xref: /freebsd/sys/netinet/sctp_timer.c (revision 26a222dc0c048fc071b548eadad7b80405a1b126)
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 #if defined(INET) || defined(INET6)
53 #include <netinet/udp.h>
54 #endif
55 
56 
57 void
58 sctp_audit_retranmission_queue(struct sctp_association *asoc)
59 {
60 	struct sctp_tmit_chunk *chk;
61 
62 	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit invoked on send queue cnt:%d onqueue:%d\n",
63 	    asoc->sent_queue_retran_cnt,
64 	    asoc->sent_queue_cnt);
65 	asoc->sent_queue_retran_cnt = 0;
66 	asoc->sent_queue_cnt = 0;
67 	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
68 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
69 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
70 		}
71 		asoc->sent_queue_cnt++;
72 	}
73 	TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) {
74 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
75 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
76 		}
77 	}
78 	TAILQ_FOREACH(chk, &asoc->asconf_send_queue, sctp_next) {
79 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
80 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
81 		}
82 	}
83 	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit completes retran:%d onqueue:%d\n",
84 	    asoc->sent_queue_retran_cnt,
85 	    asoc->sent_queue_cnt);
86 }
87 
88 int
89 sctp_threshold_management(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
90     struct sctp_nets *net, uint16_t threshold)
91 {
92 	if (net) {
93 		net->error_count++;
94 		SCTPDBG(SCTP_DEBUG_TIMER4, "Error count for %p now %d thresh:%d\n",
95 		    (void *)net, net->error_count,
96 		    net->failure_threshold);
97 		if (net->error_count > net->failure_threshold) {
98 			/* We had a threshold failure */
99 			if (net->dest_state & SCTP_ADDR_REACHABLE) {
100 				net->dest_state &= ~SCTP_ADDR_REACHABLE;
101 				net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY;
102 				net->dest_state &= ~SCTP_ADDR_PF;
103 				sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN,
104 				    stcb, 0,
105 				    (void *)net, SCTP_SO_NOT_LOCKED);
106 			}
107 		} else if ((net->pf_threshold < net->failure_threshold) &&
108 		    (net->error_count > net->pf_threshold)) {
109 			if (!(net->dest_state & SCTP_ADDR_PF)) {
110 				net->dest_state |= SCTP_ADDR_PF;
111 				net->last_active = sctp_get_tick_count();
112 				sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
113 				sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_TIMER + SCTP_LOC_3);
114 				sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net);
115 			}
116 		}
117 	}
118 	if (stcb == NULL)
119 		return (0);
120 
121 	if (net) {
122 		if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) {
123 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
124 				sctp_misc_ints(SCTP_THRESHOLD_INCR,
125 				    stcb->asoc.overall_error_count,
126 				    (stcb->asoc.overall_error_count + 1),
127 				    SCTP_FROM_SCTP_TIMER,
128 				    __LINE__);
129 			}
130 			stcb->asoc.overall_error_count++;
131 		}
132 	} else {
133 		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
134 			sctp_misc_ints(SCTP_THRESHOLD_INCR,
135 			    stcb->asoc.overall_error_count,
136 			    (stcb->asoc.overall_error_count + 1),
137 			    SCTP_FROM_SCTP_TIMER,
138 			    __LINE__);
139 		}
140 		stcb->asoc.overall_error_count++;
141 	}
142 	SCTPDBG(SCTP_DEBUG_TIMER4, "Overall error count for %p now %d thresh:%u state:%x\n",
143 	    (void *)&stcb->asoc, stcb->asoc.overall_error_count,
144 	    (uint32_t) threshold,
145 	    ((net == NULL) ? (uint32_t) 0 : (uint32_t) net->dest_state));
146 	/*
147 	 * We specifically do not do >= to give the assoc one more change
148 	 * before we fail it.
149 	 */
150 	if (stcb->asoc.overall_error_count > threshold) {
151 		/* Abort notification sends a ULP notify */
152 		struct mbuf *op_err;
153 
154 		op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION,
155 		    "Association error counter exceeded");
156 		inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_1;
157 		sctp_abort_an_association(inp, stcb, op_err, SCTP_SO_NOT_LOCKED);
158 		return (1);
159 	}
160 	return (0);
161 }
162 
163 /*
164  * sctp_find_alternate_net() returns a non-NULL pointer as long
165  * the argument net is non-NULL.
166  */
167 struct sctp_nets *
168 sctp_find_alternate_net(struct sctp_tcb *stcb,
169     struct sctp_nets *net,
170     int mode)
171 {
172 	/* Find and return an alternate network if possible */
173 	struct sctp_nets *alt, *mnet, *min_errors_net = NULL, *max_cwnd_net = NULL;
174 	int once;
175 
176 	/* JRS 5/14/07 - Initialize min_errors to an impossible value. */
177 	int min_errors = -1;
178 	uint32_t max_cwnd = 0;
179 
180 	if (stcb->asoc.numnets == 1) {
181 		/* No others but net */
182 		return (TAILQ_FIRST(&stcb->asoc.nets));
183 	}
184 	/*
185 	 * JRS 5/14/07 - If mode is set to 2, use the CMT PF find alternate
186 	 * net algorithm. This algorithm chooses the active destination (not
187 	 * in PF state) with the largest cwnd value. If all destinations are
188 	 * in PF state, unreachable, or unconfirmed, choose the desination
189 	 * that is in PF state with the lowest error count. In case of a
190 	 * tie, choose the destination that was most recently active.
191 	 */
192 	if (mode == 2) {
193 		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
194 			/*
195 			 * JRS 5/14/07 - If the destination is unreachable
196 			 * or unconfirmed, skip it.
197 			 */
198 			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
199 			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
200 				continue;
201 			}
202 			/*
203 			 * JRS 5/14/07 -  If the destination is reachable
204 			 * but in PF state, compare the error count of the
205 			 * destination to the minimum error count seen thus
206 			 * far. Store the destination with the lower error
207 			 * count.  If the error counts are equal, store the
208 			 * destination that was most recently active.
209 			 */
210 			if (mnet->dest_state & SCTP_ADDR_PF) {
211 				/*
212 				 * JRS 5/14/07 - If the destination under
213 				 * consideration is the current destination,
214 				 * work as if the error count is one higher.
215 				 * The actual error count will not be
216 				 * incremented until later in the t3
217 				 * handler.
218 				 */
219 				if (mnet == net) {
220 					if (min_errors == -1) {
221 						min_errors = mnet->error_count + 1;
222 						min_errors_net = mnet;
223 					} else if (mnet->error_count + 1 < min_errors) {
224 						min_errors = mnet->error_count + 1;
225 						min_errors_net = mnet;
226 					} else if (mnet->error_count + 1 == min_errors
227 					    && mnet->last_active > min_errors_net->last_active) {
228 						min_errors_net = mnet;
229 						min_errors = mnet->error_count + 1;
230 					}
231 					continue;
232 				} else {
233 					if (min_errors == -1) {
234 						min_errors = mnet->error_count;
235 						min_errors_net = mnet;
236 					} else if (mnet->error_count < min_errors) {
237 						min_errors = mnet->error_count;
238 						min_errors_net = mnet;
239 					} else if (mnet->error_count == min_errors
240 					    && mnet->last_active > min_errors_net->last_active) {
241 						min_errors_net = mnet;
242 						min_errors = mnet->error_count;
243 					}
244 					continue;
245 				}
246 			}
247 			/*
248 			 * JRS 5/14/07 - If the destination is reachable and
249 			 * not in PF state, compare the cwnd of the
250 			 * destination to the highest cwnd seen thus far.
251 			 * Store the destination with the higher cwnd value.
252 			 * If the cwnd values are equal, randomly choose one
253 			 * of the two destinations.
254 			 */
255 			if (max_cwnd < mnet->cwnd) {
256 				max_cwnd_net = mnet;
257 				max_cwnd = mnet->cwnd;
258 			} else if (max_cwnd == mnet->cwnd) {
259 				uint32_t rndval;
260 				uint8_t this_random;
261 
262 				if (stcb->asoc.hb_random_idx > 3) {
263 					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
264 					memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values));
265 					this_random = stcb->asoc.hb_random_values[0];
266 					stcb->asoc.hb_random_idx++;
267 					stcb->asoc.hb_ect_randombit = 0;
268 				} else {
269 					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
270 					stcb->asoc.hb_random_idx++;
271 					stcb->asoc.hb_ect_randombit = 0;
272 				}
273 				if (this_random % 2 == 1) {
274 					max_cwnd_net = mnet;
275 					max_cwnd = mnet->cwnd;	/* Useless? */
276 				}
277 			}
278 		}
279 		if (max_cwnd_net == NULL) {
280 			if (min_errors_net == NULL) {
281 				return (net);
282 			}
283 			return (min_errors_net);
284 		} else {
285 			return (max_cwnd_net);
286 		}
287 	}
288 	/*
289 	 * JRS 5/14/07 - If mode is set to 1, use the CMT policy for
290 	 * choosing an alternate net.
291 	 */
292 	else if (mode == 1) {
293 		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
294 			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
295 			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
296 				/*
297 				 * will skip ones that are not-reachable or
298 				 * unconfirmed
299 				 */
300 				continue;
301 			}
302 			if (max_cwnd < mnet->cwnd) {
303 				max_cwnd_net = mnet;
304 				max_cwnd = mnet->cwnd;
305 			} else if (max_cwnd == mnet->cwnd) {
306 				uint32_t rndval;
307 				uint8_t this_random;
308 
309 				if (stcb->asoc.hb_random_idx > 3) {
310 					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
311 					memcpy(stcb->asoc.hb_random_values, &rndval,
312 					    sizeof(stcb->asoc.hb_random_values));
313 					this_random = stcb->asoc.hb_random_values[0];
314 					stcb->asoc.hb_random_idx = 0;
315 					stcb->asoc.hb_ect_randombit = 0;
316 				} else {
317 					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
318 					stcb->asoc.hb_random_idx++;
319 					stcb->asoc.hb_ect_randombit = 0;
320 				}
321 				if (this_random % 2) {
322 					max_cwnd_net = mnet;
323 					max_cwnd = mnet->cwnd;
324 				}
325 			}
326 		}
327 		if (max_cwnd_net) {
328 			return (max_cwnd_net);
329 		}
330 	}
331 	mnet = net;
332 	once = 0;
333 
334 	if (mnet == NULL) {
335 		mnet = TAILQ_FIRST(&stcb->asoc.nets);
336 		if (mnet == NULL) {
337 			return (NULL);
338 		}
339 	}
340 	for (;;) {
341 		alt = TAILQ_NEXT(mnet, sctp_next);
342 		if (alt == NULL) {
343 			once++;
344 			if (once > 1) {
345 				break;
346 			}
347 			alt = TAILQ_FIRST(&stcb->asoc.nets);
348 			if (alt == NULL) {
349 				return (NULL);
350 			}
351 		}
352 		if (alt->ro.ro_rt == NULL) {
353 			if (alt->ro._s_addr) {
354 				sctp_free_ifa(alt->ro._s_addr);
355 				alt->ro._s_addr = NULL;
356 			}
357 			alt->src_addr_selected = 0;
358 		}
359 		if (((alt->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE) &&
360 		    (alt->ro.ro_rt != NULL) &&
361 		    (!(alt->dest_state & SCTP_ADDR_UNCONFIRMED))) {
362 			/* Found a reachable address */
363 			break;
364 		}
365 		mnet = alt;
366 	}
367 
368 	if (alt == NULL) {
369 		/* Case where NO insv network exists (dormant state) */
370 		/* we rotate destinations */
371 		once = 0;
372 		mnet = net;
373 		for (;;) {
374 			if (mnet == NULL) {
375 				return (TAILQ_FIRST(&stcb->asoc.nets));
376 			}
377 			alt = TAILQ_NEXT(mnet, sctp_next);
378 			if (alt == NULL) {
379 				once++;
380 				if (once > 1) {
381 					break;
382 				}
383 				alt = TAILQ_FIRST(&stcb->asoc.nets);
384 				if (alt == NULL) {
385 					break;
386 				}
387 			}
388 			if ((!(alt->dest_state & SCTP_ADDR_UNCONFIRMED)) &&
389 			    (alt != net)) {
390 				/* Found an alternate address */
391 				break;
392 			}
393 			mnet = alt;
394 		}
395 	}
396 	if (alt == NULL) {
397 		return (net);
398 	}
399 	return (alt);
400 }
401 
402 static void
403 sctp_backoff_on_timeout(struct sctp_tcb *stcb,
404     struct sctp_nets *net,
405     int win_probe,
406     int num_marked, int num_abandoned)
407 {
408 	if (net->RTO == 0) {
409 		net->RTO = stcb->asoc.minrto;
410 	}
411 	net->RTO <<= 1;
412 	if (net->RTO > stcb->asoc.maxrto) {
413 		net->RTO = stcb->asoc.maxrto;
414 	}
415 	if ((win_probe == 0) && (num_marked || num_abandoned)) {
416 		/* We don't apply penalty to window probe scenarios */
417 		/* JRS - Use the congestion control given in the CC module */
418 		stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout(stcb, net);
419 	}
420 }
421 
422 #ifndef INVARIANTS
423 static void
424 sctp_recover_sent_list(struct sctp_tcb *stcb)
425 {
426 	struct sctp_tmit_chunk *chk, *nchk;
427 	struct sctp_association *asoc;
428 
429 	asoc = &stcb->asoc;
430 	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
431 		if (SCTP_TSN_GE(asoc->last_acked_seq, chk->rec.data.TSN_seq)) {
432 			SCTP_PRINTF("Found chk:%p tsn:%x <= last_acked_seq:%x\n",
433 			    (void *)chk, chk->rec.data.TSN_seq, asoc->last_acked_seq);
434 			if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
435 				if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
436 					asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
437 				}
438 			}
439 			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
440 			if (PR_SCTP_ENABLED(chk->flags)) {
441 				if (asoc->pr_sctp_cnt != 0)
442 					asoc->pr_sctp_cnt--;
443 			}
444 			if (chk->data) {
445 				/* sa_ignore NO_NULL_CHK */
446 				sctp_free_bufspace(stcb, asoc, chk, 1);
447 				sctp_m_freem(chk->data);
448 				chk->data = NULL;
449 				if (asoc->prsctp_supported && PR_SCTP_BUF_ENABLED(chk->flags)) {
450 					asoc->sent_queue_cnt_removeable--;
451 				}
452 			}
453 			asoc->sent_queue_cnt--;
454 			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
455 		}
456 	}
457 	SCTP_PRINTF("after recover order is as follows\n");
458 	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
459 		SCTP_PRINTF("chk:%p TSN:%x\n", (void *)chk, chk->rec.data.TSN_seq);
460 	}
461 }
462 
463 #endif
464 
465 static int
466 sctp_mark_all_for_resend(struct sctp_tcb *stcb,
467     struct sctp_nets *net,
468     struct sctp_nets *alt,
469     int window_probe,
470     int *num_marked,
471     int *num_abandoned)
472 {
473 
474 	/*
475 	 * Mark all chunks (well not all) that were sent to *net for
476 	 * retransmission. Move them to alt for there destination as well...
477 	 * We only mark chunks that have been outstanding long enough to
478 	 * have received feed-back.
479 	 */
480 	struct sctp_tmit_chunk *chk, *nchk;
481 	struct sctp_nets *lnets;
482 	struct timeval now, min_wait, tv;
483 	int cur_rto;
484 	int cnt_abandoned;
485 	int audit_tf, num_mk, fir;
486 	unsigned int cnt_mk;
487 	uint32_t orig_flight, orig_tf;
488 	uint32_t tsnlast, tsnfirst;
489 	int recovery_cnt = 0;
490 
491 
492 	/* none in flight now */
493 	audit_tf = 0;
494 	fir = 0;
495 	/*
496 	 * figure out how long a data chunk must be pending before we can
497 	 * mark it ..
498 	 */
499 	(void)SCTP_GETTIME_TIMEVAL(&now);
500 	/* get cur rto in micro-seconds */
501 	cur_rto = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv;
502 	cur_rto *= 1000;
503 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
504 		sctp_log_fr(cur_rto,
505 		    stcb->asoc.peers_rwnd,
506 		    window_probe,
507 		    SCTP_FR_T3_MARK_TIME);
508 		sctp_log_fr(net->flight_size, 0, 0, SCTP_FR_CWND_REPORT);
509 		sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT);
510 	}
511 	tv.tv_sec = cur_rto / 1000000;
512 	tv.tv_usec = cur_rto % 1000000;
513 	min_wait = now;
514 	timevalsub(&min_wait, &tv);
515 	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
516 		/*
517 		 * if we hit here, we don't have enough seconds on the clock
518 		 * to account for the RTO. We just let the lower seconds be
519 		 * the bounds and don't worry about it. This may mean we
520 		 * will mark a lot more than we should.
521 		 */
522 		min_wait.tv_sec = min_wait.tv_usec = 0;
523 	}
524 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
525 		sctp_log_fr(cur_rto, now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME);
526 		sctp_log_fr(0, min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME);
527 	}
528 	/*
529 	 * Our rwnd will be incorrect here since we are not adding back the
530 	 * cnt * mbuf but we will fix that down below.
531 	 */
532 	orig_flight = net->flight_size;
533 	orig_tf = stcb->asoc.total_flight;
534 
535 	net->fast_retran_ip = 0;
536 	/* Now on to each chunk */
537 	cnt_abandoned = 0;
538 	num_mk = cnt_mk = 0;
539 	tsnfirst = tsnlast = 0;
540 #ifndef INVARIANTS
541 start_again:
542 #endif
543 	TAILQ_FOREACH_SAFE(chk, &stcb->asoc.sent_queue, sctp_next, nchk) {
544 		if (SCTP_TSN_GE(stcb->asoc.last_acked_seq, chk->rec.data.TSN_seq)) {
545 			/* Strange case our list got out of order? */
546 			SCTP_PRINTF("Our list is out of order? last_acked:%x chk:%x\n",
547 			    (unsigned int)stcb->asoc.last_acked_seq, (unsigned int)chk->rec.data.TSN_seq);
548 			recovery_cnt++;
549 #ifdef INVARIANTS
550 			panic("last acked >= chk on sent-Q");
551 #else
552 			SCTP_PRINTF("Recover attempts a restart cnt:%d\n", recovery_cnt);
553 			sctp_recover_sent_list(stcb);
554 			if (recovery_cnt < 10) {
555 				goto start_again;
556 			} else {
557 				SCTP_PRINTF("Recovery fails %d times??\n", recovery_cnt);
558 			}
559 #endif
560 		}
561 		if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) {
562 			/*
563 			 * found one to mark: If it is less than
564 			 * DATAGRAM_ACKED it MUST not be a skipped or marked
565 			 * TSN but instead one that is either already set
566 			 * for retransmission OR one that needs
567 			 * retransmission.
568 			 */
569 
570 			/* validate its been outstanding long enough */
571 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
572 				sctp_log_fr(chk->rec.data.TSN_seq,
573 				    chk->sent_rcv_time.tv_sec,
574 				    chk->sent_rcv_time.tv_usec,
575 				    SCTP_FR_T3_MARK_TIME);
576 			}
577 			if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) {
578 				/*
579 				 * we have reached a chunk that was sent
580 				 * some seconds past our min.. forget it we
581 				 * will find no more to send.
582 				 */
583 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
584 					sctp_log_fr(0,
585 					    chk->sent_rcv_time.tv_sec,
586 					    chk->sent_rcv_time.tv_usec,
587 					    SCTP_FR_T3_STOPPED);
588 				}
589 				continue;
590 			} else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) &&
591 			    (window_probe == 0)) {
592 				/*
593 				 * we must look at the micro seconds to
594 				 * know.
595 				 */
596 				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
597 					/*
598 					 * ok it was sent after our boundary
599 					 * time.
600 					 */
601 					continue;
602 				}
603 			}
604 			if (stcb->asoc.prsctp_supported && PR_SCTP_TTL_ENABLED(chk->flags)) {
605 				/* Is it expired? */
606 				if (timevalcmp(&now, &chk->rec.data.timetodrop, >)) {
607 					/* Yes so drop it */
608 					if (chk->data) {
609 						(void)sctp_release_pr_sctp_chunk(stcb,
610 						    chk,
611 						    1,
612 						    SCTP_SO_NOT_LOCKED);
613 						cnt_abandoned++;
614 					}
615 					continue;
616 				}
617 			}
618 			if (stcb->asoc.prsctp_supported && PR_SCTP_RTX_ENABLED(chk->flags)) {
619 				/* Has it been retransmitted tv_sec times? */
620 				if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) {
621 					if (chk->data) {
622 						(void)sctp_release_pr_sctp_chunk(stcb,
623 						    chk,
624 						    1,
625 						    SCTP_SO_NOT_LOCKED);
626 						cnt_abandoned++;
627 					}
628 					continue;
629 				}
630 			}
631 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
632 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
633 				num_mk++;
634 				if (fir == 0) {
635 					fir = 1;
636 					tsnfirst = chk->rec.data.TSN_seq;
637 				}
638 				tsnlast = chk->rec.data.TSN_seq;
639 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
640 					sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count,
641 					    0, SCTP_FR_T3_MARKED);
642 				}
643 				if (chk->rec.data.chunk_was_revoked) {
644 					/* deflate the cwnd */
645 					chk->whoTo->cwnd -= chk->book_size;
646 					chk->rec.data.chunk_was_revoked = 0;
647 				}
648 				net->marked_retrans++;
649 				stcb->asoc.marked_retrans++;
650 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
651 					sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO,
652 					    chk->whoTo->flight_size,
653 					    chk->book_size,
654 					    (uintptr_t) chk->whoTo,
655 					    chk->rec.data.TSN_seq);
656 				}
657 				sctp_flight_size_decrease(chk);
658 				sctp_total_flight_decrease(stcb, chk);
659 				stcb->asoc.peers_rwnd += chk->send_size;
660 				stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh);
661 			}
662 			chk->sent = SCTP_DATAGRAM_RESEND;
663 			SCTP_STAT_INCR(sctps_markedretrans);
664 
665 			/* reset the TSN for striking and other FR stuff */
666 			chk->rec.data.doing_fast_retransmit = 0;
667 			/* Clear any time so NO RTT is being done */
668 
669 			if (chk->do_rtt) {
670 				if (chk->whoTo->rto_needed == 0) {
671 					chk->whoTo->rto_needed = 1;
672 				}
673 			}
674 			chk->do_rtt = 0;
675 			if (alt != net) {
676 				sctp_free_remote_addr(chk->whoTo);
677 				chk->no_fr_allowed = 1;
678 				chk->whoTo = alt;
679 				atomic_add_int(&alt->ref_count, 1);
680 			} else {
681 				chk->no_fr_allowed = 0;
682 				if (TAILQ_EMPTY(&stcb->asoc.send_queue)) {
683 					chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq;
684 				} else {
685 					chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.TSN_seq;
686 				}
687 			}
688 			/*
689 			 * CMT: Do not allow FRs on retransmitted TSNs.
690 			 */
691 			if (stcb->asoc.sctp_cmt_on_off > 0) {
692 				chk->no_fr_allowed = 1;
693 			}
694 #ifdef THIS_SHOULD_NOT_BE_DONE
695 		} else if (chk->sent == SCTP_DATAGRAM_ACKED) {
696 			/* remember highest acked one */
697 			could_be_sent = chk;
698 #endif
699 		}
700 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
701 			cnt_mk++;
702 		}
703 	}
704 	if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) {
705 		/* we did not subtract the same things? */
706 		audit_tf = 1;
707 	}
708 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
709 		sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT);
710 	}
711 #ifdef SCTP_DEBUG
712 	if (num_mk) {
713 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
714 		    tsnlast);
715 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%ld\n",
716 		    num_mk, (u_long)stcb->asoc.peers_rwnd);
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:%d\n",
720 		    num_mk,
721 		    (int)stcb->asoc.peers_rwnd);
722 	}
723 #endif
724 	*num_marked = num_mk;
725 	*num_abandoned = cnt_abandoned;
726 	/*
727 	 * Now check for a ECN Echo that may be stranded And include the
728 	 * cnt_mk'd to have all resends in the control queue.
729 	 */
730 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
731 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
732 			cnt_mk++;
733 		}
734 		if ((chk->whoTo == net) &&
735 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
736 			sctp_free_remote_addr(chk->whoTo);
737 			chk->whoTo = alt;
738 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
739 				chk->sent = SCTP_DATAGRAM_RESEND;
740 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
741 				cnt_mk++;
742 			}
743 			atomic_add_int(&alt->ref_count, 1);
744 		}
745 	}
746 #ifdef THIS_SHOULD_NOT_BE_DONE
747 	if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) {
748 		/* fix it so we retransmit the highest acked anyway */
749 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
750 		cnt_mk++;
751 		could_be_sent->sent = SCTP_DATAGRAM_RESEND;
752 	}
753 #endif
754 	if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) {
755 #ifdef INVARIANTS
756 		SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n",
757 		    cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk);
758 #endif
759 #ifndef SCTP_AUDITING_ENABLED
760 		stcb->asoc.sent_queue_retran_cnt = cnt_mk;
761 #endif
762 	}
763 	if (audit_tf) {
764 		SCTPDBG(SCTP_DEBUG_TIMER4,
765 		    "Audit total flight due to negative value net:%p\n",
766 		    (void *)net);
767 		stcb->asoc.total_flight = 0;
768 		stcb->asoc.total_flight_count = 0;
769 		/* Clear all networks flight size */
770 		TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) {
771 			lnets->flight_size = 0;
772 			SCTPDBG(SCTP_DEBUG_TIMER4,
773 			    "Net:%p c-f cwnd:%d ssthresh:%d\n",
774 			    (void *)lnets, lnets->cwnd, lnets->ssthresh);
775 		}
776 		TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
777 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
778 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
779 					sctp_misc_ints(SCTP_FLIGHT_LOG_UP,
780 					    chk->whoTo->flight_size,
781 					    chk->book_size,
782 					    (uintptr_t) chk->whoTo,
783 					    chk->rec.data.TSN_seq);
784 				}
785 				sctp_flight_size_increase(chk);
786 				sctp_total_flight_increase(stcb, chk);
787 			}
788 		}
789 	}
790 	/* We return 1 if we only have a window probe outstanding */
791 	return (0);
792 }
793 
794 
795 int
796 sctp_t3rxt_timer(struct sctp_inpcb *inp,
797     struct sctp_tcb *stcb,
798     struct sctp_nets *net)
799 {
800 	struct sctp_nets *alt;
801 	int win_probe, num_mk, num_abandoned;
802 
803 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
804 		sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT);
805 	}
806 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
807 		struct sctp_nets *lnet;
808 
809 		TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
810 			if (net == lnet) {
811 				sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3);
812 			} else {
813 				sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3);
814 			}
815 		}
816 	}
817 	/* Find an alternate and mark those for retransmission */
818 	if ((stcb->asoc.peers_rwnd == 0) &&
819 	    (stcb->asoc.total_flight < net->mtu)) {
820 		SCTP_STAT_INCR(sctps_timowindowprobe);
821 		win_probe = 1;
822 	} else {
823 		win_probe = 0;
824 	}
825 
826 	if (win_probe == 0) {
827 		/* We don't do normal threshold management on window probes */
828 		if (sctp_threshold_management(inp, stcb, net,
829 		    stcb->asoc.max_send_times)) {
830 			/* Association was destroyed */
831 			return (1);
832 		} else {
833 			if (net != stcb->asoc.primary_destination) {
834 				/* send a immediate HB if our RTO is stale */
835 				struct timeval now;
836 				unsigned int ms_goneby;
837 
838 				(void)SCTP_GETTIME_TIMEVAL(&now);
839 				if (net->last_sent_time.tv_sec) {
840 					ms_goneby = (now.tv_sec - net->last_sent_time.tv_sec) * 1000;
841 				} else {
842 					ms_goneby = 0;
843 				}
844 				if ((net->dest_state & SCTP_ADDR_PF) == 0) {
845 					if ((ms_goneby > net->RTO) || (net->RTO == 0)) {
846 						/*
847 						 * no recent feed back in an
848 						 * RTO or more, request a
849 						 * RTT update
850 						 */
851 						sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
852 					}
853 				}
854 			}
855 		}
856 	} else {
857 		/*
858 		 * For a window probe we don't penalize the net's but only
859 		 * the association. This may fail it if SACKs are not coming
860 		 * back. If sack's are coming with rwnd locked at 0, we will
861 		 * continue to hold things waiting for rwnd to raise
862 		 */
863 		if (sctp_threshold_management(inp, stcb, NULL,
864 		    stcb->asoc.max_send_times)) {
865 			/* Association was destroyed */
866 			return (1);
867 		}
868 	}
869 	if (stcb->asoc.sctp_cmt_on_off > 0) {
870 		if (net->pf_threshold < net->failure_threshold) {
871 			alt = sctp_find_alternate_net(stcb, net, 2);
872 		} else {
873 			/*
874 			 * CMT: Using RTX_SSTHRESH policy for CMT. If CMT is
875 			 * being used, then pick dest with largest ssthresh
876 			 * for any retransmission.
877 			 */
878 			alt = sctp_find_alternate_net(stcb, net, 1);
879 			/*
880 			 * CUCv2: If a different dest is picked for the
881 			 * retransmission, then new (rtx-)pseudo_cumack
882 			 * needs to be tracked for orig dest. Let CUCv2
883 			 * track new (rtx-) pseudo-cumack always.
884 			 */
885 			net->find_pseudo_cumack = 1;
886 			net->find_rtx_pseudo_cumack = 1;
887 		}
888 	} else {
889 		alt = sctp_find_alternate_net(stcb, net, 0);
890 	}
891 
892 	num_mk = 0;
893 	num_abandoned = 0;
894 	(void)sctp_mark_all_for_resend(stcb, net, alt, win_probe,
895 	    &num_mk, &num_abandoned);
896 	/* FR Loss recovery just ended with the T3. */
897 	stcb->asoc.fast_retran_loss_recovery = 0;
898 
899 	/* CMT FR loss recovery ended with the T3 */
900 	net->fast_retran_loss_recovery = 0;
901 	if ((stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) &&
902 	    (net->flight_size == 0)) {
903 		(*stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) (stcb, net);
904 	}
905 	/*
906 	 * setup the sat loss recovery that prevents satellite cwnd advance.
907 	 */
908 	stcb->asoc.sat_t3_loss_recovery = 1;
909 	stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq;
910 
911 	/* Backoff the timer and cwnd */
912 	sctp_backoff_on_timeout(stcb, net, win_probe, num_mk, num_abandoned);
913 	if ((!(net->dest_state & SCTP_ADDR_REACHABLE)) ||
914 	    (net->dest_state & SCTP_ADDR_PF)) {
915 		/* Move all pending over too */
916 		sctp_move_chunks_from_net(stcb, net);
917 
918 		/*
919 		 * Get the address that failed, to force a new src address
920 		 * selecton and a route allocation.
921 		 */
922 		if (net->ro._s_addr) {
923 			sctp_free_ifa(net->ro._s_addr);
924 			net->ro._s_addr = NULL;
925 		}
926 		net->src_addr_selected = 0;
927 
928 		/* Force a route allocation too */
929 		if (net->ro.ro_rt) {
930 			RTFREE(net->ro.ro_rt);
931 			net->ro.ro_rt = NULL;
932 		}
933 		/* Was it our primary? */
934 		if ((stcb->asoc.primary_destination == net) && (alt != net)) {
935 			/*
936 			 * Yes, note it as such and find an alternate note:
937 			 * this means HB code must use this to resent the
938 			 * primary if it goes active AND if someone does a
939 			 * change-primary then this flag must be cleared
940 			 * from any net structures.
941 			 */
942 			if (stcb->asoc.alternate) {
943 				sctp_free_remote_addr(stcb->asoc.alternate);
944 			}
945 			stcb->asoc.alternate = alt;
946 			atomic_add_int(&stcb->asoc.alternate->ref_count, 1);
947 		}
948 	}
949 	/*
950 	 * Special case for cookie-echo'ed case, we don't do output but must
951 	 * await the COOKIE-ACK before retransmission
952 	 */
953 	if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
954 		/*
955 		 * Here we just reset the timer and start again since we
956 		 * have not established the asoc
957 		 */
958 		sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net);
959 		return (0);
960 	}
961 	if (stcb->asoc.prsctp_supported) {
962 		struct sctp_tmit_chunk *lchk;
963 
964 		lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc);
965 		/* C3. See if we need to send a Fwd-TSN */
966 		if (SCTP_TSN_GT(stcb->asoc.advanced_peer_ack_point, stcb->asoc.last_acked_seq)) {
967 			send_forward_tsn(stcb, &stcb->asoc);
968 			if (lchk) {
969 				/* Assure a timer is up */
970 				sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo);
971 			}
972 		}
973 	}
974 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
975 		sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX);
976 	}
977 	return (0);
978 }
979 
980 int
981 sctp_t1init_timer(struct sctp_inpcb *inp,
982     struct sctp_tcb *stcb,
983     struct sctp_nets *net)
984 {
985 	/* bump the thresholds */
986 	if (stcb->asoc.delayed_connection) {
987 		/*
988 		 * special hook for delayed connection. The library did NOT
989 		 * complete the rest of its sends.
990 		 */
991 		stcb->asoc.delayed_connection = 0;
992 		sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
993 		return (0);
994 	}
995 	if (SCTP_GET_STATE((&stcb->asoc)) != SCTP_STATE_COOKIE_WAIT) {
996 		return (0);
997 	}
998 	if (sctp_threshold_management(inp, stcb, net,
999 	    stcb->asoc.max_init_times)) {
1000 		/* Association was destroyed */
1001 		return (1);
1002 	}
1003 	stcb->asoc.dropped_special_cnt = 0;
1004 	sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0, 0);
1005 	if (stcb->asoc.initial_init_rto_max < net->RTO) {
1006 		net->RTO = stcb->asoc.initial_init_rto_max;
1007 	}
1008 	if (stcb->asoc.numnets > 1) {
1009 		/* If we have more than one addr use it */
1010 		struct sctp_nets *alt;
1011 
1012 		alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0);
1013 		if (alt != stcb->asoc.primary_destination) {
1014 			sctp_move_chunks_from_net(stcb, stcb->asoc.primary_destination);
1015 			stcb->asoc.primary_destination = alt;
1016 		}
1017 	}
1018 	/* Send out a new init */
1019 	sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1020 	return (0);
1021 }
1022 
1023 /*
1024  * For cookie and asconf we actually need to find and mark for resend, then
1025  * increment the resend counter (after all the threshold management stuff of
1026  * course).
1027  */
1028 int
1029 sctp_cookie_timer(struct sctp_inpcb *inp,
1030     struct sctp_tcb *stcb,
1031     struct sctp_nets *net SCTP_UNUSED)
1032 {
1033 	struct sctp_nets *alt;
1034 	struct sctp_tmit_chunk *cookie;
1035 
1036 	/* first before all else we must find the cookie */
1037 	TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) {
1038 		if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) {
1039 			break;
1040 		}
1041 	}
1042 	if (cookie == NULL) {
1043 		if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
1044 			/* FOOBAR! */
1045 			struct mbuf *op_err;
1046 
1047 			op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION,
1048 			    "Cookie timer expired, but no cookie");
1049 			inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_4;
1050 			sctp_abort_an_association(inp, stcb, op_err, SCTP_SO_NOT_LOCKED);
1051 		} else {
1052 #ifdef INVARIANTS
1053 			panic("Cookie timer expires in wrong state?");
1054 #else
1055 			SCTP_PRINTF("Strange in state %d not cookie-echoed yet c-e timer expires?\n", SCTP_GET_STATE(&stcb->asoc));
1056 			return (0);
1057 #endif
1058 		}
1059 		return (0);
1060 	}
1061 	/* Ok we found the cookie, threshold management next */
1062 	if (sctp_threshold_management(inp, stcb, cookie->whoTo,
1063 	    stcb->asoc.max_init_times)) {
1064 		/* Assoc is over */
1065 		return (1);
1066 	}
1067 	/*
1068 	 * cleared theshold management now lets backoff the address & select
1069 	 * an alternate
1070 	 */
1071 	stcb->asoc.dropped_special_cnt = 0;
1072 	sctp_backoff_on_timeout(stcb, cookie->whoTo, 1, 0, 0);
1073 	alt = sctp_find_alternate_net(stcb, cookie->whoTo, 0);
1074 	if (alt != cookie->whoTo) {
1075 		sctp_free_remote_addr(cookie->whoTo);
1076 		cookie->whoTo = alt;
1077 		atomic_add_int(&alt->ref_count, 1);
1078 	}
1079 	/* Now mark the retran info */
1080 	if (cookie->sent != SCTP_DATAGRAM_RESEND) {
1081 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1082 	}
1083 	cookie->sent = SCTP_DATAGRAM_RESEND;
1084 	/*
1085 	 * Now call the output routine to kick out the cookie again, Note we
1086 	 * don't mark any chunks for retran so that FR will need to kick in
1087 	 * to move these (or a send timer).
1088 	 */
1089 	return (0);
1090 }
1091 
1092 int
1093 sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1094     struct sctp_nets *net)
1095 {
1096 	struct sctp_nets *alt;
1097 	struct sctp_tmit_chunk *strrst = NULL, *chk = NULL;
1098 
1099 	if (stcb->asoc.stream_reset_outstanding == 0) {
1100 		return (0);
1101 	}
1102 	/* find the existing STRRESET, we use the seq number we sent out on */
1103 	(void)sctp_find_stream_reset(stcb, stcb->asoc.str_reset_seq_out, &strrst);
1104 	if (strrst == NULL) {
1105 		return (0);
1106 	}
1107 	/* do threshold management */
1108 	if (sctp_threshold_management(inp, stcb, strrst->whoTo,
1109 	    stcb->asoc.max_send_times)) {
1110 		/* Assoc is over */
1111 		return (1);
1112 	}
1113 	/*
1114 	 * cleared theshold management now lets backoff the address & select
1115 	 * an alternate
1116 	 */
1117 	sctp_backoff_on_timeout(stcb, strrst->whoTo, 1, 0, 0);
1118 	alt = sctp_find_alternate_net(stcb, strrst->whoTo, 0);
1119 	sctp_free_remote_addr(strrst->whoTo);
1120 	strrst->whoTo = alt;
1121 	atomic_add_int(&alt->ref_count, 1);
1122 
1123 	/* See if a ECN Echo is also stranded */
1124 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1125 		if ((chk->whoTo == net) &&
1126 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1127 			sctp_free_remote_addr(chk->whoTo);
1128 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
1129 				chk->sent = SCTP_DATAGRAM_RESEND;
1130 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1131 			}
1132 			chk->whoTo = alt;
1133 			atomic_add_int(&alt->ref_count, 1);
1134 		}
1135 	}
1136 	if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1137 		/*
1138 		 * If the address went un-reachable, we need to move to
1139 		 * alternates for ALL chk's in queue
1140 		 */
1141 		sctp_move_chunks_from_net(stcb, net);
1142 	}
1143 	/* mark the retran info */
1144 	if (strrst->sent != SCTP_DATAGRAM_RESEND)
1145 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1146 	strrst->sent = SCTP_DATAGRAM_RESEND;
1147 
1148 	/* restart the timer */
1149 	sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, strrst->whoTo);
1150 	return (0);
1151 }
1152 
1153 int
1154 sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1155     struct sctp_nets *net)
1156 {
1157 	struct sctp_nets *alt;
1158 	struct sctp_tmit_chunk *asconf, *chk;
1159 
1160 	/* is this a first send, or a retransmission? */
1161 	if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) {
1162 		/* compose a new ASCONF chunk and send it */
1163 		sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED);
1164 	} else {
1165 		/*
1166 		 * Retransmission of the existing ASCONF is needed
1167 		 */
1168 
1169 		/* find the existing ASCONF */
1170 		asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue);
1171 		if (asconf == NULL) {
1172 			return (0);
1173 		}
1174 		/* do threshold management */
1175 		if (sctp_threshold_management(inp, stcb, asconf->whoTo,
1176 		    stcb->asoc.max_send_times)) {
1177 			/* Assoc is over */
1178 			return (1);
1179 		}
1180 		if (asconf->snd_count > stcb->asoc.max_send_times) {
1181 			/*
1182 			 * Something is rotten: our peer is not responding
1183 			 * to ASCONFs but apparently is to other chunks.
1184 			 * i.e. it is not properly handling the chunk type
1185 			 * upper bits. Mark this peer as ASCONF incapable
1186 			 * and cleanup.
1187 			 */
1188 			SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n");
1189 			sctp_asconf_cleanup(stcb, net);
1190 			return (0);
1191 		}
1192 		/*
1193 		 * cleared threshold management, so now backoff the net and
1194 		 * select an alternate
1195 		 */
1196 		sctp_backoff_on_timeout(stcb, asconf->whoTo, 1, 0, 0);
1197 		alt = sctp_find_alternate_net(stcb, asconf->whoTo, 0);
1198 		if (asconf->whoTo != alt) {
1199 			sctp_free_remote_addr(asconf->whoTo);
1200 			asconf->whoTo = alt;
1201 			atomic_add_int(&alt->ref_count, 1);
1202 		}
1203 		/* See if an ECN Echo is also stranded */
1204 		TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1205 			if ((chk->whoTo == net) &&
1206 			    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1207 				sctp_free_remote_addr(chk->whoTo);
1208 				chk->whoTo = alt;
1209 				if (chk->sent != SCTP_DATAGRAM_RESEND) {
1210 					chk->sent = SCTP_DATAGRAM_RESEND;
1211 					sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1212 				}
1213 				atomic_add_int(&alt->ref_count, 1);
1214 			}
1215 		}
1216 		TAILQ_FOREACH(chk, &stcb->asoc.asconf_send_queue, sctp_next) {
1217 			if (chk->whoTo != alt) {
1218 				sctp_free_remote_addr(chk->whoTo);
1219 				chk->whoTo = alt;
1220 				atomic_add_int(&alt->ref_count, 1);
1221 			}
1222 			if (asconf->sent != SCTP_DATAGRAM_RESEND && chk->sent != SCTP_DATAGRAM_UNSENT)
1223 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1224 			chk->sent = SCTP_DATAGRAM_RESEND;
1225 		}
1226 		if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1227 			/*
1228 			 * If the address went un-reachable, we need to move
1229 			 * to the alternate for ALL chunks in queue
1230 			 */
1231 			sctp_move_chunks_from_net(stcb, net);
1232 		}
1233 		/* mark the retran info */
1234 		if (asconf->sent != SCTP_DATAGRAM_RESEND)
1235 			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1236 		asconf->sent = SCTP_DATAGRAM_RESEND;
1237 
1238 		/* send another ASCONF if any and we can do */
1239 		sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED);
1240 	}
1241 	return (0);
1242 }
1243 
1244 /* Mobility adaptation */
1245 void
1246 sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1247     struct sctp_nets *net SCTP_UNUSED)
1248 {
1249 	if (stcb->asoc.deleted_primary == NULL) {
1250 		SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n");
1251 		sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1252 		return;
1253 	}
1254 	SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary ");
1255 	SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa);
1256 	sctp_free_remote_addr(stcb->asoc.deleted_primary);
1257 	stcb->asoc.deleted_primary = NULL;
1258 	sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1259 	return;
1260 }
1261 
1262 /*
1263  * For the shutdown and shutdown-ack, we do not keep one around on the
1264  * control queue. This means we must generate a new one and call the general
1265  * chunk output routine, AFTER having done threshold management.
1266  * It is assumed that net is non-NULL.
1267  */
1268 int
1269 sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1270     struct sctp_nets *net)
1271 {
1272 	struct sctp_nets *alt;
1273 
1274 	/* first threshold managment */
1275 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1276 		/* Assoc is over */
1277 		return (1);
1278 	}
1279 	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1280 	/* second select an alternative */
1281 	alt = sctp_find_alternate_net(stcb, net, 0);
1282 
1283 	/* third generate a shutdown into the queue for out net */
1284 	sctp_send_shutdown(stcb, alt);
1285 
1286 	/* fourth restart timer */
1287 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt);
1288 	return (0);
1289 }
1290 
1291 int
1292 sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1293     struct sctp_nets *net)
1294 {
1295 	struct sctp_nets *alt;
1296 
1297 	/* first threshold managment */
1298 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1299 		/* Assoc is over */
1300 		return (1);
1301 	}
1302 	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1303 	/* second select an alternative */
1304 	alt = sctp_find_alternate_net(stcb, net, 0);
1305 
1306 	/* third generate a shutdown into the queue for out net */
1307 	sctp_send_shutdown_ack(stcb, alt);
1308 
1309 	/* fourth restart timer */
1310 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt);
1311 	return (0);
1312 }
1313 
1314 static void
1315 sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp,
1316     struct sctp_tcb *stcb)
1317 {
1318 	struct sctp_stream_queue_pending *sp;
1319 	unsigned int i, chks_in_queue = 0;
1320 	int being_filled = 0;
1321 
1322 	/*
1323 	 * This function is ONLY called when the send/sent queues are empty.
1324 	 */
1325 	if ((stcb == NULL) || (inp == NULL))
1326 		return;
1327 
1328 	if (stcb->asoc.sent_queue_retran_cnt) {
1329 		SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n",
1330 		    stcb->asoc.sent_queue_retran_cnt);
1331 		stcb->asoc.sent_queue_retran_cnt = 0;
1332 	}
1333 	if (stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1334 		/* No stream scheduler information, initialize scheduler */
1335 		stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc, 0);
1336 		if (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1337 			/* yep, we lost a stream or two */
1338 			SCTP_PRINTF("Found additional streams NOT managed by scheduler, corrected\n");
1339 		} else {
1340 			/* no streams lost */
1341 			stcb->asoc.total_output_queue_size = 0;
1342 		}
1343 	}
1344 	/* Check to see if some data queued, if so report it */
1345 	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
1346 		if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) {
1347 			TAILQ_FOREACH(sp, &stcb->asoc.strmout[i].outqueue, next) {
1348 				if (sp->msg_is_complete)
1349 					being_filled++;
1350 				chks_in_queue++;
1351 			}
1352 		}
1353 	}
1354 	if (chks_in_queue != stcb->asoc.stream_queue_cnt) {
1355 		SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n",
1356 		    stcb->asoc.stream_queue_cnt, chks_in_queue);
1357 	}
1358 	if (chks_in_queue) {
1359 		/* call the output queue function */
1360 		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1361 		if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1362 		    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1363 			/*
1364 			 * Probably should go in and make it go back through
1365 			 * and add fragments allowed
1366 			 */
1367 			if (being_filled == 0) {
1368 				SCTP_PRINTF("Still nothing moved %d chunks are stuck\n",
1369 				    chks_in_queue);
1370 			}
1371 		}
1372 	} else {
1373 		SCTP_PRINTF("Found no chunks on any queue tot:%lu\n",
1374 		    (u_long)stcb->asoc.total_output_queue_size);
1375 		stcb->asoc.total_output_queue_size = 0;
1376 	}
1377 }
1378 
1379 int
1380 sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1381     struct sctp_nets *net)
1382 {
1383 	uint8_t net_was_pf;
1384 
1385 	if (net->dest_state & SCTP_ADDR_PF) {
1386 		net_was_pf = 1;
1387 	} else {
1388 		net_was_pf = 0;
1389 	}
1390 	if (net->hb_responded == 0) {
1391 		if (net->ro._s_addr) {
1392 			/*
1393 			 * Invalidate the src address if we did not get a
1394 			 * response last time.
1395 			 */
1396 			sctp_free_ifa(net->ro._s_addr);
1397 			net->ro._s_addr = NULL;
1398 			net->src_addr_selected = 0;
1399 		}
1400 		sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1401 		if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1402 			/* Assoc is over */
1403 			return (1);
1404 		}
1405 	}
1406 	/* Zero PBA, if it needs it */
1407 	if (net->partial_bytes_acked) {
1408 		net->partial_bytes_acked = 0;
1409 	}
1410 	if ((stcb->asoc.total_output_queue_size > 0) &&
1411 	    (TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1412 	    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1413 		sctp_audit_stream_queues_for_size(inp, stcb);
1414 	}
1415 	if (!(net->dest_state & SCTP_ADDR_NOHB) &&
1416 	    !((net_was_pf == 0) && (net->dest_state & SCTP_ADDR_PF))) {
1417 		/*
1418 		 * when move to PF during threshold mangement, a HB has been
1419 		 * queued in that routine
1420 		 */
1421 		uint32_t ms_gone_by;
1422 
1423 		if ((net->last_sent_time.tv_sec > 0) ||
1424 		    (net->last_sent_time.tv_usec > 0)) {
1425 			struct timeval diff;
1426 
1427 			SCTP_GETTIME_TIMEVAL(&diff);
1428 			timevalsub(&diff, &net->last_sent_time);
1429 			ms_gone_by = (uint32_t) (diff.tv_sec * 1000) +
1430 			    (uint32_t) (diff.tv_usec / 1000);
1431 		} else {
1432 			ms_gone_by = 0xffffffff;
1433 		}
1434 		if ((ms_gone_by >= net->heart_beat_delay) ||
1435 		    (net->dest_state & SCTP_ADDR_PF)) {
1436 			sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
1437 		}
1438 	}
1439 	return (0);
1440 }
1441 
1442 void
1443 sctp_pathmtu_timer(struct sctp_inpcb *inp,
1444     struct sctp_tcb *stcb,
1445     struct sctp_nets *net)
1446 {
1447 	uint32_t next_mtu, mtu;
1448 
1449 	next_mtu = sctp_get_next_mtu(net->mtu);
1450 
1451 	if ((next_mtu > net->mtu) && (net->port == 0)) {
1452 		if ((net->src_addr_selected == 0) ||
1453 		    (net->ro._s_addr == NULL) ||
1454 		    (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1455 			if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1456 				sctp_free_ifa(net->ro._s_addr);
1457 				net->ro._s_addr = NULL;
1458 				net->src_addr_selected = 0;
1459 			} else if (net->ro._s_addr == NULL) {
1460 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1461 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1462 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1463 
1464 					/* KAME hack: embed scopeid */
1465 					(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
1466 				}
1467 #endif
1468 
1469 				net->ro._s_addr = sctp_source_address_selection(inp,
1470 				    stcb,
1471 				    (sctp_route_t *) & net->ro,
1472 				    net, 0, stcb->asoc.vrf_id);
1473 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1474 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1475 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1476 
1477 					(void)sa6_recoverscope(sin6);
1478 				}
1479 #endif				/* INET6 */
1480 			}
1481 			if (net->ro._s_addr)
1482 				net->src_addr_selected = 1;
1483 		}
1484 		if (net->ro._s_addr) {
1485 			mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_rt);
1486 #if defined(INET) || defined(INET6)
1487 			if (net->port) {
1488 				mtu -= sizeof(struct udphdr);
1489 			}
1490 #endif
1491 			if (mtu > next_mtu) {
1492 				net->mtu = next_mtu;
1493 			}
1494 		}
1495 	}
1496 	/* restart the timer */
1497 	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
1498 }
1499 
1500 void
1501 sctp_autoclose_timer(struct sctp_inpcb *inp,
1502     struct sctp_tcb *stcb,
1503     struct sctp_nets *net)
1504 {
1505 	struct timeval tn, *tim_touse;
1506 	struct sctp_association *asoc;
1507 	int ticks_gone_by;
1508 
1509 	(void)SCTP_GETTIME_TIMEVAL(&tn);
1510 	if (stcb->asoc.sctp_autoclose_ticks &&
1511 	    sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) {
1512 		/* Auto close is on */
1513 		asoc = &stcb->asoc;
1514 		/* pick the time to use */
1515 		if (asoc->time_last_rcvd.tv_sec >
1516 		    asoc->time_last_sent.tv_sec) {
1517 			tim_touse = &asoc->time_last_rcvd;
1518 		} else {
1519 			tim_touse = &asoc->time_last_sent;
1520 		}
1521 		/* Now has long enough transpired to autoclose? */
1522 		ticks_gone_by = SEC_TO_TICKS(tn.tv_sec - tim_touse->tv_sec);
1523 		if ((ticks_gone_by > 0) &&
1524 		    (ticks_gone_by >= (int)asoc->sctp_autoclose_ticks)) {
1525 			/*
1526 			 * autoclose time has hit, call the output routine,
1527 			 * which should do nothing just to be SURE we don't
1528 			 * have hanging data. We can then safely check the
1529 			 * queues and know that we are clear to send
1530 			 * shutdown
1531 			 */
1532 			sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED);
1533 			/* Are we clean? */
1534 			if (TAILQ_EMPTY(&asoc->send_queue) &&
1535 			    TAILQ_EMPTY(&asoc->sent_queue)) {
1536 				/*
1537 				 * there is nothing queued to send, so I'm
1538 				 * done...
1539 				 */
1540 				if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) {
1541 					/* only send SHUTDOWN 1st time thru */
1542 					struct sctp_nets *netp;
1543 
1544 					if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
1545 					    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
1546 						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
1547 					}
1548 					SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
1549 					SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
1550 					sctp_stop_timers_for_shutdown(stcb);
1551 					if (stcb->asoc.alternate) {
1552 						netp = stcb->asoc.alternate;
1553 					} else {
1554 						netp = stcb->asoc.primary_destination;
1555 					}
1556 					sctp_send_shutdown(stcb, netp);
1557 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
1558 					    stcb->sctp_ep, stcb,
1559 					    netp);
1560 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
1561 					    stcb->sctp_ep, stcb,
1562 					    netp);
1563 				}
1564 			}
1565 		} else {
1566 			/*
1567 			 * No auto close at this time, reset t-o to check
1568 			 * later
1569 			 */
1570 			int tmp;
1571 
1572 			/* fool the timer startup to use the time left */
1573 			tmp = asoc->sctp_autoclose_ticks;
1574 			asoc->sctp_autoclose_ticks -= ticks_gone_by;
1575 			sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb,
1576 			    net);
1577 			/* restore the real tick value */
1578 			asoc->sctp_autoclose_ticks = tmp;
1579 		}
1580 	}
1581 }
1582