = mnet->cwnd;
334 				}
335 			}
336 		}
337 		if (max_cwnd_net) {
338 			return (max_cwnd_net);
339 		}
340 	}
341 	mnet = net;
342 	once = 0;
343 
344 	if (mnet == NULL) {
345 		mnet = TAILQ_FIRST(&stcb->asoc.nets);
346 		if (mnet == NULL) {
347 			return (NULL);
348 		}
349 	}
350 	do {
351 		alt = TAILQ_NEXT(mnet, sctp_next);
352 		if (alt == NULL) {
353 			once++;
354 			if (once > 1) {
355 				break;
356 			}
357 			alt = TAILQ_FIRST(&stcb->asoc.nets);
358 			if (alt == NULL) {
359 				return (NULL);
360 			}
361 		}
362 		if (alt->ro.ro_rt == NULL) {
363 			if (alt->ro._s_addr) {
364 				sctp_free_ifa(alt->ro._s_addr);
365 				alt->ro._s_addr = NULL;
366 			}
367 			alt->src_addr_selected = 0;
368 		}
369 		/* sa_ignore NO_NULL_CHK */
370 		if (((alt->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE) &&
371 		    (alt->ro.ro_rt != NULL) &&
372 		    (!(alt->dest_state & SCTP_ADDR_UNCONFIRMED))) {
373 			/* Found a reachable address */
374 			break;
375 		}
376 		mnet = alt;
377 	} while (alt != NULL);
378 
379 	if (alt == NULL) {
380 		/* Case where NO insv network exists (dormant state) */
381 		/* we rotate destinations */
382 		once = 0;
383 		mnet = net;
384 		do {
385 			if (mnet == NULL) {
386 				return (TAILQ_FIRST(&stcb->asoc.nets));
387 			}
388 			alt = TAILQ_NEXT(mnet, sctp_next);
389 			if (alt == NULL) {
390 				once++;
391 				if (once > 1) {
392 					break;
393 				}
394 				alt = TAILQ_FIRST(&stcb->asoc.nets);
395 			}
396 			/* sa_ignore NO_NULL_CHK */
397 			if ((!(alt->dest_state & SCTP_ADDR_UNCONFIRMED)) &&
398 			    (alt != net)) {
399 				/* Found an alternate address */
400 				break;
401 			}
402 			mnet = alt;
403 		} while (alt != NULL);
404 	}
405 	if (alt == NULL) {
406 		return (net);
407 	}
408 	return (alt);
409 }
410 
411 static void
412 sctp_backoff_on_timeout(struct sctp_tcb *stcb,
413     struct sctp_nets *net,
414     int win_probe,
415     int num_marked, int num_abandoned)
416 {
417 	if (net->RTO == 0) {
418 		net->RTO = stcb->asoc.minrto;
419 	}
420 	net->RTO <<= 1;
421 	if (net->RTO > stcb->asoc.maxrto) {
422 		net->RTO = stcb->asoc.maxrto;
423 	}
424 	if ((win_probe == 0) && (num_marked || num_abandoned)) {
425 		/* We don't apply penalty to window probe scenarios */
426 		/* JRS - Use the congestion control given in the CC module */
427 		stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout(stcb, net);
428 	}
429 }
430 
431 #ifndef INVARIANTS
432 static void
433 sctp_recover_sent_list(struct sctp_tcb *stcb)
434 {
435 	struct sctp_tmit_chunk *chk, *nchk;
436 	struct sctp_association *asoc;
437 
438 	asoc = &stcb->asoc;
439 	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
440 		if (SCTP_TSN_GE(asoc->last_acked_seq, chk->rec.data.TSN_seq)) {
441 			SCTP_PRINTF("Found chk:%p tsn:%x <= last_acked_seq:%x\n",
442 			    (void *)chk, chk->rec.data.TSN_seq, asoc->last_acked_seq);
443 			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
444 			if (chk->pr_sctp_on) {
445 				if (asoc->pr_sctp_cnt != 0)
446 					asoc->pr_sctp_cnt--;
447 			}
448 			if (chk->data) {
449 				/* sa_ignore NO_NULL_CHK */
450 				sctp_free_bufspace(stcb, asoc, chk, 1);
451 				sctp_m_freem(chk->data);
452 				chk->data = NULL;
453 				if (asoc->peer_supports_prsctp && PR_SCTP_BUF_ENABLED(chk->flags)) {
454 					asoc->sent_queue_cnt_removeable--;
455 				}
456 			}
457 			asoc->sent_queue_cnt--;
458 			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
459 		}
460 	}
461 	SCTP_PRINTF("after recover order is as follows\n");
462 	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
463 		SCTP_PRINTF("chk:%p TSN:%x\n", (void *)chk, chk->rec.data.TSN_seq);
464 	}
465 }
466 
467 #endif
468 
469 static int
470 sctp_mark_all_for_resend(struct sctp_tcb *stcb,
471     struct sctp_nets *net,
472     struct sctp_nets *alt,
473     int window_probe,
474     int *num_marked,
475     int *num_abandoned)
476 {
477 
478 	/*
479 	 * Mark all chunks (well not all) that were sent to *net for
480 	 * retransmission. Move them to alt for there destination as well...
481 	 * We only mark chunks that have been outstanding long enough to
482 	 * have received feed-back.
483 	 */
484 	struct sctp_tmit_chunk *chk, *nchk;
485 	struct sctp_nets *lnets;
486 	struct timeval now, min_wait, tv;
487 	int cur_rto;
488 	int cnt_abandoned;
489 	int audit_tf, num_mk, fir;
490 	unsigned int cnt_mk;
491 	uint32_t orig_flight, orig_tf;
492 	uint32_t tsnlast, tsnfirst;
493 	int recovery_cnt = 0;
494 
495 
496 	/* none in flight now */
497 	audit_tf = 0;
498 	fir = 0;
499 	/*
500 	 * figure out how long a data chunk must be pending before we can
501 	 * mark it ..
502 	 */
503 	(void)SCTP_GETTIME_TIMEVAL(&now);
504 	/* get cur rto in micro-seconds */
505 	cur_rto = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv;
506 	cur_rto *= 1000;
507 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
508 		sctp_log_fr(cur_rto,
509 		    stcb->asoc.peers_rwnd,
510 		    window_probe,
511 		    SCTP_FR_T3_MARK_TIME);
512 		sctp_log_fr(net->flight_size, 0, 0, SCTP_FR_CWND_REPORT);
513 		sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT);
514 	}
515 	tv.tv_sec = cur_rto / 1000000;
516 	tv.tv_usec = cur_rto % 1000000;
517 	min_wait = now;
518 	timevalsub(&min_wait, &tv);
519 	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
520 		/*
521 		 * if we hit here, we don't have enough seconds on the clock
522 		 * to account for the RTO. We just let the lower seconds be
523 		 * the bounds and don't worry about it. This may mean we
524 		 * will mark a lot more than we should.
525 		 */
526 		min_wait.tv_sec = min_wait.tv_usec = 0;
527 	}
528 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
529 		sctp_log_fr(cur_rto, now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME);
530 		sctp_log_fr(0, min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME);
531 	}
532 	/*
533 	 * Our rwnd will be incorrect here since we are not adding back the
534 	 * cnt * mbuf but we will fix that down below.
535 	 */
536 	orig_flight = net->flight_size;
537 	orig_tf = stcb->asoc.total_flight;
538 
539 	net->fast_retran_ip = 0;
540 	/* Now on to each chunk */
541 	cnt_abandoned = 0;
542 	num_mk = cnt_mk = 0;
543 	tsnfirst = tsnlast = 0;
544 #ifndef INVARIANTS
545 start_again:
546 #endif
547 	TAILQ_FOREACH_SAFE(chk, &stcb->asoc.sent_queue, sctp_next, nchk) {
548 		if (SCTP_TSN_GE(stcb->asoc.last_acked_seq, chk->rec.data.TSN_seq)) {
549 			/* Strange case our list got out of order? */
550 			SCTP_PRINTF("Our list is out of order? last_acked:%x chk:%x",
551 			    (unsigned int)stcb->asoc.last_acked_seq, (unsigned int)chk->rec.data.TSN_seq);
552 			recovery_cnt++;
553 #ifdef INVARIANTS
554 			panic("last acked >= chk on sent-Q");
555 #else
556 			SCTP_PRINTF("Recover attempts a restart cnt:%d\n", recovery_cnt);
557 			sctp_recover_sent_list(stcb);
558 			if (recovery_cnt < 10) {
559 				goto start_again;
560 			} else {
561 				SCTP_PRINTF("Recovery fails %d times??\n", recovery_cnt);
562 			}
563 #endif
564 		}
565 		if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) {
566 			/*
567 			 * found one to mark: If it is less than
568 			 * DATAGRAM_ACKED it MUST not be a skipped or marked
569 			 * TSN but instead one that is either already set
570 			 * for retransmission OR one that needs
571 			 * retransmission.
572 			 */
573 
574 			/* validate its been outstanding long enough */
575 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
576 				sctp_log_fr(chk->rec.data.TSN_seq,
577 				    chk->sent_rcv_time.tv_sec,
578 				    chk->sent_rcv_time.tv_usec,
579 				    SCTP_FR_T3_MARK_TIME);
580 			}
581 			if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) {
582 				/*
583 				 * we have reached a chunk that was sent
584 				 * some seconds past our min.. forget it we
585 				 * will find no more to send.
586 				 */
587 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
588 					sctp_log_fr(0,
589 					    chk->sent_rcv_time.tv_sec,
590 					    chk->sent_rcv_time.tv_usec,
591 					    SCTP_FR_T3_STOPPED);
592 				}
593 				continue;
594 			} else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) &&
595 			    (window_probe == 0)) {
596 				/*
597 				 * we must look at the micro seconds to
598 				 * know.
599 				 */
600 				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
601 					/*
602 					 * ok it was sent after our boundary
603 					 * time.
604 					 */
605 					continue;
606 				}
607 			}
608 			if (stcb->asoc.peer_supports_prsctp && PR_SCTP_TTL_ENABLED(chk->flags)) {
609 				/* Is it expired? */
610 				if (timevalcmp(&now, &chk->rec.data.timetodrop, >)) {
611 					/* Yes so drop it */
612 					if (chk->data) {
613 						(void)sctp_release_pr_sctp_chunk(stcb,
614 						    chk,
615 						    1,
616 						    SCTP_SO_NOT_LOCKED);
617 						cnt_abandoned++;
618 					}
619 					continue;
620 				}
621 			}
622 			if (stcb->asoc.peer_supports_prsctp && PR_SCTP_RTX_ENABLED(chk->flags)) {
623 				/* Has it been retransmitted tv_sec times? */
624 				if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) {
625 					if (chk->data) {
626 						(void)sctp_release_pr_sctp_chunk(stcb,
627 						    chk,
628 						    1,
629 						    SCTP_SO_NOT_LOCKED);
630 						cnt_abandoned++;
631 					}
632 					continue;
633 				}
634 			}
635 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
636 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
637 				num_mk++;
638 				if (fir == 0) {
639 					fir = 1;
640 					tsnfirst = chk->rec.data.TSN_seq;
641 				}
642 				tsnlast = chk->rec.data.TSN_seq;
643 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
644 					sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count,
645 					    0, SCTP_FR_T3_MARKED);
646 				}
647 				if (chk->rec.data.chunk_was_revoked) {
648 					/* deflate the cwnd */
649 					chk->whoTo->cwnd -= chk->book_size;
650 					chk->rec.data.chunk_was_revoked = 0;
651 				}
652 				net->marked_retrans++;
653 				stcb->asoc.marked_retrans++;
654 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
655 					sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO,
656 					    chk->whoTo->flight_size,
657 					    chk->book_size,
658 					    (uintptr_t) chk->whoTo,
659 					    chk->rec.data.TSN_seq);
660 				}
661 				sctp_flight_size_decrease(chk);
662 				sctp_total_flight_decrease(stcb, chk);
663 				stcb->asoc.peers_rwnd += chk->send_size;
664 				stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh);
665 			}
666 			chk->sent = SCTP_DATAGRAM_RESEND;
667 			SCTP_STAT_INCR(sctps_markedretrans);
668 
669 			/* reset the TSN for striking and other FR stuff */
670 			chk->rec.data.doing_fast_retransmit = 0;
671 			/* Clear any time so NO RTT is being done */
672 
673 			if (chk->do_rtt) {
674 				if (chk->whoTo->rto_needed == 0) {
675 					chk->whoTo->rto_needed = 1;
676 				}
677 			}
678 			chk->do_rtt = 0;
679 			if (alt != net) {
680 				sctp_free_remote_addr(chk->whoTo);
681 				chk->no_fr_allowed = 1;
682 				chk->whoTo = alt;
683 				atomic_add_int(&alt->ref_count, 1);
684 			} else {
685 				chk->no_fr_allowed = 0;
686 				if (TAILQ_EMPTY(&stcb->asoc.send_queue)) {
687 					chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq;
688 				} else {
689 					chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.TSN_seq;
690 				}
691 			}
692 			/*
693 			 * CMT: Do not allow FRs on retransmitted TSNs.
694 			 */
695 			if (stcb->asoc.sctp_cmt_on_off > 0) {
696 				chk->no_fr_allowed = 1;
697 			}
698 #ifdef THIS_SHOULD_NOT_BE_DONE
699 		} else if (chk->sent == SCTP_DATAGRAM_ACKED) {
700 			/* remember highest acked one */
701 			could_be_sent = chk;
702 #endif
703 		}
704 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
705 			cnt_mk++;
706 		}
707 	}
708 	if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) {
709 		/* we did not subtract the same things? */
710 		audit_tf = 1;
711 	}
712 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
713 		sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT);
714 	}
715 #ifdef SCTP_DEBUG
716 	if (num_mk) {
717 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
718 		    tsnlast);
719 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%ld\n",
720 		    num_mk, (u_long)stcb->asoc.peers_rwnd);
721 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
722 		    tsnlast);
723 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%d\n",
724 		    num_mk,
725 		    (int)stcb->asoc.peers_rwnd);
726 	}
727 #endif
728 	*num_marked = num_mk;
729 	*num_abandoned = cnt_abandoned;
730 	/*
731 	 * Now check for a ECN Echo that may be stranded And include the
732 	 * cnt_mk'd to have all resends in the control queue.
733 	 */
734 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
735 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
736 			cnt_mk++;
737 		}
738 		if ((chk->whoTo == net) &&
739 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
740 			sctp_free_remote_addr(chk->whoTo);
741 			chk->whoTo = alt;
742 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
743 				chk->sent = SCTP_DATAGRAM_RESEND;
744 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
745 				cnt_mk++;
746 			}
747 			atomic_add_int(&alt->ref_count, 1);
748 		}
749 	}
750 #ifdef THIS_SHOULD_NOT_BE_DONE
751 	if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) {
752 		/* fix it so we retransmit the highest acked anyway */
753 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
754 		cnt_mk++;
755 		could_be_sent->sent = SCTP_DATAGRAM_RESEND;
756 	}
757 #endif
758 	if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) {
759 #ifdef INVARIANTS
760 		SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n",
761 		    cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk);
762 #endif
763 #ifndef SCTP_AUDITING_ENABLED
764 		stcb->asoc.sent_queue_retran_cnt = cnt_mk;
765 #endif
766 	}
767 	if (audit_tf) {
768 		SCTPDBG(SCTP_DEBUG_TIMER4,
769 		    "Audit total flight due to negative value net:%p\n",
770 		    (void *)net);
771 		stcb->asoc.total_flight = 0;
772 		stcb->asoc.total_flight_count = 0;
773 		/* Clear all networks flight size */
774 		TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) {
775 			lnets->flight_size = 0;
776 			SCTPDBG(SCTP_DEBUG_TIMER4,
777 			    "Net:%p c-f cwnd:%d ssthresh:%d\n",
778 			    (void *)lnets, lnets->cwnd, lnets->ssthresh);
779 		}
780 		TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
781 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
782 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
783 					sctp_misc_ints(SCTP_FLIGHT_LOG_UP,
784 					    chk->whoTo->flight_size,
785 					    chk->book_size,
786 					    (uintptr_t) chk->whoTo,
787 					    chk->rec.data.TSN_seq);
788 				}
789 				sctp_flight_size_increase(chk);
790 				sctp_total_flight_increase(stcb, chk);
791 			}
792 		}
793 	}
794 	/* We return 1 if we only have a window probe outstanding */
795 	return (0);
796 }
797 
798 
799 int
800 sctp_t3rxt_timer(struct sctp_inpcb *inp,
801     struct sctp_tcb *stcb,
802     struct sctp_nets *net)
803 {
804 	struct sctp_nets *alt;
805 	int win_probe, num_mk, num_abandoned;
806 
807 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
808 		sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT);
809 	}
810 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
811 		struct sctp_nets *lnet;
812 
813 		TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
814 			if (net == lnet) {
815 				sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3);
816 			} else {
817 				sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3);
818 			}
819 		}
820 	}
821 	/* Find an alternate and mark those for retransmission */
822 	if ((stcb->asoc.peers_rwnd == 0) &&
823 	    (stcb->asoc.total_flight < net->mtu)) {
824 		SCTP_STAT_INCR(sctps_timowindowprobe);
825 		win_probe = 1;
826 	} else {
827 		win_probe = 0;
828 	}
829 
830 	if (win_probe == 0) {
831 		/* We don't do normal threshold management on window probes */
832 		if (sctp_threshold_management(inp, stcb, net,
833 		    stcb->asoc.max_send_times)) {
834 			/* Association was destroyed */
835 			return (1);
836 		} else {
837 			if (net != stcb->asoc.primary_destination) {
838 				/* send a immediate HB if our RTO is stale */
839 				struct timeval now;
840 				unsigned int ms_goneby;
841 
842 				(void)SCTP_GETTIME_TIMEVAL(&now);
843 				if (net->last_sent_time.tv_sec) {
844 					ms_goneby = (now.tv_sec - net->last_sent_time.tv_sec) * 1000;
845 				} else {
846 					ms_goneby = 0;
847 				}
848 				if ((net->dest_state & SCTP_ADDR_PF) == 0) {
849 					if ((ms_goneby > net->RTO) || (net->RTO == 0)) {
850 						/*
851 						 * no recent feed back in an
852 						 * RTO or more, request a
853 						 * RTT update
854 						 */
855 						sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
856 					}
857 				}
858 			}
859 		}
860 	} else {
861 		/*
862 		 * For a window probe we don't penalize the net's but only
863 		 * the association. This may fail it if SACKs are not coming
864 		 * back. If sack's are coming with rwnd locked at 0, we will
865 		 * continue to hold things waiting for rwnd to raise
866 		 */
867 		if (sctp_threshold_management(inp, stcb, NULL,
868 		    stcb->asoc.max_send_times)) {
869 			/* Association was destroyed */
870 			return (1);
871 		}
872 	}
873 	if (stcb->asoc.sctp_cmt_on_off > 0) {
874 		if (net->pf_threshold < net->failure_threshold) {
875 			alt = sctp_find_alternate_net(stcb, net, 2);
876 		} else {
877 			/*
878 			 * CMT: Using RTX_SSTHRESH policy for CMT. If CMT is
879 			 * being used, then pick dest with largest ssthresh
880 			 * for any retransmission.
881 			 */
882 			alt = sctp_find_alternate_net(stcb, net, 1);
883 			/*
884 			 * CUCv2: If a different dest is picked for the
885 			 * retransmission, then new (rtx-)pseudo_cumack
886 			 * needs to be tracked for orig dest. Let CUCv2
887 			 * track new (rtx-) pseudo-cumack always.
888 			 */
889 			net->find_pseudo_cumack = 1;
890 			net->find_rtx_pseudo_cumack = 1;
891 		}
892 	} else {
893 		alt = sctp_find_alternate_net(stcb, net, 0);
894 	}
895 
896 	num_mk = 0;
897 	num_abandoned = 0;
898 	(void)sctp_mark_all_for_resend(stcb, net, alt, win_probe,
899 	    &num_mk, &num_abandoned);
900 	/* FR Loss recovery just ended with the T3. */
901 	stcb->asoc.fast_retran_loss_recovery = 0;
902 
903 	/* CMT FR loss recovery ended with the T3 */
904 	net->fast_retran_loss_recovery = 0;
905 	if ((stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) &&
906 	    (net->flight_size == 0)) {
907 		(*stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) (stcb, net);
908 	}
909 	/*
910 	 * setup the sat loss recovery that prevents satellite cwnd advance.
911 	 */
912 	stcb->asoc.sat_t3_loss_recovery = 1;
913 	stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq;
914 
915 	/* Backoff the timer and cwnd */
916 	sctp_backoff_on_timeout(stcb, net, win_probe, num_mk, num_abandoned);
917 	if ((!(net->dest_state & SCTP_ADDR_REACHABLE)) ||
918 	    (net->dest_state & SCTP_ADDR_PF)) {
919 		/* Move all pending over too */
920 		sctp_move_chunks_from_net(stcb, net);
921 
922 		/*
923 		 * Get the address that failed, to force a new src address
924 		 * selecton and a route allocation.
925 		 */
926 		if (net->ro._s_addr) {
927 			sctp_free_ifa(net->ro._s_addr);
928 			net->ro._s_addr = NULL;
929 		}
930 		net->src_addr_selected = 0;
931 
932 		/* Force a route allocation too */
933 		if (net->ro.ro_rt) {
934 			RTFREE(net->ro.ro_rt);
935 			net->ro.ro_rt = NULL;
936 		}
937 		/* Was it our primary? */
938 		if ((stcb->asoc.primary_destination == net) && (alt != net)) {
939 			/*
940 			 * Yes, note it as such and find an alternate note:
941 			 * this means HB code must use this to resent the
942 			 * primary if it goes active AND if someone does a
943 			 * change-primary then this flag must be cleared
944 			 * from any net structures.
945 			 */
946 			if (stcb->asoc.alternate) {
947 				sctp_free_remote_addr(stcb->asoc.alternate);
948 			}
949 			stcb->asoc.alternate = alt;
950 			atomic_add_int(&stcb->asoc.alternate->ref_count, 1);
951 		}
952 	}
953 	/*
954 	 * Special case for cookie-echo'ed case, we don't do output but must
955 	 * await the COOKIE-ACK before retransmission
956 	 */
957 	if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
958 		/*
959 		 * Here we just reset the timer and start again since we
960 		 * have not established the asoc
961 		 */
962 		sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net);
963 		return (0);
964 	}
965 	if (stcb->asoc.peer_supports_prsctp) {
966 		struct sctp_tmit_chunk *lchk;
967 
968 		lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc);
969 		/* C3. See if we need to send a Fwd-TSN */
970 		if (SCTP_TSN_GT(stcb->asoc.advanced_peer_ack_point, stcb->asoc.last_acked_seq)) {
971 			send_forward_tsn(stcb, &stcb->asoc);
972 			if (lchk) {
973 				/* Assure a timer is up */
974 				sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo);
975 			}
976 		}
977 	}
978 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
979 		sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX);
980 	}
981 	return (0);
982 }
983 
984 int
985 sctp_t1init_timer(struct sctp_inpcb *inp,
986     struct sctp_tcb *stcb,
987     struct sctp_nets *net)
988 {
989 	/* bump the thresholds */
990 	if (stcb->asoc.delayed_connection) {
991 		/*
992 		 * special hook for delayed connection. The library did NOT
993 		 * complete the rest of its sends.
994 		 */
995 		stcb->asoc.delayed_connection = 0;
996 		sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
997 		return (0);
998 	}
999 	if (SCTP_GET_STATE((&stcb->asoc)) != SCTP_STATE_COOKIE_WAIT) {
1000 		return (0);
1001 	}
1002 	if (sctp_threshold_management(inp, stcb, net,
1003 	    stcb->asoc.max_init_times)) {
1004 		/* Association was destroyed */
1005 		return (1);
1006 	}
1007 	stcb->asoc.dropped_special_cnt = 0;
1008 	sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0, 0);
1009 	if (stcb->asoc.initial_init_rto_max < net->RTO) {
1010 		net->RTO = stcb->asoc.initial_init_rto_max;
1011 	}
1012 	if (stcb->asoc.numnets > 1) {
1013 		/* If we have more than one addr use it */
1014 		struct sctp_nets *alt;
1015 
1016 		alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0);
1017 		if (alt != stcb->asoc.primary_destination) {
1018 			sctp_move_chunks_from_net(stcb, stcb->asoc.primary_destination);
1019 			stcb->asoc.primary_destination = alt;
1020 		}
1021 	}
1022 	/* Send out a new init */
1023 	sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1024 	return (0);
1025 }
1026 
1027 /*
1028  * For cookie and asconf we actually need to find and mark for resend, then
1029  * increment the resend counter (after all the threshold management stuff of
1030  * course).
1031  */
1032 int
1033 sctp_cookie_timer(struct sctp_inpcb *inp,
1034     struct sctp_tcb *stcb,
1035     struct sctp_nets *net SCTP_UNUSED)
1036 {
1037 	struct sctp_nets *alt;
1038 	struct sctp_tmit_chunk *cookie;
1039 
1040 	/* first before all else we must find the cookie */
1041 	TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) {
1042 		if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) {
1043 			break;
1044 		}
1045 	}
1046 	if (cookie == NULL) {
1047 		if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
1048 			/* FOOBAR! */
1049 			struct mbuf *oper;
1050 
1051 			oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
1052 			    0, M_DONTWAIT, 1, MT_DATA);
1053 			if (oper) {
1054 				struct sctp_paramhdr *ph;
1055 				uint32_t *ippp;
1056 
1057 				SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
1058 				    sizeof(uint32_t);
1059 				ph = mtod(oper, struct sctp_paramhdr *);
1060 				ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1061 				ph->param_length = htons(SCTP_BUF_LEN(oper));
1062 				ippp = (uint32_t *) (ph + 1);
1063 				*ippp = htonl(SCTP_FROM_SCTP_TIMER + SCTP_LOC_3);
1064 			}
1065 			inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_4;
1066 			sctp_abort_an_association(inp, stcb, oper, SCTP_SO_NOT_LOCKED);
1067 		} else {
1068 #ifdef INVARIANTS
1069 			panic("Cookie timer expires in wrong state?");
1070 #else
1071 			SCTP_PRINTF("Strange in state %d not cookie-echoed yet c-e timer expires?\n", SCTP_GET_STATE(&stcb->asoc));
1072 			return (0);
1073 #endif
1074 		}
1075 		return (0);
1076 	}
1077 	/* Ok we found the cookie, threshold management next */
1078 	if (sctp_threshold_management(inp, stcb, cookie->whoTo,
1079 	    stcb->asoc.max_init_times)) {
1080 		/* Assoc is over */
1081 		return (1);
1082 	}
1083 	/*
1084 	 * cleared theshold management now lets backoff the address & select
1085 	 * an alternate
1086 	 */
1087 	stcb->asoc.dropped_special_cnt = 0;
1088 	sctp_backoff_on_timeout(stcb, cookie->whoTo, 1, 0, 0);
1089 	alt = sctp_find_alternate_net(stcb, cookie->whoTo, 0);
1090 	if (alt != cookie->whoTo) {
1091 		sctp_free_remote_addr(cookie->whoTo);
1092 		cookie->whoTo = alt;
1093 		atomic_add_int(&alt->ref_count, 1);
1094 	}
1095 	/* Now mark the retran info */
1096 	if (cookie->sent != SCTP_DATAGRAM_RESEND) {
1097 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1098 	}
1099 	cookie->sent = SCTP_DATAGRAM_RESEND;
1100 	/*
1101 	 * Now call the output routine to kick out the cookie again, Note we
1102 	 * don't mark any chunks for retran so that FR will need to kick in
1103 	 * to move these (or a send timer).
1104 	 */
1105 	return (0);
1106 }
1107 
1108 int
1109 sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1110     struct sctp_nets *net)
1111 {
1112 	struct sctp_nets *alt;
1113 	struct sctp_tmit_chunk *strrst = NULL, *chk = NULL;
1114 
1115 	if (stcb->asoc.stream_reset_outstanding == 0) {
1116 		return (0);
1117 	}
1118 	/* find the existing STRRESET, we use the seq number we sent out on */
1119 	(void)sctp_find_stream_reset(stcb, stcb->asoc.str_reset_seq_out, &strrst);
1120 	if (strrst == NULL) {
1121 		return (0);
1122 	}
1123 	/* do threshold management */
1124 	if (sctp_threshold_management(inp, stcb, strrst->whoTo,
1125 	    stcb->asoc.max_send_times)) {
1126 		/* Assoc is over */
1127 		return (1);
1128 	}
1129 	/*
1130 	 * cleared theshold management now lets backoff the address & select
1131 	 * an alternate
1132 	 */
1133 	sctp_backoff_on_timeout(stcb, strrst->whoTo, 1, 0, 0);
1134 	alt = sctp_find_alternate_net(stcb, strrst->whoTo, 0);
1135 	sctp_free_remote_addr(strrst->whoTo);
1136 	strrst->whoTo = alt;
1137 	atomic_add_int(&alt->ref_count, 1);
1138 
1139 	/* See if a ECN Echo is also stranded */
1140 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1141 		if ((chk->whoTo == net) &&
1142 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1143 			sctp_free_remote_addr(chk->whoTo);
1144 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
1145 				chk->sent = SCTP_DATAGRAM_RESEND;
1146 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1147 			}
1148 			chk->whoTo = alt;
1149 			atomic_add_int(&alt->ref_count, 1);
1150 		}
1151 	}
1152 	if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1153 		/*
1154 		 * If the address went un-reachable, we need to move to
1155 		 * alternates for ALL chk's in queue
1156 		 */
1157 		sctp_move_chunks_from_net(stcb, net);
1158 	}
1159 	/* mark the retran info */
1160 	if (strrst->sent != SCTP_DATAGRAM_RESEND)
1161 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1162 	strrst->sent = SCTP_DATAGRAM_RESEND;
1163 
1164 	/* restart the timer */
1165 	sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, strrst->whoTo);
1166 	return (0);
1167 }
1168 
1169 int
1170 sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1171     struct sctp_nets *net)
1172 {
1173 	struct sctp_nets *alt;
1174 	struct sctp_tmit_chunk *asconf, *chk;
1175 
1176 	/* is this a first send, or a retransmission? */
1177 	if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) {
1178 		/* compose a new ASCONF chunk and send it */
1179 		sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED);
1180 	} else {
1181 		/*
1182 		 * Retransmission of the existing ASCONF is needed
1183 		 */
1184 
1185 		/* find the existing ASCONF */
1186 		asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue);
1187 		if (asconf == NULL) {
1188 			return (0);
1189 		}
1190 		/* do threshold management */
1191 		if (sctp_threshold_management(inp, stcb, asconf->whoTo,
1192 		    stcb->asoc.max_send_times)) {
1193 			/* Assoc is over */
1194 			return (1);
1195 		}
1196 		if (asconf->snd_count > stcb->asoc.max_send_times) {
1197 			/*
1198 			 * Something is rotten: our peer is not responding
1199 			 * to ASCONFs but apparently is to other chunks.
1200 			 * i.e. it is not properly handling the chunk type
1201 			 * upper bits. Mark this peer as ASCONF incapable
1202 			 * and cleanup.
1203 			 */
1204 			SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n");
1205 			sctp_asconf_cleanup(stcb, net);
1206 			return (0);
1207 		}
1208 		/*
1209 		 * cleared threshold management, so now backoff the net and
1210 		 * select an alternate
1211 		 */
1212 		sctp_backoff_on_timeout(stcb, asconf->whoTo, 1, 0, 0);
1213 		alt = sctp_find_alternate_net(stcb, asconf->whoTo, 0);
1214 		if (asconf->whoTo != alt) {
1215 			sctp_free_remote_addr(asconf->whoTo);
1216 			asconf->whoTo = alt;
1217 			atomic_add_int(&alt->ref_count, 1);
1218 		}
1219 		/* See if an ECN Echo is also stranded */
1220 		TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1221 			if ((chk->whoTo == net) &&
1222 			    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1223 				sctp_free_remote_addr(chk->whoTo);
1224 				chk->whoTo = alt;
1225 				if (chk->sent != SCTP_DATAGRAM_RESEND) {
1226 					chk->sent = SCTP_DATAGRAM_RESEND;
1227 					sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1228 				}
1229 				atomic_add_int(&alt->ref_count, 1);
1230 			}
1231 		}
1232 		TAILQ_FOREACH(chk, &stcb->asoc.asconf_send_queue, sctp_next) {
1233 			if (chk->whoTo != alt) {
1234 				sctp_free_remote_addr(chk->whoTo);
1235 				chk->whoTo = alt;
1236 				atomic_add_int(&alt->ref_count, 1);
1237 			}
1238 			if (asconf->sent != SCTP_DATAGRAM_RESEND && chk->sent != SCTP_DATAGRAM_UNSENT)
1239 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1240 			chk->sent = SCTP_DATAGRAM_RESEND;
1241 		}
1242 		if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1243 			/*
1244 			 * If the address went un-reachable, we need to move
1245 			 * to the alternate for ALL chunks in queue
1246 			 */
1247 			sctp_move_chunks_from_net(stcb, net);
1248 		}
1249 		/* mark the retran info */
1250 		if (asconf->sent != SCTP_DATAGRAM_RESEND)
1251 			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1252 		asconf->sent = SCTP_DATAGRAM_RESEND;
1253 
1254 		/* send another ASCONF if any and we can do */
1255 		sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED);
1256 	}
1257 	return (0);
1258 }
1259 
1260 /* Mobility adaptation */
1261 void
1262 sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1263     struct sctp_nets *net SCTP_UNUSED)
1264 {
1265 	if (stcb->asoc.deleted_primary == NULL) {
1266 		SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n");
1267 		sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1268 		return;
1269 	}
1270 	SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary ");
1271 	SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa);
1272 	sctp_free_remote_addr(stcb->asoc.deleted_primary);
1273 	stcb->asoc.deleted_primary = NULL;
1274 	sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1275 	return;
1276 }
1277 
1278 /*
1279  * For the shutdown and shutdown-ack, we do not keep one around on the
1280  * control queue. This means we must generate a new one and call the general
1281  * chunk output routine, AFTER having done threshold management.
1282  * It is assumed that net is non-NULL.
1283  */
1284 int
1285 sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1286     struct sctp_nets *net)
1287 {
1288 	struct sctp_nets *alt;
1289 
1290 	/* first threshold managment */
1291 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1292 		/* Assoc is over */
1293 		return (1);
1294 	}
1295 	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1296 	/* second select an alternative */
1297 	alt = sctp_find_alternate_net(stcb, net, 0);
1298 
1299 	/* third generate a shutdown into the queue for out net */
1300 	sctp_send_shutdown(stcb, alt);
1301 
1302 	/* fourth restart timer */
1303 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt);
1304 	return (0);
1305 }
1306 
1307 int
1308 sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1309     struct sctp_nets *net)
1310 {
1311 	struct sctp_nets *alt;
1312 
1313 	/* first threshold managment */
1314 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1315 		/* Assoc is over */
1316 		return (1);
1317 	}
1318 	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1319 	/* second select an alternative */
1320 	alt = sctp_find_alternate_net(stcb, net, 0);
1321 
1322 	/* third generate a shutdown into the queue for out net */
1323 	sctp_send_shutdown_ack(stcb, alt);
1324 
1325 	/* fourth restart timer */
1326 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt);
1327 	return (0);
1328 }
1329 
1330 static void
1331 sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp,
1332     struct sctp_tcb *stcb)
1333 {
1334 	struct sctp_stream_queue_pending *sp;
1335 	unsigned int i, chks_in_queue = 0;
1336 	int being_filled = 0;
1337 
1338 	/*
1339 	 * This function is ONLY called when the send/sent queues are empty.
1340 	 */
1341 	if ((stcb == NULL) || (inp == NULL))
1342 		return;
1343 
1344 	if (stcb->asoc.sent_queue_retran_cnt) {
1345 		SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n",
1346 		    stcb->asoc.sent_queue_retran_cnt);
1347 		stcb->asoc.sent_queue_retran_cnt = 0;
1348 	}
1349 	if (stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1350 		/* No stream scheduler information, initialize scheduler */
1351 		stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc, 0);
1352 		if (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1353 			/* yep, we lost a stream or two */
1354 			SCTP_PRINTF("Found additional streams NOT managed by scheduler, corrected\n");
1355 		} else {
1356 			/* no streams lost */
1357 			stcb->asoc.total_output_queue_size = 0;
1358 		}
1359 	}
1360 	/* Check to see if some data queued, if so report it */
1361 	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
1362 		if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) {
1363 			TAILQ_FOREACH(sp, &stcb->asoc.strmout[i].outqueue, next) {
1364 				if (sp->msg_is_complete)
1365 					being_filled++;
1366 				chks_in_queue++;
1367 			}
1368 		}
1369 	}
1370 	if (chks_in_queue != stcb->asoc.stream_queue_cnt) {
1371 		SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n",
1372 		    stcb->asoc.stream_queue_cnt, chks_in_queue);
1373 	}
1374 	if (chks_in_queue) {
1375 		/* call the output queue function */
1376 		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1377 		if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1378 		    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1379 			/*
1380 			 * Probably should go in and make it go back through
1381 			 * and add fragments allowed
1382 			 */
1383 			if (being_filled == 0) {
1384 				SCTP_PRINTF("Still nothing moved %d chunks are stuck\n",
1385 				    chks_in_queue);
1386 			}
1387 		}
1388 	} else {
1389 		SCTP_PRINTF("Found no chunks on any queue tot:%lu\n",
1390 		    (u_long)stcb->asoc.total_output_queue_size);
1391 		stcb->asoc.total_output_queue_size = 0;
1392 	}
1393 }
1394 
1395 int
1396 sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1397     struct sctp_nets *net)
1398 {
1399 	uint8_t net_was_pf;
1400 
1401 	if (net->dest_state & SCTP_ADDR_PF) {
1402 		net_was_pf = 1;
1403 	} else {
1404 		net_was_pf = 0;
1405 	}
1406 	if (net->hb_responded == 0) {
1407 		if (net->ro._s_addr) {
1408 			/*
1409 			 * Invalidate the src address if we did not get a
1410 			 * response last time.
1411 			 */
1412 			sctp_free_ifa(net->ro._s_addr);
1413 			net->ro._s_addr = NULL;
1414 			net->src_addr_selected = 0;
1415 		}
1416 		sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1417 		if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1418 			/* Assoc is over */
1419 			return (1);
1420 		}
1421 	}
1422 	/* Zero PBA, if it needs it */
1423 	if (net->partial_bytes_acked) {
1424 		net->partial_bytes_acked = 0;
1425 	}
1426 	if ((stcb->asoc.total_output_queue_size > 0) &&
1427 	    (TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1428 	    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1429 		sctp_audit_stream_queues_for_size(inp, stcb);
1430 	}
1431 	if (!(net->dest_state & SCTP_ADDR_NOHB) &&
1432 	    !((net_was_pf == 0) && (net->dest_state & SCTP_ADDR_PF))) {
1433 		/*
1434 		 * when move to PF during threshold mangement, a HB has been
1435 		 * queued in that routine
1436 		 */
1437 		uint32_t ms_gone_by;
1438 
1439 		if ((net->last_sent_time.tv_sec > 0) ||
1440 		    (net->last_sent_time.tv_usec > 0)) {
1441 			struct timeval diff;
1442 
1443 			SCTP_GETTIME_TIMEVAL(&diff);
1444 			timevalsub(&diff, &net->last_sent_time);
1445 			ms_gone_by = (uint32_t) (diff.tv_sec * 1000) +
1446 			    (uint32_t) (diff.tv_usec / 1000);
1447 		} else {
1448 			ms_gone_by = 0xffffffff;
1449 		}
1450 		if ((ms_gone_by >= net->heart_beat_delay) ||
1451 		    (net->dest_state & SCTP_ADDR_PF)) {
1452 			sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
1453 		}
1454 	}
1455 	return (0);
1456 }
1457 
1458 void
1459 sctp_pathmtu_timer(struct sctp_inpcb *inp,
1460     struct sctp_tcb *stcb,
1461     struct sctp_nets *net)
1462 {
1463 	uint32_t next_mtu, mtu;
1464 
1465 	next_mtu = sctp_get_next_mtu(net->mtu);
1466 
1467 	if ((next_mtu > net->mtu) && (net->port == 0)) {
1468 		if ((net->src_addr_selected == 0) ||
1469 		    (net->ro._s_addr == NULL) ||
1470 		    (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1471 			if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1472 				sctp_free_ifa(net->ro._s_addr);
1473 				net->ro._s_addr = NULL;
1474 				net->src_addr_selected = 0;
1475 			} else if (net->ro._s_addr == NULL) {
1476 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1477 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1478 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1479 
1480 					/* KAME hack: embed scopeid */
1481 					(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
1482 				}
1483 #endif
1484 
1485 				net->ro._s_addr = sctp_source_address_selection(inp,
1486 				    stcb,
1487 				    (sctp_route_t *) & net->ro,
1488 				    net, 0, stcb->asoc.vrf_id);
1489 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1490 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1491 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1492 
1493 					(void)sa6_recoverscope(sin6);
1494 				}
1495 #endif				/* INET6 */
1496 			}
1497 			if (net->ro._s_addr)
1498 				net->src_addr_selected = 1;
1499 		}
1500 		if (net->ro._s_addr) {
1501 			mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_rt);
1502 			if (net->port) {
1503 				mtu -= sizeof(struct udphdr);
1504 			}
1505 			if (mtu > next_mtu) {
1506 				net->mtu = next_mtu;
1507 			}
1508 		}
1509 	}
1510 	/* restart the timer */
1511 	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
1512 }
1513 
1514 void
1515 sctp_autoclose_timer(struct sctp_inpcb *inp,
1516     struct sctp_tcb *stcb,
1517     struct sctp_nets *net)
1518 {
1519 	struct timeval tn, *tim_touse;
1520 	struct sctp_association *asoc;
1521 	int ticks_gone_by;
1522 
1523 	(void)SCTP_GETTIME_TIMEVAL(&tn);
1524 	if (stcb->asoc.sctp_autoclose_ticks &&
1525 	    sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) {
1526 		/* Auto close is on */
1527 		asoc = &stcb->asoc;
1528 		/* pick the time to use */
1529 		if (asoc->time_last_rcvd.tv_sec >
1530 		    asoc->time_last_sent.tv_sec) {
1531 			tim_touse = &asoc->time_last_rcvd;
1532 		} else {
1533 			tim_touse = &asoc->time_last_sent;
1534 		}
1535 		/* Now has long enough transpired to autoclose? */
1536 		ticks_gone_by = SEC_TO_TICKS(tn.tv_sec - tim_touse->tv_sec);
1537 		if ((ticks_gone_by > 0) &&
1538 		    (ticks_gone_by >= (int)asoc->sctp_autoclose_ticks)) {
1539 			/*
1540 			 * autoclose time has hit, call the output routine,
1541 			 * which should do nothing just to be SURE we don't
1542 			 * have hanging data. We can then safely check the
1543 			 * queues and know that we are clear to send
1544 			 * shutdown
1545 			 */
1546 			sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED);
1547 			/* Are we clean? */
1548 			if (TAILQ_EMPTY(&asoc->send_queue) &&
1549 			    TAILQ_EMPTY(&asoc->sent_queue)) {
1550 				/*
1551 				 * there is nothing queued to send, so I'm
1552 				 * done...
1553 				 */
1554 				if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) {
1555 					/* only send SHUTDOWN 1st time thru */
1556 					struct sctp_nets *netp;
1557 
1558 					if (stcb->asoc.alternate) {
1559 						netp = stcb->asoc.alternate;
1560 					} else {
1561 						netp = stcb->asoc.primary_destination;
1562 					}
1563 					sctp_send_shutdown(stcb, netp);
1564 					if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
1565 					    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
1566 						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
1567 					}
1568 					SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
1569 					SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
1570 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
1571 					    stcb->sctp_ep, stcb,
1572 					    netp);
1573 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
1574 					    stcb->sctp_ep, stcb,
1575 					    netp);
1576 				}
1577 			}
1578 		} else {
1579 			/*
1580 			 * No auto close at this time, reset t-o to check
1581 			 * later
1582 			 */
1583 			int tmp;
1584 
1585 			/* fool the timer startup to use the time left */
1586 			tmp = asoc->sctp_autoclose_ticks;
1587 			asoc->sctp_autoclose_ticks -= ticks_gone_by;
1588 			sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb,
1589 			    net);
1590 			/* restore the real tick value */
1591 			asoc->sctp_autoclose_ticks = tmp;
1592 		}
1593 	}
1594 }
1595