xref: /freebsd/sys/netinet/sctp_timer.c (revision 0e97acdf58fe27b09c4824a474b0344daf997c5f)
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 couter 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 	do {
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 		/* sa_ignore NO_NULL_CHK */
360 		if (((alt->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE) &&
361 		    (alt->ro.ro_rt != NULL) &&
362 		    (!(alt->dest_state & SCTP_ADDR_UNCONFIRMED))) {
363 			/* Found a reachable address */
364 			break;
365 		}
366 		mnet = alt;
367 	} while (alt != NULL);
368 
369 	if (alt == NULL) {
370 		/* Case where NO insv network exists (dormant state) */
371 		/* we rotate destinations */
372 		once = 0;
373 		mnet = net;
374 		do {
375 			if (mnet == NULL) {
376 				return (TAILQ_FIRST(&stcb->asoc.nets));
377 			}
378 			alt = TAILQ_NEXT(mnet, sctp_next);
379 			if (alt == NULL) {
380 				once++;
381 				if (once > 1) {
382 					break;
383 				}
384 				alt = TAILQ_FIRST(&stcb->asoc.nets);
385 			}
386 			/* sa_ignore NO_NULL_CHK */
387 			if ((!(alt->dest_state & SCTP_ADDR_UNCONFIRMED)) &&
388 			    (alt != net)) {
389 				/* Found an alternate address */
390 				break;
391 			}
392 			mnet = alt;
393 		} while (alt != NULL);
394 	}
395 	if (alt == NULL) {
396 		return (net);
397 	}
398 	return (alt);
399 }
400 
401 static void
402 sctp_backoff_on_timeout(struct sctp_tcb *stcb,
403     struct sctp_nets *net,
404     int win_probe,
405     int num_marked, int num_abandoned)
406 {
407 	if (net->RTO == 0) {
408 		net->RTO = stcb->asoc.minrto;
409 	}
410 	net->RTO <<= 1;
411 	if (net->RTO > stcb->asoc.maxrto) {
412 		net->RTO = stcb->asoc.maxrto;
413 	}
414 	if ((win_probe == 0) && (num_marked || num_abandoned)) {
415 		/* We don't apply penalty to window probe scenarios */
416 		/* JRS - Use the congestion control given in the CC module */
417 		stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout(stcb, net);
418 	}
419 }
420 
421 #ifndef INVARIANTS
422 static void
423 sctp_recover_sent_list(struct sctp_tcb *stcb)
424 {
425 	struct sctp_tmit_chunk *chk, *nchk;
426 	struct sctp_association *asoc;
427 
428 	asoc = &stcb->asoc;
429 	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
430 		if (SCTP_TSN_GE(asoc->last_acked_seq, chk->rec.data.TSN_seq)) {
431 			SCTP_PRINTF("Found chk:%p tsn:%x <= last_acked_seq:%x\n",
432 			    (void *)chk, chk->rec.data.TSN_seq, asoc->last_acked_seq);
433 			if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
434 				if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
435 					asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
436 				}
437 			}
438 			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
439 			if (PR_SCTP_ENABLED(chk->flags)) {
440 				if (asoc->pr_sctp_cnt != 0)
441 					asoc->pr_sctp_cnt--;
442 			}
443 			if (chk->data) {
444 				/* sa_ignore NO_NULL_CHK */
445 				sctp_free_bufspace(stcb, asoc, chk, 1);
446 				sctp_m_freem(chk->data);
447 				chk->data = NULL;
448 				if (asoc->prsctp_supported && PR_SCTP_BUF_ENABLED(chk->flags)) {
449 					asoc->sent_queue_cnt_removeable--;
450 				}
451 			}
452 			asoc->sent_queue_cnt--;
453 			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
454 		}
455 	}
456 	SCTP_PRINTF("after recover order is as follows\n");
457 	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
458 		SCTP_PRINTF("chk:%p TSN:%x\n", (void *)chk, chk->rec.data.TSN_seq);
459 	}
460 }
461 
462 #endif
463 
464 static int
465 sctp_mark_all_for_resend(struct sctp_tcb *stcb,
466     struct sctp_nets *net,
467     struct sctp_nets *alt,
468     int window_probe,
469     int *num_marked,
470     int *num_abandoned)
471 {
472 
473 	/*
474 	 * Mark all chunks (well not all) that were sent to *net for
475 	 * retransmission. Move them to alt for there destination as well...
476 	 * We only mark chunks that have been outstanding long enough to
477 	 * have received feed-back.
478 	 */
479 	struct sctp_tmit_chunk *chk, *nchk;
480 	struct sctp_nets *lnets;
481 	struct timeval now, min_wait, tv;
482 	int cur_rto;
483 	int cnt_abandoned;
484 	int audit_tf, num_mk, fir;
485 	unsigned int cnt_mk;
486 	uint32_t orig_flight, orig_tf;
487 	uint32_t tsnlast, tsnfirst;
488 	int recovery_cnt = 0;
489 
490 
491 	/* none in flight now */
492 	audit_tf = 0;
493 	fir = 0;
494 	/*
495 	 * figure out how long a data chunk must be pending before we can
496 	 * mark it ..
497 	 */
498 	(void)SCTP_GETTIME_TIMEVAL(&now);
499 	/* get cur rto in micro-seconds */
500 	cur_rto = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv;
501 	cur_rto *= 1000;
502 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
503 		sctp_log_fr(cur_rto,
504 		    stcb->asoc.peers_rwnd,
505 		    window_probe,
506 		    SCTP_FR_T3_MARK_TIME);
507 		sctp_log_fr(net->flight_size, 0, 0, SCTP_FR_CWND_REPORT);
508 		sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT);
509 	}
510 	tv.tv_sec = cur_rto / 1000000;
511 	tv.tv_usec = cur_rto % 1000000;
512 	min_wait = now;
513 	timevalsub(&min_wait, &tv);
514 	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
515 		/*
516 		 * if we hit here, we don't have enough seconds on the clock
517 		 * to account for the RTO. We just let the lower seconds be
518 		 * the bounds and don't worry about it. This may mean we
519 		 * will mark a lot more than we should.
520 		 */
521 		min_wait.tv_sec = min_wait.tv_usec = 0;
522 	}
523 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
524 		sctp_log_fr(cur_rto, now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME);
525 		sctp_log_fr(0, min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME);
526 	}
527 	/*
528 	 * Our rwnd will be incorrect here since we are not adding back the
529 	 * cnt * mbuf but we will fix that down below.
530 	 */
531 	orig_flight = net->flight_size;
532 	orig_tf = stcb->asoc.total_flight;
533 
534 	net->fast_retran_ip = 0;
535 	/* Now on to each chunk */
536 	cnt_abandoned = 0;
537 	num_mk = cnt_mk = 0;
538 	tsnfirst = tsnlast = 0;
539 #ifndef INVARIANTS
540 start_again:
541 #endif
542 	TAILQ_FOREACH_SAFE(chk, &stcb->asoc.sent_queue, sctp_next, nchk) {
543 		if (SCTP_TSN_GE(stcb->asoc.last_acked_seq, chk->rec.data.TSN_seq)) {
544 			/* Strange case our list got out of order? */
545 			SCTP_PRINTF("Our list is out of order? last_acked:%x chk:%x\n",
546 			    (unsigned int)stcb->asoc.last_acked_seq, (unsigned int)chk->rec.data.TSN_seq);
547 			recovery_cnt++;
548 #ifdef INVARIANTS
549 			panic("last acked >= chk on sent-Q");
550 #else
551 			SCTP_PRINTF("Recover attempts a restart cnt:%d\n", recovery_cnt);
552 			sctp_recover_sent_list(stcb);
553 			if (recovery_cnt < 10) {
554 				goto start_again;
555 			} else {
556 				SCTP_PRINTF("Recovery fails %d times??\n", recovery_cnt);
557 			}
558 #endif
559 		}
560 		if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) {
561 			/*
562 			 * found one to mark: If it is less than
563 			 * DATAGRAM_ACKED it MUST not be a skipped or marked
564 			 * TSN but instead one that is either already set
565 			 * for retransmission OR one that needs
566 			 * retransmission.
567 			 */
568 
569 			/* validate its been outstanding long enough */
570 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
571 				sctp_log_fr(chk->rec.data.TSN_seq,
572 				    chk->sent_rcv_time.tv_sec,
573 				    chk->sent_rcv_time.tv_usec,
574 				    SCTP_FR_T3_MARK_TIME);
575 			}
576 			if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) {
577 				/*
578 				 * we have reached a chunk that was sent
579 				 * some seconds past our min.. forget it we
580 				 * will find no more to send.
581 				 */
582 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
583 					sctp_log_fr(0,
584 					    chk->sent_rcv_time.tv_sec,
585 					    chk->sent_rcv_time.tv_usec,
586 					    SCTP_FR_T3_STOPPED);
587 				}
588 				continue;
589 			} else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) &&
590 			    (window_probe == 0)) {
591 				/*
592 				 * we must look at the micro seconds to
593 				 * know.
594 				 */
595 				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
596 					/*
597 					 * ok it was sent after our boundary
598 					 * time.
599 					 */
600 					continue;
601 				}
602 			}
603 			if (stcb->asoc.prsctp_supported && PR_SCTP_TTL_ENABLED(chk->flags)) {
604 				/* Is it expired? */
605 				if (timevalcmp(&now, &chk->rec.data.timetodrop, >)) {
606 					/* Yes so drop it */
607 					if (chk->data) {
608 						(void)sctp_release_pr_sctp_chunk(stcb,
609 						    chk,
610 						    1,
611 						    SCTP_SO_NOT_LOCKED);
612 						cnt_abandoned++;
613 					}
614 					continue;
615 				}
616 			}
617 			if (stcb->asoc.prsctp_supported && PR_SCTP_RTX_ENABLED(chk->flags)) {
618 				/* Has it been retransmitted tv_sec times? */
619 				if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) {
620 					if (chk->data) {
621 						(void)sctp_release_pr_sctp_chunk(stcb,
622 						    chk,
623 						    1,
624 						    SCTP_SO_NOT_LOCKED);
625 						cnt_abandoned++;
626 					}
627 					continue;
628 				}
629 			}
630 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
631 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
632 				num_mk++;
633 				if (fir == 0) {
634 					fir = 1;
635 					tsnfirst = chk->rec.data.TSN_seq;
636 				}
637 				tsnlast = chk->rec.data.TSN_seq;
638 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
639 					sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count,
640 					    0, SCTP_FR_T3_MARKED);
641 				}
642 				if (chk->rec.data.chunk_was_revoked) {
643 					/* deflate the cwnd */
644 					chk->whoTo->cwnd -= chk->book_size;
645 					chk->rec.data.chunk_was_revoked = 0;
646 				}
647 				net->marked_retrans++;
648 				stcb->asoc.marked_retrans++;
649 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
650 					sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO,
651 					    chk->whoTo->flight_size,
652 					    chk->book_size,
653 					    (uintptr_t) chk->whoTo,
654 					    chk->rec.data.TSN_seq);
655 				}
656 				sctp_flight_size_decrease(chk);
657 				sctp_total_flight_decrease(stcb, chk);
658 				stcb->asoc.peers_rwnd += chk->send_size;
659 				stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh);
660 			}
661 			chk->sent = SCTP_DATAGRAM_RESEND;
662 			SCTP_STAT_INCR(sctps_markedretrans);
663 
664 			/* reset the TSN for striking and other FR stuff */
665 			chk->rec.data.doing_fast_retransmit = 0;
666 			/* Clear any time so NO RTT is being done */
667 
668 			if (chk->do_rtt) {
669 				if (chk->whoTo->rto_needed == 0) {
670 					chk->whoTo->rto_needed = 1;
671 				}
672 			}
673 			chk->do_rtt = 0;
674 			if (alt != net) {
675 				sctp_free_remote_addr(chk->whoTo);
676 				chk->no_fr_allowed = 1;
677 				chk->whoTo = alt;
678 				atomic_add_int(&alt->ref_count, 1);
679 			} else {
680 				chk->no_fr_allowed = 0;
681 				if (TAILQ_EMPTY(&stcb->asoc.send_queue)) {
682 					chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq;
683 				} else {
684 					chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.TSN_seq;
685 				}
686 			}
687 			/*
688 			 * CMT: Do not allow FRs on retransmitted TSNs.
689 			 */
690 			if (stcb->asoc.sctp_cmt_on_off > 0) {
691 				chk->no_fr_allowed = 1;
692 			}
693 #ifdef THIS_SHOULD_NOT_BE_DONE
694 		} else if (chk->sent == SCTP_DATAGRAM_ACKED) {
695 			/* remember highest acked one */
696 			could_be_sent = chk;
697 #endif
698 		}
699 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
700 			cnt_mk++;
701 		}
702 	}
703 	if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) {
704 		/* we did not subtract the same things? */
705 		audit_tf = 1;
706 	}
707 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
708 		sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT);
709 	}
710 #ifdef SCTP_DEBUG
711 	if (num_mk) {
712 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
713 		    tsnlast);
714 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%ld\n",
715 		    num_mk, (u_long)stcb->asoc.peers_rwnd);
716 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
717 		    tsnlast);
718 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%d\n",
719 		    num_mk,
720 		    (int)stcb->asoc.peers_rwnd);
721 	}
722 #endif
723 	*num_marked = num_mk;
724 	*num_abandoned = cnt_abandoned;
725 	/*
726 	 * Now check for a ECN Echo that may be stranded And include the
727 	 * cnt_mk'd to have all resends in the control queue.
728 	 */
729 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
730 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
731 			cnt_mk++;
732 		}
733 		if ((chk->whoTo == net) &&
734 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
735 			sctp_free_remote_addr(chk->whoTo);
736 			chk->whoTo = alt;
737 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
738 				chk->sent = SCTP_DATAGRAM_RESEND;
739 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
740 				cnt_mk++;
741 			}
742 			atomic_add_int(&alt->ref_count, 1);
743 		}
744 	}
745 #ifdef THIS_SHOULD_NOT_BE_DONE
746 	if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) {
747 		/* fix it so we retransmit the highest acked anyway */
748 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
749 		cnt_mk++;
750 		could_be_sent->sent = SCTP_DATAGRAM_RESEND;
751 	}
752 #endif
753 	if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) {
754 #ifdef INVARIANTS
755 		SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n",
756 		    cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk);
757 #endif
758 #ifndef SCTP_AUDITING_ENABLED
759 		stcb->asoc.sent_queue_retran_cnt = cnt_mk;
760 #endif
761 	}
762 	if (audit_tf) {
763 		SCTPDBG(SCTP_DEBUG_TIMER4,
764 		    "Audit total flight due to negative value net:%p\n",
765 		    (void *)net);
766 		stcb->asoc.total_flight = 0;
767 		stcb->asoc.total_flight_count = 0;
768 		/* Clear all networks flight size */
769 		TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) {
770 			lnets->flight_size = 0;
771 			SCTPDBG(SCTP_DEBUG_TIMER4,
772 			    "Net:%p c-f cwnd:%d ssthresh:%d\n",
773 			    (void *)lnets, lnets->cwnd, lnets->ssthresh);
774 		}
775 		TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
776 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
777 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
778 					sctp_misc_ints(SCTP_FLIGHT_LOG_UP,
779 					    chk->whoTo->flight_size,
780 					    chk->book_size,
781 					    (uintptr_t) chk->whoTo,
782 					    chk->rec.data.TSN_seq);
783 				}
784 				sctp_flight_size_increase(chk);
785 				sctp_total_flight_increase(stcb, chk);
786 			}
787 		}
788 	}
789 	/* We return 1 if we only have a window probe outstanding */
790 	return (0);
791 }
792 
793 
794 int
795 sctp_t3rxt_timer(struct sctp_inpcb *inp,
796     struct sctp_tcb *stcb,
797     struct sctp_nets *net)
798 {
799 	struct sctp_nets *alt;
800 	int win_probe, num_mk, num_abandoned;
801 
802 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
803 		sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT);
804 	}
805 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
806 		struct sctp_nets *lnet;
807 
808 		TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
809 			if (net == lnet) {
810 				sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3);
811 			} else {
812 				sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3);
813 			}
814 		}
815 	}
816 	/* Find an alternate and mark those for retransmission */
817 	if ((stcb->asoc.peers_rwnd == 0) &&
818 	    (stcb->asoc.total_flight < net->mtu)) {
819 		SCTP_STAT_INCR(sctps_timowindowprobe);
820 		win_probe = 1;
821 	} else {
822 		win_probe = 0;
823 	}
824 
825 	if (win_probe == 0) {
826 		/* We don't do normal threshold management on window probes */
827 		if (sctp_threshold_management(inp, stcb, net,
828 		    stcb->asoc.max_send_times)) {
829 			/* Association was destroyed */
830 			return (1);
831 		} else {
832 			if (net != stcb->asoc.primary_destination) {
833 				/* send a immediate HB if our RTO is stale */
834 				struct timeval now;
835 				unsigned int ms_goneby;
836 
837 				(void)SCTP_GETTIME_TIMEVAL(&now);
838 				if (net->last_sent_time.tv_sec) {
839 					ms_goneby = (now.tv_sec - net->last_sent_time.tv_sec) * 1000;
840 				} else {
841 					ms_goneby = 0;
842 				}
843 				if ((net->dest_state & SCTP_ADDR_PF) == 0) {
844 					if ((ms_goneby > net->RTO) || (net->RTO == 0)) {
845 						/*
846 						 * no recent feed back in an
847 						 * RTO or more, request a
848 						 * RTT update
849 						 */
850 						sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
851 					}
852 				}
853 			}
854 		}
855 	} else {
856 		/*
857 		 * For a window probe we don't penalize the net's but only
858 		 * the association. This may fail it if SACKs are not coming
859 		 * back. If sack's are coming with rwnd locked at 0, we will
860 		 * continue to hold things waiting for rwnd to raise
861 		 */
862 		if (sctp_threshold_management(inp, stcb, NULL,
863 		    stcb->asoc.max_send_times)) {
864 			/* Association was destroyed */
865 			return (1);
866 		}
867 	}
868 	if (stcb->asoc.sctp_cmt_on_off > 0) {
869 		if (net->pf_threshold < net->failure_threshold) {
870 			alt = sctp_find_alternate_net(stcb, net, 2);
871 		} else {
872 			/*
873 			 * CMT: Using RTX_SSTHRESH policy for CMT. If CMT is
874 			 * being used, then pick dest with largest ssthresh
875 			 * for any retransmission.
876 			 */
877 			alt = sctp_find_alternate_net(stcb, net, 1);
878 			/*
879 			 * CUCv2: If a different dest is picked for the
880 			 * retransmission, then new (rtx-)pseudo_cumack
881 			 * needs to be tracked for orig dest. Let CUCv2
882 			 * track new (rtx-) pseudo-cumack always.
883 			 */
884 			net->find_pseudo_cumack = 1;
885 			net->find_rtx_pseudo_cumack = 1;
886 		}
887 	} else {
888 		alt = sctp_find_alternate_net(stcb, net, 0);
889 	}
890 
891 	num_mk = 0;
892 	num_abandoned = 0;
893 	(void)sctp_mark_all_for_resend(stcb, net, alt, win_probe,
894 	    &num_mk, &num_abandoned);
895 	/* FR Loss recovery just ended with the T3. */
896 	stcb->asoc.fast_retran_loss_recovery = 0;
897 
898 	/* CMT FR loss recovery ended with the T3 */
899 	net->fast_retran_loss_recovery = 0;
900 	if ((stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) &&
901 	    (net->flight_size == 0)) {
902 		(*stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) (stcb, net);
903 	}
904 	/*
905 	 * setup the sat loss recovery that prevents satellite cwnd advance.
906 	 */
907 	stcb->asoc.sat_t3_loss_recovery = 1;
908 	stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq;
909 
910 	/* Backoff the timer and cwnd */
911 	sctp_backoff_on_timeout(stcb, net, win_probe, num_mk, num_abandoned);
912 	if ((!(net->dest_state & SCTP_ADDR_REACHABLE)) ||
913 	    (net->dest_state & SCTP_ADDR_PF)) {
914 		/* Move all pending over too */
915 		sctp_move_chunks_from_net(stcb, net);
916 
917 		/*
918 		 * Get the address that failed, to force a new src address
919 		 * selecton and a route allocation.
920 		 */
921 		if (net->ro._s_addr) {
922 			sctp_free_ifa(net->ro._s_addr);
923 			net->ro._s_addr = NULL;
924 		}
925 		net->src_addr_selected = 0;
926 
927 		/* Force a route allocation too */
928 		if (net->ro.ro_rt) {
929 			RTFREE(net->ro.ro_rt);
930 			net->ro.ro_rt = NULL;
931 		}
932 		/* Was it our primary? */
933 		if ((stcb->asoc.primary_destination == net) && (alt != net)) {
934 			/*
935 			 * Yes, note it as such and find an alternate note:
936 			 * this means HB code must use this to resent the
937 			 * primary if it goes active AND if someone does a
938 			 * change-primary then this flag must be cleared
939 			 * from any net structures.
940 			 */
941 			if (stcb->asoc.alternate) {
942 				sctp_free_remote_addr(stcb->asoc.alternate);
943 			}
944 			stcb->asoc.alternate = alt;
945 			atomic_add_int(&stcb->asoc.alternate->ref_count, 1);
946 		}
947 	}
948 	/*
949 	 * Special case for cookie-echo'ed case, we don't do output but must
950 	 * await the COOKIE-ACK before retransmission
951 	 */
952 	if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
953 		/*
954 		 * Here we just reset the timer and start again since we
955 		 * have not established the asoc
956 		 */
957 		sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net);
958 		return (0);
959 	}
960 	if (stcb->asoc.prsctp_supported) {
961 		struct sctp_tmit_chunk *lchk;
962 
963 		lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc);
964 		/* C3. See if we need to send a Fwd-TSN */
965 		if (SCTP_TSN_GT(stcb->asoc.advanced_peer_ack_point, stcb->asoc.last_acked_seq)) {
966 			send_forward_tsn(stcb, &stcb->asoc);
967 			if (lchk) {
968 				/* Assure a timer is up */
969 				sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo);
970 			}
971 		}
972 	}
973 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
974 		sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX);
975 	}
976 	return (0);
977 }
978 
979 int
980 sctp_t1init_timer(struct sctp_inpcb *inp,
981     struct sctp_tcb *stcb,
982     struct sctp_nets *net)
983 {
984 	/* bump the thresholds */
985 	if (stcb->asoc.delayed_connection) {
986 		/*
987 		 * special hook for delayed connection. The library did NOT
988 		 * complete the rest of its sends.
989 		 */
990 		stcb->asoc.delayed_connection = 0;
991 		sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
992 		return (0);
993 	}
994 	if (SCTP_GET_STATE((&stcb->asoc)) != SCTP_STATE_COOKIE_WAIT) {
995 		return (0);
996 	}
997 	if (sctp_threshold_management(inp, stcb, net,
998 	    stcb->asoc.max_init_times)) {
999 		/* Association was destroyed */
1000 		return (1);
1001 	}
1002 	stcb->asoc.dropped_special_cnt = 0;
1003 	sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0, 0);
1004 	if (stcb->asoc.initial_init_rto_max < net->RTO) {
1005 		net->RTO = stcb->asoc.initial_init_rto_max;
1006 	}
1007 	if (stcb->asoc.numnets > 1) {
1008 		/* If we have more than one addr use it */
1009 		struct sctp_nets *alt;
1010 
1011 		alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0);
1012 		if (alt != stcb->asoc.primary_destination) {
1013 			sctp_move_chunks_from_net(stcb, stcb->asoc.primary_destination);
1014 			stcb->asoc.primary_destination = alt;
1015 		}
1016 	}
1017 	/* Send out a new init */
1018 	sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1019 	return (0);
1020 }
1021 
1022 /*
1023  * For cookie and asconf we actually need to find and mark for resend, then
1024  * increment the resend counter (after all the threshold management stuff of
1025  * course).
1026  */
1027 int
1028 sctp_cookie_timer(struct sctp_inpcb *inp,
1029     struct sctp_tcb *stcb,
1030     struct sctp_nets *net SCTP_UNUSED)
1031 {
1032 	struct sctp_nets *alt;
1033 	struct sctp_tmit_chunk *cookie;
1034 
1035 	/* first before all else we must find the cookie */
1036 	TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) {
1037 		if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) {
1038 			break;
1039 		}
1040 	}
1041 	if (cookie == NULL) {
1042 		if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
1043 			/* FOOBAR! */
1044 			struct mbuf *op_err;
1045 
1046 			op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION,
1047 			    "Cookie timer expired, but no cookie");
1048 			inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_4;
1049 			sctp_abort_an_association(inp, stcb, op_err, SCTP_SO_NOT_LOCKED);
1050 		} else {
1051 #ifdef INVARIANTS
1052 			panic("Cookie timer expires in wrong state?");
1053 #else
1054 			SCTP_PRINTF("Strange in state %d not cookie-echoed yet c-e timer expires?\n", SCTP_GET_STATE(&stcb->asoc));
1055 			return (0);
1056 #endif
1057 		}
1058 		return (0);
1059 	}
1060 	/* Ok we found the cookie, threshold management next */
1061 	if (sctp_threshold_management(inp, stcb, cookie->whoTo,
1062 	    stcb->asoc.max_init_times)) {
1063 		/* Assoc is over */
1064 		return (1);
1065 	}
1066 	/*
1067 	 * cleared theshold management now lets backoff the address & select
1068 	 * an alternate
1069 	 */
1070 	stcb->asoc.dropped_special_cnt = 0;
1071 	sctp_backoff_on_timeout(stcb, cookie->whoTo, 1, 0, 0);
1072 	alt = sctp_find_alternate_net(stcb, cookie->whoTo, 0);
1073 	if (alt != cookie->whoTo) {
1074 		sctp_free_remote_addr(cookie->whoTo);
1075 		cookie->whoTo = alt;
1076 		atomic_add_int(&alt->ref_count, 1);
1077 	}
1078 	/* Now mark the retran info */
1079 	if (cookie->sent != SCTP_DATAGRAM_RESEND) {
1080 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1081 	}
1082 	cookie->sent = SCTP_DATAGRAM_RESEND;
1083 	/*
1084 	 * Now call the output routine to kick out the cookie again, Note we
1085 	 * don't mark any chunks for retran so that FR will need to kick in
1086 	 * to move these (or a send timer).
1087 	 */
1088 	return (0);
1089 }
1090 
1091 int
1092 sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1093     struct sctp_nets *net)
1094 {
1095 	struct sctp_nets *alt;
1096 	struct sctp_tmit_chunk *strrst = NULL, *chk = NULL;
1097 
1098 	if (stcb->asoc.stream_reset_outstanding == 0) {
1099 		return (0);
1100 	}
1101 	/* find the existing STRRESET, we use the seq number we sent out on */
1102 	(void)sctp_find_stream_reset(stcb, stcb->asoc.str_reset_seq_out, &strrst);
1103 	if (strrst == NULL) {
1104 		return (0);
1105 	}
1106 	/* do threshold management */
1107 	if (sctp_threshold_management(inp, stcb, strrst->whoTo,
1108 	    stcb->asoc.max_send_times)) {
1109 		/* Assoc is over */
1110 		return (1);
1111 	}
1112 	/*
1113 	 * cleared theshold management now lets backoff the address & select
1114 	 * an alternate
1115 	 */
1116 	sctp_backoff_on_timeout(stcb, strrst->whoTo, 1, 0, 0);
1117 	alt = sctp_find_alternate_net(stcb, strrst->whoTo, 0);
1118 	sctp_free_remote_addr(strrst->whoTo);
1119 	strrst->whoTo = alt;
1120 	atomic_add_int(&alt->ref_count, 1);
1121 
1122 	/* See if a ECN Echo is also stranded */
1123 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1124 		if ((chk->whoTo == net) &&
1125 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1126 			sctp_free_remote_addr(chk->whoTo);
1127 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
1128 				chk->sent = SCTP_DATAGRAM_RESEND;
1129 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1130 			}
1131 			chk->whoTo = alt;
1132 			atomic_add_int(&alt->ref_count, 1);
1133 		}
1134 	}
1135 	if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1136 		/*
1137 		 * If the address went un-reachable, we need to move to
1138 		 * alternates for ALL chk's in queue
1139 		 */
1140 		sctp_move_chunks_from_net(stcb, net);
1141 	}
1142 	/* mark the retran info */
1143 	if (strrst->sent != SCTP_DATAGRAM_RESEND)
1144 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1145 	strrst->sent = SCTP_DATAGRAM_RESEND;
1146 
1147 	/* restart the timer */
1148 	sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, strrst->whoTo);
1149 	return (0);
1150 }
1151 
1152 int
1153 sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1154     struct sctp_nets *net)
1155 {
1156 	struct sctp_nets *alt;
1157 	struct sctp_tmit_chunk *asconf, *chk;
1158 
1159 	/* is this a first send, or a retransmission? */
1160 	if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) {
1161 		/* compose a new ASCONF chunk and send it */
1162 		sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED);
1163 	} else {
1164 		/*
1165 		 * Retransmission of the existing ASCONF is needed
1166 		 */
1167 
1168 		/* find the existing ASCONF */
1169 		asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue);
1170 		if (asconf == NULL) {
1171 			return (0);
1172 		}
1173 		/* do threshold management */
1174 		if (sctp_threshold_management(inp, stcb, asconf->whoTo,
1175 		    stcb->asoc.max_send_times)) {
1176 			/* Assoc is over */
1177 			return (1);
1178 		}
1179 		if (asconf->snd_count > stcb->asoc.max_send_times) {
1180 			/*
1181 			 * Something is rotten: our peer is not responding
1182 			 * to ASCONFs but apparently is to other chunks.
1183 			 * i.e. it is not properly handling the chunk type
1184 			 * upper bits. Mark this peer as ASCONF incapable
1185 			 * and cleanup.
1186 			 */
1187 			SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n");
1188 			sctp_asconf_cleanup(stcb, net);
1189 			return (0);
1190 		}
1191 		/*
1192 		 * cleared threshold management, so now backoff the net and
1193 		 * select an alternate
1194 		 */
1195 		sctp_backoff_on_timeout(stcb, asconf->whoTo, 1, 0, 0);
1196 		alt = sctp_find_alternate_net(stcb, asconf->whoTo, 0);
1197 		if (asconf->whoTo != alt) {
1198 			sctp_free_remote_addr(asconf->whoTo);
1199 			asconf->whoTo = alt;
1200 			atomic_add_int(&alt->ref_count, 1);
1201 		}
1202 		/* See if an ECN Echo is also stranded */
1203 		TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1204 			if ((chk->whoTo == net) &&
1205 			    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1206 				sctp_free_remote_addr(chk->whoTo);
1207 				chk->whoTo = alt;
1208 				if (chk->sent != SCTP_DATAGRAM_RESEND) {
1209 					chk->sent = SCTP_DATAGRAM_RESEND;
1210 					sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1211 				}
1212 				atomic_add_int(&alt->ref_count, 1);
1213 			}
1214 		}
1215 		TAILQ_FOREACH(chk, &stcb->asoc.asconf_send_queue, sctp_next) {
1216 			if (chk->whoTo != alt) {
1217 				sctp_free_remote_addr(chk->whoTo);
1218 				chk->whoTo = alt;
1219 				atomic_add_int(&alt->ref_count, 1);
1220 			}
1221 			if (asconf->sent != SCTP_DATAGRAM_RESEND && chk->sent != SCTP_DATAGRAM_UNSENT)
1222 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1223 			chk->sent = SCTP_DATAGRAM_RESEND;
1224 		}
1225 		if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1226 			/*
1227 			 * If the address went un-reachable, we need to move
1228 			 * to the alternate for ALL chunks in queue
1229 			 */
1230 			sctp_move_chunks_from_net(stcb, net);
1231 		}
1232 		/* mark the retran info */
1233 		if (asconf->sent != SCTP_DATAGRAM_RESEND)
1234 			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1235 		asconf->sent = SCTP_DATAGRAM_RESEND;
1236 
1237 		/* send another ASCONF if any and we can do */
1238 		sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED);
1239 	}
1240 	return (0);
1241 }
1242 
1243 /* Mobility adaptation */
1244 void
1245 sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1246     struct sctp_nets *net SCTP_UNUSED)
1247 {
1248 	if (stcb->asoc.deleted_primary == NULL) {
1249 		SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n");
1250 		sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1251 		return;
1252 	}
1253 	SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary ");
1254 	SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa);
1255 	sctp_free_remote_addr(stcb->asoc.deleted_primary);
1256 	stcb->asoc.deleted_primary = NULL;
1257 	sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1258 	return;
1259 }
1260 
1261 /*
1262  * For the shutdown and shutdown-ack, we do not keep one around on the
1263  * control queue. This means we must generate a new one and call the general
1264  * chunk output routine, AFTER having done threshold management.
1265  * It is assumed that net is non-NULL.
1266  */
1267 int
1268 sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1269     struct sctp_nets *net)
1270 {
1271 	struct sctp_nets *alt;
1272 
1273 	/* first threshold managment */
1274 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1275 		/* Assoc is over */
1276 		return (1);
1277 	}
1278 	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1279 	/* second select an alternative */
1280 	alt = sctp_find_alternate_net(stcb, net, 0);
1281 
1282 	/* third generate a shutdown into the queue for out net */
1283 	sctp_send_shutdown(stcb, alt);
1284 
1285 	/* fourth restart timer */
1286 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt);
1287 	return (0);
1288 }
1289 
1290 int
1291 sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1292     struct sctp_nets *net)
1293 {
1294 	struct sctp_nets *alt;
1295 
1296 	/* first threshold managment */
1297 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1298 		/* Assoc is over */
1299 		return (1);
1300 	}
1301 	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1302 	/* second select an alternative */
1303 	alt = sctp_find_alternate_net(stcb, net, 0);
1304 
1305 	/* third generate a shutdown into the queue for out net */
1306 	sctp_send_shutdown_ack(stcb, alt);
1307 
1308 	/* fourth restart timer */
1309 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt);
1310 	return (0);
1311 }
1312 
1313 static void
1314 sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp,
1315     struct sctp_tcb *stcb)
1316 {
1317 	struct sctp_stream_queue_pending *sp;
1318 	unsigned int i, chks_in_queue = 0;
1319 	int being_filled = 0;
1320 
1321 	/*
1322 	 * This function is ONLY called when the send/sent queues are empty.
1323 	 */
1324 	if ((stcb == NULL) || (inp == NULL))
1325 		return;
1326 
1327 	if (stcb->asoc.sent_queue_retran_cnt) {
1328 		SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n",
1329 		    stcb->asoc.sent_queue_retran_cnt);
1330 		stcb->asoc.sent_queue_retran_cnt = 0;
1331 	}
1332 	if (stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1333 		/* No stream scheduler information, initialize scheduler */
1334 		stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc, 0);
1335 		if (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1336 			/* yep, we lost a stream or two */
1337 			SCTP_PRINTF("Found additional streams NOT managed by scheduler, corrected\n");
1338 		} else {
1339 			/* no streams lost */
1340 			stcb->asoc.total_output_queue_size = 0;
1341 		}
1342 	}
1343 	/* Check to see if some data queued, if so report it */
1344 	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
1345 		if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) {
1346 			TAILQ_FOREACH(sp, &stcb->asoc.strmout[i].outqueue, next) {
1347 				if (sp->msg_is_complete)
1348 					being_filled++;
1349 				chks_in_queue++;
1350 			}
1351 		}
1352 	}
1353 	if (chks_in_queue != stcb->asoc.stream_queue_cnt) {
1354 		SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n",
1355 		    stcb->asoc.stream_queue_cnt, chks_in_queue);
1356 	}
1357 	if (chks_in_queue) {
1358 		/* call the output queue function */
1359 		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1360 		if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1361 		    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1362 			/*
1363 			 * Probably should go in and make it go back through
1364 			 * and add fragments allowed
1365 			 */
1366 			if (being_filled == 0) {
1367 				SCTP_PRINTF("Still nothing moved %d chunks are stuck\n",
1368 				    chks_in_queue);
1369 			}
1370 		}
1371 	} else {
1372 		SCTP_PRINTF("Found no chunks on any queue tot:%lu\n",
1373 		    (u_long)stcb->asoc.total_output_queue_size);
1374 		stcb->asoc.total_output_queue_size = 0;
1375 	}
1376 }
1377 
1378 int
1379 sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1380     struct sctp_nets *net)
1381 {
1382 	uint8_t net_was_pf;
1383 
1384 	if (net->dest_state & SCTP_ADDR_PF) {
1385 		net_was_pf = 1;
1386 	} else {
1387 		net_was_pf = 0;
1388 	}
1389 	if (net->hb_responded == 0) {
1390 		if (net->ro._s_addr) {
1391 			/*
1392 			 * Invalidate the src address if we did not get a
1393 			 * response last time.
1394 			 */
1395 			sctp_free_ifa(net->ro._s_addr);
1396 			net->ro._s_addr = NULL;
1397 			net->src_addr_selected = 0;
1398 		}
1399 		sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1400 		if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1401 			/* Assoc is over */
1402 			return (1);
1403 		}
1404 	}
1405 	/* Zero PBA, if it needs it */
1406 	if (net->partial_bytes_acked) {
1407 		net->partial_bytes_acked = 0;
1408 	}
1409 	if ((stcb->asoc.total_output_queue_size > 0) &&
1410 	    (TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1411 	    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1412 		sctp_audit_stream_queues_for_size(inp, stcb);
1413 	}
1414 	if (!(net->dest_state & SCTP_ADDR_NOHB) &&
1415 	    !((net_was_pf == 0) && (net->dest_state & SCTP_ADDR_PF))) {
1416 		/*
1417 		 * when move to PF during threshold mangement, a HB has been
1418 		 * queued in that routine
1419 		 */
1420 		uint32_t ms_gone_by;
1421 
1422 		if ((net->last_sent_time.tv_sec > 0) ||
1423 		    (net->last_sent_time.tv_usec > 0)) {
1424 			struct timeval diff;
1425 
1426 			SCTP_GETTIME_TIMEVAL(&diff);
1427 			timevalsub(&diff, &net->last_sent_time);
1428 			ms_gone_by = (uint32_t) (diff.tv_sec * 1000) +
1429 			    (uint32_t) (diff.tv_usec / 1000);
1430 		} else {
1431 			ms_gone_by = 0xffffffff;
1432 		}
1433 		if ((ms_gone_by >= net->heart_beat_delay) ||
1434 		    (net->dest_state & SCTP_ADDR_PF)) {
1435 			sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
1436 		}
1437 	}
1438 	return (0);
1439 }
1440 
1441 void
1442 sctp_pathmtu_timer(struct sctp_inpcb *inp,
1443     struct sctp_tcb *stcb,
1444     struct sctp_nets *net)
1445 {
1446 	uint32_t next_mtu, mtu;
1447 
1448 	next_mtu = sctp_get_next_mtu(net->mtu);
1449 
1450 	if ((next_mtu > net->mtu) && (net->port == 0)) {
1451 		if ((net->src_addr_selected == 0) ||
1452 		    (net->ro._s_addr == NULL) ||
1453 		    (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1454 			if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1455 				sctp_free_ifa(net->ro._s_addr);
1456 				net->ro._s_addr = NULL;
1457 				net->src_addr_selected = 0;
1458 			} else if (net->ro._s_addr == NULL) {
1459 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1460 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1461 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1462 
1463 					/* KAME hack: embed scopeid */
1464 					(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
1465 				}
1466 #endif
1467 
1468 				net->ro._s_addr = sctp_source_address_selection(inp,
1469 				    stcb,
1470 				    (sctp_route_t *) & net->ro,
1471 				    net, 0, stcb->asoc.vrf_id);
1472 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1473 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1474 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1475 
1476 					(void)sa6_recoverscope(sin6);
1477 				}
1478 #endif				/* INET6 */
1479 			}
1480 			if (net->ro._s_addr)
1481 				net->src_addr_selected = 1;
1482 		}
1483 		if (net->ro._s_addr) {
1484 			mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_rt);
1485 #if defined(INET) || defined(INET6)
1486 			if (net->port) {
1487 				mtu -= sizeof(struct udphdr);
1488 			}
1489 #endif
1490 			if (mtu > next_mtu) {
1491 				net->mtu = next_mtu;
1492 			}
1493 		}
1494 	}
1495 	/* restart the timer */
1496 	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
1497 }
1498 
1499 void
1500 sctp_autoclose_timer(struct sctp_inpcb *inp,
1501     struct sctp_tcb *stcb,
1502     struct sctp_nets *net)
1503 {
1504 	struct timeval tn, *tim_touse;
1505 	struct sctp_association *asoc;
1506 	int ticks_gone_by;
1507 
1508 	(void)SCTP_GETTIME_TIMEVAL(&tn);
1509 	if (stcb->asoc.sctp_autoclose_ticks &&
1510 	    sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) {
1511 		/* Auto close is on */
1512 		asoc = &stcb->asoc;
1513 		/* pick the time to use */
1514 		if (asoc->time_last_rcvd.tv_sec >
1515 		    asoc->time_last_sent.tv_sec) {
1516 			tim_touse = &asoc->time_last_rcvd;
1517 		} else {
1518 			tim_touse = &asoc->time_last_sent;
1519 		}
1520 		/* Now has long enough transpired to autoclose? */
1521 		ticks_gone_by = SEC_TO_TICKS(tn.tv_sec - tim_touse->tv_sec);
1522 		if ((ticks_gone_by > 0) &&
1523 		    (ticks_gone_by >= (int)asoc->sctp_autoclose_ticks)) {
1524 			/*
1525 			 * autoclose time has hit, call the output routine,
1526 			 * which should do nothing just to be SURE we don't
1527 			 * have hanging data. We can then safely check the
1528 			 * queues and know that we are clear to send
1529 			 * shutdown
1530 			 */
1531 			sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED);
1532 			/* Are we clean? */
1533 			if (TAILQ_EMPTY(&asoc->send_queue) &&
1534 			    TAILQ_EMPTY(&asoc->sent_queue)) {
1535 				/*
1536 				 * there is nothing queued to send, so I'm
1537 				 * done...
1538 				 */
1539 				if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) {
1540 					/* only send SHUTDOWN 1st time thru */
1541 					struct sctp_nets *netp;
1542 
1543 					if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
1544 					    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
1545 						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
1546 					}
1547 					SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
1548 					SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
1549 					sctp_stop_timers_for_shutdown(stcb);
1550 					if (stcb->asoc.alternate) {
1551 						netp = stcb->asoc.alternate;
1552 					} else {
1553 						netp = stcb->asoc.primary_destination;
1554 					}
1555 					sctp_send_shutdown(stcb, netp);
1556 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
1557 					    stcb->sctp_ep, stcb,
1558 					    netp);
1559 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
1560 					    stcb->sctp_ep, stcb,
1561 					    netp);
1562 				}
1563 			}
1564 		} else {
1565 			/*
1566 			 * No auto close at this time, reset t-o to check
1567 			 * later
1568 			 */
1569 			int tmp;
1570 
1571 			/* fool the timer startup to use the time left */
1572 			tmp = asoc->sctp_autoclose_ticks;
1573 			asoc->sctp_autoclose_ticks -= ticks_gone_by;
1574 			sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb,
1575 			    net);
1576 			/* restore the real tick value */
1577 			asoc->sctp_autoclose_ticks = tmp;
1578 		}
1579 	}
1580 }
1581