xref: /freebsd/sys/netinet/sctp_timer.c (revision b4e38a41f584ad4391c04b8cfec81f46176b18b0)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved.
5  * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved.
6  * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions are met:
10  *
11  * a) Redistributions of source code must retain the above copyright notice,
12  *    this list of conditions and the following disclaimer.
13  *
14  * b) Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in
16  *    the documentation and/or other materials provided with the distribution.
17  *
18  * c) Neither the name of Cisco Systems, Inc. nor the names of its
19  *    contributors may be used to endorse or promote products derived
20  *    from this software without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
24  * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
26  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
27  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
28  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
29  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
30  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
31  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
32  * THE POSSIBILITY OF SUCH DAMAGE.
33  */
34 
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
37 
38 #define _IP_VHL
39 #include <netinet/sctp_os.h>
40 #include <netinet/sctp_pcb.h>
41 #ifdef INET6
42 #endif
43 #include <netinet/sctp_var.h>
44 #include <netinet/sctp_sysctl.h>
45 #include <netinet/sctp_timer.h>
46 #include <netinet/sctputil.h>
47 #include <netinet/sctp_output.h>
48 #include <netinet/sctp_header.h>
49 #include <netinet/sctp_indata.h>
50 #include <netinet/sctp_asconf.h>
51 #include <netinet/sctp_input.h>
52 #include <netinet/sctp.h>
53 #include <netinet/sctp_uio.h>
54 #if defined(INET) || defined(INET6)
55 #include <netinet/udp.h>
56 #endif
57 
58 void
59 sctp_audit_retranmission_queue(struct sctp_association *asoc)
60 {
61 	struct sctp_tmit_chunk *chk;
62 
63 	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit invoked on send queue cnt:%d onqueue:%d\n",
64 	    asoc->sent_queue_retran_cnt,
65 	    asoc->sent_queue_cnt);
66 	asoc->sent_queue_retran_cnt = 0;
67 	asoc->sent_queue_cnt = 0;
68 	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
69 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
70 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
71 		}
72 		asoc->sent_queue_cnt++;
73 	}
74 	TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) {
75 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
76 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
77 		}
78 	}
79 	TAILQ_FOREACH(chk, &asoc->asconf_send_queue, sctp_next) {
80 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
81 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
82 		}
83 	}
84 	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit completes retran:%d onqueue:%d\n",
85 	    asoc->sent_queue_retran_cnt,
86 	    asoc->sent_queue_cnt);
87 }
88 
89 static int
90 sctp_threshold_management(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
91     struct sctp_nets *net, uint16_t threshold)
92 {
93 	if (net) {
94 		net->error_count++;
95 		SCTPDBG(SCTP_DEBUG_TIMER4, "Error count for %p now %d thresh:%d\n",
96 		    (void *)net, net->error_count,
97 		    net->failure_threshold);
98 		if (net->error_count > net->failure_threshold) {
99 			/* We had a threshold failure */
100 			if (net->dest_state & SCTP_ADDR_REACHABLE) {
101 				net->dest_state &= ~SCTP_ADDR_REACHABLE;
102 				net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY;
103 				net->dest_state &= ~SCTP_ADDR_PF;
104 				sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN,
105 				    stcb, 0,
106 				    (void *)net, SCTP_SO_NOT_LOCKED);
107 			}
108 		} else if ((net->pf_threshold < net->failure_threshold) &&
109 		    (net->error_count > net->pf_threshold)) {
110 			if (!(net->dest_state & SCTP_ADDR_PF)) {
111 				net->dest_state |= SCTP_ADDR_PF;
112 				net->last_active = sctp_get_tick_count();
113 				sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
114 				sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT,
115 				    inp, stcb, net,
116 				    SCTP_FROM_SCTP_TIMER + SCTP_LOC_1);
117 				sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net);
118 			}
119 		}
120 	}
121 	if (stcb == NULL)
122 		return (0);
123 
124 	if (net) {
125 		if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) {
126 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
127 				sctp_misc_ints(SCTP_THRESHOLD_INCR,
128 				    stcb->asoc.overall_error_count,
129 				    (stcb->asoc.overall_error_count + 1),
130 				    SCTP_FROM_SCTP_TIMER,
131 				    __LINE__);
132 			}
133 			stcb->asoc.overall_error_count++;
134 		}
135 	} else {
136 		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
137 			sctp_misc_ints(SCTP_THRESHOLD_INCR,
138 			    stcb->asoc.overall_error_count,
139 			    (stcb->asoc.overall_error_count + 1),
140 			    SCTP_FROM_SCTP_TIMER,
141 			    __LINE__);
142 		}
143 		stcb->asoc.overall_error_count++;
144 	}
145 	SCTPDBG(SCTP_DEBUG_TIMER4, "Overall error count for %p now %d thresh:%u state:%x\n",
146 	    (void *)&stcb->asoc, stcb->asoc.overall_error_count,
147 	    (uint32_t)threshold,
148 	    ((net == NULL) ? (uint32_t)0 : (uint32_t)net->dest_state));
149 	/*
150 	 * We specifically do not do >= to give the assoc one more change
151 	 * before we fail it.
152 	 */
153 	if (stcb->asoc.overall_error_count > threshold) {
154 		/* Abort notification sends a ULP notify */
155 		struct mbuf *op_err;
156 
157 		op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
158 		    "Association error counter exceeded");
159 		inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_2;
160 		sctp_abort_an_association(inp, stcb, op_err, SCTP_SO_NOT_LOCKED);
161 		return (1);
162 	}
163 	return (0);
164 }
165 
166 /*
167  * sctp_find_alternate_net() returns a non-NULL pointer as long
168  * the argument net is non-NULL.
169  */
170 struct sctp_nets *
171 sctp_find_alternate_net(struct sctp_tcb *stcb,
172     struct sctp_nets *net,
173     int mode)
174 {
175 	/* Find and return an alternate network if possible */
176 	struct sctp_nets *alt, *mnet, *min_errors_net = NULL, *max_cwnd_net = NULL;
177 	int once;
178 
179 	/* JRS 5/14/07 - Initialize min_errors to an impossible value. */
180 	int min_errors = -1;
181 	uint32_t max_cwnd = 0;
182 
183 	if (stcb->asoc.numnets == 1) {
184 		/* No others but net */
185 		return (TAILQ_FIRST(&stcb->asoc.nets));
186 	}
187 	/*
188 	 * JRS 5/14/07 - If mode is set to 2, use the CMT PF find alternate
189 	 * net algorithm. This algorithm chooses the active destination (not
190 	 * in PF state) with the largest cwnd value. If all destinations are
191 	 * in PF state, unreachable, or unconfirmed, choose the desination
192 	 * that is in PF state with the lowest error count. In case of a
193 	 * tie, choose the destination that was most recently active.
194 	 */
195 	if (mode == 2) {
196 		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
197 			/*
198 			 * JRS 5/14/07 - If the destination is unreachable
199 			 * or unconfirmed, skip it.
200 			 */
201 			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
202 			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
203 				continue;
204 			}
205 			/*
206 			 * JRS 5/14/07 -  If the destination is reachable
207 			 * but in PF state, compare the error count of the
208 			 * destination to the minimum error count seen thus
209 			 * far. Store the destination with the lower error
210 			 * count.  If the error counts are equal, store the
211 			 * destination that was most recently active.
212 			 */
213 			if (mnet->dest_state & SCTP_ADDR_PF) {
214 				/*
215 				 * JRS 5/14/07 - If the destination under
216 				 * consideration is the current destination,
217 				 * work as if the error count is one higher.
218 				 * The actual error count will not be
219 				 * incremented until later in the t3
220 				 * handler.
221 				 */
222 				if (mnet == net) {
223 					if (min_errors == -1) {
224 						min_errors = mnet->error_count + 1;
225 						min_errors_net = mnet;
226 					} else if (mnet->error_count + 1 < min_errors) {
227 						min_errors = mnet->error_count + 1;
228 						min_errors_net = mnet;
229 					} else if (mnet->error_count + 1 == min_errors
230 					    && mnet->last_active > min_errors_net->last_active) {
231 						min_errors_net = mnet;
232 						min_errors = mnet->error_count + 1;
233 					}
234 					continue;
235 				} else {
236 					if (min_errors == -1) {
237 						min_errors = mnet->error_count;
238 						min_errors_net = mnet;
239 					} else if (mnet->error_count < min_errors) {
240 						min_errors = mnet->error_count;
241 						min_errors_net = mnet;
242 					} else if (mnet->error_count == min_errors
243 					    && mnet->last_active > min_errors_net->last_active) {
244 						min_errors_net = mnet;
245 						min_errors = mnet->error_count;
246 					}
247 					continue;
248 				}
249 			}
250 			/*
251 			 * JRS 5/14/07 - If the destination is reachable and
252 			 * not in PF state, compare the cwnd of the
253 			 * destination to the highest cwnd seen thus far.
254 			 * Store the destination with the higher cwnd value.
255 			 * If the cwnd values are equal, randomly choose one
256 			 * of the two destinations.
257 			 */
258 			if (max_cwnd < mnet->cwnd) {
259 				max_cwnd_net = mnet;
260 				max_cwnd = mnet->cwnd;
261 			} else if (max_cwnd == mnet->cwnd) {
262 				uint32_t rndval;
263 				uint8_t this_random;
264 
265 				if (stcb->asoc.hb_random_idx > 3) {
266 					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
267 					memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values));
268 					this_random = stcb->asoc.hb_random_values[0];
269 					stcb->asoc.hb_random_idx++;
270 					stcb->asoc.hb_ect_randombit = 0;
271 				} else {
272 					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
273 					stcb->asoc.hb_random_idx++;
274 					stcb->asoc.hb_ect_randombit = 0;
275 				}
276 				if (this_random % 2 == 1) {
277 					max_cwnd_net = mnet;
278 					max_cwnd = mnet->cwnd;	/* Useless? */
279 				}
280 			}
281 		}
282 		if (max_cwnd_net == NULL) {
283 			if (min_errors_net == NULL) {
284 				return (net);
285 			}
286 			return (min_errors_net);
287 		} else {
288 			return (max_cwnd_net);
289 		}
290 	}			/* JRS 5/14/07 - If mode is set to 1, use the
291 				 * CMT policy for choosing an alternate net. */
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_nh == 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_nh != 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 		if (net->RTO_measured) {
410 			net->RTO = stcb->asoc.minrto;
411 		} else {
412 			net->RTO = stcb->asoc.initial_rto;
413 		}
414 	}
415 	net->RTO <<= 1;
416 	if (net->RTO > stcb->asoc.maxrto) {
417 		net->RTO = stcb->asoc.maxrto;
418 	}
419 	if ((win_probe == 0) && (num_marked || num_abandoned)) {
420 		/* We don't apply penalty to window probe scenarios */
421 		/* JRS - Use the congestion control given in the CC module */
422 		stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout(stcb, net);
423 	}
424 }
425 
426 #ifndef INVARIANTS
427 static void
428 sctp_recover_sent_list(struct sctp_tcb *stcb)
429 {
430 	struct sctp_tmit_chunk *chk, *nchk;
431 	struct sctp_association *asoc;
432 
433 	asoc = &stcb->asoc;
434 	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
435 		if (SCTP_TSN_GE(asoc->last_acked_seq, chk->rec.data.tsn)) {
436 			SCTP_PRINTF("Found chk:%p tsn:%x <= last_acked_seq:%x\n",
437 			    (void *)chk, chk->rec.data.tsn, asoc->last_acked_seq);
438 			if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
439 				if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) {
440 					asoc->strmout[chk->rec.data.sid].chunks_on_queues--;
441 				}
442 			}
443 			if ((asoc->strmout[chk->rec.data.sid].chunks_on_queues == 0) &&
444 			    (asoc->strmout[chk->rec.data.sid].state == SCTP_STREAM_RESET_PENDING) &&
445 			    TAILQ_EMPTY(&asoc->strmout[chk->rec.data.sid].outqueue)) {
446 				asoc->trigger_reset = 1;
447 			}
448 			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
449 			if (PR_SCTP_ENABLED(chk->flags)) {
450 				if (asoc->pr_sctp_cnt != 0)
451 					asoc->pr_sctp_cnt--;
452 			}
453 			if (chk->data) {
454 				/* sa_ignore NO_NULL_CHK */
455 				sctp_free_bufspace(stcb, asoc, chk, 1);
456 				sctp_m_freem(chk->data);
457 				chk->data = NULL;
458 				if (asoc->prsctp_supported && PR_SCTP_BUF_ENABLED(chk->flags)) {
459 					asoc->sent_queue_cnt_removeable--;
460 				}
461 			}
462 			asoc->sent_queue_cnt--;
463 			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
464 		}
465 	}
466 	SCTP_PRINTF("after recover order is as follows\n");
467 	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
468 		SCTP_PRINTF("chk:%p TSN:%x\n", (void *)chk, chk->rec.data.tsn);
469 	}
470 }
471 #endif
472 
473 static int
474 sctp_mark_all_for_resend(struct sctp_tcb *stcb,
475     struct sctp_nets *net,
476     struct sctp_nets *alt,
477     int window_probe,
478     int *num_marked,
479     int *num_abandoned)
480 {
481 
482 	/*
483 	 * Mark all chunks (well not all) that were sent to *net for
484 	 * retransmission. Move them to alt for there destination as well...
485 	 * We only mark chunks that have been outstanding long enough to
486 	 * have received feed-back.
487 	 */
488 	struct sctp_tmit_chunk *chk, *nchk;
489 	struct sctp_nets *lnets;
490 	struct timeval now, min_wait, tv;
491 	int cur_rto;
492 	int cnt_abandoned;
493 	int audit_tf, num_mk, fir;
494 	unsigned int cnt_mk;
495 	uint32_t orig_flight, orig_tf;
496 	uint32_t tsnlast, tsnfirst;
497 	int recovery_cnt = 0;
498 
499 
500 	/* none in flight now */
501 	audit_tf = 0;
502 	fir = 0;
503 	/*
504 	 * figure out how long a data chunk must be pending before we can
505 	 * mark it ..
506 	 */
507 	(void)SCTP_GETTIME_TIMEVAL(&now);
508 	/* get cur rto in micro-seconds */
509 	cur_rto = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv;
510 	cur_rto *= 1000;
511 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
512 		sctp_log_fr(cur_rto,
513 		    stcb->asoc.peers_rwnd,
514 		    window_probe,
515 		    SCTP_FR_T3_MARK_TIME);
516 		sctp_log_fr(net->flight_size, 0, 0, SCTP_FR_CWND_REPORT);
517 		sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT);
518 	}
519 	tv.tv_sec = cur_rto / 1000000;
520 	tv.tv_usec = cur_rto % 1000000;
521 	min_wait = now;
522 	timevalsub(&min_wait, &tv);
523 	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
524 		/*
525 		 * if we hit here, we don't have enough seconds on the clock
526 		 * to account for the RTO. We just let the lower seconds be
527 		 * the bounds and don't worry about it. This may mean we
528 		 * will mark a lot more than we should.
529 		 */
530 		min_wait.tv_sec = min_wait.tv_usec = 0;
531 	}
532 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
533 		sctp_log_fr(cur_rto, now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME);
534 		sctp_log_fr(0, min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME);
535 	}
536 	/*
537 	 * Our rwnd will be incorrect here since we are not adding back the
538 	 * cnt * mbuf but we will fix that down below.
539 	 */
540 	orig_flight = net->flight_size;
541 	orig_tf = stcb->asoc.total_flight;
542 
543 	net->fast_retran_ip = 0;
544 	/* Now on to each chunk */
545 	cnt_abandoned = 0;
546 	num_mk = cnt_mk = 0;
547 	tsnfirst = tsnlast = 0;
548 #ifndef INVARIANTS
549 start_again:
550 #endif
551 	TAILQ_FOREACH_SAFE(chk, &stcb->asoc.sent_queue, sctp_next, nchk) {
552 		if (SCTP_TSN_GE(stcb->asoc.last_acked_seq, chk->rec.data.tsn)) {
553 			/* Strange case our list got out of order? */
554 			SCTP_PRINTF("Our list is out of order? last_acked:%x chk:%x\n",
555 			    (unsigned int)stcb->asoc.last_acked_seq, (unsigned int)chk->rec.data.tsn);
556 			recovery_cnt++;
557 #ifdef INVARIANTS
558 			panic("last acked >= chk on sent-Q");
559 #else
560 			SCTP_PRINTF("Recover attempts a restart cnt:%d\n", recovery_cnt);
561 			sctp_recover_sent_list(stcb);
562 			if (recovery_cnt < 10) {
563 				goto start_again;
564 			} else {
565 				SCTP_PRINTF("Recovery fails %d times??\n", recovery_cnt);
566 			}
567 #endif
568 		}
569 		if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) {
570 			/*
571 			 * found one to mark: If it is less than
572 			 * DATAGRAM_ACKED it MUST not be a skipped or marked
573 			 * TSN but instead one that is either already set
574 			 * for retransmission OR one that needs
575 			 * retransmission.
576 			 */
577 
578 			/* validate its been outstanding long enough */
579 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
580 				sctp_log_fr(chk->rec.data.tsn,
581 				    chk->sent_rcv_time.tv_sec,
582 				    chk->sent_rcv_time.tv_usec,
583 				    SCTP_FR_T3_MARK_TIME);
584 			}
585 			if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) {
586 				/*
587 				 * we have reached a chunk that was sent
588 				 * some seconds past our min.. forget it we
589 				 * will find no more to send.
590 				 */
591 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
592 					sctp_log_fr(0,
593 					    chk->sent_rcv_time.tv_sec,
594 					    chk->sent_rcv_time.tv_usec,
595 					    SCTP_FR_T3_STOPPED);
596 				}
597 				continue;
598 			} else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) &&
599 			    (window_probe == 0)) {
600 				/*
601 				 * we must look at the micro seconds to
602 				 * know.
603 				 */
604 				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
605 					/*
606 					 * ok it was sent after our boundary
607 					 * time.
608 					 */
609 					continue;
610 				}
611 			}
612 			if (stcb->asoc.prsctp_supported && PR_SCTP_TTL_ENABLED(chk->flags)) {
613 				/* Is it expired? */
614 				if (timevalcmp(&now, &chk->rec.data.timetodrop, >)) {
615 					/* Yes so drop it */
616 					if (chk->data) {
617 						(void)sctp_release_pr_sctp_chunk(stcb,
618 						    chk,
619 						    1,
620 						    SCTP_SO_NOT_LOCKED);
621 						cnt_abandoned++;
622 					}
623 					continue;
624 				}
625 			}
626 			if (stcb->asoc.prsctp_supported && PR_SCTP_RTX_ENABLED(chk->flags)) {
627 				/* Has it been retransmitted tv_sec times? */
628 				if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) {
629 					if (chk->data) {
630 						(void)sctp_release_pr_sctp_chunk(stcb,
631 						    chk,
632 						    1,
633 						    SCTP_SO_NOT_LOCKED);
634 						cnt_abandoned++;
635 					}
636 					continue;
637 				}
638 			}
639 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
640 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
641 				num_mk++;
642 				if (fir == 0) {
643 					fir = 1;
644 					tsnfirst = chk->rec.data.tsn;
645 				}
646 				tsnlast = chk->rec.data.tsn;
647 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
648 					sctp_log_fr(chk->rec.data.tsn, chk->snd_count,
649 					    0, SCTP_FR_T3_MARKED);
650 				}
651 
652 				if (chk->rec.data.chunk_was_revoked) {
653 					/* deflate the cwnd */
654 					chk->whoTo->cwnd -= chk->book_size;
655 					chk->rec.data.chunk_was_revoked = 0;
656 				}
657 				net->marked_retrans++;
658 				stcb->asoc.marked_retrans++;
659 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
660 					sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO,
661 					    chk->whoTo->flight_size,
662 					    chk->book_size,
663 					    (uint32_t)(uintptr_t)chk->whoTo,
664 					    chk->rec.data.tsn);
665 				}
666 				sctp_flight_size_decrease(chk);
667 				sctp_total_flight_decrease(stcb, chk);
668 				stcb->asoc.peers_rwnd += chk->send_size;
669 				stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh);
670 			}
671 			chk->sent = SCTP_DATAGRAM_RESEND;
672 			chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
673 			SCTP_STAT_INCR(sctps_markedretrans);
674 
675 			/* reset the TSN for striking and other FR stuff */
676 			chk->rec.data.doing_fast_retransmit = 0;
677 			/* Clear any time so NO RTT is being done */
678 
679 			if (chk->do_rtt) {
680 				if (chk->whoTo->rto_needed == 0) {
681 					chk->whoTo->rto_needed = 1;
682 				}
683 			}
684 			chk->do_rtt = 0;
685 			if (alt != net) {
686 				sctp_free_remote_addr(chk->whoTo);
687 				chk->no_fr_allowed = 1;
688 				chk->whoTo = alt;
689 				atomic_add_int(&alt->ref_count, 1);
690 			} else {
691 				chk->no_fr_allowed = 0;
692 				if (TAILQ_EMPTY(&stcb->asoc.send_queue)) {
693 					chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq;
694 				} else {
695 					chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.tsn;
696 				}
697 			}
698 			/*
699 			 * CMT: Do not allow FRs on retransmitted TSNs.
700 			 */
701 			if (stcb->asoc.sctp_cmt_on_off > 0) {
702 				chk->no_fr_allowed = 1;
703 			}
704 #ifdef THIS_SHOULD_NOT_BE_DONE
705 		} else if (chk->sent == SCTP_DATAGRAM_ACKED) {
706 			/* remember highest acked one */
707 			could_be_sent = chk;
708 #endif
709 		}
710 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
711 			cnt_mk++;
712 		}
713 	}
714 	if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) {
715 		/* we did not subtract the same things? */
716 		audit_tf = 1;
717 	}
718 
719 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
720 		sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT);
721 	}
722 #ifdef SCTP_DEBUG
723 	if (num_mk) {
724 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
725 		    tsnlast);
726 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%u\n",
727 		    num_mk,
728 		    stcb->asoc.peers_rwnd);
729 	}
730 #endif
731 	*num_marked = num_mk;
732 	*num_abandoned = cnt_abandoned;
733 	/*
734 	 * Now check for a ECN Echo that may be stranded And include the
735 	 * cnt_mk'd to have all resends in the control queue.
736 	 */
737 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
738 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
739 			cnt_mk++;
740 		}
741 		if ((chk->whoTo == net) &&
742 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
743 			sctp_free_remote_addr(chk->whoTo);
744 			chk->whoTo = alt;
745 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
746 				chk->sent = SCTP_DATAGRAM_RESEND;
747 				chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
748 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
749 				cnt_mk++;
750 			}
751 			atomic_add_int(&alt->ref_count, 1);
752 		}
753 	}
754 #ifdef THIS_SHOULD_NOT_BE_DONE
755 	if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) {
756 		/* fix it so we retransmit the highest acked anyway */
757 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
758 		cnt_mk++;
759 		could_be_sent->sent = SCTP_DATAGRAM_RESEND;
760 	}
761 #endif
762 	if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) {
763 #ifdef INVARIANTS
764 		SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n",
765 		    cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk);
766 #endif
767 #ifndef SCTP_AUDITING_ENABLED
768 		stcb->asoc.sent_queue_retran_cnt = cnt_mk;
769 #endif
770 	}
771 	if (audit_tf) {
772 		SCTPDBG(SCTP_DEBUG_TIMER4,
773 		    "Audit total flight due to negative value net:%p\n",
774 		    (void *)net);
775 		stcb->asoc.total_flight = 0;
776 		stcb->asoc.total_flight_count = 0;
777 		/* Clear all networks flight size */
778 		TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) {
779 			lnets->flight_size = 0;
780 			SCTPDBG(SCTP_DEBUG_TIMER4,
781 			    "Net:%p c-f cwnd:%d ssthresh:%d\n",
782 			    (void *)lnets, lnets->cwnd, lnets->ssthresh);
783 		}
784 		TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
785 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
786 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
787 					sctp_misc_ints(SCTP_FLIGHT_LOG_UP,
788 					    chk->whoTo->flight_size,
789 					    chk->book_size,
790 					    (uint32_t)(uintptr_t)chk->whoTo,
791 					    chk->rec.data.tsn);
792 				}
793 
794 				sctp_flight_size_increase(chk);
795 				sctp_total_flight_increase(stcb, chk);
796 			}
797 		}
798 	}
799 	/* We return 1 if we only have a window probe outstanding */
800 	return (0);
801 }
802 
803 
804 int
805 sctp_t3rxt_timer(struct sctp_inpcb *inp,
806     struct sctp_tcb *stcb,
807     struct sctp_nets *net)
808 {
809 	struct sctp_nets *alt;
810 	int win_probe, num_mk, num_abandoned;
811 
812 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
813 		sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT);
814 	}
815 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
816 		struct sctp_nets *lnet;
817 
818 		TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
819 			if (net == lnet) {
820 				sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3);
821 			} else {
822 				sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3);
823 			}
824 		}
825 	}
826 	/* Find an alternate and mark those for retransmission */
827 	if ((stcb->asoc.peers_rwnd == 0) &&
828 	    (stcb->asoc.total_flight < net->mtu)) {
829 		SCTP_STAT_INCR(sctps_timowindowprobe);
830 		win_probe = 1;
831 	} else {
832 		win_probe = 0;
833 	}
834 
835 	if (win_probe == 0) {
836 		/* We don't do normal threshold management on window probes */
837 		if (sctp_threshold_management(inp, stcb, net,
838 		    stcb->asoc.max_send_times)) {
839 			/* Association was destroyed */
840 			return (1);
841 		} else {
842 			if (net != stcb->asoc.primary_destination) {
843 				/* send a immediate HB if our RTO is stale */
844 				struct timeval now;
845 				unsigned int ms_goneby;
846 
847 				(void)SCTP_GETTIME_TIMEVAL(&now);
848 				if (net->last_sent_time.tv_sec) {
849 					ms_goneby = (now.tv_sec - net->last_sent_time.tv_sec) * 1000;
850 				} else {
851 					ms_goneby = 0;
852 				}
853 				if ((net->dest_state & SCTP_ADDR_PF) == 0) {
854 					if ((ms_goneby > net->RTO) || (net->RTO == 0)) {
855 						/*
856 						 * no recent feed back in an
857 						 * RTO or more, request a
858 						 * RTT update
859 						 */
860 						sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
861 					}
862 				}
863 			}
864 		}
865 	} else {
866 		/*
867 		 * For a window probe we don't penalize the net's but only
868 		 * the association. This may fail it if SACKs are not coming
869 		 * back. If sack's are coming with rwnd locked at 0, we will
870 		 * continue to hold things waiting for rwnd to raise
871 		 */
872 		if (sctp_threshold_management(inp, stcb, NULL,
873 		    stcb->asoc.max_send_times)) {
874 			/* Association was destroyed */
875 			return (1);
876 		}
877 	}
878 	if (stcb->asoc.sctp_cmt_on_off > 0) {
879 		if (net->pf_threshold < net->failure_threshold) {
880 			alt = sctp_find_alternate_net(stcb, net, 2);
881 		} else {
882 			/*
883 			 * CMT: Using RTX_SSTHRESH policy for CMT. If CMT is
884 			 * being used, then pick dest with largest ssthresh
885 			 * for any retransmission.
886 			 */
887 			alt = sctp_find_alternate_net(stcb, net, 1);
888 			/*
889 			 * CUCv2: If a different dest is picked for the
890 			 * retransmission, then new (rtx-)pseudo_cumack
891 			 * needs to be tracked for orig dest. Let CUCv2
892 			 * track new (rtx-) pseudo-cumack always.
893 			 */
894 			net->find_pseudo_cumack = 1;
895 			net->find_rtx_pseudo_cumack = 1;
896 		}
897 	} else {
898 		alt = sctp_find_alternate_net(stcb, net, 0);
899 	}
900 
901 	num_mk = 0;
902 	num_abandoned = 0;
903 	(void)sctp_mark_all_for_resend(stcb, net, alt, win_probe,
904 	    &num_mk, &num_abandoned);
905 	/* FR Loss recovery just ended with the T3. */
906 	stcb->asoc.fast_retran_loss_recovery = 0;
907 
908 	/* CMT FR loss recovery ended with the T3 */
909 	net->fast_retran_loss_recovery = 0;
910 	if ((stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) &&
911 	    (net->flight_size == 0)) {
912 		(*stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) (stcb, net);
913 	}
914 
915 	/*
916 	 * setup the sat loss recovery that prevents satellite cwnd advance.
917 	 */
918 	stcb->asoc.sat_t3_loss_recovery = 1;
919 	stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq;
920 
921 	/* Backoff the timer and cwnd */
922 	sctp_backoff_on_timeout(stcb, net, win_probe, num_mk, num_abandoned);
923 	if ((!(net->dest_state & SCTP_ADDR_REACHABLE)) ||
924 	    (net->dest_state & SCTP_ADDR_PF)) {
925 		/* Move all pending over too */
926 		sctp_move_chunks_from_net(stcb, net);
927 
928 		/*
929 		 * Get the address that failed, to force a new src address
930 		 * selecton and a route allocation.
931 		 */
932 		if (net->ro._s_addr) {
933 			sctp_free_ifa(net->ro._s_addr);
934 			net->ro._s_addr = NULL;
935 		}
936 		net->src_addr_selected = 0;
937 
938 		/* Force a route allocation too */
939 		RO_NHFREE(&net->ro);
940 
941 		/* Was it our primary? */
942 		if ((stcb->asoc.primary_destination == net) && (alt != net)) {
943 			/*
944 			 * Yes, note it as such and find an alternate note:
945 			 * this means HB code must use this to resent the
946 			 * primary if it goes active AND if someone does a
947 			 * change-primary then this flag must be cleared
948 			 * from any net structures.
949 			 */
950 			if (stcb->asoc.alternate) {
951 				sctp_free_remote_addr(stcb->asoc.alternate);
952 			}
953 			stcb->asoc.alternate = alt;
954 			atomic_add_int(&stcb->asoc.alternate->ref_count, 1);
955 		}
956 	}
957 	/*
958 	 * Special case for cookie-echo'ed case, we don't do output but must
959 	 * await the COOKIE-ACK before retransmission
960 	 */
961 	if (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED) {
962 		/*
963 		 * Here we just reset the timer and start again since we
964 		 * have not established the asoc
965 		 */
966 		sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net);
967 		return (0);
968 	}
969 	if (stcb->asoc.prsctp_supported) {
970 		struct sctp_tmit_chunk *lchk;
971 
972 		lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc);
973 		/* C3. See if we need to send a Fwd-TSN */
974 		if (SCTP_TSN_GT(stcb->asoc.advanced_peer_ack_point, stcb->asoc.last_acked_seq)) {
975 			send_forward_tsn(stcb, &stcb->asoc);
976 			for (; lchk != NULL; lchk = TAILQ_NEXT(lchk, sctp_next)) {
977 				if (lchk->whoTo != NULL) {
978 					break;
979 				}
980 			}
981 			if (lchk != NULL) {
982 				/* Assure a timer is up */
983 				sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo);
984 			}
985 		}
986 	}
987 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
988 		sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX);
989 	}
990 	return (0);
991 }
992 
993 int
994 sctp_t1init_timer(struct sctp_inpcb *inp,
995     struct sctp_tcb *stcb,
996     struct sctp_nets *net)
997 {
998 	/* bump the thresholds */
999 	if (stcb->asoc.delayed_connection) {
1000 		/*
1001 		 * special hook for delayed connection. The library did NOT
1002 		 * complete the rest of its sends.
1003 		 */
1004 		stcb->asoc.delayed_connection = 0;
1005 		sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1006 		return (0);
1007 	}
1008 	if (SCTP_GET_STATE(stcb) != SCTP_STATE_COOKIE_WAIT) {
1009 		return (0);
1010 	}
1011 	if (sctp_threshold_management(inp, stcb, net,
1012 	    stcb->asoc.max_init_times)) {
1013 		/* Association was destroyed */
1014 		return (1);
1015 	}
1016 	stcb->asoc.dropped_special_cnt = 0;
1017 	sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0, 0);
1018 	if (stcb->asoc.initial_init_rto_max < net->RTO) {
1019 		net->RTO = stcb->asoc.initial_init_rto_max;
1020 	}
1021 	if (stcb->asoc.numnets > 1) {
1022 		/* If we have more than one addr use it */
1023 		struct sctp_nets *alt;
1024 
1025 		alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0);
1026 		if (alt != stcb->asoc.primary_destination) {
1027 			sctp_move_chunks_from_net(stcb, stcb->asoc.primary_destination);
1028 			stcb->asoc.primary_destination = alt;
1029 		}
1030 	}
1031 	/* Send out a new init */
1032 	sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1033 	return (0);
1034 }
1035 
1036 /*
1037  * For cookie and asconf we actually need to find and mark for resend, then
1038  * increment the resend counter (after all the threshold management stuff of
1039  * course).
1040  */
1041 int
1042 sctp_cookie_timer(struct sctp_inpcb *inp,
1043     struct sctp_tcb *stcb,
1044     struct sctp_nets *net SCTP_UNUSED)
1045 {
1046 	struct sctp_nets *alt;
1047 	struct sctp_tmit_chunk *cookie;
1048 
1049 	/* first before all else we must find the cookie */
1050 	TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) {
1051 		if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) {
1052 			break;
1053 		}
1054 	}
1055 	if (cookie == NULL) {
1056 		if (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED) {
1057 			/* FOOBAR! */
1058 			struct mbuf *op_err;
1059 
1060 			op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
1061 			    "Cookie timer expired, but no cookie");
1062 			inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_3;
1063 			sctp_abort_an_association(inp, stcb, op_err, SCTP_SO_NOT_LOCKED);
1064 		} else {
1065 #ifdef INVARIANTS
1066 			panic("Cookie timer expires in wrong state?");
1067 #else
1068 			SCTP_PRINTF("Strange in state %d not cookie-echoed yet c-e timer expires?\n", SCTP_GET_STATE(stcb));
1069 			return (0);
1070 #endif
1071 		}
1072 		return (0);
1073 	}
1074 	/* Ok we found the cookie, threshold management next */
1075 	if (sctp_threshold_management(inp, stcb, cookie->whoTo,
1076 	    stcb->asoc.max_init_times)) {
1077 		/* Assoc is over */
1078 		return (1);
1079 	}
1080 	/*
1081 	 * Cleared threshold management, now lets backoff the address and
1082 	 * select an alternate
1083 	 */
1084 	stcb->asoc.dropped_special_cnt = 0;
1085 	sctp_backoff_on_timeout(stcb, cookie->whoTo, 1, 0, 0);
1086 	alt = sctp_find_alternate_net(stcb, cookie->whoTo, 0);
1087 	if (alt != cookie->whoTo) {
1088 		sctp_free_remote_addr(cookie->whoTo);
1089 		cookie->whoTo = alt;
1090 		atomic_add_int(&alt->ref_count, 1);
1091 	}
1092 	/* Now mark the retran info */
1093 	if (cookie->sent != SCTP_DATAGRAM_RESEND) {
1094 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1095 	}
1096 	cookie->sent = SCTP_DATAGRAM_RESEND;
1097 	cookie->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1098 	/*
1099 	 * Now call the output routine to kick out the cookie again, Note we
1100 	 * don't mark any chunks for retran so that FR will need to kick in
1101 	 * to move these (or a send timer).
1102 	 */
1103 	return (0);
1104 }
1105 
1106 int
1107 sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
1108 {
1109 	struct sctp_nets *alt, *net;
1110 	struct sctp_tmit_chunk *strrst = NULL, *chk = NULL;
1111 
1112 	if (stcb->asoc.stream_reset_outstanding == 0) {
1113 		return (0);
1114 	}
1115 	/* find the existing STRRESET, we use the seq number we sent out on */
1116 	(void)sctp_find_stream_reset(stcb, stcb->asoc.str_reset_seq_out, &strrst);
1117 	if (strrst == NULL) {
1118 		return (0);
1119 	}
1120 	net = strrst->whoTo;
1121 	/* do threshold management */
1122 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1123 		/* Assoc is over */
1124 		return (1);
1125 	}
1126 	/*
1127 	 * Cleared threshold management, now lets backoff the address and
1128 	 * select an alternate
1129 	 */
1130 	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1131 	alt = sctp_find_alternate_net(stcb, net, 0);
1132 	strrst->whoTo = alt;
1133 	atomic_add_int(&alt->ref_count, 1);
1134 
1135 	/* See if a ECN Echo is also stranded */
1136 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1137 		if ((chk->whoTo == net) &&
1138 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1139 			sctp_free_remote_addr(chk->whoTo);
1140 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
1141 				chk->sent = SCTP_DATAGRAM_RESEND;
1142 				chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1143 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1144 			}
1145 			chk->whoTo = alt;
1146 			atomic_add_int(&alt->ref_count, 1);
1147 		}
1148 	}
1149 	if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1150 		/*
1151 		 * If the address went un-reachable, we need to move to
1152 		 * alternates for ALL chk's in queue
1153 		 */
1154 		sctp_move_chunks_from_net(stcb, net);
1155 	}
1156 	sctp_free_remote_addr(net);
1157 
1158 	/* mark the retran info */
1159 	if (strrst->sent != SCTP_DATAGRAM_RESEND)
1160 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1161 	strrst->sent = SCTP_DATAGRAM_RESEND;
1162 	strrst->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1163 
1164 	/* restart the timer */
1165 	sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, alt);
1166 	return (0);
1167 }
1168 
1169 int
1170 sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1171     struct sctp_nets *net)
1172 {
1173 	struct sctp_nets *alt;
1174 	struct sctp_tmit_chunk *asconf, *chk;
1175 
1176 	/* is this a first send, or a retransmission? */
1177 	if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) {
1178 		/* compose a new ASCONF chunk and send it */
1179 		sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED);
1180 	} else {
1181 		/*
1182 		 * Retransmission of the existing ASCONF is needed
1183 		 */
1184 
1185 		/* find the existing ASCONF */
1186 		asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue);
1187 		if (asconf == NULL) {
1188 			return (0);
1189 		}
1190 		net = asconf->whoTo;
1191 		/* do threshold management */
1192 		if (sctp_threshold_management(inp, stcb, net,
1193 		    stcb->asoc.max_send_times)) {
1194 			/* Assoc is over */
1195 			return (1);
1196 		}
1197 		if (asconf->snd_count > stcb->asoc.max_send_times) {
1198 			/*
1199 			 * Something is rotten: our peer is not responding
1200 			 * to ASCONFs but apparently is to other chunks.
1201 			 * i.e. it is not properly handling the chunk type
1202 			 * upper bits. Mark this peer as ASCONF incapable
1203 			 * and cleanup.
1204 			 */
1205 			SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n");
1206 			sctp_asconf_cleanup(stcb);
1207 			return (0);
1208 		}
1209 		/*
1210 		 * cleared threshold management, so now backoff the net and
1211 		 * select an alternate
1212 		 */
1213 		sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1214 		alt = sctp_find_alternate_net(stcb, net, 0);
1215 		if (asconf->whoTo != alt) {
1216 			asconf->whoTo = alt;
1217 			atomic_add_int(&alt->ref_count, 1);
1218 		}
1219 
1220 		/* See if an ECN Echo is also stranded */
1221 		TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1222 			if ((chk->whoTo == net) &&
1223 			    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1224 				sctp_free_remote_addr(chk->whoTo);
1225 				chk->whoTo = alt;
1226 				if (chk->sent != SCTP_DATAGRAM_RESEND) {
1227 					chk->sent = SCTP_DATAGRAM_RESEND;
1228 					chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1229 					sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1230 				}
1231 				atomic_add_int(&alt->ref_count, 1);
1232 			}
1233 		}
1234 		TAILQ_FOREACH(chk, &stcb->asoc.asconf_send_queue, sctp_next) {
1235 			if (chk->whoTo != alt) {
1236 				sctp_free_remote_addr(chk->whoTo);
1237 				chk->whoTo = alt;
1238 				atomic_add_int(&alt->ref_count, 1);
1239 			}
1240 			if (asconf->sent != SCTP_DATAGRAM_RESEND && chk->sent != SCTP_DATAGRAM_UNSENT)
1241 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1242 			chk->sent = SCTP_DATAGRAM_RESEND;
1243 			chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1244 		}
1245 		if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1246 			/*
1247 			 * If the address went un-reachable, we need to move
1248 			 * to the alternate for ALL chunks in queue
1249 			 */
1250 			sctp_move_chunks_from_net(stcb, net);
1251 		}
1252 		sctp_free_remote_addr(net);
1253 
1254 		/* mark the retran info */
1255 		if (asconf->sent != SCTP_DATAGRAM_RESEND)
1256 			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1257 		asconf->sent = SCTP_DATAGRAM_RESEND;
1258 		asconf->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1259 
1260 		/* send another ASCONF if any and we can do */
1261 		sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED);
1262 	}
1263 	return (0);
1264 }
1265 
1266 /* Mobility adaptation */
1267 void
1268 sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
1269 {
1270 	if (stcb->asoc.deleted_primary == NULL) {
1271 		SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n");
1272 		sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1273 		return;
1274 	}
1275 	SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary ");
1276 	SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa);
1277 	sctp_free_remote_addr(stcb->asoc.deleted_primary);
1278 	stcb->asoc.deleted_primary = NULL;
1279 	sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1280 	return;
1281 }
1282 
1283 /*
1284  * For the shutdown and shutdown-ack, we do not keep one around on the
1285  * control queue. This means we must generate a new one and call the general
1286  * chunk output routine, AFTER having done threshold management.
1287  * It is assumed that net is non-NULL.
1288  */
1289 int
1290 sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1291     struct sctp_nets *net)
1292 {
1293 	struct sctp_nets *alt;
1294 
1295 	/* first threshold management */
1296 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1297 		/* Assoc is over */
1298 		return (1);
1299 	}
1300 	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1301 	/* second select an alternative */
1302 	alt = sctp_find_alternate_net(stcb, net, 0);
1303 
1304 	/* third generate a shutdown into the queue for out net */
1305 	sctp_send_shutdown(stcb, alt);
1306 
1307 	/* fourth restart timer */
1308 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt);
1309 	return (0);
1310 }
1311 
1312 int
1313 sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1314     struct sctp_nets *net)
1315 {
1316 	struct sctp_nets *alt;
1317 
1318 	/* first threshold management */
1319 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1320 		/* Assoc is over */
1321 		return (1);
1322 	}
1323 	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1324 	/* second select an alternative */
1325 	alt = sctp_find_alternate_net(stcb, net, 0);
1326 
1327 	/* third generate a shutdown into the queue for out net */
1328 	sctp_send_shutdown_ack(stcb, alt);
1329 
1330 	/* fourth restart timer */
1331 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt);
1332 	return (0);
1333 }
1334 
1335 static void
1336 sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp,
1337     struct sctp_tcb *stcb)
1338 {
1339 	struct sctp_stream_queue_pending *sp;
1340 	unsigned int i, chks_in_queue = 0;
1341 	int being_filled = 0;
1342 
1343 	/*
1344 	 * This function is ONLY called when the send/sent queues are empty.
1345 	 */
1346 	if ((stcb == NULL) || (inp == NULL))
1347 		return;
1348 
1349 	if (stcb->asoc.sent_queue_retran_cnt) {
1350 		SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n",
1351 		    stcb->asoc.sent_queue_retran_cnt);
1352 		stcb->asoc.sent_queue_retran_cnt = 0;
1353 	}
1354 	if (stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1355 		/* No stream scheduler information, initialize scheduler */
1356 		stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc, 0);
1357 		if (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1358 			/* yep, we lost a stream or two */
1359 			SCTP_PRINTF("Found additional streams NOT managed by scheduler, corrected\n");
1360 		} else {
1361 			/* no streams lost */
1362 			stcb->asoc.total_output_queue_size = 0;
1363 		}
1364 	}
1365 	/* Check to see if some data queued, if so report it */
1366 	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
1367 		if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) {
1368 			TAILQ_FOREACH(sp, &stcb->asoc.strmout[i].outqueue, next) {
1369 				if (sp->msg_is_complete)
1370 					being_filled++;
1371 				chks_in_queue++;
1372 			}
1373 		}
1374 	}
1375 	if (chks_in_queue != stcb->asoc.stream_queue_cnt) {
1376 		SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n",
1377 		    stcb->asoc.stream_queue_cnt, chks_in_queue);
1378 	}
1379 	if (chks_in_queue) {
1380 		/* call the output queue function */
1381 		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1382 		if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1383 		    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1384 			/*
1385 			 * Probably should go in and make it go back through
1386 			 * and add fragments allowed
1387 			 */
1388 			if (being_filled == 0) {
1389 				SCTP_PRINTF("Still nothing moved %d chunks are stuck\n",
1390 				    chks_in_queue);
1391 			}
1392 		}
1393 	} else {
1394 		SCTP_PRINTF("Found no chunks on any queue tot:%lu\n",
1395 		    (u_long)stcb->asoc.total_output_queue_size);
1396 		stcb->asoc.total_output_queue_size = 0;
1397 	}
1398 }
1399 
1400 int
1401 sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1402     struct sctp_nets *net)
1403 {
1404 	uint8_t net_was_pf;
1405 
1406 	if (net->dest_state & SCTP_ADDR_PF) {
1407 		net_was_pf = 1;
1408 	} else {
1409 		net_was_pf = 0;
1410 	}
1411 	if (net->hb_responded == 0) {
1412 		if (net->ro._s_addr) {
1413 			/*
1414 			 * Invalidate the src address if we did not get a
1415 			 * response last time.
1416 			 */
1417 			sctp_free_ifa(net->ro._s_addr);
1418 			net->ro._s_addr = NULL;
1419 			net->src_addr_selected = 0;
1420 		}
1421 		sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1422 		if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1423 			/* Assoc is over */
1424 			return (1);
1425 		}
1426 	}
1427 	/* Zero PBA, if it needs it */
1428 	if (net->partial_bytes_acked) {
1429 		net->partial_bytes_acked = 0;
1430 	}
1431 	if ((stcb->asoc.total_output_queue_size > 0) &&
1432 	    (TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1433 	    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1434 		sctp_audit_stream_queues_for_size(inp, stcb);
1435 	}
1436 	if (!(net->dest_state & SCTP_ADDR_NOHB) &&
1437 	    !((net_was_pf == 0) && (net->dest_state & SCTP_ADDR_PF))) {
1438 		/*
1439 		 * when move to PF during threshold mangement, a HB has been
1440 		 * queued in that routine
1441 		 */
1442 		uint32_t ms_gone_by;
1443 
1444 		if ((net->last_sent_time.tv_sec > 0) ||
1445 		    (net->last_sent_time.tv_usec > 0)) {
1446 			struct timeval diff;
1447 
1448 			SCTP_GETTIME_TIMEVAL(&diff);
1449 			timevalsub(&diff, &net->last_sent_time);
1450 			ms_gone_by = (uint32_t)(diff.tv_sec * 1000) +
1451 			    (uint32_t)(diff.tv_usec / 1000);
1452 		} else {
1453 			ms_gone_by = 0xffffffff;
1454 		}
1455 		if ((ms_gone_by >= net->heart_beat_delay) ||
1456 		    (net->dest_state & SCTP_ADDR_PF)) {
1457 			sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
1458 		}
1459 	}
1460 	return (0);
1461 }
1462 
1463 void
1464 sctp_pathmtu_timer(struct sctp_inpcb *inp,
1465     struct sctp_tcb *stcb,
1466     struct sctp_nets *net)
1467 {
1468 	uint32_t next_mtu, mtu;
1469 
1470 	next_mtu = sctp_get_next_mtu(net->mtu);
1471 
1472 	if ((next_mtu > net->mtu) && (net->port == 0)) {
1473 		if ((net->src_addr_selected == 0) ||
1474 		    (net->ro._s_addr == NULL) ||
1475 		    (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1476 			if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1477 				sctp_free_ifa(net->ro._s_addr);
1478 				net->ro._s_addr = NULL;
1479 				net->src_addr_selected = 0;
1480 			} else if (net->ro._s_addr == NULL) {
1481 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1482 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1483 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1484 
1485 					/* KAME hack: embed scopeid */
1486 					(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
1487 				}
1488 #endif
1489 
1490 				net->ro._s_addr = sctp_source_address_selection(inp,
1491 				    stcb,
1492 				    (sctp_route_t *)&net->ro,
1493 				    net, 0, stcb->asoc.vrf_id);
1494 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1495 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1496 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1497 
1498 					(void)sa6_recoverscope(sin6);
1499 				}
1500 #endif				/* INET6 */
1501 			}
1502 			if (net->ro._s_addr)
1503 				net->src_addr_selected = 1;
1504 		}
1505 		if (net->ro._s_addr) {
1506 			mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_nh);
1507 #if defined(INET) || defined(INET6)
1508 			if (net->port) {
1509 				mtu -= sizeof(struct udphdr);
1510 			}
1511 #endif
1512 			if (mtu > next_mtu) {
1513 				net->mtu = next_mtu;
1514 			} else {
1515 				net->mtu = mtu;
1516 			}
1517 		}
1518 	}
1519 	/* restart the timer */
1520 	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
1521 }
1522 
1523 void
1524 sctp_autoclose_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
1525 {
1526 	struct timeval tn, *tim_touse;
1527 	struct sctp_association *asoc;
1528 	uint32_t ticks_gone_by;
1529 
1530 	(void)SCTP_GETTIME_TIMEVAL(&tn);
1531 	if (stcb->asoc.sctp_autoclose_ticks > 0 &&
1532 	    sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) {
1533 		/* Auto close is on */
1534 		asoc = &stcb->asoc;
1535 		/* pick the time to use */
1536 		if (asoc->time_last_rcvd.tv_sec >
1537 		    asoc->time_last_sent.tv_sec) {
1538 			tim_touse = &asoc->time_last_rcvd;
1539 		} else {
1540 			tim_touse = &asoc->time_last_sent;
1541 		}
1542 		/* Now has long enough transpired to autoclose? */
1543 		ticks_gone_by = sctp_secs_to_ticks((uint32_t)(tn.tv_sec - tim_touse->tv_sec));
1544 		if (ticks_gone_by >= asoc->sctp_autoclose_ticks) {
1545 			/*
1546 			 * autoclose time has hit, call the output routine,
1547 			 * which should do nothing just to be SURE we don't
1548 			 * have hanging data. We can then safely check the
1549 			 * queues and know that we are clear to send
1550 			 * shutdown
1551 			 */
1552 			sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED);
1553 			/* Are we clean? */
1554 			if (TAILQ_EMPTY(&asoc->send_queue) &&
1555 			    TAILQ_EMPTY(&asoc->sent_queue)) {
1556 				/*
1557 				 * there is nothing queued to send, so I'm
1558 				 * done...
1559 				 */
1560 				if (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_SENT) {
1561 					/* only send SHUTDOWN 1st time thru */
1562 					struct sctp_nets *net;
1563 
1564 					if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) ||
1565 					    (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
1566 						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
1567 					}
1568 					SCTP_SET_STATE(stcb, SCTP_STATE_SHUTDOWN_SENT);
1569 					sctp_stop_timers_for_shutdown(stcb);
1570 					if (stcb->asoc.alternate) {
1571 						net = stcb->asoc.alternate;
1572 					} else {
1573 						net = stcb->asoc.primary_destination;
1574 					}
1575 					sctp_send_shutdown(stcb, net);
1576 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
1577 					    stcb->sctp_ep, stcb, net);
1578 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
1579 					    stcb->sctp_ep, stcb, NULL);
1580 				}
1581 			}
1582 		} else {
1583 			/*
1584 			 * No auto close at this time, reset t-o to check
1585 			 * later
1586 			 */
1587 			uint32_t tmp;
1588 
1589 			/* fool the timer startup to use the time left */
1590 			tmp = asoc->sctp_autoclose_ticks;
1591 			asoc->sctp_autoclose_ticks -= ticks_gone_by;
1592 			sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb, NULL);
1593 			/* restore the real tick value */
1594 			asoc->sctp_autoclose_ticks = tmp;
1595 		}
1596 	}
1597 }
1598