xref: /freebsd/sys/netinet/sctp_timer.c (revision f4b37ed0f8b307b1f3f0f630ca725d68f1dff30d)
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,
114 				    stcb->sctp_ep, stcb, net,
115 				    SCTP_FROM_SCTP_TIMER + SCTP_LOC_1);
116 				sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net);
117 			}
118 		}
119 	}
120 	if (stcb == NULL)
121 		return (0);
122 
123 	if (net) {
124 		if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) {
125 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
126 				sctp_misc_ints(SCTP_THRESHOLD_INCR,
127 				    stcb->asoc.overall_error_count,
128 				    (stcb->asoc.overall_error_count + 1),
129 				    SCTP_FROM_SCTP_TIMER,
130 				    __LINE__);
131 			}
132 			stcb->asoc.overall_error_count++;
133 		}
134 	} else {
135 		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
136 			sctp_misc_ints(SCTP_THRESHOLD_INCR,
137 			    stcb->asoc.overall_error_count,
138 			    (stcb->asoc.overall_error_count + 1),
139 			    SCTP_FROM_SCTP_TIMER,
140 			    __LINE__);
141 		}
142 		stcb->asoc.overall_error_count++;
143 	}
144 	SCTPDBG(SCTP_DEBUG_TIMER4, "Overall error count for %p now %d thresh:%u state:%x\n",
145 	    (void *)&stcb->asoc, stcb->asoc.overall_error_count,
146 	    (uint32_t) threshold,
147 	    ((net == NULL) ? (uint32_t) 0 : (uint32_t) net->dest_state));
148 	/*
149 	 * We specifically do not do >= to give the assoc one more change
150 	 * before we fail it.
151 	 */
152 	if (stcb->asoc.overall_error_count > threshold) {
153 		/* Abort notification sends a ULP notify */
154 		struct mbuf *op_err;
155 
156 		op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
157 		    "Association error counter exceeded");
158 		inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_2;
159 		sctp_abort_an_association(inp, stcb, op_err, SCTP_SO_NOT_LOCKED);
160 		return (1);
161 	}
162 	return (0);
163 }
164 
165 /*
166  * sctp_find_alternate_net() returns a non-NULL pointer as long
167  * the argument net is non-NULL.
168  */
169 struct sctp_nets *
170 sctp_find_alternate_net(struct sctp_tcb *stcb,
171     struct sctp_nets *net,
172     int mode)
173 {
174 	/* Find and return an alternate network if possible */
175 	struct sctp_nets *alt, *mnet, *min_errors_net = NULL, *max_cwnd_net = NULL;
176 	int once;
177 
178 	/* JRS 5/14/07 - Initialize min_errors to an impossible value. */
179 	int min_errors = -1;
180 	uint32_t max_cwnd = 0;
181 
182 	if (stcb->asoc.numnets == 1) {
183 		/* No others but net */
184 		return (TAILQ_FIRST(&stcb->asoc.nets));
185 	}
186 	/*
187 	 * JRS 5/14/07 - If mode is set to 2, use the CMT PF find alternate
188 	 * net algorithm. This algorithm chooses the active destination (not
189 	 * in PF state) with the largest cwnd value. If all destinations are
190 	 * in PF state, unreachable, or unconfirmed, choose the desination
191 	 * that is in PF state with the lowest error count. In case of a
192 	 * tie, choose the destination that was most recently active.
193 	 */
194 	if (mode == 2) {
195 		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
196 			/*
197 			 * JRS 5/14/07 - If the destination is unreachable
198 			 * or unconfirmed, skip it.
199 			 */
200 			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
201 			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
202 				continue;
203 			}
204 			/*
205 			 * JRS 5/14/07 -  If the destination is reachable
206 			 * but in PF state, compare the error count of the
207 			 * destination to the minimum error count seen thus
208 			 * far. Store the destination with the lower error
209 			 * count.  If the error counts are equal, store the
210 			 * destination that was most recently active.
211 			 */
212 			if (mnet->dest_state & SCTP_ADDR_PF) {
213 				/*
214 				 * JRS 5/14/07 - If the destination under
215 				 * consideration is the current destination,
216 				 * work as if the error count is one higher.
217 				 * The actual error count will not be
218 				 * incremented until later in the t3
219 				 * handler.
220 				 */
221 				if (mnet == net) {
222 					if (min_errors == -1) {
223 						min_errors = mnet->error_count + 1;
224 						min_errors_net = mnet;
225 					} else if (mnet->error_count + 1 < min_errors) {
226 						min_errors = mnet->error_count + 1;
227 						min_errors_net = mnet;
228 					} else if (mnet->error_count + 1 == min_errors
229 					    && mnet->last_active > min_errors_net->last_active) {
230 						min_errors_net = mnet;
231 						min_errors = mnet->error_count + 1;
232 					}
233 					continue;
234 				} else {
235 					if (min_errors == -1) {
236 						min_errors = mnet->error_count;
237 						min_errors_net = mnet;
238 					} else if (mnet->error_count < min_errors) {
239 						min_errors = mnet->error_count;
240 						min_errors_net = mnet;
241 					} else if (mnet->error_count == min_errors
242 					    && mnet->last_active > min_errors_net->last_active) {
243 						min_errors_net = mnet;
244 						min_errors = mnet->error_count;
245 					}
246 					continue;
247 				}
248 			}
249 			/*
250 			 * JRS 5/14/07 - If the destination is reachable and
251 			 * not in PF state, compare the cwnd of the
252 			 * destination to the highest cwnd seen thus far.
253 			 * Store the destination with the higher cwnd value.
254 			 * If the cwnd values are equal, randomly choose one
255 			 * of the two destinations.
256 			 */
257 			if (max_cwnd < mnet->cwnd) {
258 				max_cwnd_net = mnet;
259 				max_cwnd = mnet->cwnd;
260 			} else if (max_cwnd == mnet->cwnd) {
261 				uint32_t rndval;
262 				uint8_t this_random;
263 
264 				if (stcb->asoc.hb_random_idx > 3) {
265 					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
266 					memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values));
267 					this_random = stcb->asoc.hb_random_values[0];
268 					stcb->asoc.hb_random_idx++;
269 					stcb->asoc.hb_ect_randombit = 0;
270 				} else {
271 					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
272 					stcb->asoc.hb_random_idx++;
273 					stcb->asoc.hb_ect_randombit = 0;
274 				}
275 				if (this_random % 2 == 1) {
276 					max_cwnd_net = mnet;
277 					max_cwnd = mnet->cwnd;	/* Useless? */
278 				}
279 			}
280 		}
281 		if (max_cwnd_net == NULL) {
282 			if (min_errors_net == NULL) {
283 				return (net);
284 			}
285 			return (min_errors_net);
286 		} else {
287 			return (max_cwnd_net);
288 		}
289 	}
290 	/*
291 	 * JRS 5/14/07 - If mode is set to 1, use the CMT policy for
292 	 * choosing an alternate net.
293 	 */
294 	else if (mode == 1) {
295 		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
296 			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
297 			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
298 				/*
299 				 * will skip ones that are not-reachable or
300 				 * unconfirmed
301 				 */
302 				continue;
303 			}
304 			if (max_cwnd < mnet->cwnd) {
305 				max_cwnd_net = mnet;
306 				max_cwnd = mnet->cwnd;
307 			} else if (max_cwnd == mnet->cwnd) {
308 				uint32_t rndval;
309 				uint8_t this_random;
310 
311 				if (stcb->asoc.hb_random_idx > 3) {
312 					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
313 					memcpy(stcb->asoc.hb_random_values, &rndval,
314 					    sizeof(stcb->asoc.hb_random_values));
315 					this_random = stcb->asoc.hb_random_values[0];
316 					stcb->asoc.hb_random_idx = 0;
317 					stcb->asoc.hb_ect_randombit = 0;
318 				} else {
319 					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
320 					stcb->asoc.hb_random_idx++;
321 					stcb->asoc.hb_ect_randombit = 0;
322 				}
323 				if (this_random % 2) {
324 					max_cwnd_net = mnet;
325 					max_cwnd = mnet->cwnd;
326 				}
327 			}
328 		}
329 		if (max_cwnd_net) {
330 			return (max_cwnd_net);
331 		}
332 	}
333 	mnet = net;
334 	once = 0;
335 
336 	if (mnet == NULL) {
337 		mnet = TAILQ_FIRST(&stcb->asoc.nets);
338 		if (mnet == NULL) {
339 			return (NULL);
340 		}
341 	}
342 	for (;;) {
343 		alt = TAILQ_NEXT(mnet, sctp_next);
344 		if (alt == NULL) {
345 			once++;
346 			if (once > 1) {
347 				break;
348 			}
349 			alt = TAILQ_FIRST(&stcb->asoc.nets);
350 			if (alt == NULL) {
351 				return (NULL);
352 			}
353 		}
354 		if (alt->ro.ro_rt == NULL) {
355 			if (alt->ro._s_addr) {
356 				sctp_free_ifa(alt->ro._s_addr);
357 				alt->ro._s_addr = NULL;
358 			}
359 			alt->src_addr_selected = 0;
360 		}
361 		if (((alt->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE) &&
362 		    (alt->ro.ro_rt != NULL) &&
363 		    (!(alt->dest_state & SCTP_ADDR_UNCONFIRMED))) {
364 			/* Found a reachable address */
365 			break;
366 		}
367 		mnet = alt;
368 	}
369 
370 	if (alt == NULL) {
371 		/* Case where NO insv network exists (dormant state) */
372 		/* we rotate destinations */
373 		once = 0;
374 		mnet = net;
375 		for (;;) {
376 			if (mnet == NULL) {
377 				return (TAILQ_FIRST(&stcb->asoc.nets));
378 			}
379 			alt = TAILQ_NEXT(mnet, sctp_next);
380 			if (alt == NULL) {
381 				once++;
382 				if (once > 1) {
383 					break;
384 				}
385 				alt = TAILQ_FIRST(&stcb->asoc.nets);
386 				if (alt == NULL) {
387 					break;
388 				}
389 			}
390 			if ((!(alt->dest_state & SCTP_ADDR_UNCONFIRMED)) &&
391 			    (alt != net)) {
392 				/* Found an alternate address */
393 				break;
394 			}
395 			mnet = alt;
396 		}
397 	}
398 	if (alt == NULL) {
399 		return (net);
400 	}
401 	return (alt);
402 }
403 
404 static void
405 sctp_backoff_on_timeout(struct sctp_tcb *stcb,
406     struct sctp_nets *net,
407     int win_probe,
408     int num_marked, int num_abandoned)
409 {
410 	if (net->RTO == 0) {
411 		net->RTO = stcb->asoc.minrto;
412 	}
413 	net->RTO <<= 1;
414 	if (net->RTO > stcb->asoc.maxrto) {
415 		net->RTO = stcb->asoc.maxrto;
416 	}
417 	if ((win_probe == 0) && (num_marked || num_abandoned)) {
418 		/* We don't apply penalty to window probe scenarios */
419 		/* JRS - Use the congestion control given in the CC module */
420 		stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout(stcb, net);
421 	}
422 }
423 
424 #ifndef INVARIANTS
425 static void
426 sctp_recover_sent_list(struct sctp_tcb *stcb)
427 {
428 	struct sctp_tmit_chunk *chk, *nchk;
429 	struct sctp_association *asoc;
430 
431 	asoc = &stcb->asoc;
432 	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
433 		if (SCTP_TSN_GE(asoc->last_acked_seq, chk->rec.data.TSN_seq)) {
434 			SCTP_PRINTF("Found chk:%p tsn:%x <= last_acked_seq:%x\n",
435 			    (void *)chk, chk->rec.data.TSN_seq, asoc->last_acked_seq);
436 			if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
437 				if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
438 					asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
439 				}
440 			}
441 			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
442 			if (PR_SCTP_ENABLED(chk->flags)) {
443 				if (asoc->pr_sctp_cnt != 0)
444 					asoc->pr_sctp_cnt--;
445 			}
446 			if (chk->data) {
447 				/* sa_ignore NO_NULL_CHK */
448 				sctp_free_bufspace(stcb, asoc, chk, 1);
449 				sctp_m_freem(chk->data);
450 				chk->data = NULL;
451 				if (asoc->prsctp_supported && PR_SCTP_BUF_ENABLED(chk->flags)) {
452 					asoc->sent_queue_cnt_removeable--;
453 				}
454 			}
455 			asoc->sent_queue_cnt--;
456 			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
457 		}
458 	}
459 	SCTP_PRINTF("after recover order is as follows\n");
460 	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
461 		SCTP_PRINTF("chk:%p TSN:%x\n", (void *)chk, chk->rec.data.TSN_seq);
462 	}
463 }
464 
465 #endif
466 
467 static int
468 sctp_mark_all_for_resend(struct sctp_tcb *stcb,
469     struct sctp_nets *net,
470     struct sctp_nets *alt,
471     int window_probe,
472     int *num_marked,
473     int *num_abandoned)
474 {
475 
476 	/*
477 	 * Mark all chunks (well not all) that were sent to *net for
478 	 * retransmission. Move them to alt for there destination as well...
479 	 * We only mark chunks that have been outstanding long enough to
480 	 * have received feed-back.
481 	 */
482 	struct sctp_tmit_chunk *chk, *nchk;
483 	struct sctp_nets *lnets;
484 	struct timeval now, min_wait, tv;
485 	int cur_rto;
486 	int cnt_abandoned;
487 	int audit_tf, num_mk, fir;
488 	unsigned int cnt_mk;
489 	uint32_t orig_flight, orig_tf;
490 	uint32_t tsnlast, tsnfirst;
491 	int recovery_cnt = 0;
492 
493 
494 	/* none in flight now */
495 	audit_tf = 0;
496 	fir = 0;
497 	/*
498 	 * figure out how long a data chunk must be pending before we can
499 	 * mark it ..
500 	 */
501 	(void)SCTP_GETTIME_TIMEVAL(&now);
502 	/* get cur rto in micro-seconds */
503 	cur_rto = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv;
504 	cur_rto *= 1000;
505 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
506 		sctp_log_fr(cur_rto,
507 		    stcb->asoc.peers_rwnd,
508 		    window_probe,
509 		    SCTP_FR_T3_MARK_TIME);
510 		sctp_log_fr(net->flight_size, 0, 0, SCTP_FR_CWND_REPORT);
511 		sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT);
512 	}
513 	tv.tv_sec = cur_rto / 1000000;
514 	tv.tv_usec = cur_rto % 1000000;
515 	min_wait = now;
516 	timevalsub(&min_wait, &tv);
517 	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
518 		/*
519 		 * if we hit here, we don't have enough seconds on the clock
520 		 * to account for the RTO. We just let the lower seconds be
521 		 * the bounds and don't worry about it. This may mean we
522 		 * will mark a lot more than we should.
523 		 */
524 		min_wait.tv_sec = min_wait.tv_usec = 0;
525 	}
526 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
527 		sctp_log_fr(cur_rto, now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME);
528 		sctp_log_fr(0, min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME);
529 	}
530 	/*
531 	 * Our rwnd will be incorrect here since we are not adding back the
532 	 * cnt * mbuf but we will fix that down below.
533 	 */
534 	orig_flight = net->flight_size;
535 	orig_tf = stcb->asoc.total_flight;
536 
537 	net->fast_retran_ip = 0;
538 	/* Now on to each chunk */
539 	cnt_abandoned = 0;
540 	num_mk = cnt_mk = 0;
541 	tsnfirst = tsnlast = 0;
542 #ifndef INVARIANTS
543 start_again:
544 #endif
545 	TAILQ_FOREACH_SAFE(chk, &stcb->asoc.sent_queue, sctp_next, nchk) {
546 		if (SCTP_TSN_GE(stcb->asoc.last_acked_seq, chk->rec.data.TSN_seq)) {
547 			/* Strange case our list got out of order? */
548 			SCTP_PRINTF("Our list is out of order? last_acked:%x chk:%x\n",
549 			    (unsigned int)stcb->asoc.last_acked_seq, (unsigned int)chk->rec.data.TSN_seq);
550 			recovery_cnt++;
551 #ifdef INVARIANTS
552 			panic("last acked >= chk on sent-Q");
553 #else
554 			SCTP_PRINTF("Recover attempts a restart cnt:%d\n", recovery_cnt);
555 			sctp_recover_sent_list(stcb);
556 			if (recovery_cnt < 10) {
557 				goto start_again;
558 			} else {
559 				SCTP_PRINTF("Recovery fails %d times??\n", recovery_cnt);
560 			}
561 #endif
562 		}
563 		if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) {
564 			/*
565 			 * found one to mark: If it is less than
566 			 * DATAGRAM_ACKED it MUST not be a skipped or marked
567 			 * TSN but instead one that is either already set
568 			 * for retransmission OR one that needs
569 			 * retransmission.
570 			 */
571 
572 			/* validate its been outstanding long enough */
573 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
574 				sctp_log_fr(chk->rec.data.TSN_seq,
575 				    chk->sent_rcv_time.tv_sec,
576 				    chk->sent_rcv_time.tv_usec,
577 				    SCTP_FR_T3_MARK_TIME);
578 			}
579 			if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) {
580 				/*
581 				 * we have reached a chunk that was sent
582 				 * some seconds past our min.. forget it we
583 				 * will find no more to send.
584 				 */
585 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
586 					sctp_log_fr(0,
587 					    chk->sent_rcv_time.tv_sec,
588 					    chk->sent_rcv_time.tv_usec,
589 					    SCTP_FR_T3_STOPPED);
590 				}
591 				continue;
592 			} else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) &&
593 			    (window_probe == 0)) {
594 				/*
595 				 * we must look at the micro seconds to
596 				 * know.
597 				 */
598 				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
599 					/*
600 					 * ok it was sent after our boundary
601 					 * time.
602 					 */
603 					continue;
604 				}
605 			}
606 			if (stcb->asoc.prsctp_supported && PR_SCTP_TTL_ENABLED(chk->flags)) {
607 				/* Is it expired? */
608 				if (timevalcmp(&now, &chk->rec.data.timetodrop, >)) {
609 					/* Yes so drop it */
610 					if (chk->data) {
611 						(void)sctp_release_pr_sctp_chunk(stcb,
612 						    chk,
613 						    1,
614 						    SCTP_SO_NOT_LOCKED);
615 						cnt_abandoned++;
616 					}
617 					continue;
618 				}
619 			}
620 			if (stcb->asoc.prsctp_supported && PR_SCTP_RTX_ENABLED(chk->flags)) {
621 				/* Has it been retransmitted tv_sec times? */
622 				if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) {
623 					if (chk->data) {
624 						(void)sctp_release_pr_sctp_chunk(stcb,
625 						    chk,
626 						    1,
627 						    SCTP_SO_NOT_LOCKED);
628 						cnt_abandoned++;
629 					}
630 					continue;
631 				}
632 			}
633 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
634 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
635 				num_mk++;
636 				if (fir == 0) {
637 					fir = 1;
638 					tsnfirst = chk->rec.data.TSN_seq;
639 				}
640 				tsnlast = chk->rec.data.TSN_seq;
641 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
642 					sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count,
643 					    0, SCTP_FR_T3_MARKED);
644 				}
645 				if (chk->rec.data.chunk_was_revoked) {
646 					/* deflate the cwnd */
647 					chk->whoTo->cwnd -= chk->book_size;
648 					chk->rec.data.chunk_was_revoked = 0;
649 				}
650 				net->marked_retrans++;
651 				stcb->asoc.marked_retrans++;
652 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
653 					sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO,
654 					    chk->whoTo->flight_size,
655 					    chk->book_size,
656 					    (uintptr_t) chk->whoTo,
657 					    chk->rec.data.TSN_seq);
658 				}
659 				sctp_flight_size_decrease(chk);
660 				sctp_total_flight_decrease(stcb, chk);
661 				stcb->asoc.peers_rwnd += chk->send_size;
662 				stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh);
663 			}
664 			chk->sent = SCTP_DATAGRAM_RESEND;
665 			SCTP_STAT_INCR(sctps_markedretrans);
666 
667 			/* reset the TSN for striking and other FR stuff */
668 			chk->rec.data.doing_fast_retransmit = 0;
669 			/* Clear any time so NO RTT is being done */
670 
671 			if (chk->do_rtt) {
672 				if (chk->whoTo->rto_needed == 0) {
673 					chk->whoTo->rto_needed = 1;
674 				}
675 			}
676 			chk->do_rtt = 0;
677 			if (alt != net) {
678 				sctp_free_remote_addr(chk->whoTo);
679 				chk->no_fr_allowed = 1;
680 				chk->whoTo = alt;
681 				atomic_add_int(&alt->ref_count, 1);
682 			} else {
683 				chk->no_fr_allowed = 0;
684 				if (TAILQ_EMPTY(&stcb->asoc.send_queue)) {
685 					chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq;
686 				} else {
687 					chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.TSN_seq;
688 				}
689 			}
690 			/*
691 			 * CMT: Do not allow FRs on retransmitted TSNs.
692 			 */
693 			if (stcb->asoc.sctp_cmt_on_off > 0) {
694 				chk->no_fr_allowed = 1;
695 			}
696 #ifdef THIS_SHOULD_NOT_BE_DONE
697 		} else if (chk->sent == SCTP_DATAGRAM_ACKED) {
698 			/* remember highest acked one */
699 			could_be_sent = chk;
700 #endif
701 		}
702 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
703 			cnt_mk++;
704 		}
705 	}
706 	if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) {
707 		/* we did not subtract the same things? */
708 		audit_tf = 1;
709 	}
710 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
711 		sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT);
712 	}
713 #ifdef SCTP_DEBUG
714 	if (num_mk) {
715 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
716 		    tsnlast);
717 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%ld\n",
718 		    num_mk, (u_long)stcb->asoc.peers_rwnd);
719 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
720 		    tsnlast);
721 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%d\n",
722 		    num_mk,
723 		    (int)stcb->asoc.peers_rwnd);
724 	}
725 #endif
726 	*num_marked = num_mk;
727 	*num_abandoned = cnt_abandoned;
728 	/*
729 	 * Now check for a ECN Echo that may be stranded And include the
730 	 * cnt_mk'd to have all resends in the control queue.
731 	 */
732 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
733 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
734 			cnt_mk++;
735 		}
736 		if ((chk->whoTo == net) &&
737 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
738 			sctp_free_remote_addr(chk->whoTo);
739 			chk->whoTo = alt;
740 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
741 				chk->sent = SCTP_DATAGRAM_RESEND;
742 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
743 				cnt_mk++;
744 			}
745 			atomic_add_int(&alt->ref_count, 1);
746 		}
747 	}
748 #ifdef THIS_SHOULD_NOT_BE_DONE
749 	if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) {
750 		/* fix it so we retransmit the highest acked anyway */
751 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
752 		cnt_mk++;
753 		could_be_sent->sent = SCTP_DATAGRAM_RESEND;
754 	}
755 #endif
756 	if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) {
757 #ifdef INVARIANTS
758 		SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n",
759 		    cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk);
760 #endif
761 #ifndef SCTP_AUDITING_ENABLED
762 		stcb->asoc.sent_queue_retran_cnt = cnt_mk;
763 #endif
764 	}
765 	if (audit_tf) {
766 		SCTPDBG(SCTP_DEBUG_TIMER4,
767 		    "Audit total flight due to negative value net:%p\n",
768 		    (void *)net);
769 		stcb->asoc.total_flight = 0;
770 		stcb->asoc.total_flight_count = 0;
771 		/* Clear all networks flight size */
772 		TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) {
773 			lnets->flight_size = 0;
774 			SCTPDBG(SCTP_DEBUG_TIMER4,
775 			    "Net:%p c-f cwnd:%d ssthresh:%d\n",
776 			    (void *)lnets, lnets->cwnd, lnets->ssthresh);
777 		}
778 		TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
779 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
780 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
781 					sctp_misc_ints(SCTP_FLIGHT_LOG_UP,
782 					    chk->whoTo->flight_size,
783 					    chk->book_size,
784 					    (uintptr_t) chk->whoTo,
785 					    chk->rec.data.TSN_seq);
786 				}
787 				sctp_flight_size_increase(chk);
788 				sctp_total_flight_increase(stcb, chk);
789 			}
790 		}
791 	}
792 	/* We return 1 if we only have a window probe outstanding */
793 	return (0);
794 }
795 
796 
797 int
798 sctp_t3rxt_timer(struct sctp_inpcb *inp,
799     struct sctp_tcb *stcb,
800     struct sctp_nets *net)
801 {
802 	struct sctp_nets *alt;
803 	int win_probe, num_mk, num_abandoned;
804 
805 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
806 		sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT);
807 	}
808 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
809 		struct sctp_nets *lnet;
810 
811 		TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
812 			if (net == lnet) {
813 				sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3);
814 			} else {
815 				sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3);
816 			}
817 		}
818 	}
819 	/* Find an alternate and mark those for retransmission */
820 	if ((stcb->asoc.peers_rwnd == 0) &&
821 	    (stcb->asoc.total_flight < net->mtu)) {
822 		SCTP_STAT_INCR(sctps_timowindowprobe);
823 		win_probe = 1;
824 	} else {
825 		win_probe = 0;
826 	}
827 
828 	if (win_probe == 0) {
829 		/* We don't do normal threshold management on window probes */
830 		if (sctp_threshold_management(inp, stcb, net,
831 		    stcb->asoc.max_send_times)) {
832 			/* Association was destroyed */
833 			return (1);
834 		} else {
835 			if (net != stcb->asoc.primary_destination) {
836 				/* send a immediate HB if our RTO is stale */
837 				struct timeval now;
838 				unsigned int ms_goneby;
839 
840 				(void)SCTP_GETTIME_TIMEVAL(&now);
841 				if (net->last_sent_time.tv_sec) {
842 					ms_goneby = (now.tv_sec - net->last_sent_time.tv_sec) * 1000;
843 				} else {
844 					ms_goneby = 0;
845 				}
846 				if ((net->dest_state & SCTP_ADDR_PF) == 0) {
847 					if ((ms_goneby > net->RTO) || (net->RTO == 0)) {
848 						/*
849 						 * no recent feed back in an
850 						 * RTO or more, request a
851 						 * RTT update
852 						 */
853 						sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
854 					}
855 				}
856 			}
857 		}
858 	} else {
859 		/*
860 		 * For a window probe we don't penalize the net's but only
861 		 * the association. This may fail it if SACKs are not coming
862 		 * back. If sack's are coming with rwnd locked at 0, we will
863 		 * continue to hold things waiting for rwnd to raise
864 		 */
865 		if (sctp_threshold_management(inp, stcb, NULL,
866 		    stcb->asoc.max_send_times)) {
867 			/* Association was destroyed */
868 			return (1);
869 		}
870 	}
871 	if (stcb->asoc.sctp_cmt_on_off > 0) {
872 		if (net->pf_threshold < net->failure_threshold) {
873 			alt = sctp_find_alternate_net(stcb, net, 2);
874 		} else {
875 			/*
876 			 * CMT: Using RTX_SSTHRESH policy for CMT. If CMT is
877 			 * being used, then pick dest with largest ssthresh
878 			 * for any retransmission.
879 			 */
880 			alt = sctp_find_alternate_net(stcb, net, 1);
881 			/*
882 			 * CUCv2: If a different dest is picked for the
883 			 * retransmission, then new (rtx-)pseudo_cumack
884 			 * needs to be tracked for orig dest. Let CUCv2
885 			 * track new (rtx-) pseudo-cumack always.
886 			 */
887 			net->find_pseudo_cumack = 1;
888 			net->find_rtx_pseudo_cumack = 1;
889 		}
890 	} else {
891 		alt = sctp_find_alternate_net(stcb, net, 0);
892 	}
893 
894 	num_mk = 0;
895 	num_abandoned = 0;
896 	(void)sctp_mark_all_for_resend(stcb, net, alt, win_probe,
897 	    &num_mk, &num_abandoned);
898 	/* FR Loss recovery just ended with the T3. */
899 	stcb->asoc.fast_retran_loss_recovery = 0;
900 
901 	/* CMT FR loss recovery ended with the T3 */
902 	net->fast_retran_loss_recovery = 0;
903 	if ((stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) &&
904 	    (net->flight_size == 0)) {
905 		(*stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) (stcb, net);
906 	}
907 	/*
908 	 * setup the sat loss recovery that prevents satellite cwnd advance.
909 	 */
910 	stcb->asoc.sat_t3_loss_recovery = 1;
911 	stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq;
912 
913 	/* Backoff the timer and cwnd */
914 	sctp_backoff_on_timeout(stcb, net, win_probe, num_mk, num_abandoned);
915 	if ((!(net->dest_state & SCTP_ADDR_REACHABLE)) ||
916 	    (net->dest_state & SCTP_ADDR_PF)) {
917 		/* Move all pending over too */
918 		sctp_move_chunks_from_net(stcb, net);
919 
920 		/*
921 		 * Get the address that failed, to force a new src address
922 		 * selecton and a route allocation.
923 		 */
924 		if (net->ro._s_addr) {
925 			sctp_free_ifa(net->ro._s_addr);
926 			net->ro._s_addr = NULL;
927 		}
928 		net->src_addr_selected = 0;
929 
930 		/* Force a route allocation too */
931 		if (net->ro.ro_rt) {
932 			RTFREE(net->ro.ro_rt);
933 			net->ro.ro_rt = NULL;
934 		}
935 		/* Was it our primary? */
936 		if ((stcb->asoc.primary_destination == net) && (alt != net)) {
937 			/*
938 			 * Yes, note it as such and find an alternate note:
939 			 * this means HB code must use this to resent the
940 			 * primary if it goes active AND if someone does a
941 			 * change-primary then this flag must be cleared
942 			 * from any net structures.
943 			 */
944 			if (stcb->asoc.alternate) {
945 				sctp_free_remote_addr(stcb->asoc.alternate);
946 			}
947 			stcb->asoc.alternate = alt;
948 			atomic_add_int(&stcb->asoc.alternate->ref_count, 1);
949 		}
950 	}
951 	/*
952 	 * Special case for cookie-echo'ed case, we don't do output but must
953 	 * await the COOKIE-ACK before retransmission
954 	 */
955 	if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
956 		/*
957 		 * Here we just reset the timer and start again since we
958 		 * have not established the asoc
959 		 */
960 		sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net);
961 		return (0);
962 	}
963 	if (stcb->asoc.prsctp_supported) {
964 		struct sctp_tmit_chunk *lchk;
965 
966 		lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc);
967 		/* C3. See if we need to send a Fwd-TSN */
968 		if (SCTP_TSN_GT(stcb->asoc.advanced_peer_ack_point, stcb->asoc.last_acked_seq)) {
969 			send_forward_tsn(stcb, &stcb->asoc);
970 			if (lchk) {
971 				/* Assure a timer is up */
972 				sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo);
973 			}
974 		}
975 	}
976 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
977 		sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX);
978 	}
979 	return (0);
980 }
981 
982 int
983 sctp_t1init_timer(struct sctp_inpcb *inp,
984     struct sctp_tcb *stcb,
985     struct sctp_nets *net)
986 {
987 	/* bump the thresholds */
988 	if (stcb->asoc.delayed_connection) {
989 		/*
990 		 * special hook for delayed connection. The library did NOT
991 		 * complete the rest of its sends.
992 		 */
993 		stcb->asoc.delayed_connection = 0;
994 		sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
995 		return (0);
996 	}
997 	if (SCTP_GET_STATE((&stcb->asoc)) != SCTP_STATE_COOKIE_WAIT) {
998 		return (0);
999 	}
1000 	if (sctp_threshold_management(inp, stcb, net,
1001 	    stcb->asoc.max_init_times)) {
1002 		/* Association was destroyed */
1003 		return (1);
1004 	}
1005 	stcb->asoc.dropped_special_cnt = 0;
1006 	sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0, 0);
1007 	if (stcb->asoc.initial_init_rto_max < net->RTO) {
1008 		net->RTO = stcb->asoc.initial_init_rto_max;
1009 	}
1010 	if (stcb->asoc.numnets > 1) {
1011 		/* If we have more than one addr use it */
1012 		struct sctp_nets *alt;
1013 
1014 		alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0);
1015 		if (alt != stcb->asoc.primary_destination) {
1016 			sctp_move_chunks_from_net(stcb, stcb->asoc.primary_destination);
1017 			stcb->asoc.primary_destination = alt;
1018 		}
1019 	}
1020 	/* Send out a new init */
1021 	sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1022 	return (0);
1023 }
1024 
1025 /*
1026  * For cookie and asconf we actually need to find and mark for resend, then
1027  * increment the resend counter (after all the threshold management stuff of
1028  * course).
1029  */
1030 int
1031 sctp_cookie_timer(struct sctp_inpcb *inp,
1032     struct sctp_tcb *stcb,
1033     struct sctp_nets *net SCTP_UNUSED)
1034 {
1035 	struct sctp_nets *alt;
1036 	struct sctp_tmit_chunk *cookie;
1037 
1038 	/* first before all else we must find the cookie */
1039 	TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) {
1040 		if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) {
1041 			break;
1042 		}
1043 	}
1044 	if (cookie == NULL) {
1045 		if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
1046 			/* FOOBAR! */
1047 			struct mbuf *op_err;
1048 
1049 			op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
1050 			    "Cookie timer expired, but no cookie");
1051 			inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_3;
1052 			sctp_abort_an_association(inp, stcb, op_err, SCTP_SO_NOT_LOCKED);
1053 		} else {
1054 #ifdef INVARIANTS
1055 			panic("Cookie timer expires in wrong state?");
1056 #else
1057 			SCTP_PRINTF("Strange in state %d not cookie-echoed yet c-e timer expires?\n", SCTP_GET_STATE(&stcb->asoc));
1058 			return (0);
1059 #endif
1060 		}
1061 		return (0);
1062 	}
1063 	/* Ok we found the cookie, threshold management next */
1064 	if (sctp_threshold_management(inp, stcb, cookie->whoTo,
1065 	    stcb->asoc.max_init_times)) {
1066 		/* Assoc is over */
1067 		return (1);
1068 	}
1069 	/*
1070 	 * cleared theshold management now lets backoff the address & select
1071 	 * an alternate
1072 	 */
1073 	stcb->asoc.dropped_special_cnt = 0;
1074 	sctp_backoff_on_timeout(stcb, cookie->whoTo, 1, 0, 0);
1075 	alt = sctp_find_alternate_net(stcb, cookie->whoTo, 0);
1076 	if (alt != cookie->whoTo) {
1077 		sctp_free_remote_addr(cookie->whoTo);
1078 		cookie->whoTo = alt;
1079 		atomic_add_int(&alt->ref_count, 1);
1080 	}
1081 	/* Now mark the retran info */
1082 	if (cookie->sent != SCTP_DATAGRAM_RESEND) {
1083 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1084 	}
1085 	cookie->sent = SCTP_DATAGRAM_RESEND;
1086 	/*
1087 	 * Now call the output routine to kick out the cookie again, Note we
1088 	 * don't mark any chunks for retran so that FR will need to kick in
1089 	 * to move these (or a send timer).
1090 	 */
1091 	return (0);
1092 }
1093 
1094 int
1095 sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1096     struct sctp_nets *net)
1097 {
1098 	struct sctp_nets *alt;
1099 	struct sctp_tmit_chunk *strrst = NULL, *chk = NULL;
1100 
1101 	if (stcb->asoc.stream_reset_outstanding == 0) {
1102 		return (0);
1103 	}
1104 	/* find the existing STRRESET, we use the seq number we sent out on */
1105 	(void)sctp_find_stream_reset(stcb, stcb->asoc.str_reset_seq_out, &strrst);
1106 	if (strrst == NULL) {
1107 		return (0);
1108 	}
1109 	/* do threshold management */
1110 	if (sctp_threshold_management(inp, stcb, strrst->whoTo,
1111 	    stcb->asoc.max_send_times)) {
1112 		/* Assoc is over */
1113 		return (1);
1114 	}
1115 	/*
1116 	 * cleared theshold management now lets backoff the address & select
1117 	 * an alternate
1118 	 */
1119 	sctp_backoff_on_timeout(stcb, strrst->whoTo, 1, 0, 0);
1120 	alt = sctp_find_alternate_net(stcb, strrst->whoTo, 0);
1121 	sctp_free_remote_addr(strrst->whoTo);
1122 	strrst->whoTo = alt;
1123 	atomic_add_int(&alt->ref_count, 1);
1124 
1125 	/* See if a ECN Echo is also stranded */
1126 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1127 		if ((chk->whoTo == net) &&
1128 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1129 			sctp_free_remote_addr(chk->whoTo);
1130 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
1131 				chk->sent = SCTP_DATAGRAM_RESEND;
1132 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1133 			}
1134 			chk->whoTo = alt;
1135 			atomic_add_int(&alt->ref_count, 1);
1136 		}
1137 	}
1138 	if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1139 		/*
1140 		 * If the address went un-reachable, we need to move to
1141 		 * alternates for ALL chk's in queue
1142 		 */
1143 		sctp_move_chunks_from_net(stcb, net);
1144 	}
1145 	/* mark the retran info */
1146 	if (strrst->sent != SCTP_DATAGRAM_RESEND)
1147 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1148 	strrst->sent = SCTP_DATAGRAM_RESEND;
1149 
1150 	/* restart the timer */
1151 	sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, strrst->whoTo);
1152 	return (0);
1153 }
1154 
1155 int
1156 sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1157     struct sctp_nets *net)
1158 {
1159 	struct sctp_nets *alt;
1160 	struct sctp_tmit_chunk *asconf, *chk;
1161 
1162 	/* is this a first send, or a retransmission? */
1163 	if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) {
1164 		/* compose a new ASCONF chunk and send it */
1165 		sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED);
1166 	} else {
1167 		/*
1168 		 * Retransmission of the existing ASCONF is needed
1169 		 */
1170 
1171 		/* find the existing ASCONF */
1172 		asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue);
1173 		if (asconf == NULL) {
1174 			return (0);
1175 		}
1176 		/* do threshold management */
1177 		if (sctp_threshold_management(inp, stcb, asconf->whoTo,
1178 		    stcb->asoc.max_send_times)) {
1179 			/* Assoc is over */
1180 			return (1);
1181 		}
1182 		if (asconf->snd_count > stcb->asoc.max_send_times) {
1183 			/*
1184 			 * Something is rotten: our peer is not responding
1185 			 * to ASCONFs but apparently is to other chunks.
1186 			 * i.e. it is not properly handling the chunk type
1187 			 * upper bits. Mark this peer as ASCONF incapable
1188 			 * and cleanup.
1189 			 */
1190 			SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n");
1191 			sctp_asconf_cleanup(stcb, net);
1192 			return (0);
1193 		}
1194 		/*
1195 		 * cleared threshold management, so now backoff the net and
1196 		 * select an alternate
1197 		 */
1198 		sctp_backoff_on_timeout(stcb, asconf->whoTo, 1, 0, 0);
1199 		alt = sctp_find_alternate_net(stcb, asconf->whoTo, 0);
1200 		if (asconf->whoTo != alt) {
1201 			sctp_free_remote_addr(asconf->whoTo);
1202 			asconf->whoTo = alt;
1203 			atomic_add_int(&alt->ref_count, 1);
1204 		}
1205 		/* See if an ECN Echo is also stranded */
1206 		TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1207 			if ((chk->whoTo == net) &&
1208 			    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1209 				sctp_free_remote_addr(chk->whoTo);
1210 				chk->whoTo = alt;
1211 				if (chk->sent != SCTP_DATAGRAM_RESEND) {
1212 					chk->sent = SCTP_DATAGRAM_RESEND;
1213 					sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1214 				}
1215 				atomic_add_int(&alt->ref_count, 1);
1216 			}
1217 		}
1218 		TAILQ_FOREACH(chk, &stcb->asoc.asconf_send_queue, sctp_next) {
1219 			if (chk->whoTo != alt) {
1220 				sctp_free_remote_addr(chk->whoTo);
1221 				chk->whoTo = alt;
1222 				atomic_add_int(&alt->ref_count, 1);
1223 			}
1224 			if (asconf->sent != SCTP_DATAGRAM_RESEND && chk->sent != SCTP_DATAGRAM_UNSENT)
1225 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1226 			chk->sent = SCTP_DATAGRAM_RESEND;
1227 		}
1228 		if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1229 			/*
1230 			 * If the address went un-reachable, we need to move
1231 			 * to the alternate for ALL chunks in queue
1232 			 */
1233 			sctp_move_chunks_from_net(stcb, net);
1234 		}
1235 		/* mark the retran info */
1236 		if (asconf->sent != SCTP_DATAGRAM_RESEND)
1237 			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1238 		asconf->sent = SCTP_DATAGRAM_RESEND;
1239 
1240 		/* send another ASCONF if any and we can do */
1241 		sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED);
1242 	}
1243 	return (0);
1244 }
1245 
1246 /* Mobility adaptation */
1247 void
1248 sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1249     struct sctp_nets *net SCTP_UNUSED)
1250 {
1251 	if (stcb->asoc.deleted_primary == NULL) {
1252 		SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n");
1253 		sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1254 		return;
1255 	}
1256 	SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary ");
1257 	SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa);
1258 	sctp_free_remote_addr(stcb->asoc.deleted_primary);
1259 	stcb->asoc.deleted_primary = NULL;
1260 	sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1261 	return;
1262 }
1263 
1264 /*
1265  * For the shutdown and shutdown-ack, we do not keep one around on the
1266  * control queue. This means we must generate a new one and call the general
1267  * chunk output routine, AFTER having done threshold management.
1268  * It is assumed that net is non-NULL.
1269  */
1270 int
1271 sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1272     struct sctp_nets *net)
1273 {
1274 	struct sctp_nets *alt;
1275 
1276 	/* first threshold managment */
1277 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1278 		/* Assoc is over */
1279 		return (1);
1280 	}
1281 	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1282 	/* second select an alternative */
1283 	alt = sctp_find_alternate_net(stcb, net, 0);
1284 
1285 	/* third generate a shutdown into the queue for out net */
1286 	sctp_send_shutdown(stcb, alt);
1287 
1288 	/* fourth restart timer */
1289 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt);
1290 	return (0);
1291 }
1292 
1293 int
1294 sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1295     struct sctp_nets *net)
1296 {
1297 	struct sctp_nets *alt;
1298 
1299 	/* first threshold managment */
1300 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1301 		/* Assoc is over */
1302 		return (1);
1303 	}
1304 	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1305 	/* second select an alternative */
1306 	alt = sctp_find_alternate_net(stcb, net, 0);
1307 
1308 	/* third generate a shutdown into the queue for out net */
1309 	sctp_send_shutdown_ack(stcb, alt);
1310 
1311 	/* fourth restart timer */
1312 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt);
1313 	return (0);
1314 }
1315 
1316 static void
1317 sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp,
1318     struct sctp_tcb *stcb)
1319 {
1320 	struct sctp_stream_queue_pending *sp;
1321 	unsigned int i, chks_in_queue = 0;
1322 	int being_filled = 0;
1323 
1324 	/*
1325 	 * This function is ONLY called when the send/sent queues are empty.
1326 	 */
1327 	if ((stcb == NULL) || (inp == NULL))
1328 		return;
1329 
1330 	if (stcb->asoc.sent_queue_retran_cnt) {
1331 		SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n",
1332 		    stcb->asoc.sent_queue_retran_cnt);
1333 		stcb->asoc.sent_queue_retran_cnt = 0;
1334 	}
1335 	if (stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1336 		/* No stream scheduler information, initialize scheduler */
1337 		stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc, 0);
1338 		if (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1339 			/* yep, we lost a stream or two */
1340 			SCTP_PRINTF("Found additional streams NOT managed by scheduler, corrected\n");
1341 		} else {
1342 			/* no streams lost */
1343 			stcb->asoc.total_output_queue_size = 0;
1344 		}
1345 	}
1346 	/* Check to see if some data queued, if so report it */
1347 	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
1348 		if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) {
1349 			TAILQ_FOREACH(sp, &stcb->asoc.strmout[i].outqueue, next) {
1350 				if (sp->msg_is_complete)
1351 					being_filled++;
1352 				chks_in_queue++;
1353 			}
1354 		}
1355 	}
1356 	if (chks_in_queue != stcb->asoc.stream_queue_cnt) {
1357 		SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n",
1358 		    stcb->asoc.stream_queue_cnt, chks_in_queue);
1359 	}
1360 	if (chks_in_queue) {
1361 		/* call the output queue function */
1362 		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1363 		if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1364 		    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1365 			/*
1366 			 * Probably should go in and make it go back through
1367 			 * and add fragments allowed
1368 			 */
1369 			if (being_filled == 0) {
1370 				SCTP_PRINTF("Still nothing moved %d chunks are stuck\n",
1371 				    chks_in_queue);
1372 			}
1373 		}
1374 	} else {
1375 		SCTP_PRINTF("Found no chunks on any queue tot:%lu\n",
1376 		    (u_long)stcb->asoc.total_output_queue_size);
1377 		stcb->asoc.total_output_queue_size = 0;
1378 	}
1379 }
1380 
1381 int
1382 sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1383     struct sctp_nets *net)
1384 {
1385 	uint8_t net_was_pf;
1386 
1387 	if (net->dest_state & SCTP_ADDR_PF) {
1388 		net_was_pf = 1;
1389 	} else {
1390 		net_was_pf = 0;
1391 	}
1392 	if (net->hb_responded == 0) {
1393 		if (net->ro._s_addr) {
1394 			/*
1395 			 * Invalidate the src address if we did not get a
1396 			 * response last time.
1397 			 */
1398 			sctp_free_ifa(net->ro._s_addr);
1399 			net->ro._s_addr = NULL;
1400 			net->src_addr_selected = 0;
1401 		}
1402 		sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1403 		if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1404 			/* Assoc is over */
1405 			return (1);
1406 		}
1407 	}
1408 	/* Zero PBA, if it needs it */
1409 	if (net->partial_bytes_acked) {
1410 		net->partial_bytes_acked = 0;
1411 	}
1412 	if ((stcb->asoc.total_output_queue_size > 0) &&
1413 	    (TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1414 	    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1415 		sctp_audit_stream_queues_for_size(inp, stcb);
1416 	}
1417 	if (!(net->dest_state & SCTP_ADDR_NOHB) &&
1418 	    !((net_was_pf == 0) && (net->dest_state & SCTP_ADDR_PF))) {
1419 		/*
1420 		 * when move to PF during threshold mangement, a HB has been
1421 		 * queued in that routine
1422 		 */
1423 		uint32_t ms_gone_by;
1424 
1425 		if ((net->last_sent_time.tv_sec > 0) ||
1426 		    (net->last_sent_time.tv_usec > 0)) {
1427 			struct timeval diff;
1428 
1429 			SCTP_GETTIME_TIMEVAL(&diff);
1430 			timevalsub(&diff, &net->last_sent_time);
1431 			ms_gone_by = (uint32_t) (diff.tv_sec * 1000) +
1432 			    (uint32_t) (diff.tv_usec / 1000);
1433 		} else {
1434 			ms_gone_by = 0xffffffff;
1435 		}
1436 		if ((ms_gone_by >= net->heart_beat_delay) ||
1437 		    (net->dest_state & SCTP_ADDR_PF)) {
1438 			sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
1439 		}
1440 	}
1441 	return (0);
1442 }
1443 
1444 void
1445 sctp_pathmtu_timer(struct sctp_inpcb *inp,
1446     struct sctp_tcb *stcb,
1447     struct sctp_nets *net)
1448 {
1449 	uint32_t next_mtu, mtu;
1450 
1451 	next_mtu = sctp_get_next_mtu(net->mtu);
1452 
1453 	if ((next_mtu > net->mtu) && (net->port == 0)) {
1454 		if ((net->src_addr_selected == 0) ||
1455 		    (net->ro._s_addr == NULL) ||
1456 		    (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1457 			if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1458 				sctp_free_ifa(net->ro._s_addr);
1459 				net->ro._s_addr = NULL;
1460 				net->src_addr_selected = 0;
1461 			} else if (net->ro._s_addr == NULL) {
1462 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1463 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1464 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1465 
1466 					/* KAME hack: embed scopeid */
1467 					(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
1468 				}
1469 #endif
1470 
1471 				net->ro._s_addr = sctp_source_address_selection(inp,
1472 				    stcb,
1473 				    (sctp_route_t *) & net->ro,
1474 				    net, 0, stcb->asoc.vrf_id);
1475 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1476 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1477 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1478 
1479 					(void)sa6_recoverscope(sin6);
1480 				}
1481 #endif				/* INET6 */
1482 			}
1483 			if (net->ro._s_addr)
1484 				net->src_addr_selected = 1;
1485 		}
1486 		if (net->ro._s_addr) {
1487 			mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_rt);
1488 #if defined(INET) || defined(INET6)
1489 			if (net->port) {
1490 				mtu -= sizeof(struct udphdr);
1491 			}
1492 #endif
1493 			if (mtu > next_mtu) {
1494 				net->mtu = next_mtu;
1495 			}
1496 		}
1497 	}
1498 	/* restart the timer */
1499 	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
1500 }
1501 
1502 void
1503 sctp_autoclose_timer(struct sctp_inpcb *inp,
1504     struct sctp_tcb *stcb,
1505     struct sctp_nets *net)
1506 {
1507 	struct timeval tn, *tim_touse;
1508 	struct sctp_association *asoc;
1509 	int ticks_gone_by;
1510 
1511 	(void)SCTP_GETTIME_TIMEVAL(&tn);
1512 	if (stcb->asoc.sctp_autoclose_ticks &&
1513 	    sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) {
1514 		/* Auto close is on */
1515 		asoc = &stcb->asoc;
1516 		/* pick the time to use */
1517 		if (asoc->time_last_rcvd.tv_sec >
1518 		    asoc->time_last_sent.tv_sec) {
1519 			tim_touse = &asoc->time_last_rcvd;
1520 		} else {
1521 			tim_touse = &asoc->time_last_sent;
1522 		}
1523 		/* Now has long enough transpired to autoclose? */
1524 		ticks_gone_by = SEC_TO_TICKS(tn.tv_sec - tim_touse->tv_sec);
1525 		if ((ticks_gone_by > 0) &&
1526 		    (ticks_gone_by >= (int)asoc->sctp_autoclose_ticks)) {
1527 			/*
1528 			 * autoclose time has hit, call the output routine,
1529 			 * which should do nothing just to be SURE we don't
1530 			 * have hanging data. We can then safely check the
1531 			 * queues and know that we are clear to send
1532 			 * shutdown
1533 			 */
1534 			sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED);
1535 			/* Are we clean? */
1536 			if (TAILQ_EMPTY(&asoc->send_queue) &&
1537 			    TAILQ_EMPTY(&asoc->sent_queue)) {
1538 				/*
1539 				 * there is nothing queued to send, so I'm
1540 				 * done...
1541 				 */
1542 				if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) {
1543 					/* only send SHUTDOWN 1st time thru */
1544 					struct sctp_nets *netp;
1545 
1546 					if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
1547 					    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
1548 						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
1549 					}
1550 					SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
1551 					SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
1552 					sctp_stop_timers_for_shutdown(stcb);
1553 					if (stcb->asoc.alternate) {
1554 						netp = stcb->asoc.alternate;
1555 					} else {
1556 						netp = stcb->asoc.primary_destination;
1557 					}
1558 					sctp_send_shutdown(stcb, netp);
1559 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
1560 					    stcb->sctp_ep, stcb,
1561 					    netp);
1562 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
1563 					    stcb->sctp_ep, stcb,
1564 					    netp);
1565 				}
1566 			}
1567 		} else {
1568 			/*
1569 			 * No auto close at this time, reset t-o to check
1570 			 * later
1571 			 */
1572 			int tmp;
1573 
1574 			/* fool the timer startup to use the time left */
1575 			tmp = asoc->sctp_autoclose_ticks;
1576 			asoc->sctp_autoclose_ticks -= ticks_gone_by;
1577 			sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb,
1578 			    net);
1579 			/* restore the real tick value */
1580 			asoc->sctp_autoclose_ticks = tmp;
1581 		}
1582 	}
1583 }
1584